*Review* **Blood-Derived Biomarkers of Diagnosis, Prognosis and Therapy Response in Prostate Cancer Patients**

**Katalin Balázs, Lilla Antal, Géza Sáfrány and Katalin Lumniczky \***

Unit of Radiation Medicine, Department of Radiobiology and Radiohygiene, National Public Health Centre, 1221 Budapest, Hungary; balazs.katalin@osski.hu (K.B.); antal.lilla@osski.hu (L.A.); safrany.geza@osski.hu (G.S.) **\*** Correspondence: lumniczky.katalin@osski.hu; Tel.: +36-1-482-2011 or +36-30-554-9308

**Abstract:** Prostate cancer is among the most frequent cancers in men worldwide. Despite the fact that multiple therapeutic alternatives are available for its treatment, it is often discovered in an advanced stage as a metastatic disease. Prostate cancer screening is based on physical examination of prostate size and prostate-specific antigen (PSA) level in the blood as well as biopsy in suspect cases. However, these markers often fail to correctly identify the presence of cancer, or their positivity might lead to overdiagnosis and consequent overtreatment of an otherwise silent non-progressing disease. Moreover, these markers have very limited if any predictive value regarding therapy response or individual risk for therapy-related toxicities. Therefore, novel, optimally liquid biopsy-based (blood-derived) markers or marker panels are needed, which have better prognostic and predictive value than the ones currently used in the everyday routine. In this review the role of circulating tumour cells, extracellular vesicles and their microRNA content, as well as cellular and soluble immunological and inflammation- related blood markers for prostate cancer diagnosis, prognosis and prediction of therapy response is discussed. A special emphasis is placed on markers predicting response to radiotherapy and radiotherapy-related late side effects.

**Keywords:** prostate cancer; radiotherapy; liquid biopsy; circulating tumour cells; extracellular vesicles; microRNAs; immune system; inflammation

#### **1. Introduction**

Based on the 2020 cancer statistics of the International Agency for Research on Cancer (IARC) out of 19.3 million newly diagnosed cancers prostate cancer is ranked as the third most common among both sexes (constituting 7.1% of total cases). Regarding mortality rate it is on the eights place with approx. 375,000 deaths per year [1]. There are marked differences in the incidence of prostate cancer among various countries and races. It was reported that Japanese men living in Japan had very low prostate cancer incidence, while the incidence among USA resident Japanese increased and was at an intermediate level between Japanese living in Japan and European American men. Conversely, African American men have the highest incidence and mortality rates from prostate cancer within the United States [2,3]. These observations stress the importance of both genetic susceptibility and lifestyle in disease development and progression.

Among the main reasons for the increased prostate cancer mortality are the lack of reliable and effective prognostic biomarkers and methods which enable to recognise tumours in an early stage, to monitor individual therapy response more effectively, to sensitively detect minimal residual disease and development of distant metastasis as well as predict tumour relapse. These markers would allow patient stratification for optimal response rate to a certain therapy and enable the identification of those patients who are at increased risk for developing therapy-related side effects; thus, they are prerequisites for the development of efficient individualized anticancer treatment protocols.

At present diagnosis of prostate cancer is complex and is based on symptoms such as difficulties in urination, presence of blood in the urine or sperm, physical examination

**Citation:** Balázs, K.; Antal, L.; Sáfrány, G.; Lumniczky, K. Blood-Derived Biomarkers of Diagnosis, Prognosis and Therapy Response in Prostate Cancer Patients. *J. Pers. Med.* **2021**, *11*, 296. https:// doi.org/10.3390/jpm11040296

Academic Editors: Christophe Badie and Eric Andreas Rutten

Received: 17 March 2021 Accepted: 3 April 2021 Published: 13 April 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/).

(including rectal digital examination), ultrasound examination, blood test to measure the prostate specific antigen (PSA) and tissue sample testing (biopsy). Due to the rather unspecific symptoms, early diagnosis of prostate cancer is not without problems and often the disease is only diagnosed in an advanced stage, where actually symptoms related to bone metastasis (bone pain and limb weakness caused by spinal marrow compression) are already present. While early detection of prostate cancer is fully curable, the efficiency of anti-cancer treatment in an advanced stage of the disease is very low.

Therefore, regular screening protocols in asymptomatic men with the purpose of identifying early-stage prostate cancer would be very important. Though, regular screening for prostate cancer has certain caveats. One such caveat is that more than 75% of PSApositive tests (with blood PSA levels above 4 ng/mL, traditionally used as a cut-off value) are followed by a negative biopsy [4]. Biopsy can lead to infections [5], significant drop in quality of life [6] and can cause urinary, bowel and sexual dysfunctions persisting for several months [7,8]. Another caveat is the identification of indolent prostate cancer patients. Based on autopsy material, prostate cancers are identified at a much younger age (31–40 years) than clinically diagnosed in symptomatic patients. It appears that some prostate cancers may pass through a period of latency of up to 15 to 20 years, during which the disease is histologically present but it is completely asymptomatic. PSA-based screening of the population might result in over-diagnosed and therefore over-treated indolent prostate cancers resulting in serious side effects (such as incontinence and impotence), which would have caused no clinical consequences during a man's lifetime if left untreated [9–11].

The issue of regular prostate cancer screening is a dilemma all over the world. Currently, in several European countries the main indications of annual prostate monitoring are age (men over 45 years) and/or family history. Nevertheless, the influence of family history for the risk of developing prostate cancer is recently being under revision due to studies with contradictory conclusions (Selkirk, Wang et al., 2015, Abdel-Rahman 2019). Increasing number of studies are investigating benefits, harms and cost-effectiveness of prostate cancer screening based on experimental and clinical data, clinical trials and model calculations. Two big studies are especially worth mentioning. One such study called the European Randomised Study of Screening for Prostate Cancer (ERSPC) randomized more than 180,000 men from Europe to analyse the longitudinal relationship between PSA values and biopsy (biopsy was carried out if blood PSA level was 3.0 ng/mL or higher) at regular intervals (every 4 years). After 13 years of follow-up the risk of prostate cancer mortality decreased 21% in the surveyed population compared to control group [12]. The other one called the Prostate, Lung, Colorectal and Ovarian (PLCO) carried out in the USA randomised more than 76,000 men using basically similar screening principles to the ERSPC study. Their final, updated conclusion, as a result of an extended follow up was that no significant difference was found in prostate cancer mortality in the screened group compared to control [13]. Thus, regarding the primary endpoint of the two studies, namely, to evaluate the predictive value of PSA screening in reducing prostate cancer mortality the two trials seem to reach contradictory conclusions. However, a recent re-analysis of both trials showed that the discordant results were due to differences in implementation and setting. After correcting them both studies reached the conclusion that screening could significantly reduce prostate cancer death [14]. These clinical trials highlight the overall positive balance of screening even using a marker (PSA) by far not optimal in detecting those patients which indeed should be treated for prostate cancer. More efficient, costeffective and specific screening methodology is needed to reliably discriminate prostate cancer from benign alterations or other non-cancerous prostate diseases. Screening is especially important in races with an increased incidence of the disease (African American men) in order to reduce racial differences in prostate cancer survival [15].

Over the past decade, liquid biopsy investigations have received more and more attention. This minimal invasive method enables us to study a wide array of blood-based cellular and secreted soluble or vesicular markers, which offer a complex, comprehensive and real-time information on tumour stage, progression, tumour micro- and macro-environment,

including the integrity of the anti-tumour immune response. These complex indicators might serve as prognostic and/or predictive markers able to predict patients' outcomes, their response to particular therapies, forecast the formation of late therapy-related side effects (such as secondary tumours after radiotherapy) and ultimately to improve medical decision-making [16]. Blood-based markers of prognosis or therapy responsiveness are especially important in prostate cancer patients because of the high heterogeneity and molecular diversity of prostate cancer and because the prostate gland contains different subclones, which respond differently to treatments [17,18], so prostate gland biopsy can be misleading. comes, their response to particular therapies, forecast the formation of late therapy-related side effects (such as secondary tumours after radiotherapy) and ultimately to improve medical decision-making [16]. Blood-based markers of prognosis or therapy responsiveness are especially important in prostate cancer patients because of the high heterogeneity and molecular diversity of prostate cancer and because the prostate gland contains different subclones, which respond differently to treatments [17,18], so prostate gland biopsy can be misleading. In this review, we summarize the current knowledge on blood-based liquid biopsy

Over the past decade, liquid biopsy investigations have received more and more attention. This minimal invasive method enables us to study a wide array of blood-based cellular and secreted soluble or vesicular markers, which offer a complex, comprehensive and real-time information on tumour stage, progression, tumour micro- and macro-environment, including the integrity of the anti-tumour immune response. These complex indicators might serve as prognostic and/or predictive markers able to predict patients' out-

*J. Pers. Med.* **2021**, *11*, x FOR PEER REVIEW 3 of 28

In this review, we summarize the current knowledge on blood-based liquid biopsy analyses in prostate cancer focusing on disease- and therapy-related changes in PSA and related molecules, immune cells and immune- and inflammation-related secreted factors, circulating tumour cells (CTCs) and tumour-derived cell-free circulating nucleic acids as well as extracellular vesicles (EVs) and their micro-RNA (miRNA) content (Figure 1 and Table 1). We also discuss the relevance of these markers in radiotherapy-treated patients, in predicting their therapy-response or their risk for developing radiotherapy-induced late toxicities. Apart of blood-derived markers a large panel of urine and tissue-based potential biomarkers with prognostic and/or predictive value are identified. A few are already in the possession of clinical approval and some are still in an experimental stage. Though, in the present review we do not wish to focus on these. Detailed reviews such as [19–24] are available on these topics. For interested readers we advise consulting these publications. analyses in prostate cancer focusing on disease- and therapy-related changes in PSA and related molecules, immune cells and immune- and inflammation-related secreted factors, circulating tumour cells (CTCs) and tumour-derived cell-free circulating nucleic acids as well as extracellular vesicles (EVs) and their micro-RNA (miRNA) content (Figure 1 and Table 1). We also discuss the relevance of these markers in radiotherapy-treated patients, in predicting their therapy-response or their risk for developing radiotherapy-induced late toxicities. Apart of blood-derived markers a large panel of urine and tissue-based potential biomarkers with prognostic and/or predictive value are identified. A few are already in the possession of clinical approval and some are still in an experimental stage. Though, in the present review we do not wish to focus on these. Detailed reviews such as [19–24] are available on these topics. For interested readers we advise consulting these publications.

