==== Front PeerJ PeerJ peerj peerj PeerJ 2167-8359 PeerJ Inc. San Diego, USA 10171 10.7717/peerj.10171 Genomics Molecular Biology Gynecology and Obstetrics Women’s Health Data Mining and Machine Learning Identification of key genes and pathways in endometriosis by integrated expression profiles analysis Cui Ding 1 Liu Yang 1 Ma Junyan 1 Lin Kaiqing 2 Xu Kaihong 2 Lin Jun linjun@zju.edu.cn2 1 Department of Laboratory, Women’s Hospital, Zhejiang University School of Medicine, Hangzhou, China 2 Department of Gynecology and Obstetrics, Women’s Hospital, Zhejiang University School of Medicine, Hangzhou, China Uversky Vladimir 7 12 2020 2020 8 e1017115 10 2019 22 9 2020 ©2020 Cui et al.2020Cui et al.This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ) and either DOI or URL of the article must be cited.The purpose of this study was to integrate the existing expression profile data on endometriosis (EM)-related tissues in order to identify the differentially expressed genes. In this study, three series of raw expression data were downloaded from GEO database. Differentially expressed genes (DEGs) in three tissue types were screened. GO, KEGG pathway enrichment analysis, core differential genes (CDGs) protein–protein interaction (PPI) network and weighted gene co-expression network analysis (WGCNA) were performed, finally, the dysregulation of Hippo pathway in ectopic endometrium (EC) was detected by Western blotting. A total of 1,811 DEGs between eutopic (EU) and normal endometrium (NE), 5,947 DEGs between EC and EU, and 3,192 DEGs between EC and NE datasets were identified. After screening, 394 CDGs were obtained, and 5 hub genes identified in the PPI network. CDGs enrichment and WGCNA network analysis revealed cell proliferation, differentiation, migration and other biological processes, Hippo and Wnt signaling pathways, and a variety of tumor-related pathways. Western blotting results showed that YAP/TAZ was upregulated, and MOB1, pMOB1, SAV1, LATS1 and LATS2 were downregulated in EC. Moreover, CDGs, especially the hub genes, are potential biomarkers and therapeutic targets. Finally, the Hippo pathway might play a key role in the development of endometriosis. EndometriosisExpression profilesBioinformatical analysisWGCNAHippo signaling pathway National Natural Science Foundation of China81370694 81571413This work was supported by the National Natural Science Foundation of China (No. 81370694 and No. 81571413). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. ==== Body Introduction Endometriosis (EM) is a common gynecological disease and one of the leading causes of female infertility and chronic pelvic pain (Giudice & Kao, 2004). Although the prevalence of EM in fertile women is approximately 10% to 15% (Olive & Pritts, 2001), its exact pathogenesis is not yet clear. A classic hypothesis of EM etiology is the “retrograde menstruation theory” proposed by Sampson (Sampson, 1927) in 1927. According to this theory, endometrial debris enters the abdominal cavity with menstrual reflux, and then adheres in this ectopic position forming lesions. However, there are still many gaps in this theory, and it has not been proven conclusively. Subsequent studies found that EM is also associated with genetic, immunological, and inflammatory factors, along with endometrial stem/progenitor cells (Burney & Giudice, 2012; Riccio et al., 2018; Cousins, Dorien & Gargett, 2018). As a chronic disease, EM seriously affects the quality of life of patients and causes a heavy social and economic burden. Therefore, it is extremely important to explore its underlying molecular mechanisms, and discover novel markers to facilitate early diagnosis and individualized treatment. Gene chip technology is a high-throughput detection method that can analyze the expression of all genes in a given tissue/organism at one time, making it an ideal technique for screening disease-related genes and biomarkers. Since the advent of gene-chip technology, a large amount of high-throughput data has been generated, uploaded and stored in public databases. Unfortunately, the use of these data is limited by the narrow objectives of most studies, and its full potential has not been tapped yet. To understand the pathogenesis of EM, it is essential to compare the ectopic (EC), eutopic (EU) and normal (NE) endometrial tissues since all three are concurrently involved in EM development. Ectopic endometrium (EC) is defined as a condition that endometrium-like tissue grows outside the uterine cavity in patients with endometriosis, ectopic endometriotic nodules are formed due to the cyclical bleeding of EC and the fibrosis of the surrounding tissues, which could eventually lead to symptoms of dysmenorrhea, chronic pelvic pain, abnormal menstruation and infertility. Eutopic endometrium (EU) refers to the is the endometrial tissue grows in the uterine cavity in patients with endometriosis, while, the normal endometrium (NE) refers to the endometrial tissue grows in uterine cavity in patients with non-endometriosis. Some studies have reported that the EU in endometriosis patients has stronger proliferation, migration and invasion ability than the NE in non-endometriosis patients, thus, EU cells attach and colonize more easier outside of uterine cavity, thereby causing the occurrence of endometriosis. However, studies profiling the gene expression in EM have only focused on two out the three tissues, either due to the limitation of the research scale or the experimental design (Hever et