MicroRNA-31 Is a Transcriptional Target of Histone Deacetylase Inhibitors and a Regulator of Cellular Senescence*

Background: Histone deacetylase inhibitor (HDACi) inhibits expression of polycomb group proteins and can differentially regulate expression of miRNAs. Results: HDACi up-regulated expression of miR-31, which down-regulated BMI1 and induced cellular senescence. Conclusion: miR-31 is a novel target of HDACi, and is a novel regulator of cellular senescence. Significance: HDACi can be used to induce miR-31, which in turn can cause senescence in cancer cells. MicroRNAs (miRNAs) have emerged as important regulators of tumorigenesis. Several miRNAs, which can function either as oncomiRs or tumor suppressive miRs are deregulated in cancer cells. The microRNA-31 (miR-31) has been shown to be overexpressed in metastatic breast cancer. It promotes multiple oncogenic phenotypes, including proliferation, motility, and invasion of cancer cells. Using a breast cancer-related miRNA array analysis, we identified miR-31 as a novel target of histone deacetylase inhibitors (HDACi) in breast cancer cells. Specifically, we show that sodium butyrate (NaB) and panobinostat (LBH589), two broad-spectrum HDAC inhibitors up-regulate hsa-miR-31 (miR-31). The up-regulation of miR-31 was accompanied by repression of the polycomb group (PcG) protein BMI1 and induction of cellular senescence. We further show that inhibition of miR-31 overcomes the senescence-inducing effect of HDACi, and restores expression of the PcG protein BMI1. Interestingly, BMI1 also acts as a repressor of miR-31 transcription, suggesting a cross-negative feedback loop between the expression of miR-31 and BMI1. Our data suggest that miR-31 is an important physiological target of HDACi, and that it is an important regulator of senescence relevant to cancer. These studies further suggest that manipulation of miR-31 expression can be used to modulate senescence-related pathological conditions such as cancer, and the aging process.

MicroRNAs (miRNA), 3 which are evolutionarily conserved small RNA molecules of 19 -24 nucleotides in length, have recently emerged as major regulators of cancer (1,2). miRNAs control expression of target genes via base pairing to seed sequences that are found in the 3Ј untranslated region of a particular target gene. In the context of tumorigenesis, they can function either as oncogenes (onco-miRs) or tumor suppressors (3,4). Many miRNA profiling studies have documented miRNAs that are either significantly reduced or overexpressed in breast cancer cells (5,6). For example, miR-21, miR-155, miR-10b, and miR-29 are overexpressed and are described as oncomiRs, whereas let-7, miR-200 family, miR-125a/b, miR-206, miR-17, miR-34a, and miR-31 have been reported to function as tumor suppressive miRs (5,6). The miR-31, a pleiotropically acting miRNA was cloned as a breast cancer metastasis suppressor (7). It can target several cancer and metastasis relevant targets such as Fzd3, ITGA5, MMP16, RDX, and RhoA (7). In addition, miR-31 was recently shown to target NF-B inducing kinase and to be negatively regulated by the polycomb group (PcG) protein EZH2 in adult T cell leukemia cells (8). The PcG protein EZH2 is a constituent of polycomb repressive complex (PRC) 2, which is thought to work in concert with PRC1 (9,10). Another important PcG protein, BMI1, is one of the main constituents of PRC1, which promotes H2A monoubiquitination. Both BMI1 and EZH2 are overexpressed in various types of cancers, including breast, prostate, and colon cancers (11)(12)(13), and are known to promote oncogenic phenotype in in vitro and in vivo models of cancer progression (14 -17). The PcG protein BMI1 is known to regulate cellular senescence via repression of the tumor suppressor p16INK4a (herein referred to as p16) (18,19). Cellular senescence acts as a strong tumor suppressor mechanism, and is controlled by p53-p21 and p16-pRB pathways (20). In addition to cellular senescence, BMI1 also promotes cancer stem cell phenotype and therapy resistance in cancer cells (11,21). In addition to p16, BMI1 is known to regulate expression of other cancer and aging relevant genes, such as p57, and genes involved in TGF-␤ signaling, endoplasmic reticulum stress, and WNT pathways (22)(23)(24)(25). Because of their role in promoting various cancers, inhibitors of BMI1 and EZH2 are of clinical importance. Recently, we and others have shown that the expression of the PcG proteins is inhibited by histone deacetylase inhibitors (HDACi) (22). HDACi transcrip-tionally regulate expression of genes involved in proliferation control and tumorigenesis, which results in inhibition of the oncogenic phenotype and induction of cell cycle arrest, autophagy, cell death, differentiation, and cellular senescence in cancer cells (26). The exact mechanism of the inhibition of PcG proteins by HDACi is not known. HDACi can also differentially regulate expression of miRNAs in cancer cells (27)(28)(29)(30). For example, miR-125a, miR-125b, and miR-205 are known to be up-regulated by HDACi in breast cancer cells (31). We have reported that sodium butyrate (NaB), a widely used HDACi can up-regulate miR-141 in human diploid fibroblasts (HDFs) (32). In this study, using a miRNA-PCR array, we determined which of the candidate miRNAs are regulated by HDACi in breast cancer cells that may be relevant to cellular senescence, a tumor suppressive phenotype induced by HDACi. We report that miR-31 is a novel target of HDACi in breast cancer cells. We further studied whether miR-31 regulates expression of PcG proteins and cellular senescence, and determined whether HDACi induce senescence via up-regulation of miR-31.

