miRNA Profiling Reveals miRNA-130b-3p Mediates DENND1A Variant 2 Expression and Androgen Biosynthesis

Abstract

Polycystic ovary syndrome (PCOS) is a common endocrine disorder of reproductive-age women involving overproduction of ovarian androgens and, in some cases, from the adrenal cortex. Family studies have established that PCOS is a complex heritable disorder with genetic and epigenetic components. Several small, noncoding RNAs (miRNAs) have been shown to be differentially expressed in ovarian cells and follicular fluid and in the circulation of women with PCOS. However, there are no reports of global miRNA expression and target gene analyses in ovarian theca cells isolated from normal cycling women and women with PCOS, which are key to the elucidation of the basis for the hyperandrogenemia characteristic of PCOS. With the use of small RNA deep sequencing (miR-seq), we identified 18 differentially expressed miRNAs in PCOS theca cells; of these, miR-130b-3p was predicted to target one of the PCOS genome-wide association study candidates, differentially expressed in neoplastic vs normal cells domain containing 1A (DENND1A). We previously reported that DENND1A variant 2 (DENND1A.V2), a truncated isoform of DENND1A, is upregulated in PCOS theca cells and mediates augmented androgen biosynthesis in PCOS theca cells. The comparison of miR-130b-3p in normal and PCOS theca cells demonstrated decreased miR-130b-3p expression in PCOS theca cells, which was correlated with increased DENND1A.V2, cytochrome P450 17α-hydroxylase (CYP17A1) mRNA and androgen biosynthesis. miR-130b-3p mimic studies established that increased miR130b-3p is correlated with decreased DENND1A.V2 and CYP17A1 expression. Thus, in addition to genetic factors, post-transcriptional regulatory mechanisms via miR-130b-3p underly androgen excess in PCOS. Ingenuity^® Pathway Analysis Core Pathway and Network Analyses suggest a network by which miR-130b-3p, DENND1A, the luteinizing hormone/choriogonadotropin receptor, Ras-related protein 5B, and signaling pathways that they potentially target may mediate hyperandrogenism in PCOS.

Polycystic ovary syndrome (PCOS) is a common endocrine disorder that affects 5% to 10% of reproductive-age women across multiple ethnic groups worldwide (1–7) and is by far the most common disorder involving overproduction of sex steroids (8). The disorder is heterogeneous, characterized by a broad spectrum of reproductive, metabolic, and endocrinological features, including hyperandrogenism, anovulation, and infertility and the presence of multiple small, subcortical follicular cysts embedded in bilaterally enlarged ovaries (9, 10). Other endocrinopathies and metabolic features commonly observed in patients with PCOS include abnormal glucose metabolism, insulin resistance, dyslipidemia, and obesity. Women with PCOS are known to have higher prevalence rates of type 2 diabetes, metabolic syndrome, and cardiovascular disease (1, 11). Family-based studies revealed that PCOS is a complex genetic disorder (12, 13). There is also evidence that strongly suggests that epigenetic and post-transcriptional gene-regulatory factors influence the complex pathophysiology of PCOS (14–19). Most consensus diagnosis schemes recommend the presence of hyperandrogenism (clinical or biochemical) as an essential characteristic for reaching a PCOS diagnosis, and it is considered the hallmark of PCOS. Cross-sectional case-control studies suggested that hyperandrogenism is implicated in both the metabolic and reproductive morbidities of PCOS and may be a common link between the two (20–22).

Ovarian androgen production is increased in PCOS, and studies have shown that theca cells from women with PCOS secrete more androgen than theca cells from regularly ovulating women (23–28). Thus, the understanding of gene expression and its regulation in theca cells is central to the understanding of PCOS pathophysiology, or at least hyperandrogenism associated with PCOS. Wood et al. (29, 30) demonstrated that PCOS theca cells have a characteristic molecular signature compared with normal theca cells. Although these results revealed altered gene expression in PCOS theca compared with normal theca cells, it is not yet known whether the altered expression profile is the result of transcriptional or post-transcriptional regulatory mechanisms.

miRNAs are small (20 to 24 nucleotides), single-stranded, noncoding, regulatory RNA molecules. They are involved in post-transcriptional regulation of gene expression either by complimentary binding to the 3′-untranslated region (3′UTR) of their target mRNA, thereby inhibiting translation (31), or by inducing mRNA degradation (32). The post-transcriptional mechanism operationalized for downregulation of gene expression depends on the binding target sequence. mRNA degradation occurs in the case of complete or near-complete complementarity between miRNA and the target, whereas repression or inhibition of translation occurs if there is not sufficient complementarity for cleavage (32). Although not much is known about the roles of miRNA in the hyperandrogenemia of PCOS, there have been studies providing evidence of differential miRNA expression in the ovarian stroma, endometrium, follicular fluid, and granulosa cells of women with PCOS (33–45). Metformin, an insulin-sensitizing drug, which is used as a treatment to lower insulin and androgen levels and initiate ovulation in women with PCOS, changes global miRNA expression patterns (46–48). miRNAs are also reported to be stably expressed in encapsulated vesicles or free circulating in the serum, plasma, urine, and saliva, making them potential noninvasive biomarkers for PCOS diagnosis (49, 50). Although a recent study compared the differential expression of a cohort of miRNAs using a limited miRNA-microarray platform in intact theca isolated from wedge resections of ovaries of normal cycling women and women with PCOS (51), there are no reports that have explored differential miRNA expression in normal and PCOS theca cells using global miRNA deep sequencing combined with functional analyses to identify miRNA that underlies increased androgen biosynthesis and gene expression in PCOS.

Several genome-wide association studies (GWAS) have identified loci significantly associated with PCOS. In a two-part study, GWAS on Han-Chinese populations identified 11 candidate loci (52, 53). Subsequent GWAS on European populations confirmed associations of some of the Han-Chinese GWAS loci and also identified additional candidate loci (54, 55). Together, four GWAS (52, 53, 56–59) and PCOS GWAS meta-analyses (60) have identified 22 candidate loci/genes for PCOS, including differentially expressed in neoplastic vs normal cells domain containing 1A (DENND1A). Although some of these loci encompass plausible PCOS candidate genes, the differential expression and the molecular mechanisms by which most of the identified genes contribute to PCOS are not clearly understood. We previously reported that a truncated splice variant (variant 2) of DENND1A, DENND1A.V2, is increased in PCOS theca cells, with no change in (variant 1) DENND1A.V1 (full-length isoform) expression (61). Forced expression of DENND1A.V2 in normal theca cells increased cytochrome P450 17α-hydroxylase (CYP17A1) expression and androgen production, whereas knockdown of DENND1A.V2 expression in PCOS theca cells reduced CYP17A1 expression and androgen synthesis, implicating DENND1A.V2 in the regulation of steroidogenesis and the PCOS phenotype (61). The mechanisms underlying the increased DENND1A.V2 expression in PCOS theca cells, as well as the altered transcriptome signature in PCOS, remain to be elucidated. One possibility is that these alterations result, in part, from differential expression of miRNAs. However, a detailed study of miRNA expression and target gene analysis has not been performed in normal and PCOS theca cells. Moreover, the presumptive role of differentially expressed miRNAs in PCOS theca cells on increased DENND1A.V2 and CYP17A1 mRNA and augmented androgen biosynthesis has not been examined.

In the current study, miRNA expression profiles of human theca cell cultures established from women with PCOS and without the disease were determined using next-generation miR-seq. Target gene analysis of the differentially expressed miRNA was focused on PCOS candidate genes identified by GWAS (52–55). Of these, we focused on the expression profiles of miR-130b-3p, because it was highly predicted target DENND1A and specifically, the DENND1A.V2 3′UTR. These studies provide evidence that differential expression of miR-130b-3p in normal and PCOS theca cells is associated with the expression of DENND1A.V2 and CYP17A1 mRNA and androgen biosynthesis. Ingenuity® Pathway Analysis (IPA) Core Pathway and Network Analyses identified a common network, whereby the predicted interactions of miR-130b-3p with the PCOS GWAS candidates DENND1A.V2, luteinizing hormone/choriogonadotropin receptor (LHCGR), and Ras-related protein 5 (RAB5B), as well as signaling components, mediate the increased gene expression required for the hyperandrogenia in PCOS.

