{"paper_id":"2415e370-1544-4313-9e2f-db0e68879d18","body_text":"Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 1 of 26\nGenetics and Genomics\nMedicine\nSingle-cell characterization of\nmenstrual fluid at homeostasis\nand in endometriosis\nPetra C Schwalie, Cemsel Bafligil, Julie Russeil, Magda Zachara, Marjan Biocanin, Daniel Alpern,\nEvelin Aasna, Bart Deplancke, Geraldine Canny, Angela Goncalves\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life Sciences, Ecole\nPolytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland • Roche Innovation Center Basel, Pharma\nResearch and Early Development, Basel, Switzerland • Wellcome Trust Sanger Institute, Wellcome Genome\nCampus, Cambridge, UK • German Cancer Research Center (DKFZ), Somatic Evolution and Early Detection\nHeidelberg, Germany • University College Dublin, Research Office, Dublin, Ireland\nhttps://en.wikipedia.org/wiki/Open_access\nCopyright information\neLife Assessment\nThis basic research study presents useful data concerning the menstrual fluid\ncomposition and its potential for endometriosis biomarker research. However,\ndespite solid bioinformatics analyses, the choice of markers used to separate or\nidentify the different cell types needs to be justified and the results better discussed\nin relation to current knowledge of the pathophysiology of endometriosis.\nhttps://doi.org/10.7554/eLife.99558.1.sa2\nAbstract\nProgress in detecting and understanding endometrial conditions in women of fertile age,\nsuch as endometriosis, has been hampered by the invasiveness of the sample collection\nprocedure. Menstrual fluid (MF) can be sampled non-invasively and could provide a unique\nopportunity to study the physiological state of tissues in the reproductive system. Despite this\npotential, the use of MF for diagnostics and research has been limited. Here we establish\nprotocols and assess the feasibility of collecting and processing MF in an outpatient setting.\nWe characterize the cellular contents of MF from 15 healthy women using flow cytometry\nand single-cell RNA-sequencing, and demonstrate the ability to recover millions of live cells\nfrom the different cellular fractions of interest (epithelial, stromal, endothelial, perivascular\nand blood). Through computational integration of MF with endometrial samples we show\nthat MF sampling is a good surrogate for endometrial biopsy. In a proof-of-principle case-\ncontrol study, we collect MF from a further 7 women with a diagnosis of endometriosis and\n11 healthy controls. Through RNA sequencing of 93 MF samples from these women we\nhighlight important differences between ex vivo and cultured cells, identify impaired\ndecidualisation, low apoptosis, high proliferation, and both higher and lower inflammatory\nactivity in different subsets of immune cells as distinguishing features of endometriosis\nReviewed Preprint\nv1 • November 19, 2024\nNot revised\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 2 of 26\npatients. Finally, we identify potential novel pan-cell-type biomarkers for this neglected\ncondition.\nIntroduction\nMenstrual fluid (MF) is an accessible source of large numbers of live cells of high biomedical\ninterest. During menstruation, the female body sheds 86±48ml of MF (1     ) containing cervico-\nvaginal secretions, blood and cells from the lining of uterus, cervix, vagina, and possibly the\nfallopian tube and ovaries. This rich source of live cells has been previously proposed as a scalable\nsource of live uterine NK cells (2     ), mesenchymal stem cells (3     –7     ), T-cells (8     ) and\nepithelial cells (9     ,10     ) for research and therapy.\nMenstrual fluid has also been proposed as a sampling strategy for potential diagnosis of\nendometriosis (11     –14     ), a chronic, currently incurable disease that affects an estimated 10% of\nwomen of reproductive age. Endometriosis occurs when cells similar to those that line the inside\nof the uterus – the eutopic endometrium – grow outside the uterine cavity in so called ectopic\nlesions. These ectopic, endometrial-like cells can be found on the ovaries, fallopian tubes, and the\ntissue lining the pelvis. In rare cases, they may spread beyond the pelvic region. Like the\nendometrial cells in the uterus, these cells outside the uterus respond to the menstrual cycle. This\ncan cause severe menstrual pain, chronic lower abdominal and pelvic pain, painful intercourse,\nand fertility issues. The exact cause of endometriosis is not well understood and its diagnosis is\noften delayed, taking on average 8-10 years from the onset of symptoms. This delay is partly due to\nthe necessity of laparoscopic surgery for definitive diagnosis. Progress in understanding the\nmolecular basis of endometriosis has also been hampered by the invasiveness of the sample\ncollection procedure, as well as the lack of appropriate animal models. Therefore, there is a real\nneed for non-invasive, early diagnosis methods, as well as a better understanding of the disease.\nDespite its potential for research and diagnostics, quantitative parameters of MF cellular\ncomposition, including cell-count, cell-viability, cell-type representation and transcriptional\nsimilarity to endometrium remain only partially described (13     ,15     –17     ). Moreover, it is\nunclear whether endometriosis biomarkers identified in other tissue compartments are likewise\ndetectable in MF. To date, the identification of biomarkers for endometriosis has proven\nchallenging due to contradictory results that depend on the exact tissues being contrasted (e.g.\nperitoneal fluid versus endometriotic lesion versus endometrial biopsy), on the phase of the\nmenstrual cycle at which the tissues were collected, or on the cell-type analysed (see (18     ) for a\nsystematic review of these issues).