Adipokines Deregulate Cellular Communication via Epigenetic Repression of Gap Junction Loci in Obese Endometrial Cancer

The risk of developing endometrial cancer, the most common gynecologic malignancy in the United States, is associated with age, estrogen exposure, and obesity (1, 2). The rise of obesity is expected to lead to further increase in the incidence of endometrial cancer. Recent studies show that although endometrial cancer is associated with old age (i.e., >60 years), this profile characteristic is changing with increasing incidence in obese women at a younger age (3). Obesity-associated endometrial tumors fall into the endometrioid endometrial cancer Type I, which is the most common histologic type of endometrial cancer (4). Type I lesions are characterized by well and moderately differentiated endometrioid histology, early stage, and favorable prognosis. Obesity is also a risk factor for other malignancies as well, including breast and colon cancers (5, 6). Adipose stromal cells (ASC), which are adipocyte progenitors within fatty tissue, have been implicated in the development of these cancers (7, 8). Although released factors from abdominal fat can be transported systemically, identification of circulating ASCs in obese subjects and recruitment of these cells by tumor-produced chemokines suggest that ASCs are trafficked to target tissue sites and embedded within the tumor microenvironment (9–12).

The tumor microenvironment is composed of complex cell types, including cancer-associated fibroblasts and leukocytes, which contribute to malignant development and progression (13, 14). More recently, ASCs and adipocytes in the tumor microenvironment have been implicated as promoters of tumor progression (15, 16). These cells produce hormones and cytokines that stimulate the growth of many tumor types, including endometrial cancer (7, 17). Cancer-associated adipocytes may also contribute to sustenance of cancer cells by providing an energy source through lipolysis, supporting their growth (18). The interaction between the stromal compartment and cancer cells plays a role in regulating gene transcription during tumorigenesis (19). Persistent stimulation by tumor microenvironmental factors has long-term effect on gene repression through epigenetic mechanisms such as promoter CpG island hypermethylation (20, 21).

Epigenetic repression frequently takes place in tumor-suppressor genes, including those that mediate or regulate the gap-junction intercellular communication (GJIC; refs, 22, 23). The gap junction (GJ) gene family, which encodes connexin (Cx) proteins that form the gap junction channels, and their structural and kinase modulators (herein collectively referred to as GJ-associated loci) control the transfer of metabolites and ions between adjacent cells (24, 25). Gap junctions are also upregulated during normal invasive processes (e.g., oocyte implantation), as well as extravasation and distant metastatic lesion formation (26, 27). This dichotomy in gap junction function reflects the complex cellular interactions occurring during tumor development and later in aggressive disease establishing metastatic growth.

In this study, we first examined the role of ASCs in repressing GJ-associated gene expression in immortalized endometrial cells. We then determined whether this repression can be recapitulated in a mouse model of obese endometrial cancer, and whether silencing is marked by DNA methylation in primary endometrial tumors of obese patients. Potential epigenetic regulation of cell–cell communication and interactions was first examined generally by DNA demethylation treatment of cancer cells. Then, cellular growth and motility effects by specific knockdown or reactivation of the major connexin Cx43 (encoded by methylation-target GJA1) in endometrial immortalized epithelial and cancer cells were examined, respectively. Our studies suggest that epigenetic repression of GJ-associated loci is involved in stunting GJIC and development of obesity-associated endometrial cancer.

Detailed description of the experimental procedures and reagents used in this study can also be found in Supplementary Experimental Procedures.

ASCs were obtained from the Coriell Institute (Camden, NJ). The cells were isolated from fat tissue during subject's abdomen and waist ante-mortem elective cosmetic tumescent liposuction. ASCs were cultured in nondifferentiating media (DMEM supplemented with 0.5% FBS/0.2% BSA) and grown on fibronectin-coated culture dishes. For coculture, ASCs were seeded on a fibronectin-layered insert of a Boyden chamber. In the bottom well, endometrial cells were seeded for the coculture experiment. Control wells contained the fibronectin-layered insert without ASCs. Cocultures were carried out in duplicates for 3 weeks, followed by lysis of endometrial cells for RNA isolation. Ishikawa and HEC-1A cells were obtained from Millipore Sigma and ATCC, respectively. Ishikawa was derived from a well-differentiated human endometrial adenocarcinoma, whereas HEC-1A was derived from a stage IA moderately differentiated adenocarcinoma (28, 29). Both these cancer cell lines represent Type I endometrial cancer (28–30). Immortalized EM-E6/E7/TERT-1 cells, established and characterized by Kyo and colleagues, are nontransformed, nontumorigenic endometrial epithelial cell lines (31). Cell line authentication was carried out by short tandem repeat analysis at the ATCC. Cells were used within 3 passages after thawing from frozen stock. Cells were routinely checked for Mycoplasma while in use by the Mycoplasma PCR Detection Kit (ABM).

