Genome-wide methylation profiling of ovarian cancer patient-derived xenografts treated with the demethylating agent decitabine identifies novel epigenetically regulated genes and pathways
© The Author(s). 2016
Received: 13 May 2016
Accepted: 4 October 2016
Published: 20 October 2016
In high-grade serous ovarian cancer (HGSOC), intrinsic and/or acquired resistance against platinum-containing chemotherapy is a major obstacle for successful treatment. A low frequency of somatic mutations but frequent epigenetic alterations, including DNA methylation in HGSOC tumors, presents the cancer epigenome as a relevant target for innovative therapy. Patient-derived xenografts (PDXs) supposedly are good preclinical models for identifying novel drug targets. However, the representativeness of global methylation status of HGSOC PDXs compared to their original tumors has not been evaluated so far. Aims of this study were to explore how representative HGSOC PDXs are of their corresponding patient tumor methylome and to evaluate the effect of epigenetic therapy and cisplatin on putative epigenetically regulated genes and their related pathways in PDXs.
Genome-wide analysis of the DNA methylome of HGSOC patients with their corresponding PDXs, from different generations, was performed using Infinium 450 K methylation arrays. Furthermore, we analyzed global methylome changes after treatment of HGSOC PDXs with the FDA approved demethylating agent decitabine and cisplatin. Findings were validated by bisulfite pyrosequencing with subsequent pathway analysis. Publicly available datasets comprising HGSOC patients were used to analyze the prognostic value of the identified genes.
Only 0.6–1.0 % of all analyzed CpGs (388,696 CpGs) changed significantly (p < 0.01) during propagation, showing that HGSOC PDXs were epigenetically stable. Treatment of F3 PDXs with decitabine caused a significant reduction in methylation in 10.6 % of CpG sites in comparison to untreated PDXs (p < 0.01, false discovery rate <10 %). Cisplatin treatment had a marginal effect on the PDX methylome. Pathway analysis of decitabine-treated PDX tumors revealed several putative epigenetically regulated pathways (e.g., the Src family kinase pathway). In particular, the C-terminal Src kinase (CSK) gene was successfully validated for epigenetic regulation in different PDX models and ovarian cancer cell lines. Low CSK methylation and high CSK expression were both significantly associated (p < 0.05) with improved progression-free survival and overall survival in HGSOC patients.
HGSOC PDXs resemble the global epigenome of patients over many generations and can be modulated by epigenetic drugs. Novel epigenetically regulated genes such as CSK and related pathways were identified in HGSOC. Our observations encourage future application of PDXs for cancer epigenome studies.
Ovarian cancer is the fifth most common cancer type in women and is the most lethal gynecologic malignancy . The most abundant histological subtype of ovarian cancer, high-grade serous ovarian cancer (HGSOC), is characterized by mutations in a few genes, mainly TP53 and BRCA1/2 . Therefore, changes in the epigenome, like DNA methylation and histone modifications, may play an important role in the biological behavior of the disease. Aberrant DNA methylation patterns are universally observed in HGSOC and are known to frequently affect gene regulation involved in cancer-related processes [2–5]. Since epigenetic modifications, including DNA methylation, are reversible in nature, these epigenetic alterations have emerged as attractive targets for epigenetic therapy for cancer [6, 7].
Effective treatment of cancer relies on the identification of key molecular targets of cancer growth and subsequent development of therapeutic agents against these targets. This in turn mainly depends on preclinical research and predictive model systems. Recent genomic analyses have shown that most commonly used HGSOC cell lines, like SKOV3 and A2780, are less representative models of HGSOC [8, 9]. Recently, patient-derived xenografts (PDXs), i.e., patient tumor tissues transplanted directly into immune-deficient mice, have appeared as better representative preclinical models . They recapitulate the histological type and maintain the genomic features and reminiscent heterogeneity of corresponding patients’ primary tumors [11–13]. Furthermore, results from treatment of ovarian cancer PDXs have a good predictive value for standard platinum-based chemotherapy and novel therapeutic agents [14–16]. Although several comparative gene expression and mutational studies have been performed for HGSOC PDXs, comparable studies on the epigenome are not available. Until now, only a few small studies in other tumor types have compared genome-wide DNA methylation of PDXs with their corresponding solid patient tumors [17–19].
In the present study, we first compared genome-wide DNA methylation patterns in different generations of HGSOC PDX tumors and their corresponding primary tumors using Infinium 450 K methylation arrays. Furthermore, we analyzed global methylome changes after treatment of HGSOC PDXs with decitabine (DAC), a DNA demethylating agent, and cisplatin, as platinum-containing chemotherapy is standard of care in first-line treatment of HGSOC. The findings were validated and pathway analysis was performed.
PDX establishment and treatment
PDXs were established as described previously . Briefly, after patients gave informed consent, HGSOC specimens were obtained at primary debulking surgery (patient 36 and -37) or at interval surgery (patient- 56). The clinicopathological features of each patient are provided in Additional file 1: Figure S1a. Tumor fragments were cut into pieces of 3 × 3 × 3 mm3 and implanted in 6–12-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ mice (NSG mice, internal breed, Central Animal Facility, University Medical Center Groningen). Periodic two-dimensional tumor measurement was carried out using a slide vernier caliper and when the tumor size reached >1 cm3, tumors were harvested and were either directly propagated into a further generation or snap frozen in liquid nitrogen for storage along with a piece for formalin fixation. To investigate global DNA methylation changes related to establishment of PDX models from primary HGSOC, we implanted primary tumors of three different HGSOC patients (patients 36, 37, and 56) into the flanks of NSG mice (PDX-36, -37 and -56) and tumors were propagated for up to three generations (F1, F2, and F3) (Additional file 1: Figure S1b). The histology of primary tumors and PDX tumors was analyzed by an experienced gynecologic pathologist.
