**Structural Basis of Beneficial Design for E**ff**ective Nicotinamide Phosphoribosyltransferase Inhibitors**

**Sei-ichi Tanuma 1,2, \*, Kiyotaka Katsuragi 2 , Takahiro Oyama 3 , Atsushi Yoshimori 4 , Yuri Shibasaki 2 , Yasunobu Asawa 5 , Hiroaki Yamazaki 3 , Kosho Makino 6 , Miwa Okazawa 1 , Yoko Ogino 2,7 , Yoshimi Sakamoto 8 , Miyuki Nomura 8 , Akira Sato 2 , Hideaki Abe 3 , Hiroyuki Nakamura 5 , Hideyo Takahashi 6 , Nobuhiro Tanuma <sup>8</sup> and Fumiaki Uchiumi 7**


Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso Received: 22 July 2020; Accepted: 6 August 2020; Published: 10 August 2020

**Abstract:** Inhibition of nicotinamide phosphoribosyltransferase (NAMPT) is an attractive therapeutic strategy for targeting cancer metabolism. So far, many potent NAMPT inhibitors have been developed and shown to bind to two unique tunnel-shaped cavities existing adjacent to each active site of a NAMPT homodimer. However, cytotoxicities and resistances to NAMPT inhibitors have become apparent. Therefore, there remains an urgent need to develop effective and safe NAMPT inhibitors. Thus, we designed and synthesized two close structural analogues of NAMPT inhibitors, azaindole–piperidine (**3a**)- and azaindole–piperazine (**3b**)-motif compounds, which were modified from the well-known NAMPT inhibitor FK866 (**1**). Notably, **3a** displayed considerably stronger enzyme inhibitory activity and cellular potency than did **3b** and **1**. The main reason for this phenomenon was revealed to be due to apparent electronic repulsion between the replaced nitrogen atom (N1) of piperazine in **3b** and the Nδ atom of His191 in NAMPT by our in silico binding mode analyses. Indeed, **3b** had a lower binding affinity score than did **3a** and **1**, although these inhibitors took similar stable chair conformations in the tunnel region. Taken together, these observations indicate that the electrostatic enthalpy potential rather than entropy effects inside the tunnel cavity has a significant impact on the different binding affinity of **3a** from that of **3b** in the disparate enzymatic and cellular potencies. Thus, it is better to avoid or minimize interactions with His191 in designing further effective NAMPT inhibitors.

**Keywords:** nicotinamide phosphoribosyltransferase; NAD<sup>+</sup> biosynthesis; inhibitor; azacyclohexane; anticancer drug; drug design; enthalpy effect

### **1. Introduction**

Nicotinamide adenine dinucleotide (NAD+) is a critical molecule in control of numerous basic cellular processes, such as ATP production and maintenance of cellular integrity and genome stability [1–4]. It not only serves as a cofactor for redox bioreactions but can also be a substrate for multiple NAD+-consuming enzymes, such as poly(ADP-ribose) polymerases, mono-ADP-ribosyltransferases and sirtuins, participating in the epigenetic regulation of DNA transaction (transcription, replication, repair and recombination), cellular signaling processes and calcium homeostasis, and thereby in cell proliferation, differentiation and death [3–8]. Given the biological importance of this molecule, mammalian cells have evolved multiple biosynthetic pathways to produce NAD+; it is synthesized from tryptophan, nicotinic acid and nicotinamide (NAM)/nicotinamide riboside via *de novo*, Preiss–Handler and salvage pathways, respectively [1–4,9–11].

As NAD<sup>+</sup> is consumed through many enzymatic processes, accelerated NAD<sup>+</sup> depletion is often characteristic in cancer cells [9–12]. To rapidly replenish the NAD<sup>+</sup> pool, cancer cells rely heavily on the NAM salvage pathway that backs NAM to NAD<sup>+</sup> in two steps primarily catalyzed by nicotinamide phosphoribosyltransferase (NAMPT) [13–17]. Furthermore, in various types of cancer cells, NAMPT is found to be up-regulated [8,13,16,17], although the molecular mechanisms that dictate the salvage pathway choice remain elusive. Thus, NAMPT is considered an attractive target for the development of new anticancer drugs and therapies [18–31].

NAMPT, which functions as a homodimer with two unique tunnel-shaped cavities existing adjacent to each active site at the dimer interface, catalyzes the rate-limiting primary step of the salvage pathway, the transfer of phosphoribosyl residue from 5-phosphoribosyl-1-pyrophosphate (PRPP) to NAM to produce nicotinamide mononucleotide (NMN) (Figure 1) [1–4,32–34]. Many NAMPT inhibitors, such as FK866 (**1**), CSH-828, GNE0617 and STF-11880, which bind to the tunnel cavity of NAMPT, have been reported to exert anti-tumor suppression effects and have progressed to clinical trials [18,23–25,30]. However, cytotoxicities dominated by gastrointestinal symptoms and thrombocytopenia and resistance to NAMPT inhibitors have become apparent [23,35–37]. Therefore, there remains an urgent need to develop effective and safe anticancer NAMPT inhibitors for cancer chemotherapy. Thus, detailed insights into the tunnel structure of NAMPT are required to generate effective NAMPT inhibitors with a better therapeutic index and that can overcome resistance.

In the present study, we designed and synthesized two close structural analogues of NAMPT inhibitors **3a** and **3b**, which were modified from the best explored NAMPT inhibitor 1 [18,35]. Interestingly, we showed that **3a** has a considerably stronger enzyme inhibitory activity and cellular potency than does **3b** or **1**. Furthermore, using these inhibitors as chemical probes, we characterized the inhibitor-targeting tunnel cavity of NAMPT. Importantly, our in silico binding mode analyses of these inhibitors with the tunnel cavity of NAMPT revealed that there is apparent electronic repulsion only between the nitrogen atom of piperazine in **3b** and the Nδ atom of His191 in NAMPT. These findings indicate that the electrostatic enthalpy potential inside the tunnel region has a significant impact on the different binding affinity of **3a** from **3b** in the disparate cellular potencies. Thus, these results provide new insights into the design for further effective NAMPT inhibitors for cancer chemotherapies.

**Figure 1.** Overview of the human (h) nicotinamide phosphoribosyltransferase (NAMPT) as the dimer form and its enzyme reaction. The homodimer structure of hNAMPT (PDB Code: 2GVJ) [38–42] is shown in the ribbon diagram colored light blue and brown, respectively. The two tunnel-shaped cavities near the active sites in the hNAMPT molecule are shown in magenta shading. NAMPT catalyzes the conversion of nicotinamide (NAM) and phosphoribosyl pyrophosphate (PRPP) to produce nicotinamide mononucleotide (NMN).

### **2. Results and Discussion**

### *2.1. Synthesis and Biochemical Properties of 3a and 3b*

Many NAMPT inhibitors bind to the tunnel-shaped cavity of NAMPT (Figure 1), although the role of this tunnel in NAMPT function remains unknown [18–31]. These inhibitors have a unique pharmacophore consisting of three parts (Figure 2): head (a pyridine-like aromatic moiety), linker (a tunnel-interacting moiety) and tail (a solvent-exposed bulky group). To get more valuable insights into the tunnel cavity for the development of effective NAMPT inhibitors, we structure-basically designed and synthesized two close structural analogues of NAMPT inhibitors, namely azaindole–piperidine (**3a**)- and azaindole–piperazine (**3b**)-motif compounds, which were modified from the head and linker moieties of **1** (Figure 2, Scheme 1). Since the tail acts as a packing moiety against the solvent exposed tunnel exit surface, the tail structures of **3a** and **3b** were fixed as the same benzoyl group of **1**. The nitrogen-containing aromatic heads present in many NAMPT inhibitors mimic the natural NAM substrate of the enzyme [18–31]. However, the head moiety of vinylpyridine ring of **1** (Figure 2) is known to be able to non-specifically interact with proteins [43,44]. Based on this knowledge, we first replaced this head to the 5-azaindole heterocyclic ring to improve the specificity to NAMPT (Figure 2). Secondly, as the linker moiety of azacyclohexane may be considered to become a critical motif for NAMPT inhibitory activity, we synthesized a close structural set of piperidine (**3a**)- and piperazine (**3b**)-motif NAMPT inhibitors according to the methods of Bair et al. [45] and Vogel et al. [46], respectively (Scheme 1).

**Figure 2.** Chemical structures of **3a**, **3b** and **1**.

**Scheme 1.** Synthetic route of compounds **3a** and **3b**. Reagents and conditions: (**i**) NH2NH<sup>2</sup> , EtOH, r.t. to reflux, 2 h; (**ii**) 1*H*-pyrro[3,2-c]pyridine-2-carboxylic acid, EDC-HCl, DIPEA, DMF, r.t., 24 h. The targets **3a** and **3b** were prepared according to the methods of [45,46], respectively, with slight modifications.

μ μ μ μ μ μ To compare the NAMPT inhibitory activities of **3a**, **3b** and **1**, we determined the IC<sup>50</sup> values by the titration curves from the standard calorimetric enzyme-cycling assay [31,36,37]. As shown in Figure 3a, the inhibitory activity of **3a** (IC<sup>50</sup> = 0.11 µM) was approximately 5-fold stronger than that of **1** (IC<sup>50</sup> = 0.52 µM). This result suggests that the replacement of vinylpyridine with an azaindole ring could improve anti-NAMPT activity. The binding mode analyses of these inhibitors on the NAMPT molecule support this observation, as described in the next section; a hydrogen bond formed between the N1 atom of azaindole and Asp219 of NAMPT (Figure 4b). Surprisingly, the substitution of piperidine (**3a**) with piperazine (**3b**) resulted in a dramatic loss in inhibitory activity against NAMPT; the inhibitory activity of **3b** (IC<sup>50</sup> = 5.06 µM) was shown to be about 50-fold weaker than that of **3a**. The inhibition curve of **3a** was much steeper than that of **3b**, suggesting that incubation with **3a**

resulted in the formation of an entity capable of extremely tight binding to the tunnel cavity of NAMPT. From these observations, we suspect that unfavorable interactions between the piperazine moiety of **3b** and the tunnel cavity of NAMPT were responsible for the considerably weak anti-NAMPT activity.

error (SE) μ **Figure 3.** Biochemical properties of **3a** and **3b**. (**a**) The inhibitory activities of **3a** (open circle), **3b** (open triangle) and **1** (open square) against NAMPT enzyme were examined, as described previously [31,36,37]. NAMPT activity was measured using a coupled-enzyme reaction system (CycLex NAMPT colorimetric assay kit) in 96-well plate format using the one-step method. Data are the averages of three independent experiments, and the bars indicate the standard error (SE) values. (**b**) The anti-proliferative effects of **3a** (circle) and **3b** (triangle) on HCT116 cells were investigated as described previously [31,36,37]. HCT116 cells were treated with the indicated concentrations of each inhibitor with (closed symbol) or without (open symbol) 100 µM NMN for 72 h. The cell activity (% of control) was measured by use of the WST-8 assay. Data are presented as the means of three independent experiments ± SE, and the bars indicate the SE values.

To confirm the marked differences of anti-NAMPT activities between **3a** and **3b**, the anti-proliferation effects of these inhibitors on the human colon cancer cell line HCT116 were examined after continuous exposure for 72 h [31,36,37]. As shown in Figure 3b, compound **3b** displayed significantly weaker cell activity; the 50% effective concentration (EC50) value was calculated to be 87.5 nM, which was approximately 200-fold higher than that of **3a** (EC<sup>50</sup> = 0.48 nM). These results of the different inhibition degrees in cell-based assay from those in enzyme assay (Figure 3) paralleled observations by others that the degrees of potency of cell-based assessments were not always identical with those of NAMPT enzymatic inhibition [26,28,47]. In addition, there was little cytotoxicity on non-cancer cell lines, such as TIG-120, WI38 and MRC-5 cells. Both **3a** and **3b** caused more than 90% reduction of NAD<sup>+</sup> levels in the cancer cells after 24 h-treatment and thereby induced cell death. Furthermore, supplementation of the product of NAMPT, NMN, to the cell culture medium could significantly rescue HCT116 cells from cell death by treatment with **3a** or **3b** (Figure 3b). Additionally, the dose–response curves of **1** were similar to those of **3a**. These results confirmed that NAMPT is a target of both inhibitors. Collectively, these results indicate that the substitution of one carbon (C) atom of the piperidine moiety in **3a** for a nitrogen (N) atom (piperazine moiety in **3b**) exerts a critical effect on the anti-NAMPT activity.

### *2.2. Binding Modes of 3a and 3b to the Tunnel Cavity of NAMPT*

To understand the different NAMPT inhibitory activities of **3a** and **3b**, we analyzed their binding modes to NAMPT (PDB Code: 2GVJ [38–42]) by our in silico binding mode analyses using RDKit [48] and LigandScout [49]. The superimpositions of these inhibitors in the tunnel cavity revealed that they adopted almost identical chair binding forms as with the case in **1** (Figure 4a). The NAMPT residues that contacted these inhibitors were observed in very similar locations in the structure. These results

indicate that the binding modes of **3a** and **3b** in the tunnel cavity of NAMPT do not significantly alter as compared with that of **1**.

When observed closely, the position of the head regions of **3a** and **3b** in the tunnel cavity were nearly identical to that of **1** (Figure 4b). The subtle shifts observed in **3a** and **3b** from the position of **1** may have been introduced by their slightly different orientations of the head-azaindole bicyclic ring from the head-pyridine ring of **1** inside the tunnel cavity of NAMPT. The azaindole rings of **3a** and **3b** were sandwiched between Phe193A and Tyr18B, forming tight aromatic π–π stacks like the pyridine ring of **1**. Additionally, the nitrogen atoms (N1) in the azaindole rings of **3a** and **3b** could interact with Asp219 of NAMPT by a hydrogen bond (Figure 4b), thus suggesting that they might be involved in the tighter specific binding of **3a** and **3b** to NAMPT than that of **1**.

However, this specific contacts of the head-azaindole moiety with NAMPT could not explain the discrepancy in the considerably weaker inhibitory activity of **3b** than that of **3a** or **1**. Therefore, to better understand this phenomenon, we focused on the importance of linker-motifs for the potencies of NAMPT inhibitors. Thus, we analyzed closely the binding forms of the linker-azacyclohexane motifs of **3a** and **3b** in the tunnel cavity of NAMPT. The carbon atom (C4) of piperidine in **3a** and nitrogen atom (N1) of piperazine in **3b** were revealed to protrude approximately 1Å deeper into the NMA binding active site, as compared to the corresponding carbon atom of piperidine in **1**. Importantly, the electronic repulsion of the replaced nucleophilic N1 atom of piperazine in **3b** with the Nδ atom of the imidazole side chain of His191, located 3.82 Å away from the N1 atom, was revealed to occur (Figure 4c). In contrast, there was no electronic repulsion of the corresponding carbon atom of piperidine in **3a** with the Nδ atom as well as that of **1**. These results clearly show that the electronic repulsion interferes with the proper binding of **3b** to the tunnel cavity of NAMPT.

It is noteworthy that Zheng and colleagues have reported that the imidazole ring of His191 interacts with the linker aromatic group of GNE series inhibitors through a herringbone stacking [44,45,47]. Furthermore, NAMPT mutations, including H191R, D93del, Q388R, S165Y and G217V, have been reported to confer resistance to NAMPT inhibitors [50–52]. In addition, we have recently shown that an NAMPT mutation variant of HCT116 cells generated by continuously exposing cells to increasing concentrations of **1**, which confers cross-resistance to diverse NAMPT inhibitors, such as CHS-828, GNE-617 and STF-118804, was mapped to only H191R [36,37]. Thus, this His191 residue should be considered with higher priority in designing further effective NAMPT inhibitors. That is, it is preferable to avoid interactions with His191 in designing novel NAMPT inhibitors.

, x

**Figure 4.** Binding modes of **3a**, **3b** and **1** in the tunnel cavity of NAMPT. (**a**) Binding modes of **3a, 3b** and **1** in the tunnel cavity of NAMPT were analyzed by our in silico binding mode analysis using RDKit [48] and LigandScout [49]. (**b**) Close-up view of head regions of these three NAMPT inhibitors in the tunnel cavity of NAMPT using Open3DALIGN [53]. (**c**) Close-up view of the binding interactions of azacyclohexane-linker moieties of these NAMPT inhibitors with His191A located in the tunnel cavity of NAMPT. Carbons are colored green (**3a**), magenta (**3b**) and gray (**1**). Nitrogen and oxygen are colored blue and red, respectively.

### *2.3. Influence of Enthalpic E*ff*ects of 3a and 3b on the Interactions with Tunnel Cavity of NAMPT*

The C-to-N conversion in the azacylohexane-linker moiety was considered to be possible to introduce some entropic differences in the compounds' abilities to interact with the tunnel cavity of NAMPT. Although piperazine and piperidine rings are known to form various conformations from chairs to half-chair and skewed boat-to-boat forms in aqueous solution [54], piperazine-containing **3b** and piperidine-containing **3a** took nearly identical positions as a chair form to that occupied by **1** in the tunnel cavity analyzed by in silico binding modes and crystal structure analyses [38–42] (Figure 4). In addition, both C4 in **3a** (Figure 5a) and tertiary amine-type nitrogen (N1) containing **3b** (Figure 5b) might prefer equatorial conformations to axial ones by 1,3-diaxial hydrogen interaction [54], taking a stable chair conformation in the tunnel cavity of NAMPT. Therefore, such entropic effects in **3a** and **3b** in the tunnel space of NAMPT may not be so different from each other. Thus, it is possible to consider that the dramatic differences in enzymatic and cellular potencies between **3a** and **3b** may be due to enthalpic effects rather than entropic effects inside the tunnel cavity of NAMPT.

To prove this hypothesis, we attempted to analyze the binding affinity scores of **3a** and **3b** to NAMPT (PDB Code: 2GVJ) by our in silico binding mode analyses using RDKit [48] and LigandScout [49], as described in Materials and Methods. Remarkably, the binding affinity scores of **3a** and **3b** (non-protonation form) (Figure 5b) were calculated to be −29.07 and −22.13 kJ/mol, respectively, while that of **1** (−28.05 kJ/mol) was near to the value of **3b** (Table 1). Thus, the piperidine-motif was revealed to be beneficial for **3a** potency and has valuable influences on the interaction with the tunnel cavity of NAMPT.

The N1 and N4 atoms in **3a** and **3b**, respectively, would take an *sp*<sup>2</sup> -like bond, due to the effect of the proximity of the benzoyl group (R2) on the lone pair (Figure 5a,b). In contrast, although it is hard to know whether or not the N1 atom of piperazine in **3b**, which takes an *sp*<sup>3</sup> bond like the C4 atom in **3a**, would be protonated in the tunnel cavity of NAMPT, it is possible to be protonated in physiological conditions. Thus, the binding affinity score of **3b** in the protonation state was measured by LigandScout. Interestingly, it was calculated to be −23.85 kJ/mol and was not as good as that of the non-protonated form (Table 1). In the protonated N1 atom, the electrostatic interaction becomes attractive to the Nδ of His 191 but repulsive against the protonated Nδ. However, as the imidazole ring of His has aromaticity, the positive charge introduced by protonation does not tend to be localized on the Nδ. Thus, the repulsion may not be so strong. In addition, it seems that a hydrogen bond between the protonated Nδ of His191 and the N1 atom in **3b** could not be formed, because the angle that NδH-N1 (piperazine) takes is 106.4◦ . Furthermore, of course, the influence of the hydrophobic amino acids in the tunnel region on the protonated N1 in **3b** cannot be ruled out, and they may negatively influence this interaction. Accordingly, the electrostatic enthalpy effects were suggested to interfere with the proper binding of **3b** to the tunnel cavity of NAMPT. For this reason, both the non-protonated and protonated forms of **3b** are suggested to have lower binding affinities to the tunnel cavity of NAMPT than does **3a**. These outcomes indicate that the electrostatic potential (enthalpy term) inside the tunnel region of NAMPT has a significant impact on the different binding affinity of **3b** from that of **3a**.

**Figure 5.** Stable conformers of **3a** and **3b** in the tunnel cavity of NAMPT. The conformational states of **3a** (**a**) and **3b** with its protonation form (**b**) in the tunnel cavity of NAMPT were analyzed by in silico binding mode analysis and crystal structure of **1** (PDB code: 2GVJ) [38–42]. The hydrogen interactions of axial-methylene with 1,3-diaxial hydrogens are represented by dashed lines.

**Table 1.** Biochemical properties and binding affinity scores of **3a** and **3b** NAMPT inhibitors.


− <sup>a</sup> The IC<sup>50</sup> values of these three NAMPT inhibitors were determined by the titration curves in Figure 3a. <sup>b</sup> The binding affinity scores of these three inhibitors in the tunnel cavity of NAMPT were calculated by LigandScout [49].

### **3. Materials and Methods**

### *3.1. Enzymatic NAMPT Assay*

For the inhibitory effects of **3a** and **3b** on NAMPT activity, recombinant NAMPT activity was measured using a coupled-enzyme reaction system (CycLex NAMPT Colorimetric Assay Kit: MEDICAL & BIOLOGICAL LABORATORIES CO., LTD., Nagoya, Japan) according to the manufacturer's instructions, as described previously [31,36,37]. Briefly, NAMPT enzyme and tested compounds were put on the 96-well transparent plate. The mixture of the rest of contents (ATP, NAM, PRPP, nicotinamide nucleotide adenylyltransferase 1 (NMNAT1), water-soluble tetrazolium salts (WST-1), alcohol dehydrogenase, diaphorase and ethanol (EtOH)) were then added onto the wells. After a brief incubation, the absorbance of the samples was detected at 450 nm on every 10 min until 60 min using a SYNERGY HTX (BioTek Instruments, Inc.; Winooski, VT, USA). The inhibition rates were calculated by the initial slope of the absorbance of the tested samples divided by that of the control wells. The IC<sup>50</sup> values for the samples were calculated by 4-parameter logistic regression using Image J 1.48v (https://imagej.nih.gov/ij/).

### *3.2. Cells and Cell Culture*

HCT116 cell was obtained from American Type Culture Collection (ATCC) (Manassas, US) and cultured in Dulbecco's modified Eagle medium (DMEM) (FUJIFILM Wako Pure Chemical Corporation) supplemented with 10% of fetal bovine serum (Biosera Europe; Nuaillé, France), 100 Units/mL of penicillin G and 100 µg/mL of Streptomycin (FUJIFILM Wako Pure Chemical Corporation) [36,37].

### *3.3. Cell Activity Assay*

One thousand HCT116 cells were inoculated into the 96-well transparent cell culture plate and cultured for 24 h. The medium was changed to new medium that contained the tested compounds and was further cultivated for 72 h. Following the culture, 1/10 volume of the WST-8 reagent (FUJIFILM Wako Pure Chemical Corporation) was added and incubated for 1 h. After the incubation, the absorbance of the samples was detected at 450 nm using a SYNERGY HTX (BioTek Instruments, Inc.). Cell activity was expressed as percentage of control cells treated with vehicle alone [31,36,37].

### *3.4. In Silico Binding Mode and Binding A*ffi*nity Score Analyses*

The crystal structure of NAMPT was downloaded from the Protein Data Bank (PDB code: 2GVJ) for molecular binding mode analyses. In silico binding mode analyses were done using RDKit [48] and LigandScout [49]. First, 1000 conformations of each inhibitor were generated using the EmbedMultipleConfs and MMFFGetMoleculeForceField functions of RDKit. Second, the conformations of each inhibitor were aligned to the crystal structure of **1** using Open3DALIGN [53] implemented in RDKit. The conformation of each inhibitor, which has the highest alignment score, was selected. Third, energy minimization of the selected conformation of each inhibitor with NAMPT was performed, and then the binding affinity score of each inhibitor was calculated using LigandScout [49].

### *3.5. Synthetic Procedures of the Target Compounds*

### 3.5.1. General Experimental Methods

<sup>1</sup>H-NMR spectra were recorded at 400 MHz using a JEOL ECZ400 operating (JEOL Ltd.; Tokyo, Japan) at the indicated frequencies. Chemical shifts were expressed in ppm relative to the internal standard tetramethylsilane (ppm = 0.00). All reagents and solvents were purchased from FUJIFILM Wako Pure Chemical Corporation. The progress of all reactions was monitored by TLC using ethyl acetate/hexane, acetone/hexane, dichloromethane (CH2Cl2)/MeOH and CH2Cl2/MeOH/triethylamine (TEA) as the solvent system, and spots were visualized by irradiation with ultraviolet light (254 nm) or ninhydrin reaction. Column chromatography was performed using silica gel (200–300 mesh).

### 3.5.2. General Procedures for Synthesizing the Target Compounds **3a** and **3b**

Compounds 1-benzoylpiperidin-4-yl)butyl)isoindoline-1,3-dione (**1a**) and 2-(4-(4-benzoylpiperazin-1-yl)butyl)isoindoline-1,3-dione (**1b**) were synthesized according to Gilig et al. [55] with slight modifications.

### 3.5.3. (4-(4-aminobutyl)piperidin-1-yl)(phenyl)methanone (**2a**)

Hydrazine hydrate (0.3 mL, 6.17 mmol) was added under Ar<sup>2</sup> atmosphere to a solution of **1a** (1.0 g, 2.56 mmol) in EtOH at 20 ◦C. After stirring at 20 ◦C for 10 min, the mixture was heated under reflux for 2 h. After cooling to 20 ◦C, the white precipitate was filtered off and washed with EtOH (10 mL). The EtOH solutions were combined, and the solvent was evaporated in vacuo. The residue was taken in CH2Cl<sup>2</sup> (20 mL) and a saturated aqueous solution of K2CO<sup>3</sup> (20 mL). Vigorous stirring for 10 min produced two clear phases. The aqueous phase was extracted with CH2Cl<sup>2</sup> (10 mL, 3 times). The combined organic phases were washed with brine (30 mL) and dried over Na2SO4. After solvent evaporation in vacuo, **2a** was obtained (0.530 g) as a pale-yellow oil.

### 3.5.4. *N*-(4-(1-benzoylpiperidin-4-yl)butyl)-1*H*-pyrrolo[3,2-c]pyridine-2-carboxamide (**3a**)

First, 1*H*-pyrro[3,2-c]pyridine-2-carboxylic acid (0.149 g, 0.918 mmol) was added to a solution of **2a** obtained above (0.239 g, 0.918 mmol) in dry DMF (8.0 mL) in a 25 mL round-bottomed flask. HOBt (0.186 g, 1.38 mmol), EDC-HCl (0.352 g, 1.84 mmol) and N, *N*-diisopropylethylamine (0.462 mL, 2.75 mmol) were added sequentially, and the suspension was allowed to stir at room temperature overnight. The reaction mixture was then poured into a separatory funnel containing water and ethyl acetate. The aqueous phase was separated and extracted three times with EtOAc. The combined organic phase was dried over Na2SO4, filtered and concentrated in vacuo. The resulting residue was purified by column chromatography (CH2Cl2/MeOH = 9/1) to afford the target product **3a** (0.215 g, yield 58%, more than 95%, as judged by ODS(C18) HPLC) as white crystals. <sup>1</sup>H-NMR (DMSO-d6) d 11.94 (s, 1H), 8.88 (s, 1H), 8.59 (t, 1H), 8.19 (d, 1H), 7.39 (m, 3H), 7.31 (m, 3H), 7.20 (s, 1H), 4.43 (m, 1H), 3.51 (m, 1H), 3.26 (m, 2H), 2.96 (m, 1H), 2.70 (m, 1H), 1.70 (m, 2H), 1.51 (m, 3H), 1.29 (m, 4H), 1.05 (m, 2H).

### 3.5.5. 4-(4-Aminobutyl)piperazin-1-yl)(phenyl)methanone (**2b**)

Compound **2b** was obtained from **1b** by similar synthetic processes as described in Section 3.5.3.

### 3.5.6. *N*-(4-(4-Benzoylpiperazin-1-yl)butyl)-1*H*-pyroro[3,2-c]pyridine-2-carboxamide (**3b**)

The procedure was the same as described above for the synthesis of **3a**. Compound **3b** was obtained as pale-yellow powder (0.104 g, yield 28%, more than 95%, as judged by ODS(C18) HPLC). <sup>1</sup>H-NMR (DMSO-d6) δ 11.97 (s, broad, 1H), 8.90 (s, 1H), 8.61 (t, 1H), 8.19 (d, 1H), 7.33–7.43 (m, 5H), 7.23 (s, 1H), 3.26–3.58 (m, 9H), 2.30–2.38 (m, 5H), 1.48–1.55 (m, 3H).

### *3.6. Statistical Analysis*

All data were obtained from at least three independent experiments and were expressed as mean ± standard error (SE).

### **4. Conclusions**

In this study, to understand the structural basis of beneficial design for effective NAMPT inhibitors, we designed and synthesized a close structural analogue set of potent and specific azacyclohexane-motif NAMPT inhibitors, namely **3a** (piperidine-motif) and **3b** (piperazine-motif) modified from **1**. In addition, these two compounds have 5-azaindole ring as the head structure, which is replaced by the vinylpyridine ring of **1**. Through biochemical experiments and in silico binding mode analyses using these inhibitors, we clearly showed that **3a** is the most potent NAMPT inhibitor among these inhibitors, and that the azaindole-head and piperidine-linker of NAMPT inhibitors are promising

motifs that result in enthalpically higher potent inhibitory activity against NAMPT molecule and cancer cells. The main reason for this phenomenon was revealed to be due to the electronic repulsion of the piperazine-linker motif of **3b** with His191 of NAMPT. Importantly, this H191 is known to be a drug-resistance sensitive amino acid residue [36,37,50–52]. Thus, it is better to avoid interactions with His191 in designing effective NAMPT inhibitors. Further studies on the tunnel cavity of NAMPT using rigid analogues of azaindole-piperidine-motif inhibitors will allow a more definitive understanding of the inhibitor pharmacophore. Although additional confirmatory studies, such as the synthesis of novel azaindole-piperidine-motif NAMPT inhibitors and details of in vivo efficacy and resistibility are warranted, our findings provide valuable insights into the design for effective NAMPT inhibitors that offer improved therapeutic potential by making high specificity and avoiding resistance to NAMPT. Furthermore, on the basis of the characteristics of NAMPT functions and the mechanisms of action of NAMPT inhibitors, we are now trying to design additional azaindole–piperidine-motif NAMPT inhibitors available in vivo, such as the translation of a potent one into a payload for antibody–drug conjugates, which are an important therapeutic modality enabling targeted drug delivery to cancer cells [56].

**Author Contributions:** Conceptualization, S.-i.T. and N.T.; software, A.Y.; validation, T.O.; formal analysis, T.O. and A.Y.; investigation, K.K., Y.S. (Yuri Shibasaki), M.O., Y.O., Y.S. (Yoshimi Sakamoto) and M.N.; resources, Y.A., K.M., H.N. and H.T.; writing—original draft preparation, S.-i.T.; writing—review and editing, S.-i.T., A.Y., Y.A., H.Y., A.S., H.N., M.O., H.T., N.T. and F.U.; visualization, T.O., A.Y. and H.A.; supervision, S.-i.T.; project administration, S.-i.T.; funding acquisition, S.-i.T. and N.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by P-CREATE for division of Pharmaceutical Research and Development from the Japan Agency for Medical Research and Development, AMED grant (18cm0106133h0001) and the Hinoki Foundation.

**Acknowledgments:** The authors would like to thank staffs for Science and Technology, Organization for Research Advancement, Tokyo University of Science for their assistance.

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

### **References**


**Sample Availability:** Not available.

© 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* **Engineering Stem Cell Factor Ligands with Di**ff**erent c-Kit Agonistic Potencies**

**Tal Tilayov 1,**† **, Tal Hingaly 1,**† **, Yariv Greenshpan 2 , Shira Cohen 3 , Barak Akabayov 3 , Roi Gazit <sup>2</sup> and Niv Papo 1, \***


Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso Received: 21 September 2020; Accepted: 19 October 2020; Published: 21 October 2020

**Abstract:** Receptor tyrosine kinases (RTKs) are major players in signal transduction, regulating cellular activities in both normal regeneration and malignancy. Thus, many RTKs, c-Kit among them, play key roles in the function of both normal and neoplastic cells, and as such constitute attractive targets for therapeutic intervention. We thus sought to manipulate the self-association of stem cell factor (SCF), the cognate ligand of c-Kit, and hence its suboptimal affinity and activation potency for c-Kit. To this end, we used directed evolution to engineer SCF variants having different c-Kit activation potencies. Our yeast-displayed SCF mutant (SCFM) library screens identified altered dimerization potential and increased affinity for c-Kit by specific SCF-variants. We demonstrated the delicate balance between SCF homo-dimerization, c-Kit binding, and agonistic potencies by structural studies, in vitro binding assays and a functional angiogenesis assay. Importantly, our findings showed that a monomeric SCF variant exhibited superior agonistic potency vs. the wild-type SCF protein and vs. other high-affinity dimeric SCF variants. Our data showed that action of the monomeric ligands in binding to the RTK monomers and inducing receptor dimerization and hence activation was superior to that of the wild-type dimeric ligand, which has a higher affinity to RTK dimers but a lower activation potential. The findings of this study on the binding and c-Kit activation of engineered SCF variants thus provides insights into the structure–function dynamics of ligands and RTKs.

**Keywords:** receptor tyrosine kinases; protein-protein interactions; protein engineering; directed evolution; angiogenesis; binding affinity; agonistic activity

### **1. Introduction**

Receptor tyrosine kinases (RTKs) are major players in signal transduction, regulating cellular activities according to the availability and potency of their cognate ligands. RTKs and their ligands have thus been extensively studied with the dual aims of elucidating their biochemical properties and finding the means to manipulate them for clinical purposes [1–4]. However, despite the work that has been done to date, the delicate balance of RTK-ligand interactions and functional activation is not yet fully understood. It remains difficult to predict the relationship between ligand dimerization, specific RTK-ligand affinity, and functional activation potency. It has, for example, been shown for vascular endothelial growth factor (VEGF), human growth hormone (hGH), and macrophage colony stimulating factor (M-CSF) that activation of their cognate RTKs is balanced between self-dimerization and RTK affinity [5–10]. Thus, ligand 'fitness' (optimal dimerization and binding affinity) is usually not optimal, most probably due to inherent regulatory requirements for dynamic and evolutionary tuning [11–14].

Among the early attempts to engineer the conversion of RTK agonistic ligands into antagonists were those that focused on VEGF, hGH, M-CSF and stem cell factor (SCF) [11–13]. For example, abolishing ligand dimerization aimed to generate monomers that would act as receptor antagonists [15]. However, this approach was impeded by the diminished affinity for the RTK of the monomeric ligand relative to the dimeric ligand, which limited the therapeutic usage of such monomers [11,15,16]. Current methodologies aimed at generating potent RTK antagonists and agonists are thus based on protein engineering designed to modify ligands by changing both their affinities [17] and their self-dimerization status [14]. However, since affinity and self-dimerization are not mutually exclusive, the engineering of optimized protein-based antagonists or agonists must find the delicate balance between improved binding affinity and impaired self-dimerization [18]. In practical terms, combinatorial site-directed engineering of both ligand-ligand and ligand-receptor interactions provides the means to generate improved therapeutic mediators and to gain insights into the dynamics of RTK-ligand interactions.

