Basigin is necessary for normal decidualization of human uterine stromal cells

Abstract

STUDY QUESTION

Does basigin (BSG) regulate human endometrial stromal cell (HESC) decidualization in vitro?

SUMMARY ANSWER

BSG regulates HESCs proliferation and decidualization.

WHAT IS KNOWN ALREADY

Studies have shown that in the human endometrium, BSG expression is menstrual-cycle dependent and its expression was significantly lower in uterine endometrium during the luteal phase of women experiencing multiple implantation failures after IVF than in women with normal fertility.

STUDY DESIGN, SIZE, DURATION

We utilized a telomerase-immortalized HESCs in an in vitro cell culture model system to investigate whether BSG regulates decidualization of stromal cells. Further, we used microarray analysis to identify changes in the gene expression profile of HESCs treated with BSG small interfering RNA (siRNA). All experiments were repeated at least three times.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The effect of BSG knockdown (using siRNA) on HESC proliferation was determined by counting cell number and by tritiated thymidine incorporation assays. The effect of BSG on decidualization of HESCs was determined by RT–qPCR for the decidualization markers insulin-like growth factor-binding protein 1 (IGFBP1) and prolactin (PRL). Immunoblotting was used to determine the effect of BSG siRNA on the expression of MMP-2,3. Microarray analysis was used to identify BSG-regulated genes in HESCs at Day 6 of decidualization. Functional and pathway enrichment analyses were then carried out on the differentially expressed genes (DEGs). The STRING online database was used to analyze protein–protein interaction (PPI) between DEG-encoded proteins, and CytoScape software was used to visualize the interaction. MCODE and CytoHubba were used to construct functional modules and screen hub genes separately. Several BSG-regulated genes identified in the microarray analysis were confirmed by qPCR.

MAIN RESULTS AND THE ROLE OF CHANCE

Knockdown of BSG expression in cultured stromal cells by siRNA significantly (P < 0.05) inhibited HESC proliferation, disrupted cell decidualization and down-regulated MMP-2 and MMP-3 expression. Microarray analysis identified 721 genes that were down-regulated, and 484 genes up-regulated with P < 0.05 in BSG siRNA treated HESCs. GO term enrichment analysis showed that the DEGs were significantly enriched in cell communication, signaling transduction and regulation, response to stimulus, cell adhesion, anatomical structure morphogenesis, extracellular matrix organization, as well as other functional pathways. KEGG pathway analysis identified upregulated gene enriched in pathways such as the MAPK signaling pathway, colorectal cancer, melanoma and axon guidance. In contrast, downregulated genes were mainly enriched in pathways including ECM–receptor interaction, PI3K-Akt signaling pathway, pathways in cancer, antigen processing, type I diabetes mellitus and focal adhesion. The top 10 hub nodes were identified using 12 methods analyses. The hub genes that showed up in two methods were screened out. Among these genes, upregulated genes included EGFR, HSP90AA1, CCND1, PXN, PRKACB, MGAT4A, EVA1A, LGALS1, STC2, HSPA4; downregulated genes included WNT4/5, FOXO1, CDK1, PIK3R1, IGF1, JAK2, LAMB1, ITGAV, HGF, MXRA8, TMEM132A, UBE2C, QSOX1, ERBB2, GNB4, HSP90B1, LAMB2, LAMC1 and ITGA1. Hub genes and module genes involved in the top three modules of PPI analysis were analyzed through the string database. Analysis showed that hub and module genes were related mainly to the WNT signaling pathway, PI3K-AKT signaling pathway and pathways in cancer.

LARGE SCALE DATA

The microarray data set generated in this study has been published online at databank.illinois.edu.

LIMITATIONS, REASONS FOR CAUTION

Most of the findings were obtained using an in vitro cell culture system that may not necessarily reflect in vivo functions.

WIDER IMPLICATIONS OF THE FINDINGS

Our results demonstrate that BSG plays a vital role in decidualization and that downregulation of BSG in the uterine endometrium may be associated with infertility in women. The identified hub genes and pathways increase our understanding of the genetic etiology and molecular mechanisms underlying the regulation of decidualization by BSG.

STUDY FUNDING/COMPETING INTEREST(S)

This work was supported by the NIH U54 HD40093 (R.A.N.). The authors have no competing interests to declare.

Introduction

Infertility has become a common health concern in humans and affects about 10–15% of couples worldwide (Ramathal et al., 2010). Studies have shown that only 50–60% of pregnancies advance beyond 20 weeks of gestation, while the rest are lost in early stages (Norwitz et al., 2001). Decidualization, which involves the process of transformation of endometrial stromal cells into specialized secretory decidual cells, is a critical and remarkable event for establishing a successful pregnancy. In humans, the decidual reaction is initiated spontaneously during the luteal phase of the menstrual cycle and continues if pregnancy occurs. Progesterone is a key factor in initiating decidualization and is a prerequisite for successful embryo implantation (Wetendorf and DeMayo, 2012). Decidualized stromal cells (DSCs), which can maintain progesterone receptor expression in the presence of progesterone, are derived from the fibroblast-like cells within the endometrium (Okada et al., 2018). The decidualized cells release several important factors including prolactin (PRL) and insulin-like growth factor-binding protein-1 (IGFBP-1), which are well recognized markers of decidualization (Dunn et al., 2003). These various factors help to provide a suitable environment to sustain implantation and pregnancy (Okada et al., 2018). This complex process is accurately controlled both spatially and temporally, through a balance of stimulatory and inhibitory signals. Several signal pathways such as cAMP, Wnt and PI3K-Akt-mTOR signaling are involved in regulating decidualization (Tepekoy et al., 2015; Liang et al., 2016; Zhang et al., 2016; Patterson et al., 2017). Impaired endometrial decidualization is now considered a critical, yet poorly understood, etiologic factor contributing to early implantation failure, miscarriages and infertility. Therefore, it is necessary to improve our knowledge of the regulatory mechanisms in decidualization which may lead to the development of new therapeutic strategies in the reproductive clinical practice.

