{"paper_id":"affbaeec-2371-47f8-97cb-cff564c20d03","body_text":"A cellular and molecular portrait of endometriosis subtypes\nMarcos A.S. Fonseca1,2*, Kelly N. Wright3*, Xianzhi Lin 1,2*, Forough Abbasi1,2, Marcela Haro1,2,\nJennifer Sun1,2, Lourdes Hernandez1,2, Natasha L. Orr4, Jooyoon Hong4, Yunhee Choi-Kuaea5,\nHoracio M. Maluf6, Bonnie L. Balzer6, Ilana Cass2,a, Mireille Truong3, Yemin Wang4,7, Margareta D.\nPisarska8, Huy Dinh9, Amal EL-Naggar7,10, David Huntsman4,7, Michael S. Anglesio4,11, Marc T.\nGoodman5, Fabiola Medeiros6†, Matthew Siedhoff3†, Kate Lawrenson1,2,5,12† ✉\n1 Women’s Cancer Research Program at the Samuel Oschin Comprehensive Cancer Center, Cedars-Sinai\nMedical Center, Los Angeles, CA, USA\n2 Division of Gynecologic Oncology, Department of Obstetrics and Gynecology, Cedars-Sinai Medical Center,\nLos Angeles, CA, USA\n3 Division of Minimally Invasive Gynecologic Surgery, Department of Obstetrics and Gynecology, Cedars-\nSinai Medical Center, Los Angeles, CA, USA\n4 Department of Obstetrics and Gynecology, UBC, Vancouver, BC, Canada\n5 Cancer Prevention and Control Program, Samuel Oschin Comprehensive Cancer Center, Cedars-Sinai\nMedical Center, Los Angeles, CA, USA\n6 Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles, CA, USA\n7 Department of Pathology and Laboratory Medicine, University of British Columbia, and Department of\nMolecular Oncology, British Columbia Cancer Research Institute, Vancouver, BC, Canada\n8 Division of Reproductive Endocrinology and Infertility, Department of Obstetrics and Gynecology, Cedars-\nSinai Medical Center, Los Angeles, CA, USA\n9 McArdle Laboratory for Cancer Research, University of Wisconsin - Madison School of Medicine and Public\nHealth, Madison, WI, USA\n10 Department of Pathology, Faculty of Medicine, Menoufia University, Menoufia Governorate, Egypt\n11 British Columbia's Gynecological Cancer Research (OVCARE) Program, University of British Columbia,\nVancouver General Hospital, and BC Cancer, Vancouver, BC, Canada\n12 Center for Bioinformatics and Functional Genomics, Cedars-Sinai Medical Center, Los Angeles, CA, USA\na Current affiliation: Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center,\nLebanon, NH, USA\n* equal contribution\n† jointly directed the study\n✉Correspondence to: Kate Lawrenson, PhD. 290W 3rd Street, Cedars-Sinai Medical Center. Los\nAngeles, CA, 90048. Email: kate.lawrenson@cshs.org. Phone: 310-423-7935\nKeywords: endometriosis, single-cell transcriptomics, ovarian cancer, ARID1A, KRAS, SOX17,\ncomplement, inflammation\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nAbstract\nEndometriosis is a common, benign condition characterized by extensive heterogeneity in lesion\nappearance and patient symptoms. We profiled transcriptomes of 207,949 individual cells from\nendometriomata (n=7), extra-ovarian endometriosis (n=19), eutopic endometrium (n=4),\nunaffected ovary (n=1) and endometriosis-free peritoneum (n=4) to create a cellular atlas of\nendometrial-type epithelial cells, endometrial-type stromal cells and microenvironmental cell\npopulations across tissue sites. Signatures of endometrial-type epithelium and stroma differed\nmarkedly across eutopic endometrium, endometrioma, superficial extra-ovarian disease and\ndeep infiltrating endometriosis, suggesting that extensive transcriptional reprogramming is a\ncore component of the disease process. Endometriomas were notable for the dysregulation of\npro-inflammatory pathways and upregulation of complement proteins C3 and C7. Somatic\nARID1A mutation in epithelial cells was associated with upregulation of pro-angiogenic factor\nSOX17 and remodeling of the endothelial cell compartment. Finally, signatures of endometriosis-\nassociated endometrial-type epithelial clusters were enriched in ovarian cancers, reinforcing the\nepidemiologic associations between these two diseases.\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nIntroduction\nEndometriosis is characterized by endometrial-like tissue growing outside of the uterine cavity,\ncausing chronic pain, dysmenorrhea and infertility. Endometriosis is thought to occur in up to\n10% of reproductive-aged people born with a uterus, globally impacting around 85 million people.\nTrue prevalence of the disease in the general population is likely grossly underestimated as\nmany patients are asymptomatic and when symptoms do exist, they tend to be variable and\nnonspecific (e.g., pelvic pain). Consequently, a definitive diagnosis requires surgery with\npathology confirmation (Hickey et al., 2014). In addition to pain and infertility, endometriosis is\nassociated with increased risk of epithelial ovarian cancer, particularly tumor with clear cell and\nendometrioid differentiation ( Lu et al., 2015; Pearce et al., 2012; Kohl Schwartz et al., 2017).\nAlthough clear cell ovarian cancer is relatively rare, these tumors are the second leading cause\nof ovarian cancer deaths due to their poor responses to standard chemotherapy regimens\n(Anglesio et al., 2011).\nMany basic questions about endometriosis etiology remain unanswered. Endometriosis is likely a\nheterogenous disease entity, but efforts to subcategorize based on staging (which largely tracks\noverall disease burden), disease site, or infiltrative behavior have failed to identify meaningful\ncorrelates with pathologic features, severity or pathologic features (Zondervan et al., 2020).\nEndometriosis can be broadly categorized into ovarian endometriosis (endometrioma), superficial\nperitoneal endometriosis, deep infiltrating endometriosis (defined clinically as lesions that\ninfiltrate > 5 mm under the peritoneal surface) and visceral endometriosis. Endometriomas\nexhibit a distinct cystic structure and almost exclusive occurrence within the ovary. Superficial\nlesions are patches of endometriosis on the lining of the peritoneum and can vary in color from\nclear, white and red to darker lesions that are brown, black or blue-ish in appearance. Deep\ninfiltrating endometriosis is characterized by cancer-like local invasive behavior and frequent\nmutations in known cancer driver genes (Anglesio et al., 2017; Lac et al., 2019) but somewhat\nparadoxically, this subtype of endometriosis has comparatively modest associations with ovarian\ncancer risk; whereas ovarian endometrioma has been shown to enhance risk (Saavalainen et al.,\n2018).\nDiagnostic criteria for endometriosis include the presence of endometrial-type epithelium and/or\nendometrial-type stroma in ectopic locations, often accompanied by hemosiderin-laden\nmacrophages. Many lesions are microscopic, rendering traditional bulk genomic characterization\napproaches challenging as they are sensitive to isolation of uniform/pure specimens. In addition,\nendometriosis is often treated with ablation, which destroys the tissue. As such, the global\nmolecular profiles of endometriosis lesions remain uncharacterized. Excision of endometriosis\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nlesions, not only offers better outcomes for patients (Pundir et al., 2017) but enables\nhistopathologic confirmation of suspected diagnoses and provides the opportunity to explore\ncorrelations of genotype and phenotype among diverse endometriosis lesions. To circumvent the\nchallenges posed by the inter- and intra-patient cellular heterogeneity of endometriosis, we\napplied single cell RNA-sequencing (scRNA-seq) to 25 endometriosis lesions and endometriomas\nfrom seven patients and ovary specimens and eutopic endometrial samples from four women,\nincluding two patients not affected by endometriosis. We used these data to create a cellular\natlas of endometriosis and to identify the molecular hallmarks of endometrial-type epithelial and\nstroma cells in the context of eutopic endometrium, endometrioma or extra-ovarian\nendometriosis. Finally, we leveraged the profiles of endometrial-type epithelium to deconvolute\nbulk expression profiles of clear cell and endometrioid ovarian cancer.\nResults\nSurgical and pathologic characterization of human endometriosis\nTo create a cellular atlas of endometriosis, we assembled a cohort of 9 endometriosis patients\nand 2 patients without endometriosis who underwent minimally invasive gynecologic surgery at\nour institution (Table 1). From these patients we collected a total of 37 specimens, which\nincluded 24 extra-ovarian (peritoneal) endometriosis specimens collected from 7 patients, 7\novarian endometriomas from 6 patients, 4 eutopic endometrium samples (2 from endometriosis\npatients and 2 from patients without endometriosis) and 2 ovary tissues from women unaffected\nby endometriosis. Patients 7-9 and 11 had both endometriomas and endometriosis lesions\nprofiled (Figure 1, Figure S1, Table 1). Both endometriosis-free patients were post-menopausal,\nboth post-menopausal patients and one of the pre-menopausal patients were taking exogenous\nhormones at the time of surgery (Tables S1).\nA detailed pathology review was performed by a specialist gynecologic pathologist (Table 1,\nFigure 1A). A pathologic confirmation of endometriosis was defined as the presence of\nendometrial-type epithelium with endometrial-type stroma, the presence of endometrial-type\nstroma only or endometrial-type epithelium in association with hemosiderin pigmentation. All 7\nthe ovarian endometriomas had endometrial-type epithelium and stroma, plus hemosiderin\npresent (Table S2). For the extra-ovarian endometriosis, 5 out of 24 specimens were suspicious\nfor endometriosis during surgery, but had no endometriosis detected (NED) upon pathologic\nreview (patient 8 - anterior cul de sac and rectal serosa; patient 9 - perirectal fat; patient 10 -\nbladder and left uterosacral ligament). To determine whether endometriosis was present in\ndeeper levels, we examined 5 deeper sections cut at 50 μm intervals. Endometriosis was absent\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nfrom all sections from patients 8 and 10, but endometrial-type stroma was detected in deeper\nlevels for the perirectal fat specimen from patient 9. Of the remaining 16 specimens, 12 had\nepithelium, stroma and hemosiderin observed in at least one part. The remaining 4 specimens\nhad two diagnostic components present in each. Endometrial-type stroma was present in all\nextra-ovarian lesions (Table S3).\nFigure 1. A cellular atlas of human endometriosis. (A) Patient cohort and specimens\nprofiled. (B) Histologic and macroscopic features of specimens from patient 9.