**Figure 1. Figure 1.**  Overview of candidate blood-based liquid biopsy markers in prostate cancer patients. Overview of candidate blood-based liquid biopsy markers in prostate cancer patients.

This schematic picture summarizes the current knowledge on liquid biopsy analyses in prostate cancer focusing on circulating tumour cells (CTCs), immune cells and secreted factors such as tumour-derived cell-free circulating nucleic acids, cytokines and chemokines, as well as EVs released by prostate cancer cells, CTCs or various immune cells and their microRNA (miRNA) content.

#### **2. Blood-Based Liquid Biopsy Marker Candidates**

#### *2.1. PSA and Related Molecules*

At present, PSA is the most widespread and most accepted biomarker for prostate cancer monitoring; however, it lacks many of the qualities of an ideal tumour marker. PSA is a serine protease secreted physiologically by the prostate gland epithelium [25]. The total PSA (tPSA) is present in the serum in two forms, free (fPSA) and conjugated to serum proteins like alpha1 antichymotripsin, alpha2 macroglobulin and alpha1 protease inhibitor. Free PSA is comprised of pro PSA, benign PSA (BPSA) and intact PSA. Pro PSA is a zymogen precursor of PSA that comes in four different isoforms, as determined by the length of its pro leader peptide sequences [26]. Pro PSA is associated with cancer, BPSA with benign diseases whilst the association of intact PSA is currently unknown [27]. The free component represents about 5%–35% of tPSA [28]. The normal range of tPSA (commonly called as PSA) is 0–4 ng/mL in peripheral blood and levels above 10 ng/mL are considered pathological [29].

Monitoring regular variations in PSA level might serve as rough indicators of cancer progression (and regression after therapy) as well as disease recurrence. Since PSA is a tumour-associated antigen (TAA) and not a tumour-specific one, its specificity in prostate cancer diagnosis is not 100%, since inflammation, benign prostate hyperplasia (BPH) or other non-malignant disorders could also cause increased blood PSA levels, while normal PSA levels do not necessarily exclude the presence of a tumour. Furthermore, the age of cancer patients also influences the risk of cancer-specific death. With 7–10 ng/mL of PSA the risk of death is only 7% for men aged 50–59 but increases to 51% for men aged 80–89, 10 years after diagnosis [30].

As mentioned before, 75% of men with PSA levels above 4 ng/mL are not diagnosed with prostate cancer on biopsy. Only 26% of patients with PSA level within the "grey zone" of 4.1–9.9 ng/mL have cancer [31]. The PSA grey zone needs more accurate noninvasive diagnostic biomarkers to avoid the false-positive results because of benign changes. It was shown that PSA produced by prostate cancer cells escape degradation and occur in complexed form in the serum. When serum PSA is between 4 and 10 ng/mL and free to total PSA ratio (F/T ratio) is less than 25%, one should strongly suspect prostate malignancy, thus avoiding up to 20% of unnecessary prostate biopsies [32].

The so-called 'prostate healthy index' (PHI) reflects the ratio between total PSA, free PSA and pro PSA levels and prostate health index density (PHID) combines PHI parameters with prostate volume. These markers are clear improvement over PSA since they allow a better distinction between benign and malignant prostate gland hypertrophy and improve the prediction of high-grade and clinically aggressive prostatic tumours, especially in cases where PSA levels are in the grey zone [33–38].

The four kallikrein (4K) test measures tPSA, fPSA, intact PSA and human kallikreinrelated peptidase 2 (hK2) in serum and is used to get a probability score for prostate cancer [39]. Similar to PHI, it improves the identification of overall and high-grade cancer and helps with reducing unnecessary biopsies [40,41]. Combined with PHI increases their diagnostic value [37]. Both the PHI and 4K test have been approved by the US Food and Drug Administration (FDA) to be used for prostate cancer screening. Recently two novel cancer-related glycoproteins, thrombospondin 1 (THBS1) and cathepsin D (CTSD) were proposed as blood-based biomarkers that outperformed PSA in distinguishing benign disease from prostate cancer in men with enlarged prostate gland [42]. A recently commercialized product—the Proclarix—incorporating THBS1, CTSD, tPSA and % of fPSA, combined with patient age, yielded a significantly better diagnostic accuracy compared to either PSA or % of fPSA alone in discriminating clinically significant from no or insignificant prostate cancer [43,44]. Though we have not found studies comparing the diagnostic efficiency of Proclarix with either PHI or 4K test.

#### *2.2. Circulating Tumour Cells (CTCs) and Cell-Free Circulating Tumour DNA (cfDNA)*

Prostate cancer metastasis is initiated by CTCs originating from the primary tumour transported through the blood or lymphatic system [45]. Some CTCs die in the circulation, others proliferate and form metastasis in distant organs [46]. Experimental models indicate that millions of tumour cells continuously circulate through the body, although only few of them can survive by evading the immune response and systemic therapies, reach a distant organ, proliferate and ultimately form metastases [47].

The number of CTCs in peripheral blood is very low, on average one CTC per one million peripheral blood mononuclear cells (PBMCs) [48], so their isolation is challenging. Several approaches have been reported for CTC detection, isolation and characterization in the peripheral blood of cancer patients such as two-stage microfluidic chip technology [49], acoustic separation of CTCs [50] and in situ hybridization (ISH) technology combined with immunomagnetic selection [51]. There are label-free techniques as well, which include size-based and density-based approaches [52] and methods based on Ficoll-Paque centrifugation [53], electrical property-based separation [54] and leukocyte depletion (anti-CD45 immunomagnetic negative selection) [55]. At present, the CellSearch™ method is an FDA-approved technology based on CTC characterisation used to predict the outcome of prostate cancer patients. It enriches CTCs from the peripheral blood using a magnetic ferrofluid containing antibodies against epithelial cell adhesion molecule (EpCAM), which is a common CTC marker. Cells are then stained for expression of cytokeratine (CK) 8, 18, and 19, all of which are intracellular structural proteins found in epithelial cells [56].

CTCs are critical for monitoring anti-cancer therapeutic efficacy such as drug screening [57], since resistance towards various chemotherapeutic agents remains a major clinical challenge [58]. Regular monitoring of CTCs can give a more complex and more realistic view of tumour heterogeneity than conventional biopsy [59]. Furthermore, CTCs are independent prognostic factors of progression-free survival (PFS) and overall survival (OS) in metastatic breast [60], colon [61] and prostate cancers [62] and their presence was implicated in worse cancer prognosis and outcome [63]. Patients with CTC numbers higher than 5 per 7.5 mL whole blood (as compared with the group with CTC numbers lower than 5 per 7.5 mL blood) had shorter median PFS and OS [64]. Quantification of CTCs has been proposed also as a potential surrogate endpoint to promote the selection of treatment algorithms [65] especially in advanced-stage prostate cancer disease, although the small number of cells detectable in the blood of prostate cancer patients and the lack of specific molecular determinants on CTCs indicative of therapy response have limited its clinical utility [66].

Not just CTC numbers can serve as prognostic markers but also the repertoire of their cell surface molecules which could also indicate the efficacy of various therapies such as radiotherapy or immune therapy. The EpCAM glycoprotein was initially described as one of the most commonly used protein CTC markers, however its level was shown to be downregulated during the dissemination of cancer cells from primary tumour [67]. Therefore, using CTCs as biomarkers of therapy response based purely on their immune phenotypical changes might be misleading because of their dynamic evolution during cancer progression [68].

Certain studies suggest that CTCs adopt different strategies to protect themselves from therapy-induced cell death, developing an epithelial to mesenchymal transition (EMT), grouping into clusters or switching between cancer stem cell state and differentiated cancer cell state. Not only individual CTCs, but also CTC clusters, their EMT and the presence of cancer-associated macrophage-like cells (CAMLs) in the blood are indicative for an increased risk of metastatic disease. Compared to single CTCs, CTC clusters may be more aggressive in forming distal metastasis [69]. CTC cluster size and number have been associated with lower overall survival in patients with breast, pancreatic, or prostate cancer [70–72]. These additional features of CTCs apart of their immune phenotypical changes have improved their prognostic value.

The programmed cell death protein (PD-1) and its ligand 1 (PD-L1) are major targets of the immune checkpoint inhibitor therapies in metastatic cancer diseases. PD-L1 expressing circulating epithelial tumour cells (CETCs) are described in 100% of prostate, 94.5% of breast, 95.4% of colorectal and 82% of lung cancer patients. Monitoring the frequency of PD-L1 positive CTCs could reflect individual patient's response to anti-PD-1/PD-L1 therapy [59,73]. Expression of nuclear PD-L1 (nPD-L1) in the CTCs of prostate cancer patients was shown to significantly correlate with short survival rate [74]. In conclusion, the use of CTC-based models for risk assessment can improve standard cancer staging.

The effect of radiotherapy on CTCs is controversial. Martin et al. described that fractionated radiotherapy disrupted the tumour mass in non-small cell lung cancer (NSCLC), thus promoting the passage of tumour cells into the circulation [75]. In contrast, Budna-Tukan et al. analysed the number of CTCs in patients with non-metastatic high-risk prostate cancer with three different innovative CTC enumeration technologies before and after radiotherapy. They did not find any differences and significant therapy-related changes in CTC counts. These latter data do not support the hypothesis that radiotherapy leads to CTC release into the circulation in prostate cancer most probably because radiotherapy efficiently reduces tumour size in patients with prostate cancer, therefore the number of cancer cells in the circulation should also partially or totally decrease. It would be interesting to analyse this CTC number in patients with worse prognosis and metastatic disease as well. Furthermore, the reason for the difference between radiotherapy effects on CTCs in NSCLC and prostate cancer could be that these tumours respond differently to radiotherapy. Less effective radiotherapy leads to a suboptimal tumour response, and possibly an increase in CTC numbers [76].

Some patients undergoing prostate brachytherapy develop distant metastases despite the absence of local recurrence. Although micrometastases were not detected by radiographic images in these patients, cytokeratine positive or PSA positive cells were present in the bone marrow aspirates, which were considered disseminated tumour cells (DTCs) [77]. It has been proposed that in the early phases of radiotherapy, when cancer cells suffer only from sub-lethal damage, the quick increase of CTCs could, in principle, contribute to the development of distant metastases [76]. Another explanation is that surgical manipulation with needles being inserted into the prostate tissue during brachytherapy may pose a potential risk for haematogenous spillage of prostate cancer cells and play a role in distant metastases development [78].