RNA was subjected to cDNA conversion and sequencing library construction, and next-generation sequencing was carried out using the HiSeq2000 platform (Illumina). Additional sequencing-run information is reported in Supplementary Fig. S1.

The studies were approved by the Institutional Animal Care and Use Committee at the Medical College of Wisconsin (AUA2927). Breeding pairs (B6.129-Ptentm1Rps, 01XH3) were obtained from Mouse Models of Human Cancers Consortium at NCI (Frederick, MD). Mice were housed at the Medical College of Wisconsin Biomedical Resource Center facility at an appropriate temperature and a standard light-dark cycle. The mouse colony was established by breeding wild-type female mice with heterozygous Pten^+/− male mice. Pten^+/− female offspring were weaned at age 21 days. Ear or tail biopsies were obtained for genotyping using the primers specified by the website of Mouse Models of Human Cancers Consortium at the NCI. Female Pten^+/− and wild-type mice at 3 to 4 weeks old were fed diets either with 16.4% energy from fat (standard AIN-93G diet) or with 63.7% fat to induce obesity (n = 9, each group), prepared by TestDiet. The mice were kept on these diets until they were sacrificed at 12, 20, or 28 weeks of age. All mice were weighed once per week to determine changes in body weight development from weaning to euthanization. One uterine horn was fresh frozen for molecular analyses, the other formalin fixed overnight. Changes in endometrial morphology were assessed from histopathologic sections processed in paraffin blocks, following the guidelines set by Fyles and colleagues and examined by a board-certified pathologist blinded to treatment group and genotype (32).

GJIC was measured using intercellular transfer of the fluorescent dye calcein, which is permeable via gap junctions. Assays were performed in culture media supplemented with charcoal-stripped FBS (10%). Donor cells were incubated with the dye for 20 minutes at room temperature. Inside the cell, Calcein-AM is cleaved by nonspecific esterases into calcein, making it impermeable to diffusion through the cell membrane. Recipient cells are grown to confluence. Calcein-labeled donor cells were then dropped (“parachuted”) onto the recipient cell layer, and calcein transfer between donor and recipient cells was observed with fluorescent microscopic imaging. Data were expressed as # of fluorescent recipient cells/# of donor cells for each condition. Experimental conditions were performed in triplicates. Fluorescent images of 10 to 15 fields per well were captured on an Operetta-automated microscope (Perkin Elmer). A program (Perkin Elmer) allowed identification of all cells on the plate (from phase contrast image), as well as original donors (100 ± 50 per well), and dye-filled recipients due to calcein transfer over time.

Nanomechanical properties of live adherent HEC-1A cells were analyzed in vitro with a Nanoscope Catalyst (Bruker) atomic force microscope mounted on a Nikon Ti inverted epifluorescent microscope. PeakForce Quantitative Nanomechanical Mapping (PF-QNM) mode of in-liquid imaging with SCANASYST-AIR probes (Bruker) was applied. Square fields from 30 × 30 μm to 50 × 50 μm corresponding to the electronic resolution from 0.19 to 0.47 μm/pixel were scanned. To calculate nanomechanical parameters of the cells, we analyzed the retrace images with Nanoscope Analysis software v.1.7 (Bruker). The Young's modulus was calculated as a measure of cell elasticity. A force needed to separate a tip from a cell was determined to assess cell adhesion to a tip. The Sneddon model of tip–object interactions was applied in the calculations. Elasticity and adhesion mode values were calculated for each cell from corresponding distribution histograms of image pixel values and used in all the downstream statistical evaluations. Quantification of adhesion parameters governing HEC-1A cell–cell interactions was performed with the single-cell force spectroscopy atomic force microscopy (AFM) as described (33) and detailed further in Supplementary Information.

