Cellular junction and mesenchymal factors delineate an endometriosis-specific response of endometrial stromal cells to the mesothelium

To examine whether PMC exposure affects the expression profiles of ESCs, we performed co-culture studies mixing ESCs from normal and endometriosis subjects with the LP9 PMC cell line ( Fig. 1A ) for 24 h. Three sets of normal/endometriosis ESC samples matched for the menstrual phase in which the cells were harvested (167 vs. 163; 171 vs. 169; and 172 vs 170) were used. The cells were subjected to CyTOF analysis using a 28-factor antibody panel ( Table S2 ), gating on ESCs with the CD10 + KRT8/18 − profile to eliminate LP9 PMCs from analysis ( Fig. 1A ). To examine the overall differential expression profiles between PMC-exposed normal vs. endometriosis ESCs, principal component analysis was performed, based on the expression profiles in the presence or absence of LP9 PMCs ( Fig. 1B ). Exposure to LP9 PMCs caused a distinct shift in 2D principal component coordinates of ESCs from endometriosis patients compared to controls (163 vs. 167; 169 vs. 171; and 170 vs 172) ( Fig. 1B ). In contrast, endometriosis ESCs in solo culture clustered closely with ESCs from control subjects, whether in solo or heterologous culture with PMCs ( Fig. 1B ). To further illustrate the divergence of the expression profile of endometriosis ESCs co-cultured with LP9 cells, pseudotime analysis was performed for the combined normal and endometriosis ESCs (with and without LP9 co-culture) ( Fig. 1C ). In this analysis, which compares samples in a trajectory of expression profiles, there is clear overlap of ESC solo cultures, whether control (red) or endometriosis (green) derived, and control ESCs cultured with PMCs (blue). While there is some overlap in the initial part of the trajectory (upper right quadrant), the endometriosis derived ESCs cultured with PMCs (yellow) are the only ones represented in the bottom left quadrant in the trajectories ( Fig. 1C ). T-SNE was used to delineate ESC cell clusters with specific expression profiles of the 28 proteins probed ( Supplemental Fig. 1 ). While no consistent differences in expression of individual adhesion and mesenchymal factors between the normal (n = 3) and endometriosis samples (n = 3) were noted, t-SNE analysis did demarcate the normal from the endometriosis ESC cell clusters, suggesting intrinsic differences in the expression signatures of these factors between the two groups ( Supplemental Fig. 1 ). To illustrate the LP9 PMC co-culture effects on the ESC protein panel expression, we aligned the cell clusters based on expression profiles using t-SNE analysis form each normal-endometriosis set (with and without LP9 PMC co-culture) for the full 28 protein panel, shown as both collective heat maps ( Supplemental Fig. 2 panels A – C ) and for individual proteins as violin plots for each set ( Supplemental Figs. 3 – 5 , respectively). Based on this initial analysis, we identified a subset of proteins that were distinctly changed in the endometriosis ESC-LP9 PMC co-cultures compared to the other ESC cultures. These were ZO2, Cx43, SNAI1, ZO1, ZEB2, CDH2, and CAV2 for 163 vs. 167; FN1, EGFR, SNAIL1, CAV2, TGFβ, and ZEB2 for 169 vs 171; and TWIST1, CAV2, TGFBR2, SNAI1 and NOTCH1 for 170 vs. 172 ( Supplemental Figs. 3 – 5 ). Illustrating this differential expression, we used t-SNE based clustering to distinguish cell clusters with unique expression signatures ( Figs. 2 – 4 , panels A ). Heatmaps were used to show the levels of expression of each of these factors, and the corresponding size of specific cell clusters with a specific expression profile of these factors ( Figs. 2 – 4 , panels B ). For the 163 (endometriosis) vs. 167 (Control) set,163 showed an overall lower expression of five proteins ZO2, Cx43, SNAI1, ZO1, ZEB1 in ESCs co-cultured with LP9 PMCs compared to the solo ESC culture ( Fig. 2B ). This is evident in the major