Comparative Meta-analysis of Adipose Tissue Transcriptomics Data in PCOS Patients and Healthy Controls Women


 Background Women with polycystic ovary syndrome (PCOS) seem to have disturbances in lipid metabolism in subcutaneous adipose tissue. Nevertheless, gene expression in subcutaneous adipose tissue of PCOS women and its relation to other disturbances have been fragmentarily investigated. Methods and results We utilized microarray data to identify the most important up and down-regulated candidate genes in adipose tissue of PCOS women in contrast to healthy women using the meta-analysis technique. Microarray data produced from three independent experiments (n=3) conducted on adipose tissue in women with PCOS were retrieved from ArrayExpress. Then, the datasets were merged using the metaSeq package in Rstudio and differentially expressed genes (DEGs) in studies were selected. The integrative bioinformatics analyses of candidates were performed by gene ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, and Protein-protein interaction (PPI) network construction. Twelve up-regulated genes and twelve down-regulated genes were identified and assessed as the most important DEGs. These the most important DEGs that were annotated by KEGG analysis were mainly involved in PI3K-Akt signaling pathway, pathways in cancer, breast cancer, metabolic pathways, ovarian steroidogenesis, and insulin secretion. In the PPI networks, PRDM10, FGFR2, IGF1R, and FLT1 were the key nodes in the up-regulated networks, while the NDUFAB1 and NME2 proteins were key in the down-regulated networks. Conclusion Overall, these findings provide a comprehensive system biology based insight into the gene expression in subcutaneous adipose tissue of PCOS women and its relation to other disturbances.


Introduction
Polycystic ovary syndrome (PCOS) is the most common hyperandrogenic disorder affecting 5-8% of women of reproductive age [1]. PCOS is considered the foremost of ovarian diseases and usually diagnosed during the early reproductive years. Most women with PCOS display some metabolic abnormalities including dyslipidemia, obesity, hyperinsulinemia, neuroendocrine abnormalities, insulin resistance, and ovulatory dysfunction [28,47]. PCOS is considered a multifactorial disorder with various metabolic, genetic, endocrine, and environmental abnormalities [17].
Obesity, a characteristic of 60-80% of PCOS patients and plays a very important role in the development of PCOS in many women. Obesity alters levels of several hormones such as insulin, androgens, and adipocytokines [12]. This suggests that obesity modi es PCOS characteristics. It has been shown that obesity has a malignant additive effect on features of PCOS such as hyperandrogenism, insulin resistance, menstrual irregularity, ovulatory disorders, and pregnancy complications [12,20]. Hyperandrogenism or increase in secretion of androgens can induce low-grade chronic in ammation by increasing the transcription of androgen receptors in mononuclear cells [51]. Low-grade chronic in ammation itself leads to the release of adipocytokines by dysfunctional adipocytes [2]. The role of adipocyte dysfunction in women with PCOS has very importance.
Visceral adipose tissue is thought to be important for the pathogenesis of PCOS, because of its association with hyperandrogenemia and its often excessive accumulation in women with PCOS [3].
Increase in subcutaneous adipocytes and decrease in the secretion of adiponectin are major factors that strongly associated with insulin resistance [32]. Furthermore, changed adipose tissue expression of some genes such as TWIST1, CCL2, LEPR, and PPARG that may be important in the pathophysiology of PCOS have been reported [5,7,10,23,45,46]. Therefore, these ndings indicate the adipose tissue dysfunction may negatively affect the metabolic health of women with PCOS and thereby increase their risk for diabetes mellitus type 2 (DM2), hyperandrogenism, and cardiovascular disease [31].
A recent meta-analysis from ve comparative studies demonstrated that women with PCOS were three times more likely to develop endometrial cancer than healthy women [21]. Moreover, Gene expression pro les in subcutaneous fat from no obese women with and without PCOS disclosed differences in the expression of genes encoding components of several biological pathways related to insulin and Wnt signaling, lipid metabolism, immune function, in ammation, and oxidative stress [7,10]. By the fact that dysfunctional adipose tissue is increasingly considered to be important in the metabolic disorders in patients with PCOS, in the present study, we utilized microarray data to identify the most important up and down-regulated candidate genes in adipose tissue in women with PCOS in contrast to healthy women using meta-analysis technique that is an attempt to integrate multiple data in different studies.