A prototypic RTK-ligand interaction that has been extensively studied and that has generated important milestones in the field is the c-Kit/SCF interaction [19], which plays a key role in the regulation of epithelial, endothelial, neuronal, and hematopoietic stem-progenitor cells (HSPCs) [20–22]. For example, dysregulation of c-Kit/SCF signaling and gain-of-function c-Kit mutations have been implicated in different cancers [23–28], including thyroid carcinoma, oncocytic intraductal papillary mucinous neoplasms (IPMNs) of the pancreas, and lung cancer [29–31]. To understand c-Kit/SCF signaling, it is necessary to take a brief look at the two partners and at what is currently known about them. SCF is a four-helix bundle-type small protein that forms non-covalent homo-dimers, which may be membrane-anchored or soluble, depending on alternative RNA splicing and proteolytic processing [13,32]. The functional core of SCF is the N-terminal domain, which includes the dimerization interface and the portions of the molecule that bind to c-Kit [33]. Following bivalent binding of an SCF-dimer to a c-Kit monomer, the receptor dimerizes with another c-Kit protein, thereby bringing the two intracellular kinase domains into proximity and allowing them to phosphorylate one another to promote the signal transduction cascade [34–36]. As is the case for other growth factors, such as platelet-derived growth factor (PDGF) and VEGF [37], SCF dimerization is a prerequisite for the proper activation of c-Kit [35]. Indeed, mutated SCF showed reduced c-Kit activation and reduced mitogenic activity in cells, due to impaired SCF self-dimerization, reduced c-Kit affinity, or both [13]. Although the c-Kit receptor activation process has been studied extensively using monoclonal antibodies [38], recombinant protein receptors [35,39], and small molecule kinase inhibitors [36,40–44], the correlations between SCF concentration, SCF dimerization, SCF-c-Kit binding affinity and c-Kit receptor activation are not always fully understood. A particular problem is that most studies showing that wild-type SCF (SCFWT) is highly potent in activating c-Kit have been conducted with very low concentrations (low nanomolar range) of SCFWT, whereas effective SCF tissue concentrations are much higher (in the micromolar range), especially when SCF is intended for use as a therapeutic. To address this problem, it is necessary to generate and study SCF-derived proteins with altered self-dimerization properties and altered affinity for c-Kit that confer, at high concentrations, a superior agonistic function in comparison to SCFWT. Such engineered SCF-derived proteins will not only provide insights into the molecular mechanisms that mediate receptor activation but will also serve as a basis for further manipulations of therapeutic value.

In this study, we combined rational and combinatorial engineering approaches for transforming dimeric SCF into ligands with different agonistic potencies at high SCF concentrations. Specifically, we engineered variants with a reduced dimerization potential and an increased affinity for c-Kit. To this end, we screened random mutagenesis SCF monomer (SCFM) libraries to identify SCF<sup>M</sup> variants with different self-association properties and different affinities for c-Kit and hence with different degrees of activation of c-Kit. Importantly, despite the parental SCF<sup>M</sup> being the only monomer variant that binds c-Kit and despite its lower affinity for c-Kit vs. the wild-type dimeric SCF (SCFWT) and vs. the other SCF<sup>M</sup> variants, SCF<sup>M</sup> was superior to SCFWT and to the other SCF<sup>M</sup> variants in activating c-Kit. Our data shows that the impairment of SCF dimerization may increase the local SCF concentration near c-Kit and thereby induce enhanced dimerization and activation of the receptor. Studying a collection of SCF<sup>M</sup> variants with optimized functions enabled us both to identify preferential agonists and also to study the relationship between ligand dimerization and c-Kit receptor activation. We suggest that potent activators for other RTKs could be generated by applying a similar structure-based library screening approach, which would, in turn, facilitate further investigation of the mechanisms underlying the activation of RTKs.

### **2. Results**

### *2.1. Sorting of SCF<sup>M</sup> Libraries for Variants with a High A*ffi*nity for c-Kit*

Our strategy to generate SCF variants with different agonistic activities was based on modifying both SCF homo-dimerization and SCF binding affinity to c-Kit. We started by generating a double mutant variant, SCF<sup>M</sup> V49L F63L, exhibiting reduced homo-dimerization [13]. We expressed this mutant on the surface of yeast, as a platform for affinity enrichment (Figure S1 in the Supplementary Material). The first yeast surface display (YSD) library based on SCF<sup>M</sup> was then generated with an initial diversity of 4.5 × 10<sup>6</sup> . By using fluorescence-activated cell sorting (FACS), this library was enriched for clones with high expression and enhanced affinity for the soluble c-Kit extracellular domain. Six rounds of sorting were performed with decreasing concentrations of c-Kit (Figure 1A). Two SCF<sup>M</sup> clones, designated SCFM,K91E and SCFM,S64P,F126S,V131A,E134G,V139I, were chosen for purification based on their high frequency in the selected population and their enhanced affinity towards soluble c-Kit in the YSD system, respectively (Figure S2A). A second-generation library, based on SCFM,K91E, was then generated, with an initial diversity of 3 × 10<sup>6</sup> . This second library was further enriched through four rounds of sorting (Figure 1B), resulting in the identification of three high-affinity variants, designated SCFM,K91E,K24N, SCFM,K91E,D97G and SCFM,K91E,L98R, with the last of these three being the most frequent mutant in the high-affinity sorting gates (Figure S2B). SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R and SCFM,S64P,F126S,V131A,E134G,V139I were purified in non-glycosylated form, and their purity and molecular weights were determined (Figure S3). Circular dichroism (CD) spectroscopy showed that SCF<sup>M</sup> and SCFM,K91E shared the same secondary structure with SCFWT, according to our measurements and in agreement with a previous publication [33] (Figure S4A). Notably, CD spectra of the glycosylated and non-glycosylated SCFWT revealed that protein glycosylation did not affect the global secondary structure of the protein (Figure S4A).

**Figure 1.** Flow cytometry sorting of yeast-display SCF<sup>M</sup> library enriches the c-Kit binding population. Flow cytometry analysis of our first-generation (**A**) and second-generation (**B**) libraries. Yeast-displayed SCF was labeled with anti-c-myc antibody and secondary PE-conjugated antibodies (*x*-axis). Soluble c-Kit (added at 1 nM for analysis) was labeled with FITC (*y*-axis). Cells having variants with the highest affinities were sorted using the gates shown only for sort #1 in both panels for purposes of the illustration; the same gate was applied to all sorts.

### *2.2. The Dimeric State of SCF Proteins is Concentration Dependent*

– −3– −3 — — Biophysical experiments were then performed to determine the extent of dimerization of each SCF protein. The association of the soluble form of each protein, SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R, and SCFM,S64P,F126S,V131A,E134G,V139I , with the same SCF protein displayed on yeast was monitored (Figure 2A and Figure S1). In the concentration range of 62.5–500 nM (1 × 10 <sup>−</sup>3–8 × 10 <sup>−</sup><sup>3</sup> mg/mL), the dimerization of SCFWT and SCFM,K91E,L98R was stronger—by more than threefold—than that of SCFM, SCFM,K91E, or SCFM,S64P,F126S,V131A,E134G,V139I . We also used dynamic light scattering (DLS) to compare the molecular sizes (hydrodynamic radii) of these purified proteins. The hydrodynamic radius of SCFWT (0.5 mg/mL; 32.25 µM) of 3.39 ± 0.57 nm was larger than the hydrodynamic radii of SCFM, SCFM,K91E and SCFM,S64P,F126S,V131A,E134G,V139I, having values of 2.62 ± 0.17, 2.77 ± 0.25 and 2.65 ± 0.69 nm, respectively (Figure 2B). The hydrodynamic radius of SCFM,K91E,L98R (3.06 ± 0.2 nm) was larger than the hydrodynamic radii of the above three variants, but smaller than that of SCFWT, suggesting that more SCFWT and SCFM,K91E,L98R units are in dimeric form than those of SCFM, SCFM,K91E, or SCFM,S64P,F126S,V131A,E134G,V139I (Figure 2B). To gain additional independent information

about the dimerization of soluble SCF, we calculated the radius of gyration (Rg) of SCFWT and SCF<sup>M</sup> by using small angle X-ray scattering (SAXS). The measured data indicated that both SCFWT and SCF<sup>M</sup> at concentrations of 3 or 5 mg/mL (193 or 322 µM, respectively), which are 10-fold higher than the concentrations used in DLS, are dimers (Figure 2C,D). For both proteins, the Rg value was slightly higher than the theoretical Rg of the SCFWT dimer. Taken together, the above findings show that at low concentrations, SCFWT forms dimers and SCF<sup>M</sup> is monomeric, but at higher concentrations both proteins are dimers.

**Figure 2.** Mutations of SCF affect dimerization. (**A**) Homo-dimerization of soluble SCF proteins with the same yeast-surface displayed (YSD) SCF protein for: SCFWT (black columns), SCF<sup>M</sup> (white columns), SCFM,K91E (checkered columns), SCFM,K91E,L98R (dotted columns) and SCFM,S64P,F126S,V131A,E134G,V139I (SCFM,N, cross-hatched columns). The values obtained for the binding of the soluble SCF proteins to YSD SCF proteins were normalized to the value that was obtained for the binding of soluble SCFWT (at 500 nM) to the YSD SCFWT. Values are given as means of three independent experiments ± SEM, \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001. (**B**) DLS analysis of SCFWT (blue), SCF<sup>M</sup> (red), SCFM,K91E (dashed green), SCFM,K91E,L98R (pink), and SCFM,S64P,F126S,V131A,E134G,V139I (purple). (**C**) SAXS results showing the radius of gyration (Rg) of SCFWT (circles) and SCF<sup>M</sup> (squares) that was measured at 3 and at 5 mg/mL. The Rg values for the theoretical SCFWT monomer and dimer were calculated from the crystal structure (PDB: 1SCF) and are indicated as dashed lines. (**D**) The crystal structure of the SCF dimer was aligned with the SAXS models obtained for SCFWT (green) and SCF<sup>M</sup> (orange) using PyMOL.

### *2.3. SCFM,K91E and SCFM,K91E,L98R Exhibit High A*ffi*nity for c-Kit*

To directly measure the affinity of each SCF protein for c-Kit, we used surface plasmon resonance (SPR) with the receptor c-Kit protein immobilized on the chip and soluble ligand concentrations of 0.62–50 nM for glycosylated and non-glycosylated SCFWT, or 31.5–500 nM for SCFM, SCFM,K91E, SCFM,K91E,L98R and SCFM,S64P,F126S,V131A,E134G,V139I (Figure 3).

– –

**Figure 3.** Affinities of SCF variants for c-Kit. The association and dissociation of the soluble SCF variants to and from surface-immobilized c-Kit. (**A**) Non-glycosylated SCFWT; (**B**) Glycosylated SCFWT; (**C**) SCFM; (**D**) SCFM,K91E; (**E**) SCFK91E,L98R; and (**F**) SCFM,S64P,F126S,V131A,E134G,V139I .

– The K<sup>D</sup> values for the glycosylated and non-glycosylated SCFWT interacting with the immobilized c-Kit were very similar, namely, 5.82 ± 2.93 or 4.36 ± 1.46 nM, respectively (Table 1), with both being in the previously reported K<sup>D</sup> range for the SCF/c-Kit interaction (0.5–65 nM) [35,45]. As expected, SCF<sup>M</sup> showed a much lower affinity (K<sup>D</sup> = 146 ± 18.3 nM) for c-Kit. The affinities of SCFM,K91E and SCFM,S64P,F126S,V131A,E134G,V139I were 88.9 ± 19.9 nM and 74.4 ± 5.09 nM, respectively. Interestingly, the affinity of SCFM,K91E,L98R was 40.7 ± 0.7 nM, which is higher than that for the other mutants, but still lower than that of SCFWT.

**Table 1.** Binding affinities of SCF variants. The dissociation constant (KD) was determined from SPR sensograms of the equilibrium-binding phase. K<sup>D</sup> values are means ± SD of three independent experiments.


To further test the affinities of SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R and SCFM,S64P,F126S,V131A,E134G,V139I for c-Kit, we evaluated their binding to c-Kit in two different cell lines, namely, A172, a brain glioblastoma cell line, and murine HSPCs. In agreement with the above-mentioned YSD and SPR binding results, SCFM,K91E and SCFM,K91E,L98R exhibited a higher affinity for the A172 cells, as compared to SCFM, reaching the effective binding levels of SCFWT. In contrast, SCFM,S64P,F126S,V131A,E134G,V139I showed weaker binding to A172 cells, compared with SCFWT (Figure 4A). Murine HSPCs showed high expression of the c-Kit receptor, which has 83% sequence homology with human c-Kit [35,46] (Figure 4C). These cells, too, showed SCFM,K91E and SCFM,K91E,L98R to exhibit the highest binding (Figure 4D), suggesting that SCFM,K91E and SCFM,K91E,L98R bind human or murine c-Kit with similar or even higher affinities compared to the parental SCFM. The conformation that c-Kit adopts when displayed on the membrane of intact cells vs. that when it is immobilized on the SPR sensor chip may be different for the different epitopes that are exposed and available for binding SCFM,S64P,F126S,V131A,E134G,V139I vs. SCFM,K91E,L98R and/or SCFM,K91E. These differences may result in a small discrepancy between the SPR and cell binding results.

**Figure 4.** Improved and reduced binding of SCF variants to cell-surface c-Kit. (**A**) Expression levels of c-Kit in the A172 cell line. Grey and black lines represent c-Kit expression and background signals, respectively. (**B**) Binding of SCF variants to A172 cells expressing c-Kit. (**C**) Expression levels of c-Kit in murine HSPCs. (**D**) Binding of SCF variants to HSPCs. In panels (**B**,**D**) the columns are designated as follows: SCFWT (black columns), SCF<sup>M</sup> (white columns) and SCFM,K91E (checkered columns), SCFM,K91E,L98R (dotted columns) and SCFM,S64P,F126S,V131A,E134G,V139I (cross-hatched columns). The results were normalized to SCFWT binding at the highest concentration of 1000 nM. Values are means ± SEM of independent measurements performed in triplicate. \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001.

### *2.4. SCF-Dimerization is the Major Determinant for c-Kit Phosphorylation in Human Umbilical Vein Endothelial Cells*

Stimulation of c-Kit by SCF activates a wide array of signaling proteins and pathways, starting with c-Kit phosphorylation and followed by activation of phosphatidylinositide 3 ′ -kinase (PI3-kinase), Scr family kinases (SFK) and Ras-Erk pathways. In human umbilical vein endothelial cells (HUVECs), these events lead to cell proliferation and the formation of capillary-like structures by the endothelial cells [47]. To test for the activation of the c-Kit receptor by the different variants, we thus performed a phosphorylation assay in HUVECs as a cell-based model of angiogenesis. For this assay, we used concentrations of the same order as those used in the above-described biophysical and biochemical assays. As shown previously, these concentrations are representative of the local concentrations of SCF in tissues rather than of global levels in serum or other body fluids [48] (even though they are substantially higher than the concentrations used in some previous studies [13]). Western blot with a specific phospho-c-Kit showed that all the proteins induced receptor activation, but to different extents. The striking—and somewhat unexpected—finding was the difference between SCFWT and SCFM: SCFWT induced c-Kit activation by strongly binding to it as a dimer (Figure 5), but SCF<sup>M</sup> bound to c-Kit as a monomer, being the only variant to do so. Furthermore, although the affinity of SCF<sup>M</sup> for c-Kit was lower than that of SCFWT, SCF<sup>M</sup> was the only protein that increased c-Kit activation to a greater (albeit modestly) extent than SCFWT. Thus, the SCFWT homo-dimer and SCF<sup>M</sup> monomer acted in a similar way as c-Kit agonists, with the latter being slightly more potent in activating c-Kit expressed in HUVECs, as shown by its ability to induce the HUVECs to assemble into tube-like structures (Figure 6). In contrast, SCFM,K91E and SCFM,K91E,L98R were inactive in this assay. ′ — —

**Figure 5.** SCF variants activate c-Kit on primary human umbilical vein endothelial cells (HUVECs). HUVECs were incubated with SCF proteins (SCFWT, SCFM, SCFM,K91E or SCFM,K91E,L98R) at a concentration of 250 nM. c-Kit phosphorylation levels were normalized to the expression levels of c-Kit (second row) and to β-actin levels (third row). Values are means ± SEM of independent measurements performed in triplicate.

– **Figure 6.** SCF variants with different functional effects on primary human umbilical vein endothelial cells (HUVECs). (**A**) HUVECs with 2% fetal bovine serum (FBS) as the positive control (light gray column), untreated HUVECs as negative control (dark gray column), and cells treated with 4 µM of SCFWT (black columns), SCF<sup>M</sup> (white columns), SCFM,K91E (checkered columns) and SCFM,K91E,L98R (dotted columns). Tube length was normalized to untreated cells, which served as a control. Values are means ± SEM of independent measurements performed in triplicate. (**B**) Positive control. (**C**) Negative control. (**D**) SCFWT. (**E**) SCFM. (**F**) SCFM,K91E. (**G**) SCFM,K91E,L98R. In panels (**B**–**G**) all proteins were added at a concentration of 4 µM. \* *p* < 0.05; \*\* *p* < 0.01.

### **3. Discussion**

This study of engineered SCF ligands and their interaction with c-Kit, the cognate SCF receptor, provides proof of concept that receptor-activating ligands can be modulated to confer different agonistic potencies through changes of affinity for the cognate receptor and dimerization of the ligand itself. The study also showed that enhancement of the agonistic potency of SCF was concentration dependent in that the concentration in our study was very high, being in the micromolar range vs. the nanomolar range of previous studies [13]. Currently, most studies on SCF are conducted with concentrations in the range of 3–100 ng/mL; this concentration range, which is clearly appropriate for the routine

growth of cells, does not represent the actual local concentration at the cell surface. The naïve-state concentration of SCF in the circulation is 3.3 ng/mL; thus, the active concentration would certainly be higher and could reach substantially higher local concentrations at the cell surface [48]. Since it was our intention to test for robust biochemical properties of the studied mutants, we used a concentration range that would represent more faithfully the high local concentrations of SCF.

Reducing SCF dimerization while preserving or even enhancing its affinity for c-Kit will change the dynamics of binding and activation. At high concentrations, namely, at concentrations representative of tissue levels, SCFWT tends to dimerize before binding to c-Kit, with the SCFWT dimer, with its two binding sites, being likely to approach a single c-Kit receptor molecule. One binding site of the SCFWT dimer will therefore bind to one c-Kit molecule, while the other will remain free. In such a scenario, it is likely that most or all of the c-Kit molecules are each bound to a different SCFWT dimer via a single SCF monomer such that the dimerization of SCFWT (bound to c-Kit) and the subsequent c-Kit dimerization and activation would be slow or unlikely. In contrast, at low (nanomolar) SCFWT concentrations, as in previous studies [13], this dimeric ligand may bind two c-Kit molecules at the same time, resulting in c-Kit dimerization and activation. A different scenario is likely for the SCF<sup>M</sup> variants. SCF<sup>M</sup> will first bind as a monomer to c-Kit, and then dimerize with another SCF<sup>M</sup> protein attached to another c-Kit receptor, thereby bringing two c-Kit receptors into proximity and promoting c-Kit dimerization and hence activation. This scenario may explain the increase of c-Kit activation by the monomeric SCF<sup>M</sup> at the high concentrations tested (250 nM in the phosphorylation assay and 2 µM and 4 µM in the tube formation assay). In addition, intermediate levels of c-Kit activation could be obtained by enhancing the affinity of a mutated SCF<sup>M</sup> ligand for its c-Kit receptor, as was the case for SCFM,K91E and for especially SCFM,K91E,L98R, in order to compensate for changes in ligand dimerization, as these two characteristics are not mutually exclusive.

To identify SCF<sup>M</sup> mutants with enhanced affinity for c-Kit, random mutations were introduced, followed by screening yeast surface displayed SCF<sup>M</sup> libraries against c-Kit and selection of the mutants with the highest affinity in increasingly stringent sorts. The fact that during affinity maturation the SCF<sup>M</sup> proteins were displayed as monomers on the surface of yeast was important, since it meant that mutagenesis was controlled in such a way that affinity enhancement was biased towards dimerization or a higher order of multimerization. Indeed, sequencing of various clones selected from the final sorting round for the first- and second-generation libraries yielded several unique sequences, all including V49L and F63L mutations, that conferred enhanced affinity of SCF<sup>M</sup> variants for c-Kit. We found that the K91E mutation observed in the affinity-engineered SCF<sup>M</sup> clones (i.e., SCFM,K91E and SCFM,K91E,L98R) was near the SCF hot-spot residues related to the interaction of SCF with c-Kit and not to SCF-SCF dimerization (Supplementary Materials: Figure S5). These hot-spot residues may be involved in improving the affinity between SCF and c-Kit, either through direct binding interactions or by structural stabilization. Using SPR spectroscopy, we identified SCF variants with up to 3.7-fold higher affinity for recombinant c-Kit than the parental SCFM, even though the variants and SCF<sup>M</sup> possess similar receptor binding sites. SPR spectroscopy also showed that glycosylation of SCFWT did not influence its affinity for c-Kit, suggesting SCF/c-Kit interactions take place through the amino-acid backbone of SCF without a direct role of post-translational modifications, such as glycosylation.

It was found that the SCFM,K91E and SCFM,K91E,L98R variants exhibited improved binding to A172 cells vs. SCFM. Strikingly, SCFM,K91E, and to an even greater extent SCFM,K91E,L98R, also showed stronger binding than the dimeric SCFWT to the murine primary cells. It is thus likely that the L98R mutation may be involved in improving SCF homo-dimerization such that the monomeric SCFM,K91E possesses only a single receptor binding site, whereas SCFM,K91E,L98R possesses two c-Kit receptor binding sites and thereby promotes an enhancement of SCF-c-Kit interactions. Similarly, the ability of the variants to induce c-Kit activation and to activate the formation of capillary-like structures in HUVECs to different extents could be explained by differences in both their ability to self-associate (with SCF<sup>M</sup> and SCFM,K91E binding as monomers to c-Kit, and SCFWT and SCFM,K91E,L98R

binding as dimers) and in their affinity for recombinant soluble c-Kit or cell-surface-expressed c-Kit (with SCFM,K91E,L98R and SCFMK91E having higher affinity for c-Kit than SCFM).

Another factor contributing to the different effects of the mutant proteins may lie in their physical state: DLS measurements showed that SCF monomer self-association to a dimer was prevented for SCF<sup>M</sup> and SCFM,K91E, but not for SCFWT and SCFM,K91E,L98R, which did form dimers. Perhaps more importantly, SAXS measurements, which were performed to obtain low-resolution structures of SCFWT and SCF<sup>M</sup> at high concentrations in solution, demonstrated that the two proteins are similar in size and shape, both of which are compatible with a dimeric structure of SCF [49]. The mutations in the monomeric SCF<sup>M</sup> (i.e., V49L and F63L) appeared to significantly decrease its affinity for purified and cell-expressed c-Kit relative to SCFWT. This is presumably due to decreased avidity, as at low concentrations SCF<sup>M</sup> cannot bind two c-Kit molecules. Importantly, at high concentrations (3 or 5 mg/mL), SCF<sup>M</sup> can dimerize. In contrast, the mutation K91E in SCFM,K91E appears to maintain this variant in its monomeric state, while increasing its affinity for both purified and cell-expressed c-Kit relative to SCFM. The mutation L98R in SCFM,K91E,L98R compensates for the V49L and F63L mutations and restores SCF homo-dimerization at low concentrations. These biophysical observations can therefore explain the order of potency in the early and late signaling biological events (i.e., phosphorylation and tube formation, respectively) of the proteins: At high concentrations, SCF<sup>M</sup> is a monomer in solution prior to binding and becomes a dimer upon enhancement of its local concentration after binding to c-Kit expressed on cells. This makes SCF<sup>M</sup> a strong activator of c-Kit. SCFWT dimerizes better at low concentrations, such that it binds c-Kit strongly as an SCF–SCF dimer, probably capturing both binding sites (one on each monomer) on a c-Kit dimer. As a result, at the high SCF concentrations tested here, c-Kit activation following SCFWT self-association and c-Kit dimerization, is weaker than that with SCFM, but at low SCF concentrations the reverse is true, as shown in previous studies [13]. SCFM,K91E,L98R, for example, binds c-Kit as a c-Kit dimer and thus has an effect that is similar to that of SCFWT.

Our approach complements other methods for targeting SCF for therapeutic application. A recent example of that type of study was presented by a group from Novartis [50] demonstrating the potency of a c-KIT-directed ADC (a humanized anti-c-KIT antibody conjugated to a microtubule destabilizing small molecule) in models of mutant and wild-type c-KIT-positive solid tumors. In this respect, the significance of our study stems from the insight it provides into the sequence-structure–function relationships and mechanism of action of agonistic and antagonistic SCF mutants and it will therefore support engineering of further improved SCF variants as potential therapeutics. Moreover, the approach of using a natural protein ligand as a molecular scaffold for engineering high affinity agents can be applied to other ligands and to create functional protein agonists and antagonists against additional biomedical targets of interest.

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

### *4.1. Generating Random Mutagenesis SCF<sup>M</sup> Libraries in Yeast*

The gene encoding for SCF<sup>M</sup> was designed on the basis of SCFWT with two mutations, one at V49L and the other at F63L [13]. The gene was purchased from Integrated DNA Technologies (IDT, Coralville, IA, USA) and amplified by PCR with a Pfx50 polymerase (Life Technologies, Carlsbad, CA, USA). The PCR product and pCTCON (a generous gift from the laboratory of Dane Wittrup, MIT) [38] vector were digested with the restriction enzymes BamHI and Nhe1, ligated using T4 ligase (New England Biolabs, Ipswich, MA, USA), according to a standard protocol, and transformed into competent EBY100 *Saccharomyces cerevisiae* yeast cells using a MicroPulser electroporator (Bio-Rad, Hercules, CA, USA). The plasmid was extracted from the yeast using a Zymoprep™ yeast plasmid miniprep I kit (Zymo Research, Irvine, CA, USA). A first-generation DNA library originating from SCF<sup>M</sup> (used as a template for error prone PCR) was constructed using error prone PCR with low fidelity Taq DNA polymerase (New England Biolabs) and 2 uM of 8-oxo dGTP and dPTP nucleotide analogs

(Jena Bioscience, Jena, Germany). The primers with pCTCON homology used for library preparation were TAAGGACAATAGCTCGACGATTGAAG and GATTTTGTTACATCTACACTGTTG. A second amplification (to generate a second-generation library) used Phusion DNA polymerase (New England Biolabs) with SCFM,K91E as a template DNA and the primers GTTCCAGACTACGCTCTGCAGG and CAGATCTCGAGCTATTACAAG. Cloning into pCTCON was performed according to the same protocol as that used to produce the first-generation library. These reactions allowed homologous recombination of the inserts (PCR products) and the pCTCON vector. The yeast library was grown in SDCAA medium (2% dextrose, 0.67% yeast nitrogen base, 0.5% Bacto™ Casamino acids, 1.47% sodium citrate, 0.429% citric acid monohydrate, pH 4.5) at 30 ◦C with shaking at 300 rpm, until the culture reached OD<sup>600</sup> = 10.0 (10<sup>8</sup> cells/mL).

### *4.2. Screening of SCF Libraries*

Yeast cells expressing each library were grown in SGCAA medium (2% galactose, 0.67% yeast nitrogen base, 0.5% Bacto Casamino acids, 1.47% sodium citrate, 0.429% citric acid monohydrate) overnight at 30 ◦C, with shaking at 300 rpm, until the culture reached OD<sup>600</sup> <sup>=</sup> 10. Cells were washed with PBSA 1% [phosphate buffered saline (PBS) with 1% bovine serum albumin (BSA)]. The cells were double labeled with 1:50 anti c-myc antibody (9E10, Abcam, Cambridge, UK) and different concentrations of human recombinant c-Kit-Fc (Abcam) in PBSA 1% for 1 h at room temperature. Cells were washed with 1% PBSA and incubated with a sheep anti-mouse antibody conjugated to phycoerythrin (PE) (Sigma Aldrich, Rehovot, Israel) and a goat anti-human Fc conjugated to fluorescein isothiocyanate (FITC) (Sigma Aldrich, Rehovot, Israel), both at a 1:50 ratio for 30 min on ice in the dark. The first-generation library was sorted multiple times using SY3200 FACS (Sony Biotechnology, Bothell, WA, USA), and the second-generation library was sorted multiple times using FACS ARIA III (BD Biosciences, San Jose, CA, USA). For each sort round, the desired population was enriched by collecting the high-expressing (PE-labeled) and high c-Kit-binding (FITC-labeled) clones by using sorting gates that included 0.3–3.5% of the entire cell population. For affinity maturation of the SCF libraries, in each sort the c-Kit concentration was sequentially reduced. After each sort, the library was labeled according to the same protocol and analyzed using Accuri C6 flow cytometer (BD Biosciences) and FlowJo software (Tree star Inc., Ashland, OR, USA). SCF<sup>M</sup> individual clones (20 to 40) from each sort were sequenced (as above) using Geneious R7 (Biomatters, Auckland, New Zealand). The binding of individual SCF variants (in their YSD format) to soluble c-Kit was analyzed (as before), and SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R, and SCFM,S64P,F126S,V131A,E134G,V139I were chosen for production and purification as described in the Supplementary Materials.

### *4.3. Dimerization Assays Using Flow Cytometry*

Yeast cells expressing SCF proteins (SCFWT, SCFM, SCFM,K91E, SCFM,S64P,F126S,V131A,E134G,V139I and SCFM,K91E,L98R) were induced overnight in SGCAA medium, washed with 1% PBSA, and aliquoted at 1 × 10<sup>6</sup> cells/sample. Cells displaying the SCF proteins were labeled either with anti c-myc or with different concentrations of purified proteins (SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R, and SCFM,S64P,F126S,V131A,E134G,V139I) in 1% PBSA for 2 h at 4 ◦C on a rotary shaker. For secondary staining, PE-conjugated sheep anti-mouse and allophycocyanin (APC)-conjugated anti-Flag (BioLegend, San Diego, CA, USA) were used. Values are presented as mean fluorescence ± SEM; statistical significance was determined using a *t*-test. *p* value < 0.05 was considered statistically significant.

### *4.4. Dynamic Light Scattering*

DLS was used to determine the hydrodynamic radius of the purified proteins (SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R, and SCFM,S64P,F126S,V131A,E134G,V139I) in PBS in concentrations ranging between 0.3–0.5 mg/mL. Proteins were centrifuged for 1 h at 10,000 rpm and filtered through a 0.22-µm filter to remove contaminants. Spectra were collected with a CGS-3 goniometer (ALV, Munich, Germany). The laser power was 20 mW at the helium-neon laser line (633 nm). Correlograms were

calculated by the ALV/LSE 5003 correlator, which were collected 10 times, each time over 20 s, at 25 ◦C. The analysis was repeated four times at the detection angles of either 90◦ or 60◦ , in independent measurements. The correlograms were fitted by the CONTIN program [51]. DLS signal intensity was transformed to number distribution based on the Stokes–Einstein equation [52].

### *4.5. Small Angle X-ray Scattering*

SAXS data were collected on SAXLAB GANESHA 300 XL system, possessing a Genix 3D Cu-source, an integrated monochromator, three-pinhole collimation, and a two-dimensional Pilatus 300K detector. SCFWT and SCF<sup>M</sup> were measured at concentrations of 3 or 5 mg/mL. A buffer-only sample was used to set the background. The measurements were performed under vacuum at 25 ◦C. The scattering vector (*q*) ranged between 0.012 and 0.7 Å−<sup>1</sup> . The magnitude of the scattering vector is described by *q* = <sup>4</sup><sup>π</sup> *sin*<sup>θ</sup> λ , where 2θ is the scattering angle and λ is the wavelength. The values for the radius of gyration (*Rg*) were derived from a Guinier plot, namely, a linear small-angle part of the SAXS scattering curve (*qRg* < 1.0), in PRIMUS. In this region, the Guinier approximation is applicable:

$$I(q) = I(0)e^{\frac{-R\frac{2}{3}q^2}{3}}$$

where *I* is the scattering intensity and q is the scattering vector magnitude (a function of the scattering angle). *Rg* values were also derived using internal scripts [53] designed to perform an automatic search for the best fitting parameters using GNOM [54]. CRYSOL [55] was used to compute the artificial SAXS spectra based on the available crystal structure of SCF 1–141 (PDB: 1SCF) [49]. These spectra served as a reference for the reconstruction of the experimental SAXS data. The molecular envelope was reconstructed by GASBOR [54] based on the best GNON fit achieved from the internal script.

### *4.6. Surface Plasmon Resonance*

The affinity constant describing the interactions between c-Kit and soluble SCF proteins was determined on a ProteOn XPR36 instrument (Bio-Rad, Hercules, CA, USA). Recombinant c-Kit (R&D Systems) was immobilized on one channel of a ProteOn GLC sensor chip using the amine coupling reagents sulfo-NHS (*N*-hydroxysuccinimide; 10 mM) and EDC (1-ethyl-3-(3 dimethylaminopropyl)-carbodiimide; 40 mM) (Bio-Rad, Hercules, CA, USA). c-Kit, 1.2 µg or 2 µg, was covalently immobilized on the chip in 10 mM sodium acetate buffer, pH 4.0, to give 3866 or 5237 response units (RU), respectively. BSA (3 µg; 4706 RU) was immobilized on the chip as a negative control on a different channel. Unbound esters were deactivated with 1 M ethanolamine HCl at pH 8.5. The temperature was set at 25 ◦C, and the proteins were then allowed to flow over the chip in a range of concentrations (i.e., in series of threefold dilutions from 50 nM to 0.6 nM) for the glycosylated and non-glycosylated SCFWT and in a series of twofold dilutions from 500 nM to 31.25 nM for the remainder of the proteins (SCFM, SCFM,K91E, and SCFM,S64P,F126S,V131A,E134G,V139I). All the proteins were dissolved in PBST solution (PBS 0.005% Tween 20) and were then allowed to flow over the SPR chip surface at 100 µL/min for 150 s, followed by a dissociation phase of 270 s in PBST. After each run, a regeneration step was conducted with 50 mM NaOH. For each protein complex, a sensogram was generated from the RUs measured during the course of the protein–protein interactions by subtracting the background response of flow cells immobilized with BSA. The dissociation constant (KD) was determined from a sensogram of the equilibrium-binding phase. The data were analyzed with ProteOn manager 3.1.0.6 and fitted to 1:1 Langmuir binding model. To achieve statistical significance, only measurements with χ <sup>2</sup> values that were at least 12% or lower than the Rmax values were chosen for analysis [56].

### *4.7. Cell Binding Assays*

The A172 glioblastoma cell line was grown in Dulbecco's modified Eagle's medium (DMEM, Biological Industries, Beit HaEmek, Israel) with 10% FBS, 1% L-glutamine (Biological Industries) and 1% penicillin streptomycin (Biological Industries, Beit-Haemek, Israel). Cells were harvested with trypsin, and distributed to give 100,000 cells/well. Volumes of 100 µL containing different concentrations of each SCF protein were incubated in the different wells for 2 h at 4 ◦C on a rotary shaker. Cells were washed twice and then stained with APC conjugated anti-FLAG (BioLegend, San Diego, CA, USA) for 30 min on ice in the dark. The cells were analyzed using Accuri C6, and the data was analyzed using FlowJo software.

Using the same protocol, FACS analysis was performed on murine HSPCs from bone marrow. Lineage−c-Kit+Sca1<sup>+</sup> bone marrow primary cells were grown in BioTarget medium (Biological Industries, Beit HaEmek, Israel) with 1% L-glutamine, 1% penicillin streptomycin, 10 ng/mL SCF, thyroid peroxidase (TPO), IL-3, and FLT-3 (Peprotech, Rehovot, Israel). The cells were analyzed using Gallios flow cytometer (Beckman Coulter Inc., Carlsbad, CA, USA), and the data was analyzed by Kaluza (Beckman Coulter Inc., Carlsbad, CA, USA) and Prism software (GraphPad, San Diego, CA, USA). The results are presented as means ± SEM of triplicate measurements. Statistical significance was determined using *t*-test analysis. *p* value < 0.05 was considered statistically significant.

