MicroRNA‐135b regulates ERα, AR and HIF1AN and affects breast and prostate cancer cell growth

Abstract

MicroRNAs (miRNAs) regulate a wide range of cellular signaling pathways and biological processes in both physiological and pathological states such as cancer. We have previously identified miR‐135b as a direct regulator of androgen receptor (AR) protein level in prostate cancer (PCa). We wanted to further explore the relationship of miR‐135b to hormonal receptors, particularly estrogen receptor α (ERα). Here we show that miR‐135b expression is lower in ERα‐positive breast tumors as compared to ERα‐negative samples in two independent breast cancer (BCa) patient cohorts (101 and 1302 samples). Additionally, the miR‐135b expression is higher in AR‐low PCa patient samples (47 samples). We identify ERα as a novel miR‐135b target by demonstrating miR‐135b binding to the 3′UTR of the ERα and decreased ERα protein and mRNA level upon miR‐135b overexpression in BCa cells. MiR‐135b reduces proliferation of ERα‐positive BCa cells MCF‐7 and BT‐474 as well as AR‐positive PCa cells LNCaP and 22Rv1 when grown in 2D. To identify other genes regulated by miR‐135b we performed gene expression studies and found a link to the hypoxia inducible factor 1α (HIF1α) pathway. We show that miR‐135b influences the protein level of the inhibitor for hypoxia inducible factor 1α (HIF1AN) and is able to bind to HIF1AN 3′UTR. Our study demonstrates that miR‐135b regulates ERα, AR and HIF1AN protein levels through interaction with their 3′UTR regions, and proliferation in ERα‐positive BCa and AR‐positive PCa cells.

Highlights

• MiR‐135b is studied in breast (BCa) and prostate cancer (PCa).

• ERα is identified as a novel miR‐135b target by interaction to the 3′UTR in BCa.

• MiR‐135b expression inversely correlates with ERα protein in BCa patient cohorts.

• MiR‐135b regulates ERα, AR and HIF1AN protein levels through 3′UTR binding.

• MiR‐135b inhibits growth of ERα+ BCa cells and AR+ PCa cells.

1. Introduction

Breast cancer (BCa) and prostate cancer (PCa) are two of the most frequently diagnosed cancers in the developed countries (Jemal et al., 2011). The localized forms of these cancers are usually efficiently treated by surgery and/or radiation therapy, whereas the disseminated disease is a considerable clinical challenge. Many patients receiving endocrine therapy relapse with fatal therapy‐resistant disease; twenty‐five to forty percent of the BCa patients develop distant metastases after initial treatment and one third of the PCa patients develop advanced PCa after 18–36 months (Guarneri and Conte, 2004; Lamoureux et al., 2013; Wu et al., 1999). Estrogen receptor α (ERα) and androgen receptor (AR) are important oncogenic drivers in the luminal hormone receptor positive BCa and in PCa respectively, and are thus primary targets of the endocrine therapy (Althuis et al., 2004; Feldman and Feldman, 2001).

In recurrent BCa and PCa the expression of ERα can be lost or modified by alterations of the ER pathway and its cofactors, but resistance may also occur due to deregulation of cell cycle signaling molecules or growth factor receptor pathways (reviewed in Garcia‐Becerra et al., 2012; Osborne and Schiff, 2011). The therapeutic by‐pass in PCa is associated with the reactivation of AR by multiple comparable pathways (reviewed in Knudsen and Penning, 2010). Also increased AR mRNA has been described sufficient for driving resistant PCa in xenografts models (Chen et al., 2004). However, AR also regulates a distinct set of genes in the recurrent form of PCa compared to the hormone dependent form of PCa, often in a tissue specific manner (Sharma et al., 2013; Wang et al., 2009). Both ERα and AR splice variants have been identified (Zhang et al., 1996; reviewed in Dehm and Tindall, 2011). ERα and AR splice variants have been suggested to contribute to resistance to endocrine therapy in BCa and to recurrent therapy resistant PCa (Murphy et al., 1998; Kong et al., 2015). Furthermore, alternative splicing may affect microRNA (miRNA) biogenesis and expression, but may also lead to altered miRNA regulation of the target gene. MiRNAs have also been shown to play a role in regulation of alternative splicing (Boutz et al., 2007; Mayoral et al., 2009; Melamed et al., 2013).

Progression of BCa and PCa has been associated with altered miRNA regulation and the miRNAs contribute to endocrine resistance by modulating cell survival signaling, apoptosis and nuclear receptor expression (Dvinge et al., 2013; Enerly et al., 2011; Le Quesne and Caldas, 2010; Ottman et al., 2014; Tian et al., 2013). Several studies show that ERα and AR are directly targeted by miRNAs (Adams et al., 2007; Castellano et al., 2009; Leivonen et al., 2009; Östling et al., 2011; Pandey and Picard, 2009; Sikand et al., 2011; Xiong et al., 2010; Zhao et al., 2008). We have previously carried out high‐throughput miRNA overexpression screens in BCa and PCa, where we assayed the effect of up to 810 miRNAs on ERα and AR protein levels, proliferation and apoptosis (Leivonen et al., 2009; Östling et al., 2011). Our studies identified that miR‐135b was one of 13 miRNAs that downregulated AR by directly targeting the AR 3′UTR (Östling et al., 2011). Other miR‐135b targets that have been validated using luciferase reporter assays have been reported in colorectal cancer, in HEK293T cells and in non‐small‐cell lung cancer (Lin et al., 2013; Matsuyama et al., 2011; Nagel et al., 2008; Valeri et al., 2014). Arigoni and coworkers have identified midline 1 (MID1) and mitochondrial carrier 2 (MTCH2) as targets in BCa (Arigoni et al., 2013). Additionally, aberrant miR‐135b expression has been observed in a number of cancers such colorectal cancer, cutaneous squamous cell carcinoma, pancreatic ductal adenocarcinoma, papillary thyroid carcinoma, PCa and BCa (Arigoni et al., 2013; Bandres et al., 2006; Brunet Vega et al., 2013; Faltejskova et al., 2012; Munding et al., 2012; Sand et al., 2012; Tong et al., 2009; Wang et al., 2013). Negative associations of miR‐135b to ER status in BCa have been identified in two different cohorts by expression profiling (Enerly et al., 2011; Lowery et al., 2009).