All human tissue samples have been obtained under International Review Board approval. The studies were conducted with obtained written-informed consent from the patients, and the studies were conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule). Pyrosequencing of paraffin-embedded tissue of endometrial tumors (n = 141), adjacent normal tissue (n = 43), and uninvolved normal uterine control (n = 11) was performed as described previously (34). Clinical characteristics of this cohort are detailed under Supplementary Information. Briefly, tumor sites were macrodissected from paraffin-embedded tumor samples, and DNA was isolated. DNA (500 ng per sample) was processed for bisulfate conversion using the EZ DNA Methylation kit (Zymo Research). Primers used to amplify specific CpG island regions are listed in Supplementary Information. Methylated CpG sites were detected by the PyroMark Q96 MD system. The methylation percentage of each interrogated CpG site was calculated and visualized using MultiExperiment Viewer v4.8 (Dana-Farber Cancer Institute, Boston, MA) and GraphPad Prism (GraphPad Software).

The Shapiro–Wilk normality test was performed to assess parametric distribution of data. For statistical analysis, t test was used after normality of data was ensured. In cases where normality test failed, the Mann–Whitney rank sum test was used as a nonparametric analysis method. P values <0.05 were considered statistically significant.

As an in vitro model of obesity for endometrial cancer, immortalized nontumorigenic EM-E6/E7/hTERT-1 cells (referred to hereinafter as EME6/7t; ref. 31) were exposed to adiposity-derived ASCs in a Boyden chamber for 21 days. Control cells were incubated in the Boyden chamber without ASCs. RNA-seq was used to determine the coculture effects on the transcriptomes of ASC-exposed and control EME6/7t cells (Fig. 1A; Supplementary Fig. S1A). Robust differential expression was observed in EME6/7t cells exposed to ASCs compared with control, and high correlation in expression was observed between the biological replicates of each sample (Supplementary Fig. S1B and S1C). We identified 3,811 differentially expressed genes between exposed and control cells, of which 2,182 were induced and 1,629 were repressed due to ASC exposure (P < 0.05; Supplementary Fig. S1D and Supplementary Table S1). On the other hand, ASC-exposed Ishikawa cancer cell line only showed 246 genes were differentially expressed in Ishikawa cells, of which 113 genes overlapped with EME6/7t (Supplementary Fig. S1D and S1E). The data indicate that EME6/7t cells may be more susceptible to ASC-mediated effects than the transformed Ishikawa cancer cells. Differentially expressed genes in EME6/7t cells (identified by a threshold of 1.9-fold difference in expression) encoding proteins involved in intercellular junctions or invasion/migration pathways (Fig. 1B and Supplementary Table S2). A number of genes suppressed in the ASC-exposed cells belonging to these pathways were significantly enriched by pathway analysis and formed a protein–protein network that is involved in mediating cellular interactions, including gap junction communication (P < 0.001; Fig. 1C and D, respectively). These included GJ genes, encoding Cx, and GJ-associated loci, which encode structural scaffolding (e.g., tight-junction proteins) and kinase modulators of gap junctions (see the model in Fig. 1E). For validation, RT-PCR of GJs and gap junction regulators was performed using the EME6/7t cells cocultured with and without ASCs. Among these genes tested, GJA1 (which encodes Cx43) showed the most prominent expression in EME6/7t cells, and its expression was reduced in ASC-exposed cells (Fig. 1F). In addition, the expression of kinases involved in gap junction modulation (i.e., MAPK1, MAPK3, PRKACA, and PRKCA) and the tight-junction protein TJP1 were also suppressed in EME6/7t cells exposed to ASCs (Fig. 1F).