Basigin (BSG), a highly glycosylated transmembrane protein also known as CD147/EMMPRIN, is a member of the immunoglobulin superfamily containing two Ig domains (Miyauchi et al., 1990). BSG is expressed in the rodent uterus under the influence of ovarian steroid hormones and appears to play an important role during embryo implantation (Xiao et al., 2002; Li and Nowak, 2020). Studies in mouse models have reported that both adult female and male Bsg null mice are infertile, and the cause of infertility in the female mice has been attributed to defective implantation and decidualization (Igakura et al., 1998; Kuno et al., 1998). Studies in the mouse have also shown that Bsg mRNA and protein are highly expressed in the uterine epithelium during proestrus estrus and early days of gestation under estrogen regulation. However, by Days 3 and 4 of gestation, expression shifts primarily to uterine stromal cells under the influence of progesterone (Kuno et al., 1998). Our research group, using an in vitro mouse uterine stromal cell culture system, found that recombinant human BSG protein (rBSG) induced production of matrix metalloproteinases (MMPs), cytokines and chemokines that mediate uterine stromal cell decidualization and trafficking of leukocytes to the endometrium (Chen et al., 2009). In addition, we have also previously reported that conditional deletion of BSG in the reproductive tract led to subfertility in mice. The knockout mice showed impaired implantation and a defective decidualization response (Li et al., 2021). In the human endometrium, BSG expression is menstrual-cycle dependent (Noguchi et al., 2003; Braundmeier et al., 2006). A recent study from Turgut et al. (2014) reported that BSG protein expression was significantly lower during the luteal phase in the uterine endometrium of female patients experiencing repeated implantation failures after IVF when compared to the endometrium of women with normal fertility.

Previous studies have demonstrated that BSG is abundant in reproductive tissues and plays essential roles in reproduction, yet the role of BSG in stromal cell decidualization in humans and the precise mechanisms involved are far from fully understood. Therefore, in the present study, we investigated whether BSG is necessary for decidualization of human endometrial stromal cells (HESCs) using an in vitro culture model and used microarray analysis to identify genes regulated by BSG during the decidualization process. We found that small interfering RNA (siRNA) knockdown of BSG inhibited HESCs proliferation and disrupted HESCs decidualization. Microarray analysis revealed several gene pathways that were regulated by BSG including the WNT signaling pathway, ECM-receptor interactions and PI3K-Akt signaling pathways.

Materials and methods

Human endometrial stromal cell culture and in vitro decidualization model

Telomerase-immortalized HESCs were obtained from American Type Culture Collection (ATCC) (CRL-4003) and were cultured in Dulbecco’s modified Eagle’s medium (DMEM)/F12 (Life Technologies, Carlsbad, CA, USA) supplemented with 10% (v/v) fetal bovine serum (FBS) (Atlanta Biologicals, Lawrenceville, GA, USA) with 1% Penicillin–Streptomycin (Cellgro, Manassas, VA, USA) in a 5% CO₂ in air humidified incubators at 37°C. HESCs were treated with a hormone mixture containing 0.5 mM 8-bromo-cAMP, 1 μM progesterone and 10 nM 17β-estradiol (Sigma, St. Louis, MO, USA) for 0–8 days to study decidualization (Li et al., 2007).

Small interfering RNA transfection procedures

Small interfering RNA corresponding to human BSG (sense: CGUAGAUUCCCAUCAUAC ATT; antisense: UGUAUGAUGGGAAUCUACGGG) were purchased from Ambion (Life technologies, Grand Island, NY, USA). Silencer select negative control #1 (Life technologies) was used as a negative control under the same conditions. The BSG siRNA or negative control #1 were transfected into HESCs following the protocol of siLentFect™ Lipid (Bio-Rad, Hercules, CA, USA). Briefly, 3 μl of siLentFect Lipid transfection reagent were mixed with 5 nM of siRNA or negative control #1 to form complexes and were dispersed into 6-well cell culture plates at 37°C. Following complete knock down of BSG in HESCs (72 h after adding siRNA), HESCs were treated with medium containing estrogen, progesterone and cAMP to induce decidualization. BSG siRNA and negative control #1 were added to the cells every four days (Days 0, 4) over the course of the decidualization protocol. Total RNA and protein were harvested at each time point (Days 0, 2, 4, 6, 8) for either quantitative RT–PC or immunoblotting.

Cell proliferation

Human endometrial stromal cells were seeded at a density of 2.5×10⁴ cells/well in 6-well plates and were cultured in DMEM/F12 medium containing 10% FBS, 1% L-glutamine and 1% penicillin–streptomycin for 24 h. HESCs were then transfected with either BSG siRNA or negative control #1. After 48 h, the siRNA and negative control #1 were removed and cells were cultured in the culture medium for another 24 h. Cells were then harvested, and the number of cells was counted. For tritiated thymidine incorporation assays, HESCs were transfected with either BSGsiRNA or negative control #1 for 48 h. HESCs were then labeled with [3H] Thymidine (2 μCi/ml) (Perkin Elmer, Waltham, MA, USA) at 37°C for another 24 h. Cells were then harvested and counted in a scintillation counter.

RNA isolation and quantitative reverse transcription–PCR

Total RNA was extracted from HESCs using TRIzol™ (Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s instructions. Two micrograms of total RNA were reverse transcribed using High-Capacity cDNA Reverse Transcription Kit (Life Technologies) following the manufacturer’s instructions. qPCR analyses were performed using TaqMan^® Universal PCR Master Mix No AmpErase^® UNG (Life Technologies). The following 20× Assays-on-DemandTM Gene Expression Assay primer-probe sets from Applied Biosystems were used for this study: PRL (Hs00168730_m1), Insulin-like growth factor-binding protein 1 (IGFBP1) (Hs00236877_m1) and POP4 (HS00198357_m1). Briefly, 4.5 μl of a 1:10 diluted cDNA sample were mixed with 5.5 μl of master mix (5 μl of TaqMan Universal PCR Mix and 0.5 μl of 20× Assay on Demand) for a total volume of 10 μl per well in a MicroAmp optical 384-well reaction plate. Three experimental replicates were performed for each sample. qPCR amplification and quantitation were performed using the ABI 7900 sequence detection system for 40 cycles (95°C for 15 s, 60°C for 1 min). The comparative CT method (ΔΔCt) was used for quantification of gene expression. Relative fold differences in gene expression for all tested genes were normalized to the POP4 endogenous control.

Reverse transcription–real-time quantitative PCR was also used to validate some of the gene expression changes identified by RNA microarray. The mRNA expression was determined by RT–qPCR using SYBR Green master mix (Appliedbiosystems by Thermo Fisher Scientific) and primers were synthesized by Integrated DNA Technologies (primers listed in Table I). Relative mRNA expression levels were calculated using the following equation: relative gene expression = 2^{−(ΔCt sample−ΔCt control)}.

