Epigenetic Silencing of miR-137 Is an Early Event in Colorectal Carcinogenesis Free

microRNAs (miRNA) are involved in the pathogenesis of multiple types of cancers, including CRC (1, 2). Growing evidence suggests that miRNAs can act as oncogenes (oncomiR) or tumor suppressor genes (tsmiR), and they are involved in the early stages of carcinogenesis (2). The pattern of miRNA expression can be used to classify diverse types and subtypes of cancers (1, 3), and miRNA expression profiles can have prognostic and therapeutic implications (4). Compared with mRNA, a modest number of miRNAs might be sufficient for clinical purposes, and more interestingly, miRNAs remain largely intact in different tissues and are virtually unaffected by RNA degradation (5). All these features make miRNAs a very exciting and promising tool for early tumor detection, prognostication, and treatment.

The causes of the widespread differential expression of miRNAs between tumor and normal cells are still unclear. Approximately 20% of all miRNAs are embedded within CpG islands, and pharmacologic unmasking of silenced miRNAs using epigenetic drugs has revealed that several miRNAs can be inactivated by this mechanism in CRC cell lines (6–8). These studies have permitted the identification of miRNAs that are methylation-silenced in CRC patients (7, 9–11); however, the concept of epigenetic regulation of miRNAs in colorectal carcinogenesis remains largely unexplored.

miR-137 is one such miRNA, which is located on chromosome 1p22 within the non–protein-coding RNA gene AK094607 (12). This miRNA is embedded in a CpG island and is frequently silenced by methylation in several tumors (9, 13, 14). Ectopic transfection of the miR-137 precursor in oral cancer and glioblastoma multiforme inhibited cell growth, suggesting its tumor-suppressive activity (13, 14). However, the biological role of miR-137 as well as its specific downstream mRNA targets in colorectal carcinogenesis remains unknown. In addition, there are essentially no data about miR-137 disruption in adenomas, the precursor lesion of CRC. We have for the first time systematically characterized the role of miR-137 in the colon by addressing some of the issues mentioned above. We investigated the epigenetic regulation of miR-137 in a panel of six CRC cell lines and more than 400 colorectal tissues, which include both colorectal adenomas and cancers. We explored the tumor-suppressive features of miR-137 in vitro and identified potential mRNA targets using whole genome expression profiling. In addition, we have successfully validated LSD1, a key element of the epigenetic machinery, as one of the targets of miR-137.

We used six different CRC cell lines (HCT116, LoVo, RKO, SW48, HT29, and SW480) obtained from the American Type Culture Collection during the last 2 years. In our laboratory, all cells are tested and authenticated every 6 months using known genetic and epigenetic marks. Cells were grown in appropriate culture conditions. For demethylation experiments, cells were treated with 2.5 μmol/L 5-aza-2-deoxy-cytidine (5-aza-CdR; Sigma) for 72 hours, replacing the drug and medium every 24 hours.

A collection of 409 colorectal tissues were analyzed in this study, which included 113 sporadic primary CRCs with their corresponding adjacent normal colonic mucosa (N-C) and 68 colorectal adenomas obtained from the Okayama University Hospital, Okayama, Japan. Twenty-one normal colonic mucosa specimens from nontumor patients (N-N) were collected at the Hospital Clinic of Barcelona, Spain. Additionally, 47 CRC tissues and the corresponding adjacent normal mucosa from 11 patients with Lynch syndrome, 14 patients with microsatellite unstable CRCs, and 22 patients with microsatellite stable tumors were collected at Baylor University Medical Center, Dallas, Texas. All patients provided written informed consent and the study was approved by the Institutional Review Boards of all participating institutions. Clinicopathologic data of the patients are presented in Table 1 and Supplementary Table S1.

Total RNA extraction was undertaken using the miRVana RNA extraction kit and the RecoverAll kit (both from Ambion, Inc.) according to the manufacturer's instructions.