al., 2007; Burney et al., 2007). In addition, traditional expression studies ultimately focus on the structural and/or functional changes in a single or a few genes. However, interactions between genes are complex, and any pathophysiological condition is a result of the synergistic effects of a large number of genes. To overcome these limitations and gain new insights into EM, we first used a bioinformatics approach to integrate the existing expression profile data on all EM-related tissues in order to identify the differentially expressed genes. The interaction between these genes was then elucidated using biological networks. Materials & Methods Data mining The gene expression profiles used in this study were downloaded from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo) in three different series: GSE7305, GSE6364 and GSE4888 (Data S1). GSE7305 contains datasets of 10 paired EC and EU samples (Hever et al., 2007), GSE6364 contains datasets of 21 EU and 16 NE samples (Burney et al., 2007), and GSE4888 contains datasets of 27 NE samples (Talbi et al., 2006). After removing duplicate samples, 69 samples (EC = 10, EU = 31, NE = 28) were included for the samples (Table 1). 10.7717/peerj.10171/table-1Table 1 Source datasets overview. Series Samples Platform NE EU EC GSE7305 – 10 / 10 10 / 10 HG-U133 PLUS 2.0 GSE6364 16 / 16 21 / 21 – HG-U133 PLUS 2.0 GSE4888 27 / 12 – – HG-U133 PLUS 2.0 Notes. Before “/” is the number of original samples contained in the data set, and after “/” is the actual number of samples used for analysis after excluding duplicate samples. Data preprocessing and principal component analysis The raw data was analyzed by the Affy package and batch effect removal was performed by sva package in R. After obtaining the normalized gene expression profiles, the principal component analysis (PCA) of 69 samples was performed based on the matrix. Identification of differentially expressed genes The annotation for the probes was downloaded from Affymetrix official website ( http://www.affymetrix.com/support/technical/annotationfilesmain.affx). limma package was used to perform the pairwise comparisons of the three groups to screen for differentially expressed genes (DEGs) at P value < 0.01. Core differential genes (CDGs) shared by the three pairwise comparisons were thus obtained. GO and KEGG pathway analysis Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (Kanehisa & Goto, 2000) analysis of the candidate genes were performed using DAVID online tool (https://david.ncifcrf.gov/summary.jsp). The enrichment significance threshold was set to P < 0.05. Protein–protein interaction network analysis A CDGs-based protein–protein interaction (PPI) network was constructed using STRING (https://string-db.org) with high confidence >0.7) (Franceschini et al., 2013). Cytoscape (Shannon et al., 2003) was used to visualize the PPI network and analyze the connectivity degree. A node with a degree of connectivity greater than 10 was defined as a hub gene and is indicated in the figure. Weighted gene co-expression network analysis To further study the synergy between genes, a weighted gene co-expression network was constructed using WGCNA package (Langfelder & Horvath, 2008). The probe with coefficient of variation (CV) ≥ 0.2 was selected for co-expression network construction. The module was then used to correlate the source trait (EC/ EU/ NE) of the endometrial tissues with the expression value of eigengene. Finally, candidate gene/probe expression values were aggregated into 28 different co-expression modules. Clinical sample collection In order to verify the association of the Hippo pathway with EM, we analyzed protein expression in EC, EU and NE endometrium by Western blotting. All subjects gave their written informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of the Obstetrics and Gynecology Hospital of Zhejiang University School of Medicine (IRB approval number: 20160054). The patients’ age was between 25–40 years, their menstrual cycles were regular, and all tissue samples were collected in the proliferative phase. No steroids were given to the patients at least 6 months prior to study inclusion. Three paired EC and EU tissue samples were collected from patients at stage III/IV ovarian EM (diagnosed under a laparoscope according to the revised AFS classification), and three NE tissue samples were obtained from women without visible endometriosis. Protein extraction and western blot Total protein was extracted from the tissues using an immunoprecipitation lysate containing protease and phosphatase inhibitors, and measured using a BCA protein quantification kit (Thermo Fisher Scientific). Equal amounts (30 µg) of protein samples were separated in a 10% SDS polyacrylamide gel and transferred to polyvinylidene fluoride (PVDF) membranes (Thermo Fisher Scientific). The membranes were incubated overnight with primary rabbit antibodies against YAP/TAZ (1:1000, #8579; Cell Signaling Technology), SAV1 (1:1000, #8579; Cell Signaling Technology), MOB1 (1:1000, #8579; Cell Signaling Technology), phospho-MOB1 (1:1000, #8579; Cell Signaling Technology), LATS1 (1:1000, #8579; Cell Signaling Technology), LATS2 (1:1000, DF7516; Affinity Biosciences) and GAPDH (1:1000, ab9485; Abcam) at 4 °C, followed by HRP-conjugated goat anti-rabbit secondary antibody (1:1000; Abcam) for 1 h at room temperature. The signals were detected with an enhanced chemiluminescence detection reagent (Thermo Fisher Scientific), and the band densities were quantified and analyzed with Quantity One software (Bio-Rad Laboratories, Hercules, CA, USA). Relative protein levels were normalized to GAPDH. Results Principal component analysis The downloaded raw data was standardized to eliminate inter-assay differences. PCA was performed based on the expression profile of the 69 samples. The results showed that sample clustering was not related to the batch (Fig. 1A, Data S2), which eliminated the batch effect. The EC samples were significantly different from the other two groups, while the latter did not differ significantly amongst themselves (Fig. 1B, Data S2). 