Mice with F3 PDX tumors were used for treatment. When tumor size reached up to 200 mm3 in size, they were treated with either saline vehicle (n = 3), demethylating agent DAC (n = 3, 2.5 mg/kg three times/week), or cisplatin (n = 3, 4 mg/kg/week) for up to 4 weeks (Additional file 2: Figure S2a). During treatment, mice were regularly checked for welfare and tumor growth (three times a week). After completion of treatment, tumors were harvested and excised into two pieces, one of which was fixed in formalin and the other snap frozen in liquid nitrogen.
Cell line culturing
The ovarian cancer cell lines CaOV3, SKOV3, OVCAR3, PEA1, PEA2, PEO14, PEO23, A2780, C30, Cp70, and IGROV1 were used for in vitro validation. The media used and culture conditions of cell lines are described in Additional file 3: Table S1. All cells were grown at 37 °C in a humidified atmosphere with 5 % CO2 and were detached with 0.05 % trypsin in phosphate-buffered saline (PBS; 0.14 mM NaCl, 2.7 mM KCl, 6.4 mM Na2HPO4, 1.5 mM KH2PO4, pH = 7.4). The authenticity of all cell lines was verified by DNA short tandem repeat analysis (Baseclear, Leiden, The Netherlands). Cells at 40–50 % confluency were treated with DAC (1 μM) for 72 h and the medium was replenished with DAC every day. For cisplatin and carboplatin, cells were treated for 72 h without any daily media replenishment. After 72 h, cells were trypsinized and processed for RNA and DNA isolation.
DNA extraction and bisulfite modification
For DNA isolation, representative frozen blocks of each sample or cells were retrieved. Frozen sections of 10 μm were cut with periodic 4 μm sections for hematoxylin and eosin staining to evaluate the vital tumor cell percentage. DNA of all samples was isolated using standard salt-chloroform extraction and isopropanol precipitation. Precipitated DNA was resuspended in Tris-EDTA buffer (10 mM Tris, 1 mM EDTA, pH = 8.0). Genomic DNA was amplified in a multiplex PCR according to the BIOMED-2 protocol to check the structural integrity of the DNA. DNA concentrations at A260 were measured using the Nanodrop ND-1000 Spectrophotometer (Thermo Scientific, Waltham, MA, USA). A260/280 ratio of >1.8 was required for all samples. Subsequently, bisulfite conversion of all samples was done as described before  using an EZ DNA methylation™ kit (Zymo Research, Orange, CA, USA) as per the manufacturer’s protocol using 1 μg of DNA.
Genome-wide methylation Infinium 450 K array
To analyze the methylation status, the Infinium HumanMethylation450K (HM450K) platform consisting of 485,512 CpG sites was used. The assay was carried out as described . In brief, 4 μl of bisulfite-converted DNA (~150 ng) was used in the whole-genome amplification reaction. After amplification, DNA was fragmented enzymatically, precipitated, and re-suspended in hybridization buffer. All subsequent steps were performed following the standard Infinium protocol (User Guide part #15019519 A). Fragmented DNA was dispensed onto the HumanMethylation450 BeadChips and hybridization was performed in a hybridization oven for 20 h. After hybridization, the array was processed through a primer extension and an immunohistochemistry staining protocol to allow detection of a single-base extension reaction. Finally, BeadChips were coated and then imaged on an Illumina iScan. Methylation levels were computed from raw iDAT files using R (http://www.R-project.org) with different R packages, including MinFi  and ChAMP .
HM450K data processing
Besides PDX tumors, we also used SKOV3 cells treated with a high dose of DAC (1 μM) for 72 h as described in the “Cell line culturing” section. Since SKOV3 is one of the most DAC-sensitive ovarian cancer cell lines, we used it as a positive control for DAC-induced demethylation effects. The ultimate goal was to use these data as a filter to screen the DAC-mediated demethylation-sensitive genes for further in vitro validation. Results of genome-wide methylation of SKOV3 were also processed in a similar way as for PDX tumors. For annotation of probe region, we used UCSC-based annotations in the context of genomic compartment and CpG islands. Further, an additional biologically relevant probe annotation was applied based on CpG enrichment, known as “HIL” CpG classes, consisting of high-density CpG island (HC), intermediate-density CpG island (IC), and non-island (LC).
Bisulfite pyrosequencing was performed as described previously . Briefly, bisulfite-treated DNA was amplified using a PyroMark PCR kit (Qiagen, Hilden, Germany). PCR and cycling conditions were according to the kit manual. All pyrosequencing primers (PCR primers and sequencing primers) were based on the selected candidate 450 K array CpG probe using PyroMark Assay Design software (Qiagen). The amplification protocol was performed according to Collela et al.  using a universal primer approach. The biotinylated PCR products were captured using 1.0 μl streptavidin-coated sepharose high-performance beads (GE Healthcare, Little Chalfont, UK). The immobilized products were washed with 70 % alcohol, denatured with PyroMark denaturation solution (Qiagen), and then washed with PyroMark wash buffer (Qiagen). The purified PCR product was then added to 25 μl PyroMark annealing buffer (Qiagen) containing 0.3 μM sequencing primers for specific genes (all primers and their sequences are available on request). Finally, pyrosequencing was performed using the Pyromark Q24 MD system (Qiagen) according to the manufacturer’s instructions using the PyroGold Q24™ Reagent kit (Qiagen). Data were analyzed and quantified with the PyroMark Q24 software version 2.0.6 (Qiagen).
Total RNA isolation, cDNA synthesis and quantitative RT-PCR
Quantitative reverse transcriptase (qRT)-PCR was performed as described previously . Total RNA was isolated from frozen tissue blocks and cell lines similarly to as described for DNA extraction. RNA was isolated using a RNeasy mini kit (Qiagen) according to the instructions of the manufacturer. RNA was analyzed quantitatively using a Nanodrop and integrity was checked using electrophoresis on agarose gel. Total RNA (1 μg) was used for cDNA synthesis by RNase H+ reverse transcriptase using an iScript cDNA synthesis kit (BioRad, Hercules, CA, USA) as per the manufacturer’s instructions. qRT-PCR was performed in an ABI PRISM 7900HT Sequence Detector (Applied Biosystems, Foster City, CA, USA) with the iTaq SYBR Green Supermix with Rox dye (Biorad, Hercules, CA, USA). Amplification was performed with the following cycling conditions: 5 min at 95 °C, and 40 two-step cycles of 15 s at 95 °C and 25 s at 60 °C. The reactions were analyzed by SDS software (version 2.4, Applied Biosystems). The threshold cycles (Ct) were calculated and relative gene expression was analyzed after normalizing for GAPDH, a house-keeping gene. qRT-PCR primer sequences are available on request.
