Background
MicroRNAs (miRNAs) are a class of highly-conserved, noncoding 18-25- nucleotide RNAs that function as negative regulators of gene expression at the post-transcription level, binding to the 3′-untranslated region (3′-UTR) of mRNAs transcripts and targeting them for degradation [
1]. Though implicated in carcinogenesis, it is not clear how miRNAs promote tumorigenesis and metastasis or what networks regulate miRNAs expression. miRNA expression is commonly dysregulated in human cancers, [
2,
3] including esophageal cancers [
4].
miR-31 expression is altered in multiple human cancers. Depending on the cellular context, miR-31 may be up- or downregulated, acting as an oncogene or tumor suppressor, respectively. Overexpression of miR-31 has been linked to disease progression in colorectal cancer [
5], head-and-neck squamous cell carcinoma (HNSCC) [
6] and lung cancer [
7]. miR-31 is downregulated in certain T-cell leukemias [
8], breast cancer [
9,
10], melanoma [
11], ovarian cancer [
12] and prostate cancer [
13]. Downregulation and loss of miR-31 in esophageal adenocarcinoma (EAC) correlates with poor patient prognosis [
14-
16]. Additionally, miR-31 expression is reduced in EAC patients with poor histomorphologic response to neoadjuvant chemoradiation therapy [
17]. Conversely, miR-31 is upregulated in serum and tissue samples of esophageal squamous cell carcinoma (ESCC), with expression correlating to staging [
18]. Yet, in another ESCC cohort miR-31 expression was decreased, and low miR-31 expression correlated with poorly differentiated tumors and decreased survival [
19]. These reports emphasize the complexity of miR-31-associated phenotypes and the need to better define miR-31 targets, as well as pathways regulating miR-31 expression in different cancers.
SOX4 is a member of the highly conserved SoxC (SRY-related high-motility group box) transcription factors family, which contains two other members, SOX11 and SOX12 [
20]. SOX4 is a putative stem cell marker that plays a crucial role during cell fate determination [
21,
22]. SOX4-deficient mice suffer from multiple developmental defects, dying at embryonic day 14, secondary to ventricular outflow tract malformation [
23]. During embryogenesis, SoxC members are highly expressed, helping to maintain survival of pluripotent mesenchymal and neural progenitor cells [
24]. In adults, expression of SOX4 is restricted to certain cell types, including hematopoietic stem cells, mammary stem cells and hair follicle stem cells [
25-
27].
Meta-analysis has identified SOX4 as one of the 64 genes that constitute a general signature in all human cancers, and genome wide promoter analysis has shown that SOX4 regulates the transcription of genes involved in TGF-β, Wnt, Hedgehog, and Notch pathways and components of miRNA processing machinery such as Dicer, Argonaute 1 and RNA Helicase A [
28,
29]. SOX4 induces EMT and breast cancer progression by cooperating with oncogenic Ras. More recent work shows that SOX4 induces EMT via the polycomb epigenetic regulator EZH2 [
30]. miRNAs, such as miR-335, are known to target SOX4, suppressing metastasis and migration in breast cancer [
3].
Polycomb group proteins have been linked to tumor progression in many cancers. The polycomb proteins can form at least two complexes: polycomb-repressive complexes 1 and 2 (PRC1 and PRC2). PRC2 contains three core proteins, EZH2, SUZ12, and EED. The histone methyltransferase EZH2 (enhancer of zeste homolog 2) epigenetically regulates genes involved in cell fate determination. Specifically, EZH2 trimethylates nucleosomal histone H3 at lysine 27 (H3K27me3). The H3K27me3 mark is associated with gene silencing and often found in the promoter of developmental genes [
31,
32]. However, it is unclear how EZH2 is recruited to the promoters it targets. Recent studies have shown that EZH2 interacts with various transcription factors such as androgen receptor (AR), GATA4, RORα and STAT3 and may directly activate or repress these genes independent of the H3K27me3 mark or chromatin modification [
33-
36]. With respect to miRNAs, prior work demonstrates that EZH2 interacts with AR to silence miR-31 in prostate cancers, and C-MYC recruits EZH2 to the miR-29 promoter in B-cell lymphomas [
37].