Table I.

The sequences of primers used for quantitative real-time PCR.

Gene name	Primers	Sequence (5′-3′)
GAPDH	Forward	5′-TCG GAG TCA ACG GAT TTG GT-3′
	Reverse	5′-TTC CCG TTC TCA GCC TTG AC-3′
WNT4	Forward	5′-CGT GCC TGC GTT CGC T-3′
	Reverse	5′-CCT TGA GTT TCT CGC ACG TC-3′
WNT5	Forward	5′-CTC CAT TCC TGG GCG CAT C-3′
	Reverse	5′-CCA ATG GAC TTC TTC ATG GCG-3′
FZD1	Forward	5′-ATC TTC TTG TCC GGC TGT TAC A-3′
	Reverse	5′-GTC CTC GGC GAA CTT GTC ATT-3′
FZD4	Forward	5′-GGA TGC TCT GTG GCC TTT CT-3′
	Reverse	5′-GGG CAT GTG TAG CAG GAA GT-3′
FOXO1	Forward	5′-ACG AGT GGA TGG TCA AGA GC-3′
	Reverse	5′-AAT TGA ATT CTT CCA GCC CGC-3′
CDK1	Forward	5′-CTT GGC TTC AAA GCT GGC TC-3′
	Reverse	5′-GGG TAT GGT AGA TCC CGG CT-3′
EGFR	Forward	5′-AGG CAC GAG TAA CAA GCT CAC-3′
	Reverse	5′-ATG AGG ACA TAA CCA GCC ACC-3′
CCND1	Forward	5′-CAA TGA CCC CGC ACG ATT TC-3′
	Reverse	5′-AAG TTG TTG GGG CTC CTC AG-3′

Immunoblotting analysis

Cells from the decidualization experiments were trypsinized, pelleted and washed twice with PBS by centrifugation. Pelleted cells were extracted with Tris NP-40 EDTA buffer [10 mM Tris (pH 8.0), 1 mM EDTA, 0.5% Nonidet P-40, 1×protease inhibitor] for 45 min on ice with vortexing every 5 min. Samples were clarified by centrifugation (10 000g for 5 min).

Fifteen micrograms of the cell lysates from each time point were loaded onto SDS–PAGE gels and transferred to Immobilon-P membranes (PVDF) (Millipore, Billerica, MA, USA). Membranes were blocked in 5% nonfat dry milk and then probed with mouse anti human BSG (BD Pharmingen, San Jose, CA, USA) at 1:2000 dilution, mouse anti human MMP-2 (Calbiochem, Billerica, MA, USA) at 1:1000 dilution and mouse anti human MMP-3 (R&D, Minneapolis, MN) at 1:100 dilution overnight at 4°C. The membranes were washed and incubated with HRP-conjugated horse antimouse IgG antibody (Cell Signaling, Boston, MA, USA) at 1:10 000 dilution for 60 min at room temperature. The bound secondary antibody was detected using SuperSignal West Pico Chemiluminescent Substrate (Fisher Scientific, Pittsburgh, PA, USA). The same membranes were stripped and reprobed with anti-GAPDH antibody (Cell Signaling, Boston, MA, USA) as a loading control.

Densitometric analysis of immunoblotting

Films developed after exposure to chemiluminescence were scanned, and images were analyzed using the ImageJ software from NIH (available at http://rsbweb.nih.gov/ij/download.html). Quantitative measurements of BSG bands were normalized for loading differences using measurements of GAPDH. Graphs represent means and SEM.

Microarray analysis

Differentially expressed genes screening

Total RNA was purified on Day 6 of the decidualization timeline following the protocol of the RNeasy Mini Kit (Qiagen, Valencia, CA, USA). Microarray analysis was performed at the University of Illinois at Urbana-Champaign Roy J. Carver Biotechnology Center. Briefly, 0.2 micrograms of total RNA were labeled using the Agilent two color QuickAmp labeling kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocol. The optional spike-in controls were not used. Samples were hybridized to Human Gene Expression 4x44K v2 Microarray (Agilent Technologies, Santa Clara, CA, USA) in an Agilent Hybridization Cassette according to standard protocols. The arrays were then scanned on an Axon GenePix 4000B scanner and the images were quantified using Axon GenePix 6.1.

Microarray data preprocessing and statistical analyses were done in R (v3.6.2) using the limma package (3.42.0; Ritchie et al., 2015). Median foreground and median background values from the four arrays were read into R and any spots that had been manually flagged (−100 values) were given a weight of zero. The background values were ignored because investigations showed that trying to use them to adjust for background fluorescence added more noise to the data; background was low and even for all arrays, therefore no background correction was done.

The individual Cy5 and Cy3 fluorescence for each array were normalized together using the quantile method 3 (Yang and Thorne, 2003). Agilent’s Human Gene Expression 4x44K v2 Microarray has a total of 45 220 probes: 1224 probes for positive controls, 153 negative control, 823 labeled ‘ignore’ and 43 118 labeled ‘cDNA’. The pos + neg + ignore probes were used to ascertain the background level of fluorescence (6, on the log2 scale) then discarded. The cDNA probes comprise 34 127 unique 60mer probes, of which 999 probes are spotted 10 times each and the rest one time each. We averaged the replicate probes for those spotted 10 times and then fit a mixed model that had treatment and dye as fixed effects and array pairing as a random effect (Smyth et al., 2005; Phipson et al., 2016). After fitting the model but before false discovery rate (FDR) correction (Benjamini and Hochberg, 1995), probes were filtered out by the following criteria: (i) did not have at least 4/8 samples with expression values >6 (14 105 probes removed), (ii) no longer had an assigned Entrez Gene ID in Bioconductor’s HsAgilentDesign026652.db annotation package (v3.2.3; 2152 probes removed) (Huber et al., 2015), (iii) mapped to the same Entrez Gene ID as another probe but had a larger P-value for treatment effect (4141 probes removed). This left 13 729 probes representing 13 729 unique genes. We also conducted a ‘treat’ test where instead of testing whether the treatment effect, |log2(SI/NC)|, was greater than 0, we tested whether it was greater than log2(1.1), which allowed us to focus on genes with larger fold-changes while still controlling the FDR properly (McCarthy and Smyth, 2009). For visualization of top 25 differentially expressed genes (DEGs), unsupervised hierarchical clustering was performed using HemI 1.0.3.7software (http://hemi.biocuckoo.org/).