To identify the ASC paracrine factors acting on EME6/7t cells, we performed multiplexing assays for a panel of cytokines and adipokines. The major factors detected in conditioned media of ASCs were plasminogen activator inhibitor 1 (PAI-1), adipsin, MCP-1, IL6, and HGF (Fig. 2A). We chose PAI-1, the most abundant secreted factor determined in our assay, to evaluate its effect on the expression of GJ-associated genes in EME6/7t cells (Fig. 2A). These cells were treated with and without ASC conditioned media and with the ASC conditioned media in the presence of PAI-1 inhibitor tiplaxtinin for 24 hours and then subjected to single-cell expression analysis by microfluidic PCR. Expression profiling revealed 20 genes that were differentially expressed due to conditioned media and PAI-1 inhibition (Fig. 2B). t-distributed stochastic neighbor embedding (t-SNE) nonlinear dimensionality reduction analysis was used to cluster cells based on their expression profiles (Fig. 2C). Six subpopulations (a–f) were identified with unique expression profiles of these GJ-associated genes in treated and control cells (Fig. 2D). Subpopulations a–c showed higher coordinated expression levels of the abundant GJA1 and other GJ-associated genes, including GJB2, PRKACA, and CDH2, whereas subpopulations d–f had lower expression levels of GJA1 and nine other genes GJA9, GJB1, GJB3, GJB5, GJB7, GJC2, TJP2, MAPK1, and PRKCB (Fig. 2D). The cell numbers of “GJA1-high” subpopulations a–c representing the majority (76%) of control cells were drastically reduced to 18% in cells exposed to conditioned media (Fig. 2E). In general, the cell composition shifted to “GJA1-low” subpopulations d–f after the exposure (Fig. 2E). This change of population dynamics was blocked by the addition of tiplaxtinin, partially restoring the two “GJA1-high” subpopulations a and b (Fig. 2E). This result suggests that the adipokine factor PAI-1 plays a role in suppressing the expression of GJA1 and other related loci in the majority of EME6/7t cell population. These single-cell data also corroborate the finding of bulk RNA-seq that obesity-derived ASCs have paracrine influences on suppressing expression of GJ-associated genes in endometrial epithelial cells.

Because the repression of candidate GJ-associated genes was observed in this transient exposure system in vitro, we sought to solidify the role of obesity in transcription repression of GJ-associated loci in an in vivo model. Obesity was induced in an endometrial cancer mouse model with a Pten^−/+ deletion fed with high-fat diet (32). Pten^−/+ (n = 9) and wild-type (n = 9) mice fed with this diet exhibited an increase in body weight and associated phenotype such as serum estradiol and leptin levels (Supplementary Fig. S2A–S2E). Mutant mice fed high-fat diet developed earlier endometrial atypia (as early as 12 weeks) and hyperplasia (20 weeks) than the control-fed mice (Fig. 3A and B). By 28 weeks, endometrial carcinomas were observed in both control- and high fat-fed Pten^−/+ mice (Fig. 3A). Furthermore, endometrium of mice fed with high-fat diet exhibited higher Ki67 proliferation levels than control-fed mice (Fig. 3C and D). qRT-PCR analysis of GJ-associated genes showed a significant decrease in PRKCB (P < 0.01), Prkca (P < 0.05), Tjp1 (P < 0.01), Tjp2 (P < 0.05), Gja1 (P < 0.01), and Gjc2 (P < 0.01) in lesions of the high-fat Pten^−/+ group (n = 9) compared with the wild-type group (n = 9), but this difference was not observed in the control-fed control group (Fig. 3E). Interestingly, Cdh1 was the only gene we studied that showed a significant decrease in Pten^−/+ mice independent of the diet they were on. Collectively, these animal studies provide crucial evidence toward establishing a link between obesity and the silencing of key GJ-associated genes in endometrial cancer.

We then assessed profiles of promoter CpG island methylation in the GJ-associated loci in our endometrial methylome dataset from endometrioid endometrial carcinoma tumors obtained previously by MBDCap-seq analysis (n = 67 primary tumors and 10 controls; Supplementary Fig. S3A; ref. 35). Methylation in CpG island cores, as well as in CpG island shores (defined as 2 Kb upstream and downstream of the core island), has been associated with gene silencing (36). Although hypermethylation was frequently observed in CpG island cores (e.g., PRKCB, GJC2, and TJP2), shore methylation also occurred in their flanking regions (e.g., left shores of PRKCA and MAPK3). Promoter CpG islands of PRKACA, TJP1, and CDH1 showed a slight increase in DNA methylation in some tumors compared with controls (Supplementary Fig. S3A). An increase in methylation was observed in non-CpG island (DNA with a G + C content <55%) promoters of GJA1, GJB4, GJB5, and GJC2 (Supplementary Fig. S3A). Non-CpG island promoter methylation had also been implicated in gene silencing (37). Analysis of The Cancer Genome Atlas endometrial cancer cohort confirmed that DNA hypermethylation of the six aforementioned promoter CpG islands (PRKCA, PRKCB, MAPK3, TJP1, TJP2, and CDH1) and three non-CpG island loci (GJA1, GJB5, and GJC2) was associated with reduced expression (Supplementary Fig. S3B and S3C).

