Endometrial Gap Junction Expression - Early Indicators of Endometriosis and Integral to Invasiveness

Endometriosis is an invasive disease, and a leading cause of pain, infertility and disability among women, with an incidence 10 fold that of cancer. A more complete understanding of disease pathogenesis is essential for the development of non-surgical diagnostic assays and non-hormonal therapeutics. Avoidance of immune clearance and implantation of endometrial tissue on peritoneal surfaces are features of endometriosis lesion formation that overlap with cancer metastasis. Connexins, and the gap junctions they form, have been implicated in cancer progression, and may be associated endometriosis pathophysiology. Single cell transcriptomic profiling of endometrial epithelial and stromal cells from women with endometriosis reveals a striking and progressive shift in expression of connexins and related regulatory and junctional genes. We demonstrate that gap junction coupling between endometrial cells and the peritoneal mesothelium is dramatically induced, specifically in endometriosis patients, and is required for invasion by inducing breakdown of the mesothelial barrier function.

1 women, with an incidence 10 fold that of cancer. A more complete understanding of disease 2 pathogenesis is essential for the development of non-surgical diagnostic assays and non-3 hormonal therapeutics. Avoidance of immune clearance and implantation of endometrial tissue 4 on peritoneal surfaces are features of endometriosis lesion formation that overlap with cancer 5 metastasis. Connexins, and the gap junctions they form, have been implicated in cancer 6 progression, and may be associated endometriosis pathophysiology. Single cell transcriptomic 7 profiling of endometrial epithelial and stromal cells from women with endometriosis reveals a 8 striking and progressive shift in expression of connexins and related regulatory and junctional 9 genes. We demonstrate that gap junction coupling between endometrial cells and the peritoneal 10 mesothelium is dramatically induced, specifically in endometriosis patients, and is required for 11 invasion by inducing breakdown of the mesothelial barrier function. 12 Endometriosis is a chronic inflammatory disease affecting 6-10% of reproductive age 14 women (Eskenazi and Warner,1997). Characterized by the presence of endometrial tissue in 15 extrauterine locations including the pelvic peritoneum, ovary and bowel surface, endometriosis is 16 diagnosed in 35-50% of women with pelvic pain and up to 50% of women with unexplained 17 infertility (Rogers et.al., 2009). At an estimated annual cost of $12,000 per patient in terms of 18 diagnosis and treatment, and adding in the significant loss of productivity, endometriosis care 19 entails significant socioeconomic burden for both individual patients and healthcare systems 20 estimated to cost $80 billion per year for the US alone (Soliman et.al., 2016). In the absence of a 21 biomarker, laparoscopic surgery remains the gold standard for diagnosis. The requirement for 22 invasive surgery, which fails to confirm endometriosis almost half the time (Mettler et.al., 2003) 23 contributes to an average latency of 6.7 years from onset of symptoms to definitive diagnosis 24 (Bontempo and Mikesell, 2020), and results in 68% of women suffering from endometriosis being 25 incorrectly diagnosed (Hudelist et.al., 2012). Diagnostic delay allows time for disease progression, 26 and potentially worsens sequelae and prognosis. An improved understanding disease etiology is 27 critical to developing new diagnostics and therapies. 28 The original, and still most widely accepted, model for the pathogenesis of endometriosis 29 is retrograde menstruation, in which sloughed endometrial tissue during menses traverses the 30 fallopian tubes and forms invasive lesions within the peritoneal cavity (Sampson, 1927). and inflammation (Augoulea et.al., 2012) or decreased immune clearance (Oosterlynck et.al., 45 1991), although this could be due to changes in the endometrial cells themselves (Somigliana 46 et.al., 1996). This has fueled the debate over whether endometriosis originates from changes in 47 the uterus that predispose the cells to lesion formation (the "seed" model) or if it is more a property 48 of a receptive peritoneal environment (the "soil" hypothesis). 49 To understand the molecular underpinnings of endometrial-peritoneal interaction in lesion 50 formation, we have focused on a class of proteins that has been implicated in tissue invasion, 51 infertility and inflammation in other contexts, but incompletely explored in the pathophysiology of 52 endometriosis. Gap junctions, composed of connexin (Cx) proteins encoded by a family of 21 GJ 53 (A-D) genes, mediate direct contact and communication between most cells of the body via 54 exchange of ions as well as metabolites and signaling molecules <1kD (Goldberg et.al, 1999;55 Weber et.al, 2004). Gap junctions have been shown to be essential to many invasive processes, 56 both normal (e.g. blastocyst implantation -Grummer et.al., 1996)  Cx43 emerged as one of three genes (along with PDGFRA2 and CAV-1) central to cancer 66 invasion and metastasis (Cheng et.al., 2015). 