### *4.8. c-Kit Phosphorylation*

HUVECs were grown in EGM-2 medium (Lonza, Basel, Switzerland) containing serum and growth factors40. Cells were grown in 6-well plates to 80% confluence and starved for 18 h by incubation in EBM-2 medium without serum and growth factors. Cells were stimulated for 7 min with 250 nM of each of the SCF proteins (i.e., SCFWT, SCFM, SCFM,K91E and SCFM,K91E,L98R). The cells were washed twice with cold PBS and collected in 200 µL of lysis buffer [50 mM Hepes, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1.5 mM MgCl2, 1 mM EDTA, 1 mM Na3VO<sup>4</sup> and protease inhibitor cocktail (Roche)] for 10 min on ice. The samples were clarified by centrifugation at 15,000 rpm for 10 min at 4 ◦C. The quantity of total proteins in the lysed cells was determined using a BCA kit (Thermo Scientific, Waltham, MA, USA). Western blot used 20 µg of proteins per lane. After blocking the membrane with TBST (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20) supplemented with 5% BSA, primary antibodies [i.e., rabbit-anti-phospho-c-Kit (R&D Systems, Minneapolis, MN, USA) at a dilution of 1:2500; rabbit-anti-c-Kit (R&D Systems, Minneapolis, MN, USA) at a dilution of 1:1500; or β-actin antibody (Cell Signaling Technology, Danvers, MA, USA) at a dilution of 1:1000] were added for an overnight incubation at 4 ◦C. The membrane was washed three times for 10 min with TBST, and then a secondary antibody, i.e., anti-rabbit IgG-HRP (Cell Signaling Technology, Danvers, MA, USA), at a dilution of 1:1000 was added for 1 h at room temperature. Blots were developed with EZ-ECL Kit (Biological Industries, Beit HaEmek, Israel). The bands were visualized by chemiluminescence with Fusion-FX7 spectra (Vilber Lourmat, Collégien, France), and the intensity of each band was measured with the image analysis ImageJ software. Values are given as means ± SEM of triplicates.

### *4.9. Matrigel Endothelial Tube Formation Assay*

Growth factor reduced (GFR) Matrigel (Corning, New York, NY, USA) was used to coat a 96-well plate by centrifugation at 300× *g* for 10 min at 4 ◦C and incubation at 37 ◦C for 30 min. HUVEC cells (3.5 × 10<sup>4</sup> cells per well) were resuspended in EBM-2 with different concentrations of each SCF protein. After incubation of 16–18 h at 37 ◦C, the plate was monitored with an EVOS FL Cell Imaging System (Thermo Scientific, Waltham, MA, USA), at a ×2 magnification. The results were analyzed with ImageJ software and with Prism software (GraphPad, San Diego, CA, USA). Values are given as means ± SEM of quadruplicates.

### **5. Conclusions**

In summary, the overall goal of this research was to use both rational and combinatorial synthetic methods to develop a new generation of SCF-derived proteins as c-Kit agonists with potential therapeutic applications and as tools to study basic ligand–receptor recognition and receptor activation during key biological processes, in this case SCF-SCF interactions and their influence on SCF/c-Kit function. We found that reduced dimerization of the SCF variants prior to receptor binding resulted in enhanced c-Kit receptor activation, whereas improved receptor binding affinity of the engineered SCF monomers resulted in a reduced receptor activation. In seeking explanations for our findings, we fully characterized the binding and biological properties of the purified variants by employing cell-based models of receptor activation. Our data shows that the impairment of SCF dimerization may increase the local SCF concentration near c-Kit and thereby induce enhanced dimerization and activation of the receptor.

**Supplementary Materials:** The following are available online, Figure S1: YSD; Figure S2: Binding analysis of individual SCF variants to soluble c-Kit; Figure S3: Protein production and purification; Figure S4: CD spectra of the purified proteins; Figure S5: Schematic representation showing the locations of SCF mutations in the SCF/c-Kit complex (PDB: 1SCF).

**Author Contributions:** T.T., T.H. and N.P. designed the research; T.T., T.H., Y.G. and S.C. performed the research; T.T., T.H., Y.G., S.C., B.A., R.G. and N.P. analyzed the data; T.T., T.H. and N.P. wrote the manuscript with assistance from R.G. All authors edited the manuscript and approved the final version. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Research Council (ERC) grant 336041 and the ISF grant 1615/19 to N.P.

**Acknowledgments:** We thank A. Zilka (BGU) and U. Hadad (BGU) for help with SPR and FACS experiments. The DLS and SAXS experiments were conducted at the Ilse Katz Institute for Nanoscale Science & Technology with the help of Sharon Hazan.

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

### **References**


**Sample Availability:** Samples of the compounds SCFWT, SCFM, SCFM,K91E, SCFM,K91E,L98R, and SCFM,S64P,F126S,V131A,E134G,V139I are available from the authors.

**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* **The Potential Role of Cathepsin K in Non-Small Cell Lung Cancer**

#### **Hui Yang 1 , Jasmine Heyer 2 , Hui Zhao 1 , Shengxian Liang 1 , Rui Guo 1, \* and Li Zhong 1,2, \***


Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso Received: 19 August 2020; Accepted: 8 September 2020; Published: 10 September 2020

**Abstract:** (1) Background: Cathepsin K has been found overexpressed in several malignant tumors. However, there is little information regarding the involvement of Cathepsin K in non-small cell lung cancer (NSCLC). (2) Methods: Cathepsin K expression was tested in human NSCLC cell lines A549 and human embryo lung fibroblast MRC-5 cells using Western blot and immunofluorescence assay. Cathepsin K was transiently overexpressed or knocked down using transfection with a recombinant plasmid and siRNA, respectively, to test the effects on cell proliferation, migration, invasion, and on the mammalian target of rapamycin (mTOR) signaling pathway. (3) Results: Expression of Cathepsin K was increased significantly in A549 cells and diffused within the cytoplasm compared to the MRC-5 cells used as control. Cathepsin K overexpression promoted the proliferation, migration, and invasion of A549 cells, accompanied by mTOR activation. Cathepsin K knockdown reversed the above malignant behavior and inhibited the mTOR signaling activation, suggesting that Cathepsin K may promote the progression of NSCLC by activating the mTOR signaling pathway. (4) Conclusion: Cathepsin K may potentially represent a viable drug target for NSCLC treatment.

**Keywords:** NSCLC; Cathepsin K; cell proliferation; cell migration; cell invasion; mTOR

### **1. Introduction**

Non-small cell lung cancer (NSCLC) is the leading cause of cancer death among men and women worldwide, with an incidence rate of 1.3 million cases per year [1]. Since NSCLC remains asymptomatic during early stages, about 80% of patients are already in metastatic stages when they are diagnosed, and their 5-year survival rates are below 15% [2]. Despite recent progress in the development of novel medications and immunotherapies in treating NSCLC, the therapeutic efficacies still remain unsatisfactory [3]. Therefore, finding new therapeutic targets for the treatment of NSCLC has become a prioritized task [4].

Cathepsin K is a type of lysosomal cysteine protease which belongs to the papain-like cysteine peptidase family; other members include Cathepsin B, Cathepsin D and Cathepsin L etc. Physiologically, Cathepsin K functions in mediating cellular protein turnover, collagen degradation, and remodeling of the extracellular matrix, which plays an important role in pulmonary fibrosis [5–8]. Deficiency of Cathepsin K can lead to severe bone abnormalities, as it is the main peptidase involved in bone remodeling in osteoclasts [9,10]. In addition, increased expression and activity of Cathepsin K have been reported in patients diagnosed with breast cancer [11], bone cancer [12], prostate cancer [13], and many other types of epithelial-derived cell cancers [14–16]. Similar to the dysregulated Cathepsin B expression in the tumor microenvironment inducing tumor progression [17], Cathepsin K over-expression is

associated with cancer metastatic disease, indicating its potential diagnostic and prognostic value. Distinct expression patterns of Cathepsin K have been identified in lung cancer cells and stromal cells, which provide further supporting evidence for this protease's significant prognostic value [18]. However, the specific role and mechanism of Cathepsin K in NSCLC is still unknown.

In this study, we revealed the effect of Cathepsin K expression on NSCLC cells in terms of cell proliferation, migration, and invasion in vitro. In order to understand the mechanisms, we also investigated the mammalian target of rapamycin (mTOR) signaling pathway. The mTOR signaling pathway plays an important role in maintaining cell growth, proliferation, motility, and survival [19]. Upregulation of the mTOR pathway has also been reported in a large number of NSCLC tumors, with increased p-mTOR expression in up to 90% of patients with adenocarcinoma, 60% of patients with large cell carcinoma, and 40% of patients with squamous cell carcinoma [20–22]. mTOR activation may also be associated with poor prognosis in early NSCLC [23,24]. Thus, mTOR inhibitors have been widely studied and employed clinically in order to suppress tumor growth and sensitize cells to anticancer drugs. Previous studies have demonstrated that inhibition of Cathepsin K can significantly reduced the phosphorylation of mTOR at S2448 in Caki cells [25]. Therefore, Cathepsin K may mediate activation of the mTOR signaling pathway in NSCLC. Our findings indicate that Cathepsin K has the potential in developing as a therapeutic target for NSCLC.

### **2. Results**

### *2.1. Cathepsin K Was Highly Expressed in A549 Cells and Di*ff*used in the Cytoplasm*

Endogenous expression of Cathepsin K was detected in human embryonic lung fibroblasts MRC-5 and NSCLC cells A549 using a Western blot (WB) analysis and immunofluorescence (IF) assay. As shown in Figure 1a,b of WB results, compared to MRC-5 cells, the expression levels of Cathepsin K in A549 cells were significantly increased. An IF assay revealed that Cathepsin K was slightly expressed in MRC-5 cells, while Cathepsin K in A549 cells was largely expressed and diffused in the cytoplasm (Figure 1c), the cell fluorescence intensity was significantly increased (Figure 1d).

**Figure 1.** Detection of the expression and location of Cathepsin K in cells. (**a**) Western blot (WB) analysis for Cathepsin K expression in MRC-5, A549 cells. Representative gel blots of Cathepsin K and Tubulin using specific antibodies. (**b**) Cathepsin K/Tubulin; (**c**) Cathepsin K immunofluorescence (IF) staining in MRC-5 and A549 cells. (**d**) The cell fluorescence intensity was calculated using Image J software (mean ± SEM, *n* ≥ 3, \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001).

### *2.2. Cathepsin K Overexpression and Silence Models Were Successfully Reconstructed In Vitro*

≥ ≤ ≤ ≤

In order to observe the phenotypic variations mediated by Cathepsin K, it was transiently overexpressed or knocked down using transfection with a recombinant plasmid and siRNA into A549 cells respectively. As shown in Figure 2a, Cathepsin K mRNA expression levels were significantly increased in the Cathepsin K overexpression cells (*CTSK*-OE) compared to control A549 cells. The protein levels of Cathepsin K were also increased 1.5 times compared to A549 cells, and increased 1.6 times compared to the vector group (Figure 2b,c). Three Cathepsin K siRNA sequences were constructed for transfection, and the optimal sequence was selected according to its silencing efficiency. The results revealed that both siRNA 2# and siRNA 3# can effectively and markedly silence Cathepsin K expression (Figure 2d,e). The 3# sequence was selected to construct the Cathepsin K knock down (*CTSK*-KD) model.

≥ ≤ ≤ ≤ **Figure 2.** Construction of the Cathepsin K overexpression and knockdown models in vitro. (**a**) qRT-PCR analysis for *CTSK* expression in cells. *CTSK*/*GAPDH*; (**b**) WB analysis for Cathepsin K expression in cells. Representative gel blots of Cathepsin K and Tubulin using specific antibodies. (**c**) Cathepsin K/Tubulin; (**d**) WB analysis for Cathepsin K expression in cells. Representative gel blots of Cathepsin K and Tubulin using specific antibodies. (**e**) Cathepsin K/Tubulin (mean ± SEM, *n* ≥ 3, \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001).

### *2.3. CTSK-OE Promoted the Proliferation, Migration and Invasion and CTSK-KD Inhibited the Proliferation, Migration, and Invasion of A549 Cells*

In order to understand whether *CTSK*-OE and *CTSK*-KD could have a significant effect on the proliferation, migration, and invasiveness of A549 cells, cell proliferation, cell scratch repair, and Transwell assays were conducted in A549 cells as shown in Figure 3. *CTSK*-OE significantly promoted the proliferation of A549 cells from the time point 48 onward (Figure 3a), significantly promoting cell migration and invasion (Figure 3b,c). In contrast, *CTSK*-KD inhibited the proliferation of A549 cells from the time point 72 onward (Figure 3a), significantly inhibiting cell migration and invasion (Figure 3b,c).

**Figure 3.** Detection of cell behavior in vitro. (**a**) Cell Counting Kit-8 (CCK-8) analysis for A549 cell proliferation. (**b**) Cell scratch repair analysis for A549 cell migration. (**c**) Invasion experiment analysis for A549 cell invasion (mean ± SEM, *n* ≥ 3, \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001).

### *2.4. CTSK-OE Promoted While CTSK-KD Inhibited the Activation of the mTOR Signaling in A549 Cells.*

In order to understand the mechanism of Cathepsin K in NSCLC, we detected mTOR and p-mTOR expression. The results are shown in Figure 4a. With a change in Cathepsin K expression levels (Figure 4b), *CTSK-*OE significantly increased level of p-mTOR and p-mTOR/mTOR, which promoted over-activation of the mTOR signaling pathway (Figure 4c,d). In contrast, *CTSK*-KD significantly decreased levels of p-mTOR and p-mTOR/mTOR, which inhibited activation of the mTOR signaling pathway (Figure 4c,d). No significant differences were found in the mTOR/Tubulin ratio among the groups (Figure 4e).

≥ ≤ ≤ ≤ **Figure 4.** WB analysis for mTOR signaling protein expression in A549 cells. (**a**) Representative gel blots of Cathepsin K, p-mTOR, mTOR and Tubulin using specific antibodies. (**b**) Cathepsin K/Tubulin; (**c**) p-mTOR/Tubulin; (**d**) p-mTOR/mTOR; (**e**) mTOR/Tubulin. (mean ± SEM, *n* ≥ 3, \* *p* ≤ 0.05, \*\* *p* ≤ 0.01, \*\*\* *p* ≤ 0.001).

### **3. Discussion**

In addition to normal physiological functions [26–28], Cathepsin K also exhibits deleterious effects on the body, as evidenced by its role in the progression of a variety of tumors. Over the past few years, accumulated data have shown overexpression of Cathepsin K in multiple cancer types, indicating its role in tumor progression and its potential diagnostic and prognostic values. For example, strong expression of Cathepsin K has been observed in primary melanoma and melanoma metastases [29]. Colorectal cancer is associated with high LPS secretion and overexpression of Cathepsin K [30]. In order to study the expression of Cathepsin K in NSCLC, human NSCLC cell lines A549 were selected in this experiment, and human embryo lung fibroblast MRC-5 cells were used as controls. WB and IF assays showed that the expression level of Cathepsin K in A549 was higher than that in MRC-5, which is consistent with the conclusion that Cathepsin K is highly expressed in cancer tissues and cells in the literature. Compared with MRC-5 cells in which Cathepsin K was slightly expressed in cells, Cathepsin K was largely expressed in A549 cells and diffused into the cytoplasm. Cathepsin K is isolated into lysosomes through the endosome, and it can also be secreted into the other compartments of the cell and extracellular environment [31,32], which is essential for its role in promoting the development of cancer cells in tumors. It is worth noting that Cathepsin K seems to also be distributed in the nucleus of MRC-5 cells. In general, the pathway for Cathepsin proteins entering the nucleus without nuclear localization sequence (NLS) has been a controversial issue. A recent study demonstrated that Cathepsin proteins could be "chaperoned" into the nucleus from a cytoplasmic source following

possible leakages from the lysosome [33]. The mechanism of Cathepsin positioning to the nucleus and its role in the nucleus are also matters for further exploration.

Proteases play important roles in cancer initiation, development and metastasis. Cathepsin K–shRNA transfection has been demonstrated to downregulate Cathepsin K, inhibiting the proliferation and metastasis of breast cancer cells [34]. In squamous cell carcinoma, mesenchymal fibroblasts expressing Cathepsin K are secreted by tumor cells through interleukin-1 stimulation and are associated with tumor aggressiveness [35]. In vitro, Cathepsin K knockdown has been shown to inhibit migration and invasiveness of the OV-2008 cell line in epithelial ovarian cancer [36]. In order to explore the effects of Cathepsin K on the proliferation, migration, and invasion of NSCLC, this experiment used a transfection method to introduce a Cathepsin K recombinant plasmid and *CTSK*-siRNA into A549 cells to construct *CTSK*-OE and *CTSK*-KD experimental models. This research employed a two-way crossover study in order to investigate the role of Cathepsin K in A549 cells. In CCK-8 detection at 48 h, the results showed that cell proliferation in the *CTSK*-OE was significantly increased. At 72 h, *CTSK*-KD significantly decreased in the cell proliferation. From the time course prospective, it is likely because siRNA displays characteristics that are distinct from the recombinant plasmid [37,38], it plays a role at a longer time point in the experiment compared to the recombinant plasmid. This phenomenon was not observed in cell migration and invasion experiments, which might be due to the researchers' selection of a specific time point for cellular detection.

In the cell scratch repair experiment, considering the damage of the transfection reagent to the cells, 1% fetal bovine serum was added to the culture medium. The cell proliferation cycle is generally 24 h in length. In order to exclude the effect of serum on migration experiments, the time point of 24 h was selected for the detection. In the same manner that the invasion experiment was conducted, no serum was present in the upper chamber of Transwell, but the medium in the lower chamber contained a serum concentration of 20%. The large concentration gradient between the upper and lower chambers resulted in a tendency for the A549 cells to be preferentially located in the lower chamber. The time point at 24 h was selected in order to avoid the potential proliferation of cells after penetration of the polycarbonate membrane, a factor that could influence the accuracy of the experiment. The results of the migration and invasion experiments were consistent with the results of the cell proliferation experiments. *CTSK*-OE significantly enhanced the migration and invasion ability of A549 cells. In contrast, the migration and invasion ability of *CTSK*-KD was significantly reduced. This is consistent with the conclusion that Cathepsin K is highly expressed during tumor invasion and metastasis [39] and has a stimulating effect on the aggressive phenotype of various types of cancer. Soond et al. [40] reported that the findings of use of utilizes combined chemotherapeutic treatment such as with Tocilizumab or Rituximab inhibited signaling transduction pathways which up-regulate the intracellular Cathepsins look very encouraging for targeting cancer. In the gastric metastasis model, the use of the pharmaceutical agent Odanacatib significantly inhibited the metastases of cancer cells, suggesting that Cathepsin K inhibition may be employed as a new therapeutic strategy to prevent tumor metastasis [41–43].

In addition to its well-known role in extracellular matrix degradation and remodeling, Cathepsin K may mediate activation of the mTOR signaling pathway. mTOR promotes anabolism and protein synthesis by phosphorylating its substrates S6 and EIF4E1 [44–46], as shown in Figure 5. It can further activate downstream products of the eIF4 complex to promote tumor development, regulate the cell cycle, and inhibit autophagy and apoptosis [19]. Seo et al. [25] demonstrated that inhibition of mTOR enhanced the chemosensitivity of cancer cells. They also revealed that treatment with mTOR inhibitors reduced tumor size and increased apoptosis in a xenograft model. Evimus, an inhibitor of mTOR, selectively inhibits mTOR signaling. It has been comprehensively evaluated in several phase I trials of previously treated advanced NSCLC [47].

**Figure 5.** The role of mTOR in vivo. mTOR promotes protein synthesis by phosphorylating its substrates S6 and EIF4E1. It can further activate downstream products of eIF4 complex to promote tumor development, regulate the cell cycle, and inhibit autophagy and apoptosis.

In this experiment, *CTSK-*OE significantly increased levels of p-mTOR and p-mTOR/mTOR, which promoted over-activation of the mTOR signaling pathway. In contrast, *CTSK*-KD significantly decreased levels of p-mTOR and p-mTOR/mTOR, which inhibited activation of the mTOR signaling pathway. Previous studies have demonstrated that inhibition of Cathepsin K can induce proteasomal degradation of proteins associated with regulatory mechanisms on the target of rapamycin [25]. This conclusion is inconsistent with our experiment, which suggests that Cathepsin K may play a role in NSCLC through regulation of the mTOR signaling pathway. However, the mechanism of action by which Cathepsin K promotes the proliferation, migration, invasion, and other malignant behavior of NSCLC cells is unclear. Potential mediators involved in the effects of Cathepsin K could include a certain molecule's unique and powerful hydrolytic activity or its role in the mTOR signaling pathway. In addition, whether the changes of Cathepsin K in NSCLC cells are related to the levels of p-mTOR remains unclear. Therefore, it is necessary to supplement the Cathepsin K activity detection and mTOR signaling pathway inhibitor experiments in further research.

μ μ In summary, our results demonstrated that Cathepsin K was overexpressed in NSCLC cells and permeated the cytoplasm. Through the detection of cell biological functions, it was found that increased expression of Cathepsin K promoted the proliferation, migration, invasion, and activation of the mTOR signaling pathway of NSCLC cells, while silencing Cathepsin K expression reversed the above behavior. These findings suggest that Cathepsin K may potentially represent a new therapeutic target for NSCLC.

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

### *4.1. Cell Culture and Transfection*

The MRC-5 and A549 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Sangon Biotech, Shanghai, China) supplemented with 10% fetal bovine serum (FBS) (ExCell Bio, Beijing, China) and cultured in 5% CO<sup>2</sup> at 37 ◦C. The A549 cells were stably upregulated and downregulated for Cathepsin K expression using the Cathepsin K plasmid and siRNA (Genepharma, Shanghai, China), respectively. DNA transfection was performed using Lipofectamine-3000 (Thermo Fisher, Waltham, MA, USA), and siRNA was transfected at a final concentration of 50 nM using Lipofectamin-3000.

### *4.2. qRT-PCR*

Total RNA was isolated using RNAiso Plus (TaKaRa, Kusatsu-shi, Japan), and RNAs were quantified using a NanoDropTM 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Synthesis of cDNA and reverse transcription was performed using 1 µg of total RNA in a 25 µL system following the instructions of a FastQuant RT Kit (Tiangen, Beijing, China). *CTSK* and *GAPDH* primers for qPCR were designed by PrimerPremier5 software. The sequences were *CTSK*-F: 5'-CCTTGAGGCTTCTCTTGG-3' *CTSK*-R: 5'-AGGGTGTCATTACTGCGG-3'; *GAPDH-F*: 5'-AGAAGGCTGGGGCTCATTTG-3' *GAPDH-F*: 5'-AGGGGCCATCCACAGTCTTC-3'. Quantitative real-time PCR was performed for *CTSK* and *GAPDH* (housekeeping gene) using a C1000 Touch Thermal Cycler CFX96TM Real-Time System (Bio-Rad, Shanghai, China) per the Universal SYBR Green qPCR Supermix (UE, Suzhou, China) instructions. Real-time PCR was triplicated for each cDNA sample.

### *4.3. Western Blot Analysis*

Total protein was isolated using radio immunoprecipitation assay (RIPA) lysate (strong) (Solarbio, Beijing, China), and the whole process was performed on ice. Proteins were quantified using a Microplate reader (Molecular Devices, Silicon Valley, CA, USA) following the instructions of a bicinchoninic acid (BCA) protein quantification kit (Solarbio, Beijing, China). Protein samples were then separated onto SDS-polyacrylamide gels and transferred electrophoretically to polyvinylidene fluoride (PVDF) membranes. The membranes were blocked with 5% milk and incubated overnight at 4 ◦C with anti-CathepsinK (1:1000; ab19027), anti-mTOR (1:1000; CST2983), anti-p-mTOR (1:1000; CST5536), anti-GAPDH (1:10,000; Proteintech10494-1-p), and anti-Tubulin (1:1000; CST2144). Blots were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody (1:3000; CST7074). Antigens were detected using a luminescence method. Band densities were determined using Image Lab software (version 5.1, Bio-Rad).

### *4.4. Immunofluorescence Assays*

The cells were incubated successively for fixation and permeation with 4% paraformaldehyde (Leagene Biotechnology, Anhui, China) and 0.1% TritonX-100 (Leagene Biotechnology, Anhui, China). After successively incubating the cells with the configured PBST (phosphate buffer saline (PBS) (Sangon Biotech, Shanghai, China) + 0.1% Tween 20 + 1% BSA + 22.52 mg/mL Glycine) to block the non-specific binding of the antibody, the slides were rinsed and incubated with the appropriate primary antibody for Cathepsin K (1:100) overnight at 4 ◦C. The following day, the slides used for IF staining were incubated with fluorescein-conjugated goat anti-rabbit IgG and subsequently stained with nuclear dye 4′ ,6-diamidino-2-phenylindole (DAPI). Finally, the slides were then examined and imaged using an Olympus fluorescence microscope (fv3000), and the fluorescence intensity was calculated using Image J software (version 1.48v). In this experiment, 3 images were collected in each group, and each image outlines 10 cells for fluorescence intensity analysis.

### *4.5. Proliferation Assays*

A Cell Counting Kit (APE×BIO, Houston, TX, USA) was used to evaluate the variation in cell proliferation, which is based on the dehydrogenase activity detected in viable cells. The formazan dye generated by dehydrogenases absorbs light at a wavelength of 450 nm. The amount of formazan dye present in cells is directly proportional to the number of living cells. In brief, 100 µL of cell suspension (1 × 10<sup>3</sup> ) was incubated in 96-well culture plates, and 10 µL of CCK-8 solution was added at the time set by the experiment. Cells were incubated at 37 ◦C for 2 h. Absorbance was analyzed at 450 nm using a microplate reader.

### *4.6. Migration Assays*

A migration assay was used to detect the variation in cell migration. A549 cells were plated in 6-well plates at a certain amount (1 × 10<sup>6</sup> ) and allowed to form a confluent monolayer for 24 h. The monolayer was scratched with a sterile pipette tip (10 µL), followed by a wash with PBS to remove floating and detached cells. The cells were then imaged (at time 0 h and 24 h) by fluorescent inversion fluorescence microscopy (Olympus, Tokyo, Japan). Image J software was used to measure the scratch

area and length of each image at 0 h and 24 h. According to the formula area/length, calculate the scratch width of the two periods and make the difference of the corresponding image, and the scratch width difference/corresponding initial width is the respective migration of each group.

### *4.7. Invasion Assays*

The invasion assay was performed using a Transwell 24-well dish with a pore size of 8 µm (Costar, NY, USA). Mix the serum-free DMEM medium and Matrigel (Corning, Corning, NY, USA) at a ratio of 6:1 on ice, and added 60 µL to the Transwell chamber. 100 µL of DMEM serum-free medium cell suspension (5 × 10<sup>4</sup> ) was placed in the upper chamber, and 500 µL of DMEM medium containing 20% serum was placed in the lower chamber. The cells were incubated for 24 h at 37 ◦C in 5% CO2, then fixed in 4% methanol and stained with 0.1% crystal violet. Cells on the upper side of the filters were removed with cotton-tipped swabs and the filters were then washed with PBS. Cells on the underside of the filters were examined and counted under a microscope.

### *4.8. Statistical Analysis*

Results were representative of at least three independent experiments, and all values were expressed as mean ± SEM. Statistical significance (*p* < 0.05) for each variable was estimated by an unpaired *t*-test (two-tailed) or a one-way analysis of variance (ANOVA) followed by a Tukey's post hoc analysis.

**Author Contributions:** Conceptualization, L.Z. and R.G.; methodology, H.Y. and H.Z; validation, J.H., H.Y. and S.L.; formal analysis, H.Y. and S.L.; writing—original draft preparation, H.Y.; writing—review and editing, L.Z., R.G. and H.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported in part by an NSFC grant (81773178, L.Z.), the Hebei Natural Science Foundation (#C2019201349, R.G.), and the National Natural Science Foundation of China (#31900534, R.G.).

**Acknowledgments:** The authors gratefully acknowledge Guo-ping Shi, from Harvard Medical School, Brigham and Women's Hospital, for kindly providing us with the *CTSK* plasmid.

**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**


47. Vansteenkiste, J.; Solomon, B.; Boyer, M.; Wolf, J.; Miller, N.; Di Scala, L.; Pylvaenaeinen, I.; Petrovic, K.; Dimitrijevic, S.; Anrys, B.; et al. Everolimus in combination with pemetrexed in patients with advanced non-small cell lung cancer previously treated with chemotherapy: A phase I study using a novel, adaptive Bayesian dose-escalation model. *J. Thorac. Oncol. O*ff*. Publ. Int. Assoc. Study Lung Cancer* **2011**, *6*, 2120–2129. [CrossRef]

### **Sample Availability:** Not available.

© 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* **Regulation of ZMYND8 to Treat Cancer**

**Yun Chen 1,2 , Ya-Hui Tsai 1,2, \* and Sheng-Hong Tseng 1,3, \***


**Abstract:** Zinc finger myeloid, nervy, and deformed epidermal autoregulatory factor 1-type containing 8 (Zinc finger MYND-type containing 8, ZMYND8) is a transcription factor, a histone H3 interacting protein, and a putative chromatin reader/effector that plays an essential role in regulating transcription during normal cellular growth. Mutations and altered expression of ZMYND8 are associated with the development and progression of cancer. Increased expression of ZMYND8 is linked to breast, prostate, colorectal, and cervical cancers. It exerts pro-oncogenic effects in breast and prostate cancers, and it promotes angiogenesis in zebrafish, as well as in breast and prostate cancers. In contrast, downregulation of ZMYND8 is also reported in breast, prostate, and nasopharyngeal cancers. ZMYND8 acts as a tumor suppressor in breast and prostate cancers, and it inhibits tumor growth by promoting differentiation; inhibiting proliferation, cell-cycle progression, invasiveness, and metastasis; and maintaining the epithelial phenotype in various types of cancers. These data together suggest that ZMYND8 is important in tumorigenesis; however, the existing data are contradictory. More studies are necessary to clarify the exact role of ZMYND8 in tumorigenesis. In the future, regulation of expression/activity of ZMYND8 and/or its binding partners may become useful in treating cancer.

**Keywords:** ZMYND8; tumorigenesis; epigenetic regulation; pro-oncogenic effects; tumor suppression

### **1. Characteristics and Functions of ZMYND8**

Zinc finger myeloid, nervy, and deformed epidermal autoregulatory factor 1-type containing 8 (Zinc finger MYND-type containing 8, ZMYND8) is a multifunctional transcription factor harboring conserved chromatin-binding module with affinity for chromatin [1]. It was initially identified as activated protein-kinase-C (PKC)-binding protein and is a member of the receptor for activated C-kinase (RACK) family proteins that anchor activated PKC and increase its phosphorylation and duration of inactivation; it is also called RACK7 [2,3]. ZMYND8 contains a Pro-Trp-Trp-Pro (PWWP) chromatin-binding domain, a bromodomain (BRD), a plant homeodomain (PHD) type zinc finger, and a MYND domain for protein–protein interaction (Figure 1) [3,4]. The PHD–BRD–PWWP (PBP) domains are histone readers, and the proteins containing PBP domains have various chromatin-related functions [5].

ZMYND8 is a core chromatin reader/effector, with distinct affinity for histone H3 and H4 [2,4–6]. The N-terminal PHD–BRD–PWWP domains of ZMYND8 can read several acetyl and methyl lysine residues on histones, including acetyl lysine 14 of H3 (H3K14ac), H4K16ac, and di- and tri-methyl lysine 36 of H3 (H3K36me2, H3K36me3) [6,7]. The PHD–BRD–PWWP domains of ZMYND8 form a stable structural reader ensemble and simultaneously engage histones and DNA, and then ZMYND8 is recruited to the transcriptional sites of the chromatin [7]. Mutation of the reader ensemble may affect the binding of ZMYND8 to histones by disrupting the interaction interface or destabilizing

**Citation:** Chen, Y.; Tsai, Y.-H.; Tseng, S.-H. Regulation of ZMYND8 to Treat Cancer. *Molecules* **2021**, *26*, 1083. https://doi.org/10.3390/ molecules26041083

Academic Editor: Marialuigia Fantacuzzi

Received: 18 January 2021 Accepted: 15 February 2021 Published: 18 February 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

the domain topology [7]. ZMYND8 has been found to regulate gene expression by recognizing dual histone marks [8]. It regulates the expression of all-trans retinoic acid (ATRA)-responsive genes through specific recognition of H3K36me2/H4K16ac [9,10]. It also recognizes H3K4me1/H3K14ac in DU154 and CWR22Rv1 prostate cancer cells [8], as well as H3K36me2/H4K16ac in SH-SY5Y neuroblastoma cells [4,9]. The dual recognition of two different histone modification by ZMYND8 suggests that the two separate conserved domains PWWP and BRD have different affinities towards their cognate histone binding partners [11]. When both H3K36me2 and H4K16ac exist in same histone octamer, the initial binding and recognition of the ZMYND8 to chromatin is considered through H3K36me2 because of its higher association rate, and the stability of the ZMYND8–nucleosome complex relies more on the binding to H4K16ac, due to its lower association rate [11].

**Figure 1.** Schematic representation of ZMYND8. ZMYND8 domains include the N-terminal PHD/BRD/PWWP reader cassette and a C-terminal MYND interacting domain. MYND: Myeloid, nervy, and deformed epidermal autoregulatory factor 1; ZMYND8: Zinc finger MYND-type containing 8; PHD: Plant homeodomain; BRD: Bromodomain; PWWP: Pro-Trp-Trp-Pro.

ZMYND8 is involved in transcription activation and in regulating transcription initiation through its interaction with the RNA polymerase II complex [4,12]. Through its putative coiled-coil domain, ZMYND8 forms a homodimer that preferentially associates with positive transcription elongation factor b (P-TEFb) complex, whereas the monomer associates with the chromodomain helicase DNA-binding protein 4 (CHD4) subunit of repressor nucleosome remodeling and deacetylase (NuRD) complex [4,12,13]. ZMYND8 and NuRD share a large number of genome-wide binding sites, mostly in active promoters and enhancers [14]. Both ZMYND8 and CHD4 modulate the expression of many genes and maintain genome integrity; silencing any one of them can alter global gene expression [4,12]. Silencing of ZMYND8 in HeLa cells increases the expression of 331 genes and decreases that of 438 genes [4]. ZMYND8 is important in modulating chromatin integrity and DNA repair [6,15,16]. Upon DNA double-strand break (DSB), histone modifications are altered to accommodate the DNA-damage signaling and the repair [6,15], and BRD2 protein and ZMYND8 are recruited to the DNA damages sites [17]. BRD2 occupies a spatially restricted region extending 2 kb either side of the DSB, and ZMYND8 spreads along the flanking chromatin [17]. The hyperacetylated chromatin domain is required for DBS repair, and the binding of BRD2 to H4ac protects the underlying acetylated chromatin from attack by histone deacetylases, whereas ZMYND8 is a repressor factor which limits transcription during DSB repair [17]. This creates a spatially restricted H4ac/BRD domain which facilitates DSB repair [17]. ZMYND8 interacts with various chromatin-remodeling complexes, histone demethylases/deacetylases, and acetyl transferases, including lysine demethylase 1A (KDM1A), KDM5A, KDM5C, and KDM5D, as well as histone acetyltransferase Tatinteractive protein-60KDa (TIP60) [4,6,8,12,16,18]. KDM5A-dependent demethylation is crucial for the binding of the ZMYND8–NuRD complex to chromatin and its recruitment to the locations of DNA damage (Figure 2) [15]. KDM5A causes H3K4me3 demethylation within chromatin, near the sites of DSB, while ZMYND8, NuRD complex, and KDM5A interact to repress transcription upon DNA damage [6,15]. KDM5A deficiency impairs

the transcriptional silencing and the repair of DSBs by homologous recombination [15]. ZMYND8 also interacts with the NuRD complex and TIP60, to mediate DNA repair through homologous recombination [6,16]. In Xenopus embryos, ZMYND8 interacts with RE1-silencing transcription factor corepressor 2 (RCOR2), and together they function as transcriptional repressors in regulating neural differentiation [19]. Both ZMYND8 and p53 play a role in DSB repair [6,15,16]. In several breast cancer cells with distinct p53 genotypes, ZMYND8 loss induced consistent micronucleus formation and DNA-damage response [20]. Additionally, in ZMYND8-depleted human U2OS osteosarcoma cells, laser micro-irradiation induced sustained p53 phosphorylation, which is a DSB marker; in contrast, there was no sustained p53 phosphorylation in control cells [6]. These results imply ZMYND8 and p53 may function independently for repair of DNA damage.

**Figure 2.** Diagram showing the KDM5A–ZMYND8–NuRD pathway during DNA damage. Upon DNA damage, the hisTable 5. A. ZMYND8 interacts with the NuRD complex and recognizes the TIP60-dependent acetylation. Then ZMYND8–NuRD complex binds to the DNA damaged site, inhibits transcription, and promotes DNA repair. TIP60: Histone acetyltransferase Tat-interactive protein-60KDa; NuRD: Nucleosome remodeling and deacetylase; KDM5A: Lysine demethylase 5A; H4K16ac: Acetyl lysine 16 of H4; H3K4me3: Tri-methyl lysine 4 of H3.

