Epigenomics and ovarian carcinoma.

Many different genes have been identified to be hypermethylated and silenced in ovarian carcinoma. Some of these genes are located in regions with known LOH in ovarian carcinoma and/or are epigenetically regulated in other types of malignancies, for example OPCML [ 62 – 64 ], DLEC1 [ 65 ], RASSF1A [ 63 , 66 – 69 ], ARLTS1 [ 70 ], ARHI [ 71 , 72 ] and TCEAL7 [ 73 ]. Since aberrant DNA methylation silences transcription, novel TSGs can be identified by analyzing CpG island hypermethylation. The breast cancer susceptibility gene 1 ( BRCA1 ) has been one of the most comprehensively analyzed owing to its tumor suppressor function and its known role in inherited forms of ovarian cancer, where hypermethylation only occurs in ovarian and breast cancers [ 74 , 75 ]. This hypermethylation of its promoter is associated with the loss of its expression [ 76 – 78 ] and is predominantly detected in cancers that exhibit LOH at the BRCA1 locus [ 75 , 77 ]. Furthermore, the methylation-mediated silencing of BRCA1 and other repair genes, such as MGMT , could result in further inactivation of TSGs or activation of oncogenes, which further promote and drive ovarian tumorigenesis [ 79 ]. By contrast, BRCA2 promoter hypermethylation is rarely found in ovarian cancers [ 80 , 81 ]. Other genes found to be hypermethylated and downregulated in ovarian carcinoma demonstrate different properties: p16 [ 82 ], SPARC [ 83 ], ANGPTL2 [ 84 ] and CTGF [ 85 ] have tumor suppressor activity, LOT1 [ 86 ] and PAR-4 [ 87 ] have pro-apoptotic function, ICAM-1 [ 88 ] and CDH1 [ 89 ] participate in cell adhesion, and PEG31 [ 90 ] plays a role in imprinting. HOXA10 [ 91 ], HOXA11 [ 91 ], PALB2 [ 92 ] and TUBB3 [ 93 ] are other examples of hypermethylated genes, the latter having a contribution to taxane resistance. As mentioned previously, global DNA hypomethylation is another feature of cancer and in ovarian epithelial neoplasms this increases with malignancy [ 94 ]. To date, hypomethylation has been demonstrated to lead to the abnormal expression of a few genes, including maspin ( SERPINB5 ) [ 95 ], SNCG [ 96 , 97 ] and CLDN4 [ 98 , 99 ] in ovarian carcinomas. In addition, hypomethylation associated with the L1 and human endogenous retrovirus-W retrotransposons, which are repetitive sequences that are widely distributed throughout the genome, is consistent with higher expression levels that occur in malignant compared with non-malignant ovarian tissue [ 100 ]. It is hypothesized that an increase in hypomethylation promotes recombination among homologous elements, leading to chromosomal aberrations, which are associated with cancer [ 101 , 102 ]. Other overexpressed genes associated with promoter hypomethylation are BORIS , a cancer testis antigen family candidate oncogene [ 103 ], and IGF2 , an imprinted gene implicated in other malignancies [ 104 ]. In addition, DNMT1 and −3b transcript levels have been reported to be increased in some ovarian cancer cell lines [ 105 ], as well as in primary and recurrent epithelial ovarian carcinoma [ 106 ], which could contribute to methylation-induced silencing of key TSGs, and have some correlation with clinical pathology and prognosis of epithelial ovarian carcinoma. However, in one study, DNMT3A1 and DNMT2 RNAs were significantly lower in carcinomas compared with low malignant potential (LMP) tumors, but DNMT3B1/-B2 RNA had significantly higher levels in carcinomas than in the LMP tumors [ 107 ]. Interestingly, gene methylation patterns are often associated with molecular, clinical and pathological features of ovarian carcinomas. For example, aberrant methylation of the promoters of SFN (an inhibitor of cell cycle progression), TMS1 and WT1 are more frequent events in clear-cell ovarian tumors than in other histological types [ 63 , 108 – 110 ]. Furthermore, Makarla et al. reported that RASSF1A, APC, GSTP1 and MGMT show aberrant methylation exclusively in invasive ovarian carcinomas [ 111 ] when compared with LMP tumors. In addition, histone modifications contribute to ovarian cancer progression via the downregulation of different genes. GATA4 and −6 gene silencing was found to correlate with hypoacetylation of histones H3 and H4, and loss of histone H3 K4 tri-methylation at their promoters [ 112 ]. The cell cycle regulatory proteins cyclinB1 [ 113 ] and p21 cip1/waf1 [ 114 ], and ADAM19 [ 115 ] were also found to be regulated by histone modifications. These advances in the knowledge of the ovarian methylome strongly indicate that DNA hypermethylation plays a crucial role in initiation, promotion and maintenance of ovarian carcinogenesis, which may contribute and synergistically interact with other genetic alterations to induce the development and progression of ovarian cancer.