After performing probe filtering, normalization and batch effect correction, we identified the differentially methylated CpG sites using Linear Models for Microarray Data (LIMMA) analysis . Since for beta-distributed data like DNA methylation β values the variance is associated with the mean (heteroscedasticity) , we cannot apply linear model-based methods without transforming the data properly (logit transformed). Therefore, normalized 450 K probe β values were converted to M values using the beta2m function . The unpaired statistical analysis was performed using the eBayes function of the Limma package . The average DNA methylation of bisulfite pyrosequencing and RNA expression levels were presented as mean ± standard deviation (SD) using the GraphPad Prism version 6.04 (GraphPad for Science, San Diego, CA, USA). Statistical significance was calculated by two-way Student’s t-test and multiple comparisons between different groups were performed by one-way ANOVA with Bonferroni post-test, unless otherwise mentioned in the respective figure legends. For selection of differentially methylated CpG sites the cutoff was p < 0.01, while other analyses are described in the respective figure legends with appropriate symbolic representation. As a positive control for DAC-induced genome-wide demethylation, SKOV3 cells showed a higher percentage (39.3 %) of CpG sites being demethylated (Additional file 2: Figure S2b, e) in comparison with DAC-treated PDX-36. These DAC-sensitive CpG sites from SKOV3 cells were also used for identification of epigenetically regulated genes and pathways for in vitro validation.
Principal component analysis was performed on BMIQ normalized data. Pre-processed, filtered, and normalized autosomal CpG probes were used for unsupervised clustering of Illumina 450 K data. Different clustering algorithms and number of clusters were investigated extensively, including k-means and hierarchical clustering approaches using average linkage methodology. Further, supervised clustering analysis was performed on significant probes after LIMMA analysis on treatment groups using hierarchical clustering with the average linkage method.
Gene ontology analysis
Functional gene ontology (GO) term enrichment analysis was performed with the DAVID tool  using DAC-sensitive genes (n = 822) on Homo sapiens as species background. We restricted the analysis to the biological process category and selected GO terms with enrichment (p ≤ 0.01). Data visualization was carried out using REVIGO (http://revigo.irb.hr/index.jsp) .
Web-based tools for networks and pathway analysis
WebGestalt (WEB-based GEne SeT AnaLysis Toolkit)  was used as the web-based tool for prediction of associated pathway and gene function using the list of DAC-sensitive genes in PDX tumors (n = 822). Parameters used for analysis were: organism, H. sapiens; ID type, gene_symbol; reference set, Entrez gene; significance level, 0.001; statistics test, hypergeometric; multiple testing corrections, Bonferroni Hedgehog test; minimum number of genes for enrichment, 3. Pathway analyses were performed using KEGG, Wiki pathways, and pathways from common databases. Genes related to pathways found in at least two of the databases were included for the final networks using the Gene Multiple Association Network Integration Algorithm (GeneMANIA; http://www.genemania.org/). This analysis builds a gene integration network incorporating physical and predicted interactions, co-localization, shared pathways, and shared protein domains.
Prognostic evaluation of CSK methylation and expression on clinical data
Methylation data of the AOCS study group (http://www.aocstudy.org) was downloaded from the NCBI GEO portal using GEO accession GSE65820 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc =GSE65820) as mentioned in Patch et al. . The clinical data of patients was downloaded from the ICGC data portal (https://dcc.icgc.org/). Data were normalized using a BMIQ normalization as described previously . The CSK methylation probe (cg00516515) identified in the PDX methylation analysis was used for further analysis. The methylation cutoff between low and high methylation was set at 0.9 based on the median β value (0.90, range 0.78–0.96). This resulted in 89 patients (31 high and 58 low methylation) for progression-free survival (PFS) analysis (a proxy for sensitivity to platinum-containing chemotherapy) and 91 patients (32 high and 59 low methylation) for overall survival (OS) analysis using the Cox proportional hazard model.
Prognostic validation of CSK expression level was performed on publicly available datasets obtained from an online tool  for genome-wide validation that can be accessed at http://kmplot.com/ovar. This online portal only contains data from publications that comprise normalized microarray gene expression data, clinical survival information, and at least 20 patients. For our prognostic analysis, data were derived from analysis using KM plotter  in October 2015, in which we selected only advanced stage (III and IV) HGSOC cancer patients who received platinum therapy. This resulted in 633 patients for PFS analysis and 656 patients for OS analysis using a Cox proportional hazard model with CSK probe (probe ID 202329_at). With an expression range of CSK probe (74–2566), the auto cutoff value of 567 for PFS analysis and 580 for OS analysis was used, based on the computation of upper and lower quartiles with default settings of the portal .
Genome-wide DNA methylation comparison of HGSOC primary and PDX tumors
Based on global DNA methylation patterns, all PDX tumors were clustered together with their respective PDX type (PDX-36, -37, and -56), irrespective of their propagated generation (F1, F2, or F3) (Fig. 1e). Notably, unsupervised clustering revealed that the methylation patterns of primary tumors from patients 36 and 37 were more similar to each other than their corresponding PDX tumors as shown by the close hierarchical clustering between these two tumors (Fig. 1e). The reason for such clustering could be the fact that primary patient tumors include human stromal and endothelial cells as well.