\nA single-cell atlas of endometriosis\nTo map the cellular and transcriptional features of endometrioma and endometriosis, we profiled\nthese specimens using droplet based single cell RNA-sequencing (RNA-seq). In five instances,\nspecimens with low cell yields were combined with specimens from similar anatomic locations in\nthe same patient, to give a total of 32 samples sequenced - 7 endometriomas, 19 extra-\nendometriosis lesions, 4 eutopic endometrium specimens and 2 normal ovary tissues. The ovary\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nspecimen from patient 3 and left pelvic side wall specimen from patient 10 failed to meet quality\ncontrol thresholds and were removed from the subsequent analyses (see Methods). A total of\n257,255 individual cells were profiled, with a total of 6,606,535,712 reads sequenced (Table S4).\nAfter filtering out doublets, any cells with >20% mitochondrial transcripts or <200 genes\ndetected, 207,949 cells remained for analysis. The median number of captured cells was 7,498\nfor eutopic endometrium (range = 3,408-20,833), 6,265 for endometrioma (range = 1,112-\n14,943), 5,038 for extra-ovarian endometriosis (range = 1,249-11,673). The observed cell\nnumber correlated with the targeted cell number (p=0.001, Spearman's r=0.65), and was not\nsignificantly different between groups (p=0.084, one-way analysis of variance) (Figure S2A,B).\nNumbers of reads and genes per cell were also not significantly different between groups and did\nnot differ based on whether specimens were cryopreserved prior to capture and sequencing\n(Figure S2A-D).\nScRNA-seq data were integrated using Harmony, regressing out the effects of inter-patient\nvariability, sequencing batch and mitochondrial RNA content (Korsunsky et al., 2019) (Figure 2A-\nE). Cell cycle genes were not present in any of the first 20 principal components and so cell cycle\nregression was not performed. We identified 96 clusters and developed a systematic pipeline for\ncell-type assignment (see Methods and Figure S2E). First, we performed differential gene\nexpression analysis to identify genes overexpressed in each cluster relative to all other clusters\n(log2 fold change = 0.2; p < 0.05) (Figure S2F and S2G). We then defined a set of rules for cell-\ntype identification based on known hierarchies of canonical cell-type specific markers, for\nexample ACTA2 in the absence of other fibroblast markers denotes smooth muscle cells, but\nwhen co-expressed with fibroblast markers denotes activated fibroblasts (Methods, Figure S2E).\nFor 84 out of 96 clusters we were able to assign cell identities using this approach (Figure 2B-E).\nFor the remaining 12 clusters that did not overexpress canonical marker genes for any cell type,\nwe calculated pairwise comparisons of all clusters and assigned cell identities based on the most\ncorrelated cluster (Figure S2H). Correlation values for cell type assignment ranged from 0.77-\n0.97 (Pearson's correlations) and were significantly higher compared to random pairwise\ncorrelations (average random correlation r=0.019; Figure S2I), providing confidence that cell\ntype assignment can be achieved using this approach.\nFibroblasts and stromal cells, identified by expression of FAP, COL1A1 and PDGFRA/B, were the\nmost abundant cell type present (n= 74,111 cells, 35.64% of cells remaining after filtering)\n(Figure 2E-F). T-cells/NKT-cells were the second most prevalent cell type present, comprising\n69,703 (33.5% of cells). Keratin (KRT7, KRT8, KRT10, KRT18, KRT19) or EPCAM-positive epithelial\ncells (n=22,772 cells) represented 10.9% of the total population (Figure 2E-F). Other less\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\ncommon cell types included myeloid cells (n= 13,003 cells, 6.25% of the total population),\nsmooth muscle cells (n=7,970 cells, 3.83% of the total population), endothelial cells (n=6,586\ncells, 3.17% of the total population) and some erythrocytes that persisted after red blood cell\ndepletion (n= 4,857 cells, 2.33% of the total population). Other immune cells included B and\nPlasma cells (n=7,452 cells, 3.58% of total population) and mast cells (n=1,495 cells, 0.72% of\ntotal population). Fibroblasts and smooth muscle cells had the greatest number of genes\ndetected (average = 1,571.6 and 1,703.4 genes/cell), and as expected, erythrocytes had the\nlowest number of genes detected per cell (729.8 genes/cell). We compared the frequency of\neach cell type across the major tissue-type classes - endometrioma, extra-ovarian endometriosis,\neutopic endometrium, tissue with no evidence of endometriosis and unaffected ovary, to identify\ndeviations from the null distribution (the overall proportion of each cell type in the data set)\n(Figure 2G). Eutopic endometrium tissues were enriched 4.3-fold and 2-fold for epithelial cells\nand erythrocytes, respectively (p=2.2x10-16 and p=2.2x10-16, respectively, Chi-squared Test),\nand the large number of epithelial cells assigned to patient 11 (captured cell number = 38,400\nafter filtering) only partially explained the epithelial cell enrichment (after removing patient 11\neutopic epithelial cells the enrichment is 2.15 fold). Endometrioma tissues were depleted 8.5-fold\nfor epithelial cells but enriched 2-fold for B and plasma cells. Extra-ovarian endometriosis was\nenriched 1.6-fold for mast cells, 1.3-fold for myeloid cells and 1.3-fold for T/NK-T cells (Figure 2G).\nOverall, each class of tissue type had a significantly different composition of cell types compared\nto the other four classes (p=2.2x10-16, Chi-squared Test).\nWe examined the overall relationships between specimens. In principal component analysis (PCA)\nendometriomas formed a group largely separated by PC1 (10.22% variance explained) (Figure\n2H, Figure S2J) with separation of extra-ovarian endometriosis becoming evident in PC4 (7.45%\nof variance explained). We calculated pairwise correlations between all the specimens in the\ncohort based on the cell-type composition and performed unsupervised clustering. Samples\nseparated into 3 major clusters (Figure 2I), Cluster 1 contained the normal ovary specimen and 2\nof 4 eutopic endometrium samples. Cluster 2 contained the majority of the extra-ovarian\nendometriosis specimens (10 of 14) plus 3 of 4 specimens where no endometriosis was detected.\nAll 7 endometrioma specimens were found in Cluster 3, along with 4 of 14 extra-ovarian\nendometriosis samples, 2 eutopic endometrium samples and the remaining specimen with no\nendometriosis detected (rectal serosa from patient 8). In some instances, samples from the same\npatient were highly correlated. For example, in patient 8, a left endometrioma was highly\ncorrelated with a nodule on the left uterosacral ligament (R2=0.77, Pearson correlation) and\nbilateral endometrioma samples from patient 2 were highly correlated (R2=0.83, Pearson\ncorrelation). In patient 9, all the extra-ovarian lesions were tightly correlated and were found\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nwithin cluster 2, whereas the endometrioma from this patient was in cluster 3 and bore the most\nsimilarity to an endometrioma in patient 7 (R2=0.83, Pearson correlation).\nFigure 2. The cellular landscapes of endometrioma, peritoneal endometriosis,\nunaffected peritoneum, eutopic endometrium and unaffected ovary. (A) Uniform\nManifold Approximation and Projection (UMAP) visualization of all sequenced cells (after filtering\nfor quality) from 32 samples representing five major tissue-type classes. (B) UMAP plot with 96\nclusters (harmony reduction using 25 dimensions and resolution of 3). (C) Major cell types\nidentified, UMAP representation. (D) Three-dimensional UMAP representation. (E) Expression of\nrepresentative markers across 9 major cell types. (F) Representation of each major class within\neach major cell type group, contribution of each patient to each group, frequencies of each cell\ntype and number of genes detected in each cell type. ‘Total’ column represents the proportion of\neach patient or class in the major cell type overall, under the null hypothesis of no enrichment in\na specific cluster. Box and whisker plots, boxes denote the interquartile range, bar denotes\nmedian number of genes detected per cell. The limits of the whiskers represent 1.5 * IQR\n(interquartile range) and outlier cells are indicated with individual dots. (G) Fold enrichment and\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\ndepletion of each cell type across the five classes. (H) Principal component analysis. (I)\nCorrelation, based on cell type frequencies, across all specimens profiled by scRNA-seq (Pearson\ncorrelation). We used the agglomeration method “complete” and “Euclidean” distance as\nclustering parameters.\nEpithelial components of eutopic and ectopic endometrium\nEpithelial and stromal cells are the two major structural cell types present in endometriosis\nlesions, and so we characterized the heterogeneity within these populations. First, we isolated\nthe 6,850 epithelial cells and re-clustered to identify 16 clusters that exhibited different patterns\nof canonical marker gene expression and were differentially represented across the tissue types\n(Figure 3A-C). Six of these were clusters of endometrial-type epithelial cells (EnEpi_1 to 6) that\nexhibited heterogenous expression of hormone receptors ESR1 and PGR and six were\nmesothelial cell clusters (Meso_1 to Meso_6) expressing WT1 and/or other mesothelial cell\nmarkers PDPN, DES and CALB2. Four clusters did not express canonical epithelial or mesothelial\nmarkers and were denoted as ‘unclassified’ epithelial clusters (UnEpi). UnEpi_1 was mostly\nderived from eutopic endometrium and expressed high levels of hemoglobin genes (HBB, HBA1\nand HBA2), low PGR and ESR1 as well as differentiated epithelial cell markers (TFF3 and SLPI;\nTable S5) and may represent luteal-phase endometrial-type epithelium. UnEpi_2 only had one\ndefining marker (C11orf96; Table S5). UnEpi_3 and _4 were rare populations that represent\npossible immune cell contamination (Table S5).\nThe six clusters of endometrial-type epithelial cells could be broadly categorized based on\nkeratin expression: EnEpi_1, 2, 4 and 6 expressed KRT10 and/or KRT17, with EnEpi_6 also co-\nexpressing low levels of KRT8 and KRT18. EnEpi_3 and 5 exhibited low but pervasive expression\nof KRT8 and KRT18. EnEpi_1, 2 and 5 were predominantly cells from endometriosis, with 93%,\n88% and 97% of cells in these clusters derived from ectopic endometrium. The top pathways\nenriched in these clusters were distinct, with EnEpi_1 associated extracellular matrix remodeling,\nEnEpi_2 with immune cell interactions, and EnEpi_5 exhibiting a strong secretory cell signature.