cell clusters (e.g., clusters #s 10, 13, 14 and 15, indicated by the size of the circles in the balloon plots) from co-culture compared to the clusters that dominate without co-culture (e.g., clusters # 17, 18 and 22) ( Fig. 2B lower panel). On the other hand, the expression of the major cell clusters in the normal ESC sample 167 with and without PMC co-culture (cluster #s 6 and 9; and cluster #s 1, 2, 4 and 5, respectively) showed similar expression for ZO2, Cx43, SNAI1, ZO1, ZEB1, indicating little effect due to LP9 PMC exposure. For the 169 (endometriosis) vs. 171 (control) set, the major cell clusters (e.g., clusters #s 8, 9 and 10) of sample 169 exposed to LP9 PMCs exhibited low expression of TWIST1, CAV2, TGFBR2, and SNAI1 in the ESCs compared to the major clusters (e.g., clusters # 3 and 5) for solo ESC culture ( Fig. 3B ). Again, the control sample 171, while it did show some changes in dominant clusters in the presence and absence of PMCs, they did not show low expression of these genes ( Fig. 3B ). For the 170 (endometriosis) vs 172 (normal) set, major clusters #s 2, 3, 4 and 5 of the 170 ESCs exposed to PMCs showed low expression for FN1, EGFR, SNAI1, CAV2 and TGFB1 compared to solo 170 major culture cluster #s 6 and 7 ( Fig. 4B ). On the other hand, the 172 ESCs exhibited similar cell clustering patterns for the co-culture and solo cultures ( Fig. 4B ). Violin plots illustrate that in the endometriosis sample (163), ZO1, ZO2, SNAI1, ZEB2, and to a lesser extent Cx43, were decreased, while CDH2 and CAV2 showed an induction in expression in response to LP9 PMC co-culture. These effects were muted or not present in the normal sample 167 ( Fig. 5A ). For 169 ESCs endometriosis sample, SNAI1, TWIST1, TGFBR2, and CAV2 showed lower expression, and NOTCH1 higher expression, due to the LP9 PMC co-culture in the endometriosis sample ( Fig. 5B ). Again, the changes induced by LP9 PMC exposure were less evident in ESCs from the normal sample 171 ( Fig. 5B ). Expression of SNAI1, CAV2, EGFR, FN1 and TGFB1 were reduced in endometriosis 170 ESCs exposed to LP9 PMCs relative to solo ESCs ( Fig. 5C ). ZEB2 was induced in the co-cultures 170 ESCs compared to solo culture ( Fig. 5C ). These differences due to co-culture with LP9 PMCs were not observed in the control 172 ESCs ( Fig. 5C ). Consistently among the 3 endometriosis/control ESC sets, the mesenchymal transcription factor SNAI1 was significantly reduced in the endometriosis ESC sample exposed to PMCs compared to the solo culture and normal ESC samples regardless of PMC exposure ( Fig. 5 ), suggesting an overall decrease in a SNAI1-mediated mesenchymal program is affected by PMC interactions in endometriosis derived eutopic ESCs but not in ESCs from subjects without endometriosis.

As part of an IRB-approved study with patients’ informed consents, primary ESCs were obtained from endometrial pipelle biopsies from women with and without endometriosis surgically confirmed and biopsy proven. The eutopic ESCs were obtained from six subjects with and without endometriosis (n = 3, each) with characterized menstrual cycle phase on the basis of patient-reported day since last menstrual cycle, histologic analysis of biopsies using Noyes criteria, and serum estrogen/progesterone levels ( Supplemental Table S1 ). These ESC samples were divided into normal and endometriosis pairs, matched for menstrual phase [167 vs. 163 (early secretory); 171 vs. 169 (mid-secretory); and 172 vs. 170 (proliferative-early secretory)]. Primary eutopic ESCs were purified using standard protocols as described previously and were stained for vimentin to ensure stromal phenotype ( De La Garza, Binkley, Ganapathy et al., 2012 ). Human LP9 established PMCs, obtained from the Coriell Foundation (Cat # AG07086; Camden, NJ), have been validated as a PMC model in several studies ( Lucidi et al., 2005 ; De La Garza et al., 2012 ; Liu et al., 2009 ). All cells were used