Advances that aid in the understanding of ovarian cancer on a molecular level have provided important tools for molecular testing for high-risk populations, predictive markers for selecting patients for certain classes of drug therapies and molecular diagnostics for the noninvasive detection of early ovarian cancer. The epigenetic revolution that has come about in the field of cancer biology during the last few decades has provided valuable knowledge on how gene expression plays a key role in cancer formation and progression, and has created new insights helping us to understand the mechanisms involved in this process. Studies carried out in the field of ovarian cancer epigenetics have led to the realization that understanding the alterations of genes and pathways during the earliest steps of ovarian cancer development can aid clinical management of the patients in the near term. An understanding of the epigenetic signals that dictate the metastatic and/or drug-resistant phenotype will provide the information necessary to develop drugs to control or prevent advanced disease. Epigenetic therapy combined with chemotherapeutic agents holds significant promise for successful treatment of ovarian cancer in the future. Further epigenetic studies on ovarian cancer stem cells, along with development of more specific epigenetic drugs, may hold the key to our ability to successfully reprogram the abnormal ovarian cancer methylome. The considerable recent advances encourage us to believe that improvements in our knowledge of the epigenetic basis of ovarian cancer will continue to reduce the burden of this disease.

A wide array of techniques has been used to detect and understand gene-specific and genome-wide epigenetic modifications in cancer. This article presents an overview of recent epigenomic technologies that have been used to discern the ovarian cancer methylome. Table 1 summarizes the technologies used to explore the ovarian cancer methylome. Epigenetic alterations have several advantages as a means to detect and classify cancer. First, methylation analysis utilizes DNA, a more chemically stable molecule than RNA and protein. Second, aberrant DNA methylation is a binary signal, where the presence of methylation indicates the presence of malignant cells [ 4 ]. This can be detected at a low concentration in a background of excess normal DNA molecules by sensitive assays such as methylation-specific PCR (MSP) [ 116 ] and quantitative fluorogenic methylation-specific PCR [ 117 ], which allow the detection of a single methylated allele in 10,000 unmethylated alleles [ 118 ]. The third advantage of using DNA methylation to detect cancer is that assay design can focus on a single amplifiable region (e.g., CpG island) rather than scanning an entire gene for mutations. In addition, methylation biomarkers are detectable in patient serum/plasma and other bodily fluids draining or surrounding a tumor site [ 119 ]. On the other hand, studies of DNA methylation for ovarian cancer could present certain bias, which is related to sampling, technique, design and data analyses. To overcome various biases, a systematic approach is needed to identify a panel of true-positive biomarkers from the large number of biomarkers reported every year. Methylation of dozens of genes in various types of samples has been correlated with ovarian cancer, but, so far, most studies have been conducted in a single center with a limited numbers of samples. To determine the usefulness of DNA methylation in ovarian cancer detection, it is essential to conduct a specific project using a well-defined end point, the same set of blinded specimens, appropriate experimental design and data analyses, and standard technology to evaluate methylation biomarkers to generate a reliable conclusion about the usefulness of these markers for a specific clinical use in ovarian cancer. Before the development of bisulfite treatment of DNA, the methods used for detection of methylation were performed using high-performance liquid chromatography and high-performance capillary electrophoresis [ 120 ]. Nevertheless, the study of DNA methylation was initially almost entirely based on digestion with restriction-enzyme assays that can differentiate between methylated and unmethylated recognition sites in genes of interest [ 121 ]. This approach demonstrated several disadvantages that limited its use: from incomplete restriction-enzyme cutting to limitation of the regions that can be studied. The bisulfite-conversion technique, which reproducibly changes unmethylated cytosines to uracil but leaves methylated cytosines unchanged [ 122 ] was the mainstay for the creation of several sensitive DNA methylation detection techniques ( Table 1 ), including bisulfite sequencing, MSP and combined bisulfite restriction analysis [ 116 , 122 , 123 ]. MSP, although not a quantitative method, is very sensitive and has been the most widely used method for DNA methylation