\nMF can be acceptably and conveniently self-collected by study participants using menstrual cups\n(19     ). Here, we profiled self-collected MF in a research setting at single-cell resolution to ask\nwhether MF contains a faithful representation of the cell-type composition and cellular\ntranscriptional states of the endometrium. We then compared the proportion of cell-types and\ntranscriptomes of the different cellular fractions of MF between endometriosis patients and\nhealthy controls.\nResults\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 3 of 26\nMenstrual fluid collection is a viable\nsampling alternative to endometrial biopsy\nTo quantify the cellular and molecular composition of MF, we analyzed 25 self-collected fresh MF\nsamples from 15 healthy women on day 2 of the menstrual cycle (Sup. Table 1). MF samples were\nserially passed through 100μm and 70μm cell strainers and split into a flow-through and a clump\nfraction. The retained cellular clumps were gently dissociated prior to further processing. Flow\ncytometry was performed on each sample fraction staining for CD45+ (immune cells), CD45-\nEPCAM+ (epithelial) and CD45-EPCAM- (stromal) cell populations (Fig. 1A     ).\nVolume, cell-viability and cell-type composition\nby fluorescence-activated cell sorting (FACS)\nThe median volume of MF per sample was 5mL (Fig. 1B     ), with a potential donor effect (Fig. 1C-\nD     , one-way ANOVA p-value = 0.09), but no age association (generalized linear mixed model with\ndonor random effect p-value = 0.43). Three out of 25 MF samples did not yield any live cells, these\nsamples came from two individuals, both of which subsequently donated again, yielding samples\nwith enough viable cells for processing. The remaining MF samples yielded a median of 2.6 million\nlive cells per mL of flow-through (Fig. 1E     ). There was no statistical association (Pearson and\nSpearman correlation coefficient not significantly different from 0 at a 5% significance level)\nbetween cell viability and processing time, with the sample with the highest processing time delay\nafter cup removal (12h) containing 63% live cells. After gating out dead cells, flow-through MF was\nmainly composed of CD45+ cells (Fig. 1F     , median 83% of cells).\nCell-type composition and comparison to\nendometrial biopsy by single-cell RNA-seq\nTo characterize and quantify the cell-types found in MF, we then generated single-cell\ntranscriptomes from 7 MF samples of 4 individuals. For each sample we mixed the 3 sorted\npopulations at a 14/22/64% ratio (Fig. 2A     ) to obtain a balanced representation of cell-types.\nUsing established marker genes, we identified ciliated, glandular and exhausted epithelial cells,\nmesenchymal and decidualized stromal cells, smooth muscle/endothelial cells, and immune cells\nin all 7 samples (Fig. 2B-D     , Methods). The ratio of epithelial to stromal and immune cells was\nsimilar among MF samples, with epithelial cells comprising around 50% of cells (Fig. 2D     ).\nWe then asked whether the transcriptional profiles of the MF samples correspond to the\ntranscriptional profiles of freshly isolated endometrial cells by computationally integrating our\nsingle-cell data with single-cells from 3 endometrial biopsies (20     ) in the late-secretory and\nproliferative phases of the menstrual cycle (Fig. 2E-F     ). Transcriptionally, MF epithelial, stromal\nand immune cells were more similar to their corresponding cell-types in endometrial biopsies\nthan to one another, indicating that these cells are able to retain their cellular identity despite\nbeing shed (Fig. 2G     ). Epithelial and stromal cells exhibited the highest degree of transcriptomic\nconcordance between MF and endometrial biopsy cells, whereas the immune cells were more\ndivergent, indicating that immune cells are more sensitive to shedding or experimental\nmanipulation.\nDespite the overall transcriptomic concordance, we expected expression differences between MF\nand the endometrial biopsies to reflect differences in cycle stage. The endometrial biopsy samples\nused came from cycle stages with high progesterone levels, which represses, among others,\nexpression of many matrix metalloproteinases (MMPs). Progesterone withdrawal at menstruation\ntriggers expression of MMPs (21     ), an effect mediated by cytokines including tumor necrosis\nfactor-α (TNF-α). In addition, a large number of neutrophils loaded with MMPs are recruited\nthrough induction of chemokines such as interleukin 8 (CXCL8) (21     ). Differential gene\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 4 of 26Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 4 of 26\nFigure 1\nCharacterization of menstrual fluid from healthy donors. (A) Sample preparation for the FACS assays. (B) Total sample\nvolumes in mL. (C) Frequency of the number of donations per donor. (D) Volume in mL of MF across 2-4 donations from the\nsame donor. (E) Million of live cells per mL of flow-through. (F) Percentage of CD45+ cells in flow-through.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 5 of 26Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 5 of 26\nFigure 2\nCharacterisation of menstrual fluid from healthy donors using single-cell RNA-seq (scRNA-seq). (A) Sample preparation for\nthe scRNA-seq assays. (B) Single-cell UMAP of ciliated, glandular and exhausted epithelial cells, mesenchymal and\ndecidualized stroma, smooth muscle/endothelial and immune cells. (C) Marker genes used for the cell-type annotation. (D)\nCell-type abundances in each MF sample (c - clumps, f - flow-through). (E) UMAP of the MF samples integrated with\nendometrial biopsies from Wang et al. (20     ). (F) Cell-type abundances in MF and biopsies. (G) Gene expression correlation\nbetween the MF and biopsies. (H) Differential gene expression between stromal cells in MF and biopsies. Gene names are\nshown for the top 15 genes with largest absolute log2 fold changes and for genes with -log10 adjusted p-values greater than\n20. Colors indicate if genes are members of the pathways indicated. The pathways chosen were significantly over-\nrepresented (FDR q-value <= 0,00027) in gene set enrichment analysis.