The DNA methylation status of the miR-137 CpG island was established by PCR analysis of bisulfite-modified genomic DNA (EZ DNA methylation Gold Kit, Zymo Research) using three different methods. First, methylation status was analyzed by bidirectional bisulfite sequencing of HCT116 and RKO cell lines. Second, we performed methylation-specific PCR (MSP) using primers for either methylated or unmethylated DNA in the six CRC cell lines used in the study. Finally, we used bisulfite pyrosequencing for quantitative methylation analysis (PSQ HS 96A pyrosequencing system, QIAGEN). The primers used are described in Supplementary Table S2.

Expression of mature miR-137 was analyzed using the TaqMan miRNA Assay (Applied Biosystems, Inc.). Expression of RNU6B (Applied Biosystems) was used as an endogenous control. All the experiments were done in triplicate.

The in situ detection of miR-137 was performed on 5-μm formalin-fixed, paraffin-embedded (FFPE) sections from five normal colonic mucosa tissues, three adenomatous polyps, and five colorectal adenocarcinomas. Briefly, the slides were hybridized with 10 pmol probe (LNA-modified and DIG-labeled oligonucleotide; Exiqon) complementary to miR-137 and after incubation with anti–DIG-AP Fab fragments conjugated to alkaline phosphatase, and the hybridized probes were detected by applying nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate color substrate (Roche) to the slides. Positive controls (RNU6B, Exiqon) and no-probe controls were included for each hybridization procedure.

HCT116 cells were transfected with pre-miR miRNA precursor molecules (Ambion) or pre-miR miRNA negative control #1 (Ambion) at a final concentration of 100 nmol/L, using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. For microarray and reverse transcription-PCR (RT-PCR) analysis, total RNA was extracted 48 hours after transfection; for Western blot analysis, cell lysates were prepared 48 hours after transfection. To ensure transfection efficiency, we verified the protein downregulation of cyclin-dependent kinase 6 (CDK6), a previously validated target by Western blotting.

HCT116 cells were transfected with control miRNA precursor or miR-137 precursor as described above. Extracted RNA was amplified using the Illumina TotalPrep RNA Amplification Kit. RNA integrity was assessed using the Agilent 2100 Bioanalyzer. Labeled cRNA was hybridized overnight to Human HT-12 V3 chips, washed, and scanned on an Illumina BeadStation-500. Illumina BeadStudio version 3.1 was used to process signal intensity values from the scans and background subtracted. Normalization was done using quantiles with the Lumi R-package. Fold changes were calculated with respect to their respective control. miRecords (http://mirecords.umn.edu/miRecords; ref. 15) was used to predict the miRNA targeting of miR-137. To narrow down the list of predicted targets, genes found to be downregulated (>2-fold change) after transfection of miR-137 precursor in the microarray were crossed with the genes predicted to be targets based on miRecords. Genes previously found to be associated with either CRC specifically or carcinogenesis in general were selected for validation.

For RT-PCR, RNA was reverse transcribed to cDNA from 1 μg of total RNA using random hexamers and Advantage RT-for PCR Kit (Clontech Laboratories). Power SYBR Green (Applied Biosystems) RT-PCR was performed for selected targets found with the strategy described above. Results were normalized to the expression of β-actin. All the experiments were performed in triplicates. Primer sequences are listed in Supplementary Table S2.

Western blot analysis was carried out using standard methods. The following primary antibodies were used: anti-CDK6 (Cell Signaling) at 1:2,000 dilution; anti-LSD1 (Cell Signaling) at 1:250 dilution; anti-SEMA4D (BD Transduction Laboratories) at 1:250 dilution; anti-AURKA (Cell Signaling) at 1:250 dilution; anti-CSE1L (BD Transduction Laboratories) at 1:1,000; anti-BX1 (Cell Signaling) at 1:1,000 dilution; and anti–β-actin antibody (Clone AC-15) at 1:32,000 dilution.