In this study, the expression of miR‐135b in BCa and PCa was analyzed in two BCa cohorts and one PCa cohort. The miR‐135b expression inversely correlated with ERα and AR protein in these tumor samples. We show that miR‐135b decrease intracellular levels of ERα protein and mRNA by interaction with the 3′UTR of the ERα encoding transcript in BCa. Growth of hormone receptor positive BCa and PCa cells was decreased upon miR‐135b overexpression. In LNCaP cells the effect of miR‐135b was more prominent than downregulation of AR alone which led us to investigate additional targets. Here we identified miR‐135b regulation of HIF1AN in both BCa and PCa cells through the 3′UTR region. Thus miR‐135b regulates the main targets of endocrine therapies ERα and AR, but also HIF1AN thereby providing a putative link between hormone signaling and HIF1α regulation.

2. Materials and methods

2.1. Breast cancer samples and analysis of miR‐135b and its target genes

MiR‐135b and target gene expression data was obtained from 101 primary breast carcinoma samples from the MicMa cohort (Enerly et al., 2011). The data have been previously published in the Gene Expression Omnibus (GEO) with accession number GSE19536. ERα status was determined by immunohistochemistry (IHC) as described previously (Enerly et al., 2011; Leivonen et al., 2009). Additionally, miR‐135b and gene expression data from 1302 breast tumors from the METABRIC data (Dvinge et al., 2013) (European Genome‐Phenome Archive, www.ebi.ac.uk/ega, accession number EGAS00000000122) were used.

2.2. Prostate cancer tissue specimens and analysis of miR‐135b levels by qRT‐PCR

Prostatic tissues obtained by transurethral resection of the prostate (TURPs) were collected 1990–1999 in Malmö, Sweden. Results were based on histopathological diagnosis in randomly selected cases with evidence of prostate adenocarcinoma in 47 patients. Analyses of miR‐135b levels by qRT‐PCR and immunohistochemistry and scoring of AR on prostate tissue slide were done as previously described (Hagman et al., 2010; Östling et al., 2011).

2.3. Cell culture

Breast cancer cell lines were cultured as previously described (Leivonen et al., 2014). Non‐malignant hTERT‐HME1 (HMEC) cells were purchased from ATCC, (American Type Culture Collection, Manassas, VA, USA) and cultured according to ATCC instructions. Prostate cells were cultured as previously described (Östling et al., 2011). DU‐145 cells were purchased from Interlab Cell Line Collection (ICLC) and the non‐malignant PNT2 cells were purchased from Sigma–Aldrich (Sigma–Aldrich Co, St Louis, MO, USA), and grown in medium conditions recommended by the providers. All cells were cultured for less than 4 months before use in the experiments.

2.4. Quantitative real‐time‐PCR analyses

Total RNAs from the cell lines were isolated with mirVana™ miRNA isolation kit (Ambion Inc., Austin, TX, USA). For cDNA synthesis, 20 ng of total RNA was reverse transcribed with a TaqMan^® MicroRNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). TaqMan^® MicroRNA Assays were performed according to manufacturer's instructions, using 1.3 μl of the reverse transcribed RNA per reaction. All samples were run in triplicates in each technical repeat. The Taqman quantitative real‐time‐PCR analysis was carried out with an Applied Biosystems 7900HT instrument. The results were analyzed with SDS 2.3 and 2.4 and RQ manager softwares (Applied Biosystems). Samples with amplification curves differing from mean when analyzed by SDS 2.3/2.4 were excluded to avoid technical bias. The relative expression values of miR‐135b was determined by using the 2^−ΔΔCt method (Livak and Schmittgen, 2011), with RNU6B and RNU48 as endogenous controls.

For cDNA synthesis, 200 ng of total RNA was reverse transcribed with a High Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA, USA). Thereafter, the cDNAs were diluted 1/10, and 2 μl of 1/10 diluted cDNA was used per reaction. Taqman quantitative real‐time‐PCR analysis was carried out as above using specific primers for ERα (ESR1), AR, HIF1AN, β‐actin (ACTB) and GAPDH designed by the Universal Probe Library Assay Design Center (Roche Applied Biosciences, Basel, Switzerland). The sequences of the primers were: ESR1 forward 5′‐TTACTGACCAACCTGGCAGA‐3′, ESR1 reverse 5′‐ATCATGGAGGGTCAAATCCA‐3′, AR forward 5′‐GCCTTGCTCTCTAGCCTCAA‐3′, AR reverse 5′‐GTCGTCCACGTGTAAGTTGC‐3′ HIF1AN forward 5′‐GGGGAACCCACAAGAGGT‐3′, HIF1AN reverse 5′‐GTGGACGGGATAGCAGTCAC‐3′, ACTB forward 5′‐CCAACCGCGAGAAGATGA‐3′, ACTB reverse 5′‐CCAGAGGCGTACAGGGATAG‐3′, GAPDH forward 5′‐AGCCACATCGCTCAGACA‐3′ and GAPDH reverse 5′‐GCCCAATACGACCAAATCC‐3′. The fluorescent Taqman probes were obtained from Roche Human Probe Library (no 24 for ESR1, no 14 for AR, no 51 for HIF1AN, no 64 for ACTB and 60 for GAPDH). The results were analyzed as above with SDS 2.3/2.4 and RQ manager softwares (Applied Biosystems), and the relative expression of mRNA was calculated by the 2^−ΔΔCt method (Livak and Schmittgen, 2011) using ACTB and GAPDH as endogenous controls. Data were collected from at least two separate biological experiments, which were run at least twice.