\nEnEpi_3, a second secretory cluster, were mainly derived from eutopic endometrium (56%) and\n42% endometriosis (42%). EnEpi_4 and EnEpi_6 were EPCAM and PAX8-positive endometrial-type\nepithelial clusters that were detected in eutopic endometrium but were enriched 4.4 and 5.6-fold\nin peritoneal lesions. EnEpi_4 was a population of FOXJ1+ ciliated epithelial cells enriched for the\n‘cilium assembly’ pathway (P = 4.4x10-20) (Table S6). We validated EnEpi_4 by performing\nimmunofluorescent staining of eutopic endometrium and endometriosis for CAPS, identifying\nCAPS-positive ciliated cells in both tissue types (Figure 3D). EnEpi_6 expressed high levels of\nEPCAM, LGR5, low expression of keratins plus intermediate levels of PAX8, ESR1 and PGR and\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nwas enriched for cell-cell communication and splicing pathways (Figure 3A-C, Table S6). We note\nthat endometrial-type epithelial cells were detected in all the four specimens that had no\nendometriosis upon pathologic review, ranging from 24 and 26 cells in the two NED specimens\nfrom patient 8, to 48 and 232 cells in the NED specimens from patient 10, suggesting that,\nparticularly for the latter patient, the portion used for scRNA-seq may have contained\nendometriosis epithelium. We implemented SCENIC to identify the main transcription factors\ndriving the transcriptional programs of each cluster (Figure 3E). Unsupervised clustering was\nperformed based on similarity of transcriptional regulon activity. The first group contained\nEnEpi_1 which was characterized by activation of IRF1, TCF4, CEBPB and FOS regulons. EnEpi_4\nwas characterized by active RFX2, RFX3, HOXB6 and BCLAF1 regulons. EnEpi_3 and EnEpi_5\nwere characterized by GTF2F1 activity. EnEpi_2 also had activation of the GTF2F1 regulon as well\nas RUNX3, ETS1 and CREM. Finally, EnEpi_6 exhibited strong activation of KLF6, ATF3, SOX17,\nREL and MAFF.\nEndometrial-type epithelium exhibited marked differences in gene expression in the context of\nendometrioma, eutopic endometrium or extra-ovarian endometriosis. The 220 genes\noverexpressed in endometrioma (log2 FC > 0 and adjusted p < 0.05) included progestagen\nassociated endometrial protein (PAEP, FC=1.44, adjusted P = 9.9x10-33) and were enriched in\npathways associated with immune cells interactions, including PD-1 signaling (P = 2.9x10-6), and\nhypoxic insult, as reflected by an enrichment of ‘Detoxification of Reactive Oxygen Species’ (P =\n6.1x10-3) (Figure 3F,G, Table S7, Table S8). The latter pathway was enriched in both\nendometriomas and peritoneal endometriosis, as was RHO GTPase effectors (P = 9.2x10-4),\nneutrophil degranulation (P = 1.7x10-10) and cellular responses to external stimuli (P = 3.1x10-3),\nhighlighting interactions with microenvironmental cells and stimuli as unifying hallmarks of\nendometriosis independent of location. Ciliated cell pathways and markers (including CAPS) were\nspecifically enriched in peritoneal endometriosis, mirroring the enrichment of the EnEpi_4 ciliated\nepithelial cell cluster in this group. Finally, translation pathways reflecting the differentiated\nsecretory epithelial clusters were enriched in eutopic endometrium.\nSomatic mutations in “cancer driver” genes including ARID1A and KRAS (Anglesio et al., 2017;\nLac et al., 2019; Suda et al., 2020) are known to occur in endometriosis, and so we sought to\ndetermine the transcriptional consequences of these mutations when they occur in vivo.\nFormalin-fixed, paraffin embedded tissue sections adjacent to the specimens used for scRNA-seq\nwere available for 25 out of 32 specimens, of these 21 had sufficient epithelial content to\nquantify ARID1A expression using immunohistochemistry as a mutation surrogate and 10\nspecimens had sufficient epithelial content to successfully identify KRAS mutations by digital\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\ndroplet PCR (ddPCR). A summary of the mutation profiling is shown in Figure 3H. The two\nendometrium specimens profiled were wild-type for both genes. Six patients had evidence of at\nleast one somatic mutation in one or more endometriosis lesions. Two endometriomas and four\nextra-ovarian lesions exhibited heterogenous ARID1A expression indicative of heterozygous\nARID1A loss of function mutations. Two of the five endometriomas and two of the five extra-\novarian lesions subjected to KRAS genotyping harbored mutations at codon 12. ARID1A and\nKRAS expression were highest in EnEpi_4 and _6, the two clusters enriched in endometriosis\nspecimens (Figure S3C). When we compared expression by mutation status, ARID1A mRNA\nexpression was 2.1-fold lower in specimens where endometrial-type epithelium exhibited\nheterogenous protein staining compared to cases with strong ARID1A staining (Figure 3I,J; Figure\nS3E,F). By contrast, KRAS gene expression was only marginally elevated (by around 5%) in\nmutant compared to wild-type cells (Figure 3K). We then performed differential expression\nanalysis to identify genes (Figure 3K, L) and pathways (Figure S3G,H) associated with mutation\nof KRAS or ARID1A, after removing genes associated with class to minimize the impact of this\nconfounding variable. KRAS mutation was associated with 302 differentially expressed genes\n(log2 FC > 0, adjusted P < 0.05) including known KRAS target gene S100 calcium-binding protein\nA1 (S100A1; log2 FC= 0.61, adjusted P = 9.16x10-33) and transcriptional regulator nuclear protein\n1 (NUPR1; log2 FC= 0.70, adjusted P = 2.05x10-29)(Figure 3K, Table S10). 118 genes were\ndifferentially expressed in endometrial-type epithelial cells associated with heterogenous ARID1A\nstaining compared to those with homogenous positive staining (log2 FC > 0, adjusted P <\n0.05)(Figure 3L, Table S11). The most upregulated gene associated with ARID1A loss of function\nwas known ARID1A target gene IGFBP2 (log2 FC= 0.62, adjusted P = 2.24x10-68) (Suryo\nRahmanto et al., 2020). Additional targets included lysosome-associated protein\ntransmembrane-4β (LAPTM4B; log2 FC= 0.51, P = 1.24x10-44) and SRY-box 17 (SOX17; log2 FC=\n0.50, adjusted P = 1.51x10-33, Figure 3L). We have recently identified SOX17 as a novel marker\nof secretory fallopian tube epithelia and high-grade serous ovarian cancer, where it positively\nregulates angiogenesis (Chaves-Moreira et al., 2020; Dinh et al., 2021; Reddy et al., 2019).\nSOX17 protein expression was validated in the same tissues using immunohistochemistry\nperformed on 6 specimens with heterogenous ARID1A staining and 2 specimens with positive\nARID1A staining. The proportion of endometrial-type epithelial cells expressing SOX17 was\nhigher in lesions with heterogeneous ARID1A staining. Heterogeneous ARID1A staining was also\nassociated with moderate/strong SOX17 staining, whereas lesions with positive ARID1A staining\nexhibited weak SOX17 staining (Figure 3M & N). We therefore interrogated the endothelial cell\ncompartment associated with ARID1A mutation status and found that endothelial cells proximal\nto ARID1A-mutant endometriosis lesions up-regulate expression of a select handful of genes\nincluding ferritin heavy chain (FTH1, log2 FC= 0.29, adjusted P = 8.18x10-44), parathymosin\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\n(PTMS, log2 FC= 0.27, adjusted P = 1.79x10-29), Y box binding protein 1 (YBX1, log2 FC= 0.27,\nadjusted P = 2.04x10-29), galectin-1 (LGASL1, log2 FC= 0.28, adjusted P = 1.58x10-14) and\nmetallothionein 1X (MT1X, log2 FC= 0.25, adjusted P = 1.85x10-10)(Figure 3O).\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nFigure 3. Epithelial components of eutopic endometrium and endometriosis. (A) UMAP\nof all epithelial cells, (B) frequency of each cluster by class. The total column represents the\ndistribution of cells or class in all epithelial cells. (C) Marker gene expression and pathway\nanalysis. (D) CAPS expression, immunohistochemical staining of eutopic endometrium and\nendometriosis. (E) Transcription factor regulon analyses using SCENIC. (F) Differential gene\nexpression in endometrial-type epithelium in the context of endometrioma, eutopic endometrium\nor extra-ovarian endometriosis (pval < 0.05 and log2 FC > 1) (G) Pathway analysis, endometrial-\ntype epithelium in the context of endometrioma, eutopic endometrium or extra-ovarian\nendometriosis. (H) Summary of ARID1A staining status and KRAS mutations detected in each\nlesion. NP, not profiled due to insufficient epithelial material available in the specimen. (I)\nExpression of ARID1A and KRAS mRNA by mutation state. (J) ARID1A immunostaining in a\nrepresentative endometriosis lesion with heterogenous staining, positive staining for ARID1A is\nshown with the black arrow, negative epithelium is shown with the arrowhead. Posterior cul-de-\nsac lesion from Patient 5 is shown. (K) Differential gene expression in KRAS mutant versus wild-\ntype endometrial-type epithelium (P < 0.05 and log2 FC = 0.6). (L) Differential gene expression in\nARID1A heterogeneously staining versus positive endometrial-type epithelium (P < 0.05 and log2\nFC = 0.5). (M) SOX17 staining in ARID1A positive and ARID1A heterogenous staining\nendometriomas from patient 2. (N) Summary of SOX17 staining by ARID1A staining status. (O)\nDifferential gene expression in endothelial cells associated with positive or heterogenous ARID1A\nstaining in endometrial-type epithelium.