Luciferase constructs were made by ligating oligonucleotides containing the wild-type or mutant putative target site of the LSD1 3′-untranslated region (UTR) downstream of the luciferase gene in the pMIR Reporter Luciferase vector (Ambion). Primers are detailed in Supplementary Table S2. Cells were cotransfected using Lipofectamine 2000 (Invitrogen) with 400 ng of firefly luciferase reporter vector containing the wild-type or mutant oligonucleotides, 200 ng of pGal control vector (Ambion), and 50 pmol of either miR-137 or negative control precursor. The parental luciferase plasmid was also transfected as a control. Luciferase activity was measured 48 hours after transfection (Bright Glo Luciferase Assay System, Promega) using β-galactosidase for normalization (β-galactosidase Enzyme Assay System, Promega). Experiments were performed in triplicate in three independent experiments.

The proliferation index was measured by bromodeoxyuridine (BrdUrd) incorporation in colon cancer cells 96 hours after transfection of either control miRNA precursor or miR-137 precursor as described above (Cell Proliferation ELISA, BrdUrd, Roche), following the manufacturer's instructions. Experiments were performed in triplicate in three independent experiments.

All data were analyzed using the SPSS 13 and GraphPad Prism 4.0 statistical software. Quantitative variables were analyzed using Student's test, Wilcoxon test (nonparametric paired analysis), and Mann-Whitney U test (nonpaired analysis). Qualitative variables were analyzed using χ² test or Fisher's exact test. A two-sided P value of <0.05 was regarded as significant.

To determine the DNA methylation status in the miR-137 CpG island, we used a sequential approach. The CpG island of miR-137, along with the localization of PCR products for each approach, is represented in Fig. 1A. First, we performed direct bisulfite sequencing of the promoter region of miR-137 in HCT116 and RKO cell lines, and we found extensive methylation throughout its promoter region in both cell lines, which was reversible following 5-aza-CdR treatment (Fig. 1C). Second, using MSP, we found that the miR-137 CpG island was extensively methylated in all of the CRC cell lines, and as expected, treatment with 5-aza-CdR induced significant demethylation (Fig. 1B). Third, we performed methylation analysis of miR-137 CpG island by quantitative pyrosequencing (Figs. 1D and 2A). Methylation, when present, generally affected all four sites CpG sites homogeneously; thus, their average was taken as the final measurement for analysis. The average methylation levels in HCT116, LoVo, RKO, SW48, HT29, and SW480 cells were 83%, 28%, 79%, 73%, 79%, and 29%, respectively.

We next analyzed the expression of miR-137 in the same panel of CRC cell lines using TaqMan RT-PCR, and we found an inverse correlation between the degree of CpG island methylation measured by pyrosequencing and the level of expression (Fig. 2A and B). Along with demethylation, treatment with 5-aza-CdR induced upregulation of miR-137 in CRC cell lines, suggesting that the expression of miR-137 is suppressed through CpG island promoter methylation in the majority of CRC cell lines.

We next used the miR-137 pyrosequencing assay to analyze the methylation status in a cohort of colorectal cancer tissues that included 21 normal mucosa from nontumor patients (N-N), 113 CRC tissues with their corresponding adjacent normal mucosa (N-C), and 68 colorectal adenomas. The mean level of methylation (±SD) in N-N, N-C, CRC, and adenomas was 7.70% (±4.18), 10.27% (±6.88), 31.67% (±17.25), and 33.09% (±21.06), respectively (Fig. 2C). Thus, the level of methylation was significantly higher in CRC tissues compared with their corresponding histologically normal mucosa (31.67% versus 10.27%, P < 0.0001), showing the cancer specificity of miR-137 methylation. In addition, the level of miR-137 methylation in N-N or N-C did not correlate with older age (data not shown). On the other hand, we observed a significantly higher degree of methylation in N-C compared with N-N (10.2% versus 7.7%, P = 0.0352), consistent with the paradigm of methylation-related field defects in CRC. Finally, and more interestingly, methylation of miR-137 in adenomatous tissues showed the same degree of methylation as CRC tissues (P = 0.8352), suggesting that methylation of this miRNA is an early event in colorectal carcinogenesis.

