Editor's evaluation: Single-cell RNA sequencing and lineage tracing confirm mesenchyme to epithelial transformation (MET) contributes to repair of the endometrium at menstruation

Abstract

Article Figures and data Abstract Editor's evaluation eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract The human endometrium experiences repetitive cycles of tissue wounding characterised by piecemeal shedding of the surface epithelium and rapid restoration of tissue homeostasis. In this study, we used a mouse model of endometrial repair and three transgenic lines of mice to investigate whether epithelial cells that become incorporated into the newly formed luminal epithelium have their origins in one or more of the mesenchymal cell types present in the stromal compartment of the endometrium. Using scRNAseq, we identified a novel population of PDGFRb + mesenchymal stromal cells that developed a unique transcriptomic signature in response to endometrial breakdown/repair. These cells expressed genes usually considered specific to epithelial cells and in silico trajectory analysis suggested they were stromal fibroblasts in transition to becoming epithelial cells. To confirm our hypothesis we used a lineage tracing strategy to compare the fate of stromal fibroblasts (PDGFRa+) and stromal perivascular cells (NG2/CSPG4+). We demonstrated that stromal fibroblasts can undergo a mesenchyme to epithelial transformation and become incorporated into the re-epithelialised luminal surface of the repaired tissue. This study is the first to discover a novel population of wound-responsive, plastic endometrial stromal fibroblasts that contribute to the rapid restoration of an intact luminal epithelium during endometrial repair. These findings form a platform for comparisons both to endometrial pathologies which involve a fibrotic response (Asherman’s syndrome, endometriosis) as well as other mucosal tissues which have a variable response to wounding. Editor's evaluation The investigators present an important finding showing the contribution of mesenchyme to epithelial transformation (MET) to the healing of the endometrial luminal epithelium, furthering our understanding of the mechanisms underlying endometrial regeneration, with a potential impact on related pathology. https://doi.org/10.7554/eLife.77663.sa0 Decision letter Reviews on Sciety eLife's review process eLife digest The human uterus is a formidable organ. From puberty to menopause, it completely sheds off its internal lining every 28 days or so, creating what is in effect a large open wound. Unlike the skin or other parts of the body, however, this tissue can quickly repair itself without scarring. This fascinating process remains poorly understood, partly because human samples and animal models that mimic human menstruation are still lacking. This makes it difficult to grasp how various types of uterine cells get mobilised for healing. To fill this gap, Kirkwood et al. focused on fibroblasts, a heterogenous cell population which helps to support the epithelial cells lining the inside of the uterus. How these cells responded to the advent of menstruation was examined in female mice genetically manipulated to have human-like periods. A method known as single-cell RNAseq was used to track which genes were active in each of these cells before, one day and two days after period onset. This revealed the existence of a subpopulation of cells which only appeared when wound healing was most needed. These ‘repair-specific’ fibroblasts expressed a mixture of genes; those typical of fibroblasts but also some known to be active in the epithelial cells lining the uterus. This suggests that the cells were in the process of changing their identity so they could remake the uterine layer lost during a period. And indeed, labelling these fibroblasts with a fluorescent tag showed that, during healing, they had migrated from within the uterine tissue to become part of its newly restored internal surface. These results represent the first evidence that fibroblasts play a direct role in repairing the uterus during menstruation. From endometriosis to infertility, the lives of millions of people around the world are impacted by disorders which affect the uterine lining. A