MSCs were isolated from fresh umbilical cords within 4 h. The umbilical cords were obtained from the normal pregnant women undergoing cesarean section in Obstetrics and Gynecology Department of Nanjing Drum Tower Hospital. This study was reviewed and approved by the Ethics Committee of Drum Tower Hospital, Nanjing University Medical School and our research was carried out in accordance with the Helsinki Declaration. Written informed consent was obtained from all participants. The MSC separation method is based on our previous report (2). Briefly, the umbilical cord was first rinsed with PBS containing penicillin and streptomycin. The arteries and veins in the umbilical cord were then stripped and the remaining umbilical cord tissue was cut to pieces. Then, the tissues were digested at 37℃ for 3 h with enzymes. After termination of digestion with DF12 medium containing FBS, the digestive mixture was filtrated through 40 um screen mesh. Subsequently, the supernatant was centrifuged and suspended with DF12 complete medium. Then the cells were incubated at 37℃ in a 5% CO₂ saturated humidified atmosphere. When the cells were spread to 80%, they were digested and passaged with 0.25% trypsin. MSCs were characterized by flow cytometry for the expression of CD11b, CD73, CD90, CD29, CD44, CD45, CD106, HLA-DR, CD19, and CD14. MSCs were treated with E2 (100 nM) for 24 h as the E2 group; and the cells were treated with the same amount of medium for 24 h as the control group.

The methods of total RNA isolation and quantitative real-time PCR were described in our previous study (13). RNA was isolated according to a standard Trizol reagent kit (Invitrogen) protocol. cDNA was synthesized through reverse transcription according to a standard reverse transcription kit (Takara, Japan) protocol. Quantitative real-time PCR was performed on LC480 (Roche). All reactions were run in duplicate. GAPDH served as an internal reference gene and all qPCR data were normalized to GAPDH.

Trizol reagent (Cat#15596-018, CA, USA) was used to extract total RNA according to the manufacturer’s instructions. RNA Integrity Number (RIN, 1 to 10) based on the ratio of the 18S to 28S ribosomal subunits could inspect RNA integration by an Agilent Bioanalyzer 2100 (Santa Clara, CA, USA). RNA with RIN (>7.0) and 28S/18S (>0.7) is up to standard. Then RNeasy Mini Kit (Cat #74106, Qiagen, Germany) and RNase-Free DNase Set (Cat #254, Qiagen) were used for further purification.

The purified RNA was amplified and labeled by Low Input Quick Amp Labeling Kit, One-Color (Cat#5190-2305, Agilent) according to the manufacturer’s instructions. Then labeled cRNA was purified using the RNeasy mini kit based on the RNA purification.

A gene expression hybridization kit (Cat#5188-5242, Agilent) was used for hybridization of slides (Probe name version: Agilent-014850 Whole Human Genome Microarray 4x44K) with 1.65 μg of Cy3-labeled cRNA at 65℃ and 10 rpm for 17 h in a hybridization oven (Cat#G2545A, Agilent) according to the manufacturer’s instructions. Then slides were washed with a gene expression wash buffer kit (Cat#5188-5327, Agilent).

An Agilent microarray scanner (Cat#G2565CA) was used to scan slides. Subsequently, data were extracted using Feature Extraction software 10.7 (Agilent). Raw data were normalized using the Quantile algorithm with Gene Spring Software 11.0 (Agilent). Lastly, eight samples were obtained including four E2 samples (E2-1, E2-2, E2-3, E2-4) and four control samples (Ctl-1, Ctl-2, Ctl-3, Ctl-4).

The limma package (14) in R software was used to identify the DEGs between the E2 and control groups according to the expression values of the probe in the slides. The threshold value of DEGs was set as |log fold change (FC)| >1 and p value <0.05. Then, the clustering heatmap was constructed using the pheatmap function in R package (version 1.0.8, https://CRAN.R-project.org/package=pheatmap).

