Identification and validation of core genes associated with polycystic ovary syndrome and metabolic syndrome

Abstract

Polycystic ovary syndrome (PCOS) is a common endocrine and metabolic disorder affecting women of reproductive age, affecting reproductive health, and increasing the incidence of diabetes mellitus and hypertension. Metabolic syndrome (MetS) is the most common metabolic disorder. Although clinical studies have shown a close association between PCOS and MetS, the molecular mechanisms are unknown. In this study, datasets of PCOS and MetS were obtained from the Gene Expression Omnibus database; differential expression analysis and weighted gene coexpression network analysis (WGCNA) were performed; and gene ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses also performed of differentially expressed genes (DEGs). The PCOS- and MetS-coexpressed DEGs were subsequently intersected with the coexpressed genes in the WGCNA module to obtain the core genes. By constructing receiver operating characteristic curves, we verified the predictive effects of the core genes. We also validated the expression of the core genes in the datasets. Finally, we verified the expression of the core genes by quantitative polymerase chain reaction in human follicular fluid granulosa cells. In addition, we used Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts to analyze the immune infiltration of immune cells in PCOS and MetS. Finally, we obtained 52 coexpressed DEGs of PCOS and MetS and 3 coexpressed genes in the WGCNA module. By taking the intersection of coexpressed DEGs and coexpressed genes of the WGCNA module, we get ELOVL fatty acid elongase 7 (ELOVL7) as the core gene. Receiver operating characteristic curve analysis showed that ELOVL7 is a reliable biological marker for PCOS and MetS. The expression level of ELOVL7 in human follicular fluid granulosa cells from PCOS patients was significantly higher than that of controls, as verified by quantitative polymerase chain reaction. This study provides the first evidence of the role of ELOVL7 in developing PCOS and MetS. This gene may serve as a potential diagnostic marker and therapeutic target for both conditions.

1. Introduction

Polycystic ovary syndrome (PCOS) is a common endocrine disorder in women of reproductive age^[1,2]; it is a lifelong reproductive, metabolic, and psychological disorder that affects 5% to 18% of women.^[3] Women with the disease may experience reproductive symptoms, such as irregular menstruation and infertility, or skin symptoms, such as hirsutism, acne, or male pattern baldness.^[4] The etiology of PCOS is complex and involves mechanisms related to genetic and epigenetic susceptibility, ovarian dysfunction, excess androgen exposure, insulin resistance, and obesity.^[1,5] Given the limited understanding of the underlying causes of PCOS, there is still no optimal treatment for it, and therapeutic strategies have focused on reducing hyperandrogenism, restoring normal menstruation, and achieving pregnancy.^[6]

Metabolic syndrome (MetS) is a complex disease in which protein, fat, carbohydrates, and other substances are metabolized in a disordered way.^[7] It is highlighted by metabolic abnormalities, including abnormal glucose tolerance, insulin resistance, centripetal obesity, dyslipidemia, and hypertension, which increase the risk of cardiovascular disease, type 2 diabetes mellitus, and all-cause mortality.^[8,9] Women with PCOS are at increased risk of MetS.^[10,11] This association may be related to a common genetic basis. Consequently, an investigation into the coexpression of genes in PCOS and MetS will facilitate a deeper comprehension of the genetic factors that contribute to the pathogenesis of both conditions.

This study aimed to explore PCOS- and MetS-coexpressed genes. We obtained candidate genes by using comprehensive analysis. Finally, the expression of this gene was verified by quantitative polymerase chain reaction (qPCR) in PCOS and control follicular fluid granulosa cells. Our study provides possible new clues for diagnostic markers and therapeutic targets for PCOS and MetS.

2. Materials and methods

2.1. Raw data download, differential expression analysis, and visualization

Sequencing datasets GSE43322, GSE8157, and GSE200744 were downloaded for analysis from the Gene Expression Omnibus database for PCOS and MetS, respectively. The dataset was normalized and analyzed using the “limma” package of the R software. The genes with P < .05 were screened as differentially expressed genes (DEGs). The R software packages “pheatmap” and “ggplot2” were used to plot heatmaps and volcano maps, respectively. Differential genes coexpressed in the 3 datasets were obtained using the “VennDiagram” package and plotted as Venn diagrams.

