Suppression of cancer stemness p21-regulating mRNA and microRNA signatures in recurrent ovarian cancer patient samples

Murine ES (ES-E14TG2a) and EC cells (pluripotent 'SCC-PSA1' and nullipotent 'Nulli-SCC1') were purchased from ATCC, cultured on murine irradiated fibroblasts in DMEM supplemented with 10% foetal bovine serum, 4 mM L-glutamine (Invitrogen) and 100 U/ml of penicillin/streptomycin (Invitrogen Corporation, Carlsbad, CA, USA) and spontaneously differentiated via removal of feeder layer. Human EC cells were retinoic acid-differentiated as previously described [15].

Tumor sample data was previously published [5, 6]. Briefly, two cohorts of primary and recurrent samples were assessed. Cohort 1 contained 5 primary and recurrent serous papillary adenocarcinomas (Grade 3). Cohort 2 contained 3 paired ovarian cancers from the same patient but with different histologies: papillary serous, mixed mullerian and clear cell carcinomas.

RNA was isolated using the RNeasy kit (Qiagen, West Sussex, UK) as per manufacturer's protocol. Digoxigenin-UTP labelled cRNA was synthesized via the Chemiluminescent RT-IVT Labelling Kit v2.0 (Life Technologies, Foster City, CA, USA) and hybridized to Mouse Genome Survey arrays (Life Technologies) as per manufacturers' instructions. Data was filtered to a signal/noise ration threshold > 3 in at least one sample using R and further analysed using Spotfire^® (Life Technologies). Genelists were generated using cut-offs of 0.05 (p-Value) and ± 2.0 (fold change). Functional relationships were analysed using DAVID [19, 20]. Pathways associations of predicted targets of miRNAs highlighted were generated using DIANA miRPath [21] using cut-offs of ≥ 2 genes per pathway and p-value ≤ 0.05.

It is well established that ES and EC cells express similar gene profiles in the undifferentiated and well-differentiated (one week or later) states [16, 11, 17, 18]. In contrast, our understanding of the earlier, upstream regulation of differentiation is sparse. We hypothesized that comparison of early differentiation of mES and mEC cells would identify cancer-specific differences in upstream regulation of stem cell differentiation. Addressing this we used microarray analysis to assay early (three day) differentiation of mES and mEC cells.

Microarray data was validated through qPCR analysis, showing excellent correlation (Figure 1). An overview of the number of differentially expressed genes in pluripotent (SCC-PSA1) and nullipotent (Nulli-SCC) mEC and mES cells is shown in Table 1. At cut-offs of 0.05 (p-Value) and ± 2.0 (fold change) SCC-PSA1 cells alter the expression of 724 genes: 202 upregulated and 522 downregulated at fold change levels between +18 and -18 (Additional File 1). Top ten SCC-PSA1 genes are characterised by receptor activity and growth and differentiation/development roles (Table 2). Noteworthy events include upregulation of apoptosis (Bcl)-related gene Bid3 and downregulation of Cav2 tumor suppressor [24] and metastasis-linked Nupr1 [25]. Functional relationship analysis identified upregulation of developmental pathways and downregulation of transcription regulation processes and Toll-Like Receptor (TLR), Interleukin-2 and cancer pathways (Additional File 1).

Nulli-SCC cells responded to differentiation stimuli through the upregulation of 185 and downregulation of 152 genes at levels from -6.3 to 14.0 fold (Additional File 2). Top ten genes included signal transducers and regulators of development/differentiation and malignancy (Table 2). Notable genes include hypoxia and tumor growth regulator Loxl2 [26] and tumor suppressor Serpini2 [27]. Interestingly Ssa2 is downregulated, a gene that is commonly expressed on the surface of apoptotic cells. Functional analysis identified upregulation of signal transduction regulators and downregulation of growth regulators (Additional File 2).

Upstream differentiation of mES cells is characterized by substantial levels of upregulation: 554 upregulated and 832 downregulated genes at levels of 232 to -68 fold (Additional File 3). Top ten genes are populated with receptors and developmental regulators (Table 2). Tll1 is linked to cardiac development, the first organised system formed during embryogenesis. Notably, a key RNAi gene, Eif2c4, is upregulated during differentiation, perhaps reflective of involvement of the RISC complex [28]. Upregulated mES genes regulate development, signalling and gene expression while downregulated genes regulate morphogenesis, particularly growth factor binding. Stemness-linked pathways such as Wnt-catenin and Hedgehog signalling were upregulated while signalling pathways including TLR and TGF-ß were downregulated (Additional File 3).

A comparison of mES and mEC early differentiation genelists is summarised in Table 1 and detailed in additional files 1, 2 and 3. In contrast to documented undifferentiated and well-differentiated comparisons, 90% of the mES genelist differed to the mEC genelist at this earlier time point (Table 1). Similarly, almost 70% of the SCC-PSA1 genelist differed from the mES genelist (Table 1). Functional relationship analysis indicates that quite different mechanisms are activated during early differentiation of mEC and mES cells. This included mES-specific upregulation of p53 signaling pathway genes (Additional File 3). There is very little overlap between Nulli-SCC and the other cell types (Additional Files 1, 2 and 3). Only four genes are upregulated by SCC-PSA1 and downregulated by Nulli-SCC cells, while only two are downregulated by both cell types. The downregulation of symporters, signal-transducing membrane proteins, which are upregulated by pluripotent cells, may indicate a potential counteraction of differentiation. Upstream regulation of differentiation represents a substantial difference between these cell types, supporting our hypothesis. While similar genes maintain the self-renewal state in each cell, different mechanisms are employed to regulate the early events in differentiation.

