DNA Methylation Biomarkers for Prediction of Response to Platinum-Based Chemotherapy: Where Do We Stand?

Simple Summary Platinum-based agents are one of the most widely used chemotherapy drugs for various types of cancer. However, one of the main challenges in the application of platinum drugs is resistance, which is currently being widely investigated. Epigenetic DNA methylation-based biomarkers are promising to aid in the selection of patients, helping to foresee their platinum therapy response in advance. These biomarkers enable minimally invasive patient sample collection, short analysis, and good sensitivity. Hence, improved methodologies for the detection and quantification of DNA methylation biomarkers will facilitate their use in the choice of an optimal treatment strategy. Abstract Platinum-based chemotherapy is routinely used for the treatment of several cancers. Despite all the advances made in cancer research regarding this therapy and its mechanisms of action, tumor resistance remains a major concern, limiting its effectiveness. DNA methylation-based biomarkers may assist in the selection of patients that may benefit (or not) from this type of treatment and provide new targets to circumvent platinum chemoresistance, namely, through demethylating agents. We performed a systematic search of studies on biomarkers that might be predictive of platinum-based chemotherapy resistance, including in vitro and in vivo pre-clinical models and clinical studies using patient samples. DNA methylation biomarkers predictive of response to platinum remain mostly unexplored but seem promising in assisting clinicians in the generation of more personalized follow-up and treatment strategies. Improved methodologies for their detection and quantification, including non-invasively in liquid biopsies, are additional attractive features that can bring these biomarkers into clinical practice, fostering precision medicine.


Introduction
Platinum-based agents (cisplatin (CDDP), carboplatin, and oxaliplatin) are broadly used for the treatment of several cancer types. Notwithstanding their broad spectrum of clinical use, several concerns remain, especially the emergence of treatment resistance, which causes additional challenges. Over the last decades, epigenetic biomarkers, especially those related to DNA methylation, have increasingly shown their value as cancer biomarkers, amenable for simple, fast, and low-cost detection in a non-or minimally invasive way. These are highly versatile, with value for diagnosis, risk stratification, and prediction of response to a specific treatment, sparing patients from harmful and unnecessary side effects. However, the setup and confirmation of a reliable biomarker with a strong routine clinical application is a complex process, which takes many steps from in vitro experiments to in vivo pre-clinical model validation and patient tissue analysis, with further validation in independent (multi-institutional) cohorts to ensure the desired high sensitivity, specificity, and accuracy [1,2]. Indeed, there is a plethora of studies proposing new biomarkers, but very few have made their way to clinical practice for several reasons including pre-analytical issues, cohort demographic variations, and a lack of standardized reporting, among many others [3]. In this review, we focused on epigenetic-based biomarkers, specifically DNA methylation, which might be used to predict response to platinum-based chemotherapy, emphasizing their establishment and detection methods.

Platinum-Based Chemotherapy: Brief Definition and Mechanisms of Action
Platinum anticancer drugs are routinely used for the treatment of several types of malignancies, both solid and hematolymphoid. Since the discovery of CDDP anti-tumor activity in 1969 and following its approval by the United States Food and Drug Administration (FDA) in 1978, platinum-based chemotherapy has been widely used in cancer treatment, which makes research on its mechanism of action and resistance extremely relevant for contemporary oncology [4,5]. Although CDDP remains the most commonly used platinum drug in the clinic, two analogs were also approved for tumor types: carboplatin (in 1989) and oxaliplatin (in 2002) [6,7].
Platinum-based chemotherapy has been proven to be effective to treat several types of cancers [8,9] and is widely used in the treatment of very distinct tumors, including esophageal (EC), gastric (GC), lung (LC) (small cell (SCLC) and non-small cell (NSCLC)), colorectal (CRC), and head and neck (HNC) cancers [10]. It is also used in urothelial (UC) and cervical (CC) carcinomas, as well as testicular and ovarian germ cell tumors (TGCT, OC). Platinum compounds are also used to treat other malignancies including leukemias, melanoma, neuroendocrine neoplasms, sarcomas, and tumors of neuroectodermal origin (such as neuroblastoma), demonstrating the versatility of these agents [11]. It is reasonable to theorize that such wide effectiveness in very distinct tumor types (with different biology, genomic drivers, risk factors, and molecular background) might be due to multiple pathways in which platinum-based drugs interfere.