67 Gap junctions are also critical for a number of steps in human fertility (reviewed in 68 Winterhager and Kidder, 2015). Endometrial gap junctions, comprised of Cx43, are essential for 69 decidualization [Kaushik et.al., 2020], blastocyst implantation [Grummer et.al.,1996: Diao et.al., 70 2013] and vascularization and endometrial development during pregnancy [Laws et.al., 2008]. 71 Connexins have also been linked to induction of inflammatory processes that either inhibit tissue 72 repair (Willebrords et.al., 2016), such as wound healing of the skin (Montgomery et.al., 2018) and 73 cornea (Ormonde et.al., 2012) processes thought to be driven by ATP release through connexin 74 hemichannels (Mugisho et.al., 2018). 75 With regard to endometriosis, prior immunohistochemistry studies demonstrated a shift in 76 Cx expression of endometrial epithelial cells (EECs) from primarily Cx26, with some Cx32 (GJB1), 77 in the uterus to Cx43 in peritoneal (ectopic) endometriotic lesions (Regidor et.al., 1997). By 78 contrast, endometrial stromal cells (ESCs) show predominantly Cx43 expression in both eutopic 79 and ectopic locations, although at reduced levels in endometriosis patients (Yu et.al., 2014). This 80 modified expression profile of connexins seen in ectopic lesions of women with endometriosis 81 was recapitulated in eutopic endometrial tissue in an endometriosis model in baboons 82 (Winterhager et.al., 2009), suggesting that early changes in the endometrium might predispose 83 refluxed menstrual endometrial tissue within the pelvic cavity for invasiveness (Guo et the extracellular matrix (MME, NOV); housekeeping (UBB, GAPDH, ACTIN, and GUSB) and cell 115 marker (Vimentin (VIM) and cytokeratin (KRT18)) genes. The genes, in the array, in order of 116 presentation in the heat map, are provided in Table S1. Those shaded in grey were control genes 117 fpr data normalization and are not presented in the arrays. 118 To optimally understand how changes in patterns of expression might contribute to 119 disease etiology, we separated the two major endometrial cell types [epithelial or glandular 120 (EECs) and stromal (ESCs)], the purity of which was demonstrated to be ~95% by 121 immunocytochemistry (Fig. S1). To further maximize detection of micro-heterogeneities in 122 expression patterns, we used the Fluidigm C1 single-cell capture system followed by the Biomark 123 microfluidic PCR to examine expression at the single cell level within each population. Two normal 124 subjects and 6 endometriosis patients (one in early stage (I-II), and the others later stage (III-IV) 125 endometriosis) were examined, with samples taken from proliferative, early and late secretory 126 menstrual phases, plus one non-cycling patient on birth control (Table 1). Normalized expression 127 results for each cell type from each patient are displayed as a heat map, with each row 128 representing a gene, and each column a cell (Fig. 1A, B). Distinct patterns of expression in 129 stromal (ESCs) and epithelial cells (EECs) were evident, which in this patient sampling clearly 130 distinguished normal from endometriosis samples, and even showed a gradual transition as 131 disease progressed from early to late stages (Fig. 1A, B). 132 (a) Gap Junction genes: With respect to genes encoding connexins (GJ geneslocated at 133 the top of the maps), a subset of ESCs (~25% of total analyzed cell population) showed high 134 relative expression of all GJ genes in normal subjects, which diminished somewhat in early stage 135 (I/II) endometriosis samples, and drops to less than 10% in late stage (III/IV) endometriosis, 136 regardless of the menstrual phase from which the sample is collected (Fig. 1A). Even though 137 single EEC cell preparations could not be made from all patients, it was clear that the pattern was 138 completely reversed compared to ESCs, with low expression of GJ genes in normal subjects (5% 139 of cells show higher expression), which progressively increases in early (I/II) and late stage (III/IV) 140 endometriosis where ~25% of cells show high expression (Fig. 1B). It is striking that this is true 141 for all GJ genes, and not just the most abundantly expressed, like GJA1(Cx43). The combined 142 expression levels of all GJ genes is illustrated in Violin plots for ESCs ( Fig. 1C) and EECs ( Fig.  143 1D) from each patient, with similar plots for specific GJ genes shown in Fig. S2  were those taken at the mid-secretory menstrual phase when progesterone levels, known to 148 regulate connexin expression in the endometrium (Grummer et.al., 1994), are high [subjects 171 149 (normal) and 169 (late endometriosis)]. EECs show the inverse pattern that was even detectable 150 in early stage endometriosis (patient 164), In EECs, the pattern was not influenced by the stage 151 of the menstrual cycle when samples were collected, although did seem to be affected in the one 152 non-cycling patient on birth control (003). 153 Comparison of individual GJ gene expression (Fig. S1) showed that GJA1 (Cx43) was 154 expressed most highly in both ESCs and EECs, almost 10 fold that of other connexins. For ESCs, 155 all GJ genes measured, except GJC2, showed decreases in late stage endometriosis compared 156 to normal subjects and early stage patients (Fig. S1A). In EECs, most of the GJ genes show the 157 inverse behavior seen in ESCs, increasing expression in endometriosis compared to that seen in 158 normal subjects. The exceptions were GJA1, the most highly expressed connexin, GJA8, GJB7 159 and GJC2 (Fig. S1B). 