2.2. Functional enrichment analysis

To better understand the significant biological functions of the DEGs, we analyzed the gene ontology and Kyoto Encyclopedia of Genes and Genomes pathways using the “clusterProfiler” R package; P < .05 was considered statistically significant.

2.3. Weighted gene coexpression network analysis

Weighted gene coexpression network analysis (WGCNA), a systematic biological approach, was used to unveil patterns of gene associations among various samples and identify potential biomarker genes or therapeutic targets based on the interconnectedness of gene sets and their correlation with phenotypes. Using the WGCNA package in the R software, we identified disease-associated gene coexpression modules by performing WGCNA of the GSE8157, GSE43322, and GSE200744 databases. The most relevant of each module was selected as the key module for subsequent analyses.

2.4. Screening of WGCNA module sharing genes and identification of core genes

Genes from key modules in the GSE8157, GSE43322, and GES200744 datasets were selected and plotted on a Venn diagram using the “VennDiagram” package to obtain WGCNA shared genes. Subsequently, the core genes were obtained by intersecting coexpressed DEGs and WGNCA shared genes through the “VennDiagram” package.

2.5. Prediction performance of core genes

Receiver operating characteristic curves are commonly used to assess the clinical utility of diagnostic and prognostic models.^[12] We analyzed and visualized the predicted functions of the core genes using the “pROC” package. The datasets GSE8157 of PCOS and GSE200744 of MetS were selected as training sets. Another PCOS dataset, GSE43322, was used for validation. Meanwhile, we presented the expression of the core genes in the GSE8157, GSE43322, and GSE200744 datasets.

2.6. Clinical analysis: real-time quantitative PCR

In patients undergoing in vitro fertilization-embryo transfer, we collected abandoned follicular fluid after egg retrieval and extracted ovarian granulosa cells from the follicular fluid. Then, we isolated total RNA from ovarian granulosa cells with Trizol reagent. The total RNA was reverse transcribed into cDNA using ToloScript All-in-One RT EasyMix for qPCR (ToloBio, China), and all steps followed the manufacturer’s protocol. Finally, qPCR was conducted using 2 × Q3 SYBR qPCR Master mix (Universal; ToloBio, China). The primers are given as follows: ELOVL fatty acid elongase 7 (ELOVL7): forward (5′-ACGTACCTGCTGGCTTTATTACTTC-3′) and reverse (5′-GTCCACGGCATGATGGTATGATG-3′). We used glyceraldehyde-3-phosphate dehydrogenase as an endogenous control for data normalization, and data analysis was carried out using the 2-ΔΔCt method. Our study was approved by the Youjiang Medical University for Nationalities Ethics Committee and was agreed to by all participants with signed informed consent.

2.7. Assessment of immune cell infiltration

Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts (CIBERSORT) is a computational method that accurately resolves the relative proportions of different cellular subpopulations in gene expression profiles of complex tissues.^[13] We analyzed the related proportion of 22 kinds of immune cells in samples from the GSE8157, GSE43322, and GSE200744 datasets by using CIBERSORT. CIBERSORT calculations were considered statistically significant at P < .05. Finally, the “VioplotR” package was used to visualize the results.

3. Result

3.1. Identification of coexpressed DEGs

A total of 3872 DEGs were screened in the GSE8157 dataset, including 1627 upregulated genes and 2245 downregulated genes; 5671 DEGs were screened in the GSE43322 dataset, of which 2827 were upregulated genes and 2844 were downregulated genes; and 1020 DEGs were screened in the GSE200744 dataset, of which 561 were upregulated genes and 459 downregulated genes. The results are presented in the volcano and heatmap, as shown in Figure 1. The intersection of DEGs from the GSE8157, GSE43322, and GSE200744 datasets was taken by Venn diagrams, resulting in 52 coexpressed differential genes (Fig. 2A).

Figure 1.

(A) Volcano plot of differentially expressed genes (DEGs) in GSE8157. (B) Heatmap of DEGs in GSE8157. (C) Volcano plot of DEGs in GSE43322. (D) Heatmap of DEGs in GSE43322. (E) Volcano plot of DEGs in GSE200744. (F) Heatmap of DEGs in GSE200744. Volcano plots showed the upregulated or downregulated genes in the datasets, with red dots indicating significant upregulation and blue dots indicating significant downregulation. Heatmaps exhibited the expression levels of genes in each sample in the dataset. PCOS = polycystic ovary syndrome.

