Background: Polycystic ovary syndrome is a female reproductive system disease closely related to endocrine and highly correlated with the development of endometrial carcinoma in women, it is important to identify the key genes involved in the development of polycystic ovary syndrome. Methods: To identify the hub genes, microarray datasets GSE48301, GSE115810 and GSE3013 were downloaded from Gene Expression Omnibus database. We performed in-depth cross-tabulation bioinformatic analysis to identify differentially expressed genes (DEGs) among four types of endometrial cells in GSE48301 and two endometrial carcinoma datasets GSE115810 and GSE3013, followed by gene ontology, KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment, protein-protein interaction network analysis. Results: Thirteen seed DEGs and 4 significantly expressed DEGs were identified, and potential drugs and mRNAs were found. Conclusion: EDNRA, FBN1, PMP22, SPARC and IGF-1 may be potential and their miRNAs, especially hsa-miR-29a-3p and hsa-miR-29b-3p may be potential biomarkers in the progression from PCOS to endometrial carcinoma.
Polycystic ovary syndrome (PCOS) is a very complex endocrine and metabolic disorders in women of reproductive age [1]. Rotterdam diagnostic criteria is generally adopted as diagnosis of PCOS, it affects 8%–20% of women of reproductive age worldwide [2, 3]. Developmental, genetic and environmental are involved in the etiology of PCOS [4, 5], typical characters are hyperandrogenism, ovulatory dysfunction and polycystic ovarian morphology and associated with other abnormalities, such as insulin resistance, metabolic syndrome, and dyslipidemia that cause more than 75% cases of anovulatory infertility [6] which is caused by follicular arrest and ovulatory dysfunction. Despite intensive research, the mechanisms underlying aberrant follicular development and anovulation in PCOS remain largely obscure.
The dysfunction of endometrium causes endometrial hyperplasia and endometrial
carcinoma (EC) in PCOS [7]. The endometrium is an ovarian steroid
hormone-responsive tissue composed of mesenchymal stem cells, epithelial cells,
endothelial cells and stromal fibroblasts [8]. It has been shown that
17
Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/gds/) is a public repository containing kinds of gene expression data submitted by research institutions. Protein-protein interaction (PPI) network for the differentially expressed genes (DEGs) was constructed to further explore the relationships among these genes and identify hub DEGs. Gene Ontology (GO) and KEGG enrichment analyses were performed to investigate the biological role of DEGs. Dataset GSE48301 of PCOS and two EC datasets GSE115810 and GSE3013 were downloaded from GEO. The 3 datasets were analysed with GSE 2R from network of GEO. The GSE48301 contained 4 types of endometrium samples, mesenchymal stem cells, epithelial cells, endothelial cells and stromal fibroblasts, we analysed the 4 types of endometrial cells to find DEGs between PCOS and control groups. GSE115810 contained 24 EC samples and 3 normal endometrium samples. GSE3013 contained 2 samples with E2 therapy and 2 EC samples without E2 therapy.
The 3 gene expression microarray datasets obtained from GEO database were
screened by an interactive online tool, GEO2R (www.ncbi.nlm.nih.gov/geo/geo2r).
The raw data of microarray datasets were pre-processed via background correction
and normalization, then
The PPI network is predicted using STRING (http://string-db.org/) (version 11.0)
online tool. In the our study, PPI network of 4 type cells were constructed, and
combined score
GO functions including molecular function (MF), cell component (CC), biological
processes (BP), and KEGG pathway by using DAVID
(http://www.david.abcc.ncifcrf.gov/, version 6.8), p
We performed gene expression level and survival analysis with Gene Expression
Profiling Interactive Analysis (GEPIA, http://gepia.cancer-pku.cn/). Gene
expression validation involved in Uterine Corpus Endometrial Carcinoma (UCEC,
Tumor: 174 Normal: 91) and Uterine Carcinosarcoma (UCS, Tumor: 57 Normal: 78)
datasets built in TCGA/GTEx database, the thresholds with
Drug-Gene Interaction Database (DGIdb, http://www.dgidb.org) is a web resource that consolidates disparate data sources describing drug-gene interactions. We input 13 seed DEGs to find approved drugs and 4 significantly validated DEGs in DGIdb to find potential drugs.