Quantification of cell-free circulating tumour nucleotides in blood samples is another promising new molecular strategy for non-invasive tumour monitoring in prostate cancer [79]. CfDNA may be more suitable than CTCs to estimate therapy efficiency in patients with early-stage disease [80]. CfDNA originates from apoptotic or necrotic cells (including CTCs) or is actively secreted by cancer cells [81] and has been detected in human blood, urine and semen [82].

CTCs carry the complete mutation spectrum of the primary tumours and metastases, [16,83] and therefore genetic and transcriptional analysis of individual CTCs might enable personalized medical decisions for cancer therapy and provide insights into the biological processes involved in metastasis. Somatic mutations are detected in advanced or metastatic tumours, so they are unsuitable for monitoring primary, nonmetastatic disease [84,85]. Conversely chromosomal rearrangements represent an early stage of cancer pathogenesis [86]. The most common chromosomal rearrangement in prostate cancer present in approx. 50% of patients results from the fusion of the androgen-regulated gene, transmembrane protease serine 2 gene (*TMPRSS2*, chr21q22.2), with E-twenty-six (ETS) related gene (*ERG*, chr21q22.3). This rearrangement is detectable in CTCs or as cfDNA and might be considered as a highly tumour-specific, non-invasive molecular biomarker for therapy assessment, risk stratification and relapse detection [79,87,88]. Furthermore, prostate cancer carrying this rearrangement can regulate the recruitment and infiltration of regulatory T cells (Tregs) in the tumour [89].

Gene expression analysis of CTCs of prostate cancer patients have revealed altered expression levels of eight metastasis-related metabolic genes, such as phosphoglycerate kinase 1 (*PGK1*) and glucose-6-phosphate dehydrogenase (*G6PD*) responsible for optimal glucose metabolism in CTCs. Their increased expression level in CTCs was associated with advanced tumour stage and metastasis proneness [90].

The clonal evolution of cancer cells can be traced at the level of CTC DNA or cfDNA by next generation sequencing or tumour mutation allele frequency analyses. The dynamics and relative abundance of the different clones are suitable to evaluate disease and metastasis heterogeneity, to monitor the emergence of resistance mechanisms as well as therapy resistance [91]. Quantification of cfDNA can reveal treatment ineffectiveness at an early stage of the treatment protocol and also avoid toxicity of ineffective overtreatment. Thus, patients can receive alternative therapies [92]. Since both CTCs and cfDNA can be retrieved basically in a non-invasive manner by blood collection, their regular follow up allows a more precise monitoring of disease progression and a better patient care as well.

#### *2.3. Cellular and Soluble Immunological Markers*

Systemic inflammatory conditions as well as the components of adaptive and innate immunity are involved in the initiation and progression of prostate cancer [93–95]. The role of chronic inflammation in this process involves multiple mechanisms such as: (a) mutagenesis caused by oxidative stress, (b) remodelling of the extracellular matrix, (c) recruitment of immune cells including tumour-associated neutrophils (TANs), tumour-infiltrating macrophages (TIMs), myeloid-derived suppressor cells (MDSCs), mast cells, as well as fibroblasts, and endothelial cells, (d) elevated secretion of cytokines and growth factors contributing to a proliferative and angiogenic environment [96–98].

Innate immune cells (macrophages, neutrophils) are triggered by foreign microbial and viral structures, known as pathogen-associated molecular patterns (PAMPs), or normal cellular constituents released upon injury and cell death, known as damage-associated molecular patterns (DAMPs), which are recognized by pattern-recognition receptors (PRRs), like the Toll-like receptor (TLR) family [99]. Innate immune cells are the main players in the early phase of the inflammation and affect tumour progression via intercellular signalling including cytokines and chemokines [100]. Tan et al. found in animal studies that the Cystein-X-Cystein (C-X-C) motif chemokine ligand 9 (CXCL9) regulated the host's response to inflammation by recruiting leukocytes to the inflammatory environment and had important role in promoting prostate tumorigenesis [101]. Activation of the innate immune system implies upregulation of major histocompatibility (MHC) class I and II molecules on the surface of nucleated cells and antigen presenting cells (APCs) like macrophages, dendritic cells (DCs) and B lymphocytes and presentation of tumour associated antigens on their MHC molecules to naive T lymphocytes [102]. These processes induce the production of different inflammatory chemokines and cytokines, which in turn lead to the activation of both the cellular and humoral arm of the adaptive immune response [103].

An important step in prostate cancer development is tumour immune escape manifested among others in defects in antigen presentation (human leukocyte antigen (HLA) class I receptor deficiency on cytotoxic T cells), imbalance in T helper type 1 (Th1) and Th2 cytokine production leading to elevated levels of immunosuppressive cytokines such as interleukine-4 (IL-4), IL-6 and IL-10. Furthermore, induction of T cell death, T cell receptor dysfunction, prostate tumour infiltration with tolerogenic DCs and Tregs are also important indicators of an immune suppressing microenvironment and low tumour immunogenicity [104].

Granulocyte colony-stimulating factor (G-CSF), IL-1β, or tumour necrosis factor (TNF) secreted by tumour cells extend the lifespan of neutrophils and attract them to the tumour microenvironment, where they become immunosuppressive tumour-associated neutrophils, which stimulate proliferation of tumour cells and angiogenesis [105]. Macrophages are important in promoting growth and bone metastasis of prostate cancer [106]. Monocyte chemotactic protein (MCP)-1/C-C motif ligand (CCL)2 secreted by cancer cells recruits

tumour-infiltrating macrophages and induces tumour progression [107]. MDSCs are the immature form of myeloid cells and they suppress anti-tumour immune responses in the tumour microenvironment. Tregs are among the most important immune cell populations suppressing antitumour immune responses [108]. Increased Treg infiltration in the tumour microenvironment through C-C chemokine receptor 4 (CCR4) [109] with the high expression level of the cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) and PD-1 markers was linked with a poor prognosis in prostate cancer [110,111]. Several studies published that the number of Tregs showed significant correlation with the number of macrophages with tumour promoting M2 phenotype in the prostate cancer microenvironment and together they were associated with a worse clinical outcome [112,113]. Furthermore, isoforms of cluster of differentiation 44 (CD44) transmembrane glycoprotein receptor serve also as a poor prognostic factor in prostate cancer. This receptor has different isoforms. The standard isoform (CD44std) is expressed in normal epithelial cells [114], while the variant isoforms (CD44v) are highly expressed in several epithelial-type carcinomas [115]. Increased expression of CD44v (also known as CD44 variant 6) was associated with progressive disease and poor prognosis in prostate cancer [116].

Several immune and inflammatory genes harbouring single nucleotide polymorphisms (SNPs) associated with prostate cancer risk were identified, including pattern recognition receptors (macrophage scavenger receptor 1 or *MSR1*, *TLR1*, *TLR4*, *TLR5*, *TLR6*, and *TLR10*) [117–120]; antiviral genes (ribonuclease L or *RNASEL*) [121,122]; cytokines (macrophage inhibitory cytokine 1 or *MIC1*, *IL-8*, *TNF-α*, and IL-1 receptor antagonist or *IL1RN*) [123,124]; and the proinflammatory gene cyclooxygenase 2 (*COX-2*) [125].

Several of the identified gene expression signatures in prostate cancer are also immunerelated, highlighting the importance of immune system in disease pathogenesis. Liong et al. developed a blood-based biomarker panel consisting of 7 mRNAs and demonstrated in 739 prostate cancer patients that the panel could identify men with aggressive prostate cancer. It is important to highlight that the majority of the differentially expressed genes were involved in immune processes [126]. Wallace et al. reported differential gene expression signatures in prostate cancer samples from African American and European American men and the majority of the differentially expressed genes was also immune related (immune response, defence response, antigen presentation, B/T cell function, cytokine signalling, chemokines, inflammatory response) [127]. The C-X-C motif chemokine receptor 4 (CXCR4) chemokine, previously linked to tumour metastasis [128] was differentially expressed between tumour and surrounding non-tumour tissue in African American men and the CXCR4 pathway was the highest-ranked pathway showing differential expression pattern among tumour and non-tumour tissue in African American men [127,129]. Clinical application of these gene expression signatures in therapy individualisation holds great promise. Though, we should mention that the so-far tested approaches for molecular characterisation of prostate cancer are based on biopsy tissues. In order to overcome serial biopsies, analysis of CTCs holds great promise in the non-invasive molecular profiling of prostate cancer and in determining both tumour- and patient-derived heterogeneity in disease progression and therapy response. A report on CTC-based liquid biopsy signatures with prognostic relevance and with implications in therapy decision has recently been published [130].

Radiotherapy can influence immune processes by increasing the expression of TAAs and DAMPs, as well as inducing cell death and secondary release of the proinflammatory cytokines and chemokines [131,132]. Radiotherapy-induced DNA, protein and lipid damage was shown to increase free radical levels in the circulation released by directly irradiated cells [133]. Gupta et al. found that IL-6, IL-8, TNF-α and transforming growth factor-beta (TGF-β) were the major mediators of ionizing radiation response in prostate cancer after radiation therapy influencing signalling pathways targeting transcription factors such as nuclear factor kappa B (NF-kB), activator protein-1 (AP-1) and signal transducers and activators of transcription (STATs). These transcription factors further enhanced expression of IL-1β and TNF-α [134]. Furthermore, radiotherapy also upregulated MHC

class I on cancer cells, leading to the recognition of TAAs by cytotoxic T cells, enabling them to raise an antitumour response. Thus, prostate radiotherapy could potentially initiate a systemic, or 'abscopal' immune response, resulting in antitumorigenic responses in distant metastases [135]. According to Reits et al., the effect of γ-irradiation on MHC class I molecules could explain immune-mediated abscopal effects [136,137].

It was shown that, in patients responsive to androgen deprivation therapy (ADT), the baseline levels of certain immune markers such as IL-6, IL-10, granulocyte macrophage colony stimulating factor (GM-CSF) was significantly lower, and the level of certain proinflammatory cytokines (IL-5, interferon-γ (IFN-γ), TNF-α) during the course of the therapy was significantly higher than in patients resistant to ADT [138]. Both ADT and radiotherapy can damage the endothelium network in prostate cancer and vascular damage is part of radiotherapy-caused late toxicities [139]. It was reported that ADT downregulated vascular endothelial growth factor (VEGF) in normal tissue as well as in malignant prostatic tissue. In the absence of VEGF immature blood vessels underwent selective apoptosis and endothelial dysfunction [140]. Microvascular endothelial cell apoptosis after high dose irradiation constituted a primary lesion, developing into persistent endothelial dysfunction with microvessel collapse, endothelial cell activation and ultimately premature aging and senescence [139]. These effects impact normal tissue homeostasis leading to hypoxia and consequent ischemia, as well as inflammation and fibrosis. The early phases of fibrogenesis after irradiation were characterized by the upregulation of pro-inflammatory cytokines such as TNF-α, IL-1 and IL-6 and many growth factors in the irradiated tissue [141]. It was shown that a complex balance between TGF-β [142] and its downstream effector connective tissue growth factor (CTGF) [143], the antifibrotic proteins such as TNF-α and IFN-γ was important in this process [144].