EXPERIMENTAL PROCEDURES
Cells, Cell Culture Methods, and Reagents-The breast cancer cell lines MDA-MB-231 and MCF7, and 293T cells were obtained from the American Type Culture Collection (ATCC) (Manassas, VA). The cells were cultured as described previously (23,32). The MRC5 strain of HDFs was obtained from the NIA Aging Cell Repository (Coriell Institute for Medical Research, Camden, NJ), and cultured as described (23,32). HDAC inhibitor LBH-589 was obtained from Selleckchem (Houston, TX), and sodium butyrate (NaB) was from Sigma. N-Acetyl-L-cysteine (NAC) was also obtained from Sigma. HDACi and NAC were dissolved in dimethyl sulfoxide and added to the cell culture medium as described (22).
Antibodies and Western Blot Analyses-Western blot analyses were done using specific antibodies as described previously (22,34). Monoclonal antibodies (mAb) against p53, p21, and p16, and a polyclonal antibody (pAb) against pRB were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and have been described previously (32). The BMI1 mAb was from Invitrogen. The ␤-actin mAb was from Sigma. Polyclonal antibodies against total H2A, H3, and H4 were from Cell Signaling (Danvers, MA). For ChIP analysis, pAbs against H2AK119Ub and H3K27Me3, and acetylated H3 and H4 were also from Cell Signaling (Danvers, MA). The densitometric quantification of signal for each protein in the Western blot was performed using ImageJ (NIH, Bethesda, MD) software.
miRNA Array, Quantitative RT-PCR, and ChIP Analyses-A breast cancer miRNA PCR array (miScript miRNA PCR Array, MIHS-109Z), which probes 84 breast cancer-related miRNAs, was purchased from Qiagen (Valencia, CA). The array set was probed with total RNA isolated from mock (dimethyl sulfoxide)-and NaB-treated (4 mM, 48 h) MDA-MB-231 cells, and the real time qPCR results were quantified using data analysis software as recommended by the manufacturer (Qiagen, Valencia, CA). The real time RT-PCR (qRT-PCR) was performed as described (32). Briefly, total RNA was isolated using TRIzol reagent as described by manufacturer (Invitrogen), and treated with DNase (Promega). For miRNA qRT-PCR, the specific primers for miR-31 and cDNA synthesis kit were from Quanta Biosciences (Gaithersburg, MD). The PCR conditions consisted of an initial activation at 50°C for 2 min, 95°C for 20 s, followed by 40 cycles of 95°C for 1 s, and 60°C for 20 s in Step One Plus Real-Time PCR system (Applied Biosystems). The C t (threshold cycle) value of each primers was normalized to that of RNU6B for miRNA or GAPDH as internal control. For qRT-PCR of miR-31 targets, the specific primers listed in Table 1 were used. The chromatin immunoprecipitation (ChIP) assays were performed as described (23,34). The immunoprecipitated chromatin was amplified using 4 different sets of miR-31 promoter-specific primers (Table 2) by qPCR as described above.
Statistical Analysis-All experiments were performed at least three times in triplicates for each group. The results are presented as the mean Ϯ S.D. Statistical significance was determined using Student's t test, and p Ͻ 0.05 was considered significant.