Pyrosequencing analysis was then conducted to verify promoter methylation states in the major connexin-encoding gene GJA1 and three GJ-associated loci (PRKCA, PRKCB, and TJP2) in primary endometrioid endometrial carcinoma (type I) tumors from a cohort of patients with endometrial cancer (n = 141; clinical characteristics in Supplementary Table S3). These loci exhibited high levels of DNA methylation in their promoter CpG islands in primary endometrial tumors relative to the control endometrium group (Fig. 4A; see also Supplementary Fig. S3A). Promoter hypermethylation was observed in the primary endometrial tumors compared with normal tissue adjacent to tumors and uninvolved noncancerous controls (Fig. 4A and B). GJA1 promoter methylation was significantly higher in tumor compared with normal tissue (Fig. 4B; P < 0.05). This hypermethylation event was also observed for PRKCA, PRKCB, and TJP2 (Fig. 4B; P < 0.001, P < 0.01, and P < 0.001, respectively), although in the former this was only on the right shore. After stratifying the cohort based on three groups—morbidly obese [body mass index (BMI) > 35], obese (BMI 30–35), and nonobese (BMI < 30), we found higher methylation levels of the PRKCA promoter in the obese group compared with those of the nonobese group (P < 0.01; Fig. 4C). GJA1, PRKCB, and TJP2 had no significant differences between the groups (Fig. 4C). To determine whether other factors may influence the methylation profile in the obesity groups, we further selected the patients that presented with comorbidity (e.g., diabetes) from each of the nonobese, obese, and morbidly obese groups. In this selected cohort, we found higher promoter methylation levels in the morbidly obese group compared with nonobese patients for GJA1, PRKCA, and TJP2 (P < 0.01, P < 0.05, and P < 0.01, respectively; Fig. 4D). For GJA1 and PRKCA, this increase in methylation level was also evident in the obese group (P < 0.01 and P < 0.001, respectively; Fig. 4D, top), whereas PRKCB showed no significant differences in any of the groups, even in this selected subset. These changes were starker between the nonobesity and the obesity groups when GJA1, PRKCA, and TJP2 methylation profiles were compared in various combinations (P < 0.001; Fig. 4D, bottom). Taken together, these epigenetically repressed GJ-associated loci can be useful biomarkers to study pathophysiology of obesity-associated endometrial cancer.

To examine whether epigenetic actions are involved in deregulating gap junction activities, we initially determined the effect of demethylation on gene re-expression by treating endometrial cancer cell lines (HEC-1A and Ishikawa) and control EME6/7t cells with 2.5 and 5 μmol/L 5-aza-2′-deoxycytidine (DAC), a general DNA demethylator (see an example in Supplementary Fig. S4A; ref. 38). Reactivation of GJB1, GJB3, GJB4, GJB5, GJB7, and GJC2 was observed mainly in HEC-1A cells (Supplementary Fig. S4B). Partial re-expression of kinase and structural modulators (PRKACA, PRKCB, TJP2, PRKACB, MAPK3, CDH1, and TJP1) with promoter CpG islands was also observed in these cells (Supplementary Fig. S4B). Re-expression of those loci without promoter CpG islands may be attributed to an effects of DAC on chromatin openness of inactive promoters (38) or secondary effects due to demethylation of regulatory genes. Consistent with this, the re-expression of some of these genes was affected to a lesser degree in Ishikawa cells (Supplementary Fig. S4B). This demethylating treatment had no obvious effect of gene reactivation in control EME6/7t cells.