> As ZMYND8 is important for transcriptional regulation and chromatin integrity, it may have a role in oncogenesis. However, the reports regarding the influence of ZMYND8 on cancers are ambiguous [2,8,9,21–25]. In this review, we summarize the evidence for both pro-oncogenic and tumor-suppressive effects of ZMYND8 in various types of cancer. We conducted a PubMed literature search, using a combination of the following keywords and their variants: ZMYND8, RACK7, cancers, neoplasms, oncogenesis, carcinogenesis, and epigenetic regulation (up to November 30, 2020). The search covered all English articles listed in PubMed. The titles and abstracts of the identified articles concerning ZMYND8, RACK7, cancer, angiogenesis, proliferation, invasiveness, metastasis, and tumor growth were included. The selected articles were read in full, and further articles that were identified from their reference lists were also reviewed, to include studies that may have been missed in the initial search. A total of 39 references were thus included in the present review.

### **2. Association of ZMYND8 and Cancers**

Many epigenetic effectors, including ZMYND8, contain structurally conserved domains of PHD fingers, and alterations in the PHD finger-containing proteins are linked to cancer [4,6,26,27]. ZMYND8 is essential for regulating transcription during normal cellular growth and DNA repair, the perturbation of which may promote cancer initiation and progression [6,15,16]. In fact, ZMYND8 is a cutaneous T-cell lymphoma-associated antigen [28]. In addition, the ZMYND8-v-rel reticuloendotheliosis viral oncogene homolog A (avian) (RELA) chimeric transcripts were reported in a four-month-old patient with acute erythroid leukemia, a type of acute myeloid leukemia (AML) [29]. This fusion gene was thought to be a possible cause of constitutive activation of nuclear factor-kB in AML cells, since the RELA gene was under the control of the ZMYND8 promoter [29]. In breast cancer cells, ZMYND8 may be present as a fusion protein with centrosomal protein 250 (CEP250), which is required for centriole–centriole cohesion during interphase [3,30]. However, the ZMYND8-CEP250 fusion mRNA was not reported in the 111 breast cancer specimens studied [3]. Mutation frequency of *ZMYND8* was 19% in mismatch repair-deficient colorectal cancers [31]. In high-grade serous ovarian cancer, the *ZMYND8* gene is located within a region with recurrent alterations of somatic copy number [32]. An increased copy number (two to six copies) of *ZMYND8* is also found in DU-145, PC-3, LNCaP-FGC, BPH-1, and 22RV1 prostate cancer cells [2]. All of these findings indicate that ZMYND8 may be involved in development and progression of cancer.

### **3. ZMYND8 and Histone Modification in Cancer Cells**

The function of ZMYND8 in cancer cells is mainly through modulation of histone methylation and acetylation [4,9,11,25,32]. It selectively recognizes H3K36me2/H4K16ac and regulates all-trans retinoic acid (ATRA)-responsive genes in SH-SY5Y neuroblastoma cells [4]. In MDA-MB-231 breast cancer cells, ZMYND8 is recruited to its target genes by binding to H3K36me2 and H4K16Ac (Figure 3A) [11]. Furthermore, in HeLa and MCF7 breast cancer cells, ATRA induces an H3K27me3 to H3K27ac switch and upregulates ZMYND8 expression [9]. Modulation of histone methylation and acetylation in the enhancer regions by ZMYND8 is particularly important, as the dysregulation of this process may cause over-activation of transcription and contribute to tumorigenesis [25]. ZMYND8 and the KDM5 family cooperatively act on super-enhancer regions and are crucial regulators of expression and repression of oncogenes and tumor-suppressor genes in various types of cancer [8,12,16,18,25,33–35]. In ZR-75-30 breast cancer cells, ZMYND8 promotes the recruitment of KDM5C to the super-enhancer region, shown by the co-binding of ZMYND8 and KDM5C to 88.7% super-enhancers (Figure 3B) [25]. Ablation of either ZMYND8 or KDM5C in ZR-75-30 breast cancer cells results in over-activation of their target enhancers, characterized by the deposition of H3K4me3 and H3K27ac, decreased H3K4me1, and increased transcription of enhancer RNAs (eRNAs) and nearby genes [25]. In DU145 prostate cancer cells, in addition to KDM5C, ZMYND8 interacts with KDM5D, to act as transcriptional co-repressors, involved in regulating metastasis-linked genes (Figure 3C) [8]. ZMYND8 and KDM5D act as general negative regulators of enhancers in prostate cancer cells, and they antagonize the expression of these genes by recognizing the gene-activation-related dual-histone marker H3K4me1-H3K14ac [8].

**Figure 3.** Diagrams showing the mechanisms of the tumor-suppression effects of ZMYND8 and demethylases. The tumor-suppression effects may involve recognition of dual histone marks. (**A**) The all-trans retinoic acid (ATRA)-induced inhibition of transcription of proliferation genes and epithelial–mesenchymal transition (EMT). (**B**) ZMYND8 cooperates with KDM5C to suppress super-enhancer and transcription of oncogenes in breast cancer. (**C**) ZMYND8 cooperates with KDM5D to suppress transcription of metastasis genes in prostate cancer. PCNA: Proliferating cell nuclear antigen; VEGF: Vascular endothelial growth factor; EGFR: Epidermal growth factor receptor.

### **4. Tumor Suppression by ZMYND8**

The expression of ZMYND8 decreases in some cancers [10,25,36]. It is downregulated in invasive ductal and lobular breast cancer tissues, compared with normal tissues [10]. In addition, ZMYND8 expression is lower in breast cancer patients with invasive ductal carcinoma than in ductal carcinoma in situ [25]. In 190 patients with nasopharyngeal carcinoma, low ZMYND8 expression was correlated with late T stage, presence of lymph node metastasis, advanced stage, and poor overall patient survival [36]. ZMYND8 is a retinoic acid–inducible gene, and ATRA, a differentiation-inducing drug, can reprogram the epigenetic features of the upstream regulatory region of ZMYND8 and promote its expression [4,9]. On the other hand, ZMYND8 can facilitate the regulation of ATRA-responsive genes in SH-SY5Y neuroblastoma cells [4]. Differentiation of neuronal precursor cells induced by ATRA also requires transcriptional regulation, mediated by the ZMYND8–P-TEFb complex [13]. In MDA-MB-231 breast cancer cells, ZMYND8 upregulates differentiation genes and induces cellular differentiation [11]. The induction of breast cancer cell differentiation by ZMYND8 was noted through the H3K36me2/H4K16ac reader function of ZMYND8 [11]. Microarray analysis of breast cancer cells with ZMYND8 knockout by siRNA revealed that depletion of ZMYND8 reduces the expression of terminal differentiation markers, such as epithelial cell adhesion molecule (EPCAM) and cytokeratin 18 (CK18), by 80% and 86%, respectively [11]. In contrast, overexpression of ZMYND8 induces a 1.5-fold increase in EPCAM and CK18 levels [11]. In addition, ZMYND8 knockdown downregulates the stemness-related genes, prevents tumor cell differentiation, and maintains cancer cells in an undifferentiated state [11]. An in vivo study also showed that ZMYND8 overexpression significantly reduces the subcutaneous 4T1 murine breast cancer growth and increases the expression of differentiation-related genes, including *CK5*, *CK18*, *CK19*, and *EPCAM* in Balb/c mice [11]. All of these results suggest that, in cancer cells, ZMYND8 positively regulates the expression of differentiation-promoting genes and induces differentiation [11].

ZMYND8 also influences cancer cell proliferation. In HeLa and MCF7 breast cancer cells, knocking down ZMYND8 increases the proliferation by about two folds, whereas ZMYND8 overexpression reduces it by about 2.5 to 3 folds [10]. ZMYND8 can be directly recruited to proliferation-promoting genes, including *Ki67* and proliferating cell nuclear antigen (*PCNA*), and affect their expression [10,11]. ZMYND8 knockdown increases the expression of *Ki67* and *PCNA* by about 14 and 4 folds, respectively, in HeLa cells; about 8 and 3 folds in MCF7 breast cancer cells; and about 9 and 2.5 folds in T47D metastatic breast cancer cells [10]. In contrast, overexpression of ZMYND8 reduces the expression of *Ki67* and *PCNA* to 0.1–0.4 fold in these cells [10], and reduces the uptake of bromodeoxyuridine (BrdU) in HeLa cells [10]. However, deletion of ZMYND8 by short hairpin RNA (shRNA) does not affect the proliferation of DU145 prostate cancer cells significantly [8]. Knocking out ZMYND8 can enhance the tumor growth in a mammary fat pad xenograft model of ZR-75-30 breast cancer cells [25]. In contrast, invasive MCF-7 or 4T1 breast cancer cells overexpressing ZMYND8 show a reduction in the tumor size and tumor weight in mice, compared with the control [10,11]. ZMYND8 modulates cell-cycle progression, and its overexpression reduces the *Cyclin* genes, including *G1/S-specific cyclin-E1 (CCNE1)*, *CCNA2*, and *G2/mitotic-specific cyclin-B1 (CCNB1)* in HeLa cells; meanwhile, the inhibition of ZMYND8 by siRNA upregulates their expression [10]. However, in breast cancer cells, the expression of CDKN1A mRNA was upregulated by ZMYND8 loss, which suggests ZMYND8 depletion can increase the p21, which is an inhibitor of cell-cycle progression [20].

In addition to the suppression of cell proliferation, ZMYND8 also affects the tumor cell migration and invasion [8,10]. It can repress the expression of genes that promote metastasis and invasion, and enhance the transcription of epithelial genes [8,10]. In the wound-healing assay, ZMYND8 silencing causes faster wound closure, and its overexpression inhibits this process [10]. In a three-dimension-based assay, ZMYND8-null ZR-75-30 breast cancer cells show increased anchorage-independent growth, migration, and invasion, which can be reversed by restoration of ZMYND8 [25]. In the Matrigel invasion assay, ZMYND8 knockout results in a 1.6-2-fold increase in the invasiveness of HeLa and MCF7 breast cancer cells, and its overexpression reduces the invasiveness by 1.3–1.5 folds in these two cells [10]. Suppression of migration and invasion by ZMYND8 in breast cancer cells is through cooperation with KDM5C and modulation of SA100, as knockout of ZMYND8 or KDM5C can de-repress S100A [25].

The influence of ZMYND8 on cancer cell invasion and metastasis is related to the interaction between ZMYND8, TROJAN, and ZNF592 in breast cancer [24]. Both TROJAN and ZNF592 are the binding partners of ZMYND8, and the function of ZMYND8 is affected by the competitive binding of TROJAN and NF152 with it [24]. TROJAN, an endogenous retrovirus-derived long noncoding RNA, is upregulated in multiple cancer cell lines and is highly expressed in human triple-negative breast cancer [24]. ZNF592 can prevent ZMYND8 degradation, and TROJAN interferes with its ability to bind ZMYND8 [24]. In MDA-MB-231 LM2 breast cancer cells, TROJAN binds to ZMYND8, and increases its degradation through the ubiquitin-proteasome pathway, by repelling ZNF592 [24]. TROJAN knockdown increases ZMYND8 protein, while overexpression of TROJAN or knockdown of ZNF592 decreases it in MDA-MB231 LM2 cells [24]. Both TROJAN and ZMYND8 occupy epidermal growth factor receptor (EGFR), vascular endothelial growth factor-A (VEGF-A), and mouse double minute 2 homolog (MDM2) promoters, suggesting that they co-regulate these metastasis- and angiogenesis-related genes [24]. In an in vivo study, mice intravenously injected with MB-231 LM2 cells with TROJAN knockdown showed fewer metastatic lung nodules, as compared with the control, whereas mice injected with ZMYND8-inactivated cells showed more of them [24]. Increased metastatic ability of MDA-MB-231 LM2 breast cancer cells, induced by ZMYND8 knockdown, can be partially reversed by simultaneous knockdown of TROJAN [24]. In human breast cancer tissues, the expression of TROJAN negatively correlates with ZMYND8 expression, and it positively correlates with EGFR, VEGF-A, and MDM2 expression [24]. In addition, the TROJAN–ZMYND8 combined expression signature can be used to predict the relapse-free survival and overall survival of patients with triple-negative breast cancers (TNBCs); those with high TROJAN and low ZMYND8 expression have worse prognosis, and the ones with low TROJAN and high ZMYND8 levels survive better [24].

ZMYND8 also prevents metastasis of prostate cancer cells [8]. Knockdown of ZMYND8 increases the invasiveness of DU145 and CWR22Rv1 human prostate cancer cells in vitro, and this was also noted in an intravenous mouse xenograft model [8]. In mice injected with luciferase-expressing DU145 prostate cancer cells with ZMYND8 silencing by shRNA (shZMYND8), the luciferase signals were about five-fold higher than in those treated with the control shRNA, eight weeks after injection [8]. Histological examination of the lungs confirmed tumor development in the shZMYND8 group [8]. These results suggest that ZMYND8 suppresses invasiveness and metastasis of prostate cancer cells, both in vitro and in vivo [8]. The ZMYND8-induced suppression of invasiveness and metastasis in prostate cancer cells was demonstrated to be through cooperating with its transcriptional corepressor KDM5D [8]. They co-occupy multiple metastasis-linked genes, such as *SAIL2, CD44, VEGF*, and *EGFR*, and hinder the expression of these genes by interacting with the gene-activation-related dual-histone mark H3K4me1-H3K14ac [8]. Knockdown of either ZMYND8 or KDM5D can upregulate these metastasis-related genes [8]. KDM5D levels are significantly reduced in metastatic prostate tumors (100%, 6/6) and primary prostate tumors (41%, 28/68), as compared with normal tissues (10%, 2/21) [37]. These observations suggest that cooperation between ZMYND8-mediated recognition of H3K4me1-H3K14ac and KDM5D-catalyzed H3K4 demethylation can hamper the expression of metastasislinked genes [8,37]. H3K4me1-H3K14ac may act as a poised epigenetic signature, and it is converted to active dual marker H3K4me3-H3K14ac during cancer metastasis [8]. Therefore, the epigenetic switch is critical for regulating the expression of metastasislinked factors, including Slug, CD44, VEGFA, and EGFR, which, in turn, modulate the invasiveness of prostate cancer cells [8].

ZMYND8 can regulate the epithelial–mesenchymal transition (EMT), which is important for cellular invasion [10]. It regulates EMT genes through recognizing H3K36me2/ H4K16ac on respective genes [10]. The epithelial phenotype of cancer cells is governed by claudin-1 *(CLDN1),* claudin-7 *(CLDN7),* and E-cadherin *(CDH1)* genes, while the mesenchymal phenotype is regulated by *ZEB1*, *SNAI1*, *SNAI2*, and Vimentin *(VIM*) genes [10,38]. In MDA-MB-231 breast cancer cells, ZMYND8 is directly recruited to metastasis-linked genes *SNAI2* and *TWIST1*, to regulate EMT [11]. Overexpression of ZMYND8 reduces the expression of *VIM*, *SNAI1*, and *ZEB1* and increases that of *CLDN1* and *CDH1* in MCF7, HeLa, and T47D cells [10]. ZMYND8 knockdown reduces the expression of *CLDN1/7* and *CDH1* genes in MCF7 and HeLa cells, and that of *CDH1* and *CLDN7* in T47D cells, while it increases the expression of *VIM*, *SNAI1*, and *ZEB1* in MCF7, HeLa, and T47D cells [10]. These observations suggest that ZMYND8 is involved in maintaining the epithelial cell phenotype, and the deletion of ZMYND8 enhances the mesenchymal transition [10].

Overall, ZMYND8 appears to act as a tumor suppressor in breast, prostate, and nasopharyngeal cancers [8,9,24,25]. It can induce differentiation; inhibit cell proliferation, cell-cycle progression, invasiveness, and metastasis; and maintain the epithelial phenotype in these cells [8,9,24,25].

### **5. Pro-Oncogenic Effects of ZMYND8**

The level of ZMYND8 expression is crucial for proliferation and invasiveness of cancer cells, and increased ZMYND8 expression is reported in colorectal, cervical, breast, and prostate cancers [2,21,23,36]. Analyses of the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) database suggested that high expression of ZMYND8 was closely correlated with poor overall and disease-free survival in 174 colorectal cancer patients [23]. Increased ZMYND8 expression was linked to the high-grade cervical intraepithelial neoplasia and cervical carcinoma [21], zebrafish prostate cancer DU145 xenografts, and prostate cancer tissues from patients [2]. In addition, ZMYND8 expression is inversely correlated with metastasis-free survival in breast cancer patients, with high ZMYND8 expression being associated with shorter survival of patients with breast cancer [22,27]. However, increased ZMYND8 expression is associated with high mortality in patients with different subtypes and stages of breast cancer, but not with different grades of tumor [22]. ZMYND8 is a downstream target of estrogen receptor α (ERα), and it is involved in a positive feedback circuit of the ER pathway [27]. It is highly expressed in ER-positive and triple-negative breast cancers [22,27]. Amplification and overexpression of ZMYND8 are more frequent in luminal B breast cancer subtypes [22,27]. Tissue microarray analysis of 160 human TNBC specimens and 91 paired adjacent normal breast tissues revealed that 80% of tumors showed an increased expression of ZMYND8, with 50% showing moderate expression and 30% showing high expression [22]. In contrast, only 18% of the normal tissues had moderate ZMYND8 expression and none showed high expression [22]. An in vitro study also showed that ZMYND8 knockdown by small interfering RNA (siRNA) suppressed colony formation and reduced the number of MCF-7 breast cancer cell colonies; however, ZMYND8 knockdown did not affect the proliferation rate of the breast cancer cells [22]. In vivo experiments revealed that ZMYND8 increased the circulating breast cancer cells, to promote their extravasation and colonization, leading to lung metastasis in severe combined immunodeficiency (SCID) mice, whereas in ZMYND8 knockouts, these features were reversed [22].

ZMYND8 may induce angiogenesis in cancer, especially in hypoxic conditions (Figure 4). Under hypoxia, ZMYND8 expression was induced, together with hypoxia-inducible factor-1α (HIF1α) and VEGF, at 72 h post-fertilization in zebrafish [2]. In breast cancer cells, HIF-1 and HIF-2 can induce ZMYND8 expression, and HIFs and ZMYND8 co-activate oncogenes and increase RNA polymerase II phosphorylation, leading to the promotion of cell motility, tumor growth, and lung metastasis [22]. When exposed to hypoxic condition, with 1% O<sup>2</sup> for 12 days, the MDA-MB-231 breast cancer cells with ZMYND8-knockout showed decreased colony formation, migration, and invasion, compared to parental cells [22].

Furthermore, ZMYND8 can induce the expression of VEGFα mRNA and promote angiogenesis in prostate cancer xenografts in zebrafish and tube formation in human umbilical vascular endothelial cell cultures [2]. In zebrafish, ZMYND8 knockdown suppresses tumor angiogenesis in DU145 prostate cancer xenografts, and the re-introduction of ZMYND8 mRNA restores the tumor angiogenesis [2]. However, ZMYND8 overexpression does not increase VEGFα expression in DU145 xenografts, suggesting that ZMYND8 mainly acts on the surrounding zebrafish tissues, to regulate the expression of VEGFα, to induce angiogenesis [2]. ZMYND8 knockout also decreases the microvessel density in the mammary fat pad MDA-MB-231 tumor and subcutaneous MAF-7 tumor in SCID mice, which further supports the induction of angiogenesis by ZMYND8 [22]. In addition, transcriptome analyses reveal an increase in ZMYND8 expression during tumor angiogenesis in prostate cancer xenografts [2]. Furthermore, ZMYND8 is a binding partner of PKC, and PKC isozymes—in particular, PKCβ—are important mediators of VEGF signaling [2,3,39]. Inhibition of PKCβ causes decreased proliferation of endothelial cells and neovascularization of hepatocellular carcinoma, in a syngeneic xenograft model in BALB/c mice [39]. Thus, ZMYND8 can be induced in hypoxic conditions, and it can promote tumor angiogenesis, probably through HIF and VEGF activation, and enhance tumor growth [2,22]. β β

**Figure 4.** Diagram showing the possible mechanisms of the pro-oncogenic effects of ZMYND8. ZMYND8 can be activated by HIFs and induces angiogenesis, cellular proliferation, and tumor growth. The pro-oncogenic pathways of ZMYND8 are still unclarified. ZMYND8 may cooperate with demethylase, recognize the histone marks, suppress the transcription of tumor suppressor genes, or activate oncogenes, to enhance the tumorigenesis. HIFs: Hypoxia-inducible factors; KDM: Lysine demethylase.

> Overall, these results indicate that ZMYND8 may have a crucial role in the tumorigenesis of breast, prostate, colorectal, and cervical cancers [2,21–23]. In addition, ZMYND8 may promote angiogenesis in zebrafish, as well as breast and prostate cancers [2,22]. Therefore, ZMYND8 is important for tumor cell proliferation, angiogenesis, and tumor growth [23]. However, data supporting its oncogenic role are still limited and the pro-oncogenic pathways of ZMYND8 are still unclarified (Figure 4).

### **6. Conclusions**

Aberrant regulation of gene expression and epigenetic mutations play an important role in carcinogenesis. ZMYND8 is a transcription factor, a histone H3-interacting protein, and a putative chromatin reader/effector, important for regulating transcription in normal cells, and its mutation, altered expression, and fusion with other genes, are associated with development and progression of cancer. Increased expression of ZMYND8 is associated with breast, prostate, colorectal, and cervical cancers, and it is pro-oncogenic in breast and prostate cancers [2,21–23]. ZMYND8 is induced during hypoxic conditions, and can promote angiogenesis in zebrafish, as well as in breast and prostate cancers [2,22]. In contrast, lower expression of ZMYND8 is linked to breast, prostate, and nasopharyngeal cancers, and ZMYND8 exerts tumor-suppressive effects on breast and prostate cancers [8,9,24,25]. It suppresses tumor growth by promoting differentiation; inhibiting proliferation, progression of cell cycle, invasiveness, and metastasis; and maintaining the epithelial phenotype in various types of cancer [8,9,24,25]. Although all of these results indicate that ZMYND8 is important in cancer, they are contradictory, and the actual role of ZMYND8 in cancer is still unclear. In addition, the investigations are limited to a few types of cancer, like that of breast and prostate. As ZMYND8 interacts with various transcriptional corepressors, chromatin remodeling complexes, histone demethylase/deacetylase, and acetyl transferase enzymes, the role of ZMYND8 in cancers may be related to not just its direct action, but to multiple factors. Different cell types, fusion with other genes, and the interaction of ZMYND8 with various binding partners may affect the functions of ZMYND8 in cancers. The interaction between ZMYND8 and various binding partners is considered to be crucial in determining whether the function of ZMYND8 is pro-oncogenic or tumor suppressive (Figure 5).

**Figure 5.** Diagram showing the pro-oncogenic and tumor-suppressive effects of ZMYND8 in cancers. The functions of ZMYND8 were considered through the interaction between ZMYND8 and a specific ZMYND8-binding partner. The ↑ indicates enhancement, and ↓ indicates suppression.

↓

↑

β

For example, the interaction between ZMYND8 and PKCβ or TROJAN may induce pro-oncogenic effects, while interaction between ZMYND8 and KDM5C, KDM5D, or ZNF592 may cause tumor-suppressive effects. Therefore, more studies are necessary, and they need to be focused not only on ZMYND8, but also on its specific binding partners in different cancer types. Hopefully, in the future, regulation of the expression of ZMYND8 and/or its binding partners may become useful in treating cancer.

**Author Contributions:** Conceptualization, S.-H.T. and Y.C.; methodology, Y.C., Y.-H.T. and S.-H.T.; investigation, Y.C. and Y.-H.T.; writing—original draft preparation, Y.C. and Y.-H.T.; writing—review and editing, S.-H.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Science and Technology, R.O.C., grant numbers MOST 107-2314-B-002-074 and MOST 108-2314-B-002-087.

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

**Sample Availability:** Samples are not available.

### **References**


### *Article* **Comparison of Transcriptomic Profiles of MiaPaCa-2 Pancreatic Cancer Cells Treated with Different Statins**

**Silvie Rimpelová 1, \* , Michal Koláˇr 2 , Hynek Strnad 2 , Tomáš Ruml 1 , Libor Vítek 3 and Helena Gbelcová 4, \***


**Abstract:** Statins have been widely used for the treatment of hypercholesterolemia due to their ability to inhibit HMG-CoA reductase, the rate-limiting enzyme of de novo cholesterol synthesis, via the so-called mevalonate pathway. However, their inhibitory action also causes depletion of downstream intermediates of the pathway, resulting in the pleiotropic effects of statins, including the beneficial impact in the treatment of cancer. In our study, we compared the effect of all eight existing statins on the expression of genes, the products of which are implicated in cancer inhibition and suggested the molecular mechanisms of their action in epigenetic and posttranslational regulation, and in cell-cycle arrest, death, migration, or invasion of the cancer cells.

**Keywords:** statins; pancreatic cancer; DNA microarray; pitavastatin; cerivastatin; simvastatin; fluvastatin; atorvastatin; pravastatin; HMG-CoA reductase inhibitors

### **1. Introduction**

Statins (Figure 1) are the inhibitors of *de novo* cholesterol synthesis in the cell due to their ability to competitively inhibit the β-hydroxy-β-methylglutaryl coenzyme A (HMG-CoA) reductase. They represent the most prescribed drugs in the treatment of cardiovascular diseases [1]. Currently, there are eight existing statins, namely atorvastatin, cerivastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin, and simvastatin [2]. Although all statins have the same mechanism of hypolipidemic action, they differ in their chemical structure, physico-chemical properties, pharmacokinetic effects, and effects on lipid profile [3]. The inhibition of HMG-CoA conversion to mevalonate caused by HMG-CoA reductase inhibition results in the depletion of downstream intermediates of the mevalonate pathway.

Mevalonate is the precursor of, i.e., farnesyl pyrophosphate (FPP), geranylgeranyl pyrophosphate (GGPP), isopentenyl adenine, dolichol, and the polyisoprenoid side chains of heme A and ubiquinone, which are essential molecules that play a vital role in almost any cell process [3]. From these, FPP and GGPP play an important role in the posttranslational modification of cellular proteins involved in cell division and differentiation, gene expression, cytoskeleton formation, intracellular protein and lipid transport, and defense against pathogens [4]. Another intermediate in the mevalonate pathway, isopentenyl adenine, is essential for proper tRNA function and protein synthesis. Dolichol acts as an important scavenger of free radicals in cell membranes, and ubiquinone is involved in

**Citation:** Rimpelová, S.; Koláˇr, M.; Strnad, H.; Ruml, T.; Vítek, L.; Gbelcová, H. Comparison of Transcriptomic Profiles of MiaPaCa-2 Pancreatic Cancer Cells Treated with Different Statins. *Molecules* **2021**, *26*, 3528. https://doi.org/10.3390/ molecules26123528

Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso

Received: 16 March 2021 Accepted: 26 May 2021 Published: 9 June 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

mitochondrial respiration and inhibition of lipid peroxidation. A decrease in the intracellular level of ubiquinone leads to mitochondrial damage, oxidative stress, and cell damage. The involvement of the aforementioned mevalonate pathway intermediates in various cell processes explains the pleiotropic activity exhibited by statins, in addition to their hypolipidemic effect.

**Figure 1.** Chemical structures of statins.

Based on this, it is clear that statins significantly affect basic physiological processes of cells and organs that also are connected with oncogenesis [5]. The results of lipid-lowering therapy in animals initially indicated an increased risk of carcinogenesis [6]. However, the dose of statins administered in that study was very high and not applicable for humans [7]. The results of studies on tissue-specific cancer development in individuals on statins therapy (as hypolipidemic drugs) have been controversial. An increased incidence of breast cancer after statin administration was observed in one study [8], but another study did not confirm these findings [9]. Contrary to that, a reduced incidence of melanomas has been reported after statin administration [9]. A similar beneficial effect of statins was observed in connection with reduced incidence of colon carcinomas [10]. However, most of these studies were not originally designed to investigate the relationship between statin intake and cancer development, but rather between statins and cardiovascular diseases. Indeed, some reports have shown a statistically significantly lower incidence of cancer in patients receiving statin therapy, despite relatively short follow-up times and inappropriate patient selection [11–16]. Moreover, extensive studies conducted on 500,000 U.S. war veterans have shown that statin use is associated with a two- to fivefold lower incidence of lung [17], breast [18], and prostate cancers [19]. Based on other meta-analyses, statins appear to be particularly effective in the chemoprevention of colorectal cancer [10,20]. While there have been several in vitro and in vivo studies directly targeting and confirming the marked effect of statins on the growth of a wide variety of tumor types, such as hepatocellular [21], lung [22], and colorectal carcinoma [23], clinical trials targeting statin efficacy in cancer treatment have been very rare, so far. One of them showed that patients

suffering from hepatocellular carcinoma exhibited statistically significantly longer survival rates for pravastatin therapy [24] than untreated. On the other hand, beneficial effects of lovastatin therapy on the survival rate of patients suffering from glioblastoma multiforme or advanced gastric adenocarcinoma have not been observed [25,26]. The findings of the meta-analyses suggested an association between pravastatin treatment and cancer in elderly patients [27] but did not support the potential role of statins in the prevention of hematological malignancies [28] or the hypothesis that statins reduce the risk of pancreatic cancer development [29].

The setting of the experimental conditions of the present study was based on the results of our previous published studies [30,31], in which we reported on the differences in the individual anticancer potential of individual statins in cells derived from pancreatic carcinoma. Moreover, we were inspired by the aforementioned findings, as well as recent reviews shedding light on statins as cancer inhibitors with several possible mechanisms of action [32–35]. Therefore, using DNA microarray analysis, we sought after statininduced changes in the expression of genes, the products of which are connected with tumorigenesis, inhibition of tumor growth, or metastasis. We focused on the pancreatic cancer cells because, in the very recent meta-analysis of 26 studies containing more than 3 million participants and 170,000 pancreatic cancer patients, statin therapy was shown to significantly reduce the risk of pancreatic cancer [36].

### **2. Results**

From eight existing statins, only six (cerivastatin, pitavastatin, simvastatin, lovastatin, fluvastatin, and atorvastatin) significantly affected the gene expression in the MiaPaCa-2 pancreatic cancer cell line after 24 h of treatment with 12 µM statin concentration. The concentration was selected based on our previous studies [30,31], in which we determined the IC<sup>50</sup> of individual statins in cells from pancreatic carcinoma after 24, 48, and 72 h. The value 12 µM corresponds to the IC<sup>50</sup> of simvastatin in MiaPaCa-2 cells after 24 h of treatment. Other statins were compared at the same concentration since simvastatin was at the time of analysis one of the most potent clinically used statins (cerivastatin had been withdrawn from the market at the time of analysis).

### *2.1. Effect of Statins on Lipid Metabolism and Synthesis of Steroids*

Lipids, particularly cholesterol, and their derivatives play several roles in tumorigenesis and cancer cell metabolism [37,38]. In contrast to noncancerous cells, malignant cells exhibit enhanced *de novo* synthesis of fatty acids, which serve as substrates for β-oxidation, or conversion to triglycerides for storage of phospholipids for membrane formation.

DNA microarray analysis showed that out of the eight statins evaluated, six effective ones significantly affected steroid biosynthesis in MiaPaCa-2 cells. They upregulated the gene encoding HMG-CoA reductase, representing the key target of statins. In addition, they even more markedly triggered upregulation of the gene encoding HMG-CoA synthase, catalyzing the synthesis of HMG-CoA, the substrate for the aforementioned enzyme covalently inhibited by statins. The inhibition of the mevalonate pathway resulted in a switch from cholesterol synthesis to the triacylglyceroles and phospholipid synthesis via the Kennedy pathway. Genes encoding enzymes of the Kennedy pathway were directly upregulated by the action of only the most effective statins; i.e., cerivastatin, pitavastatin, and simvastatin (Table 1). However, in terms of knowledge of the lipid metabolism of tumors, this observation indicated a procarcinogenic effect rather than an antitumor effect of statins. The changes in the expression of the genes involved in the mevalonate pathway are shown in Figure 2.



**Figure 2.** Mevalonate pathway and the changes induced by statin treatment in the expression of the genes that code for enzymes of the pathway. The color fill of the nodes indicates the base-2 logarithm of the fold change in gene expression upon treatment by a statin. Different statins are shown in the distinct position of the node as indicated in the bottom key.

On the other hand, statins are responsible for a decrease in the cholesterol content in lipid rafts by inhibiting the mevalonate pathway and, thus, induce apoptosis via inhibition of the Akt signaling pathway [39]. Interestingly, in our study, the Akt signaling pathway was not affected by the evaluated statins at 12 µM after 24 h treatment of MiaPaCa-2 cells.

In tumors with deregulated Hedgehog signaling, the depletion of cholesterol results in impairment of Hedgehog signal transduction and inhibition of cancer cell growth [40]. Cholesterol itself serves as a substrate for the post-translational modification of Hedgehog ligands, which is required for their proper trafficking [41]. This means that all statins should indirectly inhibit the Hedgehog signaling by inhibiting the cholesterol synthesis. Moreover, we found that pitavastatin and cerivastatin significantly upregulated the *GAS1* (growth arrest-specific gene 1) (data not shown), the product of which positively regulates the Hedgehog signaling [42,43]. The possible explanation for *GAS1* gene upregulation after statin treatment was the effort of affected cells to reactivate the statins-suppressed Hedgehog signaling.

Finally, to maintain cancer cell proliferation, the increased activation of SREBPs (sterol regulatory element-binding proteins) is required [44]; therefore, SREBP inhibitors are used in molecular-targeted cancer therapies [45,46]. SREBPs are transcription factors that regulate the expression of genes required for the synthesis of fatty acids, triglycerides, and cholesterol. Intracellular cholesterol level is controlled by SREBP-1 and SREBP-2. These transcription factors upregulate the synthesis of enzymes involved in the sterol biosynthesis upon binding to specific sterol regulatory element DNA sequences. Sterols in turn inhibit the cleavage of SREBPs, and therefore, synthesis of additional sterols is reduced via a negative feedback loop. Statin-mediated inhibition of sterol synthesis was reported to activate the SREBPs in many tumor types [45]. In our experiment, statins at a concentration (12 µM) sufficient to induce a more or less intensive antiproliferative effect on MiaPaCa-2 pancreatic cancer cells (IC50: cerivastatin 10 µM, simvastatin 12 µM, lovastatin 13 µM, pitavastatin 20 µM, fluvastatin 26 µM, atorvastatin 27 µM, pravastatin 29 µM, rosuvastatin 36 µM) did not significantly affect the expression of genes encoding SREBPs (data not shown).

### *2.2. Statins in the Role of Epigenetic Regulators*

Recently, it was described that acetyl-CoA metabolism supports multistep pancreatic tumorigenesis. In pancreatic adenocarcinoma, both upregulated gene expression of the mevalonate pathway and histone acetylation was detected [47]. Dysregulation between activities of histone acetyltransferases (HATs) and histone deacetylases (HDACs) is frequent in human tumors [48]. The expression of genes encoding HATs or HDACs was not altered by statins in our experiments (data not shown); however, genes encoding histone H4 were downregulated, in contrast to genes for histone H2B that were upregulated by cerivastatin, pitavastatin, and simvastatin (Table 2). STRING enrichment analysis of epigenetic regulators significantly affected by statin treatment is shown in Figure 3.

**Table 2.** Epigenetic regulator genes that were affected by statin treatment of MiaPaCa-2 cells. Numeric columns display fold changes of the gene expression between cells treated by individual statins at 12 µM concentration for 24 h and untreated cells. Cer.—cerivastatin; Pit.—pitavastatin; Sim.—simvastatin; Flu.—fluvastatin. Atorvastatin, lovastatin, pravastatin, and rosuvastatin did not induce any significant change in MiaPaCa-2 cells' gene expression.