After analyzing the number of differentially methylated CpG sites among primary tumors and PDX tumors from F1 to F3, we found only 2604 CpG sites in F1, 4349 sites in F2, and 4606 sites in F3 that were significantly differentially methylated (p < 0.01) in comparison with the F0 primary tumors. These results indicate that only 0.66–1.17 % of the 392,317 CpG sites were differentially methylated in primary versus PDX tumors (Fig. 1f). Moreover, a very low number of CpG sites (0.001–0.002 % of total CpG sites analyzed) was significantly differentially methylated (p < 0.01) among different generations of PDX tumors (F1 versus F2 or F2 versus F3 tumors) (Fig. 1f). Finally, global methylation patterns of all patient tumors and PDXs were verified by bisulfite pyrosequencing of the global methylation marker ALU-Yb8 (Additional file 1: Figure S1f), showing similar genome-wide methylation patterns between F0 and F3. In addition, the global methylation patterns of biological replicates of PDX-36 tumors from generation F3 (n = 3) were compared to each other and found to be highly correlated to each other (r = 0.94–0.96, p < 0.001) (Additional file 1: Figure S1g). In conclusion, these results indicate that genome-wide methylation between PDX tumors and their corresponding primary patient tumors were very similar, with only some small changes found in F1 tumors in specific CpG- enriched regions.
Effect of treatment with demethylating agent DAC or cisplatin on the global DNA methylome of PDX tumors
No major demethylation effect in genome-wide CpG probes (mean β value) of cisplatin-treated PDX 36 tumors (cisplatin = 0.327) was observed compared to vehicle-treated PDX 36 tumors (control = 0.342) (Fig. 2a; Additional file 2: Figure S2b). Furthermore, there was no significant difference in mean DNA methylation between the probes of cisplatin-treated and vehicle-treated PDX tumors at any genomic location irrespective of CpG context and content (Fig. 2b–d; Additional file 2: Figure S2c). Bisulfite pyrosequencing of global DNA methylation surrogate marker LINE-1 and ALU Yb8 in PDX tumors confirmed our findings (Additional file 4: Figure S3a, b). Furthermore, no significant differences were observed for methylation of LINE-1 and ALU Yb8 in ovarian cancer cell lines when treated with either cisplatin or carboplatin compared to untreated controls (Additional file 4: Figure S3c, d). Notably, unsupervised cluster analysis of all CpG sites showed that PDX tumors clustered together dependent on the treatment used (Fig. 2e). This apparently indicates that DNA methylation patterns are similarly affected per specific therapy.
Methylation analysis at the single CpG probe level revealed approximately 41,491 CpG sites (10.6 % of total CpG sites analyzed) that were significantly differentially methylated (p < 0.01) in DAC-treated PDX tumors compared to control PDX tumors (Fig. 2f; Additional file 2: Figure S2e). Supervised clustering analysis of the significantly (p < 0.01) differentially methylated CpG sites (n = 41,491 sites) showed that the majority of sites (97.6 %) were demethylated in DAC-treated compared to vehicle-treated tumors (Fig. 2g). Interestingly, global DNA demethylation of PDX tumors is comparable to the demethylation effect of DAC as observed in tumor DNA from patients in a recent clinical trial with DAC  (Additional file 2: Figure S2f). In stark contrast, only 0.53 % of total analyzed CpG sites, comprising 2088 sites, were significantly differentially methylated (p < 0.01) in cisplatin-treated PDX tumors compared to vehicle-treated ones (Fig. 2f). Of 2088 CpG sites, 61 % of CpG sites showed hypomethylation and 39 % showed hypermethylation in cisplatin-treated tumors in comparison with vehicle-treated ones (Additional file 2: Figure S2g). In conclusion, these results show a marginal effect of cisplatin but a strong demethylation effect of DAC in clinically relevant PDX models.
Identification of novel epigenetically regulated genes and pathways in PDX tumors
To identify the potential biological function of these 822 genes effectively demethylated by DAC treatment, we first performed GO-based functional enrichment analysis using DAVID . The major biological process-related GO terms were metabolic process, cellular transport, biosynthetic process, mitotic cell cycle, cell locomotion, transferase activity, and post-translational modifications (Fig. 3b). Subsequently, pathway enrichment analysis using KEGG, Wiki pathways, and pathway common databases revealed several enriched pathways, including mTOR pathway, insulin signaling, cellular metabolic pathway, TGF-β signaling, Wnt pathway, cell cycle, Src family kinases signaling, DNA replication, and vesicular trafficking pathways (Fig. 3c; Additional file 7: Table S3). We selected seven genes from different pathways for further validation: CSK (Src family kinase signaling), ADCY6 (metabolic pathway), PRKCζ, AKT1, RAPTOR (insulin and mTOR pathway), SKI (TGF-β signaling), and NFATC1 (T-cell stimulation). Five out of these seven genes were successfully validated by bisulfite pyrosequencing comparing DNA from PDX-36 tumors treated with DAC or vehicle (Fig. 3d).
Validation of C-terminal Src kinase (CSK) as a candidate gene for ovarian cancer treatment
For further validation, a large panel of ovarian cancer cell lines (n = 11) was treated with DAC for three days and the methylation status of CSK was analyzed using bisulfite pyrosequencing. All cell lines showed high CSK methylation levels (72–99 %), which decreased significantly (p < 0.01–0.0001) after DAC treatment (Fig. 4d). Subsequently, we found significant upregulation of CSK expression levels (p < 0.05) in SKOV3, OVCAR3, PEA1, A2780, and IGROV1 cells (Fig. 4e). Moreover, an inverse correlation (r = −0.612, p < 0.0021) between methylation and gene expression of CSK was found in the ovarian cancer cell lines (Fig. 4f). In summary, these results show that CSK is an epigenetically regulated gene with demethylation leading to higher gene expression, both in ovarian cancer PDX models as well as in cell lines.