\nEndometriomata impact mesothelial differentiation\nFive epithelial clusters were defined as mesothelial due to expression of canonical mesothelial\ncell markers (WT1, CALB2, DES and/or PDPN). Mesothelia are a specialized type of epithelium\nthat lines the peritoneum, the pleurae and pericardium (Figure 3B,C). Modified peritoneal\nmesothelium also covers the ovary as a monolayer termed the ovarian surface epithelium, which\ncan also become trapped within ovarian inclusion cysts that develop following ovulation\n(Auersperg et al., 2001). We examined the pseudotime relationships between mesothelial cells\nclusters with Monocle3 (Qiu et al., 2017). Ovarian mesothelial cells associated with\nendometrioma were enriched at an earlier timepoint in the trajectory and exhibited lower\nexpression of differentiated markers compared to mesothelial cells from unaffected ovary tissue\n(Figure 4A). Peritoneal mesothelium sat later in the pseudotime space and was more\ndifferentiated than unaffected ovary or endometrioma-associated mesothelium. Meso_5 was\nderived from endometrioma and peritoneal lesions (both with and without pathologic\nconfirmation of endometriosis). This cluster expressed PDPN and WT1 but not CALB2 or DES, and\nwas the only mesothelial cluster that expressed ESR1 and PGR. Meso_5 expressed ACTA2 and\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nalso exhibited a markedly different keratin profile, expressing KRT10 but not KRT7, KRT8, KRT18\nand KRT19, the dominant keratins in the differentiated mesothelial clusters (Figure 3B,C). This\nsuggests that mesothelial cells associated with endometrioma tend to exhibit a modified\nuncommitted mesothelial-mesenchymal phenotype. Novel markers overexpressed by\nendometrioma-associated mesothelial cells included IGFBP1, IGFBP2, IGFBP3, RBP1, MEG3 and\nSOX4 (Figure 4B, C). Endometrioma-associated mesothelial cells exhibited lower expression of\nmarkers including MSLN, SLPI, PRG4, ADIRF and TFPI2 when compared with mesothelial cells\nassociated with extra-ovarian endometriosis (Figure 4A-C).\nFigure 4. Endometriosis affects mesothelial differentiation. (A) Pseudotime analysis for\nmesothelial cells (B) Differential gene expression in mesothelial cells adjacent to endometrioma\ncompared to extra-ovarian endometriosis (labelled genes are those where adjusted P < 0.05 and\nabsolute log2 FC > 0.8). (C) RBP1, SOX4 and KRT10 are overexpressed in mesothelial cells\nproximal to endometriomata and PRG4, ADIRF and TFPI2 overexpressed in mesothelial cells\nassociated with extra-ovarian endometriosis.\nMesenchymal components of eutopic endometrium, endometrioma and endometriosis\nWe turned our attention to the large population of mesenchymal cells in the data set. Sub-\nclustering of the 74,111 mesenchymal cells identified 20 distinct clusters that could be stratified\ninto three major groups - ‘host’ organ ovarian fibroblasts (OvF), ‘host’ organ peritoneal\nfibroblasts (PerF) and endometrial-type stroma (EnS) (Figure 5A-D, Table S11). Clusters OvF_1\nand OvF_2 were largely exclusive to the unaffected ovary (Figure 5A). Cluster OvF_1 was the\nlargest fibroblast cluster (12,910 cells), exhibiting relatively low levels of canonical fibroblast\nmarkers DCN and COL1A1. Genes over/under-expressed in this cluster were associated with\nprotein production and translation (Table S12). OvF_2 was a rare cluster (186 cells) with high\nexpression of DES and ACTA2 (Figure 5B,E). Peritoneal fibroblasts comprised 11 clusters that\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\ntended to have higher expression of THY1, COL1A1 and DCN compared to ovarian and\nendometrial fibroblasts (Figure 5E). Peritoneal fibroblasts could be broadly categorized into\nPDGFRA+ clusters (PerF_3, 4, 5, 9 and 10), PDGFRB+ clusters (PerF_1 and 2), dual\nPDGFRA+/PDGFRB+ clusters (PerF_7), and PDGFRA/B weak/absent (PerF_6, 8 and 11).\nSeven clusters were identified as endometrial-type stroma (EnS) due to expression of known\nmarkers of endometrial stromal cells MME and or IFITM1, and/or predominant occurrence in\neutopic endometrium (Figure 5E). We implemented Monocle3 to infer a pseudotime trajectory\nrooted in EnS_4, which expressed the highest levels of stem cell marker CD44 and was derived\npredominantly from eutopic endometrium (90%). This created a bifurcated trajectory (Figure 5F).\nCluster EnS_6 represented one endpoint and was composed of differentiated endometrial-type\nstroma which exhibited the highest expression of IFITM1 of all the EnS clusters, and moderate\nexpression of ESR1 and PGR. EnS clusters 1, 5 and 7 resided at intermediate pseudotime points\nin the trajectory. EnS_1 was the most abundant endometrial-type stroma cluster; 43% of cells in\nthis cluster were derived from eutopic endometrium and 55% were derived from endometriosis.\nEnS_1 expressed MME, PGR, ESR1 and PDGFRB and was enriched for heme-response pathways\n(Table S12). EnS clusters 5 (associated with extracellular matrix remodeling) and 7 (associated\nwith stress response pathways) were predominantly composed of cells from peritoneal tissues\n(96 and 89%, respectively) and expressed the highest levels of PGR and ESR1. These two\nclusters were both observed in peritoneal lesions both with and without pathologic confirmation\nof endometriosis, again suggesting the portion subjected to scRNA-seq did contain endometriosis.\nBiomarkers expressed by EnS_5 include progesterone receptor membrane component 1\n(PGRMC1) and receptor activity modifying protein 1 (RAMP1). EnS_5 and 7 both overexpressed\nmatrix metalloproteinase-11 (MMP11). EnS clusters 2 and 3 represented alternative trajectory\nendpoints, both expressed IFITM1 and modest expression of PDGFRA and ACTA2 respectively\n(Figure 5F). Cells in these clusters (98 and 96% respectively) were derived almost exclusively\nfrom endometrioma samples. EnS_2 expressed an immunomodulatory set of genes including\nCXCL12 and CXCL2 (Figure 5G). Given the indication that endometrioma-associated EnS clusters\nexhibit more activation of pro-inflammatory pathways, we tested for gene signatures that\ndiffered across endometrial-type stroma by site. Endometrial-type stroma within endometriomas\noverexpressed Phospholipase A2 (PLA2G2A), which when secreted plays a role in inflammation\nand neurological disorders. Innate immunity components complement proteins C3 and C7 were\nalso overexpressed along with transcription factor JUNB and bone marrow stromal antigen 2\n(BST2) (Figure 5H, Table S13). The noncoding RNA NEAT1, transcription factor JUND and heat\nshock protein DNAJB1 were abundantly expressed in endometrial stroma in eutopic endometrium\nbut downregulated in endometriosis. Peritoneal endometriosis was associated with matrix gla\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nprotein (MGP), hemoglobin genes (HBB and HBA2) and tumor suppression and adhesion\nmodulator SPARC-like protein 1 (SPARCL1) (Figure 5H). At the pathway level, eutopic\nendometrium and extra-ovarian endometriosis were similar, with enrichment of extracellular\nmatrix organization pathways (Figure 5I, Table S14). Endometrial type stroma within\nendometriomas was enriched for translation pathways (P=2.3x10-65 and 3.6x10-64), indicative of\na secretory phenotype (Figure 5I). We noted overlap in the genes differentially regulated in\nendometrial-type epithelium and stroma associated with eutopic endometrium or endometrioma\nbut not extra-ovarian endometriosis (Figures 3 and 5), for example, in endometriomas, C3 and\nC7 are highly expressed by both types and complement pathways were enriched (Tables S7, S8,\nS13 & S14) suggesting a coordinated transcriptional response occurs (Figure 5J).\nTranscription factor regulon analyses identified an enrichment of FOXO1, XBP1, MAFF and JUND\nregulons in the putative endometrial-type stroma progenitors. Inflammatory EnS clusters (EnS_2\nand EnS_3) and EnS_1 were associated with activation of pro-differentiation factors NFIA and\nNFIB (Chen et al., 2017) with EnS_2 showing high activation of the CREM regulon. EnS_3 shared\nactivation of FOS, IRF1 and FOSB with differentiated endometrial stroma clusters EnS_6 and\nEnS_7. EnS_5-7 exhibited the highest expression of PGR and ESR1 and were associated with\nHOXA10/11 and SOX4 activation (Figure 5K). SOX4 expression was particularly high in EnS_5 and\nEnS_6 (Figure 5L).\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nFigure 5. Signatures of endometrial-type stroma and mesenchymal cells associated\nwith endometriosis. UMAP of mesenchymal cells in (A) all tissue types (B) unaffected ovary\nfrom an endometriosis-free patient, (C) peritoneal fibroblasts (D) endometrial-type stroma, (E)\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nMarker gene expression and cluster frequencies. (F) Pseudotime analysis performed using\nMonocle3, and (G) marker gene expression. (H) Differential gene expression in endometrial-type\nstroma in the context of endometrioma, eutopic endometrium or extra-ovarian endometriosis\n(log2 fold change ≥ 0.8, P < 0.05) (I) Pathway analysis, endometrial-type stroma in the context of\nendometrioma, eutopic endometrium or extra-ovarian endometriosis. (J) Heatmap, coordinate\nexpression of genes in EnEpi and EnS across eutopic endometrium, endometrioma and extra-\novarian endometriosis. Ordering of rows and columns (genes) are supervised, genes ranked in\norder of decreasing expression within each class. (K) Transcription factor regulon analyses using\nSCENIC. (L) Differential expression of candidate TFs across EnS clusters.\nDeep and superficial peritoneal endometriosis involves transcriptional reprogramming\nin endometrial-type epithelium and mesenchymal cells\nSurgical images were available for 24 out of 25 lesions and were reviewed by three expert\nminimally invasive gynecology surgeons, and consensus classifications of endometriosis subtype\nwere assigned for extra-ovarian lesions (Table 1, Figure 1, Figure S1). Five lesions were\ncategorized as deep infiltrating endometriosis and 8 were classified as superficial endometriosis\n(n=8). We asked whether single cell transcriptome profiles indicate that deep and superficial\nendometriosis are two biologically distinct entities. We integrated data sets for all 13 extra-\novarian samples to create a data set of 69,131 individual cells. We then simulated a background\ndata set based on the frequencies of the different cell types in the actual data and measured\nEuclidian distance between deep and superficial lesions, based on frequencies of all cell types\n(Figure S4C). There was a modest suggestion that deep and superficial endometriosis were\nsignificantly different when we considered all cell types present (P=0.046, compared to a\nbackground of 1,000 randomly generated data sets, pnorm R function using mean and standard\ndeviation from background) (Figure 6A). By contrast, hemorrhage and fibrosis were also not\nassociated with cellular composition (P=0.56 and P=0.82, respectively).