Functional and pathway enrichment analysis of DEGs

Genes with FDR P-value <0.05 were tested for overrepresentation of Gene Ontology terms using limma’s goana function, which tests upregulated genes separately from downregulated genes using the Entrez ID—GO mappings from the org. Hs.eg.db annotation package (v3.10.0). Likewise, we tested for overrepresentation of KEGG pathways separately for up- and downregulated genes using limma’s kegga function and updated Entrez ID—KEGG mappings downloaded directly from KEGG9 on January 10, 2020.

Protein–protein interaction network construction and selection of modules

The protein–protein interaction (PPI) network information was explored by using the online tool Search for the Retrieval of Interacting Genes (STRING, version 11.0) database (http://www.string-db.org). We imported the DEGs to STRING to identify the interactive relationship among DEGs. Interactions with a confidence score >0.7 were retained in the network and were visualized using Cytoscape (data not shown) (version 3.7.2, http://cytoscape.org). Then we output the PPI network data and imported them to Cytoscape to organize the module analysis with the Plugin MCODE with the default parameters (Degree cutoff ≥2, Node score cutoff ≥2, K-core ≥2 and max depth = 100). Furthermore, we performed function and pathway enrichment analysis for DEGs in the modules and hub genes using the string database. FDR < 0.05 was considered to be significant.

Statistical analysis

Statistics were performed using SPSS 18.0 software. For evaluating the statistical significance of differences between two group means, paired and/or unpaired Student t-test was used, and for multiple group comparison, ANOVA was used. When an F-test result was significant (P < 0.05), a further Newman–Keuls multiple comparison test was applied to analyze differences among the means. A P-value of < 0.05 was defined as statistically significant.

Results

Silencing of BSG expression inhibits stromal cell proliferation

As stromal cell decidualization is essential for embryo implantation and mice lacking BSG in the uterus were subfertile with impaired decidualization, we hypothesized that BSG in stromal cells might play a vital role during the decidualization process in humans. We then tested this hypothesis in HESCs in vitro. Proliferation of stromal cells is the first step of decidualization. To determine the effect of BSG on human stromal cell proliferation, BSG siRNA was used to knockdown gene and protein expression in HESC cells (Fig. 1A and B). Both mRNA and protein levels of BSG were markedly reduced by BSG siRNA (5 nM) treatment. Tritiated thymidine incorporation assays showed that the cells transfected with BSG siRNA exhibited around a 50% reduction in proliferation rate compared with scrambled siRNA controls after 72 h of treatment for BSG siRNA (Fig. 1C). We found that cells transfected with BSG siRNA (5 nM) exhibited an approximately 70% reduction in cell number compared with scrambled siRNA controls after 72 h of treatment with BSG siRNA (Fig. 1D).

Figure 1.

The effect of basigin knockdown on human endometrial stromal cell (HESCs) proliferation. HESCs were transfected with either BSG siRNA (5 nM) or scrambled siRNA (control) for 72 h. Changes in cell number and incorporation of [3H]-thymidine were determined. (A) Relative levels of BSG mRNA in HESCs treated with either negative control scrambled siRNA or BSG siRNA; (B) Western blot of BSG expression in HESCs transfected with si-BSG for 72 h and relative levels of BSG protein expression. (C) DNA synthesis (thymidine incorporation) in BSG siRNA treated cells was significantly decreased in comparison to the control cells. NC: negative control (scrambled siRNA). Bars represent the fold-change±SEM of three independent experiments and the asterisk indicates statistical differences (P < 0.05). (D) Cell numbers of BSG siRNA treated cells were significantly decreased in comparison to the control cells; Y axis shows cell number as X×10⁵.

Silencing of BSG expression disrupts HESCs decidualization

To further determine the role of BSG in stromal cell differentiation (decidualization), we used a hormone combination (E2 + P + cAMP) treatment protocol to induce decidualization after 72-h transfection of BSG siRNA or scrambled siRNA. BSG protein levels in the cells transfected with BSG siRNA were knocked down to extremely low levels with little expression compared to controls (Fig. 2A). IGFBP1 and PRL are well-known markers of decidualization in human uterine stromal cells (Eyal et al., 2007; Li et al., 2007). In the control group, the expression of IGFBP1 and PRL was significantly increased during the 8-day cell culture. These results confirmed that the HESCs were decidualized successfully. In addition, following effective BSG knockdown (Fig. 2A and B), HESCs showed greatly impaired decidualization, as the decidual markers, IGFBP1 and PRL were significantly decreased starting on D4 and D6, respectively, compared to the controls (Fig. 2C and D). These results indicate that expression of BSG by human stromal cells is necessary for normal decidualization.

Figure 2.

Silencing of basigin (BSG) expression inhibits decidualization. human endometrial stromal cells (HESCs) were transfected with either BSG siRNA (5 nM) or scrambled siRNA (control) and subjected to in vitro decidualization as described in Materials and methods. (A) Immunoblotting analysis of BSG protein from HESCs transfected with BSG siRNA and scrambled siRNA. GAPDH served as the loading control. (B) Bar graph summarizing the densitometric analysis of the immunoblotting. (C, D) HESCs were harvested at indicated times after addition of the hormone regimen. Total RNA was isolated and subjected to qPCR using gene-specific primers for the decidualization markers IGFBP1(C) and PRL (D). In control group, the IGFBP1and PRL increased when decidualization inducer added into the cell culture system. Compared to control group, the expression of IGFBP1 in siBSG group was significantly decreased from D4. And the PRL expression was significantly decreased from Day 6. Y axis indicates fold expression. NC: negative control (scrambled siRNA). Bars represent the fold-change±SEM of three independent experiments. The control group with different letters is significantly different (P < 0.05); significant differences within a specified concentration are denoted with an asterisk.

Silencing of BSG expression downregulates MMPs expression

Basigin is recognized as a potential inducer of MMPs (Kataoka et al., 1993; Gabison et al., 2005). Studies have confirmed that MMP-2 and MMP-3 are expressed in human uterine stromal cells (Braundmeier and Nowak, 2006; Itoh et al., 2012). Decidualization is partially regulated by the expression of MMPs and their inhibitors, the tissue inhibitors of metalloproteinases (TIMPs) (Chen, et al., 2009). We further investigated whether MMP-2 and MMP-3 levels were altered in HESCs in response to knock down of BSG as well as disrupted decidualization. The results showed that both MMP-2 and MMP-3 protein levels steadily increased during cell culture and significantly increased on D4 and D8, respectively, compared to D0 (Fig. 3A). However, in the BSG siRNA group, MMP-2 and MMP-3 levels were significantly reduced compared to the controls on D4 and D8 (Fig. 3B and C). These results indicate that silencing of BSG leads to downregulation of MMP-2 and MMP-3.