better understanding of how the uterus can fix itself month after month could help to find new treatments for these conditions. This knowledge could also be useful for to address abnormal wound healing in the skin and other tissues, as this process often involves fibroblasts. Introduction Efficient wound repair and restoration of tissue homeostasis is essential for reproductive and general health. Repetitive injury, inflammation, and fibrosis resulting in disordered tissue architecture is a major cause of morbidity and mortality due to organ failure (Greenhalgh et al., 2015). The endometrium, a complex hormone responsive multicellular tissue, exhibits remarkable resilience in its ability to respond to the monthly wound of ‘menstruation’ without fibrosis and with a response that is both rapid and scar-free (Garry et al., 2009; Ludwig and Spornitz, 1991). A critical ‘trigger’ for menstruation in women is the rapid fall in circulating concentrations of progesterone that occurs as a result of the involution of the ovarian corpus luteum in a non-fertile cycle (Maybin and Critchley, 2011). Menstruation is considered an inflammatory event with a ‘cascade’ of changes including an increase in synthesis of pro-inflammatory factors such as prostaglandins, cytokines, and chemokines which is accompanied by infiltration of immune cells of the myeloid lineage (neutrophils, monocyte/macrophages) (Evans and Salamonsen, 2012). Increased production of prostaglandins is associated with arteriole vasoconstriction which results in local hypoxia, stabilisation of HIF1alpha and increased expression of genes such as VEGF (Critchley et al., 2006b). Whilst we have a good appreciation of the mechanisms that contribute to the tightly regulated breakdown and shedding of a portion of the endometrium at menses gaps remain in our understanding of the processes that contribute to rapid, fibrosis/scar-free restoration of an intact luminal epithelium over the tissue surface. It is also important to separate the mechanisms responsible for menstrual repair, which occurs at a time in the cycle when ovarian-derived steroids are low, from endometrial regeneration/proliferation which occurs subsequently in response to rising levels of follicle-derived oestrogens during the proliferative phase of the cycle (Maybin and Critchley, 2012). Several different mechanisms have already been proposed as contributing to the luminal repair process including, activation of progenitor/stem cells in the stroma and/or epithelium, epithelial cell proliferation/migration and transformation of stromal cells into epithelial cells (mesenchyme to epithelial transition, MET) each of which are discussed briefly below. The ability of the endometrium to rapidly regenerate during each menstrual cycle has prompted researchers to explore a role for putative stem/progenitor cells residing in either the stroma and/or the epithelial compartments in endometrial repair (Gargett et al., 2016; Cousins et al., 2021). The Gargett group have identified populations of endometrial mesenchymal stem cells; eMSC with clonogenic and broad multilineage potential in both human endometrial tissue and menstrual fluid and characterised them as PDGFRB+/CD146+/ SUSD2+ (Gargett et al., 2016). Notably in human endometrial tissue sections these cells had a perivascular location and a putative pericyte identity. A role for progenitor cells residing in the basal glandular epithelium has also been proposed with a recent review highlighting a number of marker proteins employed to establish their identity including SOX9 and SSEA1 (Cousins et al., 2021). Another source of putative stem cells that may contribute to endometrial repair processes and/or endometrial pathology is the bone marrow (Kaitu’u-Lino et al., 2012) although evidence for the contribution of bone marrow-derived stem cells is somewhat limited (Deane et al., 2019; Aghajanova et al., 2010). During the first 2 days of menstruation rapid shedding of the luminal portion of the tissue (functional layer) leaves a denuded (basal) stroma interspersed with glandular ‘stumps’. This phenomenon was beautifully documented in the elegant studies of Ludwig and Spornitz, 1991. In the 1970’s Ferenczy proposed that new epithelial cells arose