2.5. Reverse transfection of cells

Reverse transfection was done as previously described (Leivonen et al., 2009) using 20 nM of the following miRNAs and siRNAs: Pre‐miR™ miRNA Precursor Negative Control #1 or #2 (AM17110), human pre‐miR‐135b and pre‐miR‐18a miRNA precursors (AM17100, PM13044 and PM12973) (Ambion Inc., Austin, TX, USA), siAR (S1538, #4390824) and siERα (#42835) (Ambion), siAllStars Cell Death control (SI1027299), siPLK1_7 (SI02223844), siKIF11_7 (SI02653770) and siHIF1AN (SI00436324, SI00436345, SI02778090) (Qiagen GmbH, Hilden, Germany).

2.6. Immunoblotting

SDS‐PAGE and immunoblotting were done as described previously (Leivonen et al., 2009; Östling et al., 2011). The following primary antibodies were used; AR (H‐280, Santa Cruz Biotechnology Inc., Santa Cruz, CA), ERα (HC‐20, Santa Cruz Biotechnology Inc., Santa Cruz, CA), HIF1AN (12289‐100, Abcam) or HIF1AN (H‐229, sc‐48813, Santa Cruz Biotechnology Inc., Santa Cruz, CA, and β‐actin (A1978 Sigma‐Aldrich). The signals were obtained with Alexa Fluor 680 tagged secondary IgG antibodies (Invitrogen) and Odyssey Li‐Cor scanner 2.1 (LI‐COR Biosciences) or with HRP‐conjugated secondary antibodies (sc2004 or sc2005 Santa Cruz Biotechnology) and enhanced chemiluminescence (ECL).

2.7. 3′UTR luciferase assays

To study the binding of miR‐135b to the ERα 3′UTR, luciferase assays were conducted in MCF‐7 cells by co‐transfecting 50 nM miR‐135b pre‐miR construct or pre‐miR miRNA negative control (Scr) with 50 ng ERα 3′UTR reporter plasmids or HIF1AN reporter plasmids and 50 ng Renilla luciferase plasmid as previously described (Leivonen et al., 2009). The ERα 3′UTR sequence was previously cloned in four fragments into MluI/HindIII sites of pMIR‐REPORT Luciferase vector (Leivonen et al., 2009). PITA, microRNA.org (August 2010), TargetScan 6.2 and NBmiRTar (v1.0_Beta) miRNA target site prediction algorithms were used to identify putative miR‐135b target sites in the ERα 3′UTR sequence. The ERα 3′UTR sequence was furthermore manually searched for miR‐135b seed sequences or partial seed sequences. The four ERα 3′UTR reporter plasmids with mutated miR‐135b putative target sites (Supplementary Table S3) were custom ordered from Blue Heron Biotech, LLC (Blue Heron Biotechnology, Inc., Bothell, WA, USA).

For the HIF1AN reporter plasmids, shorter sequences spanning two miR‐135b binding sites (predicted by TargetScan 6.2) (Supplementary Table S4) were cloned into SpeI/HindIII sites of a pMIR‐REPORT Luciferase vector (Ambion Inc., Austin, TX, USA) downstream of a luciferase gene. HIF1AN reporter plasmids (wild‐type and mutated) were custom ordered from Blue Heron Biotech, LLC (Blue Heron Biotechnology, Inc., Bothell, WA, USA).

2.8. Illumina gene expression array analysis upon miR‐135b overexpression in LNCaP cells

Gene expression was profiled with Illumina microarrays. MiR‐135b was overexpressed in LNCaP cells by reverse transfection for 12, 24 and 36 h. Total RNA was isolated utilizing mirVana™ miRNA isolation kit (Ambion Inc., Austin, TX, USA) according to the manufacturer's protocol. Integrity of the RNA prior to hybridization was monitored using a Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA) according to manufacturer's instructions. Purified total RNA (300 ng) was amplified with the TotalPrep Kit (Ambion Inc., Austin, TX, USA) and the biotin labeled cRNA was hybridized to Sentrix HumanHT‐12 v4 Expression Bead Chips (Illumina, San Diego, CA, USA). The arrays were scanned with the BeadArray Reader (Illumina, San Diego, CA, USA) and the raw data were obtained by GenomeStudio (Illumina, San Diego, CA, USA). The array experiments were performed at the Finnish DNA Microarray Centre, Turku Centre for Biotechnology, Turku, Finland. The microarray data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE57820.

The genes identified by Illumina gene expression analysis were subjected to Gene Set Enrichment Analyses (GSEA), using the web‐based resource Molecular Signatures Database (MSigDB v4.02, Broad Institute), http://www.broadinstitute.org/gsea/msigdb/annotate.jsp, in order to identify overlaps to previously identified gene sets. Overlaps were computed to C2, C5 and C6 curated gene sets with a cut‐off of FDR q‐value <0.01 used.

2.9. Data normalization

Pre‐processing and normalization of the Illumina gene expression data was done using the lumi package as previously described (Leivonen et al., 2009). Statistical analysis of differential gene expression upon miRNA transfection was performed with R/Bioconductor (Gentleman et al., 2004) using the limma package (Smyth, 2005). Gene expression upon miRNA overexpression was compared to the scrambled miRNA negative control (Scr) in pairwise fashion using the empirical Bayes statistics implemented by eBayes function (Smyth, 2004). The threshold for differential expression was q < 0.05 after the Benjamini–Hochberg multiple testing correction.

