Epigenetic Repression of microRNA-129-2 Leads to Overexpression of SOX4 Oncogene in Endometrial Cancer

The SRY-related high-mobility group box 4 gene, or SOX4, is known to be overexpressed in prostate, hepatocellular, lung, bladder, and medulloblastoma cancers with poor prognostic features and advanced disease status (1–6). Its oncogenic potentials have been shown in knock-in cells leading to aberrant transformation, whereas the proliferation and metastatic capability has been greatly reduced in knockout cancer cells (1–4, 7). Functional analysis has shown that SOX4 belongs to the T-cell factor/lymphoid enhancer factor family of transcription factors that mediate transcription responses to Wnt signaling (2, 8). More recently, a genome-wide chromatin immunoprecipitation study has further uncovered additional transcriptional targets of SOX4 that are associated with the transforming growth factor β, Hedgehog, and Notch pathways, microRNA (miRNA) processing, and tumor metastasis (7, 9).

Genetic mechanisms leading to aberrant expression of SOX4 have been explored in cancer cells. SOX4 is mapped to chromosome 6p22, a region frequently amplified in lung, bladder, and endometrial cancers (2, 10–12). Somatic mutations have also been found in the exon region of this intronless gene in lung cancer (2). However, there is no experimental evidence to show positive correlation between these reported genetic alterations and the aberrantly increased SOX4. One emerging mechanism is miRNA-mediated oncogene expression.

miRNAs, a class of small noncoding RNAs (18–25 nunleotides), are known to form imperfect paring at the 3′-end of untranslated regions (UTR) of a target locus, resulting in mRNA degradation or translational inhibition (13). Through this posttranscriptional regulatory mechanism, miRNAs control a variety of physiologic processes in normal cells, and deregulation of miRNAs may promote tumorigenesis (14). Increasing evidence indicates that epigenetic perturbations may contribute to abnormal miRNA expression in cancer cells (15). One well-studied epigenetic phenomenon is DNA methylation frequently observed in the promoter CpG island regions of genes (16). Whereas promoter hypermethylation is associated with transcriptional silencing of coding genes for tumor suppressor functions, promoter hypomethylation can be related to activation of oncogenes in cancer cells (16, 17). We therefore speculate whether methylation alteration may commonly occur in noncoding miRNAs, resulting in deregulation of its target genes in cancer cells.

Here, we report for the first time that the SOX4 oncogene is also overexpressed in endometrial cancer. A miRNA, miR-129-2, was computationally predicted and functionally validated to be an upstream regulator of SOX4. A CpG island encompassing the miR-129-2 locus was found to be hypermethylated in endometrial cancer cell lines and primary tumors. We further show that this methylation-mediated silencing has a causal role for SOX4 activation in endometrial cancer.

Tissue specimens (117 tumors and 8 uninvolved controls) were obtained as part of our ongoing work on characterizing molecular alterations in endometrioid endometrial carcinomas. All participants consented to both molecular analyses and follow-up studies, and the protocols were approved by the Human Studies Committee at the Washington University and the Ohio State University. Clinicopathologic variables of tumors, including age, stage, grade, microsatellite instability (MSI), and MLH1 methylation, were summarized in Supplementary Table S1 and reported in our previous study (18). Human endometrial cancer cell lines, AN3CA, HEC1A, Ishikawa, KLE, RL95-2, and SK-UT-1B, were routinely maintained in our laboratory (19), and ECC-1 cells were obtained from the American Type Culture Collection. For epigenetic studies, these cells were treated with 5-aza-2′-deoxycytidine (DAC; 0.5 μmol/L; Sigma) for 48 h and/or trichostatin A (TSA; 5 μmol/L; Sigma) for 24 h. DNA and RNA from treated and untreated cells were isolated using standard protocols (20).

Tissue microarray slides, each containing a total of 74 endometrial tumors and 20 normal specimens, were obtained from US Biomax. Patients' characteristics were summarized in Supplementary Table S2. These slides were preprocessed before immunohistostaining. Antigen retrieval was done by heat-induced epitope retrieval, in which the slides were placed in Dako TRS solution (pH 6.1) for 25 min at 94°C. Slides were then placed on a Dako autostainer with primary antibody (SOX4, 1:50; Abcam) and incubated for 60 min at room temperature. Staining was visualized with 3,3′-diaminobenzidine chromogen, and slides were then counterstained and dehydrated through graded ethanol solutions. Images were digitally scanned with iScan (BioImagene) and analyzed with the BioImagene TissueMine software for discriminating immunohistochemically stained cancer cells from the surrounding stromal tissue. Intensities of nuclear staining were measured as segmented images and then quantified. The algorithm reported the number and percentage of positively stained and nonstained nuclei.