Materials and Methods

Cell culture

Human theca interna tissue was isolated from the ovaries of age-matched, normal cycling women (n = 7) and women with PCOS (n=7) using a protocol approved by the Institutional Review Board of the Pennsylvania State College of Medicine, as previously described (23, 24, 61, 62). As a standard of care, oophorectomies were performed during the luteal phase of the cycle. Theca cells from normal cycling and PCOS follicles were isolated and grown, as we have previously reported in detail (63, 64). The theca cell preparations used in these studies have been described and characterized previously (23, 24, 29, 30, 63–67). The steroidogenic phenotypes of the normal and PCOS theca cells have been reported to result from the inherent properties of the cells, rather than the cycle phase at the time that they were isolated (24, 65, 68). PCOS and normal ovarian tissue came from age-matched women, 38 to 40 years old. The diagnosis of PCOS was made according to the National Institutes of Health consensus guidelines (4, 69), which include hyperandrogenemia, oligo-ovulation, and polycystic ovaries and the exclusion of 21α-hydroxylase deficiency, Cushing syndrome, and hyperprolactinemia. All of the PCOS theca cell preparations came from ovaries of women with fewer than six menses per year and elevated serum total testosterone or bioavailable testosterone levels (24, 65). Each of the PCOS ovaries contained multiple subcortical follicles of <10 mm in diameter. The control (normal) theca cell preparations came from ovaries of fertile women with normal menstrual histories, menstrual cycles of 21 to 35 days, and no clinical signs of hyperandrogenism. Neither PCOS nor normal subjects were receiving hormonal medications at the time of surgery. Indications for surgery were dysfunctional uterine bleeding, endometrial cancer, and pelvic pain. Experiments comparing PCOS and normal theca were performed using fourth-passage (31 to 38 population doublings) theca cells isolated from individual size-matched follicles obtained from age-matched subjects in the absence of in vivo stimulation. The use of fourth-passage cells allowed us to perform multiple experiments from the same patient population and were propagated from frozen stocks of primary and second passage cells in the media described above. As such, the passaged normal and PCOS theca cells are not cell lines. The passage conditions and split ratios for all normal and PCOS cells were identical. Fourth passage normal and PCOS theca cells for these studies were grown until subconfluent and were treated with and without 20 μM forskolin for 16 hours in defined serum-free media (61, 62).

For experiments using the adrenocortical H295R cell line that were obtained from American Type Culture Collection (Manassas, VA) (70), we followed American Type Culture Collection’s recommended media and culture conditions. H295R cells were grown to subconfluency and treated with and without 20 μM forskolin in the defined serum-free media, used routinely for normal and PCOS theca cells (61, 62).

miRNA global deep sequencing in normal and PCOS theca cells

miRNA deep sequencing was performed using Illumina technology (by LC Sciences, Houston, TX; www.lcsciences.com) on theca cell RNA from normal cycling women (n = 4) and women with PCOS (n = 4), each treated with (F) and without (C) 20 μM forskolin. Total RNA was extracted using Trizol reagents. Small RNA library prep, cDNA deep sequencing, and initial data analysis, including alignment of raw reads to miRBase (22 release), and normalization of count data were performed by LC Sciences. The log₂-normalized expression levels for each miRNA were used for statistical analyses after filtering for human miRNA and can be accessed via the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=cdyxguqsljaltit&acc=GSE92862).

To perform miRNA discovery sequencing analysis for each miRNA, a one-way ANOVA model was fit to the log₂-normalized expression levels to identify miRNA differentially expressed among the four groups: normal untreated control (Normal-C), normal forskolin treated (Normal-F), PCOS untreated control (PCOS-C), and PCOS forskolin treated (PCOS-F). After the fitting of the ANOVA models for each miRNA, linear contrasts were used to test for differences between the following group comparisons: Normal-C vs Normal-F, PCOS-C vs PCOS-F, Normal-C vs PCOS-C, Normal-F vs PCOS-F, and untreated (Normal-C and PCOS-C) vs treated (Normal-F and PCOS-F). Meta-analysis was performed on miRNA, common between and being significant (P < 0.05) in both group comparisons (Table 1; Normal-C vs PCOS-C and Normal-F vs PCOS-F). Their individual P values were combined using Fisher’s method for combining P values to get an overall comparison of miRNA expression between PCOS and normal theca cells. Subsequently, the combined P values were used to obtain the false discovery rate (FDR) using the Benjamini and Hochberg method. miRNAs with a statistically significant expression difference between both pairwise comparisons (Normal-C vs PCOS-C and Normal-F vs PCOS-F), a meta-analysis combined P < 0.05, and FDR < 0.05 were considered significant. All statistical analyses for the miRNA profiling were performed in R package (The R Project for Statistical Computing; http://r-project.org/). A heatmap for this data (Fig. 1) was prepared by LC Sciences, using the mean expression data from four preparations of normal (N1, N2, N3, N4) and PCOS (P1, P2, P3, P4) theca cells, treated in the absence (C) and presence (F) of 20 μM forskolin.

Table 1.

Differentially Expressed miRNAs in Normal and PCOS Theca Cells

miRNA (hsa)	Normal C Mean	PCOS C Mean	PCOS/Normal (C)	P Value P vs N (C)	Normal F Mean	PCOS F Mean	PCOS/Normal (F)	P value P vs N (F)	Meta P Value	Meta FDR
miR-501-3p	92.43	583.10	6.31	2.13E-06	97.84	544.31	5.56	5.20e-06	2.90E-10	5.29E-07
miR-100-5p	40,687.07	198,820.24	4.89	4.74E-06	36,212.81	141,097.22	3.90	2.83e-05	3.19E-09	2.43E-06
miR-409-5p	87.38	810.37	9.27	4.66E-06	146.62	954.57	6.51	3.65e-05	4.00E-09	2.43E-06
miR-125a-3p	391.08	271.26	0.69	9.19E-06	448.17	340.06	0.76	1.17e-04	2.32E-08	1.06E-05
miR-1271-5p	47.68	354.57	7.44	8.83E-05	45.34	254.33	5.61	3.70e-04	5.95E-07	2.17E-04
miR-1301-3p	22.93	61.96	2.70	1.04E-03	21.05	64.88	3.08	2.87e-04	4.79E-06	1.45E-03
miR-130b-3p	1057.38	644.83	0.61	3.63E-04	995.84	714.19	0.72	6.70e-03	3.39E-05	8.23E-03
miR-99b-5p	15,624.57	43,641.17	2.79	7.42E-04	15,478.35	36,681.54	2.37	3.51e-03	3.61E-05	8.23E-03
miR-127-3p	24,794.80	53,267.35	2.15	1.72E-04	38,179.62	54,137.86	1.42	2.86e-02	6.45E-05	1.31E-02
miR-148b-5p	190.69	132.42	0.69	1.55E-03	268.62	200.59	0.75	6.33e-03	1.23E-04	2.18E 02
miR-654-5p	322.62	779.03	2.41	1.72E-04	409.07	561.48	1.37	6.35e-02	1.35E-04	2.18E-02
miR-195-5p	175.08	80.40	0.46	8.74E-04	122.19	75.29	0.62	1.33e-02	1.44E-04	2.18E-02
miR-744-5p	887.00	1317.24	1.48	2.10E-03	1012.80	1374.08	1.36	1.00e-02	2.48E-04	3.29E-02
miR-1293-p3	6.23	12.78	2.05	8.56E-03	6.84	15.82	2.31	2.80e-03	2.78E-04	3.29E-02
miR-410-3p	3016.03	5096.84	1.69	1.30E-03	3167.66	4458.61	1.41	1.92e-02	2.90E-04	3.29E-02
miR-4524a-5p	18.37	5.74	0.31	9.42E-04	13.80	6.82	0.50	2.67-02	2.91E-04	3.29E-02
miR-502-3p	34.94	62.00	1.77	1.52E-02	30.01	66.75	2.22	1.75e-03	3.07E-04	3.29E-02
miR-494-3p	102.78	260.35	2.53	5.50E-03	130.48	339.02	2.60	5.95e-03	3.71E-04	3.76E-02

miRNA deep sequencing on theca cell RNA from normally cycling women (n = 4) and women with PCOS (n = 4), each treated with and without 20 μM forskolin. For miRNA discovery sequencing analysis, a one-way ANOVA model was fit to the log₂-normalized expression levels to identify miRNA differentially expressed among the four groups: Normal-C, Normal-F, PCOS-C, and PCOS-F. We identified 18 miRNAs that were observed to be differentially expressed between normal and PCOS theca cells. Thirteen of these 18 miRNAs were upregulated (>1.0), and five miRNAs were downregulated (<1.0) in PCOS theca cells compared with normal cells [Fisher combined P value < 0.05, and false-discovery rate (FDR) < 0.05 are considered significant]. For each of these 18 miRNAs, we present mean expression values for normal and PCOS cells under C and F-stimulated conditions, as well as the ratio of mean expression for each and statistical significance.