**Figure 3.** STRING enrichment analysis of epigenetic regulators significantly affected by statin treatment (mainly by cerivastatin and pitavastatin) of MiaPaCa-2 cells at 12 µM concentration for 24 h. Individual nodes represent affected gene products and their interactions. Input nodes represent the genes listed in Table 2 (*HIST1H4C, HIST1H2BF, HIST2H3C, H2AFX, HIST2H2BE, HIST1H2BD, HIST1H2BK,* and *HIST3H2A*). The evidence view of the association network was generated according to the known and predicted interactions in *Homo sapiens*. Known interactions: turquoise—from curated databases; violet—experimentally determined. Predicted interactions: green—gene neighborhood; red—gene fusions; blue—gene co-occurrence. Others: grass green—text-mining; black—co-expression; light blue—protein homology. A complete description of the nodes is given in Supplementary Table S1.

### *2.3. Statins and Their Potential Role as Posttranslational Regulators*

Statins deplete the cellular pool of isoprene precursors, thereby having an impact on protein prenylation; i.e., farnesylation and geranylgeranylation. Among prenylated proteins, the low-molecular-weight guanosine triphosphate-binding proteins Ras and Ras-related growth modulators were monitored (Table 3). The overall effect on the RAS signaling pathway is shown in Figure 4.

**Table 3.** Genes involved in posttranslational regulation affected by statin treatment of MiaPaCa-2 cells. Numeric columns display fold changes of the gene expression between cells treated by individual statins at 12 µM concentration for 24 h and untreated cells. Cer.—cerivastatin; Pit.—pitavastatin; Sim.—simvastatin; Flu.—fluvastatin; Ato.—atorvastatin; Lov. lovastatin. Pravastatin and rosuvastatin did not induce any significant change in MiaPaCa-2 cell gene expression.

µ


**Figure 4.** RAS signaling pathway and the changes induced by statin treatment in the expression of the genes that code for members of the pathway. The color fill of the nodes indicates the base-2 logarithm of the fold change in gene expression upon treatment by a statin. Different statins are shown in the distinct position of the node as indicated in the key.

*RHOB*, the most dramatically upregulated gene by statins (Table 3), belongs to the Rho protein family of proteins regulating diverse cellular processes, including cytoskeletal organization, gene transcription, cell cycle progression, and cytokinesis [49,50]. While most Rho proteins have been shown to have positive roles in proliferation and malignant transforma-

tion, RhoB rather appears to act as a negative regulator of these processes [51,52]. R-Ras promotes the formation of focal adhesions, cell spreading, and activation of integrins [53]. Cerivastatin, pitavastatin, simvastatin, and fluvastatin also increased the expression of the *KRAS* gene, the most frequently mutated gene in pancreatic cancer, and frequently mutated in cancer in general. As the K-Ras is the protein related to many functional pathways, such as the mitogen-activated protein kinase (MAPK) signaling pathway, the receptor tyrosineprotein kinase (ErbB) signaling pathway, dorso-ventral axis formation, axon guidance, the vascular endothelial growth factor (VEGF) signaling pathway, tight junctions, gap junctions, natural killer cell-mediated cytotoxicity, the T-cell receptor signaling pathway, the B-cell receptor signaling pathway, the Fc epsilon RI signaling pathway, long-term potentiation, long-term depression, regulation of actin cytoskeleton, the insulin signaling pathway, and the gonadotropin-releasing hormone (GnRH) signaling pathway, the pleiotropic effect of statins is comprehensible. Moreover, the *KRAS* gene upregulation after statin treatment adverts to the unavailability of K-Ras protein for the cell signaling without its correct posttranslational modification due to the mevalonate pathway inhibition. Similarly, this mechanism could also explain the upregulation of other Ras and Ras-related proteins induced by treatment with simvastatin. Genes encoding guanine nucleotide exchange factor (GEF), GTPase-activating proteins (GAP), or guanosine nucleotide dissociation inhibitor (GDI) of Ras and Ras-related proteins were not significantly up- or downregulated by statins.

### *2.4. Statins' Effect on Cell Cycle and Cell Death*

The effect of statins on the expression of genes related to the cell cycle and cell death is of special interest concerning their cancerostatic capability. The expression levels of genes associated with DNA replication were significantly affected only by pitavastatin, cerivastatin, and simvastatin, and they were mostly downregulated (Table 4). The changes in the cell-cycle circuit are shown in Figure 5.

The origin recognition complex (ORC) is a highly conserved six-subunit protein complex essential for the initiation of DNA replication in eukaryotic cells. The ORC binds specifically to origins of replication and serves as a platform for the assembly of additional initiation factors such as minichromosome maintenance (MCM) proteins. ORC1L is the largest subunit of the ORC complex. While the levels of other ORC subunits are stable throughout the whole cell cycle, the level of ORC1L changes in various phases of the cell cycle. These changes are controlled by ubiquitin-mediated proteolysis after the initiation of DNA replication [54]. From this, it seems that statins blocked the progression of the cell cycle through the S phase. From the quantum of genes encoding proteins involved in the cell-cycle regulation, such as cyclins, cyclin-dependent kinases (CDK), and cell-cycle negative regulators, only the expression of those denoted in Table 4 were affected by statins. CDC2 (CDK1) is the catalytic subunit of a highly conserved protein kinase complex known as the M-phase promoting factor (MPF), which is essential for the G1/S and G2/M phase transitions in the cell cycle of eukaryotic cells. Mitotic cyclins stably associate with this protein and function as regulatory subunits. The kinase activity of this protein is controlled by cyclin accumulation and degradation through the cell cycle [55].

M-phase inducer phosphatase 2 (CDC25B), a member of the cell division control protein 25 (CDC25) family of phosphatases, activates the cyclin-dependent kinase (CDC2) and is required for entry into the mitosis [56]. Whereas the *CDC2* gene was downregulated by simvastatin treatment of MiaPaCa-2 cells, the *CDC25B* gene was upregulated (Table 4). Some other genes encoding proteins related to the S phase (SKP2, E2F2) were downregulated by cerivastatin and pitavastatin. The SKP2 (S-phase kinase-associated protein 2) is an essential element of the cyclin A/CDK2 S-phase kinase [57]. E2F2 is a member of the E2F family of transcription factors and plays a crucial role in the control of the cell cycle. Expression of the S-phase genes is not activated when E2F is repressed [58].

**Table 4.** Genes involved in the cell cycle and DNA replication affected by statin treatment of MiaPaCa-2 cells. Numeric columns display fold changes of the gene expression between cells treated by individual statins at 12 µM concentration for 24 h and untreated cells. Cer.—cerivastatin; Pit.—pitavastatin; Sim.—simvastatin; Flu.—fluvastatin; Ato.—atorvastatin; Lov.—lovastatin. Pravastatin and rosuvastatin did not induce any significant change in MiaPaCa-2 cell gene expression.


In our experimental setup, one of the genes most affected by statin treatment was *TNFRSF10D* (Table 5). Its product, the tumor necrosis factor receptor superfamily member 10D precursor (known also as TNF-related apoptosis-inducing ligand receptor 4, or TRAIL receptor 4) is a member of the TNF-receptor superfamily, which is closely connected with apoptosis. TNFRSF10D has been shown to play an inhibitory role in TRAIL-induced cell apoptosis [59]. Upregulation of another gene, *GABARAPL* (also known as early estrogenregulated protein; Table 5) suggests that statin treatment may induce cell death by au-

tophagy [53]. STRING enrichment analysis of products of genes involved in cell death significantly affected by statin treatment is shown in Figure 6.

**Figure 5.** Cell-cycle signalization and the changes induced by statin treatment in the expression of the genes that code for **Figure 5.** Cell-cycle signalization and the changes induced by statin treatment in the expression of the genes that code for members of the circuit. The color fill of the nodes indicates the base-2 logarithm of the fold change in gene expression upon treatment by a statin. Different statins are shown in the distinct position of the node as indicated in the key.

**Table 5.** Genes involved in cell death affected by statin treatment of MiaPaCa-2 cells. Numeric columns display fold changes of the gene expression between cells treated by individual statins at 12 µM concentration for 24 h and untreated cells. Cer.—cerivastatin; Pit.—pitavastatin; Sim.—simvastatin; Flu.—fluvastatin; Ato.—atorvastatin; Lov.—lovastatin. Pravastatin and rosuvastatin did not induce any significant change in MiaPaCa-2 cell gene expression.


µ **Figure 6.** STRING enrichment analysis of products of genes involved in cell death significantly affected by statin treatment (mainly by cerivastatin and pitavastatin) of MiaPaCa-2 cells at 12 µM concentration for 24 h. Individual nodes represent affected gene products and their interactions. Input nodes were based on the genes listed in Table 5 (*TNFRSF10D, SLC6A12, DRAM, CASP9, GABARAPL1, ATG2A, TNFAIP1, CARD10,* and *TNFRSF6B*). A complete description of the nodes is given in Supplementary Table S1. See Figure 3's caption for color coding of the edges.

### *2.5. Statins' Effect on Migration and Invasion of Cancer Cells*

Since statins are known to affect migration and invasion of various cancer cells [60–64], which is also in accordance with our preliminary unpublished data, we also concentrated on the analyses of changes in the transcription of genes involved in cell migration and cytoskeleton architecture. The pancreatic cancer cells used in our study were firmly attached to the cultivation surface. However, the statin treatment induced a change in the cell shape and facilitated their detachment (data not shown). This is consistent with the effect of statins on the expression of the genes encoding the cytoskeletal proteins (Table 6). STRING enrichment analysis of products of genes involved in cytoskeleton maintenance significantly affected by statin treatment is shown in Figure 7.

µ Intermediate filaments (IF) of the cytoplasmic cytoskeleton are composed of keratins that are the major structural proteins of epithelial cells. They interact with desmosomes and hemidesmosomes, by which they assist in cell–cell adhesion [65].

TNNT1 is a subunit of troponin, which is a regulatory complex located on the thin filament of the sarcomere. This complex regulates striated muscle contraction in response to fluctuations in intracellular calcium concentration [66]. Gelsolin is an actin-binding protein that is a key regulator of actin filament assembly and disassembly. Among the lipid-binding actin regulatory proteins, gelsolin is one of the few that exhibit preferential binding toward polyphosphoinositide. The activity of gelsolin is stimulated by calcium ions (Ca2+) [67]. Kinesins belong to a class of motor proteins found in eukaryotic cells. Kinesins move along microtubule cables powered by the dephosphorylation of adenosine triphosphate (ATP). The active movement of kinesins supports several cellular functions, including mitosis,

meiosis, and transport of cargo [68]. Catenins are proteins found in complexes with cadherin cell adhesion molecules of animal cells. Junction plakoglobin (catenin gamma) was originally identified as a component of desmosomes (a cell structure specialized for cell-to-cell adhesion) [69]. Ezrin is the cytoplasmic peripheral membrane protein that serves as an intermediate between the plasma membrane and the actin cytoskeleton. This protein plays a key role in cell-surface structure adhesion, migration, and organization, and it has been implicated in various human cancers [70]. CDH10 is an integral membrane protein that mediates calcium-dependent cell–cell adhesion [71].

**Table 6.** Genes involved in cytoskeleton maintenance affected by statin treatment of MiaPaCa-2 cells. Numeric columns display fold changes of the gene expression between cells treated by individual statins at 12 µM concentration for 24 h and untreated cells. Cer.—cerivastatin; Pit.—pitavastatin; Sim.—simvastatin; Flu.—fluvastatin; Ato.—atorvastatin; Lov. lovastatin. Pravastatin and rosuvastatin did not induce any significant change in MiaPaCa-2 cell gene expression.


µ **Figure 7.** STRING enrichment analysis of products of genes involved in cytoskeleton maintenance significantly affected by statin treatment (mainly by cerivastatin, pitavastatin, and simvastatin) of MiaPaCa-2 cells at 12 µM concentration for 24 h. Individual nodes represent affected gene products and their interactions. Input nodes were based on the genes listed in Table 6 (*LOC400578, KRT16, MGC102966, KRT15, KLHL24, JUP, TUBB1, KRT13, KRT19, GSN, KIF1A, TNNT1, ITGB4, VIL2, NAV1, GSN, CDH10,* and *SYNM*). A complete description of the nodes is given in Supplementary Table S1. See Figure 3's caption for color coding of the edges.

### **3. Materials and Methods**

### *3.1. DNA Microarray Analysis*

Description of an experimental method for the studying of the effects of individual statins on the gene-expression profile of MiaPaCa-2 cells was reported by Gbelcová et al. [30]. We compared the effect of statins to simvastatin because it was chosen as the most effective clinically used statin tested *in vitro* in our previous study. Briefly, pure forms (≥98 %) of all commercially available statins were used: atorvastatin, lovastatin, simvastatin, fluvastatin, cerivastatin, pravastatin, rosuvastatin, and pitavastatin (LKT Laboratories, USA). Statins were dissolved in methanol and tested in 12 µM concentrations, representing the IC<sup>50</sup> value for simvastatin after 24 h of treatment of MiaPaCa-2 cancer cells. Human pancreatic cancer cells MiaPaCa-2 (ATCC, Manassas, VA) were cultured in DMEM medium (Sigma Aldrich, Germany) supplemented with 10 % fetal bovine serum. Illumina HumanWG-6\_V3 chips (Illumina, USA) were used for the microarray analysis following the standard protocol. Annotation of differentially expressed transcripts was done with R/BioConductor packages [72] against the Ensembl database (version 47) [73]. The transcripts with a false discovery rate smaller than 0.05 and fold change smaller than 0.5 or greater than 2 were reported and used in the downstream analysis. Changes in the gene expression were

visualized using the Pathview package [74] on the pathways provided by the KEGG database [75].

### *3.2. RT-qPCR Analysis*

Validation of the selected gene expression changes was performed using quantitative RT-PCR. The amount of 1 × 10<sup>8</sup> cells was washed with phosphate-buffered saline and lysed using the RLT buffer (Qiagen, Hilden, Germany) supplemented with β-mercaptoethanol. The total RNA was extracted by the RNeasy Micro kit (Qiagen) according to the manufacturer's protocol. All extracts were treated with DNase I (Qiagen) to remove contaminating genomic DNA. The quality and quantity of the RNA were evaluated with an Agilent 2100 Bioanalyzer instrument using the RNA 6000 Nano kit (both Agilent Technologies, Santa Clara, CA, USA). Expression levels of the mRNA were determined using a two-step RT-qPCR method. First, cDNA was reverse-transcribed from 1 µg of total RNA in a final reaction volume of 20 µL using a QuantiTect reverse transcription kit (Qiagen) according to the manufacturer's instructions. Then, the cDNA concentration was quantified using a LightCycler 480 instrument in LightCycler 480 SYBR Green I master mix (both Roche Applied Sciences, Penzberg, Germany) with a custom primer mix (see Supplementary Table S1 for primer sequences). For each condition, two biological replicates were analyzed (three for the control group).

Data analysis was performed using the delta–delta Cq method [76] within the R statistical environment [77]. Cq values were truncated at the value of 40. Three housekeeping genes (RPS9, TBP, and GAPDH) were used as reference genes for RNA quantity normalization using the geNorm algorithm [78]. Relative expression levels were computed assuming the perfect efficiency of the PCR.

### *3.3. STRING Analysis*

A functional association network of known and predicted functional partners or selected genes identified as significantly changed in the microarray analysis in MiaPaCa-2 cells treated with 12 µM concentration of statins for 24 h was created using STRING database 11.0 [79]. STRING is an interaction network database for functional enrichment analysis of protein–protein interaction networks. Input nodes were chosen based on the most important hits identified by the microarray analysis and depicted in Figure 3, Figure 6, and Figure 7. The evidence view diagrams were generated by STRING to illustrate the known protein–protein interactions of all connected nodes. The view of the association network was done for *Homo sapiens* according to the known and predicted from curated databases, experimentally determined, gene neighborhood, gene fusions, gene co-occurrence, text-mining, co-expression, and protein homology. The confidence score was set to high, equal to 0.700, with a maximum of 30 interactions. Disconnected nodes in the networks were hidden.

### **4. Discussion and Conclusions**

Statins have been intensively studied drugs based on their deep impact on the human organism caused, in particular by their cholesterol-lowering activity. However, their effects also include remarkable and potentially clinically relevant antitumor effects.

Although lipids are not genetically encoded, the genome changes can reflect cholesterol homeostasis indirectly. The changes in gene expression following statin treatment studied by microarray technology have been reported since the year 2000. The interpretation of our microarray analysis indicates a significant correlation between our and previously reported results. On the other hand, many differences are resulting from distinct experimental conditions due to the type of statin used, its concentration and duration of its activity, the experimental model (the type of a cell line or organism), etc.

Previously, we have demonstrated substantial differences in cancer cell antiproliferative effects of all commercially available statins in an experimental model of human pancreatic cancer [31]. Various statins exhibited significantly different inhibitory efficacy,

and we have also observed notable differences in statin sensitivity between pancreatic cancer cell lines, concerning the level of differentiation and harboring of G12C-activating mutation in the *KRAS* gene. Cerivastatin, pitavastatin, and simvastatin were the most effective, whereas rosuvastatin and pravastatin were the least effective, and their effectiveness correlated with the properties of the cell lines tested (the level of differentiation, and presence of G12C *KRAS* mutation) [31]. In the present study, we observed that depletion of farnesylated K-Ras protein caused by the three most effective statins led to significant upregulation of gene-encoding K-Ras.

The pleiotropic effect of statins is comprehensible, since K-Ras is related to many functional pathways, such as the MAPK signaling pathway, the ErbB signaling pathway, dorso-ventral axis formation, axon guidance, the VEGF signaling pathway, tight junction, gap junction, natural killer cell-mediated cytotoxicity, the T-cell receptor signaling pathway, the B-cell receptor signaling pathway, the Fc epsilon RI signaling pathway, long-term potentiation, long-term depression, regulation of actin cytoskeleton, the insulin signaling pathway, and the GnRH signaling pathway. Moreover, the upregulation of KRAS after the statin treatment seemed to be a result of the unavailability of farnesylated K-Ras protein for cell signaling due to the mevalonate pathway inhibition. Similarly, this mechanism could also explain the upregulation of other Ras and Ras-related proteins induced by statin treatment (Table 3). However, the products of the mevalonate pathway are required not only for post-translational modifications of many proteins, but also for regulation of the transcription of many proteins, including proteins of the cytoskeleton. For example, expression of keratin 13 is known to be regulated by nuclear receptor ligands such as retinoids, 1α, 25-dihydroxy vitamin D3, or estrogen [65]. Except for the metabolism of lipids (biosynthesis of steroids and sphingolipid metabolism), the inhibition of the mevalonate pathway by statins directly affected the energy metabolism due to ubiquinone depletion, as ubiquinone is involved in mitochondrial respiration [80].

Genetic heterogeneity is an important factor that affects the sensitivity of particular cancer cells to the antiproliferative/proapoptotic effects of statins. Despite the level of cancer cell differentiation, pancreatic cancer cell lines harboring activating K-Ras mutation were more sensitive to the antiproliferative effect of statins than cells harboring wild-type K-Ras [31]. Similarly, Wong et al. demonstrated that only 8 of 17 multiple myeloma cell lines evaluated were sensitive to lovastatin-induced apoptosis, while resistant cell lines had different genetic profiles [81].

The first effect of statins, specifically simvastatin, studied by microarray analysis was focused on actomyosin contraction, gap formation, and barrier dysfunction produced by thrombin. The experiment was performed in human pulmonary artery endothelial cells (EC) treated with 5 µM simvastatin for 24 h. Several genes related to thrombinmediated cytoskeletal dynamics and barrier regulation, including caldesmon and the thrombin receptor PAR-1, were dramatically downregulated. In addition, ITGB4, a protein known to be involved in cell–cell adhesion, was dramatically upregulated. *RhoA* and *RhoC* genes were also upregulated similarly to Rac1 and specific GEFs, potential regulators of preferential Rho GTPase activity. In contrast, the RhoGDP dissociation inhibitor was downregulated, which was interpreted as a compensatory response [82]. Consistent with these data, our results also indicated that statins directly affected the expression of specific genes related to the Rho GTPase signaling and cytoskeletal regulation. However, from all the Rho family members, only *RhoB* was significantly upregulated by all effective statins (pravastatin and rosuvastatin were not effective in our microarray study), and *RhoA* was significantly upregulated only by the most effective statins (cerivastatin, pitavastatin, and simvastatin). No changes were observed in the expression of genes encoding specific GEFs. Also, downregulation of caldesmon or the thrombin receptor PAR-1 was not observed in our study. This could vary in different types of cell lines evaluated.

Interestingly, a gene encoding the integrin beta 4 subunit of a receptor for the laminins; i.e., *Itgb4*, was the most significantly upregulated gene (fold change—7.57) of all the tested genes [83]. The reason for the enhanced transcription of a gene involved in the biology of

invasive carcinoma is unclear. Another notable finding is that no changes were observed in the expression of the *RhoB* gene after treatment of EC with 5 µM simvastatin [84], whereas in our study, the *RhoB* gene was the more affected gene compared to the Itgb4 gene after treatment with all effective statins.

In the other study, the DNA microarrays were used to identify gene expression patterns in the cerebral cortex of mice treated with simvastatin (50 mg/kg b.wt.) by daily oral doses for 21 days. The maximum average concentration of simvastatin was determined as 600 pmol/g in brains. This study revealed the influence of simvastatin on the expression of several genes involved in cell growth and signaling. *C-fos*, *c–myc*, and *Bcl-2* were of particular interest due to the linkage of simvastatin with cell growth and apoptosis. Simvastatin significantly reduced the expression of the proto-oncogene *c-fos*. On the other hand, it significantly increased the expression of the oncogene *c-myc* and antiapoptotic gene *Bcl-2* [84,85]. In contrast to these data, neither the expression of previously mentioned genes nor of other genes attributed to apoptotic cell death (e.g., genes encoding death receptors or caspases, the only exception to which was caspase 9, which was upregulated by cerivastatin) were affected by statins in our microarray analysis. However, as demonstrated by the analysis of the most affected functional pathways, statins affected some processes related to DNA repair, such as base excision repair or mismatch repair [30]. It is known that failure of these processes could be followed by programmed necrosis [86]. Finally, the upregulation of the *GABARAPL* gene related to autophagy was observed (Table 5). Autophagy does not always result in programmed cell death, but it represents the important mechanism for catabolic production of ATP during nutrient stress, and also plays an important role in the turnover of proteins and organelles under nutrientreplete conditions [87]. The statin-treated cells were in nutrient stress due to inhibition of the mevalonate pathway. From many end products of this pathway, ubiquinone is required in a process of ATP formation during oxidative phosphorylation. Moreover, oxidative phosphorylation is also reported as one of the significantly affected functional pathways by statins [30]. Interestingly, it was published that ubiquitinated hydrophobic proteins that are prone to aggregation are kept on the surface of lipid droplets and subjected to autophagy, as well as proteasomal degradation [88].

The cell-cycle arrest represents another frequently discussed event associated with statins. Many reports describe the effect of individual statins on the expression of cell-cyclerelated genes. For example, changes in the expression of genes related to the cell cycle in chronic myelogenous leukemia cells K562 were described. Fifteen downregulated and 9 upregulated cell-cycle-related genes were observed in the presence of 20 µM simvastatin for 48 h. The results of flow cytometry showed that the cell cycle was arrested in the G1 phase [89]. Assmus et al. performed a microarray analysis of about 12,000 genes in endothelial progenitor cells (a primary cell line) treated with 0.1 µM atorvastatin for 10 h. The expression of cyclins and proliferating cell nuclear antigen (PCNA) was increased after atorvastatin treatment. Moreover, the expression of the cell-cycle inhibitory protein p27 was reduced [90]. In the next study, downregulation of cyclin D1, PCNA, and c-myc, and upregulation of p21 and p19, were provided by the treatment of human breast cancer cells with cerivastatin [91].

In our microarray analysis, the expression of genes encoding cyclin D1 was affected only by cerivastatin, cyclin A2 by pitavastatin, and cyclin E2 by both mentioned statins, similar to the *PCNA* gene. Other statins did not affect the expression of genes encoding cyclins or PCNA. The expression of the *p21* gene was increased by all effective statins. Moreover, the genes associated with DNA replication, such as *ORC1L*, *MCM2*, or *MCM3,* were downregulated by cerivastatin and pitavastatin (Table 4). This suggests that statins blocked the progress of the cell cycle through the S phase of MiaPaCa-2 cells. The downregulation of the genes encoding histone H4 and upregulation of the gene encoding histone H2B by the three most effective statins (cerivastatin, pitavastatin, and simvastatin) also was very interesting. The effect of simvastatin on expression of the gene encoding histone was also observed in a report by Johnson-Anuna et al., in which the expression level of the gene

encoding the linker histone H1.2 was increased after simvastatin therapy [84]. Likewise, other proteins related to the S phase (SKP2, E2F2) or M phase (CDC2, CDC25B) were downregulated by cerivastatin and pitavastatin treatment of MiaPaCa-2 cells in our study, indicating that except for the G1 phase, the most effective statins blocked the cell-cycle entry into the M phase. This was not surprising, as lovastatin is used in the cell-cycle synchronization protocols [92] and as a pharmacological tool for controlling the growth of neoplastic cells both in vitro and in vivo [23,93,94]. Furthermore, lovastatin is commercially available as an inhibitor of the cell cycle in the G1 and G2/M phases (Sigma, USA). The G1 block has been attributed to the inhibition of either cytokinesis or cell spreading following cytokinesis [95]. Several authors have also noted the retardation or arrest of the cell cycle at the G2/M transition [93,94,96]. However, the mechanism of the cell-cycle arrest by statins is not exactly clear. Despite a piece of evidence that lovastatin suppresses cell proliferation through inhibition of proteasome-mediated degradation of p21 and p27 [97], it was concluded that lovastatin neither synchronizes cells nor arrests the cells in the G1 phase of the division cycle [98].

To explain the reported results, the distinct impact of statins on gene expression profiles in MiaPaCa-2 cells should be related to their inhibitory activity interfering with the mevalonate pathway. Quantum biochemistry computations indicated some variations among the attractive forces of four tested statins (atorvastatin, rosuvastatin, simvastatin, and fluvastatin) and the HMG-CoA reductase binding site. The highest binding energies was determined for atorvastatin followed by rosuvastatin, while the lowest were found for simvastatin and fluvastatin; i.e., binding energies of 320, 310, 290, and 290 kcal·mol-1 , respectively [99]. However, in this study, cerivastatin, lovastatin, and pitavastatin were not included in the calculations. In another study, the crystal structures of the catalytic moiety of HMG-CoA reductase in a complex with six statins documented van der Waals interactions of the rigid hydrophobic moieties of the statins through a shallow nonpolar binding pocket and a part of the binding surface for CoA [100]. These interactions prevented the binding of the substrate HMG-CoA to the active site of the enzyme. No dramatic differences were found among the numbers of binding interactions among the statins evaluated, namely: compactin, simvastatin, fluvastatin, cerivastatin, atorvastatin, and rosuvastatin. Atorvastatin, simvastatin, lovastatin, fluvastatin, and cerivastatin are relatively lipophilic and are metabolized by the cytochrome P450 system. The other lipophilic compound pitavastatin is metabolized poorly via this pathway. Very limited P450-mediated metabolization was reported also for hydrophilic pravastatin and rosuvastatin, which were only nonsignificantly metabolized. Interestingly, high systemic bioavailability was reported for both cerivastatin and pitavastatin (60% [101] and 80% [102], respectively), which could explain their large impact on changes in gene expression. Except for pravastatin, all the other statins were efficiently bound to plasma proteins. However, the unbound pravastatin was poorly distributed in tissues due to its high hydrophilic nature [103].

Further, the distinct efficacy of individual statins on both antiproliferative activity and changes in gene expression could be also correlated with the statin levels inside cells. This correlation was the strongest for the least efficient statins (rosuvastatin and pravastatin); whereas for the most bioavailable statins (in particular lovastatin), this correlation was not so strong [30]. Thus, the statin effects on whole gene expression correlated with their bioavailability, as well as the impact on cell viability, only to a limited extent. This hypothesis held for cerivastatin; however, not so for pitavastatin and lovastatin, which led us to the conclusion that other crucial factors played an important role in the differences of statin effects on pancreatic cancer cell proliferation.

In summary, differences in the efficacy of individual statins are known depending on their structure, concentration, duration of action, or microenvironment. Although tumor cells exhibit many identical properties, the effect of statins depends also on the cell type [104]. The antitumor effect of statins is not only a function of the mechanism of their action, but also of how they are metabolized. In general, healthy cells are generally more resistant to statins than tumor cells [105]. The use of statins in the treatment of cancer as monotherapeutics is not effective enough, but their use in combination with other therapeutic approaches would significantly increase the effectiveness of cancer treatments and patient survival [106,107].

**Supplementary Materials:** The following are available online: Table S1: Legend for figures from STRING database.

**Author Contributions:** Conceptualization, L.V., H.G., S.R.; methodology, L.V., H.S., H.G. software, M.K., S.R.; validation, H.S.; formal analysis, M.K., H.G.; investigation, L.V., H.S., H.G.; resources, S.R., M.K., H.G.; data curation, M.K., H.G., S.R.; writing—original draft preparation, H.G., S.R.; writing—review and editing, M.K., T.R., L.V.; visualization S.R., M.K.; supervision, L.V., T.R; project administration, L.V.; funding acquisition, L.V., T.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by grant no. APVV-15-0217 given by the Slovak Research and Devel-opment Agency; by the Operational Programme Research, Development, and Education (Reg. No. CZ.02.1.01/0.0/0.0/16\_019/0000785); and by grant no. RVO-VFN64165/2021 given by the Czech Ministry of Health.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The gene microarray data have been deposited in the ArrayExpress database (accession number E-MTAB-3263).

**Acknowledgments:** We gratefully acknowledge Martina Krausová, Jana Šáchová, and Miluše Hradilová for excellent technical support.

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

### **References**


### *Article* **Saponins Extracted from Tea (***Camellia Sinensis***) Flowers Induces Autophagy in Ovarian Cancer Cells**

**Yaomin Wang 1,2,3,**† **, Chen Xia 1,**† **, Lianfu Chen 1 , Yi Charlie Chen 4, \* and Youying Tu 1, \***


Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso Received: 19 October 2020; Accepted: 9 November 2020; Published: 11 November 2020

**Abstract:** Tea flower saponins (TFS) possess effective anticancer properties. The diversity and complexity of TFS increases the difficulty of their extraction and purification from tea flowers. Here, multiple methods including solvent extraction, microporous resin separation and preparative HPLC separation were used to obtain TFS with a yield of 0.34%. Furthermore, we revealed that TFS induced autophagy—as evidenced by an increase in MDC-positive cell populations and mCherry-LC3B-labeled autolysosomes and an upregulation of LC3II protein levels. 3-MA reversed the decrease in cell viability induced by TFS, showing that TFS induced autophagic cell death. TFS-induced autophagy was not dependent on the Akt/mTOR/p70S6K signaling pathway. TFS-induced autophagy in OVCAR-3 cells was accompanied by ERK pathway activation and reactive oxygen species (ROS) generation. This paper is the first report of TFS-mediated autophagy of ovarian cancer cells. These results provide new insights for future studies of the anti-cancer effects of TFS.

**Keywords:** saponins; phytochemicals; tea (*Camellia sinensis*) flower; ovarian cancer; autophagy

### **1. Introduction**

Epithelial ovarian cancer continues to be a deadly disease due to its high mortality rate and poor long-term prognosis [1–4]. In recent years, much attention has been focused on studying natural products and searching for compounds which possess anti-cancer effects [5–7]. The effectiveness of tea (*Camellia sinensis*) against different types of cancer has been extensively studied owing to the numerous bioactive constituents contained in the tea plant [8–11]. Tea flowers are the reproductive organs of tea, and recent studies have demonstrated the effective anti-cancer effects of tea flowers and compounds extracted from tea flowers [12]. Tea flowers contain a number of bioactive constituents including polyphenols, polysaccharides, amino acids and saponins [13–16]. Tea flower saponins in particular have attracted attention for their anticancer effects [17]. However, the diversity and complexity of tea flower saponins increases the difficulty of their extraction and separation from tea flowers, which limits the research of tea flower saponins.

Apoptosis and autophagy are two major modalities of programed cell death (PCD). Targeting PCD pathways is becoming an important strategy for drug discovery from natural compounds [18]. Apoptosis and autophagy may share common upstream signals; thus, apoptosis and autophagy may occur simultaneously [19–21]. Natural saponins have been reported to induce combined apoptosis and

autophagy in various cancer cell lines in vitro [22–24]. Our previous study revealed that tea flower saponins (TFS) induced apoptosis in A2780/CP70 human ovarian cancer cells and OVCAR-3 cells, as well as S-phase arrest [17]. Nonetheless, an autophagic effect caused by TFS has never been reported.

In this study, we extracted and isolated TFS from tea flowers, then demonstrated for the first time that TFS induced autophagy, and evaluated the molecular mechanisms in OVCAR-3 human ovarian cancer cells.

### **2. Results**

### *2.1. Extraction of TFS from Tea Flowers*

Multiple processes were used to extract highly purified TFS from tea flowers. As shown in Figure 1, 210 g dried tea flowers were used for the extraction of TFS, which involved 70% methanol extraction followed by solvent extraction and then microporous resin separation. TFS was mainly present in the 75% and 90% ethanol fractions. The 45% and 60% ethanol fractions also contained smaller amounts of TFS (Figure S3). We thus collected the 75% and 90% ethanol fractions for subsequent HPLC separation, resulting in three TFS fractions. We previously characterized the high purity TFS fraction 2 (Figure S2) and found it contains 14 triterpenoid saponins [17]. TFS fractions 1 and 3 also have high purities (Figure S4). The total yield was 0.34%.

**Figure 1.** Extraction and isolation scheme of tea flower saponins (TFS) from tea flowers.

### *2.2. TFS Decreased Cell Viability and Induced Morphological Changes in OVCAR-3 Cells*

To test whether TFS possesses an anti-cancer effect, we determined cell viability using an MTS assay after OVCAR-3 cells were treated with TFS for 24 h. As shown in Figure 2, TFS decreased cell viability in a dose-dependent manner. We additionally observed the morphology of OVCAR-3 cells by microscopy after treatment with TFS for 24 h. As shown in Figure 2, TFS induced morphological changes in OVCAR-3 cells, as the cells exhibited round bodies and clustered in groups.

μ **Figure 2.** TFS decreased cell viability and induced morphology changes in OVCAR-3 cells. (**A**) The effect of TFS on cell viability was conducted using an MTS assay. Results were obtained from three independent experiments and expressed as mean ± SEM. Significant difference verse control (\* *p* < 0.05 and \*\*\* *p* < 0.001. (**B**) OVCAR-3 cells were treated with the indicated concentrations of TFS for 24 h. Cell morphology was observed using microscopy (magnification 10×), scale bar = 50 µm.

### *2.3. TFS Triggers Autophagy in OVCAR-3 Cells*

μ MDC staining was conducted to determine whether TFS induces autophagy and decreased viability of OVCAR-3 cells. MDC is an eosinophilic fluorescent dye that binds to the ubiquitin-like binding system Apg8, which requires the formation of autophagic vacuoles. Under excitation by ultraviolet light, the autophagic vacuoles demonstrated bright green fluorescence [25]. As shown in Figure 3, the presence of MDC-positive cells increased dose-dependently following treatment with TFS for 24 h. As indicated by the white arrows, bright green fluorescence appeared within the cells. The number of autophagic cells in the group treated with 1.5 µg/mL TFS for 24 h was significantly greater than the number of autophagic cells in the control group.

μ μ **Figure 3.** TFS induced autophagic vacuoles in OVCAR-3 cells. OVCAR-3 cells were treated with TFS (0, 0.5, 1.0 and 1.5 µg/mL) for 24 h then stained with MDC dye and visualized by fluorescence microscopy. White arrows indicate autophagic cells (magnification 20×), scale bar = 20 µm.

To further confirm the induction of autophagy by TFS, autophagic lysosomes were detected by mCherry-LC3B plasmid transfection. The mCherry-LC3B fusion fluorescent protein aggregates on autophagic lysosomes during autophagy, and specific spots of bright red fluorescence can be visualized under confocal microscopy. In this study, mCherry-LC3B plasmids were transfected into OVCAR-3 cells, and then the cells were treated with TFS for 24 h. Confocal microscopy images were captured with an inverted laser scanning confocal microscope. As shown in Figure 4A, the number of autophagic lysosomes marked by red fluorescence in the OVCAR-3 cells increased with higher concentrations of TFS.