ECC-1 and Ishikawa cells (3 × 10⁶) were transfected with mature miR-129-2 mimics (miR-129-3p and miR-129-5p, 2.5 nmol/L; Ambion), pre-miR negative control (#1, 2.5 nmol/L; Ambion), SOX4 siGenome SMART pool siRNA (2.5 nmol/L; Dharmacon), siGenome nontargeting siRNA pool (#1, 2.5 nmol/L; Dharmacon), and plasmids using the Cell Line Nucleofector Kit (Lonza) according to manufacturer's instructions.

Total RNA (1 μg) was reverse transcribed with the Superscript III reverse transcriptase (Invitrogen). PCR was done as described previously (20). Specific primers for amplification are listed in the Supplementary Table S3. The relative expression of a coding gene in cells was determined by comparing the threshold cycle (Ct) of the gene against the Ct of GAPDH (20).

For detecting mature miRNA molecules (i.e., miR-129-3p and miR-129-5p), reverse transcription was done following the Applied Biosystems TaqMan MicroRNA Assay protocol. This sensitive system has been designed to specifically detect mature miRNAs that are distinct from their precursors. In addition, the assay can often distinguish between miRNA targets that differ by only a single nucleotide (21). All reactions were done in triplicate. The expression of miR-129-2-3p or miR-129-2-5p was normalized using RNU48 or U6. The expression relative to RNU48 or U6 was determined using the 2^−ΔCt method.

Whole-cell protein lysates were extracted with the M-PER Mammalian Protein Extraction Reagent (Pierce). Western blot analysis was conducted using antibodies against SOX4 (Abcam) and β-actin (Santa Cruz).

Cell proliferation was monitored using the CellTiter 96 Aqueous solution (Promega). Endometrial cancer cells (3,000 per well) transfected with miR-129-2, negative control miRNA, SOX4 siRNA, or nontargeting siRNA pool were seeded in 96-well plates. Cell proliferation was documented every 24 h following the manufacturer's protocol. To measure cell proliferation, 20 μL of MTS labeling reagent were added to each well and incubated at 37°C for 1 h. Absorbance was measured at 490 nm in an ELISA reader (Molecular Devices).

The full-length 3′-UTRs of SOX4 and UBE2F, generous gifts from Dr. Joan Massague, were cloned into the Psicheck 2 dual luciferase reporter vector (Promega). ECC-1 or Ishikawa cells were transfected with reporter constructs and miR-129-2 and/or its antagomir targeting endogenous miR-129-2 (Ambion). Cells were lysed at 24 h after transfection, and the ratio of Renilla to firefly luciferase was measured using the dual luciferase assay (Promega). Normalized Renilla-to-firefly ratios were determined in the presence or absence of miR-129-2 inhibition on SOX4 UTR luciferase activities.

Genomic DNA (500 ng) was treated with sodium bisulfite using the EZ DNA Methylation kit (Zymo Reasearch) following the manufacturers' recommended protocols. Combined bisulfite restriction analysis (COBRA) was used to evaluate methylation of miR-129-2. Target sequences were amplified by PCR, and the products were digested with AciI (New England Biolabs) to identify methylated sequences. Primer sequences and PCR conditions are presented in Supplementary Table S3. Digested and nondigested PCR products were resolved on 2% agarose gels stained with ethidium bromide. DNA fragments digested by AciI were scored as “methylated” in a given sample.

To quantify methylation levels of the miR-129-2 CpG islands in clinical samples, the high-throughput MassARRAY platform (Sequenom) was carried out as described previously (22). Briefly, bisulfite-treated DNA was amplified with primers for the miR-129-2 CpG island. The PCR products were spotted on a 384-pad SpectroCHIP (Sequenom), followed by spectral acquisition on a MassARRAY Analyzer. Methylation data of individual units (one to three CpG sites per unit) were generated by the EpiTyper software (Sequenom).