Prostate cancer is among the more radioresistant malignant tumours [145]; the disease recurs in 30%–40% of prostate cancer patients receiving radiation therapy [146]. A retrospective study investigated the overexpression of 24 genes (DNA damage regulated autophagy modulator 1 or *DRAM1*, keratin 14 or *KRT14*, protein tyrosine phosphatase, non-receptor type 22 or *PTPN22*, zinc finger matrin-type 3 or *ZMAT3*, Rho GTPase activating protein 15 or *ARHGAP1*, *IL-1B*, anillin actin binding protein or *ANLN*, ribosomal protein S27a or *RPS27A*, melanoma associated antigen mutated 1 or *MUM1*, topoisomerase (DNA) II alpha or *TOP2A*, cyclin dependent kinase inhibitor 3 or *CDKN3*, G protein subunit gamma 11 or *GNG11*, haematopoietic cell-specific Lyn substrate 1 or *HCLS1*, denticleless E3 ubiquitin protein ligase homologue or *DTL*, IL-7 receptor or *IL-7R*, ubiquitin like modifier activating enzyme 7 or *UBA7*, NIMA related kinase 1 or *NEK1*, CDKN2A interacting protein or *CDKN2AIP*, apurinic/apyrimidinic endonuclease 2 or *APEX2*, kinesin family member 23 or *KIF23*, sulfatase 2 or *SULF2*, polo like kinase 2 or *PLK2*, essential meiotic structure-specific endonuclease 1 or *EME1*, and bridging integrator 2 or *BIN2*) related to radiotherapy and DNA damage-response and found that their expression signatures predicted response to radiotherapy and radioresistance in prostate cancer patients and helped specifically with selecting patients profiting from radiotherapy [147]. Further predictive markers of radioresistance are oxidative stress markers such as lipid peroxidation 4-hydroxylnonenal (4HNE) or 3-nitrotyrosine (3NT) [148,149]. Preclinical studies showed that hypoxia lead to a radioresistant and metastatic phenotype of prostate tumours [150]. Extracellular vesicles could mediate hypoxia-induced prostate cancer progression, enhanced the invasiveness and stemness of prostate cancer cells and increased the level of signalling molecules such as TGF-β2, TNF-lα, IL-6, tumour susceptibility gene 101 (TSG101), protein kinase B (PKB or Akt), integrin-linked kinase 1 (ILK1), matrix metalloproteinase (MMP), and β-catenin [151]. High level of the IL-8 (CXCL8) chemokine was also associated with increased radioresistance [152]. Sequence variants of several genes such as ataxia-teleangiectasia mutated (*ATM)*, breast cancer type 1 susceptibility protein (*BRCA1)*, H2A histone family member X (*H2AFX)* and mediator of DNA damage checkpoint protein 1 (*MDC1)* were linked to increased radiosensitivity and could distinguish prostate cancer patients with high radiation toxicity from those with low toxicity [153,154]. Langsenlehner et al. investigated

603 patients treated with three-dimensional conformal radiotherapy and found that single nucleotide polymorphisms in the X-ray repair cross-complementing protein 1 (*XRCC1*) gene was associated with radiation-induced late toxicity in prostate cancer patients [155].

#### *2.4. Extracellular Vesicles*

Extracellular vesicles (EVs) received considerable attention in recent years because of their role in intercellular communication. EVs are phospholipid bilayer membrane-coated vesicles released by most cell types in physiological and pathological conditions [156]. Since EVs are highly heterogeneous in size, biogenesis, function, content, membrane markers, and so forth [157], we use "extracellular vesicle" as a generic term to describe any type of membrane-coated vesicles (e.g., exosomes, microvesicles, microparticles, apoptotic bodies, etc.) released into the extracellular matrix. A common feature of all types of EVs is their complex cargo consisting of various bioactive molecules, like proteins, lipids, DNA-fragments, different species of RNAs. These molecules are protected by the lipid membrane of the EVs and thus they are transported in an intact and biologically functional form between cells [156,158,159].

The release of EVs from cells and their journey throughout the body is not random, several regulated mechanisms underlie this process. It was shown that cancer cells produced more EVs compared to normal cells [160]. This might be partly due to the acidic environment characteristic for many cancers. Several studies demonstrated that the extracellular pH was an important modifier of EV traffic, since low pH altered EV membrane fluidity [161] and increased EV release and uptake [162]. Under chronic hypoxia prostate cancer cells secreted more EVs as a survival mechanism to remove metabolic waste [163]. Tumour-derived EVs played a key role in tumour cell growth and in the crosstalk between cancer cells and the tumour microenvironment (TME), contributing to the development of a cancer-supportive microenvironment, angiogenesis [164,165] and metastasis [166,167].

Since tumour derived EVs have specific cargos, which differentiate them from EVs released under physiological conditions, and given the fact that they are released into various human body fluids (e.g., blood, saliva, urine, amniotic fluids, sperm, bile, etc.) [158], EVs represent a source of biomarkers for the early detection of cancer, therapeutic planning and monitoring. EVs can transmit their information to recipient cells in several ways, such as (a) transfer of bioactive molecules which regulate signalling pathways in recipient cells; (b) receptor shuttling to alter cellular activities; (c) delivery of fully functional proteins to accomplish specific functions in target cells; (d) and providing new genetic information with various type of nucleic acids to gain new traits [168]. Accordingly, different types of biomolecules within the EV cargo can serve as prognostic and predictive biomarkers in cancer. Increased plasma EV levels in prostate cancer patients were reported by several studies [169–172] but reports are contradictory whether prostate cancer cell-derived EVs could be distinguished from total plasma EVs based on the presence of prostate-specific membrane antigen (PSMA) on the EV membrane. Plasmatic EVs expressing both CD81 and PSA were significantly higher in prostate cancer patients compared to either healthy controls or patients with BPH, reaching 100% specificity and sensitivity in distinguishing prostate cancer patients from healthy individuals [172]. Biggs et al. reported that prostatespecific plasma EV (identified based on PSMA expression on the EV membrane) numbers were suitable to identify prostate cancer patients with high risk, and those with metastatic disease [169]. On the other hand, Joncas et al. found that PSMA expression on plasma EVs was not a reliable marker for the identification of prostate cancer cell-specific EVs. Their conclusion was based on the proteomic analysis of PSMA-enriched EVs, in which no cancer-specific proteins could be identified [170].

Investigation of plasma EV cargo, mainly their protein and RNA content is receiving much attention as diagnostic and prognostic tools in prostate cancer. EVs isolated from either plasma or urine could be utilized to monitor prostate cancer stages, to discriminate high-grade from low-grade prostate cancer and benign disease, thereby reducing the number of unnecessary biopsies. Although it is not a blood-based marker, it is important to mention that ExoDx Prostate, a commercialized, urine-based test evaluates 3 EV-derived mRNAs, used to identify high-grade prostate cancer in patients with previous negative biopsies or with low initial PSA values [173].

Survivin is an apoptosis inhibitor selectively expressed in different tumours, including prostate cancer, and its main role is to promote cancer cell survival and protect cancer cells from apoptosis. It was shown that survivin was present in EVs secreted by prostate cancer cells and survivin levels in plasma-derived EVs from newly diagnosed prostate cancer patients (both early-stage and advanced cancers) and patients who relapsed after chemotherapy were significantly increased. These findings indicate that plasma EV-derived suvivin might be a promising liquid biopsy marker for the early diagnosis and systemic monitoring of prostate cancer [174].

Lundholm et al. found that NKG2D ligand-expressing prostate tumour-derived EVs selectively induced the downregulation of NKG2D on natural killer (NK) and CD8+ T cells, leading to damaged cytotoxic T cell function in vitro. Consistently with these data, surface NKG2D expression on circulating NK and CD8+ T cells was significantly decreased in patients with castration-resistant prostate cancer (CRPC) compared to healthy individuals [175]. These findings suggest that prostate tumour-derived EVs promote immune suppression and tumour escape by acting as down-regulators of the NKG2D-mediated cytotoxic response in prostate cancer patients.

Androgen receptor splice variant 7 (AR-V7) was associated with resistance to hormonal therapy in castration-resistant prostate cancer and plasma-derived EVs were shown to contain AR-V7 RNA [176]. Validation of AR-V7 as a potential target for treatment of CRPC could make it a clinically predictive biomarker of resistance to hormonal therapy and facilitate the decision-making process and therapy planning in these patients. Additionally, another study suggested that EV AR-V7 RNA was correlated with lower level of sexual steroid hormones in CRPC patients with a poor prognosis [170].

#### *2.5. MicroRNAs*

RNA content of EVs is considerably different from their parent cells, suggesting that cells can selectively sort their species of RNA into EVs, including small non-coding RNAs such as miRNAs or miRs with important regulatory functions on protein expression. Each miRNA regulates multiple target messenger RNAs (mRNAs). They control protein expression through the degradation of mRNAs or the inhibition of protein translation of target mRNAs by binding to the 30 -untranslated region (UTR). In view of their complex regulatory ability, it is not surprising that abnormal miRNA expression has been described in the pathogenesis of several diseases including cancer. Incorporation of miRNAs into EVs or binding to RNA-binding protein complexes increases their stability and protects them from degradation by various environmental factors [177]. Therefore, miRNAs are very stable in serum, plasma and other biofluids and are resistant to boiling, pH change, repeated freeze-thaw cycles, and fragmentation by chemical or enzymes [178], making them ideal biomarker candidates for the diagnosis, prognosis, and therapeutic planning in cancer disease, including prostate cancer. So far miRNA research in the prostate cancer field mainly focused on the characterization of differentially expressed miRNAs or miRNA panels involved in tumour progression [179]. Relatively few clinical trials have been conducted to date to explore miRNAs as indicators of prognosis or prediction of therapy response (Table 1).