To find possible miRNA, we putted 4 significantly validated DEGs in miRDB (http://www.mirdb.org/), TargetScan (http://www.targetscan.org/) and miRTarBase (http://www.targetscan.org/) databases, then to find the common miRNAs of the 4 DEGs.
After standardization, there were 73, 214, 150 and 148 DEGs respectively in mesenchymal stem cells, epithelial cells, endothelial cells and stromal fibroblasts. Then 483 DEGs were identified in GSE115810, 92 DEGs were identified between estrogen therapy and control EC groups in GSE3013. 21 DEGs were found between epithelial cells and GSE115810, containing 19 consistent up-regulated and 1 consistent down-regulated genes as shown in Fig. 1A, IGF-1 was the only up-regulated DEGs among GSE48301, GSE115810 and GSE3013 in epithelial cells (Fig. 1B), DEGs behaved differently in the datasets due to the heterogeneity of the human.
Venn diagram. (A) DEGs of epithelial cells and GSE115810. (B) The DEGs of GSE115810, GSE3013 and epithelial cells.
The PPI network of DEGs were constructed in mesenchymal stem cells, epithelial cells, endothelial cells and stromal fibroblasts (Fig. 2A–D) by hiding disconnected nodes and the most significant modules were obtained by using MCODE in mesenchymal stem cells (Fig. 2A, a), epithelial cells (Fig. 2B, a–f), endothelial cells (Fig. 2C, a–d) and stromal fibroblasts (Fig. 2D, a–c), the seed genes marked with yellow in Fig. 2.
PPI network and the most significant module in mesenchymal stem cells (A, a), epithelial cells (B, a–f), endothelial cells (C, a–d), and stromal fibroblasts (D, a–c). Up-regulated genes are marked in light red, down-regulated genes are marked in light blue, and seed DEGs are marked in yellow.
In mesenchymal stem cells, MF enriched in Toll-like receptor 4 binding, arachidonic acid binding and RAGE receptor binding, CC mainly enriched in nucleus, extracellular space, and extracellular region, BP mostly enriched in neutrophil chemotaxis, inflammatory response and innate immune response. In epithelial cells, MF enriched in protein binding and integrin binding, CC mostly enriched in extracellular exosome, extracellular space and extracellular region, BP enriched in extracellular matrix organization, platelet degranulation and extracellular matrix disassembly, KEGG enriched in focal adhesion and PI3K-Akt signaling pathway. In endothelial cells, MF enriched in protein binding, CC enriched in cytosol, nucleoplasm, and membrane, BP mostly enriched in proteasome-mediated ubiquitin-dependent protein catabolic process, negative regulation of ubiquitin-protein ligase activity involved in mitotic cell cycle, and positive regulation of ubiquitin-protein ligase activity involved in regulation of mitotic cell cycle transition, pathway enriched in TGF-beta signaling pathway, cell cycle and protein processing in endoplasmic reticulum. In stromal fibroblasts, MF enriched in protein binding, ATP binding and DNA binding, CC enriched in nucleus, cytosol, and cytoplasm, BP mostly enriched in mitotic nuclear division, cell division and sister chromatid cohesion, signal pathway most enriched in cell cycle, progesterone-mediated oocyte maturation and oocyte meiosis (Fig. 3).
GO function and signal pathway in four types of PCOS.
Twenty consistent DEGs showed that MF mostly enriched in calcium ion binding, collagen binding and integrin binding, CC mostly enriched in plasma membrane, extracellular exosome, extracellular region, extracellular space and extracellular matrix, BP enriched in signal transduction, extracellular matrix organization and cell proliferation (Table 1).