analysis on clinical samples, also partly owing to its simplicity [ 116 ]. Owing to its subjectivity, several real-time MSP methods, such as MethyLight™ [ 124 ], quantitative multiplex MSP [ 125 , 126 ] or pyrosequencing [ 127 ], have been developed, improving the detection of small amounts of DNA methylation in a quantitative fashion. MethyLight is a highly sensitive assay, capable of detecting methylated alleles in the presence of a 10,000-fold excess of unmethylated alleles. The assay is also highly quantitative and can very accurately determine the relative prevalence of a particular pattern of DNA methylation [ 124 ]. Pyrosequencing is a sequencing-by-synthesis method that quantitatively monitors the real-time incorporation of nucleotides through the enzymatic conversion of released pyrophosphate into a proportional light signal [ 127 ]. MSP and quantitative fluorogenic MSP technology can only detect CpG islands that are within the primer sequences and cannot detect CpG sites outside the methylation-specific primers. Pyrosequencing, in an unbiased manner, is a newly emerging method, although its intrinsic short-read sequencing (normally only up to 30 bp at a time) presents a disadvantage in comparison with DNA sequencing [ 127 , 128 ]. Several other methylation assays have been developed for different purposes and each has certain advantages, limitations and suitability. Besides the gene-specific profiling methods described previously, several genome-wide techniques have been useful for the study of global DNA methylation patterns in normal and cancer cells ( Table 1 ). Restriction-landmark genomic scanning is one of first genome-wide methylation analyses described that can evaluate the methylation status of thousands of CG-rich sequences and simultaneously obtain information on the gene copy number [ 129 ]. However, this is a laborious technique and a comparatively large amount of DNA is needed for the assay. The application of DNA microarray technology made it possible to discover new techniques that have had an important impact on cancer epigenetics. Two examples of microarray assays are differential methylation hybridization and gene expression profiling. Differential methylation hybridization was developed by Huang et al. and allows for the detection of differential methylated CpG islands between two different samples. Differential methylation hybridization has been widely used in the identification of aberrantly methylated gene promoters that are differentially expressed in various cancers [ 130 ]. Finally, gene expression profiling assesses genome-wide DNA methylation patterns, by comparing expression levels from cancer cells before and after treatment with a demethylating drug, HDAC inhibitor (HDACI) or both [ 131 – 133 ]. The identified candidate genes are further verified by quantitative real-time PCR and promoter methylation analyses. Another important technique for DNA methylation analysis, useful because arbitrary primed PCR is carried out using DNA templates that have been enriched for methyl sequences, resulting in preferential amplification of CpG islands and gene-rich regions, is amplification of intermethylated sites [ 134 , 135 ]. However, validation by bisulfite genomic sequencing is required. In addition, other important advances made in profiling the cancer epigenome have been achieved by techniques such as HpaII tiny fragment enrichment by ligation-mediated PCR assay, which uses a modified approach to globally analyze DNA methylation patterns [ 136 ]and other methods based on chromatin immunoprecipitation (ChIP) such as methylated DNA immunoprecipitation [ 137 – 139 ]. Recently, Illumina released the new generation of BeadArrays for assessment of DNA methylation where 27,578 CpG loci, covering more than 14,500 genes, can be analyzed at single-nucleotide resolution. Very recently, data derived from this new generation of BeadArrays have been published [ 140 ]. The most commonly used technologies for DNA methylation are summarized in Table 1 . As bisulfite sequencing is the gold standard for DNA methylation studies, the gold standard for accurately assessing global levels of histone modifications is mass spectrometry; however, it requires a high degree of technical expertise and is difficult to apply to the entire genome [ 6 ]. Nowadays, one of the most powerful techniques to identify and characterize the interactions of specific genomic DNA sequences associated with a target protein, such as transcription factors, is ChIP [ 141 ]. An antibody specific to the target protein is used to immunopreprecipitate the protein–DNA complexes. After the cross-link between the two is reversed, the DNA sequences are then uncovered by amplification and sequencing. Furthermore, recent techniques that combine ChIP with serial analysis of gene expression technology and high-throughput sequencing techniques, have been developed for profiling histone modifications [ 142 – 144 ].

