**PARPs, PAR and NAD Metabolism and Their Inhibitors in Cancer**

Editors

**Nicola Curtin P ´eter Bay**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editors* Nicola Curtin Newcastle University Centre for Cancer Newcastle upon Tyne UK

Peter Bay ´ University of Debrecen Debrecen Hungary

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Cancers* (ISSN 2072-6694) (available at: https://www.mdpi.com/journal/cancers/special issues/parp2019).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-0365-8156-9 (Hbk) ISBN 978-3-0365-8157-6 (PDF)**

© 2023 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

### **Contents**


Reprinted from: *Cancers* **2019**, *11*, 1539, doi:10.3390/cancers11101539 ................ **161**


#### **Jos ´e Y´elamos, Lucia Moreno-Lama, Jaime Jimeno and Syed O. Ali**


### PARylation During Transcription: Insights into the Fine-Tuning Mechanism and Regulation Reprinted from: *Cancers* **2020**, *12*, 183, doi:10.3390/cancers12010183 ................. **435**

### *Editorial* **PARPs, PAR and NAD Metabolism and Their Inhibitors in Cancer**

**Nicola Curtin 1,\* and Péter Bai 2,3,4,\***


Received: 15 September 2020; Accepted: 19 October 2020; Published: 24 November 2020

The role of poly(ADP-ribose) polymerase-1 (PARP1) in DNA repair and as a potential target for anticancer therapy has been under investigation for more than 50 years. The field has expanded over the decades to include not only a family of ADP-ribosylating enzymes (PARPs/ARTDs), but also interacting and polymer-degrading proteins. In this special issue of *Cancers* primary research articles and reviews describe various aspects of PARP biology along with therapeutic targeting. Some historical perspective is reviewed along with the development of the PARP inhibitor (PARPi), rucaparib/Rubraca®, including the identification of the synthetic lethality with homologous recombination repair (e.g., BRCA) defects [1]. Further consideration of the mode of action of several PARPi and their interaction with cytotoxic chemotherapy is described by Min and Im [2], and the evidence for a unique mode-of-action of another PARPi, PJ34 [3] is summarized. Several other aspects of PARPi therapy are described, including research indicating the therapeutic potential of the PARPi, olaparib, as a single agent in myelodysplastic syndromes, not only through cytotoxic and cytostatic effects but also through the induction of differentiation [4], along with the augmentation of UVB-induced DNA damage and mitochondrial alterations resulting in reduced proliferation and viability of keratinocytes [5]. On the other side of the coin, deficiency of the enzyme that degrades the PARP-generated ADP-ribose polymers, PARG, in ES cells resulted delayed tumour growth when they were implanted SC and increased antitumour activity of X-rays [6].

Moving PARPi therapy beyond BRCA mutated cancers requires the use of biomarkers to predict PARPi sensitivity as reviewed by Singh et al. [7]. The role of defects in the G1 cell cycle checkpoint signaling kinase, ATM, as a determinant of sensitivity to PARPi was reviewed, with the finding that PARPi alone are cytostatic in ATM defective cancer cells but require the addition of an inhibitor of the S/G2 cell cycle checkpoint kinase, ATR, to induce cell death [8]. An investigation of the synergy between the PARPi, olaparib, and the ATR inhibitor VE-821 in a panel of neuroblastoma cells, revealed that it was independent of MYCN or ATM status in these cells [9] and similar studies identified that the synergy between an inhibitor of the S/G2 cell cycle checkpoint kinase, CHK1, and the PARPi, rucaparib was largely through the impairment of homologous recombination repair by the CHK1 inhibitor [10]. Depletion of p60/150 CAF-1 also impaired homologous recombination repair thereby inducing sensitivity to PARPi and irradiation in head and neck cancers with therapeutic implications [11]. Interestingly, depletion of NMNAT1, which is involved in the synthesis of NAD+, PARP's substrate, induced DNA damage and sensitized cells to cisplatin but exhibited redundancy with PARPi in this respect [12]. Inhibition of EGFR or Syk (which mediates EGFR signaling) was synergistically cytotoxic in combination with olaparib in squamous cell carcinoma cells suggesting therapeutic potential [13].

Of course, resistance to PARPi is an emerging problem clinically and the role of the PI3K-AKT pathway in protection from PARPi-induced cytotoxicity and its significance in shock, inflammation, ischemia-reperfusion injury and cancer is reviewed by Gallyas et al. [14]. Interestingly, gastric cancer cell lines made resistant to olaparib were found to be cross-resistant to cisplatin, but had increased sensitivity to irinotecan due to upregulation of TOP1 and TDP1, which has therapeutic implications for patients who develop PARPi resistance [15].

PARP has been investigated as a regulator of transcription and key roles regarding the role of PARP both in nucleolar function in relation to cancer biology and the role PARylation plays in the regulation of transcription were reviewed [16,17]. With original research showing that PARylation activates the histone acetyl transferase, EP300, contributing to its regulation of transcription of DNA repair and proliferation genes [18].

An important factor in cancer biology is the tumour microenvironment and the interaction between PARP and key features of the tumour microenvironment such as autophagy, hypoxia and angiogenesis was reviewed [19]. There is significant interest in the immune microenvironment and the roles of both PARP1 and PARP2 in modulating both the innate and adaptive immune system was reviewed by Yelamos et al. [20], with the therapeutic potential of the combination of PARPi with immune checkpoint inhibitors, including translational and clinical studies were reviewed by Peyraud and Italiano [21].

Feijs and colleagues [22] reviewed the roles of ADP-ribosyl hydrolases, also involved in the degradation of ADP-ribose and poly(ADP-ribose) chains, MACROD1, MACROD2 and TARG1 in carcinogenesis. The role of ARH1, another degradatory enzyme, was reviewed by Ishiwata-Endo and colleagues [23] in cancer and non-cancer diseases.

The roles for poly(ADP-ribosyl)ation in biochemistry, cell biology, physiology and pathophysiology are rapidly expanding, the findings discussed in the "PARPs, PAR and NAD Metabolism and Their Inhibitors in Cancer" special issue of Cancers will surely provide a better understanding of these processes and widen the scope of our appreciation of poly(ADP-ribosyl)ation.

**Funding:** P.B. was supported by grants from NKFIH (K123975 and GINOP-2.3.2-15-2016-00006) and by the Higher Education Institutional Excellence Programme (NKFIH-1150-6/2019) of the Ministry of Innovation and Technology in Hungary, within the framework of the Biotechnology thematic programme of the University of Debrecen. NC is funded by Newcastle University.

**Conflicts of Interest:** N.C. has worked for many years on the development of rucaparib resulting in 2 patents WO 2005/012305 A2 andWO/2006/033006 with royalties paid to CRUK and Newcastle University through which she receives a percentage. She has received research funding foe PARP-related studies from Agouron Pharmaceuticals, Pfizer, BioMarin and consulted for Abbvie, Eisai and Tesaro.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Targeting Nuclear NAD**<sup>+</sup> **Synthesis Inhibits DNA Repair, Impairs Metabolic Adaptation and Increases Chemosensitivity of U-2OS Osteosarcoma Cells**

**Alexandra Kiss 1,2, Arnold Péter Ráduly 1, Zsolt Regdon 1, Zsuzsanna Polgár 1, Szabolcs Tarapcsák 3, Isotta Sturniolo 1, Tarek El-Hamoly 1,4, László Virág 1,5,\* and Csaba Heged ˝us 1,\***


Received: 11 February 2020; Accepted: 4 May 2020; Published: 7 May 2020

**Abstract:** Osteosarcoma (OS) is the most common bone tumor in children and adolescents. Modern OS treatment, based on the combination of neoadjuvant chemotherapy (cisplatin + doxorubicin + methotrexate) with subsequent surgical removal of the primary tumor and metastases, has dramatically improved overall survival of OS patients. However, further research is needed to identify new therapeutic targets. Here we report that expression level of the nuclear NAD synthesis enzyme, nicotinamide mononucleotide adenylyltransferase-1 (NMNAT1), increases in U-2OS cells upon exposure to DNA damaging agents, suggesting the involvement of the enzyme in the DNA damage response. Moreover, genetic inactivation of NMNAT1 sensitizes U-2OS osteosarcoma cells to cisplatin, doxorubicin, or a combination of these two treatments. Increased cisplatin-induced cell death of NMNAT1−/<sup>−</sup> cells showed features of both apoptosis and necroptosis, as indicated by the protective effect of the caspase-3 inhibitor z-DEVD-FMK and the necroptosis inhibitor necrostatin-1. Activation of the DNA damage sensor enzyme poly(ADP-ribose) polymerase 1 (PARP1), a major consumer of NAD<sup>+</sup> in the nucleus, was fully blocked by NMNAT1 inactivation, leading to increased DNA damage (phospho-H2AX foci). The PARP inhibitor, olaparib, sensitized wild type but not NMNAT1−/<sup>−</sup> cells to cisplatin-induced anti-clonogenic effects, suggesting that impaired PARP1 activity is important for chemosensitization. Cisplatin-induced cell death of NMNAT1−/<sup>−</sup> cells was also characterized by a marked drop in cellular ATP levels and impaired mitochondrial respiratory reserve capacity, highlighting the central role of compromised cellular bioenergetics in chemosensitization by NMNAT1 inactivation. Moreover, NMNAT1 cells also displayed markedly higher sensitivity to cisplatin when grown as spheroids in 3D culture. In summary, our work provides the first evidence that NMNAT1 is a promising therapeutic target for osteosarcoma and possibly other tumors as well.

**Keywords:** NAD+; NMNAT1; cisplatin; chemotherapy; apoptosis; PARP1; osteosarcoma; cancer

#### **1. Introduction**

Bone tumors constitute 4–7% of all cancers in children aged 0–14 years and 7–8% in children aged 15–19 years [1]. Osteosarcoma is the most prevalent neoplasm of the bone in children and adolescents, with approximately 850,000 cases reported each year in the US. Tumors are typically diagnosed between 10 and 30 years of age and mostly arise in the femur, tibia, and humerus. Although most osteosarcoma cases are sporadic, a higher incidence of osteosarcoma has been reported in certain hereditary cancer syndromes. For example, mutations of the tumor suppressor retinoblastoma protein (Rb) in retinoblastoma [2] or p53 in Li-Fraumeni syndrome [3], as well as the DNA repair helicase Werner protein are also associated with osteosarcoma [4]. Previous radiotherapy of other tumors may also lead, in the long term, to osteosarcoma development.

Treatment of osteosarcoma usually begins with neoadjuvant chemotherapy followed by surgical removal of the primary tumor and metastases (typically localized to the lungs). Agents used for osteosarcoma chemotherapy include cisplatin, methotrexate, and doxorubicin [5]. In the European Union, the liposome-encapsulated synthetic muramyl dipeptide analog, mifamurtide, is also approved by the European Medicines Agency for the immunotherapy of osteosarcoma [6]. The five-year survival rate depends on the stage, type, and subtype of osteosarcoma and ranges from 60% to 75% [7]. The rather poor therapeutic responsiveness indicates that new treatment modalities are clearly needed to improve the disease-free survival of osteosarcoma patients.

Neoplastic transformation is accompanied by fundamental rearrangements of metabolic pathways. A key hallmark of cancer metabolism is the reliance of tumor cells on glycolysis, even if oxygen is available (aerobic glycolysis, also known as the Warburg effect) [8]. NAD<sup>+</sup> is a central metabolite of energy production that serves as an electron carrier in glycolysis and the tricarboxylic acid (TCA) cycle. In addition to its role as a metabolic cofactor, NAD<sup>+</sup> also has signaling roles [9]. NAD+-derived signaling messengers include the calcium-mobilizing agents cyclic ADP-ribose [10] and nicotinic acid adenine dinucleotide phosphate (NAADP) [11]. Moreover, members of the ADP-ribosyltransferase (ART) and sirtuin enzyme families also use NAD<sup>+</sup> as a substrate converting it to ADP-ribose and nicotinamide [12]. Sirtuins are NAD+-dependent histone deacetylases that transfer the removed acetyl group onto NAD+-derived ADP-ribose to generate O-acetyl-ADP-ribose, whereas ART enzymes cleave NAD<sup>+</sup> to nicotinamide and ADP-ribose and transfer the latter onto protein acceptors [13]. Some ARTs function as mono-ADP-ribosyltransferases while others cleave multiple NAD+s and polymerize the resulting ADP-ribose moieties onto suitable protein acceptors, resulting in protein poly(ADP-ribosyl)ation (PARylation) [14,15]. This latter group includes poly(ADP-ribose) polymerase-1 (PARP1; also known as ARTD1), PARP2 (ARTD2), and tankyrases (ARTD5-6) [16]. PARP1, a major NAD<sup>+</sup> consumer, plays a central role in tumor biology [17]. On one hand, PARP1 regulates DNA replication, gene expression, DNA repair, cell adhesion, and migration [18]. On the other hand, PARP1 is the primary target for a novel cancer therapeutic modality based on the synthetic lethality paradigm [19,20]. Clinically used PARP inhibitors (PARPi) take advantage of the vulnerability of cancer cells (e.g., ovarian and breast carcinomas) that carry mutations in the DNA repair genes BRCA1/2, rendering them sensitive to DNA repair inhibitors, such as PARPi compounds. PARP inhibitors have also been evaluated as chemosensitizers (e.g., in combination with the DNA alkylating agent temozolomide or the topoisomerase inhibitor topotecan) [21–23], but myelotoxicity prevented the successful translation of promising preclinical results into the clinic.

The question arises whether and how NAD<sup>+</sup> metabolism can be targeted for cancer therapy. NAD<sup>+</sup> is synthesized through multiple routes [24]. In humans, the central pathway relies on nucleobases with special preference for nicotinamide (Nam), although nicotinic acid (NA) can also serve as starting material for NAD<sup>+</sup> synthesis [25]. Nicotinamide phosphoribosyltransferase (NAMPT) enzymes can use Nam and phosphoribosyl pyrophosphate (PRPP) to make nicotinamide mononucleotide (NMN) [26]. In turn, nicotinamide mononucleotide adenosyltransferases (NMNAT) catalyze the final step by forming NAD<sup>+</sup> from NMN and ATP. Tissue and subcellular distribution of the three NMNAT isoforms (NMNAT1-3) show distinct patterns [27]. Besides the nuclear (NMNAT1) and cytoplasmic/Golgi

(NMNAT2) isoforms, a mitochondrial NAD<sup>+</sup> synthase (NMNAT3) has also been described; however, the mitochondrial localization and contribution of NMNAT3 to the mitochondrial NAD<sup>+</sup> pool has been questioned [28]. Nuclear NAD<sup>+</sup> synthesis is of special interest because several major NAD+-consuming enzymes (PARP1,2, SIRT1,6,7) that mediate important survival mechanisms operate in the nuclear compartment [29].

Since inhibitor molecules are available for NAMPT, our knowledge on the targetability of NAD<sup>+</sup> synthesis for cancer therapy comes from experiments with the NAMPT inhibitor compounds, FK866 and GMX1777. Although these drug candidates efficiently kill cancer cells, they also show some toxicity in non-transformed cells. In clinical trials, these drugs were relatively well-tolerated and displayed acceptable safety profiles [30]. 2,3-Dibromo-1,4-naphthoquinone (DBNQ) was recently identified as an inhibitor of purified NMNAT-1 [31]. However, this finding needs to be confirmed in cell-based assays.

The role of NMNAT1 in cancer cells is not known and its role in cancer cell chemosensitivity has not yet been investigated. In our current study, we aimed to investigate whether NMNAT1 plays a role in maintaining cell viability and whether it contributes to osteosarcoma cell survival following genotoxic stimuli. We report that U-2OS human osteosarcoma cells tolerate the genetic inactivation of the NMNAT1 gene and show increased susceptibility to cisplatin and doxorubicin. Increased chemosensitivity of the NMNAT1−/<sup>−</sup> phenotype is likely due in part to restricted DNA-damage-induced protein PARylation and consequently enhanced DNA damage. U-2OS cells unable to respond sufficiently to cisplatin treatment undergo caspase-mediated apoptotic and caspase-independent necroptotic death, suggesting that NMNAT1 may be worth investigating further as a potential target in cancer therapy with special regard to osteosarcoma treatment.

#### **2. Results**

#### *2.1. DNA Damaging Drugs Induce NMNAT1 Expression in Tumor Cell Lines*

Up to date NMNAT1 mRNA expression levels for osteosarcoma cells (U-2OS, SAOS-2) were not available. That is why we checked the mRNA levels in osteosarcoma cells and compared with other human tumor cell lines of different origins. Eleven human tumor cell lines were tested for NMNAT-1 mRNA expression (Figure 1A). The transcript could be detected, although at different levels, in all cell lines. Compared to the average expression level, A431 cells displayed significantly higher mRNA expression, while significantly lower expression was detected in A549, Capan2, MCF7 and HepG2 cell lines (Figure 1A). U-2OS osteosarcoma cells which showed average levels of NMNAT1 expression were chosen for further investigation. Of the three human NMNAT isoforms, NMNAT-1 is most highly expressed, in U-2OS cells, while NMNAT2 and NMNAT-3 are expressed at lower levels (Figure S1A). The DNA damaging chemotherapeutic agents doxorubicin and cisplatin significantly upregulated NMNAT1 mRNA expression (Figure 1B), suggesting that the enzyme may be a survival factor in DNA damage situations. Cisplatin, one of the drugs most commonly used for the treatment of osteosarcoma, caused a marked elevation in NMNAT1 protein expression in U-2OS cells (Figure S1B). Cisplatin treatment caused a concentration-dependent cytotoxicity in U-2OS cells (Figure 1C). Based on the viability data, we chose the 6.25 μg/mL concentration for the following experiments. At this concentration, cisplatin caused no significant change in NAD<sup>+</sup> levels (Figure 1D).

**Figure 1.** Nicotinamide mononucleotide adenylyltransferase-1 (NMNAT1) mRNA expression and induction by antitumor agents. The expression of NMNAT1 mRNA was determined in eleven human tumor cell lines. The y-axis shows the average expression level of the tested cell lines. Bars marked with asterisks are significantly different from the average expression (Student–Newman–Keuls method; \* *p* < 0.05) (**A**). NMNAT1 expression in the U-2OS cell line was induced 24 h after cisplatin (6.25 μg/mL) or doxorubicin (2 μg/mL) treatment. Bars marked with asterisks are significantly different from the control (Dunnett test; \* *p* < 0.05) (**B**). Calcein acetoxymethyl (Calcein AM) assay, indicating the concentration-dependent cytotoxic effect of cisplatin (3.125–50 μg/mL) on U-2OS cells, was measured 24 h after cisplatin treatment. Bars marked with asterisks are significantly different from the control (Bonferroni test; \* *p* < 0.05) (**C**). Total NAD<sup>+</sup> content was measured in cell lysates 24 h after cisplatin (6.25 μg/mL) treatment and normalized to protein content. Bars marked with asterisks are significantly different from the control (Student's *t* test; \* *p* < 0.05, N.S.: not significant) (**D**). Data plotted are means ± SEM (*n* = 3).

#### *2.2. Generation and Characterization of an NMNAT1*−/<sup>−</sup> *Cell Line*

To investigate the role of NMNAT1 in the survival of cisplatin-treated cells, we inactivated the gene for NMNAT1 using CRISPR-Cas9 technology. Single cell clones were obtained by cell sorting from cultures of NMNAT1−/<sup>−</sup> cells. We tested all the clones and all of them lacked NMNAT1 mRNA (Figure 2A). Clone 1B6 was selected for downstream experiments. Western blotting proved that NMNAT1 protein was missing from this clone (Figure 2B). Morphological properties of wild type and NMNAT1−/<sup>−</sup> cells (Figure S2A) revealed a significant reduction in the nuclear size and cell size (Figure S2B and C). The nuclear and cellular roundness was also slightly but significantly affected by the absence of a functional NMNAT1 protein (Figure S2D,E). The NMNAT1 deficient U-2OS cell line showed unaltered cell viability, as determined using the Calcein acetoxymethyl (Calcein AM) method (Figure 2C). However, clonogenic activity was impaired in the absence of a functional enzyme (Figure 2D). Despite elevated NMNAT-2 expression (Figure S1A), total cellular NAD<sup>+</sup> levels dropped to approximately one third of the control cell line (Figure 2E), indicating that NMNAT1 plays a dominant role in cellular NAD<sup>+</sup> synthesis. Interestingly, lower NAD<sup>+</sup> levels in NMNAT1−/<sup>−</sup> cells did not suppress ATP levels (Figure 2F) or impair cellular respiration, as indicated by the unchanged oxygen consumption rate (Figure 2G). Extracellular acidification rate (ECAR), a measure of glycolysis, showed higher values in the absence of NMNAT1 compared to the parent cell line (Figure 2H).

**Figure 2.** Characterization of NMNAT 1 KO cell line. NMNAT1 knockout cell lines were generated with CRISPR-CAS9 technology. Puromycin resistant cells were sorted and single cell colonies were grown. NMNAT1 mRNA levels were measured with RT-QPCR in each colony. Results are expressed as a percentage of NMNAT1 expression of the wild type U-2OS cell line (control). Bars marked with asterisks are significantly different from the control (Dunnett test; \* *p* < 0.05) (**A**). Clone 1B6 was chosen for further investigation. NMNAT1 protein was measured in cell lysates of wild type U-2OS and the 1B6 clone with Western blot (**B**). Full WB image can be found in Supplementary Material. The following experiments compare the basic characteristics of wild type (WT) and NMNAT1 knockout (KO) cells. Viability was measured with a Calcein AM viability assay. Data points marked with asterisks are significantly different from the control (Student's t test; \* *p* < 0.05, N.S.: not significant) (**C**). Clonogenic activity was assessed on day 6 by counting crystal violet stained colonies. Bars marked with asterisks are significantly different from the control (Student's *t* test; \* *p* < 0.05, N.S.: not significant) (**D**). Basal total NAD<sup>+</sup> (**E**) and ATP (**F**) levels were assayed from cell lysates and normalized to protein content. Bars marked with asterisks are significantly different from the control (Student's *t* test; \*\*\* *p* < 0.001, N.S.: not significant) Metabolic parameters such as oxygen consumption rate (OCR)/oxidative phosphorylation (**G**) and extracellular acidification rate (ECAR)/glycolytic activity (**H**) were measured with a Seahorse metabolic analyzer. Bars marked with asterisks are significantly different from the control (Student's *t* test; \*\*\* *p* < 0.001, n.s.: not significant). Data plotted are means ± SEM (*n* = 3).

#### *2.3. NMNAT-1 Inactivation Increases Chemosensitivity of the U-2OS Osteosarcoma Cells*

Next, we investigated the role of NMNAT1 in the survival/death of cisplatin-treated cells. We found that the absence of NMNAT1 sensitized cells to the toxic effects of cisplatin (Figure 3A). For example, 5 μg/mL cisplatin did not cause significant toxicity in the parent cell line but killed 45% of NMNAT1−/<sup>−</sup> cells (Figure 3A). Morphological phenotyping of cisplatin-treated cells was performed to understand the mechanism of chemosensitization in NMNAT1−/<sup>−</sup> cells.

**Figure 3.** Chemosensitivity of NMNAT-1 KO cell line. Chemosensitivity of wild type (WT) and knockout (KO) cells was compared in a Calcein-AM viability assay 24 h after treatment with different concentrations of cisplatin (0.625–10 μg/mL). A new batch of cisplatin was used for this and all subsequent experiments. Results are expressed as percentages, compared to the vehicle (medium) treated samples. Data points marked with asterisks are significantly different from the control (Bonferroni test; \*\*\* *p* < 0.001). Data points marked with hashmarks are significantly different from the corresponding treatment of the wild type cells (Bonferroni test; ### *p* < 0.001). (**A**). Cisplatin-induced DNA damage was assessed by quantifying the P-H2AX signal after immunofluorescent staining. High-content analysis detected three different cell morphologies, "normal", "spotted" and "fragmented", representative images are shown (**B**). Chart shows the percentage of the three types of morphologies in untreated and in cisplatin-treated samples at 12, 18, and 24 h after cisplatin treatment. Bars marked with asterisks are significantly different from the control (0 hours of cisplatin treatment) (Student–Newman–Keuls method; \* *p* < 0.05). Bars marked with hashmarks are significantly different from the corresponding treatment of the wild type cells (Student–Newman–Keuls method; # *p* < 0.05) (**B**). Error bars are not shown on panel B because of the 3D presentation. Lactate dehydrogenase (LDH) release was determined from the supernatants at 24 h after cisplatin treatment and expressed as a percentage of the positive control (lysed cells). Cells were treated with cisplatin or with the combination of cisplatin with the necroptosis inhibitor, concentration necrostatin-1 (NEC1). Bars marked with asterisks are significantly different from the control (not shown) (Dunnett test; \* *p* < 0.05). Bars marked with hashmarks are significantly different from the cisplatin-treated samples (Dunnett test; # *p* < 0.05). Bars marked with \$ are significantly different from the corresponding treatment of the wild type cells (Dunnett test; \$ *p* < 0.05) (**C**). Apoptotic cell death was measured with a fluorogenic caspase-3 substrate in a kinetic assay on a Perkin Elmer Opera Phenix High Content Analyzer. Representative images (**D**) show the caspase activity signal (green) with digital phase contrast (red) at 22 h after cisplatin treatment. Images were taken every hour between 12 and 22 h after cisplatin treatment. Some samples were pretreated with the caspase-3 inhibitor z-DEVD-FMK as shown. The numbers of caspase positive cells are shown as a percentage of the actual number of cells. Data points marked with asterisks are significantly different from the control (12 h of cisplatin treatment) (Student–Newman–Keuls method; \* *p* < 0.05). Data points marked with hashmarks are significantly different from the cisplatin treated samples Student–Newman–Keuls method; # *p* < 0.05). (**E**). Means ± SEM are shown (*n* = 3).

Cisplatin is a DNA crosslinking agent that does not induce DNA breaks directly. γH2AX foci, indicating DNA double strand breaks (DSB), have been reported in cisplatin-treated cells and reflect activation of DNA repair mechanisms, such as nucleotide excision repair or non-homologs end joining [32]. High-content analysis detected three different cell morphologies: "normal" cells contained no P-H2AX signal, "spotted" cell type contained P-H2AX foci, while the "fragmented" type showed a condensed morphology, with a diffuse P-H2AX signal. We measured γH2AX and analyzed staining in combination with cell fragmentation, a marker of cell death that indicates failed DNA repair. In line with previous reports [32], cisplatin caused H2AX phosphorylation (γH2AX formation) in wild type cells. Significantly more cells were positive for γH2AX formation and/or showed signs of nuclear fragmentation in cisplatin-treated NMNAT1 deficient cells compared to their wild type counterparts (Figure 3B).

Cisplatin-induced cell death occurred via multiple pathways in the absence of NMNAT1. Loss of plasma membrane integrity, a sign of necroptosis, was observed, as indicated by increased lactate dehydrogenase (LDH) release in cisplatin-treated NMNAT-1−/<sup>−</sup> cells. Necrostatin-1 (NEC1) abolished LDH release, suggesting that it was indeed the consequence of necroptosis (Figure 3C). The involvement of the caspase-mediated apoptotic cell death pathway was also clearly demonstrated with a fluorogenic caspase-3/7 substrate and inhibition of the signal by z-DEVD-FMK (Figure 3D,E).

Metabolic reprogramming is a hallmark of cancer. Moreover, anticancer agents, including cisplatin, have been reported to cause metabolic perturbations [33,34]. Considering that NAD<sup>+</sup> is a central energy metabolite and NMNAT1 knockout resulted in a significant drop in cellular NAD+, we hypothesized that reduced NAD<sup>+</sup> availability may have an impact on cell metabolism in cisplatin-treated U-2OS cells. Basal NAD<sup>+</sup> levels were significantly lower in NMNAT1−/<sup>−</sup> cells and slightly but significantly decreased by cisplatin treatment (Figure 4A). A more dramatic change in ATP levels was observed. Cisplatin caused a marked drop in cellular ATP content in NMNAT1 deficient cells while no change could be seen in the wild type cells (Figure 4B). Cell death-associated impairment of energy production was partly responsible for the drop in ATP level, as both the caspase inhibitor z-DEVD-FMK and the necroptosis inhibitor NEC1 significantly prevented ATP loss (Figure 4B).

To characterize cellular energy-producing pathways, we measured oxygen consumption rate, (OCR) as a measure of cellular respiration, and extracellular acidification rate (ECAR), as a measure of glycolysis. No major difference could be observed in basal respiration between the two cell lines (Figure 4C). A mitochondrial stress test, conducted by sequential addition of different mitochondrial toxins (see Materials and Methods section for details), also showed similar responses in both cell lines. However, cisplatin treatment completely reduced mitochondrial respiratory reserve capacity in NMNAT1−/<sup>−</sup> cells and cells were unable to recover even basal respiratory activity after oligomycin treatment (Figure 4C). Wild type cells, on the other hand, displayed unaltered respiratory adaptation in response to cisplatin. The basal glycolytic activity proved to be significantly higher in the NMNAT1−/<sup>−</sup> cell line (Figure 4D). However, no significant differences between the two cell lines could be observed in glycolytic stress tests with or without the addition of cisplatin. This finding suggests that glycolytic activity remains mostly unaffected by cisplatin regardless of the NMNAT1 status. Cellular metabolism can be characterized by the OCR/ECAR ratio, based on the basal rates of the two parameters. Knockout cells have a reduced OCR/ECAR ratio indicating that they are more glycolytic, compared to the WT cells (Figure 4E). Furthermore, cisplatin treatment caused a more pronounced drop in the basal OCR/ECAR ratios in the NMNAT1−/<sup>−</sup> cells, demonstrating even more reliance on glycolysis rather than respiratory energy production.

**Figure 4.** Metabolic alterations in NMNAT-1 KO cell line. Following cisplatin (5 μg/mL) treatment (24 h), total NAD<sup>+</sup> levels were determined from cell lysates of WT and KO cells and normalized to protein content. Bars marked with asterisks are significantly different from the control (Bonferroni test; \*\*\* *p* < 0.001). Data points marked with hashmarks are significantly different from the corresponding treatment of the wild type cells (Bonferroni test; ## *p* < 0.01) (**A**). Cellular ATP was assayed 24 h after cisplatin treatment (**B**). Cells were treated with cisplatin alone or in combination with apoptosis (z-DEVD-FMK, 100 μM) or necroptosis (NEC1, 30 μM) inhibitors. Inhibitors were added 30 minutes before cisplatin treatment. Bars marked with asterisks are significantly different from the control (not shown) (Dunnett test; \*\* *p* < 0.01, \*\*\* *p* < 0.001). Bars marked with hashmarks are significantly different from the cisplatin-treated samples (Dunnett test; ## *p* < 0.01, ### *p* < 0.001). (**B**). Oxygen consumption rate (OCR)/oxidative phosphorylation (**C**) and extracellular acidification rate (ECAR)/glycolytic activity (**D**) were determined using a Seahorse metabolic analyzer and expressed as a percentage of the WT baseline. Both OCR/oxidative phosphorylation and ECAR/glycolytic activity were monitored in specific stress tests 13 h after cisplatin treatment. Mitochondrial stress tests included oligomycin (2 μM), FCCP (0.5 μM), and antimycin A (1 μM) for OCR (**C**) and glucose (10 mM), oligomycin (2 μM) and 2-deoxyglucose (50 mM) for ECAR (**D**). Data points marked with asterisks are significantly different from the corresponding vehicle treated data points at the same phase of the graph (Student–Newman–Keuls method; \* *p* < 0.05). Data points marked with hashmarks are significantly different from the corresponding data points of wild type cells at the same phase of the graph (Student-Newman-Keuls method; # *p* < 0.05). Ratios of the two metabolic routes were determined from the baseline OCR and ECAR values 13 hours after cisplatin treatment. Bars marked with asterisks are significantly different from the corresponding vehicle treated group (Student-Newman-Keuls method; \* *p* < 0.05). Bars marked with hashmarks are significantly different from the corresponding group of wild type cells (Student-Newman-Keuls method; # *p* < 0.05) (**E**). Data plotted are means ± SEM (*n* = 3).

#### *2.4. The Role of PARylation in the Increased Cisplatin Sensitivity of NMNAT1*−/<sup>−</sup> *K.O. Cells*

Cisplatin-induced DNA damage activates PARP1 [35] and PARP1 activation contributes to the repair of cisplatin-crosslinked DNA lesions and cell survival. Thus, we hypothesized that NMNAT1 deficiency and consequent nuclear NAD<sup>+</sup> scarcity may compromise PARP1's ability to initiate the DNA damage response and prevent cell death. We found that cisplatin caused a time dependent poly(ADP-ribose) (PAR) formation in WT cells, while no PAR formation could be detected in the NMNAT1−/<sup>−</sup> cells (Figure 5A,B). Moreover, cisplatin treatment caused only a mild and non-significant decrease in the clonogenic activity of wild type cells, but significantly reduced the clonogenic activity of NMNAT1−/<sup>−</sup> cells. Combined treatment with cisplatin and the PARP inhibitor (olaparib) caused a significantly lower proliferation in wild type cells, compared to the cisplatin-treated cells, but no further decrease could be detected in the NMNAT1−/<sup>−</sup> cells (Figure 5C). These data suggest that cisplatin-induced DNA damage is unable to activate PARP1 in NMNAT1−/<sup>−</sup> cells and substrate deprivation of PARP1 limits the capacity of the cells to cope efficiently with DNA injury.

**Figure 5.** Impaired PARylation contributes to increased cisplatin sensitivity of NMNAT1−/<sup>−</sup> cells. The level of poly(ADP-ribose) (PAR) polymer formation was detected using western blot in wild type (WT) and in NMNAT1 <sup>−</sup>/<sup>−</sup> (KO) cells at the indicated timepoints (0.5, 2, 6, 10, and 15 h) after cisplatin (5 μg/mL) treatment (**A**). Full WB image can be found in Supplementary Material. Actin was used as a loading control. Relative densities (normalized to beta actin and compared to the untreated WT control) are shown on panel (**B**). Bars marked with hashmarks are significantly different from the corresponding samples of wild type cells (Bonferroni test; # *p* < 0.05). The role of possible of PARP-1-NMNAT1 interaction in clonogenic activity was determined by counting crystal violet-stained colonies at day 6 (**C**). Cells were treated with vehicle (CTL), cisplatin, or the combination of cisplatin (5 μg/mL) and olaparib (10 μM) (Cisp+ OLA) and individual colonies were counted in each sample. Results are expressed as a percentage of the number of colonies in the vehicle treated samples. Bars marked with asterisks are significantly different from the corresponding control (Bonferroni test; \*\*\* *p* < 0.001). Bars marked with hashmarks are significantly different from the corresponding cisplatin treated samples (Bonferroni test; # *p* < 0.05). Bars marked with + are significantly different from the corresponding samples of wild type cells (Bonferroni test; +++ *p* < 0.001). Data plotted are means ± SEM (*n* = 3).

#### *2.5. Chemosensitization in 3D Cultures and in Combination Treatment*

To demonstrate the role of NMNAT1 in cellular models that more closely resemble in vivo conditions or clinical settings, we investigated a) chemosensitization to cisplatin by NMNAT1 deficiency in a 3D model (as opposed to 2D cultures) and b) a treatment protocol based on a combination of chemotherapeutic drugs. Wild type and NMNAT1−/<sup>−</sup> U-2OS cells formed spheroids under favorable conditions (see "Materials and Methods" section for detail). Cell-to-cell contacts in spheroids render cells resistant to toxic stimuli [36]; therefore, we used higher concentrations of cisplatin in the spheroid experiments. While spheroids of the wild type cells only became less compact at 50 μg/mL cisplatin concentration, NMNAT1−/<sup>−</sup> U-2OS cells displayed a significant decrease in size, or a complete disintegration of the spheroids under the same conditions (Figure 6A,B). The inner region of cisplatin-treated spheroids show a marked elevation in Annexin V positivity, a more dramatic elevation can be detected in the case of NMNAT1−/<sup>−</sup> spheroids (Figure 6C).

**Figure 6.** Chemosensitivity of spheroids to cisplatin treatment. Using WT and NMNAT-1−/<sup>−</sup> U-2OS cells, 3D cell cultures (spheroids) were generated. Spheroids were treated with cisplatin (50 μg/mL) for 6 days. Cells were stained without fixation with Hoechst (red pseudocolor) and Annexin V (green). Representative images are shown in panel **A**. Images of 3 spheroids/treatment were taken and analyzed for area (**B**) and the intensity of the inner region (**C**). Results are expressed as percentages of the control (vehicle treated) samples. Bars marked with asterisks are significantly different from the vehicle treated control (Student-Newman-Keuls method; \* *p* < 0.05). Bars marked with hashmarks are significantly different from the cisplatin treated wild type spheroids (Student-Newman-Keuls method; # *p* < 0.05). Data plotted are means ± SEM (*n* = 3).

Moreover, a combination of cisplatin with doxorubicin resulted in more efficient killing of cancer cells (Figure 7). A concentration series of doxorubicin was used to explore differences in the sensitivity of wild type and NMNAT1−/<sup>−</sup> U-2OS cells. At concentrations higher than 750 ng/mL, knockout cells were significantly more sensitive to doxorubicin (Figure 7A). Cells were exposed to different concentrations of cisplatin and a fixed doxorubicin concentration of 150 ng/ml. Combination treatment was significantly more efficient in killing NMNAT1−/<sup>−</sup> cells than wild type cells (Figure 7B).

**Figure 7.** Combined treatments on U-2OS WT and NMNAT1−/<sup>−</sup> cells, chemosensitivity of SAOS-2. Wild type (WT) and NMNAT1−/<sup>−</sup> cells were treated with the indicated concentrations of doxorubicin (DOX) and viability was determined with the Calcein-AM assay 24 hours after DOX treatment. Data points marked with asterisks are significantly different from the corresponding vehicle treated data points (not shown) (Dunnett test; \* *p* < 0.05). Data points marked with hashmarks are significantly different from the corresponding data points of wild type cells (Student–Newman–Keuls method; # *p* < 0.05) (**A**). DOX was also applied in combination with cisplatin (**B**). Panel B shows the viability of WT and KO at constant (150 ng/mL) DOX and increasing cisplatin concentrations. Data points marked with asterisks are significantly different from the corresponding vehicle treated data points (Dunnett test; \* *p* < 0.05). Data points marked with hashmarks are significantly different from the corresponding data points of wild type cells (Student–Newman–Keuls method; # *p* < 0.05). SAOS-2 cells were transfected with control or NMNAT-1 specific siRNAs. Efficiency of NMNAT-1 silencing was verified by Western blot (**C**). Full WB image can be found in Supplementary Material. Cells were treated with cisplatin for 48 hours and a Calcein-AM viability assay was performed (**D**). Data points marked with asterisks are significantly different from the vehicle-treated sample (Dunnett test; \* *p* < 0.05). Data plotted are means ± SEM (*n* = 3).

To confirm the sensitizing effect of the absence of NMNAT1 in cisplatin-induced cell death, cells of another osteosarcoma cell line, SAOS-2, were used. NMNAT1 could be effectively silenced in SAOS-2 cells (Figure 7C). Similarly to NMNAT-1 knockout U-2OS cells, NMNAT-1 silenced SAOS-2 cells also showed significantly higher sensitivity to cisplatin treatment (Figure 7D).

#### **3. Discussion**

Similar to most neoplastic diseases, osteosarcoma poses a great challenge to modern medicine. The therapeutic approaches are limited and are based on the combination of chemotherapy and surgical removal of the primary tumor and metastasis. Two of the most efficient chemotherapeutics used for the therapy of osteosarcoma (cisplatin and doxorubicin) cause DNA injury. Methotrexate, the third component of the most common chemococktails, is an anti-metabolite folate analog that interferes with the synthesis of purine and pyrimidine bases and, therefore, DNA and RNA synthesis. Of note, the anticancer effect of methotrexate has also been reported to be due to DNA breakage [37]. Thus, it seems that the DNA damage response (DDR) might be the Achilles heel of osteosarcomas. Targeting the DNA damage response is a novel anticancer approach in cancers displaying DNA repair defects. The best known example is PARPi treatment of BRCA-negative ovarian cancers [38]. Nuclear NAD<sup>+</sup> synthesis may be an unlikely target of DNA repair in cancer therapy. However, as the substrate of PARP enzymes, NAD<sup>+</sup> may be a limiting factor for PARylation activity. Moreover, the nuclear SIRT enzymes, another major NAD<sup>+</sup> consumer group, are also known to promote cell survival and are potential cancer targets (see more details below).

Previous experiments with NAMPT inhibitors suggested that inhibition of the salvage pathway of NAD<sup>+</sup> synthesis may have beneficial effects for cancer therapy [39]. Due to the lack of specific NMNAT1 inhibitors, however, the role of NMNAT1 in cancer has not yet been investigated. In the present study, we have investigated the role of NMNAT1 as a potential target in osteosarcoma. All cancer cell lines that were tested expressed NMNAT1 mRNA. Compared to the other assayed cell lines, U-2OS osteosarcoma cells displayed an average level of NMNAT1 mRNA. Our mRNA data also show that in U-2OS cells, the NMNAT-1 has the highest expression, while NMNAT2 and NMNAT-3 have lower contributions. These cells tolerated inactivation of the NMNAT1 gene without showing decreased viability or marked morphological alterations. Notable phenotypic differences between NMNAT1−/<sup>−</sup> and wild type cells included reduced proliferation capacity (clonogenicity) and moderately increased glycolytic activity. The most prominent feature of the NMNAT1−/<sup>−</sup> cells was a markedly reduced NAD<sup>+</sup> level, which did not compromise cellular energy production in the resting state, as reflected by the unaltered ATP content. The two-thirds reduction in cellular NAD<sup>+</sup> in NMNAT1−/<sup>−</sup> cells highlights the dominant role of NMNAT1 in cellular NAD<sup>+</sup> synthesis, as elevated NMNAT-2 expression cannot substitute the role of NMNAT1 in NAD+ production of NMNAT1 KO cells. A cross-compartment exchange of NAD<sup>+</sup> between nuclear and cytoplasmic pools is possible, as demonstrated in a model of adipogenic signaling where NMNAT2 activation drained nuclear NAD<sup>+</sup> while competing for the common substrate NMN [40]. Similar nuclear and cytoplasmic NAD<sup>+</sup> concentrations also suggest relatively barrier-free equilibration of NAD<sup>+</sup> between these compartments [41]. However, deciphering mechanisms underlying the marked drop of total NAD<sup>+</sup> in NMNAT1−/<sup>−</sup> cells and characterization of the effect of NMNAT1 inactivation on the main NAD<sup>+</sup> compartments require further investigation.

A key question asked in our study was whether NMNAT1 contributes to the cells' ability to cope with DNA damage caused by chemotherapeutic agents used in osteosarcoma therapy. Induction of NMNAT1 by cisplatin and doxorubicin suggests that NAD+-dependent processes may be involved in the cellular survival response to these chemotherapeutics. Moreover, a major finding of our paper is that NMNAT1 targeting is a potent chemosensitizing strategy. NMNAT1−/<sup>−</sup> U-2OS cells displayed increased sensitivity to cisplatin, as indicated by reduced viability and increased cell death compared to wild type U-2OS cells. Cisplatin-induced cell death of NMNAT1 knockout cells showed features of both apoptosis and necroptosis. Activation of caspase3/7-like proteases and the cytoprotective effect of caspase-3/7 inhibition suggest that caspase-mediated apoptosis is involved in cell death in our model. Furthermore, the protective effect of the necroptosis inhibitor Nec1 indicates that the necroptotic pathway is also activated and contributes to cell death. If cisplatin-induced killing of NMNAT1 deficient cells also has a necroptosis-like component in vivo, NMNAT1 inhibition may enhance the anticancer immune response or may promote metastasis [42]. A detailed characterization of cell death modalities involved in the killing of osteosarcoma cells with reduced nuclear NAD<sup>+</sup> production and the effects on antitumor immunity versus metastasis formation goes beyond the scope of the current paper.

An important question raised by our study is whether the anticancer potential of NMNAT1 targeting is restricted to osteosarcoma. It is quite likely that targeting NMNAT1 may also be beneficial in other types of tumors. In support of this statement, we found that NMNAT1 is likely to be important for the progression of various tumors. RNAsec data from the kmplot database [43] show that low NMNAT1 expression correlates with better survival of patients with sarcomas, liver hepatocellular carcinoma, bladder carcinoma, breast cancer, esophageal adenocarcinoma, kidney renal papillary cell carcinoma, pancreatic ductal adenocarcinoma and uterine corpus endometrial carcinoma (Figure S3). Whether NMNAT1 inhibition provides therapeutic benefits in these and other tumors requires further investigation.

Our data strongly support the role of PARP1 in the chemosensitization process. We hypothesize that nuclear NAD<sup>+</sup> levels may be too low to support PARP activity in NMNAT1−/<sup>−</sup> cells. In fact, cisplatin stimulated PARylation in wild type cells. However, PAR formation was absent in NMNAT1−/<sup>−</sup> cells. Moreover, the potent PARP inhibitor olaparib enhanced the anti-clonogenic effect of cisplatin in the wild type but not the NMNAT1-deficient cells. The role of PARP-1 in the anti-clonogenic effect is also supported by our unpublished data from experiments with PARP-1 and PARP-2 silenced osteosarcoma cells. These results suggest that PARP1 activation acts as a survival factor via assisting DNA repair, in line with our current understanding of the role of PARylation in the repair of cisplatin-induced DNA damage [44]. PARP1 is unable to contribute to cell survival in NMNAT1−/<sup>−</sup> cells (due to the shortage of NAD+). Thus, a key factor in the chemosensitizing effect of NMNAT1 inactivation is reduced PARylation and, consequently, impaired DNA repair. It is tempting to hypothesize that NMNAT1 and PARP1 interact directly at sites of DNA damage, similar to the direct interaction between NMNAT1 and PARP1 on PARP1-dependent promoters [45]. This interaction might involve the binding of NMNAT1 to PAR polymers [46]. This hypothesis, however, needs experimental verification. The possible role of PARP1 in the chemosensitization of NMNAT1−/<sup>−</sup> osteosarcoma cells was also supported by previous preclinical studies. These papers reported (a) correlation between PARP1 expression level and osteosarcoma cell survival; (b) osteosarcoma cell killing or chemosensitization by PARP1 knockout or PARP inhibition [47–49]. Ongoing and finished clinical trials (e.g., ClinicalTrials.gov ID: NCT01583543, NCT01858168 NCT02044120) also set out to investigate the effects of olaparib and other PARP inhibitors in osteosarcoma. Our data suggest that, at least under our current experimental settings, much of the chemosensitization observed in NMNAT1 cells is due to indirect inhibition of PARylation. It remains to be seen, however, how potent NMNAT1 inhibitors (which are not yet available) will compare with clinically used PARPi compounds in terms of therapeutic spectrum, efficiency, tolerability and side effects in different types of tumors. We hypothesize that PARPi and NMNAT1i will necessarily differ in one or more of these aspects.

The hypothesis that impaired metabolic adaptation may also play a role in the chemosensitivity of NMNAT1-deficient cells is plausible, considering the central role of NAD<sup>+</sup> in energy metabolism. Under basal conditions, a slight shift from respiration towards glycolysis could be detected in NMNAT1−/<sup>−</sup> cells but energy production was balanced, as indicated by the normal ATP levels. DNA damage revealed the vulnerability of the respiratory system, as the mitochondrial stress test identified severely impaired respiratory reserve capacity. Multiple lines of evidence highlight the importance of metabolic regulation in tumor behavior and chemosensitivity. For example, analysis of integrated transcriptomic and metabolomics datasets identified altered glycolysis-related mRNAs and metabolites in osteosarcoma samples [50]. Moreover, various microRNAs regulate both osteosarcoma growth (proliferation) and glycolysis either positively (e.g., miRNA543) [51] or negatively (e.g., miRNA186) [52]. Furthermore, glycolysis also appears to correlate with chemosensitivity (e.g., cisplatin sensitivity [53]) or viability of osteosarcoma cells [54,55].

The extent to which SIRT inhibition contributes to chemosensitization under NAD<sup>+</sup> scarcity is worth investigating. The nucleus is home to several sirtuin (SIRT) enzymes (SIRT1,6,7). SIRT1 plays a role in the DNA damage response [56]. Furthermore, the central role of sirtuins in energy homeostasis [57] and metabolic regulation [58] support the possible involvement of sirtuins in shifting metabolic balance (e.g., respiration vs glycolysis) after DNA damage. Nuclear SIRTs facilitate cell survival under various stress conditions, underlining their potential role in chemosensitization. For example, SIRT6 is downregulated in DOX-treated liver cancer cells, contributing to DOX-induced cell death. The pro-survival role of SIRT6 in this model, as demonstrated by the overexpression of the enzyme, is mediated by Foxo3 degradation [59]. Inhibition of SIRT6 also sensitizes cancer cells to chemotherapy [60]. Similarly, SIRT7 has been identified as a survival factor via complex mechanisms including direct p53 diacetylation [61] and the promotion of autophagy [62]. In a transcription regulatory setting, a requirement for NMNAT1 and NAMPT for SIRT1 activity has already been demonstrated [63] and similar interplay between nuclear SIRTs and enzymes of the salvage pathway of NAD<sup>+</sup> synthesis is likely to take place in the DNA damage response as well. Nevertheless, clarifying the role of SIRTs in the chemosensitizing effect of NMNAT1 inactivation goes beyond the scope of this paper.

In summary, NMNAT1 has been identified as a possible target for the treatment of osteosarcoma. DNA damaging chemotherapeutic agents are likely to cause synergistic toxicity with NMNAT1 inhibitors, if NMNAT1 inhibitors are developed. Key mechanisms underlying this synergy include the impairment of PARylation-dependent DNA repair processes and metabolic adaptation. The chemosensitization by NMNAT1 inactivation could also be observed under more clinically relevant conditions, i.e., in 3D spheroids and combination chemotherapy. However, the effect of NMNAT1 targeting in non-transformed cells is a critical issue for the development of anticancer strategies targeting NMNAT1.

#### **4. Materials and Methods**

#### *4.1. D Cell Culture*

Human U-2 OS cells were grown in Dulbecco's modified Eagle's medium (DMEM, #12-604F, Lonza, Basel, Switzerland), containing 10% fetal bovine serum (#10500-064, GIBCO, ThermoFisher Waltham, MA, USA) L-glutamine, and penicillin-streptomycin, under standard cell culture conditions (humidified atmosphere, 5% CO2). The cells were routinely tested for mycoplasma contamination.

#### *4.2. D Cell Culture (Spheroids)*

Spheroids were grown as previously described [64] with modifications as follows. Flat-bottom cell culture microplates were coated with low melting point agarose to form a U-shaped, cell-repellent bottom. Cells were seeded into the wells and grown for 2 days to form spheroids.

#### *4.3. Generation of NMNAT1*−/<sup>−</sup> *Cells by Crispr-cas9*

NMNAT1 knock-out U-2OS cells were made by CRISPR-Cas9 technology using reagents from Santa Cruz Biotechnology, following the manufacturer's instructions. Cells were grown to 40–80% confluence and then trypsinized in antibiotic-free standard growth medium (DMEM, #12-604F, Lonza, Basel, Switzerland). Cells were counted, cell number was adjusted to 5.33 <sup>×</sup> 105 cells/mL, and a 2 mL cell suspension was centrifuged (100<sup>×</sup> *<sup>g</sup>*, 10 min). UltraCruz® Transfection Reagent (18 <sup>μ</sup>L; sc-395739, Santa Cruz Biotechnology, Dallas, TX, USA) was diluted with 82 μL of plasmid transfection medium (sc-108062, Santa Cruz Biotechnology, Dallas, TX, USA) to bring the final volume to 100 μL. The medium was carefully aspirated and the diluted transfection reagent was pipetted onto the cells. Next, 2.5 μL of NMNAT1 CRISPR Plasmid and 2.5 μL of HDR Plasmid were added to the tube. Cells were immediately transferred to transfection cuvettes and transfected using an Amaxa Nucleofector II (Lonza, Basel, Switzerland). Transfected cells were pipetted into 6-well plates, which contained pre-warmed cell culture media (DMEM, #12-604F, Lonza, Basel, Switzerland) supplemented with 10% fetal bovine serum (#10500-064, GIBCO, ThermoFisher Waltham, MA, USA), 5% L-glutamine, and 5% penicillin-streptomycin. Cells were incubated for 24–72 hours under regular conditions and transfection efficiency was visually confirmed by the detection of red fluorescent protein (RFP) via fluorescent microscopy. Transfected cells were selected for puromycin (2.5 μg/mL) resistance for 3 weeks and sorted for RFP fluorescence with a BD LSR II Cell Sorter (Franklin Lakes, NJ, USA. Single-cell colonies were grown and used for the experiments. The NMNAT1 expression of the clones was checked with qPCR for mRNA-level and Western blot for protein level (see below).

#### *4.4. Western Blot*

Western blotting was carried out as previously described [65–67], with modifications as follows. U-2OS cells were sonicated in RIPA lysis buffer supplemented with protease inhibitor cocktail (#M221, VWR International, Radnor, PA, USA) and phosphatase inhibitor (PMSF, 1:100; # PMSF-RO, Merck, Darmstadt, Germany). Lysates were centrifuged at 16,100× *g* for 10 min at 4 ◦C. Protein concentrations of the supernatants were measured with a Direct Detect infrared spectrometer (EMD Millipore Corporation, Burlington, MA, USA). Equal amounts of the proteins from each extract were separated on SDS-polyacrylamide gels (10%) in SDS-Tris-glycine running buffer (10× glycine, Tris, SDS). The proteins were then transferred electrophoretically to nitrocellulose membranes in transfer buffer (5× Tris, glycine, in dH2O). Membranes were blocked for 1 hour at room temperature with 5% skim milk-powder (#70166, Merck, Darmstadt, Germany) dissolved in 0.01% PBS-Tween20. Primary antibodies were applied in blocking solution overnight at 4 ◦C. Horseradish peroxidase (HRP)-coupled secondary antibodies were applied in the same type of solution as primary antibodies for 2 hours at room temperature. ECL-based chemiluminescence (Super Signal, West Pico Plus, Luminol/Enhancer #1863098, Peroxidase Solution #1863099, ThermoFisher Waltham, MA, USA) was used for detection. Images were acquired with a Bio-Rad ChemiDoc Imager (Bio-RAD, Hercules, CA, USA) and ImageLab 6.0 software was used for protein quantification. The antibodies used for western blotting are shown in Table S1.

#### *4.5. High Content Analysis (HCA)*

#### 4.5.1. HCA on Fixed Cells: Quantification of γH2AX

Cells (2 <sup>×</sup> 104, 100 <sup>μ</sup>L/well) were seeded into sterile microplates (Cell Carrier-96 ultra, PerkinElmer, Waltham, MA, USA) and grown for 24 h. After applying the indicated cell treatments (see figure legends), cells were fixed in 3% formaldehyde/PBS solution for 15 min at room temperature, washed 3× with PBS, and incubated in blocking solution (5% BSA in PBS) for 15 min at room temperature. The anti-phospho-H2AX antibody (Table S2) was diluted in blocking solution and added to wells (50 μL/well, 2 h at room temperature). The secondary antibody (Alexa Fluor 488, Table S2) was diluted in blocking solution and incubated for 1 hour. After antibody incubations, cells were washed twice with PBS and incubated for 5 min with PBS containing 4',6-diamidino-2-phenylindole dihydrochloride (DAPI, Table S2) at room temperature for nuclear staining. Cells were then washed three times with PBS and were kept in 100 μL of PBS until imaging. Images were acquired using an Opera Phenix High Content Analyzer (PerkinElmer, Waltham, MA, USA) with a 10× air objective (NA 0.3). Image analysis was performed with the built in Harmony software (version 4.8).

#### 4.5.2. HCA on Live Cells: Caspase Activation

Cells (2 <sup>×</sup> 10<sup>4</sup> cells, 100 <sup>μ</sup>L/well) were seeded into 96-well plates (Perkin Elmer, Waltham, MA, USA) and grown for 24 h. The following day, cells were pretreated with 100 μM z-DEVD-FMK caspase inhibitor (#S7312, Selleckchem, Houston, TX, USA) and/or treated with 5 μM cisplatin (Accord, Warsaw, Poland). Then, the cells were stained with 50 μL CellEvent™ Caspase-3/7 Green Detection Reagent (Table S2) in 7 μM final concentration. The plates were placed into an Opera Phenix High-Content Analyzer (Perkin Elmer, Waltham, MA, USA) with environment control (5% CO2, 37 ◦C) and fluorescence was measured every hour for 11 h at 502/530 nm. The analysis was performed with the Harmony software (Perkin Elmer, Waltham, MA, USA).

#### 4.5.3. HCA on Live Cells: Cell Death in Spheroids

Spheroids were transferred to glass bottom microplates (Cell Carrier-96 ultra, PerkinElmer, Waltham, MA, USA) and treated with cisplatin (25, 50, and 100 μg/mL) on day 2 and day 4. Afterward, the cells were stained with Hoechst (Table S2) and Annexin V-FITC (Table S2.) for 1 hour in growth medium. Images were acquired using an Opera Phenix High Content Analyzer (Perkin Elmer, Waltham, MA, USA). Fluorescence intensity was detected at 350 (Hoechst) and 488 nm (Annexin V-FITC).

#### 4.5.4. HCA on Live Cells: Cell Morphology

Cells (2 <sup>×</sup> 10<sup>4</sup> cells, 100 <sup>μ</sup>L/well) were seeded into 96-well plates (Perkin Elmer, Waltham, MA, USA) and grown for 24 h. The following day, cells were stained with 50 μL DRAQ5 (Table S2.) to obtain a final concentration of 2.5 μM. Plates were placed into an Opera Phenix High-Content Analyzer with environment control (5% CO2, 37 ◦C), fluorescence was measured at 488 to 647 nm, and the analysis was performed with the Harmony software (version 4.8).

#### *4.6. Calcein-AM Viability Assay*

This assay was carried out as previously described [68], with modifications as follows. Cells (2 <sup>×</sup> 104 cells, 100 <sup>μ</sup>L/well) were seeded into 96-well plates and were grown for 24 h. The next day, cells were treated with the indicated treatments (see figure legends). Two different batches of cisplatin were used for the experiments. The first batch was used for the experiments presented on Figure 1 (Figure 1B–D), whereas a second batch was used for the experiments presented in Figure 3, Figure 4, Figure 5, Figure 6, and Figure 7. Then the cells were stained by adding 50 μL of Calcein-AM (#17783, Merck, Darmstadt, Germany) solution in a final concentration of 1 μM. After incubating cells for 1 h at 37 ◦C, the fluorescent signal was measured with a Tecan Spark 20M (Tecan, Männedorf, Switzerland) multimode reader (Ex/Em = 485/530 nm with). Viability was expressed as a percentage of the untreated control.

#### *4.7. Clonogenic Survival Assay*

U-2OS cells were seeded into 6-well plates (#92006, TPP, Trasadingen, Switzerland) at a density of <sup>1</sup> <sup>×</sup> <sup>10</sup><sup>3</sup> cells/mL. After 24 h the cells were treated with 10 <sup>μ</sup>M PJ34 (P4365, Merck, Darmstadt, Germany) or 10 μM Olaparib (S1060, Selleckchem, Houston, TX, USA) for 30 minutes, and then with 5 μg/mL cisplatin (Accord Healthcare Inc., Durham, UK). Cells were incubated for 6 days and then counted manually after staining with 0.5% crystal violet dissolved in 20% ethanol [69].

#### *4.8. Lactate Dehydrogenase (LDH) Release Assay*

Cell death was assessed by determining the activity of LDH released into the culture medium. The LDH assay kit (#786-210, GBiosiences, St. Louis, MO, USA) was used as described by the manufacturer.

#### *4.9. Measurement of Total Cellular NAD*<sup>+</sup> *and ATP*

NAD<sup>+</sup> content was measured with an NAD<sup>+</sup> assay kit following the manufacturer's instructions (NAD+/NADH Quantitation Kit. #MAK037-1KT, Merck, Darmstadt, Germany). The NAD<sup>+</sup> content of the cells was measured with a plate reader (Tecan Spark 20M, Tecan, Männedorf, Switzerland) at 450 nm.

ATP content was determined using an ATP assay kit following the manufacturer's instructions (#110M6101, Merck, Darmstadt, Germany). Luminescence was measured with a plate reader (Tecan Spark 20M, Tecan, Männedorf, Switzerland). NAD<sup>+</sup> content and ATP content were normalized to protein content.

#### *4.10. Metabolic Analysis with Seahorse Metabolic Analyzer*

U-2OS cells were seeded (1 <sup>×</sup> 10<sup>5</sup> cells/well) in DMEM in Seahorse XF96 cell culture microplates (Agilent Technologies, Inc., Santa Clara, CA, USA) and incubated overnight at 5% CO2 and 37 ◦C. The sensor cartridge was prepared by adding 200 μL of dH2O overnight. Then, Seahorse Bioscience XF96 calibrant solution (pH 7.4) (Part No.: 100-840-000) was added to each well of a Seahorse 96-well utility plate. The sensors with the calibrant solution were incubated for 2 hours at 37 ◦C without CO2. The measurement was performed using the Seahorse XF96 Analyzer (Agilent Technologies, Inc., Santa Clara, CA, USA).

XF Cell Mito Stress analyses were performed as described in [70], with modifications as follows. The mitochondrial inhibitors were applied at the following final concentrations: 2 μM oligomycin, 0.5 μM FCCP, and 1 μM antimycin-A. Glycolysis stress was based on [70].

#### *4.11. Sulforhodamine B (SRB) Assay*

The SRB assay was performed as described in Kovács, P. et al. [71].

#### *4.12. RNA Extraction and Quantitative PCR*

RNA was purified with TRIzol reagent (Tri-RNA reagent, #FATRR001, Amplicon, Odense, Denmark) as described in the manufacturer's instructions. Quantitative real-time PCR was performed with a LightCycler 480 thermocycler (Roche, Basel, Switzerland) using SYBR Green (SyberGreen, #4472908, Applied Biosystems, Foster City, CA, USA) according to the manufacturer's protocol. Reactions were carried out in triplicate and data were normalized to the geometric mean of housekeeping genes (36B4 and cyclophilin). Sequences of primers are given in Table S3. hNMNAT1, hNMNAT2, hNMNAT3, hATP5B, hHSP60, hPKM2 primers were ordered from IDT (Coralville, IA, USA); and h36B4 and hcyclophilin from Merck (St. Louis, MO, USA).

#### *4.13. Statistical Analysis*

Experiments were repeated at least three times and data are expressed as mean ± SEM. Based on the type of experiment and the distribution of data, different kinds of statistical analysis were used. The statistical tests used are indicated in the figure legends.

#### **5. Conclusions**

Genetic inactivation of NMNAT1 sensitizes U-2OS osteosarcoma cells to cisplatin, doxorubicin, or a combination of these two treatments. Upon cisplatin treatment, an impaired PARP1 activity, insufficient DNA repair, a marked drop in cellular ATP and a limited mitochondrial reserve capacity could be observed, which all contribute to the higher chemosensitivity of NMNAT-1−/<sup>−</sup> cells. Increased cell death of NMNAT1−/<sup>−</sup> cells shows features of both apoptosis and necroptosis. NMNAT1−/<sup>−</sup> cells also displayed markedly higher sensitivity to cisplatin when grown as spheroids in 3D culture. Our results suggest that NMNAT1 may be worth investigating further as a potential target in cancer therapy.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/5/1180/s1, Figure S1: mRNA expressions of NMNAT enzymes and induction of NMNAT1 protein expression in U-2OS cells, Figure S2: Cell morphology analysis, Figure S3: Tumors with better patient survival at low NMNAT1 expression, Table S1: Antibodies used for Western blotting, Table S2: Antibodies and dyes used for High Content Analysis and Table S3: Sequences of primers, used in quantitative PCR experiments. Supplementary materials also contain the full images of Western Blots.

**Author Contributions:** Conceptualization, L.V. and C.H.; Data curation, C.H.; Formal analysis, A.K., Z.R. and C.H.; Funding acquisition, L.V. and C.H.; Investigation, A.K., A.P.R., Z.R., S.T., I.S., T.E.-H. and C.H.; Methodology, Z.P., S.T. and C.H.; Supervision, L.V. and C.H.; Writing—original draft, L.V. and C.H.; Writing—review and editing, L.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** L.V. received funding from the National Research, Development and Innovation Office grants GINOP-2.3.2-15-2016-00020 TUMORDNS", GINOP-2.3.2-15-2016-00048-STAYALIVE, OTKA K132193. CH received funding from the National Research, Development and Innovation Office grant PD 116845; Bolyai postdoctoral fellowship (BO/00468/17/8); and was supported by the ÚNKP-19-4-DE-299 New National Excellence Program of the Ministry of Human Capacities.

**Acknowledgments:** The authors thank Karen Uray for careful English language editing; Erzsébet Herbály and Erika Gulyás for their excellent technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest. The funding agencies had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **ATR Inhibition Potentiates PARP Inhibitor Cytotoxicity in High Risk Neuroblastoma Cell Lines by Multiple Mechanisms**

#### **Harriet E. D. Southgate 1, Lindi Chen 1, Deborah A. Tweddle 1,\* and Nicola J. Curtin 2,\***


Received: 30 March 2020; Accepted: 23 April 2020; Published: 28 April 2020

**Abstract:** *Background:* High risk neuroblastoma (HR-NB) is one the most difficult childhood cancers to cure. These tumours frequently present with DNA damage response (DDR) defects including loss or mutation of key DDR genes, oncogene-induced replication stress (RS) and cell cycle checkpoint dysfunction. Aim: To identify biomarkers of sensitivity to inhibition of Ataxia telangiectasia and Rad3 related (ATR), a DNA damage sensor, and poly (ADP-ribose) polymerase (PARP), which is required for single strand break repair. We also hypothesise that combining ATR and PARP inhibition is synergistic. *Methods:* Single agent sensitivity to VE-821 (ATR inhibitor) and olaparib (PARP inhibitor), and the combination, was determined using cell proliferation and clonogenic assays, in HR-NB cell lines. Basal expression of DDR proteins, including ataxia telangiectasia mutated (ATM) and ATR, was assessed using Western blotting. CHK1S345 and H2AXS129 phosphorylation was assessed using Western blotting to determine ATR activity and RS, respectively. RS and homologous recombination repair (HRR) activity was also measured by γH2AX and Rad51 foci formation using immunofluorescence. *Results: MYCN* amplification and/or low ATM protein expression were associated with sensitivity to VE-821 (*p* < 0.05). VE-821 was synergistic with olaparib (CI value 0.04–0.89) independent of *MYCN* or ATM status. Olaparib increased H2AXS129 phosphorylation which was further increased by VE-821. Olaparib-induced Rad51 foci formation was reduced by VE-821 suggesting inhibition of HRR. *Conclusion:* RS associated with *MYCN* amplification, ATR loss or PARP inhibition increases sensitivity to the ATR inhibitor VE-821. These findings suggest a potential therapeutic strategy for the treatment of HR-NB.

**Keywords:** poly (ADP-ribose) polymerase inhibitors (PARPi); ataxia telangiectasia and Rad3 related inhibitors (ATRi); replication stress; cell cycle checkpoints

#### **1. Introduction**

Inhibitors of ataxia telangiectasia and Rad3 related (ATR) kinase are currently being tested in early phase clinical trials for patients with adult cancers. ATR is a DNA damage sensor kinase which has a pivotal role in recovery from replication stress (RS). Activated ATR signals to many pathways involved in regulation of origin firing, and stabilisation and restart of stalled replication forks (reviewed in: [1,2]). ATR signals to S and G2/M cell cycle checkpoint control by inhibiting CDK2 and CDK1 via its primary target CHK1 [3,4]. ATR inhibition abrogates S and G2/M checkpoint arrest, leading to cell death by mitotic catastrophe [5,6].

Two ATR inhibitors, AZD6738 (Astra Zeneca) and M6620 (formerly VX-970, Merck), have entered phase II trials, with a third, BAY1895344 (Bayer) in phase I (listed on https://clinicaltrials.gov/). Various determinants of ATR inhibitor sensitivity have been proposed, including RS, loss of G1 checkpoint control and loss of the ataxia telangiectasia mutated (ATM) protein [7,8].

RS is a common feature of cancer cells and leads to genetic instability. There are many causes of RS including overexpression of proliferation-driving oncogenes (*Ras, Myc, cyclin E* etc.), deregulation of replication origin firing, limited nucleotide pools or essential replication factors and replication through fragile sites or damaged DNA regions [9,10].

Loss of G1 checkpoint control also contributes to RS and is common in cancer through loss of tumour suppressors such as p53, pRB and ATM, imbalance of cyclins, cyclin-dependent kinases and their inhibitors and expression of oncogenes [11]. G1 checkpoint deficiency results in a reliance on the S and G2/M checkpoints to maintain genome integrity and prevent replication of damaged DNA/mitotic catastrophe [12–14].

Neuroblastoma (NB) is a rare embryonal tumour derived from cells of the developing sympathetic nervous system. Around 100 cases are diagnosed a year in the UK, of which 50% are classified as high risk, but accounts for ~10% of paediatric cancer deaths [15,16]. Long term survival of high-risk neuroblastoma (HR-NB) (metastatic disease over 1 year of age- or *MYCN*-amplified disease) currently remains less than 50% at 5 years despite intensive high-dose multimodal treatment [17,18]. Survival of relapsed NB is particularly poor with less than 10% 5-year survival [19]. HR-NB frequently present with DNA damage response (DDR) defects including loss or mutation of key DDR genes, oncogene-induced RS and cell cycle checkpoint dysfunction, which suggest they would be sensitive to ATR inhibition [20–23].

Fifty percent of HR-NB have amplification of the *MYCN* oncogene, leading to RS. *MYCN*-amplified (MNA) tumours also show defective G1 checkpoint arrest [23]. A common genetic abnormality observed in non-MNA NB is allelic loss of chromosome 11q. Tumours with 11q deletion display a poor prognosis similar to MNA [24]. Many genes coding for proteins involved in the DDR are located on 11q including *ATM*, *CHEK1*, *H2AFX* and *MRE11* [20]. Together, MNA and 11q deletion occur in 70–80% of HR-NB tumours.

Although rare at diagnosis, defects in p53 signalling have been observed in up to 50% relapsed NB tumours [22,25], causing further G1 checkpoint dysfunction and abrogating the p53 dependent intrinsic apoptosis pathway.

Poly ADP-ribose polymerase (PARP) inhibitors also cause RS [26]. PARP is activated in response to DNA single strand breaks and orchestrates repair [27]. Several PARP inhibitors have been approved for ovarian and breast cancer with defective homologous recombination repair. There are currently seven clinical trials testing the use of PARP inhibitors for paediatric tumours of which only three include NB (https://clinicaltrials.gov/: NCT04236414, NCT03233204, NCT02392793).

Preclinical testing of the PARP inhibitor olaparib (Astra Zeneca) in NB shows that PARP inhibition potentiates the cytotoxic effect of a variety of chemotherapy agents and ionising radiation [28–31]. In addition, NB tumours with *MYCN* amplification or *ATM* deficiency have been shown to have increased sensitivity to single agent olaparib treatment [32,33].

We aimed to test if the DDR defects frequently observed in NB would be potential predictive biomarkers of sensitivity to ATR inhibition using VE-821 (the preclinical lead from which M6620 was developed).

We hypothesise that there will be mutual synergy between ATR and PARP inhibitors by further increasing RS, irrespective of *MYCN* or *ATM* status, by the accumulation of unrepaired single strand breaks, when PARP is inhibited, and failure to arrest in S-phase when ATR is inhibited.

In this study, we identify features of NB cell lines that determine sensitivity to ATR inhibition, for use as potential predictive biomarkers, and examine the effect of ATR inhibition on the cytotoxicity of the PARP inhibitor olaparib.

#### **2. Results**

#### *2.1. DDR Protein Expression in NB Cell Lines*

To reflect the variety of DDR defects observed in NB tumours, we chose a panel of NB cell line of varying *MYCN*, 11q and *TP53* status to interrogate what features would lead to sensitivity to ATR and PARP inhibitors. The genetic features of these cell lines are listed in Table 1.


**Table 1.** Cell line genetic abnormalities.

Amp: amplified, Non-amp: non-amplified, WT: wild type; \* cell line is near tetraploid [44], Chr11 LOH; \*\* derived from NGP, assume same.

We analysed endogenous expression of key DDR proteins (ATM, ATR, CHK1, CHK2, MYCN and p53) in these cell lines as well as baseline activity of ATR, ATM (phospho-CHK1S345 and phospho-CHK2T68 and expression, respectively) by Western blot (Figure 1A). Mean protein band intensities of MYCN, ATM and p53 from two independent experiments are shown in Figure 1B. ATM and p53 function after activation with doxorubicin was also examined (Figure 1C,D, respectively).

As expected, *MNA* cell lines show high MYCN protein expression compared to non-*MNA* cell lines (Figure 1A and Figure S3A), with the exception of SJNB1, which has high expression of MYCN in the absence of a gene amplification. In contrast, some cell lines with 11q deletion have baseline ATM expression, suggesting that ATM expression from the other allele is sufficient to produce a functional protein (Figure 1A,C). Cell lines with *TP53* mutations show stabilised p53 protein (NMB and Kat100) or no p53 protein expression (SKNAS and IGRN91). The *TP53* mutation in the NMB and Kat100 cell lines are point mutations leading to accumulation and stabilisation of the dysfunctional protein (Figure 1D and previously in [38,41]), whereas SKNAS and IGRN91 have a deletion and duplication, respectively, of whole exons and do not stabilise the protein. The IGRN91 cell line expresses a high molecular weight gene product after activation with doxorubicin (consistent with duplication of exons 7–9 [35]), which retains some function (Figure 1D).

#### *2.2. MYCN Amplification and Low ATM Expression are Determinants of Sensitivity to ATR Inhibition in NB Cell Lines*

Growth inhibition by the ATR inhibitor VE-821 was determined by XTT cell proliferation assay (Figure 2A). Cell lines were grouped based on genetic features, baseline MYCN, p53 and ATM protein expression above (high) or below (low) median expression (Figure 1B) and ATM and p53 response to doxorubicin. Growth inhibition at 10 μM for each cell line was analysed across groups (Figure 2B–D) by Mann–Whitney *U* test. *MNA* cell lines show high MYCN protein expression (Figure 1B) and these cell lines were significantly more sensitive to 10 μM VE-821 than non-*MNA* cell lines (*p* < 0.05). Although there was no significant difference in VE-821 sensitivity between cell lines with or without *ATM* aberration (*ATM* mutation and/or 11q deletion), cell lines with low baseline ATM protein expression were significantly more sensitive than cell lines with high baseline ATM expression (*p* < 0.05). In regression analysis, MYCN and ATM protein expression was negatively and positively correlated with sensitivity to VE-821, respectively (Figure S3B,C). Cell lines with dysfunctional ATM were more sensitive to VE-821 than cell lines with functional ATM, indicated by ATMS1981 auto-phosphorylation after treatment with 1 μM doxorubicin (Figure 1C). There was no significant difference in sensitivity to VE-821 when cell lines were grouped according to *TP53* mutation status, p53 protein expression or p53 function. These results were confirmed by clonogenic survival assay (Figure S4).

**Figure 1.** Expression and function of key DNA damage response (DDR) proteins in a panel of neuroblastoma (NB) cell lines. (**A**) Baseline protein expression of Rad3 related (ATR), ataxia telangiectasia mutated (ATM), CHK1, CHK2, phospho-CHK1S345, phospho-CHK2T68 MYCN and p53 in NB cell lines used in this study. Ponceau S stain was used as a loading control (Figure S1). \*MYCN amplified, ˆ11q deleted, †TP53 mutant. (**B**) MYCN, ATM and p53 mean protein band intensity measured by densitometry (ImageJ) and normalised to total protein (Ponceau S) from 2 replicates. Cell lines are ordered by protein expression from low to high. The dashed line indicates median expression of each protein. (**C**) ATM function was determined by phospho-ATMS1981 and phospho-CHK2T68 expression after treatment with 1 μM doxorubicin (doxo) for 4 h. (**D**) p53 function was determined by p53 and p21 expression after treatment with doxo for 24 h. Images of the uncropped Western blots for A and C, D can be found in Figures S1 and S2.

**Figure 2.** Determinants of ATR inhibitor sensitivity. (**A**) VE-821 (ATR inhibitor) dose response curves for NB cell lines, data are mean + SEM from 3 independent experiments. Cell lines were split into 2 groups based on (**B**) molecular features: *MYCN* amplification (MNA), p53 mutation status and ATM aberration (11q deleted or *ATM* mutated), (**C**) protein expression above (high) or below (low) median expression (Figure 1B) and (**D**) ATM and p53 responses after treatment with doxorubicin (1 μM). ATM and p53 response was determined by expression of pATMser1981 and p21, respectively (Figure 1C,D). Average percentage control growth at 10 μM VE-821 was plotted for cell lines belonging to each group (*n* = 3). \* *p* < 0.05 Mann Whitney U test, ns: not significant, wt: wildtype).

#### *2.3. MYCN Amplification and Low ATM Expression Are Not Significant Determinants of Sensitivity to PARP Inhibition in NB Cell Lines*

To see if the features that are associated with ATR inhibitor sensitivity also determine sensitivity to PARP inhibition, sensitivity to the PARP inhibitor olaparib was determined by XTT cell proliferation assay (Figure S5A). Cell lines were grouped by the molecular features above and growth inhibition at 10 μM olaparib for each cell line was analysed. Unlike for VE-821, there was no significant difference in olaparib sensitivity when cell lines were grouped according to *MYCN* amplification, MYCN protein expression, ATM protein expression or ATM function (Figure S5B–D). There was no significant difference in sensitivity to olaparib when cell lines were grouped according to *TP53* mutation status, p53 protein expression or p53 function. Single agent sensitivity of cell lines to VE-821 and olaparib by XTT cell proliferation assay is summarised in Table 2.


**Table 2.** Summary of VE-821 and olaparib single agent sensitivity.

However, when analysed by clonogenic assay (Figure S6), cell lines with high MYCN protein expression were significantly more sensitive to olaparib (*p* < 0.05). This was not the case when cell lines were analysed based on *MYCN* amplification and some cell lines, such as SJNB1, which are not *MNA* but express high levels of MYCN, were relatively sensitive to olaparib.

#### *2.4. PARP and ATR Inhibition Synergistically Inhibit Cell Growth*

To test if growth inhibition by olaparib and VE-821 is synergistic, an XTT cell proliferation assay was carried out testing the response of four NB cell lines to increasing concentrations of olaparib with the addition of fixed concentrations of VE-821. SHSY5Y (non-MNA, p53 wt), SKNAS (non-MNA, p53 mutant), NGP (MNA, p53 wt) and N20\_R1 (MNA, p53 mut) cells were treated with 0, 0.1, 1, and 10 μM olaparib with the addition of either 0, 0.1 or 1 μM VE-821 (Figure 3A). VE-821 significantly sensitized the SKNAS, NGP and N20\_R1 cell lines to olaparib, fold sensitisation 1.43-4.60 (Figure 3B,C). Although not statistically significant, the cytotoxicity of olaparib was increased by VE-821 in the SHSY5Y (non-*MNA*, *TP53* wt) cell line, fold sensitisation 1.41. Synergism between VE-821 and olaparib was determined by combination index analysis (calcusyn) where combination index (CI) <1 indicates synergy. PARP inhibition was synergistic with ATR inhibition in all cell lines, irrespective of *MYCN* or *TP53* status (Figure 3D). This was also confirmed by clonogenic survival assay (Figure S7).

**Figure 3.** (**A**) XTT cell proliferation of the SHSY5Y, SKNAS, NGP and N20\_R1 neuroblastoma cell lines treated with 0.1, 1 and 10 μM olaparib alone and with the addition of 0 (blue), 0.1 (red) and 1 (green) μM VE-821. Percentage control growth was normalised to effect of VE-821 alone. Data shown are the mean + SEM from 4 individual experiments. (**B**) Effect of 1 μM VE-821 on cytotoxicity of 1 μM olaparib normalised to the effect of VE-821 alone. Data shown are the mean + SEM from 4 individual experiments. T-test: ns; not significant, \* *p* < 0.05, \*\* *p* < 0.01. (**C**) Fold sensitization of 1 μM olaparib by 1 μM VE-821 for each cell line. (**D**) Combination index (CI) values were calculated using CalcuSyn and plotted in heat map.

#### *2.5. ATR Inhibition Increases Replication Stress Caused by Olaparib by Blocking S and G2 Cell Cycle Arrest and Reducing Homologous Recombination Repair*

PARP inhibition has been shown to increase RS in NB cell lines [32]. Since ATR inhibition potentiated olaparib induced growth inhibition, we investigated the effect of ATR and PARP inhibitor combination on RS markers. As most markers of RS are dependent on ATR activity, such as CHK1S345 and RPA2T21 phosphorylation, we measured phosphorylation of histone 2AX (γH2AX) as a surrogate marker of RS. γH2AX is a marker of DNA double strand breaks (DSB) and RS. In the absence of genotoxic agents, γH2AX primarily marks RS. SHSY5Y, SKNAS, NGP and N20\_R1 cells were treated with olaparib 5 μM and/or VE-821 1 μM for 24 h. pCHK1S345 (ATR activation), pCHK2T68 (ATM activation) and γH2AX (RS) protein expression was measured by Western blot (Figure 4A). Protein expression was quantified by densitometry (ImageJ) and compared to DMSO control (Figure 4B). Treatment with 5 μM olaparib resulted in activation of ATR (pCHK1S345) in all cell lines, which was reduced with the addition of 1 μM VE-821 as expected. pCHK2T68 and γH2AX expression increased after treatment with olaparib and was further increased with the addition of VE-821, suggesting

increased RS and DNA damage. Another marker of RS is phosphorylation of RPA2S4/S8 by ATM and DNA-PK [45]. Using the SHSY5Y cell line, expression of pRPA2S8 was measured after treatment with olaparib 5 μM +/- VE-821 1 μM over the course of 48 h (Figure 4C). The addition of VE-821 1 μM brought forward the RPA2S8 phosphorylation induced by olaparib 5 μM by 24 h.

**Figure 4.** Effect of VE-821 and olaparib combination on replication stress. (**A**) pCHK1S345, pCHK2T68, γH2AX protein expression of NGP, N20\_R1, SHSY5Y and SKNAS cells after incubation with 5 μM olaparib and/or 1 μM VE-821 for 24 h. (**B**) pCHK1S345, pCHK2T68, γH2AX mean fold change in protein band intensity measured by densitometry (ImageJ) and normalised to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) loading control, data are mean + SEM from 3 independent experiments. (**C**) pRPA2S8 expression in SHSY5Y cell line in response to 5 μM olaparib with or without 1 μM VE-821 over 48 h. Images of the uncropped Western blots can be found in Figure S8.

ATR signals to S and G2 cell cycle arrest [4], which are abrogated when ATR is inhibited [5,6,46]. Since olaparib is known to lead to cell cycle arrest [47,48], we investigated the effect of olaparib and VE-821 combination on the cell cycle in SHSY5Y, SKNAS and NGP cell lines (Figure 5A). Olaparib treatment alone increased the proportion of cells in S and G2 phase for each cell line, consistent with response to olaparib-induced RS. This cell cycle arrest was not seen with the addition of both inhibitors, suggesting loss of S and G2 checkpoints arrest.

**Figure 5.** Effect of VE-821 and olaparib combination on cell cycle and homologous recombination repair (HRR). (**A**) Cell cycle analysis of SHSY5Y, SKNAS and NGP cell lines treated with 5 μM (SHSY5Y and SKNAS) or 1 μM (NGP) olaparib in combination with 1 μM VE-821 for 24 h. Data are mean + SEM from 3 independent experiments. (**B**) Representative γH2AX and Rad51 foci images from SKNAS and NGP cell lines treated with 1 μM VE-821, 10 μM olaparib or both for 24 h. (**C**) Average number of Rad51 foci per cell for NGP, N20\_R1, SHSY5Y and SKNAS cell lines treated with 1 μM VE-821, 10 μM olaparib or both, *n* = 4 + sd. (**D**) Average γH2AX nuclear fluorescence intensity for NGP, N20\_R1, SHSY5Y and SKNAS cell lines treated with 1 μM VE-821, 10 μM olaparib or both, fold change from control (DMSO), data are mean + SEM from 4 independent experiments. \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001, \*\*\*\* *p* ≤ 0.0001, 2-way ANOVA difference from control (DMSO).

We also examined the effect on homologous recombination repair (HRR). PARP inhibition is synthetically lethal with HRR deficiency [49]. ATR inhibition has been shown to reduce Rad51 foci formation (an indicator of HRR activity [50]), inducing a HRR-deficient phenotype [51,52]. We analysed the effect of 10 μM olaparib and/or 1 μM VE-821 on Rad51 and γH2AX foci in SHSY5Y, SKNAS, NGP and N20\_R1 cell lines. Representative γH2AX and Rad51 foci images from SKNAS (non-*MNA*, *TP53* mutant) and NGP (*MNA*, *TP53* wt) cell lines treated with 1 μM VE-821, 10 μM olaparib or both are shown in Figure 5B. Treatment with olaparib induced an increase in Rad51 foci formation, which was suppressed with the addition of VE-821 in all cell lines except SKNAS (Figure 5C). Interestingly, SKNAS cells are relatively resistant to both olaparib and VE-821. RS was measured by γH2AX nuclear fluorescence intensity, which increased after treatment with olaparib and increased further after treatment with both inhibitors in all cell lines (Figure 5D).

Overall, the combination of olaparib and VE-821 leads to increased RS due to defective S and G2 cell cycle arrest and suppression of HRR leading to collapsed replication forks.

#### **3. Discussion**

It is widely accepted that overexpression of proliferation-inducing oncogenes leads to RS and dependency on ATR signalling [53]. In models of Ras- or MYC-driven cancer, signalling through these oncogenes has been shown to lead to sensitivity to ATR inhibition [54–56]. In NB cell lines, high MYCN expression has been shown to increase RS [32,57]. Here we have shown that amplification and overexpression of the *MYCN* oncogene in NB cell lines is associated with sensitivity to VE-821, confirming that *MYCN* driven NB cells are vulnerable to ATR inhibition.

ATM loss or dysfunction is also associated with sensitivity to VE-821. Our results confirm that reduced ATM function, which can be partially determined by protein level, confers sensitivity to ATR inhibition in NB as it does in other tumour types [5,58–60]. In addition to 11q deletion, loss of ATM can occur in *MNA* NB by upregulation of the ATM targeting micro-RNA, miR-421 [61]. ATM silencing, in addition to MYCN-driven proliferation, is likely contributing to replication stress through impaired DNA DSB signalling and repair, making these cells especially vulnerable to ATR inhibition.

MYCN and ATM protein expression levels were more powerful determinants of VE-821 sensitivity than the genetic status (amplification or deletion/mutation, respectively), and MYCN protein levels positively correlated with VE-821 sensitivity, whereas ATM protein level was negatively correlated with VE-821 sensitivity.

Mutation, expression or dysfunction of the p53 tumour suppressor protein was not associated with sensitivity to ATR inhibition. Middleton et al. also demonstrated that defective p53 signalling did not lead to sensitivity to VE-821 alone, but the addition of VE-821 showed greater potentiation of gemcitabine and ionising radiation induced cytotoxicity in p53-/- cell lines [46].

To summarise, MYCN and ATM protein expression levels, as well as genetic status, could provide useful predictive biomarkers to stratify NB patients who would benefit from an ATR inhibitor.

Further investigation into combining ATR inhibition with DNA-damaging chemotherapy may provide a novel therapeutic strategy for both *TP53* wt and *TP53* mutant NB.

In contrast, neither *ATM* nor *MYCN* status significantly determined sensitivity to PARP inhibition by olaparib. Colicchia et al. previously demonstrated that *MYCN*-amplification or overexpression renders NB cells especially sensitive to single agent olaparib [32]. Sensitivity of cell lines to PARP inhibitors by this group was analysed by MTS cell proliferation assay, an assay which measures metabolic activity in a similar way to XTT. Although we did not see a significant difference between MNA cell lines and non-MNA cell lines in olaparib sensitivity in either XTT or clonogenic survival assay, high MYCN expressing cell lines did show significantly increased sensitivity when analysed by clonogenic survival. The clonogenic survival assay measures the ability of single cells to divide indefinitely and relies on cell lines having good cloning efficiency. Unfortunately, some NB cell lines show poor cloning efficiency and some cell lines show stark differences in sensitivity between the two assays.

Our results highlight the importance of validating sensitivity found to novel agents in models of disease with other types of cell viability assays. In the case of NB, more evidence is required before the role of MYCN in PARP sensitivity can be established. It may be that in the context of complex genetic alterations, cells may have acquired a spectrum of features leading to resistance, as well as sensitivity to these inhibitors, making it difficult to predict response.

In addition to investigating features associated with sensitivity to these inhibitors as single agents, we provide evidence of synergy between ATR and PARP inhibition in NB cell lines. VE-821 enhanced olaparib-induced growth inhibition independently of *MYCN*, *ATM* and *TP53* status. ATR inhibition has been shown to enhance PARP inhibitor sensitivity in a variety of adult solid tumour types [51,62,63] and overcome resistance to PARP inhibitors [52,64].

Olaparib treatment alone induced markers of RS (γH2AX, pRPA2S8, Figures 4 and 5) and activated ATR (pCHK1S345 expression, Figure 4A). When ATR was inhibited, olaparib-induced markers of RS were increased, which suggested the inhibition of ATR exacerbated RS caused by PARP inhibition. This could be because inhibition of ATR releases olaparib-induced S and G2 arrest, shown here (Figure 5A) and previously reported [63]. Loss of cell cycle checkpoints results in DNA replication through damaged areas, fork stalling, collapse and chromosome breakage, which, if left unrepaired, leads to mitotic catastrophe and cell death [44].

Loss or inhibition of ATR has been shown to restrict HRR function [51,52,65]. We confirmed this in MNA and non-MNA NB cell lines. However, one non-MNA cell line, SKNAS, did not show increased Rad51 after PARP inhibition, and would be considered HRR-defective. HRR-defective cells are usually highly sensitive to PARP inhibitors, whereas the SKNAS cell line is the most resistant to olaparib out of the cell lines analysed in this study.

Although inhibiting ATR to specifically target RS in cancer cell looks promising, other mechanisms of tolerating RS have been identified, such as overexpression of components of the fork protection complex [66] and recruitment of Y-family (translesion synthesis) DNA polymerases [67,68], which should be considered as possible resistance mechanisms. These are all factors to be considered when interpreting data from the recently initiated trials of PARPi and ATRi combinations (NCT02723864, NCT03462342, NCT04065269, NCT03787680, NCT04149145, and NCT03682289).

In conclusion, we have shown that the combination of ATR and PARP inhibition leads to increased RS and cytotoxicity in NB cell lines. This synergy is likely to be, at least in part, not only due to abrogation of cell cycle checkpoints but also inhibition of HRR, thereby inducing synthetic lethality with PARP inhibition.

#### **4. Materials and Methods**

#### *4.1. Chemicals and Cell Lines*

VE-821 and Olaparib were purchased from Stratech Scientific Ltd. (Cambridge, UK) and stored at −20 ◦C in stock solution of 100 mM and 20 mM, respectively, in DMSO (Sigma-Aldrich, Gillingham, Dorset, UK).

Neuroblastoma cell lines used in this study were: SHSY5Y, SKNAS, NGP, N20\_R1, NMB, IMR32, Kat100, IGRN91, SJNB1, and GIMEN (Table 1). All neuroblastoma cell lines were obtained between 1996 and 2018, with the exception of N20\_R1, and were validated upon receipt using cytogenetic analysis courtesy of Dr Nick Bown (Institute of Genetic Medicine, Newcastle University). N20\_R1 (N\_N20R1) were generated from parental NGP cell line with resistance to Nutlin-3 [37]. All cell lines were cultured in RPMI-1640 (Sigma-Aldrich) supplemented with 10% (*v*/*v*) Fetal Calf Serum (Gibco, Life Technologies Ltd., Thermo Fisher Scientific, Waltham, MA, USA) and maintained at 37 ◦C in a humidified incubator with 5% CO2. Cell lines were routinely tested for Mycoplasma and were confirmed to be negative.

#### *4.2. Growth Inhibition Assay*

Cells were seeded in 96-well plates (Corning, VWR International Ltd., Lutterworth, UK), and allowed to adhere overnight before treatment with ATR or PARP inhibitors alone or in combination for 72 h. Inhibitors were added at 200× dilution to give a final DMSO concentration of 0.05%. Percentage

control growth was assessed using the XTT cell proliferation assay (Roche, Burgess Hill, UK) according to the manufacturer's instructions and using the following formula: (average absorbance test/average absorbance control) × 100. Combination Index (CI) values were determined by the Chou–Talalay method using CalcuSyn v2 (Biosoft, Cambridge, UK).

#### *4.3. Protein Extraction and Western Blotting*

Cell pellets were harvested and protein extracted using PhosphoSafe™ Extraction Buffer (Novagen, Merck Millipore Ltd. Watford, UK) following the manufacturer's protocol. Protein concentration was quantified using the Pierce BCA protein assay kit (Thermo Fisher Scientific).

Proteins were separated using 3–8% Criterion™ XT Tris-Acetate Protein Gel (Bio-Rad Laboratories Ltd., Hemel Hempstead, UK) and transferred onto Hybond-C Extra membrane (GE Life Sciences, Little Chalfont, UK). Membranes were stained with Ponceau S (Sigma-Aldrich) to control for loading, destained in tris buffered saline, 0.5% tween 20 (TBST) and blocked for 1 h in 5% milk TBST.

Primary antibodies were: mouse monoclonal MYCN 1:1000 (Santa Cruz Bio-technology (SCBT), Dallas, TX, USA: A1513), rabbit polyclonal ATM 1:500 (cell signalling technology(CST), Danvers, MA, USA: 2873S), mouse phospho-ATM (ser1981) 1:1000 (CST: 4526), mouse monoclonal CHK1 1:1000 (SCBT: sc-8408), mouse monoclonal CHK2 1:1000 (SCBT: sc-17747), rabbit polyclonal phosho-CHK1 (ser345) 1:1000 (CST:2341S), rabbit monoclonal phosphor-CHK2 (thr68) 1:1000 (CST:2197S), rabbit polyclonal ATR 1:500 (CST: 2790S), p53 1:1000 (NCL-L-p53-DO7, Leica Biosystems Ltd.),rabbit monoclonal p21 Waf1/Cip1 (12D1) 1:1000 (CST: 2947S), rabbit monoclonal phosopho-H2AX (ser139) 1:2000 (CST: 2577S), rabbit monoclonal phospho-RPA2 (ser8) 1:1000 (CST: 54762S), mouse monoclonal glyceraldehyde 3-phosphate dehydrogenase (GAPDH) 1:5000 (SCBT: sc-47724). Secondary antibodies used were peroxidase-conjugated goat anti-mouse (Dako, Glostrup, Denmark) and anti-rabbit immunoglobulins (Dako) at 1:2500.

Protein detection was performed using enhanced chemiluminescence (Bio-Rad) and imaged using the ChemiDoc imaging system (Bio-Rad Laboratories Ltd., Hemel Hempstead, UK). Densitometry was performed using ImageJ image analysis software (Version 1.52p; Java 1.8.0\_172 [64-bit]).

#### *4.4. Cell Cycle Analysis*

Cells were harvested post-treatment, fixed in ice-cold 70% (*v*/*v*) ethanol and stored at −20 ◦C. Prior to analysis, cells were washed with phosphate buffered saline (PBS), resuspended in 500 μL PBS with 50 μg/mL propidium iodide (Sigma-Aldrich) and 50 μg/mL RNAse A (Sigma-Aldrich), and incubated at 37 ◦C for 30 min. Samples were analysed on the Attune NxT Flow Cytometer using Invitrogen™ Attune NxT Software (Thermo Fisher Scientific). Data were analysed using FlowJo TM (BD Biosciences, Wokingham, UK). Experiments were at least *n* = 3.

#### *4.5. Immunofluorescence*

Cells were treated for 24 h with control vehicle (DMSO) and 1 μM VE-821 with or without 10 μM olaparib. Cells were stained with mouse monoclonal anti phospho-Histone H2A.X (Ser139) antibody (SCBT) at 1:500 and rabbit monoclonal anti RAD51 antibody (CST: 8875S) at 1:250. Secondary antibodies used were Alexa 488 conjugated goat anti rabbit and Alexa 546 conjugated goat anti mouse (Invitrogen, Thermo Fisher Scientific), both at 1:1000. Cells were imaged using a Leica DM6 microscope and Leica Application Suite (LAS) X software (Leica Microsystems, Wetzlar, Germany). The number of RAD51 foci in each cell and total nuclear fluorescence intensity for γH2AX were quantified using ImageJ software (Version 1.52p; Java 1.8.0\_172 (64-bit)) and data was plotted using GraphPad Prism v6 (San Diego, CA, USA).

#### *4.6. Statistical Analysis*

Mann Whitney U and 2-way ANOVA statistical tests were carried out using GraphPad Prism v6 software and *p* < 0.05 was taken to be statistically significant.

#### **5. Conclusions**

In the era of precision medicine, cancer specific DDR defects and RS have become attractive targets, with many novel agents entering clinical trials. With the introduction of PARP inhibitors into clinical trials for the treatment of paediatric cancers, including HR-NB, a disease with one of the worst long term prognoses, and ATR inhibitors being tested in adult trials, we sought to investigate which NB specific DDR defects, if any, would lead to sensitivity to these agents alone and in combination. We have shown that *MYCN*-amplification and protein overexpression and loss of functional ATM protein through deletion or mutation result in vulnerability to ATR inhibition by VE-821. The case for PARP inhibition is less clear and, although there is a trend towards MYCN overexpressing cells being more sensitive to olaparib, the role of MYCN in PARP inhibitor sensitivity needs to be investigated further. We also provide evidence of synergy between ATR and PARP inhibition in NB cell lines independent of *MYCN*, *ATM* or *TP53* status. This may be due to increased RS, compromised S and G2 checkpoint arrest and/or defective HRR in the presence of both PARP and ATR inhibitors.

Overall, our work gives exciting insights into the use of PARP and ATR inhibitors as a novel treatment strategy for HR-NB.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/5/1095/s1, Figure S1: Full images of baseline DDR protein expression by Western blot, Figure 1A, Figure S2: Full images of Western blots in Figure 1C,D, Figure S3: Correlation of relative baseline protein expression of MYCN and ATM and VE-821 sensitivity at 10 μM, Figure S4: Determinants of ATR inhibitor sensitivity analysis by clonogenic survival assay, Figure S5: Determinants of PARP inhibitor sensitivity by XTT cell proliferation assay, Figure S6: Determinants of PARP inhibitor sensitivity analysis by clonogenic survival assay, Figure S7: Effect of VE-821 on olaparib cytotoxicity by clonogenic survival assay, Figure S8: Full images of Western blots in Figure 4.

**Author Contributions:** Conceptualization, H.E.D.S., L.C., D.A.T. and N.J.C.; Formal analysis, H.E.D.S.; Funding acquisition, D.A.T. and N.J.C.; Investigation, H.E.D.S.; Methodology, H.E.D.S. and L.C.; Supervision, D.A.T. and N.J.C.; Writing—original draft, H.E.D.S.; Writing—review and editing, H.E.D.S., L.C., D.A.T. and N.J.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Children's Cancer and Leukaemia Group and Little Princess Trust (CCLGA-2017-12).

**Acknowledgments:** This work was funded by the Children's Cancer and Leukaemia Group and Little Princess Trust (CCLGA-2017-12)-. We thank the following for cell lines: Linda Harris (SJNB1), Penny Lovat (SHSY5Y and IMR32), Rogier Versteeg (NGP), Mirco Ponzoni (GIMEN), John Lunec (N20\_R1), Jean Bénard (SKNAS and IGRN91), Garrett Brodeur (NMB) and Michelle Haber (IMR/KAT100).