160 (b) Other genes: Several other genes involved in intercellular or cell-matrix interactions were 161 also found to change in similar pattern to that of the GJ genes. The adhesion related genes 162 EpCAM, ß-catenin (CTNNB1), as well as NOV1 show decreased expression in ESCs in 163 endometriosis (Fig. 1E) but increase in EECs (Fig. 1F), while the master transcriptional regulator 164 of epithelial to mesenchymal transitions (EMT), Snail 1 (SNAIl1), which negatively regulates 165 several adhesion molecules (Hugo et.al., 2011) and Cx43 (deBoer et.al., 2007), shows the inverse 166 pattern (increasing with endometriosis in ESCs and decreasing in EECs). Several other genes 167 change in one or other cells type. In ESCs, the metallo-endopeptidase, CD10 (Fig. S2A), NOV1 168 and the catalytic subunit of protein kinase A gamma (PRKACG) decrease, and ZO2 (TJP2), an 169 accessory protein of tight and gap junctions, increases with disease (Fig. 1E). In EECs, only 170 MAPK1 showed consistent increase in endometriosis patients (Fig. 1F). Some other genes 171 showed increases in the two endometriosis patients in the proliferative or early secretory stages 172 of their menstrual cycle [N-cadherin (CDH2), (Fig. 1F), and PRKACG (Fig. S2B)), but markedly 173 different patterns in the non-cycling patient (Fig. S2B). Independent of these overall patterns, It 174 is notable in the heat maps that the same subset of ESCs from the normal and early endometriosis 175 samples (left three maps in Fig. 1A), and EECs from endometriosis patients (right two maps in 176 Fig. 1B), that show high GJ genes expression, exhibit low expression of several of the tight 177 junction associated genes, but higher expression of adhesion related genes (upper and lower 178 rows, respectively in the group of genes labeled Adh/TJs) ( Fig. 1A and B). Similar to what was 179 observed for the tight junction associated genes, several of the regulatory genes (mostly kinases) 180 we tested show the lowest expression in the cells that exhibit the highest expression of GJ genes 181 (left three maps of ESCs (Fig. 1A) and right two maps if EECs (Fig. 1B). 182 183 Cell cluster analysis 184 As an independent means to assess changes in gene expression in endometriosis, intra-185 sample cellular heterogeneity was analyzed by a graph-based cluster discovery algorithm (Ruan,186 2009). By design, the algorithm is able to identify topologically distinct clusters and, importantly, 187 can automatically determine the most appropriate number of clusters for each dataset. Utilizing 188 this algorithm, ESCs and EECs were divided into 6 and 5 sub-populations, respectively ( Fig endometriosis 5-8% and late stage endometriosis <4%. This pattern was reversed in EECs, 197 although we will need more patient samples to establish specific ranges. 198 199

Functional assessment of GJIC in ESCs and EECs, with endometriosis progression 200
We next assessed the functional consequences of these changes in GJ gene expression by 201 measuring GJIC using an automated variant of the "parachute" technique where dye (calcein) 202 loaded donor cells (D) are dropped onto a monolayer of acceptor cells (A) and the degree of dye 203 spread to the monolayer (A/D ratio) is measured over 10-15 fields in an automated confocal 204 microscope over time ( Fig. 3 A and B). Given the changes in GJ gene expression between normal 205 and endometriosis patients in Fig. 1, it was surprising that GJIC did not change significantly in 206 either cell type, although a small decrease was seen in ESCs and increase in EECs (Fig. 3C). 207 However, as endometriosis is characterized by invasive behaviors following interactions with the 208 peritoneal mesothelium, we also assessed heterotypic GJIC between endometrial cells and a 209 peritoneal mesothelial cell (PMC) line (LP9). This revealed a dramatic induction of GJIC that was 210 most notable in ESCs from endometriosis patients (4 fold increase compared to 2 fold in normal 211 subjects). While some induction was also observed in EECs, this was not significant in either 212 normal or endometriosis samples (Fig. 3C). This induction occurred relatively rapidly, within the 213 context of the assay (~ 2 hours). 214 Given this time course, which seemed inconsistent with a transcriptional event, we 215 examined the distribution of Cx43 protein in ESCs alone, and after contact with PMCs. Alone, 216 both stromal (Fig. 3D) and mesothelial cells (not shown) show primarily intracellular Cx43 217 staining. By contrast, in heterotypic cultures, very little intracellular staining is seen within ESCs 218 (fluorescent green labelled cell in Fig. 3E), but plaques are readily observed at sites of contact 219 with PMCs (yellow arrows, Fig. 3E), although significant intracellular labeling within PMCs was 220 still evident. This suggests that the heterotypic contact may trigger enhanced trafficking of Cx43 221 to the cell surface. 222 Since the major functional difference we observed related to gap junctions in samples 223 from endometriosis patients were associated with heterotypic contacts with mesothelial cells, 224 especially by ESCs, we predicted that they may influence differential adhesion between the cells. 