2.10. Statistical analyses

The gene and miRNA expression data are shown as mean ± standard deviation of the mean. The in vitro experiments were performed at least twice, and all data are shown as mean ± standard error of the mean. The 3′UTR luciferase assays were performed at least three times. The data were analyzed for statistical significance using an unpaired Student's t‐test. Association of miR‐135b expression with the ERα or AR protein status was determined by Mann–Whitney test. Nonparametric Kruskal–Wallis test was performed for determining the association between miR‐135b expression level and ERα grading. Correlations between miR‐135b and mRNA expressions were calculated with the Spearman Rank method. The SPSS v.20 (SPSS Inc., Chicago, IL) was used for the statistical calculations. For all the experiments, a p‐value below 0.05 was considered statistically significant.

3. Results

3.1. MiR‐135b inversely associates with ERα in breast cancer specimens and cell lines and with AR in prostate tumor samples and cell lines

We wanted to study the expression of miR‐135b in clinical samples of BCa and thus analyzed its expression in relation to the ERα status in two BCa patient cohorts: the MicMa with 101 tumors (71 ERα+ and 30 ERα− tumors) (Enerly et al., 2011) and the METABRIC with 1302 tumors (1003 ERα+ and 299 ERα− tumors) (Dvinge et al., 2013). The expression of miR‐135b was significantly higher in the ERα− tumors compared to the ERα+ tumors in both datasets (p < 0.001, Figure 1A and B). Furthermore, miR‐135b inversely associated with the levels of the ERα protein in the MicMa cohort, when the tumors were divided into five groups according to the number of ERα positive nuclei (p < 0.001, Figure 1C). The information on the tumor grades for the MicMa and for the METABRIC data set is provided in Supplementary Figure S1.

Figure 1.

Characterization of miR‐135b expression and ERα/AR protein expression in breast and prostate tumor samples. Breast tumor samples were analyzed for ERα and miR‐135b expression, showing that miR‐135b is higher in ERα− tumors in two different datasets A) the MicMa dataset with 101 tumors (71 ERα+, 30 ERα−) and B) the METABRIC dataset with 1302 samples (1003 ERα+, 299 ERα−). C) Tumors within the MicMa dataset were grouped according to ERα protein grading (0–4), based on immunohistochemical staining, depending on the percentage of cell nuclei that were stained (0, 0%; 1, 1–10%; 2, 10–50%; 3, 50–75%; 4, 75–100%), clearly showing that tumors with the highest ERα protein expression (4) have the lowest miR‐135b expression (p < 0.001). D) Prostate cancer tissue specimens (47 samples) were stained for AR and scored for the intensity of AR, and analyzed for miR‐135b expression levels by qRT‐PCR, showing that miR‐135b expression is higher in the AR‐low (1 + 2) tumors (p = 0.01).

In PCa specimens, we analyzed the expression of miR‐135b by qRT‐PCR in 47 PCa samples. These samples have been stained for AR protein, scored and grouped into three groups according to their staining intensity. Statistical analyses of these samples revealed that miR‐135b had significantly lower expression in the tumors, which had high AR expression, when compared with tumors with low AR expression (p = 0.01) (Figure 1D). The information on the tumor grades for the PCa data set is provided in Supplementary Figure S1. In addition, miR‐135b showed a significant negative correlation with ERα and AR mRNA levels in the two breast cancer cohorts (Supplementary Figures S2A–B and S3A–B).

Next, we analyzed miR‐135b, ERα and AR levels by qRT‐PCR in cell lines derived from normal and malignant breast and prostate. The results showed that the endogenous miR‐135b expression was higher in those cell lines in which ERα and AR expression was low, whereas cells with high ERα or AR demonstrated lower miR‐135b expression levels (Figure 2A–D and Supplementary Figure S4A–D). Western blot analyses of ERα and AR protein show similar expression of both proteins as observed by qRT‐PCR (Figure 2E–F). The expression pattern observed in the cell lines is similar to our observations from in vivo tumor samples. The miR‐135b expression was highest in the ERα‐negative KPL‐4 and JIMT‐1 BCa cell lines, as well as in the non‐malignant HMEC (ERα‐negative) cell line (Figure 2C). In PCa cell lines, the miR‐135b expression was highest in the non‐malignant PNT2 (AR‐negative) cell line, and in the AR negative PC‐3 and DU‐145 cell lines. However, the expression of both miR‐135b and AR was high in the VCaP cell line. The VCaP cell line has the highest expression of AR of all cell lines (∼12‐fold) (Figure 2D), due to the amplification of AR in VCaP (Liu et al., 2008).

Figure 2.

Characterization of miR‐135b, ERα and AR levels in breast and prostate cell lines. Relative miR‐135b expression levels in A) breast cell lines and in B) prostate cell lines, as determined by qRT‐PCR. The expression levels were normalized to RNU6B and the 2−ΔΔCt method was used for quantification. C) Relative ERα mRNA expression in breast cell lines and D) AR mRNA expression were quantified by qRT‐PCR and the 2−ΔΔCt method. ACTB was used as a control. All error bars represent standard deviation of mean of at least two independent experiments. Endogenous E) ERα protein expression levels in breast cell lines and F) AR protein expression in prostate cell lines were analyzed by Western blot.

3.2. MiR‐135b downregulates ERα through interaction with the 3′UTR sequence

We have previously shown that miR‐135b downregulates AR protein levels in PCa cell lines by targeting the AR 3′UTR (Östling et al., 2011). To study the effect of miR‐135b on the ERα levels in BCa cells, we transiently transfected miR‐135b to ERα+ BCa cell lines, MCF‐7 and BT‐474, and assessed the ERα protein and mRNA level. MiR‐18a, which has previously been shown to downregulate ERα (Leivonen et al., 2009), and ERα siRNAs were used as positive controls. Our results show downregulation of the ERα protein level in both MCF‐7 and BT‐474 cells after 72 h (Figure 3A). The same effect was seen on ERα mRNA in MCF‐7 cells after 24 h (Figure 3B and Supplementary Figure S4G).