Figure 1.

Heatmap and hierarchical clustering of the most highly significant 18 differentially expressed miRNAs in normal and PCOS theca cells. Expression profiling of relative gene expression of the 18 miRNAs found to have the most significant differences between theca cells propagated from four normal cycling (N1 to N4) and four PCOS (P1 to P4) women (P < 0.001) treated in the absence (c) and presence (f) of 20 μM forskolin. Shown in the color key, variations in light to darker red indicate increases in relative miRNA expression (>2.0), whereas variations from light to darker green indicate decreases (<2.0) in relative miRNA expression. Hierarchical clustering of miRNA expression, shown by the bars (left), suggests that two separate related families of miRNAs are involved in differential miRNA expression in normal and PCOS theca cells.

miRNA target identification

The miRNA target prediction filter in Qiagen’s IPA tool (Qiagen, Redwood City, CA; www.qiagen.com/ingenuity), miRBase (http://www.mirbase.org), and miRNA TargetMiner (https://www.isical.ac.in/∼bioinfo_miu/targetminer20.htm) was used to identify potential target genes for the 18 miRNAs (Table 2) that were differentially expressed between PCOS and normal theca cells (FDR < 0.05). miRNA-target gene relationships, having a confidence level of “Experimentally Observed” or “High-Predicted” defined by this analyses, were primarily considered to be significant.

Table 2.

miRNA Target Identification Identified miR-130b-3p as a Possible Candidate Mediating DENND1A.V2 Expression in Normal and PCOS Theca Cells

ID (hsa)	Targeted Gene
miR-130b-3p	DENND1A***
miR-130b-3p	ZNF217
miR-130b-3p	RAB5B
miR-130b-3p	LHCGR
miR-130b-3p	ERBB3
miR-130b-3p	KCNA4
miR-195-5p	INSR
miR-195-5p	YAP1
miR-195-5p	HMGA2
miR-654-5p	GATA4
miR-1271-5p	ERBB3

Target identification analyses (miRBase, IPA target filter, and miRNA TargetMiner) predicted that 5 of the 18 differentially expressed miRNAs targeted 9 of the 18 PCOS GWAS loci that have been validated in several ethnic populations. As indicated in boldface, miR-130b-3p was found potentially to target 5 of the PCOS GWAS loci, including DENND1A, ZNF217, RAB5B, LHCGR, and ERBB3. Of particular interest to these studies, miR-130b-3p was observed to target DENND1A***, a functional PCOS GWAS candidate shown to mediate excess androgen production in PCOS theca cells.

Quantitative RT-PCR-based analysis of miRNA expression

The expression profile of miR-130b-3p in normal and PCOS theca cells was validated in vitro using the TaqMan MicroRNA Assays (Thermo Fisher Scientific, Pleasanton, CA), following the manufacturer’s instructions. For the miRNA-quantitative RT-PCR (qRT-PCR) assays, total RNA was extracted from the established theca cell cultures (normal: n = 7; PCOS: n= 7; treated with and without forskolin) using TRI Reagent (Millipore-Sigma, St. Louis, MO). In brief, 10 ng RNA was reverse transcribed using the target (miRNA)-specific stem-loop RT primer and the TaqMan MicroRNA RT kit (Thermo Fisher Scientific). The cDNA was then amplified by real-time qRT-PCR using target-specific TaqMan primer-probe mix. The qRT-PCR was performed in triplicate per sample, and RNA U6 small nuclear 1 was used for normalization of the miR-130b-3p qRT-PCR expression data. The mean expression value for each miRNA was divided by the mean RNA U6 small nuclear 1 expression value to normalize each sample (Fig. 2).

Figure 2.

miR-130b-3p expression is decreased, and DENND1A.V2 expression is increased in PCOS theca cells. (A) To validate the decreases in expression of miR-130b-3p, observed from miRNA expression-profiling analysis, miR-130b-3p expression was analyzed by real-time qRT-PCR in theca cells propagated from seven individual normal women and seven individual women with PCOS treated in the absence (c) and presence (f) of 20 μM forskolin for 16 h. miR-130b-3p expression was decreased in PCOS theca cells compared with normal theca cells under basal conditions (**P < 0.01) and forskolin-stimulated conditions (*P < 0.05). (B) The use of mRNA harvested from the identical patients examined above, as we previously reported that DENND1A.V2 expression was increased in PCOS theca cells compared with normal theca cells under both normal (**P < 0.001) and forskolin-stimulated (***P < 0.0001) conditions.

Real-time qRT-PCR analyses of DENND1A.V1, DENND1A.V2, and CYP17A1

Quantitation of DENND1A.V1, DENND1A.V2, and CYP17A1 mRNA abundance was determined using the Single Step Brilliant III Ultra Fast qRT-PCR kit (Agilent, Santa Clara, CA), using primer and probe sets, as we have previously described in detail (61). The gene-specific, one-step PCR was carried out in duplicate for each mRNA sample and for a series of dilutions in an Agilent AriaMx Real-Time PCR System, according to the manufacturer’s instructions for this instrument, as previously described (61). TATA-binding protein mRNA used for normalization and the mean-expression value for each mRNA was divided by the mean TATA-binding protein expression value to normalize each sample (61).

Quantitation of dehydroepiandrosterone

ELISA for dehydroepiandrosterone (DHEA) was performed on cell culture media collected from the identical seven normal and seven PCOS theca cell preparations (treated with and without 20 μM forskolin) that were used in parallel studies to examine miR-130b-3p, DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA accumulation. The DHEA ELISAs were performed without organic solvent extraction using kits from DRG International, Inc. (Springfield, NJ), as described by the manufacturer’s protocol, and normalized by cell count, as previously described (24, 61). The specificity and crossreactivity of the specific DHEA ELISA antibody [RRID: AB_2800459 (71)] were determined by DRG International Inc. DHEA ELISA antibody displayed 100% crossreactivity with DHEA and the following crossreactivity with 17α-hydroxy-pregnenolone, 0.072%; androsterone, 0.056%; desoxycorticosterone, 0.052%; progesterone, 0.023%; and pregnenolone, 0.013%, and <0.01% for 11-desoxycortisol, corticosterone, DHEA sulfate, testosterone, and 5α-dihydrotestosterone.

Analysis of DENND1.V2 and DENND1A.V1 mRNA accumulation in response to hsa-miR-130b-3p mimic in H295R cells

The highly androgenic human adrenocortical H295R cell line was used for our miRNA mimic studies (72), because of the limitations of using human theca cells described below. Human adrenocortical H295R cells were transfected with 75 pM mirVana hsa-miR-130b-3p mimic or negative control-1 mimic (Life Technologies, Waltham, MA) using RNAiMAX transfection reagents (Invitrogen, Waltham, MA) using the manufacturer’s protocol, in Opti-MEM serum-free medium (Thermo Fisher Scientific, Pleasanton, CA). Eight hours following transfection, the cells were treated, with and without 20 μM forskolin, and total RNA from the cells was harvested 48 hours thereafter. The effects of miR-130b-3p mimic on DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA accumulation were quantified, as described above. The human adrenocortical H295R cell line was used for these miR-mimic studies instead of normal or PCOS theca cells, as the reagents/conditions required for miR-mimic transfection were toxic to passaged theca cells when the experiments were preformed. The H295R cell line also has a relatively high transfection efficiency (>25%) compared with early-passaged human theca cells (<10% to 15%) under the most moderate of conditions. As we have previously reported, forced DENND1A.V2 mRNA and protein expression in the adrenocortical H295R cell line via adenoviral infection increased CYP17A1 gene expression and DHEA production (72), concordant with results observed in normal theca cells (61).