from proliferation of intact surface epithelium bordering denuded stroma and the basal glands that were exposed when the functional layer was shed (Ferenczy, 1976b; Ferenczy, 1976a). He noted that the re-epithelialisation of the human endometrium was rapid (~48 hr) and did not appear to involve mitosis of the epithelial cells (migration?). These mechanisms of epithelial repair were later endorsed by Ludwig and Spornitz, 1991 who also observed fibrin mesh formation on the denuded surface in the early days of menstruation and growth of an epithelial monolayer in spirals which would appear to be consistent with them being associated with the residual stumps of endometrial glands. When Garry and colleagues used a more dynamic approach involving pressure-controlled, continuous flow hysteroscopy they found evidence that loss and repair of the surface occur synchronously at different places within the endometrial cavity a process they called ‘piecemeal’ (Garry et al., 2009). They also reported epithelial cells of the glands underwent apoptosis and were shed with the decidual tissue mass. Apoptosis of glandular epithelial cells in carefully dated human endometrium has also been documented using staining for cleaved caspase 3 and shown to be high in both late secretory and menstrual phases (Armstrong et al., 2017). Taken together the data from Garry et al and Armstrong et al challenges the earlier papers citing the glands as the source of new luminal cells. There is also evidence that cells within the stromal compartment may change their identity to adopt an epithelial phenotype (mesenchyme to epithelial transition, MET) and thereby contribute to the luminal surface. This was first proposed in the 1960’s by Baggish et al who evaluated sections of menstruating human endometria and noted hyperchromatic stromal cells between gland stumps that appeared to take part in re-epithelisation by a process simulating metaplasia (Baggish et al., 1967). In the 1970’s the involvement of the stroma was challenged (Ferenczy, 1976b) but has been supported by more recent studies (Garry et al., 2009). Specifically in 2009 the authors reported that the denuded basalis endometrium was rapidly covered with a fibrinous mesh within which new surface epithelial cells were present which appeared to arise from the underlying stroma. To complement studies on human tissue samples and studies in primates (Brenner et al., 2002; Critchley et al., 2006a) we have used a highly reproducible mouse model that recapitulates key features of human menstruation in response to withdrawal of progesterone including vascular breakdown (vaginal bleeding), spatial and temporal hypoxia, cell apoptosis, expression of matrix metalloproteinases and influx of myeloid immune cells (Armstrong et al., 2017; Maybin et al., 2018; Cousins et al., 2016a; Cousins et al., 2016b; Cousins et al., 2014). Importantly, as in the human endometrium (Garry et al., 2009), restoration of an intact luminal epithelial layer occurs within two days (Cousins et al., 2014). Consistent with some of the reports on repair processes in human endometrium we found evidence for epithelial cell migration and proliferation of residual epithelial cells (Ki67+) but no significant proliferation of basal glands (Cousins et al., 2014). Other groups have also used similar mouse models to explore the dynamic changes during a menstrual-like event. Kaitu’u-Lino et al used a mouse model in combination with pulse labelling of cells with BrdU, and reported that a population of epithelial progenitor cells residing in the basal glands might contribute to postmenstrual repair (Kaitu’u-Lino et al., 2010, Kaitu’u-Lino et al., 2012). In our endometrial repair/menstrual model one of our novel findings was the existence of stromal cells that co-expressed vimentin (stromal cell marker) and cytokeratin (epithelial cell marker) specifically residing in areas denuded of epithelium which suggested to us that stromal cells might be changing their phenotype via MET (Cousins et al., 2014). In the current study, we have used single cell RNA sequencing and lineage tracing to specifically address which, if any, of the stromal mesenchymal cell subpopulations we previously identified in the mouse endometrium (Kirkwood et al., 2021) might contribute to the rapid restoration of the intact luminal epithelial cell layer observed in our mouse model (Cousins et al., 2014; Cousins et al., 2016a). To the best of our knowledge this is the first time that inducible-Cre lineage tracing approaches in adult mice have been used to explore MET in a model of menstruation. Results Single-cell RNA sequencing identifies novel mesenchymal cell subpopulations unique to ‘repairing’ tissue in a mouse model of endometrial breakdown and repair (menstruation) To identify whether changes in the transcriptome of mesenchymal cells occurred during endometrial repair we recovered tissues from mice expressing the Pdgfrb-BAC-eGFP transgene during the normal cycle (controls; n=4) and at 24 (n=4) and 48 hr (n=4) after progesterone withdrawal using a well-validated model of endometrial shedding/repair (simulated ‘menstruation’; Cousins et al., 2016b; Figure 1A i, ii). Population restricted cell sorting was used to isolate CD31-/CD45-/GFP+(PDGFRβ+) cells prior to single-cell sequencing. In cycling mice we have previously reported GFP + cells are restricted to the stromal compartment of the endometrium (Kirkwood et al., 2021). Figure 1 Download asset Open asset Single-cell RNA sequence analysis identified mesenchymal cell populations unique to ‘repairing’ tissue in a mouse model of endometrial breakdown and repair (menstruation). (A, i) Mouse model of endometrial tissue breakdown and repair, (ii) histological morphology of uterine tissues 0, 24, 48, and 72 hr progesterone withdrawal illustrating tissue breakdown, repair, remodelling and resolution. (B) UMAP visualisation: GFP +mesenchymal cells isolated from Pdgfrb-BAC-eGFP mouse endometrium (cycling plus 24/48 hr after progesterone withdrawal) cluster into eight distinct populations. (C) Dot plot: expression of canonical gene signatures associated with known cell types present in the endometrium: mesenchymal cells (Pdgfrb, Vim, Des, Thy1), perivascular cells (Mcam, Cspg4, Rgs4, Acta2) and fibroblasts (Pdgfra, Cd34, Fbln1/2, Mfap4/5, Col1a1) (dot colour: average expression per cluster; dot size: percent cluster expressing gene) (D) UMAP visualisation: mesenchymal clusters split by source/dataset identifies transient repair-specific subpopulations P2 and F4 in 24 hr tissues and F5 in 48 hr tissues. scRNAseq was used to generate a dataset for each timepoint and these were then integrated for downstream cluster and differential gene expression analysis. This strategy resulted in identification of 8 transcriptionally distinct clusters of GFP + cells (Figure 1B). The expression of canonical phenotypic gene markers for mesenchymal, perivascular and fibroblast cell populations previously identified in control/cycling mouse endometrium was used to assign putative identities to each cluster (Kirkwood et al., 2021). We identified three perivascular populations (V, P1, P2; expression of Mcam, Cspg4, Rgs4, Acta2) and five fibroblast subpopulations (F1-5; expression of Pdgfra, Cd34, Fbln1/2, Mfap4/5, Col1a1). Importantly, cells in all eight clusters expressed mesenchymal markers Pdgfrb, Vim, Des, and Thy1 (Figure 1C, Figure 2—figure supplement 1A(iii)) while no clusters expressed Pecam1 (CD31) or Ptprc (CD45) (Figure 2—figure supplement 1A(i-ii)) consistent with the cell isolation strategy employed. Analysis of the distribution of these subpopulations in the different datasets (Figure 1D) confirmed the presence of five mesenchyme subpopulations previously identified in cycling mouse uterine tissue (V, P1, F1-3; Kirkwood et al., 2021). However in tissue recovered at 24 and 48 hr three additional transcriptionally distinct repair-specific cell clusters were identified: P2 and F4 in 24 hr tissues (Figure 1D; green and purple circles respectively) and F5 present in 48 hr tissues (Figure 1D; pink ellipse). Gene expression analysis identified a unique repair-associated cluster that expressed genes associated with epithelial cell phenotype Differential gene expression analysis revealed a very high degree of similarity between P1 (previously identified as pericytes) (Kirkwood et al., 2021) and P2 which was specific to 24 hr tissue (Figure 2A). Unbiased GO analysis of the P2 transcriptome identified functions