Differential Gene Expression in Granulosa Cells from Polycystic Ovary Syndrome Patients with and without Insulin Resistance: Identification of Susceptibility Gene Sets through Network Analysis

Abstract

Context:

Polycystic ovary syndrome (PCOS) is a heterogeneous, genetically complex, endocrine disorder of uncertain etiology in women.

Objective:

Our aim was to compare the gene expression profiles in stimulated granulosa cells of PCOS women with and without insulin resistance vs. matched controls.

Research Design and Methods:

This study included 12 normal ovulatory women (controls), 12 women with PCOS without evidence for insulin resistance (PCOS non-IR), and 16 women with insulin resistance (PCOS-IR) undergoing in vitro fertilization. Granulosa cell gene expression profiling was accomplished using Affymetrix Human Genome-U133 arrays. Differentially expressed genes were classified according to gene ontology using ingenuity pathway analysis tools. Microarray results for selected genes were confirmed by real-time quantitative PCR.

Results:

A total of 211 genes were differentially expressed in PCOS non-IR and PCOS-IR granulosa cells (fold change ≥ 1.5; P ≤ 0.001) vs. matched controls. Diabetes mellitus and inflammation genes were significantly increased in PCOS-IR patients. Real-time quantitative PCR confirmed higher expression of NCF2 (2.13-fold), TCF7L2 (1.92-fold), and SERPINA1 (5.35-fold). Increased expression of inflammation genes ITGAX (3.68-fold) and TAB2 (1.86-fold) was confirmed in PCOS non-IR. Different cardiometabolic disease genes were differentially expressed in the two groups. Decreased expression of CAV1 (−3.58-fold) in PCOS non-IR and SPARC (−1.88-fold) in PCOS-IR was confirmed. Differential expression of genes involved in TGF-β signaling (IGF2R, increased; and HAS2, decreased), and oxidative stress (TXNIP, increased) was confirmed in both groups.

Conclusions:

Microarray analysis demonstrated differential expression of genes linked to diabetes mellitus, inflammation, cardiovascular diseases, and infertility in the granulosa cells of PCOS women with and without insulin resistance. Because these dysregulated genes are also involved in oxidative stress, lipid metabolism, and insulin signaling, we hypothesize that these genes may be involved in follicular growth arrest and metabolic disorders associated with the different phenotypes of PCOS.

Polycystic ovary syndrome (PCOS) is a heterogeneous, genetically complex, endocrine disorder of uncertain etiology in women. Its principal features include hyperandrogenism, ovulatory dysfunction, and polycystic ovaries (1–3). Besides the reproductive abnormalities, metabolic complications including insulin resistance, obesity, dyslipidemia, and hypertension are also evident in many PCOS women (4). Compared with the normally cycling women, PCOS patients are at a greater risk for diabetes mellitus and cardiovascular diseases (5).

We hypothesized that modest effects of individual genetic variants may lead to abnormal expression of multiple PCOS, insulin resistance, and cardiovascular disease susceptibility genes in granulosa cells during gonadotropin therapy for in vitro fertilization. Findings presented here demonstrate significant differences in granulosa cell expression of genes involved in diabetes mellitus, inflammation, cardiovascular disease, and infertility in PCOS patients with and without insulin resistance, which may relate to the different PCOS phenotypes.

Subjects and Methods

Experimental subjects

The detailed descriptions of PCOS and control subjects, methods for subject recruitment, ovarian stimulation, and granulosa cell collection are presented in Supplemental Methods (published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). We conducted our gene expression study on gonadotropin-stimulated granulosa cells from 12 controls, the normal ovulatory women; 12 women with PCOS without evidence for insulin resistance (PCOS non-IR); and 16 women with PCOS complicated by peripheral insulin resistance (PCOS-IR) who were undergoing in vitro fertilization.

Microarray hybridizations and ingenuity pathway analysis (IPA)

Microarray hybridizations on Affymetrix HG-U133 Plus 2 whole genome GeneChips (Affymetrix, Inc., Santa Clara, CA), were conducted with control and PCOS patient mRNA samples. The detailed descriptions of microarray hybridizations, molecular interaction, and functional analysis using IPA tools and real-time quantitative PCR (RT-qPCR) are presented in the Supplemental Methods. The microarray data are deposited at the Gene Expression Omnibus public repository (http://www.ncbi.nlm.nih.gov/geo) with accession number GSE34526. The criteria for selecting differentially expressed genes were preset as at least fold change (FC) ≥ 1.5 fold (P ≤ 0.001, unpaired t test).