\nWe posited that differences in deep and superficial endometriosis may more apparent in the\nexpression signature of cells rather than overall cellular composition of lesions and so we tested\nfor associations between endometrial-type epithelial or fibroblast gene expression and surgical\nannotations. Epithelial cells in deep infiltrating endometriosis were associated with upregulated\nexpression of mitochondrial genes (MT-ND4, MT-CO1 and MT-ATP6) and Nuclear Receptor\nSubfamily 4 Group A Member 1 (NR4A1) (Figure 6B,C; Table S15). Intelectin-1 (ITLN1) was\noverexpressed in superficial lesions and both at the gene and pathway level, pathways\nassociated with exposure to heme were enriched in superficial lesions (Figure 6B,C). Senescence\nand inflammatory pathways including interleukin-4 and -13 signaling were enriched in deep\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nendometriosis (P=1.1x0-2 and P=4.3x10-4, Figure 6C). Differential gene expression in deep\nversus superficial lesions were more marked in fibroblasts compared to epithelial cells (although\nthis is likely due in part to the larger number of cells). NR4A1 was overexpressed in fibroblasts\nassociated with deep endometriosis (log2 FC=0.58, P=4.1x10-42), plus BTG2, CEBPB and CEPBD,\nSERPINE1 and YAP target gene CTGF. Fibroblasts associated with superficial lesions\noverexpressed MMP11, SFRP4 and PI16 (Figure 6D, Table S16). Stress associated transcription\nwas upregulated in fibroblasts in superficial endometriosis, while no pathways were enriched in\nfibroblasts in deep endometriosis lesions (Figure 6E).\nFigure 6. Molecular correlates of peritoneal endometriosis subtypes. (A) Euclidian\ndistance between samples, compared to a simulated background distribution. Deep/superficial\nstatus, presence/absence of hemorrhage or fibrosis were all surveyed. (B) Differentially\nexpressed genes (log2 fold change ≥ 0.5, P < 0.05) and (C) pathway enrichment in endometrial-\ntype epithelial cells associated with deep or superficial endometriosis. (D) Differentially\nexpressed genes (log2 fold change ≥ 0.4, P < 0.05) and (E) pathway enrichment in fibroblasts\nassociated with deep or superficial endometriosis (F) Expression of BTG2, PDK4 and NR4A1\nmarkers were enriched in deep endometriosis and expression of SFRP4, PI16 and MMP11\nenriched in superficial endometriosis.\nLinking endometriosis signatures to endometriosis-associated ovarian cancers\nEndometrioid and clear cell ovarian cancers are associated with a personal history of\nendometriosis, suggesting that endometrial-type epithelial cells may be precursors for these\ntumors. We implemented MuSiC (Wang et al., 2019) to test whether cluster-specific signatures of\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nendometrial-type epithelium were enriched in these tumor types. Across three independent data\nsets, clear cell and endometrioid ovarian cancers (Tan et al., 2019; Tothill et al., 2008; Winterhoff\net al., 2016) consistently showed a strong enrichment of signatures for ciliated cluster EnEpi_4,\nand to a lesser extent, cell-cell communication-associated cluster EpEpi_6 (Figure 7A-C). Both of\nthese clusters of endometrial-type epithelial cells were enriched in endometriosis lesions relative\nto eutopic endometrium (Figure 3B). There was no evidence of enrichment for the signatures of\nendometrial-type epithelial cell clusters that were derived from eutopic endometrium.\nFigure 7. Deconvoluting endometriosis-associated ovarian cancers with single cell\nendometriosis signatures. Deconvolution of (A) 25 clear cell ovarian cancers (CCOC, 24\nprimary tumor specimens, one ascites sample) (B) 14 endometrioid ovarian cancers (EnOC) and\n(C) 24 CCOC and 35 EnOC tumors, based on signatures of 6 endometrial-type epithelial\nsubclusters.\nDiscussion\nEndometriosis is a common but poorly studied condition of unknown etiology and poorly\ncharacterized pathogenesis. We generated a cellular atlas of endometriosis incorporating over 30\ntissues from 11 patients, analyzing around 250,000 individual cells. In doing so we catalogued\nthe epithelial component but also the stromal cells and immune cells in the microenvironment\nwhich are not passive bystanders but play active roles in endometriosis pathogenesis. We\nleveraged the data to ask some key clinical questions. First, these analyses add to a growing\nbody of literature to suggest that endometriomas and peritoneal lesions are two distinct disease\nentities. Although heterogeneity across patients was observed, for some patients, cellular\ncomposition of lesions was highly correlated across sites. This composition may be influenced by\nunderlying clonal expansion of endometriosis epithelium as it transits from endometrium and\ncolonized ectopic sites where clonality has also been observed across lesions types through DNA-\nbased analyses (Moore et al., 2020; Praetorius et al., 2021; Suda et al., 2018).\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nDespite the small size of most endometriosis lesions, in all except one endometriosis specimen\nwe were able to capture the endometrial-type epithelial and stromal components that are\ndiagnostic for the disease. First, we asked whether the molecular profiles of endometrial-type\nepithelium and stroma differ by site. In addition to observing striking differences in gene\nexpression by context, we noted convergent expression of genes and pathways in endometrial-\ntype epithelium and stroma in the context of endometrioma or extra-ovarian endometriosis and\nalso in comparisons of deep and superficial peritoneal lesions, consistent with a recent multi-\norgan analysis of structural cells that highlighted pervasive organ-specific patterns of gene\nexpression (Krausgruber et al., 2020). Both endometrial-type epithelium and endometrial-type\nstroma exhibited greater activation of immune pathways in the context of endometriomata, with\ncomplement proteins C3 and C7 expressed by both cell types, indicative of dysregulated innate\nimmunity. This may be due to elevated apoptosis in the hypoxic microenvironment in an\nendometrioma, since cells undergoing apoptosis can activate the complement pathway, where\ndying cells opsonized by complement components help phagocytic cells such as macrophages\ndispose of the apoptotic cells. Moreover, serum and peritoneal fluid levels of C3 are elevated in\nwomen with endometriosis compared to controls (Hasan et al., 2019; Kabut et al., 2007),\nalthough endometrioma-specific analyses have not yet been performed. Interestingly, lower\nlevels of the inactive iC3b are observed in both the serum and peritoneal fluid of endometriosis\npatients, yet higher levels of C3c and SC5b-9, possibly due to cleavage of iC3b by regulatory\nprotein Factor I or complement receptor 3, releasing C3c leaving behind C3dg bound to cells.\nBound C3dg can still interact with CR3/CD11b found on phagocytes to stimulate phagocytosis,\nand with CR2/CD21 found on B cells which can augment signaling through the B cell receptor\n(Ricklin et al., 2016). Whether the interaction of complement and CD21 on B cells contributes to\nthe increase in autoantibodies seen in endometriosis patients is yet to be determined, but high\nexpression of BST2/CD317 and CXCL12 on endometrial-type stroma associated with\nendometriomata may contribute to altered innate immunity in this context specifically.\nIt has been proposed that two subtypes of peritoneal disease exist - superficial peritoneal\ndisease and deep infiltrating endometriosis (Brosens et al., 1993), and these categories are\nwidely used in endometriosis research. We asked whether these subtypes were supported by the\ncellular and molecular profiles of peritoneal endometriosis categorized as deeply infiltrating or\nsuperficial. The overall cellular landscape of endometriosis did associate with deep/superficial\nstatus, but not fibrosis or hemorrhage, although we note we were powered only to detect very\nstrong effects. We also observed that both endometrial-type epithelial cells and fibroblasts\nexhibited marked differential gene expression associated with deep or superficial status. These\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nresults may suggest that deep and superficial disease are not distinct entities but parts of the\nsame disease continuum resulting from transcriptional reprogramming, consistent with recent\ngenomic analyses (Praetorius et al., 2021). NR4A1 was identified as an epithelial and fibroblastic\nmarker of deep endometriosis and has been previously implicated in endometriosis-associated\nfibrosis (Zeng et al., 2018). NR4A1 has been proposed as a therapeutic target for endometriosis\n(Mohankumar et al., 2020), although the precise functions of their protein in disease\npathogenesis have yet to be elucidated. NR4A orphan nuclear receptors facilitate transcriptional\nand posttranscriptional responses to changes in the cellular microenvironment and may\ntherefore serve as a hub for the altered epithelial-stromal interactions specific to deep infiltrating\ndisease (Crean and Murphy, 2021).\nOur study findings are consistent with the prevailing notion that bidirectional interactions\nbetween endometrial-type epithelium/stroma and the local microenvironment play critical roles\nin endometriosis pathogenesis. We discovered context-specific features of mesothelial cells,\nwhich were less differentiated when associated with endometrioma and some peritoneal lesions.\nA set of atypical mesothelial cells that co-express mesothelial markers (WT1 and PDPN),\nhormone receptors (ESR1 and PGR), KRT10 and ACTA2 were enriched in endometrioma\nspecimens. These altered mesothelial cells may contribute to adhesion formation, for example\ndue to downregulated expression of the cell surface glycoprotein mesothelin. Arguably the\ngreatest insight into microenvironmental remodeling was revealed when we interrogated the\ntranscriptional hallmarks of ARID1A mutant and wild-type endometrial-type epithelium. We\nobserved dysregulated gene expression in endothelial associated with ARID1A-mutant epithelium,\nlikely mediated by SOX17, a transcription factor which when expressed in Müllerian epithelium,\ninduces a pro-angiogenic gene signature and altered secretion of angiogenesis-regulating\nproteins (Chaves-Moreira et al., 2020). Further investigations will be needed to functionally\nvalidate the impact of ARID1A and KRAS mutations on the behavior of endometriosis epithelial\ncells and altered interactions with microenvironmental populations, and to understand the\nspecific context in which ARID1A mutations, in particular, result in endometriosis-associated\novarian cancer (Anglesio et al., 2015; Wiegand et al., 2010).