10.7717/peerj.10171/fig-1Figure 1 Principal component analysis (PCA) of gene expression profiles. Samples are marked with different colors based on the batch or the histologic origin. (A) Green for GSE4888, blue for GSE7305, and red for GSE6364; (B) Green for NE, orange for EU, and purple for EC. Identification of DEGs between EC, EU and NE EC, EU and NE samples were compared in pairs to identify DEGs, and Venn diagrams were drawn. The threshold was set at P < 0.01. A total of 1,811 DEGs (10,78 upregulated and 733 downregulated) between EU and NE, 5947 DEGs (3014 upregulated and 2,933 downregulated) between EC and EU, and 3,192 DEGs (1,754 upregulated and 1,438 downregulated) between EC and NE were observed (Fig. 2A, Data S3, Data S4). The top 5 up- and down-regulated genes in the three pairwise comparisons based on log2FC are listed in Table 2. To summarize, the least number of DEGs was seen between EU and NE (n = 1811) and the highest between EC and EU (n = 5,947). This was contrary to our expectation that greater differences ought to be seen between EC and NE rather than EC and EU. A total of 394 CDGs were shared across all three pairs (Fig. 2B, Data S5). 10.7717/peerj.10171/fig-2Figure 2 DEGs between EC, EU and NE. (A) Statistical analysis of DEGs between the three pairwise comparisons. Orange represents the number of up-regulated DEGs, and blue represents the number of down-regulated DEGs; (B) Venn diagrams of DEGs. Red represents DEGs between EU and NE, blue represents DEGs between EC and EU, and green represents DEGs between EC and NE. Crossing areas represent shared DEGs. GO and KEGG pathway analysis of CDGs GO analysis was performed based on the 394 CDGs (Data S6), and Fig. 3A shows the top 20 entries of BP terms. The enriched CDGs included not only those involved in important cellular activities like cytokinesis, cell differentiation, and cell migration but also the regulation of the key biological processes such as the EGFR and Wnt signaling pathways, inflammatory response etc. 10.7717/peerj.10171/table-2Table 2 Differentially expressed genes between EC, EU, and NE (top five). Up-regulated Down-regulated EU-NE EC-EU EC-NE EU-NE EC-EU EC-NE RRM2 DLK1 CLDN11 ABCA8 SCGB1D2 MMP26 CDCA7 CLDN11 DLK1 CEMIP LINC01541 SCGB1D2 NDC80 GATA6 CHL1 GATA6 SCGB2A1 SCGB2A1 GDF7 SCN7A SCN7A CHI3L1 GABRP LINC01541 HOXA10 COL8A1 GATA6 LOC102723694 SLC26A7 KIAA1324 10.7717/peerj.10171/fig-3Figure 3 GO and KEGG analysis of CDGs (top 20). (A) GO (BP) enrichment analysis for 394 CDGs. (B) KEGG enrichment analysis for 394 CDGs. 10.7717/peerj.10171/fig-4Figure 4 Protein-Protein Interaction (PPI) Network of CDGs. (A) Nodes represent CDGs, and edges represent interactions. Hub genes with connectivity degree >10 are marked in orange. (B) Subnetwork constructed with PIK3CA as the core. KEGG pathway analysis was also performed on the CDGs (Data S7) and the top 20 entries with significant enrichment are shown in Fig. 3B. The enriched CDGs in pathways relevant to endometrial cancer and colorectal cancer may be related to the malignant features of EM. Furthermore, enriched CDGs in the Hippo signaling pathway suggests its involvement in the pathophysiological process of EM. Protein–protein interaction network of CDGs Using STRING database (Franceschini et al., 2013) and Cytoscape software (Shannon et al., 2003), CDGs were identified for protein-protien interactions and a PPI network was constructed. After screening, 134 CDGs formed a PPI network consisting of 134 nodes and 227 edges (Fig. 4A). The size of the node is proportional to the degree of connectivity and a node with a connectivity degree > 10 is defined as a hub gene. A total of five hub genes were obtained: PIK3CA, CDK1, BIRC5, KIF2C, ITGA1 (Table 3), and marked in orange in the Fig. 4B. Using PIK3CA, the most connected hub gene, as the core, the nodes were directly connected to it to form the subnetwork containing 20 nodes and 33 edges (Fig. 4B). Weighted gene co-expression network analysis To further understand the synergy between genes, we performed WGCNA analysis on the gene expression profiles. The gene co-expression network was constructed on the basis of the correlation between the expression values of all the sample genes. Highly related genes formed a module, and 28 co-expression modules were finally identified (Fig. 5A; marked with different colors). The association of different modules with traits (i.e., different sources of endometrium: EC/ EU/ NE) showed EC correlated most with the “indianred4” module (R = 0.32, P = 0.007), followed by the “mediumpurple2” module (R = 0.26, P = 0.03) (Fig. 5B). The high correlation of the genes in these two modules with EC suggests their involvement in EM. Function annotation of EC-related modules We conducted further analysis on the two modules showing highest correlation with EC. After annotating, the “indianred4” module contained 421 genes, and the “mediumpurple2” module contained 121 genes (Table 4, Data S8). KEGG pathway enrichment analysis revealed that the genes of the “indianred4” module were enriched in the Tight junction, RNA transport, Lysine degradation and Hippo signaling pathway entries (Fig. 6A, Table 4), while those of the “mediumpurple2” module were enriched mainly in Natural killer cell mediated cytotoxicity, Antigen processing and presentation, T cell and B cell receptor signaling pathway and other immune-related pathways, and VEGF Signaling pathway and Neurotrophin signaling pathway entries (Fig. 6B, Table 4). Therefore, the “indianred4” module may be related to the adhesion and growth of ectopic endometrium while the ”mediumpurple2” module to EMs-related immune responses, and the formation of new blood vessels and nerves. 