225 To measure the degree to which heterotypic adhesion properties with mesothelial cells differ 226 between EECs and ESCs, we used atomic force microscopy (AFM) that enables quantitative 2017). Adhesiveness between PMCs (LP9 cells), and EECs and ESCs from endometriosis 229 patients was measured by attaching a single cell to a cantilever of an AFM probe (Fig. 4A). The 230 attached cell was placed in contact with a monolayer of either the same cell, or a different cell 231 type, growing on a culture dish, and the force needed to separate the probe-attached cell from 232 the plate-attached cell was measured (Fig. 4B). Mesothelial cells show low levels of adhesion to 233 one another that was similar to their adhesion to EECs. In contrast, ESCs showed significantly 234 higher adhesion to PMCs (Fig. 4C). This data is consistent with ESCs being the cell type that is 235 the major invasive front into the mesothelium (Nair et We then directly assessed the invasiveness of primary endometrial cells from normal and 240 endometriosis patients using an established 3D-invasion assay that assessed the efficiency with 241 which ESCs or EECs can pass through a confluent monolayer of PMCs grown on a growth-242 hormone depleted Matrigel-coated membrane (Fig. 5A). Initial studies using unseparated 243 endometrial cells showed significantly greater invasiveness in samples from endometrial patients 244 than control subjects (n=3; Fig. 5B). For ESCs (n=3 and 6 for control and endometriosis, 245 respectively) and EECs (n=3 and 6 for normal and endometriosis, respectively) the differences 246 were not significant. In all cases tested the level of invasion was not seen when endometrial cells 247 were dropped on a membrane alone, so that the invasion was dependent on contact with a 248 mesothelial monolayer (data not shown). 249 Given that we had demonstrated that we had demonstrated that endometrial cell contact 250 with mesothelial cells specifically induced GJIC, we tested the degree to which invasive behavior 251 was dependent on GJIC by targeting Cx43, the dominantly expressed Cx in both cell types (Fig.  252  S2). Four approaches were used to test this. Firstly, we pre-treated both PMCs and endometrial 253 cells (not separated into ESCs and EECs) with GAP27, a peptidomimetic to the extracellular 254 domain of Cx43 that we, and many others (Evans and Leybaert, 2007), showed blocked the 255 formation of gap junctions between newly contacting cells, in this case by ~85%. This reduced 256 invasiveness of unseparated endometrial cells from control subjects by 57% (although not 257 significant at the 0.05 level), and from endometriosis patients by 70% (p<0.01) (Fig. 6A). 258 Secondly, we tested purified ESCs (as only they showed increased adhesiveness to 259 PMCs (Fig. 4)), using a combination of two siRNAs targeted to Cx43, where the PMC monolayer 260 was transfected immediately prior to the invasion assay. This caused an average of 65 ± 9 % 261 reduction in Cx43 protein (Fig. 6B) and ~70% reduction in GJIC between endometrial and 262 mesothelial cells (p<0.001 in control and <0.0005 in endometriosis) (Fig. 6C). Invasiveness was 263 reduced by 30-40% (p<0.05 in control, and <0.02 in endometrial samples) (Fig. 6D). In all cases, 264 Cx43 siRNA effects were compared to effects by either a scrambled siRNA, or siRNA directed to 265 GAPDH. 266 Thirdly, in order to avoid any negative effects of the transient transfection process for 267 siRNA on the integrity of the PMC monolayer, or invasive potential of ESCs, we prepared stably 268 expressing LP9 cells and ESCs from a normal and endometriosis patient (172 and 163, 269 respectively), by infection with lentivirus that either expressed shRNAs targeted to Cx43 that were 270 inducible by doxycycline. In the presence of doxycycline, shRNA suppressed Cx43 protein levels 271 by 35 ± 6 % (Fig. 6E), and GJIC was by 88 ± 4 % in the cell types tested (Fig. 6F). Invasive 272 behavior was also inhibited by 86 ± 2 % when Cx43 was suppressed in the ESCs from either 273 control or endometriosis patients, or in the PMCs (Fig. 6G). 274 Finally, we used the same Lentivirus system to express a dominant negative Cx43 275 construct, Cx43 T154A (DN Cx43), which we have previously shown preserves normal gap 276 junctional plaque structures, but fails to open functional channels, and prevents the opening of 277 co-expressed wt Cx43 (Beahm et.al, 2006). Expression of DN Cx43 increased total Cx43 levels 278 by 2 fold in ESCs and 1.4 fold in PMCs (Fig. 6E), decreased GJIC by 90% (Fig. 6F) and invasive 279 behavior by a remarkable 98-99%, whether expressed in ESCs or PMCs (Fig. 6G). 280 This analysis clearly shows that both EECs and ESCs can be invasive across a 281 mesothelium, and that this is even true when they are derived from control subjects, although the 282 levels are higher in cells from endometriosis patients. Three independent treatments that 283 specifically block either functional GJIC between ESCs and PMCs (GAP27 or DN CX43) or 284 expression of Cx43 in either of the cell types, greatly reduce the invasive behavior in both control 285 and endometrial samples. Tests on EEC invasiveness (data not shown) showed similar 286 dependence of invasiveness on Cx43, but the higher variability between samples from different 287 patients led to the effect not being significant at the 0.05 level.