With the acknowledgement that epigenetic mechanisms contribute to the formation and progression of tumors, efforts have been made in order to develop novel epigenetic therapies to target cancer cells. One important characteristic of epigenetic alterations is reversibility, unlike genetic mechanisms, which are are irreversible processes. This feature has promoted the development of pharmacologic inhibitors of DNA methylation and histone deacetylation [ 181 – 184 ], which have been proven to demethylate DNA and inhibit histone deacetylation, in order to reverse epigenetic silencing of key genes, leading to re-expression of these genes in cancer cells and reactivation of important cellular tumor suppression pathways. The majority of these therapeutic agents are cytosine analogs, 5-azacytidine (Vidaza®) and 5-aza-2′-deoxycytidine (decitabine) being the most extensively studied DNMT inhibitors (DNMTIs). These compounds incorporate into DNA in the place of cytosines during DNA replication, covalently attaching to DNMTs [ 185 ] and cause the depletion of active DNMT enzymes. They act primarily on DNMT1 and their demethylation activity is replication dependent, requiring several cell divisions to achieve genomic demethylation [ 186 ]. The demethylating effect of decitabine is stronger than Vidaza since the former only binds DNA, unlike 5-azacytidine, which incorporates into both DNA and RNA [ 182 ]. Both compounds have been approved by the US FDA for the treatment of myelodysplastic syndrome. Zebularine, a recently developed cytosine analog-based DNMTI, has a very stable chemical property suitable for oral administration, is less toxic and has high selectivity for tumor cells than the first DNMTIs [ 187 ]. Zebularine forms a covalent complex with DNMTs (e.g., Hha I) [ 188 ] in order to deplete them (Dnmt1) or cause partial depletion (Dnmt3a and Dnmt3b) and, importantly, it has been demonstrated to reactivate hypermethylated genes in yeast models and p16INK4a in bladder cancer cells through this mechanism [ 189 , 190 ]. Decitabine has been demonstrated to exert its effect in several cancer cell lines, including in the ovaries and was also shown to restore the expression of several tumor suppressors, such as hMLH1 [ 191 ]. Plumb et al. demonstrated that treatment of the ovarian drug-resistant cell line A2780/CP with decitabine could restore hMLH activity and cisplatin sensitivity, in both cultured cells and mouse xenografts [ 191 ]. It has also been demonstrated to reverse lysine methylation at K9 of histone H3 [ 192 ], which represents another epigenetic silencing mechanism [ 193 , 194 ]. With respect to miRNA gene regulation, a group of six miRNAs clustered on chromosome 19 and seven clustered on chromosome 14, were upregulated by the DNMTI decitabine, demonstrating that miRNAs can be regulated by DNA methylation [ 195 ]. Zebularine was able to demethylate and reactivate a silenced p16 gene in vitro and in vivo [ 189 ]. However, high levels of the drug were required to achieve efficacy of zebularine and, thus, affect its potential application in a clinical setting. Other cytidine analogs, such as arabinosyl-5-aza-cytosine (fazarabine) and dihydro-5-aza-cytidine have proved disappointing in clinical trials [ 186 ]. In addition, non-nucleotide DNMTIs have recently been developed to avoid the inherent toxicity of nucleotide analogs, even though they have the same mechanism of action of binding cytosines and interfering with DNMTs. These agents include procainamide (antiarrhythmic), procaine (anesthetic), hydralazine (antihypertensive) and epigallocathechin-3-gallate derived from green tea, and the novel compound RG108 [ 196 – 199 ]. The therapeutic agents that specifically target one type of DNMT have also been developed, for example MG98, an antisense oligonucleotide that specifically inhibits DNMT1 function [ 200 ]. From these non-nucleotide DNMTIs, hydralazine is currently being evaluated in a randomized, double-blind Phase III clinical trial in cisplatin-resistant recurrent ovarian cancer. Table 4 shows the ongoing clinical trials of epigenetic therapies in ovarian cancer. As mentioned previously, DNA-associated histone proteins are subject to modifications that modulate chromatin organization into a permissive or repressive form. Histone deacetylation is one of the modifications that is correlated with a repressive chromatin and transcriptional silencing. For this reason, HDACIs were developed to relieve gene repression and exert anticancer activity [ 184 ]. These inhibitors can upregulate specific genes, such as p21 cip1/waf1 , a p53 -induced cyclin-dependent kinase inhibitor that causes G1 cell cycle arrest [ 201 ], and apoptotic genes, such as CD95 , TRAIL , DR4 , DR5 , Bax , Bak , Bim , Bmf and Apaf1 , involved in the extrinsic death-receptor and intrinsic mitochondrial death pathways [ 181 ]. In addition, treatment with HDACIs has been found to downregulate genes required for cell cycle progression ( cyclin D1 and cyclin A) , antiapoptosis ( Bcl-2 ) and angiogenesis ( VEGF and HIF-1 α) [ 183 ] in cancer cells and xenograft models [ 181 – 184 , 202 ], with evidence that these antitumor effects involve both transcriptional and nontranscriptional mechanisms [ 181 , 183 ]. The induced hyperacetylation by HDACIs of histones and nonhistone transcription factors, such as p53 , p73 , E2F1 , STAT1 , STAT3 and NF- κ B , activates or represses their target genes [ 181 , 203 – 206 ]. More importantly, cancer cells are more sensitive to growth inhibition by HDACIs compared with