We quantified the effect of TFS treatment on the expression of LC3 protein in OVCAR-3 cells by western blotting. Under normal conditions, LC3 protein is cleaved by Atg4 at the carboxyl end to form LC3-I (16 kD) and localizes to the cytoplasm. When autophagy occurs, ubiquitin-like systems such as Atg7 and Atg3 modify LC3-I, causing it to bind with phosphatidylethanolamine found on the surface of the autophagosome membrane to produce LC3II (14 kD) where it remains localized on the autophagic membrane. LC3-II expression is directly proportional to the activity of autophagic vesicles. Therefore, the change of expression between LC3-I and LC3-II could indicate whether autophagy is occurring. As shown in Figure 4B,C, the expression of LC3-II protein in OVCAR-3 cells increased significantly after treatment with TFS in a dose-dependent manner after 24 h. Taken together, the results from MDC staining, mCherry-LC3B plasmid transfection, and Western blotting for LC3 protein clearly indicated that TFS induced autophagy in OVCAR-3 cells.

μ Next, we added the autophagy inhibitor 3-MA to determine whether TFS-induced autophagy in OVCAR-3 cells promoted cancer cell survival or caused cancer cell death. As shown in Figure 4D, cell viability increased from 47% to 79% following treatment with 1.5 µg/mL of 3-MA for 24 h, indicating that TFS-induced autophagy promoted cell death of OVCAR-3 cells. Thus, these results revealed that TFS treatment led to the autophagic death of OVCAR-3 cells.

μ μ μ μ **Figure 4.** TFS treatment induced autophagic cell death in OVCAR-3 cells. (**A**) OVCAR-3 cells were transfected with mCherry-LC3B plasmid and treated with TFS (0, 0.5, 1.0, 1.5 µg/mL) for 24 h. Cells were imaged with an inverted laser scanning confocal microscope (magnification 40×), scale bar = 50 µm. (**B**,**C**) OVCAR-3 Cells were treated with different concentrations of TFS (0, 0.5, 1.0, 1.5 µg/mL) for 24 h. Western blotting was conducted to evaluate the levels of LC3I and LC3II protein with GAPDH used as an internal control. (**D**) OVCAR-3 cells were treated with 3-MA and TFS (1.5 µg/mL) for 24 h and then detected for cell viability. Each data point represents three independent experiments, significant difference versus control (\*\*\* *p* < 0.001).

### *2.4. TFS-Induced Autophagy in OVCAR-3 Cells Independently from the Akt*/*mTOR*/*p70S6K Pathway and Was Accompanied by ERK Activation*

We further evaluated the possible molecular mechanisms of TFS-induced autophagy by Western blot analysis, first by measuring Akt/mTOR/p70S6K pathway-related proteins. The Akt/mTOR/p70S6K pathway is a major autophagy regulatory pathway which negatively regulates autophagy [24,26,27]. As shown in Figure 5A, protein levels of p-Akt and Akt remained unchanged following treatment with TFS (0, 0.5, 1.0, 1.5 µg/mL) for 24 h. Phosphorylation of p70S6K was upregulated dose-dependently by TFS in OVCAR-3 cancer cell lines. Taken together, these results demonstrated that TFS-induced autophagy in OVCAR-3 cells was independent of the Akt/mTOR/p70S6K pathway. It has been reported that MAPK signaling pathways, which include ERK1/2, p38, and JNK, play a pivotal role in autophagy induction [22,27,28]. Accordingly, we evaluated the levels of MAPK signaling pathway proteins following treatment with TFS for 24 h. As shown in Figure 5B, TFS treatment significantly upregulated protein levels of p-ERK while total ERK expression remained unchanged. No obvious changes to p-p38, p38, and JNK expression were observed. These results indicated that the ERK pathway was activated in OVCAR-3 cancer cells. μ

μ **Figure 5.** Effects of TFS on Akt pathways and MAPK pathways in OVCAR-3 cells. Protein expression of (**A**) Akt pathways related proteins p-Akt, Akt, p-P70S6K and (**B**) MAPK pathways related proteins p-ERK, ERK, p-p38, p38, JNK were analyzed by Western blot after being treated with different concentrations of TFS (0, 0.5, 1.0, 1.5 µg/mL) for 24 h, GAPDH was used as internal control.

### *2.5. TFS-Induced Autophagy in OVCAR-3 Cells Is Accompanied by ROS Generation*

Reactive oxygen species (ROS) generation is considered to be associated with the autophagy initiation and contributes to autophagy-mediated cell death [5,29]. We measured ROS generation using DCFH-DA following treatment with TFS. As shown in Figure 6, DCFH-DA positive cells increased after treatment with TFS, indicating that TFS treatment significantly enhanced ROS generation in OVCAR-3 cells. The results suggested that ROS generation is possibly involved in TFS-induced autophagy and this finding will need to be further elucidated.

μ μ **Figure 6.** Effects of TFS on reactive oxygen species (ROS) generation in OVCAR-3 cells. ROS generation in OVCAR-3 cells was evaluated by DAPI and DCFH-DA double staining. Representative images are shown. After treatment with TFS (0, 0.5, 1.0 and 1.5 µg/mL) for 24 h, cells were stained with DCFH-DA and DAPI and visualized by fluorescence microscopy (magnification 20×), scale bar = 50 µm. The top line of images (**a**–**d**) shows cells stained with DAPI. The middle line of images (**e**–**h**) shows cells which generated ROS as represented by staining with DCFH-DA. The bottom line of images (**i**–**l**) shows merged fluorescence images of cells stained by DAPI and DCFH-DA.

### **3. Discussion**

Natural saponins from tea (*Camellia sinensis*) flowers possess remarkable structural diversity and have become increasingly important compounds for the prevention and treatment of cancer and other diseases [29–31]. Several triterpenoid saponins have demonstrated the ability to induce autophagy of cancer cells in various cancer cells lines [32,33]. To better understand the mechanisms underlying this observed anticancer effect, we performed a chemical extraction of triterpenoid saponins from tea flowers. We first extracted dried tea flowers with 70% methanol, which was the most efficient extraction solution according to our previous studies. We then performed solvent extraction, microporous resin separation, and preparative HPLC separation to obtain three TFS fractions with a total yield of 0.34%. Our previous study showed that TFS induced apoptosis and S phase arrest in A2780/CP70 and OVCAR-3 cells [17]. In our present study, we have shown for the first time that TFS can induce autophagic cell death in OVCAR-3 cells.

To confirm that TFS induced autophagy in OVCAR-3 cells, MDC staining was conducted following treatment with TFS for 24 h, showing a dramatic increase of MDC-positive cells. Autophagic lysosomes in OVCAR-3 cells increased significantly after transfection with mCherry-LC3B plasmids and TFS treatment. Western blot analysis was conducted to evaluate expression changes of LC3 in OVCAR-3 cells. LC3-II is considered to be the most reliable autophagic biomarker as LC3-II expression levels are proportional to the formation of autophagosomes [18,33]. In our study, we found that TFS dose-dependently upregulated protein levels of LC3-II in OVCAR-3 cells. These results indicated that TFS induced autophagy in ovarian cancer cells. Moreover, the autophagy inhibitor 3-MA reversed the decrease of cell viability caused by TFS, which further indicated that TFS induced autophagic cell death in OVCAR-3 cells.

Inhibition of the Akt/mTOR/p70S6K pathway is crucial for the induction of autophagy. p70S6K is the downstream target of Akt, and expression levels of p-p70S6K can be used as a marker for mTOR activity [22]. The natural compound delicaflavone can induce autophagy via the Akt/mTOR/p70S6K pathway in human lung cancer cells [26]. However, in our study, proteins levels of p-Akt and Akt remained unchanged while p-P70S6K was upregulated in OVCAR-3 cells. These results suggest that TFS-induced autophagy in ovarian cancer cells does not involve the Akt/mTOR/p70S6K pathway, which is consistent with previous reports that Platycodin D triggered autophagy in HepG2 cells [34].

We further evaluated the effects of TFS on MAPK signaling pathways and ROS generation. ERK, p38, and JNK are important regulatory proteins of MAPK signaling pathways which could also regulate autophagy processes [35,36]. In our present study, TFS significantly upregulated protein levels of p-ERK in OVCAR-3 cells. Protein levels of p-p38, p38, and JNK were not affected after treatment with TFS. These results indicated that the ERK pathway was activated by TFS-induced autophagy in ovarian cancer cells. Our findings were also consistent with several reports that ERK activation occurs in autophagy induced by different compounds in various cancer cell lines [22,37,38]. To better elucidate the role of ERK pathway in TFS-induced autophagy, ERK inhibitor such as U0126 should be applied to future experiment and clarify whether ERK inhibition attenuates TFS-induced autophagy in ovarian cancer cells. Additionally, TFS treatment significantly enhanced ROS generation in OVCAR-3 cells. As ROS generation is considered to be an essential aspect of apoptosis and autophagy [29], we previously reported that TFS induces apoptosis in A2780/CP70 and OVCAR-3 cells [17]. The specific role of ROS generation in TFS-induced cell death however, needs to be further elucidated by pretreatment with ROS inhibitor N-acetyl-L-cysteine (NAC), then to detect the expression levels of autophagy related proteins to clarify the role of ROS in TFS-induced autophagy in OVCAR-3 cells. Taken together, we conclude that TFS-induced autophagy in OVCAR-3 cells is accompanied by ERK activation and ROS generation, the mechanisms of which will merit further study.

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

### *4.1. Materials and Reagents*

Dried tea flowers (Figure S1) were purchased from Zhejiang Yilongfang Co., Ltd. (Hangzhou, China). Analytical grades of methanol, ethanol, N-butanol, Ethyl acetate and other extraction chemical regents used in this study were obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Monodansylcadaverine (MDC) was purchased from Nanjing KeyGen Biotech Co., Ltd. (Nanjing, China). 3-Methyladenine (3-MA) was purchased from Dalian Meilun Biotech Co., Ltd. m-Cherry LC3B plasmid was purchased from Hanheng Biotech Co., Ltd. (Shanghai, China). Lipofectamine 2000 Reagent was purchased from Invitrogen (Grand Island, NY, USA). Primary rabbit polyclonal antibodies against LC3B, p-Akt, Akt, p-P70S6K, p-p38, p38 and JNK were obtained from Cell Signaling Inc. (Danvers, MA, USA). Secondary antibodies, p-ERK, ERK, and GAPDH were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). 2′ ,7′ -dichlorodihydrofluorescein diacetate (DCFH-DA) and CellTiter 96 Aqueous One Solution Cell Proliferation were purchased from Sigma-Aldrich (St. Louis, MO, USA) and Promega (Madison, WI, USA), respectively.

### *4.2. Extraction and Isolation of TFS*

Dried tea flowers (210 g) were pulverized and extracted with 70% methanol two separate times for 2 h under reflux at 60 ◦C. Seventy per cent methanol extracts were concentrated by a rotary evaporator to obtain a 70% methanol concentrate. The methanol extract was suspended in H2O and extracted with equal volumes of ethyl acetate and *N*-butanol three times, successively. The solvents were then evaporated to obtain the ethyl acetate fraction (12.67 g), *N*-butanol fraction (25.50 g) and water fraction (59.71 g). The 25.50 g *N*-butanol fraction was resolved in 10% ethanol and loaded in a D101 macroporous adsorption resin column and successively eluted with 0%, 15%, 30%, 45%, 60%, 75%, and 90% ethanol. The 45%, 60%, 75%, and 90% ethanol eluents were collected, then evaporated and freeze-dried to obtain the 45% ethanol fraction (3.94 g), 60% ethanol fraction (4.17 g), 75% ethanol fraction (3.24 g), and 90% ethanol fraction (0.70 g). The 75% and 90% ethanol fractions were mixed and

separated using preparative HPLC to obtain three fractions, TFS fraction 1 (216.8 mg), TFS fraction 2 (281.0 mg), and TFS fraction 3 (126.6 mg). TFS fraction 2 was identified previously and marked as "TFS" in later experiments [17].

### *4.3. Cell Lines and Cell Culture*

The human ovarian cancer cell line OVCAR-3 was provided by Dr. Jiang from West Virginia University. The cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) at 37 ◦C with 5% CO<sup>2</sup> in a humidified incubator.

### *4.4. MTS Assay*

OVCAR-3 cells were seeded in a 96-well plate at a density of 1.0 × 10<sup>4</sup> cells per well and incubated overnight for attachment. The cells were then treated with different concentrations of TFS with or without 3-MA for 24 h. MTS reagents were added to each well and absorbance was measured at 490 nm using a microplate reader.

### *4.5. Morphology Observation*

OVCAR-3 cells were exposed to different concentrations of TFS (0, 0.5, 1.0, 1.5 µg/mL) for 24 h. Cell morphology was then observed by microscopy (ZEISS, Heidelberg, Germany). Images were captured using a digital camera attached to the microscope at magnification × 10.

### *4.6. Labeling of Autophagic Vacuoles with MDC Staining*

OVCAR-3 cells were seeded in a 24-well plate at a density of 1.0 × 10<sup>5</sup> cells per well after a growth period of 24 h. The cells were then treated with different concentrations of TFS (0, 0.5, 1.0, 1.5 µg/mL) for another 24 h. After media change and wash with PBS, 50 µM MDC was added to each well and the plate was incubated for 30 min, after which fluorescence images were captured using an Olympus DP70 fluorescence microscope.

### *4.7. Transfection of mCherry-LC3B Plasmid*

OVCAR-3 cells were grown with medium (without antibiotics) in a 24-well plate (1.0 × 10<sup>5</sup> cells per well) for 24 h. The cells were ready for transfection once cell confluence reached 90%. The transfection mixture was prepared as follows: 0.5 µg mCherry-LC3B plasmid and 2 µL Lipofectamine 2000 were diluted with 50 µL OPTI-MEMI medium, respectively, and incubated at room temperature for 5 min. The diluted mCherry-LC3B plasmid was mixed with diluted Lipofectamine 2000 to prepare the transfection mixture and incubated at room temperature for 20 min. The culture medium was discarded and washed with PBS twice. Next, 400 µL of serum-free culture medium (without antibiotics) was added to each well and 100 µL of transfection mixture was then added. After a transfection period of 6 h in the incubator, the cells were then incubated with different concentrations of TFS for 24 h. Imaging slides were mounted with DAPI and images were captured using an FV1000 Laser Scanning Confocal Microscope (Olympus, Tokyo, Japan).

### *4.8. Reactive Oxygen Species (ROS) Detection*

ROS generation was measured by DCFH-DA staining. Briefly, OVCAR-3 cells were grown in a 96-well plate (1.0 × 10<sup>4</sup> cells per well) for 24 h and different concentrations of TFS (0, 0.5, 1.0, 1.5 µg/mL) were added to the indicated wells. After a 24 h incubation period, the cells were then incubated with DCFH-DA (10 µM) at 37 ◦C for 30 min. Imaging slides were mounted with DAPI and images were captured with an Olympus DP70 fluorescence microscope.

### *4.9. Western Blotting*

Cells were treated with different concentrations of TFS for 24 h, washed with PBS twice, harvested, and total protein was extracted using protein extraction reagent supplemented with 1% protease inhibitor. Protein content was detected by a BCA protein assay kit. Equal amounts of protein were loaded into 10% SDS-PAGE gels for separation and transferred onto nitrocellulose membranes. The membranes were blocked with 5% non-fat milk for 1 h, then incubated with indicated primary antibody overnight followed by secondary antibody for 2 h at room temperature. Protein bands were visualized using the ChemiDoc MP System (Bio Rad, Hercules, CA, USA). GAPDH was used to normalize relative values of each protein.

### *4.10. Statistical Analysis*

Data represents the mean of three independent experiments. Statistical significance was analyzed by Graphpad Prism software using the Newman–Keuls test. \* *p* < 0.05, \*\* *p* < 0.01 and \*\*\* *p* < 0.001 were considered as significant *p*-values.

### **5. Conclusions**

In the present study, we demonstrated an efficient extraction and purification method for TFS with a yield of 0.34%. We showed for the first time that TFS induced autophagic cell death in ovarian cancer cells. The specific autophagic effect mediated by TFS occurred independently from Akt/mTOR/p70S6K pathway signaling and the TFS-induced autophagy in ovarian cancer cells was found to be accompanied by ERK activation and ROS generation. Thus, our research provides an important basis for future study on the autophagic effects of TFS in ovarian cancer cells.

**Supplementary Materials:** The following are available online, Figure S1: Dried tea (*Camellia sinensis*) flowers, Figure S2: Saponins isolated from dried tea (*Camellia sinensis*) flowers, Figure S3: HPLC chromatograms of eluted fractions of 45%, 60%, 75%, 90% ethanol, Figure S4: Chromatograms of TFS 1, TFS 2and TFS 3 detected by LC/MS.

**Author Contributions:** Conceptualization, Y.W., Y.T. and Y.C.C.; methodology, Y.W. and C.X.; formal analysis, Y.W., C.X., and L.C.; resources, Y.T. and Y.C.; writing—original draft preparation, Y.W., C.X., and L.C.; writing—review and editing, Y.W., Y.T., and Y.C.C.; funding acquisition, Y.T. and Y.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Key research and development projects in Zhejiang Province "Industrialization model projects on exploring functional components and related products from tea flowers and fruits" (Grant No: 2018C02012), the National key research and development plan "Processing of tea products, key technology of quality control during storage and equipment development" (Grant No: 2017YFD0400803), and the Natural Science Foundation of Hubei Province (Grant No: 2019CFB187). The authors are very grateful to Yul Huh for language editing.

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

### **Abbreviations**


### **References**


**Sample Availability:** Samples of the compounds TFS are not available from the authors.

**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*

### **Standardized Saponin Extract from Baiye No.1 Tea (***Camellia sinensis***) Flowers Induced S Phase Cell Cycle Arrest and Apoptosis via AKT-MDM2-p53 Signaling Pathway in Ovarian Cancer Cells**

**Youying Tu 1 , Lianfu Chen 1,2 , Ning Ren 1,2 , Bo Li 1 , Yuanyuan Wu 1 , Gary O. Rankin 3 , Yon Rojanasakul 4 , Yaomin Wang 5, \* and Yi Charlie Chen 2, \***


Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso Received: 22 June 2020; Accepted: 29 July 2020; Published: 31 July 2020

**Abstract:** Ovarian cancer is considered to be one of the most serious malignant tumors in women. Natural compounds have been considered as important sources in the search for new anti-cancer agents. Saponins are characteristic components of tea (*Camellia sinensis*) flower and have various biological activities, including anti-tumor effects. In this study, a high purity standardized saponin extract, namely Baiye No.1 tea flower saponin (BTFS), which contained Floratheasaponin A and Floratheasaponin D, were isolated from tea (*Camellia sinensis* cv. Baiye 1) flowers by macroporous resin and preparative liquid chromatography. Then, the component and purity were detected by UPLC-Q-TOF/MS/MS. This high purity BTFS inhibited the proliferation of A2780/CP70 cancer cells dose-dependently, which is evidenced by the inhibition of cell viability, reduction of colony formation ability, and suppression of PCNA protein expression. Further research found BTFS induced S phase cell cycle arrest by up-regulating p21 proteins expression and down-regulating Cyclin A2, CDK2, and Cdc25A protein expression. Furthermore, BTFS caused DNA damage and activated the ATM-Chk2 signaling pathway to block cell cycle progression. Moreover, BTFS trigged both extrinsic and intrinsic apoptosis—BTFS up-regulated the expression of death receptor pathway-related proteins DR5, Fas, and FADD and increased the ratio of pro-apoptotic/anti-apoptotic proteins of the Bcl-2 family. BTFS-induced apoptosis seems to be related to the AKT-MDM2-p53 signaling pathway. In summary, our results demonstrate that BTFS has the potential to be used as a nutraceutical for the prevention and treatment of ovarian cancer.

**Keywords:** tea (*Camellia sinensis*) flowers; BTFS; A2780/CP70 ovarian cancer cells; apoptosis; S phase cell cycle arrest

### **1. Introduction**

Ovarian cancer is a malignant tumor of the female reproductive system that causes high mortality in women. Worldwide, 4.4% of cancer-related deaths are due to ovarian cancer [1]. Recent statistics

have shown that there were 52,100 new cases of ovarian cancer and about 22,500 deaths in China [2]. In the US, there were 22,240 new cases of ovarian cancer with a death toll of 14,070; the survival rate is only 37% [3]. Currently, the standard treatment for ovarian cancer is radical surgery combined with adjuvant chemotherapy drugs, which has certain effects on early patients. In clinic, most ovarian cancer patients are diagnosed in the middle and late stages due to the vague early symptoms and the lack of detection methods. Accordingly, ovarian cancer is easily prone to relapse and drug resistance, and ultimately leads to high mortality [4]. Thus, it is particularly necessary to find new compounds from natural products, which could effectively prevent and treat ovarian cancer.

Infinite proliferation is the main feature of cancer cells and drugs that inhibit cell division are often used in cancer treatment. Disorders of the cell cycle can cause cells to proliferate indefinitely and induce cancer. Cyclin dependent kinases (CDKs), Cyclin, and CKI are highly correlated with the occurrence, development, and prognosis of ovarian cancer [5]. CDK2, CDK4, Cyclin A, Cyclin D, and the expression of Cyclin E are upregulated significantly in ovarian cancer tissues [6], while the protein expression of p21 and p27 are significantly lower than that of non-cancerous ovarian tissue. Many studies have improved the therapeutic effect through the action of drugs on cell cycle checkpoints. Apoptosis is a kind of programmed cell under physiological conditions. The abnormal regulation of apoptosis is closely related to the occurrence and development of cancer. Targeting extrinsic (death receptor-mediated) and intrinsic (mitochondrial-mediated) apoptosis is an important strategy in cancer treatment [7].

Saponins are a class of glycoside compounds with polycyclic compounds as ligands that are widely present in ginseng [8], Chonglou [9], Camellia [10], and other plants. Recent studies demonstrate that saponins show a strong inhibitory effect against ovarian cancer. Chonglou Saponin II induced G2 phase cell cycle arrest in SKOV3 ovarian cancer cells [11]. Baiying steroidal saponin aescin up-regulates the activity of Caspase-3 and promotes the apoptosis of SKOV3 cells [12]. Moreover, several other studies have demonstrated the potency of saponins by inhibiting cell proliferation, inducing apoptosis and cell cycle arrest, as well as inhibiting angiogenesis and weakening the invasion and metastasis, which has made saponins important natural sources in the search for new chemopreventive agents [13–15].

Tea is a popular beverage globally, and many functional ingredients have multiple health functions for the human body. Tea saponins are widely distributed in the roots, stems, leaves, and flowers [13]. However, compared to tea leaves, tea flowers contain a much higher amount of saponins. Moreover, tea flowers have become important food resources and it is an urgent need to explore the health benefits of tea flowers saponins. In recent years, tea flower saponins have demonstrated strong anticancer activity, as confirmed by several research. Kitagawa et al. showed that Chakasaponin I, II, and Floratheasaponin A have inhibitory effects on human gastrointestinal cancer cells HSC-2, HSC-4, MKN-45, and Caco-2, and the effect is stronger than catechin, flavonoids, and caffeine. The inhibition of HSC-2 cell proliferation may be due to the induction of cell cycle arrest in the G2/M phase and activation of apoptosis [16]. Previously, we also reported the effects of tea flower saponins containing Chakasaponin II and other 13 monomers on ovarian cancer cells; the results demonstrated that tea flower saponins shows strong anticancer activity by inducing apoptosis and S phase cell cycle arrest in human OVCAR-3 and A2780/CP70 cancer cells [17]. Now, further-purified tea flower saponins is needed to better elucidate anticancer properties and mechanisms. Thus, the aim of this paper is to obtain high-purity tea flower saponins and further elucidate its antiproliferative effect and mechanisms on human ovarian cancer cell. Our results show that high-purity BTFS extracted from Baiye No.1 tea flowers induced cell cycle arrest in the S phase and induced apoptosis in A2780/CP70 cells by targeting both extrinsic and intrinsic apoptotic pathways.

#### **2. Results 2. Results**

#### *2.1. Analysis and Identification of BTFS 2.1. Analysis and Identification of BTFS*

We found, through UPLC-Q-TOF/MS/MS analysis, that BTFS extracted from tea flowers contained 3 triterpenoid saponins (Figure 1A,B). Based on the published literature [18–20] and combined with the retention time, molecular weight, molecular formula, and secondary ion fragment information of the corresponding peak, we inferred that the saponins in BTFS are Floratheasaponin D (58.5%), Floratheasaponin A (36.7%), and an unnamed saponin (4.9%) (Table 1) [20]. Our previous study showed that Floratheasaponin A and Floratheasaponin D were the main saponin components in Baiye No.1 tea flowers [20]. We found, through UPLC-Q-TOF/MS/MS analysis, that BTFS extracted from tea flowers contained 3 triterpenoid saponins (Figure 1A,B). Based on the published literature [18–20] and combined with the retention time, molecular weight, molecular formula, and secondary ion fragment information of the corresponding peak, we inferred that the saponins in BTFS are Floratheasaponin D (58.5%), Floratheasaponin A (36.7%), and an unnamed saponin (4.9%) (Table 1) [20]. Our previous study showed that Floratheasaponin A and Floratheasaponin D were the main saponin components in Baiye No.1 tea flowers [20].

*Molecules* **2020**, *25*, 3515 3 of 17

**Figure 1.** Typical chromatograms of BTFS. (**A**) Ultraviolet chromatograms. (**B**) Total ion **Figure 1.** Typical chromatograms of BTFS. (**A**) Ultraviolet chromatograms. (**B**) Total ion chromatograms.

chromatograms. **Table 1.** MS data in negative mode of saponins extracted from tea flower.


#### *2.2. BTFS Inhibits Ovarian Cancer Cell Proliferation 2.2. BTFS Inhibits Ovarian Cancer Cell Proliferation*

The effects of BTFS on cell viability were determined by CellTiter 96 Aqueous One Solution Cell Proliferation assay. BTFS reduced the cell viability of A2780/CP70 cells dose-dependently, while BTFS displayed less cytotoxicity to normal ovarian cell line IOSE-364 (Figure 2A). Colony formation assay is considered as an effective way to determine the long-term proliferation capacity of cancer cells. We then carried out this assay to further explore the inhibitory effect of BTFS on A2780/CP70 cells. The results found that the number of cell colonies was significantly lower for cells subjected to BTFS than for the control cell (Figure 2B,C). Besides, we also analyzed the protein expression of proliferating The effects of BTFS on cell viability were determined by CellTiter 96 Aqueous One Solution Cell Proliferation assay. BTFS reduced the cell viability of A2780/CP70 cells dose-dependently, while BTFS displayed less cytotoxicity to normal ovarian cell line IOSE-364 (Figure 2A). Colony formation assay is considered as an effective way to determine the long-term proliferation capacity of cancer cells. We then carried out this assay to further explore the inhibitory effect of BTFS on A2780/CP70 cells. The results found that the number of cell colonies was significantly lower for cells subjected to BTFS

than for the control cell (Figure 2B,C). Besides, we also analyzed the protein expression of proliferating cell nuclear antigen (PCNA) by Western blot; the results showed that PCNA was downregulated in the BTFS-treated group, compared to the vehicle group. (Figure 2D,E, the original images for Figure 2D can be seen in Supplementary Materials). In summary, the results demonstrated that BTFS exhibited an antiproliferative effect and cytotoxicity on A2780/CP70 cells. *Molecules* **2020**, *25*, 3515 4 of 17 cell nuclear antigen (PCNA) by Western blot; the results showed that PCNA was downregulated in the BTFS-treated group, compared to the vehicle group. (Figure 2D,E). In summary, the results demonstrated that BTFS exhibited an antiproliferative effect and cytotoxicity on A2780/CP70 cells.

**Figure 2.** BTFS inhibits proliferation of A2780/CP70 cells. (**A**) BTFS inhibited the cell viability of A2780/CP70 and showed moderate effect on IOSE-364 cells. (**B**) The effect of BTFS on colony formation of A2780/CP70 cells. (**C**) Statistical histogram. \**p* < 0.05; \*\**p* < 0.01 versus control. (**D**)Effects of BTFS on protein expression of PNCA in A2780/CP70 cells. (**E**) Statistical histogram of protein quantization. \* *p* < 0.05; \*\* *p* < 0.01 versus control. **Figure 2.** BTFS inhibits proliferation of A2780/CP70 cells. (**A**) BTFS inhibited the cell viability of A2780/CP70 and showed moderate effect on IOSE-364 cells. (**B**) The effect of BTFS on colony formation of A2780/CP70 cells. (**C**) Statistical histogram. \**p* < 0.05; \*\**p* < 0.01 versus control. (**D**)Effects of BTFS on protein expression of PNCA in A2780/CP70 cells. (**E**) Statistical histogram of protein quantization. \* *p* < 0.05; \*\* *p* < 0.01 versus control.

#### *2.3. BTFS Induces Cell Cycle Arrest in the S Phase in A2780*/*CP70 Cells 2.3. BTFS Induces Cell Cycle Arrest in the S Phase in A2780/CP70 Cells*

In order to elucidate the mechanism of BTFS inhibiting cell proliferation, flow cytometry was used to detect the cell cycle phase distribution of BTFS-treated human ovarian cells stained with propidium iodide (PI). The results showed that BTFS treatment induced a dose-dependent increase in the proportion of A2780/CP70 cells in the S phase. We also observed a reduction of cell proportion in the G0/G1 and G2/M phases (Figure 3A,B). When cells were treated with 1.5, 2.0, and 2.5 µg/mL BTFS for 24 h, the proportion of cells at the S phase were 29.04%, 35.60%, and 43.52%, respectively, compared with 20.30% in the vehicle-treated cells. In order to elucidate the mechanism of BTFS inhibiting cell proliferation, flow cytometry was used to detect the cell cycle phase distribution of BTFS-treated human ovarian cells stained with propidium iodide (PI). The results showed that BTFS treatment induced a dose-dependent increase in the proportion of A2780/CP70 cells in the S phase. We also observed a reduction of cell proportion in the G0/G1 and G2/M phases (Figure 3A,B). When cells were treated with 1.5, 2.0, and 2.5 μg/mL BTFS for 24 h, the proportion of cells at the S phase were 29.04%, 35.60%, and 43.52%, respectively, compared with 20.30% in the vehicle-treated cells.

*Molecules* **2020**, *25*, 3515 5 of 17

**Figure 3.** BTFS-induced cell cycle arrest at S phase in A2780/CP70 cells and regulated proteins expression related with S phase. (**A**,**B**) BTFS induced cell cycle arrest at S phase by flow cytometry. Statistical analysis bar chart, \* *p* < 0.05 and \*\* *p* < 0.01 versus control. (**C**,**D**) Effects of BTFS on the expression of cell cycle-related proteins in A2780/CP70 cancer cells. Statistical histogram of protein quantization, \* *p* < 0.05; \*\* *p* < 0.01 versus control. **Figure 3.** BTFS-induced cell cycle arrest at S phase in A2780/CP70 cells and regulated proteins expression related with S phase. (**A**,**B**) BTFS induced cell cycle arrest at S phase by flow cytometry. Statistical analysis bar chart, \* *p* < 0.05 and \*\* *p* < 0.01 versus control. (**C**,**D**) Effects of BTFS on the expression of cell cycle-related proteins in A2780/CP70 cancer cells. Statistical histogram of protein quantization, \* *p* < 0.05; \*\* *p* < 0.01 versus control.

#### *2.4. The Effects of BTFS on Cell Cycle Regulatory Protein Expression 2.4. The E*ff*ects of BTFS on Cell Cycle Regulatory Protein Expression*

We then evaluated the expression of cell cycle regulatory proteins by western blot after BTFS treatment. Cyclin and Cyclin dependent kinase (CDK) form Cyclin/CDK complex to regulate cell cycle progression. Furthermore, p21 and p27 are CDK inhibitors (CDKI), which negatively regulate cell cycle; the Cdc25 phosphatase family have a positive regulation effect on cell cycle. We found that BTFS could effectively suppress the expression of CDK2, Cyclin A, and Cdc25A and increased the expression of Cyclin E1 and p21, while showing no effect on the protein levels of p27 and Cdc25C (Figure 3C,D). The results demonstrated that the down-regulation of Cdc25A, up-regulation of p21, and reduction of kinase activities of CyclinE1/CDK2 and CyclinA2/CDK2 complexes might be responsible for S phase arrest induced by BTFS in A2780/CP70 cells. We then evaluated the expression of cell cycle regulatory proteins by western blot after BTFS treatment. Cyclin and Cyclin dependent kinase (CDK) form Cyclin/CDK complex to regulate cell cycle progression. Furthermore, p21 and p27 are CDK inhibitors (CDKI), which negatively regulate cell cycle; the Cdc25 phosphatase family have a positive regulation effect on cell cycle. We found that BTFS could effectively suppress the expression of CDK2, Cyclin A, and Cdc25A and increased the expression of Cyclin E1 and p21, while showing no effect on the protein levels of p27 and Cdc25C (Figure 3C,D). The results demonstrated that the down-regulation of Cdc25A, up-regulation of p21, and reduction of kinase activities of CyclinE1/CDK2 and CyclinA2/CDK2 complexes might be responsible for S phase arrest induced by BTFS in A2780/CP70 cells.

#### *2.5. BTFS Activates Apoptosis in A2780*/*CP70 Cells Molecules* **2020**, *25*, 3515 6 of 17

The increasing of sub-G1 phase population demonstrates that BTFS might cause cellular apoptosis in A2780/CP70 cells. Accordingly, we further explored whether BTFS caused apoptosis in A2780/CP70 cells. Hoechst 33342 staining was performed to observe the morphological changes of apoptosis. As shown in Figure 4A, after treating with BTFS, A2780/CP70 cells showed more apoptotic cells, which were brighter blue with condensed or fragmented nuclei than the untreated group. We then used quantitative fluorescence spectrophotometer to evaluate the mitochondrial membrane potential of A2780/CP70 cells. BTFS treatment resulted in a notable reduction in the red-green fluorescence ratio of JC-1 dye, which indicated that BTFS could induce apoptosis of A2780/CP70 cells (Figure 4B). Flow cytometric analysis was then conducted to further verify the pro-apoptotic effect of BTFS. As shown in Figure 4C,D, BTFS could significantly reduce the proportion of live cells and increase the proportion of apoptotic cells dose-dependently. Together, these results suggested that inducing apoptosis may be an important factor for the antiproliferative effect of BTFS on A2780/CP70 cells. *2.5. BTFS Activates Apoptosis in A2780/CP70 Cells*  The increasing of sub-G1 phase population demonstrates that BTFS might cause cellular apoptosis in A2780/CP70 cells. Accordingly, we further explored whether BTFS caused apoptosis in A2780/CP70 cells. Hoechst 33342 staining was performed to observe the morphological changes of apoptosis. As shown in Figure 4A, after treating with BTFS, A2780/CP70 cells showed more apoptotic cells, which were brighter blue with condensed or fragmented nuclei than the untreated group. We then used quantitative fluorescence spectrophotometer to evaluate the mitochondrial membrane potential of A2780/CP70 cells. BTFS treatment resulted in a notable reduction in the red-green fluorescence ratio of JC-1 dye, which indicated that BTFS could induce apoptosis of A2780/CP70 cells (Figure 4B). Flow cytometric analysis was then conducted to further verify the pro-apoptotic effect of BTFS. As shown in Figure 4C,D, BTFS could significantly reduce the proportion of live cells and increase the proportion of apoptotic cells dose-dependently. Together, these results suggested that inducing apoptosis may be an important factor for the antiproliferative effect of BTFS on A2780/CP70 cells.