289 AFM analysis of ESC effects on the PMC monolayer 290
Given the clear dependence of invasiveness of ESCs through the mesothelium on Cx43 291 mediated GJIC, we turned to AFM to provide insights into how this occurs. We specifically probed 292 the influence of ESCs, in the presence or absence of Cx43, on the "barrier function" of the 293 mesothelium (i.e. the "tightness" of contact between PMCs that prevent transmigration of cells). 294 ESCs from a control subject (172) or endometriosis patients (169 and 170) were first labeled with 295 the membrane dye DiO, and dropped onto a PMC monolayer at a ratio (ESC:PMC) of 1:20. After 296 ~3 hrs, the monolayer was imaged with AFM by pressing it with a 'sharp' conical probe at a 297 constant pressure of 1 nN to obtain a 3-D contour map of the monolayer (Figs. 7A-B). This readily 298 allowed the interfaces between cells to be identified (Fig. 7C), and to measure the depth of 299 penetrance between cells. ESCs from several patients all induced an increase in penetrance, 300 measured ~10um away (~2 cell diameters) from site of the dropped cells, identified based on prior 301 DiO labeling (Fig. 7D). This was also evident as a widening in the gap between cells (compare 302 Figs 7A and B). This effect on penetrance between cells was most pronounced in ESCs from an 303 endometriosis patient (169) compared to a control patient (171) (Fig. 7D), and also showed a 304 dependence on the density of ESCs, as penetrance was less when a 1:50 ratio of ESCs:PMCs 305 was used instead of 1:20 (data not shown). We have also conducted a similar test using an AFM 306 probe to which a 3 M diameter glass bead is attached to mimic the shape of a cell, and similarly 307 shown that ESCs induce greater ability for larger objects to penetrate the monolayer (data not 308 shown). 309 We then used the shRNA, DN Cx43 and wt Cx43 infected LP9 cells, characterized in Fig.  310 6E, to test the dependence of these changes on Cx43 GJIC. First, we observed that the "barrier 311 function" of mesothelial cells alone (i.e. in the absence of ESCs) was dependent on Cx43 312 expression, as the degree of penetrance was reduced when Cx43 was overexpressed, and 313 increased when Cx43 was inhibited by shRNA (Fig. 7E). Strikingly, this effect was exactly inverted 314 when we introduced ESCs, as now Cx43 overexpression significantly increased penetrance, while 315 Cx43 inhibition by shRNA completely eliminated the effect of ESCs, so that penetrance was the 316 same as in their absence. This effect was due to the channel forming role of Cx43, as the same 317 reduction in penetrance was observed in LP9 cells infected with a DN Cx43, which would 318 preserve, or even enhance, any adhesive functions of Cx43 between ESCs and PMCs. 319 320

DISCUSSION 321
Invasive processes of any cell into a "foreign" environment occurs in many instances, both 322 normal (during development, extravasation, blastocyst implantation) and pathological 323 (metastasis) and involve close intercellular interactions. While the need for initial adhesion events 324 has long been recognized, it has become increasingly evident that the formation of heterotypic 325 gap junctions between the invading and target tissue is an early event in implantation (Grummer disease progression were those characterized by high GJ expression with (Fig. 2). This pattern 340 has likely escaped prior screens of endometrial tissue using bulk PCR approaches, as the 341 changes in GJ gene expression in the two major cell types of the endometrium were in opposite 342 directions, with stromal cells showing decreased expression (Fig 1C), and epithelial cells 343 evidencing upregulation (Fig 1D) as disease stage increased. 344 Notably, these changes were not restricted to the connexins expressed most abundantly 345 in the endometrium (Cx43, 26 and 32), but applied to most of the GJ gene members screened. It 346 was also notable that Cx43 was the dominantly expressed Cx in both ESCs and EECs. For the 347 former, this was consistent with what has been reported in the literature, but EECs have been 348 reported to display mostly Cx26 and Cx32 expression in eutopic endometrium (Jahn et.al., 1995).

349
This has been shown to switch to Cx43 in ectopic lesions (Regidor et.al., 1997), and even 350 eutopically in a baboon model of endometriosis (Winterhager et.al., 2009). While it is possible that 351 the short time in culture modified this expression pattern of EECs in our study, it is also possible 352 that the switch to Cx43 occurs very early in release of cells from the endometrial lining, even in 353 normal subjects. In either event it is clear that the expression of Cx43 is greatly enhanced in EECs 354 from endometriosis patients compared to control, even while they are still resident in the eutopic 355 endometrium. 356 In terms of other genes involved in intercellular interactions, many do not show consistent 357 changes, at least in both cell types, but two genes associated with adhesion (EpCAM and ß-358 catenin) show the same expression shifts as the GJ genes, while Snail 1, a master transcriptional 359 repressor that promotes endothelial to mesothelial transition (EMT), shows the opposite pattern. 360 This is consistent with the demonstrated suppressive effect Snail 1 has been shown to have on 361 Cx43 (deBoer et.al., 2007), and the promotion of a migratory phenotype in ESCs, and its 362 repression in EECs. However, it is not clear if in this instance Snail1 is activating a typical EMT 363 response, as the expression of E-cadherin, which is a primary target for suppression by Snail1, 364 was not detected in the endometrium, and N-cadherin, the expression of which is normally 365 increased by Snail1, increases in endometriosis EECs coincident with a Snail 1 decrease. 366 Furthermore, the increase in Snail1 in ESCs is associated with a reduction in ß-catenin levels, 367 which may limit EMT effects as they typically depend on activation of ß-catenin signaling. This is 368 consistent with the view that EMT transitions in endometriosis are only partial (Konrad et.al.,369 2020). 370 With regard to kinases that regulate many intercellular junctional components, it is notable 371 that the cells with the highest GJ expression (control ESCs and endometriosis EECs) in general 372 show the lowest expression of most kinase genes (see heat maps in Fig. 1 A-B), suggesting an 373 inhibitory relationship. However, in terms of global kinase levels, the only kinases to show 374 significant shifts with endometriosis were PKA gamma in both ESCs and EECs, and MAPK1 in 375 EECs (Fig. 1E-F, and Fig. S2B), all of which changed in the same direction as the GJ genes. 376 This is a striking example of how overall expression levels can be misleading when compared to 377 the details of expression patterns at the single cell level. when contacting mesothelial cells that is greatly enhanced in endometriosis samples (Fig. 3C).