Microarray data
To identify key up and down-regulated candidate genes in adipose tissue in women with PCOS in contrast to healthy women (control), microarray data produced from three independent experiments (n=3) conducted on adipose tissue in women with PCOS were retrieved from ArrayExpress (https://www.ebi.ac.uk/arrayexpress/). In all used experiments fat biopsy samples obtained from morbidly obese women with or without PCOS. In the rst experiment (E-GEOD-43322) which was a casecontrol study we just exploited the data of 23 samples (sixteen PCOS patients and seven control). In the second experiment (E-GEOD-43264), we exploited the data of 15 samples (eight PCOS patients and seven control). In the third experiment (E-GEOD-5090), we exploited the data of 17 samples (nine PCOS patients which submitted to bariatric surgery because of morbid obesity and eight control) (Supplementary Table   S1). The overall scheme of data analysis and computational tools used in this study is represented in Fig.  1.

Meta-analysis
Meta-analysis is an attempt to integrate multiple data in different studies. Indeed, by meta-analysis, genes which differentially expressed in many studies are selected as DEGs by metaSeq package [48]. Microarray studies were processed separately as individual datasets with FlexArray software version 1.6.3. The raw data were normalized using Robust Multiarray Average (RMA) algorithm and then RMA signal values were transformed into log2. Then, the datasets were merged using the metaSeq package in Rstudio and differentially expressed genes (DEGs) in studies were selected. To select DEGs, P-value was adjusted in false discovery rate (FDR) as less than <0.05, and list of genes was obtained for adipose tissue in women with PCOS in contrast to control. Venn diagram of all genes in three studies were generated by (http://bioinformatics.psb.ugent.be/webtools/Venn/).

GO analysis and KEGG pathway enrichment analysis
To gene ontology (GO) analysis, genes that were the most important up and down-regulated genes in adipose tissue from women with PCOS in contrast to control in three studies selected and were conducted using Gene Ontology Consortium tools (http://www.geneontology.org/) with default signi cance levels (less than < 0.05). Then, the results were described in the forms of biological process, molecular function and cellular components [19]. In addition, KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis [25] was performed to identify the important pathways of the most important up and down-regulated genes in adipose tissue in women with PCOS. For this target, KEGG database (http://kobas.cbi.pku.edu.cn/) was used to test the enrichment of DGEs in KEGG pathways.

Protein-protein interaction (PPI) network analysis
Interaction protein networks for the mentioned up and down-regulated genes in adipose tissue in women with PCOS in contrast to control were constructed using the STRING v.10 (Search Tool for the Retrieval of Interacting Genes/Proteins) for detect functional association between those genes [11,44]; this database is based on speci c relationships between proteins and constructs associations based on distinct lines of evidence: Experimental evidence from PPI assays; recurring neighborhood of the genes in known genomes; co-expression databased on the expression data; events of fusion between those genes and a large collection of full-text articles [44].

Results
This study was conducted to identify key up and down-regulated candidate genes and obtain more information about these gens in adipose tissue in women with PCOS in contrast to healthy women.
This experiment was implemented based on microarray data retrieved from three independent studies.
The results illustrated that the study of one have 3233 DEGs, whereas DEGs in studies of two and three were 1304 and 571 genes, respectively (Fig. 2). As a result some genes were common in adipose tissue from women in three studies (Fig. 2) Number of common genes that were signi cantly up and downregulated in adipose tissue in women with PCOS in contrast to healthy women was demonstrated in Fig. 2. Additionally, the most important DEGs in adipose tissue in women with PCOS in contrast to control women represented in Table 1.

Visualization of DEG position
The genomic position of identi ed DEG on all chromosomes of human is represented in Fig. 3. compound metabolic process, cellular aromatic compound metabolic process, and regulation of biological process, following MF: RNA binding, organic cyclic compound binding, and cation binding, and following CC: nucleoplasm part, nucleoplasm, membrane-enclosed lumen, and membrane-bounded organelle. Moreover, downregulated genes were enriched in 276 GO functions (P < 0.05, Fig. 4). Of the dow-nregulated genes, enrichment was mainly involved in the following BP: nucleoside monophosphate metabolic process, peptide metabolic process, immune response, intracellular transport, and organic substance metabolic process, following MF: RNA polymerase II transcription factor binding, histone binding, RNA polymerase II regulatory region sequence-speci c DNA binding, DNA-binding transcription factor activity, and following CC: mitochondrial respiratory chain complex I, NADH dehydrogenase complex, respiratory chain complex I, ribosomal subunit (Supplementary Table S2).