Category | Term | Count | Genes |
MF | calcium ion binding | 5 | SPARC, CDH11, NID1, PAMR1, FBN1 |
collagen binding | 3 | SPARC, TGFBI, NID1 | |
integrin binding | 3 | TGFBI, IGF1, FBN1 | |
protein-glutamine gamma-glutamyltransferase activity | 2 | F13A1, TGM2 | |
laminin binding | 2 | LGALS1, NID1 | |
extracellular matrix binding | 2 | SPARC, TGFBI | |
CC | plasma membrane | 10 | EDNRA, GJA1, SPARC, MME, P2RY14, CDH11, PMP22, TGFBI, IGF1, TGM2 |
extracellular exosome | 9 | GJA1, LGALS1, MME, CDH11, ALDH1A1, TGFBI, NID1, FBN1, TGM2 | |
extracellular region | 8 | SFRP4, SPARC, F13A1, TGFBI, IGF1, NID1, PAMR1, FBN1 | |
extracellular space | 6 | SFRP4, LGALS1, SPARC, TGFBI, IGF1, FBN1 | |
extracellular matrix | 5 | LGALS1, TGFBI, NID1, FBN1, TGM2 | |
basement membrane | 4 | SPARC, TGFBI, NID1, FBN1 | |
proteinaceous extracellular matrix | 4 | LGALS1, SPARC, TGFBI, FBN1 | |
platelet alpha granule lumen | 3 | SPARC, F13A1, IGF1 | |
focal adhesion | 3 | GJA1, MME, TGM2 | |
BP | signal transduction | 5 | EDNRA, GJA1, LGALS1, SPARC, IGF1 |
heart development | 4 | EDNRA, GJA1, SPARC, FBN1 | |
extracellular matrix organization | 4 | SPARC, TGFBI, NID1, FBN1 | |
cell proliferation | 4 | EDNRA, TGFBI, IGF1, BCAT1 | |
G-protein coupled receptor signaling pathway | 4 | SFRP4, EDNRA, P2RY14, TGM2 | |
platelet degranulation | 3 | SPARC, F13A1, IGF1 | |
skeletal system development | 3 | CDH11, IGF1, FBN1 | |
positive regulation of I-kappaB kinase/NF-kappaB signaling | 3 | GJA1, LGALS1, TGM2 | |
myoblast differentiation | 2 | LGALS1, IGF1 | |
negative regulation of endothelial cell proliferation | 2 | GJA1, SPARC | |
negative regulation of neuron projection development | 2 | LGALS1, PMP22 | |
response to peptide hormone | 2 | GJA1, SPARC | |
peptide cross-linking | 2 | F13A1, TGM2 | |
positive regulation of smooth muscle cell proliferation | 2 | IGF1, TGM2 | |
positive regulation of osteoblast differentiation | 2 | GJA1, IGF1 | |
extracellular matrix disassembly | 2 | NID1, FBN1 | |
ossification | 2 | SPARC, CDH11 |
12 DEGs had significant expression in TCGA/GTEx dataset (Fig. 4A), EDNRA, FBN1, PMP22 and SPARC had significantly difference between high expression and low expression in OS (Fig. 4B).
Box plots and survival analysis plots of DEGs. (A) Significant expression of 12 DEGs in UCEC (Tumor: 174 Normal: 91) and UCS (Tumor: 57 Normal: 78) built in TCGA/GTEx database. (B) Significant analysis of OS in four DEGs between high expression and low expression on GEPIA.
Twelve seed DEGs and 4 significantly expressed DEGs were putted in DGIdb database to find drug-gene interactions. PBK, SPP1 and TUBA1A had 25 kinds of approved drugs in seed DEGs (Table 2, Fig. 5A, a), EDNRA, SPARC and PMP22 had 23 kinds of drugs in all, and 6 kinds of drugs have been approved by FDA (Table 3, Fig. 5A, b).
Gene | Drug | Interaction types | Sources |
PBK | GEFITINIB | N/A | CIViC |
SPP1 | ALTEPLASE | N/A | NCI |
GENTAMICIN | N/A | NCI | |
TACROLIMUS | N/A | NCI | |
CALCITONIN | N/A | NCI | |
TUBA1A | IXABEPILONE | inhibitor | ChemblInteractions |
TRASTUZUMAB EMTANSINE | inhibitor | ChemblInteractions | |
VINCRISTINE | N/A | DTC | |
ARTENIMOL | ligand | DrugBank | |
VINCRISTINE SULFATE | inhibitor | ChemblInteractions | |
MEBENDAZOLE | inhibitor | DrugBank—TdgClinicalTrial | |
ALBENDAZOLE | inhibitor | DrugBank | |
VINBLASTINE SULFATE | inhibitor | ChemblInteractions | |
ERIBULIN MESYLATE | inhibitor | ChemblInteractions | |
VINFLUNINE | inhibitor | ChemblInteractions | |
COLCHICINE | inhibitor | DTC|ChemblInteractions | |
CABAZITAXEL | inhibitor | ChemblInteractions | |
PACLITAXEL | inhibitor | DTC|ChemblInteractions | |
VINORELBINE | N/A | DTC | |
BRENTUXIMAB VEDOTIN | inhibitor | ChemblInteractions | |
VINBLASTINE | adduct | DTC|DrugBank | |
PODOFILOX | N/A | DTC | |
VINORELBINE TARTRATE | inhibitor | ChemblInteractions | |
DOCETAXEL | inhibitor | ChemblInteractions | |
VORINOSTAT | N/A | DTC |
Chord diagram of Drug-Gene and Venn diagrams of miRNAs. (A) Drug-gene interactions of 3 seed DEGs (a) and 3 significant expressed DEGs in OS (b). (B) The overlap miRNAs of 4 DEGs in miRDB, TargetScan and miRTarBase database.