10.7717/peerj.10171/table-3Table 3 Hub genes in PPI networks of CDGs (degree of connectivity >10). Gene symbol Connectivity degree PIK3CA 19 CDK1 18 BIRC5 17 KIF2C 12 ITGA1 12 10.7717/peerj.10171/fig-5Figure 5 Weighted gene co-expression network analysis (WGCNA) and Module trait relationship. (A) The co-expression modules for each gene below the hierarchical clustering tree are marked with different colors. Dynamic Tree Cut corresponds to the original module; Merged dynamic corresponds to the merged modules finally obtained. (B) Red indicates positive correlation and blue indicates negative correlation. The number above each cell indicates the correlation coefficient between the module and the corresponding endometrial tissue, and the number in parentheses indicates the corresponding P value. Dysregulation of Hippo signaling pathway in EMs KEGG results indicated a key role of the Hippo signaling pathway in EM pathogenesis. In order to verify the association between the Hippo signaling pathway and EM, we confirmed the expression of the proteins encoded by the key genes (indicated on the map) of this pathway by Western blotting (Fig. 7A). Compared to the EU and NE samples, YAP/TAZ expression was increased, and SAV1, MOB1, p-MOB1, LATS1, and LATS2 were decreased in EC (Fig. 7B). These result indicate that the Hippo signaling pathway is abnormally activated in EC compared to EU and NE. Discussion In order to compare the overall molecular phenotypes of ectopic (EC) and eutopic (EU) endometrial and non-endometrial (NE) tissues, and to identify key genes and pathways associated with EM, we performed bioinformatics analysis on gene-chip expression profiles from GEO database. We screened DEGs, constructed PPIs and weighted gene co-expression networks, and identified hub nodes and important co-expression modules in the networks. Finally, Western blotting was performed to validate the dysregulation of Hippo signaling pathway in EC tissues. These results provide new insights into EM. Due to the different screening thresholds set in various studies, and inconsistencies in the batches, platforms, and observation priorities, it was difficult to directly compare those studies. According to the classical implantation theory (Sampson, 1927), ectopic endometrial lesions may originate from the recurrent endometrial tissue that flows back into the abdominal cavity with menstrual blood. Subsequent studies also reported that compared to normal endometrium, ectopic and eutopic endometria show stronger proliferation, adhesion, migration and invasiveness (Lin et al., 2018; Yotova et al., 2011; Wingfield et al., 1995; Sundqvist et al., 2012), which is important for the ectopic implantation and growth of the endometrial tissues. Ideally, the EC tissues ought to be closer to EU in terms of biological behavior and overall molecular phenotype. However, our results showed the highest number of DEGs between EC and EU, which prompts us to rethink the histologic origin of the ectopic endometrium. 10.7717/peerj.10171/table-4Table 4 Genes within EC-related modules and KEGG pathway analysis (top 10). Module Gene KEGG Term Count P-value indianred4 421 Tight junction 8 2.68 ×10−3 RNA transport 7 3.07 ×10−2 Lysine degradation 4 3.43 ×10−2 Hippo signaling pathway 6 5.64 ×10−2 mediumpurple2 121 Natural killer cell mediated cytotoxicity 19 5.47 ×10−20 Antigen processing and presentation 10 2.09 ×10−9 T cell receptor signaling pathway 7 7.88 ×10−5 Fc epsilon RI signaling pathway 6 1.08 ×10−4 Leukocyte transendothelial migration 7 1.68 ×10−4 Fc gamma R-mediated phagocytosis 6 2.95 ×10−4 Chemokine signaling pathway 8 3.02 ×10−4 VEGF signaling pathway 5 8.66 ×10−4 Neurotrophin signaling pathway 5 1.00 ×10−2 B cell receptor signaling pathway 4 1.27 ×10−2 10.7717/peerj.10171/fig-6Figure 6 Functional annotation of EC-related modules. (A) KEGG pathway analysis of genes within the “indianred4” module. (B) KEGG pathway analysis of genes within the “mediumpurple2” module. 