EZH2 is upregulated in multiple cancers, promoting invasion and metastasis [
38-
40]. Genetic and epigenetic loss of miR-31 is associated with EZH2 overexpression in melanoma [
11], suggesting that miR-31 directly or indirectly regulates EZH2 expression. Interestingly, studies show that polycomb complexes silence the CDKN2A and CDKN2B loci, which encode the tumor suppressors p14 (ARF), p15 (INK4B) and p16 (INK4A) [
41] and contain the MIR31HG locus on chromosome 9 [
42]. In line with this observation, Yamagishi
et al. reported that PRC2 binds the miR-31 coding region and directly represses transcription of miR-31 in adult T-cell leukemia [
8]. SOX4 positively regulates EZH2, indicating a potential functional link between miR-31, EZH2 and SOX4.
The roles of SOX4, HDAC3 and EZH2 in microRNA regulation are largely unknown and have been poorly defined so far. In this study, we explore the role of SOX4 and EZH2 in miR-31 repression and the contribution of miR-31 to survival, migration and invasion of aggressive esophageal cancers cells. We identify SOX4 as a direct target of miR-31. Expression of miR-31 inhibits SOX4, EZH2 and HDAC3 expression. We show that miR-31 is repressed in invasive esophageal cancers cell lines and that miR-31 levels inversely correlate with SOX4, EZH2 and HDAC3 expression. Co-immunoprecipitation demonstrates that SOX4 interacts with EZH2 and that HDAC3 may be important to bridge this interaction. We show that EZH2 and HDAC3 bind to the miR-31 promoter using chromatin immunoprecipitation. Altogether, our results identify a feed-forward loop that leads to the activation of SOX4, which in turn up-regulates and binds to EZH2, cooperating with HDAC3 to repress the miR-31 promoter and advance esophageal tumorigenesis.
Discussion
Differential microRNAs expression has been linked to tumor initiation and progression. Depending on cellular context, microRNA expression in tumors may be increased or decreased, with microRNAs behaving as tumor promoters or suppressors, respectively [
2]. Moreover, multiple microRNAs have been shown to promote or inhibit metastasis [
3,
45]. Because metastasis is responsible for more than 90% of cancers-related deaths, it is important to define molecular mechanisms by which microRNAs regulate metastasis and define new therapeutic targets. Studies on the role of miR-31 in esophageal tumors have produced conflicting results [
16,
18,
19,
46]. Hence, it is important to define not only the molecular pathways regulated by miR-31 but also the factors regulating miR-31 expression and functions in various tissues and cancers. The lack of consistency in the literature with respect to miR-31 expression in esophageal cancer may be due to platform choice or normalization methods. Variability may also result due to the limited number of patients in each study. Here, we show downregulation of miR-31 in invasive and aggressive esophageal cancer cells. miR-31 has a known role in prostate cancer and melanoma, suppressing key cell cycle regulators and pro-oncogenic genes such as CDK1, E2F2, EXO1, FOXM1, MCM2, Src or MET [
11]. We show that miR-31 significantly suppresses migration and invasion
in vitro in both aggressive esophageal adenocarcinoma and squamous cell carcinoma. Ectopic expression of miR-31 did not significantly affect tumor cell proliferation, causing only a marginal decrease in colony formation in the adenocarcinoma cell line FLO1. Interestingly, Valastyan
et al. also reported that miR-31 promoted metastasis but not cell proliferation in breast cancer [
9]. A number of studies argue that epithelial-mesenchymal transition (EMT) plays a crucial role in cancer metastasis and progression, and loss of miR-200 family members drives EMT in multiple cancers [
47,
48]. We examined expression of EMT markers in miR-31-overexpressing cell lines but found no significant alterations (data not shown). We conclude that miR-31 is not a strong inhibitor of EMT.