Figure 3.

MiR‐135b regulates ERα protein and mRNA levels through interaction with the ERα 3′UTR sequence. A) Western blot showing the effect of miR‐135b on ERα protein expression 72 h post transfection in MCF‐7 and BT‐474 cells. The previously identified ERα regulating miR‐18a (Leivonen et al., 2009) was used as an additional positive control. B) ERα mRNA level is affected by miR‐135b overexpression in MCF‐7 cells, ± standard error of the mean. C) The 3′UTR sequence of the ERα is 4307 bp long, and was predicted by PITA (in grey) TargetScan 6.2 (in orange), microRNA.org (in green) and NBmiRTar (v1.0_beta) (in dark grey) to be targeted by miR‐135b at 15 sites indicated here by the colored arrows (17 arrows as two sites are predicted by two algorithms). D) The table summarizes the 29 potential target sites (predicted by algorithms and manually found), the position of them, in which construct of the ERα 3′UTR they are located and whether they are mutated in the 3′UTR assays conducted. E) Luciferase assays with wild‐type ERα 3′UTR constructs and ERα 3′UTR constructs with mutated miR‐135b binding sites were performed in MCF‐7 cells, showing that miR‐135b directly targets ERα at several binding sites. The white bars represent controls (ERα 3′UTR wild‐type and ERα 3′UTR with mutated miR‐135b target sites co‐transfected with neg. ctrl/Scr). Black bars represent ERα 3′UTR wild‐type co‐transfected with miR‐135b and grey bars represent ERα 3′UTR with mutated miR‐135b target sites co‐transfected with miR‐135b. All error bars represent standard error of the mean of at least three independent experiments *p < 0.05, **p < 0.01.

Next, we conducted luciferase reporter assays in MCF‐7 cells upon miR‐135b overexpression together with four different Luc‐ERα‐3′UTR reporter plasmids spanning the entire 4 kb 3′UTR region. MiR‐135b is predicted by TargetScan 6.2, microRNA.org (August 2010), PITA, and NBmiRTar (v1.0_beta) to bind the ERα 3′UTR at altogether 15 sites as indicated in Figure 3C. All putative sites, predicted by algorithms and manually found (adding up to a total of 29 sites), are summarized in Figure 3D including their mode of identification, location to the four reporter plasmids and their mutation status in our experiments. As our results in Figure 3E show, miR‐135b is able to interact with the ERα 3′UTR luciferase reporter at several binding sites in MCF‐7 cells, also at sites not identified by prediction algorithms.

3.3. MiR‐135b overexpression decreases the growth of hormone receptor positive cells

We wanted to follow the effect of miR‐135b overexpression on the growth of breast and prostate cell lines. We imaged the confluence of cell culture every hour using an Incucyte live content imager for up to five days during miR‐135b overexpression (Figure 4A–B). The ERα+ MCF‐7 and BT‐474 cell lines grew slower in the presence of miR‐135b (Figure 4A) compared to the ERα− KPL‐4 and HMEC cell lines, where the growth was not reduced upon miR‐135b overexpression (Figure 4B). Equally, the growth of the AR+ PCa cell lines, LNCaP and 22Rv1, were clearly decreased by miR‐135b overexpression (Figure 4A), whereas the growth of the AR‐low PNT2 and PC‐3 cell lines were not affected (Figure 4B). These data indicate that when overexpressed, miR‐135b more prominently regulates ERα/AR+ cell lines compared to the ERα/AR− cell lines. Representative cell images of the difference in the growth between MCF‐7 and HMEC, and between LNCaP and PNT2 are showed in Supplementary Figure S5A and S5B, respectively.

Figure 4.

Transient overexpression of miR‐135b more prominently regulates ERα and AR positive cell lines compared to the ERα and AR negative cell lines. A–B) Cells were transiently transfected using human pre‐miR‐135b miRNA precursor and a pre‐miR miRNA negative control miRNA (Scr) as well as with siERα or siAR. PLK1 or KIF11 and AllStars Cell Death siRNAs were used as positive controls. The cells were grown on 24‐well plates in Incucyte live content imager, with confluence imaged every hour. Growth curves are representative growth curves from one experiment. All cell lines have been followed and imaged upon transfection at least two times.

3.4. Gene expression profiling of LNCaP cells upon miR‐135b overexpression reveals putative miR‐135b target genes

Knock‐down of AR with a specific siRNA did not affect the growth of AR+ prostate cancer cells to the extent as overexpression of miR‐135b did (Figure 4A). This suggested that miR‐135b regulates the proliferation of PCa cells through additional target genes. To get a better understanding of the molecular phenotype associated with miR‐135b, we conducted gene expression analyses following miR‐135b overexpression. To this end we treated LNCaP cells with miR‐135b pre‐miR construct (and Scr) for 12, 24 and 36 h, and assessed the genome‐wide gene expression changes by microarrays. The expression of 146, 484 and 671 (in total 943 unique genes) genes (Supplementary Table S1) were up‐ or downregulated by 1.4 fold, at 12, 24, and 36 h, respectively (Figure 5A). Figure 5B lists top 25 changed genes at each time point and the Venn diagram in Figure 5C shows the overlap of 64 common genes deregulated by miR‐135b overexpression at all three time points (listed in Supplementary Table S2).

Figure 5.