Platinum compounds, especially CDDP, demonstrate remarkable clinical success in TGCT, enabling high cure rates (~80%) even in cases of heavily metastatic disease [12][13][14]. However, this specific type of testicular tumor shows hypersensitivity to CDDP, a fact that is tightly linked to its epigenetic and developmental biology background, and clinicians have not been able to reproduce such a success rate on somatic-type tumors of adulthood treated with the same platinum compounds [15][16][17]. This creates a need to fully understand the mechanism of action of this therapy and its mechanisms of resistance to improve patient care. In this review, we addressed the three platinum compounds more widely used in clinical practice, with a particular focus on the most widely used: CDDP.
CDDP has been used for years as a first-line treatment for several cancer types, either alone or in combination with other therapeutic options, such as radiation (to serve as a radiosensitizer) or other chemotherapeutics [18]. It is usually administered as either a neoadjuvant (for tumor shrinkage) or adjuvant (to lessen the risk of recurrence) therapy [19][20][21]. It may also be used in a palliative chemotherapy scenario, due to its cytotoxic activity, in an attempt to maintain patient quality of life [11]. The downside is that platinum agents demonstrate several critical side effects, such as nephrotoxicity and peripheral neurotoxicity, limiting the dose that might be used for patient treatment [22]. Additionally, cancer survivors previously treated with platinum disclose traceable levels of CDDP in urine and plasma many years after treatment, which is a major concern that may cause long-term side effects, triggering a decline in quality of life and, ultimately, resulting in death [23][24][25]. The current precision medicine paradigm is no longer compliant with sustaining such side effects either in a short and/or long term, and all efforts must

A Brief Introduction to DNA Methylation
The radical difference between genetic and epigenetic changes is that genetic lesions are irreversible whereas epigenetic lesions are potentially reversible as they are associated with gain or loss of DNA methylation or other modifications of chromatin, thus enabling therapeutic intervention [43,44]. Epigenetic mutations, also called epimutations, are heritable; they may be constitutional and derived from a germline, thus expected to be found in all of the tissues of an individual, or they may be somatic, eventually restricted to a specific somatic tissue [45]. Epigenetic aberrations may consist of abnormal patterns of DNA methylation, disrupted patterns of histone posttranslational modifications (PTMs), altered expression of small non-coding RNAs, and alterations in chromatin composition and/or organization [46]. Histone modification and DNA methylation specifically regulate gene expression at the transcriptional level, preceding morphological changes associated with neoplastic transformation and even genetic alterations [47].

DNA Methylation Regulates Transcription and Affects Protein Levels
DNA methylation mostly affects CpG dinucleotides and is involved in tumorigenesis through three main mechanisms: locus-specific (e.g., tumor suppressor genes (TSG)) hypermethylation [48], global hypomethylation of the cancer genome [49], or direct mutagenesis of 5mC sequences [50,51]. It is noteworthy that all three routes occur simultaneously, indicating the importance of methylation as an epigenetic driver in cancer development. Hypermethylation negatively impacts transcription, reducing levels of the proteins responsible for processes such as DNA damage repair, creating a fundamental replication advantage over normal cells [52]. It was demonstrated that DNA methylation alterations result from the altered expression of methyltransferases [53]. Since these enzymes are responsible for the transfer of a methyl group to DNA, their up/downregulation leads to DNA hyper/hypomethylation, thus impairing normal epigenetic regulation and enabling malignant transformation and progression together with the increase in chemoresistance [54]. Importantly, due to their significant role in the epigenome regulation, methyltransferases could serve as a suitable therapeutic target in cancer treatment [55].