HDAC Inhibitors Transcriptionally Up-regulate miR-31-To
probe a hypothesis that HDACi may up-regulate growth inhibitory miRNAs, we analyzed a RT-PCR array that is designed to probe simultaneous expression of breast cancer-related 84 miRNAs, which are deregulated in breast cancer. These miR-NAs are relevant to other cancers as well and may control various steps in oncogenesis including early steps such as bypass of senescence and acquisition of replicative immortality. We treated MDA-MB-231 cells with two HDACi, NaB and LBH589, and confirmed up-regulation of H3 and H4 acetylation as well as down-regulation of PcG proteins BMI1 and EZH2 (Fig. 1A). Next, we analyzed the breast cancer miRNA array using RNA isolated from NaB-treated cells. The array analysis indicated that whereas the majority of the miRNAs are down-regulated or not affected, 18 miRNAs are up-regulated ( Fig. 1B). Several of these miRNAs, such as miR-125a, miR-FIGURE 1. HDACi transcriptionally up-regulate miR-31. A, the total cell lysates were prepared from mock and HDACi-treated cells and the expression of BMI1, EZH2, total H2A, total H3, H3-Ac, total H4, H4-Ac, and ␤-actin was detected by Western blot analysis using a specific antibody against each protein. B, heat map of the gene expression profile of the selected miRNAs in control versus HDACi (NaB)-treated MDA-MB-231 cells. Total RNA was extracted with a RNA purification kit (Qiagen), and 10 ng of RNA was reverse transcribed using High Capacity cDNA RT kits (Qiagen). Real time PCR was performed using a StepOnePlus RT-PCR system (Applied Biosystems) with TaqMan Gene Expression Master Mix (Qiagen) as suggested by the manufacturer. Gene expression data analysis and visualization of differentially expressed genes were conducted using online Data Assist software from Qiagen-SABioscience. Unsupervised hierarchical cluster analysis was performed based on the C t values using Pearson's correlation as a measure of similarity, and Average Linkage and Global View were used for generating a heat map as described by the manufacturer (Qiagen-SABioscience). Only HDACi up-regulated miRNAs are shown in the adjacent graph. The arrow indicates hsa-miR-31-5p that was further chosen for the study.

TABLE 2
Primer sets for ChIP assay 125b, miR-205, miR-141, and miR-200c, are known to be upregulated by HDACi (31,32). One of the miRNAs, miR-31 that was up-regulated by NaB in MDA-MB-231 is a pleiotropic miRNA and a well known suppressor of breast cancer metastasis (7,39). As miR-31 is not previously reported to be regulated by HDACi, here, we focused on HDACi regulation of miR-31. First, up-regulation of miR-31 expression was confirmed by qRT-PCR of NaB-and LBH589-treated MDA-MB-231 cells (Fig. 1C). To further confirm transcriptional regulation of the miR-31 encoding gene, we cloned 671 bp upstream of miR-31 transcription start as miR-31 promoter into the pGL4.18 luciferase reporter vector and determined whether the promoter activity is up-regulated by HDACi. Results of the promoterreporter assay confirmed that miR-31 is transcriptionally upregulated by NaB (Fig. 1D).
Because HDACi led to up-regulation of acetylated H3 and H4, as well as down-regulation of the PcG proteins, we determined a time course of up-regulation of miR-31 and its correlation with changes in expression of acetylated histones and the PcG proteins. The cells were treated with NaB for 3-48 h and analyzed for the expression of Ac-H3, Ac-H4, BMI1, and EZH2 by Western blot analysis. We also analyzed expression of miR-31 and BMI1 by qRT-PCR analysis. The results indicated that significant up-regulation of miR-31 occurs after 12 h treatment of cells with NaB, which correlated with down-regulation of BMI1 by Western blot analysis and RT-PCR analysis (Fig.  1E). The data also indicated that up-regulation of Ac-H3 and Ac-H4 can be detected as early as 3 h of HDACi treatment. These results suggest that HDACi treatment possibly up-regulates miR-31 via down-regulation of BMI1. Similar results were obtained with LBH589 (not shown).