10.7717/peerj.10171/fig-7Figure 7 Dysregulation of Hippo signaling pathway in EC. (A) Schematic diagram of the Hippo signaling pathway. The key genes of interest are marked in red. (Image credit: the KEGG archive at https://www.kegg.jp). (B) Western blotting was performed to detect the expression and phosphorylation of Hippo signaling pathway proteins in EC, EU and NE. The implantation hypothesis proposed by Sampson (1927) is accepted by most researchers. However, this theory cannot explain why only 10% of menstruating women are effected by EM when most women have retrograde menstruation (Olive & Pritts, 2001), or why extra-pelvic EM and pre-menarche EM occur (Jubanyik & Comite, 1997; Schifrin, Erez & Moore, 1973). With the discovery of endometrial stem cells, Sampson’s theory has been expanded; according to the new stem cell theory, endometrial stem cells are abnormally shed during menstruation and enter the abdominal cavity with retrograde menstruation and are ectopically implanted, eventually leading to EM (Starzinski-Powitz et al., 2001; Leyendecker et al., 2002; Gargett, 2007). This theory enhances the interpretive capacity of the morbidity model of EM, and at the same time explains our results; it is likely that the EC tissue does not originate from completely differentiated EU cells, but may be formed as a separate lineage during the initial differentiation. The PPI network analysis revealed 5 hub genes: PIK3CA, CDK1, BIRC5, KIF2C and ITGA1. Anglesio et al. (2017) reported several cancer driver mutations, including PIK3CA, in 5 out of 27 non-cancer EM. Other studies have reported aberrant expression of CDK1 in gynecologic tumors such as ovarian cancer, cervical cancer and endometrial cancer (Bae et al., 2015; Luo et al., 2016; Li, Qu & Li, 2017), but its association with EM has not been verified. BIRC5, also known as Survivin, is an anti-apoptotic and pro-proliferative factor associated with various tumors (Ghaffari et al., 2016; Wang et al., 2014; Wang et al., 2016), and has recently been confirmed as a crucial factor in EM development as well as a valid diagnostic biomarker (Uegaki et al., 2015; Acimovic et al., 2016). KIF2C, the most representative member of the mitotic and cell cycle regulating Kinesin-13 family, is closely related to hepatocellular carcinoma, breast cancer and other tumors (Chen et al., 2017; Shimo et al., 2008), but its role in EMs is unclear. ITGA1 indirectly influences the proliferation, differentiation, and apoptosis of tumor cells by mediating cell adhesion processes (Yim et al., 2013; Boudjadi et al., 2016), but its association with EMs has not been reported. The Hippo signaling pathway was first discovered in 1995 (Justice et al., 1995; Xu et al., 1995). It is highly conserved in mammals and regulates organ size and homeostasis by maintaining the balance between cell proliferation and apoptosis. The key component of the human Hippo signaling pathway is the core kinase chain consisting of MST1/2, SAV1, LATS1/2, and MOB1, which inhibit cell proliferation and promote apoptosis by inhibiting the downstream YAP/TAZ transcriptional coactivator (Yu, Zhao & Guan, 2015). YAP is the main effector molecule of the Hippo signaling pathway, which can promote proliferation, inhibit apoptosis, and ultimately lead to loss of cellular contact inhibition and malignant transformation. Dysregulation of the Hippo signaling pathway in tumors has been confirmed by a large number of studies in recent years (Janse van Rensburg & Yang, 2016; Cao & Huang, 2017). Interestingly, the Hippo signaling pathway also regulates the proliferation and differentiation of stem cells in various tissues (Li et al., 2015a; Yin & Zhang, 2015; Hiemer & Varelas, 2013). However, its association with EMs has only been reported for the downstream effector molecule YAP (Song et al., 2016). Our results suggests that the Hippo signaling pathway is abnormally activated in EC and may play an important role in the occurrence of EM. Interestingly, three of the five identified hub genes are closely related to the Hippo pathway and its core factors. Studies have shown that the overexpression of PIK3CA is associated with the poor clinical prognosis in squamous cell carcinoma of the head and neck (HNSCC) (Garcia-Escudero et al., 2018). The tissue array showed that the expression of PIK3CA gene protein product (PI3Kp110α) was related to the YAP nuclear localization in HNSCC. Zhao et al. (2018) found that PIK3CA can enhance the activity of YAP and TAZ in breast tumors in vitro via multiple pathways. This indicated that YAP and TAZ were the key mediators of PIK3CA-induced breast tumorigenesis, and had a synergistic effect with PIK3CA in transforming the breast cells. Another study showed that the simultaneous activation of the PI3K and YAP pathways triggered the rapid development of liver tumors in mice, and a synergistic effect was observed between PI3K and YAP (Li et al., 2015b). The CDK1 phosphorylation by YAP can promote the mitotic defects, migration and invasion of tumor cells, which is essential for malignant transformation (Yang et al., 2013). On the other hand, CDK1 phosphorylation of TAZ may inhibit its carcinogenic activity (Zhang et al., 2015). BIRC5 is an important downstream effectors of the Hippo pathway, and its expression is regulated by the phosphorylation of YAP / TAZ, which in turn affects the cell proliferation, apoptosis, epithelial-mesenchymal transition, invasion and metastasis (Moriyama & Hori, 2019). Conclusions Through comprehensive bioinformatics analysis of common gene-chip data, we compared gene expression between EC, EU, and NE tissues. Surprisingly, the number of DEGs between EC and EU was far greater than that between EC and NE. This may have implications for our understating of the underlying mechanism of EM. A total of 394 CDGs were shared by the three pair-wise comparisons, and five hub genes in the PPI network to which the CDGs belonged were finally screened. Through weighted gene co-expression network analysis, EM-related gene modules were identified. The CDGs enrichment analysis and EM-related modules functional annotation suggest that the Hippo signaling pathway may play an important role in EM. Western blotting results confirmed