To determine whether miR-137 methylation might be different among the different molecular subtypes of CRC based on the presence of microsatellite instability (MSI), we next analyzed miR-137 methylation status in tumors and matching normal tissues (N-C) from 11 patients with Lynch syndrome, 14 patients with sporadic MSI, and 22 MSS patients. The mean levels of methylation in Lynch, sporadic MSI, and MSS were 22%, 27.3%, and 24.7%, respectively. The corresponding levels in N-C were 6.4%, 6.2%, and 7.3%, respectively. Accordingly, miR-137 methylation seems to be equally present in all molecular subtypes of CRC.

Because pyrosequencing analysis allows quantitative measurements, we next analyzed miR-137 methylation results as a categorical variable. We determined a methylation cutoff based on the average miR-137 methylation levels in normal mucosa from nontumor individuals plus 2 SD (cutoff, 15%). Based on these results, miR-137 was methylated in 4% (1 of 21) of N-N, 14.4% (16 of 111) of N-C, 81.4% (92 of 113) of CRC, and 82.3% (56 of 68) of adenomas. Clinicopathologic features of CRC and adenoma patients are listed in Table 1 and Supplementary Table S1, respectively. Patients with CRC-associated miR-137 methylation were significantly older than those without methylation (66 versus 60 years-old, P = 0.024), showed more frequent somatic KRAS mutations (37% versus 14.3%, P = 0.046), and were less frequently associated to mucinous features (2.2% versus 15%, P = 0.039). There was no association between methylation status and TNM stage. In adenoma patients, we observed a significantly higher degree of methylation in villous compared with nonvillous adenomas (44.28% versus 29.08%, P = 0.0051).

We used TaqMan RT-PCR to assess the expression of miR-137 in 15 pairs of CRC and normal colonic mucosa and found substantial downregulation of the expression in the tumor compared with the normal mucosa (Fig. 2D). Interestingly, the only two cases that did not show downregulation in the tumor tissue were those not associated with methylation (data not shown).

To investigate which specific cell types expressed miR-137 in the colon, we performed in situ hybridization using 5′-DIG–labeled LNA probes, a technique that has been applied to the detection of miRNA in situ in FFPE tissues (Fig. 3). In normal colonic mucosa, miR-137 was expressed only in the colonic epithelial cells throughout the colonic crypts. However, none of the adenomatous and CRC samples showed miR-137 expression, consistent with our observation that miR-137 is silenced in the majority of colonic neoplastic tissues.

Having discovered that miR-137 is epigenetically silenced in CRC through CpG island methylation, we next performed functional studies to determine whether miR-137 had in vitro tumor-suppressive features following transfection of miR-137 precursor into CRC cell lines. We performed BrdUrd incorporation assays after transfection of either miR-137 precursor or a negative control precursor in three different colon cancer cell lines. Restoration of miR-137 significantly reduced cell proliferation in HCT116 and RKO but not in SW480 (Fig. 5D), suggesting the specificity of miR-137 as a tumor suppressor in CRC.

To identify gene targets of miR-137, we first performed whole genome gene expression analysis in HCT116 cells after transfection of either miR-137 or negative control precursors. Four hundred ninety-one genes showed more than 2-fold decrease in their expression following miR-137 transfection, compared with the negative controls (Supplementary Table S3). We next used the miRecords resource (15) to obtain a list of predicted miR-137 targets. This bioinformatics approach integrates information of predicted miRNA targets produced by 11 established miRNA target prediction programs, thus providing a more accurate and comprehensive assessment of predicted targets compared with a single database. Selecting the predicted targets from at least 4 of the 11 prediction tools included in the website, we obtained a list of 505 potential targets (Supplementary Table S3). After cross-referencing the list of miR-137–related gene targets from our own gene expression microarray data and information gathered from the miRecords database, we determined that 32 genes met the criteria of downregulation by miR-137 transfection and being an in silico predicted target (Fig. 4A and B).