To determine phenotypic effects of reactivated GJ-associated genes, we tested the homotypic GJIC of DAC-treated and -untreated cells using the “parachuting” assay. Donor cells labeled with the gap junction–permeable dye calcein were dropped (parachuted) on a confluent layer of recipient cells (Fig. 5A). GJIC was measured through the detection of calcein transfer from donor to acceptor cells (i.e., ratio of acceptor/donor cells). Without DAC treatment, the level of GJIC in EME6/7t cells was 3-fold higher than the endometrial cancer cells, consistent with the suppression of gap junction coupling in cancer (Fig. 5B). DAC treatment led to a significantly robust increase in GJIC in HEC-1A (P < 0.001), and to a lesser, yet significant (P < 0.01), increase in Ishikawa cells (Fig. 5B and C). We also determined whether DAC had an effect on the growth and motility in these cancer cells. Reduction in motility and growth was observed in both DAC-treated HEC-1A and Ishikawa cells (P < 0.001 and P < 0.01, respectively; Supplementary Fig. S4C).

We then examined cellular nanomechanical characteristics that may contribute to the restoration of GJIC and decreased cellular motility in DAC-treated HEC-1A cells using the Peak Force Quantitative Nanomechanical Mode of AFM. This technique allows a direct assessment of several mechanical parameters of single cells by probing them with very low forces at a high spatial resolution (39). Here, we focused on cell elasticity, expressed as the Young's modulus (39), and cell adhesion (to the AFM tip), since these parameters may reflect the strength of cell–cell interactions. DAC treatment had a little effect on cell shape or elasticity, but a significant increase in cell adhesiveness was detected (P < 0.001, Fig. 5D and E).

Although initially cellular adhesiveness was assayed as a function of nonspecific cell stickiness to the AFM probe made with silicon nitride, we also examined whether DAC treatment affected cell-to-cell adhesiveness. To quantify adhesion between cells, we applied single-cell force spectroscopy (33). In this AFM-based method, an AFM probe is constructed from a tipless cantilever and a single “tester” cell attached to it (Supplementary Fig. S5A). The tester cell is then brought to a close vicinity of a cell growing in a culture dish. After establishing contact, the tester cell is pressed on the target cell until the applied force reaches its predefined maximum value, and cellular interactions are measured by the force (Newton) and work (Joule) needed to separate the cells, as described in Supplementary Methods (Supplementary Fig. S5B and S5C). Here, a single tester HEC-1A cell attached to an AFM tipless probe was used to test the interactions with target HEC-1A cells growing on a cell culture dish (Supplementary Fig. S5A). Representative data of detachment curves showing increased cell–cell adhesiveness in DAC-treated HEC-1A cells (solid red trace) compared with control cells (solid black trace) are shown in Supplementary Fig. S5D. It is of note that the detachment of control cells passed through a sharp maximum, whereas the treated cells clutched together through a flat maximum. Importantly, tester cells interacted only weakly (dotted red and black traces) with the bottom of the dish allowing for a straightforward distinction between successful and false cell–cell adhesion (Supplementary Fig. S5D).

The maximal force required to separate the tester cell from target cell measured in nano-newtons (nN) was twice as large in HEC-1A cells treated with DAC (5 μmol/L) than control cells (0.91 vs. 0.42 nN, P < 0.001, Fig. 5F). In addition, the work required to separate the cells (defined as the force applied at a specific distance) in femto-joules (fJ) was about 3-fold higher in the DAC-treated cells compared with untreated cells (3.2 vs. 1.2 fJ, P < 0.001, Fig. 5G). Interestingly, the detachment distance during which separation takes place was approximately 30% longer for the treated cells, possibly indicating a larger and more complex contact area between the interacting cells (Fig. 5H). The observed differences in the mechanical parameters strongly point to the elevated cell–cell adhesiveness and the extensiveness of cell–cell contact in DAC-treated cells relative to untreated cells.