Epigenetic-Based Cancer Biomarkers
According to the National Cancer Institute (NCI) Dictionary of Cancer Terms, a cancer biomarker is a biological molecule that is found in blood or other body fluids or tissues, which indicates an abnormal, cancer-related process or condition [56]. Biomarkers vary depending on their objective (risk assessment, diagnosis, prognosis, or prediction of response to therapy), and herein we focus on the predictive type of biomarkers, which forecast the response to a specific treatment [57]. Ideally, a biomarker should have perfect (100%) specificity and sensitivity [57,58]. We reviewed available data on cancer cell lines and patient tissues because these are critical for the development of reliable biomarkers' detection and establishment, including those based on the hypermethylation of gene promoters [59,60]. Recently, many studies focused on the need to identify and select reliable biomarkers for all kinds of cancer treatment. Sample (blood, urine, stool) collection is minimally or non-invasive and may thus be performed more frequently, allowing for easier and earlier diagnosis, disease monitoring, and easy storage [61]. Furthermore, it may assist clinicians in the decision of prescribing neoadjuvant or adjuvant treatment [59]. In this context, personalized medicine has become a priority. It is widely acknowledged that there is no universal treatment for cancer and that some patients are resistant to specific types of treatment. A perfect treatment strategy should target cancer cells in all the pathways that are crucial for their survival. Since epi-genetic aberrations may, at the least partially, contribute to cancer resistance to therapy and relapse, epigenetic modulation, such as DNA demethylating agents, may prove useful [62,63]. In this scenario, epigenetic-based biomarkers become relevant to determine whether a patient might benefit from a specific chemotherapy regimen, including those that are platinum based.
Among cancer-related epigenetic alterations, we focused our literature search on DNA methylation as a biomarker for predicting patient response to platinum-based chemotherapy ( Figure 1). Indeed, DNA methylation itself is advantageous compared to other epigenetic biomarkers not only because it can be detected in non-invasively collected body fluids but also because it is representative of tumor heterogeneity. Whereas primary tumor and metastatic deposits' tissue samples are highly heterogeneous, having several tumor cell clones, which may be missed by needle biopsy sampling, circulating tumor cells or nucleic acids are representative of the tumor bulk, either primary or metastatic [63,64]. Additionally, either in tissue or liquid biopsy specimens, DNA is much more stable and resistant to degradation (by formalin fixation, freeze, and thawing procedures) than RNA [57,65,66]. Furthermore, data obtained from the assessment of DNA methylation may be compared to absolute reference points (fully methylated/unmethylated DNA) allowing for quantification [65].
Currently, there is a large number of methods to detect DNA methylation, either target-based (e.g., methylation-specific polymerase chain reaction (PCR) (MSP), bisulfite sequencing, pyrosequencing methylation-specific restriction endonucleases, etc.) or genome wide-based (e.g., 450K or 850K array) and a vast amount of data are available publicly, enabling comparisons [67]. Finally, improvements in technology are under development to facilitate the detection of DNA methylation biomarkers in an absolute way without the need for pre-amplification reactions (e.g., droplet digital PCR (ddPCR)) [68,69].
Although such biomarkers seem auspicious due to their feasibility, they are not widely used in practice because of their limited sensitivity compared to the available standard-ofcare tools. The particular reason for this circumstance is that often the detection of a single biomarker (e.g., promoter hypermethylation of a specific gene) is not sufficient to obtain a reliable conclusion and gene panels are required to overcome this limitation. Furthermore, for validation purposes, promoter methylation status must be confirmed using multiple methods and in several cohorts with distinct demographic features before. Additionally, other environmental conditions might impact gene methylation acting as confounders in cancer biomarker studies [67,70]. Additional problems are related to biomarker results' interpretation and reporting, including normalization (appropriate normalizers, method of relative quantification), sample and DNA input conditions (which may be prohibitive in specific clinical scenarios), cost-related issues, etc To set up a reliable biomarker predictive of response to chemotherapy, one needs to identify relevant genes by compiling and testing training and validation cohorts and comparing DNA methylation status among specific cohorts of patients, specifically, in this setting, patients who responded (either completely or partially) to treatment and those who did not respond and endured a poor outcome [57,67,71]. Additionally, adjusting for demographic and clinicopathologic factors (age, gender, grade, stage, baseline characteristics of patients, etc.) is very relevant since DNA methylation biomarkers may lose their predictive value after adjustment in multivariable models. Moreover, cancer cell line testing is important to illuminate how chemotherapy-sensitive and -resistant cells react to treatment, evaluating whether the methylation profile changes over time and if the use of a demethylating agent sensitizes resistant cells [72]. For that purpose, the collection of patient tissue samples (for instance, biopsies, FFPE tissue samples, frozen samples, etc.), the extraction of DNA (assuring the best possible quality), and performing bisulfite treatment (or variations, such as with the use of methylation-sensitive endonucleases) followed by targeted MSP-based methods are required ( Figure 1). Then, if sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy reach high levels of performance, a biomarker is a candidate for further testing in body fluids to determine whether it will constitute a reliable biomarker for clinical use [65].