Our study for the first time shows that HGSOC PDX tumors are epigenetically stable comparing primary tumors with their subsequent PDX generations. Only 0.66–1.17 % of the total methylated CpG sites significantly changed in HGSOC PDX tumors during propagation. While cisplatin treatment did not alter the DNA methylation pattern, treating these PDX models with DAC significantly reduced tumor growth and was accompanied by significant changes in methylation of CpG sites. Further validation and subsequent pathway analysis revealed enrichment of several biological pathways (e.g., the Src family kinase pathway) in HGSOC that were affected by DAC treatment. Expression of CSK, a negative regulator of non-receptor Src family kinases, is epigenetically regulated and can be upregulated by DAC treatment in several HGSOC PDXs and cell lines. Moreover, we show that CSK methylation and expression have prognostic value in HGSOC patients.
There is growing evidence that HGSOC PDX models not only recapitulate the histology of patients’ tumors but also maintain the heterogeneity of them to some extent [12, 13]. However, their utility in epigenomics studies has not been assessed yet. In HGSOC, frequent aberrant epigenomic alterations, including DNA methylation, with less somatic mutations  present DNA methylation as a suitable target for future epigenetic cancer therapy. Finding novel and robust epigenetically regulated genes and pathways warrants suitable preclinical models with better prediction value for therapeutic targets and therapy response. Cell lines and cell line-based xenografts are known to be more homogenous models but with the lack of representative prediction of drug responses . Moreover, continuous propagation of cell lines induces many epigenetic changes and HGSOC cell lines are, therefore, epigenetically far from patient tumors . Until now global DNA methylome analysis has been performed on PDX models of just a few cancer types, including head and neck, small cell lung, and colon cancer and osteosarcoma [17–19]. All previously reported studies were limited by having low numbers of PDX samples and not including propagation-related, trans-generational comparisons. So far, only osteosarcoma and colon PDXs have been used to compare global trans-generational methylation patterns up to the second generation . In line with our observations, the methylome of osteosarcoma and colon cancer PDXs was very similar to the primary tumor with, on average, only 2.7 % difference in the assayed CpG sites. In this study, we used only subcutaneously implanted PDX models, which do not have the advantages of orthotopic implanted models that have the same anatomic microenvironment of patients’ tumors and resemble their metastatic behavior . However, the generation of orthotropic xenografts is more labor-intensive and expensive and complex surgery and imaging methods are required to monitor tumor growth . Therefore, we used subcutaneous implantation methodology, achieving not only high take rates but also PDX tumors that histologically and genomically mimicked the patients’ tumors . Nevertheless, it would be of interest to compare the global methylome of subcutaneously implanted and orthotropic implanted PDX models.
Current knowledge regarding the effect of epigenetic drugs like demethylating agents on DNA methylation patterns in PDX models is obscure. Our study presents the first results of DAC treatment on the methylome of HGSOC PDX tumors. We observed a global demethylation effect of DAC treatment at all CpG sites irrespective of their genomic location, with 10.6 % significantly demethylated CpG sites. We also verified these results by a decrease in DNA methylation of the global methylation markers LINE-1 and ALU Yb8. Similar global demethylation effects (e.g., LINE-1 and ALU Yb8 as well as the total percentage of demethylated CpGs) were also observed in peripheral blood mononuclear cells, ascites, and tumor DNA in platinum-resistant ovarian cancer patients treated with DAC in a phase II clinical trial . Many of the significantly demethylated genes in DAC-treated PDXs in our study were related to cell–cell adhesion, MAPK, mTOR, cytokine- and chemokine-related pathways, cell–matrix adhesion, NFKB, and other related pathways. Most of these pathways were also found to be altered in DAC-treated ovarian cancer patients. However, less similarity was observed at the gene level [43, 44]. Concisely, the effects of DAC on HGSOC PDX models resemble the effects of this agent observed in patients. Hence, HGSOC PDX models can be utilized for analyzing the effects of novel epigenetic cancer therapies.
Notably, we observed DAC-induced tumor growth inhibition in all three HGSOC PDX models. Therefore, we focused on finding putative genes and/or pathways whose demethylation might be responsible for such tumor growth inhibition. Consequently, our search for novel epigenetically regulated key genes and pathways related to DAC treatment in HGSOC PDXs led to the identification of CSK. CSK is known for its role as a negative regulator of non-receptor tyrosine Src family kinases, including c-Src, c-Fgr, Lyn, c-Yes, and others . CSK phosphorylates these kinases, leading to an inactive conformation of kinases and decreased downstream signaling [39, 45]. CSK has been found highly expressed in normal organs, while its reduced expression and concomitant increased c-Src activity were reported in many cancer types, including hepatocellular carcinoma and prostate cancer [46, 47]. We found that high expression as well as low methylation of CSK in advanced stage HGSOC patients was related to better PFS and OS. Moreover, Src and other Src kinases have been reported to be overexpressed in advanced stage ovarian cancer [48, 49]. Emerging data are supporting the key role of Src family kinases in many carcinogenic processes, including tumor growth and metastasis in ovarian and colon cancer [50, 51]. Hence, they are being considered as suitable targets for ovarian cancer treatment in combination with standard chemotherapy [51, 52]. It has, however, been reported that selective inhibition of Src in ovarian cancer could lead to enhanced expression of other Src family kinases and related pathways . CSK overexpression actually causes inhibition of in vivo tumor growth and metastasis in colon cancer cell lines . Therefore, it is tempting to speculate that reversion of epigenetically silenced CSK or induction of CSK expression in ovarian cancer might lead to an adequate suppression of Src family kinases and consequently less tumor growth. Thus, more in-depth functional validation of CSK is warranted to study how this protein is involved in chemoresponses and OS of ovarian cancer patients.
We show that genome-wide DNA methylation in HGSOC PDX models is largely stable during propagation in mice. The methylome of PDX tumors can be efficiently affected with the demethylating agent DAC. Using this model, we have identified novel epigenetically regulated genes, such as CSK, and related pathways. Our results encourage the application of PDXs for further cancer epigenomics studies.