Next, we analyzed expression of miR-31 in control, BMI1 overexpressing, and BMI1 knockdown MCF10A cells. The results indicated that miR-31 is down-regulated in BMI1 overexpressing cells, whereas it was up-regulated in MCF10A cells expressing a BMI1 shRNA (BMI1 knockdown cells) ( Fig. 2A). The knockdown of BMI1 also up-regulated miR-31 expression in MRC5 cells (Fig. 2B). We also determined whether BMI1 can negatively regulate miR-31 by measuring luciferase activity driven by the miR-31 promoter. The promoter-reporter construct was co-transfected with vector, BMI1 expression vector pcDNA-HA-BMI1, or pRS-BMI1 shRNA (32), and luciferase activity was determined 48 h after transfection. The results indicated that BMI1 overexpression led to a decrease in promoter activity, whereas BMI1 knockdown increased miR-31 promoter activity in a dose-dependent manner suggesting that BMI1 negatively regulates miR-31 expression (Fig. 2C). To further confirm that miR-31 up-regulation by HDACi is indeed due to down-regulation of BMI1, we generated MDA-MB-231 cells overexpressing retroviral long terminal repeat-driven BMI1. The control (B0) and BMI1 overexpressing MDA-MB-231 cells were either mock-treated or treated with HDACi, and the expression of miR-31 was determined by qRT-PCR analysis. The results indicated that NaB and LBH589 did not up-regulate miR-31 in BMI1 overexpressing cells (Fig. 2D).
As HDACi treatment led to down-regulation of BMI1 and EZH2, and up-regulation of acetylated histones, we surmised whether differential expression of these proteins results in their increased or decreased occupancy at the miR-31 promoter. To examine this possibility, we performed a ChIP assay using antibodies specific to BMI1, EZH2, SUZ12, H3K27me3,

Regulation of miR-31 by HDAC Inhibitors
H2AK119Ub, acetylated H3, and acetylated H4. The immunoprecipitated chromatin was amplified by qPCR using 4 sets of primers that cover 671 bp upstream of the miR-31 promoter (Fig. 3). The results indicated that the maximum binding of these factors was associated with region 4, which is next to the transcription start (Fig. 3). More importantly, the results indicated that the binding of BMI1, EZH2, H3K27me3, and H2AK119Ub was significantly decreased in all four regions in NaB-treated cells, whereas acetylated H3 and H4 exhibited increased binding to all four regions (Fig. 3). Taken together our data suggest that HDACi treatment via down-regulation of the PcG proteins, in particular BMI1, leads to up-regulation of miR-31. We also determined whether the expression of miR-31 and BMI1 negatively correlates in breast cancer cell lines. The results of qRT-PCR analysis of BMI1 and miR-31 indicated that indeed there is a negative correlation (Pearson's coefficient r ϭ Ϫ0.329, p Ͻ 0.0001), between the expression of BMI1 and miR-31 in selected breast cancer cells lines (Fig. 4).
miR-31 Regulates Cell Proliferation and Senescence-The miR-31 has been shown to regulate several oncogenic properties, such as migration, invasion, and metastasis of breast cancer cells. Its role in early steps of cancer such as bypass of senes-cence is not known. Moreover, HDACi are known to induce senescence. Hence, we hypothesized that miR-31 may regulate cell proliferation and senescence. To probe this hypothesis, we analyzed expression of miR-31 in proliferating early passage and late passage (senescent) MRC5 cells by qRT-PCR analysis. The results indicated that miR-31 is up-regulated in senescent HDFs (Fig. 5A). The up-regulation of p53, p16, and p21, and relative abundance of unphosphorylated pRb (known molecular markers of senescence), and down-regulation of BMI1 was confirmed by Western blot analysis (Fig. 5A, right panel).
Next, we stably expressed miR-31 pre-miRNA and miR-31 inhibitor in the MRC5 strain of HDFs, which are widely used in senescence research. The cells were selected in puromycin and hygromycin, respectively, and analyzed for cell proliferation, expression of BMI1, and induction or abrogation of senescence using markers of senescence. The cell proliferation assay indicated that miR-31 overexpression inhibits proliferation, whereas miR-31 inhibition promotes cell proliferation (Fig. 5B). Accordingly, we found that miR-31 overexpression leads to repression of BMI1 and its inhibitor up-regulates BMI1 (Fig.  5C). We also found that miR-31 down-regulates expression of EZH2 and cyclin D1 (Fig. 5C). To examine the expression of known targets of miR-31, we also performed qRT-PCR analysis of cells expressing miR-31 pre-miRNA or its inhibitor. The results indicated that most of the described targets of miR-31 are down-regulated upon its overexpression except for RHOA and TFDP1 in MRC5 cells (Fig. 5D).