KEGG enrichment analysis of the most important DEGs
KEGG enrichment analysis of the most important DEG targets revealed several pathways to gain greater perception into mechanisms of DEG biological functions in the adipose tissue of women with PCOS (P <0.05, Fig. 5). KEGG enrichment analysis demonstrated that the most important up-regulated expression genes were assigned into 20 pathways where the largest categories interestingly were involved in Ras signaling pathway, EGFR tyrosine kinase inhibitor resistance, MAPK signaling pathway, PI3K-Akt signaling pathway, pathways in cancer, cancers of the digestive system, long-term depression, gastric cancer, hepatocellular carcinoma, cancers of the lung and pleura, and hepatocellular carcinoma (Fig. 5).
In contrast, the most important dow-nregulated gene expression was assigned into 7 pathways where the largest categories were involved in the metabolic pathways, pyrimidine metabolism, ovarian steroidogenesis, insulin secretion, and GnRH signaling pathway (Fig. 5). The complete signi cant KEGG pathways for the most important DEGs are provided in Supplementary Table S3.

Analysis of the PPI networks of the most important DEGs
To explore the regulation mechanism of key DEGs in the adipose tissue of women with PCOS, the most important DEGs were imputed into STRING to construct a PPI network, and the network was visualized for up and down-regulated genes separately ( Fig. 6 and 7). In PPI Networks, the most important up and down-regulated genes with the highest hub scores were identi ed as hub genes that highly correlated with these pathways-related genes, suggesting their regulatory function ( Fig. 6 and 7). Hub genes play a determinant role in gene regulation because of their central position in the network. In the PPI networks of DEGs, PRDM10, FGFR2, IGF1R, and FLT1 were the key nodes in the up-regulated networks (Fig. 6), while the NDUFAB1 and NME2 proteins were key in the down-regulated networks (Fig. 7).