We first validated the microarray data by RT-PCR (Fig. 4C) for a subset of selected genes. Lysine (K)-specific demethylase 1A (KDM1A; also known as LSD1), chromosome segregation 1-like (CSE1L), Y box binding protein 1 (YBX1), semaphorin 4D (SEMA4D), and Aurora Kinase A (AURKA) were chosen for validation because of their putative role in carcinogenesis. Sequence interaction between the 3′UTR region of these genes and miR-137 is shown in Supplementary Fig. 1. RT-PCR results were highly concordant with microarray data for all selected genes. All of these genes contain well-conserved target sites in their 3′UTR region for miR-137, and Western blot analysis effectively showed different degrees of downregulation of expression in three of five selected genes after miR-137 precursor transfection (Fig. 4D).

We next explored the functional interaction between miR-137 and LSD1 (Fig. 5A–C). The expression levels of miR-137 and LSD1 mRNA were determined in six CRC cell lines by TaqMan RT-PCR and RT-PCR, respectively (Fig. 5B). In CRC cell lines with lower endogenous miR-137 expression (HCT116, LoVo, RKO, and SW48), a higher LSD1 expression level was observed, whereas in the HT29 cell line, which has higher expression levels of miR-137, the amount of LSD1 was much lower.

To determine if the 3′UTR region of LSD1 was indeed a functional target site of miR-137, a luciferase reporter plasmid harboring either wild-type or mutated predicted region of interaction at the LSD1 3′UTR region was constructed. The transient transfection of HCT116 cells with the wild-type reporter plasmid and the miR-137 precursor led to a decrease of luciferase activity in comparison with the control precursor (Fig. 5C). However, the luciferase activity remained unaffected following transfection with either the parental plasmid or the one with the mutated sequence.

In this study, we report that epigenetic silencing through promoter methylation of miR-137 is an early event in colorectal carcinogenesis. In situ hybridization analysis showed that this miRNA is constitutively expressed in the normal colonic epithelium and is silenced in neoplastic tissues. Gene expression analysis combined with in silico prediction tools allowed us to identify several potential targets of miR-137, including LSD1, a histone demethylase that plays a central role in the epigenetic machinery.

Our data indicate that methylation of the miR-137 promoter is an early event in colorectal carcinogenesis, supported by the fact that the average methylation level in adenomas was similar to that in CRC specimens (31.67% versus 33.09%, respectively; P = 0.8352). In addition, miR-137 methylation seems to be tumor specific because it preferentially occurs in neoplastic tissues and its methylation in normal colonic mucosa did not increase as a function of age. More interestingly, we found a higher degree of average methylation in histologically normal colorectal mucosa from CRC patients compared with the normal mucosa from nontumor patients (10.27% versus 7.7%, respectively; P = 0.0352), consistent with the epigenetic field defect proposed in CRC (16). The rationale for this phenomenon relies on the fact that sporadic CRCs are thought to arise from a region of cells characterized by a “field defect” (17). DNA methylation of multiple genes has been proposed as a major contributor to the field defect hypothesis (16); however, only two previous studies have suggested that miRNA methylation may have an analogous manifestation in the gastrointestinal tract. Grady and colleagues found that miR-342, an intronic miRNA encoded within the EVL gene, was methylated in 56% of normal-appearing colonic mucosa from CRC patients and in six of nine adenomas (10). In gastric cancer, Ando and colleagues recently found that methylation of miR-124a1-3 shows a similar pattern in the normal gastric mucosa from tumor and healthy subjects (18). We must mention that tissue specimens for normal colonic mucosa from CRC patients and healthy subjects were obtained from two different populations from Japan and Spain, respectively. Consequently, the differences in miR-137 methylation in these tissues could be attributed to other factors, such as ethnicity or other environmental factors that require further analysis to definitively establish the field defect associated with miR-137 methylation in the colon. Based on our findings for cancer specificity, the high level of methylation in colonic adenomas, and the potential field defect feature, the methylation status of miR-137 may serve as a potential noninvasive biomarker for CRC.

Although several studies have shown that miRNA expression profiles in CRC are significantly different than the normal colonic mucosa, it is unclear which specific cell types within the colon express various miRNAs. Elucidation of this aspect is essential to our understanding of the biological function of miRNAs in carcinogenesis, as these are known to have different roles in different cell types, and the same miRNA can act as a tumor suppressor or an oncogene depending on the tissue and cell compartment (2). In situ hybridization in FFPE tissues has been recently used for miRNA detection with great success for different cancers (19). We found that miR-137 is preferentially expressed in the epithelial cells of normal colonic mucosa, whereas no miR-137 expression was observed in any of the adenomatous polyps and CRC tissues. These findings are of significance and are consistent with our gene expression data that showed downregulation of miR-137 in CRC relative to normal mucosa.