**Conflicts of Interest:** N.J.C. has received research funding from Aguoron Pharmaceuticals and Pfizer for the development of rucaparib, BioMarin for studies on talazoparib and Tesaro for studies with niraparib and paid consultancy from Abbvie for veliparib. N.J.C. is inventor on the patent relating to the use of rucaparib in HRD cancer (WO2005012305A3) and Newcastle University receives royalty income from the sales of rucaparib that are shared with contributors to its development, including N.J.C. N.J.C. has also received research funding from Vertex Pharmaceuticals and is currently in receipt of funding from Merck KGaA for work on ATR inhibitors and holds a patent for a pharmacodynamic biomarker of ATR inhibition (WO2014055756A1). All other authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

## **Reduced Tumorigenicity of Mouse ES Cells and the Augmented Anti-Tumor Therapeutic E**ff**ects under** *Parg* **Deficiency**

**Yuki Sonoda 1,2,**†**, Yuka Sasaki 1,3,**†**, Akemi Gunji 4, Hidenori Shirai 4, Tomonori Araki 1,2, Shoji Imamichi 1,3, Takae Onodera 1,3, Anna-Margareta Rydén 4, Masatoshi Watanabe 5, Jun Itami 6, Takuya Honda 2, Kazuto Ashizawa 7, Kazuhiko Nakao <sup>2</sup> and Mitsuko Masutani 1,3,4,8,\***


Received: 7 April 2020; Accepted: 21 April 2020; Published: 24 April 2020

**Abstract:** PolyADP-ribosylation is a post-translational modification of proteins, and poly(ADP-ribose) (PAR) polymerase (PARP) family proteins synthesize PAR using NAD as a substrate. Poly(ADP-ribose) glycohydrolase (PARG) functions as the main enzyme for the degradation of PAR. In this study, we investigated the effects of *Parg* deficiency on tumorigenesis and therapeutic efficacy of DNA damaging agents, using mouse ES cell-derived tumor models. To examine the effects of *Parg* deficiency on tumorigenesis, *Parg*+/+ and *Parg*−/<sup>−</sup> ES cells were subcutaneously injected into nude mice. The results showed that *Parg* deficiency delays early onset of tumorigenesis from ES cells. All the tumors were phenotypically similar to teratocarcinoma and microscopic findings indicated that differentiation spectrum was similar between the *Parg* genotypes. The augmented anti-tumor therapeutic effects of X-irradiation were observed under *Parg* deficiency. These results suggest that *Parg* deficiency suppresses early stages of tumorigenesis and that *Parg* inhibition, in combination with DNA damaging agents, may efficiently control tumor growth in particular types of germ cell tumors.

**Keywords:** cancer; poly (ADP-ribose) glycohydrolase; radiosensitization; poly (ADP-ribose) polymerase; ES cells

#### **1. Introduction**

PolyADP-ribosylation is the post-translational modification of proteins, and poly(ADP-ribose) (PAR) polymerase (PARP) family proteins synthesize PAR using NAD as a substrate. Poly(ADP-ribose) glycohydrolase (PARG) is involved in the degradation of PAR as a main enzyme. PARP-1 is activated

upon DNA damage and is crucial for preserving genome stability [1]. Activated PARP-1 recognizes DNA strand breaks and is involved in base-excision repair, DNA strand break repair pathways [1].

The process of polyADP-ribosylation of proteins is tightly regulated by the second enzyme in the metabolic pair: PARG and ADP-ribosyl hydrolase3 (ARH3) [2]. PARG reverses the action of PARP by hydrolyzing the glycosidic bonds of PAR to produce ADP-ribose [3,4]. The catalytic capacity of PARG keeps the polyADP-ribosylation of proteins transient; the conversion rate of PAR is measured in minutes [1].

In DNA repair process, PARG is involved in the regulation of base excision repair and DNA strand break repair through the control of XRCC1 function [5]. PARG functional inhibition leads to increase in radiosensitivity in particular types of cells [6,7]. Perturbation of the PARP-1/PARG balance by the over-expression of PARG has been also shown to alter the genome methylation pattern to that of cancer cell types [8].

The importance of PARG in normal development is underscored by the fact that disruption of PARG leads to embryonic lethality [9,10]. PARG enzyme structures have been revealed [11–13]. There are several isoforms of PARG, which are produced from a single gene by alterative splicing. The longest (110 kDa, PARG110) and most abundant isoform is located in the nucleus, while the 60 kDa isoform is located in the cytoplasm and mitochondria [14–17]. Disruption of exon 4, which is common to all PARG isoforms, leads to cell lethality [9]. However, disruption of exons 2–3 abrogates expression of nuclear PARG, but still allows expression of the remaining isoforms [18]. Furthermore, deficiency of the full-length isoform of PARG leads to enhanced cytotoxic sensitivity by causing PAR accumulation, induced by menadione, alkylating agents, endo-toxic shock and γ-irradiation [18,19].

Upon extensive DNA damage, PARP-1 becomes excessively activated and depletes cellular NAD<sup>+</sup> to polyADP-ribosylate proteins. PAR accumulation and the depletion of NAD<sup>+</sup> severely perturb the energy balance of cells and lead to apoptosis. PAR accumulation also triggers AIF (apoptosis inducing factor) activation in mitochondria and induces parthanotos, which can be observed in neuronal and cancer cells [20,21]. In heterozygously knockout mice of *Parg*, suppression of lung cancer development has been reported [22].

As described above, the cytoprotective role of PARG through its involvement in DNA repair in normal cells has been suggested, whereas the functional inhibition of PARG is reported to cause sensitization to DNA-damaging agents in cancer cells. Therefore, PARG may be considered both a tumor suppressor and a therapeutic target of cancer. The accumulated evidence thus led us to hypothesize that PARG may impact both cancer development and cancer therapy. Meanwhile, embryonic stem (ES) cells of mice can be useful as experimental systems for tumorigenesis and teratocarcinoma. In this study, using hypomorphic *Parg* knockout ES cells, we investigated the effects of *Parg* deficiency on tumorigenesis and the therapeutic efficacy of DNA damaging agents using ES cell derived tumor models.

#### **2. Results**

#### *2.1. Parg*−/<sup>−</sup> *ES Cells Show Delayed Tumor Development*

We previously generated two hypomorphic *Parg*−/<sup>−</sup> ES cell clones, D79 and D122, which retained about 10% residual PARG activity compared to parental wild-type J1 ES cells [7]. The growth rates of these *Parg*−/<sup>−</sup> and J1 ES cells are similar in the absence of DNA damaging agents. To examine the effects of *Parg* deficiency on tumorigenesis, *Parg*+/+ J1 and two D79 and D122 *Parg*−/<sup>−</sup> ES cells were subcutaneously injected into the flanks of nude mice. Following injection, tumor size development was observed weekly over four weeks. An initial delay of tumor growth was observed at weeks two and three in tumors derived from *Parg*−/<sup>−</sup> ES cells (*p* < 0.01, Figure 1). This effect was observed during only the early phase, as tumor size did not differ significantly between the genotypes at week four. These results indicate that PARG deficiency delays the early onset of tumorigenesis derived from ES cells.

**Figure 1.** <sup>E</sup>ffect of *Parg* deficiency on tumorigenesis from embryonic stem (ES) cells. In total, <sup>1</sup> <sup>×</sup> <sup>10</sup><sup>7</sup> ES cells were inoculated *s.c.* into nude mice and size of tumors was measured weekly. Wild-type, J1. *Parg*−/−, D79 and D122.

#### *2.2. Characterization of Tumor Tissues*

To further characterize the tumorigenesis, histological evaluation of HE-stained tumor sections was carried out. Comparison of tissues and cell type are summarized in Table 1. Tumors observed four weeks after injection into nude mice showed ectodermal, mesodermal and endodermal tissue derivatives. Both undifferentiated and differentiated germinal components were detected. Histopathological examination at four weeks identified all tumors as immature teratoma, partially accompanying embryonal carcinoma components. Microscopically, the tumors showed heterogeneous components, especially containing immature neuroectodermal tissue in the form of primitive neuroepithelial rosettes and tubules (Figure 2B,C).

**Table 1.** Tissue components in tumors from *Parg*+/+ and *Parg*−/<sup>−</sup> ES cells.


**Figure 2.** *Cont.*

**Figure 2.** Hematoxylin-eosin staining of tumor tissues derived from mouse ES cells. (**A**) Hematoxylin-eosin staining of tumor tissues derived from mouse ES cells 1 and 2 weeks after injection. Upper panels, wild-type J1. Lower panels, *Parg*−/−. (**B**) The left graph shows the percentage of hyperchromatic area of the tumors 4 weeks after injection. Percentage of hematoxylin-positive hyperchromatic area in the total area of tumor section was measured for each tumor. \* *p* < 0.01, \*\* *p* < 0.005. Right panels show the typical hyperchromatic areas of hematoxylin-eosin staining of tumors 4 weeks after injection. Upper panels, ×20 magnification (Squares show magnified regions in the lower panels. Lower panels, ×40 magnification. The tumors showed heterogeneous cell components containing primitive neuroepithelial components and embryonal carcinoma components. (**C**) HE staining and immunostaining of the tumors at 4 weeks with antibodies against b-III-tubulin, ectoderm marker; AFP, endoderm marker (×20 magnification). Hematoxylin-eosin staining, ×10 magnification. The mixed staining pattern of ectodermal and endodermal markers was observed in hyperchromatic regions of *Parg*−/<sup>−</sup> tumors at 4 weeks. (**D**) Immunostaining of the tumors at 4 weeks after injection with antibody against anti-PAR. Right panels in D are magnified images, Bars, 50 mm (left panels in D), 20 mm (right panels in D). PAR staining was observed occasionally in the cell nuclei in the *Parg*−/<sup>−</sup> tumor but not in the *Parg*+/+ tumor.

We previously observed that tumors derived from *Parp1*−/<sup>−</sup> ES cells showed differentiation into trophoblast lineages, including trophoblast giant cells [23]. Microscopic findings from the tumors derived from *Parg*−/<sup>−</sup> ES cells showed no such components, suggesting that in the hypomorphic *Parg* deficient state, marked differentiation alterations did not occur (Table 1).

#### *2.3. Time Course Analysis of Tumorigenesis*

To evaluate the defect in early stage tumorigenesis under *Parg* deficiency, further histological analyses were performed on sections of tumor tissues (Figure 2A,B). At one and two weeks after injection, tumors derived from *Parg*−/<sup>−</sup> ES cells showed a higher tendency of necrosis. The density of tumor cells and stromal cells appeared to be lower in the *Parg*−/<sup>−</sup> tumors. As shown in Figure 2B, comparison of percentage of hematoxylin-positive regions in the tumors at four weeks (Figure S1) showed the augmented hematoxylin-positivity, namely hyperchromatic areas (typical areas are shown as Figure 2B), in *Parg*−/<sup>−</sup> tumors with a statistical significance. It may suggest that the chromatin density of the cells was higher, possibly reflecting differences in the chromatin state or cell properties.

To characterize the properties of differentiated cells and hyperchromatic components further, we performed the immunostaining analysis for the tumors at four weeks with antibodies against beta-III-tubulin, ectoderm marker; AFP, endoderm marker; TRA-1-60, pluripotent marker, and Brachyury, mesoderm marker. As shown in Figure 2C, immunohistochemical analysis showed beta-III tubulin-positive staining of immature neuroepithelial tissues in both wild-type and *Parg*−/<sup>−</sup> tumors. It is, therefore, implied that hyperchromic regions may consist of both ectodermal and endodermal differentiated tissues. The pluripotent marker TRA-1-60 showed higher tendencies of diffused staining in the stromal components of the *Parg*−/<sup>−</sup> tumor compared with the wild-type tumor. For the mesoderm marker Brachyury, a higher tendency of staining was also observed in the cell components of *Parg*−/<sup>−</sup> tumor. On the other hand, AFP-positive staining patterns of teratocarcinoma and immature glandular components were similar between *Parg*−/<sup>−</sup> and wild-type tumors.

We also analyzed whether PAR accumulation occurs in *Parg*−/<sup>−</sup> tumors, due to hypomorphic Parg activity. As presented in Figure 2D panels, PAR staining was observed occasionally in the cell nuclei of the *Parg*−/<sup>−</sup> tumors but not in the *Parg*+/+ tumor. This elevated level of PAR confirmed the defect of Parg activity.

#### *2.4. Augmented Anti-Tumor Therapeutic E*ff*ects under Parg Deficiency*

Previous reports describe that *Parg <sup>110</sup>* deficient animals are more sensitive to MMS treatment and γ-irradiation compared to wild-type mice [18,19]. Hypomorphic *Parg*−/<sup>−</sup> ES cells and particular human cancer cells with *PARG* knocked down also showed increased sensitivity to DNA-damaging agents, such as alkylating agents and γ-irradiation. Therefore, we first treated ES cells in vitro with MMS and γ-irradiation, and three hours later, the cells were subcutaneously injected to nude mice to observe the tumor growth. As shown in Figure 3A,B, reflecting the in vitro higher sensitivity of *Parg*−/<sup>−</sup> ES cells, the tumor growth is delayed for at least 2–3 weeks after treatment with both MMS and 6 Gy γ-irradiation.

**Figure 3.** Effect of *Parg* deficiency on tumorigenesis from ES cells pretreated with MMS or γ–irradiation before injection into nude mice. ES cells treated with γ–irradiation at 6 Gy (**A**) or 0.5 mM MMS (**B**) and cultured for 3 h, respectively, were inoculated *s.c.* into nude mice. Tumor growth was observed for four weeks. The bars show the mean size.

Next, we carried out therapeutic models of local X-irradiation using tumors derived from wild-type and *Parg*−/<sup>−</sup> ES cells. As shown in Figure 4A–D, when the tumors derived from *Parg*−/<sup>−</sup> ES cells D79 and D122 were X-irradiated at the single dose of 7 Gy, they showed a delay in tumor growth. In contrast, when the tumors derived from wild-type J1 ES cells were X-irradiated, delay in tumor growth was not clearly observed. It is thus suggested that the therapeutic efficacy of X-irradiation could be higher in the tumors harboring *Parg* deficiency.

**Figure 4.** Effect of *Parg* deficiency on therapeutic efficacy of X–irradiation on tumors derived from ES cells. Tumors of one of the hinder leg pair were subjected to local X–irradiation with a single dose at 7 Gy ((**A**,**C**): *n* = 7 for wild-type (J1), *n* = 8 for *Parg*−/<sup>−</sup> (D79). (**B**,**D**): *n* = 9 for wild-type (J1) and *Parg*−/<sup>−</sup> (D122)). Tumor growth of non-irradiation controls (Control) was monitored for non-irradiated side of hinder leg pair. Tumor growth was observed for 22 days. (**A**,**B**) Tumor growth curve, Mean ± S. E. (**C**,**D**) Measured tumor weight 22 days after injection of ES cells, Mean ± S. E. (**C**) Statistical analysis was carried out with Turkey's test. \* *p* < 0.05. (**D**) Statistical analysis was carried out with Mann–Whitney *U*-test. \* *p* < 0.01.

These results suggest that Parg inhibition in combination with DNA damaging agents may efficiently control tumor growth in particular types of germ cell tumors.

#### **3. Discussion**

Our findings demonstrate that *Parg* deficiency delays the early onset of cancer in vivo in teratocarcinoma model. This suggests that PARG might be an attractive therapeutic target in cancer control of germ cell tumors. The therapeutic model used in this study showed that X-irradiation is more effective in tumors with *Parg* deficiency compared to wild-type cells.

ES cells are derived from normal cells and their tumor models are phenotypically close to teratocarcinoma and teratoma. It is reported that tumorigenicity of ES cells is driven by oncogene action of *E-Ras* gene [24], and the genes including PI(3)K [24] and *c-Jun* [25], *Cox1*/*2* [26] and *Oct3*/*4* [27] are involved in the tumorigenesis process. The present study implies that *Parg* could be a candidate target for the suppression of tumorigenesis at early stages.

In the subcutaneous injection model of ES cells used herein, ES cells should be grown in spheroids or attached to extracellular matrix in subcutaneous regions. During the stressed period of early tumorigenesis, the presence of *Parg* was suggested to enhance cell survival. In cultured *Parg*−/<sup>−</sup> ES cells, the PAR degradation activity was decreased to 10% of wild-type ES cells [7]. We observed that PAR staining was observed occasionally in the cell nuclei of the *Parg*−/<sup>−</sup> tumors but not in the *Parg*+/+ tumors. This accumulation of PAR confirmed the defect in *Parg* activity. The infrequent accumulation of PAR may explain the advance of tumor growth in *Parg*−/<sup>−</sup> tumors at four weeks, as efficient PAR degradation may be necessary for proliferation. The pluripotency marker staining of *Parg*−/<sup>−</sup> tumors at four weeks was also observed in a scattered manner, suggesting that the presence of a majority of differentiated tissues could be related to the infrequent accumulation of PAR in *Parg*−/<sup>−</sup> tumors, as differentiated cells are reported to have low levels of PAR compared with proliferating cells [1]. The occasional increase in PAR staining in *Parg*−/ − tumors could be possibly due to the presence of S-phase cells with stalled replication forks [28], leading to a lower proliferation activity. Another possibility could be an increased expression of other PAR degradation enzymes, such as ARH3 in the differentiated tumors. The clarification of the relationship between PAR accumulation and cell proliferation needs to be addressed in further studies.

When we added further stress with alkylating agents and γ-irradiation to ES cells (Figure 3A,B), tumor development was further suppressed under *Parg* deficiency. This was consistent with our previous observation that when *Parg* was inhibited in ES cells, caspase-dependent apoptosis was enhanced with S-phase arrest and PAR accumulation after MMS treatment [29].

In diseases, PARP-1 has been shown to be involved in stroke, ischemia diabetes and other inflammatory diseases [30]. PARP-1 is frequently overexpressed in various types of cancers [31–33]. As PARG has an opposing enzymatic activity to PARP-1, it is reasonable to hypothesize a similar involvement in disease categories. The involvement of PARG in cancer development has not been extensively studied. However, recent genome sequencing approach accumulated the information of *PARG* gene mutations in various types of cancer. As shown in Figure 5, the data from TCGA database (A) and CanSAR database (B) showed that *PARG* gene mutations are present in particular types of cancers and of note, deep deletions could be observed in non-seminomatous germ cell tumors, melanoma, and well-differentiated thyroid cancers, possibly indicating PARG-deficient state. On the other hand, esophageal squamous cell carcinoma, cholangiocarcinoma, bladder urothelial carcinoma, thymic epithelial tumors and diffuse glioma show amplification of *PARG* gene, possibly suggesting the overexpression of PARG. In fact, PARG activity is reported to be high in C6 glioma tumor cells [34]. With further clinical and basic studies, PARG-deficiency and PARG-overexpression may become useful biomarkers for therapeutic selection and monitoring.

Functional inhibition of PARG in pancreatic cancer cells, which has p53 pathway inactivation, enhanced the necrotic cell death after MMS treatment [29], suggesting that PARG could be a therapeutic target in certain types of cancer cells. It is reported that ovarian cancer cells show differential sensitivity to PARG and PARP inhibitors, and cells with replication vulnerability show persistent replication fork stalling and replication catastrophe, with treatment of the PARG inhibitor and sensitization to CHK1 inhibitor [35].

Whilst the hypomorphic ES cells used in this study show PARG activity of residual 10%, the cell growth was not reduced compared to wild-type ES cells [29], suggesting that ES cells may not be replication vulnerable cells. *Parg* deficient ES cells also showed a higher sensitivity to cisplatin, but not to topoisomerase I inhibitor, campthothecan and hydrogen peroxide [36].

PARG is involved not only in the DNA repair pathway but also in various cell physiological processes, including epigenetic regulation [8], microRNA regulation and RNA splicing. Pancreatic ductal adenocarcinoma cells expressing pro-oncogenic mRNA stability factor HuR (ELAVL1) show sensitization to oxaliplatin and 5-fluorouracil through persistent PARylation [37]. Silencing of PARG by shRNA decreased the proliferation rate twofold over wild type LoVo colon cancer cells [38]. PARG function was also suggested to be involved in MAP kinase cascade regulation, as implicated by the synthetic lethal effect under *PARG* and *DUSP22* double dysfunction [39]. The mechanism for delayed tumor formation process under *Parg* deficiency in ES cells should be further analyzed from multiple physiological aspects.

**Figure 5.** *PARG* mutation frequencies in human cancers. (**A**) The percentage of nonsynonymous mutation of *PARG* in each cancer in the TCGA database [40]. (**B**) The percentage of nonsynonymous substitution mutation of *PARG* in each cancer (total 165 mutations were listed in tumor samples from 8018 patients.) in the CanSAR database [41].

#### **4. Materials and Methods**

#### *4.1. Cell Culture*

The cell lines used were cultured at 37 ◦C in a humidified incubator, where CO2 levels were kept at 5%. ES cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies Corp., Carlsbad, CA, USA) supplemented with 20% fetal bovine serum, non-essential amino acids (Life Technologies Corp.), 55 μM β-mercaptoethanol, 0.3 mM each of adenosine, guanosine, and thymidine, 0.1 mM uridine, and 1 <sup>×</sup> 10<sup>3</sup> U/mL mouse LIF (Chemicon International Inc., Temecula, CA, USA) on gelatin-coated dishes (Asahi Glass Co. Ltd., Tokyo, Japan). Mouse ES cell J1 established from 129Sv/J mouse was provided by Dr. Ochiya of National Cancer Center Research Institute [42]. Hypomorphic two *Parg*−/<sup>−</sup> ES cell clones, D122 and D79, which we previously established [7], were used in this study. MMS (Sigma-Aldrich Co., St. Louis, MO, USA) was prepared in saline before use and sterilized by filtration.

#### *4.2. Tumorigenesis Analysis*

ES cells (J1, D122 and D79) cultured in the absence of a STO feeder layer [42] were trypsinized and <sup>1</sup> <sup>×</sup> 107 cells (Figures <sup>1</sup> and 3) or 5 <sup>×</sup> 106 cells (Figure 4) were subcutaneously injected into both legs of male BALB/c-nu/nu mice (CLEA Japan, Tokyo). The growth of tumors was measured continuously every 3–4 days. Four weeks after injection of ES cells, mice were euthanized and each tumor was histologically analyzed. The tumor volumes were calculated with the following formula: (smallest diameter)2 <sup>×</sup> (largest diameter)/2. For the time course experiment, mice were euthanized at 7 and 14 days after subcutaneously injection of ES cell. Local X-ray irradiation of the tumors on hind legs were carried out with an X-ray machine (CP-160, 160-kVp, Faxitron X-Ray Corp., Tucson, AZ, USA) using radiation shield of lead. Tumors of one of the hind leg pair were subjected to local X–irradiation with a single dose at 7 Gy (Figure 4, A and C: *n* = 7 for wild-type (J1), *n* = 8 for *Parg*−/<sup>−</sup> (D79). B and D: *n* = 9 for wild-type (J1) and *Parg*−/<sup>−</sup> (D122)). Tumor growth of non-irradiation controls (Control) was monitored for non-irradiated side of hind leg pair. All animal experiments were approved by the Institutional Animal Experiment Committee of National Cancer Center Research Institute (T05-053-C01, A377-16, T17-053). All animal works were conducted according to relevant national and international guidelines for animal welfare.

#### *4.3. Histological Analysis*

After resection of the tumors, they were fixed in neutralized 10% formalin solution and embedded in paraffin blocks using standard procedures. Paraffin sections were stained with hematoxylin-eosin (HE), and histopathological analysis was performed under a light microscopic observation. Tissue sections mounted on slides were also subjected to immunostaining after deparaffinization and rehydration, and antigen retrieval, according to the manufacturer's protocol. Antibodies used in this study were anti-beta-III-tubulin (Abcam, ab264113, Cambridge, UK), anti-AFP (α-fetoprotein, Pierce, PA5-21004, Shreveport, LA, USA), anti-TRA-1-60 (Santa Cruz, sc-21705, Dallas, TX, USA), Brachyury (Santa Cruz, sc-374321, Dallas, TX, USA) and anti-PAR (BD Pharmingen, Franklin Lakes, NJ, USA). After several washes with PBS, bound antibodies were visualized using 3, 3 -diaminobenzidine according to the manufacturer's protocol. The sections were counterstained with hematoxylin and mounted.

#### *4.4. Statistical Analysis*

Statistical analysis was carried out by the One-way ANOVA test, Mann–Whitney *U* tests, Turkey's test and Student *t*-tests using either SPSS Statistics of Macintosh version (IBM corporation, Armonk, NY, USA), JMP (SAS Institute Inc. Cary, NC, USA) or GraphPad Prizm7 (GraphPad Software Inc., San Diego, CA, USA).

#### **5. Conclusions**

In conclusion, our results demonstrate that *Parg* deficiency delayed the onset of tumor formation of ES cells and augmented anti-tumor therapeutic effects of DNA-damaging agents, including alkylating agents and x- ray irradiation. These results suggest that the inhibition of PARG is likely to be an option for the therapeutic and prophylactic target of cancer control. Some specific inhibitors for PARG have been reported [43–45], although clinical studies for therapeutic agents are yet to take place. PARG inhibitors, in combination with DNA-damaging agents, may efficiently suppress tumor growth in particular types of germ cell tumors.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/4/1056/s1, Figure S1: Hematoxylin-eosin staining of tumor tissues derived from mouse ES cells 4 weeks after injection.

**Author Contributions:** Conceptualization, M.M.; Methodology, A.G., H.S., Y.S. (Yuka Sasaki), T.O., S.I., M.M.; Validation, A.G., H.S., Yuka Sasaki, A.-M.R., T.A., T.O., S.I., M.W., M.M.; Investigation, Y.S. (Yuki Sonoda), Y.S. (Yuka Sasaki), A.G., H.S., T.A., T.O., S.I., M.W., M.M.; Resources, J.I., K.A., K.N., M.M.; Data curation, Y.S. (Yuki Sonoda), Y.S. (Yuka Sasaki), A.G., H.S., A.-M.R., T.A., S.I., M.M.; Writing—Original draft preparation, Y.S. (Yuki Sonoda), Y.S. (Yuka Sasaki), T.A., T.O., S.I., M.M.; Writing—Review and editing, T.H., J.I., K.A., K.N.; Visualization, Y.S. (Yuki Sonoda), Y.S. (Yuka Sasaki), M.M.; Supervision, M.M., T.H., K.N.; Project administration, M.M.; Funding acquisition, K.A., Y.S. (Yuka Sasaki), M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded in part by a Grant-in-Aid for the Second Term Comprehensive 10-Year Strategy for Cancer Control, a Grant-in-Aid for Cancer Research from the Ministry of Health, Labor and Welfare of Japan, a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, Sports, and Culture of Japan (15025274) and Grant-in-Aids from Foundation for Promotion of Cancer Research and Taiju Life Social Welfare Foundation.

**Acknowledgments:** We are grateful to Toshio Imai, Tadashige Nozaki, Shozo Takayama, M. Yanagihara, Hitoshi Nakagama, and Takashi Sugimura for valuable suggestions. We thank for technical assistance by Hideki Ogino and Masako Ichishi.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Exploring the Synergy between PARP and CHK1 Inhibition in Matched** *BRCA2* **Mutant and Corrected Cells**

#### **Hannah L Smith 1, Lisa Prendergast <sup>2</sup> and Nicola J Curtin 1,\***


Received: 31 January 2020; Accepted: 31 March 2020; Published: 4 April 2020

**Abstract:** PARP inhibition results in the accumulation of DNA SSBs, causing replication stress (RS) and lesions that can only be resolved by homologous recombination repair (HRR). Defects in HRR, e.g., due to *BRCA2* mutation, confer profound sensitivity to PARP inhibitor (PARPi) cytotoxicity. In response to RS, CHK1 is activated to signal to S and G2/M cell cycle checkpoints and also to HRR. To determine the relative contribution of these two functions of CHK1 to survival following PARPi exposure, we investigated the effects of rucaparib (a PARPi) and PF-477736 (a CHK1 inhibitor) alone and in combination in cells with mutated and corrected *BRCA2*. The *BRCA2* mutated V-C8 cells were 1000× more sensitive to rucaparib cytotoxicity than their matched *BRCA2* corrected V-C8.B2 cells, but no more sensitive to PF-477736 despite having seven-fold higher levels of RS. PF-477736 caused a five-fold enhancement of rucaparib cytotoxicity in the V-C8.B2 cells, but no enhancement in the V-C8 cells. This differential sensitivity was not due to a difference in PARP1 or CHK1 expression or activity. PF-477736 increased rucaparib-induced RS (γH2AX foci) and completely inhibited RAD51 focus formation, indicating a profound suppression of HRR. Our data suggested that inhibition of HRR was the main mechanism of sensitisation to rucaparib, compounded with an inhibition of cell cycle checkpoints by PF-477736.

**Keywords:** PARP; ATR; CHK1; replication stress; homologous recombination DNA repair; cell cycle; cytotoxicity

#### **1. Introduction**

Replication stress (RS) is a key source of genomic instability, an enabling characteristic of cancer [1]. RS is increased in cancer due to an almost ubiquitous loss of G1 checkpoint control [2] and unrepaired DNA lesions encountering the replication fork. Poly(ADP-ribose) polymerase (PARP) is a key component of the DNA damage response by promoting DNA base excision repair/single-strand break repair (BER/SSBR) [3]. PARP is the first line of defence against the most common type of endogenous DNA damage, oxidative stress, a major contributor to RS. PARP inhibitors (PARPi) are a significant breakthrough in the treatment of cancer by exploiting cancer-specific defects in homologous recombination DNA repair (HRR), e.g., due to *BRCA* mutations. Toxicities associated with these drugs are generally mild [4]. Three PARPi are currently approved for the treatment of ovarian cancer, the success being largely due to the high frequency (>50%) of HRR defects in this cancer type [5–7].

The high levels of RS and loss of G1 control make cancer cells dependent on S and G2/M cell cycle checkpoint control [8]. Checkpoint kinase 1 (CHK1) is a pivotal checkpoint kinase signalling RS to cell cycle arrest through inactivation of cdc25A and cdc25C. Cdc25A and cdc25C are phosphatases that remove inactivating phosphates on CDK2 and CDK1, respectively. Since CDK2 is required for S-phase entry and progression and CDK1 is needed for mitosis, activation of CHK1 leads to S and G2/M arrest. CHK1 has also been shown to phosphorylate RAD51 and thus has key involvement in signalling to HRR, as well as halting the cell cycle to allow repair to occur [9,10].

CHK1 inhibitors have the potential to counteract HRR-mediated PARPi resistance [11]. Indeed, PARPi and CHK1 inhibitors have been shown to interact to cause increased cytotoxicity in breast and ovarian cancer cells, which was mediated by inhibition of HRR and increased DNA damage [12,13]. However, to date, no investigations have been carried out in paired HRR competent and HRR defective (HRD) cell lines to confirm this as the mechanism.

To better understand the mechanisms underlying the synergy between PARP and CHK1 inhibitors, we used paired *BRCA2* mutant (V-C8) and corrected (V-C8.B2) cells. We examined the effects of the clinically approved PARPi, rucaparib, and the CHK1 inhibitor, PF-477736, that has undergone clinical evaluation (NCT00437203) on target enzyme activity and inhibition, cell cycle control, DNA repair, and cytotoxicity. Our data suggest that CHK1 inhibition results primarily in an HRD phenotype, and this is synthetically lethal with PARP inhibition.

#### **2. Results**

#### *2.1. V-C8 Cells Are More Sensitive to Rucaparib, but not PF-477736, and PF-477736 only Sensitised V-C8.B2 Cells to Rucaparib*

Colony formation assays were used to determine the potency of rucaparib across V-C8 and V-C8.B2 cell lines. As expected, the HRD V-C8 cells were particularly sensitive to rucaparib (LC50 < 0.01 μM) and significantly more sensitive compared to matched HRR-competent V-C8.B2 cells (LC50 > 10 μM, *p* < 0.001) (Figure 1a). In contrast, no significant difference in cytotoxicity to PF-477736 (Figure 1b) was observed between the cell lines as both V-C8 and V-C8.B2 cells had similar LC50 (100.9 and 87.5 nM, respectively). This suggested that *BRCA2*/HRD status was not a determinant of sensitivity to CHK1 inhibitor PF-477736. Survival in V-C8 and V-C8.B2 cell was inhibited by less than 50% (44.0 ± 5.9 and 41.9 ± 3.5 %, respectively) at 50 nM PF-477736, so this concentration was used for further studies.

We next tested whether the CHK1 inhibitor could potentiate PARPi in HRR competent and defective cells. The survival of cells was evaluated when exposed to a range of rucaparib concentrations (V-C8, 0–0.3 μM, V-C8.B2, 0–30 μM, to account for increased sensitivity to rucaparib) with or without 50 nM PF-477736. In V-C8.B2 cells, co-incubation with PF-477736 decreased the LC50 of rucaparib 4.8-fold ± 2.7 (Figure 1c). PF-477736 did not sensitise HRD V-C8 cells to rucaparib (Figure 1d). This differential sensitisation of HRR functional and dysfunctional cells suggested PF-77736 was sensitising HRR competent cells by inhibiting HRR.

**Figure 1.** Differential cytotoxicity and synergy of rucaparib and PF-477736 in *BRCA2* mutant and corrected cells. V-C8 and V-C8.B2 cells were exposed to drugs at the indicated concentration for 24 h prior to replacement with drug-free medium for 7–10 days to allow colony formation. (**a**) Rucaparib, (**b**) PF-477736, (**c**) the combination of rucaparib with 50 nM PF-477736 in V-C8 B2 cells, and (**d**) the combination of rucaparib with 50 nM PF-477736 in V-C8 cells. Data are the mean and standard error of three independent experiments.

#### *2.2. The Di*ff*erence in Rucaparib Sensitivity Is not Due To Di*ff*erential PARP-1 Expression between Cell Lines*

To exclude differences in PARP1 expression, activity, or inhibition as factors contributing to the differential sensitisation to rucaparib by PF-477736, we measured PARP expression and activity in both cell lines. PARP1 levels were not substantially different between cell lines and were only modestly affected by rucaparib and PF-477736 (Figure 2a and Figure S1). Endogenous PAR levels were 2.3-fold higher in the HRD V-C8 cells compared to the V-C8.B2 cells (Figure 2b), possibly reflecting a higher level of endogenous DNA breakage activating PARP. There was no significant difference between the baseline PARP activity of both cell lines (Figure 2c). Rucaparib inhibited PARP activity to a similar extent in both V-C8 and V-C8.B2 cells with IC50 values of 59 and 53 nM, respectively (Figure 2d). In both cell lines, a concentration of 0.1 μM rucaparib inhibited PARP activity >95%. This concentration only killed 3% of V-C8.B2 cells compared with >99% of V-C8 cells, demonstrating that the differential cytotoxicity was determined by HRR status rather than the extent of PARP inhibition.

 (a)

(c)

#### (d)

**Figure 2.** PARP-1 expression, activity, and inhibition by rucaparib in *BRCA2* mutant and corrected cells. (**a**) V-C8 and V-C8.B2 cells were exposed to drugs at the indicated concentration for 24 h prior to Western blot. Chemiluminescence was quantified using Syngene software. Bar charts are PARP-1 expression relative to the GAPDH loading control. Data are from a single representative experiment. (**b**) Endogenous PAR was measured in the absence of oligonucleotide and NAD+ by reference to a standard curve. Data are the mean and standard deviation of three independent experiments. (**c**) PARP activity (mean pixel values) was measured in the presence of oligonucleotide and NAD+. Data are the mean and standard deviation of three individual experiments. (**d**) PARP activity following exposure to increasing concentrations of rucaparib. Data are the mean and standard error of three individual experiments.

#### *2.3. Rucaparib Activates CHK1 Similarly in Both Cell Lines, and This Activation Is Inhibited by PF-477736*

Another possible explanation for the differential sensitisation by PF-477736 could have been a difference in the activation of CHK1 by rucaparib and a difference in the extent to which it was inhibited by PF-77736. The expression of CHK1 in the native and phosphorylated forms and downstream signalling to CDK1 were determined by Western blot (Figure 3a and Figure S2). CHK1 expression was slightly lower in V-C8 cells compared to V-C8.B2 cells, but this was not significant (Figure 3b). Rucaparib (10 μM) activated checkpoint signalling through ATR and CHK1 in both cell lines (Figure 3a). CHK1S345 levels (indicating ATR activation) were increased 2.1-fold and 1.6-fold by rucaparib in V-C8 and V-C8.B2 cells, respectively. CHK1 activity (CHK1S296) was increased by rucaparib to a similar extent in V-C8 and V-C8.B2 cells (1.5-fold and 1.6-fold, respectively) (Figure 3c). In both cell lines, PF-477736 (50 nM) inhibited this activation of CHK1, by 91.4% ± 12.2 in V-C8 cells and 75.6% ± 17.6 in V-C8.B2 cells. The difference in inhibition of CHK1S296 by PF-477736 was not statistically significantly different between the two cell lines, suggesting this did not contribute towards the difference observed in sensitisation. Downstream of CHK1, PF-477736 similarly inhibited phosphorylation of CDK1 (CDK1Y15) by 69.6% ± 24.8 in V-C8 cells and 56.88 % ± 26.1 in V-C8.B2 cells. Upstream of CHK1, PF-477736 activated ATR 2.1-fold ± 1.3 in V-C8 cells and 2.4-fold ± 1.2 in V-C8.B2 cells, suggesting an increased reliance on ATR when CHK1 was inhibited.

**Figure 3.** CHK1 expression, activation by rucaparib, and inhibition by PF-477736 in *BRCA2* mutant and corrected cells. Cells were exposed for 24 h to DMSO, rucaparib, or rucaparib and PF-477736 and analysed for pCHKS345, CHK1, and pCDK1Tyr15. Vinculin was used as a loading control for pCHKS345, CHK1, pCDK1Tyr15 and CDK1, and GAPDH was used as the loading control for CHK1S296 (**a**) Western blot from a single representative experiment. (**b**) Densitometry was calculated using Syngene software of CHK1 expression normalised to Vinculin. (**c**) Normalised densitometry of phosphorylated CHK1 and CDK1. Data are the mean and standard deviation of three individual experiment.

#### *2.4. Rucaparib Causes S-phase and G2 Accumulation, Which Is Attenuated by PF-477736*

The effect of rucaparib and PF-477736 on the cell cycle was analysed to assess if the mechanism of sensitisation was by cell cycle checkpoint signalling. In V-C8.B2 cells, rucaparib caused a modest increase in S-phase (19.31%). Co-exposure to PF-477736 reduced the S-phase by 64.5% and G2/M by 33.6% compared to control cells, and this was accompanied by an increase in the sub-G1 fraction of 21.5%, suggesting that cells were progressing through S and G2/M with damaged DNA and dying, probably by mitotic catastrophe (Figure 4a).

#### *2.5. Rucaparib Inhibition Leads to Increased DNA Damage and HRR, and PF-477736 Inhibits HRR*

The induction of replication stress by rucaparib, resulting in collapsed replication forks following accumulation of endogenous DNA damage, is commonly measured by γH2AX phosphorylation. At baseline, untreated V-C8 cells had on average seven-fold more γH2AX foci/cell, compared to V-C8.B2 cells, reflecting their inability to resolve RS through HRR (Figure 4b). In V-C8.B2 cells, rucaparib increased γH2AX foci/cell 16-fold (*p* < 0.0001), PF-477736 caused a three-fold increase in γH2AX foci/cell (*p* < 0.0001), and the γH2AX foci/cell following the combination were higher still, but not significantly different from rucaparib alone.

In V-C8 cells, the already high levels of γH2AX foci were only increased by 20% by PF-477736 and 30% by rucaparib (Figure 4b, not significant). The combination increased the number of foci approximately three-fold (*p* < 0.0001). Due to the *BRCA2* mutation in these cells, they had very low levels of RAD51 foci, which were not increased by rucaparib or PF-477736 alone or in combination.

In marked contrast, in parallel with the induction of DNA breaks (γ-H2AX) in V-C8.B2 cells, rucaparib exposure caused an approximately 13-fold increase (*p* < 0.0001) in RAD51 foci in γ-H2AX-positive cells. Importantly, PF-477736 completely abrogated the rucaparib induction of RAD51 (significantly different from rucaparib alone, *p* < 0.0001), and the levels of foci were similar to control V-C8.B2 and the HRD V-C8 cells. These data indicated that the inhibition of CHK1 totally abolished HRR and led to a greater accumulation of collapsed replication forks.

**Figure 4.** *Cont.*

**Figure 4.** Impact of PF-477736 on cell cycle and homologous recombination repair (HRR) in rucaparib-treated cells. Cells were exposed for 24 h to DMSO, rucaparib, or rucaparib with PF-477736 at the indicated concentrations. (**a**) Cell cycle analysis of V-C8.B2 cells; data are from a single experiment. (**b**) Representative images are shown above quantified γH2AX and RAD51 foci, where each data point represents a single nucleus. RAD51 foci were only counted in cells with >mean γH2AX foci in untreated cells. Data are pooled from three independent experiments.

#### **3. Discussion**

Defects in either *BRCA1* or *BRCA2* have been known to result in a profound sensitivity to PARP inhibition since 2005 [14,15]. Rucaparib has previously been documented to be ≥1000× more cytotoxic to *BRCA2* mutant V-C8 cells compared to their matched *BRCA2* corrected V-C8.B2 cells (Patent WO/2005/012305), which we confirmed here. Interestingly, *BRCA2* was not a determinant of PF-477736 sensitivity. RS was thought to be a determinant of sensitivity to cell cycle checkpoint inhibitors, so this was somewhat surprising given that untreated V-C8 cell had seven-times more H2AX foci than V-C8.B2 cells. In previous studies, V-C8 cells were more sensitive to the ATR inhibitor VE-821 [16], possibly RS being a greater determinant of sensitivity to ATR than CHK1 inhibition.

Here, we confirmed that the observed difference in rucaparib sensitivity was not due to different PARP-1 expression levels or activity between the cell lines. This contrasts with previous reports of substantially higher levels of PARP activity in V-C8 compared to V-C8.B2 cells [17]. This may reflect differences in the method of detecting PARP activity. In our study, we used a GCLP-validated assay that has been used as a pharmacodynamic biomarker in PARPi clinical trials [18]. However, like this previous study, endogenous PAR levels were around two-fold higher in *BRCA2* mutant cells than wildtype, possibly indicating higher levels of endogenous DNA damage. In support of this hypothesis, there was approximately seven-times as much γH2AX foci formation in untreated *BRCA2* mutant V-C8 cells compared to *BRCA2* corrected V-C8.B2. The differential rucaparib sensitivity observed in cytotoxicity experiments was clearly due to *BRCA2* mutation and the synthetically lethal relationship between BER and HRR pathways. Therefore, it is interesting that PF-477736 only sensitised V-C8.B2 cells and not V-C8 cells to rucaparib, suggesting the CHK1 inhibitor was primarily acting on HRR.

We demonstrated that rucaparib induced RS to a higher extent in V-C8.B2 cells; however, this was most likely due to the high levels of RS in V-C8 control cells. CHK1 was activated to a similar extent by rucaparib in both V-C8 and V-C8.B2 cells, and PF-477736 caused a similar inhibition in both cell lines, so the differential sensitisation observed was not due to a greater activation or inhibition of the pathway. Rucaparib caused a similar increase in pCDK1Y15, which was inhibited to a similar degree by PF-47736 in both cell lines, indicating that the downstream checkpoint signalling was intact and equally responsive to the two inhibitors in both cell lines. Therefore, this was not a factor in the differential sensitisation of VC8.B2 cells. Increased CHKS345 phosphorylation in both V-C8 and V-C8.B2 cells was also observed in response to CHK1 inhibition with PF-477736, suggesting upstream ATR activation as previously described [19,20].

Cell cycle analysis showed only a modest S-phase accumulation caused by rucaparib, probably associated with the increase in RS that was demonstrated by the increase in γH2AX (Figure 4). PF-477736 not only abolished this rucaparib-induced S-phase arrest, but also reduced both S- and G2-phases relative to untreated controls. This reduction in S and G2 was accompanied with a corresponding increase in the sub-G1 fraction, suggesting damaged cells were forced into mitosis before they could undertake repair, as has been reported for other PARP and CHK1 inhibitor combinations [21,22].

Although these data indicated that inhibition of cell cycle checkpoints by PF-477736 made a contribution to the cytotoxicity of rucaparib, this is unlikely to explain the differential sensitisation of the *BRCA2* corrected compared to the *BRCA2* mutant cells. We therefore tested the hypothesis that the primary mechanism of sensitisation was via PF-477736-mediated inhibition of HRR. As predicted, PF-477736 profoundly inhibited HRR, completely abolishing RAD51 foci formation. In a study of breast cancer cell lines, prexasertib synergised with the PARPi olaparib by causing S-phase arrest and inhibiting HRR/RAD51 foci [23]. In contrast, in high-grade serous *BRCA*-wildtype ovarian cancer cells, prexasertib synergised with olaparib by inhibiting G2/M arrest, as well as also inhibiting olaparib-induced RAD51 foci formation [13]. Clearly, the various molecular pathologies of these cells complicate the interpretation of the data.

#### **4. Materials and Methods**

#### *4.1. Chemicals and Reagents*

All chemicals and reagents used were obtained from Sigma-Aldrich, unless stated otherwise. Rucaparib was gifted from Pfizer Global R&D. Both rucaparib and PF-477736 (Selleckchem, Houston, TX, USA) were dissolved in dry DMSO at respective concentrations of 20 mM and 5 mM and stored at −80 ◦C before use.

#### *4.2. Cell Culture*

VC-8 and VC-8.B2 cells, a gift from Malgorzata Zdzienica, Leiden University [24], were maintained in exponential growth phase (<70–80% confluence) in DMEM with 10% foetal bovine serum (FBS; Thermo Fisher Scientific, Waltham, MA, USA) and incubated at 37 ◦C, 5% CO2, and 95% humidity. VC8.B2 cells contained a BAC containing the *BRCA2* gene and maintained under selection with 200 μg/mL G-418 (Thermo Fisher Scientific, Waltham, MA, USA). Cells were mycoplasma free.

#### *4.3. Cytotoxicity Assay*

Exponentially growing cells were seeded at various densities in 6 well plates, allowing 3 different densities for each drug concentration estimated to give 20–200 colonies following drug treatment. After attachment, cells were exposed to various concentrations of rucaparib or PF-477736 alone or the combination of rucaparib ± 50 nM PF-477736 in DMSO or DMSO alone at a final concentration of 0.5% for 24 h. The medium was replaced with fresh medium and the cells left to incubate for 7–10 days to form colonies. The colonies were fixed in methanol:acetic acid (3:1) stained with 0.4% crystal violet, and colonies of >30 cells were counted by eye. Graphs were plotted using Graphpad Prism 6 software (San Diego, CA, USA).

#### *4.4. Measurement of PARP Activity and Inhibition*

Following incubation with rucaparib (0–10 μM) for 30 min, PARP activity was measured in permeabilised cells at 26 ◦C in the presence of 350 nM NAD+ and 10 mg/mL 12 mer palindromic oligonucleotide to activate the enzyme, as described previously [25,26]. Following transfer to a nitrocellulose membrane (GE Healthcare Life, Sciences, Amersham, Buckinghamshire, UK) the product, PAR, was measured with 10H anti-PAR monoclonal antibody (Enzo life sciences, Exeter, UK) overnight at 4 ◦C and HRP-conjugated goat anti-mouse secondary antibody (Dako, Santa Clara, CA, USA) diluted 1:1000 in PBS-Tween (PBS-T) 5%. Clarity Max ECL Western substrate (Bio-Rad, Hercules, CA, USA) was added to the membrane, chemiluminescence imaged using the GBox and Genesys software (Syngene, Cambridge, UK), and quantified with reference to a standard curve of 0–25 pmol purified PAR (Enzo life sciences, Farmingdale, NY, USA) after subtraction of background reactions in the absence of NAD and oligonucleotide. Baseline PAR levels were measured using 50–100-times the cells used for inhibition experiments and with no oligonucleotide or NAD+ present. Purified PAR standard was diluted accordingly to a concentration of 0–25 pmol.

#### *4.5. Measurement of CHK1 Activation and Inhibition by Western Blot*

Exponentially growing cells in 100 mm dishes were exposed to media containing DMSO, 10 μM rucaparib, or 10 μM rucaparib and 50 nM PF-477736 for 24 h before extraction with 250 μL per dish of phosphosafe extraction reagent with protease cocktail inhibitor (Thermo Fisher Scientific, Waltham, MA, USA) at a 1:100 dilution per dish at 4 ◦C for 5–7 min, then scraped on ice into Eppendorf tubes. Following centrifugation at 8000× *g* for 10 min at 4 ◦C, the protein concentration of the supernatants was measured using a BCA protein assay kit (Thermo Fisher Scientific, Waltham, MA, USA) and diluted with deionised water to 0.5–1 mg/mL. XT sample buffer (Bio-Rad, Hercules, CA, USA) and XT reducing agent (Bio-Rad, Hercules, CA, USA) were added at constants of 25% and 0.5%, respectively. Samples were heated at 90 ◦C for 5 min, then 30 μg were loaded/well of 3–8% Criterion XT tris-acetate gels (Bio-Rad, Hercules, CA, USA), and the gel ran at 150V for approximately 1 h alongside HiMark pre-stained protein standard (Thermo Fisher Scientific, Waltham, MA, USA) in diluted 20× XT tricine with deionised water. The separated proteins were transferred onto a nitrocellulose membrane (Amersham, Buckinghamshire, United Kingdom) at 100 V for 1 h. The membrane was blocked for 1 h at room temperature in 5% milk in TBS-tween (TBST). Primary antibodies PARP-1 (#ab227244 Abcam, Cambridge, UK), vinculin (#4650), GAPDH (14C10 #2118), phospho-CHK1 (Ser296) (D3O9F #90178), phospho-Chk1 (Ser345) (#2341), and phospho-CDK1 (Tyr15) (10A11 #4539) were each diluted 1:1000 in TBS-Tween (TBS-T) and 5% bovine serum albumin (BSA) and left to incubate overnight at 4 ◦C (all obtained from Cell signalling, Massachusetts, United States except PARP-1). The membrane was cut into sections (approximately 30–45 kDa, 45–70 kDa, and 70–220 kDa), washed in TBS-T, before adding anti-rabbit goat polyclonal HRP secondary antibody (Cell Signaling Technology, Danvers, MA, USA) diluted 1:2000 in 5% skimmed milk in TBST at room temperature for 1 h. After washing with TBST, Clarity Max ECL Western substrate (Bio-Rad, Hercules, CA, USA) was applied to the membrane slices and chemiluminescence measured using Syngene software on the G-box.

#### *4.6. Cell Cycle Analysis*

Exponentially growing cells were exposed to DMSO, 10 μM rucaparib or 10 μM rucaparib with 50 nM PF-477736 for 24 h. Cells were washed twice with PBS, with each washing collected to ensure no loss of cells, before being trypsinised and harvested. Following centrifugation at 1500 rpm for 5 min, the supernatant was discarded, and the remaining cell pellet was resuspended in 1 mL ice cold PBS and centrifuged (3000 rpm, 5 min). The supernatant was removed, and 1 mL of 70% ethanol was added dropwise to the cell pellet. Samples remained at 4 ◦C for a minimum of 1 h. Prior to staining, cells were washed twice in PBS to remove ethanol before eventually being resuspended in 700 μL PBS. RNase was added to cells (12 μL of 1 mg/mL stock) alongside propidium iodide stain (20 μL of 1mg/mL stock). Cells were incubated in dark conditions at 37 ◦C for a minimum of 30 min. Analysis was carried out using Attune Nxt Cytometer. De Novo FCS Express 7 software was used to analyse the data.

#### *4.7. Homologous Recombination Repair Assay*

Cells were seeded at 0.5 <sup>×</sup> 10<sup>5</sup> cells/mL onto coverslips and incubated for 24 h before being exposed to either DMSO, 50 nM PF-477736, 10 μM rucaparib, or 10 μM rucaparib with 50 nM PF-477736 for 24 h. Cells were then washed with PBS and fixed with ice cold methanol for a minimum of 1 h at −20 ◦C. Coverslips were then washed with PBS 0.2% Triton-X-100 and blocked (2% BSA, 10% skimmed milk powder (Marvel, UK), 10% goat serum) for 1 h at room temperature before

being incubated with anti-RAD51 (Abcam, Cambridge, UK) diluted 1:500 in blocking buffer at 4 ◦C overnight. After washing in PBS, 0.2% Triton X-100 anti-phospho-histone H2AX (Merck, Branchburg, NJ, USA) was added to coverslips at 1:1000 and incubated for 1 h. After washing, secondary antibodies, Alexa Fluor 488 and Alexa Fluor 546 (ThermoFisher Scientific, Waltham, MA, USA), were added and incubated for 1 h, in dark conditions. Coverslips were washed then exposed to 0.5 μg/mL DAPI solution. Coverslips were mounted onto microscope slides using H-1400 hard set mounting media (Vector Laboratories, Burlingame, CA, USA). Slides were stored in dark conditions and left to dry prior to imaging. Cells were imaged using the Leica DM6 LED fluorescence microscope (Leica microsystems GmbH, Wetzlar, Germany). Images were analysed using Fiji ImageJ software [27]. In DMSO-treated cells, the mean γH2AX foci were calculated. Cells (control and treated) with more than the mean number of γH2AX foci in control cells were deemed γH2AX positive. RAD51 foci were only counted in γH2AX positive cells.

#### **5. Conclusions**

On the basis of our data in matched cell lines that differed only in their BRCA2 status, we proposed the model shown in Figure 5. That is, the RS caused by the accumulation of unrepaired SSB when PARP was inhibited led to collapse of replication forks. These lesions triggered activation of the ATR-CHK1 pathway that signals to cell cycle arrest, but the major contribution that CHK1 makes was to protect cells from PARPi cytotoxicity by promoting HRR. The use of a CHK1 inhibitor may therefore be useful not only for tumours that are intrinsically PARPi resistant due to functional HRR, but may also overcome acquired resistance to PARPi in BRCA mutant tumours that have restored HRR function. Recent studies with ovarian cancer PDX models demonstrated that the combination of the CHK1 inhibitor prexasertib with olaparib caused greater tumour growth delay and survival in both olaparib sensitive and resistant tumours [28]. Our data along with this PDX study may help in the interpretation of the results of clinical trials of PARPi-CHK1 inhibitor combinations, such as NCT03057145, in which olaparib is being investigated in combination with prexasertib.

**Figure 5.** The DNA damage response, highlighting cross-talk between the cell cycle and DNA repair. (**a**) Base excision repair (BER) is the first line of defence when DNA damage occurs. PARP activation by SSB recruits XRCC1, Pol β, and Lig III to repair the DNA. (**b**) When PARP is inhibited, SSB accumulate, causing stalled replication forks, and single-ended DSBs, causing fork collapse. BRCA2 facilitates replacing RPA with repair protein RAD51, which subsequently forms a filament to search for the specific

homologous sequence on the sister chromatid as a template for repair and re-start. PARPi-induced replication stress activates ATR, which initiates a signalling cascade. ATR activates CHK1 by phosphorylation at serine 345, causing CHK1 to autophosphorylate at serine 296, to achieve full activation. CHK1 inactivates CDC25A/C, thereby preventing the removal of inhibitory phosphorylation on CDK2 and 1, respectively, thus preventing S-phase progression and mitosis. The pathway also promotes HRR as CHK1 activates BRCA2 and RAD51 by phosphorylation, and ATR also phosphorylates RAD51.

#### **6. Patents**

Helleday T and Curtin NJ. Therapeutic Compounds (PARP inhibitors in homologous repair/BRCA defective cancer) Patent Application Number PCT/GB2004/003183. Publication numberWO 2005/012305 A2 Divisional application 16th April 2004 GB 0408524. WO2005012305A3.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/4/878/s1: Figure S1: PARP1 expression in V-C8 and V-C8.B2 cells, Figure S2: CHK1 expression, activation by rucaaprib and inhibition by PF-477736 in V-C8 and V-C8.B2 cells.

**Author Contributions:** H.L.S. produced the data and wrote the manuscript; L.P. directed the research and edited the manuscript; N.J.C. conceived the project, directed the research and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was partly funded by the JGW Patterson foundation.

**Conflicts of Interest:** N.J.C. contributed to the development of rucaparib and obtained research funding from Agouron Pharmacurtical, Pfizer, and Clovis. By virtue of the two active patents and an original agreement between Cancer Research Technology, Newcastle University, and Agouron, she is in receipt of royalty payments, which she does not take personally; in the past, they have contributed to her research accounts, and a recent large sum has been used to set up a charitable fund within the local Community Foundation. https://www.communityfoundation. org.uk/group\_grant/passionate-about-realising-your-potential/.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **Synergistic Anti-Tumour E**ff**ect of Syk Inhibitor and Olaparib in Squamous Cell Carcinoma: Roles of Syk in EGFR Signalling and PARP1 Activation**

### **Duen-Yi Huang 1, Wei-Yu Chen 2,3, Chi-Long Chen 3,4, Nan-Lin Wu 5,6,7 and Wan-Wan Lin 1,8,\***


Received: 20 December 2019; Accepted: 17 February 2020; Published: 19 February 2020

**Abstract:** Syk is a non-receptor tyrosine kinase involved in the signalling of immunoreceptors and growth factor receptors. Previously, we reported that Syk mediates epidermal growth factor receptor (EGFR) signalling and plays a negative role in the terminal differentiation of keratinocytes. To understand whether Syk is a potential therapeutic target of cancer cells, we further elucidated the role of Syk in disease progression of squamous cell carcinoma (SCC), which is highly associated with EGFR overactivation, and determined the combined effects of Syk and PARP1 inhibitors on SCC viability. We found that pharmacological inhibition of Syk could attenuate the EGF-induced phosphorylation of EGFR, JNK, p38 MAPK, STAT1, and STAT3 in A431, CAL27 and SAS cells. In addition, EGF could induce a Syk-dependent IL-8 gene and protein expression in SCC. Confocal microscopic data demonstrated the ability of the Syk inhibitor to change the subcellular distribution patterns of EGFR after EGF treatment in A431 and SAS cells. Moreover, according to Kaplan-Meier survival curve analysis, higher Syk expression is correlated with poorer patient survival rate and prognosis. Notably, both Syk and EGFR inhibitors could induce PARP activation, and synergistic cytotoxic actions were observed in SCC cells upon the combined treatment of the PARP1 inhibitor olaparib with Syk or the EGFR inhibitor. Collectively, we reported Syk as an important signalling molecule downstream of EGFR that plays crucial roles in SCC development. Combining Syk and PARP inhibition may represent an alternative therapeutic strategy for treating SCC.

**Keywords:** Syk; PARP; EGFR; squamous cell carcinoma

#### **1. Introduction**

The epidermal growth factor (EGF) receptor/ligand system plays key roles in essential cellular functions, including proliferation and migration. Ligand binding to the epidermal growth factor receptor (EGFR) induces the formation of receptor dimerisation and stimulates multiple pathways of signal transduction. To date, EGFR has been linked to several malignant phenotypes of human cancers, including proliferation, inflammatory response, DNA repair, therapeutic resistance, and poor clinical outcomes in patients with cancer [1,2]. Currently, there are several EGFR antagonists, including

humanised neutralising monoclonal antibodies and tyrosine kinase inhibitors clinically used in patients with non-small-cell lung [3] and metastatic colorectal cancers [4].

Spleen tyrosine kinase (Syk) is a non-receptor tyrosine kinase that is widely expressed in hematopoietic cells. Syk contains two tandem NH2 terminal Src homology 2 (SH2) domains, multiple tyrosine phosphorylation sites, and a COOH terminal tyrosine kinase domain. The SH2 domains bind phosphorylated immunoreceptor tyrosine-based activation motifs (ITAM), and hence, they couple activated immunoreceptors to multiple downstream signalling pathways [5–7]. Anomalous regulation of this kinase could lead to different allergic disorders and antibody-mediated autoimmune diseases. Thus, Syk may serve as a therapeutically relevant target for rheumatoid arthritis, asthma, psoriasis, and allergic rhinitis [8–10]. It has been believed for years that Syk function is solely linked to hematopoietic cell signalling. However, more recent studies have indicated a ubiquitous pattern of Syk gene expression. In addition to functioning as the therapeutic target of haematological malignancies [11, 12], Syk can function as either a tumour suppressor in breast, colorectal, and gastric cancers or a tumour enhancer in lung, pancreatic, ovarian, and squamous cell carcinoma (SCC) of the head and neck [13]. In cancers of non-immune cells, Syk elicits a pro-survival signal, but can also suppress tumourigenesis by restricting epithelial-mesenchymal transition, enhancing cell-cell interactions, and inhibiting migration [13]. To date, only a few studies have suggested a functional link between Syk and EGFR. In our previous study on human primary normal keratinocytes, we demonstrated that Syk is a downstream signal of EGFR and is involved in negatively regulating keratinocyte differentiation [14].

SCC is an uncontrolled growth of abnormal cells arising in the squamous cells, which compose most of the skin's upper layers (the epidermis). Similar to other cancer cell types, inhibition of EGFR activity is a potential drug regimen for SCC, which overexpresses EGFR [15–17]. A recent in vivo xenograph study suggested a potential therapy involving the use of EGFR monoclonal antibodies in SCC [18]. For better therapeutic efficacy, combination regimens are increasingly under investigation. For example, the EGFR inhibitor, gefitinib (Iressa), can enhance the apoptotic effects of cisplatin in SCC [17]. Clinical trials involving the combination regimens of anti-EGFR agents with immune checkpoint inhibitors are currently in progress for SCC of the head and neck [19,20]. In this study, we explored the role of Syk in EGFR signalling in three SCC cell lines (A431, CAL27, and SAS), analysed Syk expression in oral SCC cancer tissues, and determined the relationship between Syk activity and disease prognosis. Finally, because inhibition of poly(ADP-ribose) polymerase (PARP), a key DNA repair enzyme, could improve the efficacy of radiotherapy in human head and neck cancer [21], we determined the combined effects of the Syk and PARP inhibitors on cell viability in SCC.

#### **2. Materials and Methods**

#### *2.1. Reagents and Antibodies*

Dulbecco's Modified Eagle Medium (DMEM) and trypsin-EDTA were obtained from Gibco (Carlsbad, CA, USA). Foetal bovine serum (FBS) was procured from HyClone (Logan, UT, USA). Penicillin-streptomycin-amphotericin B solution was sourced from Biological Industries (Kibbutz Beit Haemek, Israel). Dulbecco's Phosphate Buffered Saline (PBS) was obtained from Sigma-Aldrich (St. Louis, MO, USA). Gefitinib was procured from Selleckchem (Houston, TX, USA). R406, PP2, and the p-Syk antibody were purchased from Calbiochem (San Diego, CA, USA). Olaparib was procured from Selleckchem. Recombinant human EGF was obtained from PeproTech (Rocky Hill, NJ, USA). PARP, γH2AX, p-EGFR, p-JNK, JNK, p-p38, p-ERK, p-Lyn, p-Src, p-STAT1, p-STAT3, and p-STAT5 antibodies were obtained from Cell Signaling (Beverly, MA, USA). EGFR, p38, ERK, Lyn, Fyn, Fgr, Src, Syk, STAT1, and STAT3 antibodies were purchased from Santa Cruz (Santa Cruz, CA, USA). Antibody against PAR was obtained from BD Pharmingen (San Diego, CA, USA). The β-actin antibody was procured from NOVUS (Littleton, CO, USA).

#### *2.2. Cell Culture*

Human skin SCC (A431), oral SCC (CAL27), and head and neck SCC (SAS) cell lines were cultured in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin-amphotericin B. All cell lines were incubated at 37 ◦C under a humidified atmosphere of 5% CO2 in the air. Prior to stimulation of cells with EGFR ligands or otherwise specified conditions, cells were maintained in the medium without FBS for at least 1 d.

#### *2.3. Immunoblotting*

Protein expression was determined in cell lysates by electrophoresis and immunoblotting as previously described [14].

#### *2.4. Real-tTime PCR*

The primer sequence pairs used for quantitative real-time PCR were human β-actin: 5 -CGGGGACCTGACTGACT ACC-3 and 5 -AGGAAGGCTGGAAGAGTGC-3 ; and human IL-8: 5 -ACTGAGAGTGATTGAGAGTGGAC-3 and 5 -AACCCTCTGCACCCAGTTTTC-3 . The experiment and data analysis were conducted as previously described [22].

#### *2.5. Enzyme-Linked Immunosorbent Assay (ELISA)*

For the detection of secreted proteins in the cell culture supernatant, a commercial ELISA kit for IL-8 (CXCL8) (R&D Systems, Minneapolis, MN, USA) was used to determine the level of released proteins in accordance with the manufacturer's protocol.

#### *2.6. Confocal Microscopic Analysis*

All procedures were conducted at room temperature unless otherwise noted. Cells were fixed with 4% paraformaldehyde in PBS for 20 min. After being rinsed three times with PBS, the fixed cells were permeabilised with 0.2% Triton X-100 in PBS for 20 min. After fixation, the samples were blocked with 4% BSA for 1 h and then incubated with the primary antibody for 2 h at room temperature or overnight at 4 ◦C after aspiration of the blocking solution. The primary antibody was then removed and specimens were washed three times with PBS. Subsequently, the specimens were incubated with the fluorochrome-conjugated secondary antibody for 1 h in the dark. Following immunostaining, the coverslip was counterstained with 4 ,6-diamidino-2-phenylindole (DAPI) and mounted on microscope slides in the dark. Slides were analysed using an LSM 780 confocal microscope (Carl Zeiss Micro Imaging GMbH, Jena, Germany).

#### *2.7. Crystal Violet Assay*

To evaluate the relative cell numbers, cells were seeded in 96-well plates in medium followed by the indicated treatment. After washing with PBS, cells were fixed with methanol for 10 min at room temperature and stained with 0.1% crystal violet. The relative cell number was determined by measuring the absorbance of the dissolved dye at 540 nm after elution with 33% acetic acid.

#### *2.8. Intracellular ATP Assay*

Intracellular ATP was determined by a CellTiter-Glo® Luminescent Kit (G7571, Promega Heidelberg, Germany) according to the manufacturer's protocol. After the indicated drug treatment, the reagent in the kit was added to lyse the cells, and then 150 μL solution was transferred to a 96-well plate for chemiluminescence detection using a luminometer. With background subtraction, the values were normalised to the individual control group being 100%.

#### *2.9. MTT Assay*

The viability of cells was measured by the turnover of yellow 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2*H*-tetrazolium bromide (MTT, Calbiochem) to dark blue formazan by mitochondrial reductase in living cells. A sample of 1 <sup>×</sup> 10<sup>4</sup> cells was seeded in 96-well plates for at least 16 h followed by the indicated treatment. At the end of the assay, MTT solution (5 mg/mL in PBS) was added to the medium with a 5-fold dilution for 1 h at 37 ◦C until a purple precipitate was visible. Cells in the medium were then collected and centrifuged at 800 × rpm for 5 min. After discarding the supernatant, the pellet was dissolved in DMSO. The absorbance (OD550–OD630) was detected using a spectrophotometer. The absorbance of the control group represented 100% cell viability.

#### *2.10. Cell Death Assays by Annexin V*/*PI Staining and Flow Cytometry*

After the indicated treatment, cells were collected and washed with ice-cold PBS. Cells were stained with Annexin V-FITC/PI according to the manufacturer's instructions (BioLegend, San Diego, CA, USA). After incubation, the cells were measured by flow cytometry (FACSCalibur, BD, Franklin Lakes, NJ, USA).

#### *2.11. Tissue Samples and Immunohistochemistry*

Pathology files of the Wan Fang Hospital (Taipei, Taiwan) from 2000 to 2010 were searched for SCC of the oral cavity. The pathological diagnoses were microscopically reconfirmed by pathologists (WYC and CLC). Tissue microarrays of oral SCC were manually constructed. Triplicate tissue cores of SCC and one tissue core from the adjacent non-neoplastic squamous epithelium were obtained for each case. In total, 105 patients were included in this study. The clinical records were reviewed for tumour recurrence and survival. For immunohistochemistry, the sections were deparaffinised, rehydrated, and blocked with 3% hydrogen peroxide. Heat-induced antigen retrieval was performed in a citric acid buffer (pH 6.0) at 121 ◦C for 10 min using a decloaking chamber (Biocare Medical, Concord, CA, USA). The sections were incubated with rabbit polyclonal phospho-Syk (Y525/526) antibody (Catalogue No. AP3271a, 1:300; Abgent Inc., San Diego, CA, USA) at 4 ◦C overnight. Next, the sections were incubated with a prediluted biotin-conjugated secondary antibody (Starr Trek Universal HRP Detection System; Biocare Medical) at room temperature for 30 min, followed by prediluted streptavidin-horseradish peroxidase complex at room temperature for 10 min. The antigen was detected by adding 3,3 -diaminobenzidine. The sections were then counterstained by haematoxylin. The appropriate positive and negative controls were included in these assays. The intensity of p-Syk expression in carcinoma cells was evaluated by immunostaining. The scoring of expression in tumour tissue was performed according to a four-tiered approach: negative (0), weak (1+), moderate (2+), and strong (3+) staining. Based on these criteria, p-Syk expression was defined as either low (scores 0 and 1+) or high expression (scores 2+ and 3+). Tissue samples were obtained and used according to protocols approved by the Taipei Medical University-Wan Fang Hospital Joint Institutional Review Board (approval no. 99049). The study was conducted according to the Declaration of Helsinki principles.

The clinicopathological characteristics and p-Syk expression were compared using the Chi-square test for categorical data and the two-tailed Student's t test for continuous data. The overall survival and disease-free curves were calculated using the Kaplan-Meier method, and the difference between the low and high p-Syk expression groups was evaluated using the log-rank test. The Cox proportional-hazard model was utilised to identify statistically significant clinicopathological factors affecting the prognosis of patients. A *p* value < 0.05 was considered statistically significant. All statistical calculations were performed using SPSS statistics 17.0 software (SPSS Inc. Chicago, IL, USA).

#### *2.12. Statistical Analysis*

Except immunohistochemistry data as mentioned above, all other data were presented as the mean ± standard error of the mean (S.E.M) from at least three independent experiments. Comparisons between two groups were performed using an unpaired two-tailed Student's *t* test.

#### **3. Results**

#### *3.1. SFK-Dependent Syk Activation Mediates Downstream Signal Pathways of EGFR in SCC*

After observing the crucial role of Syk in skin keratinocytes [14], we further addressed the roles of Syk in skin tumour development. To this end, we first detected the expression levels of Syk and Src family kinases (SFKs) (c-Src, Lyn, Hck, and Fgr) in A431 SCC, melanoma A375, and SK-MEL-28 cells. We found that Syk expression was highly expressed in A431 cells but not in the others. Lyn and Hck expression was also higher in A431 cells than that in melanoma cells. In contrast, c-Src expression was comparable in these cell lines, whereas Fgr was non-detectable (Figure 1A). To confirm the absence of Fgr in these cells, we used U937 monocytes as a comparison. We found that Syk and Lyn were expressed in U937 cells to the same extent as in the A431 cells; however, U937 cells expressed Fgr, but not c-Src or Hck (Figure 1A).

**Figure 1.** EGF can induce Syk activation through SFK signalling in A431 cells. (**A**) Western blot analysis of Syk and SFK expression in various cancer cell lines. (**B**) Phosphorylation of EGFR, Syk, Src, and Lyn after EGF (100 ng/mL) treatment with or without PP2 (10 μM) for 0, 10, 20, 40, and 60 min in A431 cells. (**C**) Phosphorylation of EGFR downstream signalling molecules after EGF (100 ng/mL) treatment with or without R406 (1 μM) for 0, 10, 20, 40, and 60 min in A431 cells. Data are presented from three independent experiments.

Next, we found that EGF (100 ng/mL) can increase Syk and c-Src phosphorylation in a time-dependent manner. Moreover, constitutive Lyn phosphorylation was detected but was not altered upon EGF stimulation (Figure 1B). Because of the non-availability of a specific antibody for the phosphorylated Hck, we could not determine its activation status in EGF-stimulated A431 cells. Notably, non-selective SFK inhibitor PP2 (10 μM) decreased the phosphorylation of c-Src, Lyn, Syk, and EGFR (Figure 1B). Because PP2 also reduced EGFR phosphorylation, we suggest there is a feedback

loop to control EGFR activation through SFKs as previously reported in fibroblasts [23], breast [23,24], and colon cancer cells [25] in a transactivation manner. All these findings suggested that Syk is a signal molecule for EGFR, and its activation is downstream of SFKs.

Next, we used the Syk inhibitor, R406, to test its effects on EGF-induced signalling pathways. As shown in Figure 1C, R406 treatment significantly inhibited the phosphorylation extent of JNK, p38 MAPK, STAT1, and STAT3 in EGF-stimulated A431 cells. Notably, R406 also reduced EGFR phosphorylation as did PP2 (Figure 1C).

In addition to A431 cells, we also examined the role of Syk in CAL27 (oral SCC) and SAS (head and neck SCC) cells. As observed in A431 cells, EGF stimulated the phosphorylation of EGFR, c-Src, JNK, and p38 in time- and concentration-dependent manners in both the cell lines (Figure 2A). Moreover, EGF activated Syk in CAL27 (Figure 2B) and SAS cells (Figure 2C). In CAL27 cells, EGF-induced phosphorylation of EGFR, STAT3, Akt, and JNK was inhibited by R406 (Figure 2B). Similarly, EGF-induced STAT3 phosphorylation in SAS cells was reduced by R406 (Figure 2C). These findings indicated that Syk is the upstream signal molecule that transduces the downstream signalling pathways of EGFR.

**Figure 2.** Syk is also an upstream signalling molecule of EGFR in SCC. (**A**) EGF-induced phosphorylation of EGFR and downstream signalling molecules for 0, 15, 30, and 60 min after EGF (100 ng/mL) treatment compared to that of A431. Concentration-dependent EGF treatment (0, 10, 30, and 100 ng/mL) for 30 min in CAL27 and SAS cells. (**B**,**C**) Phosphorylation of EGFR and downstream signalling molecules after EGF (100 ng/mL) treatment with or without R406 (1 μM) for 0, 0.5, 1, 3, and 6 h in CAL27 (**B**) and SAS cells (**C**). Data are presented from three independent experiments.

#### *3.2. Syk Mediates EGF-Induced IL-8 Upregulation But Is Not Involved in the Regulation of Cell Fate*

Given that IL-8 is a crucial player in cancer development [26], we examined the effects of R406 on EGF-induced IL-8 expression in A431 cells. We found that EGF can upregulate IL-8 mRNA expression and increase IL-8 protein production (Figure 3A), and these actions were inhibited by R406 and gefitinib. Similarly, in SAS and CAL27 cells, IL-8 mRNA induced by EGF was markedly reduced by R406 and gefitinib (Figure 3B,C).

**Figure 3.** Syk mediates EGF-induced IL-8 mRNA and protein expression in SCC. EGF (100 ng/mL) treatment in A431 (**A**), SAS (**B**), and CAL27 (**C**) cells for 0, 6, 12, and 24 h and the addition of R406 (1 μM) or gefitinib (5 μM) as a positive control. IL-8 mRNA levels were determined by RT-PCR and protein levels by ELISA.

Using BrdU incorporation as an index of cell proliferation, our data did not show the significant effects of EGF and R406 on BrdU uptake in A431 cells (Figure 4A). We also used MTT, ATP, and crystal violet assays to assess the effects of EGF and R406 on cell viability. Our data revealed non-significant effects of both agents on the cell viability of A431 cells (Figure 4B). When Syk mRNA was silenced by the siRNA approach, EGF still could not affect cell viability in A431 cells (Figure 4C). We also tested another EGFR activator, TGFα and found that in agreement with the data for EGF, TGFα (10 and 100 ng/mL), in the absence or presence of R406, did not change cell viability as assessed by MTT (Figure 4D) and crystal violet assays (Figure 4E) in A431 cells. Similarly, MTT assays revealed that R406 did not alter the viability of SAS and CAL27 cells (Figure 4F).

**Figure 4.** *Cont.*

**Figure 4.** Syk is not involved in the proliferation and viability of SCC cells. (**A**) Effects of R406 (1 μM) with or without co-treatment with EGF (50 ng/mL) on A431 cell proliferation for 16 and 24 h were analysed by BrdU assay. (**B**) Effects of R406 (1 μM) with or without co-treatment with EGF (100 ng/mL) on A431 cell viability for 24 h analysed by MTT, ATP, and crystal violet assays. (**C**) Effects of Syk siRNA with or without EGF (50 ng/mL) for 24 h on A431 cell viability analysed by an MTT assay. (**D**,**E**) Cell viability of A431 cells after treatment with R406 (1 μM) with or without co-treatment with TGFα (10 or 100 ng/mL) for 24 h analysed by MTT and crystal violet assays. (**F**) Viability of SAS and CAL27 cells treated with R406 for 24 h determined by MTT assay. Data are expressed as the mean ± S.E.M. from three independent experiments.

#### *3.3. Syk Regulates Intracellular EGFR Movement in a Cell Type-Specific Manner*

In our previous study, in NHEK cells, we observed the co-localisation of EGFR and Syk. Moreover, Syk was moved along with EGFR trafficking intracellularly after EGF stimulation [14]. Here, we further detected the intracellular localisation of EGFR in A431 and SAS cells before and after EGF stimulation. Unlike NHEK cells, where EGFR during the resting state was apparently present in the membrane, EGFR in A431 cells was present in both the cytosol and plasma membrane during the resting state. After EGF stimulation, EGFR was translocated to the plasma membrane and peri-nuclei of a spot like. The latter site was co-stained with the early endosomal marker EEA1, suggesting the occurrence of common EGFR trafficking from the internalisation to the late endosome/lysosomal degradation and endocytotic recycling pathways. The treatment with the Syk inhibitor R406 reduced the membrane location but not the peri-nuclear translocation of EGFR induced by EGF (Figure 5A). In SAS cells, EGFR was primarily observed in the plasma membrane as NHEK and could move to the peri-nuclei early endosome upon EGF stimulation. R406 treatment appeared to induce an uneven distribution of EGFR in peri-nuclear sites after EGF stimulation in SAS cells (Figure 5B).

**Figure 5.** *Cont.*

**Figure 5.** Syk modulates EGFR localisation under EGF treatment in A431 and SAS cells. Confocal microscopic analyses of EGFR and EEA1 (early endosome marker) localisation after treatment of EGF (50 ng/mL) for 1 h with or without the addition of R406 (1 μM) in A431 (**A**) and SAS cells (**B**). Data are from three independent experiments.

#### *3.4. Syk Activity Is Associated with Clinicopathologic Features and Outcome of Patients with Oral SCC*

To better understand the prognostic role of Syk in oral SCC, an immunohistochemical analysis for p-Syk expression was performed on the tissue microarray sections of oral SCC, and the clinicopathological phenomena and p-Syk expression in tumour tissues in 105 patients were analysed. We first compared the expression levels of p-Syk in oral SCC with that of matched non-neoplastic squamous epithelium (Figure 6A). Oral SCC had stronger p-Syk expression than normal squamous epithelium. The association between clinicopathological characteristics and p-Syk expression is shown in Table 1. Oral SCC in the high p-Syk expression group tended to be poorly differentiated in grading (*p* = 0.002). Lymphovascular invasion was more commonly detected in oral SCC with high p-Syk expression (*p* = 0.018). The associations between p-Syk and other clinicopathological factors, such as gender, age, perineural invasion, tumour size and extent, lymph node metastasis, and pathologic staging were not statistically significant. In addition, the Kaplan-Meier curve analysis revealed that the overall survival of the patients with low p-Syk expression was significantly longer than that of the patients with high p-Syk expression (*p* = 0.022) (Figure 6B). Patients with low p-Syk expression tended to have long disease-free survival; however, this was not statistically significant (*p* = 0.13) (Figure 6C). Univariate and multivariate analyses using the Cox proportional-hazard model were conducted to determine prognostic factors affecting overall survival (Tables 2 and 3). In the univariate analysis, old age (>65 years), larger tumour and/or extent (T3 + T4), presence of lymph node metastasis, high pathological staging (stage III and IV), and high p-Syk expression were associated with a poorer prognosis. The multivariate analysis showed that old age (>65 years), larger tumour and/or extent, and high p-Syk expression were independent poor prognostic factors for overall survival.

**Figure 6.** Syk is important for SCC development in human patients. (**A**) Immunohistochemical analysis of p-Syk expression in tissue microarray sections of oral SCC. (**B**,**C**) Correlation of Syk activity to overall survival (**B**) and disease-free survival (**C**) in patients with oral cancer.




**Table 1.** *Cont.*

Abbreviation: AJCC, American Joint Committee on Cancer.

**Table 2.** Univariate analysis of clinicopathological factors and p-Syk expression affecting overall survival.


Abbreviations: CI, confidence interval; AJCC, American Joint Committee on Cancer.


**Table 3.** Multivariate analysis of clinicopathological factors and p-Syk expression affecting overall survival.

Abbreviation: CI, confidence interval; AJCC, American Joint Committee on Cancer.

#### *3.5. Syk and EGFR Regulate PARP1 Activation and Syk Inhibitor Exerts a Synergistic Anti-Tumour E*ff*ect with Olaparib*

Because we found that Syk is involved in EGFR signalling in SCC, we intended to determine if Syk can serve as a therapeutic target. Moreover, PARP1 is involved in DNA damage repair and its expression is negatively correlated with patient survival rate and prognosis. Given that PARP1 inhibitors (PARPi) are currently used clinically for treating ovarian and metastatic breast cancers [27,28], and PARP inhibition results in the radiosensitisation of HPV/p16-positive HNSCC cells [21,29], we also aimed to determine the combined effects of Syk and PARP inhibition. First, we found that both R406 and the EGFR inhibitor, gefitinib, could induce PARP activation and DNA damage in the three SCC cell lines as indexed by increased PAR formation and γH2AX expression, respectively. Furthermore, PARP cleavage was induced by both agents (Figure 7A). Secondly, we found that although R406 and olaparib did not significantly affect cell viability, their combination induced significant cell death (Figure 7B). Such an enhanced cytotoxic l effect was more apparent in SAS and CAL27 cells than that in A431 cells. Moreover, because targeting EGFR has been evaluated in SCC [15–17], we tested the combined effects of gefitinib and olaparib. We found that although gefitinib itself did not affect cell survival, as demonstrated in Figure 4F, co-treatment with both the agents increased cell death, whereas the efficacy was slightly weaker than that of R406/olaparib (Figure 7B). Altogether, these results suggested a promising treatment of SCC by the combined inhibition of Syk and PARP.

**Figure 7.** R406 and gefitinib induce PARP activation, and inhibition of Syk and PARP enhances cell death in SCC. (**A**) Cells were treated with R406 (1 μM) or gefitinib (5 μM) for different periods, Figure 1. γ-H2AX, and β-actin. (**B**) Cells were treated with R406 (1 μM), gefitinib (5 μM) and/or olaparib (Ola, 10 μM) for 24 h. Cell death was determined by Annexin V/PI staining and flow cytometry. Data in (**B**) are expressed as the mean ± S.E.M. from three independent experiments. \* *p* < 0.05 indicates a synergistic effect of R406 + Ola and gefitinib + Ola on cytotoxicity compared to individual drug treatments.

#### **4. Discussion**

The importance of EGFR in tumourigenesis has been well established. Excessive activation of EGFR signalling by overexpression of, or mutations in, EGFR has been found in various types of human tumours, making EGFR a widely recognised target for cancer therapy [30]. Indeed, prior studies have established enhanced anti-tumour efficacy of the EGFR inhibitor (or antibody) with radiation [31] or cytotoxic drugs for A431 cells [17,32]. Nevertheless, long-term EGF treatment also leads to drug resistance, which is an urgent issue that must be overcome. This indicates the necessity to explore new targets for anti-tumour therapy. In this study, we found that Syk is a new downstream signal molecule of SFKs in activated EGFR-transduced signalling pathways, and similar to SFKs, Syk possesses a positive feedback loop to enhance EGFR phosphorylation. Moreover, the association of higher Syk activity with poor clinicopathological features of patients with oral SCC, including poor overall survival and high lymphovascular invasion, was observed. Furthermore, a combined inhibition of Syk and PARP leading to enhanced SCC cell death may be developed into a new therapeutic regimen in the future.

The major functions of Syk in hematopoietic cells involving innate and adaptive immune systems are well characterised. Apart from the established molecular mechanisms and functional roles in autoimmunity, accumulating evidence suggests that Syk plays dual roles in cancers, i.e., either as a tumour suppressor or promoter [13]. Reportedly, the reduced expression of Syk in breast cancer and melanoma cells is associated with a higher degree of malignancy and poor prognosis. The tumour suppressor action of Syk has been associated with positive regulation of p53 activity [33] and p21 expression [34], as well as negative effects on mitotic progression [35,36]. In contrast, Syk serves as a tumour promoter in hematopoietic malignancies and could be a potential oncogenic driver in small cell lung cancer [37] and ovarian cancer [36] by inducing pro-survival signals [13]. Therefore, Syk could become a new therapeutic target in some cancer cell types. For example, a recent study indicated that a Syk inhibitor could sensitise TRAIL-induced apoptosis by downregulating Mcl-1 in breast cancer cells [38]. In this study, we strengthened our previous speculations that Syk is an upstream signalling molecule for EGFR in SCC cell lines and mediates EGF-induced IL-8 gene expression. IL-8 is a potent chemoattractant molecule that performs different pro-tumoural functions, such as proliferation, angiogenesis, and metastasis of cancer cells. Since in patients with cancer, IL-8 is mainly produced by tumour cells, its serum concentration has been shown to be correlated with the tumour burden, and IL-8 is an effective pharmacodynamic biomarker to detect an early response to immunotherapy [26]. The serum level of IL-8 is significantly increased in patients with recurrent and metastatic head and neck SCC [39]. In addition, elevated levels of IL-8 are found in the saliva of patients with oral SCC; thus, IL-8 could serve as a salivary biomarker of oral SCC [40]. In addition, using human oral SCC samples, we demonstrated the strong association of Syk activity with poor disease progression. All these data prompt us to suggest Syk inhibition as an alternative cancer treatment.

In addition to clarifying the role of Syk in EGFR signalling and the prognosis of patients with SCC, we tested the combined effects of Syk and PARP inhibition on cancer cell death. The PARP1 inhibitor has been shown to sensitise radiotherapy in head and neck SCC [21]. Moreover, because cetuximab can augment cytotoxicity with PARP inhibition in head and neck SCC [41], we also tested the anti-tumour activity of gefitinib and olaparib. Although PARP1 inhibition has been applied in clinical settings in combination with different therapeutic agents, it is still crucial to develop alternative regimens to overcome drug resistance and insufficient efficacy in cancer therapy. In addition, some recent studies have indicated the effect of EGFR activation on the regulation of PARP1 in a cell context-dependent manner. In human LNCaP prostate cancer cells, radiation-induced PARP activation is enhanced through EGFR-ERK signalling [42]. In UT-SCC5 and SAS head and neck cancer cells, PARP1 has been shown to serve as a mediator of EGFR/MEK-dependent regulation of DNA double-strand breaks [43]. In breast cancer cells, a contextual synthetic lethality may exist between combined EGFR and PARP inhibitors [44]. In glioblastoma cells, EGFRvIII overexpression causes increased ROS-dependent DNA strand break accumulation and PARP activation, and reduced DNA repair gene expression, including PARP1, results in improved patient survival [45]. In hepatocellular carcinoma, EGFR and c-MET cooperate to enhance resistance to PARP inhibitors [46]. Furthermore, a heterodimer of EGFR and MET can phosphorylate Y907 of PARP1 in the nucleus of hepatocellular carcinoma and contribute to this resistance [46]. In pancreatic cancer cells, adaptive expression of EGF following exposure to ionising radiation may induce radio-sensitisation of cells through the induction of the cyclin D1/p53/PARP pathway [47]. In contrast, EGF in combination with radiation augments the radiation effects in A431 cells by inhibiting DNA damage repair [48]. Additionally, EGFR-mutant lung cancer cells display higher sensitivity to the PARP inhibitor olaparib [49]. All these studies suggest that EGF can regulate PARP1, which might impact the efficacy of cancer therapy, particularly when DNA damage is caused by cancer therapy. However, to date, studies on the relationship between Syk and DNA breaks are

limited. The only finding reported is that DNA damage and DNA double-strand breaks can suppress the expression of Syk in B [50] and NK cells [51].

Our data revealed a time-dependent effect of Syk and EGFR inhibition on DNA damage and PARP1 activation in three SCC cell lines. R406 exerts higher responses in terms of DNA damage and PARP1 activation than those exerted by gefitinib in CAL27 and A431 cells. However, such parallel events are not observed in SAS cells where R406 induces greater DNA damage but less PARylation compared to that of gefitinib. Currently, the reasons underlying this paradoxical finding are unknown. Notably, even though PARP1 cleavage was observed upon treatment of R406 or gefitinib, the cell viability assay did not reveal significant effects for either agent on cell death. We speculate that this might be due to the insufficient activation of executive caspase 3 for cell death and/or the existence of other pro-survival pathways. Another possible reason is that the PARP1 cleavage is mediated by other proteases. Previous studies showed that PARP1 can be cleaved by caspase 1 and caspase 7 to yield 85–89 kD PARP-1 fragments [52–54], and such cleavage leads to an NF-κB-dependent inflammatory gene expression [54]. Therefore, PARP1 cleavage is a hallmark of apoptosis yet not essential. In addition, we speculate the PARP1 activation evoked by R406 and gefitinib alone might participate in the cell survival, for example as previously observed for the DNA repairing process upon moderate DNA damage. However, interestingly, enhanced cell death effect was observed upon the co-inhibition of Syk and PARP. Notably, this enhanced cell death effect is greater than that induced by gefitinib and olaparib. Therefore, we suggest that Syk inhibition leads to the interruption of EGFR-dependent and -independent cellular actions that might exert synergistic effects with olaparib in terms of cell death. It is necessary to further address the molecular mechanisms underlying the coordinative cellular events mediated by Syk and PARP1 in the future. Additionally, it will be interesting to investigate and compare the efficacies between the Syk inhibitor and EGFR mAb (e.g., cetuximab) in SCC upon co-treatment with PARP1 inhibitors.

A431 cells compose a high EGFR-activating skin SCC. Notably, it is unexpected to find EGF at concentrations of 50–100 ng/mL, which is mitogenic for general cell types and cannot induce cell proliferation and viability in A431 cells within 24 h of incubation. Supporting this notion, previous studies reported a growth arrest effect of EGF in A431 cells that was observed slowly [55,56]. To date, the mechanism through which EGF induces growth arrest in A431 cells has been ascribed to the upregulation of IRF-1 through SFR-dependent STAT1 and STAT3 activation [57,58]. Similarly, cell lines that hyperexpress EGFR, such as MDA-MB-468 cells, have been documented to undergo receptor-mediated apoptosis through STAT3 [59]. Therefore, because of the overexpression of EGFR and constitutive overactivation of EGFR in A431 cells, we speculate there might be an aberrant regulation mechanism in the EGFR-dependent cellular functions. Another distinct feature in A431 is the subcellular distribution of EGFR. Unlike primary keratinocytes and SAS cells, EGFR is not only localised in the plasma membrane but is also present in the cytosol in A431 cells. Moreover, after EGF stimulation, the plasma membrane level of EGFR is increased in A431 cells, which is opposite to the general trafficking direction of activated EGFR. We found lower amounts of EGFR expression in SAS cells compared to that of A431 cells could more rapidly undergo EGFR internalisation after EGF stimulation for 1 h. All these phenomena observed in A431 cells might result from the high abundance of EGFR expression in this specific cell line. Another cell type-specific action of the Syk inhibitor is its effects on EGF-induced EGFR trafficking. We found that in A431 cells, R406 could decrease the plasma membrane level of EGFR upon EGF stimulation, whereas in SAS cells, EGF-induced EGFR trafficking to the early endosome displayed a higher level upon Syk inhibition. Because Syk can associate with EGFR in keratinocytes [14] and altered EGFR internalisation and recycling can regulate drug sensitivity in cancer cells [60,61], we plan to further determine the role of Syk in EGFR trafficking, including receptor internalisation, receptor recycling to the plasma membrane, and degradation by lysosomes in the future.

#### **5. Conclusions**

We confirmed our previous findings in keratinocytes that Syk mediates EGFR signalling in SCC and Syk is a positive contributor to disease progression in patients with oral SCC. Even though Syk and PARP inhibition cannot induce cell death individually, their combination causes enhanced cell death. Thus, our data provide the rational for co-treatment of PARP inhibition and Syk inhibition in SCC.

**Author Contributions:** Conceptualization, D.-Y.H. and N.-L.W.; Data curation, D.-Y.H.; Methodology, W.-Y.C. and C.-L.C.; Project administration, W.-W.L.; Resources, W.-Y.C., C.-L.C. and N.-L.W.; Writing—original draft, W.-W.L. All authors have read and agreed to the published version of the manuscript.

**Acknowledgments:** This work was supported by research grants from Taipei Medical University (TMUTOP103005-5), Ministry of Science and Technology (MOST 103-2320-B-002-069-MY3), Excellent Translational Medicine Research Projects of National Taiwan University College of Medicine (NSCCMOH-94-51), and National Taiwan University (NTU-CC-108L893601).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **TDP1 and TOP1 Modulation in Olaparib-Resistant Cancer Determines the E**ffi**cacy of Subsequent Chemotherapy**

**Jin Won Kim 1,2,3,**†**, Ahrum Min 4,5,**†**, Seock-Ah Im 1,3,4,5,6,\*, Hyemin Jang 4, Yu Jin Kim 4, Hee-Jun Kim 7, Kyung-Hun Lee 1,4,5,6, Tae-Yong Kim 1,4,5,6, Keun Wook Lee 1,2, Do-Youn Oh 1,3,4,5,6, Jee-Hyun Kim 1,2,3 and Yung-Jue Bang 1,4,5,6**


Received: 31 December 2019; Accepted: 31 January 2020; Published: 3 February 2020

**Abstract:** The aim of this study was to elucidate the carryover effect of olaparib to subsequent chemotherapy and its underlying mechanisms. We generated olaparib-resistant SNU-484, SNU-601, SNU-668, and KATO-III gastric cancer cell lines and confirmed their resistance by cell viability and colony forming assays. Notably, olaparib-resistant cell lines displayed cross-resistance to cisplatin except for KATO-III. Inversely, olaparib-resistant SNU-484, SNU-668, and KATO-III were more sensitive to irinotecan than their parental cells. However, sensitivity to paclitaxel remained unaltered. There were compensatory changes in the ATM/ATR axis and p-Chk1/2 protein expression. ERCC1 was also induced in olaparib-resistant SNU-484, SNU-601, and SNU-668, which showed cross-resistance to cisplatin. Olaparib-resistant cells showed tyrosyl-DNA phosphodiesterase 1 (TDP1) downregulation with higher topoisomerase 1 (TOP1) activity, which is a target of irinotecan. These changes of TOP1 and TDP1 in olaparib-resistant cells was confirmed as the underlying mechanism for increased irinotecan sensitivity through manipulated gene expression of TOP1 and TDP1 by specific plasmid transfection and siRNA. The patient-derived xenograft model established from the patient who acquired resistance to olaparib with BRCA2 mutation showed increased sensitivity in irinotecan. In conclusion, the carryover effects of olaparib to improve antitumor effect of subsequent irinotecan were demonstrated. These effects should be considered when determining the subsequent therapy with olaparib.

**Keywords:** olaparib; carryover effect; TOP1 activity; TDP1; irinotecan

#### **1. Introduction**

Olaparib is an oral poly(ADP-ribose) polymerase (PARP) inhibitor that traps inactivated PARP onto single-strand DNA breaks (SSBs), leading to double-strand DNA breaks that induce cell death [1,2]. Olaparib showed antitumor effect alone or in combination with chemotherapy in tumors with deficiencies at the repair of double-strand breaks caused by BRCA1/2 mutations, ATM dysfunction, and RAD51C deficiency [3–6]. Recently, in germline BRCA-mutated advanced ovarian cancer patients who have been treated with three or more prior lines of chemotherapy, olaparib showed clinical efficacy and has been approved by U.S. Food and Drug Administration [4]. The Cancer Genome Atlas reported that 7 of 287 gastric patients (2.4%) showed genetic alterations in RAD51C gene, and it has been well established that loss of RAD51C is associated with increased cancer risk [7–9]. In our previous study, we found that RAD51C depletion led to hypersensitivity to olaparib in gastric cancer cells [6]. Therefore, PARP inhibition can be an attractive strategy for gastric cancer treatment.

In a randomized phase II trial of second-line metastatic gastric cancer, the addition of olaparib to paclitaxel showed much longer overall survival (OS) than paclitaxel plus placebo, although the progression-free survival (PFS) was marginally different between the two groups [10]. This finding was also identified in a randomized phase 3 trial of using olaparib in second-line gastric cancer [11]. It means that post-progression survival was longer in olaparib arm; however, the reason for this could not be fully explained. There is no preclinical evidence to explain how olaparib affects post-progression survival. In a previous study, our group showed the prolonged antitumor effect of olaparib even after stopping olaparib treatment in an in vivo xenograft model [6]. Based on this finding, carryover effect of olaparib, which has an effect on subsequent chemotherapy, was suggested as one of the potential explanations for this, particularly, in irinotecan, which is usually applied as a third line of treatment.

PARP-1 is activated by irinotecan-induced DNA breaks. PARP inhibitors share their predictive characters with platinum and enhance the cytotoxicity of topoisomerase 1 (TOP1) inhibitors [12]. Restoration of homologous recombination through a loss of 53BP1 or BRCA re-expression is an important mechanism for PARP inhibitor resistance in BRCA1-deficient mammary tumors [13]. A previous study has demonstrated that ATM regulates TOP2 expression. ATM loss results in increased TOP2 levels and enhances sensitivity to TOP2 inhibition [14]. Therefore, it could be possible to compensate for an impaired DNA repair pathway from olaparib by another repair-pathway related to SSB, including the TOP1 and tyrosyl-DNA phosphodiesterase 1 (TDP1) pathways, in resistant cell lines that have been exposed to olaparib for long periods. On the other hand, it was reported that TDP1 and TOP1 are predictive biomarkers for irinotecan and new therapeutic targets in the era of precision medicine [14,15].

On these backgrounds, we explored the carryover effect of olaparib, which influences the efficacy of subsequent chemotherapy after olaparib treatment. Additionally, we attempted to evaluate the mechanisms underlying this effect.

#### **2. Results**

#### *2.1. Altered Sensitivity of Chemotherapeutic Agents in Olaparib-Resistant Cells*

We established olaparib-resistant cell lines by long-term treatment with olaparib from SNU-484, SNU-601, SNU-668, and KATO-III. The resistance was confirmed by using cell growth inhibition and colony forming assays (Table 1 and Figure 1). Except for KATO-III, all olaparib-resistant cell lines became resistant to cisplatin. Inversely, the olaparib-resistant cell lines became more sensitive to irinotecan, except for SNU-601. Apoptosis also increased after irinotecan treatment in olaparib-resistant cell lines (Figure S1). Sensitivity toward paclitaxel was not altered after the acquisition of resistance to olaparib.


**Table 1.** Drug sensitivity in parental and olaparib-resistant cells, as assessed by using cell growth-inhibition assay.

\* MTT for 5 days.

**Figure 1.** *Cont*.

**Figure 1.** The sensitivity of chemotherapeutic agents was altered in olaparib-resistant cells. Cell viability was calculated through counting cell colonies by using GELCOUNT in SNU-484 (**A**), SNU-601 (**B**), SNU-668 (**C**), and KATO-III (**D**) cells. \* indicates *p* < 0.001.

#### *2.2. Changes of Proteins Related to the DNA-Damage Response in Olaparib-Resistant Cell Lines*

To evaluate the mechanisms underlying the altered sensitivity toward cisplatin, changes of the DNA-damage response proteins were evaluated. p-ATR was induced in all olaparib-resistant cell lines (Figure 2). p-ATM was slightly upregulated in olaparib-resistant SNU-484 and olaparib-resistant SNU-668. p-Chk1 and p-Chk2 were elevated in olaparib-resistant cell lines compared with the parental cells. Furthermore, ERCC1 was also upregulated in olaparib-resistant SNU-484, SNU-601, and SNU-668 cells, which were resistant to cisplatin. This ERCC1 induction was not detected in olaparib-resistant KATO-III cells, which were not resistant to cisplatin.


**Figure 2.** The expressions of DNA-damage response proteins were altered after acquisition of olaparib resistance. Western blot analysis for p-ATR, ATR, p-ATM, ATM, p-Chk1, Chk1, p-Chk2, Chk2, and ERCC1 was conducted to evaluate the altered expression after acquisition of resistance to olaparib. There were compensatory changes in the DNA-damage response proteins.