associated with blood vessel formation and immune cell signalling, characteristic of pericytes (Figure 2—figure supplement 1D(i)) suggesting a role for these cells in angiogenesis and regulation of the immune system in response to changes in the local environment. Figure 2 with 1 supplement see all Download asset Open asset Gene expression analysis identifies transcriptomic profiles of mesenchymal cells and the expression of genes associated with epithelial cell identity in repair-specific fibroblasts (cluster F4). (A) Heatmap (yellow, high; purple, low) displaying differentially expressed genes per cluster when compared to all other clusters (logFC >0.5, pvalue <0.05, Wilcoxon rank-sum test) top is colour coded and named by cluster; V=vascular smooth muscle cells (vSMCs), P1=pericytes 1, P2=pericytes 2, F1=fibroblasts 1, F2=fibroblasts 2, F3=fibroblasts 3, F4=fibroblasts 4, F5=fibroblasts 5. The expression of the top 5 exemplar genes in each cell cluster is illustrated. (B) Dot plot: GO enrichment terms relating to biological processes (BP) associated with the genetic signature of repair-specific cluster F4 (dot size: gene ratio, number of genes in data/number of genes associated with GO term; dot colour: p-value representing the enrichment score). (C) Gene expression plots: expression of canonical mesenchymal cell markers (i) Pdgfrb and (ii) Vim and canonical epithelial cell markers (iii) Epcam and (iv) Krt18. Note F4 fibroblasts (red box) express markers associated with both cell lineages. (D) Dot plot: expression of extended gene signatures associated with mesenchymal and epithelial cell lineages in the endometrium (dot colour: average expression per cluster; dot size: percent cluster expressing gene). (E) Dot plot: expression of genes associated with the regulation of MET/EMT taken from the MSigDB (dot colour: average expression per cluster; dot size: percent cluster expressing gene). We noted that cluster F5 (specific to 48 hr tissue) expressed a large number of mitochondrial genes and genes associated with cell death including Casp3, Casp9, Trp53 and Bax (Figure 2—figure supplement 1). GO analysis of the F5 transcriptome identified functions associated with DNA damage. Based on this in silico data and the results in our previous study using the same mouse model system that detected high levels of cleaved caspase expression in the stromal compartment at 24 hr (Armstrong et al., 2017) we suggest that F5 represents a cluster of dead/apoptotic cells and it was therefore excluded from further downstream analyses. Cluster F4 (24 hr tissue) was of particular interest as it had a transcriptomic signature that was distinct from to those of the mesenchymal cell populations (F1-3) present in cycling endometrium (Kirkwood et al., 2021; Figure 2A; red box). Unbiased GO analysis enriched for biological functions that are associated with endometrial tissue repair including response to wounding, reproductive structure development, response to oxidative stress and regulation of vasculature development. Additional functions that were identified included epithelium migration, morphogenesis of a branching epithelium and regulation of epithelial cell migration (Figure 2B). Cluster F4 was the only mesenchymal cell cluster that expressed markers known to be associated with both definitive mesenchymal (Pdgfrb, Vim) and epithelial (Epcam, Krt18) cell types (Figure 2C). Comparison across all mesenchymal clusters of the expression of both known canonical mesenchymal/fibroblast markers including Thy1, Cd34, Cdh11, Col1a1, Col3a1, Fbln1, Fbln2, Mfap4, and Mfap5 and epithelial markers such as Krt8, Krt19, Muc1, Muc4, Cd24a, Car2, Sdc4, Alcam, and Cdh1 revealed F4 to be the only cluster that expressed genes associated with both cell types. Notably, F4 did not express the full array of keratins and mucins associated with definitive epithelial lineages but preferentially expressed Alcam, Cd24a, Clu, Krt18, and Sdc4 when compared to other mesenchymal clusters (Figure 2D; red box). We considered this was evidence that these cells may be of an intermediate phenotype and represent cells undergoing a mesenchymal to epithelial transition (MET) but not yet fully committed to a definitive epithelail lineage. To