Several mechanisms could contribute to aberrant miR-31 expression in cancer. Based on data demonstrating epigenetic repression of miR-31 expression by DNA methylation and EZH2-mediated H3K27me3 epigenetic mark in melanoma, leukemia and prostate cancer [
8,
11], we examined the role of epigenetic regulation of miR-31 in invasive esophageal cancer. Our results indicate that DNA methylation and Polycomb-mediated histone methylation both contribute to miR-31 silencing, since treatment with the EZH2 inhibitor, DZNep, and the DNMT inhibitor, AZA, enhanced miR-31 expression. Additionally, we show for the first time that histone deacetylation contributes to miR-31 silencing as treatment with the pan-HDAC inhibitor SAHA restored miR-31 expression. This is in line with recent observations that EZH2 interacts with HDAC3 to downregulate miR-29 [
37] and that EZH2 can interact with DNMT to control DNA methylation [
49]. Therefore, we propose a key role for Polycomb/EZH2, HDAC and DNMT for survival and metastasis of esophageal cancers.
Recent work argues that SOX4 is a master regulator of EMT and cell invasion through its binding to the EZH2 promoter, inducing EZH2 transcription [
30]. Our results indicate that miR-31 down-regulates SOX4 by binding to its 3′-UTR in EAC and ESCC cells. Consistent with previous reports [
8,
11], we found that miR-31 negatively regulates EZH2 as well. Furthermore, we observed a significant decrease in HDAC3 expression with miR-31 overexpression. We, therefore, propose that one underlying mechanism by which miR-31 suppresses tumor cell invasion is by directly targeting SOX4 and indirectly targeting EZH2 and HDAC3. Notably, overexpression of SOX4 is observed in a variety of human cancers [
21]. Loss of SOX4 led to a significant increase in miR-31 expression and strongly inhibited tumor cells proliferation, migration and invasion. Additional studies will aim to identify a subset of patients with concomitant high SOX4, EZH2 and HDAC3 and low miR-31 to demonstrate the mechanistic and clinical correlation between these pathways.
Multiple reports show that the polycomb PRC2/EZH2 interacts with HDAC to repress the transcription of CDH1 [
38]. Moreover, MYC was recently shown to interact with EZH2 and HDAC3 to repress miR-29 in lymphomas [
37]. However, it is not clear how PRC2/EZH2 is recruited to the promoter of specific genes. Similarly, SOX4 has been shown to interact with multiple transcription factors. Despite the fact that our computational analysis did not detect any SOX4 regulatory elements upstream of the miR-31 promoter, co-IP assays show that SOX4 interacts with EZH2 and HDAC3. Based on the ChIP experiments, we propose that EZH2 and HDAC3 bind to a similar region on the miR-31 promoter, confirming that histone methylation and histone deacetylation contribute to miR-31 repression. It is also possible that other molecules such as PRC2-interacting YY1 or the SOX4-interracting protein GATA4 mediate SOX4 recruitment to the DNA. Future studies are needed to confirm such possibilities.
Materials and methods
Cell lines
The squamous cell carcinoma cell lines TE8, TE1, TE11, TE7, TE12, TE2, TE3, TE5 and TE9 were developed and characterized by Kuroki et al. [
50]. OE33 and FLO1 were established by Dr. David Beer (University of Michigan). HEK293T, CP-A and CP-B were purchased from American Type Culture Collection (ATCC). KYSE140, KYSE180, KYSE150, KYSE520, KYSE70 were developed and characterized by Shimada et al. [
51]. All cells were cultured at 37°C in 5% CO
2.
Chemicals and antibodies
Suberoyl anilide hydroxamic acid (SAHA, a pan-HDAC inhibitor) and 3-Deazaneplanocin A (DZNep, a polycomb EZH2 subunit inhibitor) were purchased from Millipore, Temecula, CA. The DNA methylation inhibitor 5-AZA-Cytidine (AZA) was from Sigma, St. Louis, MO. Cells were treated with indicated concentration and incubated for 72 hours before harvesting. Agarose A/G plus is from Santa Cruz. The following antibodies were used as primary antibodies: SOX4 (Santa Cruz; 1:1000); EZH2 (BD; 1:4000); HDAC3 (Millipore; 1:4000); total histone H3 (Abcam; 1:5000); Tri-methylated histone H3 H3K27me3 (Millipore; 1:3000); SUZ12 (Santa Cruz; 1:1,000); EZH1 (Santa Cruz; 1:1,000); β-tubulin (Sigma-Aldrich; 1:5,000).