#### *2.3. Morphological Changes in the Olaparib-Resistant Cell Lines*

With the acquisition of resistance to olaparib, the nuclear size of the resistant cells increased, as assessed by confocal imaging (Figure 3A). Increased abnormal DNA contents indicating aneuploidy was not observed in cells with increased nuclear size (Figure 3B). In the olaparib-resistant cells, increased nuclear size indicated the relaxation of DNA condensation [16,17].

**Figure 3.** *Cont*.

**Figure 3.** The nuclei were larger in olaparib-resistant cells than in parental cells. The degree of DNA condensation was examined by 4 ,6-diamidino-2-phenylindole (DAPI) staining and confocal laser microscopy. Representative images are presented (80× original magnification). Scale bars represent 50 μm. In olaparib-resistant cells, increased nucleus indicates a relaxation of DNA condensation (**A**). The proportion of cells with a DNA content of more than 4n was calculated by flow cytometry with PI staining. The increased nuclear size of olaparib-resistant cells was not related to aneuploidy (**B**).

#### *2.4. Increased TOP1 Activity and Decreased TDP1 Expression in Olaparib-Resistant Cell Lines*

To evaluate the underlying mechanism of increased irinotecan sensitivity in the olaparib-resistant cell lines, protein expression, and activity of TOP1 and protein expression of TDP1, which are the target and predictive markers of irinotecan, were measured. Olaparib-resistant cells showed higher TOP1 activity than their parental cells, although protein expression of TOP1 was not altered (Figure 4A,B). In addition, TDP1 expression was decreased in olaparib-resistant cell lines except in olaparib-resistant SNU-601, which did not exhibit sensitivity to irinotecan and had a mutation in TDP1 (Table 2).

**Figure 4.** Topoisomerase 1 (TOP1) activity was increased and tyrosyl-DNA phosphodiesterase 1 (TDP1) expression was downregulated in olaparib-resistant cells. TOP1 activity was increased in all olaparib-resistant cells compared with parental cells (**A**), although the protein expression of TOP1 was not altered (**B**). TDP1 protein expression was downregulated in olaparib-resistant SNU-484, SNU-668, and KATO-III cells compared with their parental cells (**B**). \* indicates *p* < 0.001.

98



TDP1, tyrosyl-DNA phosphodiesterase 1; TOP1, topoisomerase 1.

#### *2.5. Changes of Irinotecan Sensitivity According to Manipulated TOP1 and TDP1 Expression*

To determine if the manipulation of TOP1 and TDP1 could change the sensitivity of irinotecan, the sensitivity of irinotecan was evaluated according to manipulated expressions of TOP1 and TDP1 by using specific plasmid transfection and siRNA. In olaparib-resistant cells of SNU-484 and SNU-668 with increased activity of TOP1, irinotecan sensitivity was reduced when TOP1 expression was downregulated by siRNA (Figure 5A). In addition, in parental cells of SNU-484 and SNU-668, decreased TDP1 expression by siRNA induced the increased irinotecan sensitivity as same as olaparib-resistant cells. Olaparib-resistant SNU-601, in which TDP1 expression was not decreased, also showed an increased irinotecan sensitivity when TDP1 expression was knocked down (Figure 5B). Inversely, TOP1, or TDP1, expression was overexpressed by plasmid transfection in parental or olaparib-resistant cells of SNU-484 (Figure 5C). Although overexpressed TOP1 had slightly increased irinotecan sensitivity, simultaneous upregulation of TOP1 with downregulation of TDP1 in parental SNU-484 showed the significantly increased sensitivity to irinotecan compared with parental cells (Figure 5C). Furthermore, transiently overexpressed TDP1 had attenuated irinotecan sensitivity in olaparib-resistant SNU-484 cells, but olaparib-resistant SNU-484 with TOP1 downregulation and TDP1 upregulation was confirmed to attenuate the sensitivity to irinotecan dramatically (Figure 5C). The successful expression modulation by transfection was validated by western blot analysis (Figure 5C). Therefore, olaparib-resistant cell lines could be highly sensitive to subsequent irinotecan through TDP1 downregulation at the same time as the increased TOP1 activity. Furthermore, downregulation of TDP1 with TOP1 upregulation increased the nuclear size in parental cells, similar to olaparib-resistant cells (Figure 6). Thus, morphological changes of olaparib-resistant cells could be attributed to the relaxation of DNA condensation modulated by TOP1 hyperactivity and depletion of TDP1.

**Figure 5.** *Cont*.

**Figure 5.** The sensitivities to irinotecan with modulating TOP1 and TDP1 expressions were evaluated by MTT assay. TOP1-siRNA mix, TDP1-siRNA mix, and non-targeting siRNA (160nM each) were transfected into olaparib-resistant cells or parental cells, and then treated with irinotecan for 5 d. Cell viability percentages were then measured by MTT assay, and TOP1 or TDP1 silencing was determined by western blotting (**A**,**B**). pCMV-tag2A-TOP1 or pCMV-tag2A-TDP1 and/or siTOP1 or siTDP1 were transfected into parental or olaparib-resistant cells. The sensitivity to irinotecan was calculated by MTT assay and the efficacy of transfection was confirmed by western blotting (**C**,**D**).

**Figure 6.** Parental cells were transfected with siRNA targeting TDP1 and/or pCMV-tag2A-TOP1 for 5 d. The size of nuclei was measured by DAPI staining and confocal laser microscopy. Representative images are presented (80× original magnification). Scale bars represent 50 μm.

#### *2.6. Potent Anti-Tumor Activity of Irinotecan in Olaparib-Resistant Patient-Derived Breast Cancer Xenograft Model*

To validate the increased antitumor effect of irinotecan in olaparib resistant model, we tested the antitumor activity of irinotecan in patient-derived xenograft (PDX) models established from a breast cancer patient with BRCA2 580del4 mutation who acquired resistance to olaparib. In concordance with clinical response, IMX 181 model showed no response to olaparib, but statistically significant delay of tumor growth on irinotecan treatment (Figure 7A). Both treatments were well tolerated without any sign of toxicity (Figure 7B).

**Figure 7.** Irinotecan impeded tumor growth in a BRCA2 mutant PDX model with acquisition of olaparib resistance. (**A**) PDX model were treated with 50 mg/kg olaparib or 10 mg/kg irinotecan for 21 days (n = 4/group). The mean of tumor volumes was presented in a graph with the standard deviation (SD). \* indicates *p* < 0.001 versus vehicle and olaparib group. (**B**) Changes in body weight were measured every three days for 21 days.

#### **3. Discussion**

This study confirmed the carryover effect of olaparib-treatment to subsequent chemotherapy, particularly in irinotecan. Through manipulating gene expressions, increased sensitivity to irinotecan in olaparib-resistant cells was confirmed to be attributed to TOP1 upregulation and TDP1 downregulation, which was shown in olaparib-resistant cells. These results could explain the higher OS improvement compared with PFS prolongation in the randomized clinical study of second-line gastric cancer, evaluating the efficacy of olaparib combined with paclitaxel.

After irinotecan treatment, SSBs are repaired by a complex consisting of TDP1, which functions in the base excision repair pathway [18]. PARP inhibitors, which inhibit base excision repair, sensitize cells to TOP1 inhibitors [19]. Therefore, irinotecan and olaparib represent a potent combination. However, concurrent treatment with both PARP inhibitors and irinotecan is too toxic for clinical development, although a preclinical study demonstrated synergistic effects [20–23]. Therefore, sequential treatment might represent a promising alternative approach. Our results suggest that the application of irinotecan after olaparib treatment may be a feasible treatment option owing to the carryover effect of olaparib.

DNA-damage response proteins function in various complex and overlapping pathways. In the present study, long-term olaparib treatment induced a compensatory alteration of the ATR/ATM axis, Chk, and ERCC1 expression. The development of olaparib resistance can be ascribed to this compensatory alteration, which also results in cisplatin resistance. Furthermore, olaparib and platinum share a common mechanism of action in the DNA repair pathway and similar predictive characters, such as the presence of BRCA mutation and RAD51 deficiency [12,24–26]. For a specific example, SNU-601 was highly sensitive to olaparib due to RAD51C-deficiency, which was also identified in several cancer types [6,25,26]. Parental SNU-601 was also sensitive to cisplatin, and olaparib-resistant SUN-601 showed resistance to cisplatin likewise. Therefore, this suggests that treatment with platinum should be avoided after olaparib failure, although clinical studies have so far not shown a decreased response to platinum after the resistance of PARP inhibitor in ovarian cancer [27]. In addition, this carryover effect might be specific to tumor cells based on synthetic lethality as a cytotoxic mechanism of PARP inhibitor. Clinically, there are no studies to report that this carryover effect could exacerbate the toxicities of subsequent chemotherapy.

TOP1 is an important cellular enzyme that allows for DNA relaxation. TOP1 cleaves DNA to create a DNA single-strand break to which it remains covalently bound to, thus allowing for rotation and relaxation of DNA. Once rotated, bound TOP1 ligates the nicked DNA and is released. In olaparib-resistant cell lines, the size of the nucleus was larger than that in the parental cell lines. This finding gave an indirect clue that TOP1 activity was increased in olaparib-resistant cells [16,17]. TOP1 activity is a predictive marker of irinotecan [28,29]. In the present study, increased TOP1 activity in olaparib-resistant cells was confirmed by using TOP1 activity assay. According to changes of TOP1 expression by transfection of siRNA, the sensitivity of irinotecan was changed, and the size of the nucleus was altered corresponding to olaparib-resistant cells. These results, therefore, provide direct evidence for the alteration of irinotecan sensitivity in olaparib-resistant cells.

TOP1 activity was increased in all olaparib-resistant cell lines, although TOP1 protein expression was not altered. TDP1 expression was downregulated in olaparib-resistant SNU-484, SNU-668, and KATO-III cells, which were more sensitive to irinotecan after the acquisition of olaparib resistance. Exceptionally, olaparib-resistant SNU-601 did not show altered expression of TDP1 and sensitivity to irinotecan. SNU-601 had TDP1 mutation (A520D). It has been reported that inactive mutant TDP1 (H263A) did not reduce DNA-damage by camptothecin, although the function of this TDP1 mutation (A520D) was unknown [30]. In colon cancer, TDP1 depletion increased the sensitivity to irinotecan in a TOP1-dependent manner [31]. In our study, in parental cells, TOP1 upregulation and TDP1 downregulation resulted in the highest sensitivity to irinotecan compared with either change alone. The mechanism underlying the carryover effect of olaparib for irinotecan could be confirmed, as these changes in olaparib-resistant cells, which included an increased TDP1 downregulation along with increased TOP1 activity.

The PDX model established from the patient who acquired resistance even harboring BRCA2 mutation exhibited sensitivity to irinotecan. These results can help to explain how olaparib showed an increase in overall survival, without increasing PFS in phase 2 trial conducted in gastric cancer, by its carryover effect on irinotecan, which is used as a third-line treatment [10].

In summary, this study demonstrated that there were carryover effects after acquisition of olaparib resistance. In olaparib-resistant cells, cisplatin resistance might occur because of compensatory alterations in the ATR/ATM axis, and Chk and ERCC1 expression. Importantly, irinotecan sensitivity was enhanced through TDP1 downregulation, concomitant with increased TOP1 activity in olaparib-resistant cells. Sensitivity to paclitaxel remained unaltered after acquisition of resistance to olaparib. Based on these results, carryover effect of olaparib to subsequent therapy should be significantly considered during the clinical use of olaparib.

#### **4. Materials and Methods**

#### *4.1. Reagents*

The PARP inhibitor, olaparib, was kindly provided by AstraZeneca, Ltd. (Macclesfield, Cheshire, UK). Cisplatin and paclitaxel were obtained from Choongwoe Co., Ltd., and Samyang Genex Co., Ltd. (Seoul, Korea). Irinotecan (cas no. 100286-90-6) was purchased from Sigma Aldrich (St. Louis, MO, USA).

#### *4.2. Cell Lines and Cell Culture*

Human gastric cancer cells (SNU-484, SNU-601, SNU-668, and KATO-III) were purchased from the Korean Cell Line Bank (Seoul, Korea). The identities of cell lines were validated by DNA fingerprinting analysis [32]. The cells were cultured at 37 ◦C in an atmosphere containing 5% CO2 in Roswell Park Memorial Institute (RPMI) 1640 supplemented with 10% fetal bovine serum (FBS) and 10 μg/mL gentamicin.

#### *4.3. Generation of Olaparib-Resistant Cells*

Olaparib-resistant SNU-484, SNU-601, SNU-668, and KATO-III cells were established by continuous exposure to olaparib, starting with 1 μmol/L and incrementally increasing the concentration to 5 μmol/L over 7 months. Resistant cells were expanded in RPMI-1640 medium containing 10% FBS and 1 μmol/L olaparib.

#### *4.4. Cell Growth Inhibition Assay*

Cells (2–3 <sup>×</sup> 103 in 100 <sup>μ</sup>L/well) were seeded in 96-well plates and incubated overnight at 37 ◦C in 5% CO2, and then exposed to each drug at the indicated concentration for 5 days. After drug treatment, 50 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolim bromide solution (Sigma Aldrich) was added to each well, and the plates were incubated for 4 h at 37 ◦C before the media was removed. After dissolving the formazan crystals with 150 μL of dimethyl sulfoxide, the absorbance of each well was measured at 540 nm with a VersaMax™ microplate reader (Molecular Devices, Sunnyvale, CA, USA). Half-maximal inhibitory concentration (IC50) values were analyzed using SigmaPlot software (Statistical Package for the Social Sciences, Inc., Chicago, IL, USA).

#### *4.5. Colony-Formation Assay*

To compare the drug response between parental cells and resistant cells, the cells were seeded into six-well plates and incubated with the indicated concentration of each drug for 14 days. The cell colonies were washed in phosphate-buffered saline and stained with 0.1% Coomassie Blue solution (Sigma Aldrich) for 1 h at room temperature. The excess staining solution was then removed, and the plates were washed in PBS and air-dried. The cell colonies were counted using GELCOUNT (Oxford Optronix Ltd., Abingdon, UK), and cell viability was calculated by using SigmaPlot.

#### *4.6. Western Blot Analysis*

Whole cell proteins were extracted by using RIPA buffer, and equal amounts of protein were separated on 5%–15% SDS-PAGE gels, as described previously [33]. Primary antibodies against phosphorylated (p)-ATM, ATM, p-ATR, ATR, p-Chk1, Chk1, p-Chk2, Chk2, and ERCC1 were acquired from Cell Signaling Technology (Beverley, MA, USA). Antibodies against TOP1 and TDP1 were purchased from Abcam (Cambridge, UK). Actin antibody (Sigma Aldrich) was used as a control. The band intensity was calculated by Image J software.

#### *4.7. Immunofluorescence Assay*

The degree of DNA condensation was examined by 4 ,6-diamidino-2-phenylindole (DAPI) staining and confocal laser microscopy. Cells were plated on 0.01% poly-l-lysine (Sigma Aldrich)-coated coverslips and incubated overnight. Then, the cells were stained with DAPI (300 nM; Invitrogen, Carlsbad, CA, USA) for 1 min. The coverslips were rinsed three times for 10 min in PBS and mounted on slides using Faramount aqueous mounting medium (DAKO, Denmark). Immunofluorescence was visualized using a Zeiss LSM 510 laser-scanning microscope.

#### *4.8. Nuclear Extraction*

Cells were incubated in hypotonic buffer A (10 mM HEPES (pH 7.5), 1.5 mM MgCl2, 10 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, 2 mM phenylmethylsulfonyl fluoride (PMSF), 1 mg/mL pepstatin A, 0.2 mM leupeptin, 10 μg/mL aprotinin, 1 mM sodium vanadate, 1 mM nitrophenylphosphate, and 5 mM benzamidine) for 10 min at 4 ◦C. The cells were incubated with 0.1% NP-40 for 10 min at 4 ◦C, and then centrifuged at 2000 rpm for 5 min at 4 ◦C. The supernatant was removed, and the pellets were resuspended in 1 mL of cold TEM solution (10 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 4 mM MgCl2) with 0.5 mM PMSF and centrifuged at 2000 rpm at 4 ◦C. The nuclear pellets were resuspended in a small volume of TEM, and mixed and incubated with an equal volume of 1 M NaCl for 30 min at 4 ◦C. Then, the supernatant was kept as the nuclear extract, following centrifugation at 13000 rpm for 15 min at 4 ◦C.

#### *4.9. TOP1 Activity Assay*

The TOP1 activities of the nuclear extracts from each cell line were assessed by measuring the relaxation of the supercoiled pHOT1 plasmid (TopoGEN, Inc., Buena Vista, CO, USA). Supercoiled pHOT1 plasmid (250 ng/μL) was incubated with 1 μL of nuclear extract in 10 mM Tris-HCl (pH 7.9), 150 mM NaCl, and 1 mM EDTA for 60 min at 37 ◦C in a final volume of 20 μL [34,35]. The reaction was terminated by the addition of 4 μL of stop buffer (1% Sarkosyl, 0.025% bromophenol blue, and 5% glycerol), and the samples were loaded onto 1% agarose gels. TOP1 activities were calculated by measuring the degree of disappearance of the supercoiled DNA and are presented in a bar graph.

#### *4.10. Plasmid and siRNA Transfection*

pME18S-FL3-TOP1 and pOTB-TDP1 plasmids were purchased from the Korean human gene bank (Daejoen, Korea) and both TOP1 and TDP1 were subcloned into pCMV-Tag2A vector. SNU-484 and SNU-668 parental cells were transfected with 4 μg of pCMV-Tag2A-TOP1 plasmid and/or 160 nM of TDP1-specific siRNA, while SNU-484 and SNU-668 olaparib resistant cells were transfected with 4μg of pCMV-Tag2A-TDP1 plasmid and/or 160 nM of TOP1-specific siRNA using Lipofectamin 2000 (Invitrogen, Carlsbad, CA, USA). TOP1- and TDP1-specific siRNAs were synthesized from Genolution (Seoul, Korea), and the sequences of the TOP1-specific siRNAs were 5 -GGGAAGGACTCCATCAGATACTATA-3 , 5 -AAGTGGAAATGGTGGGAAGAA-3 , and 5 -CGAAGAAGGTAGTAGAGTC-3 , and the sequences of the TDP-specific siRNAs were 5 -GACCATA TCTAGTAGTGAT-3 , 5 -GGAGTTAAGCCAAAGTATA-3 , and 5 -CTAGACAGTTTCAAAGTG-3 . The sequence of the non-specific siRNA was 5 -AATTCTCCGAACGTGTCACG-3 .

#### *4.11. Library Preparation and Sequencing*

Library preparations, clustering, and whole genome sequencing were conducted by Macrogen (Seoul, Korea). Libraries were prepared according to the TruSeq nano DNA library prep manual (Illumina; San Diego, CA, USA). Genomic DNA (100 ng) was sheared using an LE220 Focused-ultrasonicator (Covaris; Woburn, MA, USA) with a duty factor of 15%, peak incident power of 450 W, 200 cycles per burst for 50 s. Sheared DNA fragments of around 350 bp were obtained according to the manufacturer's instructions. DNA libraries were enriched after ligating the indexing adapters to the ends of the DNA fragments. The quality of the libraries was evaluated using TapeStation 2200 (Agilent; Santa Clara, CA, USA), quantified by using PicoGreen® dsDNA quantitation assay (Thermo Fisher Scientific; Waltham, MA, USA), and measured on a Victor3 plate reader (PerkinElmer; Waltham, MA, USA). A unique "bridged" amplification reaction was utilized for sequencing. A flow cell containing the libraries was prepared and then loaded on to the HiSeq X-10 sequencer (Illumina) for automated cycles of extension and imaging.

#### *4.12. Patient-Derived Xenograft (PDX) Study*

Animal experimentation was performed in accordance with the guidelines approved by the Institutional Animal Care and Use Committee (IACUC) and IRB (20180416-1402-054-555). Fresh human breast cancer tissue was obtained through a gun biopsy from hormone-receptor-positive, HER2-negative IDC breast cancer patient with *BRCA2* 580del4 mutation, who acquired resistance to olaparib. The obtained tissue was cut ~2-mm pieces within 1 h while covered in gauze dipped in saline and transplanted into six-week-old, severe combined immunodeficient (NOG) female mice (F0). Fresh human breast cancer tissue was cut into ~2-mm pieces and transplanted into 6-week-old, severe combined immunodeficient (NOG) female mice (F0). When the tumor volumes reached to 1.5 cm in diameter, it was dissected and re-implanted into another set of mice (F1). When the tumor volume of F1 reached 150 mm3, the mice were randomly divided into three groups (4 mice per each group) and olaparib (50 mg/kg, p.o, once daily), irinotecan (10 mg/kg, i.p, twice a week), or vehicle (0.5% hydroxypropyl methylcellulose, p.o, once daily) were administered. Tumor volume was measured three times a week by caliper and calculated using ((width)2 <sup>×</sup> (height))/2. All mice were sacrificed with CO2 on 21 days after the treatment.

#### *4.13. Statistical Analysis*

All experiments were conducted independently at least three times, and statistical analyses were performed using SigmaPlot version 9.0 (Statistical Package for the Social Sciences, Chicago, IL, USA). Two-sided Student's t-test was used when appropriate. The results are expressed as the mean ± standard error (SE). A *p*-value < 0.01 was considered to represent statistical significance.

#### *4.14. Ethics Declarations*

The PDX study was approved by the Institutional Animal Care and Use Committee (IACUC) and IRB (20180416-1402-054-555).

#### **5. Conclusions**

A carryover effect of olaparib-treatment, sensitizing to subsequent treatment, was suggested in a clinical trial. To explore the possibility and the mechanism of this effect, olaparib-resistant gastric cancer cells were tested with several chemotherapeutic agents, and underlying mechanisms were explored. Olaparib-resistant gastric cancer cells show the compensatory alterations in DNA-damage response pathways, and exhibit cross-resistance to cisplatin; however, these cells are highly sensitive to subsequent treatment with irinotecan through tyrosyl-DNA phosphodiesterase 1 (TDP1) downregulation with increased topoisomerase 1 (TOP1) activity. These findings have substantial implications on subsequent therapies for patients with olaparib-resistant cancers.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/2/334/s1, Figure S1. The populations of apoptotic cells following irinotecan or olaparib treatment; Figure S2. Protein expression levels of parental and olaparib-resistant cells; Table S1. Sequences of the siRNAs.

**Author Contributions:** J.W.K. and A.M. designed the study, performed the experiments, interpreted the data, and wrote the manuscript. H.J. and Y.J.K. assisted the PDX experiment. H.-J.K., K.-H.L., T.-Y.K., K.W.L., D.-Y.O., J.-H.K., Y.-J.B., and S.-A.I. supplied the clinical sample, conceived of the study, and participated in coordination. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by Korean Cancer Foundation (K20121110), and funded by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) and by the Ministry of Health & Welfare, Korea (grant number: HI14C1277).

**Conflicts of Interest:** Seock-Ah Im is a recipient of research funds from AstraZeneca Inc. and Pfizer and has consultant and advisory role for AstraZeneca, Hanmi Corp, Novartis, Pfizer, Roche, and Spectrum. The other authors declared that they have no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **PARP1 Inhibition Augments UVB-Mediated Mitochondrial Changes—Implications for UV-Induced DNA Repair and Photocarcinogenesis**

**Csaba Heged ˝us 1, Gábor Boros 2, Eszter Fidrus 1, Gréta Nikoletta Kis 3, Miklós Antal 3, Tamás Juhász 3, Eszter Anna Janka 1, Laura Jankó 4,5, György Paragh 6, Gabriella Emri 1, Péter Bai 4,5,7,**† **and Éva Remenyik 1,\*,**†


Received: 26 November 2019; Accepted: 10 December 2019; Published: 18 December 2019

**Abstract:** Keratinocytes provide the first line of defense of the human body against carcinogenic ultraviolet (UV) radiation. Acute and chronic UVB-mediated cellular responses were widely studied. However, little is known about the role of mitochondrial regulation in UVB-induced DNA damage. Here, we show that poly (ADP-ribose) polymerase 1 (PARP1) and ataxia-telangiectasia-mutated (ATM) kinase, two tumor suppressors, are important regulators in mitochondrial alterations induced by UVB. Our study demonstrates that PARP inhibition by ABT-888 upon UVB treatment exacerbated cyclobutane pyrimidine dimers (CPD) accumulation, cell cycle block and cell death and reduced cell proliferation in premalignant skin keratinocytes. Furthermore, in human keratinocytes UVB enhanced oxidative phosphorylation (OXPHOS) and autophagy which were further induced upon PARP inhibition. Immunoblot analysis showed that these cellular responses to PARP inhibition upon UVB irradiation strongly alter the phosphorylation level of ATM, adenosine monophosphate-activated kinase (AMPK), p53, protein kinase B (AKT), and mammalian target of rapamycin (mTOR) proteins. Furthermore, chemical inhibition of ATM led to significant reduction in AMPK, p53, AKT, and mTOR activation suggesting the central role of ATM in the UVB-mediated mitochondrial changes. Our results suggest a possible link between UVB-induced DNA damage and metabolic adaptations of mitochondria and reveal the OXPHOS-regulating role of autophagy which is dependent on key metabolic and DNA damage regulators downstream of PARP1 and ATM.

**Keywords:** UVB; PARP; mitochondria; metabolism; biogenesis; autophagy; carcinogenesis; DNA repair

#### **1. Introduction**

Mitochondria regulate their shape, number, distribution, mass, content of mitochondrial DNA (mtDNA), and metabolic capacity in a process called mitochondrial biogenesis, which requires the orchestration of complex transcriptional control of both nuclear and mitochondrial genes [1,2]. The function of mitochondrial biogenesis is to provide quality control of mitochondria by regulating mitochondrial fission, fusion, and mitophagy [3,4] to maximize the energy utilization of mitochondria [5] to meet cellular and environmental demands. Imbalances or perturbations in these processes can lead to mitochondrial dysfunction [3,6].

Accumulating evidence suggests that mitochondria also play a central role in skin physiology. Although, involvement of other organ systems predominates in classical mitochondrial disorders, several cutaneous diseases can be linked to mitochondrial dysfunctions [7]. Interestingly, mitochondria lack functional nucleotide excision repair (NER) pathway [8,9] which is responsible for the removal of ultraviolet (UV)-induced DNA lesions including cyclobutane pyrimidine dimers (CPD). Accumulation of these DNA photoproducts in mtDNA leads to mutations and deletions resulting in mitochondrial alterations which have been associated with photoaging [10,11] and are present in melanoma [12], as well as in non-melanoma skin cancers [13,14]. The other types of mitochondrial alterations such as upregulated oxidative phosphorylation (OXPHOS), mitochondrial membrane hyperpolarization, and decreased mitophagy are frequently observed in patients with DNA repair deficiencies [15–18]. Even though, these DNA repair proteins are encoded by the nuclear genome, their absence lead to mitochondrial functional changes emphasizing the importance of nucleus-to-mitochondria (NM) signaling [19].

Growing evidence suggests that the key component of NM signaling is poly (ADP-ribose) polymerase 1 (PARP1) activation [19,20]. PARP1 is a multifunctional zinc-finger protein involved in the regulation of DNA repair, chromatin structure, cell cycle, calcium homeostasis, transcription regulation, cell death, immune response, and metabolism [21,22]. Through either direct interaction or via poly (ADP-ribose) (PAR) polymer formation PARP1 can modulate the activity of ataxia-telangiectasia mutated kinase (ATM) [23] and tumor suppressor protein 53 (p53) [24,25] involved in DNA damage response. Enhanced PARP1 activity also induces ATP depletion [26], which implies the activation of AMPK-activated protein kinase (AMPK) [27,28] that is responsible for the regulation of various cellular pathways including protein kinase B (AKT) and mammalian target of rapamycin (mTOR) [29–31]. This complex interplay between PARP1, ATM, AMKP, p53, AKT, and mTOR indicates that these DNA damage responders can fine-tune and modulate the interaction between DNA repair pathways with metabolism [32]. PARP1 and PARP2 also modulate mitochondrial activity through NAD<sup>+</sup> depletion [33–35]. Pharmacological inhibition or deletion of PARP1 and PARP2 improves mitochondrial function and protects against mitochondrial [36,37], metabolic [34], and neurological diseases [38]. PARP inhibitors became valuable tools in treating cancer cells harboring DNA repair defect with the combination of either radio- or chemotherapy [39–41]. In addition to several orally-administered PARP inhibitors that are under active clinical development, recently ABT-888 (veliparib) has emerged as an effective drug in treating various solid tumors [42] partially via modulating mitochondrial activity [43,44].

Although UV radiation has been shown to trigger morphological [45,46] and functional changes of mitochondria [47–49], published data led to contradictory results most likely due to the diversity of applied UV spectrum, UV dose, and cell type. Studies using UVC irradiation revealed that UVC induced mitochondrial hyperfusion and resulted in enhanced ATP synthesis via oxidative phosphorylation (OXPHOS) [50]. Other authors demonstrated that UVC irradiation caused a significant increase in mitochondrial content, oxygen consumption, and fatty acid oxidation [28]. Nevertheless, the functional consequences of mitochondrial alterations after UVB-induced DNA damage and the molecular pathways leading to mitochondrial changes remain to be elucidated.

In this study, we defined the central role of PARP1 and the linked molecular pathways in mediating UVB-induced DNA damage response and mitochondrial changes in a clinically relevant human keratinocyte cell line.

#### **2. Results**

#### *2.1. PARP Inhibition Impairs CPD Repair, Augments UVB-Induced Cell Cycle Block, Apoptosis and Reduces Keratinocyte Proliferation*

To explore the effect of UVB on PARP activation, firstly, we investigated poly (ADP-ribose) polymer (PAR) formation in a time-dependent manner in human immortalized keratinocytes exposed to mid (20 mJ/cm2) or high (40 mJ/cm2) dose of UVB (Figure 1A). In HaCaT cells, poly (ADP-ribosyl) ation (PARylation) signal was initially observed at and over 95 kDa at 5 min after UVB exposure and the extent of PARP activation was dose-dependent with higher PARP activation after 40 mJ/cm<sup>2</sup> UVB exposure. The signal was lost when cells were treated with ABT-888, a pan-PARP inhibitor (for uncut PAR western see Supplementary Figure S2). PARP1 is considered to contribute 85–90% to total PARP activity, the rest is largely the activity of PARP2 [51–53]. Moreover, the size of the PARylated band suggests PARP1 (auto) PARylation indicating the involvement of PARP1 in UVB-induced damage. Since PARP1 is involved in regulation of various DNA repair pathways, we wanted to assess how PARP inhibition (PARPi) regulates the removal of cyclobutane pyrimidine dimers (CPDs) introduced by UVB (Figure 1B). A slow decline of CPDs after UVB exposure was observed, reflective of nucleotide excision repair (NER) activity. In contrast, the relative amount of CPDs in PARP-inhibited cells remained elevated even after 24 h compared to the UVB-irradiated cells suggesting impaired efficiency of NER, similarly as described by King et al. [54]. Since unrepaired photolesions can initiate cell cycle block to prevent cells with DNA damage from entering mitosis, we performed cell cycle analysis (Figure 1C,D). A higher proportion of PARP-inhibited cells after 20 mJ/cm<sup>2</sup> UVB accumulated in G2/M phase of cell cycle characteristic of enhanced DNA damage [21]. ABT-888 treatment sensitized cells to apoptosis after UVB as reflected by decreased cell viability 24 h post-UVB compared to controls (Figure 1E). Gene silencing of PARP1 showed similar changes regarding cell viability as PARP inhibition. Furthermore, long-term keratinocyte survival using May-Grünwald-Giemsa staining in clonogenic assay showed that ABT-888 treatment led to a marked decrease in the number of keratinocyte clones after UVB-irradiation (Figure 2A,B). To find out whether PARPi-induced retention of CPDs induce higher mutation rate, we performed hypoxanthine-guanine phosphoribosyltransferase (HPRT) mutation assay (Figure 2C,D). We could not adjust the method for HaCaT keratinocytes as these cells were extremely tolerant to the 6-thioguanine selection medium. Therefore, we used Chinese hamster ovary (CHO) cells which is a widely used cell line for HPRT mutation assay [55–57] and also show high level of p53 protein level due to p53 mutation [58,59] as HaCaT cells do [60]. In this case, we used lower UVB doses to allow the accumulation of mutations compared to higher UVB which potentially triggers apoptosis. Although, both 10 and 20 mJ UVB resulted in increased number of HPRT-mutated cells, interestingly PARP inhibition caused a marked reduction in the number of mutated cell colonies suggesting that PARPi initiate apoptosis of cells with high CPD content instead of allowing the accumulation of gene mutations. Interestingly, some DNA damage markers, including cell viability, CPD and colony formation, cell cycle progression showed no significant difference between the vehicle and ABT-888 treated groups after 40 mJ/cm2. This phenomenon can be due to the fact that 40 mJ/cm2 UVB dose in our experiments represents such high DNA damage that cannot be augmented by PARP inhibition. However, PARP1 knockdown cells displayed significantly lower cell viability compared to control siRNA-transfected cells even after 40 mJ/cm2 UVB irradiation.

**Figure 1.** Poly (ADP-ribose) polymerase (PARP) inhibition impairs cyclobutane pyrimidine dimer (CPD) repair, augments ultraviolet B (UVB)-induced cell cycle block and apoptosis. (**A**) Time-course of PARP activity (PAR) after 20 and 40 mJ/cm2 UVB exposure and 25 μM ABT-888 were analyzed by Western blot (*n* = 3). (**B**) Cells were exposed to a single dose of 20 or 40 mJ/cm<sup>2</sup> UVB and collected at various time points for DNA extraction. CPD formation was determined by ELISA (*n* = 4). (**C**,**D**) Cell cycle progression was evaluated by propidium iodide staining after 24 h. DNA content was analyzed in FL2-A (*n* = 4). (**E**) Cell viability, apoptosis, and necrosis was assessed by dual labelling with Annexin V-Alexa 488 and propidium iodide 24 h post-UVB. Double negative cells are considered as viable (*n* = 5). (**F**) Cell viability was measured similarly as in Figure 1E after PARP1 knockdown (*n* = 3). −/+ represent vehicle (−) or ABT-888 (+) treatment \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. Error bars represent SEM.

**Figure 2.** Poly (ADP-ribose) polymerase (PARP) inhibition decreases clone formation and ultraviolet B (UVB)-induced mutation rate. (**A**,**B**) Colony formation assay of HaCaT cells after 10 days post-UVB exposure was assessed by clonogenic assay (*n* = 4). (**C**,**D**) HPRT mutation assay was carried out on CHO cells. Preselected hypoxanthine-guanine phosphoribosyltransferase (HPRT) mutant cells were cultured for 10 days post-UVB (*n* = 3). −/+ represent vehicle (−) or ABT-888 (+) treatment \* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.001, respectively. Error bars represent SEM.

#### *2.2. PARP Inhibition Enhances UVB-Mediated Mitochondrial Biogenesis*

Mitochondrial biogenesis, by promoting the growth, formation, and assembly of newly synthesized mitochondria, has recently been linked to cancer development [61], apoptosis [62–64], and DNA damage [18,28,65]. Accumulating evidence suggest that DNA damage can initiate mitochondrial biogenesis which is accompanied by elevation in mitochondrial number, area, and mass [18,28,65,66]. Transmission electron microscopic images revealed that UVB-treated cells contain more and longer cristae than non-irradiated cells (Figure 3A). This morphological alteration became more pronounced after PARP inhibition. Similarly, UVB treatment increased both mitochondria number and total mitochondrial area (Figure 3B,C). ABT-888 treatment resulted in further increase in these parameters suggesting that PARP inhibition may boost UVB-mediated mitochondrial response. Since mitochondrial content changes with the balance between mitochondrial biogenesis and turnover, we quantified mitochondrially encoded cytochrome C oxidase I (MTCO1)/succinate dehydrogenase complex, subunit A (SDHA) ratio that is a marker of mitochondrial biogenesis. This experiment demonstrated that the mitochondrially-encoded MTCO1 show strong induction after UVB irradiation and become even more expressed after PARPi, while the expression of the nuclearly-encoded SDHA protein is unaltered (Figure 3D,E). The higher mitochondrial mass after both UVB and PARPi (Figure 3F) and the enhanced expression of the master regulators of mitochondrial biogenesis including mitochondrial transcription factor A (TFAM), nuclear respiratory factor 2 (NRF2), and estrogen-related receptor alpha (ERRA) (Figure 3G) also suggest that PARPi augments the UVB-triggered mitochondrial biogenesis.

**Figure 3.** Poly (ADP-ribose) polymerase (PARP) inhibition enhances ultraviolet B (UVB)-mediated mitochondrial biogenesis. (**A**) Effect of UVB irradiation and PARP inhibition on mitochondrial ultrastructure visualized by transmission electron microscopy (TEM) 24 h after UVB exposure. Enlarged pictures are displayed at the right bottom corner. Scale bar is presented on the lower panels. (**B**) Mitochondrial number and (**C**) area were calculated from TEM images (minimum 7 cells were analyzed). (**D**,**E**) Mitochondrial biogenesis was quantified by the ratio of the mitochondrially encoded cytochrome C oxidase I (MTCO1) and succinate dehydrogenase complex, subunit A (SDHA) expression 24 h post UVB (*n* = 3). (**F**) Mitochondrial mass was determined by Mitotracker Green labeling 24 h after UVB irradiation (*n* = 3). (**G**) mRNA levels of master regulators of mitochondrial biogenesis were quantified by real-time PCR 24 h post-UVB (*n* = min. 3). −/+ represent vehicle (−) or ABT-888 (+) treatment. \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. Error bars represent SEM.

#### *2.3. PARP Inhibition Augments UVB-Mediated Mitochondrial Fusion*

To identify if UVB and PARPi also alters mitochondrial dynamics, we evaluated mitochondrial morphology using confocal microscopy. Non-irradiated cells mainly contained fragmented mitochondria which normally represents low metabolic activity. After 40 mJ/cm2 UVB, a statistically significant reduction in the frequency of fragmented mitochondria and an elevation in tubular mitochondria was detected compared to the non-irradiated control. The frequency of intermediate mitochondria was increased after PARPi at 0 mJ UVB compared to the vehicle control, and we also observed decreased fragmented mitochondrial frequency and higher percentage of tubular mitochondria after 20 mJ/cm2 + ABT-888 treatment compared to the 20 mJ/cm<sup>2</sup> UVB exposed cells (Figure 4A). The dose-dependent effect of UVB in the branching aspect of mitochondria (Figure 4C) as defined by form factor ((Perimeter2/(4<sup>π</sup> <sup>×</sup> area)) was also observed suggesting enhanced mitochondrial fusion after UVB and PARP inhibition. Finally, to confirm that the observed mitochondrial fusion are regulated by the dynamin-related proteins, we checked the expression of mitofusin-1 (Mfn1), mitofusin-2 (Mfn2), and optic atrophy 1 (OPA1) (Figure 4D,E). Similarly to mitochondrial morphological alterations, we detected enhanced protein expression of Mfn1, Mfn2, and OPA1 which support mitochondrial fusion at the outer and inner mitochondrial membrane. These results clearly indicate that besides enhancing mitochondrial biogenesis, UVB also triggers mitochondrial fusion and increases the complexity of mitochondrial network which is more prominent after PARP inhibition.

#### *2.4. PARP Inhibition and UVB Induces Bulk Autophagy but Not Mitophagy*

Several DNA-damaging agents were shown to initiate autophagy [67,68] and mitophagy [69] to remove damaged macromolecules or organelles including mitochondria, which prompted us to explore if UVB and PARPi-induced mitochondrial biogenesis and morphological changes of mitochondria affects autophagy or mitophagy. We used dual labelling of Mitotracker CMxROS to stain mitochondria and microtubule-associated proteins 1A/1B light chain 3B (LC3A/B) antibody to detect autophagosomes. Our results show the accumulation of LC3-positive cells after UVB exposure. PARP inhibition augmented autophagy not only in UVB-exposed cells, but also under non-irradiated conditions (Figure 5A,B). LC3B western blotting (Figure 5C) also confirmed these results indicating the autophagy inducer role of ABT-888. PARP1 knockdown also showed similar results but with much more pronounced autophagy induction after UVB (Figure 5E). Interestingly, the role of PARP1 in the regulation of autophagy is controversial. Both autophagy inductor [70–72] and inhibitor [73–75] role of PARP1 has been described suggesting that the autophagy-modulatory effect of PARylation might show DNA damage and cell type specificity. Even though, we detected an increased number of LC3 puncta, this type of macroautophagy cannot be considered as mitophagy since autophagic puncta show very mild colocalization with mitochondria (Figure 5D). Furthermore, a dose-dependent mitochondrial elongation was observed after UVB and PARP inhibition in Figure 4A,B indicating mitochondrial fusion. Since fused mitochondria are protected from mitophagy [76], and mitophagy-coupled elimination of mitochondria is usually preceded by fission [69] and decline in mitochondrial function [77], the here experienced induction in autophagy is considered as general autophagy and suggest that mitochondrial morphological rearrangements induced by UVB and PARPi may interfere with the initiation of mitophagy.

**Figure 4.** Poly (ADP-ribose) polymerase (PARP) inhibition augments ultraviolet B (UVB)-mediated mitochondrial fusion. (**A**) Effect of UVB irradiation and PARP inhibition on mitochondrial morphological subtypes (tubular, intermediate, and fragmented) are quantified based on confocal microscopic images 24 h post-UVB (minimum 29 cells). (**B**) Mitochondrial morphology visualized by confocal microscopy with Mitotracker Red CMXRos dye (*n* = 3). Enlarged pictures are displayed at the right bottom corner. Scale bar is presented on the images. (**C**) The branching aspect of mitochondria was derived from confocal microscopic images and was represented as a form factor (*n* = 3). (**D**) Protein expression of mitofusin-1 (Mfn1), mitofusin-2 (Mfn2), and optic atrophy 1 (OPA1) was visualized by Western blot. (**E**) Expression of mitochondrial fusion proteins were analyzed by Western blot 24 h post-UVB (*n* = min. 3). −/+ represent vehicle (−) or ABT-888 (+) treatment. \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. Error bars represent SEM.

**Figure 5.** Poly (ADP-ribose) polymerase (PARP) inhibition and ultraviolet B (UVB) induces bulk autophagy but not mitophagy. (**A**) Expression of microtubule-associated proteins 1A/1B light chain 3B (LC3A/B) protein marker of autophagy by confocal microcopy (*n* = 3). Scale bar is presented on the figure. (**B**) Quantification of LC3A/B puncta in cells derived from confocal images in Figure 5A (*n* = 3). (**C**) LC3A/B proteins were quantified by Western blotting. Brightness and contrast were adjusted. Protein of interest were normalized to the loading control B-actin. (**D**) Dual staining of LC3-puncta and mitochondria by confocal microscopy (*n* = 3). (**E**) Quantification of LC3A/B puncta in PARP1 siRNA and control siRNA-transfected cells derived from confocal images (*n* = 4) −/+ represent vehicle (−) or ABT-888 (+) treatment \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05, *p* < 0.01 and *p* < 0.001, respectively. Error bars represent SEM.

#### *2.5. PARP Inhibition Boosts UVB-Mediated Mitochondrial Bioenergetic Changes*

Alterations in the mitochondrial network is strictly controlled by intra- or extracellular signals connecting mitochondrial biogenesis and fusion with energy perturbations. To test whether increased mitochondrial biogenesis and fusion after UVB and PARP inhibition supports mitochondrial activity, we evaluated mitochondrial parameters. Our results show that UVB exposure increased mitochondrial membrane potential (Figure 6A) and total ATP level (Figure 6B). PARPi boosted the UVB-induced mitochondrial membrane hyperpolarization and it also raised ATP level compared to UVB-irradiated samples. Mitochondrial membrane potential and ATP levels change independently, and hyperpolarization of mitochondrial membrane potential does not necessarily correlate with ATP production but it may stem from decreased F0F1 ATP synthase activity and concomitant lower ATP production. Furthermore, increased ATP level may result either from enhanced glycolysis, increased respiration via electron flow from complex I-V, or decreased energy expenditure. In order to clarify the reasons of the elevation in ATP levels, we complemented our data with quantitative analysis of metabolic flux by XF96 oximeter and monitored extracellular acidification rate (ECAR) indicating glycolysis and oxygen consumption (OCR) representing oxidative phosphorylation with sequential addition of oligomycin and antimycin. Although basal ECAR (Figure 6C) is statistically unchanged after UVB exposure, PARP inhibition enhanced the rate of glycolysis after 20 mJ/cm2 UVB dose and at non-irradiated conditions. In contrast, UVB dose-dependently increased oxygen consumption and ABT-888 treatment significantly augmented OXPHOS compared to 0 or 20 mJ/cm2 UVB (Figure 6C). Citrate synthase (CS) activity, the initial enzyme in the tricarboxylic acid (TCA) cycle catalyzing the formation of citrate from oxalacetate and acetyl-CoA, showed significant elevation after UVB radiation, which was augmented by PARP inhibition (Figure 6C). This suggest that not only the distal part of respiratory chain is altered by UVB and PARPi, but also increased terminal oxidation is preceded by elevation in TCA activity as well. To exclude the possibility of the off-target effects of ABT-888, we confirmed the mitochondrial changes by PARP1 silencing. We detected increased mitochondrial mass reflecting mitochondrial biogenesis after UVB in control siRNA-transfected keratinocytes that was elevated by PARP1 knockdown (Figure 6D). Mitotracker Red CMXRos incorporation into mitochondria reflecting mitochondrial membrane potential also showed similar changes similarly after vehicle and ABT-888 treatment (Figure 6E). Although, we could not detect difference in basal ECAR between the control and PARP siRNA-transfected cells, CS activity and OXPHOS showed significant changes (Figure 6F) confirming the pivotal role of PARP1 in the UVB-induced mitochondrial changes. In summary, we can conclude that elevation in cellular ATP level after UVB is due to enhanced energy production and cells increase their energy reserves through both glycolysis, TCA cycle, and OXPHOS after PARP inhibition.

**Figure 6.** PARPinhibition boosts UVB-mediatedmitochondrial bioenergetic changes. (**A**) For determination of mitochondrial membrane potential cells were stained with 3,3 -dihexyloxacarbocyanine iodide (DIOC6(3)) 24 h post-UVB and measured by flow cytometry in FL1 channel (*n* = 4). (**B**) Total cellular ATP level was quantified by colorimetric assay (*n* = 3). (**C**) Basal extracellular acidification rate (ECAR) represents glycolysis after 24 h post-UVB. XF medium was supplemented with 10 mM glucose (*n* = 6). After four oxygen consumption rate (OCR) measurement oligomycin and antimycin was used to determine mitochondria-linked ATP production and basal OCR, respectively (*n* = 6). Citrate synthase activity, the initial enzyme of the tricarboxylic acid (TCA) cycle was measured by citrate synthase kit (*n* = 3). (**D**) Mitochondrial mass determined similarly as in Figure 3F (*n* = 4). (**E**) Mitotracker Red CMXRos mean fluorescence intensity was measured by flow cytometry (*n* = 4). (**F**) Metabolic parameters including glycolysis, citrate synthase activity, and oxidative phosphorylation (OPXHOS) was detected similarly as in Figure 6C (*n* = 2 for ECAR and OCR, *n* = 3 for citrate synthase (CS) activity). \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

#### *2.6. PARP Inhibition Restores NAD*<sup>+</sup> *Level and SIRTUIN Expression*

NAD<sup>+</sup> and NADH plays a central role in redox homeostasis and cellular metabolism including glycolysis via glyceraldehyde-3-phosphate dehydrogenase (GAPDH) activity [78] and OXPHOS by regulating the transfer of electrons to complex I [79]. UVB irradiation caused a slight but significant decrease in NAD<sup>+</sup> level after both 20 and 40 mJ/cm2 UVB (Figure 7A). ABT-888 treatment efficiently restored intracellular NAD<sup>+</sup> content suggesting the role of PARP1 in UVB-mediated NAD<sup>+</sup> depletion. It has been known that a decrease in NAD<sup>+</sup> by enhanced PARP activation inhibits another NAD+-consuming enzyme family, the class III histone deacetylase Sirtuins [33,34,36–38,80] which regulate diverse cellular processes [81] including mitochondrial metabolism and have intricate relationships with PARPs [82,83]. To test whether UVB modulate Sirtuin expression and PARPi can restore their expression after UVB similarly as NAD<sup>+</sup> level changes, we chose the 40 mJ/cm<sup>2</sup> UVB dose which caused a more significant increase in NAD<sup>+</sup> level after ABT-888 treatment compared to UVB-irradiated cells. We detected slight mRNA downregulation in SIRT1, SIRT2, SIRT3, SIRT4, SIRT5, and SIRT7 after 40 mJ/cm<sup>2</sup> UVB dose (Figure 7B) and PARP inhibition increased the gene expression of all Sirtuins suggesting that intracellular NAD<sup>+</sup> availability after UVB and PARPi may regulate SIRTs expression. Although, the role of Sirtuins in the UVB-mediated cellular response and skin physiology is poorly characterized [84–88], here we revealed the Sirtuin expression-modulatory effect of NAD<sup>+</sup> after UVB and PARP inhibition. To explore the potential role of NAD<sup>+</sup> and Sirtuins in supporting glycolysis or oxidative phosphorylation further experiments are needed.

**Figure 7.** PARP inhibition restores NAD<sup>+</sup> level and Sirtuin expression. (**A**) Total cellular NAD<sup>+</sup> were quantified by colorimetric assay 24 h after UVB irradiation. Absorbance was measured at 490 nm (*n* = 2). (**B**) mRNA expression of the Sirtuin enzyme family 24 h post-UVB (40 mJ, *n* = 3). \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. ## indicate significant difference at *p* < 0.01 compared to non-irradiated control. Error bars represent SEM.

#### *2.7. PARP Inhibition Enhances UVB-Mediated Upregulation of Metabolic Proteins*

To explore the potential mechanisms responsible for elevated oxidative phosphorylation after UVB and PARP inhibition, we checked the expression of metabolic proteins involved in oxidative phosphorylation by Western blot (Figure 8A,B). Consistent with their role in mitochondrial activity, we detected significant increase in ATM, p-ATM, SIRT1, peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1A), AMPK, p-AMPK, p53, p-p53, p-AKT, and p-p70S6K1 (mTOR activity) expression after UVB which was further enhanced by PARP inhibition. The most prominent and statistically significant upregulation in these proteins can be detected after 40 mJ/cm2 UVB dose similarly as OXPHOS was increased to two-fold at *p* < 0.001. PARPi induced statistically significant difference at 0 and 20 mJ/cm<sup>2</sup> compared to vehicle control similarly as ABT-888 increased OXPHOS as well. PARP1 knockdown also induced upregulation in these proteins compared to siRNA control. These results suggest that these proteins switch on oxidative metabolism and their elevated expression may be responsible for higher oxygen consumption after UVB and ABT-888 treatment.

**Figure 8.** PARP inhibition enhances UVB-mediated upregulation of metabolic proteins. (**A**) Metabolic proteins involved in oxidative phosphorylation were analyzed in total protein lysates. Protein of interest were normalized to the loading control B-actin. Cells were harvested 24 h after UVB. Brightness and contrast were adjusted. (**B**) Densitometric representation of Figure 8A proteins (*n* = min. 3). (**C**) Protein expression involved in oxidative phosphorylation was confirmed by PARP1 knockdown. \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. Error bars represent SEM.

#### *2.8. PARP Inhibition and UVB-Induced Oxidative Phosphorylation and Autophagy Are Dependent on ATM, AMPK, p53, AKT, and mTOR Activation*

To test their requirement for increased oxidative phosphorylation, we applied KU-60019 as an ATM inhibitor (Figure 9A) and ATM siRNA for gene silencing (Figure 10A), compound C as an AMPK inhibitor, Pifithrin-alpha-HBr as a p53 inhibitor, Wortmannin as a phosphoinositide 3-kinase (PI3Ki)/AKT inhibitor and Rapamycin as a mTOR inhibitor (Figure 9A). Chemical inhibition of these proteins led to significant reduction in OXPHOS highlighting their role in mediating metabolic alterations after UVB and PARP inhibition. ATM inhibition (Figure 9B) and ATM knockdown (Figure 10B) also decreased the phosphorylation of AMPK, p53, AKT, and p70S6K1 which suggest that elevation in OXPHOS after UVB is downstream of ATM, which is known as one of the most important players in DNA damage response besides PARP1. Since CPD removal is mediated by the energetically demanding nucleotide excision repair, keratinocytes may try to compensate their increased DNA damage with a more intense mitochondrial activity to facilitate CPD elimination. In this respect, OXPHOS seems to be a beneficial response after UVB. To confirm this hypothesis, we applied inhibitors that decreased (ATMi, AMPKi, p53i, PI3Ki, mTORi) or limit the activity of mitochondrial electron transport chain (oligomycin and rotenone) or inhibits mitochondrial protein synthesis (chloramphenicol) (Figure 9C). Our data show that all these inhibitions led to a more pronounced cell death after UVB which confirms that increased mitochondrial activity bears an adaptive and antiapoptotic response after UVB. We also wanted to explore if UVB and PARPi-triggered general autophagy (Figure 5A,B) by recycling damaged organelles and macromolecules serves to provide metabolites for OXPHOS. To confirm this hypothesis, we tested the ATM, AMPK, p53, PI3K, and mTOR inhibitors after 40 mJ/cm2 UVB that caused the most prominent decrease in oxygen consumption (Figure 9A). Chemical inhibition of these proteins induced significant reduction in LC3B (Figure 9D) and Parkin (Figure 9E) expression similarly as UVB and PARPi-induced OXPHOS decreased after ATM, AMPK, p53, PI3K, and mTOR inhibitors indicating a parallel reduction in oxidative phosphorylation, autophagy, and PARKIN expression (a marker for mitophagy) as well. Reduction in PARKIN expression suggests impaired mitochondrial quality control, allowing the accumulation of damaged mitochondria, corrupting mitochondrial function, eventually leading to decreased OXPHOS. These results together suggest that after UVB and ABT-888 treatment, AMPK, p53, AKT, and mTOR are the main mediators of oxidative phosphorylation downstream of ATM kinase, and autophagy by recycling damaged cellular parts after stress response may provide metabolites to mitochondria and supports OXPHOS after UVB exposure and PARP inhibition.

**Figure 9.** *Cont.*

**Figure 9.** PARP inhibition and UVB-induced oxidative phosphorylation and autophagy are dependent on ATM, AMPK, p53, AKT, and mTOR activation. (**A**) To determine key proteins involved in mediating mitochondrial changes ATMi (KU-60019), AMPKi (Compound C), p53i (Pifithrin-alpha-HBr), PI3Ki (Wortmannin), and mTORi (Rapamycin) were added to the medium and OCR was measured as in Figure 6C (*n* = min.3). (**B**) ATM downstream signaling pathway was investigated by Western blot with the addition of its pharmacological inhibitors KU-60019 (*n* = 3). Brightness and contrast were adjusted. (**C**) Cell viability was measured by flow cytometry as in Figure 1E after 40 mJ/cm<sup>2</sup> UVB with oligomycin, rotenone, chloramphenicol, ATMi, AMPKi, p53i, PI3Ki, and mTORi (*n* = min. 4). To determine the involvement of ATM, AMPK, p53, AKT, and mTOR in the regulation of (**D**) autophagy and (**E**) PARKIN expression, we applied their respective pharmacological inhibitors after 40 mJ/cm<sup>2</sup> UVB. (*n* = 4). \*; \*\* and \*\*\* indicate statistically significant difference at *p* < 0.05 and *p* < 0.01, *p* < 0.001, respectively. Error bars represent SEM. ATM: ataxia-telangiectasia-mutated kinase, AMPK: adenosine monophosphate-activated kinase, AKT: protein kinase B, mTOR: mammalian target of rapamycin, OCR: oxygen consumption rate.

**Figure 10.** ATM silencing after UVB showed similar results as ATM inhibition by KU-60019. (**A**) OCR was measured as in Figure 6C (*n* = 2) after ATM knockdown. (**B**) ATM downstream signaling pathway was investigated by Western blot with ATM silencing (*n* = 2). \* indicates statistically significant difference at *p* < 0.05. Error bars represent SEM.

#### **3. Discussion**

In this study, we identified morphological and functional changes of mitochondria after UVB and PARP inhibition. Early after UVB exposure, we detected PAR formation, reflecting PARP activity which was completely suppressed by PARP inhibition. We showed that loss of PARylation caused defective CPD repair compared to the UVB-irradiated cells suggesting impaired NER pathway as several PARP1-intercating partners (DDB1, DDB2, ATM, RAD51, ALC1, XPC) [89–91] have been described so far. Furthermore, PARP1 promotes the establishment of locally relaxed chromatin structure [92] to enable the removal of damaged DNA parts. Therefore, it is logical to assume that PARP inhibition renders chromatin to become compact, blocking not only replication, transcription but also the repair of UVB-induced photoproducts suggesting the particular importance of PARP1 in mediating transcription coupled nucleotide excision repair. Our results show a prolongation in G2/M phase of cell cycle after PARPi as described earlier [93–95]. Inhibition of PARP1 was synergistic with UVB with respect to G2/M accumulation characteristic of severe DNA damage after 20 mJ/cm2. G2/M accumulation can also be attributed to PARP trapping, in which inactivated PARP1 remains bound to DNA stalling replication, transcription leading to replicative stress and double-strand breaks [96,97] which make PARP inhibitors cytotoxic in combination with DNA damaging agents [98,99]. If DNA damage cannot be repaired, cells try to evade the accumulation of mutations by initiating apoptosis. Since apoptotic response after UVB cannot be ameliorated by PARPi, we can exclude the possibility that UVB-triggered cell death in our model system is PARP1-dependent, it may rather suggest that the accumulated and unrepaired CPDs augments the apoptotic response after PARPi. Long-term survival of keratinocytes revealed that the combined treatment of UVB and PARP inhibition severely reduced cell proliferation even at the non-irradiated conditions possibly suggesting the role of trapped PARP1-DNA complexes preventing cells division. We also detected that loss of PARylation decreased the frequency of mutations in UVB-treated cells as evidenced by decreased number of colonies after ABT-888 treatment suggesting that PARPi initiate apoptosis of cells with a high content of CPDs. Our results are in accordance with the anti-cancer effect of diverse PARP inhibitors and provide experimental evidence for the photosensitizing effect of PARP inhibitors experienced clinically [100,101]. Nevertheless, it is important to know that therapeutic application of PARP inhibitors might be associated with photosensitivity (sunburn) but not with an increased photocarcinogenesis risk.

Emerging evidence suggest that autophagy induction can be coupled to DNA damage response [67]. Accordingly, UVB triggered the accumulation of autophagosomes which corroborates with enhanced

LC3B expression. PARP inhibition resulted in enhanced autophagosome formation and higher LC3B protein level. The role of PARP1 in the regulation of autophagy is still poorly understood, both autophagy inducer [70] and inhibitory effect of PARP1 [75] can be found in the literature. In our model system, increased autophagosome formation after UVB and PARPi can be attributed to the elevated level of DNA damage (higher CPD content), which serves as a positive signal for autophagy since the initiation of global genome NER (GG-NER) subpathway is controlled by autophagy [67,68].

As Le Brace et al. [28] demonstrated, different DNA damaging agents including UVC which also induce CPD formation as UVB, led to AMPK activation and increased fatty acid oxidation. Similarly to their results, UVB radiation triggered mitochondrial fusion and induced mitochondrial biogenesis, which culminated in enhanced mitochondrial activity and complex metabolic events which were more prominent in PARP-inhibited cells. Increased number of mitochondria and mitochondrial area after UVB irradiation was more robust after PARPi. mRNA expression of mitochondrial biogenesis regulators, increased MTCO1/SDHA ratio, higher mitochondrial mass show similar trends suggesting that elevated mitochondrial content is indeed due to enhanced mitochondrial biogenesis and not defective mitochondrial turnover. By measuring mitochondrial function, we observed that PARP inhibition enhanced the UVB-mediated mitochondrial hyperpolarization, raised cellular ATP level and OXPHOS. Interestingly, we detected neither mitochondrial membrane depolarization nor ATP loss which normally occurs after PARP activation via NAD<sup>+</sup> resynthesis from ATP [102]. Since intracellular ATP concentration is also a regulator in DNA repair and a key factor in the cell's decision to die via apoptosis or necrosis [103], it is likely that this elevation in ATP level is required for the energetically costly nucleotide excision repair to remove UVB-induced photoproducts, and/or necessary for the apoptotic process as seen after staurosporine treatment [104]. Boost in mitochondrial OXPHOS was dependent on ATM, AMPK, p53, AKT, and mTOR phosphorylation since inhibition of them lead to significant reduction in oxygen consumption after both UVB and PARPi. Moreover, enhanced mitochondrial activity seems to induce a beneficial response after DNA damage, since decreasing mitochondrial biogenesis or ameliorating OXPHOS resulted in elevated cell death after UVB irradiation suggesting the protective effect of OXPHOS on cell viability. Inhibition of the ATM, AMPK, p53, PI3K, and mTOR pathways caused significant reduction in LC3A/B and Parkin expression emphasizing their role in autophagy and mitophagy. Decrease in mitophagy and the concomitant impaired mitochondrial quality control may also be an underlying cause for decreased OXPHOS after chemical inhibition of these proteins. We hypothesize that the purpose of autophagy induction after UVB-induced DNA damage is to recycle damaged molecules by fueling mitochondria with metabolites for oxidative phosphorylation as suspected by Dong et al. [105] and autophagy promotes the DNA damage recognition via NER. Since autophagy is an ATP-dependent process and cells must maintain their energy status to promote autophagy, we cannot exclude reciprocal regulation that is the shift toward oxidative metabolism regulates autophagy as described by Thomas et al. [106]. Nonetheless, in our model system it seems obvious that the purpose of autophagy and OXPHOS induction is to promote cell survival after UVB via ATM, AMPK, p53, AKT, and mTOR activation.

So how, increased mitochondrial activity is connected to DNA damage? The UVB-induced photolesions trigger local conformational changes on DNA similarly as seen after etoposide treatment [107] and UVC irradiation [108] which also resulted in increased mitochondrial biogenesis and activity [18,28]. Since, PARP-inhibited cells displayed higher level of CPDs compared to UVB-irradiated samples, it seems plausible that CPDs generated on DNA may be a trigger for mitochondrial changes. This is supported by the fact that the phosphorylation status of the main DNA damage responders (ATM, AMPK, p53, AKT, and mTOR) that are activated by UVB become more prominent after PARPi. ATM is mainly activated by DNA double-strand breaks but also a sensor of CPDs [109]. Furthermore, unrepaired CPD lesions in PARPi cells due to PARP trapping may eventually be converted into DNA breaks during replication which is an ultimate trigger for ATM activation. ATM can directly phosphorylate AMPK [110], p53 [111], and AKT [112] and via AKT and AMPK signaling ATM regulates mTOR activity [113], as well to turn on oxidative metabolism. AMPK can also phosphorylate p53 at

S15 [114], thereby potentiating its activity in enhancing mitochondrial metabolism. Furthermore, it is noteworthy that although cell type-dependently, AMPK [115], p53 [116] AKT [117], and mTOR [118] activation can trigger mitochondrial fusion which may also be a possible explanation for enhanced mitochondrial activity. Since PARP inhibition alone do not induce CPD formation, we must take into account that the PARPi-mediated changes in the non-irradiated cells are mediated by CPD-independent mechanisms either through autophagy induction and/or NAD<sup>+</sup> prevention. However, we can exclude the involvement of NAD<sup>+</sup> in the regulation of mitochondrial function after UVB, since morphological alteration of mitochondria and oxidative phosphorylation do not show similar changes as intracellular NAD<sup>+</sup> availability after UVB irradiation.

#### **4. Materials and Methods**

#### *4.1. Chemicals*

All chemicals were obtained from Sigma-Aldrich (St. Louis, MO, USA) unless stated otherwise.

#### *4.2. Cell Culture*

Human immortalized keratinocyte-derived (HaCaT) cell line were cultured in a T75 flask, as previously described [119] using 4500 mg/L Dulbecco's modified eagle media (DMEM) Glutamax supplement (Thermo Fisher Scientific, Waltham, MA, USA) containing 10% heat-inactivated fetal bovine serum (FBS) and 0.5% antibiotic/antimycotic solution. For HPRT mutation assay, Chinese hamster ovary (CHO) cells were used as described in Section 4.7. Cells were maintained in a humidified incubator at 37 ◦C with 5% CO2 atmosphere.

#### *4.3. Cell Treatment*

Cells were harvested with trypsin-EDTA (Biosera, Budapest, Hungary) then seeded in 12-well plate in 200,000 cell/well density (unless stated otherwise), and allowed to adhere for at least 12 h. At 80% confluence, cells were pretreated with the PARP inhibitor veliparib (ABT-888) (Selleckchem, Houston, TX, USA) at a final concentration of 25 μM. For detection of cell death, autophagy, mitophagy, and OXPHOS, the following inhibitors were used: ATMi: 6 μM KU-60019 (Adooq, Irvine, CA, USA), AMPKi: 2.5 μM compound C (Selleckchem), p53i: 30 μM pifithrin-alpha-HBr (AdooQ), mTORi: 300 nM rapamycin (AdooQ), PI3Ki: 300 nM wortmannin (AdooQ), mitochondrial activity inhibitors: 5 μM oligomycin, and 500 nM rotenone. To inhibit mitochondrial protein synthesis and hamper mitochondrial function chloramphenicol (AdooQ) (25 μM) was used. For UVB irradiation, cells were covered with 400 μl pre-warmed Dulbecco's phosphate-buffered saline (DPBS) (Lonza, Walkersville, MD, USA) and were subjected to 20 or 40 mJ/cm<sup>2</sup> UVB using two UVB broadband tubes (TL-20W/12 RS; Philips, Eindhoven, The Netherlands). Immediately after irradiation, the old medium was placed back on the cells to evade metabolic perturbations and cells were further cultured for 24 h. Controls were covered with tin foil during irradiation. The proper UV dose was measured with a UVX digital radiometer (UVP Inc., San Gabriel, CA, USA). Detached, dead cells were excluded from the experiments except for cell viability.

#### *4.4. Gene Silencing*

On-target plus SMARTpool PARP1 and ATM siRNA sequence was purchased from Dharmacon Research, Inc. (Lafayette, CO). Non-targeting siRNA (Dharmacon) was used as a control. HaCaT cells were seeded into 12-well plates in complete high-glucose DMEM containing 10% FBS without antibiotics. DharmaFECT transfection reagent (Dharmacon) in tube 1 and siRNA in 1 × siRNA buffer (Dharmacon) in tube 2 was diluted in serum and antibiotics-free DMEM and incubated at room temperature for 5 min. The two mixtures were combined and incubated further for 20 min and added to the cells at a final siRNA concentration of 50 nM. After 48 h, cells were washed with DPBS, irradiated with UVB, and the medium was replaced with a complete growth medium. Silencing efficiency was determined by Western blotting 24 h post-UVB.

#### *4.5. Cell Viability and Proliferation*

Over the 24 h period following UVB exposure cell viability was determined by dead cell apoptosis kit containing propidium iodide/Alexa Fluor 488-conjugated Annexin V (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instruction. Labelled cells were analyzed by flow cytometry with a FACSCalibur (Becton Dickinson, San Jose, CA, USA) measuring the fluorescence emission in FL1 (530 nm) and FL3 (>575 nm). Double negative cells represent viable cells. For data collection and evaluation CellQuest software 5.2 (Becton Dickinson) and Flowjo single cell analysis software were used.

Cell proliferation was determined by clonogenic assay. Cells were seeded in 100 mm Petri dish at 5000 cells/dish and were allowed to grow for 10 days. The medium was replaced each day to monitor the long-term effect of cell treatments. Ten days later, cells were washed with PBS, fixed with 100% methanol, and stained with May-Grünwald-Giemsa solution (Histolab Products, Västra Frölunda, Sweden).

#### *4.6. Cell Cycle Analysis*

Cell cycle progression was quantified using propidium iodide (PI) staining. Briefly, cells were trypsinized, fixed with ice-cold 96% ethanol for 10 min, washed twice with PBS, and permeabilized with 0.1% Triton X-100. After extensive washing, cells were incubated in PBS solution containing 0.5 mg/mL RNase at 37 ◦C for 1 h and counted for cell number normalization. For cell staining, PI was used a final concentration of 20 μg/mL. Unbound PI was eliminated by washing with PBS and doublet discrimination was performed. To determine DNA content samples were analyzed on the x-axis in FL2-A channel using a FACSCalibur flow cytometer.

#### *4.7. HPRT Mutation Assay*

CHO cells were cultured in a HAT (hypoxanthine–aminopterin–thymidine) medium in T75 flask for one week to remove cells with pre-existing HPRT mutation. After a week, HAT medium was replaced with complete DMEM and cells were allowed to recover for three days. Thereafter, CHO cells were seeded into a 6-well plate and allowed to adhere for 24 h. Next day, cells were cultured with or without 25 μM ABT-888 and UVB irradiation was carried out with 10 or 20 mJ/cm2 dose. CHO cells were cultured further for one week in DMEM with sub-culturing three times a week. After a week, cells were trypsinized, counted, distributed to 100 mm Petri dish at concentration of 5 <sup>×</sup> <sup>10</sup><sup>4</sup> per Petri dish in selection medium supplemented with 5 μg/mL 6-thioguanine (6-TG), and incubated for 10 days. After 10 days, cells were washed with PBS, fixed with 100% methanol for 10 min, and stained with May-Grünwald-Giemsa solution. Mutant colonies were counted.

#### *4.8. CPD-Specific Enzyme-Linked Immunosorbent Assay (ELISA)*

Genomic DNA was extracted from HaCaT cells using a Purelink Genomic DNA mini kit (Thermo Fisher Scientific) according to the manufacturer's instruction. For quantitative detection of CPDs, direct ELISA was applied as previously described [120].

#### *4.9. Real-Time Quantitative RT-PCR*

Total RNA was isolated using guanidinium isothiocyanate-phenol-chloroform extraction (TRI reagent) (MRC, Cincinnati, OH, USA) according to the protocol by Chomczynski et al. [121]. RNA concentration and purity were determined spectrophotometrically. RNA was deprived of DNA contamination by DNase I treatment (Fermentas, St. Leon-Rot, Germany). Reverse transcription (RT) was carried out using the high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster

City, CA, USA) according to the manufacturer's instruction. Real-time RT-PCR analysis was performed by the SYBR green method using a Lightcyler 480 II (Roche Diagnostics). Reactions were carried out in 384-well optical plates. RNA expression values were determined by 1.8−ΔCT method and normalized for the housekeeping gene SDHA and phosphoglycerate kinase 1 (PGK1) [122]. Primer pairs and corresponding sequences are listed in Supplementary Table S1 "Primer pairs used in the study".

#### *4.10. Western Blot*

HaCaT cells were pelleted by mild centrifugation at 1500 rpm for 5 min at 4 ◦C. The pellets were lysed on ice with RIPA buffer supplemented with a protease inhibitor cocktail in 1:1000. After 5 min incubation on ice, supernatants were obtained by centrifugation of the lysates at 15,000 rpm for 10 min at 4 ◦C. The concentration of proteins was determined using Pierce BCA assay kit (Thermo Fisher Scientific). The lysates were mixed with 5X loading buffer (Bromophenol blue (0.25%), β-Mercaptoethanol (5%), Glycerol (50%), SDS (sodium dodecyl sulfate; 10%), Tris-HCl (0.25 M, pH 6.8)), boiled for 10 min at 100 ◦C and subjected to 7.5%, 10%, or 12.5% SDS polyacrylamide gel electrophoresis. Proteins were transferred onto nitrocellulose membrane (Bio-Rad, Hercules, CA, USA). The membranes were blocked in TBST (0.05% Tween 20 in TBS buffer) containing 5% bovine serum albumin (BSA). The incubation with primary antibodies was carried out overnight at 4 ◦C followed by washing. Horseradish peroxidase (HRP)-conjugated goat anti-mouse or anti-rabbit IgG was used as a secondary antibody (Bio-Rad) at room temperature for 1 h. Proteins were visualized by ECL Prime Western blotting detection reagent (Thermo Fisher Scientific). All the antibodies used for Western blotting were listed in Supplementary Table S2 "Primer antibodies used in the study". Bands were quantified using the ImageJ open source software [123] (version 1.51k, National Institutes of Health, Bethesda, MD). Proteins of interest were normalized for β-actin. Sample uncut blots are provided on Supplementary Figure S2.

#### *4.11. Mitochondrial Mass*

24 h post-UVB, Mitotracker Green (Thermo Fisher Scientific) were added to the culture medium at a final concentration of 100 nM and cells were incubated at 37 ◦C for 30 min. Following incubation, cells were washed with PBS and harvested by trypsinization and then placed on ice. The fluorescence intensity of cells stained with Mitotracker Green was analyzed in a FL1 channel using flow cytometry.

#### *4.12. Determination of Mitochondrial Membrane Potential*

HaCaT cells were seeded in a culture plate and next day subjected to UVB. 24 h post-UVB, 3,3 -dihexyloxacarbocyanine iodide (DiOC6(3)) (Invitrogen) were added to the culture medium at a final concentration of 40 nM and cells were further incubated at 37 ◦C for 30 min. Following incubation, cells were washed with PBS and harvested by trypsinization and then placed on ice. The fluorescence intensity of cells stained with DiOC6(3) was analyzed in FL1 channel using flow cytometry.

#### *4.13. Determination of Mitochondrial Ultrastructure, Mitochondrial Number, and Area by Transmission Electron Microscopy (TEM)*

After cell treatment, cells were washed with DPBS, harvested by trypsinization, pelleted, and fixed by 3% glutaraldehyde (EMS, Hatfield, PA, USA) in 0.1 M cacodylic acid (EMS) buffer complemented by 5% sacharose for 2 h. After washing steps in 0.1 M cacodylic acid buffer, cells were osmificated in 1% OsO4, dehydrated in ascending alcohol row, namely 50% for 2 × 10', 70% for 2 × 10', 96% for 2 × 15', absolute ethanol for 3 × 20', and propylene oxide for 2 × 10'. Then, overnight a durcupan araldite treatment was used for embedding the samples. Encapsulation occurred in an incubator for 48 h and ultrathin section was made by Leica EM UC7 ultramicrotome (Leica Microsystems, GmbH, Wetzlar, Germany). Standard contrasting was performed by uranyl acetate (EMS) and Reynolds' lead citrate solution. High resolution TEM images was made by the Jeol JEM 1010 electron microscope and software (JEOL Inc. Peabody, MA, USA).

#### *4.14. Assessment of Mitochondrial Morphology and Autophagy by Confocal Microscopy*

Cells were plated on glass slides in a 24-well plate and 24 h post-UVB cells were stained with 100 nM Mitotracker Red CMXRos (Thermo Fisher Scientific) dye at 37 ◦C for 30 min. Cells were washed with PBS, fixed with 3.7% paraformaldehyde solution at room temperature for 10 min and permeabilized with 0.2% Triton X-100 for 10 min. After washing with PBS, blocking was performed by 1% BSA-containing PBS at 37 ◦C for 1 h. Cells were incubated with an Alexa Fluor 488-conjugated LC3A/B antibody diluted in 1% BSA (1:50) at 4 ◦C overnight in a humid chamber. Next day, slides were washed three times with PBS. Prior to imaging, cells were stained with a mounting medium with DAPI and analyzed by confocal microscopy using 60x oil immersion objective. Background subtraction, noise reduction, local contrast enhancement, unsharp mask, and bandpass filtering were applied on raw images for better image quality and proper image evaluation [124]. Processed images were analyzed by ImageJ software. Mitochondrial complexity was calculated from confocal microscopic images and was defined as a form factor ((Perimeter2/(4π\*area)). For flow cytometric analysis of autophagy, cells were washed with PBS and harvested by trypsinization. Fixation, permeabilization, and blocking were carried out as described above. Cells were incubated with Alexa Fluor 488-conjugated LC3A/B antibody using 1:500 dilution at 4 ◦C overnight. Next day, after extensive washing, cells were analyzed for Alexa-488 fluorescence intensity in FL1 channel by flow cytometry.

#### *4.15. Measurement of Citrate Synthase Activity*

For determination of citrate synthase activity, a citrate synthase assay kit was used according to the manufacturer's instruction. Optical density changes by kinetic program was determined at 412 nm using microplate reader. Values were normalized to total protein concentration.

#### *4.16. Analysis of Oxygen Consumption and Extracellular Acidification*

Oxygen consumption (OCR) and extracellular acidification rate (ECAR) was measured using an XF96 extracellular flux analyzer (Seahorse Bioscience, North Billerica, MA, USA). The rate of oxygen consumption indicates oxidative phosphorylation, whereas ECAR represent lactic acid formed during glycolysis. Cells were seeded in a XF96 cell culture plate at 10,000 cell/well density. One hour prior to the assay, culture medium was replaced with unbuffered DMEM (Seahorse Bioscience) supplemented with 10 mM glucose and then cells were equilibrated in a CO2-free incubator for 1 h. After four measurements of oxygen consumption, oligomycin and antimycin A were subsequently injected to determine mitochondria-linked ATP production and basal OCR, respectively. All OCR and ECAR values were normalized to the total protein obtained from cells lysed by 1 M NaOH.

#### *4.17. Determination of NAD*<sup>+</sup> *Level*

The colorimetric assay for NAD+/NADH ratio determination was purchased from Biovision (Mountain View, CA, USA). Cells were washed with PBS two times and harvested by a trypsin-EDTA solution and pelleted by mild centrifugation. The pellet was extracted with NAD<sup>+</sup> or NADH extraction buffer and exposed to freeze/thaw cycles. Samples were spun and the supernatant was transferred to a 96-well plate for NAD+/NADH assays. For NAD<sup>+</sup> determination, half of the samples were transferred into another tube and NAD<sup>+</sup> was decomposed by heating samples at 60 ◦C. Finally, working reagent was added to the samples, and optical density was determined at 450 nm. Optical density of the standard curve and samples was used to calculate NAD<sup>+</sup> content. Values were normalized to total protein concentration.

#### *4.18. Measurement of ATP Content*

For determination of ATP level, an ATP colorimetric/fluorometric assay purchased from Biovision was used according to the manufacturer. Optical density was determined at 570 nm using a microplate reader. Values were normalized to total protein concentration.

#### *4.19. Statistical Analysis*

Statistical analysis was carried out by the Kolmogorov–Smirnov test to assess the normality of the population. The frequency of mitochondrial morphological subtypes (fragmented, intermediate, and tubular) was calculated by chi2 and Ficher's exact test. To assess the statistical significance between untreated and differently treated groups, ANOVA complemented by Dunnett's post-hoc test was used. The comparison of two groups was applied by an independent *t*-test. All data are reported as mean ± SEM. A *p* < 0.05 was considered statistically significant. Statistical significance was determined by the GraphPad Prism 7 (GraphPad Software Inc., San Diego, CA, USA) and SPSS 25 software. (SPSS package for Windows, Release 25.; SPSS, Chicago, IL, USA).

#### **5. Conclusions**

Our results provide new information about the role of ATM and PARP1 proteins in the regulation of mitochondrial morphology and function after UVB irradiation. Moreover, it suggests that UVB and PARPi-caused elevation of oxidative phosphorylation is mediated by the complex interplay of metabolic sensors including ATM, AMPK, p53, AKT, and mTOR which are the main mediators that connect DNA damage with oxidative metabolism and autophagy.

**Data Availability:** All primary data is uploaded to https://figshare.com/s/49fbcd8cfd5802ea15f9 (DOI: 10.6084/m9.figshare.8107727).

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/1/5/s1, Figure S1: Uncut PAR (10H) Western blot, Figure S2: Sample uncut blots, Table S1: Primer pairs used in the study, Table S2: Primary antibodies used in the study.

**Author Contributions:** Original draft preparation, investigation, methodology, C.H.; Writing—review and editing, G.B.; Methodology, E.F.; Investigation, methodology, G.N.K.; Resources, M.A.; Investigation, methodology, T.J.; Formal analysis, E.A.J.; Methodology, L.J.; Writing—review and editing, G.P.; Writing—review and editing, G.E.; Conceptualization, writing—review and editing, resources, P.B.; Project administration, funding acquisition, É.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the European Union and the European Regional Development Fund GINOP-2.3.2-15-2016-00005 and was supported by grants from NKFIH K123975, C129074, NKFIH K120206, Momentum program and NM-26/2019 of the Hungarian Academy of Sciences and the University of Debrecen. The project is co-financed by the University of Debrecen (OTKA Bridging Fund). Bolyai Janos Research Scholarship (J.T.). Szodoray Lajos and Magyary Zoltán Funds by Hungarian Academy of Science and the European Union and the State of Hungary, co-financed by the European Social Funding. Fund in the framework of TÁMOP 4.2.4.A/2-11-1-2012-0001 "National Excellence Program" (J.T.). The research was financed by the Higher Education Institutional Excellence Programme (NKFIH-1150-6/2019) of the Ministry of Innovation and Technology in Hungary, within the framework of the Biotechnology thematic programme of the University of Debrecen.

**Conflicts of Interest:** The authors declare no conflict of interest. G.P. is a consultant for ADC Therapeutics and Buffalo Biolabs. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**



#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*

### **CAF-1 Subunits Levels Suggest Combined Treatments with PARP-Inhibitors and Ionizing Radiation in Advanced HNSCC**

**Francesco Morra 1,**†**, Francesco Merolla 2,3,**†**, Ida Picardi 1, Daniela Russo 2, Gennaro Ilardi 2, Silvia Varricchio 2, Federica Liotti 1, Roberto Pacelli 2, Luca Palazzo 1, Massimo Mascolo 2, Angela Celetti 1,\* and Stefania Staibano 2,\***


Received: 6 September 2019; Accepted: 14 October 2019; Published: 17 October 2019

**Abstract:** Oral (OSCC) and oropharyngeal (OPSCC) squamous cell carcinomas show high morbidity and mortality rates. We aimed to investigate the role of the "Chromatin Assembly Factor-1" (CAF-1) p60 and p150 subunits, involved in DNA repair and replication, in OSCC and OPSCC progression and in response to Poly(ADP-ribose) polymerase (PARP)-inhibitors and exposure to ionizing radiation (IR). We immunostained tissue microarrays (TMAs), including 112 OSCC and 42 OPSCC, with anti-CAF-1/p60 and anti-CAF-1/p150 specific antibodies, correlating their expression with prognosis. Moreover, we assessed the sensitivity to PARP inhibitors and the double-strand breaks repair proficiency by cell viability and HR reporter assays, respectively, in HPV-positive and HPV-negative cell lines upon CAF-1/p60 and CAF-1/p150 depletion. The immunohistochemical analysis revealed a significant prognostic value of both tissue biomarkers combined expression in OSCC but not in OPSCC. In in vitro studies, the p60/150 CAF-1 subunits' depletion impaired the proficiency of Homologous Recombination DNA damage repair, inducing sensitivity to the PARP-inhibitors, able to sensitize both the cell lines to IR. These results indicate that regardless of the prognostic meaning of p60/p150 tissue expression, the pharmacological depletion of CAF-1 complex's function, combined to PARP-inhibitors and/or IR treatment, could represent a valid therapeutic strategy for squamous cell carcinomas of head and neck region.

**Keywords:** head and neck squamous cell carcinoma; CAL27; SCC90; p60/p150 CAF-1 subunits; DNA repair; homologous recombination; biomarkers; personalized treatment

#### **1. Introduction**

Head and neck cancer (HNC) is the eighth most common malignancy in the world, with *squamous cell carcinoma* (*SCC*) accounting for *more than 90*% [1]. HNSCC encompass tumours of larynx, hypopharynx, paranasal sinuses, oral cavity, and oropharynx. Of all HNSCC, oral cavity SCC (OSCC) and oropharynx SCC (OPSCC) represent more than a half-million new cases per year, with an estimated incidence of 7.0 per 100,000 inhabitants, worldwide. HNSCC are characterized by a high-rate of morbidity and mortality. Furthermore, in most countries, the five-year survival rate is less than 50%

and, in the United States, more than 20,000 new cases are estimated to occur in 2019, with more than 10,000 deaths [1–3]. Multiple risk factors contribute to the initiation of HNSCC. Tobacco and heavy use of alcoholic beverages, for OSCC, and high-risk Human Papilloma Virus, mainly HR-HPV16 persistent infections, for a significant percentage of OPSCC, are the most important risk-factors, causing a substantial percentage of these tumours in the Western countries. HPV16 accounts for >80% of HPV-positive oropharyngeal SCCs compared with HPV-positive oral SCCs and laryngeal SCCs. Conversely, HPV18 is relatively rare in HPV positive oropharyngeal SCCs compared with other head and neck sites.

While OSCC are commonly observed in males, aged over 50 years, OPSCC occurs according a bi-modal pattern, respectively, in those under 40 and in elderly people [4–6], and mainly affect females. OPSCC comprise of HPV-negative (HPV−) and HPV-positive (HPV+) tumours, which represent two distinctive clinicopathological and molecular entities, with a disparate range of survival rates. HPV+ OPSCC are characterized by a significantly slow progression and a high response to chemo- and radiotherapy, while HPV− OPSCC and tobacco smoking/alcohol-related OSCC are intrinsically highly aggressive, and highly chemo- and radio-resistant when in an advanced stage. Currently, there are still no reliable prognostic and predictive biomarkers for these deadly cancers that kill about 50% of patients with metastatic disease. Genotoxic exposure to tobacco carcinogens and consequent adducts formation resulting in DNA damage [7] and genomic instability induced by the unscheduled cell replication of integrated HR-HPVs, as demonstrated in in vitro experiments [8], are thought to represent important mechanisms of carcinogenesis for HPV− and HPV+ HNSCC, respectively. At least five major DNA repair pathways—base excision repair (BER), nucleotide excision repair (NER), mismatch repair (MMR), homologous recombination (HR), and non-homologous and joining (NHEJ)—are active throughout different stages of the cell cycle, allowing the cells to repair the DNA damage in a substrate-dependent manner [9–11]. The chromatin assembly factor 1 (CAF-1), a heterotrimeric protein complex formed of three subunits (p48, p60 and p150), plays a key role in the steric organization of DNA and in the assembly of nucleosomes [12]. In particular, while p48 subunit acts on acetylation/deacetylation of histones, p150 appears to be more active in interphase DNA-damage repair process, interacting with PCNA on the damaged DNA, specifically during NER [13,14] and double-strand breaks repair [15], the CAF-1/p60 subunit is more specifically connected to controlling cell replication [16–18]. Several reports have recently shown that a deregulated expression of this subunit leads cells to incorrectly replicate DNA, consequently accumulating DNA damage [16,19]. Interestingly, CAF-1/p60 expression levels are significantly correlated with the biological aggressiveness of tumours, metastasizing behaviour and worse prognosis in breast, oral, prostate, laryngeal and salivary gland carcinomas, as well as in skin melanoma [20–25], suggesting for this protein a promising role as a new sensible prognostic marker, apparently unrelated to their histogenesis. In recent decades, despite therapeutic advances in the HNSCC treatment, patient survival has not markedly improved and the mortality is still around 40–50%. CAF-1 appears as an interesting candidate to explore in HNSCC. In the present study, we evaluated the expression of CAF-1/p60 and p150 subunits in a tissue microarray (TMA) selected series of OSCC and OPSCC. We related our data to the HPV status of primary OPSCC, evaluated by immunohistochemical expression of p16INK4a protein. We also verified the prognostic value of these tissue biomarkers through follow-up analysis. We used, as models of in vitro study, in vitro cultured cells of OSCC and HPV+ OPSCC, testing the efficiency of repair mechanisms and radiation sensitivity, upon silencing of CAF-1/p60 and p150.

#### **2. Results**

#### *2.1. Immunohistochemistry Expression of CAF-1*/*p60 and p150 Subunits in OSCC and OPSCC Tumour Samples*

To investigate the functional role of the CAF-1/p60 and p150 subunits in HNSCC, we evaluated a Tissue microarray (TMA) case study (total number of cases *N* = 154) of non-oropharyngeal (OSCC) (*N* = 112) and oropharyngeal squamous cell carcinomas (OPSCC) (*N* = 42), the latter tested for the

presence of the HPV virus (HPV+ OPSCC *N* = 8). All tumour samples were staged according to the 8th AJCC staging manual [26] (Table 1). In these samples, the immunohistochemistry staining for the p60 and p150 subunits of CAF-1 was assessed and scored as "HIGH" and "LOW" as defined in the Section 4 (representative images of staining score categories are shown in Figure 1). The p60/p150 frequency distribution data were further analysed with a classification algorithm allowing stratification of samples in three clusters, homogeneous for tissue expression of p60 and p150 according to the IHC staining. The three clusters were defined as follow: p150 HIGH/p60 HIGH; p150LOW/p60HIGH; and p150LOW/p60LOW. The category p150HIGH/p60LOW was not revealed since no p60LOW cases were present in the p150HIGH sub-group (Table 1).


**Table 1.** Descriptive statistics of the studied population. NOP, non-oropharyngeal tumours; OP, oropharyngeal tumours. HPV positivity (p16 IHC) is only reported for oropharyngeal squamous cell carcinomas.

**Figure 1.** IHC staining of OSCC FFPE tumour samples with anti-CAF-1 p60 and anti-CAF-1 p150 antibodies. The figure shows representative images of anti-CAF-1 p60 and anti-CAF-1 p150 IHC staining intensity in OSCC FFPE tumour samples grouped according to cluster classification as resulted by cluster analysis of immunohistochemistry expression data: (**a**,**b**) p60 HIGH and p150 HIGH staining category, respectively; (**c**,**d**) p60 LOW and p150 LOW staining category, respectively; and (**e**,**f**) p60 HIGH and p150 LOW staining category, respectively. Magnification: for each panel, a 5× image of the entire core is shown and the inset shows the highlighted region.

To analyse the frequency distribution of CAF-1/p60, CAF-1/p150, and overall survival (OS) variables, we set up a contingency table.

By applying Fisher's exact test, the frequency distribution of p60 and p150 positivity, crossed by OS, proved to be statistically significant in the whole study population (*p* = 0.022), statistical significance was particularly high in OSCC group (*p* = 0.013) and no significance resulted from OPSCC samples analysis (*p* = 0.485) (Table S1).

By contingency table analysis, we could observe that, in the whole tested population, p60HIGH score mostly segregates with a poor prognosis, in terms of overall survival, as expected (dead/alive ratio = 1.61 in p60HIGH, 0.68 in p60LOW). Moreover, the association of p60 with the worst outcome was even stronger in the p150LOW score group (dead/alive ratio = 1.89). The analysis of p60 and p150 frequency distribution revealed the highest dead/alive ratio in the p60HIGH/p150LOW population of OSCC samples (45/18 = 2.5), while p60 expression did not correlate with outcome in OPSCC samples (*p* = 0.485). Nevertheless, out of eight HPV+ OPSCC samples, the only one presenting a poor outcome at follow-up belonged to p60HIGH/p150LOW subgroup.

Survival curves analysis confirmed a statistically significant difference between p60HIGH and p60LOW curves in the p150LOW population, (log-rank test, *p* = 0.0034). A not-significant *p*-value was obtained evaluating the differences in OPSCC group (*p* = 0.477), irrespective of HPV status (Figure 2).

**Figure 2.** Survival analysis by Kaplan–Meier curves. The picture shows CAF-1 p60 HIGH and CAF-1 p60 LOW survival curves in the study population grouped by CAF-1 p150 staining score. (**A**) CAF-1 p60 HIGH and CAF-1 p60 LOW curves in p150 LOW group. (**B**) Since CAF-1 HIGH category is only associated with CAF-1 p60 HIGH staining score, only this curve is shown. Statistical differences between curves were assessed by log-rank test, where applicable. (**C**,**D**) Kaplan–Meier curves survival analysis of the three clusters stratified by age class were performed (**C**) for age class "41–60" and (**D**) for age class ">60". Difference between curves was statistically significant in mid-age group (40–60 years old) (Log-Rank test *p* = 0.005).

To further unravel the prognostic potential of p60 and p150 tissue expression in OSCC, we stratified the studied population by the age of patients at diagnosis, grouping the population in "young" (<40 years); "mid" (41–60 years); and "old" (>60 years). The survival data frequency distribution analysis, upon crossing the clusters by the overall survival results and the age groups, gave an extremely significant result in the mid-age population (*p* = 0.002). In the old age population, the distribution was not significant (*p* = 0.910), and the group of young age population was not big enough to allow statistical analysis (*N* = 5) (Table S2).

The tissue overexpression of CAF-1/p60 subunit associates with poor overall survival in OSCC, in the absence or with concomitant low expression of CAF-1/p150 subunit. We did not observe a significant association with the outcome in OPSCC tumour samples.

#### *2.2. The Silencing of the CAF-1 Subunits Increases the Sensitivity to Ionizing Radiation in HPV-Negative and HPV-Positive Head and Neck Cancer Cell Lines*

The protein "chromatin Assembly Factor-1" (CAF-1) plays a fundamental role in the steric organization of DNA and the assembly of Nucleosomes [12], thanks to its subunits, p60, p48 and p150 that make up the Heterotrimer CAF-1 [13,17,18]. CAF1 functions as "histone chaperone", and its activity is also required during the DNA damage repair [12,27]. While CAF-1/p150 subunit appears to be more active in interphase DNA damage repair processes, interacting with PCNA on the damaged DNA, during nucleotide excision repair (NER) [10,14] and double-strand breaks repair [15], CAF-1/p60 has been described to be more specifically connected to cell replication. Nevertheless, loss of p60 leads the cells to incorrectly replicate DNA, consequently accumulating DNA damage [16].

In this work, by utilizing in vitro cultured cells of HPV-negative and HPV-positive oral cavity squamous cell carcinoma, we investigated whether the stable silencing of the CAF1 subunits, p60, p150 or p60 and p150 together, could modulate the response to ionizing radiation (IR) therapy.

The CAL27 cells (OSCC, HPV-negative) and SCC90 cells (OPSCC, HPV-positive) were silenced for the large subunit p150 and for the small subunit p60 of CAF-1, alone or in combination. The efficiency of silencing was evaluated by quantitative Real-Time-PCR (qRT-PCR), using a specific set of primers for both subunits of CAF-1. The analysis of the transcripts demonstrated low mRNA levels for the two proteins upon single or combined silencing (Figure 3A). The CAF1/p60 protein levels were evaluated in the p60 and in the p60 and p150 stably silenced cells, with Western blot (Figure 3B). The detection of the silencing efficacy of the p150 CAF-1 subunit was not possible at the protein level due to the failure of anti-p150 antibody. The standard therapy for non-metastatic head and neck cancer is based on ionizing radiation and/or surgery. However, metastatic lesions are treated with chemo- and radiotherapy, with a poor prognosis. In recent years, the possibility to characterize the presence and activation of the HPV virus in head and neck tumours has allowed the stratification of such tumours by the onset site with the prediction of response to therapies. The HPV-positive tumours showed high sensitivity to radiation, whereas the HPV-negative tumours resulted in a more resistant phenotype.

**Figure 3.** *Cont*.

**Figure 3.** (**A**) Analysis of CAF-1 p60 and p150 relative mRNA expression by quantitative RT-PCR in human OSCC (CAL 27 HPV-negative) and OPSCC (SCC90 HPV-positive) cell lines, following stable trasnfection of the sh-CAF-1 p60, sh-CAF-1 p150 or both the sh-CAF1 p60/p150 vectors. (**B**) Immunoblot analysis of the CAF-1 p60 protein levels in OSCC- and OPSCC- derived cell lines. Anti-tubulin is shown as loading control. (**C**,**D**) (left) Clonogenic assay of CAL 27 and SCC90 cells after 0, 2, 4, and 6 Gy irradiation. Only colonies consisting of at least 30 cells were counted. (right) Clonogenic survival curve of CAL 27 and SCC90 cells. Error bars indicate the standard error mean. The sensitivity to ionizing radiation is expressed as IC50 (the value of radiation able to inhibit the cell growth of 50%).

In search of the mechanisms responsible for the different behaviours of the HPV-positive and HPV-negative tumours, we investigated the role of CAF-1 subunits in determining the sensitivity to IR in head and neck cancer. As a model, we utilized two cell lines of head and neck tumours, which were confirmed for the HPV status (CAL 27 HPV-negative and SCC90 HPV-positive).

To evaluate whether the silencing of the CAF-1 subunits p60 and p150 (as single silencing or in combination) could affect the sensitivity to IR, we treated the CAL27 and SCC90 with range doses of IR (0, 2, 4, and 6 Gy). After ten days from irradiation, we analysed the rate of survival by a Colony Forming Assay (CFA) (Figure 3C,D). A significant reduction in the doses of radiations able to decrease the survival rate in 50% of the cells was observed. These effects were more evident in the shp60/shp150-silenced cells (IC50 CAL 27: shp60/p150: 1.46 Gy vs. shCTRL: 4.04 Gy; IC50 SCC90: shp60/p150: 1.2 Gy vs. shCTRL: 2.3 Gy) (Figure 3E).

The silencing of both the subunits of CAF-1 in our HPV-negative and HPV-positive cellular models resulted in an increased sensitivity to ionizing radiations.

#### *2.3. The Silencing of the Subunits of CAF-1 Leads to Defect in DNA Repair Mediated by Homologous Recombination and Sensitizes OSCC and OPSCC Cells to PARP-Inhibitors*

CAF-1 proteins have been reported to be involved in DNA repair mechanisms [28,29]. To investigate the proficiency of Homologous Recombination DNA damage repair in CAF-1 silenced cells, CAL27 and SCC90 cells were transfected with the DR-GFP reporter plasmid alone, as a control, or together with the I-SceI plasmid, able to induce DSBs. The ability to repair the DSBs by HR was measured by flow cytometry, and the frequency of HR is reported as a percentage of GFP positive cells. The silencing of both subunits determined a significant decrease of the GFP positive cells, compared to the control, suggesting that reduction of CAF-1 levels affected the DNA repair by HR. These results were obtained in both cellular models (Figure 4A,B). The HR proficiency was also assessed in cells stably silenced with shp60 or shp150 (as single silencing), obtaining similar results. Then, as defects in DNA repair by HR are reported to increase the sensitivity to inhibitors of the Poly(ADP-ribose) polymerase 1 and 2 (PARP1 and PARP2) [30,31], we evaluated the effects of the CAF-1/p150 and p60 subunits in modulating the CAL27 and SCC90 cells' sensitivity to the PARP-inhibitor Olaparib. We treated cells with different concentrations of Olaparib and quantified the cytotoxic effect by a cell viability assay.

Silencing of both the subunits of CAF-1 significantly increased the sensitivity to Olaparib (IC50 CAL27 shp60/p150CAF-1: 1.42 μM vs. shCTRL: 3.01 μM; IC50 SCC90 shp60/p150CAF-1: 0.42 μM vs. shCTRL: 1.78 μM) (Figure 4C,D). The impact of p150/p60 silencing and Olaparib effects on DNA double strand breaks with gamma H2AX staining was also assessed in the same experimental conditions (Figure S1).

**Figure 4.** (**A**,**B**) Silencing of p60 and p150 suppresses homologous recombination. CAL 27 and SCC90 cells were transfected with DR-GFP alone, as control, or together with I-SceI. The percentage of GFP positive cells, compared to controls, is plotted on histograms representative of three independent experiments. Error bars indicate the standard error mean. (**C**,**D**) Silencing of p60 and p150 increases the sensitivity to Olaparib independently HPV-status. Survival fractions of CAL 27 and SCC90 cells treated with Olaparib, at the indicated doses, in presence (sh CTRL) or absence of CAF-1 (sh p60/p150) for 144 h. The sensitivity to the Olaparib was determined by the modified MTT assay (MTS), Cell Titer 96 AQueous One Solution assay, and expressed as IC50, i.e., 50% of the inhibitory concentration. The values are expressed as mean ± the standard deviation.

#### *2.4. Treatment with PARP-Inhibitors Increases the Radiosensitivity in HPV-Negative Head and Neck Cancer Cells*

We finally analysed whether the silencing of CAF-1 might induce a radiosensitization effect in stably silenced cells upon PARP inhibitor treatment. The CAL 27 and SCC90 silenced cells and controls were plated and exposed to arange IR doses (0, 1, 2, 3, and 4 Gy), upon treatment with Olaparib at a fixed dose [0.1 μM]. Interestingly, by transfecting control shRNAs (shCTRL), shPARP1 and shPARP2, alone or together, in CAL27 cells (wild type or p150/p60 knockdown) and in SCC90 cells (wild type or p150/p60 knockdown), we yielded a radiosensibilizing effect similar to with PARP-inhibitors, as shown in Figure 5.