Recently, plasma-derived EVs have proved to be better sources for miRNAs than unfractionated plasma/serum for certain but not all miRNAs. EV-incorporated miR-200c-3p and miR-21-5p could differentiate between prostate cancer and BPH, similarly EV-incorporated Let-7a-5p level could distinguish prostate cancer patients with Gleason score above 8 from those with Gleason score below 6. Both EV-incorporated and free miR-375 in the blood is an important miRNA biomarker candidate in prostate cancer. Huang et al. found that plasma EV-derived miR-375 and miR-1290 could predict overall survival for CRPC patients [180]. Expression of miR-141 and miR-375 increased in the

blood of high risk or metastatic CRPC patients [181,182]. Other groups also confirmed miR-375 as important diagnostic marker but only if tested in the whole plasma [183]. Blood miRNAs were shown to discriminate between prostate cancer and BPH, though studies differ on the type and source of miRNAs. In one study overexpression of plasmaderived EV-containing miR-10a-5p and miR-29b-3p, while in another one downregulation of plasma-derived free hsa-miR-221-5p and hsa-miR-708-3p were indicative of prostate cancer but not BPH [184,185]. Correlated expression levels of miR-20a, miR-21, miR-145, and miR-221 [186], miR-17, miR-20a, miR-20b, miR-106a [187] as well as miR-16, miR-148a and miR-195 [188] in the plasma could significantly distinguish high risk patients from those with low risk and some of these miRNAs were shown to confer an aggressive phenotype upon overexpression in vitro as well as an accelerated biochemical recurrence [187]. Fredsoe et al. validated a blood-based miRNA diagnostic model comprising of 4 miRNAs (miR-375, miR-33a-5p, miR-16-5p and miR-409-3p), called bCaP, in 753 patients with benign prostate lesions and multiple stages of prostate cancer and showed that combined with PSA, digital rectal examination and age bCaP predicted the outcomes of biopsies better than PSA alone [189].

An important miRNA cluster with predictive value towards therapy response is formed by miR-205 and miR-31. These miRNAs regulate apoptosis in prostate cancer cells by targeting antiapoptotic proteins Bcl-w and E2F6 and they are downregulated in prostate cancer cell lines derived from advanced metastatic cancers. It was shown that their decreased expression could contribute to resistance to chemotherapy-induced apoptosis making them key targets to improve prostate cancer response to chemotherapy [190]. In this context, upregulated plasma miR-205 expression in metastatic CRPC was associated with a lower Gleason score and a lower probability of both biochemical recurrence and clinically evident metastatic events after prostatectomy [181].

Several studies highlight the prognostic and predictive value of circulating miRNAs secreted by other than prostate cancer cells, most probably reflecting a systemic response. Bone marrow mesenchymal stem cells EV-derived miR-205 contributed to repress prostate cancer cell proliferation, invasion, migration and enhance apoptosis, which suggests that miR-205 could be a valid prognostic marker and a potential therapeutic target in prostate cancer [191]. It was shown that tumour-associated macrophages (TAM)-derived EVs with increased miR-95 content could mediate prostate cancer progression by promoting proliferation, invasion, and EMT [192].

Ionizing radiation is an important exogenous factor, which modifies miRNA expression in cells, including cancer cells. Altered miRNA expression patterns can influence cancer cell radioresistance and consequently lead to changes in radiation response. There are relatively few studies, which investigate miRNA expression profile changes induced by irradiation in prostate cancer patients, most studies are mainly in vitro investigations. A comprehensive review of miRNA expression alterations after various irradiation schedules in different prostate cancer cell lines was prepared by Labbé et al. The authors also summarize the most important radiotherapy-regulated cellular mechanisms in which miRNAs are involved [193].

MiR-106a and miR-20a overexpression conferred radioresistance to prostate cancer models by increasing clonogenic survival after radiotherapy [187,194]. In another study using prostate cancer cell lines with different intrinsic radiosensitivity miR-200, miR-221, miR-31 and miR-4284 were found to correlate with clonogenic survival of cell lines after irradiation [195]. Increased miR-21 and miR-146a/155 levels were found in radiotherapytreated prostate cancer patients with acute genitourinary side effects, indicating their potential to predict radiotherapy-related toxicities [196].

Gong et al. showed that circulating miR-145 levels were increased in prostate cancer patients responsive to neoadjuvant radiotherapy indicating that miR-145 might serve both as a predictive marker of therapy response and a novel therapeutic agent able to enhance the efficacy of radiotherapy [197]. MiR-93 and miR-221 plasma levels decreased significantly after either radical prostatectomy or radiotherapy but did not change after ADT and miR-

93 significantly correlated with Gleason score in a cohort of 68 prostate cancer patients compared to the observational cohort (*n* = 81) [198]. Two studies investigated EV-miRNAs as markers of therapy efficacy. The study by Li et al. identified a panel of 9 serum-derived EV-miRNAs, which could predict therapeutic benefit of carbon ion radiotherapy based on their baseline values. Post-therapy levels of miR-654-3p and miR-379-5p were associated with therapy efficacy [199]. Another study identified hsa-let-7a-5p and hsa-miR-21-5p as increased only in high-risk prostate cancer patients after radiotherapy compared to intermediate-risk patients [200]. It is important to highlight that candidate miRNAs in the two studies did not overlap, which might be due to different patient enrolment criteria, treatment protocol and sampling time after therapy. A further limitation of the cited studies is the low number of enrolled patients (*n* = 8 and 11, respectively), thus data must be confirmed on larger patient cohorts as well.






**Table 1.** *Cont.*





#### **3. Conclusions**

Table 1 summarizes the most important liquid biopsy-based biomarkers or biomarker candidates either already approved for clinical use or investigated in ongoing clinical trials or still in experimental phase. The high number of studies focusing on different cellular and secreted blood components as candidate liquid biopsy markers demonstrates the need for validated targets with prognostic and/or predictive value in screening prostate cancer patients.

The following biomarker categories were the most successfully tested in prostate cancer:


The major requirements and guidelines for biological parameters to be considered as biomarkers and to be clinically approved have been extensively described elsewhere [212,213]. Basically, the procedure of biomarker development should adhere to the REMARK guide-

lines. Besides their proven clinical relevance, proposed biomarkers should fulfil strict specificity and sensitivity criteria, also they have to be reproducible, easy-to-perform and to interpret and cost-effective. Despite the high number of promising biomarker candidates in prostate cancer very few have actually been approved for clinical use and their spread in the clinical routine is very slow. We believe that the most important reason for this relatively low transition of experimental biomarkers into clinical setting is that most studies stop at the discovery or qualification phase and fail to proceed to biomarker verification and validation. In Table 1 one can see that many of the listed clinical trials recruited very low patient numbers, which were not sufficient for validation purposes. A large-scale validation study of a prioritized biomarker needs substantial financial and collaborative efforts involving multi-institutional and optimally international collaboration, which might take years to be finalized. Within the European Union EU-supported multi-national collaborative projects could be one solution for wide-scale harmonized biomarker validation studies. Additionally, given the wealth of available data on biomarker candidates, a metaanalysis would help in biomarker prioritisation, highlighting the most promising targets for large-scale validation.

**Author Contributions:** Conceptualisation: K.L., K.B. and L.A. Manuscript drafting: K.B. and L.A. Manuscript editing and reviewing: K.B., G.S. and K.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Hungarian National Research, Development and Innovation Office under grant number NKFI-124879.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

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

#### **References**


## *Review* **Radiation Response in the Tumour Microenvironment: Predictive Biomarkers and Future Perspectives**

**Niall M. Byrne , Prajakta Tambe and Jonathan A. Coulter \***

School of Pharmacy, Queens University Belfast, Lisburn Road, Belfast BT9 7BL, UK; n.byrne@qub.ac.uk (N.M.B.); tprajakta05@gmail.com (P.T.)

**\*** Correspondence: j.coulter@qub.ac.uk; Tel.: +44-0-28-9097-2253

**Abstract:** Radiotherapy (RT) is a primary treatment modality for a number of cancers, offering potentially curative outcomes. Despite its success, tumour *cells* can become resistant to RT, leading to disease recurrence. Components of the tumour microenvironment (TME) likely play an integral role in managing RT success or failure including infiltrating immune *cells*, the tumour vasculature and stroma. Furthermore, genomic profiling of the TME could identify predictive biomarkers or gene signatures indicative of RT response. In this review, we will discuss proposed mechanisms of radioresistance within the TME, biomarkers that may predict RT outcomes, and future perspectives on radiation treatment in the era of personalised medicine.

**Keywords:** biomarkers; immune infiltrate; radiotherapy; stroma; tumour microenvironment

#### **1. Introduction**

Radiotherapy (RT) is a primary treatment modality for a number of cancers, offering potentially curative outcomes [1]. Radiation treatment modalities have significantly improved over the last two decades with the introduction of advanced techniques including stereotactic radiotherapy (SRT) and enhanced imaging methodologies to improve the precision of RT delivery, thus limiting damage to healthy tissue. However, despite these advancements, resistance to radiotherapy still occurs, resulting in disease recurrence. Characterisation of radioresistance has traditionally focused on the effects of RT on tumour *cells*, overlooking the impact on supporting stromal and immune *cells* that make up the tumour microenvironment (TME) [2]. Although components of the TME have been shown to regulate angiogenesis [3] and promote malignant progression and metastasis [4], their role in the response to RT and their contribution to radioresistance is less well characterised [5]. As such, a greater understanding of the TME response could identify predictive biomarkers indicative of RT success or failure.

Predictive biomarkers offer an approach for stratifying patients who will respond favourably to a particular treatment, in turn sparing those for whom the modality may be less effective. While radiotherapy is intrinsically a precision treatment, directed to the specific architecture of the patient's tumour, it has so far lacked a personalised approach, taking into consideration patient-specific genomic alterations or TME composition, factors that could predict the outcome of radiotherapy [6,7]. In this review, we summarise some of the recent advances in understanding the TME response to ionising radiation. In particular, we discuss the effect of radiotherapy on the tumour stroma and immune response, and how this may contribute to radioresistance. This review will also consider the biomarkers or gene expression signatures that have been developed to predict radiation outcomes. Lastly, we conclude by exploring how these approaches could be used to develop personalised radiotherapy treatment plans to improve patient outcomes.

**Citation:** Byrne, N.M.; Tambe, P.; Coulter, J.A. Radiation Response in the Tumour Microenvironment: Predictive Biomarkers and Future Perspectives. *J. Pers. Med.* **2021**, *11*, 53. https://doi.org/10.3390/ jpm11010053

Received: 9 December 2020 Accepted: 13 January 2021 Published: 16 January 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/).