To determine whether DAC treatment results in enhancing gap junction formation, we examined the localization of Cx43 within the cell, and the distribution of gap junction plaques at cell interfaces. More punctate Cx43 staining, indicative of gap junction formation, was observed in untreated EME6/7t cells compared with the two cancer cell lines, Ishikawa and HEC-1A. A striking redistribution of Cx43 to form gap junction plaques was observed in HEC-1A cells (Fig. 6A). Although Ishikawa expressed more Cx43 protein than HEC-1A (Fig. 6A; Supplementary Fig. S6A), and some redistribution of Cx43 to the cell surface was observed, the formation of gap junction plaques was not as organized along the cell surface as in the DAC-treated HEC-1A cells (Fig. 6A). The coincident induction of GJIC and cellular adhesiveness, restoration of Cx43 gap junction plaque formation, and decreased motility in HEC-1A cells due to DAC treatment suggest epigenetic regulation of gap junction activity. Furthermore, ASC exposure led to significantly (P < 0.01) inducing GJA1 promoter methylation in EME6/7t cells, confirming that this gene encoding Cx43 is subject to adiposity-mediated epigenetic suppression (Supplementary Fig. S6B). In this in vitro ASC exposure system, induction in PRKCA methylation was also observed but not TJP2 (Supplementary Fig. S6C and S6D). We then specifically examined whether restoration of Cx43, which is repressed in HEC-1A cells (Fig. 6A; Supplementary Fig. S6A), had an effect on cellular migration. Cx43 expression was exogenously induced by transfecting an expression vector in HEC-1A cells and selected several clones overexpressing Cx43 (Fig. 6B as an example). Two HEC-1A clones overexpressing Cx43 exhibited reduced migration compared with parental HEC-1A cells (P < 0.001; Fig. 6C), with little effect on cellular proliferation (Fig. 6C). In complementary experiments, we examined whether Cx43 knockdown by specific siRNAs affected cellular motility and growth of EME6/7t endometrial epithelial cells, which expressed Cx43 (Supplementary Fig. S6A). Cx43 knockdown, as shown in Fig. 6D, resulted in a significant increase in migration (P < 0.001) but had little effect on growth of these cells (Fig. 6E). Furthermore, treatment of EME6/7t cells with Gap27, a peptide mimicking the extracellular domain of Cx43 that has been shown to inhibit coupling (40), also diminished cellular motility in a dose-response fashion (P < 0.001, Fig. 6F), but had a limited effect on cellular growth (Fig. 6F). Combined, the data suggest that endometrial cellular coupling by Cx43 gap junctions diminishes endometrial cancer cell motility, concomitant with enhanced cell–cell adhesion. Conversely, suppression of this Cx43 coupling serves to promote motility of these endometrial cancer cells.

The present study demonstrates a strong link between obesity and epigenetic silencing of GJ-associated loci in endometrial cancer. Although the development of obesity-mediated tumors is multifaceted, overgrowth of abdominal adipose tissue is highly associated with endometrial cancer (3, 41). Abundant adiposity frequently undergoes tissue remodeling and releases ASCs into the circulation (11). These stromal cells can then home to endometrial tissue and contribute to their malignant growth via local secretion of tumor-promoting hormones, adipokines, and cytokines (7, 9, 16, 17, 42). Our studies show that ASCs may act in part by PAI-1 to suppress GJ-associated genes in the immortalized EEC line EME6/7t, suggesting that cellular communication in EECs may be epigenetically targeted promoting type I endometrial cancer observed in obesity. Although PAI-1 is known as a serine protease inhibitor, other functions of PAI-1 include regulation of gene expression via Stat-1 and enhancing motility in cancer cells (43, 44). PAI-1 is elevated in visceral fat and may contribute to morbidity associated with obesity (45).

Although transcriptional repression of candidate GJ-associated loci observed in vitro and in vivo may be transient, we further demonstrated that epigenetic mechanisms reinforce long-term repression in endometrial cancer development. Tumor DNA hypermethylation of these loci was observed in our endometrial cancer cohort, and ASC exposure induced GJA1 promoter methylation in vitro. Thus, although initially gene suppression may be reversible, de novo DNA methylation of the GJ-associated loci can spread to promoter CpG island shores or cores for permanent silencing.

The obesity-mediated epigenetic repression of GJ-associated loci has a profound effect on gap junction homeostasis. GJIC is regulated through multiple layers, which has been mainly learned from studies of the major connexin Cx43. Some kinase modulators (e.g., PKA and PKC) can affect connexin trafficking, whereas others (e.g., ERK and SRC) serve to directly modulate channel function. Scaffolding proteins (e.g., ZO1 and ZO2) can regulate the assembly of Cx43 gap junction plaques at the cell surface (46, 47) by enhancing cell–cell adhesion as shown by the AFM studies (48, 49). Disruption of the major connexin Cx43 gap junction activity plays a role in endometrial cancer cell migration potentially associated with tumor promotion. Reestablishment of gap junction activity, by re-expression of connexins and/or gap junction modulators, may be required for metastatic invasion during disease progression (50). This functional redundancy among members of gap junction proteins can be exploited by aggressive cancer cells to gain metastatic potential (50). Comprehensive analysis of these gap junction genes and their modulators is needed to provide a better understanding of their roles in endometrial cancer development and advanced disease.