Given the evidence that miR-137 is commonly silenced in CRC cell lines and CRC tissues due to promoter methylation, we performed functional studies to explore the potential tumor suppressor features of miR-137 in vitro. We discovered that the restoration of miR-137 expression in HCT116 and RKO cells, two cell lines with lower constitutive miR-137 expression, indeed resulted in a significant decrease in proliferation. Although further studies are needed, our results suggest that miR-137 may act as a tumor suppressor miRNA in CRC.

A critical and challenging step in understanding miRNA function is the identification of its gene targets. Lack of reliable and specific methods for target identification limits our understanding of the biological role of miRNAs. Computational methods, the most commonly used approach, are based on base pairing between the miRNA seed region (first 2–8 bases of the mature miRNA) and the target, but suffer from specificity issues because predicted targets are usually in the hundreds or thousands, making it difficult to ascertain the authentic targets of a given miRNA. Whereas miRNAs were believed to act mainly through translational inhibition rather than mRNA cleavage, it is increasingly being realized that miRNAs may downregulate a much larger number of transcripts than previously appreciated (7, 20, 21). Thus, gene expression microarray analysis has been proposed as a useful strategy to identify physiologically relevant miRNA:mRNA interactions. In this study, candidate miR-137 targets were selected using a combination of bioinformatic and transcriptomic approaches. We used stringent criteria to decrease the false positivity rate and identified 32 potential targets that were downregulated after miR-137 transfection at the mRNA level. These same gene targets were also predicted targets by at least four of the available computational-based prediction tools. It was encouraging to observe that our protein expression analysis revealed that three of the five selected genes also showed corresponding downregulation of protein expression following miR-137 transfection, thus reinforcing the value of this strategy.

Interestingly, we found that LSD1, a histone demethylase, is a target of miR-137. LSD1 is part of a new class of histone demethylating enzymes that, in addition to demethylating H3K4 and H3K9 (22, 23), are essential for the maintenance of global DNA methylation through demethylation of a nonhistone substrate, DNMT1, by increasing its stability (24). Because many human cancers show increased expression of DNMT1, it is plausible to speculate that LSD1, which is often upregulated in cancer cells, might be partly responsible for this epigenetic defect. Overexpression of LSD1 has been documented in prostate cancer (22) and in neuroblastoma (25), where LSD1 is involved in maintaining the undifferenciated phenotype and inhibition of its function inhibits tumor xenograft growth. In line with this argument, we found an inverse correlation between miR-137 and LSD1 in CRC cell lines and validated this functional interaction using luciferase experiments. Because LSD1 seems to play a central role in the epigenetic machinery and has a potential role in cancer therapy (26), future experiments will reveal additional roles of LSD1 in CRC carcinogenesis and its relationship with modulation of miR-137 expression.

In summary, this study reports that miR-137 acts a tumor suppressor in the colon and is frequently silenced in CRC through promoter hypermethylation and that its restoration inhibits cell proliferation in vitro. Because miR-137 methylation is an early event in colorectal carcinogenesis, the potential use of miR-137 methylation as a CRC biomarker is very promising. We have identified multiple potential mRNA targets of miR-137, including LSD1 in CRC, which provides novel evidence for the cross-talk between miRNAs and other components of the epigenetic machinery. These findings raise the possibility that in the future, miR-137 precursors may have potential therapeutic value in CRC patients.

No potential conflicts of interest were disclosed.

Grant Support: National Cancer Institute, NIH, grants R01 CA72851 and CA129286 and funds from the Baylor Research Institute (C.R. Boland and A. Goel), Fundación Alfonso Martín Escudero (F. Balaguer), and Societat Catalana de Digestologia (F. Balaguer).

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.