at passage ≤5. ESCs were grown in Dulbecco’s Modified Eagle Medium: Nutrient Mixture F-12 (DMEM/F-12) supplemented with 10% fetal bovine serum (FBS) and 100 U/mL penicillin/streptomycin. LP9 PMCs were maintained in M199:MCDB131 media with 15% FBS, 0.4 μg/ml hydrocortisone, 1 mM sodium pyruvate, 1 mM glutamine, 0.2 μg/ml human epidermal growth factor (hEGF), and 100 U/mL penicillin/streptomycin. Cytometry by time-of-flight (CyTOF) ready antibodies were obtained from commercial sources. The antibodies and their sources are listed in Supplemental Table S2 . Metal conjugation to antibodies was carried out by the Maxpar ® X8 Antibody Labeling Kit per manufacturer’s protocol (Fluidigm Corporation, San Francisco, CA) to generate the 28-marker adhesion and mesenchymal antibody panel. CyTOF was performed using the Helios third-generation mass cytometer (Fluidigm Corporation). CyTOF allows high parameter analysis of cellular protein factors, each targeted with a unique heavy-metal isotope conjugated to a specific antibody. After staining the cell suspensions with the antibodies, factor abundance in each cell is then measured through time-of-flight mass spectrometry in the Helios system after disrupting the cells with high temperature plasma. ESCs alone, and ESCs mixed with LP9 cells (10:1 ratio), were incubated in cell culture for 24 h prior to preparing single-cell suspension, which was stained with Cell-ID Cisplatin solution (Fluidigm, Cat# 201064) to identify dead cells. Cells were then fixed and permeabilized by Fix I Buffer (Fluidigm, Cat# 201065), washed and incubated with the 28 metal-conjugated antibody panel ( Supplemental Table S2 ) for 1hr at room temperature prior to fixation with MaxPar ® Fix and Perm Buffer (Fluidigm Corporation, Cat #201067). Cell-ID Intercalator-Ir 191/193 (0.125 mM - Fluidigm Corporation, Cat #201192A) was added, and cells were incubated at 4 ◦ C overnight to stain the nuclei. Cells were then subjected to CyTOF in the Helios platform to generate the single-cell proteomic profiles based on the antibody staining. Data from at least 15,000 cells were analyzed for each condition. Raw data was normalized by CyTOF software (Version 6.7.1014, Fluidigm) and live cell gating was executed in Cytobank ( https://www.cytobank.org/ , Cytobank Inc.). For co-culture (ESC-LP9 cells) samples, additional gating was performed for the ESC marker profile CD10 + /KRT18 − to select for ESC data points ( Sumathi and McCluggage, 2002 ), and exclude LP9 PMCs. After data extraction and transformation, qualitative assessments, visual illustrations and t-SNE maps were prepared by an automated clustering algorithm PhenoGraph in Cytofkit ( Chen et al., 2016 ). Differences in expression and cluster size in different treatment groups were revealed by violin plots and balloon plots (R package: ggplot2). Heatmaps (MeV) were generated to display differences in factor expression in cell clusters across each sample condition. A principal components analysis (R package: stats) reduced the dimensionality of normalized CyTOF data, and the first two principal components were plotted. The output was then hierarchically clustered (R package: stats) to produce 3 groupings of factors. These groupings were then visualized with a variable correlation plot (R package: factoextra) and used to separate factors for distinct principal component analyses. The two principal components were plotted to demonstrate potential effects of LP9 PMCs on ESC expression profiles (i.e., ESCs in co-culture with LP9 cells versus ESCs alone). For pseudotime analysis, 200 cells were randomly chosen from each patient’s data in the normalized CyTOF FCS files from the six subject ESC samples (with and without LP9 PMC co-culture) using R package. Pseudotime ordering of the single cells was visible as the diffused colored map and built by R package Destiny (version 3.8) analysis through selected cell files ( Angerer et al., 2016 ).