Western blot analysis of DENND1.V2 and DENND1A.V1 protein in miR-130b-3p mimic transfected H295R cells

The adrenocortical H295R cell line was grown until subconfluent and transfected with hsa-miR-130b-3p and negative control mimic (Neg-mimic) and treated, with or without, forskolin, as described above. Forty-eight hours thereafter, the cells were harvested in radioimmunoprecipitation assay, and Western blot analyses were performed using 30 μg protein per lane, as we have previously described (61). An N-terminal Sigma antibody for an intrapeptide sequence of DENND1A [RRID: AB_2092106 (73)], which binds both DENND1A.V1 (119 kDa) and DENND1A.V2 (62 kDa), and a rabbit polyclonal specific for the unique 32 amino acid C-terminal sequence of DENND1A.V2 [RRID: AB_2800456 (74)] were used as primary antibodies, as we have previously reported (61). DENND1A.V2 and DENND1A.V1 protein was visualized using enhanced chemiluminescence (Rockland Immunochemicals) and quantitated using a FluorChem R (ProteinSimple, Pottstown, PA). Both DENND1A.V2 and DENND1A.V1 proteins were normalized using total mammalian target of rapamycin (mTOR) antibody (75), which is not significantly different in normal and PCOS theca cells nor regulated by forskolin treatment (61). These experiments were repeated approximately four times, and the data are presented as cumulative results.

IPA Core Network Pathway Analysis

The miRNA target-prediction filter in Qiagen’s IPA tool was used to identify potential target genes for miR-130b-3p. Core pathway and network analyses using the “Grow” and “Build” network functions available in IPA use dropdown analysis tools and options designed within the IPA analysis tool. This analysis takes into account any known or predicted connections present in the Ingenuity Knowledge Base, which took into account the set parameters defined to be associated with PCOS and miR-130b-3p in an extensive IPA library of previously reported gene/protein/function interactions.

Statistical analysis

For comparisons of miR-130b-3p, DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA in normal and PCOS theca cells, one-way ANOVA was performed to test for differences, and P values were determined by the Bonferroni method for multiple comparisons and between groups Normal-C vs PCOS-C and Normal-F vs PCOS-F, and P < 0.05 was considered statistically significant. Correlations were determined from experiments where miR-130b-3p, DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA and steroid (DHEA) were examined from multiple individual patients under treated and untreated conditions. The Spearman nonparametric correlation coefficient (ρ) was calculated to estimate the degree of correlation between all pairwise comparisons (ρ was considered significant when P < 0.05), using JMP software (SAS, Cary, NC; see Fig. 4). The mRNA and Western analyses data for DENND1A.V2 and DENND1A.V1, following miR-130b-3p mimic transfection, were analyzed by one-way ANOVA, and P values were determined by the Bonferroni method for multiple comparisons.

Figure 4.

Decreased miR-130b-3p expression is correlated with increased DENND1A.V2 mRNA expression and augmented CYP17A1 mRNA expression and DHEA biosynthesis . Real-time qRT-PCR analyses of miR-130b-3p for each of the individual normal (n = 7) and PCOS (n = 7) theca cell preparations were compared with (A) DENND1A.V2, (B) DENND1A.V1, (C) CYP17A1 mRNA expression, and (D) DHEA biosynthesis. (A) There was a significant negative correlation between miR-130b-3p and DENND1A.V2 expression (Spearman ρ = −0.6796, P < 0.001). (B) There was a significant positive relationship between miR-130b-3p and DENND1A.V1 mRNA expression (Spearman ρ = 0.3853, P = 0.0429). (C) There was significant negative correlation between miR-130b-3p and CYP17A1 mRNA abundance (Spearman ρ = −0.6034, P = 0.0001). (D) A significant negative correlation between miR-130b-3p levels and DHEA accumulation was observed in PCOS theca cells. (Spearman ρ = −0.6836, P < 0.0001).

Results

Expression profiling of differentially expressed miRNAs in theca cells isolated from normal cycling women and women with PCOS

For miRNA profiling, miR-seq was performed using libraries prepared from RNA isolated from fourth-passage theca cells, from four individual normal cycling women and four individual women with PCOS treated for 16 hours under control nonstimulated (C) and 20 μM forskolin-stimulated (F) conditions, by LC Sciences (see “Materials and Methods”). A total of 1823 miRNAs were identified from miR-seq to be differentially expressed in normal and PCOS theca cells. After meta-analysis of the Normal-C vs PCOS-C and Normal-F vs PCOS-F pairwise comparisons (described in detail in “Materials and Methods”), 18 miRNAs were found to have the highest statistical differences in expression between theca cells obtained from four normal cycling women and four women with PCOS (Fisher combined P < 0.05) with an FDR <0.05 (Table 1). For each of these 18 miRNAs in Table 1, we present the mean expression values for normal and PCOS cells under control (C) and forskolin (F)-stimulated conditions, as well as the ratio of mean expression for each and statistical significance. Of the most significant 18 miRNAs observed to be differentially expressed, 13 had increased expression, and 5 had decreased expression in PCOS theca cells compared with normal theca cells.

As presented in Fig. 1, a heatmap with hierarchical clustering was prepared using the global normalization expression data and the relative copy numbers of the 18 miRNAs found to have the highest statistical differences between theca cells propagated from four normal cycling women and women with PCOS (P < 0.001) treated in the absence (C) and presence (F) of 20 μM forskolin. Red-color variations from light to darker red indicate increased relative miRNA expression (>2.0), whereas green-color variations from light to darker green indicate decreased (<2.0) relative miRNA expression. As shown in Fig. 1, hierarchical clustering of the 18 differentially expressed miRNAs (indicated on the left) suggests that there are two families of related miRNAs that mediate miRNA expression in normal and PCOS theca cells. Each individual group appears to mediate increased expression in PCOS theca cells in contrast to increased expression in normal theca cells.

Target analyses predict that several of differentially expressed miRNAs in normal and PCOS theca cells target PCOS candidate genes, including DENND1A

The 18 miRNA (listed in Table 1 and Fig. 2) were used for bioinformatics analysis, including target prediction in IPA. The miRNA target-prediction filter in IPA was selected to identify the validated and predicted PCOS GWAS target genes of the differentially expressed miRNA. As shown in Table 2, target prediction revealed that 4 of the 18 (∼22%) of the miRNA that are differentially expressed between normal and PCOS theca cells (i.e., miR-130b-3p, miR-195-5p, miR-654-5p, and miR-1271-5p) were predicted to target 10 of the 22 (45%) of the PCOS GWAS candidate genes (i.e., DENND1A, INSR, LHCGR, RAB5B, ZNF217, YAP1, GATA4, HMGA2, ERBB3, and KCNA4), identified in Han-Chinese (52, 53) and European populations (54, 55) following both IPA analyses, miRBase, and miRNA TargetMiner database analyses. A total of 8366 additional interactions were also identified by IPA. Noticeable in Table 2, miR-130b-3p was predicted to have target interactions with PCOS GWAS candidates DENND1A, ZNF217, RAB5B, LHCGR, ERBB3, and KCNA4. Of particular interest, miR-130b-3p was observed to target DENND1A, which we previously identified as a functional GWAS candidate mediating excess androgen production in PCOS theca cells (61).

miR-130b-3p expression is decreased, and DENND1A.V2 expression is increased in PCOS theca cells compared with normal theca cells

Based on target analyses that predicted that miR-130b-3p targets DENND1A.V2 (Table 2), we used qRT-PCR to examine the differential expression of both miR-130b-3p and DENND1A.V2 in theca cells isolated from seven different normal cycling women and seven different women with PCOS, treated with either 20 μM forskolin or under control, nonstimulated conditions for 24 hours, as described in “Materials and Methods.” As shown in Fig. 2A, miR-130b-3p was observed to be significantly decreased in PCOS theca cells under both control (P < 0.01) and forskolin-treated (P < 0.05) conditions compared with normal theca cells. As shown in Fig. 2B, an examination of DENND1A.V2 mRNA expression in the same seven normal and seven PCOS mRNA samples used to examine miR-130b-3p demonstrated a statistical increase in DENND1A.V2 mRNA abundance in PCOS theca cells compared with normal theca cells. Forskolin treatment had no effect on miR-130b-3p or DENND1A.V2 expression in normal or PCOS theca cells.