C-terminal Src kinase
High-grade serous ovarian cancer
Highly methylated site
Partially methylated site
Quantitative reverse transcriptase polymerase chain reaction
SdJ and GBAW are members of the EurOPDX Consortium.
This work was supported by the Jan Kornelis de Cock foundation grant 2012, 2014 and the Van Der Meer-Boerema Stichting grant 2014. Further, this work was also supported in part by Dutch Cancer Society (KWF) grants (RUG 2010–4833, RUG 2011–5231, RUG 2012–5477).
Availability of data and materials
The processed 450 K data have been deposited in the Gene Expression Omnibus under accession number GSE81438.
TT, SdJ and GBAW conceived and designed the experiments. TT, NGA, RH, GJM, and HK performed the experiments. TT, SdJ, RH, and GBAW analyzed the data. TT, SdJ, AGJvdZ and GBAW wrote the paper. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Ethics approval and consent to participate
For tumor samples from patients, this study was reviewed by the medical ethics committee of the University Medical Centre Groningen and no approval was needed in compliance with Dutch law. The study conformed to the principles of the Declaration of Helsinki. All patients gave written informed consent. All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Groningen (Groningen, the Netherlands).
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Siegel RL, Miller KD, Jemal A. Cancer statistics, 2015. CA Cancer J Clin. 2015;65:5–29.View ArticlePubMedGoogle Scholar
- Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474:609–15.View ArticleGoogle Scholar
- Barton CA, Hacker NF, Clark SJ, O’Brien PM. DNA methylation changes in ovarian cancer: implications for early diagnosis, prognosis and treatment. Gynecol Oncol. 2008;109:129–39.View ArticlePubMedGoogle Scholar
- Shih I-M, Chen L, Wang CC, Gu J, Davidson B, Cope L, Kurman RJ, Xuan J, Wang T-L. Distinct DNA methylation profiles in ovarian serous neoplasms and their implications in ovarian carcinogenesis. Am J Obstet Gynecol. 2010;203:584.e1–22.View ArticleGoogle Scholar
- Huang R-L, Gu F, Kirma NB, Ruan J, Chen C-L, Wang H-C, Liao Y-P, Chang C-C, Yu M-H, Pilrose JM, Thompson IM, Huang H-C, Huang TH-M, Lai H-C, Nephew KP. Comprehensive methylome analysis of ovarian tumors reveals hedgehog signaling pathway regulators as prognostic DNA methylation biomarkers. Epigenetics. 2013;8:624–34.View ArticlePubMedPubMed CentralGoogle Scholar
- Tsai H-C, Baylin SB. Cancer epigenetics: linking basic biology to clinical medicine. Cell Res. 2011;21:502–17.View ArticlePubMedPubMed CentralGoogle Scholar
- Azad N, Zahnow CA, Rudin CM, Baylin SB. The future of epigenetic therapy in solid tumours--lessons from the past. Nat Rev Clin Oncol. 2013;10:256–66.View ArticlePubMedPubMed CentralGoogle Scholar
- Ince TA, Sousa AD, Jones MA, Harrell JC, Agoston ES, Krohn M, Selfors LM, Liu W, Chen K, Yong M, Buchwald P, Wang B, Hale KS, Cohick E, Sergent P, Witt A, Kozhekbaeva Z, Gao S, Agoston AT, Merritt MA, Foster R, Rueda BR, Crum CP, Brugge JS, Mills GB. Characterization of twenty-five ovarian tumour cell lines that phenocopy primary tumours. Nat Commun. 2015;6:1–14.View ArticleGoogle Scholar
- Beaufort CM, Helmijr JC, Piskorz AM, Hoogstraat M, Ruigrok-Ritstier K, Besselink N, Murtaza M, van IJcken WF, Heine AA, Smid M, Koudijs MJ, Brenton JD, Berns EM, Helleman J. Ovarian cancer cell line panel (OCCP): clinical importance of in vitro morphological subtypes. PLoS One. 2014;9, e103988.View ArticlePubMedPubMed CentralGoogle Scholar
- Hidalgo M, Amant F, Biankin AV, Budinska E, Byrne AT, Caldas C, Clarke RB, de Jong S, Jonkers J, Maelandsmo GM, Roman-Roman S, Seoane J, Trusolino L, Villanueva A. Patient-derived xenograft models: an emerging platform for translational cancer research. Cancer Discov. 2014;4:998–1013.View ArticlePubMedPubMed CentralGoogle Scholar
- Ricci F, Bizzaro F, Cesca M, Guffanti F, Ganzinelli M, Decio A, Ghilardi C, Perego P, Fruscio R, Buda A, Milani R, Ostano P, Chiorino G, Bani MR, Damia G, Giavazzi R. Patient-derived ovarian tumor xenografts recapitulate human clinicopathology and genetic alterations. Cancer Res. 2014;74:6980–90.View ArticlePubMedGoogle Scholar
- Alkema NG, Tomar T, Duiker EW, Jan Meersma G, Klip H, van der Zee AGJ, Wisman GB, de Jong S. Biobanking of patient and patient-derived xenograft ovarian tumour tissue: efficient preservation with low and high fetal calf serum based methods. Sci Rep. 2015;5:14495.View ArticlePubMedPubMed CentralGoogle Scholar
- Colombo P, Manoir S, Orsetti B, Bras R. Ovarian carcinoma patient derived xenografts reproduce their tumor of origin and preserve an oligoclonal structure. Oncotarget. 2015;6:28327–40.View ArticlePubMedPubMed CentralGoogle Scholar