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 6 of 26\nexpression comparison between MF and the endometrial biopsy stromal cells revealed much\nhigher expression in MF of MMPs, members of the TNF-α signaling pathway and CXCL8 (Fig. 2H     ,\nSup. Table 2).\nIn sum, our data show that sampling self-collected MF is a viable strategy for obtaining large\nnumbers of live epithelial, stromal and immune cells from the human endometrium in an\noutpatient setting without prior scheduling. MF samples capture the main cell types of the\nendometrium and are transcriptionally similar to cells collected via invasive biopsy.\nProperties of MF from patients with\nendometriosis and healthy controls\nTo assess the potential of MF to provide insight into female reproductive tract disorders, we\ncollected 36 fresh MF samples from 7 women with a confirmed diagnosis of endometriosis and 11\nage-matched healthy volunteers (Fig. 3A     , Table 1     , Sup. Table 3). Samples were processed by\ndensity gradient to remove red blood cells. High volume samples were split into CD45+ and CD45-\nfractions by MACS. For a subset of samples, we used FACS to estimate proportions of CD45+ vs.\nCD45-cells and the CD45-fraction was further sorted into CD105+/- cells for a estimation of the\nfraction of putative mesenchymal stem cells present (22     ).\nOnce more, we observed a donor effect on volume (Fig. 3B     , ANOVA p-value 0.054). However, we\ndid not observe any statistically significant differences between endometriosis and healthy\nvolunteer-derived MF samples in volume, cell number, cell viability, fraction of CD45+/- cells, or\nfraction of mesenchymal stem cells (CD45-CD105+) (Fig. 3B,C     ). Moreover, MF volume was not\nstatistically significantly different between endometriosis stages at this sample size (Fig. 3B     ).\nThe vast majority of samples could be successfully cultured over multiple passages, under\nstandard mesenchymal cell culture conditions (Methods), irrespective of whether they came from\ncases or controls.\nCell-type composition of ex vivo and in vitro MF cells by RNA-seq\nTo molecularly characterize both freshly isolated (ex vivo) and cultured (in vitro) samples, as well\nas identify putative differences between endometriosis patients, we performed 3’ bulk RNA-\nsequencing on 93 samples (Fig. 4A     ). A principal component analysis of gene expression patterns\nrevealed two main sample groups: samples that were processed directly after isolation (ex vivo)\nand those that were cultured (in vitro) (Fig. 4B     ). An over-representation analysis of genes\nnegatively correlated with PC1 (i.e. genes with higher expression in ex vivo samples) revealed\nthese to be highly enriched for blood cell-type markers, i.e. monocytes, dendritic and macrophage-\nspecific genes (Fig. 4C     , Sup. Table 4). Cultured samples were transcriptionally more\nhomogeneous than freshly isolated samples, likely due culture-induced selection for CD45-cells.\nIndeed, we observed that the CD45+ cell fraction typically failed to attach, in line with previous\nreports that have shown preferential attachment of menstrual stromal cells (so-called menstrual-\nderived stem cells, or colony-forming cells) upon culture (23     ). Nevertheless, when comparing in\nvitro cells to CD45-ex vivo cells, we observed that the transcriptional profiles remain distinct,\nsuggesting major transcriptional shifts induced by cell culture (Fig. 4D     , Sup. Table 5). In\ncontrast, ex vivo CD45+ samples showed higher similarity to ex vivo non-sorted (unsorted) samples\n(Fig. 4E     ), in line with the observation that most MF cells are CD45+ as determined by FACS.\nTo confirm the differences in cell-type composition, we used an in silico transcriptome\ndeconvolution method (24     ) and a single-cell reference dataset (13     ) to estimate cell-type\nfractions from our bulk RNA-seq samples. We found that indeed in vitro samples are\npreponderantly composed of stromal, smooth muscle and/or endothelial-like cells, whereas ex vivo\nsamples are preponderantly composed of myeloid cells (Fig. 4F     ).\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 7 of 26Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 7 of 26\nFigure 3\nComparison of menstrual fluid volume, cell number and cell-type proportions between endometriosis patients and healthy\ncontrols. (A) Sample preparation procedure. (B) MF volume in controls - C, and endometriosis stages 2 to 4 (E2-E4). Points are\ncoloured by donor. (C) Cell number, viability, fraction of CD45+ cells and fraction of CD45+CD105+ cells in endometriosis - E,\nand controls - C.\nTable 1\nDemographic and disease characteristics of the endometriosis and healthy controls.\n\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 8 of 26Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 8 of 26\nFigure 4\nCell-type composition of MF samples by cell-type enrichment method. (A) Sample preparation procedure. (B) Principal\ncomponent analysis of all samples using the full transcriptome. Shown in parentheses on the axes is the percentage of\nvariance explained by each of the principal components. (C) Over-representation analysis of the top 30 genes positively\nassociated with PC1 in the cell-type marker gene-sets from PanglaoDB(54     ). P-values were calculated using Fisher’s exact\ntest and corrected for multiple testing using the Benjamini-Hochberg procedure. (D) Gene expression heatmap of the 10\nmost significantly differentially expressed genes between CD45-ex vivo and all in vitro samples. (E) Principal component\nanalysis of ex vivo samples. (F) In silico cell-type deconvolution of the bulk RNA-seq samples using MuSiC. pDC - plasmacytoid\ndendritic cells, uNK – uterine NK cells. (G) Spearman correlation of the whole CD45+ ex vivo transcriptomes for donors that\ndonated more than once. The row-side dendrogram (identical to the column-side one) was omitted.