As miR-31 is up-regulated in senescent cells, we hypothesized that its overexpression may cause premature senescence in HDFs. To support this hypothesis, we analyzed vector control and miR-31 overexpressing cells for the markers of cellular senescence. First, we determined the expression of p53, pRb, p21, p16, and BMI1 in these cells. The result of Western blot analysis indicated that miR-31 overexpression led to an increase in p53, p16, p21, and unphosphorylated pRb levels (Fig. 6A). miR-31 overexpression also led to significant downregulation of BMI1 (Fig. 6A). Next, cells were analyzed by SA-␤gal and EdU co-staining to determine whether miR-31 overexpression induced senescence by increasing the percentage of SA-␤-gal positive cells and corresponding decrease in EdU positive cells (proliferating cells). The data showed that indeed, miR-31 overexpression increased SA-␤-gal positive cells and decreased EdU positive cells indicating that miR-31 is a potent inducer of cell senescence (Fig. 6B). The role of miR-31 in senescence was further confirmed by examining ␥H2AX foci formation, which is another marker of cell senescence. The data indicated that miR-31 overexpression leads to an increase in ␥H2AX foci formation in MRC5 cells (Fig. 6C). Induction of senescence by miR-31 overexpression was also confirmed in MDA-MB-231 and MCF7 cells (Fig. 6D). Because these cell lines do not express p16, and MDA-MB-231 express mutant p53, we hypothesized that senescence induction in these cell lines may depend on p21 and not p53, p16, or pRb. Results indicated that miR-31 overexpression indeed led to p21 induction (Fig. 6E). Collectively, these results strongly suggest that miR-31 is a physiological regulator of cell senescence.
HDAC Inhibitors Regulate Proliferation and Senescence via miR-31-As miR-31 is induced by HDACi and these agents exhibit anti-tumorigenic activities, including induction of cellular senescence, we determined whether induction of cellular senescence by HDACi is mediated via induction of miR-31. To examine this possibility, we generated MRC5 cells stably expressing a control and miR-31 inhibitor, and determined the proliferation and senescence of these cells with or without HDACi treatment. The cells were mock, NaB, and LBH589 treated and the expression of miR-31 was determined by qRT-PCR. The results indicated that the induction of miR-31 by NaB and LBH589 was much less in miR-31 inhibitor expressing cells (Fig. 7A). Next, the control and miR-31 inhibitor expressing cells were examined for cell proliferation and expression of the molecular markers of cellular senescence. The results of the Western blot analysis showed that NaB and LBH589 induced p53, p21, and p16, and reduced comparative levels of phosphorylated pRb in control but not in miR-31 inhibitor expressing cells (Fig. 7B). The data also indicated that the miR-31 inhibitor overcomes the proliferation inhibitory effect of NaB and LBH589 in MRC5 cells, and that the miR-31 inhibitor promotes cell proliferation (Fig. 7C).
Next, we examined the ability of HDACi to induce senescence in control and miR-31 inhibitor expressing MRC5 cells using SA-␤-gal/EdU co-staining and ␥H2AX foci formation assay. The results indicated that NaB and LBH589 induced a robust senescent phenotype in control but not in miR-31 inhibitor expressing cells as indicated by an increase in SA-␤-galand decrease in EdU-positive cells (Fig. 8A). There was a substantial decrease in SA-␤-gal-positive cells in miR-31 inhibitor expressing cells, indicating that miR-31 expression can partially overcome the senescence-inducing effect of HDACi in MRC5 cells (Fig. 8A). Similar results were obtained when cells were immunostained for ␥H2AX to determine whether NaB and LBH589 increased ␥H2AX foci formation in miR-31 inhibitor expressing cells. The results indicated that NaB and LBH589 increased ␥H2AX foci formation in control but not in miR-31 inhibitor expressing cells (Fig. 8B). Taken together, our data suggest that the HDACi-induced senescence is mediated via up-regulation of miR-31 and that its abrogation can overcome HDACi-induced senescence. Thus, miR-31 is a physiological target of HDACi and an important regulator of cellular senescence.

FIGURE 8. miR-31 inhibition overcomes senescence induction by HDAC inhibitors.