**Figure 4.** BTFS-induced apoptosis in A2780/CP70 cells. (**A**) Hoechst 33342 staining confirmed the apoptotic effect induced by BTFS in A2780/CP70 cells. BTFS-induced apoptosis in A2780/CP70 cells evidenced by (**B**) Flow cytometry. (**C**) Statistical analysis bar chart, \*\* *p* < 0.01 versus control. **Figure 4.** BTFS-induced apoptosis in A2780/CP70 cells. (**A**) Hoechst 33342 staining confirmed the apoptotic effect induced by BTFS in A2780/CP70 cells. (**B**) The effect of BTFS on mitochondrial membrane potential in A2780/CP70 cells was determined by JC-1 staining. \*\* *p* < 0.01 versus control. (**C**,**D**) BTFS-induced apoptosis in A2780/CP70 cells evidenced by flow cytometry. Statistical analysis bar chart, \*\* *p* < 0.01 versus control.

### *2.6. BTFS Mediates Apoptosis via Intrinsic and Extrinsic Apoptotic Pathways*

Apoptosis could be classified into the intrinsic (mitochondrial mediated) pathway and the extrinsic (death receptor mediated) pathway. To investigate the specific pathway of BTFS in inducing apoptosis in A2780/CP70 cells, caspase-Glo assay was used to detected the caspase activation. As shown in Figure 5A, caspase-3/7, -8 and -9 activities in BTFS-treated cells were increased significantly compared with vehicle treated cells, which hinted that BTFS might trigger both the intrinsic and extrinsic apoptotic pathway. Furthermore, western blot assay was then conducted to examine the key proteins in both apoptotic pathways. After BTFS treatment, cleaved PARP protein, served as a marker of apoptosis, increased significantly, further confirming that BTFS could induce apoptosis in A2780/CP70 cells. Meanwhile, extrinsic pathway related proteins were investigated to confirm whether the extrinsic pathway was involved in BTFS-induced apoptosis. The results showed that BTFS could markedly up-regulate the expression of death receptors, Fas and DR5, and dose-dependently increased the cell death adaptor protein FADD (Figure 5B,C). With regard to the intrinsic apoptosis-related proteins, BTFS obviously decreased the protein expression of Bcl-2 and Bcl-xL, while Bax, Cytochrome C, and Apaf apoptotic protease activating factor-1(Apaf-1) were significantly upregulated and had no effect on Bad expression (Figure 5D,E). These results suggested that both intrinsic and extrinsic pathways were involved in BTFS-induced apoptosis in A2780/CP70 cells. shown in Figure 5A, caspase-3/7, -8 and -9 activities in BTFS-treated cells were increased significantly compared with vehicle treated cells, which hinted that BTFS might trigger both the intrinsic and extrinsic apoptotic pathway. Furthermore, western blot assay was then conducted to examine the key proteins in both apoptotic pathways. After BTFS treatment, cleaved PARP protein, served as a marker of apoptosis, increased significantly, further confirming that BTFS could induce apoptosis in A2780/CP70 cells. Meanwhile, extrinsic pathway related proteins were investigated to confirm whether the extrinsic pathway was involved in BTFS-induced apoptosis. The results showed that BTFS could markedly up-regulate the expression of death receptors, Fas and DR5, and dosedependently increased the cell death adaptor protein FADD (Figure 5B,C). With regard to the intrinsic apoptosis-related proteins, BTFS obviously decreased the protein expression of Bcl-2 and Bcl-xL, while Bax, Cytochrome C, and Apaf apoptotic protease activating factor-1(Apaf-1) were significantly upregulated and had no effect on Bad expression (Figure 5D,E). These results suggested that both intrinsic and extrinsic pathways were involved in BTFS-induced apoptosis in A2780/CP70 cells.

*Molecules* **2020**, *25*, 3515 7 of 17

extrinsic (death receptor mediated) pathway. To investigate the specific pathway of BTFS in inducing apoptosis in A2780/CP70 cells, caspase-Glo assay was used to detected the caspase activation. As

*2.6. BTFS Mediates Apoptosis via Intrinsic and Extrinsic Apoptotic Pathways* 

**Figure 5.** BTFS trigged both extrinsic and intrinsic pathways. (**A**) Effects of BTFS on the activity of caspase-3/7, 8 and 9. \* *p* < 0.05 and \*\* *p* < 0.01 versus control. (**B**,**C**) Effects of BTFS on extrinsic **Figure 5.** BTFS trigged both extrinsic and intrinsic pathways. (**A**) Effects of BTFS on the activity of caspase-3/7, 8 and 9. \* *p* < 0.05 and \*\* *p* < 0.01 versus control. (**B**,**C**) Effects of BTFS on extrinsic apoptosisrelated protein expression in A2780/CP70 cells. (**D**,**E**) Effect of BTFS on intrinsic apoptosis-related protein expression in A2780/CP70 cells.

controls (Figure 7A,B).

related protein expression in A2780/CP70 cells.

#### *2.7. BTFS Induces DNA Damage and A*ff*ects the Expression of Upstream Regulators AKT, MDM2, and P53 2.7. BTFS Induces DNA Damage and Affects the Expression of Upstream Regulators AKT, MDM2, and P53*

*Molecules* **2020**, *25*, 3515 8 of 17

ROS and DNA damages could mediate cell cycle arrest and apoptosis [21,22]. Firstly, DCFH-DA fluorescence probe assay was conducted to assess the effect of BTFS on intracellular ROS production. ROS generation was obviously elevated when A2780/CP70 cells were exposed to BTFS, compared to the vehicle-treated group (Figure 6A). Next, western blot assay was then performed to investigate the effect of BTFS on DNA damage response. The phosphorylation of histone H2A.X at serine 139 is considered as a reliable marker of DNA damage [23]. BTFS dramatically increased the phosphorylation of histone H2A.X. ATM/Chk is one of the important signaling pathways mediating DNA damage. After exposure to BTFS, both the phosphorylation of ATM at Ser1981 as well as the phosphorylation of Chk2 at Thr68 were up-regulated (Figure 6B,C), suggesting that ATM-Chk2 pathway was activated. These results suggested that BTFS promoted ROS generation and DNA damage, which is responsible for the cytotoxicity of BTFS. Moreover, BTFS induced DNA damage may be mediated by ROS generation, the specific mechanism of which needs to be further elucidated. ROS and DNA damages could mediate cell cycle arrest and apoptosis [21,22]. Firstly, DCFH-DA fluorescence probe assay was conducted to assess the effect of BTFS on intracellular ROS production. ROS generation was obviously elevated when A2780/CP70 cells were exposed to BTFS, compared to the vehicle-treated group (Figure 6A). Next, western blot assay was then performed to investigate the effect of BTFS on DNA damage response. The phosphorylation of histone H2A.X at serine 139 is considered as a reliable marker of DNA damage [23]. BTFS dramatically increased the phosphorylation of histone H2A.X. ATM/Chk is one of the important signaling pathways mediating DNA damage. After exposure to BTFS, both the phosphorylation of ATM at Ser1981 as well as the phosphorylation of Chk2 at Thr68 were up-regulated (Figure 6B,C), suggesting that ATM-Chk2 pathway was activated. These results suggested that BTFS promoted ROS generation and DNA damage, which is responsible for the cytotoxicity of BTFS. Moreover, BTFS induced DNA damage may be mediated by ROS generation, the specific mechanism of which needs to be further elucidated.

**Figure 6.** BTFS caused DNA damage in A2780/CP70 cells. (**A**) The effect of BTFS 3 on ROS production in A2780/CP70 cells. \* *p* < 0.05; \*\* *p* < 0.01 versus control. (**B**,**C**) The effect of BTFS on the expression of DNA damage-related proteins in A2780/CP70 cells. \* *p* < 0.05; \*\* *p* < 0.01 versus control. **Figure 6.** BTFS caused DNA damage in A2780/CP70 cells. (**A**) The effect of BTFS 3 on ROS production in A2780/CP70 cells. \* *p* < 0.05; \*\* *p* < 0.01 versus control. (**B**,**C**) The effect of BTFS on the expression of DNA damage-related proteins in A2780/CP70 cells. \* *p* < 0.05; \*\* *p* < 0.01 versus control.

AKT, MDM2, and p53 are upstream regulators of the cell cycle and apoptosis related proteins mentioned above. To further explore the mechanism underlying BTFS-induced S phase cell cycle arrest and apoptosis, we evaluated the protein expression levels of these upstream regulators in A2780/CP70 cells after BTFS exposure. Western blot analysis illustrated that p-AKT and MDM2 were dramatically decreased by BTFS, meanwhile p53 and p-p53 were increased compared with the AKT, MDM2, and p53 are upstream regulators of the cell cycle and apoptosis related proteins mentioned above. To further explore the mechanism underlying BTFS-induced S phase cell cycle arrest and apoptosis, we evaluated the protein expression levels of these upstream regulators in A2780/CP70 cells after BTFS exposure. Western blot analysis illustrated that p-AKT and MDM2 were dramatically decreased by BTFS, meanwhile p53 and p-p53 were increased compared with the controls (Figure 7A,B).

*Molecules* **2020**, *25*, 3515 9 of 17

**Figure 7.** Effect of BTFS on AKT/MDM/P53 pathway related proteins. (**A**) Western blot image (**B**) Statistical histogram of protein quantization. \* *p* < 0.05; \*\* *p* < 0.01 versus control. **Figure 7.** Effect of BTFS on AKT/MDM/P53 pathway related proteins. (**A**) Western blot image (**B**) Statistical histogram of protein quantization. \* *p* < 0.05; \*\* *p* < 0.01 versus control.

#### **3. Discussion 3. Discussion**

Ovarian cancer is considered to be one of the deadliest gynecological cancers in the world. Although platinum-based chemotherapy is the most broadly used treatment for this disease, chemoresistance and adverse side-effects still remain major obstacles to successful treatment [24]. Compounds derived from natural products have recently gained much attention in cancer therapy due to the higher biological activity and lower toxicity [25]. It has been proved that various triterpenoid saponins have extensive anticancer activity [14,15] and some triterpenoid saponins, such as Albizia gummifera saponins [26], ginsenoside 20(S)-Rg3 [27], ginsenoside Rh2 [28], Camellia oleifera Abel seed saponins [29], and *Camellia sinensis* seed saponins [30], exhibited obvious antiproliferative effect against ovarian cancer cells. In this study, we extracted and isolated the characteristic saponin complex BTFS from Baiye No.1 tea flowers, which mainly contained Floratheasaponin A and Floratheasaponin D, and found that BTFS strongly inhibited cell Ovarian cancer is considered to be one of the deadliest gynecological cancers in the world. Although platinum-based chemotherapy is the most broadly used treatment for this disease, chemoresistance and adverse side-effects still remain major obstacles to successful treatment [24]. Compounds derived from natural products have recently gained much attention in cancer therapy due to the higher biological activity and lower toxicity [25]. It has been proved that various triterpenoid saponins have extensive anticancer activity [14,15] and some triterpenoid saponins, such as Albizia gummifera saponins [26], ginsenoside 20(S)-Rg3 [27], ginsenoside Rh2 [28], Camellia oleifera Abel seed saponins [29], and *Camellia sinensis* seed saponins [30], exhibited obvious anti-proliferative effect against ovarian cancer cells. In this study, we extracted and isolated the characteristic saponin complex BTFS from Baiye No.1 tea flowers, which mainly contained Floratheasaponin A and Floratheasaponin D, and found that BTFS strongly inhibited cell proliferation of human ovarian cancer cells A2780/CP70 by induction of S phase cell cycle arrest and apoptosis.

proliferation of human ovarian cancer cells A2780/CP70 by induction of S phase cell cycle arrest and apoptosis. This research measured the cytotoxic effect of BTFS on ovarian cells and found that BTFS significantly reduced the viability of A2780/CP70 cells, whereas it hardly affected the normal human immortalized ovarian surface epithelial IOSE 364 cells at indicated concentrations. Ovarian cancer cells have unlimited proliferation ability, while healthy cells normally proliferate. BTFS can inhibit cell proliferation by inducing apoptosis and S-phase cycle arrest in A2780/CP70. BTFS may not cause apoptosis and cycle arrest of normal cells. This may be why BTFS is selective towards ovarian cancer cells. The IC50 value of BTFS on A2780/CP70 cells was calculated to be 2.6 μg/mL, lower than theaflavin-3,3′-digallate (IC50 was about 20.7 μg/mL on A2780/CP70 cells) [31], and tea seed saponins (IC50 was about 5.9 μg/mL on OVCAR-3 and A2780/CP70 cells) [30]. Besides, the results obtained This research measured the cytotoxic effect of BTFS on ovarian cells and found that BTFS significantly reduced the viability of A2780/CP70 cells, whereas it hardly affected the normal human immortalized ovarian surface epithelial IOSE 364 cells at indicated concentrations. Ovarian cancer cells have unlimited proliferation ability, while healthy cells normally proliferate. BTFS can inhibit cell proliferation by inducing apoptosis and S-phase cycle arrest in A2780/CP70. BTFS may not cause apoptosis and cycle arrest of normal cells. This may be why BTFS is selective towards ovarian cancer cells. The IC50 value of BTFS on A2780/CP70 cells was calculated to be 2.6 µg/mL, lower than theaflavin-3,3′ -digallate (IC50 was about 20.7 µg/mL on A2780/CP70 cells) [31], and tea seed saponins (IC50 was about 5.9 µg/mL on OVCAR-3 and A2780/CP70 cells) [30]. Besides, the results obtained from colony formation assay and the Western blot of PCNA further demonstrated that BTFS had obvious inhibitory effects on A2780/CP70 cells.

from colony formation assay and the Western blot of PCNA further demonstrated that BTFS had obvious inhibitory effects on A2780/CP70 cells. The cell cycle disorder leading to the abnormal cell proliferation is one of the main mechanisms of tumorigenesis [32]. The S phase in the DNA synthesis phase is when DNA is replicated; thus, it is a pivotal part of the cell cycle [33]. The cell cycle is a sequential process and any phase of arrest will cause cell proliferation stagnation or death. Numerous chemotherapeutic agents exert anticancer properties by interfering with the cell cycle. In our research, analysis of the distribution of cell cycle showed that BTFS could cause S phase cell cycle arrest in A2780/CP70 cells. This result was consistent with studies that many saponins extracted from natural plants also inhibit cancer cell proliferation by arresting cells in S phase, such as Astragalus saponins [34], Albiziae Cortex total saponins [35], and ginsenoside Re [36]. Cell cycle progression is regulated by cyclins, CDKs, and other regulatory The cell cycle disorder leading to the abnormal cell proliferation is one of the main mechanisms of tumorigenesis [32]. The S phase in the DNA synthesis phase is when DNA is replicated; thus, it is a pivotal part of the cell cycle [33]. The cell cycle is a sequential process and any phase of arrest will cause cell proliferation stagnation or death. Numerous chemotherapeutic agents exert anticancer properties by interfering with the cell cycle. In our research, analysis of the distribution of cell cycle showed that BTFS could cause S phase cell cycle arrest in A2780/CP70 cells. This result was consistent with studies that many saponins extracted from natural plants also inhibit cancer cell proliferation by arresting cells in S phase, such as Astragalus saponins [34], Albiziae Cortex total saponins [35], and ginsenoside Re [36]. Cell cycle progression is regulated by cyclins, CDKs, and other regulatory proteins [37]. Cyclin E/CDK2 complex actively participates in G1/S transition and plays important role in the initial stage

proteins [37]. Cyclin E/CDK2 complex actively participates in G1/S transition and plays important role in the initial stage of S phase. S phase progression is impacted by Cyclin A/CDK2 complex, and of S phase. S phase progression is impacted by Cyclin A/CDK2 complex, and the S/G2 transition also requires Cyclin A/CDK2 complex [38,39]. Moreover, Cyclin/CDK complexes are modulated by Cyclin-dependent kinase inhibitors and the Cdc25 phosphatase family. The p21 protein is considered as a Cyclin dependent kinase inhibitor, which inhibits the formation of Cyclins/CDK2 complex, ultimately blocking cell cycle progression from S to G2/M phase [40,41]. CDC25A, a dual specificity phosphatase, activates the Cyclin/CDK complexes, which promotes cell cycle progression, and overexpression of CDC25A will lead to abnormal cell cycle regulation and lead to tumorigenesis [42,43]. In this study, after BTFS treatment, Cyclin A2, CDK2, and Cdc25A were downregulated significantly, while p21 and Cyclin E1 proteins were significantly upregulated. Thus, these data suggest that BTFS-induced S phase arrest might be due to changes in the S phase related proteins in A2780/CP70 cells.

Apoptosis is considered one of the most widespread forms of programmed cell death, playing a pivotal role in various physiological processes and pathological conditions [44]. Inducing apoptosis is regarded as one of the major mechanisms for cancer treatments with natural compounds. In this study, the results obtained from Hoechst 33342 staining, JC-1 fluorescent staining, and flow cytometry assay suggest that BTFS induced apoptosis in A2780/CP70 cells as evidenced by much brighter and more condensed nuclei within cells, notable reduction of mitochondrial membrane potential, and higher percentage of apoptotic cells. Two well-known pathways to trigger apoptosis are the extrinsic and intrinsic apoptotic pathways. Activation of caspase-8 and caspase-9 are regarded as essential markers of the extrinsic and intrinsic apoptotic pathways, respectively, and can further activate downstream effector caspase-3 and caspase-7, which are able to mediate the cleavage of PARP, and then regulate apoptosis [45,46]. In our research, BTFS obviously activated caspase-3/7, -8 and -9 and increased the level of cleaved PARP-1, indicating that BTFS triggered both the extrinsic and intrinsic apoptotic pathways in A2780/CP70 cells. For the extrinsic pathway, tumor necrosis factor-related apoptosis-inducing ligands (TRAIL), for instance, Apo2L/TRAIL and Fas ligand (FasL), bind with their respective death receptors, such as DR4/DR5 or Fas, and subsequently interact with the adaptor molecule, Fas-associated death domain (FADD) [47]. FADD combines with caspase-8, followed by initiating the caspase-3/7 cascade reaction and ultimately cell death [48]. The current research found that BTFS evidently up-regulated DR5, Fas, and FADD, suggesting activation of the extrinsic apoptotic pathway. Previous research reported that tea seeds saponins could also increase protein expression of DR5 and FADD and induced apoptosis in ovarian cancer cell, which are consistent with our results [30]. In the intrinsic pathway, the Bcl-2 family proteins, which include pro-apoptotic (Bax, Bad, and Bak) and anti-apoptotic (Bcl-2, Bcl-xL and Bcl-B) members, are key regulators by governing the mitochondrial outer membrane permeabilization (MOMP) [30]. After the decrease of mitochondrial membrane potential caused through the increase of MOMP, cytochrome c is released from the mitochondria into the cytosol, and then binds with Apaf-1 to form the apoptosome complex, which will activate pro-caspase 9 and trigger an enzymatic cascade leading to cell death [49,50]. In this case, BTFS increased the ratio of pro- versus anti-apoptotic Bcl-2 family protein and the release of cytochrome C and Apaf-1. A similar study reported that Paris saponin II induced apoptosis in ovarian cancer cell might also result from its effects on the expression of Bax, Bcl-2, and cytochrome c proteins [11]. These results demonstrate that induction of apoptosis through both extrinsic and intrinsic pathways might also account for the anti-proliferation activity of BTFS on A2780/CP70 cells. Previously we reported the anti-cancer effect of saponin extract, which contained 14 triterpenoid saponins [17]. In this paper, a higher-purity BTFS, which mainly contained Floratheasaponin D and Floratheasaponin A, was obtained and it showed strong antiproliferative effect against A2780/CP70 cells. Though BTFS could also induce apoptosis and cell cycle arrest, the specific contribution of Floratheasaponin D and Floratheasaponin A remains to be further elucidated. Moreover, structure-activity relationship may play a pivotal role in its anticancer effect. It has been reported that the differences of saponin type, position, and sugar moieties affect anticancer activity [51,52]. BTFS was classified as triterpenes saponins in our study; we obtained a mixture of three saponin monomers, which makes it complex to evaluate the anticancer effect and

the chemical structures. However, the specific structure-activity relationship of tea flower saponins is worth investigation in future work after we obtain the monomers.

Inducing intracellular ROS generation is considered as an important mechanism of the anticancer effect of most drugs [53]. Excessive ROS generation could cause DNA double-strand break, DNA locus mutation, and other forms of DNA damage [54]; and elevated ROS levels have been regarded as a causative trigger for DNA damage [55]. Cell cycle checkpoints and effector kinases could be activated to regulate cellular decision among cell cycle arrest, apoptosis, or other cell death modalities in response to ROS-mediated DNA damage [56]. In our research, we demonstrated ROS generation and DNA damage after treatment with BTFS in A2780/CP70 cells. Ataxia telangiectasia mutated (ATM) kinase is an important DNA damage sensor for oxidative stress response. When responding to DNA damage, ATM is activated by autophosphorylation and further phosphorylates downstream substrates such as Chk1 and Chk2 [57,58]. Furthermore, activated Chk1 and Chk2 cause the phosphorylation of their downstream effectors, for instance Cdc25A, which could mediate the transformation of cell cycle S/G2 by regulating CDK2 activity [59]. Moreover, Chk2 could phosphorylate and activate p53 protein directly [60]. This experiment revealed that p-ATM and p-Chk2 were remarkably increased by BTFS in A2780/CP70 cells, suggesting that the ATM-Chk2 pathway was activated following DNA damage.

A serine/threonine kinase named AKT can be activated by autophosphorylation at Ser473 and Thr308, thereby inhibiting apoptosis and promoting cell survival. MDM2 is one of the downstream proteins of AKT, which can promote cell proliferation and growth, and is related to the development of a variety of tumors [61]. MDM2 is also a ubiquitin ligase of p53, which can be phosphorylated by p-AKT at Ser166 and Ser186 and enter the nucleus to form a complex with p53, thereby inhibiting p53 activity [62]. P53 is an important tumor suppressor, which plays an important role in cellular response. DNA damage can induce p53 phosphorylation at Ser15 and Ser20 sites, weakening the interaction between p53 and MDM2 and inducing apoptosis, cell cycle arrest, etc. [62,63]. P53 mainly induces apoptosis by activating the death receptor pathway and the mitochondrial pathway; meanwhile, p53 can promote apoptosis by activating death receptors such as Fas, DR4, and DR5 located on the cell membrane. On the other hand, Bcl-2 family proteins as well as mitochondria proteins are also associated with p53-dependent apoptosis [64]. It has been reported that triterpenoid saponin ginsenoside Rh2 can activate the p53 pathway to induce apoptosis in colorectal cancer cells [64], and can induce apoptosis in human epidermal cancer cells by inhibiting AKT activity [65]. Based on our study, we found that BTFS can significantly reduce p-AKT and MDM2 and increase p53 and p-p53 protein levels in A2780/CP70 cells, suggesting that BTFS can reduce the cross-linking of MDM2 and p53 by inhibiting AKT autophosphorylation, and thus effectively enhance p53 activity and promote cell apoptosis. Our finding is consistent with the study that the active tea component theaflavins can induce apoptosis of ovarian cancer cellA2780/CP70 through AKT-MDM2-p53 pathway [31].

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

### *4.1. Materials and Reagents*

RPMI-1640 medium, Fetal bovine serum (FBS), and DMSO were purchased from Sigma (St. Louis, MO, USA). Caspase-Glo 3/7, Caspase-8, Caspase-9, and CellTiter 96 Aqueous kit were purchased from Promega (Madison, WI, USA). Propidium iodide (PI) and Alexa Fluor 488 Annexin V were purchased from Invitrogen (Waltham, MA, USA). Antibodies against PCNA, p21, p27, CDK2, Cdc25A, Cdc25C, CyclinA2, Cyclin E1, ATM, p-ATM (Ser1981), γ-H2AX (Ser139), PARP, Fas, DR5, FADD, Bax, Bcl-2, Bcl-xL, Apaf-1, Cytc, p53, p-p53 (Ser15), AKT, p-AKT (Ser473), and MDM2 were purchased from Cell Signaling Technology (Beverly, MA, USA). Antibodies against Chk2, p-Chk2 (Thr 68), Bad, and GAPDH were purchased from Santa Cruz Biotechnology (Dallas, TX, USA).

### *4.2. Extraction and Identification of Baiye No.1 Tea Flower Saponins (BTFS)*

Dried tea flowers of tea Baiye No.1 variety were obtained from Zhe-jiang Yilongfang Co., Ltd. (Kaihua, Zhe-jiang, China). After smashing, the powder was distilled at 80 ◦C for 45 min. The aqueous extract was separated by AB-8 macroporous adsorption resin, followed by preparative HPLC to obtain BTFS. BTFS was then identified by UPLC-Q-TOF/MS/MS system with ultra-high performance chromatography (Agilent, Santa Clara, CA, USA) coupled with electro-spray ionization quadrupole time-of-flight mass spectrometry (AB SCIEX, Framingham, MA, USA) according to the methods we described previously, with certain modifications [20]. BTFS was dissolved in dimethyl sulfoxide (DMSO) to make a 20 mg/mL stock solution and stored at −20 ◦C before use.

### *4.3. Cell Lines and Cell Culture*

Human ovarian cancer cell line A2780/CP70 was provided by Dr. Bing-Hua Jiang from Thomas Jefferson University, and the normal ovarian surface epithelial cell line IOSE-364 was provided by Dr Auersperg from the University of British Columbia. The cell lines were cultured in RPMI-1640 medium with 10% FBS, and incubated at 37 ◦C in a humidified incubator with 5% CO2.

### *4.4. Cell Viability Assay*

A2780/CP70 and IOSE-364 cells were seeded in 96-well plates at a density of 2 × 10<sup>4</sup> cells/well incubated overnight, and treated with BTFS (1.0, 1.5, 2.0, 2.5, 3.0, 3.5 µg/mL) or an equal concentration of DMSO (as vehicle) for 24 h. CellTiter 96 Aqueous kit was then used to assess the cell viability according to the manufacturer's protocol. Results were presented as percentage of control.

### *4.5. Colony Formation Assay*

A2780/CP70 cells were seeded into 6-well plates at 6 × 10<sup>5</sup> cells/well incubated overnight and treated with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h. The cells were then cultured for 7 days with drug-free medium at 2 × 10<sup>3</sup> cells/well. Colonies were fixed with ice-cold methanol for 10 min, then stained with 0.5% crystal violet solution in 25% methanol for another 10 min. The plates were rinsed with distilled water carefully several times and directly photographed. Image J software (Bethesda, MD, USA) was used to count the number of colonies; the results were adjusted to the percentage of control.

### *4.6. Cell Cycle Analysis by Flow Cytometry*

After treating with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h, the cells were washed with cold PBS twice and then fixed with ice-cold 70% ethanol at 4 ◦C overnight. Afterwards, the cells were washed twice with PBS and incubated with 180 µg/mL RNase (Invitrogen) for 15 min at 37 ◦C, then incubating with 50 µg/mL propidium iodide (PI) solution (Sigma) for 15 min in the dark at 37 ◦C. The cells were then analyzed by FACSCalibur flow cytometry (BD Biosciences, San Jose, CA, USA) and data were analyzed by FCS software (De Novo Software, CA, USA).

### *4.7. Hoechse 33342 Staining*

The cells were seeded in 96 well plates at density of 2 × 10<sup>4</sup> cells/well and incubated overnight, after treating with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h. The cells were stained with 10 µg/mL Hoechst 33342 (Sigma) in PBS for 15 min in the dark at 37 ◦C. Fluorescence microscope (ZEISS, Heidelberg, Germany) was used to examine the cellular morphology.

### *4.8. Evaluation of Mitochondrial Membrane Potential*

The cells were seeded in black 96 well plates at density of 2 × 10<sup>4</sup> cells/well and incubated overnight, and treated with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h. After washing twice, the cells were incubated with 10 µg/mL JC-1 solution for 30 min at 37 ◦C. The fluorescent intensity was measured using Synerg HT Multi Mode Microplate reader (BioTek, Winooski, VT, USA) at excitation: emission of 485/590 and 485/528 for red aggregates and green monomers, respectively.

### *4.9. Apoptosis Analysis by Flow Cytometry*

The apoptotic cells were evaluated using an Alexa Fluor 488 Annexin V/Dead Cell Apoptosis Kit (Invitrogen, Waltham, MA, USA). After treating with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h, the cells were harvested and centrifuged for 8 min at 1000 rpm; after washing twice with PBS, the cells were suspended in Annexin-binding buffer with Alexa Fluor 488 Annexin V and PI solution for in dark 15 min at room temperature. The stained cells were then analyzed by FACSCalibur flow cytometry (BD Biosciences) with fluorescence emission at 530 nm and excitation at 488 nm.

### *4.10. Cellular Caspase Activity Assay*

The cells were seeded in 96 well plates at density of 2 × 10<sup>4</sup> cells/well and incubated overnight, after treating with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h. Caspase Glo-3/7, -8 and -9 regents (Promega, Madison, WI, USA) were added to each well and incubated for 30 min at 37 ◦C. Luminescence was measured by Synerg HT Multi Mode Microplate reader (BioTek). Total protein levels were detected with a BCA assay kit to normalize the caspases activities. The results were presented as percentage of control.

### *4.11. Detection of Intracellular ROS Production*

Intracellular ROS production was measured by peroxide-sensitive fluorescent probe DCFH-DA. The cells were treated with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h, followed by treatment with 10 µM DCFH-DA for 30 min at 37 ◦C. The fluorescence intensity was measured with excitation at 485 nm and emission at 528 nm by Synergy HT Multi-Mode Microplate Reader (BioTek). Total protein levels were detected by BCA assay kit to normalize the ROS generation. The results were presented as percentage of control.

### *4.12. Western Blotting*

Cells were treated with BTFS (0, 1.5, 2.0, 2.5 µg/mL) for 24 h and then harvested and extracted with Mammalian Protein Extraction Reagent (Pierce, Rockford, IL, USA) supplemented with Halt Protease and Phosphatase Inhibitor (Pierce). BCA protein assay kit (Pierce) was used to determine the total protein concentrations of cell lysates. Protein samples of cell lysates were mixed with equal amount of SDS loading buffer and denatured by heating for 8 min, then separated by SDS-polyacrylamide gel electrophoresis before transferring onto nitrocellulose membranes. The membranes were blocked with 5% non-fat milk in Tris-buffer saline, which contained 0.1% Tween 20 (TBST) for 1 h at room temperature. Specific primary antibodies were used and the membrane was incubated at 4 ◦C overnight. After washing with TBST, appropriate secondary antibodies conjugated with horseradish peroxidase were used. The antigen-antibody complex was visualized with the ECL Western blot detection kit (Bio-Rad) and ChemiDoc MP System (Bio Rad, Hercules, CA, USA). Image J software was used to quantitate the protein bands; the indicated protein was then normalized by GAPDH.

### *4.13. Statistical Analysis*

Data were presented as mean ± standard deviation (SD) for at least three independent experiments. One-way ANOVA followed with Dunnett′ s test was performed by SPSS software (IBM, Version 22.0, Armonk, NY, USA). The criterion for statistical significance were *p* < 0.05 and *p* < 0.01.

### **5. Conclusions**

The high-purity standardized saponin extract we obtained from Baiye No.1 tea flower, namely BTFS, shows a strong antiproliferative effect against A2780/CP70 cells at low concentrations, while causing

less cytotoxicity to normal cells. BTFS caused cellular DNA damage in A2780/CP 70 cells and induced cell cycle arrest in the S phase via regulating ATM-Chk2 signaling pathway related proteins. BTFS also trigged both intrinsic and extrinsic apoptosis in A2780/CP70 cells through the AKT-MDM2-p53 signaling pathway. Based on our results, BTFS has the potential to serve as a nutraceutical for the prevention and treatment of ovarian cancer. The specific contribution of Floratheasaponins A and D to the antiproliferative activity of A2780/CP70 cells needs further investigation.

### **Supplementary Materials:** The Supplementary Materials are available online.

**Author Contributions:** Conceptualization, L.C., Y.T. and Y.C.C; methodology, L.C. and Y.C.C.; formal analysis, L.C., N.R. and Y.W. (Yuanyuan Wu); resources, Y.T., Y.R. and G.O.R.; writing—original draft preparation, L.C., B.L. and Y.W. (Yaomin Wang); writing—review and editing, Y.W. (Yaomin Wang), Y.T. and Y.C.C.; funding acquisition, Y.T. and Y.C.C., Y.W. (Yaomin Wang); All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Key Research and Development Projects in Zhejiang Province "Industrialization Model Projects on Exploring Functional Components and Related Products from Tea Flowers and Fruits" (2018C02012), the National Key Research and Development Plan "Processing of Tea Products, Key Technology of Quality Control during Storage and Equipment Development" (2017YFD0400803), the Natural Science Foundation of Hubei Province (Grant No: 2019CFB187). This research was also supported by NIH grants P20RR016477 from the National Center for Research Resources and P20GM103434 from the National Institute for General Medical Sciences (NIGMS) awarded to the West Virginia IDeA Network of Biomedical Research Excellence. This study was also supported by Grant Number P20GM104932 from NIGMS, a component of the National Institutes of Health (NIH) and its contents are solely the responsibility of the authors and do not necessarily represent the official view of NIGMS or NIH. This study was also supported by COBRE grant GM102488/RR032138, ARIA S10 grant RR020866, FORTESSA S10 grant OD016165, and INBRE grant GM103434.

**Acknowledgments:** We would like to thank Kathy Brundage from Flow Cytometry Core at West Virginia Excellence.

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

### **References**


**Sample Availability:** Samples of the compounds BTFS are not available from the authors.

© 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* **Targeting Toxins toward Tumors**

**Henrik Franzyk and Søren Brøgger Christensen \***

Department of Drug Design and Pharmacology, University of Copenhagen, Universitetsparken 2, DK-2100 Copenhagen Ø, Denmark; henrik.franzyk@sund.ku.dk

**\*** Correspondence: soren.christensen@sund.ku.dk

**Abstract:** Many cancer diseases, e.g., prostate cancer and lung cancer, develop very slowly. Common chemotherapeutics like vincristine, vinblastine and taxol target cancer cells in their proliferating states. In slowly developing cancer diseases only a minor part of the malignant cells will be in a proliferative state, and consequently these drugs will exert a concomitant damage on rapidly proliferating benign tissue as well. A number of toxins possess an ability to kill cells in all states independently of whether they are benign or malignant. Such toxins can only be used as chemotherapeutics if they can be targeted selectively against the tumors. Examples of such toxins are mertansine, calicheamicins and thapsigargins, which all kill cells at low micromolar or nanomolar concentrations. Advanced prodrug concepts enabling targeting of these toxins to cancer tissue comprise antibody-directed enzyme prodrug therapy (ADEPT), gene-directed enzyme prodrug therapy (GDEPT), lectin-directed enzymeactivated prodrug therapy (LEAPT), and antibody-drug conjugated therapy (ADC), which will be discussed in the present review. The review also includes recent examples of protease-targeting chimera (PROTAC) for knockdown of receptors essential for development of tumors. In addition, targeting of toxins relying on tumor-overexpressed enzymes with unique substrate specificity will be mentioned.

**Keywords:** chemotherapy; prodrug; drug targeting; overexpressed enzymes; ADC; ADEPT; GDEPT; LEAPT; PROTAC

### **1. Introduction**

According to the International Union of Pure and Applied Chemistry (IUPAC), prodrugs are defined as chemically modified drugs that undergo biological and/or chemical transformation(s) before eliciting pharmacological responses [1]. Drugs may be converted into prodrugs in order to: (i) increase their bioavailability, (ii) target the drugs toward tissues such as tumors, (iii) decrease toxicity, (iv) increase chemical stability, (v) increase solubility, or (vi) mask unpleasant taste [2,3]. Prodrugs are formed by covalent attachment of the drug to a carrier, also often termed as the promoiety, which is subjected to cleavage within the body to release the active drugs. Promoieties and their degradation products should be nontoxic and nonimmunogenic [2]. Pharmacologically inactive compounds, which in the organism are modified into active drugs, are known as bio-precursor prodrugs. Examples of such prodrugs are proguanil that in the liver is converted into the antimalarial drug cycloguanil [4], salicin that is converted into salicylic acid [5], and acetanilide that is converted into acetaminophen [3]. Both salicylic acid and acetaminophen are antipyretics and analgesics. Another class of prodrugs is the co-drugs that consist of two drugs covalently attached to each other—either directly or via a linker in a way so that they act as promoieties for each other [2,6]. Examples of co-drugs comprise sulfasalazine, which in the body is degraded to 5-aminosalicylic acid and sulfapyridine [6,7], and benorylate, which is an ester of acetylsalicylic acid with paracetamol [6].