nontransformed cells, suggesting that HDACIs have tumor-specific properties [ 207 ]. As mentioned previously, nontranscriptional mechanisms serve as mediators of the antitumor effects of HDACIs [ 181 , 183 , 184 ]. For example, HDACIs could induce defective mitosis in tumor cells and, in turn, trigger cell death through changes in chromatin conformation caused by hyperacetylation of centromeric histones [ 208 ]. Furthermore, HDACIs deplete protein levels of many oncoproteins, whose stability is regulated by heat-shock proteins [ 181 , 183 ], and enhance acetylation of tubulin, increasing the effects of microtubule-stabilizing drugs such as paclitaxel [ 209 ]. This range of antitumor effects, make HDACIs a very attractive and potentially effective antineoplastic alternative. Different HDACIs have been developed: trichostatin A and butyric acid have been used in numerous studies but showed limited clinical activity, and the high cytotoxicity of trichostatin A has limited its use in the clinic [ 210 , 211 ]. A member of the cyclic peptides, depsipeptide, demonstrated clinical efficacy in APL and human lymphoma xenograft models; however, it was not tested in ovarian cancer [ 212 – 215 ]. Vorinostat (suberoylanilide hydroxamic acid), a hydroxamate-based HDACI, is an oral drug that has demonstrated excellent bioavailability in Phase I trials but has major toxicities including anorexia, dehydration, diarrhea and fatigue [ 216 , 217 ]. In a Phase II trial of vorinostat as a single agent in patients with recurrent ovarian cancer, only one of 27 patients experienced a partial response [ 218 ], making it ineffective as a single-agent therapy. Vorinostat has been the only HDACI approved by the FDA for treatment of cutaneous T-cell lymphoma [ 219 ]. A recent hydroxamic acid HDACI, belinostat (PDX101), has revealed potent anti-proliferative and HDAC inhibitory activities in vitro and in xenograft ovarian and colorectal cancer models [ 220 ]. Authors of a preclinical ovarian cancer study, in which belinostat resensitized platinum-resistant xenografts in mice [ 221 ], and a Phase I trial, in which belinostat was administered intravenously in patients with advanced solid tumors, reported dose-limiting toxicities including grade 3 fatigue, diarrhea and cardiac arrhythmia, and concluded that the maximum tolerated dose was 1000 mg/m 2 daily for 5 days and disease stabilization was observed in patients with different types of cancer (sarcomas, renal cancer, melanoma and thymoma) [ 222 ]. In another preclinical study, the combination of decitabine with belinostat elicited greater resensitization of platinum-resistant ovarian cancer xenografts than decitabine alone [ 223 ], making this combination a potentially effective approach to use in the clinic. DNA methyltransferase inhibitors and HDACIs have shown promising efficacy against multiple types of cancers, both in the laboratory and in clinical trials [ 181 – 183 ]. Since ovarian tumorigenesis is driven by DNA methylation and chromosomal remodeling, it is reasonable to think that a combination of both DNTMIs and HDACIs could produce a greater effect in reactivating silenced TSGs and, thus, antitumor activity, than a single-agent therapy [ 224 ]. These epigenetic modifications also contribute to the silencing of genes related to chemosensitivity. Therefore, the reversal of these alterations in order to reactivate TSGs could be a potential target for ovarian cancer treatment. An example of this hypothesis is the study by Azar et al. where 5-aza-2-deoxycytidine pretreatment increased cytotoxicity of the topoisomerase inhibitor topotecan in vitro and in vivo [ 225 ]. As mentioned previously, treatment of cisplatin-resistant A2780/CP ovarian carcinoma cells with 5-aza-2-deoxycytidine induced expression of the mismatch repair enzyme hMLH1 and resensitized these cells to cisplatin in a mouse xenograft model [ 191 , 226 ]. Demethylating agents (e.g., decitabine), HDACIs or combinations may allow for the re-expression of silenced tumor suppressors such as hMLH1 and RASSF1A . hMLH1 plays a role in platinum resistance [ 191 , 226 – 228 ] and RASSF1A silencing may contribute to taxol resistance [ 167 ] and, thus, epigenetic re-expression of these genes might allow for resensitization of ovarian tumors to those conventional first-line therapies. Similarly, as several HDACIs enhance tubulin acetylation [ 209 ], these could conceivably augment sensitivity to taxanes. A list of the main DNMTIs and HDACIs is shown in Table 5 . Radiation therapy represents a therapeutic alternative for ovarian cancer treatment. It has been demonstrated that trichostatin A could activate the ataxia telangiectasia-mutated p53 DNA damage signaling pathway, thereby enhancing ionizing-radiation-induced ataxia telangiectasia mutation activation [ 229 ], suggesting that HDACIs may override the DNA damage defense response and facilitate radiation-induced mitotic cell death.

Ovarian cancer-specific genes discovered through the study of their DNA methylation and histone modification profiles, using the diverse technologies that have been developed, have sped up the discovery of new potential biomarkers for the diagnosis, prognosis and prediction of therapy. Through the knowledge and understanding of the ovarian cancer epigenome, it has been possible to develop epigenetic therapies that have had an enormous benefit in the prevention of chemoresistance by sensitizing tumors to conventional chemotherapeutics and abolishing cancer progression by reactivating the expression of TSGs.