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 9 of 26\nIn sum, MF collected ex vivo and in vitro display transcriptional changes consistent with changes\nin the cell-type enriched, as well as a culture induced phenotype. In vitro samples only capture a\nsubset of the cell populations initially present in MF, preponderantly stromal-like cells.\nDonor effect on transcriptomic profiles\nCycle-to-cycle variability in MF transcriptomic profiles has not been previously characterised.\nThere is a concern that samples might be too variable due to unpredictability in tissue breakdown\nand exact timing of sample collection. Leveraging on having multiple longitudinal samples from\nthe same donor (Fig. 3B     ), we asked if MF samples are transcriptionally stable across cycles of\nthe same donor. To do so, we looked at transcriptomic concordance among CD45+ ex vivo samples,\nas these were the most homogeneous in terms of their composition (Fig. 4F     ). Samples from the\nsame donor clustered together more often than expected by chance, suggesting that donor effects\ncan be observed over multiple cycles (Fig. 4G     ).\nIdentification of potential transcriptional\nbiomarkers of endometriosis in MF\nTo identify MF gene expression markers that might distinguish endometriosis from healthy\ncontrols, we used three different statistical models on various subsets of the 93 RNA-seq assays\nperformed. First, we identified endometriosis biomarkers that are detectable in stromal-like CD45-\ncells. Second, we identified biomarkers that are detectable in immune CD45+ cells. Finally, we\nimplemented a model to identify biomarkers that are detectable irrespective of cell-type enriched\nor whether the sample is freshly isolated or cultured.\nDifferentially expressed genes in CD45-stromal-like cells\nIn this model, we sought to identify differentially expressed genes that can be found in in vitro\nCD45-samples. We found 256 genes differentially expressed at a 5% false discovery rate (Fig. 5A     ,\nSup. Table 6). These included the down-regulation of stromal decidualization genes (20     ,25     –\n27     ) in endometriosis patients when compared to healthy controls, supporting findings that\ndecidualization is compromised in cultured (28     ,29     ) and freshly isolated stromal cells from\nendometriosis patients (11     ,13     ,14     ). Likewise, among genes downregulated in endometriosis\npatients we found an enrichment for targets of estrogen receptor alpha (ESR1), whose expression\nhas been previously found to be suppressed in endometric stromal cells (30     ). Additionally, we\nfound that inflammation related pathways (TNFa signaling and IL-2/STAT5), were down-regulated\nin endometriosis patients, whereas epithelial to mesenchymal transition (EMT) and angiogenesis\nwere upregulated (Sup. Table 7).\nAmong down-regulated genes we found multiple endometriosis previously implicated in the\ndisease including PENK, MLLT11, HAND2-AS1 and PDE4B.\nThe endogenous opioid peptide precursor proenkephalin (PENK) has been previously reported to\nbe up-regulated in whole-endometrium (31     ,32     ) of endometriosis patients when compared to\nhealthy controls. In contrast, PENK expression was found to be down-regulated in ectopic stroma\nwhen compared to eutopic stroma of patients (33     ). The discrepancy in the direction of\nregulation may have to do with the specific cell-types assayed, as well as the cycle phase at which\nendometrial samples were obtained, as PENK expression is modulated by the menstrual cycle\n(31     ). Interestingly, a decrease in opioid peptides has been previously suggested to be involved in\nthe maintenance of chronic inflammation in endometriosis (34     ).\nConsistent with our findings, the expression of mixed-lineage leukemia translocated to 11\n(MLLT11) was previously found to be reduced in the ectopic stromal cells of women with\nadvanced endometriosis compared to healthy controls, and its down-regulation found to be\nassociated with an increased stromal cell adhesion phenotype (35     ).\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 10 of 26Petra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 10 of 26\nFigure 5\nComparison of the transcriptomes between endometriosis patients and healthy controls. (A) Gene expression heatmap of\n150 genes differentially expressed genes in CD45-in vitro samples (columns). The genes shown are either 1) members of\npathways significantly enriched among differentially expressed genes, or 2) top most differentially expressed genes\naccording to p-value and absolute log2 fold-change (“Other”). (B) Gene expression heatmap of 150 genes differentially\nexpressed genes in CD45+ ex vivo samples (columns). The genes shown are either 1) members of pathways significantly\nenriched among differentially expressed genes, or 2) top most differentially expressed genes according to p-value and\nabsolute log2 fold change (“Other”). (C) Log2 fold-change of the inflammatory signaling pathway activity between\nendometriosis patients – E, and controls – C in the different cell-types defined in Shih. et al. Horizontal bars show standard\nerrors and asterisks indicate significance level of a t-test performed for each cell-type independently. P-values were adjusted\nfor multiple testing with the method from Benjamini-Hochberg. (D) Log2 fold-change of the expression of selected ligand-\nreceptors involved in inflammatory pathways between endometriosis – E and controls – C. (E) Volcano plot of the differential\nexpression of endometriosis - E - versus healthy controls – C using all available samples. Points above the horizontal blue line\nindicate genes with multiple-testing adjusted p-values below 0.05. (F) Normalised expression levels of MTRNR2L1 in\nendometriosis - E and control - C samples. (G) Normalised expression levels of HBG2 in endometriosis - E and control - C\nsamples.