A review focused on the prodrug approach revealed the state of the art in 2017 [8]. In the present review new developments in the fields of boronic acids as prodrugs, of anthracyclines, and of antibody–drug conjugates, targeting of paclitaxel and refined use of

**Citation:** Franzyk, H.; Christensen, S.B. Targeting Toxins toward Tumors. *Molecules* **2021**, *26*, 1292. https:// doi.org/10.3390/molecules26051292

Academic Editors: Marialuigia Fantacuzzi and Alessandra Ammazzalorso

Received: 9 January 2021 Accepted: 22 February 2021 Published: 27 February 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

prostate-specific membrane antigen (PSMA) for delivery of payloads are described. The problem of targeting oncogenes for neoplastic tissue in gene-directed enzyme prodrug therapy is discussed (which often has been omitted in previous reviews). Finally, a new group of prodrugs, which link two pharmacophores, i.e., PROtease Targeting Chimeras (PROTACs; [9,10]) and lectin-directed enxym activated prodrugs, are discussed. One of these pharmacophores has a high affinity for an E3 ubiquitin ligase (an enzyme that, with assistance of an E2 ubiquitin-conjugating enzyme, transfers ubiquitin to its protein substrate), while the other has an affinity for the targeted protein e.g., a receptor or ion channel. After binding of the protein and the ubiquitin ligase, the biomolecule will be modified with ubiquitin and subsequently cleaved by a 26S proteasome, which degrades ubiquitinated proteins [9,10].

Thus, the present review comprises the following: (i) prodrugs cleaved in the acidic microenvironment of cancer cells (Section 2.1.1), (ii) prodrugs cleaved by reactive oxygen species (ROS) in cancer cells (Section 2.1.2), (iii) prodrugs cleaved by glutathione (Section 2.1.3), (iv) prodrugs cleaved by enzymes overexpressed in cancer cells ( Section 2.1.4), (v) prodrugs cleaved by glucuronidase (Section 2.1.5), (vi) prodrugs cleaved by prostatespecific antigen (PSA) or PSMA (Section 2.1.6), (vii) antibody–drug conjugates (Section 2.2), (viii) antibody-directed enzyme prodrug therapy (Section 2.3), (ix) gene-directed enzyme prodrug therapy (Section 2.4), (x) lectin-directed enzyme-activated prodrug therapy (Section 2.5), and (xi) protease-targeting chimeras (Section 3).

### **2. Prodrugs**

### *2.1. Targeting by Selective Cleavage of Prodrugs in the Microenvironment of Cancer Cells*

The metabolism in cancer cells involves a high rate of anaerobic glycolysis resulting in an overproduction of lactic acid and carbonic acid. Since the acid protons of these acids are exported into the extracellular medium, intracellular pH of cancer cells typically is higher (pH 7.4 versus 7.2 in normal cells). [8,11,12]. As a consequence of the proton transport the microenvironment surrounding cancer cells has a pH of 6.8 in contrast to the is estimated pH of blood 7.4 [13]. A high rate of glycolysis without oxygen supply causes hypoxia inside cancer cells [8,11,14,15]. The ensuing rise in the intracellular level of reactive oxygen species (ROS), such as hydrogen peroxide, formed during hypoxia conditions may be employed as a mode of targeting drugs toward cancer cells. Cancer cells also have an increased level of glutathione [16], β-glucuronidase [17], and some specific proteolytic enzymes [17–19]. β

### 2.1.1. Prodrugs Cleaved in Acidic Media

Salts of dithiocarbamates (e.g., **1**) are stable at physiological pH, but they will after protonation in the acidic microenvironment surrounding cells undergo cleavage to release the free amine (**2**) and carbondisulfide (Scheme 1). In a similar way the emitine monoamide (**3**) with 2-methylmaleic acid was used to simultaneously increase the solubility of emitine in water and enable its release from the prodrug in slightly acidic media [20].

**Scheme 1.** Dithiocarbamates and methylmaleic amides cleaved at low pH [20].

A prodrug of doxorubicin (DOXO-EMCH, **4**) is linked via a hydrazone promoiety to a maleimide *N*-substituted with a 6-aminohexanoic spacer. This prodrug is designed to enable reaction with Cys-34 of serum albumin (to give **5**), present as the most abundant protein in blood. The resulting albumin-linked drug (**5**) cannot penetrate into cells until the hydrazone is hydrolyzed in the acidic environment around cancer cells to provide the free doxorubicin (**6**; Scheme 2). The clinical phase III trial, however, did not enable registration of the compound as a drug [14,21,22]. Despite these preliminary results, attempts to overcome the cardiotoxicity of DOX-EMCH continue.

**Scheme 2.** Linkage of doxorubicin to serum albumin to prevent its penetration into benign cells; only in the acidic microenvironment of cancer cells is doxorubicin released [14,21,22].

Polyethylene glycol-coated liposomes encapsulating doxoxrubicin (Caelyx) show less cardiotoxicity than doxorubicin itself. The drug is used for treatment of breast cancer and ovarian cancer [23].

Paclitaxel is widely used for treatment of various cancers; however, poor solubility limits its use. The formulation vehicle Cremophor EL (CrEL; macrogolglycerol ricinoleate; polyethylene glycol (PEG)-35 castor oil) and ethanol are used to increase its solubility, but unfortunately CrEL causes side effects like hypersensitivity, neurotoxicity and nephrotoxicity [24]. These side effects have been overcome by a new formulation of lyophilized paclitaxel with serum albumin (i.e., Abraxane®) that provides nanoparticles (average size: 130 nm). This drug formulation has been approved by the U. S. Food and Drug Administration (FDA) for treatment of pancreatic cancer and non-small cell lung cancer (NSCLC) [25]. Moreover, paclitaxel (**7**) has been attached to nanoparticles via acetal linkages (**8**). Acetals are cleaved in the acidic environment of cancer cells (Scheme 3) [26]. This method is also used for conjugation of paclitaxel to nanoparticles prepared from polyethylene glycols (PEGs) [13].

### 2.1.2. Prodrugs Cleaved by ROS

NH

O O

O Polymer O

O

O O O O OH O O O HO O O O H3O + O O O O OH O O O HO NH OH O O O O OH O Polymer O O H + + **8 7** In normal cells, ATP is primarily produced by oxidative phosphorylation, whereas in cancer cells ATP primarily is produced by anaerobic glycolysis (the Warburg effect) [27,28]. The anaerobic pathway stimulates generation of ROS such as hydrogen peroxide [28]. The presence of hydrogen peroxide under certain physiological conditions can be used to facilitate cleavage of arylboronic acids, or esters thereof, to give phenols and boric acid [15]. The concentration of hydrogen peroxide in benign cells is estimated to approx. 1 µM, but in cancer cells it may reach even 10 µM [15]. Some boronic acids are oxidized by cytochrome P450 [15]. Boronic acids may also be oxidized by peroxynitrite [15]. It may be questioned whether arylboronic functionalities act as true promoieties, since the oxidation provides

boric acid and the corresponding phenol. As an example, camptothecin-10-boronic (**9**) acid is oxidatively cleaved by hydrogen peroxide to give 10-hydroxycamptothecin (**10**; Scheme 4) [29]. The resulting 10-hydroxycamptothecin (**10**) proved to be a more potent topoisomerase inhibitor, and to be more cytotoxic in a number of cell systems than the original drug. Furthermore, this hydroxylated derivative exhibited tumor growth inhibition in xenograft models [29].

**Scheme 3.** Conjugation of paclitaxel (**7**) to polymers that increase solubility [26].

**Scheme 4.** Oxidation of camptothecin-10-boronic acid (**9**) to give 10-hydroxycamptothecin (**10**) [15].

Arylboronic acid prodrugs of doxorubicin (e.g., **11**) have also been reported. Oxidative cleavage of the boronic acid moiety releases a phenol that spontaneously cleaves itself from the self-immolative spacer 4-hydroxybenzyl carbamate (**12**; Scheme 5) [15]. Doxorubicin (**6**) very efficiently kills cancer cells, however, a severe cardiotoxicity limits its use as a chemotherapeutic drug [30]. Targeting of the drug may reduce this side effect. The prodrug was found to induce regression of pancreatic tumors in mice, and further analysis revealed that the prodrug was cleaved to doxorubicin inside tumors (Scheme 5) [31].

Similarly, an aryl boronic acid prodrug (**13**) of paclitaxel (**7**), also containing a selfimmolative linker, has been reported (Scheme 6) [15]. The size of the PEG moiety was adjusted so that the prodrug self-assembles into micelles with a size of ca. 50 nm. Native paclitaxel (**7**) was only released in the acidic microenvironment of the cells containing a high level of ROS. Consequently, reduced toxicity of the prodrug as compared to treatment with paclitaxel was observed in mice while retaining similar tumor regression [32]. At present no boronic acid prodrugs have been approved, despite intensive research being performed in the field [15].

### 2.1.3. Prodrugs Cleaved by Glutathione

Glutathione (H-γGlu-Cys-Gly-OH) is present in almost all mammalian tissues, but usually it is overexpressed in cancer cells [33]. The active functionality of glutathione is the thiol, which enables the molecule to participate in redox reactions, and may thus protect the cell from a high level of ROS [33]. The molecule is able to cleave disulfides including linkages within prodrugs. Hence, this feature has been utilized in the construction of prodrugs attached to a promoiety via a disulfide linkage, and e.g., camptothecin (**14**) has been linked to a near-infrared (NIR) dicyanomethylenebenzopyran fluorophore (**15**; Scheme 7). An in vivo experiment using a mouse BCap-37 tumor xenograft model showed a significantly improved regression of tumors on mice treated with the prodrug as compared

to that found for those treated with camptothecin or another prodrug in which the linkage consisted of a stable carbon–carbon bond instead of the disulfide bond [16].

**Scheme 5.** Prodrug of doxorubicin (**6**) cleaved by ROS [15].

**Scheme 6.** Polyethylene glycol (PEG)-containing prodrug (**13**) of paclitaxel (**7**) that is cleaved by reactive oxygen species (ROS) [13].

### 2.1.4. Prodrugs Cleaved by Expressed Enzymes

A number of enzymes are overexpressed in cancer cells. These enzymes include oxidoreductases, hydrolases, and matrix metalloproteinases (MMPs) [8]. A prodrug based on the overexpression of MMPs was designed for doxorubicin (Scheme 8) [34]. The peptide promoiety conjugated to the amine in doxorubicin prevents entry into cells, and consequently the compound is harmless to benign cells. By contrast, in the microenvironment of cancer cells the peptide is cleaved at the Gly-hPhe position. Proteases subsequently remove

the remaining amino acids. In a HT1080 xenograft mouse preclinical model, the prodrug was more efficient in reducing tumor growth than doxorubicin itself, and less undesired toxicity was observed [34].

γ

**Scheme 7.** Prodrug (i.e., **15**) of camptothecin (**14**) that is cleaved by glutathione [16].

**Scheme 8.** A doxorubicin (**6**) prodrug (i.e., **16**) cleaved by matrix metalloproteinase (MMP) and other proteases (Cit = citrulline; hPhe = homophenylalanine) [34].

β

β

Cathepsin B is involved in cancer invasion and metastasis, and it is overexpressed in cancer tissue [35]. A prodrug (**17**) consisting of doxorubicin conjugated via the selfimmolative linker 4-aminobenzyl alcohol to a dipeptide fragment (Ac-Phe-Lys-OH), which is a substrate for cathepsin B, has been designed (Scheme 9). In a mouse model this prodrug inhibited development of peritoneal carcinomatosis as well as its progression more efficiently and with fewer side effects than doxorubicin itself [35].

**Scheme 9.** Prodrug of doxorubicin (i.e., **17**) that is cleaved by cathepsin B [35].

#### β 2.1.5. Prodrugs Cleaved by β-glucuronidase

β Glucuronides are formed as a phase II metabolism of drugs. β-Glucuronidases are expressed excessively in a number of tumors such as breast, lung and gastrointestinal tract carcinomas as well as in melanomas, where they are particularly abundant in necrotic areas [17,36,37]. Prodrugs based on this selective enzyme distribution include glucuronides of doxorubicin (**6**) and 4′ -epi-doxorubicin (**19**) [17]. A self-immolative linker was introduced, since the β-glucuronide (i.e., **18**) conjugated directly to the sugar part of doxorubicin was not a substrate for β-glucuronidase (Scheme 10) [17]. ′ β β

′ **Scheme 10.** Glucuronides (**18** and **20**) of doxorubicin (**6**) and 4′ -epi-doxorubicin (**19**) [17].

In an attempt to circumvent the poor solubility of paclitaxel (**7**) in water a glucuronide (i.e., **21**) was made. The resulting glucuronide (**21**) proved indeed to be soluble in water,

β

and it was rapidly converted into paclitaxel (**7**) in the presence of high concentrations of β-glucuronidase (Scheme 11) [38]. β

′

′

β

β

**Scheme 11.** Water-soluble glucuronide-based prodrug (**21**) of paclitaxel (**7**) [38].

Likewise, 7-Aminocamptothecin was conjugated to glucuronic acid via another selfimmolative linker (Scheme 12).

β **Scheme 12.** Glucuronide of 7-aminocamptothecin (**23**) cleaved by β-glucuronidase [39].

### 2.1.6. Prodrugs Cleaved by PSA or PSMA

Prostate cancer is a slowly developing cancer disease. In high-income countries it is the cancer disease that causes second-most deaths among men. In its initial stages the disease can be treated with androgen ablation therapy. However, if progression of disease occurs

during treatment with anti-androgens, resulting in development of distant metastasis, the prostate cancer is defined as metastatic Castration-Resistant Prostate Cancer (mCRPC) which is not sensitive to hormone ablation [40]. All tumors of mCRPCs secrete the enzyme prostate-specific antigen (PSA) into their microenvironment. PSA is a chymotrypsin-like protease with a unique substrate specificity [18]. PSA also diffuses into the bloodstream, but PSA in the blood is inactivated by complexing with blood proteins like serum albumin [40]. Conjugation of cytotoxins with different selectively labile peptides has been used for targeting of mCRPCs. Thus, *O*-desacetylvinblastine (**24**) has been targeted to mCRPCs by conjugation to a PSA-specific peptide substrate to give a prodrug (i.e., **25**) (Scheme 13) [19].

**Scheme 13.** Prodrug (i.e., **25**) of desacetylvinblastine (**24**) cleaved by prostate-specific antigen (PSA) [19].

PSA cleaves the peptide between the two Ser residues adjacent to the C-terminal Pro residue, whereupon a spontaneous intramolecular attack of the amino group of the terminal Ser on the Pro carboxylate affords desacetylvinblastine (**24**) and a diketopiperazine. This intramolecular diketopiperazine formation appeared to depend on the presence of the Pro residue, as it did not occur when Leu was incorporated instead. Desacetylvinblastine (**24**) proved equally efficient in inducing mitotic arrest as the original vinblastine (Scheme 13) [19].

A number of peptides have been conjugated to doxorubicin (**6**) [41]. Among these different promoities the peptide H2N-Glu-Hyp-Ala-Ser-Chg-Ser-Leu-OH was found to afford a prodrug that was efficiently cleaved by PSA, and it showed a dramatically increased activity in reducing LNCaP xenografts in mice as compared to that of native doxorubicin (**6**). The released active drug consists of a mixture of doxorubicin and H-Leu-doxorubicin (Scheme 14) [41].

A drawback of using vinblastine or doxorubicin (**6**) as drugs for treatment of prostate cancer is that both compounds cause mitotic arrest, and consequently they primarily target proliferating cells. A more pronounced cell death is expected when toxins capable of killing cells in all stages are used. Thapsigargin (**27**) is a cytotoxic compound that kills cells in all states by blocking the sarco/endoplasmic reticulum Ca2+ ATPase (SERCA), thereby inducing the unfolded protein response leading to apoptosis [42]. The general toxicity of the compound requires targeting via a prodrug in order to avoid general systemic toxicity [43]. No obvious anchoring point for conjugation to a peptide exists in native thapsigargin, but replacement of the butanoyl moiety with a 12-aminododecanoic acid spacer (to give **28**) allows for introduction of an amine functionality that may serve as attachment point for a peptide promoiety (Scheme 15).

H-Glu-Hyp-Ala-Ser-Chg-Ser-Leu-OH H-Glu-Hyp-Ala-Ser-Chg-Ser-OH

**Scheme 14.** Doxorubicin prodrug (**26**) cleaved by PSA [41].

ω **Scheme 15.** Prodrug of ω-aminododecanolydesbutanoylthapsigargin (**28**) cleaved by PSA [43].

Conjugation of a thapsigargin analogue to a polar peptide was expected to inhibit diffusion into cells, thereby preventing the prodrug from reaching the intracellular SERCA pump. The hexapeptide H-His-Ser-Ser-Lys-Leu-Gln-OH is very efficiently released from

β

β

γ

the prodrug by PSA, but only to a limited extent by any other proteases in the human body [18]. The C-terminal Leu residue was introduced in order to make the prodrug a substrate for PSA. In vivo experiments in mice confirmed that the prodrug was only cleaved in the blood to a limited extent, but very efficiently cleaved by PSA to release the active Leu derivative within tumors. Thus, this prodrug prevented growth of tumors in mice [44].

KLK2 is another enzyme (previously named human kallikrein 2, hK2) that may be used for targeting drugs toward prostate cancers [45,46]. Similarly to PSA, KLK2 is also secreted from the prostate and prostate cancer cells. The level of KLK2 in blood can be used as a biomarker for prostate cancer, and KLK2 is inactivated upon entering the bloodstream by binding to blood proteins [47].

An alternative enzyme, characteristic for the prostate glandule, is prostate-specific membrane antigen (PSMA). The catalytic site of this enzyme extends outwards into the extracellular environment. The enzyme is not only expressed in the prostate glandule, but also in human prostatic carcinoma and in neovascular tissue of a number of tumors [48,49]. In a healthy individual the enzyme is exclusively expressed in the prostate, ensuring that cleavage of the peptide conjugated to its payload solely occurs in the prostata glandula [49–53]. As the only enzyme outside the central nervous system PSMA cleaves the amide linkage in the γ-Glu tetramer [54]. Taking advantage of this feature, a prodrug of the 12-aminododeanoate of desbutanoylthapsigargin (**28**), (i.e., mipsagargin, **31**) was made (Scheme 16). The C-terminal βAsp residue was introduced in order to make the prodrug a substrate for PSMA [49]. This prodrug, which has been prepared by solid-phase synthesis [55], was cleaved rapidly in tumors to release the βAsp derivative, which slowly was cleaved to provide the free mipsagargin. A solid-phase synthesis of these guaianolide prodrugs were developed [55].

In the clinical phase II trial this prodrug conferred a prolonged stabilization of the disease in patients with hepatocellular carcinoma [56]. Hepatocellular cancer also expresses PSMA in neovascular tissue, which thus should be sensitive to mipsagargin [57]. However, the results obtained in clinical phase II trial did not meet the expectations, and hence the drug was not marketed [58]. A major problem might be that the prodrug, despite its five negative charges on its peptide side chains, has been found to be able to penetrate cell membranes in benign cells, thereby causing unspecific toxicity [59].

### *2.2. Antibody–Drug Conjugates (ADCs)*

An antibody–drug conjugate (ADC) is a prodrug consisting of a monoclonal antibody conjugated to a cytotoxin (i.e., the payload) via a linker. The prodrug is designed based on the hypothesis that appropriate antibodies that preferentially bind to cancer-specific antigens (located on the surface of cancer cells) can be obtained. Indeed, the two first ADCs, Adcetris® (brentuximab vedotin) for Hodgkin lymphoma and Kadcyla® (trastuzumab emtansine) for breast cancer, were approved by the FDA in 2011 and in 2013, respectively. Since then six additional ADCs have been approved by the FDA (i.e., Besponsa® (inotuzumab ozogamicin), Mylotarg® (gemtuzumab ozogamicin), Polivy®, Enhertu®, Padcev®, and Trodelvy®) [61]. More than 60 ADC prodrugs are under clinical development [62–64].

The linker should be stable in circulation to avoid release of free cytotoxin causing systemic toxicity, as is often seen with conventional chemotherapeutics (Figure 1).

Monoclonal antibodies of rodent origin may cause severe immunogenic reactions in humans. The use of chimeric humanized antibodies has to some extent solved this problem [65].

The ideal cytotoxin should be very toxic since only a limited number of antigens are present on the surface of malignant cells, and consequently only a limited number of ADCs can be internalized to exert the cytotoxic effect [66]. A number of toxins have been used as payloads, e.g., the peptide monomethyl auristatin E, the polyketide macrolides ansamitocin (as a mixture of aliphatic alkyl esters) and maytansins (= maitansins) as well as doxorubicin

(**6**), duocarmycin (**32**), and enediynes like calicheamicin γ<sup>1</sup> I (**33**) have been converted into ADCs (Figure 2) [61,63].

**Scheme 16.** The PSMA-sensitive prodrug mipsagargin (**31**) [56,60].

**Figure 1.** Cartoon illustrating the structure of an antibody–drug conjugate (ADC) drug. γ

α β

β

γ α β

**Figure 2.** Examples of payloads in ADCs.

γ

The calicheamycins is a group of extremely potent cytotoxins. Originally the compounds were tested as antibiotics, and their minimum inhibitory cancentration (MIC) values were found to be from 0.5 µg/mL toward *Eschericia coli* to less than 0.2 ng/mL against *Bacillus subtilis* [67]. Antitumor activity was tested against P388 leukemia and B16 melanoma in mice by intraperitoneal injection. The optimum dose was found to be 5 µg/kg as compared to 1.6 mg/kg for cisplatin against P388 cells, and 1.25 µg/kg as compared to 300 mg/kg for cisplatin against B16 cells [67]. All mice died, indicating a high general toxicity. Calicheamicin γ<sup>1</sup> I is an interesting payload because of its extreme cytotoxicity [66]. This molecule represents an extraordinary example of natural bioengineering that has occurred during evolution. The sugar part including the iodinated aromatic residue confers affinity for DNA. After complexing with DNA, a nucleophilic attack, e.g., by glutathione on the central sulfur in the trisulfide, leads to the cleavage of this linkage, thereby releasing the free thiol, which undergoes an intramolecular thiol Michael addition to the α,β-unsaturated ketone. By changing the trigonal bridgehead β-carbon to a tetragonal carbon, sufficient tension is induced in the 10-membered ring to initiate a Bergman cyclization. The resulting intermediate diradical finally cleaves the DNA [66] (Scheme 17). γ β

γ

α β

**Scheme 17.** Diradical formation from calicheamicin γ1γ α β <sup>I</sup> after nucleophilic attack on the α,βunsaturated ketone [66].

In particular two functional groups are used for conjugation of drugs via a linker to the antibody, namely the thiol group of cysteine and/or the amino groups of lysine residues [61]. Traut's reagent or carbodiimides with or without hydroxysuccinimide have been used for coupling of a carboxylic acid to lysine side chains [63]. As an example, Scheme 18 depicts how calicheamicin γ<sup>1</sup> I (**33**) has been coupled to an antibody in Mylotarg (gemtuzumab ozogamicin) [63].

After internalization of the antibody–antigen complex, enzymes within the cell facilitate hydrolysis of the hydrazone moiety [63]. The disulfide will be cleaved by glutathione, enabling Bergman cyclization. Other linkers have been designed to be cleaved by intracellular proteases like cathepsin B (e.g., Adcetris). After internalization of the ADC Kadcyla, in which mertansine (**34**) is linked to an antibody, the linker including the lysine residue remains attached to the payload after decomposition of the antibody (Scheme 19). This extended linker moiety appears not to compromise the effect of the drug. Other examples of ADCs have been reported by Nicolaou and Rigol [63].

γ

**Scheme 18.** Coupling of a calicheamicin to antibodies via lysine residues (the O-Sugar moiety is defined in Scheme 17) [63].

**Scheme 19.** Internalization of an antibody–mertansine conjugate occurs while the payload remains attached to the antibody via a lysine residue [63].

### *2.3. Antibody-Directed Enzyme Prodrug Therapy*

Similarly to ADCs, the concept of antibody-directed prodrug therapy (ADEPT) is based on the ability of antibodies to selectively target antigens expressed abundantly on the surface of cancer cells [68,69]. The principle involves a preferential binding of a non-human enzyme to the surface of cancer cells via an antibody–antigen complex. The choice of a nonhuman enzyme makes it possible to choose a linker which solely is cleaved by this enzyme and not by any endogenous enzymes. On the other hand, a potential drawback of using a non-human enzyme may be a strong allergic reaction due to unforeseen immunogenicity. In contrast to the ADC approach, non-internalizing antigens can be targeted. The enzyme is linked to the antibody by using a bisfunctional linker, where one functionality can be linked to the lysine side chains present on the enzyme, while the other functionality can be linked to thiols of cysteines on the antibody (Figure 3) [68–70]. After administration to the patient the antibody binds to the surface of the cancer cell. When the excess free antibody–enzyme conjugate is completely cleared from the body the prodrug is administered.

**Figure 3.** Antibody–enzyme conjugate bound to antigens on the surface of malignant cells.

As mentioned above, a prerequisite for the use of enzymes not present in the human body is that they are non-immunogenic [70]. Enzymes belonging to the families of alkaline phosphatases (cleaving phosphate from prodrugs), peptidases, sulfatases (for cleavage of sulfate monoesters), carboxylesterases and carboxypeptidases (for cleavage of e.g., glutamic amides), have been investigated in this respect [70].

Some antibodies themselves possess catalytic properties, e.g., the antibody 38C2 catalyzes retroaldol and retro-Mannich reactions; for example, a prodrug of doxoxrubicin is cleaved by 38C2 (Scheme 20). [70,71].

**Scheme 20.** Doxorubicin prodrug (**35**) cleaved by 38C2 [70,71].

Even though some promising clinical results have been obtained, no drugs based on ADEPT are in clinical use at present [68,69].

### *2.4. Gene-Directed Enzyme Prodrug Therapy (GEPDT)*

In gene-directed enzyme prodrug therapy (GEPDT), a gene encoding for a unique enzyme is introduced into the tumor cells by using a vector. The technique was already introduced in 1986, and the gene introduced into cells is called a suicide gene [72]. Upon expression of the enzyme on the cancer cell surface, the enzyme enables cleavage of the linkage between the payload and the promoiety, after which the payload may diffuse into the cancer cell (Figure 4) [69,70]. A major drawback in GDEPT is the prerequisite of achieving selective transfer of a gene into malign cells. Retroviruses, adenoviruses and herpes viruses have been studied as potential vectors [73]. Retroviral vectors have some selectivity, since they are only incorporated into the genome of actively dividing cells [73]. Attachment of tissue-specific promoters may allow for transgenic expression only in neoplastic cells. The use of receptor-specific vectors has also been proposed [73]. Mesenchymal stem cells, exhibiting strong tropism toward tumors and metastases expressing receptors on their surface, can efficiently be transduced with vectors [72]. Virus-like particles have also been used to internalize the gene into cells [74]. No drugs based on the principle of GDEPT have been approved so far.

**Figure 4.** Cartoon illustrating the principle of gene-directed enzyme prodrug therapy (GEPDT) [69,70].

### *2.5. Lectin-Directed Enzyme-Activated Prodrug Therapy (LEAPT)*

In lectin-directed enzyme-activated prodrug therapy (LEAPT), a drug or an enzyme is targeted toward cancer cells by using sugar–protein recognition, whereas antigen–antibody recognition is used for targeting in ADEPT and GDEPT. Lectins are proteins involved in biological carbohydrate recognition comprising cellular processes such as growth, differentiation, proliferation or apoptosis [75]. In order to improve the selectivity of doxorubicin, a galacturonamide derivative (i.e., **36**) was prepared (Figure 5) [76]. Here, the expression of asialoglycoprotein receptors (ASGPRs) 1 and 2 on the surface of hepatocytes with a high affinity for D-galactose and L-rhamnose was exploited. After binding to the receptor, the appropriate carbohydrate-containing ligand is internalized. The ASGPRs are expressed on the surface of HT-29, MCF-7 and A549 cells to a much higher extent than in normal liver cells [76]. The Gal-Dox derivative proved to exhibit higher selectivity toward cancer cell lines than doxorubicin (**6**) itself. In S180 tumor-bearing mice, the Gal-Dox-treated group had a higher accumulation of the drug in the malignant tissue than the doxorubicin-treated group as well as an improved survival rate [76]. However, in this case the doxorubicin

derivative may in fact not be a true prodrug, since probably the Gal derivative may also interact with the topoisomerase target.

Another approach utilizes the overexpression of glucose transporters (GLUT) in cancer tissue. By preparing glucose or glucuronic acid derivatives of paclitaxel, two advantages are obtained: (i) the compounds become more soluble in water, and (ii) increased uptake through the GLUT into the cancer tissue (Figure 6) [77].

**Figure 6.** Glucose and glucuronic acid derivatives (i.e., **37** and **38**) of paclitaxel (**7**) displaying improved selectivity via LEAPT [77].

β α The glucose and glucuronic acid derivatives (**37** and **38**) were found to exert a low cytotoxicity on benign cells, but an activity similar to that of paclitaxel (**7**) itself on cell lines expressing GLUT. It is assumed that the prodrug is cleaved by intracellular β-glucosidases. A mechanism involving cleavage of the methyl glucoside followed by self-immolative cleavage to give paclitaxel (**7**) has been proposed [77]. However, the glucose derivative reported was an α-glycoside.

A two-phase LEAPT mechanism overcoming the requirement for intracellular cleavage of the prodrug has been suggested. In the first phase a glycosylated enzyme interacts with a carbohydrate-recognizing lectin on the surface of the cells in the targeted tissue. After similar interactions, a glycosylated prodrug, which is a substrate for the glycosylated enzyme, becomes internalized as well. Inside the cells, the internalized enzyme cleaves the glycosylated prodrug to liberate the active drug (Scheme 21).

A procedure for pergalactosylation of a naringinase produced by *Penicillium decumbens* has been developed to give a pergalactosylated enzyme, which showed high affinity for ASGPRs on the surface of hepatocytes. Binding to ASPGR triggers internalization of the bound ligand. The naringinase possesses α-rhamnosidase and β-glucosidase activities. Thus, a rhamnose derivative of doxorubicin (**6**) was prepared, and the stability of this derivative was tested [78]. At present, no drugs based on the LEAPT principle have been approved.

**Scheme 21.** Two-phase LEAPT. First a glycosylated enzyme binds to the cell surface and becomes internalized. Secondly, a glycosylated prodrug binds to the surface, and is then internalized whereafter it is cleaved by the internalized enzyme [78].

### **3. Protease-Targeting Chimeras (PROTAC)**

α β In living cells, misfolded, damaged or mutated proteins are removed from the cells by natural processes, in which the protein first becomes covalently bound to one of a number of ubiquitins, which are highly conserved 76-residue peptides [79]. This conjugation process involves ubiquitin-activating enzymes E1, which transfer ubiquitin to E2 from where ubiquitin is transferred to a von Hippel–Lindau(VHL)-cullin-RING ligase complex including E3 that conjugates ubiquitin to the target protein mainly via lysine residues [9,10]. Subsequently, the ubiquitin-modified protein is degraded by a 26S proteasome to give a number of small peptides and a number of lysine-modified ubiquitins [9,10,80]. Other proteases finally cleave the oligopeptides into free amino acids [81].

Advantages of this system are that drugs may be developed to selectively remove intracellular proteins. A chimera consisting of a residue with high affinity for the VHL complex was via a linker attached to a moiety with high affinity for the estrogen-related receptor α (ERRα). After complexing with VHL, knockdown of the ERRα level was observed. The first experiment was performed in MCF7 cells after incubation with the chimeras to knock down ERRα [9]. Moreover, a serine-threonine kinase (RIPK2) was knocked down after incubation of MCF7 cells with a chimera consisting of a moiety with high affinity for the VHL complex (Figure 7) [9].

α α α The effect of small-molecule drugs as ligands for biomolecules in treatment of cancer diseases can be limited by mutations in the gene encoding the biomolecule, thereby making the modified target insensitive to the agent. Such mutations are observed for the epidermal growth factor receptors and androgen receptors [82]. PROTAC has been used to enable knockdown of steroid receptors and for non-small lung cancer by knockdown of epidermal growth factor. In addition, the anaphylactic lymphoma kinase can similarly be removed as a possible treatment of different types of human cancers [82]. The PROTAC technique is still at an early stage, and at present no such drugs are currently in clinical use.

α

α α **Figure 7.** Cartoon illustrating the principle of protease-targeting chimera (PROTAC) [9,10,80]. A compound with affinity for the Hippel–Lincau-cullin-Ring (VHL) complex and the target protein (e.g., estrogen-related receptor α (ERRα)) is used to attach the protein to the E3 ligase in the VHL complex. After complexation the E3 conjugates ubiquitin to the protein making it a target for proteasomal degradation.

### **4. Conclusions**

In recent decades, several diverse methods have been developed for the targeting of toxins to cancer tissue to avoid their general systemic toxicity. In the present review, the initial sections concern new attempts developed for prodrugs to be cleaved predominantly in the microenvironment of cancer cells and tumors. Thus, the lower pH characterizing cancer tissue has been explored for selective cleavage of prodrugs based on amides of a substituted maleic acid [20], a hydrazone promoiety [14,21,22], and labile acetal linkage to polymers [26]. A prodrug of doxorubicin (Aldoxorubicin) designed to prevent cardiotoxicity expected to be cleaved by the acidic microenvironment of cancer cells failed in clinical trial III [22]. Moreover, proof-of-concept studies of prodrugs relying on selective cleavage due to the increased ROS production in cancer cells comprise examples of arylboronic acid derivatives [13,15,29]. An example of glutathione-promoted cleavage of a disulfide-based prodrug has also appeared [16].

Next, enzymes, overexpressed by cancer cells or neovascular tissue in tumors, capable of selective cleavage of prodrugs carrying a peptide substrate moiety, offer several examples: e.g., MMP [34], cathepsin B [35], β-glucuronidase [17,38,39], and PSA [19,41,43] and PSMA [56,60]. One prodrug, mipsagargin, actually went into clinical phase III trials but, despite the polarity of the γGlu-γGlu-γGlu-γGlu-βAsp peptide moiety, the compound appeared to be able to penetrate cell membranes of benign cells also, and thus cause general toxicity [58].

In addition, the progress within the field of ADCs (with an anticancer drug as payload) comprise >60 entities in clinical development [62–64]. In total eight ADCs have been approved by the FDA as new and improved cancer therapies, albeit not in the period 2014 to 2019 [83–88]. Calicheamicin and maytansine have been used as the payload in many of these drugs [63].

β γ γ γ γ β Another approach also involving antibodies is ADEPT [70,71]; however, even though promising clinical results have been reported, no drugs based on ADEPT are currently approved for clinical use [68,69].

In addition, an advanced approach requiring selective introduction of a gene, coding for an enzyme capable of cleaving a prodrug, into cancer cells (i.e., GDEPT) [69,70] is considered a promising approach, but so far no drugs based on this concept have been approved. Similarly, targeting to cancer cells via sugar–protein recognition processes involving lectins

present on the surface of cancer cells (i.e., LEAPT) have been explored [76–78]. Nevertheless, no drugs based on these principles have been approved as yet.

Finally, recent examples of protease-targeting chimeras (PROTACs) involve ubiquitination enzyme complexes that undergo proteolytic degradation to release the drug [9,10,80], however, this technique is at an early stage, and no examples of its clinical use have appeared.

Even though the described techniques have been utilized to improve the solubility of paclitaxel and selectivity of doxorubicin, the associated prodrugs have not as yet shown sufficiently improved properties to convince medical agencies that they can be approved as drugs. In conclusion, the new approaches reviewed here may indeed lead to future new anticancer drugs that are urgently needed for treatment of cancer diseases for which no cure exists. Nevertheless, most of these recently developed targeting principles remain to result in approved drugs, which emphasizes the need for further research to unleash the full potential of these concepts currently considered for experimental therapies.

**Author Contributions:** Both authors have contributed to the manuscript. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** No supporting information for this work.

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

### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Molecules* Editorial Office E-mail: molecules@mdpi.com www.mdpi.com/journal/molecules

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-4919-4