#### **2. Radiation Response in the Tumour Microenvironment** RT can be a cure for many; however, for some patients, the treatment fails or re-

RT can be a cure for many; however, for some patients, the treatment fails or resistance occurs. Though ionizing radiation can induce DNA damage in tumour *cells*, a potential barrier to the success of RT may be its effects on the other components of the local TME, including the vasculature, stroma and the immune infiltrate (Figure 1). These components can influence tumour progression and response to treatment. Understanding how they are influenced by RT may be critical in predicting disease outcomes. Extracellular vesicles (EVs) including exosomes have also been shown to play a role in cancer progression, immunomodulation and importantly, in modifying the response to radiation; key examples of which are below. However, recent detailed articles focusing on the role of EV-modulated radiation response exist; as such, EVs will not form a primary focus of this review [8,9]. sistance occurs. Though ionizing radiation can induce DNA damage in tumour *cells*, a potential barrier to the success of RT may be its effects on the other components of the local TME, including the vasculature, stroma and the immune infiltrate (Figure 1). These components can influence tumour progression and response to treatment. Understanding how they are influenced by RT may be critical in predicting disease outcomes. Extracellular vesicles (EVs) including exosomes have also been shown to play a role in cancer progression, immunomodulation and importantly, in modifying the response to radiation; key examples of which are below. However, recent detailed articles focusing on the role of EV-modulated radiation response exist; as such, EVs will not form a primary focus of this review [8,9].

**Figure 1.** The effect of radiation on the TME. Schematic showing the role of ionizing radiation on components of the TME and predictive biomarkers of radiation response. DAMPs, damage-associated molecular patterns; EC, endothelial cell; ECM, extracellular matrix; ICD, immunogenic cell death; MHC, major histocompatibility complex; PD-1, programmed cell death protein 1; PD-L1, programmed death ligand-1; RT, radiotherapy; TAM, tumour-associated macrophage; TCR, T-cell receptor; TGFβ, transforming growth factor beta; TME, tumour microenvironment. **Figure 1.** The effect of radiation on the TME. Schematic showing the role of ionizing radiation on components of the TME and predictive biomarkers of radiation response. DAMPs, damage-associated molecular patterns; EC, endothelial cell; ECM, extracellular matrix; ICD, immunogenic cell death; MHC, major histocompatibility complex; PD-1, programmed cell death protein 1; PD-L1, programmed death ligand-1; RT, radiotherapy; TAM, tumour-associated macrophage; TCR, T-cell receptor; TGFβ, transforming growth factor beta; TME, tumour microenvironment.

#### *2.1. Tumour Immune Microenvironment 2.1. Tumour Immune Microenvironment*

Immune evasion, the process by which tumour *cells* can avoid immune recognition and destruction, has become one of the hallmarks of cancer [10]. Subsequently, more recent therapeutic developments have focused on shifting the TME from an immunosuppressive environment to an immune-activated one through the use of immunotherapeutics: treatments that can effectively remove the brakes on immune signals mounting an anti-tumour response. RT has been shown to have contradictory immunomodulatory effects, influencing both proinflammatory and immunosuppressive responses, which likely influence response to treatment [5]. The inflammatory milieu of the TME, or the tumour immune microenvironment (TIME), is composed of T *cells*, natural killer (NK) *cells*, dendritic *cells* (DCs) and tumour-infiltrating myeloid *cells* (TIMs) including tumour-associated macrophages (TAMs), myeloid-derived suppressor *cells* (MDSCs) and dendritic *cells* (DCs), all of which are recruited into the TME through altered chemokine and cytokine signalling [11]. The extent and relative proportion of immune infiltration can also influence the response to treatment and progression. Tumours can be broadly separated into two categories based on their TIME: those that are immune "hot", being infiltrated with T lymphocytes; and those that are immune "cold", with poor infiltration [12]. In immune Immune evasion, the process by which tumour *cells* can avoid immune recognition and destruction, has become one of the hallmarks of cancer [10]. Subsequently, more recent therapeutic developments have focused on shifting the TME from an immunosuppressive environment to an immune-activated one through the use of immunotherapeutics: treatments that can effectively remove the brakes on immune signals mounting an anti-tumour response. RT has been shown to have contradictory immunomodulatory effects, influencing both proinflammatory and immunosuppressive responses, which likely influence response to treatment [5]. The inflammatory milieu of the TME, or the tumour immune microenvironment (TIME), is composed of T *cells*, natural killer (NK) *cells*, dendritic *cells* (DCs) and tumour-infiltrating myeloid *cells* (TIMs) including tumour-associated macrophages (TAMs), myeloid-derived suppressor *cells* (MDSCs) and dendritic *cells* (DCs), all of which are recruited into the TME through altered chemokine and cytokine signalling [11]. The extent and relative proportion of immune infiltration can also influence the response to treatment and progression. Tumours can be broadly separated into two categories based on their TIME: those that are immune "hot", being infiltrated with T lymphocytes; and those that are immune "cold", with poor infiltration [12]. In immune "hot" tumours, regulatory

T *cells* (Tregs) and TAMs cooperate to support the immunosuppressive TME and may be more susceptible to the immunomodulatory effects of radiotherapy [13]. Furthermore, these immune-inflamed tumours, including non-small cell lung cancer and melanoma, are more likely to respond favourably to immune checkpoint inhibitors in comparison to immune "cold" tumours, including pancreatic and prostate tumours [14]. Lack of tumour antigens, defects in antigen presentation and poor T-cell homing to the TME by the stroma may all contribute to a "cold" tumour immune phenotype; mechanisms to modulate immune infiltration and turn these tumours "hot" could improve response to therapy [14–16].

The ability of radiotherapy to modulate systemic immune responses may contribute towards the observations of tumour regression at non-irradiated sites, an effect described as an abscopal response. Abscopal effects are particularly relevant when RT is combined with immune checkpoint blockade. In preclinical syngeneic models of prostate cancer, a combination of radiotherapy (20 Gy in two fractions) with antibodies against programmed death-1 (anti-PD-1) or programmed death ligand-1 (anti-PD-L1) (iRT) significantly increased median survival (70–130%) in comparison to anti-PD-1 monotherapy, contributing to an abscopal response in which the unirradiated tumours responded similarly to the irradiated tumours. Importantly, this effect was shown to be mediated through antitumour CD8+ (cytotoxic) T *cells* [17]. Clinical observations of the abscopal effect have been rare in radiation oncology; however, with the development and advancement of immunotherapeutics, these observations are becoming more frequent across a variety of tumour types [18]. Clinically, in patients with unresectable melanoma combining anti-PD-1 therapy with hypofractionated RT (typically 26 Gy in 3–5 fractions) resulted in abscopal treatment responses in 36% of patients [19,20]. Targeting of another immune checkpoint, cytotoxic Tlymphocyte antigen 4 (CTLA-4), with the monoclonal antibody ipilimumab in combination with RT has also been shown to result in abscopal responses both preclinically in models of breast cancer and clinically in melanoma and lung cancer patients [21–24]. Interestingly, EVs isolated from irradiated tumour *cells* (H22 cells and 4T1 *cells*; 8 Gy) in vitro were shown to have immunomodulatory effects when mice were inoculated in vivo, enhancing CD8+ and CD4+ T-cell infiltration in lung metastasis in comparison to nonirradiated EVs [25]. Dose and fractionation are likely to play a critical role in the immunological responses to RT; however, the molecular and cellular mechanisms underpinning this immune-priming effect are still poorly understood [26].

RT-induced cell death is typically thought to occur through DNA damage, particularly in the form of double-strand breaks (DSB). Subsequently, the tumour cell response to radiationinduced DNA damage (RIDD) is dependent on its DNA damage response (DDR), which can activate downstream signalling to repair damage, thus contributing to radioresistance [27]. While the immune cell compartment, including lymphocyte and myeloid populations, may be more resistant to RIDD, RT can modulate immune signalling within the TME, promoting immune cell recruitment and activation and triggering immunogenic cell death [28]. RTinduced immunogenic cell death results in a cascade of events, starting with the release of damage-associated molecular patterns (DAMPs) (Figure 1) [29]. These "danger" signals released by tumour *cells* include high-mobility group box 1 (HMGB1) and ATP, triggering innate and adaptive immune responses through the expression of major histocompatibility complex (MHC) class I and MHC-II molecules. These antigen-presenting *cells* (APCs) can in turn can prime CD8+ T *cells* to induce an antitumour response [28]. In fact, RT has been shown to upregulate MHC-I expression preclinically in tumour cell lines in vivo, an observation that has been recapitulated in ex vivo-irradiated tumour biopsies [30]. Cytosolic double-stranded DNA (dsDNA) released as a result of RIDD can also promote dendritic cell activation through guanosine monophosphate–adenosine monophosphate synthase (cGAS)/stimulator of IFN *genes* (STING)/interferon (IFN) signalling, leading to CD8+ T-cell activation [31].

TIM populations, including TAMs, form another important component of the TIME and although they have a complex plasticity, they are usually organised as classically activated (M1) or alternatively activated (M2) *cells*. Numerous stimuli including chemokines can influence TAM polarisation from a proinflammatory (antitumour) M1 to an antiinflammatory (protumour) M2 phenotype, which promotes tumour angiogenesis, tissue remodelling and tumour progression [32]. Interestingly, the frequency of TAMs has also been associated with clinical treatment response and disease progression [33,34]. In murine tumour models, low-dose gamma irradiation (LDI; 2 Gy) has been shown to promote repolarisation of M2-like TAMs towards M1-like inducible nitric oxide synthase (iNOS) expressing TAMs, contributing to T-cell recruitment and tumour regression (Figure 1) [35]. TAMs and MDSCs are dependent on colony-stimulating factor (CSF1) signalling for recruitment into the TME. In murine models of breast cancer, blocking CSF1/CSF1R signalling inhibited TAM recruitment and delayed tumour regrowth following RT (5 Gy), an effect associated with an increase in CD8+ T *cells* and a reduction in CD4+ (helper) T *cells* [36]. Similar effects were observed following CSF1R signalling blockade in combination with RT (3 Gy, five fractions) in syngeneic models of prostate cancer in vivo. Furthermore, serum levels of CSF1 were also shown to be elevated in prostate cancer patients following RT [37]. Clinically, in patients with T3 rectal cancer, a short course of radiotherapy (neoadjuvant hyperfractionated 25 Gy in 10 fractions; surgery performed on day 2–5) promoted TAM repolarisation towards an M1-like proinflammatory phenotype. Interestingly, ex vivo modelling of this response suggested that HMGB1 in EVs from irradiated tumour *cells* could be responsible for this effect on TAM polarisation [38].