Results

Clinical characteristics of PCOS and control subjects

The clinical characteristics of the control, PCOS non-IR (without insulin resistance), and PCOS-IR (with insulin resistance) are shown in Supplemental Table 1. Levels of LH, LH/FSH, and free testosterone were similar in IR and non-IR groups but significantly higher than controls. Levels of insulin, body mass index, and homeostasis model assessment of insulin resistance were significantly higher in PCOS-IR patients than in the PCOS non-IR group and controls. The PCOS non-IR group and controls had normal insulin levels. All subjects had normal glucose, TSH, and prolactin levels.

Differentially expressed genes in granulosa cells of PCOS women with and without insulin resistance

To identify differentially expressed genes, microarray hybridizations were carried out for three control, three PCOS non-IR, and four PCOS-IR subjects. Classical multidimensional scaling was applied in three dimensions to the 17,958 probe sets that remained after filtering, and a 3-dimensional scatterplot showed that the three groups (Fig. 1A and Supplemental 3D Plot) were well separated. Of 17,958 analyzed probe sets, 217 probe sets were identified as differentially expressed (P ≤ 0.001). The Venn diagram (Fig. 1B) showed 103 probe sets in control vs. PCOS non-IR, 61 probe sets in control vs. PCOS-IR, and 69 probe sets in PCOS non-IR vs. PCOS-IR to be differentially expressed. Supervised hierarchical clustering was performed on the dataset of 217 unique probe sets that were significantly expressed in at least one of the three pairwise comparisons, resulting in three distinct clusters on the heat map (Fig. 1C). A total of 211 genes were differentially expressed in PCOS non-IR and PCOS-IR granulosa cells (FC ≥ 1.5; P ≤ 0.001) vs. matched controls. The differentially expressed genes for each pairwise comparison are presented in Supplemental Table 2 (control vs. PCOS non-IR), Supplemental Table 3 (control vs. PCOS-IR), and Supplemental Table 4 (PCOS non-IR vs. PCOS-IR).

Fig. 1.

A, Global gene expression profiles of the control and the two phenotypes of PCOS—PCOS non-IR and PCOS-IR granulosa cells. The 17,958 probe sets left after excluding all control probe sets and probe sets present in less than 80% of the samples were reduced to three dimensions using classical multidimensional scaling (MDS). The figure represents the gene expression data in three dimensions, with each plotted point representing one sample and with color used to represent the sample type: red, control (n = 3); blue, PCOS non-IR (n = 3); and green, PCOS-IR (n = 4). B, Venn diagram illustrating the number of probe sets identified as significantly differentially expressed (FC ≥ 1.5; P ≤ 0.001) in each pairwise comparison (control vs. PCOS non-IR, control vs. PCOS-IR, and PCOS non-IR vs. PCOS-IR) and common between them. A Venn diagram was constructed using limma Bioconductor package. C, Supervised hierarchical clustering for 217 unique probe sets significant (P ≤ 0.001) in at least one of the three pairwise comparisons was performed by Ward's method and 1-(Pearson's correlation) as the distance measure. Each row represents a single gene; each column represents a sample. Expression values were color coded: blue, transcript level below the median; green, equal to median; and yellow, greater than median. D, To compare the results of all PCOS comparison with each pairwise comparison (control vs. PCOS non-IR, control vs. PCOS-IR, and PCOS non-IR vs. PCOS-IR), the mean probe set expression values (on a log₂ scale) for all probe sets were plotted. In each scatterplot, 222 differentially expressed probe sets (P ≤ 0.001) in all PCOS groups were overlapped and displayed in red. The distribution of red points showed the specific differences between the differentially expressed probe sets of all PCOS groups and the pairwise comparisons indicating the importance of studying the gene expression profiles of the two phenotypes of PCOS separately.

Differential outcomes of pairwise and all PCOS comparisons

Statistical comparison was also performed for all PCOS (n = 7) vs. control (n = 3) subjects. Among the 17,958 probe sets tested, 222 were differentially expressed in all PCOS groups (P ≤ 0.001) corresponding to 185 unique genes (Supplemental Table 5). To compare the results of all PCOS comparisons with each pairwise comparison (control vs. PCOS non-IR, control vs. PCOS-IR, and PCOS non-IR vs. PCOS-IR), the mean probe set expression values (on a log₂ scale) for all probe sets were plotted. In each scatter plot, 222 differentially expressed probe sets (FC ≥ 1.5; P ≤ 0.001) in all PCOS groups were overlapped and displayed in red (Fig. 1D). The distribution of red points showed the specific differences between the differentially expressed probe sets of all PCOS groups and the pairwise comparisons.