The epigenetically repressed GJ-associated loci can be useful biomarkers to study pathophysiology of obesity-associated endometrial tumorigenesis, which is on the rise (3, 41). There is an epidemic of obesity-related endometrial cancers in Hispanics in our South Texas population, which represents a health care disparity. These patients with endometrial cancer tend to be considerably younger (∼50 years) than the typical patient with endometrial cancer (∼60 years; refs. 3, 4). The association of endometrial cancer in obese women with aberrant DNA methylation of GJA1, a known tumor suppressor, and at least two GJ-associated loci PRKCA and TJP2 provides mechanistic clues to these clinical observations. Similarly, endometrial lesions (atypia and hyperplasia) were observed at a younger age in the obese high-fat–fed Pten mice compared with control-fat diet mice, whereas by 28 weeks of age, both groups developed endometrial carcinomas, suggesting the contribution of obesity to early stages of tumor development, including initiation or promotion.

Our present study indicates that obesity, in part due to ASC paracrine actions, leads to downregulation of GJ-associated loci. We further confirmed the role of this epigenetic mechanism in the regulation of gap junction activity is in part mediated by effects on Cx43 assembly into functional gap junction plaques between cells. Future studies will examine the regulatory mechanisms leading to GJ-associated gene regulation and the functional changes in GJIC in obesity-associated endometrial cancer. These studies will improve our understanding of how obesity contributes to the development of endometrial cancer.

No potential conflicts of interest were disclosed.

Conception and design: S.R. Polusani, G. Huang, C.-W. Chen, E.R. Kost, D.G. Mutch, B.J. Nicholson, T.H.-M. Huang, N.B. Kirma

Development of methodology: S.R. Polusani, Y.-W. Huang, G. Huang, C.-M. Wang, P. Osmulski, N.D. Lucio, E.R. Kost, E.Y. Shim, M.E. Gaczynska, T.H.-M. Huang, N.B. Kirma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.R. Polusani, Y.-W. Huang, G. Huang, C.-W. Chen, C.-M. Wang, L.-L. Lin, P. Osmulski, N.D. Lucio, I. Aguilera-Barrantes, P.T. Valente, E.R. Kost, S.E. Lee, P. Yan, P.J. Goodfellow, D.G. Mutch, T.H.-M. Huang, N.B. Kirma

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.R. Polusani, G. Huang, C.-W. Chen, L.-L. Lin, P. Osmulski, L. Liu, Y. Zhou, C.-L. Lin, E.R. Kost, J. Ruan, M.E. Gaczynska, V.X. Jin, B.J. Nicholson, N.B. Kirma

Writing, review, and/or revision of the manuscript: S.R. Polusani, G. Huang, P. Osmulski, E.R. Kost, C.-L. Chen, P.J. Goodfellow, D.G. Mutch, B.J. Nicholson, T.H.-M. Huang, N.B. Kirma

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.R. Polusani, Y.-W. Huang, G. Huang, C.-W. Chen, Y.-T. Hsu, Y. Zhou, P. Yan, D.G. Mutch, B.J. Nicholson, N.B. Kirma

Study supervision: D.G. Mutch, B.J. Nicholson, T.H.-M. Huang, N.B. Kirma

The authors acknowledge the assistance of the Next-Generation Sequencing Core for RNA-seq analysis, the Bioanalytics and Single-Cell Core for atomic force microscopy studies and single-cell expression, and the High Throughput Screening Facility of the Center for Innovative Drug Discovery (partially supported by funds from the National Center for Advancing Translational Sciences, NIH, through Grant UL1 TR001120) for the analyses of gap junction cell coupling studies, at the University of Texas Health San Antonio, San Antonio, TX. This work was supported by NIH grants R01CA172279 and P30CA054174, the Cancer Prevention and Research Institute of Texas (CPRIT) grant RP150600, San Antonio Cancer Council, and the Max and Minnie Tomerlin Voelcker Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.