Oligonucleotides and plasmids
The pBABE-miR-31 plasmid (Plasmid#26088), pWPXL-SOX4 (plasmid#36984), pCMVHA-hEZH2 (plasmid#24230), Psicheck2 SOX4 full-length 3′UTR (Plasmid#26989) were purchased from Addgene (Cambridge, MA).
A 71 bp WT fragment of the SOX4 3′UTR (SOX4 WT OLIGO) was created by overlapping extension PCR and cloned between the XhoI and NotI site of the psicheck2 plasmid. Similarly, the mutant construct of SOX4 3′UTR (SOX4 Mutant OLIGO) which carried a substitution of four nucleotides within the core seed sequence of miR-31, was carried out using overlapping extension PCR and cloned between the XhoI and NotI site of the psicheck2 plasmid.
The two set of plasmids containing shRNA specific to SOX4 and EZH2 were purchased from OriGene (Rockville, MD).
Primers used for SOX4, forward: 5′-AGCGACAAGATCCCTTTCATTC-3′, reverse: 5′-CGTTGCCGGACTTCACCTT-3′; EZH2, forward: 5′-GTACACGGGGATAGAGAATGTGG-3′, reverse: 5′GGTGGGCGGCTTTCTTTATCA-3′, for EZH1, forward: 5′-ATGCGACTTCGACAACTTAAACG-3′, reverse: 5′-GGCTTCATTGACTGAACAGGTT-3′, HDAC3, forward: 5′-CCTGGCATTGACCCATAGCC-3′, reverse: 5′-CTCTTGGTGAAGCCTTGCATA-3′, GAPDH, forward: 5′-GCGACACCCACTCCTCCAC-3′, reverse: 5′-TCCACCACCCTGTTGCTGTAG-3′.
3′UTR Luciferase Reporter assays
HEK293, TE8 and FLO1 cells were plated in triplicate in a 24 well plate. One day after plating, cells were transfected with the dual Renilla and Firefly luciferase reporter plasmid (psiCHECK-2) containing the full length 3′UTR of SOX4 (plasmid# 26989), the short WT oligo or mutant oligo along with a pBABE-miR-31 or pBABE Empty Vector expressing plasmid using FuGene HD (Promega). 48 hr post-transfection, cells were lysed using 1X passive lysis buffer and lysates were analyzed using the Dual-Glo Luciferase Reporter Assay System (Promega) on the Synergy4 multi-mode microplate reader (BioTeK).
Quantitative real time PCR
Total RNA was isolated from cells with Qiazol reagent (Qiagen) and reverse transcribed into cDNA using the miRNeasy mini kit and miScript miRNA Reverse Transcription kit (Qiagen). qRT-PCR was performed according to the manufacturer’s instructions (Applied Biosystems).
Proliferation assay
WST-1 reagent (Roche) was used according to manufacturers’ protocols to assess cell viability.
Cell migration and invasion assays
Migration and Invasion assays were performed as previously described [
52]. Migrating or invading cells were then photographed using the Zeiss Axioskop Plus or eVos microscope and quantified with ImageJ software.
For clonogenic assay, 500 transfected cells were seeded in six-well plates and maintained in complete medium for 2 weeks. Colonies were fixed with ice-cold methanol and stained with crystal violet. Colonies were photographed and counted using GelCount (Oxford Optronix).
Western blot analysis
Western blot was performed as previously described [
52].