Dysregulated interactions in PCOS non-IR and PCOS-IR granulosa cells

The gene lists for each pairwise comparison (P ≤ 0.001; FC ≥ 1.5), control vs. PCOS non-IR (Supplemental Table 2), and control vs. PCOS-IR (Supplemental Table 3) were analyzed by IPA tools. Six networks for PCOS non-IR and four networks for PCOS-IR were generated (Supplemental Table 6). The highest scoring network of PCOS non-IR (Supplemental Fig. 1A) and PCOS-IR (Supplemental Fig. 1B) are given in the Supplemental Data. The major “hubs” dysregulated were NFkB, HNF4A, MAPK1, CAV1, TP53, CTNNB1, MYC, and HTT in PCOS non-IR and NFkB, HNF4A, TNF, HNF1A, and NFYB in PCOS-IR.

PCOS non-IR and PCOS-IR-specific gene ontology

Functional classification of differentially expressed genes identified diabetes mellitus, inflammation, and cardiovascular disease-related genes in granulosa cells from PCOS patients with (Fig. 2, A–C) and without (Fig. 2, D and E) insulin resistance. Microarray results of selected genes were confirmed in additional samples of nine control, nine PCOS non-IR, and 12 PCOS-IR subjects by RT-qPCR (Supplemental Tables 7 and 8).

Fig. 2.

Functional analysis of microarray data by IPA tools indicated differential expression (P ≤ 0.001) of diabetes mellitus, inflammatory response, and cardiovascular disease genes in the granulosa cells of PCOS-IR (n = 4) (A–C) and PCOS non-IR (n = 3) subjects (D and E) vs. matched controls (n = 3). A, Histogram analysis of diabetes mellitus genes in PCOS-IR with at least 1.8-fold change in expression. B, Histogram analysis of inflammation response genes in PCOS-IR with at least 1.58-fold change in expression. C, Histogram analysis of cardiovascular disease genes in PCOS-IR with at least 1.59-fold change in expression. D, Histogram analysis of inflammation response genes in PCOS non-IR with at least 3.11-fold change in expression. E, Histogram analysis of cardiovascular disease genes in PCOS non-IR with at least 1.98-fold change in expression.

Diabetes mellitus genes

Nine up-regulated and three down-regulated diabetes mellitus genes were identified in stimulated granulosa cells of PCOS-IR patients vs. controls (Fig. 2A). The genetic variants of the transcription factor 7-like 2 (TCF7L2) gene are possibly the strongest known genetic risk factor for type 2 diabetes mellitus. Therefore, the expression of TCF7L2 was also studied. Higher expression of neutrophil cytosolic factor 2 (NCF2) (2.13-fold) and TCF7L2 (1.92-fold) was confirmed in PCOS-IR granulosa cells (Supplemental Table 8).

Inflammation genes

Twelve inflammation genes were identified in PCOS-IR granulosa cells, 10 were up-regulated, and two were down-regulated in PCOS-IR (Fig. 2B). RT-qPCR confirmed higher expression of SERPINA1 (5.35-fold) and NCF2 (2.13-fold) (Supplemental Table 8). Four inflammation response genes were up-regulated, and two down-regulated in PCOS non-IR (Fig. 2D). Significantly increased expression of integrin, α X (complement component 3 receptor 4 subunit) (ITGAX) (3.68-fold) and TGF-β-activated kinase 1/MAP3K7 binding protein 2 (TAB2) (1.86-fold) in PCOS non-IR was confirmed (Supplemental Table 8).

Cardiovascular disease genes

Thirteen cardiovascular disease genes were identified in PCOS-IR (seven up-regulated and six down-regulated; Fig. 2C) and the decreased expression of osteonectin (SPARC) (−1.88-fold) was confirmed (Supplemental Table 8). Seventeen cardiovascular disease genes were identified in PCOS non-IR patients (eight up-regulated and nine down-regulated; Fig. 2E), and decreased expression of caveolin-1 (CAV1) (−3.58-fold) was confirmed (Supplemental Table 8).

Differential expression of genes involved in TGF-β signaling [IGF-II receptor (IGF2R), 5.10-fold and 4.30-fold; hyaluronan synthase 2 (HAS2), −5.31-fold and −3.53-fold] and oxidative stress [thioredoxin-interacting protein (TXNIP), 3.32-fold and 2.81-fold] was confirmed in PCOS non-IR and PCOS-IR granulosa cells, respectively (Supplemental Table 8).