Figure 2.

(A) Overlapping differentially expressed genes (DEGs) of GSE8157, GSE43322, and GSE200744. (B) Results of biological process (BP), cellular component (CC), and molecular function (MF) of DEGs’ gene ontology (GO) term enrichment analysis. (C) Results of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs.

3.2. Enrichment analysis of coexpressed DEGs

For gene ontology analysis (Fig. 2B), in biological process, actin filament organization, regulation of actin filament organization, regulation of actin filament polymerization, regulation of actin polymerization or depolymerization, and regulation of actin filament length were identified as the top 5 most enriched functions. In cellular components, membrane raft, membrane microdomain, the cluster of actin-based cell projections, adherens junction, and brush border were identified as the top 5 most enriched functions. In molecular function, cadherin binding, actin filament binding, phosphatidic acid binding, metallo-aminopeptidase activity, and protein tyrosine kinase activator activity were identified as the top 5 most enriched functions. Actin filaments involve nuclear positioning, germinal vesicle breakdown, spindle migration, chromosome segregation, spindle rotation, and polar body extrusion in oocyte mammalian meiosis.^[14] Kyoto Encyclopedia of Genes and Genomes analysis primarily targeted amebiasis, diabetic cardiomyopathy, and Advanced Glycation End-products - Receptor for Advanced Glycation End-products signaling pathway in diabetic complications signaling pathway (Fig. 2C).

3.3. WGCNA analysis and identification of shared genes

From the WGCNA analysis, 10 modules from the GSE8157 dataset, 10 modules from the GSE43322 dataset, and 4 modules from the GSE200744 dataset were filtered. Finally, it was determined that 4 modules in the GSE8157 dataset (black, red, blue, and pink), 5 modules in the GSE43322 dataset (brown, magenta, pink, blue, and green), and 1 module in the GSE200744 dataset (blue) were statistically different from each other (Fig. 3). The modular genes obtained from the GSE8157, GSE43322, and GSE200744 datasets with screening were taken to intersect using a “Venndiagram” package to obtain 3 shared genes: PRKG1, SLC6A4, and ELOVL7 (Figure S1, Supplemental Digital Content, http://links.lww.com/MD/N770).

Figure 3.

Weighted gene coexpression network analysis (WGCNA). (A) Determination of soft threshold power for the GSE8157 dataset. (B) The coexpression gene cluster dendrogram in the GSE8157 dataset. (C) Heatmap of the correlation between module eigengenes and the occurrence of polycystic ovary syndrome (PCOS) in the GSE8157 dataset. (D) Determination of soft threshold power for the GSE43322 dataset. (E) The coexpression gene cluster dendrogram in the GSE43322 dataset. (F) Heatmap of the correlation between module eigengenes and the occurrence of PCOS in the GSE43322 dataset. (G) Determination of soft threshold power for the GSE200744 dataset. (H) The coexpression gene cluster dendrogram in the GSE200744 dataset. (I) Heatmap of the correlation between module eigengenes and the occurrence of metabolic syndrome (MetS) in the GSE200744 dataset.

3.4. Identification and validation of core gene

By taking the intersection of the 3 shared genes identified from the WGCNA analysis and the 52 coexpressed DEGs, the core gene ELOVL7 can be identified (Fig. 4A). Receiver operating characteristic curve analysis showed that ELOVL7 was a good predictor relative to disease in datasets GSE8157, GSE43322, and GSE200744, with areas under the curve of 0.6064, 0.934, and 0.7109, respectively (Fig. 4C–4E). Then, the expression of the core gene was presented in the GSE8157, GSE43322, and GSE200744 datasets, and it showed that in GSE43322 and GSE200744, the expression level of this gene was significantly higher in PCOS and MetS compared to the control (Fig. 4B).

Figure 4.