that compared to EU and NE, the expression and phosphorylation of the Hippo signaling pathway proteins were abnormal in EC, suggesting that the dysregulation of this pathway is of great significance in endometrial development. These above findings will help us better understand the pathological basis of EM and provide new directions for discovering diagnostic biomarkers and therapeutic targets. Supplemental Information 10.7717/peerj.10171/supp-1Supplemental Information 1 Each sample ID in data set and each sample information Click here for additional data file. 10.7717/peerj.10171/supp-2Supplemental Information 2 Continuous screenshots of the PCA of gene expression profiles Click here for additional data file. 10.7717/peerj.10171/supp-3Supplemental Information 3 Continuous screenshots of the PCA of gene expression profiles Click here for additional data file. 10.7717/peerj.10171/supp-4Supplemental Information 4 The volcano plot of DEGs for each pair. Click here for additional data file. 10.7717/peerj.10171/supp-5Supplemental Information 5 The volcano plot of DEGs for each pair. Click here for additional data file. 10.7717/peerj.10171/supp-6Supplemental Information 6 The volcano plot of DEGs for each pair. Click here for additional data file. 10.7717/peerj.10171/supp-7Supplemental Information 7 Detailed analysis results of the three pairs of comparisons Click here for additional data file. 10.7717/peerj.10171/supp-8Supplemental Information 8 Detailed analysis results of the three pairs of comparisons Click here for additional data file. 10.7717/peerj.10171/supp-9Supplemental Information 9 Detailed analysis results of the three pairs of comparisons Click here for additional data file. 10.7717/peerj.10171/supp-10Supplemental Information 10 Gene symbols of 394 CDGs. Click here for additional data file. 10.7717/peerj.10171/supp-11Supplemental Information 11 GO analysis of 394 CDGs. Click here for additional data file. 10.7717/peerj.10171/supp-12Supplemental Information 12 KEGG pathway analysis of 394 CDGs. Click here for additional data file. 10.7717/peerj.10171/supp-13Supplemental Information 13 Genes within the EC-related modules. Click here for additional data file. 10.7717/peerj.10171/supp-14Supplemental Information 14 Full-length blots of Fig. 7. Click here for additional data file. Abbreviations EMendometriosis GEOGene Expression Omnibus database DEGsdifferentially expressed genes GOGene Ontology KEGGKyoto Encyclopedia of Genes and Genome CDGscore differential genes PPIprotein–protein internaction WGCNAweighted gene co-expression network analysis ECectopic endometrium EUeutopic endometrium NEnon-endometriosis endometrium PPIprotein–protein internaction YAPYes Associated Protein TAZTafazzin MOB1Mps One Binder Homolog A SAV1Salvador Family WW Domain Containing Protein 1 LATS1Large Tumor Suppressor Kinase 1 LATS2Large Tumor Suppressor Kinase 2 PCAprincipal component analysis GAPDHglyceraldehyde phosphate dehydrogenase BPbiological process PIK3CAPhosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha CDK1Cyclin Dependent Kinases 1 BIRC5Baculoviral Inhibitor of Apoptosis Repeat Containing 5 KIF2CKinesin Family Member 2C ITGA1Integrin Subunit Alpha 1 Additional Information and Declarations Competing Interests The authors declare there are no competing interests. Author Contributions Ding Cui conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, and approved the final draft. Yang Liu performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, and approved the final draft. Junyan Ma performed the experiments, authored or reviewed drafts of the paper, and approved the final draft. Kaiqing Lin, Kaihong Xu and Jun Lin conceived and designed the experiments, authored or reviewed drafts of the paper, and approved the final draft. Human Ethics The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers): The study was approved by the Ethics Committee of the Obstetrics and Gynecology Hospital of Zhejiang University School of Medicine (Ethical Application Ref: 20160054). Data Deposition The following information was supplied regarding data availability: The raw data are available in the Supplemental Files. ==== Refs References Acimovic et al. (2016) Acimovic M Vidakovic S Milic N Jeremic K Markovic M Milosevic-Djeric A Lazovic-Radonjic G 2016 Survivin and VEGF as novel biomarkers in diagnosis of endometriosis Journal of Medical Biochemistry 35 63 68 10.1515/jomb-2015-0005 28356866 Anglesio et al. (2017) Anglesio MS Papadopoulos N Ayhan A Nazeran TM Noe M Horlings HM Lum A Jones S Senz J Seckin T Ho J Wu RC Lac V Ogawa H Tessier-Cloutier B Alhassan R Wang A Wang Y Cohen JD Wong F Hasanovic A Orr N Zhang M Popoli M McMahon W Wood LD Mattox A Allaire C Segars J Williams C Tomasetti C Boyd N Kinzler KW Gilks CB Diaz L Wang TL Vogelstein B Yong PJ Huntsman DG Shih IM 2017 Cancer-associated mutations in endometriosis without cancer The New England Journal of Medicine 376 1835 1848 10.1056/NEJMoa1614814 28489996 Bae et al. (2015) Bae T Weon KY Lee JW Eum KH Kim S Choi JW 2015 Restoration of paclitaxel resistance by CDK1 intervention in drug-resistant ovarian cancer Carcinogenesis 36 1561 1571 10.1093/carcin/bgv140 26442525 Boudjadi et al. (2016) Boudjadi S Carrier JC Groulx JF Beaulieu JF 2016 Integrin alpha1beta1 expression is controlled by c-MYC in colorectal cancer cells Oncogene 35 1671 1678 10.1038/onc.2015.231 26096932 Burney & Giudice (2012) Burney RO Giudice LC 2012 Pathogenesis and pathophysiology of endometriosis Fertility and