\nThere are caveats to this study. With an overall cohort size of nine endometriosis patients and\ntwo unaffected women we were underpowered to test for confounding effects of age or identify\nassociations between molecular features and patient symptoms or outcomes. Representative\n‘normal’ tissue from women with and without endometriosis is challenging to obtain and\ntherefore underrepresented in this study, particularly uninvolved peritoneum and pre-\nmenopausal ovary tissue. We attempted to profile uninvolved peritoneum from endometriosis\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\npatients by including four samples from two endometriosis patients where no endometriosis was\ndetected upon pathologic review, however in these cases we saw some evidence of endometrial-\ntype epithelium and endometrial-type stroma, suggesting the portion of the tissue used for\nscRNA-seq did indeed contain endometriosis tissue. While scRNA-seq is unlikely to have utility as\na diagnostic tool, this illustrates the challenge of obtaining a pathologic confirmation of\nendometriosis, particularly in cases with only a few small lesions suspected at laparoscopy that\nmay be ‘missed’ upon pathologic review due to intrinsic limitations of the embedding and\nsectioning processes.\nEndometriosis research has been substantially hindered by challenges in generating global\nmolecular profiles of tissues. This single cell atlas of endometriosis therefore represents a\nvaluable and timely resource for the endometriosis research community. Continued large-scale\nsomatic profiling efforts are clearly warranted, as these data indicate that endometriosis likely\ncomprises multiple subtypes that will likely require different approaches to treatment and\ndiagnosis.\nMethods\nReagent or Resource Source Identifier\nAntibodies\nAnti-SOX17 R&D Systems AF1924; RRID:\nAB_355060\nAnti-CAPS Sigma-Aldrich HPA043520;\nRRID:AB_10964138\nAnti-ARID1A Abcam EPR13501 (ab182560)\nBiological Specimens\nHuman endometriosis,\nendometrium and ovary\ntissues (fresh)\nCedars-Sinai Medical Center – Divisions of\nMinimally Invasive Gynecologic Surgery and\nGynecologic Oncology\nThis study\nHuman endometriosis\nand endometrium\n(formalin fixed)\nCedars-Sinai Medical Center – Divisions of\nMinimally Invasive Gynecologic Surgery and\nGynecologic Oncology and Department of\nPathology and Laboratory Medicine\nThis study\nChemicals, peptides, and recombinant proteins\n10X\nCollagenase/Hyaluronida\nse in DMEM\nSTEMCELL Technologies 07912\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nMinimal Essential\nMedium (MEM)\nCorning 10-010-CV\nDeoxyribonuclease I Sigma-Aldrich DN25-1G\nCritical commercial assays\nDead Cell Removal Kit Milteny Biotec 130-090-101\nChromium Single Cell 3′\nGEM, Library & Gel Bead\nKit v3\n10x Genomics 1000075\nChromium Single Cell A\nChip Kits\n10x Genomics 120236\nChromium Chip B Single\nCell Kit\n10x Genomics 1000153\nSoftware and algorithms\nCell Ranger version 3.1.0 https://support.10xgenomics.com/single-cell-gene-\nexpression/software/overview/welcome\nRRID:SCR_017344\nR package Seurat version\n3.2.2\nhttps://satijalab.org/seurat/ RRID:SCR_007322\nR package DoubletFinder https://github.com/chris-mcginnis-\nucsf/DoubletFinder\nRRID:SCR_018771\nR package MAST https://github.com/RGLab/MAST RRID:SCR_016340\nR package Monocle3 https://cole-trapnell-\nlab.github.io/monocle3/docs/introduction/\nRRID:SCR_018685\nR package SCENIC\nversion 1.2.0\nhttps://rawcdn.githack.com/aertslab/SCENIC/6aed5\nef0b0386a87982ba4cc7aa13db0444263a6/inst/doc\n/SCENIC_Running.html\nRRID:SCR_017247\nR package - MUSIC https://github.com/xuranw/MuSiC/ RRID: SCR_008792\nR package ReactomePA https://bioconductor.org/packages/release/bioc/htm\nl/ReactomePA.html\nN/A\nR package Harmony https://portals.broadinstitute.org/harmony/articles/\nquickstart.html\nN/A\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nResource availability\nLead contact - Further information and requests for resources and reagents should be directed to\nand will be fulfilled by the Lead Contact, Kate Lawrenson (kate.lawrenson@cshs.org).\nMaterials availability - This study did not generate new unique reagents.\nData and code availability - The data generated during this study are available at NCBI GEO\nunder the following accession number: XXX (to be provided by time of publication).\nPatient specimens\nThis project was performed with approval of the Institutional Review Board at Cedars-Sinai\nMedical Center. All patients provided informed consent. Human endometriosis, endometrial or\novarian tissues were placed in sterile serum-free MEM at 4°C and transferred to the tissue\nculture laboratory.\nMethod Details\nSurgical and pathologic review\nAll patients presented to our minimally invasive gynecologic surgery division at Cedars-Sinai\nMedical Center for consultation. The division comprises three fellowship-trained minimally\ninvasive gynecologic surgeons who perform over 150 advanced endometriosis procedures\nannually. Patients were evaluated by the surgeon with a detailed history and physical taken as\nwell as all imaging reviewed, and the decision was made for surgery. Surgery was performed\neither laparoscopically or robotically, with identification and excision of any obvious or suspected\nlesions. Lesions were excised using either ultrasonic energy or monopolar energy. Wide margins\nwere attempted for each excision. For example, a lesion in the ovarian fossa would lead to the\nfull peritoneum in the ovarian fossa being removed. Deep infiltrating endometriosis resections\nwere performed until normal anatomy was restored, leaving the endometriosis lesion intact.\nUreterolysis and mobilization of the rectosigmoid colon were performed when necessary. Areas\nwere separately labeled and passed off to nursing for the research study.\nTissue collection\nA collection protocol was created and implemented in the pathology laboratory to ensure\ncollection of endometriosis samples from consented patients without risk of compromising the\nability to provide clinical histopathologic diagnoses on resected tissue (Figure S6). Examination\nof 5 deeper sections cut at 50 μm intervals was performed in all cases in which endometriosis\nwas not identified in the first level.\nTissue processing\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nTissues were minced into ~1-2 mm pieces and digested with 1× Collagenase/Hyaluronidase\n(STEMCELL Technologies) and 100 μg/mL DNase I (Sigma-Aldrich) in 7mL serum-free MEM. The\nsample was incubated at 37°C with constant rotation for 90 mins. The supernatant was\nharvested, and the cell suspension was spun at 300 g for 10 mins at 4°C. To lyse red blood cells,\nthe cell pellet was resuspended in an RBC lysis buffer (0.8% NH4Cl, 0.1% KHCO3, pH=7.2) and\nincubated for 10 mins at room temperature. Cell suspensions were spun again at 300 g for 10\nmins at 4°C and the cell pellet was resuspended in phosphate buffered saline (PBS), or, if >5%\ndead cells were observed by trypan blue staining, cells were resuspended in dead cell removal\nbuffer and dead cell removal was performed according to manufacturer's instructions.\nRemaining cells were frozen in 90% fetal bovine serum with 10% DMSO in a Mr. Frosty container\nplaced at -80°C. Frozen cell vials were stored in LN2. Cells were thawed and transferred into a\nconical tube with 7 mL serum free media and spun at 300 g for 10 mins at 4°C. The cell pellet\nwas resuspended in 100 µL PBS. Cells were counted using a hemocytometer and the sample\nvolume adjusted to achieve a cell concentration between 100/µL to 2,000/µL.\nSingle cell capture, library preparation and next-generation sequencing\nSingle cells are captured and barcoded using 10X Chromium platform (10X Genomics). scRNA-\nseq libraries were prepared following the instructions from Chromium Single Cell 3ʹ Reagent Kits\nUser Guide (v2 or v3). Briefly, Gel Bead-In EMulsions (GEMs) are generated using single cell\npreparations. After GEM-RT and cleanup, the cDNA from barcoded single cell RNAs were\namplified before quantification using Agilent Bioanalyzer High Sensitivity DNA chips. The single\ncell 3’ gene expression libraries were constructed and cDNA corresponding to an insertion size of\naround 350 bp selected. Libraries were quantified using Agilent Bioanalyzer High Sensitivity DNA\nchip and pooled together to get similar numbers of reads from each single cell before sequencing\non a NovaSeq S4 lane (Novogene).\nSingle cell data processing and filtering\nRaw reads were aligned to the hg38 reference genome, UMI (unique molecular identifier)\ncounting was performed using Cell Ranger v.3.1.0 (10X Genomics) pipeline with default\nparameters. Normal ovary from patient 3 was removed due to a low fraction of reads in cells\n(sample proportion: 35.3%, ideal fraction >70%). Left pelvic side wall from patient 10 was\nremoved due to a low fraction of reads mapped to the transcriptome (sample proportion: 17.0%,\nideal fraction >30%) and a low fraction of reads in cells (sample proportion: 38.7%, ideal\nfraction >70%). For each individual sample we removed cells with high mitochondrial content\n(>20%) and cells with less than 200 genes. We applied DoubletFinder (McGinnis et al., 2019)\nfollowing the expected percentage of doublets for each sample (0.8% per 1000 cells) to remove\npotential doublets based on the expression proximity of each cell to artificial doublets. Combined\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nSeurat objects were adjusted for bias aiming to remove confounding factors that are potential\nsources of variation. We considered: sequencing batch, number of reads, mitochondrial mapping\npercentage and sample as parameters for Seurat SCTransform function. We applied the\nCellCycleScoring Seurat procedure to check if genes related to cell cycle are guiding the PCs and\nfound that none of the 20 first PCs had cell cycle genes within the top 20 positive and negative\ngenes. To integrate the 30 samples we used Harmony (Korsunsky et al., 2019), with lambda =\n0.2, to reduce technical batch effects. We then used the reduced Seurat object to define an initial\ncluster considering Seurat’s FindNeighbors (using 25 dimensions as parameter) and FindClusters\nfunction with a resolution of 3.