The current CyTOF analysis revealed that specific adhesion and mesenchymal factors are altered in ESCs derived from eutopic endometrium of women with endometriosis upon exposure to PMCs. Thus, ESCs from endometriosis patients are inherently susceptible in their response to mesothelial cell exposure to a much higher degree than ESCs from women free of endometriosis. Consistent with our findings, previous studies showed that the source of eutopic endometrial cells (i.e., whether derived from endometriosis or non-endometriosis subjects) impacts the extent of implantation in an endometriotic invasion in vitro model ( Lucidi et al., 2005 ). The communication between ESCs and mesothelial cells within the peritoneal cavity may be essential in lesion establishment, which can be probed mechanistically by the response of invading ESCs when they encounter PMCs. The decrease in the levels of mesenchymal factors in the primary eutopic ESCs from women with endometriosis upon co-culture with PMCs suggests enhanced plasticity in ESC differentiation that is characteristic of the disease state. Specifically, we observe changes generally consistent with a mesenchymal to epithelial transition (MET), such as decreases in expression of SNAI1 and one or other of ZO 1 or 2 in all pairs tested. Less consistent changes in ZEB2 (a transcription factor often co-regulated with SNAI1) and CDH2 (N-cadherin), which each decreased in one pair, but increased in another, and CAV 2 (involved in regulating signaling levels associated with surface receptors), which increased in two and decreased in one pair, suggest that ESCs are not undergoing canonical MET, but a more complex hybrid state. Similar plasticity has been reported in metastatic circulating tumor cells when seeding a distant site ( Lambert et al., 2017 ). While metastatic spread involves EMT leading to dissemination of circulating tumor cells, a reverse process, MET, is needed for re-differentiation and establishment of lesions at the distant site ( Bilyk et al., 2017 ; Lambert et al., 2017 ). Indeed, hybrid EMT-MET cellular states correspond to higher metastatic potential ( Liao and Yang, 2020 ; Armstrong et al., 2011 ; Cheung and Ewald, 2016 ). MET has also recently been implicated in endometrial regeneration and re-epithelialization from a subset of mesenchymal stem cells within the stromal fraction ( Patterson et al., 2013 ; Owusu-Akyaw et al., 2019 ). Relevant to our study, ESCs also undergo MET during decidualization via modulation of mesenchymal and epithelial factors and induction of epithelial related adhesion molecules ( Zhang et al., 2013 ; Yu et al., 2016 ). Here, we show that mesothelial cells can also elicit an MET-like program in ESCs, which is likely to be required for the establishment of endometriotic lesions. As in metastatic models, the MET changes are not canonical, but nonetheless, the mesenchymal plasticity specific to endometriosis derived ESCs is likely to allow them to integrate with, infiltrate and colonize the mesothelial lining during endometriosis lesion formation. Our data confirm that alterations within the stromal compartment of the endometrium in women with endometriosis could predispose ESCs to initiate endometriotic lesion formation, which is consistent with previous studies ( Houshdaran et al., 2020 ). While CyTOF analysis has been used to profile the immune cellular components in peritoneal fluid of endometriosis ( Guo et al., 2020 ), here we examine the differential effects of mesothelial cells on ESCs from endometriosis patients using high-resolution single-cell CyTOF analysis. A limitation to the study is that the in vitro ESC-PMC co-culture only partially reflects the more complex cell type composition in endometriotic lesions and hormonal milieu. For example, the influence of endometrial epithelial cells in combination with ESCs and PMCs would more closely recapitulate these cellular interactions in vivo . In addition, the effects of steroid hormones on this cellular crosstalk will also need to be elucidated. Nonetheless, our studies begin to lay the groundwork to understanding the influence of heterotypic ESC-PMC interactions that may affect the development of endometriotic lesions. Additionally, lesion formation is a continuum of sequential steps that include endometrial cell attachment to PMCs, invasion through the mesothelial layer, and establishment of the lesion. Our CyTOF analysis was performed for a single time point of co-culture after 24 h of exposure, a period in which there would likely be little activation of inflammatory responses that strongly influence later phases of lesion formation. A time-course co-culture, associated with estrogen/progesterone doses corresponding to menstrual cycle progression, with PMCs may delineate further temporal changes in the expression profiles. Examining endometrial mesenchymal plasticity profiles (i.e., concurrence of epithelial and mesenchymal marker expression in the stromal compartment) in endometriotic lesions, compared to eutopic endometrium, is needed to validate our in vitro findings. These studies would also delineate potential effects of these steroid hormone fluctuations on the mesenchymal phenotype, as little is known on hormonal effects on endometrial stromal cell expression profiles ( Houshdaran et al., 2020 ). A recent study demonstrated that estrogen and progesterone have differential genomic and epigenomic effects on endometrial stromal fibroblasts from women with and without endometriosis ( Houshdaran et al., 2020 ). In summary, our data reveal that expression changes of mesenchymal and cell junction components in ESCs, elicited by interactions with PMCs, were unique to women with endometriosis. These proteomic alterations appear to predispose ESCs towards a plastic mesenchymal phenotype that may facilitate lesion establishment in the peritoneum. Future studies should include a larger number of patient-control samples to verify the differential regulation by, and temporal effects of, PMCs on the endometrial compartment, as well as inclusion of markers for endometrial mesenchymal stem cells. Delineating the contribution of mesenchymal plasticity to endometriosis lesion formation will enhance our understanding of the etiology of this disease.