The best approach when dealing with ovarian cancer, as with other cancers, is early detection. Methylation profiling could be an important tool to evaluate the applicability of genes as potential biomarkers for cancer diagnosis, prognosis and response to therapy. For a methylation-based diagnostic assay to be reliable (i.e., sensitive and specific), it is imperative to use those potential biomarkers that are found to be hypermethylated in cancer cells/tissues but unmethylated in normal cells/tissues ( Table 2 ). The best studied serum biomarker for ovarian cancer is CA-125, which is elevated in women with advanced disease in 80% of cases, but only in 50–60% of patients with early-stage disease [ 145 ]. The problem with CA-125 is its lack of specificity, especially in premenopausal women, where other conditions can elevate this marker (e.g., endometriosis and adenomyosis) [ 146 ]. As a result, cancer-specific hypermethylated genes are being considered as potential and promising biomarkers for early detection of ovarian cancer. In one study, tumor-specific hypermethylation of at least one of a panel of six TSG promoters, including RASSF1A , BRCA1, APC, CDKN2A and DAPK , could be detected in the serum or plasma of ovarian cancer patients with 100% specificity and 82% sensitivity, including 13 out of 17 cases of stage I disease [ 67 ]. Methylation was observed in only one peritoneal fluid sample from 15 stage IA or -B patients, but 11 out of 15 paired sera were positive for methylation [ 67 ]. Consistent with previous studies, these data indicate that circulating ovarian tumor DNA is more readily accessible in the bloodstream than in the peritoneum [ 147 ]. In another study, DAPK methylation could be detected in the peripheral blood of 14 out of 16 patients with DAPK -methylation-positive primary tumors, with the peripheral blood of ten out of ten being negative when the primary tumor was negative for DAPK methylation [ 148 ]. These studies demonstrate that it is feasible to detect specific methylation markers in the circulation of patients, thus representing a promising new screening method for the detection of early-stage ovarian cancer. All the methylation biomarker studies performed in bodily fluids for screening and prognosis purposes are small and retrospective, and clinical utility cannot be determined until they are tested in larger prospective studies. No single gene in ovarian cancer has been identified as being methylated in more than a relatively small proportion of cancers. Although new genome-wide approaches may aid in discovering such genes, it is likely that a panel of methylated genes will be necessary to detect ovarian cancer with sufficient specificity and sensitivity. A combination of genes that are commonly methylated in cancer and genes that are methylated specifically in ovarian cancer is the most likely methylation signature capable of distinguishing ovarian cancers from other type of cancers and from benign disease. Besides cancer detection, DNA methylation assays might be used for risk evaluation and prognosis of ovarian cancer. Several epigenetically regulated genes have been assessed for their prognostic prediction potential in ovarian cancer. For example, IGFBP-3 hypermethylation was associated with disease progression and death in ovarian cancer, particularly in patients with early-stage disease; methylation was associated with a threefold higher risk of disease progression and a fourfold higher risk of death [ 149 ]. When IGFBP-3 methylation was combined with methylation in the promoter regions of CDKN2A, BRCA1 or MLH1 , the risk of disease progression in patients with at least three methylated genes was increased sevenfold [ 150 ]. Conversely, hypermethylation of 18S and 28S ribosomal DNA is associated with prolonged progression-free survival of ovarian cancer patients [ 151 ]. Hypomethylation of certain chromosomal regions also appears to have prognostic power; patients who demonstrated little or no hypomethylation of Chr1 Sat2 or Chr1 Sat α had a significantly longer relapse-free survival compared with patients with strong hypomethylation of these regions [ 152 ]. Recently, DNA methylation of SFRP1, -2, -4 and 5, SOX1 , PAX1 , and LMX1A was analyzed by MSP in primary tumor samples from 126 patients with ovarian cancer, 75 with a benign tumor, 14 with borderline malignancy and in the serum from 26 patients with ovarian cancer and 20 with a benign tumor [ 153 ]. Six of the seven genes analyzed had higher methylation levels in the ovarian cancer cases than in borderline malignancy or benign tumors. The methylation of SFRP1, SFRP2, SOX1 and LMX1A genes correlated with recurrence and overall survival of ovarian cancer patients [ 153 ]. Combining the data for SFRP1, SFRP2 and SOX1 genes gave a relative risk for recurrence of 3.19 (p = 0.013) in patients with at least one gene methylation, and combining the data for SFRP1, SOX1 and LMX1A gave a relative risk for cancer-related death of 6.09 (p = 0.01) [ 153 ]. Fiegl et al. identified that hypermethylation in HOXA11 (a polycomb group target) is strongly associated with the residual tumor after cytoreductive surgery and is a marker that indicates poor prognosis [ 91 ]. HOXA11 DNA methylation was