- Topp MD, Hartley L, Cook M, Heong V, Boehm E, McShane L, Pyman J, McNally O, Ananda S, Harrell M, Etemadmoghadam D, Galletta L, Alsop K, Mitchell G, Fox SB, Kerr JB, Hutt KJ, Kaufmann SH, Australian Ovarian Cancer Study, Swisher EM, Bowtell DD, Wakefield MJ, Scott CL. Molecular correlates of platinum response in human high-grade serous ovarian cancer patient-derived xenografts. Mol Oncol. 2014;8:656–88.View ArticlePubMedPubMed CentralGoogle Scholar
- Weroha SJ, Becker MA, Enderica-Gonzalez S, Harrington SC, Oberg AL, Maurer MJ, Perkins SE, AlHilli M, Butler KA, McKinstry S, Fink S, Jenkins RB, Hou X, Kalli KR, Goodman KM, Sarkaria JN, Karlan BY, Kumar A, Kaufmann SH, Hartmann LC, Haluska P. Tumorgrafts as in vivo surrogates for women with ovarian cancer. Clin Cancer Res. 2014;20:1288–97.View ArticlePubMedPubMed CentralGoogle Scholar
- Scott CL, Becker MA, Haluska P, Samimi G. Patient-derived xenograft models to improve targeted therapy in epithelial ovarian cancer treatment. Front Oncol. 2013;3:1–8.View ArticleGoogle Scholar
- Hennessey PT, Ochs MF, Mydlarz WW, Hsueh W, Cope L, Yu W, Califano JA. Promoter methylation in head and neck squamous cell carcinoma cell lines is significantly different than methylation in primary tumors and xenografts. PLoS One. 2011;6, e20584.View ArticlePubMedPubMed CentralGoogle Scholar
- Poirier JT, Gardner EE, Connis N, Moreira AL, de Stanchina E, Hann CL, Rudin CM. DNA methylation in small cell lung cancer defines distinct disease subtypes and correlates with high expression of EZH2. Oncogene. 2015;34:5869–78.View ArticlePubMedPubMed CentralGoogle Scholar
- Guilhamon P, Butcher LM, Presneau N, Wilson GA, Feber A, Paul DS, Schütte M, Haybaeck J, Keilholz U, Hoffman J, Ross MT, Flanagan AM, Beck S. Assessment of patient-derived tumour xenografts (PDXs) as a discovery tool for cancer epigenomics. Genome Med. 2014;6:1–10.View ArticleGoogle Scholar
- Eijsink JJH, Lendvai A, Deregowski V, Klip HG, Verpooten G, Dehaspe L, De Bock GH, Hollema H, Van Criekinge W, Schuuring E, Van Der Zee AGJ, Wisman GBA. A four-gene methylation marker panel as triage test in high-risk human papillomavirus positive patients. Int J Cancer. 2012;130:1861–9.View ArticlePubMedGoogle Scholar
- Sandoval J, Heyn H, Moran S, Serra-Musach J, Pujana MA, Bibikova M, Esteller M. Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome. Epigenetics. 2011;6:692–702.View ArticlePubMedGoogle Scholar
- Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry RA. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;30:1363–9.View ArticlePubMedPubMed CentralGoogle Scholar
- Morris TJ, Butcher LM, Feber A, Teschendorff AE, Chakravarthy AR, Wojdacz TK, Beck S. ChAMP: 450 k chip analysis methylation pipeline. Bioinformatics. 2014;30:428–30.View ArticlePubMedGoogle Scholar
- Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, Beck S. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013;29:189–96.View ArticlePubMedGoogle Scholar
- Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics. 2012;28:882–3.View ArticlePubMedPubMed CentralGoogle Scholar
- Clausen MJ, Melchers LJ, Mastik MF, Slagter-Menkema L, Groen HJ, van der Laan BF, van Criekinge W, de Meyer T, Denil S, Wisman GB, Roodenburg JL, Schuuring E. Identification and validation of WISP1 as an epigenetic regulator of metastasis in oral squamous cell carcinoma. Genes Chromosomes Cancer. 2016;55:45–59.View ArticlePubMedGoogle Scholar
- Colella S, Shen L, Baggerly KA, Issa JPJ, Krahe R. Sensitive and quantitative universal PyrosequencingTM methylation analysis of CpG sites. Biotechniques. 2003;35:146–50.PubMedGoogle Scholar
- Joseph JV, Conroy S, Tomar T, Eggens-Meijer E, Bhat K, Copray S, Walenkamp AME, Boddeke E, Balasubramanyian V, Wagemakers M, den Dunnen WF, Kruyt FA. TGF-β is an inducer of ZEB1-dependent mesenchymaltransdifferentiation in GBM. Cell Death Dis. 2014;5, e1443.View ArticlePubMedPubMed CentralGoogle Scholar
- Smyth GK. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004;3:1–25.Google Scholar
- Du P, Zhang X, Huang C-C, Jafari N, Kibbe WA, Hou L, Lin SM. Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics. 2010;11:1–9.View ArticleGoogle Scholar
- Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43, e47.View ArticlePubMedPubMed CentralGoogle Scholar
- Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J, Stephens R, Baseler MW, Lane HC, Lempicki RA. The DAVID Gene Functional Classification Tool: a novel biological module-centric algorithm to functionally analyze large gene lists. Genome Biol. 2007;8:R183.View ArticlePubMedPubMed CentralGoogle Scholar
- Supek F, Bošnjak M, Škunca N, Šmuc T. Revigo summarizes and visualizes long lists of gene ontology terms. PLoS One. 2011;6, e21800.View ArticlePubMedPubMed CentralGoogle Scholar
- Wang J, Duncan D, Shi Z, Zhang B. WEB-based GEne SeT AnaLysis Toolkit (WebGestalt): update 2013. Nucleic Acids Res. 2013;41:W77–83.View ArticlePubMedPubMed CentralGoogle Scholar