Gene | Drug | Types | Approved | Sources |
EDNRA | AMBRISENTAN | antagonist | yes | DrugBank|TdgClinicalTrial|ChemblInteractions|TEND|GuideToPharmacology|TTD |
BOSENTAN | antagonist | yes | DrugBank|TdgClinicalTrial|ChemblInteractions|TEND|TTD | |
ATRASENTAN | antagonist | no | DrugBank|TdgClinicalTrial|GuideToPharmacology|TTD | |
MACITENTAN | antagonist | yes | DrugBank|TdgClinicalTrial|ChemblInteractions|GuideToPharmacology|TTD | |
ENRASENTAN | antagonist | no | DrugBank|ChemblInteractions | |
APROCITENTAN | antagonist | no | GuideToPharmacology | |
SITAXENTAN | antagonist | no | DrugBank|TdgClinicalTrial|GuideToPharmacology | |
NEBENTAN | antagonist | no | TdgClinicalTrial|ChemblInteractions|TTD | |
AVOSENTAN | antagonist | no | TdgClinicalTrial|ChemblInteractions|GuideToPharmacology | |
SPARSENTAN | antagonist | no | DrugBank|ChemblInteractions|GuideToPharmacology|TTD | |
DARUSENTAN | antagonist | no | DrugBank|TdgClinicalTrial|ChemblInteractions|GuideToPharmacology|TTD | |
PD-156707 | antagonist | no | GuideToPharmacology | |
ZIBOTENTAN | antagonist | no | TALC|TdgClinicalTrial|ChemblInteractions|GuideToPharmacology | |
ATRASENTAN HYDROCHLORIDE | antagonist | no | ChemblInteractions | |
TEZOSENTAN | antagonist | no | DrugBank|ChemblInteractions | |
BQ-123 | antagonist | no | GuideToPharmacology|TTD | |
CLAZOSENTAN | antagonist | no | DrugBank|TdgClinicalTrial|ChemblInteractions|TTD | |
CHEMBL1232243 | N/A | no | DrugBank | |
IRL-1620 | N/A | no | TdgClinicalTrial | |
SPARC | CALCIUM PHOSPHATE | ligand | no | DrugBank |
TRIBASIC | ||||
CALCIUM CITRATE | ligand | yes | DrugBank | |
PMP22 | PROGESTERONE | N/A | yes | NCI |
KETOPROFEN | N/A | yes | DTC |
EDNRA had 2 miRNAs: hsa-miR-200c-3p and hsa-miR-27b-3p, FBN1 had 5 miRNAs: hsa-miR-29b-3p, hsa-miR-29a-3p, hsa-miR-29c-3p, hsa-miR-486-5p and hsa-miR-767-5p, PMP22 had 6 miRNAs: hsa-miR-1233-5p, hsa-miR-4769-5p, hsa-miR-299-3p, hsa-miR-4648, hsa-miR-6778-5p and hsa-miR-4654, and SPARC had 5 miRNAs: hsa-miR-3149, hsa-miR-29b-3p, hsa-miR-29a-3p, hsa-miR-591 and hsa-miR-29c-3p (Fig. 5B). hsa-miR-29a-3p and hsa-miR-29b-3p were common miRNAs between FBN1 and SPARC.
In summary, protein binding may be common MF in PCOS. GO analysis of most significant modules in mesenchymal stem cells and epithelial cells were associated with formation of extracellular substances, such as extracellular exosome, extracellular space and extracellular region. Endothelial cells and stromal fibroblasts majored in function of intranuclear substances and cell dividing, such as cytosol, nucleoplasm and cell division. KEGG pathway showed that epithelial cells may highly related to the progression of carcinoma, and stromal cells may be associated with development and maturation of the oocyte, this may help us understand the mechanism of ovarian polycystic changes and ovulation failure.