Student's t test or Wilcox test was used to compare the immunohistostaining, cell proliferation, and reverse transcription and quantitative PCR (RT-qPCR) results in different treatment groups. Significance was assigned at P < 0.05 (*). The relationship between methylation levels of miR-129-2 and relevant categorical clinicopathologic covariates was done using the Wilcox rank sum test for binary variables. The Kruskal-Wallis test was used followed by a pairwise Wilcoxon rank sum test. Overall survival was defined as the time interval from the date of diagnosis to the date of death or latest follow-up if alive. Recurrence-free survival was defined as the time interval from surgery to recurrence, disease progression, or latest follow-up. The Kaplan-Meier product limit method was used to estimate the empirical survival functions for categorical covariates accompanied with P value from the log-rank test. Univariate Cox proportional hazard models were used to access the effect of a continuous covariate on survival outcomes. Multivariate Cox proportional hazard models were fitted to examine the potential predictive effect of covariates of interest on survival outcomes after adjustment for confounding factors. All tests were two-sided and all analyses were done using R.

To determine whether SOX4 is aberrantly expressed in endometrial tumors, we conducted tissue microarray analysis in a panel of 74 endometrioid endometrial carcinomas and 20 uninvolved controls (see representative immunohistostaining images in Fig. 1A,, left). The nuclear staining intensities of SOX4 were significantly higher in tumor sections than those of normal tissue sections (P < 0.005; Fig. 1A,, right). This finding is consistent with the RT-qPCR results in which the levels of SOX4 mRNA were higher in tumors (n = 31) compared with the adjacent normal counterparts (P < 0.001; Fig. 2D , left).

In a knockout study, we assessed the role of SOX4 in endometrial cancer cells. ECC-1 and Ishikawa cells were transiently transfected with SOX4 siRNA or a nontargeting control. Both protein and mRNA levels of SOX4 were found to be reduced to ≤50% in transfectants (Fig. 1B and C), resulting in attenuation of cell growth (Fig. 1D). This SOX4 knockout, however, had no effects on cell invasion or migration (data not shown).

Because miRNA may have a potential role in mediating oncogene repression (14), we searched potential target sequences at the 3′-UTR of SOX4 using three software programs: PicTar, TargetScan, and miRanda (23–25). Putative binding sites were found in a miRNA locus, miR-129-2, which is the precursor of two mature forms, miR-129-3p and miR-129-5p (Fig. 2A and B). To further validate this computational finding that miR-129-2 may negatively regulate SOX4, we assessed the expression of 3′-UTR of SOX4 in luciferase reporter assays (Fig. 2C). The expression of the SOX4 reporter was significantly reduced to ≤55% in miR-129-3p– or miR-129-5p–transfected ECC-1 and Ishikawa cells, but not in control cells. Transfection with either miRNA did not affect the reporter activity of a negative control gene, UBE2F, which has no known miR-129-2 binding sites on its UTR. Moreover, inhibition of miR-129-3p or miR-129-5p by antagomirs slightly enhanced the expression of SOX4, suggesting that this gene is a direct target of miR-129-2. We additionally confirmed this inverse relationship between miR-129-3p and SOX4 mRNA and expression using the aforementioned 31 paired samples (Fig. 2D).

Because the expression of miR-129-2 is frequently lost in endometrial tumors (see Fig. 2D) and the 5′-end of this locus has a canonical CpG island (Fig. 3A), we determined whether this downregulation is mediated by epigenetic mechanisms. Hypermethylation of this promoter CpG island was detected in six endometrial cancer cell lines, ECC-1, HEC1A, Ishikawa, KLE, RL95-2, and SK-UT-1B, based on a COBRA assay (Fig. 3B). When these cells were treated with a demethylating agent, DAC (0.5 μmol/L), a histone deacetylase inhibitor, TSA (5 μmol/L), or their combination, reactivation of miR-129-3p was observed in four (Ishikawa, KLE, RL95-2, and SK-UT-1B) of the six cell lines analyzed that were treated with DAC (Fig. 3C). More profound effects of this reexpression were found in all these cell lines treated with TSA and the combination (Fig. 3C; Supplementary Fig. S1). These results suggest that the loss of miR-129-2 expression is associated with promoter hypermethylation in endometrial cancer cells. Interestingly, these epigenetic treatments might lead to the suppression of SOX4 in endometrial cancer cells. The effect occurred at the mRNA level in ECC-1, HEC1A, KLE, RL95-2, and SK-UT-1B cells (Fig. 3D,, left). This suppression, however, was more prominent at the posttranslational level for Ishikawa and RL95-2 cells (Fig. 3D , right). In addition, we showed that the expression of four other miR-129-2 targets, EIF2C3, PLCG1, RAB21, and STAT5B, was repressed in cancer cells treated with DAC and/or TSA (Supplementary Fig. S2). Taken together, the observation indirectly indicates that these epigenetic treatments may lead to reactivation of miR-129-2, which in turn represses the expression of SOX4.