Discussion
Given the central role of adipose tissue in the development of PCOS in overweight and obese women, this comparative meta-analysis was performed to show changes in the expression patterns of genes between adipose tissue of PCOS women and control samples. Our result showed changes in the expression patterns of several genes in adipose tissue of PCOS women and its relation to future possible disturbances. In our meta-analysis study, the majority of signi cantly enriched pathways are involved in immune diseases, cancer, and insulin secretion emphasizing a critical role in the pathogenesis of PCOS.
In this study, we found that TNF-α and ANGPT2 genes were dramatically up-regulated in the adipose tissue of PCOS women in contrast to healthy women ( Table 1). The role of these up-regulated genes has been demonstrated in pathogenesis of PCOS and cancer. Tumor necrosis factor α (TNF-α) is a proin ammatory cytokine that is thought to play a role in the pathogenesis of PCOS [35]. Although TNF-α is involved in adipocyte metabolism [6], it has been reported that mRNA expression of TNF-α is similar levels in adipose tissue of women with and without PCOS [31]. Meanwhile, TNF-α causes insulin resistance in adipose tissue [6], and may affect the onset of type 2 diabetes mellitus (T2DM) [43]. T2DM and insulin resistance stimulate ovarian and adrenal androgen production and lead to PCOS [13]. Although the ovaries are the main source of increased androgen (hyperandrogenism) in PCOS [16], adrenal androgen excess can be present in approximately 20-25% among women with PCOS [15]. Thus, hyperandrogenism has a multifactorial origin that overexpression of TNF-α also is one of them. Of note, overexpression of TNF-α can trigger the delivery of Angiopoietin-2 (ANGPT2) into the blood [41]. ANGPT2 is a growth factor regulating vessel growth and maturation during angiogenesis [37]. ANGPT2 is expressed by activated endothelial cells under usual conditions, can also be produced by mesenchymal stem cells and by tumor cells in hypoxia and cancers conditions. In addition to its proangiogenic role during cancer progression, ANGPT2 contributes to metastatic formation [40]. It has been previously that ANGPT2 is a prognostic factor in localized metastatic colorectal cancer (CRC) [22]. Therefore, upregulation of ANGPT2 in adipose tissue of PCOS women might be considered as a biomarker in the early detection of cancer.
Our meta-analysis result showed that the expression of HPGDS and TSLP genes were dramatically downregulated in the adipose tissue from women with PCOS (Table 1). HPGDS is a Sigma-class glutathione transferase expressed in peripheral tissues such as the placenta, intestine, and adipose tissue [24] and catalyzes the isomerization of prostaglandin H2 (PGH2) to prostaglandin D2 (PGD2) [50]. Inhibiting the production of PGD2 by inhibiting HPGDS may make it an interesting target to treat allergic in ammation [39]. HPGDS may play a role in the regulation of in ammation and epithelial cell health within the oviduct, an effect likely required to maintain homeostasis and function of this key reproductive organ [4]. In our meta-analysis study, HPGDS expression was down-regulated in adipose tissue from women with PCOS; however, investigation of HPGDS-regulated in ammation in PCOS disorder should be examined.
Another interesting gene among the most important down-regulated in adipose tissue in women with PCOS was thymic stromal lymphopoietin (TSLP), which acts as a co-stimulator for thymocyte proliferation [9]. Recent studies have been reported an expanding role for TSLP in in ammatory diseases and cancer [9]. Moreover, It has been found that overexpression of TSLP (in K14-TSLP transgenic mice) can inhibit the development of early breast cancer [14]. Expression of TSLP has been reported in visceral human adipose tissue [30,49]. However, the level of mRNA expression of TSLP is lower in obese women with metabolic syndrome (state of insulin resistance associated with central obesity) compared to those without metabolic syndrome [49]. The association of abdominal central obesity with insulin resistance and T2DM characterize many patients with PCOS [34]. Therefore, activation of TSLP signaling may be a therapeutic immunotarget for improving insulin sensitivity and preventing T2DM [8].
Of the most important DEG, two up-regulated genes (FGFR2 and IGF1R) in the adipose tissue of women with PCOS were the key nodes in the up-regulated PPI networks (Fig. 6). Interestingly, KEGG analysis results showed that these genes are related to pathways in cancer, MAPK signaling pathway, and PI3K-Akt signaling pathway. Fibroblast growth factor receptor 2 (FGFR2) and insulin-like growth factor receptor (IGF1R) as speci c receptor tyrosine kinases (RTKs) [ . Dysregulation of the IGF-1/insulin/IGF-1R system may contribute to the pathophysiology of PCOS [33]. Therefore, overexpression of FGFR2 and IGF1R genes in the adipose tissue of women with PCOS may be associated with risk for breast cancer and PCOS.
In this meta-analysis study, KEGG analysis results showed that ADCY2 in the adipose tissue from women with PCOS was identi ed to play a potential role in metabolic pathways, ovarian steroidogenesis, insulin secretion, and GnRH signaling pathway. Adenylyl cyclase type 2 (ADCY2) encodes the adenylate cyclase, which can catalyze ATP to form the second messenger cyclic adenosine monophosphate (cAMP) [42]. Recently, has been reported that the expression of ADCY2 promote by estradiol (E2) [52]. Therefore, impaired E2 secretion in women with PCOS might decrease ADCY2 expression in the adipose tissue.

Conclusion
In conclusion, as subcutaneous adipose tissue is important for the pathogenesis of PCOS, this study provides more insight into the gene expression in subcutaneous adipose tissue of PCOS women and its relation to other disturbances. We showed that PCOS is associated with aberrant subcutaneous adipose tissue genes expression with dysregulated pathways including pathways in cancer, cancers of the digestive system, breast cancer, metabolic pathways, ovarian steroidogenesis, insulin secretion, and GnRH signaling pathway. This is a meta-analysis study and the cause-and-effect relationship was not clear; therefore, it is not possible to infer that PCOS alter the subcutaneous adipose tissue genes expression or contrariwise. Nevertheless, our ndings need to be con rmed in prospective studies. Therefore, further studies are recommended.

Declarations
Authors' contributions

Funding
This study was conducted without external funding.

Con ict of interest
The authors do not have any con ict of interest to declare.
14. Schematic representation of data analysis and computational tools used in this study.

Figure 2
Venn diagrams of common differential expressed genes (DEG) in three studies through individual analysis based on FlexArray.
Page 15/19   Gene network of the most important Up-regulated genes in adipose tissue in women with PCOS in contrast to control women. The connection difference colors represent the types of evidence for inferring association: recurring Neighborhood in different genomes (green line), Co-occurrence of those genes in the same organisms (dark blue), Experimental protein-protein interaction data (pink), events of Gene Fusion (red), co-expression (black), pathway described by other databases (light blue), literature textmining (yellow), and homology (purple lines).