Illumina gene expression identifies putative miR‐135b target genes in LNCaP cells. Figure A) shows the number of genes with a changed expression of 1.4 fold upon miR‐135b overexpression for 12 h, 24 h and 36 h. The top down or upregulated genes upon miR‐135b overexpression are listed in B). The Venn‐diagram C) indicates the 64 genes that were deregulated at all time points. D) Among these 64 genes, gene set enrichment analyses (MSigDB, C2, C5 and C6 gene sets, v4.02, Broad Institute) identified overlapping genes to only one gene set at a cut‐off of FDR q‐value <0.01. E) Among the 64 deregulated genes three prediction programs commonly predicted five genes to be miR‐135b target genes. F) A significant negative correlation between miR‐135b and HIF1AN was identified in the breast tumor samples in the MicMa dataset (R = −0.355, p < 0.001).

By Gene Set Enrichment Analysis (GSEA, Broad Institute) (Mootha et al., 2003; Subramanian et al., 2005) we identified a postradiation tumor escape signature with seven overlapping genes to our 64 common miR‐135b deregulated genes (FDR q‐value < 0.01, Figure 5D) (Monnier et al., 2008). The seven genes were SNTB2, RAB8B, ZMAT3, TRAF4, MYD88, PPP1R12B and PRMT6 (syntrophin beta 2, Ras‐related protein Rab‐8B, zinc finger matrin type 3, TNF receptor‐associated factor 4, myeloid differentiation primary response gene 88, protein phosphatase 1 regulatory inhibitor subunit 12B, protein arginine methyltransferase 6). This signature contains genes that have been upregulated in recurrent tumors growing in preirradiated tissue and genes upregulated by in vitro selection through repeated cycles of sever hypoxia.

To identify putative miR‐135b interaction sites among the 64 miR‐135b deregulated genes we turned to miRNA target prediction programs. TargetScan 6.2, miRBase and PicTar commonly identified putative miR‐135b binding sites on CPLX1, HIF1AN, HOXA10, KCNMA1 and SSR1 (Figure 5E). The GSEA signature (Monnier et al., 2008) and the predictions of miR‐135b interaction sites on the HIF1AN 3′UTR suggested that miR‐135b might play a role in the regulation of hypoxia signaling, where transcription factor hypoxia inducible factor 1α (HIF1α), functions as a key driver, promoting for example invasion and metastasis of hypoxic tumors. Furthermore, statistical analyses revealed an inverse correlation between miR‐135b and HIF1AN in the MicMa BCa cohort (R = −0.355, p < 0.001, Figure 5F), and this prompted us to continue studies on the link between miR‐135b and HIF1AN.

3.5. Overexpression of miR‐135b downregulates HIF1AN

We first analyzed the endogenous HIF1AN expression across breast and prostate cell lines (Figure 6A–B and Supplementary Figure S4E–F). No clear correlation between miR‐135b levels and HIF1AN was identified under normal cell culture conditions. Neither did HIF1AN levels vary according to hormone receptor status of the cell lines (see Figure 2C and D). To verify the effect of miR‐135b on the expression of HIF1AN, we overexpressed miR‐135b in breast and prostate cell lines and assessed the effect on protein level by immunoblotting. HIF1AN protein levels are affected in all tested cells (MCF‐7, KPL‐4, LNCaP and PC‐3) as showed by the results in Figure 6C. The interaction to the HIF1AN 3′UTR was studied with luciferase assays in MCF‐7 cells upon miR‐135b overexpression. Two sites (predicted by TargetScan 6.2) that were also mutated were tested. The results in Figure 6D shows that miR‐135b is able to interact with these sites. To assess the effect of HIF1AN siRNAs on the growth of breast and prostate cell lines, MCF‐7 and LNCaP cells were transfected with siRNAs and miR‐135b. The confluence was imaged from live cells as previously. siHIF1ANs seem to affect growth of LNCaP cells more potently than siAR (Figure 6D) thus suggesting that HIF1AN and miR‐135b play an important role in regulating cell proliferation in LNCaP cells. In MCF‐7 cells miR‐135b, siERα and siHIF1AN_5 diplay a similar effect on cell growth (Figure 6E).

Figure 6.

MiR‐135b regulates HIF1AN protein levels in BCa and PCa cells. Figures A) and B) show the endogenous HIF1AN expression in breast and prostate cell lines, respectively as quantified by qRT‐PCR and the 2−ΔΔCt method. ACTB was used as a control. Error bars represent standard deviation of mean of at least two independent experiments. C) MiR‐135b overexpression at 72 h decreases the HIF1AN protein level in MCF‐7, KPL‐4, LNCaP and PC‐3 cells. D) Luciferase assays with two miR‐135b target sites predicted by TargetScan 6.2 were conducted. The wild‐type sequence was transfected together with a neg. ctrl/Scr (white bar) or miR‐135 (black bar). Transfections of miR‐135b with a sequence having mutated miR‐135b binding sites confirm interaction (grey bar). All error bars represent standard error of the mean of at least three independent experiments *p < 0.05. E) The effect of miR‐135b overexpression and siHIF1AN transfection was followed as previously in an Incucyte live content imager, showing that the miR‐135b, siERα and siHIF1AN_5 affects growth of MCF‐7 cells similarly. In LNCaP cells the effect of miR‐135b and siHIF1ANs seems more potent than siAR.

4. Discussion

The questions of hormone receptor status and responsiveness to endocrine therapy in BCa and PCa are addressed in the clinic at the time of diagnosis and during treatment. Further understanding of the regulation of ERα and AR in BCa and PCa are thus of utmost importance. We have previously studied the miRNA regulation of these hormone receptors by systematic functional high‐throughput miRNA overexpression analyses (Leivonen et al., 2009; Östling et al., 2011). We identified miRNAs that regulate ERα and AR through interaction with 3′UTR regions as measured by luciferace assays. MiR‐135b was identified as one such regulator of AR, but it was not among the miRNAs that we previously identified to directly regulate ERα.