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 11 of 26\nHAND2 Antisense RNA 1 (HAND2-AS1), whose expression is coordinated with HAND2 in human\nendometrial stromal cells, was also previously found to be reduced in ectopic lesions from patients\ncompared controls, primarily expressed by stromal cells as opposed to epithelial, and that its\nsilencing impairs stromal cell decidualization (20     ,26     ,27     ).\nFinally, genetic polymorphisms in cAMP-specific 3’,5’-cyclic phosphodiesterase 4B (PDE4B) were\nrecently proposed to be implicated in the etiology of endometriosis (36     ).\nRelatively fewer genes were found to be up-regulated in endometriosis patients when compared to\ncontrols. Notably, the most significant up-regulated gene was fibronectin 1 (FN1). A genome-wide\nmeta-analysis and a subsequent targeted study have implicated single nucleotide polymorphisms\nin FN1 with moderate to severe endometriosis (37     ,38     ), and relaxed fibronectin is being\ninvestigated as a potential target for imaging endometriotic lesions (39     ). At the protein level,\nstudies comparing eutopic and ectopic endometrium from patients have yielded contradictory\nresults (39     ). However, fibronectin levels in plasma and peritoneal fluid have consistently\nreported elevated levels in endometriosis patients when compared to healthy women (40     ,41     ).\nDifferentially expressed genes in CD45+ immune cells\nWe found 792 differentially expressed genes among the CD45+ ex vivo samples (Fig. 5B     , Sup.\nTable 8, the one sample with an abnormally high percentage of stromal cells in Fig. 4D      was\nexcluded). We found a trend for lower expression of immune-related genes in CD45+ MF samples\noriginating from endometriosis patients, in particular for genes related to inflammatory pathways\n(TNFa signaling, Inflammatory response, IFN-gamma response, Sup. Table 9).\nThis was intriguing, as the production of potentially pro-inflammatory cytokines, such as IL-6, has\nbeen previously found to be up-regulated in cultured and fresh stromal cells from ectopic and\neutopic endometrium from patients compared to controls (13     ,42     ). To explore the activation of\ninflammatory pathways in specific cell-types from endometriosis patients and controls, we used\nAUCell (43     ) to score the activity of the MsigDB “inflammatory response” pathway in each\nindividual cell in the dataset from Shih et al. (13     ) (Methods).\nIn agreement with our findings, we confirmed that the activity of the inflammatory signaling\npathway was significantly downregulated in myeloid (the most abundance cell-type in CD45+ ex\nvivo samples) and CD4+ T cells, and in a subtype of stromal cells of endometriosis patients. In\ncontrast, the pathway was upregulated in uNK cells and two subtypes of CD8+ T cells of\nendometriosis patients (Fig. 5C     ). A similar pattern emerged when looking at the expression of\nspecific inflammatory ligand receptors (Fig. 5D     ).\nAt the opposite end of the spectrum, we identified several genes significantly more highly\nexpressed in endometriosis patients versus healthy controls, many of which had also previously\nshown to have increased expression in the eutopic endometrium of endometriosis patients\ncompared to healthy controls (e.g. CALD1 (44     ) and SEMA3A (45     ), which has been proposed to\nbe involved in macrophage recruitment to endometriotic lesions (46     )), or increased expression\nin the eutopic endometrium of severe compared to mild endometriosis patients (TEAD2 (47     )).\nUnlike the downregulated genes, these genes were typically not immune-cell specific, but rather\nconnected to tissue invasiveness, epithelial to mesenchymal transition or myc targets (Fig. 5B     ).\nC-myc expression has been previously reported to be altered in the eutopic endometrium of\nendometriosis patients when compared to controls which, together with a reduction in cell death\nby apoptosis, is thought to play a role in facilitating the invasive features of endometriotic\nendometrium (48     ).\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 12 of 26\nInterestingly, among immune-related genes, endometriosis patients also had significantly higher\nexpression of CD8A (Fig. 5B     ). The abundance and activation levels of CD8 T cells has been a\ntopic of considerable interest in endometriosis research. However, results vary depending on the\ntissue analysed (49     ). Our study supports the notion that CD8+ T cells (or subsets thereof (13     ))\nare enriched in the menstrual fluid of endometriosis patients.\nDifferentially expressed genes in all cell and sample types\nIn the model designed to identify biomarkers that are detectable irrespective of cell type enriched\nor whether the sample is freshly isolated or cultured, we identified a large number of\ndifferentially expressed genes (Fig. 5F     , Sup. Table 10). Many of these genes, such as MTRNR2L1\nand HBG2, were not, to our knowledge, previously described as biomarkers of endometriosis.\nMTRNR2L1 was detected by all 3 models as being substantially upregulated in healthy controls\ncompared to endometriosis patients (Fig. 5G     ). As high MTRNR2L1 expression in human\nendometrium is highly specific to endometrial epithelial glandular cells (50     ), the lower\nexpression of this gene in endometriosis patients might be driven either by lower expression\nlevels in their epithelial cells, or by a proportion of epithelial cells present in their MF.\nInterestingly, despite not having been previously linked to endometriosis, MTRNR2L1 was found to\nbe downregulated in the eutopic endometrium of adenomyosis patients compared to healthy\ncontrols (51     ).