Background
Epithelial ovarian cancer is the most lethal gynecologic malignancy [
1]. High-grade serous ovarian cancer (HGSOC), the most abundant histological subtype of ovarian cancer, is generally diagnosed at an advanced stage. Standard care of advanced stage patients includes debulking surgery in combination with platinum-based chemotherapy in an adjuvant or neoadjuvant setting. Unlike many other epithelial cancers, HGSOC is initially hypersensitive to platinum chemotherapy. However, up to 75% of responding patients relapse with platinum-resistant disease, resulting in a 5-year survival rate of below 40% [
2,
3]. Furthermore, if a relapse occurs within 6 months after initial treatment (progression-free survival (PFS) ≤ 6 months), the patient is regarded as ‘platinum resistant’ [
4,
5]. Based on clinicopathological parameters, it is difficult to identify patients who will respond to platinum chemotherapy. As a surrogate indicator for platinum sensitivity, robust biomarkers associated with very short PFS might help identify relapse-prone patients. Instead of undergoing platinum-based chemotherapy, they could be selected for other novel treatment regimes.
HGSOC differs from other malignancies regarding the prevalence of somatic gene mutations. Except for the frequent inactivating mutation of tumor suppressor
TP53 (96%) and mutations of the
BRCA1/2 (20%) from the DNA damage repair pathway, mutations in other genes are rare [
6,
7]. However, progression of HGSOC involves abundant epigenetic alterations, mainly DNA methylation redistribution, which is characterized by global genomic hypomethylation and localized hypermethylation [
6,
8]. Besides the relative stability of DNA methylation, hypermethylation is functionally related to gene expression and can be easily analyzed in body fluids [
9,
10]. Therefore, DNA methylation can be used as a clinical biomarker.
To date, several studies have been conducted to find robust DNA methylation biomarkers for ovarian cancer. Many specific hypermethylated genes have been reported as potentially useful for diagnosis, prognosis, and/or sometimes for chemoresponse [
11‐
13]. However, most of these studies included all histological subtypes of ovarian cancer and were predominantly based on a single candidate gene approach. Only a few studies have described the identification of platinum chemoresponse methylation markers in HGSOC [
14‐
16]. These studies were based on customized or commercially available methylation array-based platforms, and were limited by the number of CpG sites or to genes of specific pathways.
The aim of the present study was to identify putative methylation markers for chemoresponse in HGSOC. We took an unbiased genome-wide approach and determined the methylation status of PFS-based extreme chemoresponder and non-responder HGSOC patients by performing enrichment of methylated DNA using the methyl-CpG binding domain of MeCP2 protein followed by next generation-sequencing (MethylCap-seq). The differentially methylated profile between extreme responders and non-responders was integrated with microarray expression data to identify putative methylation markers for chemoresponse in HGSOC. In addition, our findings were validated in an independent patient cohort of extreme responders and non-responders, which resulted in FZD10, FAM83A, MYO18B, and MKX as candidate chemoresponse markers. In silico validation of candidate genes was performed using publicly available DNA methylation and expression datasets of unselected advanced stage HGSOC patients to assess their predictive value. Finally, we functionally validated FZD10 involvement in platinum sensitivity using in vitro models.
Discussion
Despite increased understanding of the molecular characteristics of ovarian cancers, no validated clinically relevant markers for platinum chemoresponse in ovarian cancer are currently available. In this study, we identified novel epigenetically-regulated chemoresponse markers for extreme HGSOC platinum responder and non-responder patients by genome-wide DNA methylation-enriched sequencing (MethylCap-seq). We discovered that four genes (FZD10, FAM83A, MYO18B, and MKX) were differentially methylated and expressed between extreme responders and non-responders. In silico analysis on publicly available DNA methylation and expression datasets of unselected advanced stage HGSOC patients showed that DNA methylation of FZD10 and MKX was independently prognostic for improved chemoresponse, as reflected by PFS. In accordance with high FZD10 methylation, low FZD10 expression was associated with a better chemotherapy response and overall survival. Functional analyses of FZD10 established its clear role in cisplatin sensitivity and migration of ovarian cancer cells.
Previously, the identification of epigenetic platinum chemoresponse markers in HGSOC was performed on customized or commercially available methylation array-based platforms with a limited number of CpG probes [
14‐
16]. In the current study, the overall genome-wide DNA methylation profile information was obtained using MethylCap-seq. A recent study has shown that MethylCap-seq technology is a promising unbiased approach for genome-wide DNA methylation profiling that outperforms other methylated DNA capturing techniques [
39]. Furthermore
, MethylCap-seq has comparable coverage of CpG sites at promoter region and CpG islands to whole-genome bisulfite sequencing [
40]. Moreover, MethylCap-seq has been shown to be sensitive in various cancer types, including head and neck, non-small cell lung cancer and cervical cancer [
24,
41‐
44]. Thus far, only one study reported a comprehensive analysis on a large ovarian cancer patient cohort (
n = 101; 75 malignant, 20 benign and 6 normal) using MethylCap-seq [
45]. The DMRs of malignant tumors were compared to benign or normal samples. However, platinum chemotherapy response was not included in the analysis.