independently associated with poor outcome (relative risk for death 3.4; 95% CI: 1.2–9.9; p = 0.03) [ 91 ]. Finally, another epigenetically upregulated gene strongly linked to tumor metastasis, is SNCG , also known as BCSG [ 97 ]. None of the studies that correlate hypermethylation with poor prognosis have been confirmed by subsequent studies and they need to be validated in large independent studies. In one study of cervical samples, SOX1 and HOXA11 [ 154 ] were reported to discriminate between high-grade squamous intraepithelial lesions and normal cervical controls, suggesting that these genes are also predictors of disease progression and, hence, poor prognosis. Nonetheless, as mentioned previously, it is imperative to further validate and confirm these studies, as well to analyze previously reported prognosis-related genes from other tumors into ovarian carcinoma samples. As an example, we have recently reported that TIMP3 methylation is related to poor prognosis in bladder cancer [ 155 ]. Although all these studies demonstrated promising methylated genes as prognostic markers, multicentered, blinded and standardized identical methods need to be developed before its clinical use. Even if the biological basis of a given biomarker is not elucidated, it still can be used in a clinical setting, as long as it is well validated in appropriate cases and controls (e.g., for a prognostic marker, good prognosis and poor prognosis). By contrast, it would be helpful for targeted therapy to understand the biological basis to target that molecule and/or pathways by which the molecule (marker) would be exerting its effect. A summary of ovarian cancer-specific methylated genes with their potential clinical correlation are shown in Table 3 . Overall, 80% of patients with ovarian cancer respond to first-line chemotherapy following surgical debulking. Despite the apparent efficacy of this treatment, up to 75% of these patients will relapse within a few years. These recurrent patients may still be chemosensitive but, ultimately, the vast majority of cases succumb to chemoresistance [ 156 ]. Ovarian cancer drug resistance can be intrinsic (tumors that do not respond to first-line chemotherapy and are conferred by the genotype of pretreatment clones) or acquired (similar to intrinsic resistance, except that it is primarily caused by mutations in progeny tumor cells after the initiation of therapy) [ 157 ]. Both intrinsic and acquired mutations can manifest themselves under various temporal conditions, such as pharmacokinetic alterations (resulting in inadequate drug exposures), variations in tumor cell microenvironments (e.g., hypoxia and altered cell–cell interactions) and differential chemosensitivity during various stages of the cell cycle [ 158 ]. One phenomenon common to both intrinsic and acquired resistance is altered gene expression in the drug-resistant tumor, compared with the drug-sensitive tumor [ 157 ]. In general, ovarian tumors have been demonstrated to upregulate a number of genes, including those involved in cell proliferation, DNA repair, angiogenesis and cell migration, which may play roles in drug resistance and also the downregulation of genes associated with cell adhesion, pro-apoptotic, anti-proliferative and DNA mismatch repair proteins [ 157 ]. Epigenetic alterations represent one of the mechanisms for differential expression of genes that correlate with clinical outcome and thus, have an impact on clinical outcome. A summary of epigenetic alterations and correlation with various clinical parameters is given in Table 3 . The net effect of the platinum drugs and taxanes on sensitive cells is cell death, predominantly through the activation of apoptotic pathways. For example, taxanes stabilize tubulin, resulting in defective spindle formation, G2/M arrest and apoptosis, probably by p53-dependent cascades [ 159 ]. Similarly, platinum compounds are incorporated into DNA, inducing inter- and intra-strand platinum adducts [ 160 ]. The mismatch repair system recognizes such adducts and activates the apoptotic program [ 156 ]. Several genes in ovarian cancer, including TSGs and genes involved in apoptotic pathways, are downregulated by epigenetic mechanisms, as mentioned previously. One well-documented example is the gene encoding the DNA mismatch repair enzyme, hMLH1 . Methylation-induced silencing of hMLH1 has been demonstrated in a number of tumors, including ovarian tumors [ 161 , 162 ]. Loss of hMLH1 expression is strongly associated with microsatellite instability [ 163 , 164 ], a tumor marker that has been linked to genetic hypermutability [ 165 ]. In addition, silencing of hMLH1 has been linked with resistance to platinum drugs [ 166 ], as this results in a decrease in the apoptotic response through p53 phosphorylation and subsequent activation of the MAPK pathway [ 160 ]. Another TSG found to be methylated and silenced in ovarian cancer is the gene encoding the Ras homolog RASSF1A [ 66 , 69 ]. RASSF1A has been reported to bind to tubulin and stabilize microtubules [ 167 ], and this protein might assist chemotherapeutics such as paclitaxel in mediating the prevention of spindle assembly. Teodoridis et al. demonstrated that methylation of at least one of three genes