\nHBG2 is more highly expressed in endometriosis patients than in controls (Fig. 5H     ). As high\nexpression of HBG2 is very highly specific to macrophages and erythroid cells and lowly expressed\nin all cell-types of the endometrium (50     ), we hypothesize that the origin of this signal may be in\na higher proportion of macrophage cells in endometriosis patients, or in a higher amount of\nambient RNA from higher levels of cell death in the samples of endometriosis patients.\nIntriguingly, HBG2 was previously found to be over-expressed in the menstrual endometrium of\nwomen with heavy menstrual bleeding when compared to controls (52     ).\nDiscussion\nThis study demonstrates the feasibility of using self-collected MF to obtain a comprehensive\ncellular and molecular snapshot of the human endometrium. Our data show that MF collection is\nnot only viable but also yields a substantial number of live cells, including epithelial, stromal, and\nimmune cells. The similarity in transcriptional profiles between MF and endometrial biopsies and\nrecapitulation of menstrual cycle dynamics reinforces the reliability of MF as a representative\nsample of the endometrial environment. This representativeness opens avenues for exploring\nmenstrual fluid in research settings and potential clinical diagnostics.\nOur pilot study reports on useful parameters for future large-scale studies. Our data show that MF\nis primarily composed by CD45+ myeloid cells, which dominate the transcriptomic signal.\nUnsorted ex vivo samples vary widely in cell-type composition with no association with disease\nstate, highlighting the importance of enriching for the target cell-type in both bulk or single-cell\ntranscriptomics assays. We report differences between ex vivo and in vitro (cultured cells) cells,\nwith the former containing a large number of immune-related cells, and the latter being composed\nmainly of stromal-like cells. This stresses the importance of probing fresh cells directly or having\nalternative specialized culture conditions if the immune compartment is to be assessed. We also\nreport on genes whose transcription in stromal cells has been previously reported to be\ndysregulated in endometriosis patients, but which we found to be affected by cell culture (e.g.\nMARCH4, INMT or GABRP (32     ,53     )) or menstrual cycle stage (18     ) (e.g. PENK).\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 13 of 26\nIn patients with endometriosis, our MF analysis revealed distinctive gene expression patterns.\nNotably, the differential expression of genes such as FN1, PDE4B, PENK, MLLT11 and HAND2-AS1\nin stromal-like cells of endometriosis patients aligns with previous findings and underscores their\npotential roles in the disease’s pathogenesis. Furthermore, the dysregulation of immune-related\ngenes in endometriosis patients suggests an altered immune landscape in endometriosis, possibly\ncontributing to the disease’s inflammatory component. The identification of MTRNR2L1 and HBG2\nas potential biomarkers that are robust to cell culture and cell-type enrichment is intriguing. These\nfindings necessitate further research to validate and understand the roles of these biomarkers in\nendometriosis.\nThe use of MF for diagnosing endometriosis offers a promising non-invasive alternative to current\ndiagnostic methods, which are often invasive and delayed. Our pilot study was limited by a\nrelatively small sample size. Further research with larger cohorts with standardized collection\nprocedures is essential to validate our findings and establish MF-based diagnostics in clinical\npractice.\nIn conclusion, our study highlights the potential of menstrual fluid as a rich, non-invasive source\nfor studying the endometrium and diagnosing endometriosis. This work paves the way for further\nstudies to explore the full diagnostic potential of menstrual fluid and its application in women’s\nhealth. Combined with banking stem cells isolated from MF, MF sampling can enable powerful\nfunctional studies of gynecological disease.\nMethods\nSample collection for the healthy donor cohort\nThe study was approved by the South West - Frenchay Research Ethics Committee (IRAS project ID\n221351). Fifteen healthy participants (menstruating, self-defined healthy females with no other\nselection criteria) were recruited from employees, visitors and students at the Wellcome Trust\nSanger Institute. All samples and data collected were solely identified by a randomly assigned\nidentifier and the research team had no access to the participant names. No links were kept\nbetween A) the consent forms containing the participant name and B) the samples and sample\nassociated data collected. Participants were between 24 and 47 years old with self-reported\nregular menstrual cycles. Two of the 15 participants were users of combined oral contraceptive\npills, and a third one of a contraceptive implant. All women were requested to collect their\ncomplete menstrual fluid during day 2 of their period using a silicone menstrual cup (Mooncup®).\nThese cups were purchased by the investigation team for the purpose of this investigation. We do\nnot hold any commercial link with this provider. Size A was recommended if the person was aged\n30 and over and/or has given birth vaginally at any age. Size B was recommended if the person\nwas under 30 and had not given birth vaginally. Participants were asked to pour and/or pipette the\nsample from the Mooncup into a sample collection tube and to deliver the sample tube to the\nstudy collection boxes within 6 hours of menstrual cup removal. Donors were able to donate\nsamples up to 6 times each and at each donation filled in a questionnaire with demographic and\nbiologic data (age, time of cup removal, current use of contraception, medication and ongoing\nillness). All sample collection tubes and questionnaires within a kit were labeled with the same\nrandomly assigned identifier, thereby completely anonymising the samples but allowing multiple\ndonations from the same participant to be linked. Sample delivery was unscheduled and samples\nwere processed upon receipt from the collection boxes (within 30m to 12h from the time of\nmenstrual cup removal, of this time up to 3h at RT and the remainder in the fridge at 5°C).