The GO and KEGG pathway enrichment analysis of DEGs was used to assess the Molecular Function (MF), Biological Process (BP), Cellular Component (CC) and pathway of DEGs via DAVID 6.8. p value <0.05 was significantly different.

PPI network was constructed with DEGs using STRING (Version 10.0, http://www.string-db.org/) under a threshold of combined score >0.4 (required confidence), and was visualized by Cytoscape software (Version 3.4.0, http://cytoscape.org/). In addition, CytoNCA (Version 2.1.6, http://apps.cytoscape.org/apps/cytonca) was used to analyze the network topology, and the parameter was set as “without weight” (15). The importance of each node within the network was analyzed according to degree centrality (16), betweenness centrality (17) and closeness centrality (18). In the PPI network, proteins with similar function were clustered together, and the distance between nodes was related to their functions. In this way, MCODE module analysis could predict the unknown functions of proteins in the network (19).

Webgestalt was used to identify the miRNA based on DEGs. The miRNAs with the count of DEGs ≥5 and the adjust p value <0.001 were identified. The integrated regulatory network of DEGs-miRNA was constructed based on the PPI network, and miRNA-regulatory networks.

MSCs cells were washed two times with PBS and suspended with ice-cold lysis buffer supplemented with 1% PMSF (protease inhibitor) incubated on ice for 30 minutes. The total protein concentration was tested by a BCA Protein Assay Kit. Subsequently, loading buffer was added to each sample and boiled at 100℃ for 10∼15 minutes. Protein samples (30 μg) were resolved by 10% SDS-PAGE and transferred to poly vinylidene difluoride membranes at 100 V for 1 h. The membranes were blocked using 5% skimmed milk in TBST for 2 hours at room temperature and subsequently incubated overnight at 4℃ with diluted primary antibodies against ADCY2 (Santa Cruz, sc-514938) and β-actin (Cell Signaling Technology, #3700), followed by incubation with the appropriate horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology) at room temperature for 40∼60 minutes. The immunoreactive blots were visualized using an ECL-PLUS Kit according to the manufacturer’s instructions. The relative protein expression levels were normalized to that of β-actin.

cAMP levels in culture supernatants from E2 treated MSCs were quantified by an enzyme-linked immunosorbent assay (ELISA) kits (Elabscience, China) according to the manufacturer’s instructions. All assays were duplicated.

The small interfering RNA specific to ADCY2 (siADCY2) was transfected by Lipofectamine 3000 reagent (Invitrogen, USA) according to the manufacturer’s instructions. Briefly, when the MSCs reached about 50% confluence in 24-wells plate, they were transfected with siADCY2 (100 nM) and negative control siRNA (100 nM) in serum-free medium using Lipofectamine 3000. After incubation for 4 h at 37℃, 400 ul DF12 complete medium was added.

The luciferase reporter plasmid containing the 3’UTRs of ADCY2 was constructed by GENEray (GENEray, Biotechnology, Shanghai). MSCs were co-transfected with 50 nM miR-148a or miR-152 mimics or negative control and 200 ng luciferase reporter plasmid in each well. After 4 hours incubation, the cells were maintained in DF12 complete medium for 24 hours and used for luciferase reporter assay. The Dual-Luciferase Reporter Assay was performed following the kit’s instruction (Promega, Mannheim, Germany) using GloMax 96 Microplate Luminometer (Promega, Mannheim, Germany).

The experiments were performed at least in triplicate, and all data are expressed as mean±S.E.M. All data were analyzed using Graphpad prism version 5.01. Differences between two groups were analyzed by a two-tailed Student’s t-test. Multiple comparisons were assessed by ANOVA. p values <0.05 were considered significant.

First, we determined whether E2 changed the phenotype of MSCs. The expression of CD11b, CD106, CD45, CD19, CD14, HLA-DR, CD44, CD73, CD29 and CD90 was analyzed after 24 h of treatment with 100 nM E2. The results showed that these antigens on the surface of MSCs were not changed after treatment with E2 (Supplementary Fig. S1). Then the DEGs in the MSCs exposed to E2 were analyzed and 390 DEGs were identified, including 164 upregulated DEGs (e.g. ADCY2, VEGFA and PPY) and 226 downregulated DEGs (e.g. KNG1, AGT and NPY). The clustering heatmap is shown in Fig. 1A. These DEGs could be well distinguished between the MSCs treated with E2 and that without E2.