Research Methodology
For the purposes of this review, a PubMed database search was conducted with the query (cisplatin OR carboplatin OR oxaliplatin) AND (DNA methylation OR epigenetics) AND (resistance OR chemoresistance). The search only considered original records published in English (i.e., reviews were excluded) and no restricted time interval was considered. Initially, the articles were chosen through comprehensive abstract analysis, and the final count was reached after a critical, full-text read of those that conveyed sig-nificant information for the topic. For this review, only studies analyzing the role of DNA methylation in platinum-based chemotherapy resistance using human cell lines or patient samples were considered. Figure 2 depicts the flow diagram representing the methodology used to reach the final set of selected sources of information. The information collected is summarized in Tables 1 and 2, which depict epigenetically regulated genes associated with platinum-based chemotherapy resistance in cell lines (Table 1) or patient tissues ( Table 2).  There are global and individual loci methylation changes in resistant cells. Expression of Casp8AP2 in oxaliplatin-resistant cells was reduced and associated with increased promoter methylation. After exposure to the demethylating agent, the sensitivity of resistant S3 cells was restored to the same level as in untreated SiHa cells. [76] GPx3 CDDP, oxaliplatin CRC RKO, SW48, LOVO, HCT116, SW480, SW620, COLO205, CACO2, and HT29

MSP
Cell lines tested showed different sensitivity to CDDP, and MSP analysis disclosed a correlation between GPx3 methylation status and mRNA expression levels. Cell lines with GPx3 promoter methylation and downregulation had increased sensitivity to platinum.
[77] FGF5 methylation is associated with response to chemoradiotherapy with CDDP. FGF5 expression was induced by CDDP treatment in three unmethylated cell lines, but not in two methylated cell lines. Exogenous FGF5 overexpression in a cell line with FGF5 promoter methylation conferred resistance to CDDP. [81] PAX5 CDDP ESCC NUEC1 and TE3 qMSP PAX5-silenced cells showed relatively higher cell proliferation and cell cycle promotion, suggesting acquisition of CDDP resistance due to methylation-associated gene silencing. BeadChip; bisulfite pyrosequencing Expression of KLK13, ETV7, FSCN1, CPT1C, and NOTCH3 before and after CDDP chemotherapy differed due to promoter methylation. These alterations may be associated with mechanisms of GC drug resistance and may be used as biomarkers to predict drug sensitivity. [84] GTSE1 CDDP GC AZ521, OCUM-1, SNU610, and SNU719 MSP All hypomethylated cell lines depicted higher GTSE1 expression. Loss of GTSE1 expression significantly enhanced sensitivity to CDDP treatment as shown by a~5-fold decrease in IC50 values in AZ521-kd cells. GTSE1 knockdown in GC cells disclosed it as the major cause of CDDP resistance. [85]

SLFN11
CDDP GC NUGC3, SNU5, SNU16, PHM82, NCI-N87, BGC823, MCG803, and AGS MSP; bisulfite sequencing SLFN11 loss of expression was associated with promoter hypermethylation. Seven out of eight cell lines expressed SLFN11 and the promoter region was methylated. SLFN11 re-expression suppressed proliferation in SNU16 and MGC803 cell lines and sensitized cells to CDDP. [86] HumanMethylation27 DNA Analysis BeadChip (high-throughput methylation profiling) Global methylation changes were determinant of the acquisition of resistance to CDDP, but methylation of some genes (CFLAR, ERBB2, KLF11) stood as the most promising markers to predict drug resistance.