\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 14 of 26\nSample processing for the healthy donor cohort\nThe total volume, appearance (color, haemolysis and viscosity scores) and presence of\nclumps/mucus were assessed for each sample. Five of the samples had high mucous content and\nwere centrifuged at 300g for 5 minutes. The supernatant was transferred to a new tube, spun at\n2000g for 3 minutes and resuspended with 50mL PBS. All samples were then serially filtered with\n100μm and 70μm nylon mesh sieves (Falcon). Clumps were backwashed from the sieves with PBS\nand digested with collagenase for 20–60 minutes at 37°C. Flowthrough and clumps were briefly\ncentrifuged and subject to red blood cell lysis (Biolegend Inc.). Cells were counted with Trypan\nblue. Samples with more than 1 million live cells were processed by flow cytometry.\nFlow cytometry staining and acquisition\nfor the healthy donor cohort\nAfter digestion, cells were washed and centrifuged at 2000rpm for 3 minutes. Cells were\nresuspended in PBS plus Human TruStain FcX Blocking solution (BioLegend) and incubated for 5\nminutes at room temperature in the dark. 5μl/million cells of CD45 (clone 2D1, BioLegend) and\nEPCAM (clone 3C4, BioLegend) antibodies were added and incubated for 1h at 4°C in the dark. Post\nstaining, cells were washed and centrifuged at 2000rpm for 3 minutes. 1μl/mL of DAPI was added\nand incubated for 5 minutes at RT. Cells were washed twice, filtered with a 50μm FACS filter and\nanalyzed using SH800S, MoFlo XDP or BD Influx cell sorters, depending on instrument availability,\naccording to manufacturer’s instructions and gating strategies. After sorting, the CD45+, CD45-\nEPCAM- and EPCAM+ fractions were collected for 10x single-cell RNA-sequencing.\nGeneration of single cell transcriptomes\nSingle-cell suspensions were loaded per channel of ChromiumTM Single Cell Chips (10X\nGenomics® Chromium Single Cell 3’ Reagent Kits v2.0), aiming for a recovery of 4,600 cells.\nReverse transcription and library construction were carried out according to the manufacturer’s\nrecommendations. Libraries were sequenced on an Illumina HiSeq 4000 using paired-end runs of\n150bp.\nComputational analysis of the single-cell RNA-seq data\nRaw sequencing reads were processed using the Cell Ranger analysis pipeline (55     ). The\n“cellranger count” command was used to generate filtered and raw matrices. Reads were aligned\nagainst the human genome version GRCh38. Raw gene barcode count matrices were processed\nusing CellBender for unsupervised denoising (56     ) and further analyzed using the R package\nSeurat (57     ). To remove low quality cells, an adaptive filtering threshold approach was used\nbased on extreme numbers of counts (count depth) and extreme numbers of genes per barcode.\nCells were filtered when counts or genes per barcode were less than 99% of all cells, or when the\nmitochondrial content was higher than 10%.\nTo annotate cell types, we used two strategies: annotation of individually processed samples, and\nannotation of all samples integrated together. For the individually processed samples, counts were\nnormalized using the SCT normalization approach of Seurat. Sample-specific UMAPs were\nconstructed using a subset of genes exhibiting high cell-to-cell variation which were identified by\nmodeling the mean-variance relationship. The top 3000 features were used to perform PCA\nanalysis. To cluster the cells, a K-nearest neighbor (kNN) graph based on the euclidean distance in\nPCA space was first constructed using the first 30 PC components as input. Next, the Louvain\nalgorithm was applied to iteratively group cells. We identified the cell types in each cluster using a\ncombination of manual and automated approaches using known marker genes (Fig. 1E     ). First,\nclusters were assigned to known cell populations using cell type– specific markers obtained\nthrough the FindAllMarkers function. Multiple testing correction was performed using Benjamini-\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 15 of 26\nHochberg procedure. Second, the R package Garnett (58     ) in cluster extension mode was used to\nannotate cells in a semi-automated manner. Integration of samples was performed using\nreference-based canonical correlation analysis together with SCT normalization (Seurat), and\nmarker based annotation was performed as above. A consensus of the three annotation strategies\nwas used to annotate each sample.\nIntegration of the MF samples with the endometrial biopsies was performed using reference-based\ncanonical correlation analysis together with SCT normalization (Seurat) using 3000 integration\nfeatures.\nDifferential expression analysis of menstrual fluid versus the endometrial biopsies was performed\nby generating pseudo-bulks for each of the main cell-types: stromal, epithelial and immune.\nPseudo-bulking was performed by summing up the counts of all cells of each cell-type by sample.\nOnly genes that had non-zero counts in at least one MF and one biopsy sample were kept. Samples\nwere then used as replicates in a DESeq2 analysis using default parameters. Log fold change\nshrinkage using the lfcShrink function with apeglm was performed on the results.\nGene set enrichment analyses were performed using the GSEA tool (v4.3.3, (59     )). Sorted\ndifferentially expressed genes by log fold change were used as input to a weighted analysis using\nthe following gene-sets: Hallmarks, GO Biological Process, Reactome and KEGG.