#### *2.2. Cancer-Associated Fibroblasts*

The stromal compartment of the TME plays an integral role in the response to treatment, including RT (Figure 1). Radiotherapy-induced tissue fibrosis is a late side effect where myofibroblast transformation leads to the excess production of collagen and deposition of components of the extracellular matrix (ECM) [39]. RT can also lead to the release of the pleotropic cytokine transforming growth factor beta (TGFβ), which can modulate fibroblast phenotype and function [40]. Fibroblasts recruited into the TME are transformed into cancer-associated fibroblasts (CAFs), where they play a role in regulating the extracellular matrix [41]. Furthermore, CAFs are responsible for the secretion of a number of cytokines (including interleukin 6 (IL6) and IL8), chemokines (including C-X-C motif ligand 12 (CXCL12)) and growth factors (including TGF-β and platelet-derived growth factor (PDGF)) that can influence immune cell fate and tumour progression, often contributing to the immunosuppressive TIME [42]. However, the effects of RT on the stromal compartment of the TME including CAFs are less well understood and they appear to have contradictory roles, contributing to both tumour growth and suppression [43]. Coimplantation of A549 lung tumour xenografts with preirradiated CAFs (at both 18 Gy × 1 fraction or 6 Gy × 3 fractions) abrogated the protumour growth effect observed in tumours coimplanted with nonirradiated CAFs [44]. In contrast, irradiated fibroblasts (1, 6 or 12 Gy) have been shown to express high levels of TGF-β1 and promote human T3M-1 squamous cell carcinoma (SCC) invasion and growth [45]. Furthermore, EVs derived from CAFs were shown to contribute to colorectal cancer cell stemness and radioresistance (6 Gy) in vitro, through the activation of the TGF-β signalling pathway [46]. It is therefore clear that more work is needed to understand the complex role of CAFs in the tumour response to RT.

#### *2.3. Tumour Vasculature*

The integrity of the tumour vasculature differs significantly from that of physiologically normal vessels, characterised by abnormal recruitment of pericytes, leading to increased tortuosity and porosity. This, in part, contributes to treatment failure through poor drug penetration into the TME, establishing local hypoxia gradients and increasing the yield of reactive oxygen species [47]. The effect of RT on the tumour vasculature has been well studied, with tumour blood vessels and their endothelial *cells* proven to exhibit increased sensitivity to radiation, a response likely dependent on total radiation dose and fractionation schedule [5,48,49]. Vascular damage is mainly witnessed at radiation

doses exceeding 5 Gy. Conversely, individual, low-dose fractions have been shown to temporarily stimulate blood flow, while at higher or cumulative doses, the vascular network is disrupted, promoting hypoxic stress that can trigger tumour cell death [50,51]. In a recent dose-escalation study, single administration of 2, 4 or 8 Gy doses were shown to compromise the tumour vasculature in a dose-dependent manner, prolonging the survival of mice bearing CT-2A (high-grade glioma) tumours. Interestingly, this was also associated with changes in the TIME, promoting an increase in CD8+ T *cells* and a reduction in M2-like TAMs [52]. Potiron et al. [53] reported that RT (at both 10 × 2 Gy and 2 × 12 Gy) induces tumour vasculature normalisation and remodelling, thus improving the distribution and efficacy of the anticancer drug doxorubicin (DOX) [53]. Further evidence of the effects of RT effects on endothelial cell permeability has been demonstrated in vitro. Monotherapy radiation doses to primary human umbilical vein endothelial *cells* (HUVECs) increased permeability and transmigration of tumour *cells*, owing to altered metalloprotease ADAM10 expression and degradation of VE-cadherin, both of which play an integral role in maintaining intercellular junctions and vascular integrity [54]. High radiation doses (>20 Gy) were also found to cause transient endothelial dysfunction, platelet leukocyte adhesion and increased expression of hypoxia-inducible factor-1α (HIF-1α) in pancreatic tumours [55]. However, a recent study indicated that high-dose RT (>8 Gy) induced expression of Notch1 signalling in HUVEC monolayers. Consequently, in vivo high-dose RT, in combination with inhibition of Notch1 signalling, resulted in a significant reduction in tumour vessel endothelial cell coverage in comparison to high-dose RT alone, suggesting Notch1 signalling may protect tumour vessels from radiation-induced damage [56]. Furthermore, it is also well understood that oxygenated tumour *cells* are preferentially killed by RT, due to oxygen-induced fixation of radiation-induced DNA damage. However, this effect has been proven to accelerate the production of proangiogenic cytokines, inhibiting treatmentinduced apoptosis, stimulating a postradiotherapy angiogenic burst that can contribute to eventual tumour regrowth [57].

#### **3. Predictive Biomarkers of Radiation Response**

Precision medicine based on common tumour-specific alterations, emerging from high-throughput molecular profiling, has become a reality in recent years. This approach underpins the discovery of clinically validated prognostic and/or predictive biomarkers, allowing for stratification of patients based either on those most likely to derive benefit or have treatment-related harm limited. This strategy gained significant momentum in the chemotherapy field with the development of various commercially produced kits such as Prosigna (NanoString Technologies, Inc., Seattle, USA) and MammaPrint (Agendia, Amsterdam, The Netherlands), designed to aid clinical decision-making [58,59]. However, equivalence in radiotherapy has not yet been achieved due to the variability in radiation response, an effect attributed to tumour heterogeneity. Heterogeneity is an umbrella term used to describe both intra- and intertumour variability at the morphological, physiological and more recently, genetic levels. Divergence of these features exerts a profound influence on localised factors such as vascular integrity, tumour oxygenation and immune infiltrate, ultimately influencing treatment outcome (detailed in Section 2 [5,13,48]). In an effort to address the issue of heterogeneity, research efforts have shifted from focusing on macroscopic phenotypic or environmental variation to the identification of commonality at the molecular level. Table 1 provides an outline of biomarkers for radiotherapy response in a number of tumour types (summarised in Figure 1); these are discussed further in the sections below.


**Table 1.** Biomarkers of radiotherapy response.


#### **Table 1.** *Cont*.

#### *3.1. Gene Signatures of Radiation Sensitivity*

An early example of this approach used the clonogenic assay to profile radiation sensitivity, based on survival fraction data at 2 Gy (SF2), of the NCI-60 cancer cell line panel [60]. This was then correlated against gene expression data from four published microarray platforms, identifying significant alterations in expression profiles for 31 *genes*, common to each microarray dataset. Unsurprisingly, significant suppression of *genes* which regulate cell cycle progression (*CCNA2*, *CDK6*, *CCND1)* and DNA damage repair were associated with increased radiosensitivity. *CCND1*, the gene encoding for cyclin D1, stalls cell cycle progression, providing time for DNA damage repair, ultimately suppressing radiation-induced apoptosis [72]. Therefore, suppressed *CCND1* and other cell cycle regulatory *genes* may contribute, in part, towards a genetic signature for identifying radiosensitive tumours. A second set of *genes* common to the top 10% most radiosensitive (SF2 < 0.2) *cells*, and totally absent from the most radioresistant (SF2 > 0.8), were those involved in integrin signalling, cell adhesion and cytoskeletal remodelling. Cell-adhesion complexes and integrin signalling act both directly and indirectly to influence radiation response [73]. Cell-to-cell contact and adhesion with the extracellular matrix are central features of the protumour phenotypes of migration and invasion. Along with integrin β1, the 31-gene profile identified downregulation of *ITGB5*, the gene encoding integrin β5, as a highly significant indicator of radiosensitivity [60]. Indeed, radiosensitisation achieved through the antagonism of αvβ5 integrin using a cyclic-RGD (arginine-glycine-aspartate) containing peptide was the focus of a large phase III clinical trial for the treatment of glioblastoma multiforme [74]. This was based on the rationale that αvβ5 antagonism suppresses tumour angiogenesis and metastasis, an effect in part attributed to the dampening of major cancer-related signalling pathways, including *Wnt* and *PI3K* [75]. Developed as a universal predicator of radiation sensitivity, independent of tumour type, many of the 31 *genes* identified likely hold predictive value in relation to radiation response. However, stringent application using only the most radiosensitive or radioresistant *cells* again highlights the problem of heterogeneity, where 80% tumour models analysed exhibited intermediary gene expression alterations, diluting the predictive power of the signature.

Recent approaches adopting a similar strategy tend to focus on a specific disease type. Breast cancer radiotherapy is most commonly used in the adjuvant setting to improve treat-

ment outcomes, forming a core strategy of breast conservation surgery and mastectomy. However, not all patients benefit from adjuvant radiotherapy and some experience significant debilitating late effects [76]. The importance of identifying those who will benefit most from adjuvant radiotherapy was neatly demonstrated in a study using FFPE tumour tissue from the Danish Breast Cancer Cooperative Group (DBCG82bc) cohort. Applying a sevengene signature to stratify patients into either high-risk loco regional recurrence (LRR) or low-risk LRR, the authors were able to establish that postmastectomy radiotherapy would benefit only those identified as high risk, providing no benefit to low-risk patients [61]. Adopting a similar strategy to the 31-gene signature, Speers et al. [62] correlated the radiation sensitivity (SF2) of a panel of breast cancer models against gene expression changes, developing a radiation sensitivity signature (RSS), which was subsequently shown to be the most significant factor in prediction of loco-regional recurrence, beating all clinicopathologic features used in clinical practice [62]. While a clear step forward, RRS remains a prognostic signature for loco-regional control, and not predictive of radiation response. Similar predictive gene signatures have been developed, including a six-gene signature (including *genes* such as HOXB13 and NKX2-2) that was also shown to predict radiotherapy sensitivity in breast cancer [77]. Applying a 24-gene signature to prostate cancer patients who had undergone radical prostatectomy to identify those most likely to benefit from postoperative radiotherapy similarly found that those with a high PROTOS (postoperative radiation therapy outcomes score), indicative of radiation-sensitive tumours, significantly benefited from radiotherapy, with a 10-year metastasis rate of 4% (95% CI 0–10) versus 35% (CI 7–54) for those not receiving radiotherapy. However, in the low PROTOS score group, radiotherapy proved detrimental (HR 2.5 (CI 1.6–4.1); *p* < 0.0001) in the 157-patient cohort training group and of no benefit in the 248-patient validation cohort [63]. Liu et al. [64] recently used multiple omics data to develop a prediction model of sensitivity to radiation in head and neck squamous cell carcinoma (HNSCC) tumours. A 12-gene signature was established from differentially expressed *genes* in patients treated with or without RT and used to develop a scoring system. Those HNSCC patients with a low score had a higher radiosensitivity and were shown to benefit from RT [64].