To further prove this inverse relationship, we conducted a functional knock-in study in two endometrial cell lines, ECC-1 and Ishikawa, harboring the epigenetically silenced miR-129-2. Transient transfection of these cells with either miR-129-3p or miR-129-5p resulted in reduction of both SOX4 mRNA and protein (Fig. 4A and B) (Note: Transfection efficiency was examined by measuring each mature miRNA; see Supplementary Fig. S3.) Moreover, the knock-in of one of these mimics, miR-129-5p, greatly reduced the proliferation of these cancer cells (P < 0.05; Fig. 4C). In addition, the expression of three SOX4-regulated genes (DHX9, FZD5, and SEMA3C; refs. 3, 9) was partially reduced in ECC-1 cells treated with DAC and/or TSA or ectopically transfected with miR-129-3p or miR-129-5p (Supplementary Fig. S4). Taken together, these in vitro studies suggest that miR-129-2 negatively regulates SOX4 and that promoter hypermethylation of this miRNA derepresses its expression.

To confirm our in vitro findings in primary tumors, we first studied the methylation patterns of the miR-129-2 CpG island (see Fig. 3A) in eight tumors with matched normal endometria by COBRA and quantitative MassARRAY. We showed that the tumors were hypermethylated compared with the corresponding adjacent normal tissues (Supplementary Fig. S5). We then extended this methylation study to 117 (34 recurrent and 83 nonrecurrent) primary endometrioid endometrial tumors and 8 uninvolved controls. Quantitative analysis indicated that 80 of these primary tumors exhibited extensive methylation in 14 units (one to three CpG sites per unit) of the miR-129-2 CpG island relative to those of uninvolved controls (P < 0.0005; Fig. 5A). This methylation survey in the patient cohort also uncovered a pattern in which methylation accumulation may begin at the flanking regions and progressively invade to the core of this CpG island, consistent with the so-called methylation spread theory (26).

The mean methylation level of each CpG unit was used to compare between the recurrent and nonrecurrent groups. More accumulation of miR-129-2 methylation was found in the former group, but the sample size of this cohort could be too small to reach statistical significance (P = 0.066; Fig. 4B). Detailed analysis of individual CpG units is shown in Supplementary Fig. S6. We also evaluated the association between miR-129-2 hypermethylation and patient survival. On univariate analysis, miR-129-2 hypermethylation was correlated with shorter overall survival (Cox hazard ratio, 1.02; P = 0.018), but not with recurrence-free survival (Cox hazard ratio, 1.02; P = 0.067), when the mean methylation level of normal controls was used to set the threshold. The Kaplan-Meier survival analysis indicated that patients with this hypermethylation had poor overall survival (Fig. 5C; P = 0.039, log-rank test). Statistical analysis further revealed that miR-129-2 hypermethylation was significantly associated with MSI and MLH1 methylation (P < 0.0001; Fig. 5D). Endometrial tumors exhibiting the MSI phenotype usually have high replication error rates and genomic instability (18, 27). This defect is attributed, in part, to epigenetic repression of MLH1, which is responsible for DNA mismatch repair in normal cells (27).

In addition to genetic alterations, promoter hypomethylation has been recognized as an epigenetic mechanism that contributes to oncogene activation in cancer cells (16, 17). In this case, a demethylation event is supposed to occur in an inactive, methylated promoter, leading to transcriptional reactivation of an oncogene. However, experimental proof for genuine promoter hypomethylation is frequently difficult and inconclusive because the outgrowth of a subpopulation of cancer cells may confound this epigenetic observation. For example, the oncogene of interest may have never been silent in a minor population of cancer-initiating cells, whereas the majority of other cells display promoter hypermethylation of the gene. The increased expression of this oncogene may simply result from rapid expansion of these few cells that eventually take over the whole population during tumor progression. If this scenario indeed occurs, it cannot be a bona fide epigenetic event for oncogene activation.