miRNA target analysis provided evidence that miR-130b-3p targets the DENND1A.V2 3′UTR

In our initial analysis, miRNA TargetMiner predicted an miR-130b-3p target seed sequence at ∼330 bp in the 3′UTR of DENND1A. However, the DENND1A.V2 3′UTR sequence used in this first approach was not found to be homologous to the DENND1A.V2 splice variant and associated 3′UTR. With the use of a more specific approach, using the miR-130b-3p target sequence and correct DENND1A.V2 C-terminal CDS and 3′UTR, we were able to identify the miR-130b-3p seed sequence (blue), ∼527 bp past the DENND1A.V2 CDS stop codon (ATTGA) using a combination of TargetScan, IPA Confidence, and TargetMiner analyses (Fig. 3). The DENND1A.V2 polyadenylation signal AAAATAAAA is indicated in purple, ∼1221 bp 3′ of the DENND1A.V2 CDS stop codon.

Figure 3.

miRNA target analyses provided evidence that miR-130b-3p targets the DENND1A.V2 3′UTR. An miR-130b-3p seed sequence (blue) was identified ∼527 bp past the DENND1A.V2 CDS stop codon (ATTGA), within its distinct 3′UTR using a combination of TargetScan, IPA Confidence, and TargetMiner analyses. The DENND1A.V2 polyadenylation signal AAAATAAAA is indicated in purple, ∼1221 bp 3′ of the DENND1A.V2 CDS stop codon.

Decreased miR-130b-3p expression is correlated with increased DENND1A.V2 and CYP17A1 mRNA expression and androgen biosynthesis in PCOS theca cells

Based on the miR130b-3p DENND1A.V2 3′UTR target analyses (Fig. 3) and data demonstrating decreased miR-130b-3p expression (Fig. 2A) and increased DENND1A.V2 mRNA expression (Fig. 2B) in PCOS theca cells (Fig. 2), experiments were performed to evaluate the extent that miR-130b-3p expression is correlated with the expression patterns of truncated DENND1A.V2, full-length DENND1A.V1 and CYP17A1 mRNAs, as well as DHEA accumulation. Real-time PCR for DENND1A.V1 and CYP17A1 mRNA was performed using RNA harvested from the same seven normal and seven PCOS theca cell preparations used for the experiments and results presented in Fig. 2, which were treated in the presence (F) and absence (C) of 20 μM forskolin for 16 hours. Parallel experiments were also performed to examine the relationship between miR-130b-3p and DHEA biosynthesis in the same seven normal and seven PCOS theca cell preparations. In Fig. 4A, the comparison of DENND1A.V2 mRNA expression with miR-130-3p for each preparation of normal and PCOS theca cells established that there was a significant negative correlation between miR-130b-3p and DENND1A.V2 mRNA. We also observed a positive relationship between miR-130b-3p and DENND1A.V1 mRNA expression, as shown in Fig. 4B. In addition, there was a significant negative correlation between miR-130b-3p and CYP17A1 mRNA abundance (Fig. 4C), and a significant negative correlation between miR-130b-3p levels and DHEA accumulation was observed in PCOS theca cells (Fig. 4D). These data support the notion that decreased miR-130b-3p expression in PCOS theca cells is correlated with augmented DENND1A.V2 and CYP17A1 mRNA expression and downstream increases in androgen biosynthesis in PCOS theca cells.

Overexpression of an miR-130b-3p mimic in the androgen-producing adrenocortical H295R cell line downregulates DENND1A.V2 and CYP17A1 mRNA expression but has no effect on DENND1A.V1 mRNA expression

As mentioned in “Materials and Methods,” our attempts to transfect human theca cells with miR-130b mimic were not successful, because the transfection reagents were toxic to the cells. Consequently, to explore whether miR-130b-3p could mediate a reciprocal decrease in DENND1A.V2 mRNA expression, we used a surrogate cell system: androgen-producing human H295R adrenocortical cells, which can be transfected with high efficiency. H295R cells express high levels of endogenous DENND1A.V2 mRNA under control nonstimulated conditions (72). We have also previously reported that forced overexpression of DENND1A.V2 in H295R cells results in an increase in CYP17A1 gene expression and androgen biosynthesis (72). The H295R cells were transfected with 75 pM miR-130b-3p mimic and nonspecific Neg-mimic, as described in “Materials and Methods.” These experiments were repeated at least four times. Forty-eight hours thereafter, mRNA was harvested, and DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA expression was quantitated by qRT-PCR. In Fig. 5A, cumulative results from adrenocortical H295R cells transfected with miR-130b-3p mimic demonstrate that DENND1A.V2 mRNA was significantly decreased in cultures treated under both control and forskolin-stimulated conditions compared with Neg-mimic transfected cells. In contrast, the miR-130b-3p mimic had no substantial effect on DENND1A.V1 mRNA (Fig. 5B). CYP17A1 mRNA expression (Fig. 5C) was substantially suppressed in miR-130b-3p transfected cells compared with Neg-mimic in forskolin-treated cells. Forskolin treatment also increased CYP17A1 mRNA in Neg-mimic cells.

Figure 5.

miRNA-130b-3p overexpression in H295R cells downregulates DENND1A.V2 and CYP17A1 mRNA expression but has no effect on DENND1A.V1 mRNA expression. To evaluate the effects of miR-130b-3p mimic on DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA expression, highly androgenic adrenocortical H295R cells were transfected with 75 pM of either miR-130b-3p mimic or Neg-mimic and treated with and without 20 μM forskolin for 48 h. (A) Real-time qRT-PCR analyses of total mRNA, harvested following transfection, demonstrated that miR-130b-3p overexpression results in the downregulation of DENND1A.V2 mRNA expression in control (c *P < 0.05) and forskolin-treated cells (f ***P < 0.001) compared with Neg-mimic control. (B) No significant differences in DENND1A.V1 mRNA expression were observed between Neg-mimic or miRNA-mimic transfected cells. (C) CYP17A1 mRNA expression was decreased in miR-130b-3p transfected cells compared with Neg-mimic control in forskolin-treated cells (****P < 0.0005). Forskolin treatment also increased CYP17A1 mRNA in Neg-mimic cells (^δP < 0.0001).

Differences in the time courses of DENND1A.V2 and CYP17A1 mRNA accumulation in adrenocortical H295R cells transfected with miR-130b-3p mimic

Whereas our initial IPA and miRNA TargetMiner analyses predicted that miR-130b-3p targets DENND1A (and specifically DENND1A.V2), we found no evidence to suggest that miR-130b-3p directly targets CYP17A1. To explore the possibility that miR-130 mimic has temporal effects on DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA expression in H295 cells, H295R cells were transfected with 75 pM miR-130b-3p mimic or nonspecific Neg-mimic for 8 hours and then subsequently treated in the absence (C) and presence (F) of 20 μM forskolin for 24 and 48 hours. mRNA was harvested from the cells at both time points, and DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA accumulation was measured by qRT-PCR, as described in “Materials and Methods.” In Fig. 6, we present cumulative data from four replicate experiments in H295R cells. In Fig. 6A, at both 24 and 48 hours following treatment, miR-130b-3p mimic significantly suppressed DENND1A.V2 mRNA under both control nonstimulated and forskolin-stimulated conditions. In contrast, the time course of DENND1A.V1 accumulation (Fig. 6B) was unaffected by miR-130-3p mimic transfection. Of significant interest, CYP17A1 mRNA accumulation (Fig. 6C) was increased by forskolin at 24 and 48 hours in Neg-mimic transfected cells. Yet, whereas miR-130 mimic transfection had no substantial effect on CYP17A1 mRNA following 24 hours of treatment, at 48 hours, CYP17A1 mRNA was substantially suppressed (Fig. 6C). The finding that miR-130b-3p transfection suppresses DENND1A.V2 mRNA at an earlier time point (24 hours) than CYP17A1 mRNA (48 hours) might suggest that miR-130b-3p mimic has direct effects on DENND1A.V2 mRNA, whereas CYP17A1 mRNA accumulation may be a result of indirect regulation via other intermediary factors and possibly via the effects of miR-130b-3p on DENND1A.V2.