\nIdentification of major cell types and epithelial subgroups\nTo define the major cell type for each cluster we divided our procedures in two steps. On step\none we performed differential expression analysis (one versus all) using MAST (Finak et al., 2015),\nimplemented in the FindAllMarker and FindMarker functions in Seurat. Then, we checked the\npresence of the following marker genes for global annotation of cell types that were differentially\nexpressed at log2 fold change 0.2 and adjusted p-value 0.05. Epithelial cells (EPCAM, KRT8,\nKRT18, KRT19, KRT7, KRT10), Fibroblasts (DCN, COL11A2, FAP, PDGFRA, COL11A1, COL1A1,\nPDGFRB), Myeloid cells (LYZ, CD14, MME, C1QA, CLEC10A), Endothelial cells (CLDN5, PECAM1,\nCD34, ESAM), Plasma cells (JCHAIN plus CD79A), B cells (JCHAIN), Smooth muscle cells (ACTA2),\nMast (TPSB2), Erythrocytes (HBB, GYPA), T cells (CD2, CD3D, CD3E, CD3G, CD8A, CCL5) and\nNatural Killer cells (TYROBP, FCGR3A). For each cluster we build a matrix of differentially\nexpressed gene counts by normalizing each count with the total markers in each cell type (Figure\nS2). To assign the cell type based on the matrix of DEG counts we applied the following rules.\nFirst, clusters that only had cell type specific genes for one cell type contained within the DEG list\nwere assigned to the corresponding cell type. If the cluster i has >35% of the cells expressing at\nleast one keratin gene and the average of scaled expression is greater than 1 then the Epithelial\ncell type was assigned to cluster i. If the cluster i has multiple cell type markers contained within\nthe DEG list, we first checked if ACTA2 was expressed, if so we checked if the proportion of\nFibroblast markers was greater than 25%, then assigned the cluster i as Fibroblast otherwise as\nSmooth muscle cells. For the case where we have multiple markers but no ACTA2 we then\nchecked which marker has max proportion and assigned the correspondent cell type to cluster i,\nif the multiple markers had the same proportion we skipped the assignment for cluster i. Clusters\nwith no markers counts were also not assigned a cell type with this decision tree.\nThe next step aimed to identify cell types for the clusters that did not express canonical patterns\nof expression of known cell-type specific genes. We use the 84 clusters for which we could\nsuccessfully assign a cell type in step one as a reference panel. We selected up to 100 of the top-\nranked DEG (log2 FC > 0; p <0.05) for each cluster to calculate pairwise Pearsons’ correlations\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nacross all clusters (to create a union set of 1,960 genes). Clusters with no cell markers were\nassigned the identity associated with the most correlated cell type with known identity. Cellular\ntranscriptomes of cells in C1 correlated with cells in C0 - Fibroblasts (r=0.91, Pearson\ncorrelation), C12 correlated with 41 - Epithelial cells (0.97), C29 correlated with C0 - Fibroblasts\n(0.89), C37 correlated with C65 - Epithelial cells (0.96), C41 correlated with C12 - Epithelial cells\n(0.97), C55 correlated with C50 - Smooth muscle cells (0.8), C63 correlated with C7 - Epithelial\ncells (0.8), C65 correlated with C37 - Epithelial cells (0.96), C83 correlated with C37 - Epithelial\ncells (0.92), C92 correlated with C37 Epithelial cells (0.88), C95 correlated with C83 - Epithelial\ncells (0.77), C84 20% of cells expressed KRT10 and it is correlated with C88 - Epithelial cells\n(0.82), C4: 23% cells expressed KRT10 and it is correlated with C88 - Epithelial cells (0.81) C12\n21% cells expressed KRT10 and it is correlated with C88 - Epithelial cells (0.77) (Figure S2).\nUMAP (uniform manifold approximation and projection)(Becht et al., 2018) was used for\nvisualizing cell types and clusters with representative markers.\nEpithelial and fibroblast subgroup analyses\nEpithelial and fibroblast clusters were identified from the parent cluster and analysed in isolation.\nWe defined the cell clusters considering Seurat’s FindNeighbors (using 20 dimensions as\nparameter) and FindClusters function with a resolution of 0.5. The eutopic endometrium sample\nfrom patient 11 exhibited a markedly different epithelial profile that dominated many of the\nsubclusters when included, and so this sample was removed from the epithelial-specific analyses.\nPathway analyses were performed using Reactome (Yu and He, 2016).\nTranscription factor analyses using SCENIC\nKey transcription factors and their gene regulatory network (GRN) were identified using SCENIC R\npackage v.1.2.0 (Single Cell rEgulatory Network Inference and Clustering (Aibar et al., 2017))\nconsidering hg38 reference genome from RcisTarget database. We applied SCENIC to\nendometrial-type epithelium and stroma using the normalized expression matrix from Seurat as\nthe input matrix.\nImmunohistochemistry for ARID1A, CAPS and SOX17\nImmunohistochemistry assays for ARID1A were performed on 5µm tissue sections on\nSuperfrostplus slides and used as surrogate for somatic loss-of-function alterations following\nestablished standards for staining and scoring (Khalique et al., 2018). ARID1A staining was\nperformed on a Leica Bond Rx (Leica Biosystems) using rabbit monoclonal antibody EPR13501\n(Abcam) at 1:3000 dilution. Slides were scored by pathologist A.E-N, assessed for (loss of)\nnuclear staining in epithelium with retained stromal nuclear staining serving as an obligate\ninternal control. CAPS staining was performed using HPA043520 (Sigma Aldrich) at a 1:5,000\ndilution and SOX17 staining performed using goat polyclonal antibody AF1924 (R&D Systems).\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nSOX17 and CAPS staining was performed on a Ventana Discovery Ultra autostainer (Roche\nVentana). CAPS and SOX17 stained slides were scored by pathologist F.M. For SOX17 the\nproportion of positively stained epithelium was estimated, and the predominant straining\nintensity categorized as negative, weak, moderate or strong.\nKRAS mutation testing by ddPCR\nEndometrial glands and stroma were enriched from 10% dilute H&E stained, 5-7µm sections by\nneedle macrodissection with a 20-gauge needle. DNA was then extracted using the Arcturus\nPicoPure DNA Extraction Kit (Thermo) and quantified using the Qubit 2.0 Fluorometer (Thermo).\n2ng of DNA was pre-amplified from the KRAS G12 codon region in a 20uL reaction vol with\nTaqMan Genotyping Mastermix (Thermo); forward and reverse KRAS primers (IDT; Table S16).\nThe following pre-amplification conditions were used: 95ºC - 10 min; 10 cycles of 94ºC/30s,\n60ºC/4min on an AC4 thermal cycler (FroggaBio). 5-fold diluted pre-amplified DNA was used in\nmultiplex ddPCR and subsequently individual-variant ddPCR for validation (if positive). All ddPCR\nreactions used the same flanking primers (Table S16) and ddPCR Supermix for Probes (no dUTP;\nBiorad) in a 25µl reaction vol with cycling parameters 95ºC/10 min, followed by 40 cycles of\n94ºC/30s, 60ºC/90s, on an AC4 thermal cycler (FroggaBio). Droplets were generated on the\nBioRad QX200 Automated Droplet Generator and read on the Biorad QX200 Droplet Reader.\nMultiplex ddPCR included an equimolar mix of probes for KRAS G12C/D/R and was used for\ndetection of KRAS G12C/D/R/V/A/S based on counts and cluster position of fluorescence signal.\nAny positive multiplex assay was then validated with individual variant ddPCR reactions (see\nTable S16 for probes).\nThe following limits of detection thresholds were applied in the multiplex assay: variant allele\nfrequency (VAF) threshold for KRAS G12C/D/R/A/S allele needed to be 3x average of negative\ncontrol reactions for the given allele. For KRAS G12V allele (VAF) was required to be 1x average\nof negative controls. For individual allele variants, ddPCR 3x average of negative control\nreactions was used universally as the minimum detection threshold.\nDefining pseudotime cell trajectories using Monocle 3\nWe used Monocle 3 (Cao et al., 2019; Qiu et al., 2017; Trapnell et al., 2014), implemented in R,\nto infer cell trajectories. We extract counts, phenotype data, and feature data from the Seurat\nobject after filtering the cells of interest. To normalize and pre-process the data we considered\nnum_dim = 100. We cluster the cells using Monocle procedures and applied the learn_graph\nprocedure using use_partition as FALSE to keep a unique trajectory. To define \"roots\" of the\ntrajectory we adapted the helper function, available on Monocle 3 website, to identify the root\nprincipal points by informing a cluster of interest.\nDeconvolution analysis using MuSiC\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nTo compare the profile of bulk tumor tissues and the histologic components present in each\nspecimen of this study, we used the multi-subject single-cell deconvolution - MuSiC method\n(Wang et al., 2019). First, we selected all epithelial cells present in Endometrioma, Extra-ovarian\nEndometriosis and Eutopic Endometrium specimes. The gene signature was based on the\ncommon genes expressed on both scRNA-seq and bulk RNA datasets. We downloaded data from\nthe following GEO datasets: GSE129617 (n=25): 24 primary tumors and 1 ascites samples;\nGSE73614 (n=107): 24 CCOC and 35 Endometrioid; GSE9899 (n=285): from 243 Ovary samples\nwe selected 20 endometrioid samples.\nAcknowledgements\nSome of the specimens were collected as part of the Biologic and Epidemiologic Markers of\nEndometriosis (BEME) study. We wholeheartedly thank all the patients who donated the\nspecimens used in this study, and the team at the Biobank and Translational Research\nLaboratory who supported tissue procurement and histologic analyses. We thank Christine Chow\nand Monica Ta at the Genetic Pathology Evaluation Centre (GPEC) for technical support for the\nmutation analyses.\nFunding\nThis study was supported by a Leon Fine Translational Science Award from Cedars-Sinai Medical\nCenter. K.L is supported by a Liz Tilberis Early Career Award (599175) and a Program Project\nDevelopment (373356) from the Ovarian Cancer Research Alliance, plus a Research Scholar’s\nGrant from the American Society (134005). The research described was supported in part by\nNIH/National Center for Advancing Translational Science (NCATS) UCLA CTSI Grant Number\nUL1TR001881 and in part by Cedars-Sinai Cancer, Canadian Cancer Society Research Institute\nimpact grant (#705647, to D.G.H) and Canadian Institute of Health Research foundation grant (to\nD.G.H.). M.A. receives funds through the Canadian Institutes of Health Research (Early Career\nInvestigator Grant in Maternal, Reproductive, Child & Youth Health), a Michael Smith Foundation\nfor Health Research Scholar Award and the Janet D. Cottrelle Foundation Scholars program\n(managed by the BC Cancer Foundation). The GPEC receives core support from BC's\nGynecological Cancer Research team (OVCARE), and The VGH+UBC Hospital Foundation. Y. W. is\na recipient of the North Family Health Research Award (administered by the VGH+UBC Hospital\nFoundation).