Next, the GO terms of these DEGs were enriched. As shown in Fig. 1B, these DEGs were enriched to cytokine-cytokine receptor interaction, TGF-beta signaling pathway, chemokine signaling pathway, and other pathways. Additionally, these DEGs were mainly enriched to signal transduction, phosphatidylinositol−mediated signaling after BP term analysis; plasma membrane, integral component of plasma membrane, and extracellular space after CC term analysis; growth factor activity, Ras guanyl−nucleotide exchange factor activity, and protein tyrosine kinase activity after MF term analysis. We also constructed the PPI network and module. There were 170 nodes (individual proteins) and 278 interconnected partners in the PPI network (Fig. 1C). The hub proteins with higher degrees were highly interconnected with other proteins in the PPI network, such as KNG1 (degree: 24), AGT (degree: 23), NPY (degree: 16) and ADCY2 (degree: 14) (Table 1). In the module (score=10), there were 10 nodes and 45 interconnected partners (Fig. 1D).

A total of 10 miRNAs were identified in the miRNA regulatory network (Table 2): miR-148A, miR-152, miR-148B, miR-27A, miR-27B, miR-182, miR-145, miR-153, miR-34B and miR-511. The miRNA-gene network is shown in Fig. 2.

The results of the gene expression profile analysis given above showed that the level of adenylyl cyclase type 2 (ADCY2) had changed significantly. Our in vitro experiments also showed that E2 increased the mRNA and protein levels of ADCY2 in MSCs (Fig. 3A and 3B). ADCY2 is an enzyme that catalyzes the formation of the secondary messenger cyclic adenosine monophosphate (cAMP) from ATP. We further examined the effect of E2 on the level of cAMP in MSCs. We assessed the concentration of cAMP in the culture medium of MSCs that had been treated with 100 nM E2 for 48 h. As shown in Fig. 3C, E2 promotes the production of cAMP in MSCs. To analyze whether ADCY2 mediates the promotion of E2 on cAMP, we transfected the small interfering fragment of ADCY2 (siADCY2) into MSCs before E2 treatment and then checked the level of cAMP. The results showed that the concentration of cAMP decreased significantly in siADCY2 pretreated MSCs (Fig. 3D).

Through bioinformatics analysis, we found a regulatory relationship between the differentially expressed genes and microRNAs after E2 treatment. miRNA has been widely reported to be involved in E2 regulation (20-22). Through miRNA-target gene prediction, we found miR-148a and miR-152 have binding sequences on the 3’UTR of ADCY2 (Fig. 4A). To determine whether miR-148a/miR-152 could regulate ADCY2 expression, we constructed a luciferase report plasmid including 3’UTR of ADCY2 and co-transfected it with miR-148a or miR-152 into MSCs. The results showed that both miR-148a and miR-152 inhibited the luciferase activity (Fig. 4B). Both miR-148a and miR-152 mimics were found to inhibit the protein levels of ADCY2 (Fig. 4C). On the contrary, after blocking miR-148a or miR-152 using miR-148a or miR-152 inhibitors, the expression of ADCY2 and cAMP was increased in MSCs (Fig. 4D and 4E). Furthermore, we analyzed the effects of E2 on the expression of miR-148a and miR-152 in MSCs. As shown in Fig. 4F and 4G, 100 nM E2 significantly inhibited the expression of miR-148a and miR-152. To determine whether miR-148a/miR-152 mediates the promotion of E2 on the expression of cAMP, we transfected miR-148a or miR-152 mimics into E2-treated MSCs, and then assessed the expression of cAMP. These results showed that miR-148a or miR-152 mimics reversed E2-induced upregulation of cAMP (Fig. 4H).