385
This induction, linked to increased trafficking of Cx43 to the cell surface ( Fig. 3D and E The importance of the enhanced heterotypic coupling between ESCs and PMCs, unique 392 to endometriosis, to their invasiveness across the mesothelium was demonstrated using four 393 different modes of inhibition of Cx43. Specifically, invasion was not only inhibited by suppression 394 of Cx43 expression ( Fig. 6D and G), but also by a peptide blocker GAP27 (Fig. 6A), which inhibits 395 all channel function, but leaves protein expression unaffected, and by DN Cx43 expression (Fig.  396  6G), which actually increases protein levels (Fig. 6E) and gap junction structures (Beahm et.al.,397 2006), but blocks all channel activity. Notably, inhibition of invasion is the same whether Cx43 398 function is ablated in ESCs or PMCs (cf. shRNA and DN Cx43 treatments for 163S and 169S in 399 Fig. 6G). This suggests that heterotypic GJIC likely triggers the invasive behavior, which, as we 400 demonstrate with AFM (Fig. 7), is associated with increased separation between PMCs, likely 401 resulting from a disruption of the adhesion and tight junctions between the cells (i.e. breakdown 402 of the "barrier function"), akin to what happens during extravasation (Ito et.al., 2000;Reymond 403 et.al., 2013). 404 In Fig. 8 we present a summary of how interactions between the mesothelium and 405 endometrial cells arriving in the peritoneum by retrograde menstruation differ in normal and 406 endometriosis patients, resulting in lesion formation. The enhanced GJIC seen in endometriosis 407 ESCs (Fig. 8B) would allow signals, indicated by green triangles, to pass effectively from 408 endometrial cells to the mesothelium, where they are further propagated through GJIC, to 409 promote disruption of "barrier function" between PMCs, allowing invasion. Interestingly, in the 410 absence of signals from ESCs, GJIC was shown to be important to maintenance of the barrier 411 function (Fig. 7E). This may also be why DN Cx43 was more effective in preventing invasion than 412 Cx43 knock-down by sh-or si-RNA (Fig. 6G), as it may reinforce the barrier functions by 413 enhancing adhesiveness between PMCs. The model emphasizes how increased receptiveness 414 of the ESCs to form heterocellular gap junctions in endometriosis can lead to invasive lesion 415 formation. However, it is also possible that they endometriotic ESCs may make more of the 416 hypothetical signals that induce disruption of the barrier function, emphasizing the importance in 417 identifying such factors. It should be noted that we cannot definitively exclude involvement of 418 factors released through Cx43 hemichannels in the invasive process, but there are no obvious 419 signals that would induce their opening in these experiments, and effects on invasion are the 420 same independent of the cell in which Cx43 is inhibited. 421 Much remains to be done to identify the intercellular signals that mediate barrier function 422 disruption, and also to understand reciprocal effects of the mesothelium on the ESCs, aside from 423 the demonstrated induction of heterotypic gap junction formation. However, the current study 424 clearly implicates changes in gap junction expression in the earliest phases of the development 425 of endometriosis, and their critical role in initiating lesion formation. They also demonstrate that 426 changes within the uterine endometrium prime these cells to be invasive once they reach the 427 peritoneal cavity, in much the same way that the metastatic potential of cancer cells is determined 428 as they leave the primary tumor, suggesting that understanding each of these processes will 429 inform the other. 430

Primary endometrial epithelial cell isolation from endometrial biopsies 433
Primary ESCs and EECs were isolated from endometrial biopsies obtained from women with and 434 without endometriosis. All women provided informed consent prior to participating in this 435 Institutional Review Board approved protocol. Study subjects were premenopausal women 436 between 20 and 45 years of age with regular menstrual cycles undergoing laparoscopic surgery 437 for gynecologic indications ( Table 1). Women with pelvic inflammatory disease/hydrosalpinx, 438 endometrial polyps, or submucosal fibroids were excluded. Endometriosis was staged according 439 to the revised ASRM criteria and confirmed by histopathologic review of peritoneal or cyst wall 440 biopsy in all cases. Fertile women undergoing tubal sterilization and without endometriosis at 441 surgery were considered healthy controls. Menstrual cycle phase (proliferative or secretory) was 442 determined by cycle history and confirmed by serum estradiol and progesterone levels.

443
Endometrial tissue was obtained by pipelle biopsy at the time of laparoscopic surgery. 444 The biopsy material was dissociated by shaking in 5mg/ml collagenase and 2.5mg/ml 445 DNase in Hanks Balanced Salt Solution at 37 0 C for 1 hour. Isolation of primary ESCs and EECs 446 from the biopsies was performed using a combination of straining (45uM nylon filter) and 447 differential sedimentation (EECs cluster and sediment faster), followed by differential attachment 448 (EECs adhere less well to culture plates), in a modification of the method developed by Kirk and 449 Irwin (1980)  permeable dye that on cleavage by intracellular esterases becomes membrane impermeable, but 497 permeable to gap junctions. After washing, trypsinization and addition of assay media, ~2500 498 calcein-labeled donor cells per well are dropped ('parachuted') onto the recipient cell layer, and 499 calcein transfer between donor and recipient cells observed by fluorescent microscopic imaging 500 (Fig. 3A). For homotypic interactions, ESCs, EECs or LP9 donor cells were parachuted onto 501 recipient cells of the same type. For heterotypic GJIC assays, ESCs or EECs were parachuted 502 onto LP9 recipient cells. Fluorescent, bright field and digital phase contrast images of 10-15 fields 503 per well were captured on an Operetta automated microscope (Perkin Elmer) at 30 min intervals 504 for approximately 2 hours. A program (developed in consultation with Perkin Elmer) allowed 505 identification of all cells on the plate, (from phase contrast image), original donors (5-15 per field), 506 and dye-filled recipients (based on calcein intensitysee Fig. 3B). Data are expressed as # of 507 fluorescent recipient cells/# of donor cells for each condition (R/D ratio), plotted over time, and a 508 linear regression line drawn through the data, with the slope used as a measure of coupling (Fig.  509 3C) and regression coefficient (mostly >0.8) used as a measure of assay reliability. 