Sterility 98 511 519 10.1016/j.fertnstert.2012.06.029 22819144 Burney et al. (2007) Burney RO Talbi S Hamilton AE Vo KC Nyegaard M Nezhat CR Lessey BA Giudice LC 2007 Gene expression analysis of endometrium reveals progesterone resistance and candidate susceptibility genes in women with endometriosis Endocrinology 148 3814 3826 10.1210/en.2006-1692 17510236 Cao & Huang (2017) Cao J Huang W 2017 Two faces of Hippo: activate or suppress the Hippo pathway in cancer Anti-Cancer Drugs 28 1079 1085 10.1097/CAD.0000000000000559 28926422 Chen et al. (2017) Chen J Li S Zhou S Cao S Lou Y Shen H Yin J Li G 2017 Kinesin superfamily protein expression and its association with progression and prognosis in hepatocellular carcinoma Journal of Cancer Research and Therapeutics 13 651 659 10.4103/jcrt.JCRT_491_17 28901309 Cousins, Dorien & Gargett (2018) Cousins FL Dorien FO Gargett CE 2018 Endometrial stem/progenitor cells and their role in the pathogenesis of endometriosis Best Practice & Research Clinical Obstetrics & Gynaecology 50 27 38 10.1016/j.bpobgyn.2018.01.011 29503126 Franceschini et al. (2013) Franceschini A Szklarczyk D Frankild S Kuhn M Simonovic M Roth A Lin J Minguez P Bork P Von Mering C Jensen LJ 2013 STRING v9.1: protein–protein interaction networks, with increased coverage and integration Nucleic Acids Research 41 D808 D815 10.1093/nar/gks1094 23203871 Garcia-Escudero et al. (2018) Garcia-Escudero R Segrelles C Duenas M Pombo M Ballestin C Alonso-Riano M Nenclares P Alvarez-Rodriguez R Sanchez-Aniceto G Ruiz-Alonso A Lopez-Cedrun JL Paramio JM Lorz C 2018 Overexpression of PIK3CA in head and neck squamous cell carcinoma is associated with poor outcome and activation of the YAP pathway Oral Oncology 79 55 63 10.1016/j.oraloncology.2018.02.014 29598951 Gargett (2007) Gargett CE 2007 Uterine stem cells: what is the evidence? Human Reproduction Update 13 87 101 10.1093/humupd/dml045 16960017 Ghaffari et al. (2016) Ghaffari K Hashemi M Ebrahimi E Shirkoohi R 2016 BIRC5 genomic copy number variation in early-onset breast cancer Iranian Biomedical Journal 20 241 245 27372966 Giudice & Kao (2004) Giudice LC Kao LC 2004 Endometriosis Lancet 364 1789 1799 10.1016/S0140-6736(04)17403-5 15541453 Hever et al. (2007) Hever A Roth RB Hevezi P Marin ME Acosta JA Acosta H Rojas J Herrera R Grigoriadis D White E Conlon PJ Maki RA Zlotnik A 2007 Human endometriosis is associated with plasma cells and overexpression of B lymphocyte stimulator Proceedings of the National Academy of Sciences of the United States of America 104 12451 12456 10.1073/pnas.0703451104 17640886 Hiemer & Varelas (2013) Hiemer SE Varelas X 2013 Stem cell regulation by the Hippo pathway Biochimica et Biophysica Acta 1830 2323 2334 10.1016/j.bbagen.2012.07.005 22824335 Janse van Rensburg & Yang (2016) Janse van Rensburg HJ Yang X 2016 The roles of the Hippo pathway in cancer metastasis Cellular Signalling 28 1761 1772 10.1016/j.cellsig.2016.08.004 27519476 Jubanyik & Comite (1997) Jubanyik KJ Comite F 1997 Extrapelvic endometriosis Obstetrics and Gynecology Clinics of North America 24 411 440 10.1016/S0889-8545(05)70311-9 9163774 Justice et al. (1995) Justice RW Zilian O Woods DF Noll M Bryant PJ 1995 The Drosophila tumor suppressor gene warts encodes a homolog of human myotonic dystrophy kinase and is required for the control of cell shape and proliferation Genes & Development 9 534 546 10.1101/gad.9.5.534 7698644 Kanehisa & Goto (2000) Kanehisa M Goto S 2000 KEGG: kyoto encyclopedia of genes and genomes Nucleic Acids Research 28 27 30 10.1093/nar/28.1.27 10592173 Langfelder & Horvath (2008) Langfelder P Horvath S 2008 WGCNA: an R package for weighted correlation network analysis BMC Bioinformatics 9 559 10.1186/1471-2105-9-559 19114008 Leyendecker et al. (2002) Leyendecker G Herbertz M Kunz G Mall G 2002 Endometriosis results from the dislocation of basal endometrium Human Reproduction 17 2725 2736 10.1093/humrep/17.10.2725 12351554 Li et al. (2015a) Li C Kan L Chen Y Zheng X Li W Zhang W Cao L Lin X Ji S Huang S Zhang G Liu X Tao Y Wu S Chen D 2015a Ci antagonizes Hippo signaling in the somatic cells of the ovary to drive germline stem cell differentiation Cell Research 25 1152 1170 10.1038/cr.2015.114 26403189 Li, Qu & Li (2017) Li L Qu YW Li YP 2017 Over-expression of miR-1271 inhibits endometrial cancer cells proliferation and induces cell apoptosis by targeting CDK1 European Review for Medical and Pharmacological Sciences 21 2816 2822 28682437 Li et al. (2015b) Li X Tao J Cigliano A Sini M Calderaro J Azoulay D Wang C Liu Y Jiang L Evert K Demartis MI Ribback S Utpatel K Dombrowski F Evert M Calvisi DF Chen X 2015b Co-activation of PIK3CA and Yap promotes development of hepatocellular and cholangiocellular tumors in mouse and human liver Oncotarget 6 10102 10115 10.18632/oncotarget.3546 25826091 Lin et al. (2018) Lin A Yin J Cheng C Yang Z Yang H 2018 Decreased expression of FOXA2 promotes eutopic endometrial cell proliferation and migration in patients with endometriosis Reproductive Biomedicine Online 36 181 187 10.1016/j.rbmo.2017.11.001 29233503 Luo et al. (2016) Luo Y Wu Y Peng Y Liu X Bie J Li S 2016 Systematic analysis to identify a key role of CDK1 in mediating gene interaction networks in cervical cancer development Irish Journal of Medical Science 185 231 239 10.1007/s11845-015-1283-8 25786624 Moriyama & Hori (2019) Moriyama K Hori T 2019 BCR-ABL induces tyrosine phosphorylation of YAP leading to expression of Survivin and Cyclin D1 in chronic myeloid leukemia