**Figure 5.** (**A**,**B**) The Dose Enhancement Ratio (DER) at 50% of the effect refers to the ratio between the dose with radiation alone and the dose with radiation + shPARP1 alone or in combination with shPARP2 for the same biological effect. If the DER is greater than one, then the silencing of PARP1/2 is functioning as a radiosensitizer. If the DER is less than one, the PARP silencing acts as a radioprotector. In the table are shown DER values at 50% effect for CAL 27 (sh CTRL vs. sh p60/p150). (**C**) The PARP1 depletion was assessed by Western blot. Tubulin is shown as loading control.

We calculated cell survival by viability assay at 144 h (Figure 5). The results of the combined treatment were extremely relevant, as analysed by the Dose Enhancement Ratio (DER) calculated at 50% of cells' survival [32]. The HPV-negative CAL27 cells showed a DER > 1 after cells exposure to different doses of ionizing radiation (0–4 Gy) in the presence of Olaparib [0.1 μM], which suggested a radiosensitization effect upon CAF-1 subunits depletion. However, the HPV-positive SCC90 cells, showed a moderate radiosensitization effect following a combined treatment, either in CAF-1 subunits depleted cells or in control cells (Figure 6). These data seem very promising as the SF2Gy, a standard measure of cells sensitivity to IR highly utilized in the field of radiobiology, indicates a great radiosensibilization effect upon CAF1 subunits silencing (Table S3).






**Figure 6.** (**A**,**B**) The Dose Enhancement Ratio (DER) at 50% of the effect refers to the ratio between the dose with radiation alone and the dose with radiation + drug for the same biological effect. If the DER is greater than one, then the addition of the drug is functioning as a radiosensitizer. If the DER is less than one, the drug acts as a radioprotector. In the table are shown DER values at 50% effect for CAL 27 (sh CTRL vs. sh p60/p150) and SCC90 (sh CTRL vs. sh p60/p150).

#### **3. Discussion**

Despite the recent advances in early diagnosis and surgical management of HNSCC, the outcome of these tumours has not substantially changed during the last decades, and the prognosis, for each HNSCC patient, remains linked to the tumour's stage at presentation. The standard of care for these patients consists of radical surgery, complemented by radiotherapy, chemotherapy and, recently, immunotherapy in the case of advanced disease. The five-year survival rate for HNSCC patients remains poor, with up to 50% of mortality rates [1]. The most relevant limits of radio- and chemotherapy are represented by systemic and/or local toxicity, and by the frequent radio/chemoresistance of these tumours with particular attention to OSCC [33].

The limited information available on the molecular determinants of the biology of HNSCC indicates the urgent need to identify new prognostic markers and/or molecular targets for personalized therapeutic strategy [34].

Oral/oropharyngeal carcinogenesis mostly depends on environmental factors, such as tobacco and alcohol abuse, particularly in the oral cavity, and persistent infection from HR-HPV in a rising percentage of oropharyngeal cancers. Thus far, only a few reliable prognostic markers have been reported, alone or in combination, for OSCC and OPSCC [35–39].

The HPV status (persistent infection by HR-HPV, mostly HPV-16), only for oropharynx SCC, has been indicated, by the new TNM classification (AJCC 8th edition) as a significant marker of a more favourable biological behaviour of tumours, as has been shown by several studies in the last decade [40–44]. In HPV-negative tumours, mutations of TP53 and amplification of the Epidermal Growth Factor Receptor (EGFR) have been reported and used for prognostic and intervention indications with dismal results [45–49].

Chromatin remodelling proteins play an important role in genome maintenance processes, including DNA repair and replication, also involved in the development and progression of several human malignant tumours. Chromatin Assembly Factor 1 (CAF-1) is a "histone chaperone", which delivers newly synthesized H3/H4 dimers to the replicative fork during the DNA synthesis phase (S) of the cell cycle [50]. CAF-1 is a heterotrimeric protein, of which p48 subunit cooperates with the Retinoblastoma protein (Rb) [51] and p150 and p60 subunits are involved in DNA repair and replication processes. Noteworthy, a leading role of p60 in sustaining the proliferative activity of cells in different malignancies has been described in the last two decades [14,52–54].

Recently, the presence of defects in genes involved in DNA damage repair by homologous recombination (HR) opened the way to the use of PARP inhibitors, alone or in association with genotoxic agents, such as ionizing radiation, in some malignant tumours [55]. The inhibition of PARP enzymes as an anticancer strategy has been established because of the biological concept of synthetic lethality, for which two genomic events, individually not lethal, become lethal when occurring together. When PARP enzymes are pharmacologically inhibited, the DNA single-strand breaks cannot be repaired and eventually progress to toxic double-strand breaks (DSBs), which result in being lethal in cells that lack HR repair capacity or have lost DNA repair genes [56–58]. In head and neck cancer cell lines characterized by a different ability to repair the DNA double-strand breaks through HR, the effectiveness of combined irradiation and PARP-inhibitor treatment in HR-deficient cells has been evaluated [59]. Very interestingly, it has been demonstrated that HPV positive head and neck tumours, carrying defects in DNA damage repair by HR, result sensitive to ionizing radiation [60].

Since HNSCC rarely carry mutations in DNA repair genes, there is considerable interest in finding alternative determinants of PARPi sensitivity. Our in vitro study, carried out by depleting CAF-1/p60 and p150 subunits (singularly or together), showed impairment of DSBs DNA repair and increased sensitivity to PARP inhibitor Olaparib in both cell lines used. This suggests that the presence of HPV, in the OPSCC cell line, did not modify the sensitivity of tumour cells with respect to the HPV-negative OSCC cell line. As a further step, we evaluated whether the treatment with Olaparib could modulate the response to ionizing radiation in the CAF-1 depleted cell lines. By combining ionizing radiation and PARP inhibitors drug, we observed an increased radio-sensitivity. Of particular interest, this effect resulted more evident in the OSCC CAL27 cells depleted of the CAF-1 p60/p150 subunits, as indicated by the DER that resulted greater than 1 (DER > 1). These data suggest that the combined treatment with ionizing radiation and PARP inhibitors can lead to the increase of the radio-sensitivity of OSCC, characteristically resistant to standard radio-treatment regimens.

All of the tumour samples included in the present study were re-examined and re-staged according to the eighth edition of the cancer staging manual of AJCC, and a "stage migration" was observed, as reported in Table S4 [61].

Classification through cluster analysis of immunohistochemical data allowed us to stratify the patients' outcome in different subgroups: (i) a p60low/p150low subgroup, mainly characterized by a good prognosis in both OSCC and OPSCC; (ii) a p60high/p150low group which showed, mainly in OSCC, the worst outcome prediction; and (iii) a p60high/p150high subgroup with an intermediate behaviour. These results confirm our preliminary data, obtained in the tongue tumour samples, of a prognostic role of the CAF-1/p60 subunit, which correlated with poor outcome and the importance of p60/p150 dual assessment as a prognostic determinant in OSCC tumours [20,24,25]. Interestingly, the survival data frequency distribution analysis, upon crossing the clusters by the overall survival results and the age groups, gave an extremely significant result in the mid-age population. Nevertheless, the Cox multivariate analysis, performed including age and stage variables together with p60/p150 clusters, revealed that p60/p150 staining is an independent prognostic factor (Figure S2).

Our data let us envisage novel potential therapeutic approaches for OSCC by blocking the p60 protein, hampering the CAF-1 complex function and inducing HR defects that would sensitize tumour cells to Olaparib and/or in association with ionizing radiation.

We did not obtain significant results in OPSCC cohort, probably due to the small number of HPV+ in our series of cases. Interestingly, the only tumour with poor outcome among the HPV+ OPSCC showed the more aggressive immunophenotype (p60high/p150low) observed in OSCC.

However, this finding is not sufficient, at present, to propose the evaluation of CAF-1/p60 and p150 protein expression for prognostic and predictive stratification also of this tumours' subset. For this reason, the evaluation of CAF-1 proteins' expression needs to be evaluated on a larger, multi-institutional case study before we can lead to a definitive conclusion. This study is currently in progress.

In conclusion, our data confirm the reliability of CAF-1/p60 subunit as prognostic marker for OSCC, indicating in addition that the combined evaluation of p60 and p150 subunit may be of particular utility in stratifying the different prognostic classes of OSCC.

In addition, we showed that CAF-1/p60 and p150 subunits are involved in HR-DDR, thus indicating the chance to induce a radiosensitizing synthetic lethality mechanism by treating tumour cells pharmacologically inhibited for CAF-1/p60 and p150 with PARP inhibitors, in OSCC patients in the worse prognostic group, in the direction of a truly personalized therapy.

#### **4. Materials and Methods**

#### *4.1. Cell Lines and Drugs*

Experiments were carried out using two human head and neck squamous cell lines, the CAL27 (Oral Squamous Cell Carcinoma, OSCC, HPV-negative) and SCC90 cells (Oropharyngeal Squamous Cell Carcinoma, OPSCC, HPV-positive). CAL27 cells were obtained by the "American Type Culture Collection" (ATCC) and SCC090 cell line was obtained by Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures, Germany.

Cell lines were cultured in the DMEM plus 10% of fetal bovine serum (Gibco, Paisley, UK).), L-Glutamine (2mM) and 100 U/mL of penicillin-streptomycin (Gibco, Paisley, UK) in 5% of CO2 at 37 ◦C. Olaparib (AZD2281) was provided by SelleckChem (Houston, TX, USA).

#### *4.2. Real Time PCR*

PCR reactions were performed on RNA isolated from cell lines using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and reverse-transcribed using MuLVRT (Invitrogen, Carlsbad, CA, USA). The qRT-PCR analysis was performed with Syber Green (Agilent, Santa Clara, CA, USA). Primer sequences are reported in Table S5. The relative expression levels were calculated by the 2-ΔΔCT method.

#### *4.3. Western Blotting and Antibodies*

Western blotting was performed as described (Figures S3 and S4) [62,63]. Immunoblotting experiments were carried out according to standard procedures and visualized using the ECL chemiluminescence system (Amersham/Pharmacia Biotech, Little Chalfont, UK). Anti-CHAF1B (HPA021679) and anti-Tubulin were provided by SIGMA-Aldrich, Inc. Secondary antibodies were from Biorad (Hercules, CA, USA).

#### *4.4. Plasmids and Transfection*

MISSION shRNA (pLKO.1) NM\_005441 (CHAF1B MISSION shRNA Plasmid DNA, cod: TRCN0000074278, TRCN0000074279, and TRCN0000074281] and NM\_005483 [CHAF1A MISSION shRNA Plasmid DNA, cod: TRCN0000074273 and TRCN00000234596) and shRNA of control (sh CTRL) were utilized for stable transfection and were provided by Sigma-Aldrich, Inc. (St. Louis, Missouri, USA) For transfection, the FuGENE 6 Transfection Reagent was provided by Promega Italia S.r.l. (Milano, Italy).

#### *4.5. Sensitivity Test and Design for Drug Combination*

Antiproliferative activity was determined by a modified 3-(4,5-dimethylthiazole-2-yl)-2-5 diphenyltetrazolium bromide assay, CellTiter 96 AQueous One Solution Assay (Promega, Milano, Italy), calculated as 50% inhibitory concentration (IC50) values, according to the manufacturer instructions.

Briefly, cells were plated in quintuplicate in 96-well plates at a density of 1000 cells per well, and continuously exposed to each drug for 144 h. Each assay was performed in quintuplicate and IC50 values were expressed as mean ± standard deviation.

The results of the combined treatment were expressed as a DER, calculated at 50% of survival. DER is a measure of how many folds each radiation dose may be reduced when administered in combination with a drug to obtain a given effect, in comparison to the same dose of radiation when given alone. A value of DER greater than one (DER > 1) indicates that a fixed dose of the drug used, in association with a range doses of IR was able to act as radiosensitizer, while a value of DER lower than one (DER < 1) indicates that a fixed dose of the drug used, in association with a range of IR doses, works as a radioprotector [32,64].

#### *4.6. Colony Forming Assay*

One-hundred-millimeter-cubed dishes of proliferating cells were exposed to 0, 2, 4, and 6 Gy of IR. After cells counting, a pre-defined number of viable cells were plated in 6-well plates, in triplicate. To receive a sufficient colony count, a different number of cells were plated for each dose of irradiation (0 Gy, 200 cells; 2 Gy, 500 cells; 4 Gy, 600 cells; and 6 Gy, 800 cells) [65]. After 10 days of incubation, prior to counting colonies, cells were stained with 0.5% Crystal Violet (10 min at room temperature). A population of at least 30 cells was scored as one survivable colony and considered for the count. The colonies' counting was performed at the optic microscope and through the open source software ImageJ-NIH [66]. The relative colony formation (surviving fraction) was expressed as the number of colonies per treatment level versus colonies that appeared in the untreated control. (mean colony counts ± standard errors are reported).

#### *4.7. TMA and IHC*

One hundred fifty-four HNSCC FFPE tumour samples (112 OSCC and 42 OPSCC, of which 8 HPV-positive), from surgical resections, were used to build tissue micro-arrays (TMAs). The study was performed according to the guidelines of the Institutional Ethic Committee, which, in agreement with the Italian law, with reference to the topics of the current research and according to the Declaration of Helsinki require, for studies based only on retrospective analyses on routine archival FFPE-tissue, a written informed consent from the living patient, following the indication of Italian DLgs No. 196/03 (Codex on Privacy), as modified by UE 2016/679 law of European parliament and Commission at the time of surgery.

The HPV positivity was confirmed through p16 immunostaining and HPV genotyping by INNO Lipa [67]. Seven TMAs were built selecting the most representative areas from each selected paraffin block, at least in duplicate. Three-millimeter tissue cores were punched from morphologically representative tissue areas of each donor block and placed into one recipient paraffin block (3 cm × 2.5 cm) using a semi-automated tissue arrayer (Galileo TMA, Milan, Italy). To ensure a sufficient representation of the tissue composition of individual cancer cases, in view of the known frequency of tumour heterogeneity in H&N SCC, we built TMAs taking 3-mm cores in at least duplicate or triplicate (in the case of bigger tumours). We already challenged our TMA protocol with whole slide assessment of tissue biomarkers expression (please see reference mentioned below) finding a high degree of agreement between the CAF-1/p60 assessment on TMAs and on routine tissue sections [54]. One section of each TMA (4 μm) was stained with hematoxylin and eosin (H&E) to check the adequacy of cores. The immunohistochemical staining with anti-CAF-1/p60 and CAF-1/p150 (ab8133 and ab126625 obtained from AbCam, Cambridge, UK) [68] were evaluated semi-quantitatively as the percentage of positive cells (with either nuclear or cytoplasmic localization) among the total number of tumour cells and classified as LOW staining (including 0 (<5% di cellule positive); and + (5% to <15%) scores) and HIGH staining (including ++ (15% to <30%); +++ (>30%) scores). All immunoassayed TMA glass slides were digitalized with an Aperio AT2 digital pathology slide scanner (Leica Biosystems Nussloch GmbH, Heidelberger, Germany) and visualized with QuPath image software analysis [69–71].

#### *4.8. Statistical Analysis*

Statistical analysis was performed using SPSS software (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY, USA). Tissue biomarker expression scores' correlation with the overall survival variable was performed by Fisher exact test. A K-mean cluster analysis was performed to sort relatively homogeneous groups of cases based on selected characteristics (CAF-1/p60 and CAF-1/p150 IHC expression). Survival analysis was performed by test Kaplan–Meier survival curves' differences by log-rank test. *p*-value was considered significant at *p* = 0.05.

#### *4.9. HR Reporter Assay*

The DR-GFP reporter and pCAGGS-I-SceI plasmids were used to verify the functionality of DNA repair mechanism by HR. The DR-GFP reporter plasmid, based on a construct developed by M. Jasin [72], consists of two mutated and GFP negative expression cassettes (GFP-I, GFP-II). In Cassette I, there is a unique cutting site recognized by the restriction enzyme I-SceI. In cells transfected with the DR-GFP and the I-SceI plasmids, the expression of I-SceI induces a double-stranded break (DSB) in Cassette I. This damage can be repaired through HR, using the GFP gene present as a template in Cassette II and thus restoring the expression of GFP (Figure 1). The ability of cells to repair double-stranded damage by HR was measured by flow cytometry reporting the percentage of positive GFP cells as a measure of HR proficiency.

#### **5. Conclusions**

The immunohistochemical analysis of CAF-1/p60 and CAF-1/p150 tissue expression in OSCC revealed a significant prognostic value of the combined expression of both tissue biomarkers, while no significant prognostic value was demonstrated in OPSCC. In HPV-positive and HPV-negative cell lines, the p60/150 CAF-1 subunits depletion impaired the proficiency of Homologous Recombination (HR) DNA damage repair, inducing sensitivity to the PARP-inhibitors drugs able to sensitize both the cell lines to ionizing radiations. These results indicate that, regardless of the prognostic meaning of the expression of the two markers, the pharmacological depletion of CAF-1 complex's function, combined to PARP-inhibitors and/or radiotherapy, could represent a valid therapeutic strategy for squamous cell carcinomas of head and neck region, by the mean of a synthetic lethality mechanism.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/11/10/1582/s1, Table S1: Contingency table showing frequency distribution of IHC expression data crossed with overall survival, Table S2 (A): The table shows the distribution of the three clusters frequency counts stratifying according to stage, age class and outcome, Table S2 (B): Demographic table showing the distribution of cases count frequency grouping by age class and outcome and stratifying by CAF p60/p150 clusters. The distribution was statistically significant in mid age class at Chi-squared test of significance (*p* = 0.002), Table S3: Type one (α) and type two (β) DNA damage in Cal27 Sh CTRL and CAL27 Sh p60/p150, using linear quadratic radiobiological model, Table S4: 34 out of 154 samples were upstaged in the staging score, 17/34 (50%) of the upstaged patients died at follow-up. The table shows the cunt of upstaged samples at the restaging according to the 8th edition of AJCC TNM manual. A&W = alive and well. DOD = death of disease, Table S5: Oligi sequence of the utilized primers, Figure S1: The percentage of g-H2AX positive nuclei at 1h and 4h from irradiation are in A and B. Error bars represent standard error mean. Results are representative of at least two independent experiments, Figure S2: p60/p150 tissue exprssion is an independent prognosti factor in a Cox multivariate analysis including also age and stage variables (global significance of the model *p* = 0,008, HR of p60high/p150low equal to 1789.

**Author Contributions:** F.M. (Francesco Morra), F.M. (Francesco Merolla), I.P., S.S. and A.C. made substantial contributions to the conception and design of the experiments; F.M. (Francesco Morra), I.P., F.L. and R.P. performed the in vitro experiments; F.M. (Francesco Merolla), D.R., G.I., and S.V. contributed to methodology and software employment for data acquisition, validation and formal analysis (as in M&M); F.M. (Francesco Morra), F.M. (Francesco Merolla), I.P., G.I., D.R., F.L., S.S. and A.C. actively participated in the interpretation of data; and F.M. (Francesco Morra), F.M. (Francesco Merolla), I.P., L.P., M.M., A.C. and S.S. were involved in drafting the manuscript or revising it critically for important intellectual content. All authors gave final approval of the version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

**Funding:** Our research was supported by POR Campania FESR 2014-2020 grant; "Technological Platform: eMORFORAD-Campania "grant PG/2017/0623667; POR Campania FESR 2014-2020 "SATIN" grant; POR Campania FESR 2014-2020 "RARE.PLAT.NET" grant; and "POR Campania FESR 2014-2020 "grant.

**Acknowledgments:** We are grateful to Mario De Felice for his countnuous support and encouragement, and to Rosaria Catalano for technical help.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


72. Jasin, M. Genetic manipulation of genomes with rare-cutting endonucleases. *Trends Genet.* **1996**, *12*, 224–228. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **PARP1 Co-Regulates EP300–BRG1-Dependent Transcription of Genes Involved in Breast Cancer Cell Proliferation and DNA Repair**

#### **Maciej Sobczak 1, Andrew R. Pitt 2, Corinne M. Spickett <sup>2</sup> and Agnieszka Robaszkiewicz 1,\***


Received: 17 August 2019; Accepted: 5 October 2019; Published: 11 October 2019

**Abstract:** BRG1, an active subunit of the SWI/SNF chromatin-remodeling complex, enables the EP300-dependent transcription of proliferation and DNA repair genes from their E2F/CpG-driven promoters in breast cancer cells. In the current study, we show that BRG1–EP300 complexes are accompanied by poly-ADP-ribose polymerase 1 (PARP1), which emerges as the functional component of the promoter-bound multiprotein units that are capable of controlling gene expression. This enzyme is co-distributed with BRG1 at highly acetylated promoters of genes such as *CDK4*, *LIG1,* or *NEIL3*, which are responsible for cancer cell growth and the removal of DNA damage. ADP-ribosylation is necessary to maintain active transcription, since it ensures an open chromatin structure that allows high acetylation and low histone density. PARP1-mediated modification of BRG1 and EP300 does not affect the association of enzymes with gene promoters; however, it does activate EP300, which acetylates nucleosomes, leading to their eviction by BRG1, thus allowing mRNA synthesis. Although PARP1 was found at BRG1 positive/H3K27ac negative promoters of highly expressed genes in a transformed breast cancer cell line, its transcriptional activity was limited to genes simultaneously controlled by BRG1 and EP300, indicating that the ADP-ribosylation of EP300 plays a dominant role in the regulation of BRG1–EP300-driven transcription. In conclusion, PARP1 directs the transcription of some proliferation and DNA repair genes in breast cancer cells by the ADP-ribosylation of EP300, thereby causing its activation and marking nucleosomes for displacement by BRG1. PARP1 in rapidly dividing cells facilitates the expression of genes that confer a cancer cell phenotype. Our study shows a new mechanism that links PARP1 with the removal of DNA damage in breast cancer cells via the regulation of BRG1–EP300-dependent transcription of genes involved in DNA repair pathways.

**Keywords:** poly-ADP-ribose polymerase 1 (PARP1); brahma-related gene 1 (BRG1); histone acetyltransferase p300 (EP300); gene transcription; cancer cell

#### **1. Introduction**

The pharmacological effects of inhibitors of poly-ADP-ribose polymerases (PARPs) in anticancer therapies are attributed to impairing DNA damage removal, since PARP1 plays a crucial role in the recruitment of repair machinery, mainly in an ADP-ribosylation-dependent manner [1,2]. Lesion recognition followed by the recruitment of poly-ADP-ribose polymerase 1 (PARP1) to sites of DNA damage and ADP-ribosylation of its automodification domain are prerequisites for the binding of XRCC1, POLB, LIG3, or ALC1, which are involved in base excision repair (BER), single-strand break repair (SSBR), and nucleotide excision repair (NER) [3,4]. However, PARP1 also facilitates alternative and conventional non-homologous end-joining (NHEJ) as well as homologous recombination (HR), therefore helping to protect genome integrity and preventing destabilization of the genome resulting from double-strand breaks [5–7]. According to the "access–repair–restore" model, nucleic acid repair is preceded by chromatin reorganization, since DNA lesions are curtained by DNA-associated proteins [8]. Thus, local chromatin rearrangements are required to allow the assembly of the multiprotein machinery that removes lesions. Recent findings identified a link between PARP1 activity, nucleosome density, and the efficiency of some repair pathways, including HR [9]. In the study referred to, auto-ADP-ribosylated PARP1 serves as an indispensable anchor that provides a platform at the damage site for the functional interaction between the nucleosome-evicting brahma-related gene 1 (BRG1; the SWI/SNF chromatin-remodeling enzyme) and the NAD-dependent deacetylase sirtuin-1 (SIRT1), which activates BRG1 by erasing lysine acetylation, thus promoting DNA end-joining. BRG1 interacts with the poly-ADP-ribose polymer through its ATPase domain rather than the N- or C-terminal tails, and is recruited to genomic regions enriched in PARP1. A previous paper reported the interaction of BRG1 with PARP1 and other histone-remodeling enzymes at the genomic level, where nucleosome-evicting ATPase cooperated with PARP1 and histone deacetylases (HDACs: HDAC2 and HDAC9) at the promoters of α*-MHC* and β*-MHC*, thereby preventing cardiac differentiation and maintaining the proliferation potential of the precursors [10]. However, the molecular mechanism that drives PARP1/BRG1-dependent up- or down-regulation of gene transcription has not yet been identified. The suggestion that PARP1 enables the binding of EP300 to the promoters of cell cycle-dependent genes in proliferating cells in an ADP-ribosylation-independent manner focused our attention on the possible role of PAR-synthesizing enzymes in the transcriptional regulation of genes controlled by BRG–EP300–HDAC1 complexes [11]. Our recent discoveries regarding the above-mentioned chromatin-remodeling units showed that these enzymes control the transcription of proliferation and DNA repair genes in two considerably different breast cancer cell lines, MCF7 and MDA-MB-231, which differ in terms of their expressed hormone and HER2 receptors [12]. The nucleosomes of E2F/CpG-driven promoters in the two studied gene groups were acetylated by histone acetyltransferase p300 (EP300), causing them to be marked for BRG1-mediated eviction, and enabling paused RNA polymerase II to become active, leading to active gene transcription. This mechanism operates only in proliferating cells, which are also characterized by a high abundance of PARP1. This is because the *PARP1* promoter is controlled by BRG1–EP300–HDAC1 complexes and is repressed with respect to the growth arrest seen in the great majority of normal primary cells [11].

Based on these premises, we aimed to discover whether PARP1 co-regulates BRG1–EP300-dependent transcription, and if PARP1 can be considered an active component of such multiprotein complexes in the studied breast cancer cell lines. We also aimed to uncover the molecular mechanism that links PARP1 with BRG1-dependent transcription and verify possible PARP1 selectivity toward functionally related genes.

#### **2. Results**

#### *2.1. PARP1 Physically Interacts with SWI*/*SNF in Breast Cancer Cells*

Data from three biological replicates run in duplicate for the PARP immunoprecipitates (IP) and two biological replicates in duplicate for the control IP were analyzed. A total of 76 interacting proteins were identified that fulfilled the selection criteria (confidence scores >50, fold change >2 and *p*-values < 0.05; the full list of PARP1-interacting proteins and associated data are shown in Table S1). Since PARP1 has been previously identified as a cofactor of the transcriptional machinery that cooperates with histone-remodeling enzymes and transcription factors [13–16], we focused on the identification of proteins that physically interacted with PARP1 within the cell nucleus, and initially assessed the interaction data for new, previously unidentified chromatin-associated proteins that could be involved in the regulation of gene expression in a PARP1-dependent fashion. The analysis of PARP1 co-immunoprecipitated proteins by mass spectrometry identified DNA-bound subunits of RNA

polymerase and mediator complexes, as well as subunits of chromatin-remodeling complexes, such as Tip60, p400 (EP400), and SWI/SNF (ARID1A, SMARCC1, BRG1; Table S1), which were significantly overrepresented in PARP1 versus IgG pull-downs. Interestingly, brahma (BRM), the other ATPase subunit in SWI/SNF, was not detected among the significant number of interacting proteins identified in PARP1 immunoprecipitates in any of three biological replicates from peptide identification in Mascot. This finding may suggest that PARP1 cooperates only with BRG1-based SWI/SNF complexes. However, to conclude on PARP1–BRM physical and functional interaction, further examination is needed.

Among other histone writers, erasers, and readers, we also found HDAC1, which was recently reported by us to be a constitutive component the BRG1–EP300–HDAC1 complex and assembles at the cell cycle-driven gene promoters of, for example, DNA repair genes in human macrophages and breast cancer cells [11,12]. To verify that formaldehyde fixation of the nuclei did not lead to false positive readouts with mass spectrometry and whether PARP1 is a bona fide member of the SWI/SNF complexes, PARP1 was immunoprecipitated from intact cells, and pull-downs were tested using Western blotting for the presence of SWI/SNF components previously detected by mass spectrometry in fixed nuclei. Western blotting of PARP1 co-immunoprecipitated proteins confirmed the direct interaction of PARP1 with ARID1A and BRG1, but also with other subunits of SWI/SNF, such as SMARCC1 and SMARCC2, in the studied breast cancer cell lines (Figure 1B). Similarly, PARP1 was detected in BRG1 pull-downs (Figure 1C).

**Figure 1.** Poly-ADP-ribose polymerase 1 (PARP1) physically interacts with the brahma-related gene 1 (BRG1)-based SWI/SNF chromatin-remodeling complex. (**A**) Exemplar mass spectrometry (MSMS; representative peptides from three biological replicates visualized in Mascot) data for the identification of ARID1 and SMARCA4 as PARP1 interactors using PARP1 immunoenrichment. (**B**) PARP1 interaction with ARID1A, SMARCC1, SMARCC2, and SMARCA4/BRG1 was confirmed by PARP1 pull-down and protein detection by Western blotting in cell lysates of two breast cancer cell lines, MCF7 and MDA-MB-231. IgG served as an isotypic control. (**C**) PARP1 was also identified in BRG1 immunoprecipitates. Western blotting images show representative images of three biological, fully reproducible replicates.

#### *2.2. PARP1 Is Co-Distributed with BRG1 at the Actively Transcribed Gene Promoters*

To confirm the possible role of PARP1 in the regulation of SWI/SNF-dependent transcription, we first tested whether PARP1 occurrence in the genome of breast cancer cells was accompanied by BRG1. The residual signal from PARP1 and BRG1 is randomly distributed across genomes, and may reflect the antibody specificity (or lack thereof) or experimental challenges rather than (or in addition to) true protein occurrence. However, the local enrichment of these proteins can be observed, and PARP1 and BRG1-rich regions can be identified by Model-based Analysis of ChIP-Seq data (MACS). As described in the Methods section, for these proteins, we set a *p*-value cutoff for peak detection at 10-3 with three levels of regions around the peak region (1, 5, and 10 kbp) assessed.

BRG1 was distributed and perfectly centered on genomic regions enriched in PARP1 (Figure 2A). BRG1/PARP1 peaks appeared predominantly at the gene regulatory regions, i.e., at the promoters and enhancers (Figure 2B; definitions of terms are given in Section 4.9), where BRG1 showed a relatively strong correlation with histone modifications, which marked open chromatin and transcriptionally active regions (Figure 2C). Focusing only at the E2F/CpG-driven gene promoters (Table S2), most of the PARP1 peaks (80.2%) were localized at the BRG1/H3K27ac-featured regions adjacent (±2 kbp) to the transcription start site (Figure 2D). This agrees with our and others' reports, where BRG1 has been documented to be associated with promoters of actively transcribed genes, characterized by histone marks, which are permissive for transcription, and the presence of CpG and E2F binding motifs. Our previous findings ascribed BRG1 as a master regulator of some H3K27ac positive promoters, which is essential for the initiation of mRNA synthesis [11,12].

PARP1/BRG1/H3K27ac-positive promoters represent genes that are functionally assigned to numerous processes that are crucial for the maintenance of intracellular homeostasis (Figure 2E; Table S3). To further investigate the molecular mechanisms that underlie the possible functional cross-talk between PARP1 and the BRG1-based SWI/SNF complex, we chose genes attributed to two groups, namely, DNA repair and positive regulation of the cell cycle. These groups were chosen because we recently selected them for an in-depth examination and explained a mutual interdependence between gene expression, BRG1 activity, promoter features, and cell proliferation status [12]. Most genes in the considered groups were over-expressed in at least one of the two fast-proliferating breast cancer cell lines when compared to normal cells derived from primary breast tissue (Figure 2F and Table S4). Particular attention was paid to three of our previously studied genes, *cyclin dependent kinase 4* (*CDK4*), *DNA ligase 1* (*LIG1*), and *nei like DNA glycosylase 3* (*NEIL3*). The transcription of these genes was driven by E2F/CpG/H3K27ac-positive promoters, which were found to be enriched in PARP1 in addition to BRG1 (Figure 2G). The molecular mechanism driving the transcription in MCF7 and MDA-MB-231 cells involved cooperation between EP300, which acetylated histones at the studied gene promoters, and BRG1, which evicted marked nucleosomes in proliferating cells [11,12].

**Figure 2.** PARP1 co-occurs with BRG1 at the promoters of transcriptionally active genes in MDA-MB-231 cells. (**A**) An example computeMatrix/plotProfile plot of ChIP-Seq data for BRG1. BRG1 peaks are centered at PARP1-enriched regions in the genome of the MDA-MB-231 cell. (**B**) Association of PARP1/BRG1 peaks with gene regulatory regions shows a prevailing occurrence at the active gene promoters. Enhancers emerged based on H3K27ac/H3K4me1 status, whereas promoters were assumed as TSS ± 2 kbp. PARP1/BRG1-enriched regions were assigned to particular groups and quantified by bedtools Intersect intervals. (**C**) BRG1 distribution at the PARP1 peaks reveals a relatively strong correlation with histone markers and with POLR2A, which is typical for actively transcribed genes (multiBamSummary/plotCorrelation; PARP1 peaks in bed for scoring). (**D**) Venn diagram showing that a high proportion of PARP1 positive gene promoters are characterized by the presence of BRG1 and strong acetylation of H3K27. (**E**) Triple-positive gene promoters identified in (**D**) represent a functional association with numerous intracellular processes. GO-enriched terms were identified in AmiGO2. (**F**) Differential expression of genes from two selected GOs, i.e., DNA repair and positive regulation of the cell cycle (marked in red in (**E**)) in cancer cells (DCIS – ductal carcinoma in situ, pre-invasive malignancy of the breast and two breast cancer cell lines: MCF7 and MDA-MB-231) versus normal cells, as quantified by Cuffdiff. The heatmap shows Log2 of the calculated fold change (Log2FC; cancer versus normal cells). (**G**) UCSC Genome browser visualization of PARP1, BRG1, and H3K27ac enrichment using Model-based Analysis of ChIP-Seq data (MACS) (bigwig) at the NEIL3 promoter.

#### *2.3. PARP1 Conditions Transcriptionally Permissive Chromatin Structure at BRG1*/*EP300-Dependent Genes*

To verify the PARP1 contribution to the transcription of genes that are concomitantly controlled by BRG1–EP300 complexes (described by Sobczak et al. [12]), we targeted PARP1 with siRNA in both studied breast cancer cell lines and measured the mRNA levels of the selected genes representing two gene ontologies (Figure 3A). PARP1 silencing resulted in considerable suppression of most genes, with *XRCC2* being the only exception that responded to PARP1 deficiency, but not to inhibition with olaparib (iPARP, pan-PARP inhibitor) with increased transcription in MCF7 cells. Furthermore, to check whether the observed PARP1 impact on gene transcription required enzymatic activity, cells were treated with olaparib, a PARP inhibitor. Loss of this enzymatic activity phenocopied PARP1 protein deficiency for *CDK4*, *BRCA1*, *LIG1*, and *NEIL3*, indicating that poly-ADP-ribosylation plays a key role in maintaining a high transcription rate of the considered genes (Figure 3A,B, Table S5). Enzyme inhibition led to a dramatic decline in the cellular abundance of CDK4, LIG1, and NEIL3 proteins (Figure 3C). PARP1 deficiency repressed the transcription of the three studied genes: *CDK4*, *LIG1,* and *NEIL3* comparably to iSWI/SNF and iEP300 (no synergistic effect was observed according to Supplementary Figure S1 and Table S5; only *LIG1* in MDA-MB-231 cells responded with enhanced gene repression after the treatment of siPARP1 transfected cells with iEP300 and iSWI/SNF), suggesting that this enzyme operates with the same, previously studied regulatory mechanism that utilizes the activity of BRG1 and EP300 at the three gene promoters considered [11,12], and that PARP1 may positively affect at least one of the chromatin-remodeling enzymes. This set of data suggests that PARP1 may also operate independently of EP300 and BRG1 (e.g., as a repressor of *XRCC2* in MCF7 cells).

Since EP300 and BRG1 drive gene transcription by respectively acetylating and displacing histones, to allow assembly of the transcriptional machinery, we focused on nucleosome acetylation status and density as possible readouts of PARP1 activity to identify the molecular basis of the observed effect of poly-ADP-ribosylation on BRG1–EP300-dependent gene expression. PARP inhibition with olaparib led to a substantial loss of histone acetylation and was associated with an increase in histone density (Figure 3D; H3 enrichment and status of H3K27ac for each of the studied promoters can be found in Table S5: sheet: LIG1, NEIL3, CDK4 ChIP)); the *XRCC1* promoter was used as a negative control since it lacks PARP1 (Figure S2; Table S5: sheet: XRCC1 ChIP). This finding confirmed that ADP-ribosylation impacts BRG1–EP300 complexes in rapidly proliferating cells and defines the output of the considered chromatin-remodeling functional unit.

Knowing that BRG1 and EP300 co-occur at the studied gene promoters with HDAC1, the observed PARP1 effect on histone acetylation and gene transcription may result from PARP1 interaction with either of the two enzymes, since the subtle balance between acetylase and deacetylase activity determines the BRG1-dependent chromatin structure [11,12]. Thus, we tested whether poly-ADP-ribosylation inhibited HDAC1 activity at the gene promoters by comparing gene expression in the presence of HDAC and PARP inhibitors (Figure 3E). First, HDAC1 did not reduce the transcription of any of the three genes, and second, cell treatment with a mixture of both inhibitors suppressed *CDK4*, *LIG1*, and NEIL3 in a similar way to iPARP only (PARP and HDAC activities had no synergistic impact on the gene expression; Table S5). This indicates that olaparib does not inhibit HDAC1 (or any other histone deacetylase, since we used a pan-HDAC inhibitor) activity from poly-ADP-ribosylation-dependent inhibition at the studied gene promoters.

**Figure 3.** ADP-ribosylation confers open chromatin structure at the gene promoters. (**A**) PARP1 silencing leads to the suppression of most genes in MCF7 and MDA-MB-231 cells that feature PARP1/BRG1/H3K27ac-positive promoters. mRNA was compared 48 h after cell transfection with siCTRL and siPARP1. Log2 of the calculated fold change (Log2FC) shows gene expression in cells treated with inhibitors and normalized to untreated cells. The silencing of PARP1 was confirmed by Western blotting (below heatmap), and H3 was used as a loading control. A similar effect was observed upon PARP1 inhibition with olaparib (iPARP; 48 h) at both the mRNA (**B**) and protein level. (**C**) Representative pictures of protein detection by Western blotting. (**D**) Analysis of structure of selected PARP1-dependent gene promoters revealed a considerable loss of histone acetylation, but increased nucleosome density upon PARP1 inhibition for 24 h. Quantification was carried out by ChIP-qPCR, and data for specific antibodies were normalized first to 10% of the corresponding input and then to untreated control cells. (**E**) The iPARP effect on gene transcription with HDAC activity deficiency (cells were treated with both inhibitors for 48 h) was studied by real-time PCR. Results are presented as Log2 of the calculated fold change (inhibitor versus untreated; Log2FC).

#### *2.4. Poly-ADP-Ribosylation of EP300 Drives BRG1–EP300-Dependent Gene Transcription*

Identified PARP1 interaction with BRG1 frequently occurred at highly acetylated gene promoters (Figure 2B). The fact that at least some of them were previously confirmed to be enriched in BRG1–EP300 functional complexes [12] prompted us to check if EP300 interacted with PARP1 and if any of BRG1–EP300 components could undergo ADP-ribosylation in proliferating breast cancer cells.

Analysis of PARP1 pull-downs by Western blotting confirmed the physical interaction between PARP1 and EP300 (Figure 4A). Poly-ADP-ribose chains were detected in both immunoprecipitated enzymes, i.e., BRG1 and histone acetyltransferase, in the studied cancer lines (Figure 4B,C), thereby providing further evidence for a possible PARP1 role in the regulation of transcriptional activity of BRG1–EP300 complexes. Bearing in mind that the dependence of gene transcription on BRG1 and EP300 is conditioned by the association of enzymes with their gene promoters and then by the catalytic activity of chromatin-bound enzymes, we tested whether poly-ADP-ribosylation affected BRG1 and EP300 levels at the investigated promoters (Figure 4D). None of the studied genomic regions responded to PARP1 inhibition with substantial displacement or recruitment of chromatin remodeling enzymes, suggesting that poly-ADP-ribosylation determines the activity of enzymes rather than their occurrence at the gene promoters (Figure 4D, Table S5). The poly-ADP-ribosylation of EP300 enabled acetyltransferase activity that led to intensive nucleosome acetylation and eviction by BRG1, since cell treatment with a PARP inhibitor resulted in a dramatic loss of EP300-dependent acetylation of the studied gene promoters (Figure 3D, Table S5), without an apparent effect on the association of EP300 with the gene promoters and the HDAC1 role in gene transcription (Figures 4D and 3E).

To check if poly-ADP-ribosylation of BRG1 directly conditioned BRG1 activity and BRG1-dependent transcription, we tested the impact of the PARP1 inhibitor on the transcription of genes that are over-expressed in cancer cells and characterized by the occurrence of PARP1 and BRG1 at their promoters, but without considerable nucleosome acetylation (Figure 4D and Table S2). *IL1RL1* served as an example of repressed genes in MCF7 cells. Surprisingly, all of the genes found to be over-expressed that were considered in this experiment responded to SWI/SNF inhibition and silencing with increased transcription (Figure 4E, Table S5), suggesting that EP300 co-distribution with BRG1 might be a hallmark of gene promoters characterized by the pro-transcriptional activity of BRG1-based SWI/SNF complexes. However, this hypothesis requires further examination of a wider range of genes, especially because the considerable inhibitory role of BRG1 on gene transcription was observed in only one cell line. This finding also stresses the differences in gene transcription control in the two chosen cell lines. In any case, PARP1 was not involved in the transcriptional regulation of genes suppressed by the SWI/SNF complex; only one repressive effect was found for *RAD1* in MDA-MB-231 cells. Together, these data indicate that PARP1 co-regulates activity of promoter-bound BRG1–EP300 complexes, and that poly-ADP-ribosylation of EP300 is required to enable the BRG1-dependent eviction of acetylated nucleosome, and therefore the transcription of genes involved in key intracellular processes, such as cell division and the removal of DNA damage in breast cancer cells. This molecular mechanism of PARP1 action in BRG1–EP300 complexes is further supported by our previous findings, in which BRG1 emerged as a reader of nucleosome acetylation [11,12]. Thus, a low histone acetylation caused by PARP inhibition prevents histone eviction by BRG1.

**Figure 4.** ADP-ribosylation of EP300 drives BRG1-dependent gene transcription. (**A**) The PARP1 interaction with EP300 was confirmed by visualization of EP300 by Western blot in PARP1 immunoprecipitates. (**B**) ADP-ribosylation of BRG1 was detected in BRG1 pull-downs by Western blot. (**C**) ADP-ribosylation of EP300 was detected as in (**B**) in EP300 pull-downs. (**D**) The effect of ADP-ribosylation on EP300 and BRG1 association with gene promoters in breast cancer cells was studied by ChIP-qPCR. Cells were supplemented with iPARP for 24 h prior to cell fixation. Data for anti-EP300 and anti-BRG1 were normalized to 10% input and then to untreated cells (Log2 enrichment vs. control cells). (**E**) The lack of contribution of ADP-ribosylation to the transcription of genes characterized by BRG1/PARP1 (no EP300) positive promoters was verified by comparing mRNA (real-time PCR) between the studied groups. For inhibitors, Log2FC was normalized to untreated cells, while data for BRG1 silencing were normalized to cells transfected with control siRNA. The efficiency of BRG1 silencing is confirmed by Western blot (below heatmaps). The column with differential expression shows gene transcription in MCF7 and MDA-MB-231 cell lines normalized to normal cells (data from RNA-Seq).

#### **3. Discussion**

Malignant transformation directs cellular changes and adaptive responses toward new requirements that cancer cells face during growth and metastasis. As long as the activation of oncogenes, loss of cell cycle checkpoint control, and impaired DNA repair capacity favor carcinogenesis, genome

integrity and cell cycle re-entrance will eventually become threatened due to increased energy demand, mild but persistent oxidative stress, and the modulation of signaling cascades necessary for tumor growth and the invasion of other tissues [17,18]. The altered expression of many gene products in response to cell transformation is affected by reprogramming the epigenome, resulting in changes in the transcription and reconstruction of the cellular proteome to meet emerging needs [19]. According to our findings, the activation of *PARP1* transcription as a consequence of the transition of a cell from quiescence to proliferation may help the cancer cell gain the necessary adaptive physiology by acting at the genomic level in two ways: directly, by contributing to DNA repair machineries, and indirectly, by affecting the transcription of BRG1–EP300 targets, among others, which enable the cancer cell to rapidly divide and resist DNA damaging agents [12,20–22]. Although the first aspects of this mechanism have been relatively deeply explored and the regulatory roles of PARP1 regarding the removal of DNA lesions and the transition between consecutive cell cycle phases in various modes—for example, the modulation of SP1 activity, H1 displacement at proliferation-relevant genes, and nuclear retention of PKM2—have been documented, the detailed molecular mechanism in regard to functionally linked genes has not been identified until now [23–25]. In this study, PARP1 was shown to be an active component of the transcription machinery that drives BRG1-EP300-dependent gene expression by the poly-ADP-ribosylation of EP300 in breast cancer cells. PARP1 was highly enriched at gene promoters characterized by the occurrence of not only BRG1 and EP300, but also E2F binding motif(s) and CpG islands. Since these features apply to many functionally linked genes, the role of PARP1 in defining the breast cancer phenotype at the transcription level is likely to go far beyond cell cycle progression and the removal of DNA damage. As shown in Figure 2E, triple-positive PARP1/BRG1/H3K27ac promoters represent the genes responsible for signal transduction and autophagy. Once poly-ADP-ribosylation is proved to be a co-activator of these genes, PARP inhibitors may be important to consider for pharmacological interventions that target and suppress mediators of pro-survival cascades at the genomic level. The role of poly-ADP-ribosylation in the fine-tuning of numerous intracellular processes simultaneously allows the maximization of the effectiveness of PARP1 inhibitors in rendering cancer cells vulnerable to anticancer drugs, which challenge PARP1-dependent or concurrent intracellular routes. Bearing in mind that the BRG1/EP300 complex was shown to operate at gene promoters in proliferating breast cancer cells and human macrophages due to an association with E2F transcription factors [12,26], the same PARP1-dependent mechanism of transcription control likely applies to other tumors, since the cell cycle status conditions both PARP1 expression and the activity of BRG1–EP300 complexes. Similarly, the energy status of proliferating cells demands a high ATP concentration; thus, the NAD+/NADH redox ratio determines, for example, that PARP1 activity is five times higher in cancer cells than in normal, non-transformed cells [27]. Therefore, it might be possible to adapt PARP1 inhibitors for the modulation of intracellular processes in a wider range of cancers.

This study revealed a new mechanism that defines cancer cell responses to DNA lesions in a poly-ADP-ribosylation-dependent manner. Although the direct contribution of PARP1 to pathway repair at sites of damage is well acknowledged [28,29], we showed for the first time an impact on the repair mechanisms that is distant, gene promoter-related, and independent of lesion location. In this context, PARP inhibitors could be used to suppress the transcription of genes characterized by promoters enriched in PARP1, BRG1, and EP300, which represent executors that are crucial for BER, SSBR, NER, MMR, HR, and alt-NHEJ, since products of only the considered genes, i.e., *BRCA1*/*2*, *LIG1,* and *NEIL3,* contribute to more than one repair pathway [30,31]. Another benefit of using PARP inhibitors as DNA repair modulators acting at the level of the epigenome comes from the observation that poly-ADP-ribosylation is a co-activator of cell cycle-dependent genes that are simultaneously controlled by BRG1 and EP300, and are mostly over-expressed in the studied breast cancer cells, thus providing some selectivity toward this group. PARP1 was also found at gene promoters occupied by BRG1 alone, but cell treatment with olaparib did not reveal a considerable impact on the expression of these genes. Notably, the PARP1/BRG1 promoter response to an SWI/SNF inhibitor that is capable of impeding the activity of both ATPases (Figure 4D) suggested that (a) EP300 has a co-activating role with BRG1 at the

BRG1/EP300 promoters, and/or (b) the role of particular ATPases may be determined by EP300, and the lack of acetyltransferase may switch BRG1 to BRM activity. Although the scope of iSWI/SNFs is broader than that of iPARPs, the unexpected reaction of the chromatin to the loss of BRG1 and BRM activity may result in a coinciding repression and up-regulation of genes assigned to one regulatory circuit. Discrete inhibitors of the two ATPases mentioned, as well as further study on the mutual interdependence between these two enzymes, are needed for future clinical applications. Similarly, inhibitors of EP300 modulate the transcription of a significant number of genes at different levels; at the genome level by preventing acetylation of transcription co-factors and nucleosomes, and at the signaling cascade level by modifying signal transducers [32,33]. This all results in even less specificity toward the desired gene pool, and underlines the importance of the possible applications of PARP1 inhibitors. Some PARP-1 and PARP-2 inhibitors, such as olaparib, niraparib, rucaparib, veliparib, and talazoparib, which are small-molecule NAD<sup>+</sup> mimetics, are currently being studied in later-stage clinical trials or are already approved for breast and ovarian cancer treatment with deleterious germline *BRCA1* and *BRCA2* mutations, which predispose women to develop triple-negative and hormone-receptor-positive, human epidermal growth factor receptor 2-negative breast cancers, respectively [34,35]. Since PARP inhibitor monotherapy strategies are effective in cancers with homologous recombination repair defects and are relatively well-tolerated by patients, they can be considered for the treatment of a wider range of cancers, both in combined therapies, due to the well-established fact that these drugs sensitize cells to DNA-damaging chemotherapy and radiation therapy, or as an alternative to taxanes and a supplement to anthracyclines [36–38]. However, numerous phase I clinical trials utilizing a combination of cytotoxic chemotherapy with PARP inhibitors failed to confirm any beneficial effects of such combinations. Therefore, the use of these drugs in adjuvant or neoadjuvant settings may need substantial revision, while also taking into consideration the myelosuppressive effects of PARP inhibitors and the careful selection of anticancer agents in combination with DNA repair inhibitor(s). Nevertheless, the long list of promoters of functionally related genes that are enriched in PARP1 presented in this manuscript suggest the likely involvement of this enzyme regarding the modulation of other intracellular processes at the transcription level. These findings open the gate for new ideas and concepts regarding anticancer approaches, which require verification first in cell and animal models.

The described contribution of PARP1 to the regulation of BRG1–EP300 activity emphasizes the role of PARP1 in chromatin remodeling. Although a number of papers have documented this enzyme as a direct or indirect regulator of chromatin structure in a context-dependent fashion, none have provided an overall mechanism for functionally linked gene sets in particular. Transcription in cells is shaped by the poly-ADP-ribosylation of nucleosomes, histone writers and erasers (KDM5B), transcription factors (e.g., C/EBPβ), or POL2 regulating co-factors (NELF), as well as physical, activity-independent interactions with gene promoters that define chromatin composition (LSD1, EP300) [13–16,39,40]. However, DNA motifs or chromatin signatures, which determine PARP1 distribution in the genome, remain unidentified. According to Gibbson, PARP1 and ADP-ribosylation correlated with histone markers (H3K4me3 and H3K27ac) featuring actively transcribed genes and with POL2 pausing machinery in embryonic fibroblasts of mice [40]. These findings agree with our own, in which PARP1 was enriched at highly acetylated CpG islands, allowing immediate POL2 pausing or release in response to received signals [11,12]. The association of PARP1 with GC-rich regions impedes the identification of the PARP1-specific motif in promoter sequences. Since only 19% of PARP1 peaks in the genome of MDA-MB-231 cells occurred at BRG1 and H3K27ac negative promoters, and less than 3% outside of BRG1 peaks, these two features of promoters, together with E2F-binding motifs and CpG islands, seem to direct the enzyme to its destination regarding chromatin, whereby the poly-ADP-ribosylation of BRG1 and EP300 enables gene expression.

#### **4. Materials and Methods**

#### *4.1. Materials*

Two epithelial, breast cancer cell lines, derived from metastatic sites, MCF7 (estrogen and progesterone receptors-positive, HER2-negative) and MDA-MB-231 (triple negative) were purchased from ATCC and Sigma Aldrich (Poznan, Poland), respectively. DMEM high glucose with L-glutamine with sodium pyruvate for MCF7, fetal bovine serum, and antibiotics (penicillin and streptomycin) were from Biowest (CytoGen, Zgierz, Poland), L15 Medium for MDA-MB-231, iSWI/SNF (PFI-3), iHDAC (sodium butyrate), anti-rabbit IgG (A0545) and anti-mouse IgG (A4416) (whole molecule)–peroxidase antibody produced in goat, BLUeye prestained protein ladder (#94964), oligonucleotides for real-time PCR, SIGMA*FAST*™ Protease Inhibitor Tablets (PIC) were from Sigma Aldrich (Poznan, Poland). iPARP (olaparib, AZD-2281) was from Cayman Chemical (Biokom, Janki/Warsaw, Poland). Lipofectamine RNAiMAX, OptiMem, Dynabeads™ Protein G, glycogen, High-Capacity cDNA Reverse Transcription Kit, SuperSignal™ West Pico Chemiluminescent Substrate, TRI Reagent™, and RNase A were from Thermofisher Scientific (Thermofisher Scientific, Warsaw, Poland). KAPA HiFi™ HotStart ReadyMix (2×) from KapaBiosystems and Takyon™No ROX SYBR Core Kit blue dTTP from Eurogentec were purchased from Polgen (Lodz, Poland). EvaGreen® Dye, 20X in water was purchased from Biotium (Corporate Place Hayward, Fremont, CA, USA). WB antibodies: anti-DNA Ligase I (sc-271678), anti-CDK4 (sc-23896), anti-NEIL3 (sc-393703), anti-pADPr (sc-56198), siPARP1 (sc-29437), and gallotannin were purchased from Santa Cruz Biotechnology (AMX, Lodz, Poland). ChIP grade antibodies: normal rabbit IgG (#2729), anti-ARID1A (#12354), anti-SMARCC2 (#12760), anti-BRG1 (#49360), anti-H3K27ac (#4353), anti-histone H3 (#4620), and anti-PARP1 (#9532) were purchased from Cell Signaling Technology (LabJOT, Warsaw, Poland). Human Cytokine and Chemokine Receptor Primer Library (HCCR-I) were from RealTime Primers (Prospecta, Warsaw, Poland). For the mass spectrometric analysis, all materials were from Thermo Fisher Scientific (Loughborough, UK) unless otherwise indicated. Porcine Trypsin (Trypsin Gold Mass Spectrometry Grade) was from Promega (Southampton, UK), and general use Protease Inhibitor Cocktail (P2174) was from Sigma-Aldrich (Poole, UK).

#### *4.2. Cell Culture and Treatment with Inhibitors*

MCF7 were cultured in DMEM supplemented with 10% FBS, penicillin/streptomycin (50 U/mL and 50 μg/mL, respectively) in 5% CO2, whereas MDA-MB-231 was cultured in F15 medium supplemented with 15% FBS and penicillin/streptomycin (50 U/mL and 50 μg/mL, respectively) without CO2 equilibration. After freezing and thawing, cells were cultured in DMEM as described for MCF7 cells. iSWI/SNF (10 μM; PFI-3), iPARP1 (olaparib, 1 μM), and iHDAC1 (sodium butyrate, 250 μM) were added to cells 48 h prior to analysis.

#### *4.3. Quantification of Gene Expression*

mRNA quantification was conducted as described in Pietrzak et al. [11] using Takyon™ No Rox SYBR® MasterMix dTTP blue (Eurogentec—from local distributor—Polgen, Lodz, Poland) and CFX96 C1000 Touch (BioRad, Warsaw, Poland) for real-time PCR. The median average of ACTB, GAPDH, and B2M were used for normalization. Data in figures are shown as Log2FC with respect to untreated control or to siCTRL (indicated in figures or figure legends).

For protein detection, cell lysates were processed as previously described and visualized using SuperSignal™ West Pico Chemiluminescent Substrate. Pictures were acquired with ChemiDoc-IT2 (UVP, Meranco, Poznan, Poland).

The following primer sets were used for the quantification of gene expression: *CDK2*, 5 -CAGGATGTGACCAAGCCAGT-3 (forward) and 5 -TGAGTCCAAATAGCCCAAGG-3 (reverse); *CDK4*, 5 -CTGGTGTTTGAGCATGTAGACC-3 (forward) and 5 -AAACTGGCGCATCAGATCCTT-3 (reverse), *XRCC2*, 5 -TCGCCTGGTTCTTTTTGCA-3 (forward) and 5 -TCTGATGAGCTCGAGGCTTTC-3 (reverse), *BRCA2*, 5 -CTTGCCCCTTTCGTCTATTTG-3 (forward) and 5 -TACGGCCCTGAAGTACAGTCT-3 (reverse), *LIG1*, 5 -CAGAGGGCGAGTTTGTCTTC-3 (forward) and 5 -AGCCAGTTGTGCGATCTCTT-3 (reverse), *EXO1*, 5'-AAACCTGAATGTGGCCGTGT-3 (forward) and 5 CCTCATTCCCAAACAGGGACT-3 (reverse), *NEIL3*, 5 -GGTCTCCACCCAGCTGTTAAAG-3 (forward) and 5 -CACGTATCATTTTCATGAGGTGATG-3' (reverse), *PCNA*, 5'-TCTGAGGGCTTCGACACCTA-3' (forward) and 5 - TTCTCCTGGTTTGGTGCTTCA-3 (reverse); *BRG1*, 5 -AAGAAGACTGAGCCCCGACATTC-3 (forward) and 5 -CCGTTACTGCTAAGGCCTATGC-3 (reverse), *BRCA1*, 5 -TGCCCACAGATCAACTGGAA-3 (forward) and 5 - CACAGGTGCCTCACACATCT-3 (reverse); *ACTB*, 5 - TGGCACCCAGCACAATGAA-3 (forward) and 5 -CTAAGTCATAGTCCGCCTAGAAGCA-3 (reverse); PARP1, 5 - AAGCCCTAAAGGCTCAGAACG-3 and 5 -ACCATGCCATCAGCTACTCGGT-3 . *GAPDH* and *B2M* were from the Human Toll-like Receptor Signaling Primer Library (HTLR-I, RealTime Primers – from local distributor - Prospecta, Warsaw, Poland)).

#### *4.4. PARP1 Co-Immunoprecipitation for Mass Spectrometry*

<sup>2</sup> <sup>×</sup> <sup>10</sup><sup>7</sup> MCF7 cells were trypsinized, washed 3<sup>×</sup> with cold PBS, and lysed on ice in hypotonic buffer (50 mM HEPES-KOH, 10 mM NaCl, 1 mM EDTA, 10% glycerol, and 0.5% NP-40, 0.25% TritonX-100). Nuclei were washed twice in PBS supplemented with protease inhibitor cocktail (10× stock added at 10% of volume, Sigma P2714 General Purpose) and fixed in 1% formaldehyde in PBS on stirrer for 10 min at room temperature. Formaldehyde was quenched by the addition of 125 mM oglycine, and after 20 min of incubation, nuclei were washed twice in 10 mM Tri-HCl (pH = 8.0), 100 mM NaCl, and 1 mM EDTA, and finally resuspended in 10 mM Tri-HCl (pH = 8.0), 150 mM NaCl, 1 mM EDTA, 0.05 Na-deoxycholate, and 0.25% N-laurosarcosine. Nuclei were sonicated until the solution was transparent, TritonX-100 was added to 1% and then centrifuged (10,000× g, 4 ◦C, 10 min) to remove insoluble material. The supernatant was split into two samples, to which were added either control IgG–Dynabead or anti-PARP1–Dynabead conjugates (prepared by the incubation of 5 μg of antibody and 10 μL of magnetic beads in 10 mM Tri-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 0.05 Na-deoxycholate, and 0.25% N-laurosarcosine for at least 15 min). After overnight incubation on a roller shaker at 4 ◦C, the supernatant was removed, and the beads were washed 5× with 50 mM HEPES-KOH (pH = 7.5), 500 mM LiCl, 1 mM EDTA, 1% NP-40, and 0.7% Na-deoxycholate; twice with 50 mM NaCl in TE buffer (10 mM Tris pH = 8.0, 1 mM EDTA); and twice with 50 mM ammonium bicarbonate using a magnetic stand. Then, the beads were incubated with trypsin (Promega Gold, 1 μg/μL) in 50 mM ammonium bicarbonate for 6 h at 36 ◦C, after which the supernatant was transferred to new tubes and dried in a SpeedVac vacuum concentrator (Eppendorf, Stevenage, UK).

#### *4.5. Mass Spectrometry Analysis of PARP1 Co-Immunoprecipitates*

Tryptic digest samples were resuspended in 25 μL of 2% acetonitrile in water and 0.5% formic acid. Peptides were separated and analyzed using a U3000 nanoflow LC system (Thermo) connected to a 5600 Triple ToF mass spectrometer (Sciex, Warrington, UK). Then, 10 μL of sample was loaded onto a 0.5 × 5 mm PepMap C18 trap, washed with buffer A (2% acetonitrile 98% water containing 0.5% formic acid) for 4 min, and then separated by a 90-minute gradient from 2% to 40% buffer B (98% acetonitrile, 2% water containing 0.5% formic acid) on a 0.075 × 150 mm PepMap C18 column (Thermo Fisher Scientific, Loughborough, UK)). MSMS data were collected for precursors of 2<sup>+</sup> to 5<sup>+</sup> charge state in the range m/z 350–1250 Th using a top 10 data dependent acquisition method, collecting MS data for 200 ms and MSMS data for 100 ms, with dynamic exclusion for 15 s and a standard rolling collision energy settings. MSMS data was collected in the range of m/z 50–2000 Th. All other settings were optimized for peptides using a standard mixture. Samples were run as biological replicates. MSMS data was analyzed using Progenesis QIP (Waters, Manchester, UK) for label-free quantitative analysis and Mascot (Matrix Science, London) for protein identification. Data was loaded as .wiff

files into Progenesis QIP, automatically aligned, and peak picked using the default settings; then, the alignment was manually improved where necessary. Default settings were used for peak picking, and Hi-N was used for quantification. Data was exported to Mascot using the default settings. For the Mascot analysis, the SwissProt Mammalian database (2018\_09) was searched, allowing 50 ppm error for MS and 100 ppm for MSMS data, two missed cleaves, methionine oxidation as a variable modification, and an overall FDR of <1%. Data was re-imported into Progenesis QIP for further quantitative analysis. Protein identifications were deemed significant if more than two peptides were identified with an overall confidence score greater than 50, but more stringent criteria were applied for proteins to be further investigated. Quantification data was considered significant where the ANOVA *p*-value was less than 0.05, the fold change was greater than 2, and the highest mean was in the PARP immunoprecipitation.

#### *4.6. Co-Immunoprecipitation and Western Blot*

MCF7 and MDA-MB-2331 cells were lysed in 20 mM Tris-HCl (pH = 7.5), 75 mM KCl, 5 mM MgCl2, 0.2 mM EDTA, 10% glycerol, 0.1% Tween20, and PIC (IP buffer); sonicated with the ultrasonic homogenizer Bandelin Sonopuls (HD 2070; 10 impulses, 60%); and centrifuged (10,000× *g*, 4 ◦C, 10 min). Supernatant was incubated with anti-PARP1, anti-BRG1, anty-EP300, and corresponding IgG at 4 ◦C for 2 h. For another 1 h, lysates were added with Dynabeads (5 μL); then, they were washed 5× with the IP buffer and once in 50 mM NaCl in TE buffer (10 mM Tris pH = 8.0, 1 mM EDTA). Beads were suspended in gel loading buffer supplemented with 5% β-mercaptoethanol, and heated at 70 ◦C for 10 min. Beads were collected on a magnetic stand and supernatant was separated by PAGE. BRG1, ARID1A, SMARCC1, SMARCC2, PARP1, EP300, H3, and poly-ADP-ribose were detected on nitrocellulose membranes after overnight staining with corresponding antibodies. For the detection of poly-ADP-ribosylation, cells were lysed and processed in the presence of a PARG inhibitor: 0.5 mM gallotannin.

Each immunoprecipitation followed by Western blot was repeated in three biological replicates. Each time, the striking difference in protein level being detected was observed between IgG (weak or lack of signal) versus anti-PARP1 (or anti-BRG1; strong and clear bands). Representative images were taken for figures.

#### *4.7. Chromatin Immunoprecipitation*

The immunoprecipitation of chromatin-bound proteins and histones was carried out according to the protocol previously described [11]. For the quantification of H3K27 acetylation, cells were lysed and processed in the presence of iHDAC (0.5 mM). Fragments spanning PARP1/BRG1/H3K27ac sites in selected gene promoters were amplified using KAPA HiFi™ HotStart ReadyMix supplemented with EvaGreen® Dye and 4% DMSO. Primers for *CDK4*, *LIG1, NEIL3*, and *XRCC1* promoters were as follows: *CDK4* prom, 5 -ATAACCAGCTCGCGAAACGA-3 and 5 -AGAGCAATGTCAAGCGGTCA-3 , *LIG1* prom, 5 -AACACACTCAGATCCGCCAG-3 and 5 -GCTTCCACCGATTCCTCCTC-3 , *NEIL3* prom, 5 -GTAGGGAGCGACCTCAACAG-3 and 5 -AGTACAGCCTGGTCCTTCCA-3 , *XRCC1* prom, 5 - TGGCCAGAAGGATGAGGTAG-3 (forward) and 5 -AGGAAACGCTCGTTGCTAAG-3 (reverse).

#### *4.8. Transient Gene Silencing*

For PARP1 and BRG1 silencing, MCF7 and MCD-MB-23 were seeded at the density of 100,000 cells per well, transfected on the following day with RNAiMAX-siRNA complexes (3 μL of transfection reagent and 20 nmol siRNA incubated in OptiMem for 20 min). The silencing was confirmed by real-time PCR and Western blot 48 h after cell transfection.

#### *4.9. ChIP-Seq Analysis in Galaxy Version 19.05.dev [41]*

The following, publically available, generated by other groups and deposited in the *PubMed* Central Database data from MDA-MB-231 cells were taken for ChIP-Seq analysis: PARP1—GSM1517306 (SRR1593959), BRG1—GSM1856026 (SRR2171350), GSM1856027 (SRR2171351), and GSM1856028 (SRR2171352), H3K27ac—GSM1855991 (SRR2171311) and GSM1855992 (SRR2171312); H3K4me3—GSM1700392 (SRR2044734), H3K4me1—GSM2036932 (SRR3096750 and SRR3096751), H3K27me3—GSM949581 (SRR513994), H3K9ac—GSM1619765 (SRR1820123 and SRR1820124), H3—GSM2531568 (SRR5332805), POLR2A—GSM2309434 (SRR4240635), and Input—GSM1964894 (SRR2976843). FASTQ quality formats were unified to Sanger formatted with a FASTQ Groomer [26]. Reads were aligned to Human Genome version 19 using a Map with Bowtie for Illumina, and unmapped reads were filtered out. ChIP-seq peaks were called in MACS with a *p*-value cutoff for peak detection set at 10<sup>−</sup>3. BRG1 co-occurrence at the PARP1 peaks was monitored by computeMatrix/plotProfile (PARP1 peaks in bed were used as regions to plot, while mapped BRG1 reads were used for scoring) [42]. The co-distribution of BRG1, POLR2A, and histone modifications at the PARP1-enriched regions was studied by MulitBamSummary/plotCorrelation (regions of the genome were limited to PARP1 peaks in bed, mapped reads for scoring) [42]. Regions simultaneously enriched in BRG1, PARP1, H3K4me1, and H3K27ac were identified by returning intersects of the peaks in bed using bedtools Intersect intervals [43]. Regions localized outside of gene promoters and double positive for H3K4me1 and H3K27ac were assigned as active enhancers high in H3K4me1 and low in H3K27ac as inactive enhancers, while gene promoters were recognized by returning intersects of BRG1/PARP1 peaks and genomic regions ±2000 bp centered on TSS (overlapping intervals of both datasets). Promoters defined by high H3K27ac and associated with the presence of gene transcripts (RNA-Seq results for MDA-MB-231) downstream of corresponding TSS were assumed as active, while the lack of promoter acetylation marked inactive gene promoters. Genomic intervals for E2F (overlaps of E2F1 and E2F4) and CpG islands were taken from the UCSC main tables wgEncodeRegTfbsClusteredV3 and cpgIslandExt, respectively. Intersects of TSS ±2 kbp and CpG or E2F intervals were compared using bedtools Intersect intervals. The characteristics of gene promoters (occurrence of particular proteins and CpGs, histone modifications, E2F binding sites) were studied using Venn diagrams, which were created in http://www.interactivenn.net/ from gene lists. The annotation of PARP1/BRG1/H3K27ac promoters to biological processes was carried out in AmiGO2 (test type—binominal, correction—FDR) [44]. PARP1, BRG1, and H3K27ac peaks were visualized in the UCSC Genome Browser.

The following, publically available datasets for various breast cells, which have been generated by other groups, were downloaded from the PubMed Central Database data and used for ChIP-Seq analysis: data from normal breast, ductal carcinoma in situ (DCIS; pre-invasive malignancy of the breast), MCF7 and MDA-MB-231 cells were taken for RNA-Seq analysis: normal breast—GSM1695870 (SRR2040339), GSM1695872 (SRR2040341), GSM1695873 (SRR2040342), GSM1695874 (SRR2040343), GSM1695877 (SRR2040346), and GSM1695878 (SRR2040347); DCIS—GSM1695891 (SRR2040360), GSM1695898 (SRR2040367), GSM1695899 (SRR2040368), GSM1695882 (SRR2040351), GSM1695890 (SRR2040359), and GSM1695894 (SRR2040363); MCF7—GSM2422725 (SRR5094305), GSM2422726 (SRR5094306), GSM2422727 (SRR5094307), GSM2422728 (SRR5094308), GSM2422729 (SRR5094309), and GSM2422730 (SRR5094310); MDA-MB-231—GSM2422731 (SRR5094311), GSM2422732 (SRR5094312), GSM2422733 (SRR5094313), GSM2422734 (SRR5094314), GSM2422735 (SRR5094315), and GSM2422736 (SRR5094316). All samples were processed as described in Sobczak et al. [12]. Differential gene expression in cancer versus normal breast cells was calculated with Cuffdiff and shown as a heatmap for two selected GOs (positive regulation of cell cycle and DNA repair) [45].

#### **5. Conclusions**

In conclusion, our study describes a new mechanism regarding the regulation of the transcription of functionally linked genes that are significant for cancer cell physiology. Poly-ADP-ribosylation emerges as a chromatin remodeler that is capable of defining the activity of BRG1–EP300 complexes at the promoters of genes encoding cell cycle and DNA repair-promoting proteins. Although the PARP1 inhibitor olaparib emerges as a promising tool to modulate PARP1/BRG1/EP300-dependent gene expression due to its safety and well-established in vivo effects in cancer treatment, the functional impact of DNA repair gene repression in anticancer therapies requires further investigation. In any case, our study provides a basis for the search for new combinations of iPARPs with other compounds to increase the beneficial effects of anticancer approaches.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/11/10/1539/s1, Table S1. Protein List identified in PARP1 immunoprecipitates by Mass Spectrometry. Table S2. PARP1, BRG1, and H3K27ac distribution at the CpG/E2F positive gene promoters. Table S3. Full list of gene enrichment analysis for PARP1/BRG1/H3K27ac/E2F/CpG positive promoters. Table S4. Differential expression of proliferation and DNA repair genes in DCIS, MCF7 and MDA-MB-231 cells versus normal breast tissue. Table S5. Statistical analysis of data. Figure S1. Effect of PARP1 silencing, SWI/SNF inhibition, and EP300 inhibition on gene expression. Figure S2. Chip-qPCR data for effect of iPARP on the *XRCC1* promoter.

**Author Contributions:** A.R.: conceptualization, supervision, bioinformatic analysis of DNA-bound proteins, experiments (PARP1 co-immunoprecipitation for mass spectrometry, Figures 2–4), writing; A.R.P. and C.M.S.: optimization and mass spectrometry analysis of PARP1 co-immunoprecipitates, M.S.: bioinformatic analysis of differentiation gene expression and experiments (Figures 1C and 2, Figures 3 and 4). Data shown in Figure 1A and Table S1 were collected at the Aston University in Birmingham, while all other experiments were conducted in the Department of General Biophysics, University of Lodz.

**Funding:** This research was funded by the Polish National Science Center, grant number DEC-2013/11/D/NZ2/00033" and A.R. was funded by Ministry of Science and Higher Education (776/STYP/11/2016).

**Acknowledgments:** A.R. acknowledges grants from the Polish National Science Center (DEC-2013/11/D/NZ2/00033) and Ministry of Science and Higher Education (776/STYP/11/2016); The visit of A.R. to the School of Life & Health Sciences at Aston University was funded by a FEBS Short-Term Fellowship.

**Conflicts of Interest:** Authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Cytotoxicity and Di**ff**erentiating E**ff**ect of the Poly(ADP-Ribose) Polymerase Inhibitor Olaparib in Myelodysplastic Syndromes**

**Isabella Faraoni 1,\*, Maria Irno Consalvo 2, Francesca Aloisio 1, Emiliano Fabiani 2, Manuela Giansanti 1, Francesca Di Cristino 1, Giulia Falconi 2, Lucio Tentori 1, Ambra Di Veroli 2, Paola Curzi 3, Luca Maurillo 3, Pasquale Niscola 4, Francesco Lo-Coco 2, Grazia Graziani 1,**† **and Maria Teresa Voso 2,\*,**†


Received: 10 August 2019; Accepted: 10 September 2019; Published: 16 September 2019

**Abstract:** Myelodysplastic syndromes (MDS) are highly heterogeneous myeloid diseases, characterized by frequent genetic/chromosomal aberrations. Olaparib is a potent, orally bioavailable poly(ADP-ribose) polymerase 1 (PARP1) inhibitor with acceptable toxicity profile, designed as targeted therapy for DNA repair defective tumors. Here, we investigated olaparib activity in primary cultures of bone marrow mononuclear cells collected from patients with MDS (*n* = 28). A single treatment with olaparib induced cytotoxic effects in most samples, with median IC50 of 5.4 μM (2.0–24.8 μM), lower than plasma peak concentration reached in vivo. In addition, olaparib induced DNA damage as shown by a high proportion of γH2AX positive cells in samples with low IC50s. Olaparib preferentially killed myeloid cells causing a significant reduction of blasts and promyelocytes, paralleled by an increase in metamyelocytes and mature granulocytes while sparing lymphocytes that are not part of the MDS clone. Consistently, flow cytometry analysis revealed a decrease of CD117+/CD123+ immature progenitors (*p* < 0.001) and induction of CD11b+/CD16+ (*p* < 0.001) and CD10+/CD15+ (*p* < 0.01) neutrophils. Morphological and immunophenotypic changes were associated with a dose-dependent increase of PU.1 and CEBPA transcription factors, which are drivers of granulocytic and monocytic differentiation. Moreover, the combination of olaparib with decitabine resulted in augmented cytotoxic and differentiating effects. Our data suggest that olaparib may have therapeutic potential in MDS patients.

**Keywords:** MDS; PARP inhibitors; olaparib; hematopoietic differentiation; PARP1; AML

#### **1. Introduction**

Myelodysplastic syndromes (MDS) are a group of highly heterogeneous diseases characterized by peripheral blood cytopenia, dysplasia in one or more hematopoietic cell lineages and a differentiation defect, with an increased risk of evolving to acute myeloid leukemia (AML) [1]. To date, only few treatment options are available in these diseases, including growth factors in lower-risk MDS and the hypomethylating agents decitabine and azacitidine in intermediate/high-risk MDS. Both agents delay progression to AML by exerting cytotoxic and differentiating effects. Moreover, azacitidine has been shown to prolong overall survival [2]. However, especially in patients with higher-risk MDS,

ineligible for allogeneic stem cell transplantation, life expectancy remains dismal. Therefore, innovative treatment strategies are warranted for elderly and unfit patients.

Poly(ADP-ribose) polymerase inhibitors (PARPi) are compounds with a favorable tolerability profile, designed as targeted therapy for homologous recombination (HR)-defective tumors. The FDA and EMA have approved some PARPi as maintenance therapy for platinum-sensitive relapsed (olaparib, rucaparib, and niraparib) or newly diagnosed (olaparib) high-grade epithelial ovarian, fallopian tube or primary peritoneal cancers. Olaparib and rucaparib received market authorization also for the treatment of relapsed BRCA1 or BRCA2 mutated advanced ovarian cancer, after previous lines of chemotherapy and independently of platinum sensitivity [3]. Recently, olaparib and talazoparib were FDA approved for patients with epidermal growth factor receptor 2 (HER2)-negative metastatic breast cancer with BRCA mutations, relapsing after previous chemotherapy. These and other PARPi are also under clinical development as monotherapy or in combination with targeted agents or chemotherapy for different types of cancer. In fact, accumulating evidence suggests a potential therapeutic role for PARPi in tumors characterized by mutations in genes involved in the HR repair of DNA double strand breaks (DSBs), including RAD51, ATM, ATR, CHEK1/2, FANC family members [4,5] (reviewed by Gadducci and Guerrieri [6]), or HR upstream modulators as PTEN [7] or IDH1 and IDH2 [8] (reviewed by Nickoloff et al. [9]). Therefore, regardless of BRCA mutational status, other defects in DNA repair may induce a "BRCAness" phenotype that would increase tumor sensitivity to PARPi.

We and others recently demonstrated that olaparib and other PARPi exert anti-proliferative and proapoptotic effects in human AML blasts at clinically achievable concentrations [10–18] that do not affect the viability of normal bone marrow (BM) stem cells [10,12,18]. The low BRCA1/2 RNA and protein expression levels detected in AML blasts [10] and myeloproliferative neoplasms [19] might account for myeloid malignancies sensitivity to synthetic lethality induced by PARPi. Also, some chromosomal translocations which are recurrent in leukemia including t(8;21) (RUNX1-RUNX1T1), t(15;17) (PML-RARA) [12] and t(17;19) (TCF3-HLF) [14] can weaken the HR repair activity and sensitize AML cells to PARPi treatment (reviewed by Faraoni et al. [20]).

Besides its involvement in DNA repair, through post-translational protein modifications (PARylation), the founding member of the PARP family PARP1 has a crucial role in the regulation of chromatin structure and gene transcription [21–23] (reviewed by Hottiger [24]). Recently, we suggested that olaparib cytotoxicity on primary AML blasts is the result of drug-induced DNA damage as well as of alternative mechanisms, such as upregulation of death receptor RNAs and proteins (FAS, DR5 and DR4), which in turn requires NF-κB activation [16]. The involvement of PARP1 in modulation of gene expression suggests additional therapeutic implications of PARPi in hematological diseases. In particular, deregulated gene expression in hematopoietic stem cells is regarded as a hallmark of MDS, with involvement of different pathways depending on the disease stage [25]. Interestingly, higher-risk MDS are characterized by a decrease of apoptosis and high levels of genomic instability, likely as a consequence of the observed alterations in DNA damage response pathways [25,26].

On this basis, we investigated the effects of PARP1 inhibition in primary cultures of BM mononuclear cells (BM-MNC) collected from MDS patients. Our results indicate that olaparib is not only cytotoxic, but importantly, also stimulates differentiation of immature MDS myeloid cells, as single agent or in combination with decitabine.

#### **2. Results**

#### *2.1. MDS Mononuclear Cells Are Sensitive to Olaparib In Vitro*

To assess the sensitivity of MDS to olaparib, experiments were performed using short-term cultures of MNC freshly isolated from the BM samples of 28 MDS patients. Table 1 lists the patients' characteristics.


**Table 1.** Patient Characteristics.

MDS-SLD: MDS with single-lineage dysplasia; MDS-MLD: MDS with multiple-lineage dysplasia; MDS-EB: MDS with excess of blasts in BM, MDS-EB-1: 5 to 9% blasts, MDS-EB-2: 10 to 19% blasts. \* inv,ins(11;9); \*\* 7q-,12p-,22q-. IPSS-R: Revised International Prognostic Scoring Systems.

The proliferation rate of untreated or olaparib-treated primary cells was monitored for 7 days and the results are shown by grouping samples according to the 2016 WHO classification of MDS [1]. Some of the untreated MDS cultures showed a slight increase in cell number (*n* = 12), while the remaining samples either did not proliferate (*n* = 7) or showed a reduction in cellularity (*n* = 9) (Figure 1, left panels). Nevertheless, viability was constantly ≥70% in control cells during the culture. No significant correlation was found between the cell growth in culture and MDS risk according to the Revised International Prognostic Scoring Systems (R-IPSS). Treatment with olaparib induced a dose-dependent decrease of cell survival in all MDS samples with a median IC50 of 5.5 μM (range 2.0–24.8 μM) (Figure 1, right panels). The median IC50 values were comparable in the four MDS subgroups analyzed (MDS-SLD, 6.1 μM; MDS-MLD, 5.4 μM; MDS-EB-1, 5.3 μM; MDS-EB-2, 3.8 μM). No statistically significant correlation was detected between the proliferation indexes of MDS cultures and olaparib IC50s, suggesting that the drug sensitivity did not depend on cell ability to grow in vitro. Representative growth curves of olaparib-treated samples with comparable olaparib IC50 values but different proliferation rates are shown in Figure S1A. Also, we found no significant correlation between cell sensitivity, expressed as olaparib IC50 values at 7 days, and the MDS prognostic variables listed in Table 1. Notably, the olaparib IC50 values were in most cases largely below the steady-state plasma peak concentrations (Cmax = 16–22 μM), measurable in patients with solid tumors receiving 300 mg olaparib twice daily [27,28]. Conversely, olaparib sensitivity of normal BM, CD34-enriched mobilized peripheral blood and purified CD34+ samples (IC50 range: 18.5–27.0 μM) was markedly higher than that of MDS cells (Figure S1B), in agreement with previous findings [5,10,12,29].

**Figure 1.** Olaparib exerts cytotoxic effects in primary MDS cultures. BM-MNC collected from MDS patients were cultured with IL-3, SCF and FLT3LG and treated (time 0) with increasing concentrations of olaparib. For each primary culture, cell proliferation was evaluated by counting viable cells using trypan blue exclusion at 3, 5 and 7 days. Standard deviation (SD) of four replicate counts was ≤20% and is not shown in the figure. MDS samples were grouped according to morphology, as MDS-SLD, MDS-MLD, MDS-EB-1 and MDS-EB-2. Left graphs represent the proliferation pattern of untreated primary MDS cells during 7 days of culture. Right graphs show the surviving fractions after 7 days of treatment, and the olaparib IC50s for each sample calculated with respect to untreated cells cultured for the same time period.

We then investigated whether the growth inhibitory activity of olaparib in MDS cells was associated with cytotoxic effects. Apoptosis was evaluated by cell staining with annexin V/PI and FACS analysis after 7 days exposure to graded concentrations of olaparib. Cells were gated in order to separately analyze apoptosis induction within the myeloid and lymphocyte populations. Dose-dependent apoptosis was observed in the myeloid compartment of MDS samples characterized by IC50 values ≤ 6.1 μM (Figure 2A), without major differences among cells at different maturation stages. Pooled statistical analysis of data referring to samples with olaparib IC50 values ≤ 6.1 μM indicated a significant increase in the percentage of apoptotic cells at 5–10 μM olaparib concentrations (Figure 2B). On the other hand, a negligible percentage of apoptotic cells was detected in the lymphocyte population (Figure 2B). Representative plots in Figure S2A demonstrate the lack of annexin V/PI staining in CD45-positive and CD33-negative lymphocytes. Lack of apoptosis detection in the lymphocyte population was not due to a faster killing kinetics in lymphoid cells, since apoptosis induction was not observed at an earlier time point (i.e., 3 days) (Figure S2B). These data suggest that olaparib preferentially kills myeloid precursors, but spares lymphocytes that are not part of the MDS clone. To further defining the targets of olaparib cytotoxic effects, two MDS samples with cytogenetic abnormalities were FISH analyzed after 7 days exposure to olaparib. The PARPi induced a 22% and 34% decrease in the number of cells with trisomy 8 and 7q deletion, respectively. Figure 2C shows representative FISH images of MDS cells with deleted chromosome 7. These data suggest a preferential killing of malignant cells by olaparib.

**Figure 2.** *Cont.*

**Figure 2.** Apoptosis and DNA damage in primary MDS in vitro cultures treated with olaparib. (**A**) Percentages of apoptotic cells (Annexin V positive/PI negative), detected by FACS analysis, in gated myeloid cells after exposure to graded concentrations of olaparib (7 days) compared to untreated cultured cells. Samples were classified in order of decreasing olaparib IC50 values. (**B**) Histograms indicate the mean percentage ±SD of apoptotic and necrotic cells (Annexin V positive/PI positive), detected within the myeloid and lymphocytic populations, respectively. Data refer to six MDS samples after exposure to graded concentrations of olaparib for 7 days compared to untreated cultured cells. Olaparib induces apoptosis in the myeloid MDS cell compartment, while sparing lymphocytes. Statistical analysis was performed on apoptotic cells (Annexin V positive/PI negative cells) by Friedman followed by Dunn's test. (**C**) Interphase FISH analysis of representative MDS cells with 7q deletion (ID-75). Images were taken at 100× magnification. Nuclei with deletions involving both critical regions at 7q22 and 7q36 show one red and one green signal only, whereas cytogenetically normal cells show two red and two green signals. In the control group, 75% of cells showed the 7q deletion, whereas in the olaparib-treated group this percentage dropped to 50%. (**D**) Analysis of γH2AX foci by immunofluorescence after 72 h of exposure to 0 and 10 μM olaparib. Values are the mean percentage ± SD of the cells with ≥5 γH2AX foci in four quadrants, containing at least 50 cells each. Statistical analysis was performed by Mann-Whitney test. Representative nuclei of MDS cells untreated or exposed to 10 μM olaparib are shown. Cells were probed with γH2AX (green stain) and DAPI (blue). Images were taken at 40× magnification. \* *p* <0.05, \*\* *p* <0.01, \*\*\* *p* <0.001 vs. untreated control.

We then assessed whether the cytotoxic effects of olaparib in MDS might be attributable to DNA damage, using immunofluorescence analysis of γH2AX foci, which are markers of DSBs. The histogram in Figure 2D represents the percentage of γH2AX positive cells in five MDS samples untreated or treated with 10 μM olaparib for 72 h. The results indicate that the PARPi induced a significantly higher percentage of γH2AX positive cells in samples characterized by low IC50s (ID-21, ID-42 and ID-41), as compared to the untreated control. Conversely, a negligible increase in DNA damage was observed in two other MDS samples (ID-43 and ID-61) that were more resistant to the PARPi. These data indicate that treatment with single agent olaparib induces DNA damage that accounts, at least in part, for olaparib cytotoxicity.

#### *2.2. Di*ff*erentiating E*ff*ects of Olaparib on MDS Immature Myeloid Cells*

Flow cytometry analysis by side scatter (SSC) vs forward scatter (FSC) and SSC vs CD45 showed an increase in the mature cells (high-SSC) after treatment with olaparib (Figure S2A, left and middle panels), suggesting a potential PARPi effect on myeloid cell differentiation. Based on these findings, we studied morphological changes in MDS samples upon treatment with 10 μM olaparib for 7 days, after May Grunwald-Giemsa staining of cytospins. We observed a significant reduction of the proportion of blasts and promyelocytes, paralleled by an increase of metamyelocytes and mature granulocytes (Figure 3A,B). In cells from one case of MDS-EB-1, olaparib treatment induced outgrowth of 22% monolobulated megakaryocytes, as depicted in Figure 3C.

Differentiation was also analyzed by means of CD117 and CD123 antigen expression, which are markers of myeloid progenitors. The percentage of CD117+/CD123+ immature progenitors was significantly reduced in treated cells, as compared to untreated controls in 12 of 13 MDS samples (Figure 4A). Representative dot-plots of the olaparib-induced decrease in CD117+/CD123+ cell population are shown (Figure 4A, right panels). The lack of CD34 expression is known to be a common aberrancy in MDS immature myeloid compartment [30]. Indeed, the percentages of CD34+ stem cells were scored ≥1% in four of 12 samples only. Treatment with olaparib resulted in a marked reduction of CD34+ cells in three out of four samples, with no effects observed in the ID-02 sample that was the most resistant to the toxic effects of olaparib (Figure S3). Looking at neutrophil maturation patterns, samples treated with olaparib showed a significant induction of CD11b+/CD16+ cells (9 of 10) and CD10+/CD15+ (6 of 6) double positive cells, indicating differentiation towards mature neutrophils (Figure 4B). Representative dot-plots show a dose-dependent increase of CD45-positive/high SSC cells in the samples exposed to the PARPi. This effect was accompanied by a shift of the double-negative immature cells towards the double-positive mature population (Figure 4B).

In six MDS samples, we also tested whether modulation of differentiation markers was dose-dependent. Olaparib induced a reduction of CD117+/CD123+ MDS cells together with an increase of mature CD11b+/CD16+ cells that reached statistically significance at the concentration of 10 μM (Figure 4C). These data strongly suggest that inhibition of PARP1 can stimulate maturation of MDS myeloid precursors into neutrophils. A similar differentiating effect was also observed in one sample from a patient with low-blast count AML (Figure S4), which was also highly susceptible to olaparib cytotoxic effects (IC50 = 1.6 μM).

**Figure 3.** Morphological analysis of MDS samples treated with olaparib. May Grunwald-Giemsa staining of cytospins from six primary MDS samples after 7 days of culture. Untreated cultured cells and those treated with 10 μM olaparib are shown. (**A**) Representative pictures are shown (sample ID-54). Images were taken at 100× magnification. (**B**) Differential counts using standard microscopy were performed by two specialized hematologists (ADV and MTV). Statistical analysis by paired Wilcoxon test: \* *p* < 0.05. (**C**) Sample ID-79 after 7 days of 10 μM olaparib treatment. Images were taken at 100× magnification.

**Figure 4.** Immunophenotypic analysis of myeloid cells in MDS samples treated with olaparib. (**A**) Flow cytometry analysis of MDS samples treated with olaparib for 7 days vs untreated control cells cultured for the same time period. Box-whisker plots show the proportion of CD117+/CD123+ immature progenitors in MDS samples (*n* = 13; paired Wilcoxon test: \*\*\* *p* <0.001). Representative plots show the changes in the expression of CD117+/CD123+ markers (red) in sample ID-76 after treatment with 5 or 10 μM olaparib. (**B**) Box-whisker plots show the proportion of CD11b+/CD16+ (*n* = 10) and CD10+/CD15+ (*n* = 6) mature neutrophils in MDS samples (paired Wilcoxon test: \*\* *p* < 0.01, \*\*\* *p* < 0.001). Representative plots of flow cytometry analysis of CD45+ cells, after lymphocyte (black) exclusion, showing changes in SSC and increased expression of CD11b+/CD16+ or CD10+/CD15+ cells in ID-76 sample after treatment with 5 or 10 μM olaparib. (**C**) Box-whisker plots showing the olaparib dose-dependent effect in CD117+/CD123+ (left panel) and CD11b+/CD16+ (right panel) cells from six samples treated with 5 or 10 μM olaparib. Statistical analysis by Friedman followed by Dunn's test: \* *p* < 0.05, \*\* *p* < 0.01.

#### *2.3. Olaparib Modulates the Expression of PU.1 and CEBPA Transcription Factors*

To confirm the ability of olaparib to stimulate myeloid differentiation, we studied the expression of the lineage-specific transcription factors PU.1 and CEBPA, drivers of granulocytic and monocytic differentiation. BM-MNC from six MDS patients were exposed to olaparib (5 and 10 μM), and total RNA analyzed by qRT-PCR after 7 days of treatment. We found that olaparib induced a statistically significant and dose-dependent upregulation of PU.1 RNA expression, while CEBPA was significantly upregulated at the 10 μM concentration (Figure 5A). This was associated with a dose-dependent increase of the corresponding proteins, as assessed by western blot analysis (Figure 5B).

**Figure 5.** Olaparib modulates the expression of transcription factors associated with myeloid differentiation. (**A**) PU.1 and CEBPA mRNA expression were evaluated by qRT-PCR after 7 days of culture and treatment with 5 or 10 μM olaparib. Histograms show the mean ±SD relative expression in six MDS samples. Statistical analysis by Kruskal-Wallis followed by Dunn's test: \* *p* < 0.05, \*\*\* *p* < 0.001. (**B**) Representative immunoblot showing PU.1 and CEBPA protein expression from one MDS sample exposed to the indicated concentrations of olaparib (i.e., ID-54) for 7 days. Numbers below blots refer to the densitometric analysis of the immunoreactive bands and represent the fold-increase of protein expression normalized to GAPDH.

#### *2.4. Cytotoxic and Di*ff*erentiating E*ff*ects of Olaparib and Decitabine Combined Treatment*

To investigate the possible synergism between olaparib and decitabine, we initially used the OCI-AML2 leukemia cell line as a model. OCI-AML2 cells were treated with increasing concentrations of both drugs as single agents or in combination. After 3 days of culture, cells were analyzed by the MTS assay. Olaparib (1 to 8 μM) and decitabine (0.1 to 0.8 μM) exerted synergistic cytotoxicity on OCI-AML2 cells as assessed by the CompuSyn model [31] at all concentration ratios (Figure S5).

To assess the susceptibility to decitabine and olaparib combination (1:10), four MDS samples were exposed to equitoxic drug concentrations, including drug IC50 values and cytotoxicity was assayed by MTS after 7 days of treatment. In all samples, the two drugs exhibited synergistic cytotoxic effects at most of the concentrations tested (Figure 6A). In particular, for all samples the combination index (CI) values were <1 at the effective dose 50 (ED50). Moreover, analysis of the dose reduction index (DRI) indicated that addition of the PARPi to decitabine allowed 3.5–20.8 fold decrease of decitabine dose at 0.5 fraction affected (Fa). The highest DRI value was observed in ID-90 cells that exhibited the lowest sensitivity to decitabine. Cytotoxicity was associated with DNA damage, as indicated by the statistically significant increase of γH2AX positive cells in co-treated samples as compared to the untreated controls (Figure 6B).

To investigate whether the synergistic activity of decitabine plus olaparib did also impact on myeloid differentiation, we evaluated by FACS analysis the percentage of CD117+/CD123+ hematopoietic progenitors and CD11b+/CD16+ neutrophils in four MDS samples, untreated or treated with the indicated concentrations of the drugs. We found that the 0.5 μM decitabine and 5 μM olaparib combination significantly decreased the proportion of immature progenitors and increased the percentage of neutrophils (Figure 6C). The expression of PU.1 and CEBPA RNA was studied in three MDS samples and at least additive effects were observed in samples treated with the drug combination (Figure 6D).

**Figure 6.** The olaparib and decitabine combination impairs survival and induces differentiation in primary MDS. (**A**) Primary MDS cells were treated with olaparib (ola, 0–10 μM) or decitabine (dac, 0–1 μM) or with different drug combinations at time 0. After 7 days of culture, cells were analyzed by the MTS assay to study cytotoxicity. Plots show the fraction of affected cells (Fa, *x* axis) and the combination index (CI, *y* axis) values as assessed by the CompuSyn model. Each plot indicates the CI values of a MDS samples treated at decitabine and olaparib 1:10 ratio. CI < 1, synergistic activity; CI = 1, additive effect; CI > 1, antagonism. (**B**) Analysis of γH2AX foci by immunofluorescence after 72 h of exposure of ID-90 cells to 5 μM olaparib and 0.5 μM decitabine, as single agents or in combination. Values are the mean percentage ±SD of the cells with ≥5 γH2AX foci in four quadrants, containing at least 50 cells each. Statistical analysis was performed by Kruskal-Wallis followed by Dunn's test. Representative nuclei of untreated or drug-treated MDS cells are shown. Cells were probed with γH2AX (green stain) and DAPI (blue). Images were taken at 40× magnification. (**C**) Histograms show the mean percentage ± SD of CD117+/CD123+ and CD11b+/CD16+ cells in 3 MDS samples untreated or treated with 5 μM olaparib and 0.5 μM decitabine, as single agents or in combination. Statistical analysis by Friedman followed by Dunn's test: \* *p* < 0.05, \*\* *p* < 0.01. Representative dot plots showing data from a MDS sample (ID-88) untreated or treated with olaparib, decitabine or both drugs. A decrease of CD117+/CD123+ immature progenitors (upper plots) and increase of CD11b+/CD16+ (lower plots) are observed in treated samples as compared to the untreated control. (**D**) PU.1 and CEBPA mRNA expression evaluated by qRT-PCR after 7 days of treatment with olaparib and decitabine, as single agents or in combination. Histograms show the mean ±SD relative expression in three MDS samples. Statistical analysis by Kruskal-Wallis followed by Dunn's test: \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

#### **3. Discussion**

In the present study, we demonstrate for the first time that the PARPi olaparib used at clinically achievable concentrations [27,28] exerts cytotoxic and differentiating effects in short-term primary cultures of BM-MNC obtained from MDS patients.

The apoptotic effects of olaparib were observed both in actively proliferating and resting MDS cultures. These findings support the activity of PARPi also on quiescent malignant stem cells [10,15,16], that are generally more resistant to cytotoxic treatments. As recently suggested, the inefficient HR repair of DSBs in non-proliferating cells characterized by a "BRCAness" phenotype, does not account for the observed PARPi-mediated synthetic lethality, suggesting the involvement of additional repair mechanisms [15].

Interestingly, in our model, olaparib mainly targeted myeloid cells including the dysplastic cell lineages, while it spared lymphocytes, a cell subset that is not involved in the MDS clone. Consistently, normal hematopoietic cells showed lower sensitivity to olaparib as compared to MDS cells. This may indicate that PARPi-induced synthetic lethality targets MDS cells that are frequently characterized by DNA damage response defects which contribute to chromosome alterations involved in disease progression [25,32].

In our study, olaparib induced MDS cell differentiation, as indicated by the results of morphological and immunophenotypic analysis, and increased the expression of myeloid-specific transcription factors. In fact, the PARPi caused a reduction of CD117+ or CD34+ hematopoietic progenitors and an increase in maturing CD11b+/CD16+ and CD10+/CD15+ cells in almost all MDS analyzed samples. Since BM samples isolated from patients contain a mixture of normal and diseased cells with differentiation block, the observed increase in neutrophils might derive either from killing of immature MDS cells and consequent relative increase of healthy cells and/or from maturation of MDS hematopoietic progenitors. Actually, we found that olaparib-differentiated cells, as assessed by morphology, were still characterized by significant dysplastic changes, and the observed decrease of trisomy 8 and del(7)–positive cells in the two FISH-analyzed MDS samples was moderate. These data argue in favor to the hypothesis that the increase of differentiated cells in olaparib-treated cultures involves malignant cells and is not merely due to the killing of the latter and recovery of normal hematopoietic progenitors.

The regulatory effect of transcription factors determines the fate of hematopoietic stem cells from self-renewal towards generation of mature blood cells [33–35]. In our study, olaparib upregulated RNA and protein expression of the granulocytic and monocytic transcription factors PU.1 and CEBPA, both proteins reported to be frequently downregulated in myeloid malignancies [33–35]. The modulation of transcription factors involved in myeloid cell differentiation induced by olaparib might depend on the known ability of PARP1 to affect chromatin structure and RNA transcription by PARylation or direct interaction with target proteins [24,36,37]. In fact, PARP1 overexpression has been found to block differentiation in response to all-trans retinoic acid (ATRA) in leukemia cell lines [38] and a reduction of PARP1 expression was observed during differentiation of neutrophils or monocytes [39,40]. Interestingly, higher PARP1 RNA levels were observed in higher-risk MDS and correlated with inferior survival [41]. However, we cannot exclude that transcription factor modulation might be the consequence, rather than the cause, of differentiation induced by excessive DNA damage, as previously described in murine models of hematopoietic stem cells [42,43].

One clinical concern with olaparib is the risk of therapy-related MDS or AML associated to high-doses [44]. This risk has not been confirmed by a recently published phase 3 clinical trial which reported the occurrence of MDS or AML in 2% of olaparib-treated (300 mg, twice daily) vs 4% of placebo-treated patients [45]. Similar results were reported for the PARPi niraparib and rucaparib [46,47], suggesting that the burden of cytotoxic treatment prior to PARPi may be responsible for the development of therapy-related myeloid neoplasms.