involved in DNA repair/drug detoxification, BRCA1 , GSTP1 and MGMT , is associated with improved response to chemotherapy of patients with late-stage epithelial ovarian tumors [ 63 ]. More recently, HSulf-1 expression has been demonstrated to influence response to chemotherapy. Patients with advanced-stage primary epithelial ovarian tumors that express high levels of HSulf-1 demonstrated an increased response rate to chemotherapy compared with patients whose tumors express low or moderate levels of HSulf-1 [ 168 ]. HSulf-1 is often downregulated in ovarian cancer by methylation-associated silencing, and this downregulation leads to the attenuation of cisplatin-induced cytoxicity [ 168 , 169 ]. Recent studies suggest that epigenetic inactivation of genes plays an important role in acquiring chemoresistance at disease relapse. For example, matched cell line models of acquired resistance have shown that chemotherapy can select for common patterns of CpG island methylation in vitro [ 170 ]. There is an increasing volume of evidence from clinical studies that supports this hypothesis. In the study by Wei et al. , patients stratified as having a short progression-free survival with a high degree of CpG island methylation had a worse response to second-line cytotoxic therapies compared with patients with a longer progression-free survival and low CpG island methylation, suggesting that patients with high CpG island methylation acquire resistance to chemotherapy more readily [ 171 ]. Several genes involved in cell proliferation and survival have been found to be upregulated through epigenetic alterations in ovarian cancer. An example of this is FANCF , which is crucial for the activation of the DNA repair complex containing BRCA1 and −2 . Inactivation of FANCF is associated with increased sensitivity to cisplatin in ovarian cancer cells with a defective BRCA2 pathway [ 172 ]. By contrast, demethylation and re-expression of FANCF , which is thought to occur early in tumor progression, is associated with acquisition of cisplatin resistance in ovarian cancer cell lines [ 172 ]. Methylation-controlled J protein (MCJ), which is required to repress the expression of the drug transporter ABCB1 (P-glycoprotein) [ 173 ], was identified as a gene that, when active, sensitized epithelial cells to cisplatin and paclitaxel, the mainstay of chemotherapy for ovarian cancer patients [ 174 ]. MCJ has been found to be methylated and silenced in normal cells, including normal ovarian surface epithelium, which is unusual for a CpG island-associated gene [ 175 ]. Nevertheless, the majority of late-stage ovarian cancers also exhibit MCJ methylation; however, many of these have undergone a partial demethylation of the MCJ gene promoter, with only 17% of cancers maintaining very high (>90%) methylation, which correlates with a poor response to chemotherapy and decreased survival [ 175 , 176 ]. Hence, MCJ methylation may be a useful marker of response to chemotherapy in ovarian cancer. In terms of ovarian cancer chemoresistance and miRNAs, miR-199a, -200a and -214 were found to be upregulated in ovarian tumors, and miR-214 was demonstrated to target the tumor suppressor PTEN and to have an association with resistance to platinum therapy [ 177 , 178 ]. The miRNA let-7i, found to have tumor suppressor activity, was significantly downregulated in platinum-resistant ovarian tumors and when its function was restored, it led to chemosensitivity of ovarian cancer cells, thereby becoming a potential biomarker and therapeutic target candidate [ 179 ]. Other miRNAs have also been correlated with chemotherapy response [ 180 ] demonstrating that miRNAs could represent potential prognostic and diagnostic biomarkers for ovarian cancer. The study of promoter region methylation of miRNAs is an emerging field and DNA-based assays can be develop considering the dysregulation of these molecules in ovarian cancer. In summary, in order to translate the diverse epigenetic biomarkers identified into clinics, several key points need to be considered: ■ Biomarkers should be related to early detection of disease, prognosis and/or therapeutic response; ■ The biomarker discoveries should be reproducible, consistent and supported by various laboratories; ■ Candidates should be validated through resources that could provide suitable specimens and infrastructure, such as the Early Detection Research Network; ■ There must be longitudinal follow-ups in a screening cohort; ■ The process from biomarker discovery to clinical application should follow all required processes, regulations and standards for the biomarkers to become commercially viable. ■ Biomarkers should be related to early detection of disease, prognosis and/or therapeutic response; ■ The biomarker discoveries should be reproducible, consistent and supported by various laboratories; ■ Candidates should be validated through resources that could provide suitable specimens and infrastructure, such as the Early Detection Research Network; ■ There must be longitudinal follow-ups in a screening cohort; ■ The process from biomarker discovery to clinical application should follow all required processes, regulations and standards for the biomarkers to become commercially viable.