cells International Journal of Hematology 110 591 598 10.1007/s12185-019-02726-7 31428968 Olive & Pritts (2001) Olive DL Pritts EA 2001 Treatment of endometriosis The New England Journal of Medicine 345 266 275 10.1056/NEJM200107263450407 11474666 Riccio et al. (2018) Riccio L Santulli P Marcellin L Abrao MS Batteux F Chapron C 2018 Immunology of endometriosis Best Practice & Research Clinical Obstetrics & Gynaecology 50 39 49 10.1016/j.bpobgyn.2018.01.010 29506962 Sampson (1927) Sampson JA 1927 Metastatic or embolic endometriosis, due to the menstrual dissemination of endometrial tissue into the venous circulation The American Journal of Pathology 3 93 110 19969738 Schifrin, Erez & Moore (1973) Schifrin BS Erez S Moore JG 1973 Teen-age endometriosis American Journal of Obstetrics and Gynecology 116 973 980 10.1016/S0002-9378(16)33845-5 4718224 Shannon et al. (2003) Shannon P Markiel A Ozier O Baliga NS Wang JT Ramage D Amin N Schwikowski B Ideker T 2003 Cytoscape: a software environment for integrated models of biomolecular interaction networks Genome Research 13 2498 2504 10.1101/gr.1239303 14597658 Shimo et al. (2008) Shimo A Tanikawa C Nishidate T Lin ML Matsuda K Park JH Ueki T Ohta T Hirata K Fukuda M Nakamura Y Katagiri T 2008 Involvement of kinesin family member 2C/mitotic centromere-associated kinesin overexpression in mammary carcinogenesis Cancer Science 99 62 70 10.1111/j.1349-7006.2007.00635.x 17944972 Song et al. (2016) Song Y Fu J Zhou M Xiao L Feng X Chen H Huang W 2016 Activated hippo/yes-associated protein pathway promotes cell proliferation and anti-apoptosis in endometrial stromal cells of endometriosis The Journal of Clinical Endocrinology and Metabolism 101 1552 1561 10.1210/jc.2016-1120 26977530 Starzinski-Powitz et al. (2001) Starzinski-Powitz A Zeitvogel A Schreiner A Baumann R 2001 In search of pathogenic mechanisms in endometriosis: the challenge for molecular cell biology Current Molecular Medicine 1 655 664 10.2174/1566524013363168 11899254 Sundqvist et al. (2012) Sundqvist J Andersson KL Scarselli G Gemzell-Danielsson K Lalitkumar PG 2012 Expression of adhesion, attachment and invasion markers in eutopic and ectopic endometrium: a link to the aetiology of endometriosis Human Reproduction 27 2737 2746 10.1093/humrep/des220 22767451 Talbi et al. (2006) Talbi S Hamilton AE Vo KC Tulac S Overgaard MT Dosiou C Le Shay N Nezhat CN Kempson R Lessey BA Nayak NR Giudice LC 2006 Molecular phenotyping of human endometrium distinguishes menstrual cycle phases and underlying biological processes in normo-ovulatory women Endocrinology 147 1097 1121 10.1210/en.2005-1076 16306079 Uegaki et al. (2015) Uegaki T Taniguchi F Nakamura K Osaki M Okada F Yamamoto O Harada T 2015 Inhibitor of apoptosis proteins (IAPs) may be effective therapeutic targets for treating endometriosis Human Reproduction 30 149 158 10.1093/humrep/deu288 25376458 Wang et al. (2016) Wang R Kang Y Lohr CV Fischer KA Bradford CS Johnson G Dashwood WM Williams DE Ho E Dashwood RH 2016 Reciprocal regulation of BMF and BIRC5 (Survivin) linked to Eomes overexpression in colorectal cancer Cancer Letters 381 341 348 10.1016/j.canlet.2016.08.008 27539959 Wang et al. (2014) Wang H Zhang X Wang L Zheng G Du L Yang Y Dong Z Liu Y Qu A Wang C 2014 Investigation of cell free BIRC5 mRNA as a serum diagnostic and prognostic biomarker for colorectal cancer Journal of Surgical Oncology 109 574 579 10.1002/jso.23526 24338523 Wingfield et al. (1995) Wingfield M Macpherson A Healy DL Rogers PA 1995 Cell proliferation is increased in the endometrium of women with endometriosis Fertility and Sterility 64 340 346 10.1016/S0015-0282(16)57733-4 7542208 Xu et al. (1995) Xu T Wang W Zhang S Stewart RA Yu W 1995 Identifying tumor suppressors in genetic mosaics: the Drosophila lats gene encodes a putative protein kinase Development 121 1053 1063 7743921 Yang et al. (2013) Yang S Zhang L Liu M Chong R Ding SJ Chen Y Dong J 2013 CDK1 phosphorylation of YAP promotes mitotic defects and cell motility and is essential for neoplastic transformation Cancer Research 73 6722 6733 10.1158/0008-5472.CAN-13-2049 24101154 Yim et al. (2013) Yim DH Zhang YW Eom SY Moon SI Yun HY Song YJ Youn SJ Hyun T Park JS Kim BS Lee JY Kim YD Kim H 2013 ITGA1 polymorphisms and haplotypes are associated with gastric cancer risk in a Korean population World Journal of Gastroenterology 19 5870 5876 10.3748/wjg.v19.i35.5870 24124332 Yin & Zhang (2015) Yin MX Zhang L 2015 Hippo signaling in epithelial stem cells Acta Biochimica et Biophysica Sinica 47 39 45 10.1093/abbs/gmu111 25476205 Yotova et al. (2011) Yotova IY Quan P Leditznig N Beer U Wenzl R Tschugguel W 2011 Abnormal activation of Ras/Raf/MAPK and RhoA/ROCKII signalling pathways in eutopic endometrial stromal cells of patients with endometriosis Human Reproduction 26 885 897 10.1093/humrep/der010 21303778 Yu, Zhao & Guan (2015) Yu FX Zhao B Guan KL 2015 Hippo pathway in organ size control, tissue homeostasis, and cancer Cell 163 811 828 10.1016/j.cell.2015.10.044 26544935 Zhang et al. (2015) Zhang L Chen X Stauffer S Yang S Chen Y Dong J 2015 CDK1 phosphorylation of TAZ in mitosis inhibits its oncogenic activity Oncotarget 6 31399 31412 10.18632/oncotarget.5189 26375055 Zhao et al. (2018) Zhao Y Montminy T Azad T Lightbody E Hao Y SenGupta S Asselin E Nicol C Yang X 2018 PI3K positively regulates YAP and TAZ in mammary tumorigenesis through multiple signaling pathways Molecular Cancer Research 16 1046 1058 10.1158/1541-7786.MCR-17-0593 29545474