In this study, we show that promoter hypermethylation can be directly associated with the activation of an oncogene. Specifically, this epigenetic event occurs in an upstream regulator, miR-129-2, which was shown to negatively regulate SOX4 oncogene in both knock-in and knockout assays. miR-129-2 is embedded in a canonical CpG island on chromosome 11, which was found to be frequently hypermethylated in endometrial cancer. This epigenetic event results in miR-129-2 silencing, which in turn derepresses SOX4 expression. Although we still cannot rule out hypomethylation of the SOX4 promoter CpG island as one of the causes, our present observation conclusively suggests that promoter hypermethylation of miR-129-2 is a common mechanism leading to the SOX4 overexpression in endometrial cancer.

It should be noted that a second CpG island is found 1.2 kb upstream from the first one analyzed in this study. In addition, the 5′-ends of two transcripts, BG120451 and BI964058, are located in this upstream region. It is possible that these transcripts are the primary RNAs for miR-129-2 (28). Future mapping of putative transcription start sites located in these 5′-end areas will provide insight into the transcriptional control of this miRNA locus. Additional methylation analysis may further determine the role of this second CpG island in the silencing of miR-129-2 during endometrial tumorigenesis.

We have additionally searched the miRBase database and found that the expression of SOX4 may be regulated by at least 16 putative miRNAs, including miR-129-1 located on chromosome 7. Similar to miR-129-2, the miR-129-1 precursor produces mature miR-129-5p, which negatively regulates SOX4, as described in this study. Because there is no known CpG island located near or within the miR-129-1 locus, it remains to be determined whether this miRNA is transcriptionally silenced by other epigenetic mechanisms (e.g., EZH2-mediated histone modifications; ref. 29) in endometrial cancer.

Five (miR-203, miR-335, miR-219-2, miR-596, and miR-618) of the other 15 miRNAs are located close to CpG islands based on our computational analysis (data not shown). Among these loci, miR-335 is the only one currently reported to be lost in primary breast tumors of recurrent patients (7). However, it remains to be determined whether promoter hypermethylation plays a role in the silencing of this locus. Future studies will therefore be important to study the coordinate regulation of these miRNAs on SOX4 repression. It is also possible that concurrent hypermethylation of these loci contributes to a CpG island methylator phenotype (30) and may improve the statistical power for predicting disease recurrence in our endometrial patient cohort (see Fig. 5B).

Extending from our present observation, epigenetically mediated silencing of other miRNAs that lead to tumor progression has recently been reported in the literature (15, 31–33). For example, ABL1 was showed to be a direct target of miR-203 (32). This miRNA was genetically and epigenetically downregulated in leukemia cells expressing ABL1 or BCR-ABL1 fusion protein (32). Restoration of miR-203 in vitro led to reduced ABL1 and BCR-ABL1 expression and decreased proliferation of malignant cells (32). Taken together, these and our studies clearly indicate that epigenetic silencing of tumor-suppressive miRNAs can be an important constituent in cancer epigenomes and that the event is as significant as hypomethylation of oncogenes and hypermethylation of tumor suppressor genes.

In conclusion, our findings support a comprehensive screening of miRNA regulators at the 3′-UTR regions of all known oncogenes. High-throughout functional studies can be developed to establish the inverse relationship between these tumor-suppressive miRNA loci and their target oncogenes. This type of omics study may find that epigenetically mediated silencing of these miRNAs can be as common as genetic alterations that contribute to oncogene activation in cancer cells. As such, the combined epigenetic and miRNA-based therapies can be feasible options for future treatments in cancer patients.

All authors have no potential conflicts of interest.

Grant support: NIH grants R01 CA069065, U01 ES015986, and U54 CA113001 (T.H-M. Huang); NIH grants R01 CA071754 and P50 CA134254 (P.J. Goodfellow); and funds from the Ohio State University Comprehensive Cancer Center (T.H-M. Huang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Chieh Ti Kuo, Xiao-Hong Gu, Geoffrey Tsoi, Mary Ann Mallon, and Drs. Kurtis H. Yearsley and Yu-I Weng for their technical assistance, and Drs. Joan Massague and Sohail F. Tavazoie (Memorial Sloan-Kettering Cancer Center, New York, NY) for providing SOX4 and UBE2F 3′-UTR plasmids.