Figure 3.

Silencing of basigin (BSG) expression reduces matrix metalloproteinase (MMP) production. Human endometrial stromal cells (HESCs) were transfected with either BSG siRNA (5 nM) or scrambled siRNA (control) and subjected to in vitro decidualization as described in Materials and methods. (A) Immunoblotting analysis of MMP-2 and MMP-3 proteins from HESCs transfected with BSG siRNA or scrambled siRNA. GAPDH served as the loading control. In control group, the expression of MMP-2 was significantly increased from Day 4 and the MMP-3 was significantly increased on Day 8. (B) Bar graph summarizing the densitometric analysis of MMP-2 immunoblotting. (C) Bar graph summarizing the densitometric analysis of MMP-3 immunoblotting. NC: negative control (scrambled siRNA). Bars represent the fold-change±SEM of three independent experiments. The control group with different letters is significantly different (P < 0.05); significant differences within a specified concentration are denoted with an asterisk.

Global gene expression analysis of BSG SiRNA-treated HESCs

Although the results above and in previous studies both show that BSG is important for decidualization, the precise mechanism involved in this process needs to be further explored. To characterize the genes that are regulated by BSG during decidualization of HESCs, we examined the gene expression profile of BSG siRNA treated and scrambled siRNA treated HESCs using microarrays. Using the analysis described in the Materials and methods section, we determined that 721 genes were downregulated and 484 genes were upregulated in the BSG siRNA treated group (P < 0.05) compared to scrambled siRNA controls (https://doi.org/10.13012/B2IDB-5457341_V1). Further, unsupervised hierarchical clustering analysis showed distinct patterns of up- and downregulated genes in the BSG siRNA treated HESCs (Fig. 4).

Figure 4.

Heat map of differentially expressed genes (DEGs) in siBSG group compared with control group. Heat map of the top 25 differentially expressed genes. Red, upregulation; blue, downregulation.

Functional annotation for the DEGs

The identified DEGs in the BSG siRNA treated HESCs were further analyzed via gene ontology (GO) and KEGG pathway analysis. For GO terms, biological process (BP), cellular component (CC) and molecular function (MF) were analyzed and the top five of each category are shown in Fig. 5A (top 5). GO term enrichment analysis showed that the upregulated genes were significantly enriched in cell communication, signaling regulation of cell communication, response to stimulus, signal transduction in the BP category; cell periphery, plasma membrane, cellular component, cell surface, cytoplasm in the CC category; and thrombin-activated receptor activity, transmembrane signaling receptor activity, TRAIL binding, signaling receptor activity, protein binding in the MF category.

Figure 5.

Function and pathway enrichment analysis of up- and downregulated genes in siBSG treated human endometrial stromal cells (HESCs). (A) The gene ontology enrichment analysis of up differentially expressed genes (DEGs) covered three categories: biological process, cellular component and molecular function. (B) The gene ontology enrichment analysis of down DEGs. (C) The top 10 enriched KEGG pathways for upregulated genes. (D) The top 10 enriched KEGG pathways for downregulated genes.

As depicted in Fig. 5B (top 5), GO term enrichment analysis showed that the downregulated genes were significantly enriched in biological adhesion, cell adhesion, anatomical structure morphogenesis, extracellular matrix (ECM) organization, multicellular organismal process in the BP category, ECM, collagen-containing ECM, extracellular region part, extracellular region, extracellular space in the CC category and ECM structural constituent, signaling receptor binding, growth factor binding, glycosaminoglycan binding, molecular transducer activity in the MF category.

Furthermore, KEGG pathway analysis showed that upregulated genes in the BSG siRNA treated group were enriched in pathways such as MAPK signaling pathway, Colorectal cancer, Pathways in cancer, Melanoma and Axon guidance (top 10 shown in Fig. 5C and Supplementary Table SI). In contrast, genes downregulated by knock down of BSG were mainly enriched in pathways including ECM–receptor interaction, PI3K-Akt signaling pathway, WNT signaling, pathways in cancer, antigen processing, type I diabetes mellitus and focal adhesion (Fig. 5D and Supplementary Table SII).

Protein–protein interaction networks and module analysis

To analyze the interaction among DEGs and acquire hub genes, we used STRING to construct a protein–protein interactome. We then output the genes with combined score ≥0.7 and imported them into Cytoscape for further analysis. The PPI network identified 1166 nodes and 2933 interactions (data not shown). The top 10 hub nodes using 12 methods analyses are shown in Table II. The hub genes that appeared in two methods were screened out (in bold). Among these genes, upregulated genes included EGFR, HSP90AA1, CCND1, PXN, PRKACB, MGAT4A, EVA1A, LGALS1, STC2, HSPA4; downregulated genes included CDK1, PIK3R1, IGF1, JAK2, LAMB1, ITGAV, HGF, MXRA8, TMEM132A, UBE2C, QSOX1, ERBB2, GNB4, HSP90B1, LAMB2, LAMC1 and ITGA1. Pathway enrichment analysis for all these hub genes using string database showed that the top three pathways are the PI3K-AKT signaling pathway, focal adhesion and pathways in cancer (shown in Table III). GO term enrichment analysis showed that these hub genes were significantly enriched in cellular protein modification processes, posttranslational protein modification, cell morphogenesis in the BP category, endoplasmic reticulum lumen, endomembrane system, cytoplasmic part in the CC category and signaling receptor binding, integrin binding, phosphotransferase activity, alcohol group as acceptor in the MF category (Table III).

Table II.

Hub genes in differentially expressed genes (DEGs): 10 up red and 17 down bold.