Figure 6.

Differences in the time courses of DENND1A.V2 and CYP17A1 mRNA accumulation in adrenocortical H295R cells transfected with miR-130b-3p mimic. To explore the possibility that miR-130 mimic has temporal effects on DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA expression in H295 cells, H295R cells were transfected with 75 pM miR-130b-3p mimic or nonspecific Neg-mimic and then subsequently treated in the absence (c) and presence (f) of 20 μM forskolin. Following 24 and 48 h of treatments, mRNA was harvested, and DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA accumulation was assessed by qRT-PCR. The results are cumulative data from four replicative experiments in H295R cells. (A) At both 24 and 48 h after treatment, miR-130b-3p mimic significantly suppresses DENND1A.V2 mRNA under both control nonstimulated (24 h: ****P < 0.0001; 48 h: *P < 0.01) and forskolin-stimulated (24 h: *P < 0.01; 48 h: ***P < 0.001) conditions. (B) The time course of DENND1A.V1 accumulation was unaffected by miR-130-3p mimic transfection. (C) Of significant interest, CYP17A1 mRNA accumulation was increased by forskolin at both 24 and 48 h (^δP < 0.001) in Neg-mimic transfected cells. miR-130 mimic transfection had no significant effect on CYP17A1 mRNA following 24 h of treatment; however, following 48 h of treatment, CYP17A1 mRNA was significantly suppressed (****P < 0.0001). These data demonstrate that miR-130b-3p mimic suppresses DENND1A.V2 mRNA at an earlier time point (24 h) than CYP17A1 mRNA (48 h).

Western blot analysis of DENND1A.V2 and DENND1A.V1 protein in H295R cells transfected with miR-130b-3p mimic

To explore the effects of overexpressing miR-130b-3p on DENND1A.V2 and DENND1A.V1 protein expression, adrenocortical H295R cells were transiently transfected with 75 pM miR-130b-3p mimic and nonspecific Neg-mimic for 48 hours in the presence (C) and absence (F) of 20 μM forskolin. In Fig. 7A, we present representative Western analysis of ∼62 kDa DENND1A.V2 and ∼119 kDa DENND1A.V1 in whole-cell extracts isolated from H295R cells following 75 pM Neg-mimic and miR-130b-3p mimic transfections or nonspecific Neg-mimic. mTOR was used for total protein normalization. Quantitative Western blot data from four replicate experiments of H295R cells transfected with 75 pM miR-130b-3p and Neg-mimics for 48 hours presented as the means ± SEM of the relative protein abundance normalized by mTOR of DENND1A.V2 (Fig. 7B) and DENND1A.V1 (Fig. 7C). DENND1A.V2 protein was significantly reduced following transfection with miR-130b-3p compared with Neg-mimic under control and forskolin-stimulated conditions (Fig. 7B). In contrast, DENND1A.V1 protein was not significantly different following miR-mimic transfection in nonstimulated cells and was significantly increased by miR-130b-3p in forskolin-treated cells (Fig. 7C). In Fig. 7D, the ratio of DENND1A.V2/-V1 was observed to be significantly reduced following transfection with miR-130b-3p mimic under control and forskolin-stimulated conditions compared with Neg-mimic. Thus, overexpression of miR-130b-3p in H295R cells converts the cells from a “PCOS phenotype” of increased DENND1A.V2 to that of a “normal cell phenotype” of lower DENND1A.V2 expression. In Fig. 7D, both the relative abundance of DENND1A.V2 protein (Fig. 7B) and the ratio of DENND1A.V2/V1 protein decreased forskolin-stimulated H295R cells transfected with control Neg-mimic. Forskolin treatment significantly increased DENND1A.V1 protein under both Neg-mimic and miR-130b-3p mimic conditions (Fig. 7C). Forskolin treatment similarly augments DENND1A.V1 protein expression in normal and PCOS theca cells (61).

Figure 7.

Western analysis of DENND1A.V2 and DENND1A.V1 protein expression following hsa-miR-130b-3p mimic transfection of the human adrenocortical H295R cell line. (A) Representative Western analysis of ∼62 kDa DENND1A.V2 and ∼119 kDa DENND1A.V1 in whole-cell extracts isolated from H295R cells transfected with 75 pM miR-130b-3p mimic, or nonspecific Neg-mimic, treated in the absence (c) and presence (f) of 20 μM forskolin for 48 h. mTOR was used for total protein normalization. These data are representative of data obtained from four individual replicate experiments. Quantitative Western blot data from four replicate experiments of H295R cells transfected with 75 pM miR-130b-3p and Neg-mimic for 48 h, presented as the means ± SEM of the relative protein abundance normalized by mTOR of (B) DENND1A.V2 and (C) DENND1A.V1. (B) DENND1A.V2 protein was reduced following transfection with miR-130b-3p compared with Neg-mimic under control (**P < 0.001) and forskolin-stimulated (**P < 0.005) conditions. (C) DENND1A.V1 protein was not significantly different following miRNA mimic transfection in nonstimulated cells but increased by miR-130b-3p in forskolin-treated cells (**P < 0.001). (D) The ratio of DENND1A.V2/-V1 was reduced following transfection with miR-130b-3p mimic under control (****P < 0.001) and forskolin-stimulated (**P < 0.01) conditions compared with Neg-mimic. Both the relative abundance of (B) DENND1A.V2 protein and the (D) ratio of DENND1A.V2/-V1 protein decreased in forskolin-stimulated H295R cells (^δP < 0.001) transfected with control Neg-mimic. In contrast, (C) forskolin treatment significantly increased DENND1A.V1 protein under both Neg-mimic (^δP < 0.005) and miR-130b-3p mimic (^εP < 0.0001) conditions.

Ingenuity core pathway and network analyses suggest a common network between miR-130b-3p and PCOS GWAS candidate genes and signaling pathways expressed in normal and PCOS theca cells

The core pathway analyses tool in IPA identifies gene-expression changes and the relationship to cellular processes. In contrast, IPA network analyses use the data derived from core analyses to predict signaling pathways. To explore whether there is a common network among the three PCOS GWAS candidates that miR-130b-3p is known or highly predicted to target, we used the Grow option, which allowed us to identify any known or predicted connections in the current literature present in the Ingenuity Knowledge Base. We used the Build network function in IPA core analysis to create a cellular network. Figure 8 depicts a diagram of the resulting IPA network among miR-130b-3p, DENND1A (dark purple), two other PCOS GWAS candidates (i.e., LHCGR and RAB5B; light purples), as well as the signaling component MAPK1, known or highly predicted to be targets of miR-130b-3p. Also included are signaling components [protein kinase A (PKA) and protein kinase B (AKT)/phosphatidylinositol 3-kinase (PI3K)] and transcription factors (GATA4/6) that have been previously reported to be involved in downstream overexpression of steroidogenic enzyme gene expression [i.e., CYP17A1, cytochrome P450 cholesterol side-chain cleavage (CYP11A1), and 3β-hydroxysteroid dehydrogenase (3β-HSD) type II (HSD3B2); in red], which have been confirmed to be involved in excess androgen biosynthesis in PCOS theca cells. We have included the insulin receptor (INSR) in our diagram in Fig. 8, given that INSR-mediated signaling has be widely accepted to be dysregulated in PCOS (76–78) and the strong genetic association studies with the INSR (77, 79). The resulting theca cell network provides an innovative representation of the putative mechanisms by which decreased miR-130b-3p in PCOS theca cells may regulate the PCOS GWAS candidates DENND1A (DENND1A.V2), LHCGR, and RAB5B and potentially mediate increased CYP17A1, CYP11A1, and HSD3B2 gene expression and hyperandrogenism in PCOS theca cells. As presented in Table 3, IPA core pathway and network analysis of the predicted interactions of miR-130b-3p with the PCOS GWAS candidates (DENND1A.V2, LHCGR, RAB5B), as well as the signaling components MAPK1; PKA; AKT; and CYP17A1, HSD3B2, and CYP11A1, enabled us to identify important functional and biological features with relevance to PCOS. For the associated network functions, including endocrine system development and function, lipid metabolism, and small molecule biochemistry, the IPA network score was 21 (Table 3).