(A) Overlapping genes of shared genes of differentially expressed genes (DEGs) and shared genes of weighted gene coexpression network analysis (WGCNA). (B) ELOVL7 expressions in the GSE8157, GSE43322, and GSE200744 datasets. (C) Receiver operating characteristic (ROC) curves of ELOVL7 in the GSE8157 dataset. (D) ROC curves of ELOVL7 in the GSE43322 dataset. (E) ROC curves of ELOVL7 in the GSE200744 dataset. (F) Relative expression of ELOVL7 in follicular fluid granulosa cells of patients with polycystic ovary syndrome (PCOS) and controls. AUC = area under the curve, ELOVL7 = ELOVL fatty acid elongase 7.

3.5. Validation by qPCR with clinical samples

The clinical features of the participants are presented in Table S1, Supplemental Digital Content, http://links.lww.com/MD/N770; 4 patients suffered from PCOS (diagnosed according to Rotterdam diagnostic criteria); and the others had fallopian tube infertility. There was a significant increase in the body mass index in PCOS individuals compared to the controls. Our study found that in human follicular fluid granulosa cells, the mRNA expression level of ELOVL7 was significantly higher in PCOS samples than in controls (P < .05; Fig. 4F).

3.6. Assessment of immune cell infiltration

CIBERSORT calculated the relative proportions of 22 different immune cell types in the disease. As shown in the figure, the level of neutrophil infiltration in the GSE8157 dataset was significant in the PCOS group compared to the control (P < .01). In contrast, the infiltration levels of naive CD4 T cells and gamma delta T cells were higher in the control group compared to the PCOS group (Fig. 5A). In the GSE43322 dataset, the level of CD8 T cells infiltration in PCOS was lower than that in the control group (Fig. 5B). In the GSE200744 dataset, the level of immune infiltration of macrophage M2 and memory CD4 T cells resting in PCOS was lower than in control (Fig. 5C).

Figure 5.

(A) The composition of the immune cell infiltrates in polycystic ovary syndrome (PCOS) of the GSE8157 dataset. (B) The composition of the immune cell infiltrates in PCOS of the GSE43322 dataset. (C) The composition of the immune cell infiltrates in metabolic syndrome (MetS) of the GSE200744 dataset.

4. Discussion

PCOS is a common endocrine disorder with high prevalence and comorbidities, and proper diagnosis and management are essential for the prevention and treatment of long-term complications.^[15] Although some of the pathogenesis of PCOS have been identified, its exact etiology and pathophysiological mechanisms are not fully understood.^[2] In fact, no medication is currently approved for the treatment of PCOS.^[16] Therefore, exploring available biomarkers is significant for the diagnosis and next steps in treating this disease. Women with PCOS are at increased risk of developing MetS.^[11,17] Diagnosis and treatment of MetS in women with PCOS can have a significant impact on patient health by reducing comorbidities and long-term mortality.^[17] Many studies have used 2 disease colocalization methods based on coexpression to screen genes. Reliable results have been obtained.^[18] In the current study, we combined multiple bioinformatics tools and successfully screened for the ELOVL7 gene, which is differently expressed in both PCOS and MetS. Notably, this gene showed high expression in both PCOS and MetS, and the gene was confirmed to predict PCOS by dual validation of predictive models and clinical tissues. Hence, it is promising as a novel diagnostic biomarker and therapeutic target for PCOS.

ELOVL7 is a rate-limiting enzyme for synthesizing ultralong-chain fatty acids, originating from chromosome 5, which extends saturated and monounsaturated fatty acids and assists in forming lipids.^[19,20] It has been demonstrated that single-nucleotide polymorphisms in ELOVL7 are associated with elevated waist circumference (odds ratio [OR], 1.716), high body mass index (OR, 1.964), elevated insulin (OR, 3.126), high triglycerides (OR, 4.452), high total cholesterol (OR, 7.125), and high levels of low-density lipoprotein (OR, 8.111).^[21] Overexpression of the ELOVL7 gene in a mouse model causes lipid overaccumulation in mice, leading to impaired neuronal function.^[22] Consequently, ELOVL7 has an important role in body weight control, insulin regulation, and fat distribution, which are also strongly associated with developing PCOS.^[23] Some studies have confirmed that obesity promotes the onset and progression of PCOS and that weight loss improves metabolic status and reproductive abnormalities in women with PCOS.^[24] Lipid abnormalities such as elevated low-density lipoprotein and high triglyceride levels are common in women with PCOS and play an important role in the disease progression of PCOS.^[25] Insulin resistance plays a central role in the pathogenesis of PCOS and its associated metabolic abnormalities.^[10] Thus, ELOVL7 may be involved in the pathogenesis of PCOS. Unfortunately, there are currently no reports on ELOVL7 in PCOS.