PCOS patients often have abnormally endometrium, our study suggested that stromal fibroblasts were highly associated with mitotic nuclear division and cell division, it may be the major proliferation of cells in endometrium. KEGG pathway indicated that stromal cells involved in procession of P4-mediated oocyte maturation and oocyte meiosis, this may implied that the development of oocytes were correlated to stromal fibroblasts. A study showed that stromal fibroblasts transformation process secrete pro-gestational proteins including IGFBP-1 and PRL, IGFBP-1 was first observed during Days 1–3, and high levels of IGFBP-1 were detected from Day 4 through at least Day12, PRL was first detected on Days 4–6, and the peak occurred on Days 26–28 [20]. And we conjectured that the IGFBP-1 and PRL may be involved into the procession of maturation and meiosis of oocyte. Until now there were no studies showing the secretion of endometrial stromal fibroblasts on the development of oocytes. The exact mechanism of action is not yet known, but it could be a new direction for study of ovarian maturation disorders, anovulation and polycystic ovarian changes. In our study, PBK, SPP1 and TUBA1A had 25 kinds of approved drugs, it may be potential drugs for PCOS.
Moreover, our study showed that epithelial cells may related to the progression of EC. Endometrial hyperplasia and EC are due to steroid hormone-driven endometrial gene transcription and cellular function resulted in tissue dyshomeostasis [21, 22]. Several studies indicate that androstenediol with estrogenic activity [23], that can be highly generated from the precursor dehydroepiandrosterone in the endometrium [24]. Higher concentration of androstenediol was detected in women with PCOS and EC, suggesting a potential role of the pathophysiology of carcinoma [25]. The androgens on EC risk has come from studies in women with PCOS where the risk of EC is higher in women with symptoms of androgen excess [14]. Our study found that PI3K-Akt signaling pathway may be the most important pathway in epithelial cells from PCOS to EC. Synthesis or the activation of certain molecules with stimulus of androstenediol or estradiol, which in turn activate PI3K-AKT pathway [26]. EDNRA, FBN1, PMP22 and SPARC were validated both significant in both expression and survival analysis, and EDNRA, SPARC and PMP22 had 23 kinds of drugs in all, and 6 kinds of drugs have been approved by FDA, which may be potential drugs for stopping the progression of PCOS to endometrial cancer or EC. mRNAs were identified for the 4 DEGs, which may be potential mRNA biomarkers for PCOS.
In our research, IGF-1 was the only up-regulated DEGs among the epithelial cells in PCOS and two EC datasets, but IGF-1 is not a cancer gene. Several studies showed a significant correlation between components of the IGF system and EC risk [27, 28]. IGF-1 is produced by the liver under the stimulation of growth hormone, it have autocrine, endocrine, and paracrine actions, and takes significantly roles in growth, development, and metabolism [29]. Growth hormone-IGF-1 endocrine axis to carcinoma development [30]. In patients with PCOS, elevated IGF‑1 levels in insulin resistance, obesity and hyperinsulinemia is a higher risk of EC [31]. Impaired glucose management, hyperglycemia and expression of insulin receptor trigger carcinoma cells proliferation and inhibit carcinoma cells apoptosis [32]. The heightening circulating levels of IGF-1 caused by hyperinsulinemia, IGF-1 binding to IGF-1 receptor (IGF-1R) leading to proliferative and anti-apoptotic events [33, 34]. IGF-1 and its receptor IGF-1R may be the new therapeutic targets for both PCOS and EC.
In our research, stromal fibroblasts may be involved in the procession of maturation and meiosis of oocyte, EDNRA, FBN1, PMP22, SPARC and IGF-1 may be potential DNA biomarkers, and their miRNAs, especially hsa-miR-29a-3p and hsa-miR-29b-3p may be potential miRNA biomarkers in the progression from PCOS to endometrial carcinoma.
We certify that YG has participated sufficiently in the intellectual content, and ZZL involved in work of conception and design of this research or the analysis and interpretation of the data, as well as the writing of the manuscript. All author contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.
Not applicable.
Thanks to Yanfang Lu and Bin Zhao for their help in the data analysis process.
This research received no external funding.
The authors declare no conflict of interest.