Figure 8.

IPA Network. Ingenuity pathway analyses suggest a common network among miR-130b-3p and PCOS GWAS candidate genes DENND1A, LHCGR, and RAB5B and various signaling pathways in normal and PCOS theca cells that regulate androgen biosynthesis. We present a diagram of the resulting IPA network among miR-130b-3p, DENND1A (dark purple), and the two other PCOS GWAS candidates (i.e., LHCGR and RAB5B) and the signaling component MAPK1, which were predicted to be targets of miR-130b-3p and were previously reported to contribute to a PCOS theca cell phenotype. IPA core pathway and network analyses also identified several signaling pathway components [protein kinase A (Pka) and phosphatidylinositol 3-kinase (PI3K)]; transcription factors (GATA4/6), previously reported to be associated with augmented CYP17A1, CYP11A1, and HSD3B2 gene expression (red); and excess androgen production in PCOS theca cells. The resulting theca cell cellular network provides an illustration of the possible mechanism(s) by which decreased miR-130b-3p in PCOS theca cells may mediate the PCOS GWAS candidates DENND1A, LHCGR, and RAB5B, in addition to their associated signaling components and transcription factors, resulting in increased CYP17A1, CYP11A1, and 3β-hydroxysteroid dehydrogenase (HSD3B2) gene expression and hyperandrogenism in PCOS theca cells. INS, insulin; LHB, luteinizing hormone β.

Table 3.

IPA Core Pathway of Highly Predicted Targets of miR-130b-3p That Include DENND1A.V2, LHCGR, RAB5B, and MAPK1 Signaling Pathways

Top Networks
		Score
Endocrine system development and function, lipid metabolism, small molecule biochemistry		21
Organ development, cellular growth and proliferation, organ morphology and development		7
Top Canonical Pathways
	P Value	Overlap, %
Gs signaling	1.17E-05	2.8
Ovarian cancer signaling	3.59E-05	1.9
cAMP-mediated signaling	1.09E-04	1.3
G-Protein-coupled receptor signaling	1.94E-03	0.7
Caveolar-mediated endocytosis signaling/synaptogenesis signaling pathway	2.48E-03	0.6
Top Diseases and Bio-Functions
	P Value	Molecules
Endocrine system disorders	1.29E-02–9.26E-06	7
Organism injury/abnormalities	4.97E-02–2.67E-06	9
Reproductive system disease	4.87E-02–2.67E-06	7
Cancer	4.87E-02–2.67E-06	9
Tumor morphology	4.48E-02–2.67E-06	3
Physiological System Development and Function
	P Value	Molecules
Embryonic development	8.04E-03–4.48E-04	3
Organ development	2.48E-02–4.48E-04	3
Organismal development	1.56E-02–4.48E-04	4
Endocrine system development and function	1.51E-02–3.21E-06	4
Organ morphology	2.00E-02–4.48E-04	3

Key findings from Core Pathway Analysis (performed in IPA) of the predicted interactions of miR-130b-3p with the three PCOS GWAS candidates (DENND1A.V2, LHCGR, RAB5B), as well as the signaling component MAPK1.

Discussion

PCOS is a multifactorial disorder that has been proposed to involve complex genetic and epigenetic components that can mediate changes in gene expression via transcriptional, post-transcriptional, translational, and post-translational mechanism(s). Recent GWAS in both Han-Chinese and European populations identified candidate loci, some of which have been replicated by multiple laboratories, providing a basis for the molecular dissection of the pathophysiology of PCOS (52–55). Theca cells, the androgen-producing cells of the ovary and central to PCOS pathophysiology, also have evidence of an altered transcriptome profile in PCOS, which could be regulated, at least in part, by miRNAs (29, 30). Although there is one report of an examination of differential miRNA expression in normal and PCOS thecal tissues from wedge resections using a limited microarray platform (51), a detailed study of global miRNA expression using next-generation sequencing high-throughput miR-seq combined with target gene analyses has not been performed in normal and PCOS theca cells. Altered miRNA levels have been associated with several phenotypes in the PCOS spectrum, including diabetes, insulin resistance, and inflammation. An increasing number of studies have shown differential miRNA expression in the ovarian stroma, follicular fluid, cumulus cells, granulosa cells, and endometrium of women with PCOS (33–35, 43–45, 49, 80–82). Our study examines differences in global miRNA expression in miRNA libraries that were sequenced and validated in detail from well characterized theca cell cultures from PCOS and normal cycling women.

The GWAS conducted on Han-Chinese populations (52, 53) and subsequent GWAS on European populations (54, 55) identified additional candidate loci. Together, four GWAS (52, 53, 56–59) and PCOS GWAS meta-analyses (60) have identified 22 candidate loci/genes for PCOS, including DENND1A. However, the functional relevance of the remaining loci with PCOS is not clear, and the exact molecular mechanism by which any of these candidate genes may contribute to the disorder is not known. Genotype–phenotype correlation studies of single-nucleotide polymorphisms identified by GWAS have yielded low-to-modest relative risk ratios (56, 83). This is not unexpected for a complex disorder, such as PCOS, and the establishment of a causal relationship between the genetic variants identified in GWAS and the pathophysiological phenotype is challenging. However, it is important to recognize that whereas some PCOS GWAS gene loci consist of plausible PCOS candidate genes, no one has characterized the differential expression of the enormous number of mRNA transcripts and mRNA isoforms (>150) that is potentially encoded by the 22 GWAS loci in relevant normal and PCOS cells. Moreover, the molecular mechanisms by which most of the identified genes contribute to PCOS are not clearly understood.

In the current study, miRNA expression profiles of human theca cell cultures established from women with PCOS and without the disease were determined using next-generation miR-seq. The data derived from these studies demonstrated that >1800 miRNAs are in normal and PCOS theca cells and identified the 18 most statistically relevant, differentially expressed miRNA in normal and PCOS theca cells (Table 1 and Fig. 1) Of the statistically relevant 18 miRNA that we observed to be differentially expressed in our normal and PCOS theca cells in culture (Table 1), only miR-502-3p was observed to overlap with those identified by the miRNA microarray platform used by Lin et al. (51), which may have not included the 18 miRNA that we found to be important. There was also no overlap between the 18 miRNA that we identified and the differentially expressed miRNA reported in granulosa cells from normal cycling women and women with PCOS (16, 39–41, 43, 84–88) nor was there overlap with miRNA in human follicular fluid that was reported to be associated with PCOS (34). miR-23a and miR-23b were reported to be differentially expressed in plasma from normal cycling women and women with PCOS (89); however, these miRNAs were not significantly associated with PCOS, based on our deep-sequencing study. Whereas there are studies demonstrating that miR-93 is differentially expressed in many PCOS cells and plasma from normal cycling women and women with PCOS, with evidence for decreased expression in PCOS blastocysts and increased expression in PCOS adipocytes (90), granulosa cells (84), as well as in the circulation (91), we did not find substantial differences in miR-93 expression levels between normal PCOS theca cells in our deep sequencing or by TaqMan-based qRT-PCR assays.

Subsequent target gene analysis of the differentially expressed miRNA was focused on PCOS candidate genes identified by GWAS (52–55). In Table 2, target prediction analyses (following both IPA analyses, miRBase, and miRNA TargetMiner Database analyses) revealed that 4 of the 18 or 22% of these miRNAs (i.e., miR-130b-3p, miR-195-5p, miR-654-5p, and miR-1271-5p) were shown to target 10 of the 22 (45%) of the PCOS GWAS candidate genes (i.e., DENND1A, INSR, LHCGR, RAB5B, ZNF217, YAP1, GATA4, HMGA2, ERBB3, and KCNA4), identified in Han-Chinese (52, 53) and European populations (54, 55). miR-130b-3p was predicted to have target interactions with PCOS GWAS candidates DENND1A, ZNF217, RAB5B, LHCGR, ERBB3, and KCNA4 (Table 2).