510 511 Trans-mesothelial Invasion Assay 512 The 3-D invasion assay modeling trans-mesothelial invasion (Fig. 5A)  growth-factor-reduced Matrigel TM , coated on 8-µm pore membranes (Corning, NY). ESCs or 516 EECs were then labeled with CellTracker Green® or DiO (Invitrogen), trypsinized and counted, 517 prior to dropping onto the confluent layer of LP9 PMCs in the prepared inserts (~20,000 cells per 518 insert), at which time the media below the insert was changed to the appropriate media for the 519 invading cell (ESC or EEC). After 24 hr. incubation, non-invading cells on the upper surface of the 520 insert were mechanically removed. Invading cells on the bottom of the membrane insert, were 521 stained with DAPI, and 10 fields counted using an Inverted Nikon 2000 fluorescence microscope 522 with 20x objective. Invasion assays for each cell type were performed in triplicates. 523 To test the role of gap junctions in the invasive process, we initially pretreated both the 524 monolayer and dropped cells for 24 hours with 300uM GAP27 (Zealand Pharma, Copenhagen, 525 Denmark), a peptidomimetic of the extracellular loops of Cx43 which competitively blocks Cx43 526 gap junction formation. Due to difficulties in obtaining consistently active peptide preparations, we 527 subsequently shifted to 24 hour pre-treatment of the LP9 monolayer with a combination of two 528 siRNAs to Cx43 (10 pmoles/well or 5nM final concentration) -Ambion™ Silencer™ Select) in Opti 529 MEM (Gibco, NY) with RNAi MAX (1/100 dilution, Invitrogen), diluted 1:1 with assay media, per 530 manufacturer's instructions. As the siRNA transfections done immediately prior to invasion could 531 impact cell behavior, we also knocked-down Cx43 expression using 3 different shRNAs 532 (Dharmacon, UK) introduced into either Endometrial or LP9 Mesothelial cells via Lentivirus 533 (carrying an RFP reported driven by a separate promoter) infection. Average infection efficiency 534 was 58% (range 30-80%). In assessing invasion, cells expressing shRNA were identifiable as 535 RFP positive, while unlabeled cells in the same experiment served as internal controls. 536 537

AFM measurements of cell-cell adhesion and mesothelial integrity. 538
We applied a Nanoscope Catalyst AFM (Bruker) interfaced with an epifluorescent inverted 539 microscope Eclipse Ti (Nikon). AFM images were acquired with the Peak Force Quantitative 540 Nanomechanical Mapping (QNM) mode with cells immersed in appropriate culture media. 541 ScanAsyst probes (Bruker) with the nominal spring constant 0.4 N/m were used for imaging. The 542 exact spring constant for each probe was determined with the thermal noise method (Butt and 543 Jaschke, 1995). For each cell culture dish, at least 5 fields 100 by 100 m were collected with the 544 Peak Force set point of 2nN, and electronic resolution of 256 by 256 pixels. Nanomechanical data 545 were processed with Nanoscope Analysis software v.1.7 (Bruker) using retrace images. Then, 546 the Sneddon model (Sneddon, 1965) and the rules established by Sokolov (Sokolov and Dokukin,547 2014) were applied for calculations of mechanical parameters. 548 Cell to cell adhesion: We attached a tester cell to a tipless probe MLCT-O10 (Bruker, cantilever 549 A, spring constant 0.07N/m) using polyethyleneimine (PEI) as a glue (Friedrichs et. al., 2013) 550 (Fig. 4). Briefly, the probes were immersed in 0.01% PEI in water for 30 min. Tester cells 551 attachment to a culture dish was weaken by replacement of the culture medium with a non-552 enzymatic cell dissociation solution (Millipore) for 15-30 min in a cell culture incubator (37°C, 553 5%CO). Next, a single tester cell loosely attached to a culture dish was attached to a PEI covered 554 cantilever by pressing it at 1 nN for 5-10 min. After visual inspection of successful cell attachment, 555 the tester cell was lifted and transferred to a dish containing single tested cells. Then the tester 556 cell was positioned over a tested cell and the cantilever slowly lowered till detection of cell-cell 557 interactions with a force plot. The cells were left interacting for 30 to 180 sec at forces 0.5 to 5 nN 558 and then the tester cell was lifted. During this step a force plot was recorded and the collected 559 data applied to calculate cellcell adhesion parameters. The force plots were baseline corrected 560 and a maximum of adhesion between cells during their detachment was calculated (units of force, 561 Newton) (Taubenberger, Hutmacher, and Muller 2014; Dufrêne et al. 2017). 562 Integrity of LP9 layer: LP9 cells were grown to confluence in a 60mm culture dish (Fig. 7) except the 170S-LP9 (Cx43) data set in Fig. 7F, which did pass the Kolgomorov test of normality. 588 P values were corrected using the more stringent Shapiro method. 589 590 Cell Cluster Analysis 591 Single cells were clustered based on the normalized expression values using an in-house graph-592 based community discovery algorithm. Briefly, the algorithm starts by computing the Euclidean 593 distances between pairs of cells, and then constructs k nearest neighbor graph, where each cell 594 is connected to k cells that are closest to it. The best k is chosen by the algorithm with a 595 topologically inspired criterion (Ruan, 2009). Finally, a community discovery algorithm optimizing 596 the well-known modularity function is applied to find dense subgraphs as cell clusters (Ruan and 597 Zhang, 2008 Table S1. (C-F) Violin plots showing the distribution of RNA expression levels (log scale) for selected genes, or groups of genes, in stromal (C, E) and epithelial cells (D, F) derived from each of the patients studied (indicated by numbers on the X-axis). C and D show the aggregate of combined GJ gene expression, while E and F show 6 genes with the most consistent changes with endometriosis. Green outlines indicate increased expression and red outlines decreased expression in endometriosis patients compared to control (EECs) or control and early endometriosis, excluding the sample taken during mid-secretory phase of the menstrual cycle (ESCs), indicated by the red horizontal line in each plot. Horizontal bulges in each plot indicate larger numbers of cells with that expression level. The black dot in the middle of the plot indicates mean expression. Significance was determined by Duncan's multiple range test and statistical differences (P<0.05) between samples, as indicated by letters (e.g. a vs. b; b vs. d). In addition, 'a' indicates the highest mean expression and 'd' indicates the lowest mean expression value for each factor (i.e. a>b>c>d). Violin plots of the remaining genes are shown in Fig. S2.