MiR‐135a and miR‐135b belong to the same family of evolutionary conserved miRNAs but are encoded by separate genes. Mir‐135a is encoded by two copies located on chromosome 3 (3p21) and 12 (12q23) whereas the mir‐135b gene is located on chromosome 1 (1q32.1). Although their seed sequences are identical, we did not observe any effect of miR‐135a in our previous studies (Leivonen et al., 2009; Östling et al., 2011). Thus we focused our studies on miR‐135b.

Here, our analyses identified inverse correlation between miR‐135b, ERα and AR in tumor samples. Our findings are consistent with previously published data. Lowery and colleagues identified a six miRNA‐signature (miR‐342, miR‐299, miR‐217, miR‐190, miR‐135b, miR‐218), which classified tumors into ER‐positive and ER‐negative phenotypes (Lowery et al., 2009). Additionally, the association of miR‐135b with the ER status has been demonstrated in a study by Enerly and colleagues utilizing the MicMa cohort (Enerly et al., 2011). However, the novel aspect in our study is, that we further characterized the association of miR‐135b with ERα by dichotomizing the tumors according to the level of the ERα protein and showed that the levels of miR‐135b gradually decrease as the levels of ERα protein increase. Our analyses also included the METABRIC data set (Dvinge et al., 2013), further confirming our findings.

When exploring the expression of miR‐135b, ERα and AR in cell lines we observed that the expression of ERα and AR is lower in breast and prostate cell lines with a higher miR‐135b expression, which nicely correlated with our observations from in vivo tumor samples. In the VCaP cell line both AR and miR‐135b expression was high and we can only speculate that the regulation may be different. In contrast to the other cell lines, the VCaP cells harbor an amplified AR gene and its expression is consequently much higher (Liu et al., 2008). Thus, in VCaP cells the regulation of AR by miR‐135b might be influenced by the high target:miRNA ratio and the threshold for regulation different (Mukherji et al., 2011). Additional reasons might also include structural accessibility of miRNA target site as miR‐135b is not expected to be the only regulator of the receptor (Wan et al., 2014).

MiR‐135b has been shown to be regulated by DNA demethylation and NF‐kB signaling (Lin et al., 2013). He and colleagues have showed a time‐dependent decrease in miR‐135b levels upon estradiol treatment in COLO205 cells (a colorectal cell line), and suggested that the estrogen regulation of miRNA expression is via ERβ (He et al., 2012). We did an attempt to explore whether the hormone receptors play a role in the regulation of miR‐135b levels by overexpressing ERα/AR and conducting qRT‐PCR analyses on miR‐135b. We did not observe effects of ERα overexpression or knockdown on miR‐135b levels, and these data are in concordance with results from He and colleagues, where also MCF‐7 cells were treated with estradiol for 12 h, and no significant changes in miR‐135b expression were observed (data not shown) (He et al., 2012). In concordance with previous studies, we also observed that miR‐135b is not regulated by AR at 18 h and 24 h time points (data not shown) (Mo et al., 2013).

Our study is the first to show that miR‐135b targets ERα 3′UTR in BCa. To our knowledge miR‐135b has been characterized to directly interact, through their 3′UTR sequence, with only two genes, MID1 and MTCH2, in BCa previously (Arigoni et al., 2013). Other direct targets of miR‐135b that have been reported are adenomatous polyposis coli (APC), transforming growth factor, beta receptor II (TGFBR2) and death‐associated protein kinase 1 (DAPK1) in colorectal cancer, leucine zipper putative tumor suppressor 1 (LZTS1), large tumor suppressor kinase 2 (LATS2), and forkhead Box O1 (FOXO1) in HEK293T cells, and LZTS1 in non‐small‐cell lung cancer (Lin et al., 2013; Matsuyama et al., 2011; Nagel et al., 2008; Valeri et al., 2014). MiR‐135b has through the regulation of its target genes, MID1 and MTCH2 in BCa, been indicated as a mediator of BCa progression, but this study was not able to show an effect on proliferation in BALB‐neuT line 1 (TUBO) cells in vitro 24, 48, 72, and 96 h after downregulation or overexpression of miR‐135b (Arigoni et al., 2013).

Our data suggest, that when overexpressed, miR‐135b decreases the growth of ERα+ BCa cells and AR+ PCa cells. MiR‐135b overexpression does not however seem to affect the growth of hormone negative breast or prostate (cancer) cells although conclusive results would require additional cell line to be tested. These data are in concordance with a previous study where miR‐135a, was reported to decrease proliferation in renal cell carcinoma at 72 h post‐transfection by targeting the c‐Myc oncogene (Yamada et al., 2013). In contrast, Zhang and colleagues showed that miR‐135b slightly increased cell proliferation in vitro by overexpressing miR‐135b in CAL27 cells (head and neck squamous cell carcinoma, HNSCC), and following growth for five days (Zhang et al., 2013). MiR‐135b inhibition has been shown to reduce tumor angiogenesis and growth in vivo of Karpas 299 (anaplastic large cell lymphoma, ALCL) subcutaneous tumors, transplanted into SCID mice (Matsuyama et al., 2011). Thus the effects of miR‐135b be can be very context dependent and our results show that miR‐135b reduces growth in vitro in this setting. This effect seems to correlate with hormone receptor expression further highlighting the context specific functions of miRNAs. The discrepancy between these different results of miR‐135b effect on growth may reflect different receptor status of cells and growth conditions, such as 2D cell culture and in vivo.