ELOVL7 is not only associated with metabolic diseases. It has also been reported in several other diseases. In Crohn disease, ELOVL7 acts as a locus associated with response to adalimumab treatment.^[26] ELOVL7 also has an association with early onset Parkinson disease^[27] and tuberculosis infection.^[28] Furthermore, RNA-seq analysis of plasmacytoid dendritic cells (pDCs) showed that ELOVL7 was upregulated in pDCs treated with TLR7 and TLR9 agonists. Knockdown of ELOVL7 reduces interleukin (IL)-6 and IL-12/IL-23 p40 production. Thus, ELOVL7 is a novel proinflammatory gene that is upregulated in response to inflammatory stimuli and regulates the function of M1-like macrophages and pDCs.^[19]

Immunotherapy is widely used to treat various diseases, especially oncological diseases.^[29] However, there is still a lack of effective immunotherapy in treating PCOS. We compared the immune cell group contents in different groups to investigate the link between PCOS and immune cell infiltration. We discovered that neutrophils were significantly higher in the PCOS group than in the control, while gamma delta T cells, naive CD4 T cells, and CD8 T cells were higher in the normal group. Excess androgen levels often accompany patients with PCOS, and excess androgens cause the bone marrow precursor cells to produce neutrophils at a faster rate; thus, patients with PCOS can develop excess neutrophils.^[30] The excess neutrophils, in turn, produce a variety of cytokines, such as tumor necrosis factor-alpha, transforming growth factor beta, IL-6, IL-1α, and IL-1β, which contribute to the pathophysiology of PCOS.^[30,31] Moreover, neutrophils consume and eliminate damaged and dead cells in excess fat, which ultimately causes chronic low-grade inflammation.^[30,32] Localized ovarian inflammation impairs ovulation and induces or exacerbates systemic inflammation. Patients with PCOS have significantly higher concentrations of circulating inflammatory cells, such as lymphocytes and neutrophils, and a lower percentage of Treg cells than non-PCOS women.^[33] Neutrophil levels can be used clinically to predict PCOS and the low-grade inflammation that goes along with it, as well as to evaluate the effectiveness of the medication.^[34]

This study has the following limitations. First, the sample size of the dataset used for analysis is relatively small. Second, the dataset used in our study was analyzed by microarrays, which lag behind today’s advanced sequencing technologies, thus influencing the accuracy of our data. In addition, the clinical samples that we used for validation were limited to follicular fluid granulosa cells and small sample sizes and have not yet been validated in human blood and ovarian tissues. Hence, clinical trial analyses with more tissues and sample sizes are still needed.

5. Conclusion

This study provides the first evidence of the role of ELOVL7 in developing PCOS and MetS. This gene may be important in the pathogenesis of PCOS and may be a diagnostic marker and therapeutic target for PCOS.

Acknowledgments

Data on genes associated with polycystic ovary syndrome and metabolic syndrome were obtained from the Gene Expression Omnibus databases. The authors thank these researchers for their selfless sharing.

Author contributions

Conceptualization: Shaohua Ling, Liying Huang.

Data curation: Shaohua Ling, Liying Huang, Chun Tian.

Formal analysis: Shaohua Ling.

Funding acquisition: Li Qin.

Investigation: Shaohua Ling, Liying Huang.

Methodology: Shaohua Ling, Liying Huang, Thongher Lia, Chun Tian.

Project administration: Shaohua Ling, Liying Huang, Thongher Lia, Chun Tian.

Resources: Shaohua Ling.

Software: Shaohua Ling, Liying Huang, Thongher Lia.

Supervision: Shaohua Ling.

Validation: Shaohua Ling, Delong Xie.

Visualization: Shaohua Ling, Xiao Qin, Li Qin.

Writing – original draft: Shaohua Ling, Liying Huang.

Writing – review & editing: Shaohua Ling, Xiao Qin, Li Qin.