In our study, the cytotoxic effects of the olaparib/decitabine combination were synergistic or at least additive in the MDS cell cultures. These findings are in line with the recently reported synthetic lethality induced by the association of the two drugs in a panel of AML cell lines [48,49]. Muvarak et al. demonstrated that low dose decitabine or azacitidine plus the PARPi talazoparib had synergistic cytotoxic effects in AML cell lines, as a consequence of increased DNA damage and delayed DNA repair [13]. In keeping with these results, the combination of the PARPi with a hypomethylating agent caused a marked antitumor response in an AML xenograft model [13]. The synergistic activity of PARPi with decitabine may also be explained by the involvement of PARP1 in the repair of decitabine-induced DNA lesions represented by randomly incorporated 5-aza-dC and trapped DNA methyltransferase 1 (DNMT1). In the presence of PARPi, DNA repair is impaired as shown by an increase of DBSs [13,48,49]. Noteworthy, the combination of olaparib and decitabine significantly decreased the proportion of immature myeloid progenitors and increased that of differentiated neutrophils. This effect was paralleled by the up-regulation of PU.1 and CEBPA transcription factors. The inhibition of DNA methylation caused by decitabine likely contributes to enhance the differentiating effect induced by olaparib. These results are of particular interest, considering that decitabine is approved by FDA for all subtypes of MDS and low-blast count AML (blasts < 30%), and by EMA for AML patients aged ≥ 65 years not candidates for standard induction chemotherapy, due to its manageable toxicity [50]. Moreover, in a clinical study with ovarian cancer patients there were no differences in the olaparib toxicity between younger and older women [51]. Thus, the favorable safety profile and good oral bioavailability of olaparib makes it a suitable agent to be combined with decitabine for MDS patients ineligible for chemotherapy or BM transplantation. This novel combination may allow dose reduction and overcome some of the limits of hypomethylating agents, by deepening the response and decreasing the burden of treatment-induced cytopenias with the related complications, ultimately leading to improved survival in MDS.

#### **4. Materials and Methods**

#### *4.1. MDS Samples*

Freshly isolated primary cells were obtained from BM aspirates of 28 adults with newly diagnosed MDS and one patient with low-blast count AML. Normal BM cells were collected from one hematopoietic stem cell donor and normal CD34+ enriched cells were obtained from the peripheral blood (PB) of three healthy donors following G-CSF-mobilization. MNC were isolated by Lynpholyte-H (Cederlane, Burlington, Canada). In the case of one CD34-enriched mobilized peripheral blood sample, CD34+ cells were purified by the MACS CD34 progenitor isolation kit using immunomagnetic beads (Miltenyi Biotech, Bergisch Gladabach, Germany). All patients and donors gave written informed consent according to institutional guidelines and the study was approved by the local institutional review board (Tor Vergata Hospital, Registry N. 12/16). Routine morphological, immunophenotypic and genetic analysis were carried out at presentation. Conventional karyotyping was performed on BM diagnostic aspirates after short-term culture and analyzed after G-banding.

#### *4.2. Cell Culture and Drug Treatment*

Freshly isolated BM-MNC were cultured at 37 ◦C in a humidified atmosphere of 5% CO2 for 1 or 2 days before starting the chemosensitivity assays. Briefly, cells were seeded into culture flasks in RPMI medium (Sigma-Aldrich, St. Louis, MO, USA), supplemented with 2 mM L-glutamine (EuroClone, Pero, Milan, Italy), 1% penicillin/streptomycin (Euroclone), 20% fetal bovine serum (FBS) (Gibco, ThermoFisher Scientific, Waltham, MA, USA) and 10 ng/mL of each IL-3, SCF and FLT3LG (PeproTech, Rocky Hill, NJ, USA), which do not have differentiating effects. The OCI-AML2 leukemia cell line (DSMZ-German Collection, Braunschweig, Germany) was cultured in RPMI plus 20% FBS.

Olaparib stock (AstraZeneca, London, UK) was prepared by dissolving 10 mg in 200 μL of DMSO (Sigma-Aldrich) and then diluted with RPMI to the concentration of 2 mM. The final DMSO concentration in the cultures was ≤0.01% (*v*/*v*). The stock solution of decitabine (Cayman Chemical, Ann Arbor, MI, USA) at 2 mM concentration was prepared by dissolving the drug in PBS. Drugs were added at the beginning of the experiments and left in the medium during the entire period of

incubation. Since in vitro culture of primary MDS samples could be maintained for 7–10 days only, the assays were performed at up to 7 days after drug exposure. Experiments with normal cells were performed with the same procedure.

For survival assay, cells were seeded in 48 or 24 wells culture plates (10<sup>6</sup> cells/mL, unless otherwise specified) in duplicate and treated with graded concentrations of olaparib (1–10 μM) or decitabine (0.1–1 μM) for the indicated times. Cells were counted by trypan blue dye exclusion in quadruplicate, and the survival fraction was calculated as proportion of the untreated control. The drug concentration capable of inhibiting 50% of cell growth (IC50), compared to untreated control, was calculated with the GraphPad Prism 5 software (San Diego, CA, USA) by using linear regression. The combined effect of olaparib and decitabine was evaluated by the MTS viability test (Promega, Madison, WI, USA), according to the manufacturer's instructions. Cells were seeded in triplicate (10<sup>4</sup> cells/mL for OCI-AML2 and 5 <sup>×</sup> 10<sup>5</sup> cells/mL for the MDS samples), in 96 wells plates (2 <sup>×</sup> 103 cells/well and 10<sup>5</sup> cells/well, respectively) and treated with graded concentrations of olaparib and decitabine. Synergism was assessed by calculating the proportion of cell growth using the CompuSyn software (ComboSyn Incorporated, Paramus, NJ, USA).

#### *4.3. Flow Cytometry and Apoptosis Assays*

Apoptosis in primary MDS samples was assayed using an annexin-V apoptosis kit (GFP Certified™Apoptosis/Necrosis Detection Kit, Enzo Life Sciences, Farmingdale, NY, USA), according to the manufacturer's instructions, and analyzed by flow cytometry.

In MDS patients, BM-MNC cells are characterized by a decrease of neutrophil granularity and increased or decreased expression of differentiation antigens or their ratios, although no single specific marker is diagnostic for MDS [30]. Thus, we performed a series of analyses with different antibody combinations to evaluate the maturation pattern induced by olaparib. For immunophenotype analysis, cells were incubated with fluorochrome-tagged monoclonal antibodies anti-CD45 (# 560777), -CD33 (561157), -CD34 (#345804), -CD117 (#339217), -CD123 (#588714), -CD10 (#332776), -CD11b (#333142), -CD15 (#332778), -CD16 (#335035) (BD Biosciences). In particular, we studied the neutrophilic population by analyzing CD11b/CD16 and CD10/CD15 differentiation markers. Samples (5 <sup>×</sup> 104 cells) were acquired on a multicolor BD FACSCanto II flow cytometer and evaluated using DIVA and FlowJo softwares (BD Biosciences, San Jose, CA, USA).

#### *4.4. Immunofluorescence, May-Grunwald Giemsa Staining and FISH Analysis*

MDS cells were cytocentrifuged (10<sup>5</sup> cells), fixed with 4% paraformaldehyde, permeabilized in 0.3% tryton and blocked in 2% bovine serum albumin (Sigma-Aldrich). Slides were incubated with mouse anti-phospho-H2A histone family member X (γH2AX) (#05-636, Millipore, Burlington, MA, USA) followed by goat anti-mouse IgG DyLight 488 (#35502, Thermo Fisher Scientific, Rockford, IL, USA), stained with DAPI (4,6 diamidino-2-phenylindole) and mounted in Fluoromount (Sigma-Aldrich). Analysis was performed using a Leica CTR 6000 fluorescence microscope and LAS AF Lite software (Leica, Wetzlar, Germany).

After 7-day culture, cytospins containing 105 BM-MNC were stained using the May-Grunwald solution. Cells were then observed under a BX61 microscope (Olympus, Tokyo, Japan). and differential counts were performed by two experienced hematologists (A.D.V. and M.T.V.).

Fluorescence in situ hybridization (FISH) was performed on MDS cells cultured for 7 days in the presence of olaparib. Cytogenetic preparations were performed according to the Kreatech FISH probes manufacturer's instructions (Leica). Slide preparations were hybridized with the Kreatech FISH probes: KBI-2008G, targeting the centromere of chromosome 8, and KBI-10202 targeting chromosome 7q22/7q36 (Leica). A minimum of 200 cells were scored for each preparation by two experienced operators (P.C. and M.G.). Evaluation of the FISH signals was performed using the BX61 Olympus fluorescence microscope.

#### *4.5. mRNA and Protein Expression*

Total RNA was isolated by Trizol reagent (Invitrogen, Carlsbad, CA, USA). One μg of total RNA was reverse-transcribed (reagents from Applied Biosystem, Warrington, UK) and quantitative real-time PCR (qRT-PCR) was carried out using iQ SYBR Green Supermix, Bio-Rad (BioRad, Hercules, CA, USA). The sequence of the primers used is as follows: PU.1, FW 5 -CAGGGGATCTGACCGACTC-3 and RV 5 -GCACCAGGTCTTCTGATGG-3 ; CEBPA, FW 5 -TTGTTTGTACTGTATGCCTTC-3 and RV 5 -GCCAGATACAAGTGTTGATAT-3 ; GAPDH, FW 5 -CAGCCGAGCCACATCG-3 and RV5 -TGAGGCTGTTGTCATACTTCTC-3 . A melting curve (62–95 ◦C) was generated at the end of each run to verify specificity of the reactions. Analysis was performed in triplicate on an ABI-7300 instrument (Applied Biosystems). The 2-ΔΔCt relative quantification method was used to calculate mRNA expression and RNA from untreated cells was used as calibrator. Total proteins were extracted from MDS cells as previously described [16] Immunoblot analysis was performed using anti-CEBPA (sc-166258, Santa Cruz Biotechnology, CA, USA), anti-PU.1 (PA5-17505, Thermo Fisher Scientific), and anti-GAPDH (#2118, Cell Signaling Technology, Danvers, MA, USA) antibodies. Horseradish peroxidase-conjugated IgGs were used as secondary antibodies. The autoradiograms were analyzed by densitometric analysis by ImageJ 1.45s software (National Institutes of Health, Bethesda, MD, USA).

#### *4.6. Statistical Analysis*

Statistical analysis was performed using the GraphPad Prism 5 software (GraphPad Software, San Diego, CA, USA) and data were reported as mean ± SD (unless otherwise specified). Correlations between olaparib IC50 values and proliferation indexes of MDS samples or the prognostic indexes listed in Table 1 were examined using the non-parametric Spearman's rank test. Differences between two groups were analyzed by the non-parametric Mann-Whitney and Wilcoxon (for matched samples) tests. For multiple comparisons, the non-parametric Kruskal-Wallis or Friedman tests followed by Dunn's post hoc test were used, as specified in the figure legends. All statistical tests were two-sided. *p* values below 0.05 were considered statistically significant.

#### **5. Conclusions**

Overall, our data demonstrate that single-agent olaparib, at clinically achievable concentrations, induces cell death of primary MDS cells, while sparing lymphocytes, which are not part of the MDS clone. Furthermore, in this short-term culture model, olaparib favors maturation of myeloid cells towards the neutrophilic lineage, as shown by induction of myeloid-specific transcription factors and increase of the metamyelocyte and neutrophil percentages. Our preclinical data, together with the acceptable toxicity profile of olaparib in vivo, an issue particularly relevant to the elderly MDS population, support the design of clinical trials with PARPi monotherapy or in combination with hypomethylating agents for the management of MDS patients.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/11/9/1373/s1, Figure S1: Growth curves of untreated and olaparib treated primary MDS and normal progenitor cell cultures, Figure S2: Analysis of apoptosis in the myeloid and lymphocytic populations of MDS samples, Figure S3: Olaparib induces a decrease in the proportion of CD34+ cells in primary MDS samples, Figure S4: Olaparib induces myeloid differentiation in a primary low blast count AML sample, Figure S5: Olaparib and decitabine combination synergistically impairs survival in the OCI-AML2 leukemia cell line.

**Author Contributions:** I.F. designed the work, performed experiments, interpreted the data and wrote the paper; M.I.C., F.A. and E.F., M.G., F.D.C., G.F., L.T., P.C. performed experiments; A.D.V. read morphological slides; L.M., P.N., M.T.V., F.L.-C. followed MDS patients and provided clinical data on MDS sample; G.G., F.L.C. and M.T.V. planned the research strategy and wrote the paper. All authors reviewed and approved the manuscript.

**Funding:** This work was supported in part by AstraZeneca and by grants from "Associazione Italiana per la Ricerca sul Cancro" (AIRC IG N.16952 to M.T.V.; AIRC IG N.20353 to G.G. and AIRC 5 × 1000 call "Metastatic disease: the key unmet need in oncology" to MYNERVA project, #21267 http://www.progettoagimm.it). The funding sponsor had no role in the design of the study, in the collection, analyses or interpretation of data, and in the writing of the manuscript.

**Acknowledgments:** Dedicated to our generous mentor and friend Francesco Lo-Coco.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **The Nucleolus and PARP1 in Cancer Biology**

#### **Marina Engbrecht and Aswin Mangerich \***

Molecular Toxicology Group, Department of Biology, 78457 Konstanz, Germany; marina.engbrecht@uni-konstanz.de

**\*** Correspondence: aswin.mangerich@uni-konstanz.de; Tel.: +49-7531-88-4067

Received: 24 May 2020; Accepted: 1 July 2020; Published: 6 July 2020

**Abstract:** The nucleolus has been known for a long time to fulfill crucial functions in ribosome biogenesis, of which cancer cells can become addicted to in order to produce sufficient amounts of proteins for cell proliferation. Recently, the nucleolus has emerged as a central regulatory hub in many other cancer-relevant processes, including stress sensing, DNA damage response, cell cycle control, and proteostasis. This fostered the idea that nucleolar processes can be exploited in cancer therapy. Interestingly, a significant proportion of poly(ADP-ribose) polymerase 1 (PARP1) molecules are localized in the nucleolus and PARP1 also plays crucial roles in many processes that are important in cancer biology, including genome maintenance, replication, transcription, and chromatin remodeling. Furthermore, during the last years, PARP1 came into focus in oncology since it represents a promising target of pharmacological PARP inhibitors in various types of cancers. Here, we provide an overview of our current understanding on the role of PARP1 in nucleolar functions and discuss potential implications in cancer biology and therapy.

**Keywords:** nucleolus; poly(ADP-ribosyl)ation; PARP; ARTDs; cancer

#### **1. Introduction into the Biology of Nucleoli**

Nucleoli are self-organizing, membrane-less sub-compartments of the nucleus, which are formed around tandemly repeated clusters of 200 to 400 ribosomal DNA (rDNA) genes known as nucleolar organizing centers (NORs) on the short arms of the five human acrocentric chromosomes 13, 14, 15, 21, and 22 [1,2]. In humans, nucleoli are surrounded by peri-nucleolar heterochromatin (PNH) derived from DNA sequences located distal and proximal to NORs [3]. The main function of nucleoli is ribosome biogenesis, which is one of the most energy-demanding and highly controlled processes in a cell. It has been estimated that in proliferating cells ribosome biogenesis consumes up to 80% of the cellular energy [4]. Therefore, it is not surprising that ribosome biogenesis is tightly coupled to the availability of growth factors, nutrients and cellular energy supply [5]. Nucleolar size positively correlates with rRNA synthesis and cell proliferation, as dividing cells often possess large nucleoli, while downregulation of rRNA gene transcription is associated with a reduction in nucleolar size [2]. Interestingly, an increased number and size of nucleoli have historically been used as a biomarker for tumor development [6].

Ribosome biogenesis starts with RNA polymerase I (Pol I)-driven synthesis of 47S pre-ribosomal RNA (pre-rRNA), which is rapidly processed by more than 200 non-ribosomal proteins and small nucleolar RNAs (snoRNAs) to mature 18S, 5.8S and 28S rRNAs [7]. These RNAs are assembled with ribosomal proteins (RPs) and 5S rRNA to form the pre-40S and pre-60S ribosomal subunits that are subsequently exported to the nucleoplasm to produce mature 80S ribosomes. Since the 5S rRNA is synthesized by RNA polymerase III (Pol III) and transcription of RP genes is mediated by RNA polymerase II (Pol II), ribosome biogenesis requires the activities of all three cellular RNA polymerases.

Light and electron microscopy have revealed the tripartite structure of the nucleolus, which reflects the different stages of ribosome biogenesis and is dependent on ongoing rDNA transcription [3,8]. During interphase, nucleoli consist of one or a few fibrillar centers (FC), of which each is surrounded by a dense fibrillar component (DFC) [9]. The FC contains non-transcribed rDNA sequences and transcription factors, e.g., the upstream binding factor (UBF), which recognizes the rRNA gene promotor. Transcription by Pol I occurs at the interface between FCs and the DFCs. DFCs are enriched in pre-rRNA processing factors, e.g., small nucleolar ribonucleoproteins (snoRNPs) and fibrillarin, and are the nucleolar sites, where early rRNA processing takes place. Both the FCs and the DFCs are embedded in a large granular component (GC), which is associated with late rRNA processing. Apart from this, mature rRNAs assemble in the GC with RPs into pre-ribosomal subunits [9].

Nucleoli are membrane-less biomolecular condensates thought to be organized by liquid-liquid phase separation [10,11]. This is supported by the finding that nucleoli of *Xenopus laevis* germinal vesicles fuse and turn over rapidly, thereby displaying liquid-like properties [12,13]. Mechanistically, molecular interactions between low-complexity, disordered protein regions and RNA molecules appear to contribute to such phase separation processes [14]. For a long time, an unresolved question has been, how the three nucleolar components can coexist without fusing to a single liquid phase. Recently, this issue has been addressed by Feric et al. [15] showing that nucleolar sub-compartments represent distinct liquid phases that can co-exist due to differences in the biophysical properties of their components, in particular due to distinct droplet surface tensions. Current evidence supports a model in which nucleoli are formed by a combination of active recruitment processes of proteins, active transcription of rDNA and formation of the respective rRNA, as well as phase separation processes involving both protein and RNA components [14].

In humans, nucleoli undergo structural reorganization throughout the cell cycle resulting in a transient loss of the tripartite structure. As cells enter mitosis, transcription of rDNA genes pauses, and nucleoli disperse. Interestingly, some nucleolar proteins, e.g., UBF, remain associated with NORs throughout the cell cycle, thus functioning as a mitotic bookmark and binding platform for the remaining Pol I transcription machinery [10,16]. Components of the DFC and the GC, such as fibrillarin and nucleophosmin, dissociate from the nucleolus during mitosis [17]. Before nucleoli reassemble in late anaphase or early telophase, these processing factors coalesce in foci designated as prenucleolar bodies (PNBs) [18]. The reassembly of nucleoli is initiated when pre-rRNA processing factors associate with NORs. First, multiple small nucleoli are formed, which as the cell cycle progresses coalesce to larger mature nucleoli [19].

Intriguingly, over the past two decades the nucleolus has emerged as a regulatory hub for multiple nuclear functions [20]. Over 4500 proteins were identified to be localized in nucleoli, of which only 30% are primarily associated with ribosome biogenesis [21,22]. This led to the assumption that the nucleolus has non-canonical regulatory functions beyond ribosome biogenesis. At present, it is well accepted that the nucleolus is a multifunctional nuclear sub-compartment, with additional roles, e.g., in stress response, DNA damage signaling, telomere maintenance, cell cycle control, cell proliferation, and proteostasis [23]. Furthermore, links between nucleolar functions and complex physiological and pathophysiological cellular processes have been established. For instance, there is increasing knowledge that the nucleolus is implicated in the aging process. In this context, several studies have postulated that rDNA instability could be a causative factor for aging [24]. Recently, the nucleolar size was shown to inversely correlate with longevity, thereby identifying small nucleoli as a potential, visible hallmark for longevity and metabolic health [4,25]. Moreover, there is growing evidence, that the nucleolus is involved in the development of neurodegenerative diseases, such as Parkinson's disease [26]. Importantly, dysregulation of nucleolar functions has been linked to carcinogenesis, and consequently the nucleolus has emerged as a new target in cancer therapy [27–30]. This aspect will be discussed in more detail in Section 5 below. Before returning to this issue, in the next sections, we discuss how nucleoli respond to stress and DNA damage (Section 2), give a brief introduction into PARylation and PARP1 (Section 3), and review the role of PARP1 in nucleolar functions (Section 4).

#### **2. How do Nucleoli Respond to Stress and DNA Damage?**

Cells are constantly exposed to exogenous and endogenous sources of DNA damage [31]. To maintain genomic integrity, they possess a repertoire of repair proteins, which detect specific DNA lesions and initiate the appropriate repair pathway. Intriguingly, over 150 DNA repair proteins were identified in nucleoli under non-stress conditions, pointing to a role of nucleoli in DNA damage response [32].

As there is no structural barrier between the nucleolus and the surrounding nucleoplasm, proteins can, in principle, freely traffic from the nucleolus to the nucleoplasm and vice versa. The mechanisms by which proteins are retained in the nucleolus are still not fully understood. In general, it is assumed that nucleolar accumulation is a consequence of affinity interactions with RNA or nucleolar proteins. Some nucleolar proteins, e.g., nucleophosmin, exhibit a nucleolar localization sequence (NoLS) [33]. While there is no highly-conserved NoLS consensus sequence, about half of the amino acids (aa) inside NoLS are lysines and arginines, which render NoLS highly positively charged [23]. Since the nucleolus harbors high numbers of negatively charged RNA molecules, electrostatic interactions between NoLS and RNA molecules were postulated to be responsible for nucleolar accumulation of NoLS-bearing proteins. However, other nucleolar proteins lack a well-defined NoLS and are thought to be targeted to the nucleolus through protein-protein interactions with nucleolar anchored hub proteins, such as nucleophosmin or nucleolin [23].

The nucleolar accumulation of a variety of DNA repair proteins raised the question whether DNA repair proteins are merely sequestered in the nucleolus for storage reasons until they are required for their functional role in DNA repair, or if they also fulfill nucleolar-specific functions. Indeed, at least for some DNA repair proteins, including the base excision repair (BER) protein apurinic/apyrimidinic endonuclease 1 (APE1) and the RecQ helicase WRN, nucleolar-specific functions were postulated [32,34]. APE1 was not only shown to reside in the nucleolus and interact with nucleophosmin, there is also growing evidence that it is involved in quality control of the transcribed rRNA [35]. The gene encoding for the Werner protein (WRN) harbors mutations in Werner syndrome, an adult onset progeria characterized by premature aging and cancer development [36]. WRN is involved in a number of DNA repair pathways, including homologous recombination (HR), non-homologous end joining (NHEJ) and BER [37,38]. Under non-stress conditions, it localizes to transcriptionally active nucleoli, giving rise to the idea that WRN might also play a role in ribosome biogenesis [39]. Further studies supported this notion in demonstrating that WRN co-immunoprecipitates with the RNA Pol I subunit RPA40 and that in the absence of WRN, 18S and 28S RNA levels are reduced [40]. Although the findings require further elucidations, these data provide convincing evidence that bona fide DNA repair proteins might play dual roles in DNA repair on the one hand and ribosome biogenesis on the other hand.

Nucleoli are highly dynamic and undergo dramatic structural changes upon suffering various types of DNA damage. A plethora of cellular stressors, including UV and γ radiation, oxidative stress, genotoxic chemotherapeutic agents, hypoxia, as well as nutrient and growth factor deprivation can induce nucleolar re-organization or disruption [41]. Any stress-induced perturbation in ribosome biogenesis that ultimately leads to disruptions in cell homeostasis through activation of p53 or other stress signaling is referred to as nucleolar—or ribosomal stress [42]. Induction of nucleolar stress can occur at various steps of ribosome biogenesis, from Pol I transcription to pre-RNA processing, and eventually to assembly of ribosomal subunits and their release from the GC [42]. For instance, treatment with actinomycin D (ActD), which at low doses acts as a selective inhibitor of RNA Pol I by intercalating with G/C-rich rDNA, was shown to induce nucleolar segregation and the formation of so called nucleolar caps [14]. Nucleolar caps are bipartite structures that are excluded from the GC and reside at the nucleolar surface. They consist of FC and DFC components, with the DFC facing the nucleolar interior and the FC facing the nucleoplasm [3]. Interestingly, a similar reorganization of nucleolar structure was observed upon targeted induction of DSBs in rDNA repeats, which was shown to result in ataxia telangiectasia mutated (ATM)-dependent inhibition of RNA Pol I transcription [43–45]. Other studies reported that rDNA DSB-induced Pol I inhibition and nucleolar cap formation can also

be regulated by the DNA-dependent protein kinase (DNA-PK) [46,47]. The formation of nucleolar caps was proposed to render damaged rDNA accessible for repair factors that are normally excluded from nucleoli under non-stress conditions [48]. In three recent studies, the *Physarum polycephalum* homing endonuclease I-*Ppo*l, which can cleave within the 28S rRNA coding region in human cells, was used to investigate whether rDNA DSBs are repaired by HR or NHEJ [49]. Harding et al. identified NHEJ as the predominant pathway in the repair of DSBs within rDNA, since depletion of NHEJ, but not HR factors, leads to accumulation of DSBs in the nucleolus [44]. Instead, van Sluis et al. [45] demonstrated that HR is also involved in rDNA DSB repair. Interestingly, HR factors are recruited to stress-induced nucleolar caps even in G1 cells, suggesting that in rDNA, HR occurs even in the absence of sister chromatids, possibly by using other rDNA repeats as a template. Therefore, the formation of nucleolar caps could facilitate repair of rDNA via HR by concentrating high levels of homologous sequences in close proximity [50]. Interestingly and in contrast to the general notion that HR acts as an error-free repair mechanism, it was reported that the HR-mediated repair of rDNA is error-prone, leading to a reduction of rDNA repeats, thus promoting rDNA instability [51].

Ribosomal DNA sequences are especially vulnerable to DNA damage due to their repetitive nature, the unique organization of rDNA in clusters on five different chromosomes and high transcription rates [17]. These special features render the rDNA prone to unscheduled DNA recombination events and frequent formation of DNA:RNA hybrids, which are thought to favor the generation of DSBs [52,53]. If not properly repaired, rDNA damage can give rise to rDNA instability, and therefore contribute to premature onset of disease and carcinogenesis. Indeed, 50% of lung and colorectal cancers were shown to harbor rDNA gene rearrangements [54]. The pathways involved in rDNA damage repair and their regulatory mechanisms still require further investigations to gain a deeper insight into the mechanisms, which drive rDNA instability and ultimately carcinogenesis.

Perturbations in nucleolar functions are often accompanied by the release of nucleolar and ribosomal proteins into the nucleoplasm, where they are proposed to take on secondary functions in stress response. For instance, nucleophosmin was shown to shuttle between the nucleolus, the nucleoplasm, and the cytoplasm upon induction of stress [55,56]. Nucleolar release of nucleophosmin can be induced by a variety of DNA-damaging agents, including ionizing radiation (IR), cisplatin or etoposide treatment [23]. Interestingly, it was postulated that nucleophosmin takes over moonlighting functions in the nucleoplasm by participating in several DNA repair pathways, including BER and nucleotide excision repair (NER) [57]. Additionally, nucleophosmin drives p53 signaling upon genotoxic stress. Nucleophosmin constitutively interacts with the key activator of the p53 signaling pathway ARF, thus targeting it to the nucleolus [58]. Upon genotoxic stress, modification of nucleophosmin leads to the release of ARF into the nucleoplasm, where it inhibits HDM2, the E3 ligase that negatively regulates p53 [17]. Additionally, nucleophosmin regulates the HDM2-p53 pathway in an ARF-independent manner, by direct protein-protein interactions with HDM2 [59]. It is important to note that while most proteins are released from nucleoli under stress conditions, there are some notable exceptions. For example, upon stress, the promyelocytic leukemia tumor suppressor (PML) protein sequesters Mdm2 to the nucleolus by nucleolar translocation, thereby stabilizing p53 in the nucleoplasm, and promoting apoptosis [60]. Furthermore, stress-induced ubiquitination of the NF-κB subcomponent RelA leads to its translocation to the nucleolus, by which this process elicits pro-apoptotic effects [61]. Interestingly, also Hsp70 and other chaperone proteins accumulate in the nucleolus during stress conditions, suggesting to protect nucleolar proteins from aggregation [62–65]. In this regard, a recent study by Frottin et al. [63] addressed the role of liquid-liquid demixing and phase transition processes during nucleolar stress response. The authors reported that the liquid-like GC phase of the nucleolus can act as a quality control compartment by exerting transient chaperone-like activity for nuclear proteins entering the nucleolus during stress conditions [63].

Apart from proteins that have predominantly nucleolar functions, several DNA repair factors, including APE1 and WRN, have been shown to undergo DNA damage-induced nucleolar-nucleoplasmic shuttling [23,40,66]. Despite extensive research the exact molecular mechanisms, by which proteins

are released from the nucleolus remain poorly understood. In general, it is assumed that such protein dynamics are regulated by a network of PTMs and protein-protein interactions, which is likely to occur in a stress- and protein-specific manner. For instance, Karmakar et al. [67] showed that WRN is acetylated upon treatment with mitomycin C and the alkylating agent methyl methane-sulfonate (MMS), but not after UV irradiation. In addition, phosphorylation was proposed to modulate WRN's subnuclear localization [39,68]. Consistent with the idea that such DNA damage-induced protein dynamics are regulated in a stress-specific manner, we recently found that after treatment with H2O2 and the alkylating agent 2-chloroethyl ethyl sulfide (CEES), but not the topoisomerase inhibitor camptothecin (CPT), nucleolar-nucleoplasmic translocation of WRN was dependent on the PARP1 protein, yet independent of its enzymatic activity [69].

#### **3. Introduction into the Biology of PARP1 and Poly(ADP-ribosyl)ation**

Post-translational modifications (PTMs), such as phosphorylation, acetylation or ubiquitination, represent an extremely versatile and fast means of regulating the complexity and dynamics of cellular processes. In particular, ADP-ribosylation, which can be found in most eukaryotes and which has been shown to be essential during mammalian development [70], represents a highly dynamic and fully reversible PTM [70–73]. This modification is catalyzed by ADP-ribosyl transferases to which the family of poly(ADP-ribose) polymerases (PARPs) belongs. The PARP gene family comprises 17 members in humans and based on structural homology of their catalytic domain with the diphteria toxin, they are also referred to as ADP-ribosyltransferases diptheria toxin-like (ARTDs) [74]. By using nicotinamide adenine dinucleotide (NAD+) as a substrate, PARPs covalently attach ADP-ribose units to a variety of aa residues of acceptor proteins, including serines, glutamates, aspartates, lysines, and tyrosines [75–82]. While most PARPs possess mono(ADP-ribosyl)transferase activity (mono(ADP-ribosyl)ation, MARylation) or are catalytically inactive, at least four of them, i.e., PARPs 1, 2, 5a, 5b, are known to synthesize negatively charged polymers of poly(ADP-ribose) (PAR), which earned them the term 'writers' of poly(ADP-ribosyl)ation (PARylation)—(PARPs 5a and 5b are also known as tankyrases 1 and 2). The resulting PAR chains are either linear or branched and may consist of more than 200 ADP-ribose moieties [83–85]. Apart from covalent modification, certain proteins—so called 'readers' of PARylation—can non-covalently interact with PAR chains via distinct PAR binding modules, which, among others, include loosely conserved PAR binding motifs (PBM), zinc finger-type structures, macrodomains, WWE domains and OB folds [86].

PARP1—the founding member of the PARP family—was extensively studied since the discovery of PARylation in 1963 [87]. It plays pivotal roles in various cellular processes, including chromatin remodeling, replication, transcription, RNA biology, energy metabolism, immunity and inflammation, as well as cell death [88–92]. In addition, one of the most critical functions of PARP1, is its role in DNA repair and genome maintenance. Thus, PARP1 participates in several DNA repair pathways, including BER, NER, as well as the two double-strand break (DSB) repair pathways, i.e., NHEJ and HR [88,93,94]. Furthermore, PARP1 was shown to recognize unligated Okazaki fragments and to stabilize stalled replication forks, owing to promote repair during replication and replicative stress [95,96]. The importance of PARP1 in DNA repair is impressively exemplified by the fact that *Parp1*-deficient mice are hypersensitive towards DNA damaging agents, which is accompanied by increased spontaneous as well as induced genomic instability and carcinogenesis [97]. On a molecular level, PARP1 acts as a sensor of DNA damage, in particular DNA single and double strand breaks, and detects those via certain zinc finger motifs of its N-terminal DNA binding domain. Binding to DNA strand breaks induces allosteric conformational changes in the protein structure, which allow binding of NAD<sup>+</sup> to the C-terminal catalytic domain, leading to the enzymatic activation of PARP1 [71,85,98]. Indeed, PARP1 was reported to be responsible for over 90% of DNA damage-induced PARylation, whereupon cellular PAR levels dramatically increase 10- to 500-fold compared to the PAR levels under non-stress conditions [88,99,100]. Apart from DNA damage-dependent activation, PARP1 activity is regulated by other PTMs, such as acetylation, phosphorylation and SUMOylation [101–104] as well as

physical protein-protein interactions [85]. The first and main target of PARP1-dependent PARylation is PARP1 itself [105,106]. Such automodification of PARP1 provides a platform for the recruitment of downstream factors, i.e., 'readers' of PARylation, of which the DNA repair protein XRCC1 represents a prime example [107–109]. There is now considerable evidence that high-affinity, non-covalent interaction of proteins with auto-PARylated PARP1 mediates substrate specificity of PARP1 and targets these PAR-binding proteins to subsequent covalent modification by PARP1 [110–112]. In total, several hundred PARylated proteins have been identified, which are involved in diverse cellular functions, ranging from genome maintenance, DNA damage response and chromatin organization to transcription, RNA metabolism and cell cycle regulation [80,82,113].

Eventually PARP1 is released from DNA, which is assumed to occur due to steric and electrostatic repulsion of the automodified protein, yet other, more specific and so far largely unexplored, mechanisms may be conceivable as well [88,114,115]. Importantly, in many instances and in particular upon genotoxic stress, PARylation is transient, highly dynamic, and fully reversible, since after being synthesized, PAR chains are rapidly degraded in a two-step process, due to the enzymatic activities of certain 'eraser' enzymes [116,117]. Thus, the bulk of DNA damage-induced PAR can be degraded by poly(ADP-ribose) glycohydrolase (PARG), which harbors exo- as well as endo-glycosidic activities for PAR. The most proximal, protein-bound ADP-ribose moiety, however, cannot be released by PARG. This is instead carried out by several other eraser enzymes, which possess individual specificities for certain ADP-ribose acceptor sites, e.g., ARH3 can remove ADP-ribose from serine, and MacroD1/MacroD2 from glutamate or aspartate [116–121]. Moreover, the terminal ADP-ribose protein glycohydrolase 1 (TARG1/C6orf130) was shown to act on glutamate and aspartate residues, as well, by removing and releasing not only ADP-ribose, but also entire PAR chains [116,122].

In general, PARylation is thought to regulate the physicochemical properties, localization, and enzymatic activities of its target proteins in a highly controlled, spatio-temporal manner. Furthermore, PAR can serve as a signal to target proteins for degradation through the ubiquitin-proteasome system [123], which can, therefore, act as a 'kiss of death' for certain factors during DNA damage response. Another layer of complexityin the functions of PARylationis given by findings demonstrating that PAR can act as a seed for liquid-liquid demixing processes, thereby triggering the formation of biomolecular condensates [124,125]. In this regard, it has also been reported that such processes contribute to the formation of dynamic protein foci at sites of DNA damage, thereby facilitating DNA repair processes through transient and functional compartmentalization of DNA damage sites [110,124]. Moreover, there is mounting evidence that PARylation and PAR-binding regulate liquid-liquid phase separation and aggregation of several neurodegenerative disease-associated RNA-binding proteins, including α-synuclein, TDP-43 and hnRNP A1 [126]. In summary, it is assumed that PARP1 orchestrates and supports DNA damage response mechanisms and local chromatin dynamics. In addition, beyond its role in the control of protein localization and biochemistry, PARP1 is involved in the regulation of cell death and cellular energy metabolism, i.e., by using NAD<sup>+</sup> as a substrate, for which reason it proves itself as a global regulator of cellular physiology and pathophysiology [92,127].

Over the past decades, PARP1 came into focus as a target in clinical oncology, since PARP inhibitors were identified to act as chemosensitizers in combination with classical DNA-damaging therapies or as monotherapeutic agents to treat cancers with defects in HR repair according to the concept of synthetic lethality. In 2005, two independent groups discovered the synthetic lethal interaction between PARP1 inhibition and loss of BRCA, i.e., *BRCA1* or *BRCA2* [128,129], which spurred the development of clinical PARP inhibitors. By now, four of such small-molecule PARP inhibitors (i.e., olaparib, rucaparib, niraparib, and talazoparib) have been approved by authorities such as the EMA and the FDA to treat certain types of ovarian, breast or pancreatic cancer with germline loss-of-function mutations of *BRCA* genes [130,131]. Functional BRCA1 and BRCA2 are of critical importance for the repair of DSBs via HR [132]. Inhibition of PARP not only leads to accumulation of DNA single-strand breaks (SSBs), but also to trapping of PARP at the DNA, which may result in toxic manifestations of the damage and stalled replication forks [133]. Due to PARP inhibition and the absence of BRCA, the stalled

replication forks cannot be restarted properly, leading to replication fork collapse and ultimately DSBs, which are lethal for HR-deficient cancer cells [134]. By selectively targeting certain types of cancers, PARP1 inhibitors provide a successful step towards precision medicine in oncology. In addition to their use in *BRCA*-mutated cancers, there is meanwhile good evidence for further synthetic lethal interactions of PARP inhibitors in combination with other genetic constellations. Moreover, the use of PARP inhibitors as chemosensitizers represents a promising strategy for their use in cancer treatment [130–132]. Taken together, it is expected that the area of application for the use of PARP inhibitors will further expand during the next decade and that PARP inhibitors will find their place in the range of chemotherapy regimens.

#### **4. On the Role of PARP1 and PARylation in the Biology of the Nucleolus**

Under non-stress conditions a substantial percentage of cellular PARP1 molecules (i.e., ~40%) reside within nucleoli [135]. Nucleolar accumulation of PARP1 was firstly documented in the late 1980s by using immunolabeling and was later on confirmed in proteomic studies [136,137]. PARP1, as well as PARP2, are retained in nucleoli via interaction with the multifunctional nucleolar hub protein nucleophosmin, which is implicated in multiple steps of ribosome biogenesis, including rDNA transcription and elongation, as well as rRNA processing [32,138]. Treatment with the RNA Pol I inhibitor ActD resulted in nucleolar release of PARP1 and PARP2, indicating that active nucleolar transcription is required for PARP1 and PARP2 to reside in nucleoli [139]. The presence of PARP1, particularly in transcriptionally active nucleoli, gave rise to the idea that PARP1 might be involved in canonical nucleolar functions, e.g., in regulating ribosomal biogenesis. Indeed, there is a growing body of evidence, that PARP1 and PARylation play important roles in nucleolar biology, which will be discussed in the following.

First evidence for PARP1 to play a role in nucleolar biology was proposed by Tulin et al. [140] in 2002 by using *Drosophila melanogaster* as a model system. Unlike mammals, the *Drosophila* genome contains only two PARP encoding genes, i.e., one that is highly related to mammalian PARP1, as well as one homolog of tankyrases [141,142]. Therefore, *Drosophila* is a powerful model organism to study PARP biology. In the study of Tulin et al. [140], it was shown that many of the PARylated proteins in *Drosophila* are enriched in nucleoli and in the heterochromatic chromocenter regions. Furthermore, disruption of PARP1 expression resulted in abrogated formation of nucleoli and larval lethality, suggesting that PARP1 is required for the formation of nucleoli during development [140]. As previously discussed, it was shown that liquid-liquid phase separation plays an important role in internal organization of nucleolar architecture [3]. Interestingly, PAR can function as a seed for liquid-liquid demixing in intrinsically disordered proteins, which are abundant in nucleoli [125]. Therefore, it is conceivable, that PARylation regulates the general biophysical state of nucleolar architecture by liquid-liquid demixing [124] (Figure 1). In this regard it is interesting to note that nucleoli in yeast and other lower eukaryotes exhibit a bipartite structure, i.e., lacking the FC in nucleoli [143]. The fact that PARylation is missing in yeast [144] makes this an interesting correlation, which needs to be analyzed in detail for any causative relationship in future studies.

Meanwhile, several studies have revealed an important role of PARP1 in regulating multiple steps of ribosome biogenesis. Thus, PARP1 was proposed to participate in ribosome biogenesis by controlling pre-rRNA processing, post-transcriptional modification, and assembly of pre-ribosomal subunits [145]. Disruptions of PARP1 enzymatic activity led to nucleolar disintegration and aberrant localization of nucleolar-specific proteins, such as fibrillarin, nucleophosmin and nucleolin. Therefore, the authors of this study concluded that PARP1 and PARylation are important for nucleolar integrity and the localization of nucleolar-specific proteins in proximity to pre-rRNA. Furthermore, PARP1 mutants displayed a delay in rRNA processing and increased levels of rRNA intermediates, such as 47S and 36S, resulting in decreased ribosome levels [145]. In mammalian cells, it was demonstrated that PARP1 participates in nucleolar remodeling complex (NoRC)-mediated rDNA silencing during replication [146] (Figure 1).

**Figure 1.** Functional roles of PARP1 in nucleolar biology. DDX21, DEAD-box helicase 21; DNA-PK, DNA-dependent protein kinase; NoRC, nucleolar remodeling complex; PAR, poly(ADP-ribose); PARP1, poly(ADP-ribosyl) polymerase 1; Pol I, polymerase I; pRNA, promotor-associated RNA; rDNA, ribosomal DNA; rRNA, ribosomal RNA; snoRNA, small nucleolar RNA; TIP5, TTF-1-interacting protein 5; UBF, upstream binding factor; WRN, Werner syndrome protein; XRCC1, X-ray repair cross complementing 1. For details see Section 4.

In mid-late S phase, the NoRC mediates heterochromatin formation and silencing of rDNA transcription via the recruitment of the histone acetylase HDAC1 and DNA methyltransferase (DNMT) to the rDNA promoter [147]. PARP1 associates with the rDNA repressor TTF-1-interacting protein 5 (TIP5), which is part of NoRC, at silent rRNA genes during replication [146].

Furthermore, association of PARP1 with TIP5 is mediated by the NoRC-associated noncoding RNA (pRNA, promoter-associated RNA). Interestingly, PARP1 PARylates silent chromatin and components of the NoRC complex, including TIP5. Thus, it is likely that pRNA stimulates PARP1 enzymatic activity, whichis necessary to establish rDNA silencing. These findingsindicate that PARP1 canmodulate chromatin structure and gene expression in nucleoli, revealing a mechanism by which PARP1 ensures that silent rDNA regions are properly inherited after their disruption during DNA replication [148,149]. In a recent study, another pathway was identified by which nucleolar actions of PARP1 participate in the control of rDNA transcription and ribosome biogenesis during cell proliferation [112]. Thus, Kim et al. [112] demonstrated that PARP1's binding to snoRNAs leads to its catalytic activation in a DNA damage-independent manner. Upon activation of its catalytic activity, PARP1 ADP-ribosylates the nucleolar RNA helicase

DDX21, which results in enhanced rDNA transcription and proliferation of breast cancer cells (Figure 1). As discussed below in more detail, the pharmacological inhibition of this pathway may contribute to the effectiveness of PARP inhibitors in the treatment of certain cancers [112].

Importantly, there is emerging evidence that PARP1 and its enzymatic activity are involved in the regulation of ribosome biogenesis under certain pathological conditions. For instance, in hippocampal pyramidal neurons in Alzheimer's disease (AD) nucleolar PARP1 is significantly decreased compared to control cells [150]. It was proposed that under physiological conditions, PARP1 ADP-ribosylates DNMT1, thereby preventing rDNA methylation, which results in upregulation of rDNA transcription. Thus, in AD neurons, PARP1 mislocalization leads to hypermethylation of rDNA, reduced rDNA transcription and impaired ribosome biogenesis, which ultimately results in disruption of long-term memory formation. Interestingly, such a decrease in PARP1 staining was revealed in neurons of individuals with mild cognitive impairment (MCI), suggesting that decreased nucleolar PARP1 could act as an early biomarker of cognitive impairment [151]. For a more detailed discussion on this topic and the role of PARPs and PARylation in RNA biology in general the reader is referred to a comprehensive recent review by Kim et al. [91].

Apart from its contribution to ribosome biogenesis under non-stress conditions, PARP1 was shown to participate in the nucleolar stress response, in particular DNA damage response. A study by Calkins et al. [152] demonstrated that PARP1 regulates rDNA transcription in response to DNA damage (Figure 1). Induction of DNA damage, by using γ irradiation, UV light or the cross-linking agent cisplatin, resulted in inhibition of rDNA transcription, as well as cell cycle arrest in S phase. Inhibition of PARP1 or DNA-PK prevented silencing of rRNA synthesis, yet not the accumulation of cells in S phase. These results indicate that PARP1 and DNA-PK are involved in DNA damage-induced inhibition of rDNA transcription, however not in the accompanying cell cycle arrest. Loss of DNA-PK function prevented PARP1 from being activated and recruited to chromatin, suggesting that DNA-PK acts upstream of PARP1 to block rRNA synthesis upon DNA damage. While the exact mechanistic details by which PARP1 and DNA-PK contribute to DNA damage-induced inhibition of rDNA transcription remain to be elucidated, PARP1 may also facilitate DNA damage-induced block of rRNA synthesis through the recruitment of nucleolar proteins with roles in ribosome biogenesis to the nucleoplasm. For instance, fluorescence loss in photobleaching (FLIP) experiments showed that the macrodomain-containing protein TARG1/C6orf130 continuously undergoes nucleolar-nucleoplasmic shuttling [153]. Interestingly, the distribution of TARG1/C6orf130 between these two compartments was regulated by PARylation. Thus, in the absence of PAR, TARG1 localizes to transcriptionally active nucleoli, while in response to DNA damage-induced PAR formation, e.g., upon H2O2 treatment, it re-localizes to the nucleoplasm. Since TARG1 was reported to bind to RNA, ribosomal proteins, as well as proteins associated with rRNA proteins and ribosomal assembly factors, it is conceivable that in nucleoli, TARG1 plays a role in ribosome assembly or quality control, which is stalled when TARG1 is recruited to sites of DNA damage in a PAR-dependent manner [153]. With regards to a role of PARP1 and PARylation in nucleolar DNA damage response, we recently showed that upon H2O2 treatment of HeLa cells, WRN and XRCC1 translocate from nucleoli to the nucleoplasm, however interestingly enough, probably by different mechanisms [69]. Thus, while the release of WRN from nucleoli was purely dependent on the presence of PARP1 protein without any obvious involvement of its catalytic activity, we found that relocalization of XRCC1 upon H2O2-induced DNA damage was dependent on both PARP1 protein and its enzymatic activity (Figure 1). We hypothesized that in case of WRN, PARP1 and an additional unknown factor mediate the release of WRN from nucleoli, while XRCC1 requires nucleoplasmic DNA damage-bound and PARylated PARP1 as a loading platform, which leads to XRCC1 retention in the nucleoplasm until its tasks in BER are completed. This notion is supported by findings revealing that in cells without PARP1 activity, XRCC1 relocates quickly to nucleoli upon DNA damage induction [69].

Similarly to XRCC1, upon induction of DNA damage the ribosomal protein L6 (RPL6) is recruited to sites of DNA damage in a PARP-dependent manner [154]. At sites of DNA damage RPL6 appears to regulate the DNA damage response. Thus, RPL6 directly interacts with histone H2A and depletion of RPL6 impairs the recruitment of the mediator of DNA damage checkpoint 1 (MDC1), reduces ubiquitination of H2A and phosphorylated histone H2AX (γH2AX). These results exemplify that ribosomal proteins can exert PARP-dependent extraribosomal functions in DNA damage response. In general, these results support the notion that nucleolar-nucleoplasmic shuttling mechanisms are mediated by several different processes, which are highly dependent on the specific stress condition as well as the specific protein.

Another layer of complexity is added by a study of Leger et al. [155], which showed that H2O2 and N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) induce PAR formation in nucleoli. Interestingly, the combination of H2O2 or MNNG treatment with low doses of ActD revealed a synergistic effect on nucleolar PAR formation. At low concentrations ActD intercalates with GC-rich sequences of rDNA downstream of rDNA transcription start sites. Thus, ActD prevents transcription at the elongation step, resulting in accumulation of short rRNA transcripts. Interestingly, this study reported that PARP2, but not PARP1, binds through its N-terminal SAP domain to these short rRNA transcripts and thereby becomes activated, which contributes to the enhanced PAR formation inside nucleoli [155].

#### **5. Implications for Cancer Biology**

The first link between nucleoli and cancer was established over a century ago, when pathologists noticed that nucleoli in cancer cells are often enlarged, irregularly shaped, as well as increased in number and therefore could serve as markers of aggressive malignancies [156]. In general, the roles of nucleolar processes in cancer biology are manifold and for a more in-depth discussion on this topic the reader is referred to comprehensive recent review articles [6,17,29,30,157]. Broadly speaking, the role of nucleoli in carcinogenesis falls into two interdependent categories: ribosome biogenesis and stress response. The current consensus is that structural abnormalities of nucleoli in cancer cells are the direct consequence of increased ribosome biogenesis, which goes along with aberrant Pol I transcription [157]. Thus, in a way, tumor cells depend on increased ribosome biogenesis to reach their demand for newly synthesized proteins during cell proliferation [6]. Interestingly, rDNA gene clusters represent hot spots of recombination in human cancer [6]. Thus, more than half of solid tumors such as lung or colorectal cancer exhibit rDNA rearrangements [54] and alterations are also frequently observed in Hodgkin's lymphoma [158].

Usually cells precisely monitor the accuracy of ribosome biogenesis, as well as nucleolar integrity, and disruption at any step of ribosome biogenesis results in activation of cellular checkpoints [27]. Thus, while the increased ribosome biogenesis was originally thought to merely reflect the increased growth and proliferation rates in cancer cells, today it is well accepted that dysregulation in ribosome biogenesis is a result of increased activity of oncogenes or inactivated tumor suppressors [28]. One of the best understood cellular surveillance mechanisms in this context is the nucleolar stress response, also referred to as the ribosomal surveillance pathway, that often results in activation of p53 [28]. The tumor suppressor p53 restrains tumor growth by inducing cell cycle arrest, senescence, or apoptosis [159]. Therefore, it is not surprising that p53 is mutated or functionally inactivated in most human cancers [160]. Targeting p53 by activating wildtype p53 or restoring the activity of loss-of-function mutants has become a promising strategy in cancer treatment. Classical chemotherapeutics, such as alkylating agents or platinum-based drugs, in general target cancer cells by directly or indirectly inducing DNA damage in rapidly proliferating cells, which leads to activation of p53 [161]. Interestingly, 21 of 36 screened chemotherapeutics, including doxorubicin and camptothecin, were shown to additionally inhibit various steps of ribosome biogenesis, leading to p53 stabilization [162]. Since classical chemotherapeutics discriminate between cancer cells and healthy cells by their different proliferative index, patients frequently suffer from side effects on other proliferating cells in the body, which includes bone marrow suppression, alopecia, mucositis, as well as toxicity to the gastrointestinal tract, skin, and heart [163]. In addition, long-term adverse effects, such as infertility or development of secondary cancers were observed in patients following chemotherapy [164]. The finding that p53 can also

be activated in a non-genotoxic manner by disrupting nucleolar function and the aim to overcome adverse effects, gave rise to the development of a series of small molecule inhibitors, which selectively inhibit RNA Pol I transcription [165,166]. One of the most promising of such inhibitors is CX-5461, which impedes the selectivity factor SL1 from binding to the rDNA promoter, thereby preventing recruitment of the Pol I complex to the rDNA and initiation of transcription [167]. Apart from this, CX-5461 stabilizes G-quadruplex structures with increased toxicity in *BRCA*-deficient or PARP inhibition resistant cancer cells [168]. In vitro studies have revealed a high antiproliferative effect of CX-5461 in a wide range of human cancers, with those derived from p53 wildtype hematological malignancies being the most sensitive [167]. By using the Eμ-*Myc* mouse model of Burkitt's lymphoma, which is an aggressive type of lymphoma affecting B lymphocytes, it was shown in vivo that treatment with CX-5461 has the potential to lead to an almost complete disease remission and a significantly increased survival [165]. Importantly, the normal B cell population in those mice was maintained, suggesting that CX-5461 selectively targets cancer cells. Due to the promising results of this preclinical data, CX-5461 is currently tested in several clinical trials. A phase I clinical trial has successfully been completed in patients with hematological cancers and at present phase I/II trails are undergoing for solid tumors, including metastatic breast cancer, ovarian, and pancreas cancer [169]. Thus, selectively targeting ribosome biogenesis could provide a novel and efficient strategy in cancer therapy. Most cancer types rely on increased levels of ribosomes and therefore display higher sensitivity towards inhibition of ribosome biogenesis compared to normal cells. Thus, in contrast to existing chemotherapies, inhibition of ribosome biogenesis is likely less genotoxic to the non-tumor population of cells, which is associated with a reduced risk of adverse effects. The previously mentioned findings from studies with CX-5461 strongly support this notion. Given the high complexity of ribosome biogenesis and the variety of factors, which are involved in this process, it is likely that in future even more effective small molecule inhibitors will be identified.

As described in Section 3, PARP inhibitors have successfully entered the clinic as monotherapeutic agents, as well as in combination with cytostatic chemotherapy or radiotherapy. PARP inhibitors can act by inducing synthetic lethality in cancers that are deficient in HR-mediated DNA repair, e.g., loss-of-function mutation in *BRCA* [128,129]. More recent studies have suggested that PARP inhibitors may also induce replication stress and subsequent DNA damage [95,170,171]. Development of chemo-resistance has been proven to be a major problem in the clinical efficacy of PARP inhibitors [172]. Extensive in vitro and in vivo studies have identified several potential resistance mechanisms, including reactivation of HR, upregulation of drug efflux pumps and stabilization of replication forks [131].

Interestingly, some recent studies suggested that targeting nucleolar proteins/processes in combination with PARP inhibitor treatment may be beneficial for the treatment of some cancers, e.g., by overcoming PARP inhibitor resistance mechanisms. Thus, in a recent study CX-5461 was demonstrated to induce replication stress and activate the DNA damage response in high-grade serous ovarian cancer (HGSOC) cells [173]. CX-5461 showed significant therapeutic benefit as a single agent in HGSOC-patient-derived xenografts with reduced sensitivity to PARP inhibitors by overcoming replication fork protection, which is a well-known PARP inhibitor resistance mechanism. Importantly, the combination of CX-5461 and PARP inhibitors resulted in enhanced replication stress, DNA damage and cell death and exhibited great therapeutic efficacy, especially in HR-deficient HGSOC-patient-derived xenografts. Thus, there is evidence that combining PARP inhibitors with CX-5461 could improve treatment of HR-deficient HGSOC.

Furthermore, a novel mechanism for PARP resistance development has been reported by Sun et al. [174], which involves increased phosphorylation of the ribosomal protein S6 (RPS6). RPS6 is a component of the 40S ribosomal subunit and a well-known downstream effector of the mammalian target of rapamycin (mTOR) signaling pathway, which is the major nutrient-sensitive regulator of eukaryotic cell growth, metabolism, proliferation and survival [175]. RPS6 is phosphorylated by ribosomal protein S6 kinases (S6Ks) at five C-terminal serine sites and this modification is crucial for regulation of cell size, cell proliferation and glucose homeostasis [176]. Phosphorylation of RPS6 is greatly increased in *BRCA1*-deficient cancer cells, which are resistant to PARP inhibition [174]. Importantly, in *BRCA1*-deficient cells RPS6 phosphorylation promoted loading of the HD marker RAD51 onto DNA following IR-induced DNA damage. Thus, RPS6 phosphorylation might play a key role in PARP inhibitor resistance by regulating HR. Intriguingly, rapamycin, which is a clinically used selective inhibitor of mTOR and S6 phosphorylation, could restore sensitivity towards PARP inhibition, suggesting that combined inhibition of S6 phosphorylation and PARP could be efficient in cancers with PARP inhibitor resistance and HR defects, including *BRCA1*-deficient breast and ovarian cancers.

Several studies have reported that *BRCA* mutations or other HR-mediated DNA repair deficiencies are not mandatory for the clinical effectiveness of PARP inhibitors in cancer therapy [177–179]. In accordance with this, Kim et al. [112] revealed an alternative working mechanism of PARP inhibitors, which is independent of HR, DNA damage and replication stress. Mechanistically, it was shown that snoRNAs can stimulate PARP1 enzymatic activity in the nucleolus, resulting in ADP-ribosylation of DDX21. The DEAD-box RNA helicase DDX21 was previously reported to directly interact with rRNA and snoRNAs at the transcribed rDNA locus, thereby promoting rRNA synthesis, rRNA processing and modification [180]. Furthermore, analyses of gene expression profiles from breast cancer patients identified DDX21 as a prognostic marker in breast cancer [181]. Kim et al. [112] demonstrated that DDX21 promotes rDNA transcription and breast cancer growth upon PARP1-mediated ADP-ribosylation. PARP inhibition on the other hand resulted in reduced tumor cell growth by modulating rRNA levels, DDX21 ADP-ribosylation and DDX21 localization. These findings could be explained by the fact that some cancer types are "addicted" to ribosome biogenesis, whereas reducing ribosome biogenesis can counteract cancer cell growth. The finding that PARP inhibitors can reduce tumor growth by targeting ribosome biogenesis provides a mechanistic explanation for efficacy of PARP inhibitors in cancer cells lacking deficiencies in HR. In addition, these results further suggest that DDX21 nucleolar localization could be a predictive biomarker of clinical responses to PARP inhibition.

#### **6. Concluding Remarks and Perspectives**

Nucleolar localization of PARP1 was demonstrated almost four decades ago. For many years, scientists have been puzzled whether PARP1 accumulation in the nucleolus merely occurs for storage reasons or if PARP1 might also be implicated in nucleolar functions. As summarized in this review, there is growing evidence, supporting a role of PARP1 and PARylation in nucleolar biology. PARP1 is implicated in multiple areas of nucleolar function, including maintenance of nucleolar integrity and structure, regulation of Pol I transcription, establishment of silent rDNA chromatin, as well as regulation of DNA damage-induced nucleolar-nucleoplasmic shuttling processes of key genome maintenance factors, e.g., WRN and XRCC1 (Figure 1). Previous studies have demonstrated that PAR can function as a seed of liquid-liquid demixing processes, thereby triggering the formation of biomolecular condensates [124,125]. Therefore, it can be assumed that PARP1 and PARylation might regulate the general biophysical state of nucleolar structure. Since PARP1 plays a crucial role in several DNA repair pathways, of which at least NHEJ and HR have been reported to take place in the nucleolus, it can be anticipated that PARP1 might also be involved in the repair of rDNA [88,182]. Repetitive rDNA sequences represent one of the most unstable regions in the genome [6]. Instability of rDNA is associated with severe pathological conditions, including cancer, premature aging and neurological impairments [182]. Yet, the DNA damage repair mechanisms that govern genomic stability and maintenance in the nucleolus remain elusive. In future studies, recent advancements in CRISPR-genome engineering could provide deeper insights into the mechanisms that help cells keep their rDNA intact. Until now, in large, only PARP1 was reported to play a role in nucleolar biology. In future studies it will be interesting to investigate to what extent other members of the PARP family, e.g., PARP2, also contribute to nucleolar functions. As previously described, the tripartite organization of the nucleolus reflects the different stages of ribosome biogenesis. Therefore, identifying the exact localization of PARP1 and other PARPs, as well as PAR, in nucleolar substructures could provide further evidence for a role in certain steps of ribosome biogenesis. Importantly, PARP1 inhibitors

have entered the clinic as promising chemotherapeutic agents in the treatment of various cancer types, mainly by exploiting synthetic lethal interactions between PARP inhibition and defects in genes that are responsible for HR-mediated DNA repair. As discussed above alternative working mechanisms of PARP inhibitors are conceivable, which act by preventing tumor growth through inhibition of ribosome biogenesis [112]. These data indicate that PARP inhibitors might be effective in a broader spectrum of cancer types than originally anticipated. In addition, in this study, the DEAD-box RNA helicase DDX21 was identified as a potential marker to predict sensitivity towards PARP inhibitors, which needs to be further evaluated in clinical trials. In future studies it will be important to elucidate further mechanisms by which PARP inhibitors contribute to inhibition of nucleolar functions, to pave the way for the identification of other biomarkers, which can predict response to PARP inhibitors in cancer patients.

**Author Contributions:** Conceptualization of the manuscript, M.E. and A.M.; writing and editing of the manuscript, M.E. and A.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** Experimental research of our lab that is associated with the research topic reviewed in this article is funded by University of Konstanz and the German Research Foundation (DFG, grant MA4905/4-1). Funding for open access charge: University of Konstanz.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design and the in the writing of the manuscript.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **The Modified Phenanthridine PJ34 Unveils an Exclusive Cell-Death Mechanism in Human Cancer Cells**

#### **Malka Cohen-Armon**

Sackler Faculty of Medicine, and Sagol School of Neuroscience, Tel-Aviv University, Tel-Aviv 69978, Israel; marmon@tauex.tau.ac.il

Received: 8 June 2020; Accepted: 15 June 2020; Published: 19 June 2020

**Abstract:** This overview summarizes recent data disclosing the efficacy of the PARP inhibitor PJ34 in exclusive eradication of a variety of human cancer cells without impairing healthy proliferating cells. Its cytotoxic activity in cancer cells is attributed to the insertion of specific un-repairable anomalies in the structure of their mitotic spindle, leading to mitotic catastrophe cell death. This mechanism paves the way to a new concept of cancer therapy.

**Keywords:** the PARP inhibitor PJ34; distorted mitotic spindles; exclusive eradication of human cancer cells

#### **1. Background**

In the last twenty years the modified phenanthridine PJ34 has been known for its activity as a PARP (polyADP-ribose polymerase) inhibitor [1,2] (Figure 1). Recently, PARP inhibitors attract the attention of researchers and clinicians due to their FDA approval for cancer therapy [3–6]. Several comprehensive overviews on the family of PARP proteins and their inhibitors have been published [3,7–11]. This overview is dedicated to the recently disclosed exceptional cytotoxicity of the PARP inhibitor PJ34 in human cancer cells, which does not affect human healthy cells.

**Figure 1.** The chemical structure of PJ34, *N*-(6-Oxo-5,6-dihydrophenanthridin-2-yl)-(*N*,*N*-dimethylamino)acet amide hydrochloride.

The modified phenanthridine PJ34 is a stable molecule, fairly soluble in water (22 mg/mL), and permeable in the cell membrane. The potency of PJ34 to inhibit PARP proteins has been

measured [8]. PJ34 is a potent inhibitor of PARP1 and PARP2 (approximate IC50, 20 nM). In addition, PJ34 inhibits tankyrase-1 and tankyrase-2 (approximate IC50, 1 μM). PJ34 hardly inhibits other PARP proteins [8]. PJ34 was originally invented to protect cells from cell death imposed by pathological stress conditions, such as ischemia [1,2,12–14].

The common concept in utilizing PARP inhibitors for cancer therapy is based on evidence associating PARP1 inhibition with the prevention of DNA break repair causing apoptotic cell death [6,15–18]. PARP1 is activated and polyADP-ribosylated in response to DNA single strand breaks frequently caused under stress conditions, including X-ray irradiation and application of DNA-damaging agents [3–6]. The binding of PARP1 to DNA single-strand breaks and its activation (polyADP-ribosylation) initiates their repair [3–6]. In addition, PARP1 is implicated in the alternative non-homologous end joining (A-NHEJ) mechanism of double strand DNA break repair [19,20]. Cancer cells have a higher percentage of DNA breaks relative to healthy cells, due to failure in arresting mitosis of cancer cells with damaged DNA [15,16]. In addition, DNA-damaging treatments are used for cancer therapy [6,16,17]. Thus, prevention of DNA repair by PARP inhibitors can be beneficial for cancer therapy, either as a monotherapy, or in combination with treatments causing DNA damage [11,16–19]. According to this concept, PARP inhibitors are offered to cancer patients carrying *BRCA* gene mutations (about 2% of the western world population). The BRCA protein is implicated in the repair of double-strand DNA breaks [17,21]. Mutations in BRCA impact the functioning of the BRCA protein, and increase DNA breaks in the cells of BRCA mutant carriers. This frequently increases the probability of mutations associated with malignancy [17,21]. Treatment with PARP inhibitors in BRCA mutant carriers is based on the interference of PARP1 inhibition with DNA repair in cells carrying a damaged DNA.

The partial efficacy of PARP inhibitors in the small population of *BRCA* mutant carriers, the resistance of a variety of cancer types to the currently approved PARP inhibitors, their side effects and the unclear impact of their chemical structure on their potency [21–26], urged a further investigation of the activity of PARP inhibitors in cancer therapy. It was observed that apart from PARP inhibition, some of these molecules target a variety of kinases implicated in signal transduction pathways in both healthy and malignant cells [27]. Unexpectedly, this research also disclosed that a group of phenanthrene derivatives acting as PARP inhibitors, exclusively kill human cancer cells without affecting benign cells [28–32]. Unlike other PARP inhibitors, these small molecules exclusively eradicated a variety of human cancer cells without affecting proliferating and non-proliferating healthy somatic cells. They did not affect human epithelial, mesenchymal and endothelial cells [28–35], nor healthy cells of mouse origin, including mouse embryonic fibroblasts (MEF), fibroblasts, neurons in the central nervous system and neuronal progenitor stem cells [28,29,31–34]. Their exclusive cytotoxic activity in human cancer cells was not shared by other potent PARP inhibitors [29–31]. Moreover, their toxicity in human cancer cells was independent of the expression of PARP1 and P53, PARP1 activity and DNA damage [29,34,35]. On the other hand, their exclusive cytotoxic activity in human cancer cells resembles the cytotoxic activity of other phenanthridines [36–38]. The modified phenanthridine PJ34, one of the molecules in this group, was the most potent in a variety of human cancer cells, including cells that are resistant to currently offered therapies [28–32]. Its specific cytotoxic activity in human cancer cells is summarized in this overview.

#### **2. PJ34 E**ffi**ciently Eradicates a Variety of Human Cancer Cells in Tissue Cultures**

After years of research based on PJ34-induced PARP inhibition in a variety of cell types under pathological conditions [1,2,8], additional activities of PJ34 have been disclosed. It was observed that PJ34 causes an irreversible cell growth arrest in cancer cells, that it interferes with angiogenesis, and, most interestingly, that PJ34 exclusively eradicates human cancer cells [29,30,39,40].

Incubation with PJ34 at higher concentrations than those inhibiting PARP1 (10–20 μM PJ34), completely eradicated within 48 h human MCF-7 breast cancer cells that are resistant to doxorubicin [28]. Furthermore, PJ34 (20–30 μM) eradicated within 72–96 h cancer cell types that are resistant to other

therapies, including types of triple negative breast cancer, pancreatic cancer, ovary cancer, colon cancer and non-small lung cancer [28–32].

Gangopadhyay and colleagues found that incubation with 30 μM PJ34 for 72 h eradicates several human metastatic lung cancer cell lines: Calu-6, A549 and H460 [41]. In addition, PJ34, at higher concentrations than those inhibiting PARP1, arrested the growth of human liver cancer cell lines (HepG2 and SMMC7721) [42], and the human multiple myeloma RPMI8226 cell line [43]. PJ34 acts as a potent anti-proliferating agent in human leukemia cell lines (ATLL and transformed HTLV-I) [44], and in human ovarian cancer epithelial cells (C13 cell line) [45]. The cell death-inducing efficacy of PJ34 at higher concentrations than those inhibiting PARP activity has been also reported in a variety of breast cancer cell-lines, carrying or not BRCA mutations, and in a variety of triple-negative breast cancer cell-lines [46], as well as in melanoma cell lines and melanoma metastases [47], thyroid cancer cell lines [48], HeLa cells [49] and several glioblastoma cell lines [50]. PJ34 also efficiently prevented Helicobacter-induced gastric pre-neoplasia [51]. On the other hand, healthy proliferating cells treated with PJ34, at the same concentrations eradicating cancer cells, continued to proliferate in the presence of PJ34 as untreated cells for weeks [28–32]. Furthermore, incubation with PJ34 (20 μM) did not affect retinoic acid-induced differentiation in the human neuroblastoma cell line SHSY5Y [52], nor impaired the neuronal excitability of mouse hippocampal neurons [34].

Notably, mice treated with PJ34 for 2–3 weeks, continued to gain a similar amount of weight as untreated mice, and did not exert any visible stress or discomfort signs, as described below.

#### **3. PJ34 Causes Eradication of Human Cancer Cells in Xenografts**

The cytotoxicity of PJ34 in human cancer cells was tested in animal models as well. PJ34 (10 mg/kg) intraperitoneal (IP) injection 3 times a week for 3 weeks attenuated the growth of intracranial tumors of glioblastoma in nude mice [50]. The efficacy of daily treatment of PJ34 (30 mg/kg) using intraperitoneal injection over 14 days has been tested in xenografts of ovarian cancers. Treatment with PJ34 at this dosage most efficiently decreased the tumors' size [53]. In contrast, treatment with PJ34 in xenografts prepared from uterus and ovarian cancer of BRCA carriers showed insufficiency when PJ34 was applied per os (10 mg/kg twice a day for 16 days), in comparison to chemotherapy with carboplatin (80 mg/kg) and paclitaxel (24 mg/kg) injected IV (intravenous) once a week [54].

PJ34 prevented the development of human breast cancer MCF-7 and triple negative breast cancer MDA-MB-231 tumors in immunocompromised mice treated with a slowly released PJ34 from subcutaneous osmotic pump (Alzet) over 14 days [28]. The PJ34-treated mice that were injected subcutaneously with MCF-7, 5 <sup>×</sup> 10<sup>6</sup> cells before treatment, did not develop breast cancer tumors, and remained tumor-free during the following four months. In mice subcutaneously injected with human MDA-MB-231 triple negative breast cancer cells (5 <sup>×</sup> 106), PJ34 slow release for 14 days prevented the development of tumors in three out of five mice, and those mice remained tumor-free during the following four months. Importantly, the treatments with PJ34 did not affect the vitality, growth and weight-gain of the treated mice during the follow-up periods [28]. The therapeutic potency of PJ34 was also tested in triple negative MDA-MB-231 breast cancer xenografts after tumor development. In these experiments PJ34 (60 mg/kg) was injected IP after the tumors reached a volume >100 mm3. PJ34 injected daily for 14 days efficiently suppressed tumor growth [31].

The efficacy of PJ34 (60 mg/kg, IV injected daily, 5 days a week, 14 intravenous injections) was tested in xenografts of human pancreas ductal adenocarcinoma PANC1 developed in immunocompromised mice. A substantial reduction of 80–90% in human cancer cells in the tumors was measured by immunohistochemistry in slices prepared from the excised tumors (16 mice) 30 days after the treatment with PJ34 has been terminated. One of the tumors disappeared after the treatment with PJ34 [32]. Benign fibroblasts infiltrated into the PANC1 tumors (stroma cells) were not impaired by the treatment with PJ34. Growth, weight-gain and behavior of the treated mice were not impaired during, and 30 days after the treatment with PJ34 has been terminated [32]. A similar cytotoxic activity of PJ34 was observed in patients-derived pancreatic cancer cells, and in patient-derived xenografts [32].

A combined treatment of PJ34 (IV) with other agents was examined, as well. A combined treatment of PJ34 with agents inducing TRAIL-mediated apoptosis in glioma xenografts reduced glioma tumor growth, and revealed minimal cytotoxicity in non-neoplastic astrocytes [55]. The combined treatment of PJ34 with TRAIL agonists was non-toxic to normal human primary glial and neuronal cells, anticipating minimal side-effects of this treatment in patients [55]. A similar effect was achieved in pancreas cancer xenografts treated with PJ34 in combination with agonists activating TRAIL-mediated apoptosis [56].

A combined treatment with PJ34 and the HDAC (histone-deacetylase) inhibitor SAHA blocked the growth of liver tumors [42]. In these experiments, HepG2 (5 <sup>×</sup> 106) cancer cells were injected subcutaneous into immunocompromised mice. These mice were treated with a combination of PJ34 (IP, 10 mg/kg) and SAHA (IP, 25 mg/kg), 3 times a week for 3 weeks. Under these conditions, tumor growth was substantially suppressed without affecting human hepatocytes [42]. Notably, SAHA at higher doses was cytotoxic to normal liver human fetal hepatocytes, while high dosage of PJ34 did not harm these cells [42].

Combining PJ34 with other anti-cancer treatments enabled reducing their cytotoxic doses and achieved efficient eradication of HL-60, Jurkat-T cells, multiple myeloma cellsPRMI8226/R and B16F10 melanoma cells [57–59].

In view of these findings, the possibility that common mechanisms are targeted by PJ34 in a variety of human cancer cell types has been examined.

#### **4. The Mechanism of Action of PJ34 in Human Cancer Cells**

In a search for the mechanism of action of PJ34 in human cancer cells, several mechanisms have been considered. The activity of PJ34 in cancer cells has been suggested to promote cell death by preventing PARP1-mediated DNA repair. This suggested mechanism is in line with PARP inhibition sensitizing cancer cells to apoptotic cell death by DNA-damaging agents [53,58–60] and by blocking DNA double-strand break repair [42]. In view of the decreased expression of PARP1 and NFkappaB in several cancer cell types eradicated by PJ34, it has been suggested that eradication by PJ34 can be attributed to the attenuation of PARP1-dependent NF-kappaB activity that promotes proliferation [45]. However, the suggested mechanisms of PARP1-dependent activity of PJ34 in cancer cells are inconsistent with its exclusive PARP1 independent cytotoxic activity in human cancer cells at a higher concentration range than that causing PARP1 inhibition [29–31,35]. Its exclusive cytotoxic activity in human cancer cells was not shared by other potent PARP inhibitors [29–31,61]. PJ34 applied at the concentration range eradicating cancer cells did not cause breaks in the DNA, nor impaired healthy cells, and its activity was not restricted to cancer cells exposed to DNA-damaging agents nor BRCA mutants [28–32,35,46].

Flow cytometry measurements revealed that PJ34 exclusively arrests mitosis in human cancer cells [28,29,32]. Incubation with PJ34 causes cell-cycle arrest in a variety of healthy proliferating and cancer cells within 3–6 h [28]. However, healthy cells of mouse and human origin overcame the imposed cell-cycle arrest [28,29,31,33], and continued to proliferate in the presence of PJ34 similarly to untreated cells for weeks [28,29] (Figure 2). The tested healthy proliferating human cells included human epithelial, mesenchymal and endothelial cells [28,29,31]. Proliferating healthy cells of mouse origin included mouse embryonic fibroblasts, fibroblasts [28,29,32], and neuronal progenitor cells and astroglia [33].

**Figure 2.** *Cont*.

**Figure 2.** G2/M arrest and cell death in human cancer cells treated with PJ34. PJ34 does not impair the cell-cycle of human healthy proliferating cells. Flow cytometry of the indicated human cancer cells (**a**) and human healthy proliferating cells (**b**) is displayed. Cells were treated with PJ34 at the indicated concentrations and incubation periods. From: Castiel et al., 2011, *BMC Cancer*.

In contrast, the cell-cycle arrest in mitosis which was imposed by PJ34 in human cancer cells was irreversible in all the tested human cancer cells, and the mitosis arrest was accompanied by cell death [28–32] (Figure 2). Cell-cycle arrest preceding cell death was measured in a variety of human cancer cell types, including cancer cells prepared from human colon, ovary, lung, pancreas, T-cell leukemia and breast cancers, including triple negative breast cancer [28,29,31,32,44].

Flow cytometry measurements at the measured time intervals did not reveal G1 or S-phase arrest in the tested human healthy and cancer cells treated with PJ34 (10, 20 and 30 μM) ([28,29], Figure 2)).

Since many of the tested malignant cell types included a high percentage of extra-centrosomal cells [28–30], the cytotoxic activity of PJ34 was attributed to de-clustering of extra-centrosomes in the mitotic spindle poles of multi-centrosomal cancer cells [29,30]. According to several reports, de-clustering of centrosomes in multi-centrosomal cancer cells leads to aberrant multi-polar spindles with un-assembled chromosomes [30,62–64]. Thus, it has been suggested that multi-centrosomes de-clustering in the spindle poles activates the spindle assembly check-point (SAC) proteins, which arrest mitosis [65,66], subsequently leading to mitotic catastrophe cell death [65–68].

Since mitotic arrest was induced by PJ34 in a variety of cancer cells, including cell types with a low percentage of extra-centrosomal cells [69,70], other mechanisms inducing mitotic catastrophe cell death were examined.

In one approach, all proteins currently known to be implicated in mitosis were screened in a group of human cancer and healthy cells, in an attempt to identify different effects of PJ34 (eradicating only cancer cells) on the post-translational modifications of proteins in the cancer versus healthy cells. Changes induced by PJ34 in their post-translational modifications were measured by the shift in their isoelectric point (IP) in two-dimensional (2-D) gel electrophoresis [31]. This approach was used because PARP inhibitors may modify a variety of proteins in both healthy and malignant cells [27,61,71].

Among other proteins associated with mitosis, proteins that are implicated in the structure of the mitotic spindle were examined for a possible interference with their post-translational modification by PJ34 in human cancer cells versus the effect of PJ34 on their post-translational modification in healthy epithelial cells [31]. This analysis was conducted in four types of human cancer cell lines (glioblastoma U87, PANC1 pancreas cancer cells, lung cancer cells A549 and triple-negative breast cancer cells MDA-MB-231). All these cells are eradicated by PJ34 [28–32]. A similar analysis was used to identify possible effects of PJ34 on the post-translational modifications of the same proteins in healthy human epithelial cells, which are not impaired by PJ34 [28,29,31].

This analysis identified only three proteins in the tested cancer cells with isoelectric point significantly shifted by PJ34, while not affected in healthy cells [31]. These proteins included two motor proteins [72], human kinesins 14/HSET/kifC1 and kif18A, and the non-motor protein NuMA (nuclear mitotic apparatus protein) [73–77]. For comparison, the IP shift of these proteins in the tested cells was also measured when these cells were incubated with the PARP inhibitor ABT-888 (Veliparib), which is a potent inhibitor of PARP-1,-2,-3 proteins (with a similar affinity for PARP1 and PARP2, *Kd* = 2.9 and 5.2 nM, respectively) [8,31]. Unlike PJ34, ABT-888 did not affect the isoelectric points of human kinesins 14/HSET/kifC1 and kif18A and of NuMA. There was no difference between the IP of these proteins in ABT-888-treated versus untreated cancer cells [31]. This result may exclude possible involvement of PARP-1,-2,-3 inhibition in the IP shift induced by PJ34 in the tested cancer cells [31]. In accordance, ABT-888 neither arrested mitosis, nor eradicated the tested human cancer cells [31]. These findings suggested a possible interference of PJ34 with the construction of the mitotic spindle in the tested human cancer cells.

Numerous findings indicate the essential role of HSET/kifC1 in the spindle structure of human cancer cells [73–75,78–81]. Differences in the expression and function of HSET in cancer versus healthy cells have been reported [74,78–80]. HSET/kifC1 inhibition or silencing causes small aberrant spindles in human malignant cells [79].

The kinesin Kif18A is implicated in microtubules de-polymerization, necessary for the binding of the duplicated chromosomes to kinetochores in the spindle mid-zone. Loss of its function results in the formation of long microtubules with chromosomes un-attached to kinetochores in the mid-zone [76].

The third identified protein, NuMA, is essential for mitosis in both malignant and benign cells [77,82,83]. Recently, several reports indicated some differences between NuMA proteins and their expression in benign versus malignant cells [84,85]. In this analysis, a clear-cut difference has been disclosed by the effect of PJ34 on the post-translational modification of NuMA only in the variety of human malignant cells ([31] and Figure 3). PJ34 did not affect the isoelectric point of NuMA in the healthy epithelial cells [31]. Concomitantly, the ability of NuMA to bind proteins was lost in the PJ34-treated malignant cells [31]. Moreover, the lost ability of NuMA to bind proteins was accompanied by un-clustering of NuMA in the spindle poles of the malignant cells treated with PJ34, as disclosed by confocal imaging [31] (Figure 3). In contrast, the bi-polar clustering of NuMA in the mitotic spindles of the healthy benign cells was not affected by the treatment with PJ34 ([31] (Figure 3).

Previous findings indicated the post-translational modification of NuMA by both polyADP-ribosylation and phosphorylation, both modifications promoting the binding of NuMA to proteins [82–89]. NuMA is phosphorylated by serine-threonine kinase pim1, and NuMA phosphorylation by serine threonine kinases at a specific site is crucial for its ability to bind proteins [87,89]. In addition, polyADP-ribosylation of NuMA by tankyrase1 in cancer cells promotes the ability of NuMA to bind other proteins [31,88].

Pim kinases and tankyrase1 are both inhibited by PJ34 at the concentrations range PJ34 causes cell death in human cancer cells (measured IC50 = 3.7 μM for pim1 inhibition by PJ34, and for tankyrase1 inhibition by PJ34, IC50 = 1 μM) [90–92]. Furthermore, tankyrase1 and pim kinases are hardly expressed in healthy somatic cells, while they are highly expressed in human cancer cells [92,93].

Clustered NuMA in the spindle poles and tethering of microtubules to the clustered proteins in the spindle poles are essential for the construction of stable poles required for the binding of the chromosomes to kinetochores in the spindle mid-zone [31,77–83]. Thus, PJ34 blocking of the post-translational modification of NuMA concomitantly prevents the clustering of NuMA in the spindle poles, causing mitotic arrest and cell death by preventing the binding of chromosomes to kinetochores in the spindle mid zone [31,65–68,80–82].

The causal association of the post-translational modification of NuMA with the ability of NuMA to bind proteins [31,87–89] could explain the interference of PJ34 with NuMA clustering and binding to HSET in the spindle poles of cancer cells treated with PJ34 [31,80–83]. NuMA clustering in the poles and the tethering of microtubules to the spindle poles implicates that HSET and NuMA binding in the protein clusters causes a stable structure that is lost in the spindle poles of human cancer cells treated with PJ34, at the concentration range inhibiting pim1 and tankyrase1 [31,91,92].

An unstable structure of the spindle poles due to impaired function of HSET and NuMA, and impaired function of kif18A in the binding of chromosomes to the kinetochores may result in scattered chromosomes instead of chromosomes bound to the kinetochores in the mid-zone (Figure 3). This activates the spindle assembly control (SAC) proteins, which leads to mitosis arrest followed by mitotic catastrophe cell death when the structural anomaly is not amended [65–68]. This is exactly the phenomenon observed by confocal imaging in a variety of human cancer cells treated with PJ34 [29–31]. De-clustering of centrosomes observed in multi-centrosomal cancer cells treated with PJ34 (20 μM) could result from the un-stable aberrant spindle poles [31,78]. Thus, prevention of the post-translational modification of NuMA and kinesins HSET and Kif18A by PJ34, could evoke mitotic arrest and mitotic catastrophe cell death in human cancer cells [30–32] (Figure 2) just by inserting specific anomalies in their mitotic spindle structure ([31] and Figure 3). This effect of PJ34 was not copied by ABT-888 at concentrations inhibiting PARP-1,-2,-3 [8,31]. It was achieved by inhibiting the activity of serine-threonine kinases, and by inhibition of tankyrase polyADP-ribosylation [31]. Other tankyrase inhibitors and tankyrase1 silencing caused similar faults in the structure of spindle poles as well as G2/M arrest, which could be attributed to inhibition of NuMA polyADP-ribosylation [31,88,94,95]. These results are consistent with the consequences of the treatment with PJ34 causing aberrant spindles with dispersed chromosomes (Figure 3) arresting mitosis and killing cancer cells without impairing healthy proliferating cells (Figure 2) [28–32]. In consistence, the findings of Leber and collaborators implicated NuMA and HSET in the construction of aberrant spindle poles with un-clustered multi-centrosomes in human cancer cells [96].

**Figure 3.** (**a**) Confocal images of mitotic spindles in human triple negative breast cancer cells (MDA-MB-231) and in human healthy breast epithelial cells (MCF10A), untreated or incubated with PJ34. Incubation of human cancer cells with PJ34 (20 μM, 27 h) impaired spindle poles (labeled by immunolabelling, γ-tubulin in the centrosomes—red), microtubules (labeled by immunolabelling—kinesin HSET or by immunolabelling α-tubulin—green), segregation and alignment of chromosomes (labeled by DAPI—blue), and NuMA clustering in the spindle poles (Immunolabeled NuMA—red). **Column 1**: Microtubules in spindles of healthy and cancer cells immunolabeled by the kinesin HSET in cancer and healthy cells, untreated and treated with PJ34. **Column 2**: Spindle poles labeled by γ-tubulin in healthy and cancer cells, untreated and treated with PJ34. HSET is immunolabeled in the microtubules. **Column 3**: Clustered NuMA in bipolar spindles of healthy cells either treated or not with PJ34, and in untreated cancer cells. Un-clustered NuMA in spindles of cancer cells treated with PJ34. **Column 4**: upper frame: In cancer cells—clustered NuMA in spindle poles and aligned chromosomes in the midzone of untreated cancer cells. Aberrant spindles, un-clustered NuMA

and scattered chromosomes in cancer cells treated with PJ34. Lower frame: In healthy cells—clustered NuMA in the spindle poles and segregated chromosomes aligned in the mid-zone of the mitotic spindle of healthy cells either untreated or treated with PJ34. (**b**) A schematic presentation indicating the effect of PJ34 on the spindle structure in human cancer cell. In the untreated cancer cell, normal bipolar spindles with clustered NuMA, clustered multi-centrosomes, and aligned chromosomes in the spindle mid-zone. In the PJ-34 treated cancer cell, aberrant microtubules (**green**), aberrant spindle poles, un-clustered NuMA (as indicated), dispersed chromosomes (**blue**) and un-clustered multi-centrosome (**red**), From: Visochek et al., 2017, Oncotarget.

#### **5. The Potency of PJ34 and Other PARP Inhibitors in Preventing Metastases**

PJ34 and additional PARP inhibitors inhibit the activity of matrix metalloproteinase-2 in a range of concentrations higher than that inducing PARP inhibition [97]. The measured IC50 was about 56 μM for PJ34 inhibition of MMP-2 [97], lower than the IC50 values of other tested PARP1 inhibitors [97].

In healthy tissues, matrix metalloproteinases (MMPs) are key enzymes in the development and remodeling of tissues, including in wound healing [98]. However, MMPs are also dominant in tumor angiogenesis, in tumor cells escape from the primary tumor and their dissemination to secondary sites [98,99]. Elevated expression of MMPs, including MMP-2, has been implicated in metastasis [98–100]. Metastases predict high invasive stage of the malignancy, and a poor prognosis [99,100]. MMP inhibitors prevent metastases in a variety of solid cancer types [99,100]. MMP inhibition by PARP inhibitors is their additional advantage in cancer therapy. ABT-888 hindered cancer cell migration rate measured in vitro by the scratch assay [101]. PJ34 at the concentration range inhibiting MMP-2 prevented metastasis in melanoma xenografts [102].

#### **6. Conclusions and Future Perspectives**

The potency of PJ34 to exclusively eradicate human cancer cells without impairing healthy cells can be attributed to anomalies exclusively inserted in the structure of the mitotic spindle of human cancer cells. Blocking the post-translational modification of NuMA and kinesins HSET and kif18A by PJ34 causes these anomalies, which subsequently prevent the alignment of chromosomes bound to kinetochores in the spindle mid-zone. This structural anomaly in the mitotic spindle arrests mitosis and kills the cancer cell in the pre-anaphase stage by mitotic catastrophe cell death.

Notably, the modified phenanthridine PJ34, which has been invented for PARP inhibition, exclusively eradicates a variety of human cancer cells without impairing normal healthy somatic quiescent and proliferating cells. Thus, in spite of the permeability of PJ34 in the cell membrane and its rapid distribution in the animal's tissues, treatment with PJ34 did not impair healthy tissues in the tested animals, nor their development and weight-gain.

On the basis of these findings, we hope that cell death evoked by structural faults in the mitotic spindle of human cancer cells will pave the way to a new concept in cancer therapy. Inserting specific structural anomalies in the mitotic spindle of human cancer cells may specifically eradicate cancer cells while saving healthy cells and physiological functions frequently lost during the currently offered DNA-damaging cancer therapies.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The author declare no conflict of interest.

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Combined PARP Inhibition and Immune Checkpoint Therapy in Solid Tumors**

### **Florent Peyraud 1,2 and Antoine Italiano 1,2,3,\***


Received: 30 April 2020; Accepted: 7 June 2020; Published: 9 June 2020

**Abstract:** Genomic instability is a hallmark of cancer related to DNA damage response (DDR) deficiencies, offering vulnerabilities for targeted treatment. Poly (ADP-ribose) polymerase (PARP) inhibitors (PARPi) interfere with the efficient repair of DNA damage, particularly in tumors with existing defects in DNA repair, and induce synthetic lethality. PARPi are active across a range of tumor types harboring *BRCA* mutations and also *BRCA*-negative cancers, such as ovarian, breast or prostate cancers with homologous recombination deficiencies (HRD). Depending on immune contexture, immune checkpoint inhibitors (ICIs), such as anti-PD1/PD-L1 and anti-CTLA-4, elicit potent antitumor effects and have been approved in various cancers types. Although major breakthroughs have been performed with either PARPi or ICIs alone in multiple cancers, primary or acquired resistance often leads to tumor escape. PARPi-mediated unrepaired DNA damages modulate the tumor immune microenvironment by a range of molecular and cellular mechanisms, such as increasing genomic instability, immune pathway activation, and PD-L1 expression on cancer cells, which might promote responsiveness to ICIs. In this context, PARPi and ICIs represent a rational combination. In this review, we summarize the basic and translational biology supporting the combined strategy. We also detail preclinical results and early data of ongoing clinical trials indicating the synergistic effect of PARPi and ICIs. Moreover, we discuss the limitations and the future direction of the combination.

**Keywords:** PARP inhibitor; DNA damage response; immunotherapy; immune checkpoint inhibitor; PD-1; PD-L1; CTLA-4; combination therapy; solid tumors

#### **1. Introduction**

Over the past decade, poly(ADP-ribose) polymerase (PARP) inhibitors (PARPi) and monoclonal antibodies that block immune checkpoints, such as programmed cell-death 1 (PD-1) and cytotoxic T lymphocyte antigen 4 (CTL-4), have transformed the treatment of multiple types of cancers. Immune checkpoint inhibitors (ICIs) used as stand-alone therapeutic interventions give rise to durable objective responses in patients affected by a variety of cancers and have been approved for an ever-growing list of malignancies, including melanoma, non-small cell lung carcinoma (NSCLC), renal cell carcinoma, head and neck squamous cell carcinoma (HNSCC), and Hodgkin's lymphoma [1–7]. More recently, monotherapy with PARPi as a maintenance strategy showed significant clinical activity in several cancer types harboring germline loss-of-function BRCA mutations such as ovarian, breast and pancreatic cancer [8–11]. However, despite these substantial advancements in clinical care, the majority of patients receiving either PARPi or ICIs alone do not provide benefit and a rationale to combine these treatments has emerged [12,13].

The groundbreaking success of anticancer immunotherapy is primarily based on the features of cancer cells and their ability to potentially initiate an antitumor immune response. These notable features include gene mutations resulting in abnormal protein expression patterns, such as neoantigens or tumor-associated antigens (TAAs) [14]. TAAs represent self-antigens that are aberrantly expressed or overexpressed in tumor cells, whereas neoantigens refer to non-self-antigens arising as a result of somatic mutation [14,15]. The formation of mutation-derived TAAs and neoantigen, reflecting the mutational burden of the tumor, allow the immune system to recognize tumor cell and initiate the cancer-immunity cycle [16,17]. The subsequent antitumor immune process against neoantigens relies on several steps, including the release and presentation of cancer antigens by antigen-presenting cells (APC), priming and activation of T cells, trafficking and infiltration of T cells, and recognizing and killing cancer cells [18]. However, a subset of cancer cells can escape host immune destruction by impairing one or more steps and result in tumor progression [19]. The role of immunotherapy is to reinvigorate antitumor immune response by disrupting co-inhibitory T cell signaling, transferring additional tumor-specific T cells clones and reshaping the immunosuppressive microenvironment [20–23]. Several strategies, including ICIs, adoptive T cell transfer, and vaccination, have been put to use in multiple cancers [24–27]. Nevertheless, due to complex and constantly evolving interactions between cancer cells and the immune system, both primary and acquired resistance with ICIs monotherapy are observed [28,29]. Therefore, combination treatment with ICIs is an attractive strategy to potentiate efficacy and lower resistance.

Recent molecular profiling of DNA damage repair genes has allowed the implementation of novel therapeutic strategies. By interfering with efficient DNA damage repair, the inhibition of PARP that target the base excision repair (BER) pathway leads to insufficient DNA repair, with subsequent unsustainable DNA damage, and thus represents a synthetic lethal therapeutic approach for the treatment of cancers with compromised ability to repair double-strand DNA breaks by homologous recombination (HR), including those with defects in *BRCA1*/*2* [30]. The unrepaired-DNA promotes immune priming through a range of molecular mechanisms and also leads to adaptative upregulation of programmed death ligand 1 (PD-L1) expression [31]. Moreover, PARPi modulates the inflammatory immune microenvironment of tumors and reinstates a productive TH1 immune response [31]. This multifaceted immunological effect of PARPi might be favorable to boost an antitumor immune response and enhance the efficacy of ICIs. In this review, we summarize the basic and translational biology supporting the combined strategy and provide a focus on preclinical studies and ongoing clinical trials of ICIs combined with PARPi, as well as perspectives and potential challenges of this combination strategy.

#### **2. DNA Damage and PARP Inhibition**

#### *2.1. Role of PARP in DNA Damage Response*

Cells are continuously faced with endogenous and exogenous stress that can ultimately lead to DNA damage. To preserve genomic integrity and prevent emergence of cancer, detection and repair of DNA is a critical process, managed by multiple pathways [32]. DNA single-strand break (SSB) damage is fixed by three main pathways: (1) BER, (2) nucleotide excision repair (NER), and (3) mismatch-repair (MMR). Possibly more dangerous DNA double-strand breaks (DSB) are restored by two additional pathways: (1) HR and (2) non-homologous end joining (NHEJ) [33]. Anomalies observed in DNA damage response (DDR) key genes, such as BRCA1/2 or TP53, are associated with cancer-prone phenotypes [34]. As a consequence, failure in DDR in an accurate and well-timed manner can result in the defective elimination of genome mutations and increases the risk of oncogenesis after established DNA damage events [35]. Depending on the context, cancer cells often harbor a lessened repertoire of DDR signaling competences, rendering them more reliant on a subset of DNA repair pathways and therefore more susceptible to DDR inhibition than normal cells [36].

PARP1/2 enzymes are core DNA-damage sensor and signal transducer in DDR, which bind to DNA breaks and catalyze the synthesis of poly(ADP-ribose)(PAR) chains on target proteins (PARylation) in the vicinity of the DNA break and itself (autoPARylation) [37]. These negatively charged PAR chains promote chromatin remodeling, recruit DNA repair-related protein complex and affect the

replication fork progression speed [38,39]. The binding of PARP1 via zinc finger domains to sites of DNA-damage carries a conformational change in the PARP1 proteins and relieves the autoinhibitory interaction between the catalytic domain and helical domain. Then, the PARP1 co-factor nicotinamide (β-NAD+) is used as a substrate at the active site of the enzyme to catalyze the transfer of ADP-ribose moieties onto target proteins. The synthesis of ADP-ribose polymeric chains on proteins in the vicinity of DNA breaks, called PARylation, likely mediates DNA repair by modifying chromatin structure and by localizing DNA repair effectors [40]. Thereafter DNA-damage restore, autoPARylation occur that rapidly dissociate PARP from damage site [41]. The role of PARP has been well identified in BER-mediated SSB repair pathways, as well as other DDR pathways [42].

#### *2.2. The Lethal Synthetic E*ff*ect of PARP Inhibitors*

#### 2.2.1. Mechanism of Action of PARPi

Although the precise mechanism by which PARPi kill tumor cells remains to be fully clarified, the anticancer effect is attributed to catalytic inhibition of PARP that block repair of DNA SSB [43] (Figure 1a). While PARPi is well-tolerated by normal cells, this effect of PARPi is more likely observed in tumor cells with a *BRCA*-deficient background [43]. As a result of defective enzymatic function induced by PARPi, the accumulation of SSB is subsequently encountered by replication forks and generates potentially lethal DSBs that need to be fixed [43,44]. In normal cells, the accumulation of DSBs are repaired preferentially by HR rather than NHEJ [45]. HR is a high-fidelity repair pathway that utilizes the sister copy of the damaged DNA as a template, leading to the reconstitution of the original sequence [46]. In contrast, NHEJ is intrinsically error-prone, modifies the broken DNA ends, and ligates them together with little or no homology, generating deletions or insertions [47]. However, in some cancer cells lacking BRCA1 or BRCA2, two key tumor suppressor proteins involved in DSB repair by HR, loss of PARP function leads to the accumulation of DSBs that are unrepaired or unsustainably repaired by NHEJ which results in cell death [44,48]. Based on the discovery of this synthetic lethality between *BRCA* and PARP, numerous PARPi have been developed, including olaparib, rucaparib, niraparib, talazoparib, and veliparib, which are mainly applied in cancer patients with *BRCA1*/*2* mutations [8–11,42].

Although the greatest efficacy of PARPi has been observed in tumors with *BRCA1*/*2* mutations, accumulating data indicate that synthetic lethality is inadequate to explain the whole antitumor activity. First, the ability of PARPi to inhibit PARP catalytic activity is poorly correlated to its cell-killing ability in HR deficiencies (HRD) cells [43,49]. In addition, PARPi induces cytotoxicity to a greater extent than PARP depletion [50,51]. Furthermore, PARP itself is essential to the cytotoxic effects of PARPi [52]. Actually, these facts may be attributed in part to the PARP trapping potency of PARPi (Figure 1b). Although the precise mechanisms of PARP trapping remains unclear, it has been proposed that PARPi could either prevent the release of PARP1 from DNA by inhibiting autoPARylation [53]. Likewise, PARPi binding to the catalytic site could cause allosteric changes in the PARP1 structure enhancing DNA avidity [49]. The trapping DNA-PARP complex stalls the progress of replication fork and elicit cytotoxic effects primarily through the conversion of unrepaired SSBs into lethal DSBs [43,49]. Moreover, PARPi could also act via the upregulation of NHEJ pathway, which presumably leads to genomic instability and eventual lethality [54]. Finally, PARPi could suppress the role of PARP in reactivating DNA replication forks and cause cell death [43]. Additional studies further demonstrated that loss of other tumor suppressor DNA repair proteins, many of which are involved in HR, such as RAD51, ataxia telengiectasia Rad3-related (ATR), ataxia telangiectasia mutated (ATM), checkpoint kinase 1 (CHK1), checkpoint kinase 2 (CHK2), and partner and localizer of *BRCA2* (PALB2) also caused sensitization to PARPi [49,55]. These results suggested that PARPi might be a useful therapeutic strategy not only for the treatment of *BRCA*-mutated tumors but also for the treatment of a wider range of non-*BRCA*-mutated tumors that are inherently HR deficient (HRD) or "BRCAness/HRDness" [56].

**Figure 1.** Mechanisms of action of PARP inhibitors (PARPi): (**a**) PARPi impedes PARP enzyme activity (or catalytic inhibition) and interferes with repair of single strand breaks (SSB) by disrupting the base excision repair (BER) pathway; (**b**) PARPi also causes trapping of PARP proteins on DNA by inhibiting autoPARylation. In homologous recombination (HR) proficient normal cells, DNA is repair and cell survive. The result in unresolved DNA double strand breaks (DSB) in HR deficient cells leads to cell death.