Overexpression of miR‐135b reduced the growth of AR+ PCa cells more potently than siRNA for AR did, and thus the role of other miR‐135b targets intrigued us. Consequently, we performed Illumina gene expression array analyses upon miR‐135b overexpression to identify additional miR‐135b target genes. Through these analyses we identified 64 deregulated genes by miR‐135b overexpression at all three time points. These genes linked to a set of genes upregulated by postradiation tumor escape (Monnier et al., 2008). The gene set includes MYD88 and TRAF4, linking to NF‐kB signaling. As NF‐kB has been shown to regulate the expression of miR‐135b (Lin et al., 2013), this may suggest a potential feedback‐loop, if miR‐135b would downregulate NF‐kB activating factors. This would be an interesting path to study, knowing the central role of NF‐kB in regulation of apoptosis, proliferation, inflammatory and immune responses, and the fact that an inverse correlation between ERα and NF‐kB signaling has been described (reviewed by Sas et al., 2012).

To narrow down the list for validation studies we however turned to commonly used prediction programs, which could agree on five putative miR‐135b targets. This strict criteria however, might leave out many true targets because miRNA target prediction algorithms tend to miss miRNA regulation of longer 3′UTR sequences as well as miRNAs binding to 5′UTRs and coding regions, but also because different prediction programs use different settings (Lytle et al., 2007; Ritchie et al., 2009).

We analyzed the expression of the five predicted genes in BCa tumors, and identified specifically that there was an inverse correlation of miR‐135b and HIF1AN expression. As HIF1α is negatively regulated by HIF1AN (also called FIH1) (Mahon et al., 2001), we wanted to continue our studies on the link between miR‐135b, the hormone receptors and the HIF1α regulator HIF1AN. As tumors growing in preirradiated tissues have been shown to be more metastatic and to have an increased fraction of hypoxic cells (Monnier et al., 2008), HIF1AN appeared as an interesting target of miR‐135b. Moreover the modulation of HIF1AN has been suggested as a potential treatment of HIF1α driven aggressive PCa (Shaida et al., 2011). HIF1AN expression and differential cellular localization has been shown to be associated with poor prognosis in PCa and with shorter survival in BCa (Shaida et al., 2011; Tan et al., 2007). The expression of miR‐135b has been shown to increase in response to oxidative stress, in the primary hippocampal neurons in Alzheimer's disease, suggesting that miR‐135b is involved in the stress response and regulation of oxygen homeostasis (Xu et al., 2012). MiR‐135b has also been indicted as a tumor promoter in a genetically defined mouse model of HNSCC by activating the HIF1α pathway (Zhang et al., 2013). As estrogen and androgen furthermore have been shown to regulate HIF1α (George et al., 2012; Sheflin et al., 2004) the role of miR‐135b on many of the central players in these pathways could be of importance.

We identified miR‐135b as a regulator of HIF1AN proteins levels in both BCa and PCa. These results are in concordance with the results of Valeri and co‐workers who identified HIF1AN as a direct target of miR‐135b in colorectal cancer (Valeri et al., 2014) Our experiments were, as theirs, conducted under normoxic condition, when HIF1α is negatively regulated by HIF1AN (Mahon et al., 2001). They show that miR‐135b is positively regulating the growth of colorectal cancer in vitro and in vivo (Valeri et al., 2014), while Pelletier and co‐workers show that silencing of HIF1AN decreases cell proliferation in vitro and tumor growth in vivo of colon adenocarcinoma (LS174) and melanoma (A375) cells (Pelletier et al., 2012). Our results in ERα+ BCa cells and AR+ PCa cells indicate a growth reducing effect of miR‐135b in 2D cell culture. MiR‐135b is furthermore, predicted to directly target HIF1α (TargetScan 6.2), and that could also play a role on the miR‐135b effect on growth under different conditions. Further indications of the differential effects under different conditions come from Umezu and co‐workers, who during the revision of this manuscript published a direct link between exosomal miR‐135b and HIF1AN, as confirmed by luciferase assays. They show that exosomal miR‐135b from hypoxic multiple myeloma enhanced endothelial tube formation and thus angiogenesis (Umezu et al., 2014).

Androgen withdrawal has been shown to increase ERα expression in human prostatic stromal cells (Kruithof‐Dekker et al., 1996) and thus, it has been suggested that also ER/estrogens would be important for the malignant tumor progression (Bonkhoff et al., 1999; Harkonen and Makela, 2004; Santti et al., 1994). The androgen signaling pathway is critical in breast carcinogenesis (Yeh et al., 2003) and studies have indicated a role for AR also in normal and malignant breast (reviewed in Hickey et al., 2012). The role of miR‐135b regulation of ERα in PCa and AR in BCa and the interplay between all these factors would thus need further attention. Additional studies are also necessary to unravel the role of miR‐135b on these signaling pathways in different settings and for example under hypoxic conditions, but are beyond the scope of this study. Taken together, our results provide novel insight and a potential mechanism of cross‐talk between the hormonal and hypoxic pathways, with major significance for tumor biology and clinical applications for cancer therapy.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

The following are the supplementary data related to this article:

Supplementary Table S1 The table shows the genes with a changed expression upon miR‐135b overexpression in LNCaP cells at three time points (12, 24 and 36 h), compared to control. Up‐ and downregulated genes at each time point are listed according to ratio in separate sheets. Values are presented as normalized log2 (miR‐135b/Scr).

Supplemantary Table S2 The table lists the overlapping up‐ or downregulated genes at all time points (12, 24, and 36 h). Up‐ and downregulated genes are listed according to ratio in separate sheets. Values are presented as normalized log2 (miR‐135b/Scr).

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data

Supplementary data 1.

1.1.

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2015.03.001.

Aakula Anna, Leivonen Suvi-Katri, Hintsanen Petteri, Aittokallio Tero, Ceder Yvonne, Børresen-Dale Anne-Lise, Perälä Merja, Östling Päivi, Kallioniemi Olli, (2015), MicroRNA-135b regulates ERα, AR and HIF1AN and affects breast and prostate cancer cell growth, Molecular Oncology, 9, doi: 10.1016/j.molonc.2015.03.001.