A, mock, NaB-, or LBH589-treated control (Ctrl IH) and miR-31 inhibitor (miR-31 IH), expressing MRC5 cells were studied for senescent phenotype using SA-␤-gal/EdU co-staining as described in the legend to Fig. 6. B, MRC5 cells with or without miR-31 inhibitor were mock, NaB-, or LBH589-treated, and analyzed for ␥H2AX foci formation as described in the legend to Fig. 6. The experiment was done in triplicate. Error bars represent Ϯ S.D., *, p Ͻ 0.05; NS, p Ͼ 0.05. BMI1 down-regulation. We generated MDA-MB-231 cells overexpressing BMI1 using pMSCV-Hygro vector. We then overexpressed miR-31 in control and exogenous BMI1 expressing MDA-MB-231 cells. The resulting set of cells were either mock treated or treated with HDACi (NaB and LBH589), and analyzed for the expression of BMI1 and markers of senescence (Fig. 9). The data indicated that miR-31 overexpression does not induce senescence in exogenous BMI1 overexpressing cells (Fig. 9B). The Western blot analysis also indicated that HDACi treatment did not induce p21 in exogenous BMI1 overexpressing MDA-MB-231 cells (Fig. 9A), confirming the earlier observation that p21 is the primary mediator of senescence in MDA-MB-231 cells.
miR-31 Overexpression Induces mtROS-As miR-31 overexpression leads to ␥H2AX foci formation, which is indicative of DNA damage and a well known marker of cellular senescence, next, we determined the possible mechanism of increased ␥H2AX foci formation in miR-31 overexpressing cells. The PcG protein BMI1 has been shown to regulate mtROS and play a role in DNA damage (40 -42). mtROS is known to induce DNA damage and cellular senescence, hence we hypothesized that miR-31 overexpression may lead to increased mtROS, leading to DNA damage and induction of cellular senescence. To test this hypothesis, we analyzed control, miR-31 overexpressing, and BMI1 and miR-31 co-overexpressing MDA-MB-231 cells for mtROS using MitoTracker Red CM-H 2 XROS dye. We also analyzed cells that were treated with a generic ROS scavenging agent NAC. The results indicated that miR-31 overexpression in control but not in exogenous BMI1 overexpressing cells led to a significant increase in mtROS (Fig. 10A). Furthermore, our results indicated that treatment of cells with NAC significantly reduced mtROS and ␥H2AX foci formation in miR-31 overexpressing cells (Fig. 10, A and B).

DISCUSSION
PcG proteins are important regulators of proliferation, stem cell phenotype, and senescence. Aberrant expression of BMI1 is associated with the bypass of senescence, increased proliferation, and oncogenic phenotypes such as increased migration, invasion, and metastasis of cancer cells (14,15,18,25,43). BMI1 is also known to promote CSC properties and therapy resistance in breast and prostate cancers (17,44). Hence, the therapeutic targeting of BMI1 can potentially help in the prevention, treatment, and recurrence of breast, prostate, and possibly other cancers. We and others have shown that BMI1 and PcG proteins can be targeted by pharmacological inhibitors of HDACs (22,45). The mechanism of targeting of PcG proteins by HDACi is not very well understood. The studies reported here suggest a novel mechanism of regulation of expression of PcG proteins, in particular BMI1 via induction of miRNAs. As reported in previous studies, HDACi modulated expression of several miRNAs that are known to be differentially expressed in cancer cells. Interestingly, most of the miRNAs that are up-regulated by HDACi appear to be silenced in cancer cells via promoter methylation such as miR-141, miR-200c (46,47), miR-205 (48,49), and miR-31 (50 -52) suggesting a cross-talk between HDACi and DNA methylation. These miRNAs are likely to function as tumor suppressors and regulate cellular senescence, which is a natural tumor suppressor mechanism (20). Along these lines, we have shown that miR-141 and miR-200c are important senescence-regulatory miRNAs (32). Interestingly, these miRNAs are also induced by various pharmacological agents that induce senescence such as HDACi and BI 2536, a PLK1 inhibitor (32,53). As miR-31 is described as a tumor suppressive miRNA in breast and prostate cancers, and is induced by HDACi, we focused our studies in defining its regulation by HDACi and the potential novel role of miR-31 in cellular senescence.