#### 2.2.2. Clinical Applications of PARPi

The early development of PARPi focused initially on their use in combination with cytotoxic chemotherapy agents and radio-sensitizing drugs, but this was rapidly rejected because of excessive toxicity [57,58]. The potent antitumor effect of PARPi was originally observed in tumors harboring germline *BRCA1*/*2* mutations (*gBRCA1*/*2*m), such as familial breast and ovarian cancer [59]. This rapid translation of preclinical studies into promising clinical data triggered the development of several PARPi in different tumors types. Initially, PARPi in the clinic improved clinical benefits for germline or somatic *BRCA*-deficient ovarian cancer [60,61]. Subsequently, breast, pancreatic and prostate cancers that harbors defects in *BRCA* also demonstrated to be PARPi responsive [9–11,60–62]. More recently, it has been suggested that patients without *BRCA* mutations shared therapeutic vulnerabilities, especially tumors with deficiencies in HR. Indeed, the activity of PARPi is based on the concept of synthetic lethality, where an underlying HRD in tumor cells makes the cells highly susceptible to PARP inhibition [42]. This hypothesis has been further confirmed with multiple clinical studies showing that sensitivity to PARPi occurs in tumors beyond those with *BRCA* mutations, especially in HRD-positive tumors [63–67]. To date, five PARPi have been approved or orphan drug designed by the FDA (veliparib, rucaparib, talazoparib, niraparib, olaparib) and applied in clinical practice [42].

Despite the advances of PARPi in a particular population, acquired resistance is a common clinical phenotype. Owing to extensive preclinical studies, several resistance mechanisms have been identified that can be classified into four main categories. Firstly, numerous different mechanisms result in the reactivation of HR function. For example, secondary reversion mutations in several key HR repair (HRR) genes, such as *BRCA1*/*2*, *RAD51C*/*D,* and *PALB2*, restore the open reading frame and thus HR competency [68]. Moreover, the loss of p53-binding protein 1 (53BP1), a protein promoting NHEJ, is associated with PARPi resistance by recovery of HRR in *BRCA1*-deficient tumors [69]. By directly impacting the activity and abundance of PAR chains that decreased PARP trapping, mutations in DNA-binding domains of PARP1 and mechanisms that increase PARylation of PARP1 could also lead to PARPi resistance [70,71]. Furthermore, the cellular availability of the inhibitor is a critical step for successful therapy, as illustrated by the upregulation of ATP-binding cassette (ABC) transporters, such as the P-glycoprotein (PgP) efflux pump that have been described to reduce the efficacy of PARPi [72]. At last, restoration of replication fork protection that induces the stabilization of stalled forks may lead to PARPi resistance [73]. Indeed, fork degradation induced by PARPi is mediated by PTIP and EZH2 proteins, which upon loss lead to protection of the fork from nucleases and thereby resistance [74,75].

Intense preclinical and clinical research are ongoing in order to broadening responding patients, overcoming acquired resistance and enhancing the efficacy of PARPi [73]. The development of combination therapy encompassing PARPi is a potential approach to address these objectives. In addition to the hypothesis that patients with HRD tumors are more prone to produce neoantigen and exhibit higher mutational load, there is a preclinical rational suggesting that PARPi may promote the formation of neoantigen and generate tumor cell recognition by the immune system, making this class of drugs a potential partner for combination with ICIs [76,77].

#### **3. The Revolution of Cancer Immunotherapy and Immune Checkpoint Inhibitors**

Immunotherapy is proving to be an effective therapeutic approach in a variety of cancers [78]. In the last decade, the use of therapeutic antibodies that disrupt negative immune regulatory checkpoints and unleash pre-existing antitumor immune responses have achieved impressive clinical successes. Among the different types of cancer immunotherapy, ICIs have demonstrated the broadest impact by leveraging the cytotoxic potential of the human immune system, especially tumor-specific cytotoxic T cells. The role of T cells is critical to adaptative immunity and contribute to improved outcomes in a large range of cancers [79]. The activation of naïve T cells requires two distinct signals [80]. The generation of the first signal occurs by binding of major histocompatibility complex (MHC)-presented immunogenic peptide antigen to the heterodimeric T cell receptor (TCR). The transduction of the second signal, also referred as co-stimulation signal, arises through ligation of the T cell co-stimulatory surface receptor CD28 to its ligand CD80 (also known as B7-1) or CD86 (also known as B7-2) on the surface of professional antigen-presenting cells (APCs). Subsequent to both these signals, activated T cells begin to express co-inhibitory cell surface receptors that control T cell function, such as CTLA-4 and PD-1. The balance between co-stimulatory and co-inhibitory signals is crucial for the activation and tolerance of T cells [81]. Importantly, targeting these co-inhibitory pathways with ICIs in the context of cancer effectively shifts that balance toward activation, thereby overcoming tumor immune subversion [82].

#### *3.1. CD80*/*86-CTLA-4 Signaling Pathway*

The first negative regulator of T cell activation to be identified was CTLA-4, a co-inhibitory receptor that is constrictively expressed on Tregs and transiently upregulated during the course of T cell activation in peripheral lymphatic organs [83]. Bound by the same ligands (CD80/86) that provide co-stimulatory signals through CD28 but with higher affinity, CTLA-4 mainly impedes acquisition of T cell effector function by mediating transendocytosis and degradation of the ligands [84]. In addition, CTLA-4 delivers inhibitory signals that block T cell proliferation and secretion of IL-2, leading to T cell tolerance through induction of energy [85,86]. Moreover, CTLA-4 counterbalance TCR/CD3-mediated phosphorylation through immunoreceptor tyrosine-based inhibitory motif (ITIM) and impede the signal transduction of TCR [87]. Therefore, CTLA-4 engagement in numerous T lymphocyte populations operates as a cardinal immune checkpoint that ultimately hampers the acquisition of T cell effector functions and dampens the antitumor immune response.

#### *3.2. PD-1*/*PD-L1 Signaling Pathway*

The expression of PD-1 on activated immune cells is ubiquitous, including T cells, B cells, natural killer (NK) cells, and dendritic cells (DC), and yields inhibitory signals through binding of its two ligands, namely PD-L1 and PD-L2 [88]. Moreover, both PD-1 ligands are expressed on a wide variety of immune and non-immune cells [88]. More particularly, PD-L1 is found on a broad range of tissues and could be upregulated under inflammatory conditions such as cancers [89]. The expression of PD-L1 on the surface of tumors underlies the crucial relevance of the PD-1/PD-L1 pathway to neoplasm. Upon binding of TCR with antigen presented by MHC, PD-1 is engaged with its ligand and becomes functional. PD-1 activation leads to phosphorylation of the ITIM and immunoreceptor tyrosine-based switch motif (ITSM) in the PD-1 cytoplasmic tail and subsequently drive the recruitment of protein tyrosine phosphatase, such as Src homology region 2 domain containing phosphatase 1/2 (SHP1/2). As a consequence of dephosphorylation, these phosphatases antagonize positive signals that occur through the TCR nad CD28, affecting downstream signaling pathways. For example, TCR signaling molecules, such as Lck and ZAP-70, and co-stimulatory signaling cascades, such as PI3K-Akt-mTOR and Ras-MEK-ERK pathways, are inhibited. The impairment of these crucial signaling pathways alters the activation, proliferation, survival, cytokine production, metabolism, and epigenetic programs in T cells [90–92]. Tumors can exploit this pathway to escape T cell-mediated tumor-specific immunity.

#### *3.3. Clinical Application of ICI*

Since the recent success of antibodies targeting checkpoint molecules CTLA-4, PD-1, and PD-L1, the field of cancer immunotherapy has been experiencing a renaissance. The anti-CTL-4 inhibitor ipilimumab was the first ICI to obtain approval in 2011 for the treatment of metastatic melanoma [1,93]. Thereafter, ICIs have yielded broad clinical activity, leading to regulatory approval of several monoclonal antibodies in a variety of advanced and up-front disease settings, including melanoma, non-small cell lung, renal cell carcinoma, urothelial carcinoma, HNSCC, Merkel cell carcinoma, gastric carcinoma, hepatocellular carcinoma, Hodgkin's lymphoma, as well as for any MMR-deficient/microsatellite instability (MSI) positive tumors [2–6,94–105]. However, as with PARPi, only a subset of patients derives benefit and a series of biomarkers have been developed to predict efficacy of ICI and select patients before treatment beginning.

Although the current understanding of the clinical response of ICI therapy indicates that there cannot be a single predictive biomarker, several factors have been identified as the core determinants of the efficacy of ICIs, such as tumor mutation burden (TMB) and particular mutational signature, the number of tumor-infiltrating lymphocytes (TILs), PD-L1 expression, immunosuppressive microenvironment, and MMR deficiency (MMRd) [106]. For example, tumors that harbor MMRd or some specific defects in DDR pathways beyond MMR demonstrated a higher ICI response [103,104,107]. The improve outcome in these patients is believed to be a result of increased mutational load, leading to greater immunogenicity. In addition, a novel perspective has arisen with the development of ICI-based combination therapy in order to improve ICI efficacy and overcome resistance. These include combinations with other checkpoint inhibitors, radiation therapy, chemotherapy, and targeted therapies, so as to foster antigen presentation, broadening T cell repertoire, impairing immunosuppressive elements, and increasing antitumor immune response [108,109].

#### **4. Combination of PARPi and ICI Therapy**

#### *4.1. A Rational to Combine PARPi and ICI*

#### 4.1.1. Tumor Mutation Burden and Neoantigen

The mutational load in a tumor, termed as TMB and determined by the number of non-synonymous single nucleotide variants (nsSNVs), may impact the odds of generating immunogenic peptides and has been significantly correlated with ICI response in previous studies [110–112]. Even if the optimal TMB cut-off remains blurred across tumor types, the relationship with efficacy of ICI is robust [113,114]. TMB is considered as a surrogate of neoantigen load which predicts the therapeutic response of ICI [14,15,115]. Likewise, a growing amount of data indicate a closely association between TMB and DDR deficiency [116]. Highly mutated tumors often exhibit one or several mutations in key

components of DDR or replicative pathways, including *MSH2* for MMR, *BRCA1*/*2* for HR and *POLE* for DNA replication, and correlate with ICI response [116]. In addition, patients with cancer harboring innate deficiencies in DDR genes, including MMR and HRR genes, achieved durable benefit from ICIs compared with patients without these deficiencies [116,117]. These results suggest that loss of normal DNA repair fidelity, such as DDR phenotype, may contribute to increased mutational load and neoantigens burden which affect the response to immunotherapy in these tumors. One relevant strategy in patients with HRD or others defects in DDR would be to combine PARPi and ICI.

Direct evidence that the targeting of DSB repair proteins with DDR inhibitors provoked and increased TMB is only beginning to emerge [118]. By affecting the HR pathway in tumor cells, impaired DNA repair induced by PARPi could subsequently generate catastrophic DNA damage that would increase the neoantigen load and TMB, thus driving a response to ICI and theoretically broadening the responding population (Figure 2a). Although tumors with non-MMR DDR genes deficiency, such as *BRCA1*/2 and other HR-related genes, have increased TMB, the association is weaker than that observed in MMR deficiency. Thereby, other fundamental links in tumor immunogenicity may be involved to explain the higher response rate to ICI in these patients [119,120].

**Figure 2.** Biological effect of PARPi on cancer cells. (**a**) PARPi delays DNA repair that generates double-strand breaks (DSB) and catastrophic unrepaired DNA damages in tumor cells, increasing neoantigens load and tumor mutation burden; (**b**) DSB induced by PARPi in tumor cells produce double-strand DNA (dsDNA) fragments that, through cyclic GMP-AMP synthase (cGAS) binding, lead to stimulator or interferon genes (STING) activation and the generation of a type I interferon (IFN) response. This upregulates chemokines CCL5 and CXCL10 leading to T cell recruitment; (**c**) Upregulation of PD-L1 via STING pathway lead to T cell exhaustion.

#### 4.1.2. DNA Damages and cGAS-STING Pathway

Aside from TMB, an emerging body of evidence supports a role for non-neoantigen-based mechanisms of tumor cell recognition and targeting by the host immune system. Genomic instability in tumor cells leads to the accumulation of incompletely repaired DNA damage, generating tumor-derived double-strand DNA (dsDNA) in the cytoplasm [121]. The sensing of tumor-derived dsDNA by cytosolic DNA sensor cyclic GMP-AMP synthase (cGAS) plays a major role in the activation of the stimulator of interferon (IFN) genes (STING) signaling pathway [122]. After the recognition of tumor-derived DNA within the cytosol, cGAS activates STING via the generation of 2 -5 cyclic GMP-AMP (cGAMP). In turn, STING prompts phosphorylation and nuclear translocation of type I IFN transcriptional regulatory factors TANK-binding kinase 1 (TBK1) and IFN regulatory factor 3 (IRF3) [123,124]. Moreover, STING activates NF-κB pathway which cooperates with IRF3. As a result, the upregulation of type I IFN promotes systemic immune response and regulates multiples components in anticancer immunity, especially T cells, NK cells and DCs [125]. According to recent studies, DNA damages and DDR deficiencies induce the activation of STING and NF-κB pathways, leading to inflammation and infiltration of tumors by immune cells across multiple types of cancers, a prerequesite of tumor-killing effect of ICI [126–130].

In clinical practice, the antitumor activity of PARPi has been observed in patients with platinum-sensitive tumors regardless of BRCA1/2 mutation or HRD status, suggesting an alternative mechanism unrelated to conventional lethal synthetic-mediated cytotoxic effects [60]. The use of PARPi treatment leads to unresolved DNA lesions and to the production of cytosolic dsDNA fragments. The accumulation of cytosolic DNA activates in turn the DNA sensing cGAS-STING pathway and boosts production of type I interferon to induce antitumor immunity independently of DNA repair deficiency [131–133] (Figure 2b). These critical changes amplify STING signaling and its associated-transcription programs, thereby promoting TILs and antitumor immunity [132,133]. Moreover, it leads to increased levels of chemokines, such as CXCL10 and CCL5, that induce the activation and function of cytotoxic CD8<sup>+</sup> T cell [132,133]. In addition, these effects of PARPi are further enhanced by ICI, providing a mechanistic rationale for the use of PARPi as immunomodulatory agents to harness the therapeutic efficacy of immunotherapy [134].

#### 4.1.3. PD-L1 Upregulation by PARPi

A key mechanism underlying cancer immune evasion is the expression of inhibitory ligands, notably PD-L1, on the surface of cancer cells. Despite the approval of PD-L1 expression on tumor cells as a companion diagnostic for anti-PD1 therapy for patients with NSCLC, it remains an imperfect predictor of ICI response [4,135,136]. Via STING pathway, tumor-associated inflammation mainly drives the upregulation of immunosuppressive PD-L1 expression, thus reflecting the status of tumor immune microenvironment [137] (Figure 2c). In addition, defects in BRCA1/2 correlates to higher levels of PD-L1 expression [126,138]. Furthermore, the serine/threonine protein kinase glycogen synthase kinase 3β (GSK3β), a regulator of glycogen metabolism, interacts with PD-L1 and modulates its expression by inducing proteasome degradation of PD-L1 [139]. Based on the latter observation, preclinical models have unveiled that PARPi upregulates PD-L1 expression primarily through GSK3β inactivation in a dose-dependent manner, suppressing T-cell activation and increasing tumor cell killing [76]. Further explorations have shown that targeting DDR proteins PARP with PARPi significantly increased expression of PD-L1 [133]. Another report has demonstrated that PARPi-induced DSBs upregulate PD-L1 by ATM-ATR-CHK1 pathway independently of the IFN pathway [140]. Interestingly, subsequent combination therapy with ICI induced PARPi sensitization and led to a greater antitumor activity than either drug alone, putting forward a rational for combining PARPi with ICI as a useful therapeutic strategy [76].

#### 4.1.4. Reprogramming of Immune Microenvironnement by PARPi

In addition to altering the intrinsic immunogenicity of tumor cells through modulation of surface phenotype and intracellular pathways, DNA damage and deficient DDR pathways also modify the extrinsic immunogenicity of tumors at the level of microenvironment. As aforementioned, tumors with existing defects in DNA repair promote inflammation and TH1 immune response through a range of molecular mechanisms, leading to extrinsic tumor suppression [19]. However, despite the ability of DNA damage to contribute to tumor immune elimination, sustaining low-level DNA damage continues to foster inflammatory signaling that stimulates the infiltration by suppressive immune cells, like myeloid-derived suppressor cells (MDSCs) or tumor-associated macrophages (TAMs), which leads to further DNA damage via free radical release. This transformation boosts chronic inflammation, immunosuppression, and cancer progression [141,142]. PARPi may have the potential to shift from chronic, low level, DNA damage to more significant TH1 immune response and create a more susceptible tumor microenvironment [143]. Nevertheless, the self-sustaining cycle of DNA damage and chronic inflammation, which is challenging to overcome with PARPi single therapeutic approach, could potentially be addressed through combination with ICIs.

In the wake of these biological findings, deficiencies in the HRR pathway and/or the use of DDR agents such as PARPi appear to activate immunosuppressive pathways, thus offering targetable immunological vulnerabilities in tumors. The interaction between DDR and immune response provides the basis of the combination therapy of ICI and PARPi. Thereby, combining ICIs and PARPi, which target HRD or induce a state of "BRCAness" in HR-proficient tumors, is a thrilling strategy, particulary as these agents have distinct and mostly non-overlapping toxicities [144,145]. Based on these assumptions, combining PARP inhibition with agents that have complementary mechanisms, such as ICIs, is currently subject to clinical testing.

#### *4.2. Preclinical Data and Clinical Studies*

#### 4.2.1. Combination of PARPi with Anti-PD1/PD-L1 ICIs

The first preclinical study evaluating PARPi veliparib in combination with anti-PD1/PD-L1 in the BRCA1-deficient BR5 mouse ovarian cancer model observed no significant boost in T cell activity and no improvement in survival [146]. However, the disappointing lack of activity for the combination with anti-PD1/PD-L1 in this preliminary work contrasts with more recent studies. In other preclinical studies conducted on breast cancer cell lines and xenograft models, PARPi olaparib significantly upregulated PD-L1 expression independently of cGAS-STING-IFN pathway and decreased antitumor immunity by attenuating the cell-killing activity of activated human peripheral blood mononuclear cells [76]. Further investigation in EMT6 syngeneic murine models demonstrated more potent antitumor effect and higher TILs infiltration with combined PARPi olaparib with anti-PD-L1 compared to either therapy alone [76]. In another mice model bearing BRCA1-deficient ovarian tumors, anti-PD1 monotherapy exhibited a non-significant effect and PARPi monotherapy delayed tumor progression compared to control group, whereas combination therapy significantly slowed the tumor growth and prolonged survival time [147]. A recent report indicated that coadministration of PARPi niraparib and anti-PD-1 enhanced the infiltration of immune cells into tumor microenvironment and increased synergistic antitumor activities in both immunocompetent BRCA-proficient and BRCA-deficient models, including breast cancer, lung squamous cell carcinoma, colon adenocarcinoma, bladder cancer, and sarcoma [148]. Similarly, additionally to augment CD8<sup>+</sup> T cell infiltration, the association of PARPi olaparib and PD-L1 blockade induced complete tumor regression in multiple immunocompetent small cell lung carcinoma (SCLC) mice models [133]. These contrasting results may be explained by the use of different models, with disparities in immune contexture. While differences in anti-PD1/PD-L1 activities cannot be excluded, the use of PARPi with differential catalytic inhibition and various PARP trapping potencies may also explicate these discrepancies [52]. Taken together, the available translational and preclinical data clearly support the combination of PARPi and ICI.

Based on these encouraging preclinical studies, several clinical trials have been conducted and some data are available to date (Table 1). In metastatic castration-resistant prostate cancer (mCRPC), the combination therapy of olaparib and durvalumab induced PSA responses (reduction ≥ 50%) in eight out of 17 patients (47%)(NCT02484404) [149]. Patients with DDR mutations exhibited greater benefit than those without known alterations (12-month progression-free survival probability of DDR-deficient vs. DDR-proficient, 83.3% vs. 36.4%, *p* = 0.031), suggesting DDR-deficient as a predictive biomarker of response [149,150]. In heavily pretreated platinum-resistant recurrent ovarian cancer, durvalumab and olaparib had clinical activity, irrespective of BRCA mutation status (NCT02484404) [151]. Interestingly, correlative analysis of paired pre- and on-therapy fresh core biopsy and blood samples collected on the latter cohort of recurrent ovarian cancer found that

treatment enhanced IFN-γ and CXCL9/CXCL10 expression, systemic IFN-γ/TNF-α production and TILs, creating a more immunostimulatory milieu [152]. While tumoral and peripheral IFN-γ increases was correlated with durable clinical benefit from combined therapy, elevated circulating VEGFR3 levels were associated with worse progression-free survival (PFS), suggesting that VEGF/VEGFR pathway may act to counterbalance immunostimulatory changes induced by PARPi and would serve as a target to further improve efficacy of the combination [152]. Despite major limitations surrounding this exploratory analysis, no significant changes in TMB, PD-L1 or STING expression were noted in ovarian cancer patients treated with olaparib and durvalumab combination, thus warranting further investigation on the underlying biological mechanism [152]. On the other hand, the results of relapsed SCLC cohort of phase 2 NCT02484404 basket study durvalumab in combination with olaparib did not meet the preset bar for efficacy [153]. The preexisting TILs level, assessed by immunohistochemistry, seemed to predict tumor responses, suggesting a contribution from an immune-mediated response. The inflamed-phenotype at baseline, defined by high TILs infiltration, may help to identify patients who are most likely to respond to ICIs [153].

The phase 2 MEDIOLA basket trial assessed the efficacy and safety of chemo-free combination of olaparib and durvalumab in patients with solid tumors, including ovarian cancer, breast cancer and gastric cancer (NCT02734004). In gBRCAm platinum-sensitive relapsed ovarian cancer, the effect of the latter combination demonstrated an overall response rate (ORR) of 63% and a 12-week disease control rate (DCR) of 81% [154]. In gBRCAm HER2 negative metastatic breast cancer, the DCR was 80% at 12 weeks and 50% at 28 weeks, with ORR of 63% [155]. Median PFS (mPFS) was 9.2 months and median overall survival (mOS) was 21.5 months [155]. Moreover, patients with no prior line of chemotherapy had higher ORR and longer OS than those with two prior lines (respectively 78% vs. 50% for ORR and 21.3 vs. 16.9 months for OS) [155]. In platinum-resistant relapsed gastric cancer, the ORR was 10% with the 12-week DCR was 26% [156].

In the phase 2 TOPACIO trial (NCT02657889), niraparib and prembrolizumab combination therapy has demonstrated clinical benefit in platinum-resistant ovarian cancers and triple-negative breast cancers, with numerically higher response rates in those with BRCA-mutated tumors only in breast cancer cohort (ORR of BRCAm vs. BRCA wild-type in breast cancer cohort, 47% vs. 11%) [157,158]. However, the ovarian cohort of the TOPACIO study did not meet its primary endpoint of ORR.

The phase 1a/b PARPi pamiparib combined with anti-PD1 tislelizumab in patients with advanced solid tumors was associated with antitumour responses and clinical benefit (ORR of 20%)(NCT02660034) [159]. Similarly, the phase 1b/2 JAVELIN PARP Medley (NCT03330405) of avelumab plus talazoparib in advanced solid tumors is ongoing but showed preliminary antitumor activity and a manageable safety profile [160].

In the recent phase 2 NEODURVARIB trial (NCT03534492), durvalumab plus olaparib administered prior surgery of resectable muscle-invasive bladder cancer (MIBC) demonstrated a pathological complete response rate of 44.5%, suggesting an active and well-tolerated neoadjuvant strategy [161].

Except for the pamiparib–tislelizumab association, where an increased rate of immune-related hepatitis was observed, all combinations were well tolerated, with toxicities in line with those detected for the relevant drugs in monotherapy settings [159]. Numerous clinical trials are ongoing in a broad range of cancers that will help to decipher the exact role of PARPi with anti-PD1/PD-L1 combination strategy (Table 2).


Acceptable toxicity

 **1.** Recent clinical trials of combination of PARPi and ICIs.

**Table**



poly(ADP)-ribose

survival; RR: response rate according to RECIST v1.1, SD: stable disease; SCLC: small-cell lung carcinoma.

 polymerase inhibitor; PR: partial response; Q2W: every 2 weeks; Q3W: every 3 weeks; Q4W: every 4 weeks; QD: once daily; rPFS: median

radiographic

progression-free


*Cancers* **2020** , *12*, 1502


**PARPi**






#### 4.2.2. Combination of PARPi with Anti-CTLA-4 ICIs

Contrary to the in-depth attention paid to anti-PD1/PD-L1, the association of PARPi with anti-CTLA-4 is less studied, probably due to the misunderstood biological effect of PARPi on CTLA-4 signaling pathway and T cell effector functions. However, a previous study has unveiled that increased tumor immunogenicity modulates the response to CTLA-4 blockade [110]. Furthermore, BRCA dysfunction is associated with increased T cell recruitment to tumour site and higher expression of immune response genes [162–164]. Moreover, targeting DDR proteins through PARP inhibition may stimulate antigen presentation and immunogenicity via increased TMB and T cell cytotoxic activity [118]. Hence, one might surmise that tumors harboring BRCA dysfunction and treated with PARPi could increase tumoral immunogenicity, thus sensitizing the tumor to anti-CTLA-4 antibodies. All together, these data provide a rationale for the combination of PARPi with anti-CTLA-4 in BRCA-deficient tumours.

Initial preclinical study conducted on an immunocompetent BRCA1-deficient BR5 murine ovarian cancer model revealed that anti-CTLA-4 combined with PARPi veliparib enhanced IFN-γ production and effector/memory T cell infiltration [146]. In addition, CTLA-4 antibody synergized therapeutically with the PARPi, resulting in long-term survival in a majority of mice [146]. To date, no other preclinical study employing this combination in solid tumors has been released, and based on these data, clinical studies were developed.

Preliminary results from a phase 1 study combining olaparib and tremelimumab for the treatment of women with BRCA-deficient recurrent ovarian cancer demonstrated evidence of therapeutic effect with acceptable tolerability (NCT02571725) [165]. Ongoing clinical trials will help to figure out the promising antitumor activity of PARPi with anti-CTLA-4 monoclonal antibodies (Table 2).

#### 4.2.3. Combination of ICI with Others DDR Inhibitors: Moving beyond PARP in Targeting the DDR

In light of the evidence that unrepaired DNA damage induced by PARPi expands the anti-tumor activity of the ICI, the therapeutic landscape of DDR-targeting agents has promptly unfolded to include inhibitors of other key mediators implied in DNA replication and repair, such as ATM, ATR, Chk1, Chk2, DNA-PK, and WEE1 [166]. The crucial roles of ATR and ATM protein kinases in DDR signaling involve the maintenance of replication fork stability and the regulation of cell cycle control checkpoints by operating together via downstream targets Chk1 and Chk2, respectively [167,168]. Additionally, the kinase activity of DNA-PK is required for NHEJ and a distinct nuclear kinase WEE1 controls mitotic entry as well as nucleotide pools in coordination with DNA damage response, making these kinase as potential targets for cancer therapy [168,169].

The role of DDR inhibitors as immunomodulatory agents that possibly potentiate ICIs has recently emerged. Recent preclinical evidence suggested that ATR or ATR inhibitors exhibit immunomodulatory functions and enhances antitumor efficacy to immune checkpoint therapy. The combination of selective ATR inhibitor, avelumab, and platinum-based chemotherapy resulted in antitumor effect in syngeneic tumor models, leading to overall survival benefit compared to any dual-combination group, and also provided protective antitumor immunity with immunological memory in cured mice [170]. Likewise, a recent study demonstrated that pharmacological ATM inhibition induced a type I IFN-mediated innate immune response in pancreatic cancer model that is further enhanced by radiation and led to increased sensitivity to anti-PD1 therapy [171]. Moreover, a preclinical model of immunocompetent SCLC in vivo observed that Chk1 inhibition, a protein kinase implicated in DSB repair, potentiated the antitumor effect of PD-L1 blockade and augmented cytotoxic T cell infiltration [133]. Moreover, a potent and selective DNA-PK inhibitor, that selectively blocks the NHEJ for repair of DSB, induced an immunomodulatory phenotype and elevated the expression of PD-L1 protein via cGAS-STING pathway activation in irradiated p53-mutant cancer cells [172]. Concordantly, combination of DNA-PK inhibitor, radiotherapy and avelumab in syngeneic mice with p53-mutant cancer cells demonstrated a superior benefit and offers a new approach to combination radio-immunotherapy of cancer [172]. All this evidence together provides a clear rationale to combine other DDR pathway inhibitor with immunotherapies.

Outside of PARPi, other DDR inhibitors are currently clinically tested in combination with ICI in several tumor types. A phase 1 modular study of ceralasertib, a potent and selective ATR inhibitor in combination with durvalumab is being evaluated in patients with advanced or metastatic cancers, including NSCLC and HNSCC (NCT02264678) [173]. Preliminary results of this study indicated acceptable tolerance with signals of activity [173]. Similarly, in the phase 1b BISCAY study, patients with metastatic MIBC with any HRD detected are being treated with durvalumab and olaparib or the WEE1 inhibitor adavosertib (NCT02546661) [174]. These ongoing trials will provide new insights into a combinatorial approach.

#### *4.3. Future Perspectives*

Although preclinical and clinical studies revealed interesting tumor responses with PARPi and ICIs in different tumors types, these combinations did not markedly improved antitumor effect compared to the individual agents alone. These disappointing outcomes suggest a lack of synergistic interaction of PARPi and ICIs. Several points could eventually explain these discrepancies.

Foremost, current animal models probably do not recapitulate the whole genomic heterogeneity or tumor microenvironment of human cancer. Indeed, data in mouse models do not predict response to the combination of PARPi and ICIs, thus limiting the transfer to Human. Better in vitro and in vivo models are needed to translate preclinical findings into clinical results. For example, the use of humanized mouse models could be a relevant strategy.

Concerning the nature and the magnitude of combination versus monotherapy benefit, most studies to date have relied on early endpoints such as ORR or DCR. The latter endpoints would be informative in case of patients with limited expected response to PARPi, such as tumors with DDR-proficient. However, in tumors where a high response to PARPi is expected, such as BRCA-mutated and other HRD phenotypes, it would be more relevant to assess the combination benefit in terms of the duration of response or survival, thus necessitating prolonged monitoring in such studies.

One another limitation in the interpretation of available data are the format of clinical trials. Indeed, current clinical results are only provided by non-randomized trials, which only allow cross-trial study comparisons. This approach of comparison is not methodologically and statistically acceptable to distinguish the specific role of each drug in terms of efficacy. To overcome this problem in the clinic, treatment strategies using DDR inhibitors with ICIs should be optimized through the use of randomized controlled multi arms phase III trials designed to enable the interpretation of the effect of each agent alone or in combination. Moreover, additional effort is required to determine the dose and schedule dependency of DNA repair modulation on the immune system.

Based on the synthetic lethality effect, the use of PARPi have been approved preferentially in tumors that harbor deficiency in the HR pathway, such as BRCA-mutated tumors and in a subsets of BRCA-negative HRD-positive cancer [175]. The assumptions that BRCA dysfunction is associated with the recruitment of T cell to tumor sites, and that PARPi may increase the immunogenicity of tumor cells have paved the way for combined strategy of ICIs and PARPi in BRCA-associated or more largely to HRD cancers. However, the combinatory effect of PARPi and ICIs in tumors without HR deficiency remains unknown. While in vitro studies in non-HRD cancer cell lines provide the rationale for a combined strategy, in vivo preclinical evidence suggesting that PARPi might increase the efficacy of ICIs has been conducted preferentially in BRCA-deficient tumor models, thus limiting the translational relevance. The question of whether PARPi and ICI should be restricted to non-HRD tumors or should be used more broadly has to be elucidated through the understanding of underlying biological mechanisms. Further work is needed to uncover the target population who are most likely to benefit from the combination strategy.

It is of note that most tumor types where the combination strategy was evaluated already demonstrated significant benefit of PARPi monotherapy, but limited activity for ICIs. Thereby, it would be more pertinent to evaluate the association in a population which cancer treatment represent a critical unmet medical need. For example, it would be interesting to focus on a subgroup which does not derive benefit or is primary/secondary resistant to either PARPi or ICI alone. Furthermore, dosing and scheduling of drugs largely differed across studies. The optimal dose and schedule of each agent needs to be determined with empirical clinical method as well as correlative analysis, including sequential tumor biopsies and serial blood collection.

In the future, a critical next step is to understand and identify the optimal patient population that will benefit the most from this combination. Biomarkers will likely play an even greater role in identifying those patients most likely to respond to PARPi. Whereas tumors with BRCA1/2 mutation or HRD are more likely to benefit from PARPi and ICI, an unmet medical need remains in the HR-proficient populations, so it is important to evaluate whether PARPi can sensitize these tumors to ICI in clinical settings. Moreover, obtaining contemporaneous tumor tissues and matched blood samples, associated with the integration of precision medicine, are key steps to better understand the mechanisms of action and the resistance pathways, and to identify novel predictive biomarkers of response. These crucial strides will allow a deeper comprehensive landscape of the interface between DNA damage and tumor immunity.

At last, although the current focus is on a combination of PARPi with anti-PD1/PD-L1 or anti-CTLA-4, other targeted agents moving beyond PARP in targeting DDR pathways as well as other promising immune-directed therapies are under development, and should be considered in the near future.

#### **5. Conclusions**

Genomic instability is a key hallmark of cancer that arises notably owing to DDR deficiencies. Major breakthroughs have been made with the successful targeting of DNA repair in clinical oncology [166]. Alike, immune-modulating therapies have also reshaped the landscape of cancer medicine [176]. However, treatment with either PARPi or ICIs alone often do not translate into benefit. While defects in DDR pathways might potentially be considered as predictive biomarkers of ICIs response, compelling evidence has provided a biological rationale, and demonstrated synergistic benefit, for combining ICIs with DDR inhibitors such as PARPi. To date, many preclinical and clinical researches focus on the identification of other tumors or molecular subtypes of cancers in which this combination will ultimately have a clinical impact, and thus turning more non-responders into responders with a strikingly boosted depth and duration of response.

While PARPi-induced a tumor HRD phenotype as well as immune modulation represent a rational approach for the association with ICIs, the combination did not markedly enhance the antitumor effect compared with individual agents in to-date clinical trials. To bring forward a critical change and improvements in patient outcomes, the development of accurate and predictive biomarker should become a priority. Moreover, unraveling the different mechanisms of resistance to PARPi and ICIs is required to pave the way for novel combination strategies. In addition, the appropriate dosing and scheduling of each agent should be determined in order to minimize adverse events while maximizing benefit and outcomes. Finally, elucidating the role of and interplay between DDR pathways, the tumor immune microenvironment and inhibitor agents, such as PARPi and ICIs, will be critical to the success and future development for this combination.

**Author Contributions:** F.P. and A.I. wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

1. Hodi, F.S.; O'Day, S.J.; McDermott, D.F.; Weber, R.W.; Sosman, J.A.; Haanen, J.B.; Gonzalez, R.; Robert, C.; Schadendorf, D.; Hassel, J.C.; et al. Improved survival with ipilimumab in patients with metastatic melanoma. *N. Engl. J. Med.* **2010**, *363*, 711–723. [CrossRef] [PubMed]


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Therapeutic Strategies and Biomarkers to Modulate PARP Activity for Targeted Cancer Therapy**

**Naveen Singh 1, S. Louise Pay 1, Snehal B. Bhandare 1, Udhaya Arimpur <sup>1</sup> and Edward A. Motea 1,2,\***


Received: 18 February 2020; Accepted: 7 April 2020; Published: 14 April 2020

**Abstract:** Poly-(ADP-ribose) polymerase 1 (PARP1) is commonly known for its vital role in DNA damage response and repair. However, its enzymatic activity has been linked to a plethora of physiological and pathophysiological transactions ranging from cellular proliferation, survival and death. For instance, malignancies with *BRCA1*/*2* mutations heavily rely on PARP activity for survival. Thus, the use of PARP inhibitors is a well-established intervention in these types of tumors. However, recent studies indicate that the therapeutic potential of attenuating PARP1 activity in recalcitrant tumors, especially where PARP1 is aberrantly overexpressed and hyperactivated, may extend its therapeutic utility in wider cancer types beyond BRCA-deficiency. Here, we discuss treatment strategies to expand the tumor-selective therapeutic application of PARP inhibitors and novel approaches with predictive biomarkers to perturb NAD<sup>+</sup> levels and hyperPARylation that inactivate PARP in recalcitrant tumors. We also provide an overview of genetic alterations that transform non-BRCA mutant cancers to a state of "BRCAness" as potential biomarkers for synthetic lethality with PARP inhibitors. Finally, we discuss a paradigm shift for the use of novel PARP inhibitors outside of cancer treatment, where it has the potential to rescue normal cells from severe oxidative damage during ischemia-reperfusion injury induced by surgery and radiotherapy.

**Keywords:** PARP Inhibitors; beta-lapachone; NQO1; PARG; NAMPT; cancer therapeutics; DNA repair; cMET

#### **1. Introduction**

Maintenance of genomic integrity is vital to achieve normal cellular function and to prevent the development of diseases such as cancer [1]. At the heart of this intricate biological process are DNA repair factors that work harmoniously to scan, detect, and repair potentially deleterious damage to cellular genetic information. Indeed, the disruption of one or more DNA repair pathways compromises genetic stability and is a known mechanism of cancer initiation, development, and progression. Poly-(ADP-ribose) polymerase 1 (PARP1) is the classical and founding member of at least 17 human PARP enzymes that share the ability to catalyze the transfer of ADP-ribose units to target proteins to modulate chromatin structure, transcription, replication, DNA damage response and repair [2]. PARP1 is an abundant nuclear protein that acts as a DNA damage sensor and a facilitator of DNA repair pathway choice in response to cellular stress [3–6]. Specifically, it is involved in the repair of single-stranded DNA breaks (SSB) via the base-excision repair (BER) pathway. In BER, PARP1 functions to recruit other repair factors by binding to single-stranded DNA break intermediates. Additionally, it catalyzes the synthesis of poly(ADP)-ribose (PAR) chains to the acceptor proteins (e.g., histones and XRCC1), including itself, using nicotinamide adenine dinucleotide (NAD+) as the substrate and source

of energy. The PARylated proteins then recruit and retain critical processing factors to the site of the lesion to facilitate the efficient repair of the SSB (Figure 1A) [7].

**Figure 1.** The role of PARP1 in DNA damage response and repair and cancer therapy. (**A**) PARP1 binds to single-strand breaks (SSB) for activity to target, recruit and retain critical DNA repair proteins at the sites of DNA lesions. (**B**) PARP inhibitors convert SSBs to lethal double-stranded breaks (DSBs) that are left unrepaired in BRCA-deficient cells due to a compromised homologous recombination (HR) repair consequently leading to cell death.

In the absence of PARP1, unrepaired SSBs are converted to double-stranded breaks (DSBs) during replication or the S-phase of the cell cycle [8]. Double-stranded DNA breaks are one of the most lethal forms of DNA damage induced by exogenous DNA damaging agents (e.g., ionizing radiation (IR) and chemotherapeutic agents) or endogenous replicative stress in fast proliferating cancer cells. Thus, accurate repair of DSBs is paramount to the growth and survival of all cells. Indeed, cells have evolved a range of DSB repair mechanisms. The two major repair pathways that have been studied extensively are homologous recombination (HR) and non-homologous end-joining (NHEJ) [9]. HR is largely error-free and more prevalently activated during and after DNA duplication when an identical chromatid is accessible as a template for repair. In contrast, NHEJ is active throughout the cell cycle and promotes direct ligation of DSB ends at the cost of small insertions, deletions, substitutions at the break, and even translocations that arise if DSBs from different parts of the genome are combined [10]. Cells utilize two mechanistically distinct end-joining pathways to process DNA DSBs [10–12]: Classical-NHEJ (c-NHEJ) leads to a minimal sequence alteration at the repair junctions, whereas alternative-NHEJ (alt-NHEJ, also known as microhomology-mediated end-joining (MMEJ) or back-up NHEJ) causes extensive genetic changes (deletions and insertions) that scar the break sites following ligation of DSB ends.

Several studies have shown that loss-of-function mutation of canonical HR factors – such as breast cancer type 1 and 2 (BRCA1/2) susceptibility proteins that are commonly associated with breast and ovarian cancer [13–15] – promotes PARP1 hyperactivation in fast replicating cancer cells [16,17]. This suggests that hyperactivation of PARP1 is essential to facilitate the repair of potentially lethal DNA breaks for the survival of HR-defective (HRD) cancers. Indeed, *BRCA1*/*2* mutant cancers are selectively killed by PARP inhibitors (Figure 1B), which has led to the approval of these agents to treat HR-deficient ovarian and breast cancers. However, several studies have reported that certain cancers without a BRCA deficiency have significant clinical benefits to PARP1 inhibitors positing that PARP

inhibitors could be expanded to a target population beyond BRCA-deficiency (e.g., *gBRCA* mutation carriers) [18–21].

This review article highlights treatment strategies to selectively target BRCA-proficient cancers by modulating PARP activity that alters PARP binding, PAR and NAD<sup>+</sup> levels to induce tumor-selective cell death using predictive biomarkers for therapeutic response. We summarize how aberrant alterations of PARG, NAMPT, NQO1, cMET and "BRCAness" genes that have been shown to affect PARP activity in cancers could serve as prognostic biomarkers for targeted therapy. Finally, we briefly discuss innovative approaches for the use of novel PARP inhibitors to rescue injured normal cells from severe oxidative damage during ischemia-reperfusion injury that could be induced by surgery and radiotherapy.

#### **2. Mechanism of Action for PARP Inhibitors**

PARP inhibitors are designed to mimic the substrate-protein interactions of NAD<sup>+</sup> within the ADP-ribose transferase (ART) catalytic core of PARP1-3, which are key DNA damage response sensors and transducers. The inhibitors compete with NAD<sup>+</sup> binding site of PARP to inhibit PAR polymerization, which then hinders the recruitment and regulation of DNA repair factors and the eventual release of PARP from DNA damage. Two mechanisms are proposed to induce the lethality of PARP inhibitors: PARP catalytic inhibition and PARP trapping [22,23]. However, the relative contributions of these two pathways in mediating the lethality of PARP inhibitors remain enigmatic. For example, there is evidence suggesting that the differences in the trapping potential of PARP1 to the DNA are more efficient at killing HR-deficient cells [22,23]. While there has been no evidence that this mechanism exists with clinically-used PARP inhibitors, a recent study has demonstrated that certain non-hydrolysable NAD<sup>+</sup> analogs (e.g., benzamide adenine dinucleotide, BAD) binding at the catalytic site of PARP1 could greatly enhance the binding affinity of PARP1 at sites of DNA damage that prevents its release [24]. Mechanistically, this analog competes with NAD<sup>+</sup> at the catalytic binding site that then stabilizes a conformation that induces the DNA binding domain of PARP1 to be locked on a DNA break with a significantly better binding affinity (~10-fold) – a phenomenon known as "reverse allostery." While most PARP inhibitors (e.g., Rucaparib) bind at the catalytic site, not all of them are "created equal." In fact, the relative contribution of DNA trapping induced by PARP inhibitors contributes significantly to the toxicity induced by these agents in cancers that are deficient in HR pathways [23,25–27]. Some of them have variable inhibitory effects on the PARP isoforms other than PARP1, and some may have more off-target effects on kinases than others. Regardless, PARP inhibitors are now being explored for use in several cancers, even in cancers beyond *BRCA* mutations [25,28–30]. Interestingly, a recent study has reported a correlation between therapeutic response to PARP inhibition and the patterns of ADP-ribosylation (i.e., amount of PARylation) in a panel of ovarian cancers, suggesting that ADP-ribosylation may be a useful biomarker for HR deficiency and sensitivity to PARP inhibitors regardless of BRCA status [31]. While most of the PARP inhibitors have been suggested to specifically inhibit PARP1 and/or PARP2, our limited understanding of the overlapping and non-overlapping functions and cross-talks among all of the PARP family members still raises the problem of target specificity and the risk of unintended effects and consequences promoted by the targeting of other PARP family members.

#### **3. Approved PARP Inhibitors for Targeted Cancer Therapy**

PARP inhibitors such as Olaparib [32,33], Rucaparib [34], and Niraparib [35] have been approved by the Food and Drug Administration (FDA) and the European Medicine Agency (EMA) for the treatment of adults with deleterious or suspected deleterious germline *BRCA*-mutated ovarian cancer in different settings [36]. Moreover, these agents have also been approved for maintenance therapy in platinum-sensitive high-grade ovarian cancer (HGOC), which has a high recurrence rate and an extremely poor prognosis following relapse [37]. Currently, PARP inhibitors are being evaluated in clinical trials in other cancer settings such as nonsmall cell lung cancer, pancreatic cancer, and gastric cancer (ClinicalTrials.gov Identifiers NCT01082549, NCT02184195, NCT03427814, respectively).

Olaparib (LYNPARZA, AstraZeneca and Merck) was the first PARP inhibitor to receive FDA approval for indications in breast cancer and pancreatic cancer, and has since been approved for metastatic breast cancer in patients with germline *BRCA1* or *BRCA2* (*gBRCA1*/*2*) mutations. In 2009, Olaparib was deemed safer than conventional chemotherapeutics, with mild and reversible side effects during clinical trials [26]. It was subsequently determined to have anti-tumor activity in advanced ovarian cancer patients with *gBRCA1*/*2* mutations, particularly with primary tumors that are sensitive to platinum [18,32,38–40]. The phase III OlympiAD trial found a 60% reduction in metastatic *BRCA1*/*2* mutant breast tumor size with Olaparib in comparison to 29% in patients receiving conventional chemotherapy. The FDA approved Olaparib in ovarian and breast patients with *gBRCA* mutations in 2014 and 2018, respectively [27,41–43].

Rucaparib (RUBRACA, Clovis Oncology, Inc.) was granted accelerated approval by the FDA in 2016 for the treatment of adult ovarian cancer patients with germline and/or somatic *BRCA*-mutations previously treated with two or more prior lines of chemotherapy. This agent is also approved as a maintenance monotherapy for adult patients with recurrent ovarian cancer who have established complete or partial response (CR/PR) to platinum-based chemotherapy. In 2018, Rucaparib gained FDA approval for the maintenance treatment of advanced or recurrent ovarian, fallopian tube, and primary peritoneal cancer in patients who have received platinum-based chemotherapies, as well as in patients with germline or somatic *BRCA1*/*2* mutations following two or more treatments with conventional chemotherapy [44,45].

Niraparib (ZEJULA, Tesaro, Inc.) was approved by the FDA and EMA in 2017. Compared to Olaparib, this agent results in longer progression-free survival in patients with *BRCA1*/*2*-mutant tumors [35]. Niraparib also increases progression-free survival in patients with wild-type *BRCA1*/*2* ovarian cancers, though less effectively than in patients with *BRCA* mutations [35]. Niraparib, however, is associated with more severe side effects than other PARP inhibitors, with patients experiencing thrombocytopenia (33.8%), anemia (25.3%), and neutropenia (19.6%) [35]. In 2019, the FDA approved Niraparib for homologous recombination-deficient (HRD) patients with advanced ovarian, fallopian tube, or primary peritoneal cancer previously treated with three or more chemotherapies [46].

Talazoparib (TALZENNA, Pfizer, Inc.) is a new active PARP inhibitor approved by the FDA in 2018 for patients with deleterious or suspected deleterious *gBRCA*-mutated, HER2-negative local advanced or metastatic breast cancer [47]. Compared to other PARP inhibitors, Talazoparib is still at an early stage of clinical development in gathering evidence to support its use on the treatment of epithelial ovarian cancer.

#### **4. Harnessing the Power of PAR and PARG Inhibition for Cancer Therapy**

Poly(ADP)-ribosylation (PARylation) is a covalent and reversible posttranslational modification (PTM) of acceptor proteins catalyzed by PARPs, particularly in response to DNA damage and oxidative stress [48,49]. PARylation is also a critical signaling PTM for other biological transactions such as transcription, cell cycle regulation, and genome maintenance [50–52]. Intracellular PAR levels are very low in unstressed cells due to low enzymatic activity of PARP. When PARP1 is activated following DNA damage, PAR formation increases and consequently depletes intracellular NAD<sup>+</sup> and ATP levels [53]. Accumulation of PAR and loss of NAD<sup>+</sup> and ATP can lead to severe metabolic dysfunction and eventual cell death. Thus, the balance between ADP-ribose synthesis (i.e., PAR writers) and degradation (i.e., PAR erasers) is critical for the coordination of various cellular response pathways for survival [54].

Poly(ADP-ribose) glycohydrolase (PARG) is an endo- and exo-glycohydrolase that rapidly catalyzes the degradation of PAR polymerized by PARP1 to coordinate DNA repair [48,49]. It is comprised of an N-terminal regulatory domain required for recruitment to DNA damage sites [55], a C-terminal catalytic domain, and a central mitochondrial-targeting sequence [56]. PARG hydrolyzes the glycosidic bonds between ADP-ribose units producing free ADP-ribose and PAR oligomers (Figure 2A) [57], which play distinct biological roles in cellular processes. Free ADP-ribose is

predominantly involved in energy catabolism, calcium signaling, and protein glycation [57]. The long-chain of free PAR formed by PARG cleavage has been shown to interact with the mitochondrial proteins triggering a unique intrinsic PAR-mediated cell death program known as PARthanatos (Figure 2A**) [58,59]**. Mechanistically, a cell commits to PARthanatos when free PAR (> 60 ADP-ribose units) migrates from the nucleus to the cytosol, which then triggers the translocation of PAR-bound apoptosis-inducing factor (AIF) from the mitochondria to the nucleus to activate the cell death process (Figure 2A) [60]. Early events in PAR-mediated cell death include loss of mitochondrial membrane potential and mitochondrial permeability transition [61]. While caspase activation has been demonstrated to act as a bystander in PAR-dependent cell death, this caspase-independent process shares cytological and morphological characteristics of both necrosis and apoptosis [58,59].

**Figure 2.** Pharmacological modulation of PAR ad NAD<sup>+</sup> in cancers for therapy. (**A**) Mechanism of PARthanatos mediated by the translocation of cleaved PAR from the nucleus to the cytosol and mitochondria to induce the release of AIF that translocates into the nucleus to initiate death in PARG-overexpressing cancers. Alternatively, breakdown and recycling of PAR can be prevented by inhibition of PARG to enhance NAD<sup>+</sup> depletion caused by PARP hyperactivation, ultimately starving the cell of ATP needed for various critical cellular processes (e.g., DNA repair). (**B**) In cancer cells that overexpress NAMPT, the use of NAMPT inhibitors interfere with generation of intracellular NAD<sup>+</sup> levels that compromise PARP activity, consequently resulting in impaired repair of DNA damage and NAD+-Keresis.

ADP-ribose hydrolases (ARH) can catalyze PAR hydrolysis but to a lesser extent than PARG, which catalyzes the majority (~90%) of PAR catabolism [62]. PARG is aberrantly overexpressed in most human cancers, suggesting that PARG is required for tumorigenesis and cancer survival [63]. Therefore, inducing cell death via PARG inhibition and manipulation of the PAR cycle through PARP1/PARG inhibition is an attractive target for cancer therapy (Figure 2). Available PARG inhibitors consist of DNA intercalators, tannins, and substrate analogs. DNA intercalators (e.g., Tilorene, GPI116552, GPI18214, Ethacridine) indirectly inhibit PARG by forming complexes with its substrate, PAR [64–66]. Hydrolyzable tannins (e.g., Nobotanin B, Oenothenin B, gallotannin) inhibit

PARG in vitro by competing with the PAR substrate [67–69]; however, poor membrane permeability, poor bioavailability, and toxicity of these compounds hinder their usage. Adenosinediphosphate (hydroxymethyl) pyrrolidinediol (ADP-HPD), a nonhydrolyzable-analogue of ADP-ribose, has been shown to be a potent competitive PARG inhibitor [66]. Other reversible PARG inhibitors under development include modified salicylanilide pharmacophore [70] and rhodanine-based small molecule analogs (RBPIs) [70]. Recently, Pillay et al. [71] demonstrated that PARG inhibitor, PDD00017273, mimics poly(ADP-ribose) polymerase (PARP) inhibitor therapy in ovarian cancer cells by exacerbating the formation of hyperPARylated-PARP1 that severely compromises PARP activity. Mechanistically, PARG inhibitors prevent the recycling of hyperPARylated-PARP1 to its more active form for efficient DNA damage response and repair. In support of this, survival studies of certain human cancer cell lines with *PARG* knockdown synergistically enhances the lethality to DNA-damaging agents such as alkylating agents [72,73] and cisplatin [72]. Overall, pharmacological targeting of PARG is a promising therapeutic target and future investigations are required to develop effective strategies to manipulate the dynamic PAR formation and break-down process in response to DNA damage. The relative low abundance of PARG in normal cells, compared to tumors with aberrant PARG overexpression, makes it an attractive target in cancer therapeutics for tumor-selective drug response. Moreover, unlike the human PARP family with 17 related enzymes with almost similar catalytic subunit – PARG is a unique mammalian protein without any paralogs, which may offer fewer off-target effects than PARP inhibitors.

#### **5. Targeting NAD**<sup>+</sup> **for Cancer Therapy**

NAD<sup>+</sup> is a co-enzyme that mediates redox reactions by acting as an electron carrier and a substrate in metabolic pathways including glycolysis, the tricarboxylic acid cycle (TCA), oxidative phosphorylation, and serine biosynthesis [74]. It is a cofactor for PARP and plays a key role in energy production, cell signaling, redox homeostasis, DNA repair, gene expression, and the stress response. Many of these processes are disrupted in cancer, which alters NAD<sup>+</sup> production and consumption [75,76]. Nicotinamide phosphoribosyltransferase (NAMPT) and nicotinamide mononucleotide adenylyltransferase (NMNAT) enzymes regulate the salvage pathway critical for controlling intracellular NAD<sup>+</sup> levels [77]. Increased NAD<sup>+</sup> and aberrant NAMPT overexpression promotes glycolysis, which fuels the growth and survival of cancer cells [78,79]. Indeed, NAMPT inhibitors suppress cancer cell proliferation by inhibiting glycolysis [79–83].

Overexpression of NAMPT is observed in several types of malignant tumors, which supplies back-up NAD<sup>+</sup> to sustain cellular proliferation and promote resistance to therapeutic agents [84–89]. NAMPT as a predictive biomarker for tumor-selective targeting of NAD<sup>+</sup> metabolism is, therefore, a promising therapeutic strategy (Figure 2B). Mechanistically, NAMPT inhibitors induce cell death in cancer cells by depleting intracellular NAD<sup>+</sup> and ATP levels and inhibiting glycolysis and glucose uptake [90,91]. Examples of NAMPT inhibitors include FK866, GMX-1777/8, STF-31, STF-118804, GNE-617, GNE-618, LSN3154567, KPT-9274. Vacor adenine dinucleotide (VAD) is an NAD<sup>+</sup> analog that has been shown to inhibit both NAMPT and NMNAT, leading to necrotic cell death through NAD<sup>+</sup> depletion (NAD+-Keresis, Figure 2B), glycolytic block, and energy failure [82,92–98]. Several NAMPT inhibitors have been investigated in the clinic as a monotherapy such as FK866 and GMX-1777/8. However, thrombocytopenia was demonstrated to be the dose-limiting toxicity in these clinical trials [99,100]. Though NAMPT inhibitors as a monotherapy are promising candidates for cancer therapy, drug resistance and toxicity to normal tissues when used at high concentration remain a major hurdle [99,101–103]. Perhaps, a combination approach with tumor-selective DNA damaging agents (e.g., NQO1-bioactivatable agents, vide infra in Section 6) may overcome this limitation to improve the tumor-selectivity of NAMPT inhibitors by severely depleting the NAD<sup>+</sup> needed by PARPs for efficient DNA repair and survival [82,104].

#### **6. Leveraging NQO1 as a Biomarker for Tumor-Selective Use of PARP Inhibitors with NQO1-Bioactivatable Drugs**

NAD(P)H:Quinone Oxidoreductase 1 (NQO1, DT-diaphorase) is a cytoplasmic Phase II detoxification enzyme that utilizes NAD(P)H to reduce certain quinones to stable hydroquinones via two-electron reductions that are readily conjugated by Phase III transporters for cellular efflux (Figure 3A) [105]. NQO1 is abnormally overexpressed in most malignant tumors, which contributes to drug resistance by metabolizing xenobiotics to their inactive forms or chemical structures that are excreted out of the cell for chemoprotection. Indeed, the addition of an NQO1-inhibitor (e.g., dicoumarol) in cancers could sensitize the anti-cancer effects of certain agents by inhibiting resistance due to drug efflux.

**Figure 3.** Strategy for tumor-selective use of PARP inhibitors in solid tumors. (**A**) The role of NQO1 in detoxifying certain quinones for cellular efflux. (**B**) The role of NQO1 in bioactivating certain quinones to induce toxicity in NQO1(+) cancer cells. Note that the stability of hydroquinones determine whether bioreduction by NQO1 leads to detoxification or toxicity. (**C**) Mechanism of tumor-selective synergistic cell death induced by combination of PARP inhibitor and NQO1-bioactivatable agents in NQO1(+) cells.

Certain quinones are bioactivated by NQO1 to create a rapid increase in reactive oxygen species (ROS), particularly hydrogen peroxide (H2O2), which can permeate through the nucleus to induce the formation of toxic DNA damage leading to cell death (Figure 3B) [106–110]. For example, β-lapachone (β-lap) is a soluble ortho-naphthoquinone with potent anti-tumor and radiosensitizing activity only in the presence of high NQO1 activity [108–114], which is only noted in most malignant tumors but not in normal tissues [109]. Mechanistically, the two-electron reductase capacity of NQO1 catalyzes the oxidoreduction of β-lap to an inherently unstable hydroquinone, which automatically and rapidly gets reverted back to its original quinone form as it undergoes a two-step oxidation (Figure 3B) [108]. This produces what is termed as a "futile cycle" of oxidation and produces a significant level of reactive oxygen species such as superoxide radicals (O2 −) that are rapidly catalyzed by superoxide dismutases to hydrogen peroxide in the cytoplasm (Figure 3B) [115]. This stable form of ROS can easily penetrate through the nucleus and is converted to genotoxic hydroxyl radicals via the Fenton reaction process, which causes massive formation of oxidative DNA base damage and SSBs (Figure 3B) [109,116]. The surge of this specific types of DNA damage in the nucleus causes hyperactivation of PARP1, which is an NAD+-dependent enzyme, to a level that consequently leads to NAD+-Keresis death due to the depletion of critical biological energy sources, NAD<sup>+</sup> and ATP [82,107,109]. Another possibility is that PARP hyperactivation could lead to an NAD+-independent glycolytic and bioenergetics crisis that is presumably caused by PAR-dependent inhibition of glycolysis that occurs through the inhibition of hexokinase leading to a specific form of cell death termed "PARthanathos" [117]. Future studies are required to firmly establish the contribution of PARP hyperactivation, NAD+, or free PAR in specific types of cell death. Regardless, the use of NQO1-bioactivatable agents to modulate PARP activity, NAD<sup>+</sup> and PAR to induce cell death is a viable strategy to treat cancer. As NQO1 is typically overexpressed in tumors but not at a significant level in normal healthy cells [109], the application of NQO1-bioactivatable agents is an attractive possibility for tumor-selective generation of H2O2-induced DNA damages that hyperactivate PARP and deplete NAD<sup>+</sup> to cause eventual cell death, which is vital for developing safe and effective drugs with minimal off-target effects.

While β-lap (ARQ761 in clinical form) shows potential as a single-agent therapeutic [107,118], it has also been associated with hemolytic anemia at a dose of 450 mg/m<sup>2</sup> [119]. Therefore, increasing its efficacy at a lower dose is therapeutically advantageous. Due to the role of PARP hyperactivation in the mechanism of action of β-lap, PARP inhibitors have been a central focus in improving the efficacy and reducing the toxicity of β-lap for use in NQO1(+) cancers. PARP inhibition followed by β-lap treatment alters the mechanism of β-lap function by blocking the repair of damaged DNA and inhibiting PARP1 hyperactivation, slowing down the process of NAD<sup>+</sup> depletion as a result, and extending the cycling of β-lap by NQO1 to maximize hydrogen peroxide production and amplifying DNA damage in a tumor-selective manner (Figure 3C) [109]. Instead of dying by caspase-independent NAD+-Keresis, cell death from exposure to β-lap following PARP1 inhibition occurs through caspase-dependent apoptosis, presumably due to the availability of NAD<sup>+</sup> and ATP (Figure 3C) [109]. As NQO1 will continue cycling β-lap until the drug is depleted or the required source of NAD(P)H is depleted, blocking the DNA repair process which leads to NAD<sup>+</sup> depletion following β-lap treatment prolongs the efficacy of the drug, allowing for lower doses to be used with the synergistic effect, minimizing the potential for hemolytic anemia.

The use of NQO1-bioactivatable drug (e.g., β-lap) with PARP inhibitors have been shown to significantly expand the clinical application of PARP inhibitors outside of *BRCA1*/*2* mutant cancers for which the majority of PARP inhibitors are currently approved. The prevalence of NQO1 overexpression in such a wide variety of malignant cancers (>90% in pancreatic cancers with KRAS mutation, >80% nonsmall cell lung cancer (NSCLC), >60% breast cancer, >60% prostate cancer, >45% head and neck, and >60% colon cancers [109]) suggests that this combination therapy may have significant future potential, especially in cancers in which prognosis is currently poor, such as head and neck cancer, pancreatic cancer and NSCLC [117,120–126].

#### **7. Targeting cMET to Attenuate PARP1 Activity**

Many studies have shown that most malignant tumors have increased levels of reactive oxygen species (ROS), particularly hydrogen peroxide (H2O2), compared to their normal counterparts. These genotoxic reactive species cause oxidative DNA damage and single-stranded DNA breaks that stimulate PARP activity in DNA repair [127]. Interestingly, Du and colleagues [128] have shown that cMET, which is a receptor tyrosine kinase, can further stimulate the activity of PARP1 in cancers to survive the lethal effects of ROS-induced DNA damage. In cancer, cMET is overexpressed and its abnormal activation can promote the development and progression of multiple cancers [129].

Mechanistically, the high level of ROS in cancer promotes the activation and translocation of cMET into the nucleus where it directly binds and phosphorylates the tyrosine residue 907 (Y907) of PARP1 to enhance its enzymatic activity (Figure 4A) [128,130,131]. Indeed, Du and colleagues have demonstrated that phosphorylated PARP1 (pY907) showed a higher level of PAR production than non-phosphorylated PARP1 in vitro, thereby highlighting the critical role of cMET in enhancing PARP1 activity in DNA repair [128,130,131]. Moreover, the phosphorylation of this specific residue (pY907) has been suggested to significantly decrease the binding of clinically-relevant PARP inhibitors due to a potential steric hindrance (Figure 4B) [128]. Indeed, c-MET-mediated phosphorylation of PARP1 at Y907 leads to PARP inhibitor resistance, which could be potentiated when combined with cMET inhibitors (e.g., crizotinib and foretinib) [128]. Thus, inhibition of cMET is an attractive strategy to indirectly block PARP1-mediated DNA repair and enhance the therapeutic effects of PARP inhibitors in cancer cells [128]. However, the mechanistic basis for requiring Y907 phosphorylation to promote H2O2-induced PARP1 activity remains to be firmly established. A recent report also suggests that phosphorylation of PARP1 by EGFR and cMET heterodimer contributes to PARP1 inhibitor resistance. Hence, combination treatment consisting of EGFR, cMET, and PARP1 inhibitors posits a novel therapeutic strategy in cancer treatment involving overexpression of PARP1/EGFR/cMET, which are frequent alterations in most solid tumors [132].

**Figure 4.** Attenuation of PARP Activity by cMET inhibition. (**A**) cMET enhances PARP activity during oxidative stress. Inhibition of cMET is an excellent strategy for inhibiting PARP1-mediated DNA repair and synergy with PARP inhibitors. (**B**) A snapshot of non-hydrolyzable NAD<sup>+</sup> analog (in cyan) bound to residues in PARP1 catalytic active site within 6Å from the ligand (PDB ID: 6 bhv, processed via PyMOL Molecular Graphics System). Phosphorylation of PARP1 residue Y907 (shown in green) inhibits binding of PARP inhibitors.

#### **8. Targeting of PARP Activity in Non-Oncological Events**

PARP inhibition or gene deletion has been shown to attenuate tissue injury associated with ischemia-reperfusion injury and inflammation that could arise during tumor surgical resection and radiation therapy [133,134]. During this non-oncological event, the catalytic activity of PARP1 becomes hyperactivated and consequently leads to NAD<sup>+</sup> depletion. This process forces NAD<sup>+</sup> replenishment through the salvage pathway, which decreases cellular ATP levels and results in bioenergetic crisis, eventually leading to necrotic cell death that is typically associated with inflammation [3,135]. Thus, inhibition of PARP activity could have protective effects by dramatically reducing NAD<sup>+</sup> consumption and preventing energetic failure and the consequent necrotic cell death. While the trapping mechanism induced by specific PARP inhibitors might be advantageous for the treatment of certain cancers, this mechanism of action might not be optimal to rescue normal cells from severe oxidative damage during ischemia-reperfusion injury (e.g., ischemia of the lung due to ionizing radiation [136] and cerebral ischemia during surgery [137]).

For non-oncological indications, the safety profile of most FDA-approved PARP inhibitors would be expected to be better if it had less PARP trapping activity to restore the viability of normal tissue damage from injury caused by ischemia-reperfusion. To circumvent the cytotoxic effects of PARP trapping, the Yu group [138] developed a strategy to decrease PARP activity using a novel lead compound, iRucaparib-AP6 (Figure 5), which promotes PARP degradation upon binding that mimics PARP1 genetic deletion. Mechanistically, the Rucaparib component of the small molecule specifically binds PARP at the site of damaged DNA, but it is attached to a linker with a ligand that brings the E3 ligase in close proximity to ubiquitinate PARP and subsequently gets degraded through the proteasome pathway (Figure 5C). Unlike PARP trapping, this strategy of modulating PARP activity protects cells against genotoxic stress-induced cell death [138] that could have potential clinical applications for the treatment of ischemia-reperfusion injury and neurodegeneration. However, further development is needed to firmly establish the mechanism of action and evaluate its protective effects to treat the aforementioned non-oncological indications.

**Figure 5.** Development of a novel "PARP degrader" molecule as a biochemical probe and therapeutic agent. (**A**) Chemical structure of Rucaparib. (**B**) Chemical structure and elements of iRucaparib-AP6. (**C**) Simplified representation of the mechanism of action for a novel proteolysis targeting chimeric (PROTAC) molecule, iRucaparib-AP6.

To summarize, the enzymatic function of PARP1 – which is to catalyze the PARylation of a large number of PAR-acceptor proteins (including itself) – becomes hyperactivated upon detection of genotoxic stresses associated with various pathological conditions such as cancer, ischemia-reperfusion injury (e.g., myocardial infarction), inflammatory diseases (e.g., colitis, arthritis, asthma), vascular diseases (e.g., diabetic complications, atherosclerosis), and neurodegeneration. Thus, modulation of PARP activity using a novel class of PARP activity inhibitors (e.g., iRucaparib-AP6) with little to no "trapping" potential may be a promising strategy to ameliorate certain pathological conditions that are aimed to heal instead of killing the severely injured normal tissues (e.g., inflamed tissues). Indeed several studies have looked at the protective effects of PARP inhibitors in lung transplantation [139], neurodegeneration [140], renal injury [141], aging [142], and acute pancreatitis [143]. While these studies offer critical insights toward the potential therapeutic repurposing of PARP inhibitors in treating non-oncological indications, it is important to learn more about the possible side-effects of long-term treatment and whether PARP inhibition may increase the risk of mutagenesis and carcinogenesis.

#### **9. Perspectives and Conclusions**

Individualized treatments to patients with recalcitrant cancers in order to achieve cure are urgently needed. Currently, cancer patients have many therapeutic options for treatment; however, the lack of effective and tumor-selective treatment options still remains a major hurdle in the fight against cancer today. This is due to very limited predictive biomarkers that define individual risks of disease recurrence or sensitivity to treatments. For example, a significant number of cancer patients are over-treated in order to possibly attain an improved overall survival in early cancer stage. This "shotgun" approach, however, could have life-threatening consequences to cancer patients due to the emergence of drug resistance and secondary malignancies. Clearly, there is an urgent need for (1) a clear understanding of normal versus cancer cell biology to gain more informative decisions regarding the most effective therapeutic interventions, and (2) an arsenal of new treatment options based on predictive biomarkers for this disease to eliminate toxic effects to normal cells, drug resistance, and development of secondary cancers.

In less than two decades, studies on the use of PARP inhibitors for the treatment of *gBRCA1*/*2* mutant cancers and beyond BRCA-deficiency have increased tremendously. Of course, these were all made possible by the collaborative effort of the scientific and medical communities to understand the functional roles of breast cancer associated genes 1 and 2 (BRCA1/2) in normal physiology and breast cancer development. Accordingly, deficiencies in BRCA1 and BRCA2 genes can compromise HR-mediated repair, conferring hypersensitivity to PARP inhibitors by chemical synthetic lethality. The rarity of breast (5–10%) and ovarian (15–20%) cancers with germline *BRCA1*/*2* loss-of-function mutations, however, currently restricts the therapeutic utility of PARP inhibitor monotherapy. This is further complicated by several reports of potential molecular mechanisms of resistance for PARP inhibitors in BRCA-deficient breast cancers [144]. Interestingly, approximately 24% of high-grade triple negative breast cancers (TNBC) without a *BRCA* mutation showed great response to PARP inhibitor in Phase II clinical trials suggesting that the use of PARP inhibitors can be expanded beyond BRCA deficiency, and this has been reinforced by several studies in other cancers [18–21]. Indeed, genetic and pharmacological alterations of specific proteins (Table 1) or long non-coding RNAs (lncRNAs, e.g., PCAT-1 [145]) that can induce a state of "BRCAness" or HR-deficiency in BRCA-proficient cancers have the potential to broaden the application of PARP inhibitors beyond the current protocols approved by the FDA and EMA.

In conclusion, a serious concern for the use of PARP inhibitors – as with all chemotherapeutic agents – is the development of acquired drug resistance and de novo malignancies [146] due to limited information and often unclear understanding of the full extent of the agents' specificity and mechanism of action. A complete and clear knowledge of how different PARPs are activated to perform overlapping and non-overlapping functions or how PAR and NAD<sup>+</sup> levels are modulated to alter specific biological events toward cellular survival or death are essential to: (1) design novel PARP inhibitors that are more specific and tumor-selective; (2) develop better mechanistic-based strategies using reliable predictive biomarkers for PARP inhibitor monotherapy as well as combination treatments; and (3) identify situations that can re-sensitize recalcitrant tumor cells to PARP activity inhibition or modulation to treat cancer patients more safely and efficiently.


**Table 1.** Representative list of genes/proteins that have been shown to cause "BRCAness" or HR defect in BRCA-proficient tumors when expression is lost or activity is inhibited. These deficiencies could be exploited as predictive biomarkers for precision treatment with PARP inhibitors.


**Table 1.** *Cont*.

**Author Contributions:** Conceptualization: E.A.M.; resources: E.A.M.; writing–original draft preparation: E.A.M., N.S., S.L.P., S.B.B., and U.A.; writing–review, and editing: E.A.M., N.S., and S.L.P.; visualization: E.A.M.; supervision: E.A.M.; funding acquisition: E.A.M. All authors have read and agree to the final version of the manuscript.

**Funding:** This work was supported by Biomedical Research Grant from IUSM, Indiana CTSI Showalter Trust and NIH/NCI R01 CA210489 to E.A.M. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **The Multifactorial Role of PARP-1 in Tumor Microenvironment**

#### **Juan Manuel Martí, Mónica Fernández-Cortés, Santiago Serrano-Sáenz, Esteban Zamudio-Martinez, Daniel Delgado-Bellido, Angel Garcia-Diaz and Francisco Javier Oliver \***

Instituto de Parasitología y Biomedicina López Neyra, CSIC and CIBERONC, Instituto de Salud Carlos III, 18016 Granada, Spain; jmmmc@ipb.csic.es (J.M.M.); monica.fernandez@ipb.csic.es (M.F.-C.); s.serrano@ipb.csic.es (S.S.-S.); estebanzm95@ipb.csic.es (E.Z.-M.); ddelgado@ipb.csic.es (D.D.-B.); agdiaz@ipb.csic.es (A.G.-D.)

**\*** Correspondence: joliver@ipb.csic.es

Received: 18 February 2020; Accepted: 15 March 2020; Published: 20 March 2020

**Abstract:** Poly(ADP-ribose) polymerases (PARPs), represent a family of 17 proteins implicated in a variety of cell functions; some of them possess the enzymatic ability to synthesize and attach poly (ADP-ribose) (also known as PAR) to different protein substrates by a post-translational modification; PARPs are key components in the cellular response to stress with consequences for different physiological and pathological events, especially during neoplasia. In recent years, using PARP inhibitors as antitumor agents has raised new challenges in understanding their role in tumor biology. Notably, the function of PARPs and PAR in the dynamic of tumor microenvironment is only starting to be understood. In this review, we summarized the conclusions arising from recent studies on the interaction between PARPs, PAR and key features of tumor microenvironment such as hypoxia, autophagy, tumor initiating cells, angiogenesis and cancer-associated immune response.

**Keywords:** Tumor microenvironment; PARPs; PARylation; hypoxia; autophagy; PARP inhibitors

#### **1. Introduction**

#### *1.1. PARP Family of Proteins*

Poly (ADP-ribose) polymerases (PARPs), more recently named ADP-ribosyl transferases (ARTs) [1] are a family of enzymes characterized by the presence of a 50 amino acid sequence called "*PARP signature*". This allows some members of the family to synthesize and transfer ADP-ribose to a large number of substrates through a process referred as poly(ADP-ribosyl)ation or PARylation [2]. The most representative member of this family is PARP-1, firstly described by Chambon et al. in 1963 [3]. Since then, many studies have deciphered both structural and biological aspects of this family, as well as the pathological consequences of the misregulation of its members.

The best studied PARP enzyme is the original constituent of the family PARP-1. Encoded by the region 1q41-q42, this protein has a molecular weight of 114 kDa and it is constitutively expressed. In fact, PARP-1 promoter owns characteristics usually revealed in housekeeping genes. Consistently, in all tissues the mRNA of PARP-1 is present, albeit at varying levels [4]. Nevertheless, PARP-1 modulation is mainly settled at the protein level and not at the mRNA level.

The PARP superfamily contains 17 members sharing on their catalytic domain the conserved "PARP signature" (Figure 1).

**Figure 1.** Poly(ADP-ribose) polymerases (PARP) family. The structure of the different members of the PARP family is described. Different domains are detailed in different colors. Brighter blue shows the PARP signature sequence, common throughout all the members of the family [5].

As this signature is not sufficient to provide a functional classification, the categorization must be established regarding their architecture and different enzymatic functions. Although firstly classified into three subgroups [6,7], this family is now grouped into five subfamilies [8,9]:


Being the most active member of this family, the PARP-1 structure was firstly described in 1984 [11] and it is composed of three domains (Figure 2):

**Figure 2.** PARP1 structure. The different domains of PARP1 are presented here. The zinc fingers responsible for the DNA binding capacity, the automodification domain thanks to which PARP1 is modified by polymer and the catalytic domain containing the "PARP signature".


In order to perform its PARylation activity, PARP-1 and PARP-2 performs both poly(ADP-ribosyl) synthetase and transferase enzymatic activity. The PAR life cycle is described as follows [7,14] (Figure 3).

**Figure 3.** PARP metabolism. Different steps in polymer formation are shown in green. In contrast, the proteins involved in polymer degradation are shown in red and yellow [14,15].

• Initiation phase: Firstly, poly(ADP-ribose) synthetase activity catalyzes the breakage of the glycosidic bond between nicotinamide and ribose on the NAD+ molecule. Through this process of oxidation, ADP-ribose is formed. Subsequently, ADP-ribose is covalently attached to different acceptor proteins via formation of an ester bond between the protein (through glutamate, aspartate or lysine residues) and ADP-ribose.


As explained previously, the best described effect of PARP activation is PARylation, or covalent protein modification by PAR. This process may affect PARP itself or other proteins that become PARylated. Nevertheless, other mechanisms underlie PARP activation and PAR synthesis. They are summarized below:


These molecular events are responsible for the different events and function alterations that take place after PARP activation (Figure 4).

**Figure 4.** Molecular events following PARP activation. Once PARP is activated, downstream events of PARP signaling take place, involving either covalent PARylation of substrates, non-covalent binding of PAR polymer to proteins bearing a PAR-binding motif, release of free PAR to the cell or lowering cellular NAD+/ATP levels. Via these pathways, PARP/PARylation regulate functions such as transcription, replication, DNA repair, protein degradation and cell cycle, mediating various cellular phenomena such as proliferation, cell survival and cell death or differentiation.

The "central dogma" of PARylation asserts that during DNA damage PARP-1 is activated, on both single strand breaks (SSB) and double strand breaks (DSB). In fact, DNA strand breaks rises up to 500 times the basal activity of PARP-1 [29]. Consistently, DNA alkylating agents as well as reactive species of both oxygen and nitrogen (ROS and RNS, respectively) have been proved to act as triggers of PAR accumulation in different cell types [30]. However, different results have shown that PARP-1 is also stimulated in the absence of DNA damage. This has been observed in the presence of non-B DNA structures like bent, cruciform DNA or stably unpaired DNA [31].

In addition, PARP-1 activity can be modified in addition to DNA-related status; several post-translational modifications also alter PARP-1 activity. PARP-1 is implicated in multiple signaling pathways often involving a kinase phosphorylating different aminoacids of PARP-1, causing its activation or inhibition. Examples of this topic include PARP-1 being phosphorylated by ERK1/2 at serine 372 and threonine 373, deriving in PARP-1 activation [32], Also c-Met phosphorylates PARP-1, specifically at the Tyr907 causing its activation. [33–35]. Other phosphorylations can have the opposite effect, leading to PARP-1 inhibition, this is the case of the one performed by AMPK at Ser177 [36] and the one mediated by the protein kinase C [37]. PARP1 regulates the c-Jun N-terminal kinase (JNK) pathway, which is a driver of tumor development and treatment response. Based on that, PARP inhibition could be potentially therapeutically beneficial in ovarian cancer taken the elevated JNK activity. Furthermore, PARP1 inhibitors promote Akt [38].

Additionally, PARP-1 is regulated through other post-translational modifications different to phosphorylation but equally necessary for PARP to develop its function. PARP-1 is acetylated by p300/CBP at the lysines 498, 505, 508, 521 and 524 [39]. Also remarkable is the ubiquitination at the lysine 486 and 203 performed by the E3 ligase regulating PARP1 activity [40]. PARP-1 even undergoes ADP-ribosylation, as the described within the serine S499 [30,41].

#### *1.2. Tumor Hypoxic Response and PARP-1*

When an avascular tumor grows over a few cubic millimeters (depending on the tissue density and irrigation), the central parts of the mass experience low oxygen concentration and nutrients deprivation [42]. The response to this hypoxic environment has proven to be a key factor characterizing the tumor microenvironment and the disease outcome. It induces metabolic adaptations like glycolysis activation [43], cancer stem cell regulation [44], tumor exosome production [45] and the crucial tumor neovascularization [46]. New vessel formation supplies the hypoxic tumor with nutrients and oxygen, allowing the stressed cells to survive and divide, promoting mass formation and increasing cell plasticity, migration and aggressiveness [47–49]. The relevance of this pathway was vindicated when its discoverers were awarded the Nobel Prize of Medicine in 2019.

Furthermore, hypoxia generates resistance to different therapies. On the one hand, poorly vascularized areas make chemotherapy and immunotherapy difficult to disseminate. On the other hand, low oxygen levels make radiotherapy less effective at generating massive levels of toxic reactive oxygen species (ROS). This is the reason why hypoxic areas are known to survive treatments and are the ones to generate relapses and develop metastasis [50].

The hypoxic response is mediated by the stabilization of the hypoxia inducible factors (HIFs), a family of transcription factors composed of three alpha chains: HIF1α, HIF2α and HIF3α which are stable only during hypoxia, and one beta chain: HIF1β, which is constitutively stable. These proteins are active as heterodimers of one of the HIFα chains with the HIF1β. The heterodimers bind to the hypoxic response elements (HRE), a sequence of nucleotides located at the promoters of hypoxic inducible genes causing their overexpression. These different α/β heterodimers present similar but not identical targets and are differently expressed among tissues [51].

PARP-1 has been shown to interact with both HIF1α and HIF2α, affecting their stability and activity (Figure 5).

**Figure 5.** Schematic representation of HIF-α and PARP1 association throughout the adaptation of tumoral cells during hypoxia. The union of PARP1 to HIF-a helps to its stabilization and allows the transcription of several genes involved in tumor growth, resistance, migration and metastasis.

The crosstalk between HIF-signaling pathways and PARP-1 has been described in models of skin carcinogenesis where a reduction in HIF1α protein and mRNA is observed during PARP-1 inhibition or knockdown [52]. In the same way, in mice brains, a reduction in HIF1α accumulation during hypoxia is observed after PARP inhibition, having an impact reducing the expression of hypoxic genes like adrenomodulin or erythropoietin [53]. This dependence of HIF1α on PARP1 status can be explained by the fact that PARP-1 and HIF1α are known to form a complex during hypoxia [54]. Moreover, during hypoxia, the presence of nitrosative and oxidative stress induced via oxidative phosphorylation [55] is relevant, leading to mitochondrial inhibition and overactivation of PARP-1, promoting the stability of HIF1α [53]. It has also been described how the PARP inhibitor veliparib can sensitize both oxic and hypoxic cells in prostate and lungs to radiotherapy [56].

HIF2α is also important during tumorigenesis; tissues where HIF1α is the predominant protein become much more dependent on HIF2α during cancer progression [57]. HIF2α is related with aggressiveness and neovascularization as well [58]. Moreover, HIF2α is known to interact with PARP-1, resulting in changes in its stability and activity. HIF2α is similar to HIF1α but has different expression patterns among tissues and some of their targets do not overlap [51]. The relation between HIF2α and PARP1 has not been deeply studied but it has been proved in several cell types that PARP-1 knockdown or inhibition reduces HIF2α accumulation in the hypoxic context [59]. It has also been described that PARP-1 binds to the HIF2α promoter (but not to HIF1α promoter) controlling its transcriptional induction [60]. However, more research is needed in order to fully comprehend their interaction and its consequences.

#### *1.3. Angiogenesis, Vasculogenic Mimicry and PARP-1*

As previously described, blood vessels formation is a central aspect during tumor development. Vascularization not only distributes nutrients and oxygen, it also removes metabolic sub products and provides access for the immune system and different treatments.

The name 'angiogenesis' was used for the first time around 1900 and it was not used to refer to the tumoral context until the 1960s [61]. Since then, different varieties of angiogenesis have been depicted in cancer: Sprouting angiogenesis, micro vessel growth and microvascular proliferation [62]. Emerging studies showed the central role of this process in tumor progression [63]. Just a few tumors are able to progress without angiogenesis induction, while the vast majority of tumors present a combination of angiogenic with non-angiogenic areas [64,65]. Traditionally, treatments against angiogenesis have focused blocking new vessel growth and also trying to dismantle the existing ones, hence starving the tumors by depleting their nutrients and oxygen supply [66]. Interestingly, it has recently being reported in the PAOLA1 study (ESMO 2019) a Progression Free Survival benefit for ovarian cancer patients treated with a combination of olaparib with bevacizumab versus bevacizumab alone [67].

It is known that PARP inhibitors present antiangiogenic activity both in vitro and in vivo [52,68–70].

However, blood supply study in the cancer context became more complex when Maniotis et al. presented, in 1999, new findings describing cancer cells that coated vascular channels; these structures were composed of non-endothelial cells that contained erythrocytes and immune cells. This process was defined as vasculogenic mimicry (VM), and it is described as the de novo generation of a network created in a 3D matrix in vitro, composed by perfusable vasculogenic-like vessels, with properties similar to the matrix rich network described in aggressive tumors [71].

Through different studies performed on models of human uveal melanoma and cutaneous melanoma, the initial characterization of VM was performed. Since then, VM has been described in other malignancies like those of the kidney, lung, bladder, pancreas, prostate, gliomas, sarcoma, ovary, breast, head and neck cancer, and the list is still growing. In survival analyses, patients carrying VM in their tumors have been seen to present a reduced clinical outcome when compared with patients not expressing VM [72]. In 2000, a seminal study using arrays of gene expression in highly malignant melanoma cells (C8161, MUM 2B) vs poorly malignant melanoma cells (MUM 2C, C81-61) showed differences in gene expression among both groups [73]. Some of the many genes that showed significant differences were VE-cad. The depletion of VE-cadherin with specific siRNA and VE-Cad antibody abolished VM in a 3D in vitro model [74]. VE-cadherin is a trans-membrane protein usually located in the endothelium, necessary for the endothelial stability through its activity in cell–cell adhesion (Figure 6).

This protein presents the activities described for classic cadherins. Since the description of VM, VE-cadherin cannot be considered a marker of exclusively endothelial cells. VE-cadherin expression has been associated with aggressiveness on different tumors carrying VM. In this direction VE-cadherin can be found in highly aggressive VM tumor cells but not in poorly aggressive ones. Moreover, its down regulation in melanoma implied the loss of capacity to form VM [74].

It has been shown that activation through VEGFR-1 or through its co-receptor NRP-1 contribute to VM [75,76] while the potent inhibitor of αν integrins cilengitide displays anti-melanoma activity through the inhibition of VM [75]. A contrary role of VEGF signaling in VM has been also proposed. While is commonly accepted that VEGF and receptors like VEGFR1 or VEGFR co-receptor (NRP-1) promote VM, it has been also proposed that VM could also appear in contexts lacking this pathway: In this situation VEGF would promote angiogenesis, and its inhibition would potentiate other survival strategies including VM [77]. A study in 2013 by Rodriguez et al. gave a closer view into VM complexity, studying the interplay between PARP inhibition, epithelial-mesenchymal transition and VM. The use of the PARP inhibitors PJ-34 and KU0058948 in a murine melanoma cell line decreased the expression of VE-cadherin, as well as its phosphorylation in the residue of tyrosine 658. This resulted in the depletion of VM formation in in vitro assays [78].

**Figure 6.** Vasculogenic mimicry pathway and PARP. VE-cadherin can be phosphorylated by stimuli of VEGF by VEGFR2 activity and the co-receptor (NRP-1). The soluble factor (VEGF) increases the phosphorylation of Y658 of VE-cadherin and the consequent internalization. Inhibition of PARP decreases the phosphorylation Y658 of VE-cadherin and finally the possibility to form VM in aggressive melanoma cells (B16F10 cells).

The use of angiogenesis inhibition in combination with other treatments has been approved by the FDA in some scenarios; this is the case of an anti-VEGF specific antibody (bevacizumab, Avastin, Roche) used in combination with chemotherapy or different cytokines therapies for late-stage advanced metastatic cancers (including renal cell cancer, colorectal cancer, non-squamous non-small cell lung cancer and breast cancer) [79]. On the other hand, angiogenesis inhibitors in combination with immune checkpoint in the treatment of breast cancer showed no clear benefits and are still under evaluation, so there are no current FDA-approved indications for their use in breast cancer [80]. Current studies show how the combination of angiogenesis and PARP inhibitors will be likely safe due to non-overlapping toxicities, and it might be expected that PARP inhibitors could be used in this context at full mono-therapy dosages.

#### **2. Immuno-Response Modulation by PARP**

There is increasing knowledge describing the immunological role of the PARP family that supports the combinatorial uses of PARP inhibitors (PARPi) and immunotherapies against cancer. It is known that PARPs have a role during inflammation, innate immunity and in immune cells. PARP enzymes interact with transcription and adhesion factors which are involved in the regulation of cytokines and inflammatory mediators related to different aspects of inflammation [81]. Several studies showed a PARP-1-dependent activation of NFkB, a major transcription factor during the inflammation process [82–84]. PARP-1 is required to trigger NEMO SUMOylation and monoubiquination which is necessary for NEMO and NFkB activation [84]. There are a complete set of other transcription factors involved in inflammation which are also modulated by PARPs like sp1 [85], NFAT [86] or SIRT1 [87] among others.

Otherwise, PARP-1 and PARP-2 regulate several common inflammatory factors and cytokines including Tumor Necrosis Factor-α, inducible Nitric Oxide Synthase or Interleukin1-β suggesting an overlapped mechanisms or regulation [88]. In addition, PARP enzymes are involved in regulating the expression of cytokines and chemo attractants like IL-6, IL-12 or CCL2. PARP-14 enhances the STAT6 regulation of the expression in Th2 cells through IL-4 induction, which is important for the immune function in the lung [89].

PARP-1 is involved in development and activation of different immune cell types like macrophages, microglia or dendritic cells. Moreover, PARP-1 and PARP-2 induce pro-inflammatory effects not only restricted to innate immune system cells but also important in dendritic cells and fibroblasts [90]. PARP enzymes are also involved in T-cells development. PARP-1 and PARP-2 expression is especially high in the lymphocyte-proliferative areas of the thymus. Moreover, inactivation of PARP-2 decreases the size of the thymus while reducing the number of CD4+/CD8+ (double positive) thymocytes, due to an affected survival of double positives [91]. In addition, other studies showed an impaired capacity to differentiate into Th2 cells in PARP-1 KO cells [92].

The stimulator of the interferon genes (STING) pathway was primarily described as a mechanism activated in response to microbial infections and DNA viruses, but also has relevance under certain autoimmune and inflammatory conditions. There is currently a list of new evidence suggesting a role of the STING pathway in tumor detection [93,94]. Activation of STING is produced by the accumulation of cytosolic DNA fragments which interact with the cGAMP synthase (cGAS), catalyzing the formation of the second messenger GAMP to activate STING [95]. After its induction, STING activates TBK1 which, in turn, phosphorylates STING and the interferon regulatory factor 3 (IRF3). Then IRF3 migrates to the nucleus, causing the overexpression of type I interferon genes, including interferon beta (IFNβ). STING and type I interferon beta signaling pathways are involved in T-cell priming and activation against tumor-associated antigens in the tumor microenvironment [96,97].

After the detection of cytosolic DNA and the activation of the cGAS-STING-TBK1-IRF3 axis leading to the activation of type I interferons, there is an observed induction of cytokines involved in T cell chemotaxis, as CCL5 or CXCL10. It has been reported that type I IFN production and CCL5 or CXCL10 expression correlate with the infiltration of cytotoxic lymphocytes CD8<sup>+</sup> in several cancers [98]. Jianfeng Shen et al. reported a mechanism describing PARPi modulation of immune responses against cancer, even independently of *BRCA1*/*2* mutational background, through the STING pathway, further enhanced by blocking different immune checkpoints [99].

It is also relevant that tumor mutational burden (TMB) serves as a good predictor of response to immunotherapy, since it correlates with the sensitivity of tumors to the immune checkpoint blockade on immunotherapies like antiPD-1/PD-L1 or CTLA4 [100]. In other words, the immunogenicity of a given tumor depends, in part, on the mutational load and subsequently on its neoantigen repertoire. Recognition of such neoantigens is considered a major factor in the efficacy of clinical immunotherapies [101]. Nonetheless, cancers presenting elevated copy numbers such as ovarian cancer and small cell lung cancer are not immunologically hot but have extraordinary levels of damaged DNA/chromosomes and there exists controversy about the neoantigen hypothesis. Some groups have shown that the vast majority of mutations within expressed genes in cancers do not lead to the formation of neoantigens that are recognised by T cells [102,103].

Mutational load range varies over several orders of magnitude between different types of tumors [104,105], but ultimately, it will be the consequence of a balance between different factors including DNA damage and DNA repair function. Conditions affecting these factors are therapeutically exploited with different approaches, using DNA chemotherapy damaging agents or radiation. This is especially important with genetic backgrounds comprising an impaired DNA repairing machinery or inhibitors of DNA repair proteins involved in DNA damage response pathways as single agents or in combination with the DNA-damaging agents [106].

As previously indicated, PARP-2, PARP3, and especially PARP-1 became catalytically active in response to DSB, recruiting proteins that are involved in chromatin remodeling and DNA repair [107]. If the tumor immune response is modulated by the mutational load of the cells and the subsequent presence of neoantigens, it seems reasonable to hypothesize that PARPi could increase the tumor mutational burden due to an impaired DNA repair pathway function, contributing to the immunotherapy efficacy. There are currently several clinical trials exploring this possibility mainly but not exclusively in ovarian cancer [106,108].

Programmed cell death-1 (PD-1) is an immune receptor mainly expressed on activated CD4+ and CD8+ T cells or peripheral B cells [109]. Interaction of PD-1 and its ligand PD-L1 is critical to control the immune response, providing its binding constitutes an immune inhibiting checkpoint which leads to immune evasion. PD-L1 can be induced in cancer cells by a variety of stimuli such as T cell interferon gamma production or ionizing radiation (IR) [110] among others. The radiation-dependent PD-L1 activation seems to be related to DSBs and DNA repair response pathways and synergistically enhances antitumor immunity if applied together with immune checkpoint inhibitors [111]. This enhanced antitumor immunity seems to be related to increased mutational burden which increases neoantigen repertoire and tumor infiltrating lymphocytes (TILs) [112].

Interestingly, several studies also reported increased intratumoral CD8<sup>+</sup> T cell infiltration and interferon production after PARP inhibition [99,113]. Nevertheless, this increased presence of antitumor immunity and TILs can be counterbalanced by PARPi-induced expression of PD-L1 which subsequently activates the PD-1/PD-L1 immune checkpoint pathway [114].

Taken together, these observations provide further rationale for the combinatorial uses of DNA damaging agents, PARP inhibitors, and immune checkpoint blockade, in order to increase the benefits of the enhanced antitumor immunity avoiding the immunosuppressive effects of PD-L1 overexpression (Figure 7).

**Figure 7.** PARP inhibitors together with immune checkpoint inhibitors potentiate antitumor immune-mediated response. Activated T lymphocytes react against tumor antigens. The inhibition of PARP increases the number of lymphocytes infiltrating the tumor after the upregulation of chemokines, promoting an immune response mediated by CTLs. In spite PARP inhibitors modulate positively the upregulation of PDL-1 (favouring tumor scape from immune control) the anti-CTLA4 activates T cells to promote an antitumor response. Anti PD1/PDL-1 reverses CTL inhibition provoked by PARP inhibitor –induced PDL-1 expression. In this way, anti-PD1/PDL-1 can synergize with PARP inhibitors to ammeliorate antitumor immune response.

#### **3. PARylation in Autophagy**

One of the main characteristics of the metabolically hyperactive cancer cells is their high demand for nutrients and oxygen from their microenvironment. This makes them vulnerable to the deficiency of both oxygen (hypoxia) and nutrients (starvation). Both situations can induce metabolic stress leading to PARP-1 over activation; during these situations, cells must activate the autophagic pathway as a response to adapt and survive. In this context it has been shown that one of the first consequences of autophagy is the activation of poly-(ADP-ribosyl)ation. The combined effect of PARP-1 activity leading to PAR modification of AMPK (the main cell energy sensor) and the nutritional status (measured via mTORC1 activation) are fundamental during the first stages of autophagy [115].

The autophagyc pathway is present in all eukaryotic cells. It consists on a "self-eating" process necessary for the maintenance of the cell homeostasis. Through this lysosomal-dependent pathway different cellular organelles, portions of the cytosol and chaperone-associated cargoes are enfolded in double-membrane spheroids called autophagosomes, being then hydrolyzed by lysosomal enzymes [116]. Different cellular stresses can fire the autophagic response: hypoxia, pH variation, mitochondrial ROS production, DNA damage, intracellular pathogens, unfolded proteins or endoplasmic reticulum (ER) stress among others.

Expanding evidence indicates how autophagy is induced after DNA breakage. Ataxia Telangiectaxia Mutated (ATM) has been described as an important link between DNA Damage Response (DDR) and the induction of autophagy [117]. In response to DNA damage by mitochondrial ROS, external toxins or irradiation, ATM is auto phosphorylated within a Mre11, Rad50 and Nbs1 (MRN) multiprotein complex that binds DSBs. Once active ATM induces the activity of AMPk and its target tuberous sclerosis protein (TSC2), this leads to the inhibition of mTORC1, promoting the formation of autophagosomes dependent of Unc-51 like autophagy activating kinase (ULK1) [118]. This is how autophagy acts as a catabolic pathway of rapid energy recovery that the cell will use to grow and expand. However, this rapid acquisition of energy has two sides; on the one hand, it responds to a pro-survival role as long as the levels of autophagy remains biologically sustainable, this way, the degradation of excess vital components of the cell is avoided; but autophagy might also become a process of cell death due to the accumulation of autophagosomes degrading essential cellular structure;, this process is called autophagic cell death (ACD).

As we know, PARPs are important guardians of the genome integrity and they have been shown to link DNA damage response with the activation of autophagy [115,119]. Autophagy is also initiated in several settings of response to chemotherapy and mediated by PARP-1. Doxorubicin treatment leads to over-activation of PARP-1, followed by ATP and NAD+ depletion, triggering the non-toxic accumulation of autophagosomes [19]. In response to methylnitronitrosoguanidine (MNNG) double knockout Bax−/−Bak−/−MEFs activate PARP-1, reducing intracellular ATP levels and activating AMPk pathway and mTORC1 down-regulation. Suppression of AMPk pathway blocks MNNG-induced autophagy and enhances cell death [19]. The same results were found in a nasopharyngeal carcinoma model exploring over-activation of PARP-1, upon ionizing radiation there is a PARylation-dependent energy depletion and up-regulation of AMPk and ULK1 pathways [120]. All these studies have demonstrated that autophagy should be contemplated as a target in cancer during the induction of DNA damage and consequently, new strategies based on the synthetic lethality concept during PARPi must be explored.

In addition to DNA damage, nutrient deprivation can be considered as the most physiological stimulus to induce reversible autophagy. Recent studies of our group have implicated PARP-1 in autophagy induced by nutrient starvation [115,121]. The absence of PARP-1 or the use of PARP inhibitors (PJ34, DPQ and Olaparib) delay starvation-induced autophagy. Mitochondrial ROS accumulation derived from starvation promoted DNA damage and PARylation signaling from PARP-1 activation, which triggers ATP depletion, AMPk activation and mTORC1 inhibition. The lack of PARP-1 activity compensates ATP depletion compromising the cascade AMPk/mTORC1 inhibition/ autophagy. Starved PARP-1-deficient/inhibited cells showed increased apoptotic cell death [115]. In order to analyze the in vivo consequences of PARP-1 ablation on autophagy both PARP-1+/+ and PARP-1−/−pups were starved for short periods of time, concluding that PARP-1−/−neonates display a deficient liver autophagy response following acute starvation, showing that PARP-1 activity and PAR formation are key players in the decision of the cell to engage autophagy [115].

The absence of PARP-1 compromises the activation of AMPk, more precisely the isoform AMPkα, reducing phosphorylation on Thr172 [121,122]. A nuclear subpopulation of AMPkα was detected in the breast cancer cell line MCF7, where it forms a stable complex with PARP-1; moreover the activation of nuclear AMPkα requires two essential events: Firstly, starvation-induced ROS must be imported to the nucleus to generate DNA damage and PARP-1 activation. Secondly, PARP-1 modifies by PARylation the nuclear population of AMPkα. Energy depletion associated with starvation was increased during non-lethal over-activation of PARP-1 in response to ROS, resulting in a feedback loop that favored the interaction between PARP-1/AMPkα and promoted the activation of LKB1. The mechanism of modulation by PARylation of LKB1 is not clearly understood but using PARPi or during specific assays with siRNA PARP-1, LKB1 activation is compromised [121].

In non-starved cells, PARP-1 forms a complex with AMPK; after starvation PARP-1 is activated and PARylation disrupt the PARP-1/AMPK complex, this triggers the nuclear export of PARylated AMPkα to the cytoplasm. This significant PARylated AMPkα population, is recognized by the active form of LKB1 promoting the phosphorylation on regulator site Thr172 of AMPkα. Again, a positive feedback loop takes place between the cytosolic energy depletion and PARylated AMPkα to potentiate the activation of the cytosolic AMPkα population by LKB1. The presence of covalently PARylated LKB1 has not been demonstrated but the effect of PARPi using PJ34 or olaparib suggested the implication of PAR on LKB1. These data place the nucleus (a classical cellular component not associated with autophagy) and the interaction PARP-1 and AMPk as an initial sensor of the metabolic alterations derived from perturbations in the nutritional extracellular status, not necessarily related with the alterations in genomic integrity. PARylation could be considered as an alert indicator of changes in the energy balance in cancer cells (Figure 8).