investigate this further we accessed annotated gene sets from the MSigDB (Subramanian et al., 2005) that are known to be associated with the regulation of MET/EMT. This analysis identified that F4 also expressed key transcription factors associated with MET/EMT including Cemip, Cited1, Ctnnb1, Grem1, Hmga2, Klf4, and Twist2 (Figure 2E). Cells that co-express both mesenchymal and epithelial cell markers can be detected in the uterus during endometrial tissue repair To validate this finding and to analyse the topography of putative MET cells in the mouse endometrium, sections from Pdgfrb-BAC-eGFP mice were co-stained with antibody directed against EPCAM. Cells that expressed both GFP and EPCAM were identified in endometrial tissue undergoing active repair (24 hr; Figure 3A (i); white arrows) specifically in regions where (a) the stromal surface was denuded after decidual shedding and (b) a new epithelial cell layer was being re-instated. Some GFP + EPCAM + cells persisted in the luminal layer at 48 hr but were not detected in 72 hr tissue (Figure 3A ii, iii). Consistent with previous data in control/cycling endometrium expression of the GFP reporter was exclusive to the stromal compartment and epithelial cells were EPCAM + (yellow arrows, Figure 3A iv; Kirkwood et al., 2021). Figure 3 Download asset Open asset Analysis of Pdgfrb-BAC-eGFP uterine tissue during endometrial tissue repair and remodelling identifies transient expression of GFP in EPCAM + epithelial cells. (A) Immunohistochemical analysis of GFP reporter protein (green) and epithelial cell marker EPCAM (red) in uterine tissues 24, 48, and 72 hr following progesterone withdrawal. (i) In 24 hr tissues cells that are GFP + EPCAM + are detected in (a) regions of denuded stromal surfaces where the decidualised tissue has detached (white arrows) and (b) regions where a new epithelium has formed (white arrows). (ii) At 48 hr GFP + EPCAM + cells were detected within the renewed luminal epithelium (white arrows) adjacent to GFP- EPCAM + epithelial cells (yellow arrows). (iii) At 72 hr no GFP + EPCAM + cells were detected. (iv) In control/cycling tissues epithelial cells are EPCAM + (yellow arrows), stroma is GFP+. (B) Flow cytometry (FC): analysis of EPCAM in Pdgfrb-BAC-eGFP uterine tissues at 24, 48, and 72 hr following progesterone withdrawal; note detection of a new population of GFP + EPCAM + cells at 24 hr (boxed lower panel) with decreased numbers at 48 hr. (C) Bar plot: quantification of FC data analysing the expression of EPCAM by GFP reporter positive cells in the uterus (%) - control (n=21) 0.53 ± 0.46%; 24 hr (n=30) 12.78 ± 2.49%; 48 hr (n=17) 3.71 ± 1.36%; 72 hr (n=6) 0.96 ± 0.47%; one-way ANOVA, Sidak’s multiple comparisons. Figure 3—source data 1 Summary statistics for flow cytometry analyses performed in Figure 3C. https://cdn.elifesciences.org/articles/77663/elife-77663-fig3-data1-v2.docx Download elife-77663-fig3-data1-v2.docx Figure 3—source data 2 One-way ANOVA with Sidak’s multiple comparisons test for Figure 3C. https://cdn.elifesciences.org/articles/77663/elife-77663-fig3-data2-v2.docx Download elife-77663-fig3-data2-v2.docx Analysis of the same tissues using flow cytometry confirmed the presence of a new population of cells that expressed both GFP and EPCAM at 24 hr with a reduction in their number at 48 hr and levels that appeared comparable to cycling/control tissues in the fully remodelled tissue (72 hr) (Figure 3B–C). These data were in agreement with the results of scRNAseq analysis where Epcam was detected in the F4 cluster in the 24 hr dataset (Figure 2C) with analysis of the tissue sections suggesting the F4 cluster may represent a mixture of cells in the stroma in the process of undergoing MET (Figure 3Aa) as well as those which are already incorporated into the luminal epithelium (Figure 3Aa) where they are still detectable at 48 hr (Figure 3Aii). Whilst promising studies on cell fate using mice from the Pdgfrb-BAC-eGFP ‘knock-in’ line has limitations as expression of GFP is limited to cells in which the Pdgfrb is active and immunohistochemistry has revealed that the PDGFRB protein is only present in endometrial mesenchyme (see Figure 1 in Kirkwood et al., 2021). With high expression of Pdgfrb in the mesenchyme