We previously reported that a splice variant of DENND1A, one of the candidate genes identified in the Han-Chinese GWAS and replicated in studies of women of European ancestry, is a functional GWAS candidate that mediates augmented androgen biosynthesis and associated gene expression in PCOS theca cells (61, 72). More specifically, we found that the alternative truncated splice variant of DENND1A, termed DENND1A.V2, was augmented in PCOS theca cells and was found to mediate both increased CYP17A1 and CYP11A1 gene expression and excess androgen production in PCOS theca cells.

In this report, we focused our studies on miR-130b-3p, given that it was predicted to target DENND1A, combined with our continued interest in the examination of the molecular basis of increased expression of the DENND1A isoform DENND1A.V2 in PCOS theca cells. Moreover, we were interested in the exploration of the post-transcriptional mechanism(s) that underlie augmented hyperandrogenism in PCOS and the possible role that miR-130b-3p may play in DENND1A, as well as other relevant PCOS GWAS candidates. We were unable to extend our studies to the other five GWAS candidates that miR-130b-3p was predicted to target, because of the lack of information known about the differential expression profiles (of numerous mRNA and protein isoforms) of the other GWAS candidates in normal and PCOS theca cells or in other relevant cells.

The subsequent analyses of 3′UTR of DENND1A.V2, which identified an miR-130b-3p seed target sequence ∼527 bp past the DENND1A.V2 CDS stop codon (ATTGA; Fig. 3), provided further convincing evidence that miR-130b-3p might play a role in translational regulation and expression of DENND1A.V2. Our miRNA-specific qRT-PCR analyses established that miR-130b-3p expression was downregulated, whereas DENND1A.V2 expression was augmented in PCOS compared with normal theca cells (Fig. 2). Decreased miR-130b-3p was also found to be statistically correlated with increased DENND1A.V2 and CYP17A1 mRNA expression, as well as DHEA biosynthesis in PCOS theca cells, yet was not correlated with DENND1A.V1 mRNA expression (Fig. 4).

Our miR-130b-3p mimic studies confirmed the functional reciprocal relationship between miR-130b-3p and DENND1A.V2 and CYP17A1 mRNA expression (Fig. 5). Examination of the time course of miR-130b-3p transfection on DENND1A.V2, DENND1A.V1, and CYP17A1 mRNA (Fig. 5) demonstrated that miR-130b-3p transfection suppresses DENND1A.V2 mRNA at an earlier time point (24 hours) than CYP17A1 mRNA (48 hours), suggesting that miR-130b-3p mimic has direct effects on DENND1A.V2 mRNA, whereas CYP17A1 mRNA accumulation may be a result of indirect regulation via other intermediary factors and possibly via the effects of miR-130b-3p on DENND1A.V2. The miR-130b-3p mimic studies also confirmed that the interaction of miR-130b-3p is specific for DENND1A.V2 and not DENND1A.V1 mRNA and protein expression (Figs. 5 and 6). Decreased miR-130b-3p expression can possibly explain, at least in part, the DENND1A.V2 overexpression in PCOS theca cells, as documented previously, which was not explained by any of the other genetic mechanisms (61, 72). These results suggest a potential translational mechanism for DENND1A.V2 overexpression in PCOS theca cells via miR-130b-3p regulation.

It is widely accepted that miRNAs are usually involved in complex regulatory networks and their expression patterns are themselves regulated by several factors. As presented in Fig. 7, our ingenuity pathway analyses suggest that there is a common network among miR-130b-3p, DENND1A, LHCGR, and RAB5B (depicted in shades of light purple); the signaling component MAPK1; and several signaling components (PKA, AKT/PI3K, cAMP) and transcription factors (GATA4/6), previously reported to mediate changes in gene expression in PCOS theca cells, as well to be involved in downstream overexpression of steroidogenic enzyme gene expression (i.e., CYP17A1, CYP11A1, and HSD3B2), which have been identified to mediate excess androgen biosynthesis in PCOS theca cells (24, 61, 64, 65, 67). The signaling components PKA, MAPK1, and PI3K have been widely reported to be associated with LHCGR- and INSR-dependent signaling. The transcription factors GATA4/6 included in our IPA network have also been reported to be pertinent to steroidogenic enzyme gene expression and androgen biosynthesis (48). Collectively, these results also support our hypothesis that miR-130b-3p and DENND1A are involved in a complex network in combination with PCOS GWAS candidates LHCGR and RAB5B and can explain the ovarian hyperandrogenemia associated with and central to PCOS.

A more in-depth IPA core pathway and network analysis enabled us to identify important functional and biological features with relevance to PCOS (Table 3). Several PCOS-relevant features were identified in the top five: networks, canonical pathways, diseases and bio-functions, and physiological system development and functions. Included in the top five networks are endocrine system development and function, lipid metabolism, and small molecule biochemistry (Table 3). Four of the top five canonical pathways are key signaling pathways central to ovarian function and PCOS pathophysiology, including Gs signaling, cAMP signaling, G-protein-coupled signaling, and ovarian cancer signaling (Table 3). Although there is no firm evidence for associations between PCOS and ovarian cancers because of limited sample sizes and confounding risk factors, perturbation of the signaling pathways relevant in ovarian cancer may be important for normal ovarian function (92). The remaining top canonical pathways include caveolar-mediated endocytosis signaling and the newly defined synaptogenesis signaling pathway. The top diseases and bio-functions include an over-representation of diseases and disorders that are highly concordant with PCOS, including endocrine system disorders, organismal injury and abnormalities, and reproductive system disease (Table 3). The predicted relevance of the physiology system development and function, presented in Table 3, with respect to embryonic, organ, organismal development, endocrine (reproductive) system development and function, and organ morphology, is highly reflective of the pathophysiological phenotypes associated with PCOS. These results suggest that more than one PCOS candidate gene is involved in the pathophysiology of PCOS, which warrants a detailed study of other GWAS candidate genes in a PCOS context (93).

In this report, we presented data to demonstrate that miR-130b-3p regulates the expression of a truncated isoform of the PCOS GWAS candidate DENND1A; DENND1A.V2 in normal and PCOS theca cells. The IPA theca cell cellular network that we have presented provides an innovative representation of the putative mechanism by which decreased miR-130b-3p in PCOS theca cells may regulate DENND1A.V2; LHCGR, RAB5B, and the MAPK1, PI3K, and cAMP/PKA signaling pathways mediate increased CYP17A1, CYP11A1, and HSD3B2 gene expression and increased androgen biosynthesis in PCOS theca cells. The mechanisms by which miR-130b-3p is involved in epigenetic and translational regulation, which mediates DENND1A, LHCGR, and RAB5B expression in normal and PCOS theca cells, remain to be examined. Further study of the functional roles of other miRNAs that we identified as differentially expressed in normal and PCOS theca cells is needed to assess their relevance to the pathophysiology of PCOS.

Glossary

Abbreviations:

3′UTR

3′-untranslated region

3β-HSD

3β-hydroxysteroid dehydrogenase

AKT

protein kinase B

C

control (without forskolin)

CYP11A1

cytochrome P450 cholesterol side-chain cleavage

CYP17A1

cytochrome P450 17α-hydroxylase

DENND1A

differentially expressed in neoplastic vs normal cells domain containing 1A

DENND1A.V1

differentially expressed in neoplastic vs normal cells domain containing 1A variant 1

DENND1A.V2

differentially expressed in neoplastic vs normal cells domain containing 1A variant 2

DHEA

dehydroepiandrosterone

F

forskolin treated

FDR

false discovery rate

GWAS

genome-wide association study

hsa

Homo sapiens

INSR

insulin receptor

IPA

Ingenuity^® Pathway Analysis

LHCGR

luteinizing hormone/choriogonadotropin receptor

miR-seq

small RNA deep sequencing

mTOR

mammalian target of rapamycin

Neg-mimic

negative control mimic

Normal-C

normal untreated control

Normal-F

normal forskolin treated

PCOS

polycystic ovary syndrome

PCOS-C

polycystic ovary syndrome untreated control

PCOS-F

polycystic ovary syndrome forskolin treated

PI3K

phosphatidylinositol 3-kinase

PKA

protein kinase A

qRT-PCR

quantitative RT-PCR

RAB5B

Ras-related protein 5B