**Figure 8.** Regulation of autophagy by PARylation. (1) PARP1 forms a complex with AMPKα in nucleus. (**A**) PARP activation conditions: During starvation-induced autophagy, ROS release leads DNA damage and PARP1 overactivation. Self-PARylated PARP1 interact with AMPKα1 subunit (2). The complex is disrupted and PAR-AMPKα is exported to cytosol (3). The presence of PAR-AMPK and the continuous absence of amino acids and ATP depletion favor total activation of AMPKα population by LKB1, inhibition of mTORC1, interaction PAR-phospho-AMPK/ULK1, and autophagosome formation (4). LKB1 activity is presumably modified in a PARylation-dependent manner. (**B**) PARP inhibited: Starvation-induced ROS production was abrogated after PARP inhibition. Following AMPKα1/PARP1 interaction (1) the AMPKα1 subunit is not PARylated and AMPK nuclear export is inhibited (2 and 3). In spite of nutrient and energy depletion, AMPKα is inactive; mTORC1 is partially activated and interacts with ULK1, favoring its inhibition (4). Finally, the autophagosomes production will be delayed.

#### **4. Cancer Initiating Cells**

Cancer Stem Cells (CSCs) are defined as the tumor cell subpopulation that present remarkably high plasticity, considering "plasticity" as the ability of pluripotent cells to induce trans-differentiation. The first evidence of cancer initiating stem-like cells was described in 1994 by Lapidot et al. and Caceres-Cortes et al. after injecting different leukemia cell subpopulations in mice [123]. Through the years, CSCs were determined in a wide multitude of cancers like breast [124], gliomas [125], prostate [126], melanoma [127], lung [128,129], colon [130,131], pancreas [132], head and neck squamous cell [133], liver [134] and renal carcinoma [135].

CSCs are a very aggressive subpopulation of cancer cells characterized by their self-renewal capacity, tumor initiation ability, resistance to chemo- and radiotherapy and multi-lineage differentiation (stemness) [125,136,137]. The presence of this subpopulation in tumors is considered to be a marker of bad prognosis and therefore they have been proposed as target in cancer therapy.

For several years, it has been known that reprogramming processes are led by specific genetic programs. Yamanaka et al. demonstrated that the ectopic expression of four genes (Oct4, Sox2, c-Myc and KLF4) in mouse fibroblast transforms them in embryogenic stem cells. These cells, denominated by the authors as induced-pluripotent stem cells (iPSCs), were able to generate cells from any germ layer [138]. This breakthrough made Shiya Yamanaka win the Noble Price of Medicine in 2012. Today, these genes are known as Yamanaka factors and several researches link the expression of these genes with cancer [138–140]. In fact, the risk of tumor development is the most important limitation in the application of iPS cell-based therapy. Tumor cells present processes of trans-differentiation and dedifferentiation or reprogramming acquiring features of CSCs [138].

In 2012, it was demonstrated that PARP-1 is involved in the reprograming process, promoting iPSCs. An intense PARylation was detected in iPSCs. Moreover, PARP-1 knockdown reduced the capacity of iPSCs generation after Yamanaka factors overexpression. It has been found that c-Myc directly regulates PARP-1 expression and PARylation. Over-expression of PARP-1 compensates the knockdown of c-Myc in reprogramming in MEFs [141]. In 2009, Gao et al. demonstrated that PARP-1 PARylates and controls Sox2 levels, regulating its activity. They found that PARylated Sox2 increases its stability and therefore induces the expression of its target gene FGF4. PARP inhibition has been observed reverting this scenario [142,143]. To reprogram cells to iPSCs by Yamanaka factors, an epigenetic remodeling is also necessary. PARP-1 regulates modification in histones that alter the chromatin pattern driving to pluripotent cell phenotype. In this line, PARP-1 is necessary to promote the access to chromatin of Oct4 [144].

Telomerase (TERT) is a reverse transcriptase necessary for telomeres elongation in embryonic stem cells. This retrotranscriptase is highly expressed in tumors and the expression is related with stemness capacity. KLF4 directly interacts with the promoter of TERT. PARP-1 interacts with KLF4 and mediates the expression of TERT in CSCs. In fact, PARP-1 suppression dramatically reduces the recruitment of KLF4 to the promoter of TERT, reducing its expression [145]. C-Myc has also been related with TERT expression, suggesting another possible relationship between PARP-1 and telomerase [146].

Due to the high resistance to radio and chemotherapy presented by CSCs, many studies focused their interest in this subpopulation as a target for future treatment. In glioblastoma (GBM), one of the most aggressive malignancies, the presence of CSCs was proposed as the main cause of tumor relapse [138]. Furthermore, reprogramming events driving the formation of endothelial cells from tumor cells have been described [147]. The importance of CSCs in this kind of tumors led Vescovi el al. to propose the differentiation of CSCs as a target, with the aim of reducing the malignancy of this brain cancer [148]. They found that glioma stem cells treated with bone morphogenetic proteins (BMPs), cytokines belonging to TGF-β superfamily, differentiated CSCs to non-stem glioma cells. BMPs interact with membrane receptors inducing differentiation through SMADs signaling pathway. It has been demonstrated that PARP-1 negatively regulates this pathway at different levels. Ectopic expression of PARP-1 suppresses the signaling mediated by BMP. On the contrary, knockdown of PARP-1 promotes BMP-mediated differentiation. Furthermore, PARG, an enzyme responsible for PAR degradation, plays a positive role in the pathway [149,150]. On the other hand, SMAD signaling activation can be a double-edged sword in GBM because the stimulation of this pathway with TGF-β induces the expression of leukemia inhibitory factor (LIF), a cytokine that induces maintenance of "stemness" capacity [151].

#### **5. Conclusions and Perspectives**

The PARP superfamily consist of a group of proteins characterized by the presence of what is called a "PARP signature" on their sequences. Their main characteristic activity is the process referred to as PARylation. Through NAD+ and ATP consumption, ADP-Ribose is generated and then transferred as poly or mono ADP-ribose to different target factors. These modifications alter their activity or stability having relevant implications on the cellular metabolism. Moreover, excessive PARylation has an impact itself: free PAR can function as a signaling molecule and its synthesis may blunt NAD+ and ATP levels producing cell death.

PAR synthesis is activated mostly during DNA damage, altered DNA configuration or when PARP present some posttranscriptional modifications. As we summarized in this review, PARP activation has key repercussions on the cell fate affecting processes like DNA repair, transcriptional regulation, DNA remodeling, hypoxic response, epithelial mesenchymal transition, angiogenesis, autophagy, inflammation and cancer stem cell programming. All these processes lead to changes in survival, proliferation, differentiation, or even malignant transformation. Considering the relevance of the previously enumerated processes, it is easy to understand the relevance of PARP during cancer development. Cells overexpressing PARP will be more likely to repair DNA damage induced by genotoxic agents, they will adapt better to hypoxia and will be prone to produce metastasis through angiogenesis and EMT. Four PARP inhibitors have been already approved by the FDA (olaparib, rucaparib, niraparib and talazoparib) and they are used today as a result of their ability to generate "synthetic lethality" on BRCA 1/2 mutated tumors (see [152] for a review). However, ongoing advanced clinical trials will most likely expand their prescription as is the case for the combination of PARP inhibitors with classical therapies (Table 1) and with anti-angiogenic treatment (Table 2).

Currently, one of the limitations facing this therapeutic option is considering mostly on the BRCA mutated cells as HR deficient. Knowing that more than 100 genes are involved on this DNA repair pathway, it looks likely that BRCA proficient cells could be still HR defective due to other mutations. Measuring the whole genomic instability within a tumor by surveying the loss of genetic heterozygosity, telomeric allelic imbalance or the extent of somatic mutations, could be a more precise approach to define PARP inhibition sensibility, making more patients candidate of this therapeutic approach.

Another important limitation that needs to be worked out is the emergence of resistances, especially during PARP inhibition in monotherapy (Figure 9).

**Figure 9.** Summary of PARP inhibition approaches during cancer treatments, different possible outcomes and ways designed to overcome the possible appearance of resistances.


*Cancers* **2020**, *12*, 739




**Table 2.** Di fferent clinical trials on cancer treatment combining PARP inhibition with anti-angiogenic strategies.

The generation of a selective force (like the caused by a drug treatment) in a highly mutagenic context can lead to the selection of resistant clones. They will then reproduce, forming a relapsed tumor resistant to the original treatment. To avoid this undesired consequence, the selective pressure can be reduced by spacing the treatments or by combining it with inhibitors for different targets, generating a constant pressure over the tumor while generating reduced selective forces.

Another way to avoid resistance is to study the genetic background and predict the more probable resistance mechanisms to arise, then combining the PARP inhibitor with secondary inhibitors for the factors related with the expected resistance. It is known that HR restoration, checkpoints activation, stalled DNA protection or drug efflux among others, can lead to PARP inhibition resistance. Combining PARP blockade with inhibitors for the possible ways of scape the treatment is being tested in clinical trials in order to make PARP inhibition more effective.

Based also in the studies summarized here, we propose that PARPi is an expansive field that may have therapeutic value beyond synthetic lethality. To this end, a precise comprehension of the implications of the different PARPs with PARylation in the complex tumor ecology is needed, including analysis of PARylation in both tumor and associated non-tumor cells, using single cell analysis and its consequences in tumor adaptation to hostile conditions. In this context, we can affirm that the use of PARP inhibitors against cancer treatment is not just a promising field but a reality and the challenge exists to widen their use by identifying new properties and deepening the role of PARP in tumor biology.

**Funding:** This work was supported by Ministerio de Economía y Competividad, Spanish Ministry of Economy and Competitiveness SAF2012-40011-C02-01, SAF2015-70520-R, RTICC RD12/0036/0026, RTI2018-098968-B-I00, CIBER Cáncer ISCIII CB16/12/00421 and Fundación Domingo Martínez to FJO. JMMC, is recipient of a FPI fellowship from Spanish Ministerio de Ciencia; MFC and EZM are recipients of FPU fellowships from the Spanish Ministerio de Ciencia.

**Conflicts of Interest:** The authors declare "no conflict of interests".

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **ATM-Deficient Cancers Provide New Opportunities for Precision Oncology**

**Nicholas R. Jette 1, Mehul Kumar 1, Suraj Radhamani 1, Greydon Arthur 1, Siddhartha Goutam 1, Steven Yip 2, Michael Kolinsky 3, Gareth J. Williams 1, Pinaki Bose <sup>1</sup> and Susan P. Lees-Miller 1,\***


Received: 14 February 2020; Accepted: 12 March 2020; Published: 14 March 2020

**Abstract:** Poly-ADP ribose polymerase (PARP) inhibitors are currently used in the treatment of several cancers carrying mutations in the breast and ovarian cancer susceptibility genes *BRCA1* and *BRCA2*, with many more potential applications under study and in clinical trials. Here, we discuss the potential for extending PARP inhibitor therapies to tumours with deficiencies in the DNA damage-activated protein kinase, Ataxia-Telangiectasia Mutated (ATM). We highlight our recent findings that PARP inhibition alone is cytostatic but not cytotoxic in ATM-deficient cancer cells and that the combination of a PARP inhibitor with an ATR (ATM, Rad3-related) inhibitor is required to induce cell death.

**Keywords:** ATM; olaparib; ATR; PARP; PARP inhibitor; prostate cancer; pancreatic cancer; lung cancer

#### **1. PARP and PARP Inhibitors**

Genome instability, characterized by the accumulation of mutations and chromosomal alterations in the genome, is both a hallmark and a driver of cancer [1,2]. Yet the same genomic alterations that predispose a cell to cancer may also render cells susceptible to targeted therapies. Accordingly, a goal of precision oncology is to achieve better cancer control by targeting therapy to specific genetic defects or aberrations in the tumour, while causing less damage to normal tissue and consequently, fewer side-effects. One of the most dramatic examples of success in this area has been the use of poly-ADP-ribose polymerase (PARP) inhibitors in the treatment of patients with tumours that harbour inactivating mutations in the breast and ovarian cancer susceptibility genes, *BRCA1* and *BRCA2*.

PARP was identified in the 1960s as an enzyme that metabolizes nicotinamide adenine dinucleotide, NAD<sup>+</sup> [3]. Early studies indicated that inhibition of PARP blocked repair of DNA strand breaks and PARP inhibitors were soon considered as potential radiation sensitizers [4]. The first PARP inhibitor, a simple analogue of nicotinamide was generated in 1971 [5] and, given the reported roles of PARP in cell death and ischemia as well as DNA repair, there was increasing interest in the clinical applications of PARP inhibition [6,7]. PARP is now established as one of a family of poly-ADP polymerases of which PARPs 1, 2 and 3 have roles in the DNA damage response [8]. Of these, PARP 1 is the most well-studied and the one to which we will refer in this review. PARP1 binds avidly to ends of DNA that occur at single and double DNA strand breaks and then, in a process called PARylation, auto-modifies to create long polymers of poly-ADP-ribose (PAR). PAR chains interact with proteins involved in DNA

repair and other pathways, recruiting them to sites of DNA damage [9]. Importantly, PARylation also serves to disengage PARP from DNA ends [10]. Most PARP inhibitors in use today prevent PARylation, leading to trapping of PARP at DNA ends [11]. It was originally proposed that PARP inhibition blocked base excision repair and single strand break repair pathways thus increasing reliance on BRCA-dependent repair. However, this model has been challenged [12] and more recent studies have shown that olaparib reduces cell proliferation by inducing replication stress [13] and that olaparib sensitivity is due to engagement of homologous recombination repair (HRR) at replication forks [14].

In 2005, two seminal papers were published demonstrating that breast cancer cells with siRNA depletion of BRCA1 or BRCA2 were exquisitely sensitive to the PARP inhibitor NU1025 [15,16]. *BRCA1* and *BRCA2* are cancer predisposition genes that are inactivated in ~25% of inherited breast cancers, ~15% of all ovarian cancers and several other cancers, suggesting that PARP inhibitors might have potential in treating a wide-range of patients with BRCA-deficient tumours [17]. The PARP inhibitor AZD2881, also known as olaparib or LynparzaTM, showed promise in mouse models of breast cancer [18] and quickly moved into clinical trials, showing anti-tumour activity in BRCA-mutated cancers, even in phase I studies [19]. Olaparib, the first PARP inhibitor to gain regulatory approval, is now FDA-approved in advanced ovarian [20], breast [21], pancreatic [22] and prostate cancers [23], with the PARP inhibitors rucaparib [24], niraparib [25] and talazoparib [26] also FDA-approved in varying indications [17,27].

BRCA1 and 2 play critical roles in detection, signalling and repair of DNA double strand breaks (DSBs) via the HRR pathway. HRR is active in S phase at stalled replication forks and in G2 phase of the cell cycle after DSBs have been resected to contain long ssDNA overhangs on their 3' ends [28]. These long regions of ssDNA are bound by replication protein A (RPA) and BRCA2 plays a role in the replacement of RPA with RAD51, the protein that initiates strand invasion and the search for a homologous DNA sequence during HRR [28]. BRCA1 interacts with BRCA2 via the PALB2 protein, and is recruited to DNA damage-induced foci where it participates in activating DNA repair and cell signalling pathways [29]. Given the encouraging early results showing PARP inhibitor sensitivity in BRCA-deficient cells, screens were initiated to identify other proteins that when knocked down with siRNA might confer sensitivity to PARP inhibitors [30–32]. One of these was Ataxia Telangiectasia Mutated (ATM).

#### **2. ATM**

ATM is a member of the phosphatidylinositol-3 kinase-like (PIKK) family of serine/threonine protein kinases with critical roles in the cellular response to DNA damaging agents, such as ionizing radiation (IR), that produce DSBs [33]. Like other members of the PIKK family, ATM is a large protein of over 350 kDa that is composed of an extended N-terminal region containing multiple HEAT (Huntingtin, Elongation factor 3, A subunit of protein phosphatase 2A and mammalian Target of rapamycin) repeats and a C-terminal kinase domain that has amino acid similarity to phosphatidyl inositol-3 kinase (PI3K) and is flanked and stabilized by conserved FRAP–ATM–TRRAP (FAT) and FAT-C domains. Generation of DSBs and/or changes in chromatin structure lead to activation of ATM and its autophosphorylation on serine 1981 [34]. Activated ATM phosphorylates a multitude of downstream targets including p53, checkpoint kinase 2 (Chk2) and histone H2AX [35]. Indeed, phospho-proteomics studies have identified hundreds of PIKK-dependent, DNA damage-induced phosphorylation events in cells [36,37]. Consistent with its role in the repair of IR-induced DSBs, cell lines with loss or inactivation of ATM are radiation sensitive, have cell cycle checkpoint defects [38] and have defects in slow repair of complex DNA damage lesions and DSBs in the context of heterochromatin [39]. Recently, roles in preventing premature aging [40] and in reactive oxygen sensing [41] have also been reported.

Germline inactivation of both copies of the *ATM* gene causes Ataxia-Telangiectasia (A–T), a devastating childhood condition characterized by ataxia (wobbly gait), telangiectasia (blood vessel abnormalities) and progressive neurodegeneration, particularly in the cerebellum, that renders its victims wheelchair-bound. A–T patients also have immune defects and cancer predisposition and usually succumb to their condition in their early twenties [42]. Accordingly, cell lines derived from A–T

patients and ATM knock out mice are hypersensitive to IR and other chemotherapeutic agents [43,44], raising the possibility that cancers with loss of ATM may be more sensitive to DNA damaging agents than their ATM-proficient counterparts [45].

#### **3. Targeting ATM-Deficient Cancers**

Genome sequencing has revealed that *ATM* is mutated in a variety of human cancers, including mantle cell lymphoma (MCL), colorectal, lung and prostate cancers. Analysis of *ATM* mutation frequency in The Cancer Genome Atlas (TCGA) cohort using c-Bioportal [46,47] indicates that ATM is mutated in approximately 5% of all cancers, with some, such as MCL, with a much higher mutation frequency of ~40% (Figure 1A). Similarly, ATM is mutated in ~20% of colorectal and uterine cancers and approximately 10% of prostate and lung cancers (Figure 1A). The vast majority of these mutations are missense mutations and are scattered throughout the coding region (Figure 1B and [45]). An exception is R377C/H, which occurred in 74 of the 2263 (~3%) of the cancers queried (Figure 1B), and has been identified as a cancer mutation hotspot [48,49]. The R337C/H mutation was prominent in colorectal cancer, but not in prostate, lung or pancreatic cancer (Figure 1C–F). Although the functional consequences of this and most other *ATM* mutations is not known, given that in A–T many mutations in *ATM* induce protein truncation, protein destabilization and resulting loss of function [50], combined with the fact that siRNA-mediated loss of ATM in cancer cell lines results in PARP inhibitor sensitivity [30–32], it seems likely that many cancers with *ATM* mutation that lead to loss of function could be candidates for PARP inhibitor treatment.

Given that *ATM* is mutated or lost in over 40% of MCL [51], we and others examined the effects of ATM loss on PARP inhibitor sensitivity in human lymphoma cell lines that lack ATM protein expression. These cell lines were more sensitive to olaparib than their ATM-proficient counterparts in both cell line and animal models [52–54]. Moreover, the ATM kinase inhibitor KU55933 enhanced sensitivity to olaparib in ATM-proficient cells indicating that ATM kinase activity protects from PARP inhibitor sensitivity [53]. Similar results were observed in gastric cancer cell lines [55], and in colorectal cancer cell lines with shRNA depletion of ATM [56]. Deletion of ATM in mouse models of lung cancer and pancreatic cancer also induced sensitivity to PARP inhibitors and/or DNA damaging agents, as did inhibitors of the related protein kinase ATR (ATM and Rad3-related) [57,58].

We observed that in MCL, gastric and colorectal cancer cell lines with loss or down regulation of ATM, sensitivity to olaparib was enhanced when *TP53* was also mutated or deleted [52,53,55,56]. However, in mouse models, both *TP53*-proficient and deficient cells were sensitive to olaparib [58]. Therefore, the effect of p53 status on PARP inhibitor sensitivity requires further clarification. Although co-mutation of both *ATM* and *TP53* is rare [59], co-mutation has been shown to occur in 2–3% of non-small cell lung cancer where it increases tumour mutation burden and correlates with better response to immune checkpoint therapy [60], suggesting additional opportunities for targeted therapy for ATM-deficient tumours.

To address the mechanism of olaparib-induced cell sensitivity in human cells lacking ATM, we used CRISPR/Cas9 to delete ATM from the p53-proficient lung adenocarcinoma cell line, A549. In keeping with recent findings, olaparib alone reduced cell proliferation [13], but surprisingly, did not induce cell death [61]. Rather, olaparib was found to induce reversible G2 arrest in ATM-deficient A549 cells [61]. Since the related protein kinase ATR plays a critical role in the G2 checkpoint [62], and given ATM-deficient cells are sensitive to ATR inhibitors [58,63–65], we asked whether inhibition of ATR using VE-821 [66,67], would ablate G2 arrest and induce cell death in olaparib-treated ATM-deficient cells. This was indeed found to be the case. Combined treatment with olaparib plus ATR inhibition with VE821 induced cell death only in ATM-deficient A549 cells, suggesting that patients with ATM-deficient tumours could benefit from a combination of PARP and ATR inhibitors [61].

**Figure 1.** Frequency of Ataxia-Telangiectasia Mutated (*ATM)* mutations in human cancer. (**A**) *ATM* was queried against all entries in the curated non-redundant data set on c-Bioportal (references [46,47]) accessed January 2020. Duplicate studies were removed and copy number variations are not included. The frequency of *ATM* alteration in various cancers is shown. (**B**) ATM is a 3056 amino acid protein consisting of a N-terminal TAN (telomere length maintenance and DNA damage repair) domain (residues 7–165), and a C-terminal kinase domain (residues 2714–2961) flanked by FAT (2097–2488) and FATC (3205–3055) domains. The location of mutations in ATM from all samples in the curated non-redundant data site available on c-Bioportal (references [46,47]), accessed January 2020 (duplicate sets removed and copy number variation not include) is shown. Mutations were distributed across the entire the coding region however, one mutation R337C/H was detected in 74 out of 2263 samples, across all cancers. (**C**) The R337C/H mutation was frequent in bowel cancer (22 out of 331 samples) but less so in lung (panel **D**), prostate (panel **E**) and pancreatic cancers (Panel **F**).

*ATM* is frequently mutated in prostate cancer [68], as well as somatic and hereditary forms of pancreatic cancers [69,70], suggesting that patients with these cancers might also benefit from treatment with a PARP inhibitor. Nevertheless, PROFOUND, a phase III trial that examined the clinical efficacy of olaparib versus standard treatment (abiraterone acetate/enzalutamide) in patients with metastatic, castration-resistant prostate cancer (mCRPC) and HRR gene alterations, revealed less impressive rates of radiographic progression free survival benefit with olaparib in patients with *ATM* alterations, in contrast to patients with other HRR gene alterations (e.g., *BRCA2*), based upon exploratory, hypothesis-generating gene-by-gene subgroup analysis [23]. Thus, it appeared necessary to consider therapeutic approaches that may enhance the efficacy of PARP inhibition in ATM-deficient cancers.

We therefore examined whether the combination of olaparib plus an ATR inhibitor would be effective in cell line models of prostate and pancreatic cancer. We depleted ATM from the prostate cancer cell line PC-3 using CRISPR/Cas9 and found that although olaparib reduced cell proliferation, ATM-deficient cells did not undergo apoptosis unless olaparib was combined with an ATR inhibitor, either VE-821, as in [61] or AZD6738 [71], an ATR inhibitor in clinical trials [72]. Significantly, olaparib and AZD6738 had little effect on ATM-proficient cells either alone or in combination [71]. Similar results were seen with the pancreatic cancer cell line, Panc 10.05, in which ATM was depleted by shRNA [71]. Thus, the combination of PARP and ATR inhibitors could be beneficial in a number of ATM-deficient cancers, including lung, prostate and pancreatic [71]. While A549 has wild type p53 [73], PC-3 are *TP53* null [74] and Panc 10.05 contain a homozygous mutation at I225N [75,76], therefore these data suggest that the sensitivity of ATM-deficient cells to the combination of PARP inhibitor and ATR inhibitor is not dependent on p53 status.

Our results also highlight the importance of the method used to assess cell viability in determining sensitivity of a cell line to a particular agent. Although ATM-deficient cells were highly sensitive to olaparib in clonogenic survival assays and the number of viable cells was decreased compared to ATM-proficient cells measured using the trypan blue exclusion assay, analysis of sub-G1 DNA or annexin staining did not reveal evidence of cell death [61,71]. Rather, olaparib-treated, ATM-deficient cells underwent reversible G2 arrest, and did not undergo cell death until an ATR inhibitor was also added [61,71]. Moreover, similar results have been seen in a patient-derived xenograft model of *BRCA*-mutant high-grade serous ovarian cancer, suggesting that PARP inhibition also increases reliance on ATR-dependent G2 arrest in BRCA-deficient cells [77], thus the combination of PARP and ATR inhibition may have benefits in other HDR-deficient cancers.

#### **4. ATM Mutation Versus Loss of Function: Identifying Patients Who May Benefit from PARP Inhibitor Treatment**

To date, most work from our lab and others has centred on the effects of olaparib on cell lines or mice in which ATM has been deleted [52–54,57,58,61,71] or inhibited with KU55933 [53,55]. An important difference is that in cancer, ATM is mutated, but the effects of these mutations on ATM function are, for the most part, unknown. As shown in Figure 1, literally hundreds of mutations have been identified in ATM and these mutations are scattered throughout the coding region. Apart from R337H/C, a hotspot mutation prevalent in colorectal cancer, other individual mutations are seen less frequently and their effects on ATM function is not known. Recently, three papers describe cryo-electron microscopy structures of Tel1, the well-conserved ATM homolog in lower eukaryotes, showing that ATM forms an autoinhibited dimer [78–80], providing insight into the conformational changes necessary for ATM activation. Three of these structures provide atomic models of the C-terminal kinase domain, and one, from a thermophilic fungus, also provides an atomic model of the majority of the N-terminal heat repeat domain in open and closed conformations. These structures provide the molecular basis to begin to understand cancer-associated *ATM* mutations. Initial analyses predict many cancer-associated mutations in the kinase domain are likely to impact ATM activity or protein folding [78,79], and the equivalent residue to R337 appears to stabilize the packing of two helices in the N-terminal domain, with the R337C/H mutation possibly destabilizing this region of ATM [78]. The methods developed to purify wildtype ATM homologs pave the way for more comprehensive studies to test the effect of cancer-associated mutations on ATM activities and stability.

For the treatment of cancer patients based on *ATM* status, it will be critical to identify those patients who have mutations that impact ATM activity or stability and are therefore most likely to benefit from PARP/ATR inhibitor combination treatment. This may be challenging by DNA sequencing alone, given the number of mutations identified and that their effects on ATM function are, for the most part, not known. That being said, preliminary clinical trials have shown promising, if not controversial results. In an initial trial of 60 men with mCRPC, 5 were shown to have mutation of *ATM* and 4 responded to olaparib [81]. A subsequent larger trial in a similar patient population demonstrated 7 of 19 patients with ATM mutation met at least one response criteria [82]. In contrast, a multicentre retrospective review of patients with mCRPC treated with olaparib showed no patients with *ATM* mutations achieved a PSA response and had significantly worse progression-free survival (PFS) and overall survival compared to *BRCA*-mutated patients. The addition of olaparib to paclitaxel failed to improve overall survival over paclitaxel alone in Asian patients with recurrent gastric cancer, both in the overall patient population as well as those with low or absent ATM expression, as determined by immunohistochemistry (IHC) [83]. Our findings that olaparib alone is cytostatic in ATM-deficient cells and that PARP and ATR inhibitors need to be combined to kill ATM-deficient cells is in line with these findings and suggest that this combinatorial approach could improve long-term survival in patients.

Despite the large number of uncharacterized mutations in *ATM* in cancer, many mutations known to impact ATM function, such as those resulting in A–T, frequently induce protein destabilization [50]. Therefore, determining the presence of ATM protein in cancer patient biopsy samples through IHC or other approaches may prove useful. Interestingly, although data from TCGA reveals that the *ATM* gene is mutated in 12–14% of patient samples with lung adenocarcinoma, another study has reported that ~40% of lung adenocarcinoma patients have low ATM protein expression by IHC [84]. A possible explanation for the apparent difference between *ATM* gene alteration and ATM protein expression could be methylation of the *ATM* promoter leading to transcription silencing [85]. Indeed, our analysis of methylation data in TCGA datasets revealed significant negative correlations between *ATM* promoter methylation and *ATM* gene expression in prostate adenocarcinoma (*p* = 1.962 <sup>×</sup> 10<sup>−</sup>7, Spearman's rho = −0.23), lung adenocarcinoma (*p* = 0.001159, Spearman's rho = −0.15) and colon adenocarcinoma (*p* = 0.0372, Spearman's rho = −0.13) (Figure 2). These findings suggest that ATM promoter methylation may play a role in the regulation of *ATM* gene expression in these cancer types.

Another possibility for stratifying patients who will benefit from a combination of PARP and ATR inhibitors would be to use a surrogate marker of ATM functionality, either through the use of phosphospecific antibodies to ATM itself or its downstream targets or an RNA signature specific to loss of ATM functionality. Indeed, the search for identifying tumours which exhibit "BRCAness" or HRR deficiency is an area of active investigation [27,86]. Answers to these questions and more may become apparent over the next few years as PARP inhibitor therapy is tested in more patients with defects in DNA damage response genes.

**Figure 2.** *ATM* promoter methylation and *ATM* gene expression in adenocarcinomas. Scatter plots showing the correlation between methylation beta values of the *ATM* promoter probe cg01756564 and *ATM* gene expression in TCGA datasets of (**A**) prostate adenocarcinoma (PRAD, *n* = 496), (**B**) colon adenocarcinoma (COAD, *n* = 276) and (**C**) lung adenocarcinoma (LUAD, *n* = 454). Spearman's rho values are indicated in the top right. Grey region of linear fit indicates 95% confidence interval. Asterisks in the top left indicate significance. \* = *p* < 0.05, \*\* = *p* <0.01, \*\*\* = *p* < 0.001.

#### **5. Concluding Remarks**

In conclusion, basic research into DNA damage repair biochemistry led to the identification of PARP, the inhibition of which, almost 40 years later, is showing great promise in the treatment of *BRCA*-deficient ovarian, breast, prostate and pancreatic cancer patients [87–89]. Work from our lab and others discussed here has shown that PARP inhibitors may also have potential in treating patients with ATM-deficient tumours. Our recent studies have revealed that in ATM-deficient cancer cell lines, olaparib is cytostatic not cytotoxic and that combination of olaparib with an ATR inhibitor is needed

to induce cell death [61,71]. Moreover, similar results were seen in models of BRCA-mutant ovarian cancer, suggesting that inhibition of ATR potentiates the effects PARP inhibition in *BRCA*-deficient cells [77], thus the combination of PARP and ATR inhibition may have benefits in other HDR-deficient cancers. It will be exciting to see whether this finding will also apply to other PARP inhibitors and, most importantly, in the clinic for improving outcomes for cancer patients. Indeed, several clinical trials using a PARP inhibitor in combination with an ATR inhibitor are ongoing (Table 1).


**Table 1.** List of ongoing clinical trials combining a poly-ADP ribose polymerase (PARP) inhibitor with an ATR inhibitor. Information obtained from https://clinicaltrials.gov, accessed March 9 2020.

It is also possible that ongoing preclinical studies will reveal new synthetic lethal interactions within the DNA damage response. Indeed, inhibitors to other proteins in the DNA damage response are being developed [27]. Identification of alternative DNA damage response genes to target in cancer could be useful as therapies in their own right and also in cases where tumours become resistant to other therapies, such as is currently observed in *BRCA*-deficient tumours treated with PARP inhibitors [90,91]. Finally, ATM has widespread cellular roles outside the DNA damage response, including roles in cellular redox signalling [41] and regulation of autophagy [92,93], apoptosis and other cell death pathways [94–98], further increasing potential opportunities to target ATM-deficient cells with novel therapies.

**Author Contributions:** Conceptualization, N.R.J., M.K. (Michael Kolinsky), S.P.L.-M.; investigation, N.R.J., M.K. (Mehul Kumar), S.R., G.A., S.G., S.Y., M.K. (Michael Kolinsky), G.J.W., P.B., S.P.L.-M.; writing—original draft preparation, N.R.J., S.R., G.A., S.G., S.Y., M.K. (Michael Kolinsky), S.P.L.-M.; writing—review and editing, N.R.J., M.K. (Mehul Kumar), S.R., G.A., S.G., S.Y., M.K. (Mehul Kumar), G.J.W., P.B., S.P.L.-M.; funding acquisition, S.P.L.-M. All authors have read and agreed to the published version of the manuscript.

**Funding:** S.P.L.-M. thanks the Alberta Cancer Foundation, the Cancer Research Society and the Pancreatic Cancer Society of Canada for support. Work in the S.P.L.-M. and G.J.W. laboratories is also supported by NIH program project grant CA92584.

**Conflicts of Interest:** S.Y. has received honoraria and/or consulting fees from Janssen, Pfizer, Roche, BMS, Merck, AstraZeneca, Bayer, and Novartis. M.K. (Michael Kolinsky) has accepted honoraria and/or consulting fees from Janssen, Ipsen, Astellas, BMS, Merck, AstraZeneca, Bayer, and travel support from Novartis. N.R.J., M.K. (Mehul Kumar), S.R., G.A., S.G., G.J.W., P.B. and S.P.L.-M. declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **The Controversial Roles of ADP-Ribosyl Hydrolases MACROD1, MACROD2 and TARG1 in Carcinogenesis**

#### **Karla L.H. Feijs 1,\*, Christopher D.O. Cooper <sup>2</sup> and Roko Žaja <sup>1</sup>**


Received: 15 January 2020; Accepted: 27 February 2020; Published: 5 March 2020

**Abstract:** Post-translational modifications (PTM) of proteins are crucial for fine-tuning a cell's response to both intracellular and extracellular cues. ADP-ribosylation is a PTM, which occurs in two flavours: modification of a target with multiple ADP-ribose moieties (poly(ADP-ribosyl)ation or PARylation) or with only one unit (MARylation), which are added by the different enzymes of the PARP family (also known as the ARTD family). PARylation has been relatively well-studied, particularly in the DNA damage response. This has resulted in the development of PARP inhibitors such as olaparib, which are increasingly employed in cancer chemotherapeutic approaches. Despite the fact that the majority of PARP enzymes catalyse MARylation, MARylation is not as well understood as PARylation. MARylation is a dynamic process: the enzymes reversing intracellular MARylation of acidic amino acids (MACROD1, MACROD2, and TARG1) were discovered in 2013. Since then, however, little information has been published about their physiological function. MACROD1, MACROD2, and TARG1 have a 'macrodomain' harbouring the catalytic site, but no other domains have been identified. Despite the lack of information regarding their cellular roles, there are a number of studies linking them to cancer. However, some of these publications oppose each other, some rely on poorly-characterised antibodies, or on aberrant localisation of overexpressed rather than native protein. In this review, we critically assess the available literature on a role for the hydrolases in cancer and find that, currently, there is limited evidence for a role for MACROD1, MACROD2, or TARG1 in tumorigenesis.

**Keywords:** ADP-ribosylation; cancer; macrodomain; ADP-ribosyl hydrolase; PARP; ARTD; MACROD1; MACROD2; TARG1

#### **1. ADP-Ribosylation Reactions**

Post-translational modifications of proteins can have a myriad of consequences: changing interactomes, stability, localisation, or activity are just a few examples. To date, more than 200 types of modification are known [1]. The addition of ADP-ribose moieties to a protein was first reported in the 1960s when chains of (poly)ADP-ribose (PAR) were identified on target proteins, termed poly(ADP-ribosyl)ation (PARylation) [2]. Since then, PARylation has been intensively studied, leading to the identification of crucial roles in the repair of DNA damage and to the development of specific inhibitors that are utilised in the clinic to treat diverse cancers [3]. In contrast, much less is known about its smaller sibling, mono(ADP-ribosyl)ation (MARylation), where only single ADP-ribose units (MAR) are conjugated into proteins. In 2008, it was first noted that PARP10/ARTD10 has MARylation rather than PARylation activity [4] and, lately, it has become clear that only the minority of PARP enzymes are capable of PARylation [5]. Moreover, not only protein, but also DNA has been identified as a target for ADP-ribosylation catalysed by PARP3 [6,7]. Following PARP1 and PARP2, PARP3 has also been described as a potential therapeutic target in certain cancers [8–11]. The most recent addition to the spectrum of ADP-ribosylated molecules is RNA, which can be modified in vitro by multiple PARPs [12].

One of the best-studied but still barely understood MARylating enzymes is PARP10. PARP10 was initially identified as an interactor of MYC [13] and, later, of ubiquitinated proliferating cell nuclear antigen (PCNA) [14,15]. Its overexpression leads to cell death in HeLa cells through an unknown mechanism [16]. In contrast, in non-transformed RPE-1 cells, overexpression leads to stimulation of cell growth [17]. It has been given a role in replication and DNA damage [15,17,18] as well as mitochondrial metabolism [19], but it is not clear what the relevant substrates of this enzyme are. Potential substrates were identified using protein microarrays [20], which was performed for the PARylating PARP2 [21]. However, partially due to a lack of antibodies at that time, these modifications were not verified to occur in cells. PARP10 was also reported to be a regulator of NF-kB signaling [22]. PARP14 functions as a co-activator of STAT6 and has been described as an anti-cancer and anti-inflammatory target, reviewed in more detail elsewhere [23,24]. PARP16 is located at the endoplasmic reticulum (ER) and plays a role in the unfolded protein response [25,26]. The most recent function identified for MARylation is its role in the anti-viral defence. PARP7, PARP10, and PARP12 gene expression is induced by interferon-α and ADP-ribosylation mediated by these PARPs may function to counteract viral proliferation [27,28]. In addition to enzymes of the PARP family, several sirtuins (SIRTs) appear to be able to ADP-ribosylate substrates in addition to their better-studied NAD+-dependent deacylase activity [29,30]. The first research showing that SIRTs possess MARylation activity described it as a very weak activity [31]. Other papers, focusing on single enzymes, showed that the mitochondrial SIRT4 MARylates glutamate dehydrogenase to downregulate insulin secretion [32]. The nuclear SIRT6 modifies PARP1 [33] and BAF10 [34] to regulate their activities and, lastly, the nucleolar SIRT7 was reported to MARylate histones p53 and ELK4 [35]. To date, very little information is available about the MARylating activity of the SIRTs. The functional consequences of MARylation in normal cell physiology, thus, appear to be very diverse, reviewed extensively elsewhere [36]. Herein, we focus on the enzymes reversing the intracellular MARylation reaction: the macrodomain-containing hydrolases.

#### **2. Macrodomains**

Intimately linked to ADP-ribosylation is a protein structural module known as the macrodomain, named for the unusually large histone macroH2A from which it was identified [37]. The macrodomain is a protein fold formed around a central β-sheet, flanked by α-helices that forms a pocket where ADP-ribose can bind [38,39]. These pockets are slightly different between the macrodomain-containing proteins with the consequence that the macrodomain-containing protein family can be subdivided. Some macrodomains bind to PAR and others bind to MAR [40–43].

Macrodomains not only serve as binding modules to mediate protein interactions depending on ADP-ribosylation, but a number of them have catalytic activity. Poly(ADP-ribosyl)glycohydrolase (PARG) is able to degrade PAR chains by cleaving the glycosidic bond between adjacent ADP-ribose moieties. However, it cannot remove the final ADP-ribose attached to the protein [44]. Three other proteins, MACROD1, MACROD2, and TARG1, are not active toward PAR but are capable of removing the ADP-ribose moiety connected to the protein partner, presumably by cleaving an ester bond between ADP-ribose and an acidic receptor amino acid [45–47]. They can, therefore, both reverse any MARylation as generated by PARP enzymes, and also remove the last ADP-ribose left behind by PARG. Of the 12 human macrodomain-containing proteins known today, 11 proteins were identified by sequence homology [39,43]. The macrodomain of PARG was only recognised after the structure had been solved [44], which indicates the possibility that more macrodomain-containing proteins remain to be identified. A recent bioinformatics approach has suggested that the largely uncharacterised protein C12orf4 harbours a macrodomain, or at least a macro-like domain, but this remains to be confirmed experimentally [48].

The macrodomain is an evolutionary well-conserved module [39]. For instance, recent work has shown that certain alpha-viruses encode a protein with a macrodomain-like fold [49], which displays catalytic activity towards MARylated proteins [50–52]. The existence of such proteins enhances the evidence connecting existing links between MARylation and immunity [53]. In this review, we will focus specifically on potential oncogenic functions of the mono(ADP-ribosyl)hydrolases MACROD1, MACROD2, and TARG1. It should also be noted, however, that other macrodomain-containing proteins such as PARG and CHD1L (ALC1) have been previously linked closely to cancer [54,55].

#### **3. Macrodomain-Containing Hydrolases: MACROD1, MACROD2, and TARG1**

MACROD1, MACROD2, and TARG1 are relatively small proteins (Table 1) without any notable other domains beside the macrodomain. The first catalytic activity of MACROD1, MACROD2, and TARG1 that was recognised was their deacetylation of *O-*acetyl-ADP-ribose (*O*AADPR) to generate acetyl and ADP-ribose [56,57]. *O*AADPR arises as a by-product of SIRT-mediated deacetylation in which the acetyl group is transferred onto ADP-ribose while releasing nicotinamide, and may have important biological functions [58]. Two years later, three labs reported the activity of MACROD1, MACROD2, and TARG1 in removing ADP-ribose from protein substrates, which was proposed to be limited to removal of ADP-ribose from acidic amino acids [43,45–47]. Serine and arginine MARylation were suggested to be removed by ARH3 and ARH1, respectively [59–62], which are structurally different proteins that do not contain a macrodomain. In 2017, MACROD1, MACROD2, and TARG1 were reported as being capable of removing ADP-ribose from both modified single-stranded DNA [63] and, in 2018, also being capable of removing ADP-ribose from MARylated single-stranded RNA [12]. Lastly, in 2019, the hydrolysis of α-NAD was added to the list of their in vitro activities [64].

To date, it is not known which of these activities is relevant for their functionality in cells, either in normal physiology nor in pathological conditions. This is predominantly due to the fact that it is not well known how prevalent their in vitro substrates are in cells or what their cellular functions may be. This is partially due to the fact that antibodies or modules recognising MARylation have been developed only recently [65–67]. Before these technical developments, it was not possible to confirm with antibodies that MARylation of proteins takes place in cells, and, hence, reversal mediated by these macrodomain-containing proteins could also not be assessed. Currently, it is still not straightforward to estimate endogenous MARylation levels in cells, as no inhibitors were available for MACROD1, MACROD2, or TARG1. During experimental procedures such as cell fixation or lysis, MARylation could possibly be removed by these enzymes and, therefore, go undetected. One study was performed to identify MACROD1 inhibitors using an AlphaScreen assay, where inhibitors in the micromolar range were identified. These can serve as lead compounds for further optimisation, but are likely not specific enough yet to be used for studies in cells due to their other activities at this concentration [68]. Another study identified an allosteric inhibitor of macrodomain 2 of PARP14, which blocks its binding to ADP-ribose. However, it was not tested whether the identified compound might inhibit one of the macrodomains with catalytic activity [69].



 upon (WB) is smaller [70].

The roles of MACROD1, MACROD2, and TARG1 in normal cell biological processes are not well understood. They appear to be expressed and localised in different tissues and intracellular locations, which clearly gives them unique roles to play despite their biochemical similarity. This also explains why loss of TARG1 alone can lead to a neurodegenerative phenotype [45]. MACROD1 resides in mitochondria [40,70,71], but, depending on the tag used for overexpression, will reside exclusively in the nucleus instead [70]. MACROD2 displays a more diffuse cytoplasmic/nuclear distribution [70,75] and overexpressed TARG1 appears to be present in nucleoplasm, nucleoli, and cytoplasmic stress granules [14,70]. TARG1 and MACROD1 are expressed throughout the different tissues with an enrichment of MACROD1 in skeletal muscle at both the protein and RNA level [70]. MACROD2 has, thus far, only been detected in human neuroblastoma cells and in mouse cortical neurons [70,76]. Several reports have correlated polymorphisms or deletions within the *MACROD2* gene with autism-spectrum disorders [77–80]. It is not clear, however, whether the *MACROD2* gene product itself or surrounding genes are responsible for this association even though elevated protein expression in neurons supports a potential brain-specific function of MACROD2 [76]. However, a long non-coding RNA has been identified within an intron of the *MACROD2* gene, which is more highly expressed in most tissues investigated than the *MACROD2* mRNA, and, thereby, potentially confuses correlations of mutations in the *MACROD2* gene and phenotypes [81]. In another genome-wide association study, *MACROD2* was identified as a factor influencing vascular-adhesion protein-1 (VAP-1) levels, which the authors confirmed by knockdown of MACROD2. This leads to lower VAP-1 expression in adipocytes, presumably through the transcriptional regulation of VAP-1 [82]. MACROD2 was also reported to leave the nucleus upon DNA damage, dependent on phosphorylation by ATM [75], whereas TARG1 localises from nucleoli to nuclear sites of damage [14], which both have an unclear functional relevance. More data may be found on MACROD1, which is also known as leukaemia-related protein 16 or LRP16. MACROD1 has been attributed to a number of functions in the nucleus, such as co-activation of the androgen receptor [83], counteracting PARP7-mediated MARylation in the nucleus [84,85], and activation of NF-kB signalling [86–88]. MACROD1 was also reported as an enhancer of oestrogen receptor signalling [89], and is upregulated after stimulation of cells with oestrogen [72,74,90]. No regulatory factors have been identified yet for MACROD2 and TARG1. A BioID interaction screen, which identifies proteins in close proximity to the protein of interest [91], identified many proteins involved in nuclear/cytoplasmic and mitochondrial nucleic acid metabolism as interactors of TARG1 and MACROD1, respectively [70]. Whether these proteins are MARylated and serve as a substrate of TARG1/MACROD1 remains to be determined. It is possible that, despite spatio-temporal restrictions, MACROD1 and TARG1 in their respective compartments are involved in similar signalling networks, converging on the regulation of cellular nucleic acids. The physiological functions of the three enzymes remains elusive. Most puzzling perhaps is that, at the moment, despite the mitochondrial localisation of MACROD1, the majority of reports describe nuclear functions. Future work will need to address this apparent discrepancy.

#### **4. Mono(ADP-ribosyl)ation in Cancer**

The post-translational modification poly(ADP-ribosyl)ation has been intimately linked to cancer before [92,93], as have other post-translational modifications such as phosphorylation [94]. In BRCA1/BRCA2-deficient patients, PARP1 inhibitors have been shown effective specifically against the tumour cells applied in the clinic. However, this is one of the rare examples of a synthetic lethal interaction [95]. Better understanding of the processes regulated by MARylation will provide opportunities for further drug development, as is exemplified by current research into the potential of PARP14 as a drug target [24]. Little is known about the potential role of the mono(ADP-ribosyl)hydrolases in cancer. *MACROD1*, *MACROD2*, and *OARD1* (TARG1) exhibit mutations only in 0.9%, 2.6%, and 1% of cancer patient samples, respectively, from over 1000 samples in the cBioPortal curated dataset [96]. The fact that *MACROD2* mutation rates are twice the rates seen for *MACROD1* and *OARD1* likely reflects that the *MACROD2* gene is larger (Figure 1), and is

located at a known fragile region [97]. A specific recurring deletion of exon 6 has been observed in esophageal squamous cell carcinoma and gastric cancer [98]. For *MACROD1*, a *RUNX-MACROD1* fusion was identified in leukaemia [99] and, for *MACROD2*, a *PDGFRA-MACROD2* fusion in a pleomorphic sarcoma [100]. Despite the low number of identified mutations in patient tumour samples, several reports have correlated *MACROD1* or *MACROD2* expression levels with the clinical outcome, as described in the next paragraphs.

**Figure 1.** Overview of the *MACROD1*, *MACROD2*, and *OARD1* gene structure.

The gene structure of *MACROD1*, *MACROD2*, and *OARD1* is shown schematically. More lncRNAs are present. However, only RPS10P2-201 is displayed since it has been shown to be relevant RPS10P2-201 [81]. CDS = coding sequence. UTR = untranslated region.

#### **5. MACROD1 in Cancer**

A number of studies have addressed a potential oncogenic function of MACROD1 and started deciphering the molecular mechanism underlying observed effects. MACROD1 expression was shown to be upregulated by oestrogen [73,74], which leads to several studies of the role of MACROD1 in tumours with a differential oestrogen status. MACROD1 overexpression in the oestrogen and progesterone receptor positive Ishikawa cells, derived from an endometrial cancer, had no effect on cell proliferation. It did, however, enhance the invasiveness of these cells as measured by transwell assays [101]. Mechanistically, the authors propose a mechanism wherein MACROD1, dependent on oestrogen, blocks recruitment of ERα to the E-cadherin promoter, which lowers E-cadherin expression and, through this, enhances invasiveness. shRNA-mediated knockdown of MACROD1 achieved the opposite effect by enhancing E-cadherin expression [101]. This implies that MACROD1 can be an important factor in metastasis. These findings appear to contradict an earlier report where overexpression of MACROD1 in MCF7 cells, which are oestrogen-responsive breast cancer cells, showed an effect on cell growth. It led to enhanced proliferation [74]. A later report studying MACROD1 in 293T cells showed that knockdown of MACROD1 sensitized cells to TNFα-induced

apoptosis [86]. From these studies, it is, thus, not clear what effect loss or gain of MACROD1 has on cell physiology. Recent work has demonstrated that fusing an N-terminal tag, such as GFP, to MACROD1 leads to a nuclear localisation, whereas C-terminally labelled and an endogenous protein appear to be exclusively localised in mitochondria [70]. Mass spectrometry datasets have also detected MACROD1 in the mitochondria [102]. This does not exclude the possibility that, under specific circumstances, MACROD1 may re-localise to the nucleus. In pathogenic conditions, such as the presence of a *RUNX-MACROD1* fusion protein that was identified in leukaemia [99], the protein likely also localises to the nucleus instead of the mitochondria, as the RUNX fusion will mask the mitochondrial targeting sequence, comparable to the localisation after labelling with an N-terminal GFP tag. Unfortunately, some of the studies investigating MACROD1 have either used N-terminally tagged fusion constructs [84,85,88], or have not stated clearly how the fusion proteins were generated [103]. Furthermore, the majority of applied antibodies show multiple bands in the western blot and, hence, are not suitable for immunohistochemistry (IHC) or immunofluorescence (IF) (Table 2). It will be worthwhile to repeat some of the studies of MACROD1 s molecular function in carcinogenesis to clarify whether unlabeled MACROD1 overexpression leads to enhanced cell growth, or whether this effect depends on the tumour background and also to study knockdown/knockout systems more thoroughly.



Despite these technical challenges, multiple studies have correlated tumour growth with MACROD1 expression in either xenograft models or in patient cohorts. Overexpression of MACROD1 in the hepatocellular carcinoma lines HepG2 and MHCC-97L leads to a decrease in cell growth and metastatic potential, as measured by transwell assays [103]. Furthermore, when delivered into nude mice, cells overexpressing MACROD1 lead to a decreased tumour volume and have a lower metastatic potential compared to cells without overexpressed MACROD1 [103]. With the information available, it cannot be distinguished whether this is a genuine effect or an artefact due to forced nuclear localisation of the overexpression construct [103]. In a similar set of experiments in the pancreatic carcinoma cell lines, Panc1, CFPAC1, Bxpc3, SW1990, AsPC1, and HPDE6-C7, opposite results were achieved [105]. Knockdown of MACROD1 leads to enhanced apoptosis and decreased cell growth. However, only one shRNA construct was used, so any effects seen could be potentially off-target. Cells overexpressing the same MACROD1 construct as in the study described before [103] grew faster and were more resistant to apoptosis. Xenograft experiments show the same trend. Cells lacking MACROD1 had a lower tumour-forming potential and higher survival rate. Cells overexpressing MACROD1 had a higher tumour volume and a lower survival rate [105]. The authors do not comment on the opposing effects in these two tumour types, but agree that larger-scale studies are required to verify these findings. A third study shows IHC of lung tumour samples with an antibody of an unknown source and specificity [104]. High MACROD1 expression, as measured by IHC, was reported to correlate with a negative outcome in colorectal carcinoma and in gastric carcinoma. However, the antibody used was generated by the authors and not validated by a western blot [106,107]. CRISPR-mediated *MACROD1* knockout rhabdomyosarcoma cells [70] and *MACROD1* knockout mice appear viable [71], which makes it unlikely that loss of MACROD1 has a drastic growth inhibitory or developmental effect.

In conclusion, most of the data available on an oncogenic function of MACROD1 rely on poorly characterised antibodies, unclear overexpression constructs or a single shRNA construct. The majority of these studies would need to be reproduced with a more thorough characterisation and description of the materials used to be able to draw deeper conclusions. Several studies agree that MACROD1 expression can be induced by oestrogen [73,74,89,90,101], but it remains unclear what the effect of this overexpression is on cells. The *RUNX-MACROD1* fusion identified in leukaemia may provide an important hint at a potential pathologic function. It is possible that the physiological mitochondrial localisation of MACROD1 can turn into a pathogenic nuclear one, where it aberrantly acts as a transcriptional activator. It will be interesting to see whether more instances can be identified where such fusions are present. Alternatively, other masking events may occur in cells, such as binding by interaction partners or PTMs, which, thereby, redirects the protein to the nucleus for a physiological function.

#### **6. MACROD2 in Cancer**

A number of analyses suggest MACROD2 may play a potential role in cancer. MACROD2 copy number is increased in three different tamoxifen-resistant MCF7 breast cancer cell lines, prompting the authors to analyse MACROD2 expression in patient samples. In oestrogen receptor-positive tissues of breast cancer patients with a recorded tamoxifen-resistance, however, three patients displayed a decreased copy number, whereas the other two patients showed an increase. Using IHC with a custom antibody, varying levels of MACROD2 were detected in primary and secondary tumour tissues of these patients, collectively showing that MACROD2 may be overexpressed in cancer tissues. MCF7 and T47D cells with exogenously overexpressed MACROD2 grow faster than control cells in media containing tamoxifen, which implies that MACROD2 confers tamoxifen-resistance to the cells and appear to be stimulated by tamoxifen, as demonstrated by faster growth in medium with tamoxifen than without [110]. Conversely, tamoxifen-resistant MCF7 clones with shRNA-mediated knockdown of MACROD2 become more sensitive even though it does not completely reverse resistance [110]. This may be mediated by the activation of oestrogen-regulated genes in response to tamoxifen in cells overexpressing MACROD2. Lastly, cells stably expressing shRNA to knockdown MACROD2

grow markedly slower in nude mice [110]. The authors argue that, as a cancer specific fragile site, the MACROD2 gene can be lost, but this fragility also allows for amplification in the specific case of ER-positive breast cancers treated with tamoxifen to incur resistance. A more recent study showed that, in approximately one-third of colorectal carcinomas investigated, heterozygous or homozygous losses were mapped within the *MACROD2* gene in which the majority are intragenic microdeletions mapping to a region in exons 4 to 5 [111]. To test whether MACROD2 deficiency can promote tumourigenesis, *MACROD2* knockout was introduced into an adenomatous polyposis coli protein (Apc) mouse model. Mutations in the APC tumour suppressor are a major driver of sporadic colorectal cancers, where it could be shown that even haploinsufficiency of *MACROD2* leads to more and larger adenoma formation [111]. Furthermore, human cells transplanted into nude mice displayed increased tumour growth when lacking *MACROD2*, but a reduced tumour growth when MACROD2 was overexpressed [111]. The underlying mechanism was suggested to be impaired PARP1 activity in the *MACROD2*-/- cells, which leads to increased sensitivity of DNA damage and, ultimately, causes enhanced chromosomal instability [111,112].

These findings appear paradoxical with the previous report, where loss of MACROD2 impairs cell growth from a loss of resistance to tamoxifen. *MacroD2* knockout mice do not show altered survival rates after sub-lethal irradiation compared to wildtype mice, which indicates that loss of MACROD2 alone is not sufficient to drive tumourigenesis triggered by DNA damage [109]. It is not clear whether loss or overexpression of MACROD2 contributes to tumourigenesis, or whether both can drive tumour growth dependent on conditions, such as the functionality of the DNA damage repair systems or the oestrogen receptor status. In an investigation of stage-III colon cancer, MACROD2 expression determined by immunohistochemistry was found to correlate with poor survival [113]. A human protein atlas antibody (HPA049076) was used for this work, which has been retracted by the company in the meantime and did not appear to be validated in any way, such as western blotting or siRNA-mediated knockdown. It is, thus, not clear whether this antibody recognises MACROD2 at all, or whether it recognises additional proteins, which may be upregulated in the samples analysed. Altogether, it appears that loss of MACROD2 as such is not sufficient to drive tumourigenesis, but may have an additive effect in models prone to tumour formation such as loss of APC in colorectal cells. More studies are urgently needed to clarify the exact role of MACROD2 in both the onset of cancer as well as in the response of existing tumours to therapies.

#### **7. TARG1 in Cancer**

The only phenotype associated with TARG1 expression is neurodegeneration, occurring due to a mutation that leads to a truncated protein with a disrupted macrodomain, and, thereby, loss of catalytic activity [45]. It is not clear whether this neurodegeneration is a result of a potentially toxic truncated or unfolded protein or from loss of hydrolase activity. Knockdown of TARG1 leads to a decrease in 293T cell proliferation and a slight increase in senescence in U2OS cells, which are derived from an osteosarcoma [45]. CRISPR-mediated knockdown of TARG1 does not influence HeLa or U2OS proliferation [14,70], which leaves it unclear in which setting TARG1 is required for cell growth. Overexpression does not lead to changes in cell proliferation [14,70]. We could not identify any data linking TARG1 to cancer, nor see elevated expression or mutation in databases such as COSMIC. Based on expression levels in databases and these experimental results, at this stage, it does appear unlikely that TARG1 is involved in cancer even though further experimentation is required.

#### **8. Conclusions**

Despite the presence of several publications reporting a correlation between MACROD1 or MACROD2 levels and cancer development or progression, it is not clear at the moment that they have a causative role. MACROD2 is potentially relevant in ER-positive, tamoxifen resistant breast cancers where it may confer resistance to treatment. However, larger cohorts of patient samples need to be analysed to further substantiate these initial findings. Loss of *MACROD2* in an APC null background potentially stimulates tumour formation. However, loss of *MacroD2* alone does not have any effect in a knockout mouse model. Detailed studies are required to determine how loss of MACROD2 might cooperate with loss of APC to drive cancer. MACROD1 levels are upregulated by oestrogen with unclear consequences for cells. Yet, it may down-regulate E-cadherin and, thereby, promote metastasis. If confirmed, this may be an attractive therapeutic target intending to keep the cancer dormant and prevent metastasis.

Conflicting data show that overexpression and knockdown of MACROD1 have no effect, no stimulus, and no inhibition of cell growth, which implies that, perhaps, inhibiting MACROD1 may not have detrimental effects for the whole organism but rather may be dependent on the tumour background and, thus, may represent a valid drug target. The effects observed may, however, be dependent on the constructs used as well as on the specific cell-types and have to be studied in more detail. Future work will have to dissect in which context loss or gain of MACROD1 may be driving aspects of cancer growth. The recent development and characterisation of more specific antibodies, the ongoing improvements of mass spectrometric measurements of MARylation, and also the attempts at making specific inhibitors for both transferases and hydrolases should allow a more detailed analysis of their (patho-)physiological function in the near future. The partially paradoxical findings described will undoubtedly be clarified with better validated tools to determine the extent to which MACROD1, MACROD2, and TARG1 are relevant for tumourigenesis in order to establish their potential as drug candidates.

**Author Contributions:** Writing, review, and editing: K.L.H.F., C.D.O.C., and R.Ž. All authors have read and agreed to the published version of the manuscript.

**Funding:** The START Program of the Medical Faculty, RWTH Aachen University to K.L.H.F. (10/18) and to R.Z (13/20) funded this research.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **The Development of Rucaparib**/**Rubraca**®**: A Story of the Synergy Between Science and Serendipity**

#### **Nicola J Curtin**

Professor of Experimental Cancer Therapeutics, Translational and Clinical Research Institute, Newcastle University Centre for Cancer, Faculty of Medical Sciences, Medical School, Paul O'Gorman Building, Newcastle University, Newcastle upon Tyne NE2 4HH, UK; nicola.curtin@newcastle.ac.uk or nicola.curtin@ncl.ac.uk; Tel.: +44-0-191-208-4415

Received: 15 January 2020; Accepted: 26 February 2020; Published: 29 February 2020

**Abstract:** The poly(ADP-ribose) polymerase (PARP) inhibitor, Rubraca®, was given its first accelerated approval for BRCA-mutated ovarian cancer by the FDA at the end of 2016, and further approval by the FDA, EMA and NICE followed. Scientists at Newcastle University initiated the early stages, and several collaborations with scientists in academia and the pharmaceutical industry enabled its final development to the approval stage. Although originally considered as a chemo- or radiosensitiser, its current application is as a single agent exploiting tumour-specific defects in DNA repair. As well as involving intellectual and physical effort, there have been a series of fortuitous occurrences and coincidences of timing that ensured its success. This review describes the history of the relationship between science and serendipity that brought us to the current position.

**Keywords:** PARP; drug development; synthetic lethality; clinical trials

#### **1. Introduction: Rationale for the Development of PARP Inhibitors**

Poly(ADP-ribose) polymerase (PARP) inhibitors (PARPi) have been the most significant addition to the armoury for the treatment of ovarian cancer since the introduction of platinum therapy in the 1970s and represent a paradigm shift in the way cancers may be treated. This is the story of the development of the first PARPi to enter anticancer clinical trials, Rubraca®, formerly known as rucaparib or AG014699, the nomenclature used here largely reflects the name of the drug at the time of use. The development of Rubraca was the result of a substantial intellectual and physical effort by many individuals working as a team but it is hard to avoid the conclusion that luck has played a significant role, and this story describes how serendipity has contributed to the final success. This review describes the development of rucaparib in the context of other advances in the understanding of DNA repair and highlights the lucky breaks along the way.

The development of PARPi that ultimately led to rucaparib began in Newcastle University in 1990 but the story begins before that. The first observation that suggested the existence of PARP was the profound NAD<sup>+</sup> depletion following exposure of cells to ethyleneimine in 1956 [1]. Coming so soon after the discovery of DNA, this effect was not attributed to the effect of DNA methylation, rather an impact on cellular metabolism. The product ADP-ribose polymers, or poly(ADP-ribose) (PAR), was identified in 1963 and at first thought to be some new kind of nucleic acid [2], with elegant experiments a few years later showing the time course of the disappearance of NAD<sup>+</sup> and the corresponding appearance of polymers and nicotinamide [3]. The first suggestion of the involvement of PARP in DNA repair was made by Edward Miller in 1975 [4]. To test this hypothesis, Purnell and Whish [5] developed the first inhibitors, which were based on the catalytic mechanism of PARP and the observation that the by-product, nicotinamide, exerts some feedback inhibition. These early benzamide inhibitors included 3-amino benzamide (3AB), which is still used as a PARPi today. The pivotal study, conducted by Barbara Durkacz in Sydney Shall's lab, was published in 1980, showing that 3AB prevented the repair of DNA and increased cytotoxicity following exposure to the DNA methylating agent, DMS [6].

#### **2. Early Studies in Newcastle**

Barbara Durkacz moved to Newcastle University in the mid-1980s to establish PARP-related studies, where she was joined by Mike Purnell for the synthesis of inhibitors. However, it was the appointment of Hilary Calvert as the new director of the Cancer Research Unit in Newcastle, and his collaboration with Bernard Golding, Professor of Organic Chemistry at Newcastle University, that really got inhibitor development going. They established a drug development programme in October 1990, with funding from the North of England Cancer Research Campaign. Other members of the team were Roger Griffin (chemist), Herbie Newell (pharmacologist) and me (biologist/biochemist). PARP was one of three initial targets for drug discovery, tapping into Barbara Durkacz's expertise. A succession of post-graduate chemistry students embarked on inhibitor synthesis guided by Roger Griffin and Bernard Golding. However, to establish any kind of structure–activity relationship (SAR) a robust, reproducible and quantitative activity assay was needed and herein lay the problem. The results using a published assay [7,8] were variable in the extreme, with replicates wildly different from each other, and no better than random. Months were spent trying to generate inhibition data until finally, in exasperation, I decided to try to get to the bottom of the problem. The assay involved incubating permeabilised cells with 32P-NAD<sup>+</sup> and broken DNA to activate the cellular PARP then precipitating the polymer with TCA, collecting it onto filters, washing surplus NAD<sup>+</sup> away and counting the acid-precipitated radioactivity. Investigating each step of the assay to identify the source of the irreproducibility by taking replicates at different stages in the assay revealed that it was nothing to do with the cell permeabilisation or collection of the product but was the reaction itself. A eureka moment led me to discover a very simple solution: switching to plastic tubes. Because of the charged nature of NAD<sup>+</sup> and the polymer we had been siliconising glass tubes for the reaction to negate the charge on the glass and prevent binding. It turned out that the siliconising agent itself was causing highly variable interference with the activity of the enzyme. By switching to plastic this interference was removed but another problem was identified, i.e., the poor thermal conductivity of the plastic, which was overcome by a pre-warming step. Thereafter, this assay provided the robust, reproducible and quantitative data needed for SAR generation. This is the first example of the synergy of science and serendipity (#1).

The next example of serendipity (#2) came on the chemistry side. Inhibitors were designed to incorporate the nicotinamide pharmacophore, which was thought to make major interactions within the catalytic site of PARP. Studies with other NAD-dependent enzymes suggested that binding was favoured when the carboxamide was anti to the aromatic ring. In 3AB, it can freely rotate to the cis-orientation, meaning that binding is relatively weak (Ki = 10 μM, IC50 = 30 μM). Parallel SAR studies led by Judy Sebolt-Leopold at Warner Lambert and an "analogue by catalogue" approach in Kunihiro Ueda's lab identified that compounds where the carboxamide group was held anti to the aromatic ring, by constraining it through incorporation into a ring structure, had increased potency [9,10]. Our approach was to hold the carboxamide in the anti- conformation through hydrogen bonding with an oxazole or imidazole group. However, during the synthesis of the target compound, 2-methylbenzoxazole-4-carboxamide, a molecular rearrangement occurred generating 8-Hydroxy-2-methylquinazolin-4[3H]-one, NU1025, where the carboxamide was constrained in the desired orientation through incorporation into a ring [11] (Figure 1). NU1025 (Ki = 48 nM IC50 = 400 nM) had comparable activity to compounds identified by the Sebolt-Leopold and Ueda groups and was substantially more potent than the compound we had been aiming to make, which had an IC50 of 10 μM. This lucky accident therefore gave us our first hit compound and allowed us to explore the cellular effects of PARP inhibition.

**Figure 1.** Accidental synthesis of first "hit" compound, NU1025. During the attempted synthesis of 2-methylbenzoxazole-4-carboxamide, which could exist in the desired anti- conformation or the syn- conformation (left), a molecular rearrangement occurred, resulting in NU1025, a much more potent compound.

My first PhD student, Karen Bowman, was tasked not only with determining the potency of the compounds against PARP activity (aided and abetted by Louise Pemberton who also synthesised some of the early inhibitors) but also with evaluating the ability of NU1025 and the benzimidazole, NU1064 (2-methylbenzimidazole-4-carboxamide, IC50 = 1.1 μM) to increase the cytotoxicity of a variety of anticancer drugs. She confirmed that PARP inhibition increases the cytotoxicity of DNA alkylating agents using temozolomide and its active derivative, MTIC. She identified that it was the DNA repair phase that was critical, rather than the DNA damage induction phase as NU1025 was as active when it was added after a 20 min pulse with MTIC as when the 2 compounds were given simultaneously. Karen also confirmed the radiosensitisation data that had been generated with the early benzamides. She showed that PARP inhibition was more effective at preventing recovery from potentially lethal damage than radiosensitising exponentially growing cells. Potentially lethal recovery is when cells are given a high dose of irradiation and their survival is estimated immediately or after a recovery period where they are allowed to repair the DNA damage. This is significant because the non-replicating fraction of a tumour is generally radioresistant. We were unable to see any sensitisation effect of NU1025 or NU1064 on antimetabolite-induced cytotoxicity [12]. Using NU1025, we were the first to demonstrate that PARP inhibition increased the DNA damage and cytotoxicity caused by topoisomerase I poisons, but not topoisomerase II poisons [13]. We continued synthesising new compounds and testing them for PARP activity with the next major step in potency being the synthesis of the benzimidazole NU1085 (2-(4-hydroxyphenyl)-benzimidazole-4-carboxamide Ki = 6 nM, IC50 = 80 nM) by Sheila Srinivasan [14]. Since our compounds were numbered sequentially, this means that our 25th compound NU1025 was 200x more potent than 3AB and our 85th compound was 1000x more potent.

The generation and characterization of PARP-knockout mice that both were viable and fertile by three groups independently in the late 1990s [15–17], initially caused us to think that PARP was non-essential, and so inhibiting it would not compromise the viability of cancer cells. However, I recalled the data of Satoh and Lindahl [18] demonstrating that nuclear extracts depleted of PARP were able to repair nicked DNA but nuclear extracts with PARP but incubated either in the absence of NAD<sup>+</sup> or presence of 3AB could not. These authors proposed that binding of PARP to the DNA break caused an obstruction unless it could auto poly (ADP-ribosylate) and dissociate to allow access of repair proteins to fill in and re-seal the gap. This was the first demonstration of PARP "trapping", although this phrase was not coined until at least 20 years later [19]. These observations meant inhibited PARP was likely to be more effective in inhibiting DNA repair than no PARP enzyme. The ability of cells derived from these mice to generate PAR led to the discovery of PARP2 [20] and subsequently other

PARPs by structural analogy [21], with the original most abundant PARP subsequently being called PARP1. It is important to note here that PARP1 is the most abundant PARP and most important in terms of DNA repair, with PARP2 having a similar role. Most inhibitors inhibit both PARP1 and 2, and the double knockout of PARP1 and PARP2 is lethal [22].

#### **3. Advanced Compound Development and Pre-clinical Studies Resulting from the Newcastle–Agouron Collaboration**

Our next piece of good luck (serendipity #3) was to meet Zdenek Hostomsky at the 9th NCI/EORTC annual meeting in Amsterdam in 1996. He was proposing to initiate a PARP programme at Agouron Pharmaceuticals (in La Jolla, California, USA), a company specialising in crystal structure-based drug design. He had obtained the plasmid containing the construct for the PARP1 catalytic domain from Gilbert deMurcia and the co-ordinates of its crystal structure from Georg Schultz [23]. All he needed was highly potent inhibitors. We struck up an agreement and gave him NU1085 to co-crystalise with the PARP1 catalytic domain. Initial promising results meant that their 6 month pilot funding to Newcastle was extended for a further 2 years to support two post-doctoral chemists and a technician, Alex White, Sarah Mellor and Richard Davies, and two post-doctoral biologists and a technician, Carol Delaney, Chris Calabrese and Lan Zhen Wang. At Agouron, analysis of the interactions made between the inhibitor and the PARP catalytic domain by Bob Almassy led the lead chemist on the project, Steve Webber, to propose tricyclic compounds with the carboxamide constrained within a 7 membered ring. Synthetic medicinal chemistry at Agouron was undertaken by Don Skalitzky and Stacie Canan, working with Steve Webber. Karen Maegley ran the biochemical evaluation, with cell-based target inhibition and cytotoxicity assays being done by Jianke Li. Ted Boritzki and Bob Kumpf led the conventional and computational pharmacology, respectively at Agouron in collaboration with Chris Calabrese and Huw Thomas in Newcastle. At the end of the 2 year period, several potent tricyclic indole and benzimidazole inhibitors had been made at Agouron and they extended the funding to Newcastle for a further 3 years to cover the in vitro and in vivo pre-clinical evaluation of these inhibitors, as well as the development of pharmacodynamic biomarkers until the clinical trial started in 2003.

Extensive investigation of the cellular activity of these tricyclic compounds in terms of their ability to enhance the cytotoxicity of the DNA methylating agent, temozolomide, and the topoisomerase I poison, topotecan was undertaken in Newcastle, principally in two colorectal cell lines, LoVo and SW620 by Lan Wang and Suzanne Kyle, with Jianke Li investigating the effects in A549 lung cancer cells. My group have always maintained the highest possible stringency regarding the cultivation of cells, confirming they are mycoplasma free by routine regular testing (2 monthly intervals) and handling each cell line separately (including reagents) to eliminate the possibility of cross-contamination. Although these studies preceded routine authentication, our stocks have been authenticated since and all human cell lines are either purchased new and placed in an authenticated bank at first or second passage or authenticated and then placed in the authenticated bank. In vivo efficacy and pharmacokinetic/pharmacodynamic (PK/PD) studies were also undertaken by Chris Calabrese working with Huw Thomas and Mike Batey [24]. The first potential clinical candidate was AG14361 (1-(4-((dimethylamino)methyl)phenyl)-8,9-dihydro-2,7,9a-triazabenzo[cd]azulen-6(7H)-one: Ki < 5 nM) [25–29]. AG14361 caused profound sensitisation of temozolomide and was able to overcome temozolomide resistance due to defects in DNA mismatch repair [30]. The enhancement of topotecan and ionising radiation cytotoxicity in vitro was less pronounced but again the prevention of recovery from potentially lethal irradiation was substantial. In vivo radiosensitisation studies were conducted in Manchester by Kaye Williams and Ian Stratford [29]. Investigations in NIH 3T3 cells from PARP1-knockout mice enabled my PhD student, Lisa Smith, co-supervised by a topoisomerase expert, Caroline Austin, to confirm that topotecan was more cytotoxic in PARP-null cells and sensitisation by AG14361 was indeed due to PARP1 inhibition [31].

Investigation of in vivo activity against SW620 and Lovo xenografts revealed some interesting data. AG14361 caused an approximately 3-fold increase in temozolomide-induced anticancer activity against Lovo xenografts, but the combination caused sustained complete tumour regressions of SW620 xenografts. This was curious because in cell cultures, Lovo cells were the most sensitised to temozolomide by AG14361 and SW620 cells were not sensitised at all. SW620 cells were particularly sensitive to temozolomide alone (due to functional mismatch repair and lack of methylguanine methyltransferase). This led to another example of the synergy of science of serendipity: the discovery of the vasoactivity of this class of compounds. I reasoned that the discrepancy between the in vivo and in vitro effects meant that the tumour regression had to be something to do with the tumour micro-environment. Nicotinamide was known to be a vasodilator [32], and so I hypothesised that AG14361, which, like all PARPi contains a nicotinamide pharmacophore, might improve the delivery of temozolomide to more of the tumour cells. In collaboration with Kaye Williams and Ian Stratford in Manchester, we showed that indeed AG14361 did increase the areas perfused in the tumour, a further example of the synergy of science and serendipity (#4). [29]. This effect may have also contributed to the enhancement of the anticancer activity of the topoisomerase I poison, irinotecan, by AG14361, which was greater in vivo than in vitro. Similar vasoactive effects were reported by others with olaparib, which was accompanied by increased radiosensitivity of tumour xenografts due to better oxygenation [33]. Vasoactivity may contribute to the in vivo radiosensitisation by AG14361 but we have not investigated whether this is the case. It is possible therefore that it is a class effect and that PARPi may increase the anticancer activity of other drugs by increasing delivery to the tumour. It should be noted that these experiments were done with subcutaneous xenografts and the effect on in situ tumours may be somewhat different, especially as PARPi in the clinic have not been reported to cause hypotension except rarely in the elderly population [34], as might be expected from a vasodilator. Nevertheless, there may be effects that are specific to the tumour vasculature, but this has never been investigated.

The exciting data with AG14361 led to a proposal for a clinical trial with this compound being designed at a meeting of the team early 2000 and drafted for submission to CRUK's New Agent Committee. The teams in Newcastle (CR Calabrese, AH Calvert, NJ Curtin, BW Durkacz, BT Golding, RJ Griffin and DR Newell), Manchester (J Monaghan, I Stratford, and K Williams) and Agouron (R Almassy, T Boritzki, S Canan-Koch, L DiMolfetto, G Furman, Z Hostomsky, A Johnston, R Kumpf, J Li, K Maegley, D. Skalitzky, SE Webber and K Zhang) all contributed to the proposal. Meanwhile, there had been continued evaluation of a number of potential inhibitors. Many of the studies that had originally been conducted with AG14361 were also replicated with several other compounds, in vitro by Suzanne Kyle and Lan-Zhen Wang and in vivo by Chris Calabrese and Huw Thomas. Computational studies by Bob Kumpf suggested a fluoro substituent at the 8 position would improve activity with the ultimate compound being 8-fluoro-5-(4-((methylamino)methyl)phenyl)-2,3,4,6-tetrahydro-1*H*-azepino[5,4,3-*cd*]indol-1-one, or AG014447. AG014447 and/or the phosphate salt—called AG014699—had been included in the pre-clinical in vitro and in vivo studies and was selected for clinical investigation [35]. AG014699 (rucaparib) was indeed more potent than AG14361 in in vivo studies and, whereas 10 mg/kg AG14361 caused complete tumour regressions in combination with temozolomide, 1 mg/kg AG014699 was equally effective [29,35]. It also had a similar vasoactive effect [36]. Pharmacokinetic (PK) studies with AG14699, conducted by Huw Thomas and Mike Batey, showed that although it was cleared quite quickly from the bloodstream, it accumulated in tumour xenografts and, at the efficacious dose, it suppressed PARP activity in the tumour by >50% for 24 h. The clinical trial proposal was re-written with AG014699 data, with input from the team above. The involvement of the team who had developed and studied the drug pre-clinically ensured that the first clinical trial was as rationally designed, with a full understanding of the pre-clinical data as it is possible to have and supported by translational pharmacodynamic, as well as PK measurements. The proposal was submitted to CRUK's New Agent

committee in 2001, with a successful outcome. AG014699, now called rucaparib, was the first PARPi to be given to a patient with cancer in May 2003.

#### **4. The First Clinical Trial and Supporting Science**

Based on the pre-clinical data, the target dose clinically (PARP inhibitory dose: PID) was one that inhibited PARP activity by >50% in a surrogate tissue (lymphocytes) for at least 24 h. To support this pharmacodynamically guided clinical trial, we started adapting the 32P NAD<sup>+</sup> incorporation assay we had used to measure PARP activity and inhibition in permeabilised cells as part of the early inhibitor development. In these endeavours, we were blessed with two more items of good fortune. Firstly, experiments conducted early in 2002 by Suzanne Kyle revealed that PARP inhibition was not lost as soon as the drug was removed Figure 2A), meaning that pre-treated cells could be harvested and processed without loss of inhibition. This was an instance of real serendipity (# 5), as it is quite rare to see such persistence. Secondly, at the same time, Suzanne showed that PARP activity was maintained in cells that had been cryopreserved at -80 ◦C for up to 14 weeks. More importantly, rucaparib's inhibition of PARP activity was incredibly stable: cells that had been pre-treated with rucaparib could be harvested and cryopreserved for up to 14 weeks without loss of inhibition, an even more remarkable example of serendipity (#6); Figure 2B.

**Figure 2.** Inhibition of cellular poly (ADP-ribose) polymerase (PARP) activity. (**A**) PARP inhibition by increasing concentrations of AG14699 (rucaparib) following pre-treatment of L1210 cells (30 min) with drug prior to harvesting (red symbols and line) compared with drug added to the permeabilised cells in the reaction mixture (blue). Data are the mean and SD of three replicates from a representative experiment. (**B**) Stability of PARP inhibition by rucaparib with storage. Cells were exposed to rucaparib for 30 min prior to harvesting and cryopreservation. PARP activity was measured by 32P NAD<sup>+</sup> incorporation. Previously unpublished data.

We then started harvesting lymphocytes from healthy volunteers (i.e., our colleagues in the lab). However, we soon discovered that PARP activity in human lymphocytes is orders of magnitude lower than in cultured cells and we would have needed approximately 50 ml blood pre- and post-treatment with rucaparib to be able to detect inhibition reliably. This is where yet another example of good fortune came in (serendipity #7). Alex Burkle had recently moved to Newcastle with his post-doc Ragen Pfeiffer and they had developed a much more sensitive immunoblot assay using the 10H antibody to the product, PAR [37]. They shared this technology with us and Ruth Plummer, who was my clinical fellow at that time, validated this pharmacodynamic (PD) assay to GCLP standard. It was initially intended that she would carry out the rucaparib Phase I trial. However, there was a substantial delay in the start of the trial, partially due to the take-over of Agouron by Warner Lambert (in January 1999) that was followed less than a year later by the take-over of Warner Lambert by Pfizer, who did not at that time have an oncology programme. Instead of conducting the rucaparib trial, Ruth used the PD assay to monitor the effects of temozolomide on PARP activity in patients with metastatic melanoma instead [38]. It was my next clinical fellow, Chris Jones, who conducted the trial under the guidance of Ruth Plummer and Hilary Calvert, the principal and chief investigators, respectively. Chris made further improvements to the assay and monitored PARP inhibition in the lymphocytes and tumour biopsies of patients treated with rucaparib [39]. Using this assay, another student of mine, Thomas Zaremba, set out to study PARP1 genomics, expression and activity in healthy volunteers and cancer patients with the aim of determining a) whether PARP genotype contributed to reduced PARP activity and the likelihood of developing cancer and b) whether PARP1 genotype, expression or activity contributed to unexpected toxicities in patients treated with temozolomide or radiotherapy. He did not manage to answer these questions but what he did find was that a) only 20% of the variation in PARP activity is due to the variation in PARP1 expression, b) that men have on average 40% more PARP activity than women and c) that it is androgen-driven, suggesting post-translational modification or co-activators/repressors that are androgen regulated [40]. The question this raises is: when used as a chemo- or radiosensitiser, should PARPi doses be lower in women?

The first trial of rucaparib was a dose escalation trial in cohorts of 3 patients with a variety of solid tumours receiving 1, 2, 4, 8 and 12 mg/m<sup>2</sup> in combination with half-dose temozolomide (100 mg/m2). There was a dose-dependent decrease in PARP activity, with the desired ≥50% inhibition of PARP activity for 24 h being consistently achieved above 8 mg/m2. In the absence of increased toxicity, 12 mg/m<sup>2</sup> was established as the safe dose to begin escalation of the temozolomide up to full dose (200 mg/m2) in melanoma patients consenting to a pre-treatment and post-treatment biopsy. There was a dose-dependent decrease in PARP activity in the tumours too, with approx. 90% inhibition 4–8 h after administration of 12 mg/m2. Escalating the dose to 18 mg/m2 led to increased haematological toxicity and 12 mg/m2 was established as the recommended Phase 2 dose to take forward in combination with temozolomide. Unfortunately, although this combination may have shown an improvement on historical data of response to temozolomide alone in patients with melanoma, there was also more haematological toxicity [41]. This may have doomed rucaparib to the same failure to progress beyond phase II studies, as with other DNA repair inhibitors such as the MGMT inhibitors O6 benzylguanine and Patrin [42], had it not been for the next and most significant example of serendipity on this project.

#### **5. Discovery of the Synthetic Lethality of PARPi**

The synergy between science and serendipity (#8) reached its peak with the discovery of synthetic lethality in BRCA mutant and other cancers defective in homologous recombination DNA repair (HRR). This came about through lucky coincidences with timing of various discoveries and meetings as described below. At the end of 2001, Herbie Newell was invited to speak at Sheffield University about the drug development programme in Newcastle University. Present at this meeting was Thomas Helleday, who had been recruited to Sheffield from Tomas Lindahl's lab. Tomas Lindahl (co-recipient of the Nobel prize for chemistry 2015 for his work on DNA repair) had a long-standing interest in PARP, its function and the effects of inhibitors. Tomas Lindahl had noted that in the presence of PARP inhibitors, there was an increase in recombination events. This led him to propose that the negative charge on the ADP-ribose polymer helped to repel negatively charged DNA from sites of DNA breakage to avoid unwanted recombination events [43]. Thomas Helleday had investigated further and found that cells lacking XRCC1 (the scaffold protein recruited to DNA breaks by ADP-ribose polymers [44], as well as those lacking PARP1, accumulated more γH2AX foci (a marker of DNA damage) and RAD51 foci (a marker of HRR). As a result of Herbie's talk in Sheffield Thomas requested some of our more potent and selective PARPi. We gave him some NU1025 and invited him to give a talk in Newcastle to present his work in January 2002. He convincingly showed that XRCC1 defects had similar effects in increasing HRR compared to PARP1 deletion or inhibition with NU1025. This suggested the polymers were not acting to inhibit recombination but instead a failure of BER led to an increase in HRR. He proposed that cell lacking BER/PARP function would be dependent on HRR for survival and as such cells lacking HRR function may be dependent on PARP activity for survival. He tested his hypothesis in cells defective in HRR (IRS1 lacking XRCC2 and irs1SF lacking XRCC3), which were killed by concentrations of NU1025 that did not kill wt or XRCC2/3 corrected cells.

By this time, I had widened my interest in DNA repair and as well as leading the pre-clinical biological studies on PARPi, I also led the DNA-PK and ATM inhibitor projects at Newcastle. As a result of this, I attended a meeting early in 2002, where Ashok Venkitaraman described the data supporting the key role of BRCA2 in HRR. BRCA 1 and BRCA 2 were discovered in the early 1990s as breast cancer susceptibility genes but at first it was not clear why. Their function in DNA repair began to emerge towards the end of that decade and it was confirmed that they played key roles in HRR around the turn of the millennium [45–48]. More reading around the topic convinced me that our PARP inhibitors might have therapeutic potential in BRCA mutant hereditary breast cancer, which I put to the drug development team in May 2002. The BRCA2 mutant Capan-1 pancreatic cells along with the BRCA wt BxPC-3 pancreatic cells, as control, were duly purchased and their sensitivity to AG14361 was determined by Suzanne Kyle. Frustratingly, these initial experiments, conducted in July 2002, did not show that the BRCA mutant cells were more sensitive. However, experiments that had been conducted 3 to 4 months previously in wt (AA8) and the HRD derivative (XRCC3 mutant: irs1SF) Chinese hamster ovary cells had indeed shown a very marked difference in the sensitivity of the wt and HRD cells (Figure 3). To this day, I am not sure why the data were so disappointing with the Capan-1 cells as we got more impressive results both in vitro and in vivo with this cell line using rucaparib (see later). Maybe it was the design of the experiment and the exposure period (24 h) was insufficient, as less than a complete cell cycle, or the moribund cells were lost during harvesting or maybe the BxPC-3 cells were also HRD.

**Figure 3.** AG14361 is more cytotoxic in XRCC3 mutant irs1SF cells compared to wild-type AA8 Chinese hamster ovary cells (**A**) but BRCA2 mutant Capan-1 cells are not more sensitive than BRCA wild-type BxPC-3 pancreatic cancer cells (**B**). Cells were exposed to increasing concentrations of AG14361 for 24 h prior to seeding for colony formation in fresh medium. The insert shows an extended concentration range for pancreatic cell lines. Previously unpublished data pooled from three independent experiments.

We communicated our findings to Thomas Helleday, with whom we had established a collaboration, in the autumn of 2002. By Spring 2003, he had obtained the VC8 BRCA2 mutant derivatives of Chinese hamster lung fibroblast V79 cells, which he later shared with us. Suzanne got her first result with these cells in June 2003 and they were so impressive we could hardly believe it. She replicated the result several times and confirmed that 98.5% of BRCA2 mutant cells were killed by 10 μM AG14361, a concentration that had no significant effect on the viability of either parental wt cells or the BRCA2-corrected cells. Conducting in vivo experiments with these cells proved more of a

challenge. The V-C8 cells would not grow subcutaneously and, conversely, the V79 cells generated very haemorrhagic tumours that rapidly made the mice unwell such that they had to be humanely killed. However, Huw Thomas eventually succeeded in growing the V-C8 cells and the BRCA2 corrected V-C8.B2 cells intramuscularly. These experiments showed that treatment of mice with AG14361 caused a complete regression of the BRCA2 mutant tumour but growth of the BRCA2 corrected tumour was unaffected. The data generated by Suzanne and Huw went into the paper that was published in Nature in 2005 [49]. These key experiments, which have led to the current use of PARPi clinically, were actually undertaken outside of the remit of the drug development programme. This was because at Herbie Newell's suggestion, the rest of the drug development management team at Newcastle had decided that PARP was no longer to be considered a drug development project to be discussed at subsequent meetings in June 2003 as the clinical trial (albeit in combination with temozolomide) was underway.

The data with the BRCA mutant and HRD models, where a tumour-specific defect in DNA repair is exploited by inhibiting a complementary DNA repair pathway were indeed ground-breaking. The findings completely changed the way we think about treating cancer, from trying to overcome some property of the tumour that gave it an advantage to exploiting a vulnerability. Indeed, the search for additional examples of synthetic lethality has become somewhat of a holy grail. However, I believe PARP inhibition in HRD is unlikely to be bettered as an example of synthetic lethality for 2 reasons: (1) there is a very high level of DNA SSB, that depend on PARP for their repair, due to the endogenous ROS produced by normal metabolism, and this is increased in cancer by virtue of the associated inflammation, and (2) HRR defects are relatively common in cancer, not only due to BRCA mutations. This is yet another example of the synergy of science and serendipity (#9)

Understanding the importance of the synthetic lethality of PARPi in HRD cancers, Thomas Helleday was keen to patent our findings before publication. Following meetings with Cancer Research Technology, we undertook work to protect rucaparib and these data were included in the patent (WO 2005/012305 A2) filed 16th April 2004.

#### **6. The Clinical Development of Rubraca**®**as A Single Agent and Predictive Biomarker Development in Ovarian Cancer**

During this time, the Phase I clinical trial of rucaparib with temozolomide was ongoing and there was clinical interaction with Pfizer who later conduct a single agent dose escalation study with rucaparib. This study was led by Hilary Calvert and Ruth Plummer, who recruited Yvette Drew, a clinical fellow, in 2006 to undertake the trial. In parallel, Yvette conducted pre-clinical and translational studies with me. The translational studies included taking blood samples from the patients before and at intervals after rucaparib adminiatration. These samples were used to determine the rucaparib PK and also PARP1 genotype, protein levels and inhibition by rucaparib in the patients' PBMCs (lymphocytes).She also undertook studies in a panel of BRCA wt and mutant human cancer cell lines to confirm the differential sensitivity we had seen in the matched Chinese hamster cells (AA8/irs1SF and V-C8/V-C8.B2). This panel included the UACC3199, which had epigenetically silenced BRCA1 due to promoter methylation and the Capan-1 cells that had been disappointing in the AG14361 experiments. Such studies were needed because there were reports that other BRCA mutant human cancer cells were insensitive to PARPi, including AG14361 [50,51]. Fortunately, rucaparib sensitivity was found to be substantially greater in the BRCA mutant cells (including the Capan-1 cells) than the wt and, importantly, the BRCA silenced cells were comparable with the mutant ones in this respect [52].

At the same time, I had established an interaction with a gynaecological surgeon, Richard Edmondson, who had a number of trainees keen to undertake lab-based projects for MD or PhD. This turned out to be yet another example of the synergy between science and serendipity (#10). I was aware that BRCA mutation/HRD was a determinant of sensitivity to cisplatin/carboplatin from discussions I had had with Paul Harkin at a meeting in 2003. The standard of care for ovarian cancer is surgery followed by carboplatin + paclitaxel chemotherapy with an initial response rate of approximately 60%. Knowing that only approximately 15%–25% of ovarian cancer is associated with BRCA mutations, that left 35%–45% of the responses unaccounted for. I reasoned that they must be HRD for some other reason and that we could test this hypothesis by measuring γH2AX and RAD51 foci. γH2AX foci are formed at collapsed replication forks and DNA DSBs and RAD51 coats the SS DNA to form the nucleoprotein filament needed to invade the complementary DNA on the sister chromatid during HRR resulting in RAD51 foci. Therefore, γH2AX foci indicate the lesions generated by PARP inhibition and RAD51 foci indicate when HRR occurs. We first met with Pfizer representatives (including Zdenek Hostomsky and Gerrit Los) in September 2006 to suggest that Yvette Drew, the clinical fellow undertaking the trial of single agent rucaparib, would not only work with cell lines to establish the spectrum of rucaparib sensitivity (described above) but also to develop an assay for HRR function, based on measuring γH2AX and RAD51 foci. We also proposed to evaluate the HRR status in primary cultures of ovarian cancer ascites collected during primary/debulking surgery with another clinical fellow Asima Mukhopadhyay, a trainee gynaecological oncology surgeon supervised by Richard Edmondson and me. Thanks to Zdenek Hostomsky, ever our PARP champion even though he had been diverted to other projects, I secured funding from Pfizer in 2008 to support these studies with an additional technician, Evan Mulligan, to help with planned in vivo efficacy studies..

These studies led to two key papers published in 2010/11 [52,53]. The data Yvette generated in the cell lines not only demonstrated the difference in rucaparib cytotoxicity between HRD and HRR functional cells but also that RAD51 foci were a useful discriminant/potential biomarker of HRR function. That is, the induction of RAD51 foci in the HRD cells was not significant whereas there was a significant increase in the BRCA wt cells. Looking at γH2AX foci really brought it home to us how much endogenous damage cells sustain without the addition of exogenous genotoxic agents. The level of the foci after only 24 h of inhibition of endogenous damage repair with rucaparib was the same as the standard radiotherapy fraction of 2 Gy. After some significant work to establish the optimum conditions to grow primary cultures from ascites, Asima successfully translated this HRR function assay to the ascites cultures and her initial studies showed that approximately 60% of the primary cultures were indeed HRD and that this corresponded to a greater rucaparib-induced inhibition of cell growth. This was the first demonstration that >50% of ovarian cancers are HRD and, at significantly less expense than similar findings by the TCGA, which was published the following year [54].

The in vivo experiments in mice bearing BRCA1 mutant (MDA-MB-436) or BRCA2 mutant (Capan-1) xenografts, conducted by Yvette Drew and Evan Mulligan, showed that rucaparib caused significant tumour growth delay and this was more pronounced when the treatment period was 5x weekly for 6 weeks compared with daily x10. This was not entirely unexpected from an understanding of the underlying mechanism whereby to kill the cells PARP must be inhibited as the cell progresses through S-phase. These findings were recapitulated in the clinical study that Yvette was conducting where responses were seen when the patients switched from daily treatment for 5 days every 3 weeks to continuous dosing [55]. This study, initiated in 2007, was the first single-agent trial of rucaparib, sponsored by CRUK and Pfizer. It began very cautiously, starting at low dose (escalating from 4 to 18 mg/m2 i.v) on an intermittent schedule (daily for 5 days every 3 weeks) based on the trial in combination with temozolomide. During the study, there was a switch to an oral formulation and daily dosing, starting at 92 mg (equivalent to 18 mg/m2 based on an oral bioavailability of approximately 30% and average body surface area of 1.7 m2) once a day for 7 days and escalating to 600 mg 2x/day continuously. Many more responses were seen once dosing was continuous. It is not altogether clear whether it was the increase in the dose or the intensity of the schedule that was critical. Unfortunately, this rather slow start allowed rucaparib to be overtaken by olaparib, which went into its first clinical trial as a single agent and the dose was escalated more quickly, as there was no history of combination toxicity data to suggest a cautious approach was necessary. As a result, the first publication showing that the pre-clinical data on the synthetic lethality of PARPi in HRD/BRCA mutant cancer was translatable to the clinic was with olaparib [56].

In a way, this result could have been avoided with more careful scrutiny of the pre-clinical data. Pre-clinically much lower (at least 10x) concentrations (in cell cultures) and doses (in tumour xenograft

studies in mice) were needed to sensitise cells and tumours to temozolomide than were needed for single agent activity in BRCA mutated/HRD cancer cells. Furthermore, increasing the dose of rucaparib to one that was active against BRCA mutant xenografts was so toxic that the mice had to be humanely killed a few days into the drug administration [29,35]. These differences are easily understood by reference to the mode of action of the PARPi in the different scenarios. Because DNA integrity is so important, the cell has more than adequate levels of repair enzymes to cope with endogenous and most environmental DNA damage. Couple this with the fact that single base lesions and SSB are the commonest form of endogenous lesions, and these are dependent on PARP for their repair, and it is easy to see that cells will have a surplus of PARP to cope with daily levels of DNA damage. In order for the damage to overwhelm this repair capacity and cause death in cells lacking the back-up pathway of HRR, there has to be a total suppression of PARP activity and relatively high levels of drug are needed to achieve this. In addition, PARP must be inhibited long enough for all tumour cells to enter S-phase with unrepaired DNA lesions and so long/continuous exposures are necessary. In the alternative scenario where, e.g., temozolomide, induces orders of magnitude greater numbers of base lesions and SSBs, it is not necessary to inhibit their repair completely to render the remaining DNA damage sufficient to kill cells and so much lower levels and shorter duration of PARP inhibition are needed. Also, when used as a single agent, the PARPi is exploiting a tumour-specific defect, and side effects are not expected so higher doses can be used. When the PARPi is synergistic with the cytotoxic in cancer cells the combination is likely to also be synergistically toxic to normal tissues so only lower doses are tolerated.

The observation that continuous therapy was much better than intermittent led us to investigate how durable PARP inhibition by rucaparib was. This study, initiated in 2011 in collaboration with and support from Zdenek Hostomsky and Gerrit Los, was partially based on our observations in the first clinical trial: as the dose of rucaparib was increased PARP activity was suppressed in PBMCs for 24 h after the 30 min infusion, even though the drug was no longer detectable in the plasma. Indeed, at the 12 mg/m2 dose, PARP activity was still way below pre-dose levels on the Monday after the last Friday dose in many patients [39]. Pre-clinically we found that rucaparib accumulated above the extracellular concentration and was retained after drug removal in cancer cells and that PARP was inhibited for at least 72 h after a 30 min exposure. In in vivo studies, a single oral dose of rucaparib inhibited PARP activity in xenografts for at least 7 days even though it was cleared from the blood within 48 h and was undetectable in the tumour within 72 h. Moreover, tumour growth inhibition following weekly dosing with rucaparib was equivalent to daily dosing [57]. This, I believe, is another example of serendipity (#11) ensuring the continuous suppression of PARP activity necessary for antitumour activity as a single agent. It is possible that this might be exploited clinically by weekly dosing schedules.

Pfizer initiated a Phase I study in combination with carboplatin alone or with paclitaxel or pemetrexed or a combination of cyclophosphamide and epirubicin NCT01009190 in November 2009 [58]. There was little pre-clinical rationale for such a study, as PARPi do not sensitise cells to either pemetrexed (an antifolate), cyclophosphamide (DNA cross-linking agent), epirubicin (DNA intercalating topoisomerase II poison) or paclitaxel (an antitubulin agent, not a DNA damaging agent) and the data with cisplatin or carboplatin are variable and cell line dependent, most probably additive in HRD cells only [59]. The triple combination arms were dropped during the study and the MTD in combination with carboplatin (AUC 5) was 240 mg one daily (carboplatin day 1, rucaparib days 1–14 of a 21 day cycle) due largely to haematological toxicities [58]. Thus, the MTD for rucaparib in combination with carboplatin (with which is does not synergise) is approximately four times higher than with temozolomide and 2.5x lower than as a single agent.

#### **7. Studies That Led to FDA Approval**

In June 2011, Pfizer licenced rucaparib to Clovis oncology and in November, study 10, (NCT01482715) a phase I–II trial of oral rucaparib as a single agent, was initiated [60,61]. By this time, Hilary Calvert had moved from Newcastle to UCL and he supported another clinician, Rebecca

Kristeleit, who led the study. The Phase I component was a dose escalation of the oral formulation, based on the intermittent vs. continuous dosing study described above, in patients with solid tumours to establish the recommended phase II dose (RP2D) based on tolerability and efficacy. The RP2D of 600 mg 2x daily was then used to treat women with BRCA1/2 mutated high-grade serous ovarian cancer as the Phase II part of the study. The objective response rate was 60% and the toxicities were generally mild with only 5 out of 98 patients having to discontinue treatment due to an adverse event.