Endometriosis is a gynecologic disorder arising from the implantation and growth of endometrial tissue in the peritoneal lining and organs within the pelvic cavity, which can result in pelvic pain and infertility. Retrograde menstruation of shed endometrial tissue was initially hypothesized by Sampson as the source of endometriotic lesions ( Burney and Giudice, 2012 ). Studies from humans and baboons suggest that the eutopic endometrial lining is altered at the molecular level in endometriosis ( Tamaresis et al., 2014 ; Hastings and Fazleabas, 2006 ; Kao et al., 2003 ; Klemmt et al., 2006 ). This is also associated with decidualization defects that contribute to the high rates of infertility in endometriosis patients. The altered expression profiles in eutopic endometrium specific to endometriosis patients suggest that inherent changes programmed into endometrial cells prime them to invade the peritoneal mesothelial lining, explaining why 95% of women experience retrograde menstruation, but only 10% develop disease. While aberrant properties of endometrial tissue may initiate the disease, changes in the environment within the peritoneal cavity, including a heightened inflammatory response, are likely to also contribute to endometriosis lesion development ( Burney and Giudice, 2012 ; Tamaresis et al., 2014 ; Hastings and Fazleabas, 2006 ). Peritoneal fluid in endometriosis not only has an altered complement of immune cells, but also a cytokine profile promoting inflammatory responses and immune suppression ( Guo et al., 2020 ; Oral et al., 1996 ; Forster et al., 2019 ), although it remains unclear how much this contributes to, or is a consequence of, the disease process. Most central to lesion formation, however, is the direct interaction between the peritoneal mesothelial lining and invading endometrial cells, and whether refluxed endometrial tissue from normal and endometriosis subjects differ in their responsiveness to the mesothelium. The primary component of the endometrium thought to initiate lesion formation is the endometrial stromal cells (ESCs) ( Burney and Giudice, 2012 ). ESC interactions with the substrate mesothelial cell lining are required as an initial step of invasion, which may trigger infiltration of the mesothelial lining ( Lucidi et al., 2005 ; Nair et al., 2008 ). This endometrial-mesothelial interaction requires intercellular adhesion molecules such as cadherins and specialized structures like gap junctions, which are essential components of heterotypic cell-cell signaling during invasive processes, including extravasation by immune cells, embryo implantation into the endometrium, and cancer metastasis ( Winterhager et al., 1993 ; el-Sabban and Pauli, 1991 ; Ito et al., 2000 ). Modulation of the mesenchymal and epithelial characteristics of invading cells, a process known as epithelial-mesenchymal transition (EMT), is observed as an initiating event in metastatic spread and during embryo implantation into the endometrial lining ( Kalluri and Weinberg, 2009 ; Kalluri, 2009 ; Liao and Yang, 2020 ). The reverse process, mesenchymal to epithelial transition (MET), is required for the seeding and integration of metastatic cells at distant sites ( Liao and Yang, 2020 ). Recent studies have suggested that MET also occurs in the regeneration of the endometrial lining post menses by endometrial mesenchymal stem cells ( Patterson et al., 2013 ; Owusu-Akyaw et al., 2019 ; Bilyk et al., 2017 ; Lambert et al., 2017 ; Cousins et al., 2014 ). ESCs also undergo MET by reduction of mesothelial factors (N-Cadherin, SNAIL1, and Vimentin) and increasing epithelial markers (E-cadherin and cytokeratin) during decidualization in preparation for embryo implantation ( Zhang et al., 2013 ; Yu, Berga, Johnston-MacAnanny et al., 2016 ). Thus, the plasticity in mesenchymal characteristics of ESCs essential for endometrial maintenance and function may also be relevant to the pathological process of endometriotic lesion formation. This study focuses on delineating differential regulation of cellular interactions and mesenchymal factors in eutopic ESCs from women with and without endometriosis when exposed to the mesothelium. Single-cell based proteomic analysis by mass cytometry was used to provide high-resolution expression profiles of ESC subpopulations in the presence and absence of peritoneal mesothelial cells (PMCs). Our data suggest that PMC exposure leads to a differential response in cell adhesion and mesenchymal signatures in eutopic ESCs from women with endometriosis.