\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nBibliography\nAibar, S., González-Blas, C.B., Moerman, T., Huynh-Thu, V.A., Imrichova, H., Hulselmans, G.,\nRambow, F., Marine, J.-C., Geurts, P., Aerts, J., et al. (2017). SCENIC: single-cell regulatory\nnetwork inference and clustering. Nat. Methods 14, 1083–1086.\nAnglesio, M.S., Carey, M.S., Köbel, M., Mackay, H., Huntsman, D.G., and Vancouver Ovarian Clear\nCell Symposium Speakers (2011). Clear cell carcinoma of the ovary: a report from the first\nOvarian Clear Cell Symposium, June 24th, 2010. Gynecol. Oncol. 121, 407–415.\nAnglesio, M.S., Bashashati, A., Wang, Y.K., Senz, J., Ha, G., Yang, W., Aniba, M.R., Prentice, L.M.,\nFarahani, H., Li Chang, H., et al. (2015). Multifocal endometriotic lesions associated with cancer\nare clonal and carry a high mutation burden. J. Pathol. 236, 201–209.\nAnglesio, M.S., Papadopoulos, N., Ayhan, A., Nazeran, T.M., Noë, M., Horlings, H.M., Lum, A.,\nJones, S., Senz, J., Seckin, T., et al. (2017). Cancer-Associated Mutations in Endometriosis without\nCancer. N. Engl. J. Med. 376, 1835–1848.\nAuersperg, N., Wong, A.S., Choi, K.C., Kang, S.K., and Leung, P.C. (2001). Ovarian surface\nepithelium: biology, endocrinology, and pathology. Endocr. Rev. 22, 255–288.\nBecht, E., McInnes, L., Healy, J., Dutertre, C.-A., Kwok, I.W.H., Ng, L.G., Ginhoux, F., and Newell,\nE.W. (2018). Dimensionality reduction for visualizing single-cell data using UMAP. Nat. Biotechnol.\n37, 38–44.\nBrosens, I., Donnez, J., and Benagiano, G. (1993). Improving the classification of endometriosis.\nHum. Reprod. 8, 1792–1795.\nCao, J., Spielmann, M., Qiu, X., Huang, X., Ibrahim, D.M., Hill, A.J., Zhang, F., Mundlos, S.,\nChristiansen, L., Steemers, F.J., et al. (2019). The single-cell transcriptional landscape of\nmammalian organogenesis. Nature 566, 496–502.\nChaves-Moreira, D., Mitchell, M.A., Arruza, C., Rawat, P., Sidoli, S., Nameki, R., Reddy, J., Corona,\nR.I., Ma, S., Winterhoff, B., et al. (2020). PAX8 orchestrates an angiogenic program through\ninteraction with SOX17. BioRxiv.\nChen, K.-S., Lim, J.W.C., Richards, L.J., and Bunt, J. (2017). The convergent roles of the nuclear\nfactor I transcription factors in development and cancer. Cancer Lett. 410, 124–138.\nCrean, D., and Murphy, E.P. (2021). Targeting NR4A nuclear receptors to control stromal cell\ninflammation, metabolism, angiogenesis, and tumorigenesis. Front. Cell Dev. Biol. 9, 589770.\nDinh, H.Q., Lin, X., Abbasi, F., Nameki, R., Haro, M., Olingy, C.E., Chang, H., Hernandez, L.,\nGayther, S.A., Wright, K.N., et al. (2021). Single-cell transcriptomics identifies gene expression\nnetworks driving differentiation and tumorigenesis in the human fallopian tube. Cell Rep. 35,\n108978.\nFinak, G., McDavid, A., Yajima, M., Deng, J., Gersuk, V., Shalek, A.K., Slichter, C.K., Miller, H.W.,\nMcElrath, M.J., Prlic, M., et al. (2015). MAST: a flexible statistical framework for assessing\ntranscriptional changes and characterizing heterogeneity in single-cell RNA sequencing data.\nGenome Biol. 16, 278.\nHasan, A., Rahim, A., Afzal, M., Naveed, A.K., Ayub, S., and Jahan, S. (2019). Serum albumin and\nC3 complement levels in endometriosis. J. Coll. Physicians Surg. Pak. 29, 702–705.\nHickey, M., Ballard, K., and Farquhar, C. (2014). Endometriosis. BMJ 348, g1752.\nKabut, J., Kondera-Anasz, Z., Sikora, J., and Mielczarek-Palacz, A. (2007). Levels of complement\ncomponents iC3b, C3c, C4, and SC5b-9 in peritoneal fluid and serum of infertile women with\nendometriosis. Fertil. Steril. 88, 1298–1303.\nKhalique, S., Naidoo, K., Attygalle, A.D., Kriplani, D., Daley, F., Lowe, A., Campbell, J., Jones, T.,\nHubank, M., Fenwick, K., et al. (2018). Optimised ARID1A immunohistochemistry is an accurate\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\npredictor of ARID1A mutational status in gynaecological cancers. J. Pathol. Clin. Res. 4, 154–166.\nKohl Schwartz, A.S., Wölfler, M.M., Mitter, V., Rauchfuss, M., Haeberlin, F., Eberhard, M., von\nOrelli, S., Imthurn, B., Imesch, P., Fink, D., et al. (2017). Endometriosis, especially mild disease: a\nrisk factor for miscarriages. Fertil. Steril. 108, 806-814.e2.\nKorsunsky, I., Millard, N., Fan, J., Slowikowski, K., Zhang, F., Wei, K., Baglaenko, Y., Brenner, M.,\nLoh, P.-R., and Raychaudhuri, S. (2019). Fast, sensitive and accurate integration of single-cell\ndata with Harmony. Nat. Methods 16, 1289–1296.\nKrausgruber, T., Fortelny, N., Fife-Gernedl, V., Senekowitsch, M., Schuster, L.C., Lercher, A., Nemc,\nA., Schmidl, C., Rendeiro, A.F., Bergthaler, A., et al. (2020). Structural cells are key regulators of\norgan-specific immune responses. Nature 583, 296–302.\nLac, V., Verhoef, L., Aguirre-Hernandez, R., Nazeran, T.M., Tessier-Cloutier, B., Praetorius, T., Orr,\nN.L., Noga, H., Lum, A., Khattra, J., et al. (2019). Iatrogenic endometriosis harbors somatic\ncancer-driver mutations. Hum. Reprod. 34, 69–78.\nLu, Y., Cuellar-Partida, G., Painter, J.N., Nyholt, D.R., Australian Ovarian Cancer Study,\nInternational Endogene Consortium (IEC), Morris, A.P., Fasching, P.A., Hein, A., Burghaus, S., et al.\n(2015). Shared genetics underlying epidemiological association between endometriosis and\novarian cancer. Hum. Mol. Genet. 24, 5955–5964.\nMcGinnis, C.S., Murrow, L.M., and Gartner, Z.J. (2019). DoubletFinder: Doublet Detection in\nSingle-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Syst. 8, 329-337.e4.\nMohankumar, K., Li, X., Sung, N., Cho, Y.J., Han, S.J., and Safe, S. (2020). Bis-Indole-Derived\nNuclear Receptor 4A1 (NR4A1, Nur77) Ligands as Inhibitors of Endometriosis. Endocrinology 161.\nMoore, L., Leongamornlert, D., Coorens, T.H.H., Sanders, M.A., Ellis, P., Dentro, S.C., Dawson, K.J.,\nButler, T., Rahbari, R., Mitchell, T.J., et al. (2020). The mutational landscape of normal human\nendometrial epithelium. Nature 580, 640–646.\nPearce, C.L., Templeman, C., Rossing, M.A., Lee, A., Near, A.M., Webb, P.M., Nagle, C.M., Doherty,\nJ.A., Cushing-Haugen, K.L., Wicklund, K.G., et al. (2012). Association between endometriosis and\nrisk of histological subtypes of ovarian cancer: a pooled analysis of case-control studies. Lancet\nOncol. 13, 385–394.\nPraetorius, T.H., Lac, V., Tessier-Cloutier, B., Senz, J., Nazeran, T.M., Köbel, M., Grube, M., Krämer,\nB., Yong, P.J., Kommoss, S., et al. (2021). Is endometriosis metastasizing? Shared somatic\nalterations suggest common origins across endometriotic lesions. MedRxiv.\nPundir, J., Omanwa, K., Kovoor, E., Pundir, V., Lancaster, G., and Barton-Smith, P. (2017).\nLaparoscopic Excision Versus Ablation for Endometriosis-associated Pain: An Updated Systematic\nReview and Meta-analysis. J. Minim. Invasive Gynecol. 24, 747–756.\nQiu, X., Mao, Q., Tang, Y., Wang, L., Chawla, R., Pliner, H.A., and Trapnell, C. (2017). Reversed\ngraph embedding resolves complex single-cell trajectories. Nat. Methods 14, 979–982.\nReddy, J., Fonseca, M.A.S., Corona, R.I., Nameki, R., Segato Dezem, F., Klein, I.A., Chang, H.,\nChaves-Moreira, D., Afeyan, L., Malta, T.M., et al. (2019). Predicting master transcription factors\nfrom pan-cancer expression data. BioRxiv.\nRicklin, D., Reis, E.S., Mastellos, D.C., Gros, P., and Lambris, J.D. (2016). Complement component\nC3 - The “Swiss Army Knife” of innate immunity and host defense. Immunol. Rev. 274, 33–58.\nSaavalainen, L., Lassus, H., But, A., Tiitinen, A., Härkki, P., Gissler, M., Pukkala, E., and\nHeikinheimo, O. (2018). Risk of gynecologic cancer according to the type of endometriosis.\nObstet. Gynecol. 131, 1095–1102.\nSuda, K., Nakaoka, H., Yoshihara, K., Ishiguro, T., Tamura, R., Mori, Y., Yamawaki, K., Adachi, S.,\nTakahashi, T., Kase, H., et al. (2018). Clonal Expansion and Diversification of Cancer-Associated\nMutations in Endometriosis and Normal Endometrium. Cell Rep. 24, 1777–1789.\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint \n\nSuda, K., Cruz Diaz, L.A., Yoshihara, K., Nakaoka, H., Yachida, N., Motoyama, T., Inoue, I., and\nEnomoto, T. (2020). Clonal lineage from normal endometrium to ovarian clear cell carcinoma\nthrough ovarian endometriosis. Cancer Sci.\nSuryo Rahmanto, Y., Shen, W., Shi, X., Chen, X., Yu, Y., Yu, Z.-C., Miyamoto, T., Lee, M.-H., Singh,\nV., Asaka, R., et al. (2020). Inactivation of Arid1a in the endometrium is associated with\nendometrioid tumorigenesis through transcriptional reprogramming. Nat. Commun. 11, 2717.\nTan, T.Z., Ye, J., Yee, C.V., Lim, D., Ngoi, N.Y.L., Tan, D.S.P., and Huang, R.Y.-J. (2019). Analysis of\ngene expression signatures identifies prognostic and functionally distinct ovarian clear cell\ncarcinoma subtypes. EBioMedicine 50, 203–210.\nTothill, R.W., Tinker, A.V., George, J., Brown, R., Fox, S.B., Lade, S., Johnson, D.S., Trivett, M.K.,\nEtemadmoghadam, D., Locandro, B., et al. (2008). Novel molecular subtypes of serous and\nendometrioid ovarian cancer linked to clinical outcome. Clin. Cancer Res. 14, 5198–5208.\nTrapnell, C., Cacchiarelli, D., Grimsby, J., Pokharel, P., Li, S., Morse, M., Lennon, N.J., Livak, K.J.,\nMikkelsen, T.S., and Rinn, J.L. (2014). The dynamics and regulators of cell fate decisions are\nrevealed by pseudotemporal ordering of single cells. Nat. Biotechnol. 32, 381–386.\nWang, X., Park, J., Susztak, K., Zhang, N.R., and Li, M. (2019). Bulk tissue cell type deconvolution\nwith multi-subject single-cell expression reference. Nat. Commun. 10, 380.\nWiegand, K.C., Shah, S.P., Al-Agha, O.M., Zhao, Y., Tse, K., Zeng, T., Senz, J., McConechy, M.K.,\nAnglesio, M.S., Kalloger, S.E., et al. (2010). ARID1A mutations in endometriosis-associated\novarian carcinomas. N. Engl. J. Med. 363, 1532–1543.\nWinterhoff, B., Hamidi, H., Wang, C., Kalli, K.R., Fridley, B.L., Dering, J., Chen, H.-W., Cliby, W.A.,\nWang, H.-J., Dowdy, S., et al. (2016). Molecular classification of high grade endometrioid and\nclear cell ovarian cancer using TCGA gene expression signatures. Gynecol. Oncol. 141, 95–100.\nYu, G., and He, Q.-Y. (2016). ReactomePA: an R/Bioconductor package for reactome pathway\nanalysis and visualization. Mol. Biosyst. 12, 477–479.\nZeng, X., Yue, Z., Gao, Y., Jiang, G., Zeng, F., Shao, Y., Huang, J., Yin, M., and Li, Y. (2018). NR4A1\nis involved in fibrogenesis in ovarian endometriosis. Cell. Physiol. Biochem. 46, 1078–1090.\nZondervan, K.T., Becker, C.M., and Missmer, S.A. (2020). Endometriosis. N. Engl. J. Med. 382,\n1244–1256.\n.CC-BY-NC 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 21, 2021. ; https://doi.org/10.1101/2021.05.20.445037doi: bioRxiv preprint","source_license":"CC0","license_restricted":false}