Previously it has been reported that miR-31 is negatively regulated by EZH2 and PRC2 (8,50). It has been also reported that the miR-31 expression is lost in melanomas, and that its genetic and epigenetic loss in cancers, in particular melanoma may lead to a feed-forward up-regulation of EZH2 (50). Because PcG proteins form PRCs and function in concert, we were interested to determine whether similar to EZH2, BMI1 is also negatively regulated by miR-31, and if so, whether it can lead to the induction of cellular senescence. Furthermore, both BMI1 and EZH2 are down-regulated by HDACi. Hence, it is possible that miR-31, which we showed is induced by NaB and LBH589, is a potent inhibitor of BMI1. Indeed, further mechanistic dissection of the possible mechanism of HDACi-induced senescence suggests that miR-31 negatively regulates expression of both BMI1 and EZH2. Negative regulation of BMI1 and EZH2 by miR-31 appears to be indirect, as none of the miRNA target prediction bioinformatics programs predict the presence of complementary seed sequences of miR-31 in BMI1 or EZH2 mRNAs. miR-31 is a pleiotropically acting miRNA with more than 200 predicted mRNA that can be targeted by miR-31 (7,8,50,54). We experimentally confirmed down-regulation of several possible targets of miR-31, such as NF-B inducing kinase, ITGA5, SEPHS1, RSBN1, and TFDP1. In addition, SRC and MET kinases are also described as possible targets of miR-31 (50). It is possible that some of these miR-31 target(s) regulate expression of BMI1. We favor the possibility of miR-31 regulating BMI1 via down-regulation of NF-B inducing kinase, which will lead to up-regulation of NF-B activity that has been shown to positively regulate BMI1 expression (55). However, it is possible that BMI1 regulation involves other novel targets of miR-31. Because miR-31 regulates senescence, it is also possible that a subset of miR-31 targets play a role in regulation of cellular senescence. While further studying the possible mechanism of miR-31-induced senescence, our results indicated that miR-31 is a strong inducer of mtROS, which may lead to induction of cellular senescence. Consistent with the role of BMI1 in suppressing mtROS, BMI1 overexpression leads to a decrease in mtROS and abrogation of senescence induction by miR-31. Along these lines, antioxidant NAC also leads to abrogation of miR-31-induced mtROS and cellular senescence suggesting that antioxidants can inhibit tumor suppressor function and in some cases promote oncogenic phenotypes by suppressing senescence. These data are consistent with the findings that support the oncogenic role of antioxidants (56).
The tumor suppressor role of miR-31 is context dependent (57). In the case of adult T cell leukemia, melanoma, breast, and prostate cancers, it is down-regulated or deleted and clearly acts as a tumor suppressive miRNA, whereas in other cancers such as colorectal cancers, it may act as an oncomiR (57). At this point, it is not clear what determines the context-dependent tumor suppressor or oncogenic role of miR-31. It is possible that some of the downstream oncogenic targets of miR-31 are differentially expressed in different types of cancers or miR-31 could target cell type-specific tumor suppressors in certain cell types, such as colorectal cancer or lung cancer cells. In these cases, overexpression of miR-31 could conceivably promote oncogenesis and function as an oncomiR. Indeed it is known that miR-31 can target tumor suppressors LATS2 and PPP2R2A in lung cancer cells (58). Nonetheless, a vast majority of miR-31 targets are pro-oncogenic factors including cell cycle promoting factors such as E2Fs (52), oncogenic tyrosine kinases such as MET and SRC (50), and genes that are up-regulated in metastatic cancer cells (7,8,50,54).
Although, for some cancer types, where miR-31 functions as an oncomiR, the HDACi regulation of miR-31 may not be relevant, the regulatory network described here is clearly important for breast, prostate, and other cancers, where miR-31 is unambiguously shown to function as a tumor suppressive miRNA. Up-regulation of miR-31 and induction of senescence is important for the physiology of HDAC inhibitory agents, including pharmacological inhibitors, many of which are in clinical trials and have shown great potential in cancer treatment. Similar to the EZH2 feed forward model (50), where miR-31 loss promotes EZH2 expression, our data clearly suggest that miR-31 negatively regulates BMI1 expression. Taken together, our data suggest a model, whereby miR-31 inhibits expression of PcG protein BMI1 (studies reported here) and EZH2 (50), and these PcG proteins in turn act as repressors of the hsa-miR-31 (MIR31HG) gene, which encodes miR-31 (Fig.  11). We speculate that a similar negative feedback loop model regulates expression of other HDACi-induced miRNAs such as miR-141/200c, which share similar properties, such as tumor suppressive functions including regulation of cellular senescence via targeting of expression of PcG protein BMI1, and are repressed in cancer cells via promoter methylation. In summary, our studies suggest that miR-31 is an important regulator of senescence and that miR-31-inducing agents including HDACi could be used to treat a subset of cancers where PcG proteins are aberrantly overexpressed.