Rank/method	Betweenness	BottleNeck	Closeness	ClusteringCoefficient	Degree	DMNC
1	EGFR	EGFR	EGFR	MGAT4A	EGFR	MGAT4A
2	HSP90AA1	HSP90AA1	PIK3R1	EVA1A	PIK3R1	EVA1A
3	CCND1	PIK3R1	HSP90AA1	MXRA8	CDK1	MXRA8
4	PIK3R1	CCND1	CDK1	TMEM132A	HSP90AA1	TMEM132A
5	CDK1	NEDD4L	ERBB2	LGALS1	UBE2C	LGALS1
6	PRKACB	QSOX1	IGF1	STC2	IGF1	STC2
7	PXN	HSPA4	JAK2	GPER1	GNB4	PRKCSH
8	ITGAV	PRKCA	HGF	GPR183	JAK2	HSP90B1
9	PRKCA	TUBB	CCND1	AMPH	LAMB1	OIP5
10	RHOC	PXN	PXN	FAM49B	ITGAV	IGFBP1

Rank/Method	EcCentricity	EPC	MCC	MNC	Radiality	Stress
1	SYNJ2	EGFR	LAMB1	EGFR	EGFR	EGFR
2	ADCY3	PIK3R1	CDK1	PIK3R1	HSP90AA1	CCND1
3	HSPA4	JAK2	LAMB2	CDK1	PIK3R1	PIK3R1
4	PRKACB	IGF1	LAMC1	UBE2C	ERBB2	HSP90AA1
5	MGAT4A	LAMB1	VCAN	HSP90AA1	IGF1	CDK1
6	EVA1A	HGF	QSOX1	GNB4	JAK2	ITGAV
7	MXRA8	HSP90AA1	UBE2C	IGF1	PXN	PRKACB
8	TMEM132A	CDK1	FBN1	JAK2	CCND1	PXN
9	LGALS1	LAMB2	HSP90B1	LAMB1	ITGA1	ITGA1
10	STC2	LAMC1	IGFBP3	HGF	CDK1	IGF1

10 upregulated genes in red and 17 downregulated genes in bold. Up genes: EGFR(9), HSP90AA1(8), CCND1(5), PXN(5), PRKACB(3), MGAT4A(3), EVA1A(3), LGALS1(3), STC2(3), HSPA4(2). Down genes: CDK1(8), PIK3R1(7), IGF1(5), JAK2(5), LAMB1(4), ITGAV(3), HGF(3), MXRA8(3), TMEM132A(3), UBE2C3), QSOX1(2), ERBB2(2), GNB4(2), HSP90B1(2), LAMB2(2), LAMC1(2), ITGA1(2).

Table III.

KEGG and GO enrichment analysis for hub genes (top 5).

Term ID	Term description	Observed gene count	FDR	matching proteins in network (labels)
KEGG enrichment analysis
hsa04151	PI3K-Akt signaling pathway	16	5.98e−19	CCND1, EGFR, ERBB2, GNB4, HGF, HSP90AA1, HSP90B1, IGF1, ITGA1, ITGAV, JAK2, LAMB1, LAMB2, LAMC1, PIK3R1, PRKCA
hsa04510	Focal adhesion	13	4.57e−17	CCND1, EGFR, ERBB2, HGF, IGF1, ITGA1, ITGAV, LAMB1, LAMB2, LAMC1, PIK3R1, PRKCA, PXN
hsa05200	Pathways in cancer	16	8.45e−17	CCND1, EGFR, ERBB2, GNB4, HGF, HSP90AA1, HSP90B1, IGF1, ITGAV, JAK2, LAMB1, LAMB2, LAMC1, PIK3R1, PRKACB, PRKCA
hsa05205	Proteoglycans in cancer	10	5.83e−12	CCND1, EGFR, ERBB2, HGF, IGF1, ITGAV, PIK3R1, PRKACB, PRKCA, PXN
hsa05165	Human papillomavirus infection	10	4.88e−10
Functional enrichment analysis
MF
GO:0005102	Signaling receptor binding	13	5.44e−06	EGFR, ERBB2, HGF, HSP90B1, IGF1, ITGA1, ITGAV, JAK2, LAMB2, PIK3R1, PRKCA, PXN, STC2
GO:0005178	Integrin binding	6	5.44e−06	EGFR, IGF1, ITGAV, LAMB2, PRKCA, PXN
GO:0016773	Phosphotransferase activity, alcohol group as acceptor	10	5.44e−06	CCND1, CDK1, EGFR, ERBB2, HGF, HSP90AA1, JAK2, PIK3R1, PRKACB, PRKCA
GO:0019899	Enzyme binding	15	5.44e−06	CCND1, EGFR, ERBB2, HGF, HSP90AA1, HSP90B1, ITGA1, ITGAV, JAK2, PIK3R1, PRKACB, PRKCA, PXN, STC2, UBE2C
GO:0016301	Kinase activity	10	6.41e−06	CCND1, CDK1, EGFR, ERBB2, HGF, HSP90AA1, JAK2, PIK3R1, PRKACB, PRKCA
CC
GO:0005788	Endoplasmic reticulum lumen	11	5.68e−11	EVA1A, HSP90B1, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, QSOX1, STC2, TMEM132A
GO:0012505	Endomembrane system	22	5.12e−08	CDK1, EGFR, ERBB2, EVA1A, HGF, HSP90AA1, HSP90B1, IGF1, ITGA1, ITGAV, JAK2, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, PIK3R1, PRKCA, QSOX1, STC2, TMEM132A
GO:0044432	Endoplasmic reticulum part	14	5.12e−08	CDK1, EGFR, EVA1A, HSP90B1, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, PIK3R1, QSOX1, STC2, TMEM132A
GO:0044444	Cytoplasmic part	28	7.50e−08	CCND1, CDK1, EGFR, ERBB2, EVA1A, GNB4, HGF, HSP90AA1, HSP90B1, HSPA4, IGF1, ITGA1, ITGAV, JAK2, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, PIK3R1, PRKACB, PRKCA, PXN, QSOX1, STC2, TMEM132A, UBE2C
GO:0005783	Endoplasmic reticulum	15	1.75e−07	CDK1, EGFR, EVA1A, HSP90B1, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, PIK3R1, PRKCA, QSOX1, STC2, TMEM132A
BP
GO:0006464	Cellular protein modification process	24	6.27e−13	CCND1, CDK1, EGFR, ERBB2, EVA1A, HGF, HSP90AA1, HSP90B1, IGF1, ITGAV, JAK2, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, PIK3R1, PRKACB, PRKCA, QSOX1, STC2, TMEM132A, UBE2C
GO:0043687	Posttranslational protein modification	11	5.67e−10	EVA1A, HSP90B1, LAMB1, LAMB2, LAMC1, LGALS1, MGAT4A, MXRA8, QSOX1, STC2, TMEM132A
GO:0000902	Cell morphogenesis	11	1.21e−07	EGFR, ERBB2, HGF, HSP90AA1, ITGA1, ITGAV, LAMB1, LAMB2, LAMC1, PIK3R1, PRKCA
GO:0051347	Positive regulation of transferase activity	11	1.21e−07	CCND1, CDK1, EGFR, ERBB2, HGF, HSP90AA1, IGF1, ITGA1, JAK2, PRKACB, UBE2C
GO:0007169	Transmembrane receptor protein tyrosine kinase signaling pathway	10	2.09e−07	EGFR, ERBB2, HGF, HSP90AA1, IGF1, ITGAV, JAK2, PIK3R1, PRKCA, PXN