By combining the genome-wide methylation and expression data of HGSOC patients and subsequent validations, we identified four novel epigenetically regulated candidate genes (
FZD10,
FAM83A,
MYO18B, and
MKX) that were differentially methylated between extreme responders and non-responders. In silico analysis of unselected advanced stage HGSOC patients showed that DNA methylation of
FZD10 and
MKX was independently associated with a better chemoresponse. Because
FZD10 was the only gene showing both methylation and expression to have prognostic value for the response to platinum-based chemotherapy, this study focused further on
FZD10 for functional validation. However, it is possible that the other genes also play a role in platinum chemoresponse in HGSOC. FAM83A, also known as BJ-TSA-9, is highly expressed in lung cancer [
46] and is highly amplified in many cancer types including breast, ovarian, lung, liver, prostate, and pancreas [
47]. Recently, FAM83A has been found to be a key mediator of resistance to many EGFR tyrosine-kinase inhibitors in breast cancer by causing phosphorylation of c-RAF and PI3K p85, thus promoting proliferation of and invasion by breast cancer cells [
48].
MYO18B has been reported to be hypermethylated in ovarian cancer and important for carcinogenesis [
11]. MKX (IRXL1) is known for its role in muscle development [
49]; recently, it has been identified as an epigenetically regulated gene by microRNA 662 in ovarian cancer [
50], but its role in ovarian cancer is unknown. Interestingly, we previously identified
MKX hypermethylation as an early detection biomarker for cervical cancer [
24]. None of these four genes has been associated with chemo-resistance or sensitivity in HGSOC, indicating that all four might be novel chemoresponse markers for platinum-based chemotherapy.
FZD10 is a member of the Frizzled family of seven-transmembrane WNT signaling receptors [
51]. FZD10 overexpression has been reported in primary cancers such as colon, sarcomas, endometrial, gliomas, and ovarian cancer [
35‐
38,
46,
51] (Additional file
3: Figure S8). FZD10 is assumed to play a role in invasion and metastasis via either the canonical (in colon, endometrial, and breast cancer) or non-canonical WNT pathway (in sarcomas) in a cancer type-dependent manner [
36,
38,
52,
53]. In the present study, we showed that downregulation of FZD10 causes a less migratory phenotype in ovarian cancer cell lines. Moreover, using a FZD10 silencing approach, we showed that FZD10 expression is not only involved in promoting migration, but also causally related to cisplatin resistance of ovarian cancer cells. In agreement with these in vitro results, we found that high
FZD10 expressing HGSOC tumors were worse responders to platinum-based chemotherapy. In a study on ovarian vascular markers, Buckanovich et al. [
54] showed that low expression of
FZD10 in ovarian cancer is significantly associated (
P = 0.001) with better prognosis, which is in line with our findings of significantly high
FZD10 methylation and low
FZD10 expression in the responder patient group in comparison to non-responders. In addition, our previously published study [
18] on global gene expression analysis of HGSOC patients (
n = 156) also showed that high
FZD10 expression was associated with poor overall survival (HR 1.57,
P = 0.0086). Since
FZD10 expression is absent or hardly detectable in any normal organs except placenta [
55] and highly expressed in ovarian cancer (Additional file
3: Figure S8), our results indicate that
FZD10 is an interesting therapeutic target in ovarian cancer. Furthermore, considering the expression of
FZD10 in other tumor types (Additional file
3: Figure S8),
FZD10 may play a role in other tumor types like uterine corpus endometrial cancer and cervical cancer, which are treated with platinum-based chemotherapy often in combination with radiotherapy. Notably, FZD10 has been shown to be a therapeutic target in synovial sarcomas; these sarcomas displayed attenuated growth when targeted by a polyclonal FZD10 antibody [
52]. In addition, a radiation-labeled humanized monoclonal antibody against FZD10 (
OTSA101) has been recently developed, and is currently in phase I clinical trials for synovial sarcoma [
56]. This approach might also be interesting in the context of chemoresistant ovarian cancer.
Although HGSOC is known for bearing mutations in a limited number of genes, aberrant DNA methylation has been found, which might have an effect on platinum-based chemotherapy response [
19,
45,
57]. In addition to the four novel epigenetically regulated genes, we also found other known genes that have been reported for chemoresponse in ovarian cancer or other cancer types. For instance, Survivin (
BIRC5) was among the top 45 gene list from our analysis and has been reported to be involved in platinum sensitivity in HGSOC [
58]. Another gene from our analysis,
GLI3 (a gene of Hedgehog signaling) has been mentioned as being epigenetically regulated and linked with platinum response in HGSOC [
45]. However,
GLI3 could only be verified with pyrosequencing but failed during further validation in our study (Table
1). Previous reports described several hypermethylated genes that we also found in our initial MethylCap-seq analysis list (4541 DMRs) (Fig.
1a). For instance,
BRCA1 hypermethylation was found to be positively associated with chemosensitivity [
6,
19,
59]. Furthermore, hypermethylation of other DNA damage repair pathway-related genes, like
GSTP1,
FANCF, and
MGMT, has been described to be positively associated with chemosensitivity in ovarian cancer patients [
13,
60]. Hypermethylation of genes like
ASS1,
MLH1, and
MSX1, and WNT pathway-related genes including
DVL1,
NFATC3, and
SFRP5 was related to poor outcome of ovarian cancer patients treated with platinum-based chemotherapy [
13,
14,
61,
62]. These genes were omitted from the gene list, since we only included genes that were significantly differentially methylated as well as expressed between responders and non-responders.
Acknowledgements
Not applicable.