Figure 3: Heterotypic coupling between ESCs and PMCs is dramatically induced in endometriosis.
Gap junction intercellular coupling (GJIC) was measured by a modified "parachute assay" where calcein ester loaded donors are dropped onto a monolayer of acceptors, and dye transfer measured over time as a linear increase in fluorescent acceptor/donor ratio (A). This can measure homotypic (donor and acceptor the same), or heterotypic (endometrial donors and mesothelial acceptors) GJIC (B). Although there are slight changes in homotypic GJIC in ESCs and EECs consistent with the changes in expression levels of connexins, they were not significant (C-green bars). By contrast, heterotypic interactions with mesothelial cells (PMCs) caused a significant induction of GJIC, most notably in ESCs from endometriosis (C-black bars). Immunocytochemistry showed this was likely due to a redistribution of Cx43 (red) from primarily cytoplasmic in homotypic ESCs (D), to the interfaces of ESCs (green) and PMCs in heterotypic cultures (E). Samples tested: ESC control (8) and endometriosis (11); EEC control (8) and endometriosis (5), which included 2-4 biological replicates for each patient. Significance level was calculated by two tailed t-test. Cells attached to an AFM tip are bought into contact with PMCs growing on a dish (A), and the force needed to separate them is measured (B). PMCs (LP9 cells) show similar adhesion to one another as to EECs, but much stronger adhesion to ESCs (C). Significance determined by two tailed t-test Using a 3-D model of endometrial cell invasion across a peritoneal cell monolayer (A), endometrial cells from patients show greater invasion than those from control subjects that was significant with a mix of ESCs and EECs (B). Number of independent patients tested -All Controls (3); Endometriosis -ESC+EEC (3 patients); ESC (5); EEC (4). Significance was determined by twotailed t-test. (A) Averaging ESCs from all control (black bars, n=3) and endometriosis patients (grey bars, n=6), invasion was inhibited by a peptide inhibitor of GJ channels, GAP27 (percent GJIC compared to untreated shown below each bar).
(E -G) Stable, doxycycline inducible shRNA infections of ESCs from a normal (172) and endometriosis (163) patient, as well as PMCs (LP9), reduced levels of Cx43 (arrow) compared to Laminin controls (E), inhibited GJIC (n=3-7 independent experiments with 3 different shRNAs) (F) and invasiveness by ~90% [10 technical replicates of all samples; 3-4 biological repeats for +/-shRNA using 3 different shRNAs (controls were internal)] (G). The block of invasiveness was observed independent of whether Cx43 expression was inhibited in ESCs or PMCs. DN Cx43 (which blocks channel function, but not protein assembly into GJs) was also expressed in ESCs and PMCs at approximately equal levels to wt Cx43 (total Cx43 levels doubled -E). This caused an even greater inhibition of invasiveness (~98% -G). All statistical significance based on onetailed t-tests. The topology of the mesothelial monolayer surface shows a significant increase in the spacing between cells in the presence of ESCs (A-B). Using constant pressure, the tip can measure the depth of penetration between cells (C). This was increased from that measured in anLP9 PMC monolayer alone (n=25) when ESCs were dropped onto the monolayer, particularly from an endometriosis patient (169, n=19)) compared to control (172, n=13) (D). Cx43 appears to help preserve the barrier function of the mesothelium alone, as overexpression reduced penetration, while Cx43 shRNA infection increased penetration in the PMC monolayer in the absence of ESCs (E). By contrast, in the presence of ESCs from a patient (170), Cx43 overexpression in PMCs enhances penetration, while blocking Cx43 with either shRNA or DN Cx43 neutralize the effect of ESCs (F). Each dot represents a cell measurement from plates prepared on the same day for comparison. Statistical tests of significance by two-tailed t-test (see methods). (A) In healthy patients, when endometrial cells (brown) encounter a mesothelium (following arrival in the peritoneum via retrograde menstruation), modest GJIC provides limited exchange of signals from ESCs to PMCs (green triangles) or PMCs to ESCs (purple dots). (B) In endometriosis, endometrial cells have Cx43 mostly in intracellular stores, but encounters with mesothelial cells triggers Cx43 trafficking to the cell surface. The increased GJIC that results mediates transfer of signals to PMCs (green triangles), which propagate through the mesothelium, inducing disruption of the adhesive and tight junctions between PMCs, facilitating invasion of the ESCs.