Discussion

Our results from whole genome microarray analysis show the differences in gene expression profiles of gonadotropin-stimulated granulosa cells of PCOS patients with and without insulin resistance. The comparison of the differentially expressed probe sets in all PCOS groups with those in the pairwise comparisons demonstrated the specific differences in the outcome. Because the differentially expressed probe sets from pairwise comparisons are not highlighted by all PCOS comparison, it was important to study gene expression profiles of the two phenotypes of PCOS separately. Network analyses of the differentially expressed genes identified “dysregulated hubs” and their interacting molecules in the two phenotypes of PCOS granulosa cells. Microarray analysis and RT-qPCR data demonstrated significant differences in the expression of diabetes mellitus, inflammation, and cardiovascular disease-related genes in PCOS non-IR and PCOS-IR granulosa cells. The identified subsets of genes are also related to oxidative stress, insulin signaling, and infertility. This is the first study to compare the expression profiles of stimulated granulosa cells from PCOS women with and without insulin resistance.

Significantly higher expression of diabetes mellitus genes SERPINA1, NCF2, and TCF7L2 and a significant decrease in the expression of cardiovascular disease-related gene SPARC was confirmed in the PCOS-IR group by RT-qPCR. SERPINA1 encodes for a serine protease inhibitor, an acute phase protein that increases during inflammation (6). NCF2 is a rate-limiting cofactor of NADPH oxidase, and increased expression of NCF2 potentiates NADPH oxidase activity, leading to oxidative stress and hypertension (7). The increased expression of TCF7L2 is reported in adipocytes of insulin-resistant subjects (8), and its overexpression (2-fold) was shown to increase insulin expression in islet cells (2.3-fold) (9). Although the role of SERPINA1, NCF2, and TCF7L2 is not well understood in granulosa cells, we hypothesize that alteration in these genes could contribute to oxidative stress and abnormal glucose metabolism in PCOS follicles. Decreased expression of SPARC, an extracellular matrix glycoprotein, may result in myocardial dysfunction (10). Additionally, SPARC is increased during luteinization and plays an important role in corpus luteum development (11). Its decreased expression in the granulosa cells of PCOS-IR patients may cause defects in proliferative or reorganizational activity during luteinization.

Increased expression of ITGAX and TAB2 and decreased expression of CAV1 were confirmed in PCOS non-IR patients. ITGAX encodes an integrin α X chain protein that mediates cell-cell interactions during inflammation. Increased level of ITGAX is associated with hypertriglyceridemia, hypercholesterolemia and atherosclerosis (12, 13). TAB2 may act as a sensor of inflammatory signals and functions as an adaptor in the IL-1 signaling pathway leading to JNK and nuclear factor κ-B activation (14, 15). CAV1 gene encodes a membrane protein that is involved in lipid and cholesterol transport and steroid and insulin signaling (16). LH/human chorionic gonadotropin is known to increase the expression of CAV1 in granulosa cells of ovulatory follicles (17), and its decreased expression may be deleterious to granulosa cell differentiation. Altered expression of ITGAX, TAB2, and especially CAV1, a dysregulated hub, may contribute to oxidative stress, abnormal lipid metabolism, and follicular growth arrest in PCOS non-IR patients.

Increased expression of IGF2R and TXNIP and decreased expression of HAS2 were confirmed in both PCOS non-IR and PCOS-IR patients. IGF2R encodes IGF-II receptor that activates TGF-β1 and also transports IGF-II to lysosomes, thereby limiting the activation of IGF1R and insulin receptor by IGF-II (18). TXNIP encodes a cytoplasmic protein that inhibits the activity of thioredoxin disulfide reductase and thereby increases oxidative stress (19). A member of the α-arrestin gene family, TXNIP also inhibits glucose uptake and lactate production by destabilizing the hypoxia-induced transcription factor (HIF1α) that regulates the most glycolytic target genes (20). Overexpression of TXNIP is linked to the dysregulation of cardiomyocyte viability (19). HAS2 encodes an enzyme that is stimulated by GDF-9, a TGF-β family member, in cumulus cells leading to hyaluronan-rich matrix during cumulus expansion in the periovulatory period (21, 22). We hypothesize that altered expression of IGF2R, TXNIP, and HAS2 in both groups of patients may contribute to abnormal ovarian follicular growth and cumulus expansion, respectively, in PCOS follicles.

In summary, the current study discloses specific gene expression patterns related to diabetes mellitus, inflammation, cardiovascular diseases, and infertility in the granulosa cells of PCOS women with and without insulin resistance. Because these dysregulated genes are also linked to oxidative stress, lipid metabolism, and insulin signaling, we hypothesize that different genes may be involved in follicular growth arrest and metabolic disorders associated with the different phenotypes of PCOS.