- Patch A-M, Christie EL, Etemadmoghadam D, Garsed DW, George J, Fereday S, Nones K, Cowin P, Alsop K, Bailey PJ, Kassahn KS, Newell F, Quinn MCJ, Kazakoff S, Quek K, Wilhelm-Benartzi C, Curry E, Leong HS, Hamilton A, Mileshkin L, Au-Yeung G, Kennedy C, Hung J, Chiew Y-E, Harnett P, Friedlander M, Quinn M, Pyman J, Cordner S, O’Brien P, et al. Whole–genome characterization of chemoresistant ovarian cancer. Nature. 2015;521:489–94.View ArticlePubMedGoogle Scholar
- Gyorffy B, Lánczky A, Szállási Z. Implementing an online tool for genomewide validation of survival-associated biomarkers in ovarian-cancer using microarray data from 1287 patients. Endocr Relat Cancer. 2012;19:197–208.View ArticlePubMedGoogle Scholar
- Tentler JJ, Tan AC, Weekes CD, Jimeno A, Leong S, Pitts TM, Arcaroli JJ, Messersmith WA, Eckhardt SG. Patient-derived tumour xenografts as models for oncology drug development. Nat Rev Clin Oncol. 2012;9:338–50.View ArticlePubMedPubMed CentralGoogle Scholar
- Price ME, Cotton AM, Lam LL, Farré P, Emberly E, Brown CJ, Robinson WP, Kobor MS. Additional annotation enhances potential for biologically-relevant analysis of the Illumina Infinium HumanMethylation450 BeadChip array. Epigenetics Chromatin. 2013;6:4.View ArticlePubMedPubMed CentralGoogle Scholar
- Okada M. Regulation of the Src Family Kinases by Csk. Int J Biol Sci. 2012;8:1385–97.View ArticlePubMedPubMed CentralGoogle Scholar
- Sen B, Johnson FM. Regulation of Src family kinases in human cancers. J Signal Transduct. 2011;1–14
- Alkema NG, Wisman GBA, van der Zee AGJ, van Vugt MATM, de Jong S. Studying platinum sensitivity and resistance in high-grade serous ovarian cancer: different models for different questions. Drug Resist Updat. 2016;24:55–69.View ArticlePubMedGoogle Scholar
- Houshdaran S, Hawley S, Palmer C, Campan M, Olsen MN, Ventura AP, Knudsen BS, Drescher CW, Urban ND, Brown PO, Laird PW. DNA methylation profiles of ovarian epithelial carcinoma tumors and cell lines. PLoS One. 2010;5, e9359.View ArticlePubMedPubMed CentralGoogle Scholar
- Matei D, Fang F, Shen C, Schilder J, Arnold A, Zeng Y, Berry WA, Huang T, Nephew KP. Epigenetic resensitization to platinum in ovarian cancer. Cancer Res. 2012;72:2197–205.View ArticlePubMedPubMed CentralGoogle Scholar
- Fang F, Zuo Q, Pilrose J, Wang Y, Shen C, Li M, Wulfridge P, Matei D, Nephew KP. Decitabine reactivated pathways in platinum resistant ovarian cancer. Oncotarget. 2014;5:3579–89.View ArticlePubMedPubMed CentralGoogle Scholar
- Chong Y-P, Mulhern TD, Cheng H-C. C-terminal Src kinase (CSK) and CSK-homologous kinase (CHK)—endogenous negative regulators of Src-family protein kinases. Growth Factors. 2005;23:233–44.View ArticlePubMedGoogle Scholar
- Masaki T, Okada M, Tokuda M, Shiratori Y, Hatase O, Shirai M, Nishioka M, Omata M. Reduced C-terminal Src kinase (Csk) activities in hepatocellular carcinoma. Hepatology. 1999;29:379–84.View ArticlePubMedGoogle Scholar
- Yang C-C, Fazli L, Loguercio S, Zharkikh I, Gleave ME, Wolf DA. Downregulation of c-SRC kinase CSK promotes castration resistant prostate cancer and pinpoints a novel disease subclass. Oncotarget. 2009;6:22060.View ArticleGoogle Scholar
- Wiener JR, Windham TC, Estrella VC, Parikh NU, Thall PF, Deavers MT, Bast RC, Mills GB, Gallick GE. Activated SRC protein tyrosine kinase is overexpressed in late-stage human ovarian cancers. Gynecol Oncol. 2003;88:73–9.View ArticlePubMedGoogle Scholar
- Huang Y-W, Chen C, Xu M-M, Li J-D, Xiao J, Zhu X-F. Expression of c-Src and phospho-Src in epithelial ovarian carcinoma. Mol Cell Biochem. 2013;376:73–9.View ArticlePubMedGoogle Scholar
- Sirvent A, Bénistant C, Pannequin J, Veracini L, Simon V, Bourgaux J-F, Hollande F, Cruzalegui F, Roche S. Src family tyrosine kinases-driven colon cancer cell invasion is induced by Csk membrane delocalization. Oncogene. 2010;29:1303–15.View ArticlePubMedGoogle Scholar
- Han LY, Sood AK. Inhibition of the Src oncogene: therapeutic potential in ovarian carcinoma. In Kaye S, Brown R, Gabra H, Gore M, editors. Emerging therapeutic targets in ovarian cancer. New York: Springer; 2011. p. 95–107.View ArticleGoogle Scholar
- Secord AA, Teoh DK, Barry WT, Yu M, Broadwater G, Havrilesky LJ, Lee PS, Berchuck A, Lancaster J, Wenham RM. A phase I trial of dasatinib, an Src-family kinase inhibitor, in combination with paclitaxel and carboplatin in patients with advanced or recurrent ovarian cancer. Clin Cancer Res. 2012;18:5489–98.View ArticlePubMedPubMed CentralGoogle Scholar
- Kim HS, Han HD, Armaiz-Pena GN, Stone RL, Nam EJ, Lee JW, Shahzad MMK, Nick AM, Lee SJ, Roh JW, Nishimura M, Mangala LS, Bottsford-Miller J, Gallick GE, Lopez-Berestein G, Sood AK. Functional roles of Src and Fgr in ovarian carcinoma. Clin Cancer Res. 2011;17:1713–21.View ArticlePubMedPubMed CentralGoogle Scholar
- Nakagawa T, Tanaka S, Suzuki H, Takayanagi H, Miyazaki T, Nakamura K, Tsuruo T. Overexpression of the csk gene suppresses tumor metastasis in vivo. Int J cancer. 2000;88:384–91.View ArticlePubMedGoogle Scholar