Immunoprecipitation (IP)
Cells were collected and lysed in IP lysis buffer (150 mM NaCL, 50 mM Tris pH8, 1% Triton X-100, 1% NP-40) supplemented with protease and phosphatase inhibitors, incubated on ice for 20 min, and cleared by centrifugation at 13,200 rpm at 4 C for 20 min. Total protein lysate (500 μg) was immunoprecipitated with the agarose-immobilized antibody (6 μg of anti-SOX4, EZH2, HDAC3 or isotype control antibodies) and incubated overnight at 4°C. Immune complexes were eluted from the agarose beads and analyzed by SDS-PAGE followed by immunoblot analysis. For co-IP in 293 T, cells were transfected with plasmids using FuGene HD (Promega). Cells were collected 48 hours post-transfection and analyzed as described above.
Chromatin Immunoprecipitation (ChIP)
The ChIP assays were adapted and performed according to previous publications [
53]. Briefly, cells were fixed using 1% formaldehyde for 15 minutes and quenched using 125 mM glycine for 5 minutes at room temperature. After centrifugation the cell pellet was re-suspended in the cell lysis buffer (150 mM NaCL, 50 mM Tris pH8, 1% Triton X-100, 1% NP-40, 0.01% SDS, 1.2 mM EDTA pH 8.0, 1 mM PMSF). Protein-bound chromatin was fragmented by sonication. Equal volumes of chromatin were immunoprecipitated with anti-HDAC3, anti-SOX4, anti-EZH2, anti-trimethyl-Histone H3 Lys27 or normal IgG as a negative control (Millipore). Following extensive washing the immunoprecipitated DNA was treated with RNase (Qiagen) for 30 min at 37°C and proteinase K (Roche) for an hour at 45°C. The DNA was eluted using 100 mM NaHCO3 and 1% SDS and the crosslinks were reversed using 300 mM NaCl at 65°C for 16 hours. Immunoprecipitated DNA and whole cell extract DNA were purified by Qiaquick PCR purification kit (Qiagen). The purified DNA was amplified by real-time quantitative PCR with Qiagen QuantiTech SYBR Green PCR master mix and analyzed for enrichment. Real-time qPCR amplification was performed with Applied Biosystems StepOnePlus real time PCR system.
Primers used as described in Lin et al. [
44]: -1,000bp forward: CCGATGACCTAGCCAGAAGT, reverse: CCCCACCCTTCAACTCGTAG; -500 bp, forward: TATCCTCAACCCTCCGTGTC, reverse: CATACACCTGAAGGGGCAGT; +500 bp, forward: CAATTTTGGCCCAGGAGATA, reverse: TTTCCGGGGACCTCTAGTTT; +42,500 bp, forward: TGGCCTATTTGCTGTTCTAATGAC, reverse: GCAAGCCAACCCCAACA; +45,000 bp, forward: AATGGGCCCTGCATTCTCT, reverse: AAAACCCACACCCTCACCAC; +47,500 bp, forward: CATCTTCAAAAGCGGACACTCT, reverse: ACAATACATAGCAGGACAGGAAG; MYT, forward: AGGCACCTTCTGTTGGCCGA, reverse: AGGCAGCTGCCTCCCGTACA; GAPDH, forward: CGGCTACTAGCGGTTTTACG, reverse: AAGAAGATGCGGCTGACTGT.
Dataset analysis
Datasets made publicly available from GEO Datasets (
http://www.ncbi.nlm.nih.gov/gds/). The collected information from each dataset was analyzed and visualized in Prism version 6.00 for Mac (GraphPad software, La Jolla, California).
Statistical analysis
Each experiment was repeated at least three times. Numerical data are presented as mean ± standard deviation. The differences between two groups were analyzed using a Student’s t-test (two-tailed) or two-way ANOVA. Differences were considered statistically significant at p < 0.05. All statistical analysis was performed on GraphPad Prism 6.0c software (La Jolla, CA).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
RBK carried out the experimental approach with assistance from KJT, STZ, and CJT. HAL performed the statistical analysis. RBK, TA and CA conceived of the study, and participated in its design and coordination and wrote the manuscript with editing assistance from KJT and CJT. All authors read and approved the final manuscript.