We have previously published microarray analysis of primary versus recurrent tumor samples [5, 6]. The study contained two cohorts. Cohort 1 represents a group of matched primary and recurrent tumors while cohort 2 represents primary and recurrent tumor samples from the same patients. In this study, raw microarray data from the primary versus recurrent study was reanalysed in an identical fashion to mES and mEC data described above (Additional File 4). Primary versus recurrent disease and mEC genelists were then compared. Genes altered similarly in mEC and mES data were not considered to be cancer-specific and were removed from this analysis. Comparison of mEC and tumor data identified 16 SCC-PSA1 genes expressed in tumor samples (Figure 2, Table 3). These genes group into those that are A) expressed in both primary and recurrent tumors and B) those expressed in primary but not recurrent tumors (Table 3). Many of these genes have links to stemness and malignancy. Tmprss2 is a transmembrane signalling protein that is upregulated in prostate cancer [29]. Cthrc1 is a Smad2/3 (TGF-β signalling) inhibiting Wnt signalling modulator that is differentially expressed in invasive breast cancer and several solid tumors [30]. Nkx3-1 is a metastatic marker transcription factor expressed in prostate cancer [31]. Pdgfc is a cisplatin-associated growth factor [32]. Col4a5 is linked to several cancers while Plaur is a regulator of tissue reorganisation [33]. Ndufs6 is an oxidative phosphorylation enzyme linked to cervical cancer [34]. Pdzk1 is linked to oestrogen-sensitivity in breast and ovarian cancers [35]. Sdsl is a cancer-specific metabolic enzyme [36]. Only one gene, Gata2, an endodermal differentiation marker [37] was upregulated by mEC cells and expressed higher in primary tumors than in recurrent.

When scrutinised, we noted that several of the genes highlighted above have been defined as p53 regulators in various models, as now described. Dusp26 is a p53-inhibiting phosphatase that negatively regulates proliferation of epithelial cells [38]. Stemness gene Sox4 is a p16 and p53 regulator in cancer cells [39] while Hsf2 is a regulator of p53 stability [40]. Hoxb2 has been linked to p205 regulation of p53 and is a well known regulator of EC differentiation [41]. Collectively, our analysis indicates that both primary and recurrent ovarian tumors express this 'p53-regulating stemness signature'.

Despite the reduced genelist size, 14 Nulli-SCC genes were expressed in A) both primary and recurrent tumors or B) primary tumors only (Figure 3, Table 3). These genes related to apoptosis/cellular proliferation, signaling and regulation. Dgcr8 (Pasha) is a key miRNA biosynthesis gene [42], while Tirap is a regulator of TLR signaling [43]. TNF-family related Ltbr and hypoxia-linked Egln3 are apoptosis regulators [44, 45]. Gpr6 is a development regulator expressed in umbilical cord cells [46]. Ndufab1 is a TGF-β signaling related NADPH enzyme [47]. Slc15a1 is involved in drug absorption in the small intestine and has been linked to several cancers and metastasis [48]. Coupled with this is the recurrent suppression of apoptosis regulators Bnip3 and Stau2 [49, 50]. Notably, two p21 regulators are expressed higher in primary tumors compared to recurrent: Cask mediates the expression of p21 to control cell proliferation [51] while Pak6 is a p21 interacting kinase that is a required for chemoresistance in prostate cells [52]. Collectively, an EC cancer stemness signature expressed in tumor samples is linked to maintained p53 regulation and suppression of p53's main target, p21, in recurrent disease.

Having identified gene level overlaps, we next conducted overlap meta-analysis of our previously published miRNA data for primary and recurrent patient samples and human EC (hEC) early (three day) differentiation [5, 6, 15]. The earlier study identified cancer stemness signature miRNAs: those miRNAs involved in the differentiation of hEC cells. Specifically, our previous tumor study highlighted 60 miRNAs (52 up and 8 downregulated) in recurrent disease [6]. Of these, 55 miRNAs (92%) are expressed in hEC cells (Additional File 5). 21 recurrent disease-specific miRNAs are linked to differentiation of pluripotent NTera2 hEC cells (Figure 4). We have previously shown that nullipotent 2102Ep hEC cells express a large number of miRNAs at substantially higher levels than NTera2 cells [15]. Here we report that 26 (43%) recurrent disease-specific miRNAs are expressed at higher levels in 2102Ep cells than in NTera2 (Figure 5). Thus, development of recurrent tumors involves recruitment of cancer stemness signature miRNAs. Specific examples include miR-9, which is the most downregulated miRNA in recurrent tumors and is > 1000% higher expressed in undifferentiated 2102Ep cells compared to NTera2, and miR-206, which is in the top ten miRNAs upregulated by recurrent tumors and downregulated during NTera2 differentiation. Molecular pathway relationships between predicted gene targets of the miRNAs highlighted were identified using DIANAmirPATH (Additional file 5). While little pathway overlap was observed in gene array data, miRNA data showed strong pathway associations. Pathway analysis highlighted alteration of several cancer pathways (miRs-10b, -100, -106, -107, -128 and let-7g) as well as Wnt and TGF-β stemness signaling pathways (mirs-10b, -100, -106, -128a and 137). Finally, we assessed the expression of p53-p21 regulating miRNAs in these datasets. Two miRNAs, miRs-106a and b, are validated targets of p21 [52] that are upregulated in recurrent disease and expressed in hEC cells. Notably, miR-106b expression in 2102Ep cells is double that of NTera2 cells. In contrast, miR-155, the only validated p53-regulating miRNA, is unaltered in recurrent tumors. We note that the p53 signaling pathway was highlighted for let-7g and miRs-106b and -107 in pathway analysis (Additional file 5). In overview, we find that miRNAs linked to 2102Ep malignancy are highly relevant to primary and recurrent tumors.