[87] MT1E promoter methylation is common in human melanoma and might be considered a biomarker. Gene silencing was suggested to play a role in the resistance of melanoma to chemotherapy. Hypermethylation of RIP3 promoter region was detected in all LC cell lines but not in primary human bronchial epithelial cells. RIP3 mRNA and protein expression increased after demethylating agent treatment in LC cell lines with methylated promoters, but not in those without methylation.
Restored RIP3 expression sensitized cells to CDDP.
[99] p57Kip2 is epigenetically downregulated in a carboplatin-resistant cell line (PEO1CarbR). MSP analysis of the CpG island located at the 5' region of the p57Kip2 gene disclosed that methylation level was significantly higher for PEO1CarbR than for PEO1. Microarray-based methylation analysis Analysis of TRIB2 confirmed an indirect contribution of hypermethylation to gene silencing and the functional impact of this gene on A2780 chemosensitivity. TRIB2 overexpression in resistant cells led to reduced IC50, and shRNA-mediated silencing of TRIB2 in parental sensitive A2780 cells increased their resistance.
[116]    The genes listed were found to disclose higher methylation levels in CDDP-resistant NSCLC samples compared to sensitive tumors.
[100] The quantitative methylation data for probes located within the RIP3 promoter CpG island revealed significantly higher methylation in 25% of tumors. When RIP3 promoter was methylated, protein expression was suppressed, correlating with increased resistance to CDDP. [99]

MethylCap-seq;
bisulfite pyrosequencing FZD10 was the most differentially methylated gene among two of the chemoresponsive-related groups. FZD10 expression was significantly lower due to promoter methylation in the extreme responder HGSOC patient group compared to the non-responder group. [127]

DNA Methylation and Platinum Resistance
Resistance to platinum treatment can be divided into two main types: intrinsic and acquired. Many patients that initially are sensitive to the treatment often develop resistance to it during their treatment course, causing relapse and reducing its overall clinical efficacy [29]. The development of new CDDP analogs, with fewer side effects, also aimed to tackle resistance to platinum-based chemotherapy, a goal which was not fully achieved and that also met with decreased effectiveness.
Changes in the epigenetic landscape, a cancer hallmark [129], appear to be nonrandom and are associated with the acquisition of chemoresistance to platinum in various types of cancers [130,131]. Indeed, platinum-based chemotherapy seems to induce changes in DNA methylation patterns [33,77,96,101,132]. This epigenetic mechanism plays a substantial role in the platinum resistance mechanism, affecting the transcription and translation of genes involved in reduced platinum influx to the cell or increased export (e.g., ABCB1), increased DNA damage repair routes (e.g., BRCA1, ERCC1, MLH1, hMSH2), inactivation of apoptosis pathways (e.g., Casp8AP2, GULP1, p73, RIP3), or increased platinum detoxification (e.g., MT1E) (Tables 1 and 2). DNA damage repair pathways, for instance, have been proven to be of critical importance in the process of resistance to platinum compounds due to the DNA adducts these create [8,26,29,62,133]. The mismatch repair pathway (MMR) is a vital tool to keep genome stability, and a deficiency in this system has been shown to cause CDDP resistance in cells, associated with poor prognosis in some tumors [26,133]. Several previous studies have shown that promoter hypermethylation of genes involved in this pathway, such as MLH1 and hMSH2, is associated with the acquisition of resistance to platinum therapy [72,92,105,110,121,124]. The NRF2/KEAP1 pathway plays a key role in the chemoresistance process of different tumor types and is capable of inhibiting apoptosis, promoting cell proliferation, and chemoresistance [134]. This pathway has already shown to be regulated by epigenetic modifications, including DNA methylation [135]. Additionally, genes such as ERα and ABC transporters, proven to be involved in the chemoresistance process, are regulated by the NRF2/KEAP1 pathway [136].