Ongoing studies with other PARPi clearly demonstrated that BRCA mutation was not synonymous with a response in ovarian cancer [62], with cases with BRCA mutations in the non-responder group as well as cases without BRCA mutations in the responder group. Indeed, our earlier studies (described above) clearly indicated that BRCA mutation significantly underestimated the HRD and PARPi sensitive population. Several companies were developing predictive biomarkers of HRD and we shared some of our samples that we had previously characterised by the functional assay (by this time, our collection had grown substantially) with Tom Harding at Clovis and Foundation Medicine from 2011 to 2013. However, it seemed that the normal cellular component (that died off during culture leaving us with just the cancer cells to classify) diluted the cancer component too much for the samples to be of much use. Nevertheless, based on the observation that BRCA mutations result in substantial genome-wide loss of heterozygosity (LOH), Foundation Medicine [63] developed a genomic profiling test to identify HRD independent of BRCA mutation to be used as a companion diagnostic in the Ariel 2 trial NCT01891344 [64]. Similar tests have been developed by Myriad. Although these tests represent genetic "scarring" caused by an HRR defect at some time during the cancer's development, but not necessarily at the time of treatment, they are substantially less challenging than functional assays, which by their nature require viable fresh tumour tissue. Screening for mutations in known HRR genes has also been used.

Ariel 2 was an innovative and pivotal trial that led to the rucaparib being given breakthrough status with the FDA. In this trial, 198 patients with high-grade serous ovarian cancer (HGSOC) were characterised: 40 had BRCA mutations (germline or somatic), 80 had high levels of LOH (presumed HRD) and the remaining 70 had neither BRCA mutation nor LOH. The median progression-free survival for the BRCA mutant and high LOH groups was significantly longer than the biomarker negative group. Accelerated approval was given by the FDA in December 2016. The follow-on Ariel 3 study led to the FDA approval of Rubraca (in April 2018) as maintenance therapy in platinum-sensitive ovarian, fallopian tube or primary peritoneal cancer, along with Foundation Medicine's complementary diagnostic CDx BRCA LOH in 2016. Ariel 3 (NCT01968213) was a randomised double-blind placebo-controlled trial in 564 patients that had responded to platinum-based chemotherapy. This trial showed a significant improvement in median progression-free survival in all patients treated with rucaparib (10.8 months) compared to placebo (5.4 months), but particularly the 130 BRCA mutated (16.6 months) and 106 BRCA wild-type, LOH-high HRD patient (9.7 months months) subgroups compared to the BRCA wt LOH low (6.7 months) [65]. More recently, Rubraca has been approved by the European Medicines Agency in January 2019 and by the National Institute for Health and Care Excellence in October 2019 for maintenance therapy in epithelial ovarian, fallopian tube, or primary peritoneal cancer who are in a complete or partial response to platinum-based chemotherapy, without the need for a companion diagnostic. Foundation Medicine's CDx BRCA LOH may still be used as a complementary diagnostic. This is because the Ariel 3 trial showed that platinum sensitivity was a sufficiently good indicator of response to rucaparib, which was indeed the basis of our hypothesis that led us to investigate HRD status in ovarian cancer in 2008.

Rubraca is in several advanced clinical trials as a single agent in other cancers associated with BRCA mutations: breast, pancreatic and prostate cancer. It has breakthrough designation with the FDA for castrate-resistant prostate cancer. It seems as though the best responses are observed in the ovarian setting. Whether this is because BRCA mutations, or even LOH are not entirely synonymous with HRD [66], or whether something within the tumour microenvironment influences the outcome is anybody's guess at the current time. If it is the former, then our small studies on the frequency of cancers lacking HRR function, as measured in viable tumour tissue, may be relevant [67] and suggest more complex readouts of HRR function may be necessary. Combinations with other molecularly targeted agents are also being investigated in other tumour types. For most up to date data, visit https://www.clovisoncology.com/pipeline/rucaparib/ and https://clinicaltrials.gov/.

#### **8. Conclusions**

In conclusion, the story of the development of rucaparib/Rubraca shows not only the necessary input of intellectual (creative, deductive and analytical) and physical effort needed over a protracted period of time for the successful development of a drug (summarised in the timeline, Figure 4) but also how luck played a significant role all the way through from the initial hit right through to the clinical development in ovarian cancer. The role of luck is by no means unique to the development of Rubraca®, luck has probably contributed to many of the scientific advances that we take for granted. One prime example was the production of penicillin during the Second World War, which would not have been possible without the necessary contacts and proximity of the potteries, a fact acknowledged by Norman Heatley himself in his lecture "Penicillin and Luck" [68]. There have been some setbacks in the development of Rubraca®but, on the whole, luck has been on our side—most significantly, the fact that PARPi block the repair of the most common forms of endogenously generated DNA damage and that this is synthetically lethal with a relatively common DNA repair defect in cancer, HRR. This has caused a paradigm shift in how we think about treating cancer and there are numerous investigations to identify other instances of synthetic lethality. It seems unlikely that anything as profound as the synthetic lethality of PARPi against HRD cancers will be identified. Some examples of serendipity are yet to be exploited to the full—in particular, the discovery of the durability of PARP inhibition by rucaparib warrants further exploration clinically in pharmacodynamically guided intermittent schedules. The outcome of such a study may prove to be another paradigm shift in our handling of molecularly targeted drugs, where the biologically effective dose rather than the tolerable dose is adopted as the appropriate measure to decide the recommended phase II dose.

**Figure 4.** Timeline of the development of Rubraca®. Key milestones are indicated at the top of the timeline, the nature of the work undertaken at Newcastle, the funding sources and pharmaceutical companies involved are shown below the timeline.

#### **9. Patents**

#### *9.1. Expired Patents*

Griffin, R. J., Calvert, A. H., Curtin, N. J., Newell, D. R., and Golding, B. T. Benzamide Analogues ADPRT (PARP) Inhibitors. Patent Application Number PCT/GB95/00513 (1995)

Griffin, R. J., Calvert, A. H., Curtin, N. J., Newell, D. R. and Golding, B. T. Benzimidazole PARP Inhibitors.Patent Application Number PCT/GB96/01832.

Griffin, R. J., Calvert, A. H., Curtin, N. J., Newell, D. R. and Golding, B. T. Prodrugs of PARP Inhibitors.Patent Application Number PCT/GB97/02701.5

#### *9.2. Active Patents*

Helleday T and Curtin NJ. Therapeutic Compounds (PARP inhibitors in homologous repair/BRCA defective cancer) Patent Application Number PCT/GB2004/003183. Publication numberWO 2005/012305 A2 Divisional application 16th April 2004 GB 0408524. WO2005012305A3

Boritzki TJ, Calvert AH, Curtin NJ, Dewji MR, Hostomsky Z, Jones C, Kaufman R, Klamerus KJ, Newell DR, Plummer ER , Reich SD, Steinfeldt HM, Stratford IJ, Thomas HR Williams KJ.. Therapeutic Combinations Comprising PARP inhibitor US application No. 60/612,458 Filed 22nd September 2004 WO/2006/033006) THERAPEUTIC COMBINATIONS COMPRISING POLY(ADP-RIBOSE) POLYMERASES INHIBITOR

**Funding:** This research has been funded by several sources over the nearly 30 years it has been going—all of which have been acknowledged in the primary research papers cited here.

**Acknowledgments:** People who have contributed to the body of work have been acknowledged in the text but special thanks go to my students, clinical fellows and technical staff who worked so hard to generate the data and to Zdenek Hostomsky who initiated the collaboration with Agouron and continued to champion my work until he left Pfizer.

**Conflicts of Interest:** I have had funding from Agouron, Pfizer and Clovis, for some of the work described in this review. By virtue of the two active patents and an agreement between Cancer Research Technology, Newcastle University, and Agouron, now Pfizer, and Clovis, I am in receipt of royalty payments. I do not take these personally, modest sums have contributed to my research accounts and a recent large sum has been used to set up a charitable fund within the local Community Foundation. https://www.communityfoundation.org.uk/ group\_grant/passionate-about-realising-your-potential/

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Role of Akt Activation in PARP Inhibitor Resistance in Cancer**

**Ferenc Gallyas Jr. 1,2,3,\*, Balazs Sumegi 1,2,3,**† **and Csaba Szabo 1,4**


Received: 2 January 2020; Accepted: 24 February 2020; Published: 25 February 2020

**Abstract:** Poly(ADP-ribose) polymerase (PARP) inhibitors have recently been introduced in the therapy of several types of cancers not responding to conventional treatments. However, *de novo* and acquired PARP inhibitor resistance is a significant limiting factor in the clinical therapy, and the underlying mechanisms are not fully understood. Activity of the cytoprotective phosphatidylinositol-3 kinase (PI3K)-Akt pathway is often increased in human cancer that could result from mutation, expressional change, or amplification of upstream growth-related factor signaling elements or elements of the Akt pathway itself. However, PARP-inhibitor-induced activation of the cytoprotective PI3K-Akt pathway is overlooked, although it likely contributes to the development of PARP inhibitor resistance. Here, we briefly summarize the biological role of the PI3K-Akt pathway. Next, we overview the significance of the PARP-Akt interplay in shock, inflammation, cardiac and cerebral reperfusion, and cancer. We also discuss a recently discovered molecular mechanism that explains how PARP inhibition induces Akt activation and may account for apoptosis resistance and mitochondrial protection in oxidative stress and in cancer.

**Keywords:** PARP-Akt interplay; PI3K; mTOR; cytoprotection; apoptosis resistance; oxidative stress; mitochondrial protection

#### **1. Introduction**

Inhibitors of the constitutive mammalian enzyme poly(ADP-ribose) polymerase (PARP)1 received renewed interest in recent years due to their importance in cancer therapy [1]. Mostly, their use is based on the synthetic lethality theory [2], i.e., to block the single-strand DNA break repair process in double-strand DNA break repair (BRCA1/2) deficient cancer cells that selectively eliminates the cancer cells via DNA damage accumulation while not affecting normal cells possessing intact double-strand DNA break repair system [3,4]. Accordingly, review articles focusing on the role of PARP inhibitors in DNA repair vastly dominate over those that at least mention the effect of PARP inhibition on kinase signaling, although the latter effect is an important limiting factor in the cancer therapy [5], and a pivotal mechanism in non-cancerous therapeutic applications of the PARP inhibitors [6].

The phosphatidylinositol-3 kinase (PI3K)-Akt-mammalian target of rapamycin (mTOR) pathway is very often activated in human cancer [7]. The pathway is often utilized by various cytokine and growth factor receptor signaling, and mutation, expressional change or amplification of any of its elements could cause resistance against tumor therapy [7,8]. Accordingly, all members of this pathway (including Akt) were suggested as targets in monotherapy or combination therapy for various cancers. In turn, several inhibitors of the pathway were approved by the Food and Drug Administration [9]. In previous research papers, we demonstrated that PARP inhibition leads to Akt activation resulting in cyto- and mitochondria protecting actions [10]. In turn, we proposed that Akt inhibitors may be used to prevent limitation of PARP inhibitor based anti-cancer therapy [11]. Moreover, we have recently unveiled a novel molecular mechanism for the Akt activating effect of PARP inhibition [12]. In the present review, we seek to increase awareness of the therapeutic limitations that the PARP-Akt interplay represent. In addition, we discuss the role of the PARP-Akt interplay in non-oncological disease conditions and their experimental models.

#### **2. The PARP-Akt Interplay**

#### *2.1. Cellular Role and Regulation of Akt*

Protein kinase B/Akt, a 57-kDa serine/threonine kinase, is regarded as a pivotal mediator of cell survival promoting effects of various growth factors and other factors, including cAMP, hypoxia, and cytokines (Figure 1). In mammals, three Akt genes were identified that are differentially expressed at both the mRNA and protein levels [13]. Akt1 can phosphorylate thereby inactivate various components of the apoptotic machinery, therefore, it is involved in cellular survival pathways. Akt2's function seems to be closely associated with the insulin receptor signaling pathway while the specific role of Akt3 is less clear [14]. Structure of Akts' amino terminus pleckstrin homology (PH) domain and central kinase domain is conserved across evolution, while the carboxyl terminus hydrophobic and proline-rich domain is more variable. The kinase activity of Akt is regulated by phosphorylation of Thr<sup>308</sup> and Ser<sup>473</sup> situated in the PH domain, although phosphorylation of Ser<sup>124</sup> and Thr<sup>450</sup> sites are needed for sensitizing Akt toward regulatory stimuli [13]. In the classical activation pathway of Akt (Figure 1), the upstream element PI3K phosphorylates phosphatidylinositol-4,5-bisphosphate to phosphatidylinositol-3,4,5-trisphosphate (PIP3) that binds to PH domain of the constitutively active 3-phosphoinositide-dependent protein kinase (PDK)1. Binding of PIP3 to PDK1 results in translocation of the enzyme to the cell membrane's inner surface, a crucially important step in Akt's activation process. PIP3's binding to Akt's PH domain renders Akt more accessible to the PDK1 mediated phosphorylation of Thr308 regulation site [13]. Negative regulation of the pathway is mediated by the phosphatase and tensin homolog PTEN that removes the 3-phosphate from PIP3 [15].

Besides the classical pathway, Akt can be activated by various signaling processes. PDK1 can interact with the downstream element protein kinase C related kinase2, rendering it capable to phosphorylate both Thr308 and Ser473 of Akt1 in a PIP3 dependent manner [16]. The mammalian target of rapamycin complex 2 (mTORC2) that phosphorylates Ser473 is a PDK1 independent crucial mechanism of Akt activation (Figure 1). Additionally, mTORC2 and cyclin-dependent kinase 2/cyclin can phosphorylate Ser<sup>477</sup> and Thr479, also resulting in Akt activation [17]. Integrin-linked kinase and calcium/calmodulin-dependent kinase-kinase that responds to the increases in intracellular calcium, activate Akt also by phosphorylating Ser<sup>473</sup> [18,19]. When activated, Akt can phosphorylate various downstream targets that contain the consensus phosphorylation sequence RXRXXS/T, although Akt's substrate specificity also depends on sequence determinants outside the consensus site.

By phosphorylating its downstream targets, Akt can regulate metabolism, apoptosis, transcription factors, angiogenesis, and cell cycle progression (Figure 1). The constitutively active glycogen synthase kinase (GSK)-3 phosphorylates thereby inactivates glycogen synthase, the enzyme responsible for the synthesis of glycogen from excess glucose in the liver. When high glucose levels in the blood triggers insulin secretion, the insulin signaling activates Akt that inactivate GSK-3 by Ser<sup>9</sup> phosphorylation enabling glycogen synthesis that contribute to decreasing the blood glucose level [20]. In parallel, Akt activation via Akt-targeted Rab GTPase-activating proteins and downstream Rab GTPases along with the input of Rac1 and actin filaments, molecular motors, and membrane fusion regulators leads to cell surface expression of glucose transporters GLUT4 that allow the entry of excess blood glucose into muscle and adipose tissues (Figure 1). Attenuation of these processes results in symptoms of metabolic syndrome and type 2 diabetes [21].

**Figure 1.** Schematic representation of the Akt signaling pathway. Pointed arrows denote activation while arrows with flat end represent inhibition. P indicates phosphorylation. AS160, Akt substrate of 160 kDa; Bad, Bcl-2 associated agonist of cell death; Deptor, dishevelled, Egl-10, and Pleckstrin domain-containing mTOR-interacting protein; FOXO, class O of forkhead box transcription factors; Gαβγ, G protein subunits; GF, growth factor; GRF, growth-related factor; GPCR, G protein-coupled receptor; GS, glycogen synthase; GSK, GS kinase; mLST8, mammalian lethal with Sec13 protein 8; mSINI, mammalian stress-activated protein kinase interacting protein; mTOR, mammalian target of rapamycin; mTORC, mTOR complex; Nrf2, nuclear factor erythroid 2-related factor 2; PDK, 3-phosphoinositide-dependent protein kinase; PI3K, phosphatidylinositol-3 kinase; PIP2 phosphatidylinositol-bisphosphate; PIP3, phosphatidylinositol-3,4,5-trisphosphate; PKB, protein kinase B; PRAS40, proline-rich Akt substrate of 40 kDa; Protor, protein observed with rictor; PTEN, phosphatase and tensin homolog; Raptor, regulatory protein associated with mTOR; Rheb, Ras homolog enriched in brain; Rictor, rapamycin-insensitive companion of mTOR; RTK, Tyrosine kinase receptor; SHIP, phosphoinositide phosphatase; TBC1D7, TBC1 domain family member 7; TSC, tuberous sclerosis protein.

Activated Akt regulates various downstream signaling proteins that promote cell survival [22]. Phosphorylation thereby inactivation of the pro-apoptotic B-cell lymphoma 2 (Bcl-2) family member Bcl-2 associated agonist of cell death (Bad) by Akt prevents Bad to form heterodimers with other pro-apoptotic Bcl-2 family members, translocate to the mitochondria, and, by labilizing its outer membrane, facilitate the release of cytochrome C, apoptosis inducing factor and endonuclease G from the mitochondrial intermembrane space (Figure 1). Instead, phosphorylated Bad forms a complex with the cytoplasmic scaffolding protein 14-3-3, shifting the balance toward dominance of anti-apoptotic Bcl-2 family members, stabilization of mitochondrial integrity, and promoting cell survival [23]. Besides blocking cytochrome C mediated activation of caspase 9, Akt effectively inactivates it by directly phosphorylating its Ser196 [24].

Inactivation of GSK-3 by Akt is another major survival pathway since, in addition to its metabolic role, GSK-3 extensively participates in signaling pathways. Under hypoxic condition, its Tyr216 is phosphorylated, resulting in activation of the enzyme [25]. The activated GSK-3 contributes to hypoxic-ischemic tissue injury by multiple mechanisms. It downregulates nuclear factor erythroid 2-related factor 2 (Nrf2) expression, Nrf2 translocation from the cytoplasm to the nucleus, and Nrf2 binding to antioxidant response element (ARE) DNA sequence, thereby limiting expression of antioxidant proteins encoded in Nrf2/ARE regulated genes (Figure 1) [26]. Apart from its pro-oxidative activity, GSK-3 activation causes decreased nuclear translocation of cAMP response element-binding protein (CREB) transcription factor [27], leading to increased expression of pro-inflammatory cytokines via altering the interaction of CREB and the pro-inflammatory transcription factor nuclear factor (NF)κB with the co-activator CREB binding protein (CBP) in the nucleus [28].

After activation, Akt translocates to the nucleus where it modulates by phosphorylation the activity of various transcription factors [29]. When phosphorylated by Akt, forkhead family transcription factors are retained in the cytoplasm in form of a complex with the 14-3-3 protein and cannot promote expression of genes encoding growth factors, stress response proteins, and enzymes of carbohydrate and lipid synthesis [30]. Akt not only prevents GSK-3's negative effect on CREB activation [28], it phosphorylates CREB at its Ser<sup>133</sup> that enhances CREB's binding to CBP thereby increases CREB-mediated transcription of critical survival genes [31]. Inhibitor of NFκB (IκB) retains NFκB in the cytoplasm by forming a complex with it. Phosphorylation by its kinase (IKK), causes the release of NFκB from the complex, and marks IκB for ubiquitination and proteosomal degradation. Akt regulates the process by phosphorylating IKK at Thr23 thereby activating it that results eventually in expression of NF-κB dependent anti-apoptotic genes [32,33]. Additionally, Akt activates mTOR, a major node of the PI3K-Akt pathway, by direct phosphorylation. Activation of mTOR trigger translation initiation factors, hypoxia and angiogenesis associated factors, and cell survival and growth associated transcription factors [34]. All in all, the PI3K-Akt pathway promotes cell proliferation and survival by various mechanisms, therefore, hyperactivation of the pathway leads to apoptosis resistance and malignant transformation [13,35].

#### *2.2. Nuclear E*ff*ects of PARP Activation*

PARP1, the original (and principal) member of a family of 18 enzymes [36], binds to various DNA structures, including single- and double-strand breaks, crossovers, cruciforms, supercoils, and some specific double-stranded sequences [37]. This binding stimulates PARP1's very low basal enzymatic activity. It cleaves its substrate, NAD<sup>+</sup> to nicotinamide and an ADP-ribose unit, which it covalently attaches on glutamate, aspartate, arginine, lysine, or serine residues of target proteins. PARP1 can further catalyze formation of 2 -1-O-glycosidic or α(1–2)-ADP-ribose bonds, resulting in linear or branched ADP-ribose polymer (PAR) chains, respectively [38]. PAR chains are heterogeneous either in length (up to 200 ADP-ribose units) or in the extent of branching (one branch per 20–50 ADP-ribose units), however the role of this heterogeneity has not yet been elucidated [39]. PAR chains are catabolized very shortly after their synthesis to free ADP-ribose units by exo- and endoglycosidase activities of poly-(ADP-ribose) glycohydrolases (PARGs) [39]. The mono(ADP-ribose) unit bound directly to the protein that cannot be removed by PARG is cleaved off by terminal ADP-ribose protein glycohydrolases (TARGs) [40]. The PARP1 enzyme exists in a million-copy number in the nucleus and is responsible for the majority of PARylation occurring in mammalian cells [41]. In addition to the nuclear PARP, there is also a mitochondrial form of this enzyme in many cell types, the functional role of which is currently less understood [42]. The targets of PARP1's enzymatic activity include core histones, the linker histone H1, a variety of transcription-related factors, and PARP1 itself, which is the primary target in vivo [43]. Due to its localization, activation characteristics, and enzymatic activity, PARP1 is considered to play a pivotal role in various DNA repair systems, and in the maintenance of genome integrity [44]. When PARP1 is knocked out, PARP2 takes over its role and vice versa, however the PARP1/PARP2 double knock out mutation results in embryonic lethality [41].

Recent reports indicate a scaffolding role for PARP1 in genotoxic stress induced apoptotic response [45,46]. PARP1 activated by DNA strand breaks PARylates itself and ataxia telangiectasia mutated kinase (ATM) to recruit a small ubiquitin-like modifier (SUMO)-1 ligase, PIASy, and nuclear IKKγ (NEMO). Concertedly, PIASy SUMOylates and ATM phosphorylates NEMO, leading to IKK activation and NFκB regulated resistance toward apoptosis [45].

Because of PARP1's high sensitivity toward DNA breaks and its copiousness, genotoxic stress can induce hyperactivation of the enzyme leading to deleterious consequences for the organism [47]. Experimentally induced massive DNA damage triggers excessive PARP1 activation that causes depletion of the substrate NAD<sup>+</sup> within minutes [47]. Therefore, the metabolic pathways that utilize NAD<sup>+</sup> as substrate such as glycolysis, citric acid cycle and mitochondrial respiration become compromised, resulting in reduced ATP production and cellular dysfunction. To aggravate the precarious metabolic situation, phosphoribosyl pyrophosphate synthetase and nicotinamide mononucleotide adenylyl transferase consume ATP to resynthesize NAD<sup>+</sup> leading to ATP depletion and predominantly necrotic cell death [48,49]. This mechanism (Figure 2) explains the detrimental effect of DNA-damaging agents, such as high dose of N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) on cultured cells, and is consistent with the therapeutic efficacy of PARP-1 inhibition in animal models of oxidative stress related diseases. However, some studies indicated the necessity of including other mechanisms to explain the cytoprotective effect of PARP inhibition. Namely, MNNG was reported to cause similar NAD<sup>+</sup> depletion in PARP1−/<sup>−</sup> and PARP1+/+ fibroblasts [50], and PARP1 inhibition failed to prevent NAD<sup>+</sup> and ATP depletion in carcinogen-treated hamster cells [51].

**Figure 2.** Schematic mechanisms of the PARP1/Akt interplay in oxidative stress. (**a**) Effects of oxidative stress in the absence of PARP inhibition. Oxidative stress causes DNA strand breaks that activate nuclear PARP. Excess PARP activity exhausts the substrate NAD<sup>+</sup> that impairs ATP production. Attempted restoration of NAD+ levels leads to ATP depletion and eventually cell death. ATM-NEMO complexes, even if formed, are retained in the nucleus, therefore, Akt activated cytoprotective mechanisms are inactive. The oxidative stress is free to damage the mitochondria resulting in the release of pro-apoptotic intermembrane proteins. (**b**) Effects of oxidative stress in the presence of PARP inhibition. The PARP inhibitor blocks excess PARP activity, thereby prevents NAD<sup>+</sup> and consequent ATP depletion. Activated ATM-NEMO complex is able to translocate to the cytoplasm and associate with the mitochondrial outer membrane. Akt and mTOR are recruited to form the ATM-NEMO-Akt-mTOR cytoprotective signalosome that prevents oxidative stress induced mitochondrial damage and activates various cell survival pathways. Pointed arrows denote activation while arrows with flat end represent inhibition. P indicates phosphorylation.

#### *2.3. Role of the PARP-Akt Interplay in Shock and Inflammation*

PARP is known to regulate a host of signal transduction pathways and PARP inhibitors have been shown to have the ability to suppress the production of various pro-inflammatory mediators (including cytokines and chemokines). These effects are thought to be important contributors to the beneficial effects of PARP inhibitors in various experimental models of local and systemic inflammation, as well as in various models of critical illness [52–59]. However, in addition to effects on pro-inflammatory signaling, PARP inhibitors' anti-inflammatory and beneficial effects in various models of inflammation and shock may also be related, at least in part, to activation of Akt. In 2003 we reported that the protective effects of the PARP inhibitor PJ34 in a model of endotoxic shock in mice (which included improved survival and protection against multiorgan dysfunction) were associated with Akt activation in various organs including the liver and the gut [60], and similar results were subsequently obtained with another PARP inhibitor, 4-hydroxyquinazoline [61]. Moreover, in a model of neuroinflammation and demyelination in mice (induced by the mitochondrial toxin cuprizone), the PARP inhibitor 4-hydroxy-quinazoline protected against many of the observed pathophysiological changes (weight loss, demyelination): these effects were associated with a modulation of Akt activation [62]. In this model, the effect of cuprizone alone was associated with an increase in Akt phosphorylation; however, Akt phosphorylation was further and markedly enhanced in the presence of the PARP inhibitor [62]. It is conceivable that Akt activation contributes to the beneficial effect of PARP inhibition in these models; however, this remains to be directly tested in future experiments.

Additional inflammatory conditions where PARP-inhibitor/Akt "connection" has been evaluated include various in vitro and/or in vivo models of arthritis and dermatitis. In an experiment by Garcia and colleagues utilizing CD95/Apo-1 (Fas)-induced apoptosis in fibroblast-like synoviocytes (FLS) from rheumatoid arthritis patients, PARP-1-deficient FLS were protected from cell death compared to wild-type cells. This was associated with the activation of the Akt-GSK survival pathway in the PARP1-deficient cells, but inhibition of PI3K/Akt pathway did not affect the difference between PARP1-competent or -deficient cells in terms of viability [63]. Thus—although in this experimental setting the same authors have demonstrated that Akt activation can, in fact confer cytoprotection [64]—in these experimental conditions, pathways other than Akt must be responsible for the beneficial effect of PARP1 deficiency. In contrast, in models of sulfur mustard induced dermatitis, the beneficial effects of the PARP inhibitor ABT-88, as well as the beneficial effects of PARP1 deficiency, appears to involve a significant modulation of the Akt pathway. In these experiments, sulfur mustard exposure decreased Akt phosphorylation in the control group, while in the PARP1/deficiency + sulfur mustard groups, a marked increase in Akt phosphorylation was observed [65].

#### *2.4. Role of the PARP-Akt Interplay in Reperfusion Injury*

In one of the earliest studies related to the subject of PARP-Akt interplay, PARP inhibitors of two different structural class (HO-3089 and 4-hydroxyquinazoline) were evaluated on energy metabolism, contractile function, and cell death of isolated rat hearts during ischemia-reperfusion. In line with earlier studies demonstrating the beneficial effects of PARP inhibitors (or genetic PARP1 deficiency) in myocardial ischemia-reperfusion [66–68], both of the tested PARP inhibitors significantly improved the recovery of high-energy phosphate intermediates in the heart [69], directly confirming the well-known PARP-activation-cellular energetic failure concept. Both of the tested PARP inhibitors also improved cardiac contractility [69]. In addition, ischemia-reperfusion induced a slight, but detectable phosphorylation of Akt at Ser437 (i.e., evidence of Akt activation). Importantly, treatment with either of the two tested PARP inhibitors induced a marked further enhancement of Akt activation [69].

Subsequent studies, utilizing similar experimental designs, confirmed and expanded these findings and directly tested the functional importance of Akt activation in the cardioprotective effect of these two PARP inhibitors (HO-3089 and 4-hydroxyquinazoline). Pharmacological inhibition of PI3-kinase by wortmannin or LY294002 reduced the PARP inhibitor-elicited increase of Akt phosphorylation during ischemia-reperfusion, and significantly diminished the PARP-inhibitor-mediated recovery of ATP and creatine phosphate in the reperfused hearts. Moreover, pharmacological inhibition of PI3-kinase also diminished the protective effect of PARP inhibitors on infarct size and the recovery of cardiac function in this model [70]. In the same study, the Akt downstream target GSK-3β was also evaluated; this effector was found to be activated by the PARP inhibitors during ischemia-reperfusion [70]. Interestingly, wortmannin or LY294002, on their own (i.e., in the absence of PARP inhibitor treatment) failed to increase infarct size or suppress cardiac function during ischemia-reperfusion, nor did they affect the basal level of Akt phosphorylation suggesting that (a) the basal and the PARP-inhibitor-induced Akt phosphorylation involves fundamentally different mechanisms; the former not being dependent on PI3-kinase, while the latter is PI3-kinase dependent and (b) the low basal level of Akt activation that occurs during ischemia-reperfusion is unable to significantly counteract cardiac reperfusion injury (as opposed to the PARP-inhibitor-induced activation of Akt, which is clearly cardioprotective) [70].

Similar findings were obtained with PARP inhibitors of various structural classes in hypoxic-reoxygenated cardiac myocytes (in vitro) and in ischemic-reperfused hearts (in vitro or in vivo) in several additional studies. For example, Palfi and colleagues evaluated the effect of the PARP inhibitor L-2286 in an ischemia-reperfusion model in Langendorff perfused rat hearts and in an isoproterenol-induced myocardial infarction model in rats. In addition to cardiac energy metabolism, oxidative damage and (in the in vivo experiment, infarct size), and the phosphorylation state of Akt and MAPK cascades were also monitored. As expected, L-2286 exerted a significant protective effect against ischemia-reperfusion-induced myocardial injury in both experimental models. In addition, in line with prior observations made with other PARP inhibitors, L-2286 enhanced the ischemia-reperfusion-induced activation of Akt. Moreover, it also induced an activation of extracellular signal-regulated kinase and p38-MAPK, while c-Jun N-terminal kinase activation was repressed [71]. Similarly, the PARP inhibitor DPQ's protective effect in another rat isolated heart model against ischemia-reperfusion was found to be associated with a marked increase in Akt phosphorylation [72]. Moreover, in H9c2 cardiomyocytes in vitro, the PARP inhibitor 5-AIQ enhanced the oxidative stress-induced activation (phosphorylation) of Akt (as well as of GSK-3β) and these actions conferred a cytoprotective effect; inhibiting the Akt/GSK-3β pathway by LY294002 significantly attenuated the cytoprotective effect of 5-AIQ [73] and similar effects were subsequently reported with the PARP inhibitor KR-33889 in the same cardiomyocyte cell line in vitro; the data presented in the paper show that presence of the PARP inhibitor extended the duration of Akt activation after hydrogen peroxide exposure [74].

The cardioprotective effect of PARP inhibitors goes beyond protection in ischemic or hypoxic cardiomyocytes or hearts, and includes storage-mediated injury, as well as various forms of chronic heart failure (including septic, diabetic, and cardiotoxic drug-induced) [75–85]. Several studies have evaluated whether this protective action in these models is also associated with (and/whether it is dependent on) Akt activation. In a model of cardiac dysfunction induced by cold storage (6 h of storage in Celsior solution), the presence of the PARP inhibitor INO-1153 in the storage fluid afforded a better subsequent functional recovery, and this effect was associated with an increased Akt activation in these hearts [86]. The cardioprotective effect of the PARP inhibitor was abrogated by wortmannin co-treatment [86]. Moreover, in a model of doxorubicin-induced heart failure, the protective effect of L-2286 was associated with an activation of Akt [87], and similar effects were noted with a PARP inhibitor of a different structural class (BPG-15) in a model of imatinib-induced cardiomyopathy [88]. In addition, in a model of "hyperglycemia" in vitro (elevated extracellular glucose exposure of H9c2 cells), siRNA-mediated PARP1 depletion was found to increase the levels of phosphorylated Akt (PARP1 depletion also provided significant cytoprotective effects) [89]. The beneficial cardiac effects of L-2286 discussed earlier were also found to extend to the post-infarction cardiac remodeling stage (8 weeks of follow-up was performed); however, at these later time points, the Akt activating effect was no longer detectable (in contrast, the protective effects were attributable to effects on various protein kinase C isoforms) [90]. However, in a model of hypertension-induced cardiac remodeling in the rat, the beneficial effects of the PARP inhibitor L-2286 against cardiac remodeling and vascular

dysfunction could be attributed, at least in part, to Akt activation (and in part to effects of MAP kinase pathways) [91,92].

It should be stressed that the phenomenon of Akt-activation-mediated cardioprotection is not at all unique to PARP inhibitors. The fact that Akt activation, in general, activates cytoprotective pathways is a well-known phenomenon (see prior sections); and specifically, the cardioprotective (or, more generally, organ-protective as it also extends to various organs including the brain, kidney, and others) effect of a wide set of therapeutic interventions, including ischemic pre- and post-conditioning, calcium antagonists, beta blockers, caffeine, erythropoietin, insulin, vitamin E, and many others has been attributed, partially or wholly, to a set of cellular effects (including protection against mitochondrial dysfunction, reduction of cellular reactive oxygen species generation, induction of cytoprotective, and antiapoptotic pathways, etc.) that are, wholly or partly, the consequences of Akt activation [93–96].

The phenomenon of PARP-inhibitor mediated Akt activation (and associated cytoprotective and organ-protective effects) has also been demonstrated in various non-cardiac experimental systems exposed to ischemia or hypoxia. For instance, the first-generation PARP inhibitor 3-aminobenzamide (3-AB) was found to exert beneficial effects that are (at least in part) dependent on Akt activation in a rodent model of ischemic stroke. In Sprague-Dawley rats subjected to middle cerebral artery occlusion, 3-AB increased and prolonged Akt activation in the brain, and this effect was associated with smaller infarct volume, lower brain water content, less TUNEL staining, less cleavage of caspases, and lower induction of various adhesion molecules [97]. Similarly, in an in vitro study, nicotinamide—which is a PARP inhibitor of low potency, but it also has a host of additional pharmacological actions [98]—exerted cytoprotective effects in an anoxia model in cultured rat hippocampal neurons; its beneficial effects were associated with Akt activation, and were attenuated by inhibition of Akt phosphorylation using pretreatment with wortmannin or LY294002 [99]. Application of wortmannin (500 nM) or LY294002 (10 μM) without anoxia was not toxic to the neurons, suggesting that basal Akt activity does not regulate cell viability or survival (in contrast to the PARP-inhibitor-inducible increase in Akt activity, which does) [99].

Acute urinary retention is a condition with a significant hypoxia-associated component. In a rat model of acute urinary retention induced by bladder distension, bladder apoptosis was reduced by inhibition of PARP activation with 3-aminobenzamide treatment. The PARP inhibitor also increased the levels of ATP and NAD+, phosphorylation of Akt, and Bcl-2/Bax ratio, and significantly reduced the activation of caspase 3 [100]. Since in this study it was not evaluated whether pharmacological inhibition of Akt activation counteracts the effects of 3-aminobenzamide, it is not possible to determine to what extent Akt activation contributed to the observed bladder-protective effects.

The cytoprotective effect of PARP inhibition (using 4-hydroxy-quinazoline) was also associated with Akt activation in an acute rat kidney rejection model [101]. Moreover, the cytoprotective effect of a clinically approved (for oncological indications) PARP inhibitor (olaparib) was also associated with Akt activation in hypoxia-reoxygenation induced acute retinal injury model [102]. However, similarly to the acute urinary retention experiments discussed above, the functional importance of Akt activation was not directly tested in these studies. In the latter model, olaparib treatment induced alterations in a host of additional pathways (MAPKs, HIF1α, Nrf2, and NFκB) [102]; thus, it is conceivable that these effects may have also contributed to its beneficial effects.

#### *2.5. Mechanism of PARP Inhibition Induced Akt Activation*

All aforementioned in vitro, ex vivo, and in vivo results indicate that inhibition of nuclear protein PARylation reaction catalyzed by PARP1 somehow triggers the activation of cytoplasmically localized Akt. Recently, we provided in vitro experimental evidences for a mechanism (Figure 2), which may account for the Akt activating effect of PARP1 inhibition under oxidative stress conditions, and is consistent with the experimental and clinical data on PARP inhibitors [12]. According to it, PARP1 activated by DNA strand breaks PARylates itself, thereby creating a scaffold that recruits ataxia telangiectasia mutated kinase (ATM) and NEMO [45]. Binding to PAR strands activates ATM that phosphorylates itself and NEMO [46]. If excess PARP activation is prevented by a pharmacological inhibitor or genetic manipulation, ATM is not PARylated [103], translocates to the cytoplasm in the form of a complex with NEMO, and at least in part associates with the mitochondrial outer membrane. There, the complex recruits Akt and mTOR, and together with them forms a signalosome that activates Akt, leading to activation of downstream pro-survival pathways (Figure 2) [12].

Several elements of this model have been described previously. Independent studies reported that ATM mediates Akt activation in cancer cell lines, in insulin treated myocytes, and under cellular stress conditions [104–106]. Defective ATM linked activation of Akt was found to be involved in diabetes and neurodegenerative diseases [106]. Furthermore, direct PARylation of ATM by PARP1 was indicated in independent reports [103,107]. PARP activity dependent formation and nuclear-to-cytosolic translocation of ATM-NEMO complex was suggested as the molecular mechanism underlying genotoxicity induced NFκB regulated apoptosis [45,108]. As for the model's remaining elements, we demonstrated an at least partial mitochondrial localization for the proposed ATM-NEMO-Akt-mTOR signalosome, and proved that PARP1 inhibitor's cytoprotective effect is abolished by silencing any components of the signalosome [12]. As a partial independent confirmation of latter results, PARP1 inhibition was shown to induce synthetic lethality in ATM deficient cells [109]. However, by showing that continuously active Akt rescued these cells [12], we demonstrated that the said synthetic lethality is predominantly resulted from insufficient Akt activation in ATM suppressed cells. These data emphasize the crucial importance of Akt activation in PARP1 inhibition's cytoprotective property that can lead to undermining PARP1 inhibition's cytostatic effect on breast cancer gene (BRCA) mutations carrying cancers [110]. Independent reports support this notion by demonstrating the increase of PARP1 inhibitors' cytotoxicity by Akt inhibitors [8,11,111].

#### **3. PARP-Akt Interactions in Cancer Biology**

Based on PARP1's role in DNA repair, early prognoses for the role of its inhibitors in cancer therapy were to enhance antitumor activity of radio- or chemotherapy [112]. Although these prognoses have proved to be correct for combination therapy, present use of PARP inhibitors is mostly based on their synthetic lethality with homologous recombination repair deficiency of DNA double strand breaks [113,114]. PARP1 binds to DNA strand breaks, and is trapped on the DNA when inactivated by the inhibitor causing replication fork collapse [115]. The resulting DNA double strand breaks are repaired by the error prone non-homologous end joining repair system when homologous recombination is defective due to mutation in its key elements, such as BRCA1 and BRCA2 [113,114]. Both mechanisms are deleterious for homologous recombination deficient cancer cells. Accordingly, in 2014, the United States Food and Drug Administration (FDA) approved the first PARP inhibitor, olaparib monotherapy for treating germline BRCA1/2 mutated, advanced stage ovarian cancer [116]. Since that time, three additional PARP inhibitors were approved and the therapy was extended to other types of BRCA deficient cancers (Table 1). Clinical trials expanded the use of PARP inhibitor monotherapy to cancers with defects other than BRCA mutation in the double-strand repair pathway and other forms of genomic instability, since these cancers seem to be more dependent on PARP1 to maintain genomic integrity [117].

Combination therapy uses PARP inhibitors together with radiotherapy, immune therapy, a cytotoxic agent, an angiogenesis blocker, a signaling pathway inhibitor, or a blocker of the DNA damage-response pathway [118]. In these therapeutic strategies, the PARP inhibitor sensitizes the cancer cells for the cytostatic effect of the agent(s) used in combination with it by limiting DNA damage repair. Alternatively, a targeted treatment, such as a signaling pathway inhibitor, could relieve the resistance of tumor cells toward the PARP inhibitor [119]. Up to now, no combination therapy with PARP inhibition was approved by FDA for the clinical practice, although a number of ongoing clinical trials utilize this strategy for various types of cancers [118].


**Table 1.** Food and drug administration (FDA) approved inhibitors of enzymes relevant for the PARP-Akt interplay.

Solid tumors grow in hypoxic condition and undergo metabolic reprograming to provide energy and nutrients for proliferation and survival. Accordingly, increased anaerobic glycolysis, modified oxidative phosphorylation, and mitochondrial remodeling are hallmarks of cancer cells' phenotype [120]. This metabolic reprograming is often accompanied by an elevated production of reactive oxygen and nitrogen species (RONS) that can directly or indirectly activate Akt [121]. Additionally, various genetic events lead to activation of this pro-survival pathway. Activating mutations in PI3K, mTOR, and the Akt isoforms, and loss-of-function mutations and deletions in PTEN occur in about 27%, 8%, 3%, and 19% of all cancer cases, respectively [122]. PIK3CA encodes the p110α catalytic subunit of PI3K, and is the most frequently altered oncogene in human cancers, including endometrial, ovarian, colorectal, and breast cancers [123]. Mutations of mTOR and Akt are much less frequent, although, they were demonstrated in melanoma, renal carcinoma, bladder tumor, lung cancer, breast cancers, head and neck squamous cell carcinomas, and endometrial cancer [124]. PTEN loss represent the second most mutated tumor suppressor gene frequently occurring in various human cancers including colorectal cancer, breast cancer, glioblastoma, endometrial cancer, malignant melanoma, and prostate cancer [125]. Although their mutation frequency is quite low, gene amplification and overexpression of Akt isoforms occur more frequently. Akt1 gene amplification was found in breast, colon, pancreatic, gastric, esophageal, ovarian, and thyroid cancers, and glioblastoma while amplification of Akt2 occurred in pancreatic, ovarian, and breast cancers [126–128]. Akt2 overexpression was present in 40% of hepatocellular carcinoma and 57% in colorectal cancers, while Akt3 overexpression was found in prostate and breast cancers [126–128]. Furthermore, Akt was hyperactivated in many human cancers due to activating mutations of upstream elements such as PI3K and RAS, or loss of function mutations of p27, PTEN [129,130], and inositol polyphosphate-4-phosphatase2 [131]. An additional activating mechanism, N6-methyladenosine mRNA methylation of PTEN, leads to Akt activation in many cancers [132]. Certainly, Akt activation can also be triggered by upstream signaling elements, such as growth factors or cross-talking signaling pathways [130], or via the retrograde activating mTOR-PI3K feedback loop [133]. However, PI3K/Akt pathway's oxidative regulation via glutathion, peroxiredoxins, glutaredoxins, and oxidative inactivation of PTEN remains a significant regulatory factor of Akt mediated malignant transformation [134,135]. Although inhibitors of the pathway are attractive agents in the cancer therapy, so far no Akt inhibitor has been approved by FDA. On the other hand, FDA approved four PI3K and three mTOR inhibitors (Table 1) for clinical cancer therapy, and two Akt inhibitors are in phase III clinical trials for combination therapy of breast and colorectal cancers [124].

One of the most important limitations of PARP inhibition based cytostatic therapy is the resistance of cancer cells against either PARP inhibition monotherapy or combination therapy. Various mechanisms can account for this resistance, including down-regulation of PARP1 expression, reduced cellular availability of the inhibitor, reactivation of the homologous recombination DNA repair pathway, exploiting altered cell cycle regulation, and changing the miRNA environment [136]. Additionally, activation of the PI3K-Akt pathway by PARP inhibition, a mechanism originally suggested for reducing side effects of platinum compounds and alkylating agents [10], represent a significant limitation of PARP inhibitor utilizing chemotherapy by protecting mitochondrial integrity and function, and mitigating apoptosis in the cancer cells (Figure 2). To support this notion, activation of Akt increased paclitaxel resistance in vitro [11], and negative synergism was found between the PARP inhibitor PJ-34 and cisplatin or temozolomide during a short-term combination treatment of B16F10 metastasizing melanoma cell line [137]. Additionally, in clinical trials, inhibitors of the P13K/AKT pathway potentiated cytostatic effect of PARP inhibitors in a combination therapy, indicating the clinical relevance of the concept [8,111].

#### **4. Open Questions and Future Directions**

At present, using PARP inhibitors in homologous recombination deficient malignancies is a well-established practice. Judging from the increasing number of clinical trials, extension of the strategy of synthetic lethality to more types of cancers can be expected. There are a number of ongoing clinical trials aiming at improving the efficacy of PARP inhibitors by combining them with platinum, taxane or alkylating chemotherapy, radiotherapy, and inhibitors of signaling pathways [138]. They intend to exploit synergistic effects of the agents or to overcome cytostatic resistance of the cancer cells. In the latter respect, ataxia telangiectasia and Rad3-related protein, mTOR, and NFκB pathway inhibitors were found to be effective [119,139,140].

Another line of improving therapeutic efficacy of PARP inhibition addresses removal of PAR polymers. Importance of the process is highlighted by the finding that complete PARG deficiency results in embryonic lethality [141]. Furthermore, PARG inhibition together with homologous recombination deficiency results in synthetic lethality [142], the same way as PARP1 inhibition does. Mutation of TARG1, the other enzyme that together with PARG removes PAR polymers from the target proteins, was indicated in a neurodegenerative and seizure disorder, and its deficiency sensitizes the cells to DNA alkylating agents [143]. These data suggest that removal of PAR polymers could be essential in DNA double strand break repair, and the PARG and TARG enzymes may represent targets for cancer therapy. More research and specific inhibitors are needed to elucidate the particulars concerning these enzymes.

Recent clinical trials aim to expand therapeutic application of PARP inhibitors to cancers of intact homologous recombination repair. To this end, the PARP inhibitors are proposed to be applied in combination with immune checkpoint blockade. The cancer cells are frequently resistant against immune checkpoint blockade that could be alleviated by the PARP inhibitor via promoting cross-presentation and modifying immune microenvironment. The PARP inhibitors are expected to increase T cell's tumor-killing efficacy, and to activate the cancer-immunity cycle thereby increasing the sensitivity of tumor cells toward immune checkpoint blockade [144]. However, considering the diversity of the involved mechanisms, extensive research is needed to elucidate the possible interference by various networks including the PI3K-Akt cytoprotective signaling pathway.

With the advancement of whole exome sequencing, screening of cancer-related genetic aberrations is possible at a reasonable price. The technique provides genome-wide, and high-throughput results that can be used to identify synthetic lethal pairs involving epigenetic-related synthetic lethal genes. For the development of novel rational combination therapies, enhanced genetic-interaction screens are needed that assumes integration of functional interaction data with orthogonal methods [145]. Additionally, heterogeneity of the patient population limits therapy development, and questions optimality of the applied treatment, a problem which can be amended by determining microsatellite instability and tumor mutation burden [146]. All these techniques, as they become affordable, pave the way toward the realm of personalized medicine [147].

#### **5. Conclusions**

Accumulating clinical evidence indicates that PARP inhibitors can be successfully applied in cancers not responding to conventional treatments. To date, therapeutic application of the FDA approved PARP inhibitors utilize homologous recombination repair deficiency in these tumors. At the same time, ongoing clinical trials combining PARP inhibitors with agents targeting non-homologous recombination DNA repair systems, signaling pathways, angiogenesis, or immune checkpoint mechanisms intend to extend the therapeutic potential of PARP inhibitors well beyond their present one. However, because of complexity and redundancy of the mechanisms that regulate PARP activity, we still fail to comprehend fully the processes leading to de novo and acquired resistance toward PARP inhibitor therapy. As in vitro and in vivo experimental evidences indicate, the cytoprotective PI3K-Akt pathway is activated by PARP1 inhibition, and via multiple mechanisms, its activation induce apoptosis resistance and mitochondrial protection that limit cytostatic efficacy of PARP inhibitor monoor combination therapy. Because of the importance of the PI3K-Akt pathway in all branches of medicine associated with oxidative stress including shock, inflammation, cardiac, and cerebral reperfusion injury besides cancer, extensive research is going on to fully comprehend its mechanisms and its interplay with other signaling pathways including those regulated by PARP inhibition. The availability of FDA approved PI3K-Akt pathway inhibitors should certainly facilitate these research efforts. Additionally, recent advancements in the field of next generation sequencing and bioinformatics could contribute to improving treatment strategies and outcomes of PARP inhibitor therapy.

**Author Contributions:** All authors have read and agreed to the published version of the manuscript. Conceptualization, S.B.; writing—original draft preparation, F.G.J. and C.S; writing—review and editing, F.G.J.; funding acquisition, F.G.J., B.S. and C.S.

**Funding:** This research was funded by Hungarian grants GINOP-2.3.3-15-2016-00025, GINOP-2.3.2-15-2016-00049, EFOP-3.6.1-16-2016-00004, and Higher Education Institutional Excellence Program to F.G.J. and B.S. and by the Swiss National Foundation (SNF 310030\_188525) to C.S.

**Acknowledgments:** We apologize to those authors whose study we could not cite due to space restriction reasons.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **ARH1 in Health and Disease**

#### **Hiroko Ishiwata-Endo, Jiro Kato, Linda A. Stevens and Joel Moss \***

Pulmonary Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD 20892-1590, USA; endohiro@mail.nih.gov (H.I.-E.); katoj@nhlbi.nih.gov (J.K.); stevensl@nhlbi.nih.gov (L.A.S.)

**\*** Correspondence: mossj@nhlbi.nih.gov; Tel.: +301-496-1597

Received: 27 December 2019; Accepted: 15 February 2020; Published: 19 February 2020

**Abstract:** Arginine-specific mono-adenosine diphosphate (ADP)-ribosylation is a nicotinamide adenine dinucleotide (NAD)+-dependent, reversible post-translational modification involving the transfer of an ADP-ribose from NAD<sup>+</sup> by bacterial toxins and eukaryotic ADP-ribosyltransferases (ARTs) to arginine on an acceptor protein or peptide. ADP-ribosylarginine hydrolase 1 (ARH1) catalyzes the cleavage of the ADP-ribose-arginine bond, regenerating (arginine)protein. Arginine-specific mono-ADP-ribosylation catalyzed by bacterial toxins was first identified as a mechanism of disease pathogenesis. Cholera toxin ADP-ribosylates and activates the α subunit of Gαs, a guanine nucleotide-binding protein that stimulates adenylyl cyclase activity, increasing cyclic adenosine monophosphate (cAMP), and resulting in fluid and electrolyte loss. Arginine-specific mono-ADP-ribosylation in mammalian cells has potential roles in membrane repair, immunity, and cancer. In mammalian tissues, ARH1 is a cytosolic protein that is ubiquitously expressed. ARH1 deficiency increased tumorigenesis in a gender-specific manner. In the myocardium, in response to cellular injury, an arginine-specific mono-ADP-ribosylation cycle, involving ART1 and ARH1, regulated the level and cellular distribution of ADP-ribosylated tripartite motif-containing protein 72 (TRIM72). Confirmed substrates of ARH1 in vivo are Gαs and TRIM72, however, more than a thousand proteins, ADP-ribosylated on arginine, have been identified by proteomic analysis. This review summarizes the current understanding of the properties of ARH1, e.g., bacterial toxin action, myocardial membrane repair following injury, and tumorigenesis.

**Keywords:** arginine-specific mono-ADP-ribosylation; bacterial toxin; cholera toxin; ART1; ARH1; tumorigenesis; loss of heterozygosity; membrane repair; gender bias

#### **1. ARH Family**

#### *1.1. Properties of ARHs*

ADP-ribosylarginine hydrolase 1 (ARH1) is a member of an ADP-ribosyl-acceptor hydrolase (ARH) family, which is composed of three 39-kDa proteins, ARH1-3, based on sequence and size similarity. These hydrolases differ in enzymatic activities and biological functions [1]. Human ARH1 is a 357 amino acid (aa) protein, which shares significant sequence and size conservation with ARH2 (354 aa protein, 47% identity, 68% similarity) and ARH3 (363 aa protein, 22% identity, 41% similarity). The mouse ARH1 amino acid sequence shows 83% identity and 91% similarity to human ARH1 [2]. ARH1 requires Mg2<sup>+</sup> for catalytic activity [1,3,4]. The active sites of ARH1 (human, mouse, or rat) contain aspartates residues for coordination of Mg2<sup>+</sup> binding [5,6]. Critical residues involved in ARH1 enzymatic activity and binding of ADP-ribose include the conserved vicinal aspartates 55 and 56 in humans [5,6] and 60 and 61 in rats [7]. Replacement of Asp 55 and/or Asp56 in humans and Asp60 and/or Asp61 in rats with alanine, glutamine, or asparagine significantly reduced hydrolase activity (ADP-ribosylarginine hydrolase, α-NADase) [5–7].

#### *1.2. Function and Substrates of ARHs*

Mono- and poly-ADP-ribosylation are reversible post-translational modifications of proteins, DNA and RNA [8,9]. The ARTs in mammals, i.e., ART1, ART2, ART5, are confirmed mono-ADP-ribosyltransferases that modify only arginine residues [10–13]. Target residues of the poly (ADP-ribose) polymerase (PARP) family include aspartate, glutamate, and serine [14–16]. Hydrolases, such as the ARHs and macrodomain proteins, also show amino acid-specific hydrolytic reactions. Mono-ADP-ribosylation on arginine, serine, and glutamate, respectively, was hydrolyzed by ARH1, ARH3, and macrodomain proteins, i.e., MacroD1, MacroD2, and terminal ADP-ribose protein glycohydrolase (TARG1)/C6orf130 [5,17]. Poly(ADP-ribose) glycohydrolase (PARG) cleaves poly(ADP-ribose) (PAR) chains of PARP-1 but did not cleave terminal ADP-ribose linked directly to amino acids of PARP-1. ARH3 releases ADP-ribose from serine on PARP-1 [18,19], but did not hydrolyze ADP-ribose-glutamate/aspartate linkages [16,20]. ADP-ribosylated DNA and RNA were hydrolyzed by ARH3, but not ARH1 [5,8,9]. Furthermore, ARH1 and ARH3 hydrolyzed the α-*O*-glycosidic bonds of the poly-ADP-ribose polymer attached to PARP1, although ARH3 has approximately 50 times more activity than ARH1 [1,21,22]. In addition, ARH1 and ARH3 hydrolyzed *O*-acetyl-ADP-ribose (*O*AADPr), the product of sirtuin deacetylases [23]. Recent reports showed that ARH2 is primarily expressed in the heart and appears to be involved in the regulation of heart chamber outgrowth [24]; although ARH2 has significant amino acid identity and homology to ARH1, it has not been shown to be an active hydrolase [2].

*Arh1*-deficient mouse embryonic fibroblasts (MEFs) and mouse tissues are unable to hydrolyze the ADP-ribose-arginine bond, supporting the view that ARH1 is the only mammalian arginine-specific hydrolase [25,26]. ARH1 has a role in controlling TRIM72 ADP-ribosylation, and thereby TRIM72 activity in membrane repair: Membrane injury may cause the mixing of intracellular with extracellular factors, e.g., small molecule such as NAD+, and proteins [27]. Another opportunity for ARH1 interaction with ART1 substrates might be through lipid raft-mediated endocytosis, which is enhanced by UV light and H2O2-induced stress [28]. Interestingly, glycosylphosphatidylinositol (GPI)-anchored ARTs and caveoline-3, a cofactor of TRIM72, exist within lipid rafts [10,29]. To address the question of how cytoplasmic protein ARH1 contributes to the reversion of arginine-ADP-ribosylation catalyzed by extracellular enzyme ARTs, localization of arginine ADP-ribosylated proteins needs to be determined.

Arginine-specific ADP-ribosylation has been identified in the endoplasmic reticulum (ER) protein binding immunoglobulin protein (BIP) also known as 78-kDa glucose-regulated protein (GRP78) [30], mitochondrial protein heat shock protein 75 (HSP75) also known as tumor necrosis factor receptor-associated protein 1TRAP1 [31], secreted protein tumor necrosis factor-α [32], and external plasma membrane protein integrin α7 [31], although, none of them has been identified as a substrate of ARH1. Under stress conditions, BIP/GRP78 and HSP75/TPAP1 are released to the cell surface and cytoplasm, respectively [33,34]. By this mechanism, cytoplasmic protein ARH1 might be able to interact with its target proteins. However, there is still the question as to how ecto-enzyme ARTs modify intracellular proteins. Further studies are needed to understand (1) whether extracellularly modified substrates can be internalized into cytoplasm and cleaved by ARH1, and (2) whether an intracellular transferase catalyzes arginine ADP-ribosylation.

Not all arginine ADP-ribosylated proteins have been reported to be hydrolyzed by ARH1. For example, human neutrophil peptide-1 (HNP-1) is ADP-ribosylated by ART1. HNP-1 is an antibacterial peptide contributing to host defense immunity, as well as being toxic for host epithelial cells. HNP-1 ADP-ribosylated on arginines 14 and 24 shows reduced antimicrobial and cytotoxic activities [35]. Non-enzymatic replacement of the ADP-ribosylated arginines of HNP-1 with ornithine resulted in a peptide with less cytotoxicity than unmodified HNP-1 but with retention of its antibacterial activity [36]. Thus, the ornithine-containing HNP-1, which is no longer a substrate for ARH1, may have therapeutic potential as an ARH1-resistant molecule.

Proteomic analysis has revealed that more than 1000 proteins are ADP-ribosylated on arginine residues by ARTs [31,37–39], the confirmed substrate proteins of ARH1 in vivo are only ADP-ribose-Gαs [26] and ADP-ribose-TRIM72 [27]. The amount and duration of cellular ADP-ribosylation may be controlled by substrate-specific transferases and hydrolases, which catalyze opposing arms of ADP-ribosylation cycles. Both ADP-ribosylation and de-ADP-ribosylation alter protein activities involved in cell signaling and viability [10,27,40].

#### **2. Structure and Enzymatic Activity of ARH1**

#### *2.1. Structure and Stereospecific Activity of ARH1*

ARH1 hydrolyzes the *N*-glycosidic linkage of ADP-ribosyl-arginine (e.g., arginine-Gαs, arginine-TRIM72) [10,26,27] and also cleaves the *O*-glycosidic linkage of poly-ADP-ribose, *O*-acetyl-ADP-ribose, and α-NAD<sup>+</sup> [1,21–23]. The crystal structure of *h*ARH1 is available in a complex with ADP-ribose [5,6]. ARH1 requires Mg2<sup>+</sup> for maximal catalytic activity [1,4,7]. The rat, mouse, and turkey ARH1 activities were stimulated by Mg2<sup>+</sup> and dithiothreitol (DTT), whereas pig and calf ARH1 showed Mg2+, but not DTT dependence. Human ARH1 was DTT independent, as well [3,25]. The difference in DTT dependence appears to reside in the number of cysteine residues in the ARH1 proteins. ARH1 has a high structural similarity with ARH3 [6]. One major difference between ARH1 and ARH3 is the binding of the adenosine ribose moiety of ADP-ribose, resulting in a difference in ARH1 binding for ADP-ribose, which is 70× less than ARH3 [6]. ARH1 has four phosphorylation sites, tyrosine (Tyr)-4, Tyr-19, Tyr-20, and Tyr-205 [41]. Phosphorylation of ARH1 leads to conformational changes of the catalytic pocket, facilitating ADP-ribose-arginine binding to ARH1, which may affect ARH1 hydrolytic activity [41]. We speculate that the different affinities for ADP-ribose between ARH1 and ARH3 might be due to phosphorylation at the ADP-ribose binding pocket. ARH1 phosphorylation, at Tyr-4 and Tyr-19, was identified in a highly metastatic hepatocellular carcinoma (HCC) cell line, MHCC97H, but not in a nonmetastatic HCC cell line, Hep3B, implying that tyrosine phosphorylation of ARH1 was associated with HCC metastasis [42].

Mammalian ART1 catalyzes the transfer of ADP-ribose from β-NAD<sup>+</sup> to proteins in a stereospecific manner, forming α-ADP-ribosylated-(arginine) proteins, which alters the protein's function in cellular biological processes [43]. Arginine-specific ADP-ribosylation is a reversible modification, however, in vitro, α-ADP-ribosyl-arginine anomerizes to the β-form, which is not hydrolyzed by ARH1. Anomerization of α- to β-ADP-ribosyl-arginine interrupts a stereospecific ADP-ribosylation cycle in vitro [17,44–46]. Mono-ADP-ribosylated proteins are de-modified by the stereospecific α-ADP-ribose acceptor hydrolases, e.g., ADP-ribosyl-acceptor hydrolases (ARHs), MacroD1, MacroD2, Af1521, TARG1/C6orf130 [5,17]. ARH1 is the only ADP-ribosylated arginine-specific hydrolase identified in mammals that cleaves the *N*-glycosidic bond at the C1 linkage of α-anomeric ADP-ribose-arginine [4,17,25]. Currently, ARH1 has four stereospecific hydrolytic activities using as substrates α-NAD, α-*O*AADPr, poly(ADP-ribose), and α-ADP-ribose-(arginine) protein and generating free ADP-ribose as a product.

#### *2.2. ARH1 Protein Expression and Cellular Distribution*

Among rat tissues, ADP-ribosylarginine-specific hydrolase activity was greatest in the brain, spleen, and testis [25]. ARH1 is a cytosolic protein, the product of a single gene that is ubiquitously expressed in mammalian tissues [25]. Similar to the finding with glycosylphosphatidylinositol (GPI)-linked ART1 protein expression [47], the amount of ARH1 protein increased during C2C12 myoblast differentiation into myotubes, indicating that it may play a role in myoblast differentiation [27]. ARH1 protein levels in lung adenocarcinoma and lymphoma in *Arh1*+/<sup>−</sup> mice were lower than detectable levels by Western blotting. However, ARH1 existed in surrounding *Arh1*+/<sup>−</sup> nontumorous lung tissue [40], suggesting that the loss of ARH1 activity enhanced tumor formation. These data are consistent with a role for inactivation or loss of the functioning *Arh1* gene or protein in the mouse tumorigenesis model. According to the human cancer database Oncomine (www.oncomine.org) [48], *ARH1* mRNA expression in human lung adenocarcinoma was significantly lower than in that of normal lung tissue [40,49], consistent with a tumor-suppressor function of ARH1.

#### *2.3. Arginine-Specific Mono-ADP-Ribosylation Cycle*

Arginine-specific ADP-ribosyltransferases and ADP-ribosylarginine hydrolase 1 (ARH1) are opposing arms of a mono-ADP-ribosylation cycle [50]. ADP-ribosylation of arginine was first discovered as a mechanism of action of bacterial toxins, e.g., *vibrio cholerae* cholera toxin, *Escherichia coli* (*E. coli*) heat-labile enterotoxin, *pseudomonas aeruginosa* exoenzyme S (ExoS), which catalyze the NAD+-dependent disruption of the signal transduction pathway by ADP-ribosylation of critical proteins, e.g., G protein alpha subunit that is stimulating for adenylyl cyclase (Gαs), rat sarcoma viral oncogene homolog (Ras), and Ras-related protein in brain (Rab) [51–54]. Cholera toxin produces arginine ADP-ribosylated Gαs, resulting in intoxication, which is terminated by ARH1 cleaving ADP-ribose from ADP-ribosylated Gαs. Thus, arginine-specific mono-ADP-ribosylation is a reversible reaction.

Other possible substrates of ARH1 might be Ras and Rab. It has been reported that ExoS catalyzes arginine ADP-ribosylation of Ras and Rab inhibiting nerve growth factor-stimulated neurite formation of PC-12 cells and disrupting normal vesicle trafficking, respectively, however, it has not known whether ARH1 cleaves ADP-ribose from ADP-ribosylated Ras and Rab [53,54].

In mammalian cells, endogenous ADP-ribosyltransferases (ARTs), extracellular GPI-anchored ART1 and ART2, and secreted ART5, catalyze arginine-specific ADP-ribosylation similar to those of the bacterial toxins, e.g., cholera toxin, *E. coli* toxin [10–13]. Ecto-ART proteins show tissue-specific expression such as heart and skeletal muscle for ART1, lymphocytes for mouse ART2, and testis for ART5. In humans and chimpanzees, however, ART2 is a pseudogene [10]. Furthermore, ART5 is primarily an NAD<sup>+</sup> glycohydrolase; NADase activity of ART5 is 10x higher than its ADP-ribosyltransferase activity [10,55]. Therefore, GPI-linked ART1 seems to be a primary contributor to arginine-specific ADP-ribosylation in humans [35,56]. Under normal conditions, there is no difference in NAD<sup>+</sup> levels of lung, heart, and brain between wild-type and *Arh1*-deficient mice, suggesting that ARH1 does not consume NAD<sup>+</sup> [49]. However, ARH1 decreased the levels of ADP-ribose-(arginine) content [27]. Arginine-specific ADP-ribosyltransferase activity in mouse heart did not differ between wild-type and *Arh1*-deficient mice, suggesting that the accumulation of ADP-ribosylarginine content is due to ARH1 deficiency in *Arh1*−/<sup>−</sup> mice [27]. These data imply that cardiomyocytes are undergoing an arginine-specific ADP-ribosylation cycle in vivo. It is not surprising that this hypothesis raises questions regarding how an extracellular protein, GPI-linked ART1, catalyzes ADP-ribosylation in the extracellular space where the NAD<sup>+</sup> concentration is 0.1 μM [11,13,57] and how cytoplasmic protein ARH1 hydrolyzes ADP-ribosyated proteins synthesized by an extracellular enzyme ART1. There is evidence that cellular NAD<sup>+</sup> may be released into the extracellular matrix during inflammation and under pathological circumstances where cells may be killed by ischemic stress, thus providing substrate for the ADP-ribosyltransferases [58–60]. Additional evidence suggests that serum TRIM72 is released from injured or dead cells and is detectable following muscle injury induced by cardiac ischemia-reperfusion and treadmill exercise [61,62]. Furthermore, ART1, ARH1, caveolin-3, and cytoplasmic membrane repair protein TRIM72 were detected in macromolecule complexes [27]. Altogether, cytoplasmic ARH1 may leak with TRIM72 and NAD<sup>+</sup> into the extracellular space where ART1 resides. In the last step of the cycle, the release of ADP-ribose from ADP-ribosylated TRIM72 by ARH1 promotes oligomerization of TRIM72 and recruitment of TRIM72 to the site of injury [27]. Thus, ART1-TRIM72-ARH1 appears to constitute an ADP-ribosylation cycle.

#### **3. Functions of ARH1 in Disease**

#### *3.1. Defense Mechanism against the Action of Cholera Toxin*

According to the World Health Organization (WHO), each year, there are still about 1.3 million to 4 million cases of cholera and 21,000 to 143,000 deaths worldwide [63]. Cholera toxin produced by *Vibrio cholerae* consists of a catalytic A-subunit, which dissociates from its B-subunits in the ER and catalyzes ADP-ribosylation of the α subunit of the intestinal Gs protein (Gαs) [64,65]. ADP-ribosylated Gαs at arginine 187 is incapable of hydrolyzing GTP and remains in an active state, resulting in stimulation of adenylyl cyclase (AC) and increased cyclic AMP (cAMP) formation. In this model, increased cAMP activates the cystic fibrosis transmembrane conductance regulator (CFTR) chloride channel, which leads to a loss of Cl−, Na+, and water in the intestinal lumen, causing the devastating diarrhea characteristic of cholera [26,64,66,67].

Cholera toxin ADP-ribosylates arginine moieties in a number of proteins, e.g., unidentified 18-, 98-, and 200-kDa proteins, however, Gαs including Gαs-S and Gαs-L appears to be the predominant protein that is ADP-ribosylated on arginine by cholera toxin [68]. The amount of ADP-ribosyl Gαs in the presence of cholera toxin was greater in *Arh1*-deficient mice, which suggests that, in wild-type mice, ARH1 cleaves ADP-ribose from Gαs, thereby generating unmodified Gαs and reducing fluid accumulation caused by cholera toxin [26]. Indeed, *Arh1*-deficient mice were more sensitive to cholera toxin-stimulated fluid accumulation in intestinal loops than wild-type mice [69]. In addition, the ARH1-based host defense mechanism occurs in a gender-specific manner. Female *Arh1*-deficient mice were more sensitive to the cholera toxin than were male mice [69]. Male and female wild-type mice, however, did not show a difference in cholera toxin sensitivity. The knockout mice but not the wild-type mice data supported the finding that women had a higher prevalence of cholera than men [70]. In addition to differences in ARH1 in humans and mice, these gender effects in humans may result from other factors such as societal norms (e.g., domestic responsibility for caring of the sick, time spent at home, and accessibility to health care), rather than biological differences in reaction to cholera and/or cholera toxin.

#### *3.2. Tumor-Suppressor Function of ARH1*

#### 3.2.1. Increased Tumor Formation in *Arh1*-Deficient and *Arh1*-Heterozygous Mice

Increased tumorigenesis was seen in *Arh1*-deficient and *Arh1*-heterozygous mice [2,40]. During a 24-months observation period, 20.5% (32 out of 156 mice) of *Arh1*-deficient and 11% (19 out of 169 mice) of *Arh1*-heterozygous mice showed increased frequency and extent of tumors in multiple organs, e.g., adenocarcinoma in lung, uterus, and mammary gland; hepatocellular carcinoma; hepatic and gastrointestinal lymphoma; hemangiosarcoma [40]. Tumors between *Arh1*−/<sup>−</sup> and *Arh1*+/<sup>−</sup> mice differed in the age of appearance with *Arh1*−/<sup>−</sup> and *Arh1*+/<sup>−</sup> mice, showing tumors at 3 months and 6 months, respectively [40]. Consistent with increased in vivo tumorigenesis, mouse embryonic fibroblasts (MEFs) generated from *Arh1* knockout and heterozygous mice showed increased cell proliferation and tumor formation in nude mice compared to wild-type MEFs [40]. Furthermore, *Arh1*-knockout MEFs transformed with an inactive double-mutant (D60, 61A) *Arh1* gene [7,40] did not rescue the *Arh1* knockout MEFs and showed increased cell proliferation as well as tumor formation in nude mice [2]. In agreement, overexpression of active ARH1 protein in *Arh1*-deficient MEFs partially reversed the tendency to develop tumors [2]. Other *Arh1*+/<sup>−</sup> MEFs that developed tumors in nude mice showed loss of heterozygosity of the remaining *Arh1* gene. Tumorigenic MEFs with *Arh1* gene heterozygosity showed a mutation in the remaining allele and expressed a low level of ARH1 activity [2]. These data are consistent with a tumor-suppressor function of ARH1.

The transmembrane ecto-enzyme CD38 functions as a NAD glycohydrolase and an ADP-ribosyl cyclase and is an NAD+-dependent oncogene [71]. Consistent with this hypothesis, CD38 was overexpressed in 41% (11 out of 27 human lung tumor samples) of tumor cells [49]. The anti-CD38 monoclonal antibody, daratumumab (DARA), is an approved treatment for patients with multiple myeloma [72]. CD38 activities were inhibited by ADP-ribosylation on arginine [71]; it is not known whether ADP-ribosylated CD38 is a substrate of ARH1. Deletion of the *Cd38* gene reduced tumor formation in both *Arh1*-deficient and wild-type mice [49], with significant reductions in the incidence of lymphomas, adenocarcinoma, and hemangio/histolytic sarcomas [49]. Knockout of CD38 in A549 human adenocarcinoma cells inhibited anchorage-independent cell growth, cell invasion, and xenograft growth in nude mice [49]. In contrast, *Arh1*-deficiency in MEFs affected cell cycle progression, resulting in increased cell proliferation [40]. These data suggest that ARH1 affects

the cell cycle, preventing tumor formation, rather than control cell migration, as is the case with CD38-mediated metastasis [22,49]. In addition, estrogen promoted the survival rate of *Arh1*-deficient MEFs in the murine circulation and increased tumor metastasis to the lung [73]. As described above, increased tumor formation was dependent on the loss of ARH1 activity. Thus, ARH1 plays a role in cell proliferation in response to modifiers of tumorigenesis, e.g., CD38, estrogen.

#### 3.2.2. ARH1 Heterozygosity and Tumorigenesis

As noted earlier, a 24-month observation of ARH1 littermates from birth revealed that *Arh1*-deficient mice showed a 1.8× higher incidence of tumor formation than *Arh1*-heterozygous mice. However, between the ages of 24 and 33 months, the frequency of tumors seen in *Arh1*-deficient and heterozygous mice was similar, 31% and 28%, respectively. This age-dependent increased occurrence of malignancy in *Arh1*-heterozygous mice resulted in a loss of heterozygosity (LOH) and an absence or mutation of the *Arh1* gene. Mutation of the good allele in the ARH1 heterozygous mice resulted in an ARH1 protein whose activity was between 4% to 55% of the wild-type ARH1 [2,40]. *Arh1* gene mutation in MEFs and *Arh1*+/<sup>−</sup> heterozygous mice tended to be in exons 2 and 3 that is comparable to the human ARH1 catalytic site in exons 3 and 4 [2,40]. In the human cancer database, LOH of the *ARH1* gene was identified in the lung (15%) and kidney (18%) [2]. According to the human somatic tumor mutation database, human *ARH1* gene mutations observed in cancer were also located in the human ARH1 catalytic region that corresponds to the mutation sites in mouse tumors [2]. Based on these findings, ARH1 in the murine model appears to be applicable to human cancer studies. Together, these data support the hypothesis that ARH1 is a tumor-suppressor gene that participates in the pathogenesis of both human and mouse cancers.

#### *3.3. Membrane Repair Function of ARH1*

*Arh1*-deficient 8-month-old mice developed cardiomyopathy with myocardial fibrosis. Cardiac fibrosis occurred in a gender-specific manner with fibrosis in *Arh1*-knockout male mice being 10× greater than in female mice [27]. Cardiac fibrosis is characterized by increased collagen type I deposition due to aging or as a result of injury, e.g., myocardial infarction, hypertensive heart disease, idiopathic dilated cardiomyopathy, and diabetic hypertrophic cardiomyopathy [74]. In contrast, regardless of sex, during dobutamine-induced stress, *Arh1*-knockout mice showed significantly lower ejection fraction and fractional shortening than wild-type mice, consistent with systolic dysfunction [27]. The membrane repair protein TRIM72 was identified as a substrate for ARH1 and ART1 [27]. TRIM72 had been described as an essential molecule of the membrane repair process, recruiting intracellular vesicles to sites of membrane disruption [75,76]. Cytoplasmic protein TRIM72 leaks from injured cardiac tissue into serum [61], thus serum TRIM72 may be a potential biomarker of acute cardiac injury. TRIM72 was ADP-ribosylated on arginines 207 and 260 [37,39]. The endogenous ADP-ribosylated TRIM72 level was elevated in *Arh1*-deficient mice following cardiac ischemia-reperfusion injury.

In rat cardiac myocytes, greater than 80% of cellular NAD<sup>+</sup> is in mitochondria [77]. Indeed, NAD<sup>+</sup> release from mitochondria in cytosol protects myocytes from post-ischemic reperfusion injury [77]. The importance of ARH1, ART1, and TRIM72 ADP-ribosylation cycle in plasma membrane repair and wound healing was demonstrated using a laser injury model and scratch wound-healing assay in C2C12 myotubes after stable transformation with TRIM72, ARH1, ART1, and ARH1 plus ART1 shRNA, and transient transformation with wild-type TRIM72-GFP and double mutant TRIM72 (R207K, R260K)-GFP that is not ADP-ribosylated [27]. In addition, heterogeneous complexes containing TRIM72 with components of a reversible ADP-ribosylation cycle included ART1, ARH1, caveolin-3 [27,75,78,79]. Notably, the mono-ADP-ribosyltransferase inhibitors vitamin K1 and novobiocin, as well as the loss of ARH1 activity, inhibited the oligomerization of TRIM72, the essential mechanism by which TRIM72 is recruited to the site of membrane injury [27]. Taken together, the arginine mono-ADP-ribosylation cycle controlled by ART1 and ARH1 is fundamental to the oligomerization of TRIM72 during the membrane repair process in cardiomyocytes (Figure 1).

**Figure 1.** Tripartite motif-containing protein 72 (TRIM72) adenosine diphosphate (ADP)-ribosylation cycle in membrane repair. Ischemia-reperfusion-induced membrane disruption increased the ADP-ribosylation of TRIM72 by ADP-ribosyltransferase (ART) 1 at the sites of membrane damage, facilitating binding of TRIM72 and caveolin-3 to the membrane. Oligomerization of TRIM72 is essential for acute membrane repair and involves the recruitment of TRIM72 and intracellular vesicles at the injury sites [78]. ADP-ribosylarginine hydrolase (ARH) 1 catalyzes de-ADP-ribosylation of modified TRIM72, cleaving the ADP-ribose from ADP-ribosylated (arginine)TRIM72, promoting oligomerization of TRIM72 at the sites of injury.

#### **4. Proteomics**

Reanalysis of phosphoproteomic data identified arginine mono-ADP-ribosylation sites on 79 proteins [37]. Using a peptide-based enrichment strategy, 830 proteins containing ADP-ribosylated arginine were identified in mouse liver after H2O2-induced oxidative stress [31]. Arginine was the major (86%) ADP-ribosylated amino acid in mouse liver lysed in RIPA buffer with 40 μM PJ-34 and 1 μM adenosine diphosphate (hydroxymethyl) pyrrolidinediol (ADP-HPD), a specific PARP inhibitor (IC50 120 nM) [6,31]. Following overexpression of murine arginine-specific ADP-ribosyltransferase 2 (ART2) in microglia, 33 arginine ADP-ribosylated proteins were identified [38]. Most recently, 354 proteins ADP-ribosylated on arginine were identified from wild-type mouse heart, mouse skeletal muscle, and C2C12 myotubes (Figure 2) [39]. Six ADP-ribosylated proteins overlapping with mouse heart, mouse skeletal muscle, and C2C12 myotubes were identified (Figure 2), e.g., Golgi apparatus protein 1 (Glg1), basement membrane-specific heparan sulfate proteoglycan core protein (Hspg2), integrin α-7, nidogen-1, protein disulfide-isomerase A3, and TRIM72 (Table 1). HNP-1, integrin α-7, and TRIM72 represent the only previously reported ART1 target proteins [5,27,36,80,81]. Among these 354 proteins, only 6 proteins were identified in *Art1*-deficient mouse skeletal muscle (Table 2) [39]. These data support the hypothesis that ART1 is the main contributor to skeletal muscle and heart arginine ADP-ribosylation.

Together, recent advances in proteomic analysis uncovered more than 1000 proteins ADP-ribosylated on arginine residues [31,32,37–39]. ADP-ribosylome data, analyzed by the gene ontology, suggested that ADP-ribosylated proteins of the wild-type mouse heart, but not *Art1*-deficient mouse heart (i.e., ART1-independent ADP-ribosylated proteins), are involved in the regulation of muscle contraction and apoptotic processes [39]. Further analysis by STRING-based protein-protein interaction analysis identified arginine ADP-ribosylated protein interaction networks that are involved in stress response, wounding response, and regulation of the heart rate [39]. Thus, arginine-specific ADP-ribosylation cycle controlled by ART1 and ARH1 is important in muscle physiology and pathophysiology.

**Figure 2.** Venn diagram of proteins ADP-ribosylated on arginine in wild-type mouse heart, skeletal muscle, and C2C12 myotubes. Data are modified from Leutert et al. [39].

**Table 1.** Proteins adenosine diphosphate (ADP)-ribosylated on arginine residues in C2C12 myotubes, mouse skeletal muscle, and mouse heart. Data derived from Leutert et al. [39].


**Table 2.** Characterization of ADP-ribosylation by ADP-ribosyltransferase (ART)1 of mouse heart and skeletal muscle proteins. *Art1*-deficient mouse heart and skeletal muscle contained very few proteins ADP-ribosylated on arginine residues compared to the wild-type, which suggested that ART1 is a major contributor to ADP-ribosylation in skeletal muscle and heart. Data derived from Figure 3B from Leutert et al. [39].


#### **5. Conclusions**

Increasing evidence from knockout mouse models, where the ADP-ribosylation cycle is disrupted, shows the importance of arginine-specific ADP-ribosyltion cycles in disease, e.g., cancer [2,40], bacterial toxin-mediated infection [26,69], cardiomyopathy with myocardial fibrosis [27], and muscle weakness [39]. Data are consistent with ART1 and ARH1 serving as opposing arms of an arginine-specific ADP-ribosylation cycle. The COSMIC database analysis of human somatic mutations in cancer revealed 32 *ARH1* mutations in human lung, breast, and colon cancers, overlapping with the mutations found in *Arh1*-heterozygous mice after 6 months of age [2], demonstrating that ARH1 is an age-related cancer risk factor. In addition, *Arh1* deficiency resulted in gender-biased phenotypes. *Arh1*-deficient mice showed a female-biased increase in tumorigenicity and susceptibility to cholera [26,49,82]. In contrast, cardiomyopathy with myocardial fibrosis was seen in male more than female *Arh1*-deficient mice [27]. Recently, advances in mass spectrometry-based proteomics identified numerous arginine ADP-ribosylated proteins as well as the location of their modification sites in vitro. The role of the modification on function has been demonstrated for a limited number of proteins such as Gαs, TRIM72, and HNP-1, probably due to difficulties in reproducing mono-ADP-ribosylation, in vivo. Finding substrates of ARH1 can lead to studies of physiologically and pathologically relevant conditions for deciphering ARH1 functions in health and disease.

**Author Contributions:** Writing—original draft preparation, H.I.-E.; Writing—review and editing, H.I.-E., L.A.S., J.K. and J.M.; Supervision—J.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Intramural Research Program of the National Institutes of Health, National Heart, Lung, and Blood Institute.

**Conflicts of Interest:** All authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **PARP Inhibitors as Therapeutics: Beyond Modulation of PARylation**

**Ahrum Min 1,2 and Seock-Ah Im 1,2,3,4,\***


Received: 30 December 2019; Accepted: 5 February 2020; Published: 8 February 2020

**Abstract:** Poly (ADP-ribose) polymerase (PARP) 1 is an essential molecule in DNA damage response by sensing DNA damage and docking DNA repair proteins on the damaged DNA site through a type of posttranslational modification, poly (ADP-Ribosyl)ation (PARylation). PARP inhibitors, which inhibit PARylation through competitively binding to NAD+ binding site of PARP1 and PARP2, have improved clinical benefits for BRCA mutated tumors, leading to their accelerated clinical application. However, the antitumor activities of PARP inhibitors in clinical development are different, due to PARP trapping activity beyond blocking PARylation reactions. In this review, we comprehensively address the current state of knowledge regarding the mechanisms of action of PARP inhibitors. We will also discuss the different effects of PARP inhibitors in combination with cytotoxic chemotherapeutic agents regarding the mechanism of regulating PARylation.

**Keywords:** PARP; PARP inhibitors; PARylation; trapping; cancer therapeutic strategy

#### **1. Introduction**

All cells have more than tens of thousands of events that damage DNA in multiple ways, ranging from single base mismatches, bulky adducts in DNA bases, intra- and inter-strand DNA crosslinks, to single- and double-strand breaks (SSBs and DSBs) [1,2]. This DNA damage threatens genomic stability. There are DNA damage responses (DDRs), the sophisticated mechanisms of genome protection in cells, that function to activate the cell cycle checkpoint pathway to maintain genome stability by stopping or delaying the cell cycle during DNA damage or unstable DNA replication to allow the repair of damaged DNA lesions. DDRs activate transcription of a repair molecule or pro-apoptotic molecule to cause overexpression of the related molecule. DDRs activate mechanisms to remove uncontrolled damaged cells and to repair DNA damage from apoptosis caused by chromatid instability [3]. Regulation of the DDR pathway is induced by post-translational modification (PTM); poly ADP-Ribosylation (PARylation) is the pivotal PTM that occurs rapidly at the damage site during DDR [4,5].

PARylation is the reaction of transferring ADP-ribose residues to target substrates by ADP-ribosyl transferase using NAD+. It rapidly recognizes multiple types of DNA damage, including SSBs, and is recruited to the damaged site to induce the recruitment of DDR molecules so that the poly (ADP-Ribose) polymerase (PARP), specifically PARP1, PARP2, PARP3, PARP5a, and PARP5b, which are known as the major molecules of DDR, performs poly (ADP-ribose) (PAR) synthesis in humans [6]. PARP1 was proposed as a new treatment for cancer, as the synthetic lethality concept suggested that its depletion in breast-cancer patients with germline mutations in the *BRCA1* or *BRCA2* genes, key molecules in the homologous recombination (HR) pathway, could cause cancer cell death [7,8]. Since it was proven to be true, PARP inhibitors that inhibit DDR resulted in improved clinical benefits and became standard

therapy [9–11]. To date, four PARP inhibitors have been approved by the FDA and are being applied clinically. However, while all PARP inhibitors inhibit PARP catalytic activities, they have different cytotoxicities. Therefore, the anti-tumor effects of the PARP inhibitors have been suggested to be due to PARP trapping, as well as the inhibition of the enzymatic activities [12,13].

The catalytic inhibition and trapping effects of PARP are tightly regulated, and the cytotoxicity of each mechanism can cause different reactivities. Therefore, in this review, based on mechanisms of PARP, we intend to examine the difference of anti-tumor effect of the PARP inhibitors and the current aspect of the roles in combination treatment.

#### **2. PARPs and PARylation**

Poly (ADP-ribose) polymerase (PARP) is a family of 17 proteins in mammals, encoded by different genes, but with a conserved catalytic domain. Other than the catalytic domain, PARP family members contain one or more other motifs or domains, including zinc fingers, a breast cancer-susceptibility protein (BRCA) C-terminus-like (BRCT) motifs, ankyrin repeats, macro domains, and WWE domains [14] (Figure 1A). PARP1 was the first family member identified and has a critical role in SSB repair through the metabolism of recruiting and dissociating repair proteins by PARylation. In addition to DNA damage repair, PARP1 has important roles in a various range of cellular processes from cell proliferation to cell death, due to having diverse substrates like nuclear proteins involved in transcriptional regulation, apoptotic cell death, chromatin decondensation, inflammation, and cell cycle regulation [15,16]. PARP1 has a total molecular weight of 113 kDa and contains seven independent domains (Figure 1B) [5,17]. The N-terminus is the DNA binding domain (residues 1-353), which contains three zinc-finger DNA-binding domains, ZnFI, ZnFII, and ZnFIII, which are responsible for recognizing sites of damaged DNA and binding through allosteric activation. In the N-terminus there is a nuclear localization sequence (NLS) that places PARP1 in the nucleus with the KRK-X(11)-KKKSKK sequence. Between residues 211 and 214, there is a DEVD site that is cleaved by caspase into fragments of 23 and 89 KDa during apoptosis [18]. Residues 373 to 662 are the auto-modification domain consists of BRCA C-terminus-like (BRCT) domain serving sites of auto-ADP ribosylation and functioning in protein-protein interaction, and a WGR domain which roles in activating DNA damage repair by interaction with ZnFI, ZnFII, and catalytic domain. The auto-modification domain is rich in glutamate and lysine residues and is the site of self-PARylation. Finally, the C-terminus (residues 662–1014) is the catalytic domain, and the (ADP-Ribosyl) transferase (ART) domain is a NAD+ acceptor site where the His-Try-Glu residues called ART signatures are preserved well [19–21]. The helical subdomain (HD), an auto-inhibitory domain in the C-terminus, inhibits the binding of PARP1 and β-nicotinamide adenine dinucleotide without binding to DNA. When PARP1 binds to the DNA damage site, the auto-inhibitory function of HD is removed. The activation of the catalytic activity of ART and the generation of PAR chains in the target protein lead to the recruitment of DNA repair molecules. Thereafter, PARP1 is dissociated from DNA by auto-PARylation of PARP1, resulting in DNA repair [22].

**Figure 1.** PARPs structure (**A**) The PARP family consists of 17 members, divided into five subgroups according to domain structure and function: DNA damage-dependent PARPs (PARP1, PARP2, and PARP3), tankyrases (tankyrase1/PARP5 and tankyrase2/PARP5b), CCCH-type PARPs (PARP7, PARP12, and PARP13), macro-PARPs [B-aggressive lymphoma 1 (BAL1)/PARP9, BAL2/PARP14, and BAL3/PARP15], and other PARPs (PARP4, PARP6, PARP8, PARP10, PARP11, and PARP16). The catalytic domain at the C-terminus is conserved in all members and contains additional zinc fingers, BRCA C-terminus-like (BRCT) motifs, ankyrin repeats, macro domains, and WWE domains. (**B**) The seven major domains of PARP1 include three zinc-finger domains in the DNA binding domain, the BRCT domain in the auto-modification domain, and the pADPr accepting WGR domain (W), located centrally. The C-terminus has two catalytic domains: ART and a helical domain (HD).

This series of reactions is caused by PARylation. While the catalytic domain is conserved in the PARP family, only PARP1/2/3/4/5a/5b activates PARylation by possessing the His-Tyr-Glu motif called the "ART signature" [5,15,23–26]. The role of PARP3 as an (ADP-ribosyl) transferase is controversial. PARP4 is the largest protein in the PARP family, and PARP5a and PARP5b, classified as tankyrase1/2, have a SAM (Sterile Alpha motif) domain that interacts between proteins with the ability to homo- and hetero-oligomerize PARP1 and 2. PARP1 transfers ADP-ribose residues from NAD+ to acidic amino acid residues such as glutamates (E), lysine (K), arginine (R), serine (S), and aspartate (D), forming the negative poly (ADP-ribose) (PAR) chain [5,24,26]. PARP1 is believed to perform more than 90% of total PARylation in response to DNA damage. As soon as DNA damage occurs, ADP-ribosylation is covalently bound to the carbonyl group of the acidic residues of the target protein via ester bonds. PARP then forms a PAR chain by cleaving the glycosidic bond between nicotinamide and ribose of NAD+ by catalytic activity and binding ADP-ribosylation to the target protein via a 2 ,1"-O-glycosidic bond [22,27]. PARylation in the DNA damage repair pathway plays a role throughout DNA strand breaks repair through rapid DNA repair molecules recruitment to the DNA damage site, DNA damage signal transduction, causing apoptosis and protein degradation. Typically, the BRCT domain of X-ray repair cross-complementing protein 1 (XRCC1) binds directly to the PAR chain to be recruited to the DNA damage site. Upon XRCC1 binding to DNA damage sites, the PAR formation is increased by sequestering poly (ADP-ribose) glycohydrolase (PARG) from the interaction with PARP1 and PARG, resulting in causing dissociation of PARP1 from DNA damage site and increasing repair signal transduction [26–28]. In addition, PARP1 has been reported to promote DNA repair by interacting with DNA glycosylase 8-oxoguanine glycosylase 1 (OGG1), XRCC1, DNA polymerase (DNAP) β, DNA ligase III, proliferating cell nuclear antigen (PCNA), aprataxin, and condensin I involved in BER and SSBR through PARylation of PARP1 [29,30].

#### **3. Clinical Development of PARP Inhibitors**

PARP inhibitors are nicotinamid analogs that inhibit PARylation through competitively binding to the NAD+ binding sites of PARP1 and PARP2. As a cancer treatment drug, olaparib is first defined as the HR deficient tumor treatment, and it is fully approved by FDA for serous ovarian cancer and breast cancer treatment with germline *BRCA1* or *BRCA2* mutations [9,31,32]. To date, four PARP inhibitors (olaparib, niraparib, rucaparib, and talazoparib) have been FDA approved, and veliparib is waiting for FDA approval with promising results of phase III trial showing significantly extended Progression-free survival (PFS) in combination with carboplatin plus paclitaxel in serous ovarian cancer, and the drugs are compared in Table 1.


**1.**Comparison of available PARP inhibitors in the clinics.

**Table** 


*Cancers* **2020**, *12*, 394

It has been reported that the PARP inhibitor olaparib causes cell death by synthetic lethality in BRCA-deficient breast, ovarian, and prostate cancer. In a phase 2 trial, olaparib maintenance treatment in platinum-sensitive ovarian cancer patients improved progression-free survival (PFS) by 7 months [46]. In the SOLO-2 study, median PFS increased from 5.5 to 19.1 months over placebo in patients with BRCA1/2 mutations [11]. The OlympiAD trial, which was the global phase 3 trial for metastatic HER2-negative breast cancer with germline *BRCA1* or *BRCA2* mutations, confirmed that olaparib increased PFS by 2.8 months to 7.0 months (hazard ratio 0.58; P<0.001) and received FDA approval [9]. Not only for BRCA-deficient tumors, in the phase 2 STUDY-19 showed olaparib maintenance prolonged PFS in BRCA wild-type relapsed, platinum-sensitive serous ovarian cancer patients, extending the FDA approval to olaparib maintenance treatment in platinum-sensitive patients regardless of the BRCA status [47,48]. Moreover, PARP inhibitors were applied in prostate and pancreatic cancer, as well beyond the breast and ovarian cancers. The developed genomics technology verified that about 20 percent of prostate cancers have defects in DNA repair genes, resulting in a good candidate for PARP inhibitors [49]. In the phase II TOPARP-A trial, olaparib showed an 88 percent response rate in metastatic castrate-resistant prostate cancer (mCRPC) with BRCA1/2, ATM, or PALB2 mutation [50]. The PFS of patients with DNA repair gene defects increased from 2.7 to 9.8 months, and overall survival was also extended from 7.5 to 13.8 months. Based on these results, the FDA granted olaparib in breakthrough status in prostate cancer treatment, and several clinical trials are conducted in prostate cancer using PARP inhibitors. In addition, the median PFS was increased from 3.8 to 7.4 months with olaparib maintenance in the phase 3 POLO trial for metastatic pancreatic cancer patients with germline BRCA mutations who were sensitive to the first-line platinum-based therapy [37]. Recently, the FDA approved olaparib plus bevacizumab for the maintenance treatment of advanced ovarian cancer patients, who showed a response to the first-line platinum-based chemotherapy regardless of BRCA mutation. It is based on phase III PAOLA-1 trial results, in which the addition of olaparib to bevacizumab improved PFS significantly, to 37.2 months compared with 17.7 months in the placebo group among Homologous recombination deficiency (HRD)-positive ovarian cancer patients (HR, 0.33; 95% CI, 0.25–0.45) [51].

Rucaparib received FDA approval by demonstrating efficacy in a phase 2 trial of relapsed platinum-sensitive ovarian cancer patients who had previously received at least two platinum-based chemotherapies [38]. In addition, based on the phase 3 ARIEL3 trial, it received FDA approval for maintenance treatment of recurrent epithelial ovarian, fallopian tube, or primary peritoneal cancer with a partial or complete response to platinum-based chemotherapy [40].

In the case of niraparib, a randomized and double-blind phase 3 trial in 553 platinum-sensitive, recurrent ovarian cancer patients showed an increase in PFS from 5.5 to 21.0 months relative to placebo in the presence of germline BRCA mutations [41]. The FDA approval was granted based on the phase 3 trial on patients with recurrent epithelial ovarian, fallopian tube, or primary peritoneal cancer who were in complete or partial response to platinum-based chemotherapy [41].

Talazoparib has been FDA-approved for its efficacy against cancers with germline *BRCA1* or *BRCA2* mutations and HER2-negative metastatic breast cancer [52]. In the phase 3 EMBRACA trial, upon which FDA approval was based, talazoparib significantly increased median PFS from 5.6 to 8.6 months compared to physician's choice standard-of-care chemotherapy. The objective response rate was more than doubled over that of the control arm (62.6% for talazoparib vs. 27.2% for chemotherapy [OR: 4.99 (95% CI: 2.9–8.8), *p* < 0.0001]) [53].

Veliparib is another potent inhibitor of PARP1 and PARP2 in the developmental stage. In the phase III VELIA trial, which involved adding veliparib to first-line induction chemotherapy with carboplatin and paclitaxel followed by maintenance monotherapy in serous ovarian cancer increased median PFS from 17.3 to 23.5 months regardless of BRCA or HRD status [54]. Besides, veliparib 120mg bid plus carboplatin AUC6 and paclitaxel 200 mg/m2 treatment showed improved PFS from 4.2 to 5.8 months and OS from 8.4 to 10.3 months in phase II trial of Non-small-cell lung cancer (NSCLC) [55]. These results contributed that FDA grants orphan drug designation to veliparib for advanced NSCLC.

Veliparib, in addition to carboplatin and paclitaxel, also significantly improved PFS to patients with HER2-negative advanced or metastatic breast cancer with germline BRCA-mutation in a phase III trial. The rate of 3-year PFS was 26 percent on veliparib addition group when the placebo group showed 11 percent [45]. Although veliparib is not yet approved to FDA on any indications, these data could accelerate the application of veliparib in the clinics.

PARP inhibitors in the clinic showed improved clinical benefits, not only for tumors with *BRCA* mutations, but also for platinum-sensitive tumors caused by HRD, leading to their accelerated clinical application. However, the effects of these PARP inhibitors are difficult to understand as an inhibition of catalytic activity that simply inhibits PARylation. Although all four PARP inhibitors can inhibit the catalysis by PARP1 and PARP2, as shown in Table 1, each PARP inhibitor shows different cytotoxicity. PARP inhibition by PARP inhibitors induces a cytotoxicity far superior to the cytotoxicity induced by the knockout of PARP genes, suggesting that their antitumor effects are due to mechanisms other than the catalytic inhibition of PARP [12,13,34,56].

This difference can be conceptualized as PARP trapping: the ability of PARP inhibitors to trap PARP-DNA complexes while increasing the stability of the binding between PARP and DNA. As shown in Table 1, each PARP inhibitor has a different cytotoxicity that correlates with its PARP trapping activity. In other words, talazoparib, with the strongest PARP trapping effect, also is the most cytotoxic. Therefore, PARP trapping should be considered as a mechanism for the application of PARP inhibitors in clinical trials [34,57,58]. These differences in PARP trapping capacity may have different effects on combination therapy, as well as on monotherapy. Indeed, reactivities differ with different combination partners for each drug.

#### **4. Combination E**ff**ect of Conventional Chemotherapy according to the Mechanism of Action of PARP Inhibitors**

PARP inhibitors inhibit the catalytic activity of NAD+ depletion through competitive binding with NAD+, thereby inhibiting PARP itself as well as PARylation of target proteins in the nucleus. Not only do they cause cytotoxicity via irreparable damage caused by inhibiting repair protein recruitment to DNA damage sites, but also, they block cellular replication by inducing stalling or collapsing of the replication fork, as the PARP protein is continuously trapped in SSBs due to suppressed dissociation of PARP from DNA. This results in more deleterious DSBs, which in turn leads to cell death (Figure 2) [59–61].

PARP inhibitors were originally developed to sensitize tumors to the effects of DNA damaging agents, including ionizing radiation, temozolomide, and topotecan. In fact, PARP inhibitors have successfully sensitized tumors to radiation and to camptothecin, a topoisomerase I inhibitor [62,63]. However, the combination of gemcitabine, doxorubicin, and taxan showed no significant synergies [64,65]. It was observed that the efficacy of combination therapy with the same chemoagent was dependent on PARP inhibitors [66]. PARP inhibitors are thought to have different synergies in combinations with different chemoagents, depending on the mechanism and activity of cytotoxicity, as well as the regulation of PARylation. The combined effect of PARP inhibitors with different classes of chemoagents in the clinic is summarized in Table 2, in accordance with the mechanism of action of PARP inhibitors.

**Figure 2.** Mechanisms of PARylation and PARP inhibition in the DNA damage response. When a DNA single-strand break occurs, PARP quickly binds to the damage site using a zinc finger domain. It causes recruitment of DNA repair proteins to DNA damage sites by catalyzing PARylation between PARP and its target proteins XRCC1, DNA ligase III, etc., by the ART catalytic domain using NAD+ as a substrate. PARP auto-PARylation then decreases the affinity for DNA, resulting in dissociation from DNA so that the repair protein can bind. At this time, the PARP inhibitor binds to the pocket instead of NAD+, causing PARP to be trapped in the DNA. It encounters a replication fork, causing it to stall, and is converted into double-strand breaks (DSBs) leading to cell death; alternatively, it blocks the recruitment of repair proteins by blocking enzymatic activity where PARP PARylation occurs.