#### *3.2. DNA Damage Response Biomarkers*

The antitumour effects of radiotherapy are directly proportional to the degree to which potentially lethal DNA DSBs are both induced by radiation and are sustained by the cell following activation of DDR processes. Continual refinements to the delivery of radiotherapy have ensured that the DNA-damaging properties of the most commonly utilised radiation sources, such as X-rays and γ-rays, minimise dose to surrounding healthy tissue, while focusing dose on the target volume. In parallel, intensive research efforts have led to the development of numerous small-molecule inhibitors targeting key DNA damage repair proteins, thus sustaining radiation-induced damage, resulting in increased tumour cell death. This is the fundamental basis of many radiosensitising strategies. Key targets of the DNA damage response pathways for which clinically utilised inhibitors have been developed include the ATM/ATR (ataxia–telangiectasia mutated and Rad3-related) signalling pathways, PARP (poly (ADP-ribose) polymerase), DNA-PKcs (DNA-dependent protein kinase, catalytic subunit), BRAC1 (breast cancer1 C terminal) and HIF-1, amongst others. While reviewing the full therapeutic potential of these inhibitors is beyond the scope of the current article, several recent publications provide comprehensive details of this field [27,78,79]. Herein, we aim to focus on the utility of gene expression alterations in DDR *genes* as prognostic/predictive indicators of radiation response. Piening et al. [65] developed an early radiation-derived gene signature, evaluated for prognostic utility in breast cancer. The signature was derived from gene expression alterations following a 5 Gy dose across a panel of nontumour lymphoblast *cells*, a relevant point given that genomic instability in tumours support aberrant DDR activity. Expression levels of 219 *genes* were altered with 160 being induced and 59 repressed by radiation. Using a gene set enrichment algorithm [80], the prognostic utility of the signature was evaluated against publicly avail-

able breast cancer microarray data. With respect to the repressed *genes*, tumour samples neatly clustered into two groups, aligning with gene repression or not, where the former strongly correlated with increased proliferation and poor overall treatment outcomes. Similarly, *genes* induced by radiation correlated positively with those who responded favourably to radiation treatment, promoting the expression of *genes* involved in negative regulation of the cell cycle, apoptosis (e.g., caspases) and DNA damage repair proteins. Importantly, applying the same approach but using the NCI-60 cancer cell line panel to derive the radiation signature failed to discriminate between favourable and poor outcomes, with no overlap between the altered gene set signature [65]. This clearly illustrates the impact of genomic instability in influencing the DDR response and an important point for consideration in the development of radiation biomarkers.

Another study exploited the overlapping DNA damage responses activated by both chemotherapy and radiotherapy, producing a radiation-induced 30-gene signature. This signature was proven capable of discriminating between breast cancer patients likely to achieve a pathological complete response (pCR) to neoadjuvant chemotherapy and poorresponding patients. Importantly, pCR represents the most relevant clinical end point for predicting improved overall and disease-free survival [81]. In addition to *genes* clearly linked to DNA damage pathways, such as the extracellular signal-regulated kinase (*ERK*) pathway, *AKT*, *mTOR* and *NF-kB*, radiation significantly elevated the expression of metabolism processing *genes*, in particular *PDHA1* and *LDHB*. These genes encode for key proteins driving pyruvate metabolism and energy production, along with the catalytic conversion of pyruvate to lactate, thus indicating that tumours with a high metabolic demand are more likely to prove sensitive to the effects of chemo- and radiotherapy [66].

#### *3.3. Hypoxia Biomarkers*

As outlined previously, hypoxia resulting from aberrant tumour vasculature can influence RT resistance. As such, there is a strong rationale for identifying robust biomarkers of tumour hypoxia that predict response to RT [82]. Traditionally, tumour hypoxia was measured using oxygen electrode probes, endogenous HIF-1α levels, physiological markers such as pimonidazole staining or other imaging methodologies (MRI). However, gene signatures may better represent the nuances of hypoxia within the TME that might predict response to RT. To this end, Eustace et al. [67] developed a 26-hypoxia gene signature (informed by a 121-gene hypoxia meta-signature derived from datasets of head and neck, breast and lung cancers [83]) predicting treatment response in laryngeal cancer. This hypoxia signature, composed of *genes* involved in glucose metabolism (*ALDOA*, *ENO1*, *LDHA*), cell proliferation (*CDKN3*, *FOSL1*) and angiogenesis (*VEGFA*), could predict those patients receiving RT for whom hypoxia-modifying ARCON (accelerated radiotherapy with carbogen and nicotinamide) therapy would be of benefit in laryngeal carcinomas [67]. The approach of stratifying patients for hypoxic modification of RT has also been performed by Troustrup et al. [68] to classify HNSCC tumours as "more" or "less" hypoxic [84]. A 15-gene hypoxic signature including *genes* for stress response (ADM, HIG2), cell proliferation (FOSL2, IGFBP3) and glucose metabolism (ALDOA, FKBP3) was developed from HN-SCC cell lines under hypoxic conditions, and subsequently validated in patients that had previously been hypoxia-evaluated [85,86]. The predictive power of this gene signature was validated in a clinical cohort of HPV-negative HNSCC tumours, with those classified as having "more" hypoxic tumours having more favourable outcomes (loco-regional tumour control and disease-specific survival) after combining RT with hypoxia modification using nimorazole [68].

#### *3.4. Liquid Biopsies*

Minimally invasive liquid biopsies represent an area of intense research interest. While the field is in relative infancy, with no commercially validated tests, the identification of circulating biomarkers predicative of radiation response holds tremendous potential. MicroRNAs (miRNAs) are differentially regulated in a number of disease types and fol-

lowing exposure to ionizing radiation; they therefore offer a potential biomarker to predict treatment response in cancer [87–89]. A radiotherapeutic response predication was developed for patients with lower-grade glioma (LGG), based on the expression of five miRNAs. The signature was capable of classifying those as low-risk or high-risk in terms of survival and radiation response, based on the analysis of miRNA expression profiles in 624 patients. This signature was found to be superior to isocitrate dehydrogenase (IDH) mutational status in predicting survival in LGG [90]. Of particular interest is free plasma or exosome secretion of miRNAs predictive of radiation response: Li et al. [69] linked low-level miR-221 expression with increased radiation sensitivity, a finding subsequently correlated with several patient studies reporting that low serum levels of miR-221 and miR-125b are indicative of low-risk prostate cancer [69,91,92]. Furthermore, Li et al. [70] associated the levels of three miRNAs (miR-374a-5p, miR-342-5p and miR-519d-3p) with radiation responses in the plasma of patients with nonmetastatic rectal cancer and head and neck cancers. Prediction classifiers were developed from miRNA signatures in pre- and postradiotherapy samples and could significantly distinguish between radiation responders and poor responders 6 months postradiotherapy [70]. The importance of effective biomarkers, particularly in the prostate cancer setting, is evident considering that prostate-specific antigen (PSA) screening has formed the bedrock of prostate cancer diagnosis for over 25 years—a test lacking in specificity—resulting in significant treatment related morbidities from overdiagnosis and overtreatment [93]. Given the role of the TIME in influencing tumour fate postradiotherapy (detailed in Section 2), immune infiltrate composition in the TME may predict radiotherapy response and prognosis in cancer patients [5,94]. Cui et al. [71] pioneered a combined radiation sensitivity (RS) gene signature with an antigen-presentation (AP) immune signature, establishing a dual-modality approach with predictive capabilities of radiation response. Independently, both RS and AP signatures were proven capable of predicting increased disease-specific survival (DSS) in patients identified with either radiosensitive or immune-effective tumours, with the reverse observed in radioresistant and immunedefective individuals. Importantly, integration of both signatures further strengthened the predictive capabilities of either signature used independently [71].

#### **4. Conclusions and Future Perspectives**

RT is the treatment of choice for a number of cancer, designed to target and kill tumour *cells*; however, it triggers a myriad of effects on other components of the TME, including the vasculature, stroma and the immune compartment [5]. The immunomodulatory effects of RT are complex, with reported changes to the proportions and functionality of T *cells* and antigen-presenting dendritic *cells*, and effects on TAM polarisation within the TME. This effect is further complicated by clinical observations of an increase in the abscopal effect reported in patients receiving RT in combination with immunotherapeutics. RT has also been shown to affect tumour vascular architecture, inducing tissue fibrosis. It is important to note that the majority of responses to RT in the TME reported above are in the context of conventional X-ray or photon radiation therapy. Recent advances in the clinical delivery of RT, including high-energy proton beam therapy and heavy ion therapy, have the improvement of delivering more dose in the Bragg peak with a lower dependence on tissue oxygenation and improved biological effectiveness [95]. While these newer treatment modalities are likely to have biological effects on the components of the TME outlined in this review, their response has been less well characterised [96,97]. Therefore, it is of critical importance to take into consideration the role of the TME when considering radiobiological responses and disease recurrence. As RT techniques have evolved over the last two decades, so too have their physical precision, aided by improved imaging guidance and technological advancements. However, genomic precision has lagged, as most RT treatment planning is designed around the tumour and local tissue architecture, with the aim to deliver the maximum dose to the tumour while sparing healthy tissue. However, as highlighted above, genomic signatures could allow for a greater prediction of those patients for whom RT would be of benefit as a single therapy or in combination with

radiation sensitizers or hypoxia modifiers [6]. Yet, of critical importance, these findings further stress the necessity for a precision medicine approach, in that not only do patients with radioresistant tumours fail to experience radiotherapy benefit, but that treatment is actually detrimental both in terms of DSS and toxicities associated with radiation-induced late effects [71]. Taking a more "personalised" approach to RT could ensure patients receive the most benefit from their treatment.

**Author Contributions:** Conceptualization, N.M.B. and J.A.C.; writing—original draft preparation, review and editing, N.M.B., P.T. and J.A.C. All 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:** Data sharing not applicable.

**Acknowledgments:** Figures created using MedART (creative commons license): https://smart. servier.com/.

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

#### **References**


*Communication*