Our review of the literature disclosed several publications on the relevance of DNA methylation-based biomarkers for the prediction of response to platinum therapy, notwithstanding their heterogeneity concerning methodological settings (cell lines vs. tumor tissues). However, no such biomarker has been approved so far [57]. There are several critical steps in the validation of biomarkers as well as several hurdles that make the process of approval very strict, complex, time-consuming, and expensive, justifying why so very few of these markers make it all the way to clinical practice [2,57,65,71,137]. All these limitations make DNA methylation predictive biomarkers for platinum-based chemotherapy still relatively unexplored. One of the main problems observed in this type of study is the size of the validation cohort. If the cohort of patients treated with platinum compounds is not large enough, reliable conclusions about the predictive value of biomarkers cannot be drawn [74,92,99,105]. Thus, sample size estimation is mandatory to assure that the study cohort(s) enable the identification of significant differences between responders and nonresponders if they exist. Additionally, multiple clinical variables must be considered, and the results should be adjusted/stratified according to these parameters, such as tumor type, stage, the platinum compound used for treatment, and treatment response, among others, as these are highly relevant clinical factors that may significantly influence methylation levels as well as the likelihood of response to therapy [57,71].
Another relevant issue is the estimation of tumor cell density in tissue samples tested. For example, the percentage of tumor in tissue sections chosen for DNA purification varies widely among published reports, from >30% of tumor cells [117] to >70% [82,125] or even >90% [98,122]. This variability is very likely to influence the determination of methylation levels (even considering that normalization for input has been made), jeopardizing the comparison of results and the reproducibility of experiments, undermining the possibility of biomarker validation [138].
An important and very common limitation in the publications assessed is the lack of information in clinical studies concerning biomarker performance parameters, such as sensitivity, specificity, accuracy, etc. [81,92,123]. These parameters are critical to evaluating the potential of the new DNA methylation biomarkers and comparing them with conventional methods or other markers that are routinely used in clinical care at present. Considering published data reviewed and depicted in Table 2, reported sensitivity or specificity values are modest (e.g., 67% [92] or 45% [81] sensitivity), probably owing to the very small size of initial tissue biopsies, which may not provide enough DNA for the experiments using several replicates, or contamination of purified DNA with tissue residues such as proteins, complicating the determination of correlation between gene methylation and expression [57,65,139,140]. To overcome this problem, optimal sample processing should be ensured and cohort size should be increased to account for variations in DNA concentration and purity among samples, enabling a more robust analysis of results [65].
Interestingly, differences in tissue condition regarding treatment are also apparent among the studies on methylation analysis. Whereas, in most studies, the tumor tissue analyzed for DNA methylation was collected after platinum treatment [108,109,117], in some assessed tumor tissue samples collected from untreated patients, primary cell cultures were established and were exposed to platinum prior to methylation analyses [88,98]. Results from these two strategies must be compared with caution because the presence or absence of the tumor microenvironment and altered cell communication derived from culture conditions is likely to entail the activation of different pathways [141].
Most clinical studies (i.e., those based on patient cohorts) performed DNA methylation analysis in tissues, either fresh, frozen, or formalin-fixed and embedded in paraffin. Indeed, very few have used liquid biopsies [75,103], which seem advantageous considering they are easier, less invasive, faster, and more comfortable to obtain compared with conventional tissue biopsies. Importantly, liquid biopsies allow for real-time monitoring, as blood or urine may be drawn periodically and biomarkers assessed over shorter or longer periods of time [137,142,143].
In addition to in vitro and clinical studies, in vivo animal models are very useful in cancer research as they more closely replicate the complexity and heterogeneity of cancer tissues compared to in vitro cell line studies [144]. Nonetheless, they are much more expensive and represent a superior work burden and some of these models may not very precisely mimic the human tumor microenvironment [145]. From our search, very few studies on DNA methylation biomarkers or therapeutic targets of platinum-based chemotherapy have used animal models and the ones that did mostly used those in in vivo assays to complement the in vitro cell studies [100,108,115].