MCODE was used to identify the module genes of the PPI network. We selected the top three modules (Fig. 6) and then used the STRING database to do further functional annotation of these genes involved in the three modules. The analysis showed that module genes were related mainly to PI3K-AKT signaling pathway, pathways in cancer and ECM (data shown in Supplementary Tables SIII, SIV and SV).

Figure 6.

The clusters identified in the protein–protein interaction (PPI) network of the differentially expressed genes (DEGs) from siBSG treated human endometrial stromal cells (HESCs). (A) Cluster 1 with the highest degree (19), include 20 nodes and 190 edges. (B) Cluster 2, include 37 nodes and 326 edges, average node degree: 17.6. (C) Cluster 3, include 13 nodes and 78 edges, average node degree: 12. All three clusters PPI enrichment P < 1.0e-16. The nodes stand for the gene, and edge stands for the interaction of genes. Black edges are derived from STRING. Red, represents upregulated gene; green, represents downregulated genes.

Validation of DEGs via RT–PCR

To confirm vital genes identified by the above analysis, three hub genes (EGFR, CCND1 and CDK1) and five reported decidualization associated genes (FOXO1, WNT4, WNT5A, FZD1 and FZD4) were evaluated using RT–PCR. Our microarray data showed that in the BSG siRNA group, CDK1, FOXO1, WNT4, WNT5A, FZD1 and FZD4 were downregulated DEGs, while CCND1 and EGFR were upregulated DEGs. In the RT–PCR results (Fig. 7), the relative expression levels of CCND1 were increased, while those of CDK1, FOXO1, WNT4, WNT5A, FZD1 and FZD4 were significantly decreased in BSG siRNA treated cells compared with controls, which was consistent with the DEG analysis. Although the expression level of EGFR in the two groups showed no significant difference, it was still higher in the BSG siRNA treated group than in the control scrambled siRNA group.

Figure 7.

Gene expression validation by real-time PCR. Triplicate assays were performed for each sample, and the relative level of each gene was normalized to GAPDH. Error bars represent SEM. Statistically significant differences (P < 0.05) are indicated by asterisks.

Discussion

Several studies have confirmed that endometrial stromal cells play a critical role in the implantation process (Okada, Tsuzuki and Murata, 2018). Studies on primary endometrial stromal cells have provided evidence that programming defects are linked to failed reproductive ability (Chen et al., 2017; Soczewski et al., 2020). In order to explore the potential role of BSG in endometrial stromal cells in human, we used telomerase-immortalized HESCs as a cell model and found that downregulation of BSG in the cells resulted in disruption of decidualization and inhibition of cell proliferation and differentiation.

Since the discovery of BSG three decades ago, it has been considered as a major inducer of MMPs in stromal cells, including endometrial stromal cells, by many investigators. Recombinant BSG protein treatment can stimulate the production of specific MMPs, including MMPs-1, -2, and-3 in human uterine stromal cells and MMPs-3 and -9 in mouse stromal cells (Belton et al., 2008; Braundmeier et al., 2012). In our study, we found that BSG siRNA treatment significantly reduced the production of MMP-2 and MMP-3. As MMPs are vital components of the ECM, and the degradation and remodeling of the ECM is a critical important event in all processes involved in normal human reproduction (Hulboy et al., 1997) including endometrial decidualization, we propose that BSG regulates HESCs decidualization at least partially through regulating MMPs.

Our results demonstrated that the knock down of BSG was associated with defective decidualization of stromal cells. This is consistent with the previous report that loss of BSG expression in the uterus caused impaired implantation and decidualization leading to subfertility in mice (Li et al., 2021). To investigate the potential mechanism, we used mRNA microarray to identify genes regulated by BSG by comparing BSG siRNA treated HESCs to scrambled siRNA treated control cells. In total, there were 721 genes downregulated and 484 genes upregulated in the BSG siRNA treatment group with P < 0.05. Also consistent with previous results, marker genes of decidualization such as IGFBP1, PRL and MMPs were significantly downregulated (as shown in Figs. 2 and 3).

WNTs are highly conserved secreted glycoproteins and more than 15 different ligands have been reported in vertebrates (Li et al., 2013). WNTs modulate developmental and other processes through interaction with Frizzle homology (FZD) receptors. Studies have shown that WNT signaling is crucial for implantation and decidualization (Ramathal et al., 2010; Li et al., 2013; Shukla et al., 2019; Zhou et al., 2019). Our microarray analyses revealed that the expression of WNT4 was significantly decreased in siBSG group. Moreover, the expression of other WNT ligands, including WNT2, WNT3, WNT5A WNT5B and WNT7A were also markedly downregulated (Fig. 7 and Supplementary Table SIII). Several Frizzled genes, including FZD1, FZD4, FZD9 and FZD10, were also significantly decreased in response to silencing of BSG (Fig. 7 and Supplementary Table SII). Forkhead box protein O1 (FOXO1), an evolutionary conserved transcription factor, is one of the earliest induced transcription factors in HESCs in response to cAMP signaling and is required for decidualization. In our study, we found significantly downregulated levels of FOXO1 in the BSG siRNA group (Fig. 7 and Supplementary Table SII). It has been reported that the WNT4/β-catenin pathway controls the expression of FOXO1 and its downstream genes that regulate decidualization (Li et al., 2013; Ring et al., 2014; Ujvari et al., 2017). FOXO1 has also been shown to be involved in the regulation of ECM remodeling (Takano et al., 2007). Taken together, our data suggests that BSG may regulate decidualization of HESCs through the WNT4/β-catenin and FOXO1 pathway.