and some expression persisting in F4 (Figure 2C) we speculated that presence of GFP in the luminal epithelial cells at 24 and 48 hr represented mesenchyme cells that underwent a rapid MET. It was not possible to determine whether these cells did/or did not persist at 72 hr as absence of GFP might be due to protein turnover and therefore additional studies with new mouse lines were necessary to resolve this question. Comparison between fibroblast and epithelial cell scRNA clusters confirmed the repair-specific cells (F4) had a unique transcriptome To compare the transcriptome of the F4 cluster (putative MET mesenchyme/epithelium) to differentiated endometrium epithelial cells scRNAseq datasets were generated by from CD31-/CD45-/GFP-/EPCAM + epithelial cells obtained from the Pdgfrb-BAC-eGFP mouse uterus (control/cycling and 48 hr). Bioinformatic analysis combining GFP + datasets (cycling/24 hr/48 hr) and GFP-EPCAM + epithelial cell data (cycling/48 hr) identified 16 distinct cell clusters (Figure 4A and C). Based on expression of Pdgfrb (Figure 4Bii) and canonical markers (Figure 4D) seven of these had a mesenchyme phenotype: three perivascular (V, P1-2), four fibroblast (F1-4). Nine of the clusters expressed classical markers of epithelial cells including Epcam (E1-9) (Figure 4A and B (i, ii), C). Notably, the previously identified repair-specific fibroblasts (F4) formed a cluster which appeared in close proximity to both fibroblast and epithelial cell clusters but remained distinct from definitive perivascular cells (V, P1, P2). Doublet analysis was also conducted to confirm this was a unique population (Figure 4—figure supplement 1A). Figure 4 with 1 supplement see all Download asset Open asset In silico analysis of scRNAseq reveals transcriptomic similarity between repair-specific mesenchymal cells and subpopulations of EPCAM + endometrial epithelial cells. (A) UMAP visualisation: GFP + mesenchymal cells and GFP-EPCAM + epithelial cells isolated from Pdgfrb-BAC-eGFP mouse endometrium (cycling/24/48 hr and cycling/48 hr, respectively) cluster into 16 distinct populations: 7 mesenchymal, 9 epithelial. Repair-specific fibroblasts (F4) cluster in close proximity to epithelial cell clusters but remain a distinct population of cells. (B) (i) UMAP visualisation: three groups of cell populations are detected in the data: perivascular, fibroblast, and epithelial cells (add colours) (ii) gene expression of Pdgfrb is restricted to mesenchymal populations (perivascular/fibroblast) while Epcam is restricted to epithelial populations (and repair-specific fibroblasts; F4), confirming the isolation strategy adopted. (C) Scaled heatmap (yellow, high; purple, low) displaying differentially expressed genes per cluster when compared to all other clusters (logFC >0.5, pvalue <0.05, Wilcoxon rank-sum test) top is colour coded and named by cluster; V=vascular smooth muscle cells (vSMCs), P1/2=pericytes 1/2, F1−4=fibroblasts 1–4, E1−9=epithelial cells 1–9. The expression of 3 exemplar genes in each cell cluster is displayed. F4 shows genetic similarity to E1/2 (red box). (D) Dot plot: expression of canonical genes associated with mesenchymal, perivascular, fibroblast and epithelial lineages, F4 is the only cluster that expressed genes from multiple lineages (red box) (dot colour: average expression per cluster; dot size: percent cluster expressing gene). (E) Dot plot: expression of gene signatures associated with known epithelial cell subtypes present in the endometrium as per the literature: canonical epithelial, luminal, differentiating, ciliated, SCs, glandular, basal epithelial cells (dot colour: average expression per cluster; dot size: percent cluster expressing gene). Red box F4 compared to E1, E2 highlighting expression of genes found in the luminal and differentiating epithelial cell populations. A high degree of similarity was observed between the transcriptomic signature of F4 and the definitive epithelial cell clusters E1 and E2 (Figure 4C, red box, Figure 4—figure supplement 1B). Extended analysis of canonical markers of mesenchymal (perivascular/fibroblast) and epithelial cell types confirmed that cluster F4 uniquely expressed genes associated with both fibroblast and epithelial lineages (Fig