Notwithstanding the hypothesis that DNA methylation biomarkers might help to predict a response to platinum-based chemotherapy, they may also represent important therapeutic targets that might help sensitize tumor cells to platinum compounds [72,74,80,86,95,99,107]. For instance, a previous study showed that the impairment of ABCB1 expression due to promoter hypermethylation caused a reduction in the ABCB1 transporter and lessened CDDP resistance [95]. Another study showed that promoter methylation levels of BRCA1, a key gene involved in DNA repair, were higher in CDDP-resistant ovarian cancer cell lines and that exposure to a demethylating agent sensitized those cells to platinum treatment [107]. Thus, if further demonstrated in clinical studies, DNA methylation patterns might allow for the improvement of therapeutic strategies [57,65,71]. Figure 3 illustrates the ideal process of how a biomarker of platinum-based agent resistance (in this case, hMSH2 promoter hypermethylation) could be validated and confirmed as a predictive biomarker, assisting in the therapeutic decision for OC patients, improving survival and quality of life.
1 Figure 3. Schematic representation of pipeline for validation of DNA methylation-based biomarker to predict resistance to platinum-based chemotherapy in ovarian cancer (OC) patients. After a clinical diagnosis of OC (top of the picture), if the disease was staged as IC or higher, the recommended treatment is adjuvant chemotherapy with a platinum agent (CDDP, carboplatin, or oxaliplatin), eventually in combination with Taxol. However, there is a 20% probability that the patient will be resistant to platinum agents, which complicates the choice of treatment [66]. To select the best treatment method, biomarker validation could be performed. This follows with non-invasive patient sample collection (for instance, blood plasma), which can be used for circulating tumor DNA methylation analysis, focusing on platinum agent resistance. In this case, gene promoter hypermethylation indicating platinum resistance in OC was detected (e.g., hMSH2) [124], indicating that the patient will likely endure platinum resistance. Thus, not only may the side effects of ineffective treatment [22] be avoided but alternative treatments, eventually including epi-drugs, should be considered. Created with BioRender.com (accessed on 1 February 2022).
Epi-drugs, which may be inhibitors of DNA methyltransferases, histone deacetylases, histone acetyltransferases, histone methyltransferases, or histone demethylases, may play an important role in cancer treatment by enhancing the effects of combinational therapy with platinum-based compounds as sensitizers [146,147]. This was shown in several clinical trials [73,148,149] and opens the way for a wider use of predictive DNA methylation-based biomarkers in tumors candidating for treatment with platinum compounds. Although holding substantial potential for the enactment of precision medicine, more robust validation studies are required to provide definitive evidence.

Conclusions
Presently, cancer is a leading cause of death worldwide and its incidence and mortality are increasing. Thus, in parallel with the implementation of preventive and early diagnosis measures, the development of effective and patient-specific therapeutic strategies is required to tackle this growing public health problem. Platinum-based chemotherapy, in use for more than 40 years, remains the first-line treatment for many types of cancer; resistance to this therapy is a major concern. Importantly, epigenetic dysregulation, specifically aberrant DNA methylation, plays an important role in the resistance process. Thus, biomarkers based on DNA methylation might enable the identification of those tumors more prone to demonstrate or acquire resistance to platinum compounds as well as constituting therapeutic targets enabling the sensitization of tumors. Thus, many studies have been undertaken to unveil and validate candidate biomarkers. Our review disclosed several mechanistic studies with cell lines and animal models, as well as some clinical studies, using patient samples, which identified some promising DNA methylation biomarkers predictive of response/resistance to platinum treatment. However, none of these biomarkers has been validated yet since most clinical studies analyzed small cohorts and the heterogeneity of patients, samples, and analytical methods precludes a meaningful and decisive conclusion. Hence, there is an urgent need to set up clinical validation studies, with adequate statistical power to enable the identification of the added value of those epigenetic biomarkers. This requires a joint effort from basic scientists and clinicians, departing from the more robust pre-clinical and clinical data available and bridging the gap that will lead to biomarker-assisted therapeutic decisions for patients who are candidates for platinum-based chemotherapy.