To gain insight into the function of the DEGs, GO terms and KEGG pathway annotation analysis were applied to identify the BSG significant enriched pathways and GO terms. KEGG annotation revealed that DEGs involved in important signaling pathways (TOP 10, see in Fig. 5A and B) related to proliferation and differentiation (MAPK, and PI3K-Akt signaling pathway), adhesion (ECM–receptor interaction, axon guidance, focal adhesion, cell adhesion molecules (CAMs) and Rap1 signaling pathway), oncogenesis (colorectal cancer, pathways in cancer, melanoma, transcriptional misregulation in cancer, proteoglycans in cancer, chronic myeloid leukemia and endometrial cancer) and metabolism (mineral absorption type I diabetes mellitus, ovarian steroidogenesis and calcium signaling pathway) were significantly enriched among the up- or downregulated genes.

Several of these pathways have already been reported to be regulated by BSG. For example, BSG has been reported to induce angiogenesis via upregulation of the PI3K-AKT and IGF-1 pathways (Yin et al., 2017). Tang et al. (2006) also showed that, in both tumor and fibroblast cells, BSG regulates VEGF production through the PI3K-Akt pathway. Moreover, another previous study demonstrated that BSG secreted from lung carcinoma cell microvesicles promoted ECM degradation through the MAPK pathway (Sidhu et al., 2004). Lim et al. (1998) reported that in human lung fibroblasts, BSG stimulates the production of MMP-1 through the p38 MAPK signaling pathway. Meanwhile, one study reported that decidualization of mouse endometrial stromal cells was regulated via the phosphoinositide 3 kinase/mammalian target of rapamycin (PI3K-Akt-mTOR) signaling pathways (Zhang et al., 2016). In addition, EP-induced decidualization of endometrial stromal cells was alleviated by metformin through activation of p38-MAPK signaling in PCOS patients (Xiong et al., 2019). In accordance with the KEGG analysis, the significantly enriched GO terms were related to signal transduction, biological adhesion, cell adhesion, anatomical structure morphogenesis, ECM organization, growth and metabolism. These GO terms are consistent with the knowledge that defective function of BSG is a primary cause for reduction of signal transduction and ECM production. The functional details revealed through bioinformatic analyses and published studies of others indicate that the DEGs regulated by siBSG target a variety of signaling pathways involved in the regulation of HESCs decidualization.

We also analyzed the interaction and acquired hub genes of DEGs (Fig. 6). Among these hub genes, upregulated genes included EGFR, CCND1 (also known as cyclin D1), downregulated genes included CDK1 and these were chosen for further validation. Previous studies reported that the expression of EGFR was progesterone and cAMP-dependent and induced during in vitro decidualization of HESCs (Lockwood et al., 2000; Chobotova et al., 2005). In this study, we also used progesterone and cAMP to induce stromal cell decidualization. In the BSG siRNA treated group, the expression of EGFR was increased more evidently than in the scrambled siRNA controls, suggesting that the downregulation of BSG might increase EGFR production, but the precise mechanism needs further exploration. Cellular differentiation usually commences after cell cycle arrest. Cyclin D1 can form a complex with CDK4 or CDK6, which is required for the transition from G1 to S phase. Wang et al. (2018) found that both in vitro and in vivo, the expression level of cyclin D1 was significantly lower in DSCs than in control cells. These results indicate that during decidualizaton of endometrial cells, In our study, we found that when decidualization of HESCs was inhibited by BSG siRNA, the CCND1 gene was upregulated, suggesting that knockdown of BSG might result in an increase in cyclin D1, preventing the stromal cells exiting from G1/G0 cell arrest and thereby suppressing decidualization.

CDK1 is also a pivotal cell cycle regulator. It binds to G1/S cyclin and is critical for the preparation of S phase (e.g. duplication of centromeres or the spindle pole bodies). Ullah et al. (2018) determined that during the formation of syncytiotrophoblast in the placenta, CDK1 and cyclin B1 are both downregulated at the mRNA and protein levels. They demonstrated that direct inhibition of CDK1 could induce placental cell fusion (Ullah et al., 2018). Consistent with these studies, we found that when HESCs proliferation and decidualization were inhibited, CDK1 expression was significantly downregulated. As CDK1 and cyclin D1 are both cell cycle regulators, in some situations, they regulate the cell cycles together and have a similar expression pattern (Wang et al., 2018), but in our study, they displayed an opposite expression pattern. Cdk1 and cyclin B1 are required for the transition from the G2 phase into mitosis and this is when expression peaks (Malumbres, 2014), while cyclin D1 regulates the transition from G1 to S phase with the cyclin-dependent kinase 4 (CDK) and CDK6 (Montalto and De Amicis, 2020). Although they are both cell cycle regulators, they control different cell cycle phases, and their expression may be regulated through different signaling pathways or mechanisms. They may therefore show different responses to knockdown of a specific gene, in this case BSG. Taken together, the hub genes identified in our studies indicate that BSG siRNA negatively impacts endometrial stromal cell proliferation and differentiation, and these genes play important roles in decidualization of HESCs.

In conclusion, our results showed that BSG is essential for decidualization in HESCs. Knockdown of BSG inhibited HESCs proliferation and MMP production, and disrupted HESCs decidualization. Microarray analysis revealed that several pathways such as ECM–receptor interactions, WNT4/β-catenin and FOXO1 signaling pathways, and PI3K-Akt signaling pathway are involved in HESCs decidualization. Our study using bioinformatics analysis identified a cluster of potential decidualization-related genes and molecular pathways for future investigation and may be useful for exploring the molecular mechanisms by which BSG regulates decidualization. While the present studies utilized an in vitro cell culture system, the results parallel findings reported in an in vivo mouse model. Further experiments are needed to determine these hub genes functions during the process of BSG regulated decidualization.

Data Availability

All data underlying this article are available in the article and have also been published online at databank.illinois.edu as: https://doi.org/10.13012/B2IDB-5457341_V1

Authors’ roles

S.Y. and J.B. made substantial contributions to the design of experiments, acquisition of data, interpretation of data and writing of manuscript; K.L. contributed to data interpretation and writing of manuscript; J.D. carried out computational analysis of all microarray data; R.A.N. made major contributions to conception and design of all studies, analysis and interpretation of data and writing of the manuscript.

Funding

This work was supported by the National Institutes of Health U54 HD40093 (R.A.N.).

Conflict of interest

The authors have no competing interests to declare.