\nSample collection and processing for the endometriosis cohort\nWe recruited age-matched (20     –45     ) healthy female volunteers and endometriosis patients\nthrough advertising at endometriosis support groups as well as the EPFL campus. The study\nreceived ethical approval by the Swiss Cantonal Authorities Vaud, CER-VD, project nr. 2016-00770.\nAll samples and data collected were solely identified by a randomly assigned identifier. Samples\nwere processed within 6h by a standard density gradient-based cell isolation protocol\n(Histopaque). The viability and the number of nucleated cells in the cell suspension was\ndetermined using a Nucleostainer and when possible, further processed by MACS into a CD45+\nand CD45-fraction. For a subset of samples, surface marker gene expression was determined by\nFACS. The isolated single cell suspension was diluted to 1 × 107 cells/ml with FACS buffer (DPBS −/−)\nwith 1% human platelet serum and the following fluorophore-conjugated antibodies were added:\nanti-human CD45 for identifying hematopoietic cells, and anti-human CD105 for identifying\nmesenchymal stromal cells. 7-Aminoactinomycin D (7-AAD) was used for assessing viability and\nSyto40 was used for discerning nucleated cells. Cell culture was performed using high glucose\nMEMalpha medium supplemented with 5% platelet serum and 50 ug/ml Primocin. TrypLE Select\nreagent was used to collect the cells from the cell culture plates.\nGeneration of bulk RNA-seq for the endometriosis cohort\nRNA was collected from fresh or cultured samples into Tri-Reagent and bulk RNA-seq was\nperformed as previously described, using BRB-seq, a highly multiplexed 3’ end protocol (60     ).\nComputational analysis of the RNA-\nseq for the endometriosis cohort\nSequencing results were processed in the Deplancke laboratory by a standard in-house pipeline,\nconsistent with published methodology to obtain digital gene expression values for each cell and\nestimated expression for each sample. In brief, sequenced tags were demultiplexed (at sample and\nlibrary level) and fastq files containing 62-bp-long single-end sequenced tags (reads) were were\ntrimmed and filtered using prinseq 0.20.3 and cutadapt 1.5 and subsequently aligned to the\nEnsembl 84 gene annotation of the hg19 human genome using STAR 2.4.0g. The number of tags per\ngene was calculated using htseq-count 0.6.0 with the parameters ‘htseq-count -m intersection-\nnonempty -s no -a 10 -t exon -i gene_id’.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 16 of 26\nThe over-representation analysis in Fig. 4C      was performed by taking the top 30 genes most\nanticorrelated with PC1 from Fig. 4B      and testing their enrichment in the cell-type marker gene\nsets from PanglaoDB (54     ), using the hypergeometric test implemented in Enrichr (61     ).\nThe decomposition of the bulk-RNA-seq data into cell-type compositions was performed using raw\ncounts and cell-type annotations from the dataset in Shih et al.(13     ) as reference, and the sample\ndeconvolution method implemented in the MuSiC R package (24     ), with default parameters.\nThe differential gene expression analyses of endometriosis versus healthy controls were\nperformed using the standard DESeq2 workflow (62     ). For the CD45-in vitro and CD45+ ex vivo\ncell analyses we used endometriosis (yes or no) as a fixed effect. For the “all cell and sample types”\nanalyses we used type (ex vivo or in vitro), cell-type (unsorted, CD45+ or CD45-) and endometriosis\n(yes or no) as fixed effect terms.\nOver representation analysis for Fig. 5A-B      was performed by taking all genes significantly up or\ndown regulated in each analysis and testing for their enrichment in the MSigDB Hallmarks gene\nsets (63     ) using the hypergeometric test implemented in Enrichr (61     ) and correcting for\nmultiple testing. Additionally, in Fig. 5A      we included differentially expressed decidualization-\nassociated genes. We defined decidualization-associated genes as genes whose expression peaks in\nstromal cells during decidualization in Wang et al. (20     ) or that are known from the literature.\nFor the scoring of the “Inflammatory signaling” pathway in the single-cells from Shih et al.(13     ),\nwe downloaded the gene names for this pathway from MSigDB (63     ) and used AUCell (43     )\nwith maxrank set to 800. To test for significance of differential scoring between cell-types we\ntransformed the scores with log2(score+0.0000001) and then used a linear model with\nendometriosis status (yes or no) as predictor.\nData Availability\nAll data produced in the present work are contained in the manuscript and available online at\n10.5281/zenodo.11105267\nhttps://zenodo.org/records/11105267     \nAcknowledgements\nThis work was supported by an Early Detection Gynaecological Cancers Pump Priming award by\nCRUK to AG. PCS was supported by an HFSP Postdoctoral Fellowship and a Diversa Foundation\nGrant. We thank Suzanne Ratte for help with experimental technical assistance, Johanne Doleman\nfor support with the ethical application and Daniel Gaffney for support with resources. We thank\nall volunteers for their participation in this study.\nAdditional information\nAuthor contributions\nAG and PCS: designed the study.\nGC: provided critical input on study design and volunteer recruitment.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 17 of 26\nCB, JR, MB, MZ and DA: collected, processed samples and performed experiments. AG, PCS and EA:\nanalysed data.\nAG: wrote the manuscript.\nBD: provided resources and input on data interpretation. PCS: edited the manuscript.\nAll authors: read, commented and approved the manuscript.\nDeclaration/Conflict of interest\nThe cups (LadyCup) for the endometriosis cohort were donated by Ladyplanet GmbH.\nDA, PCS and MB conducted this work while employed at EPFL in Switzerland. DA is\ncurrently employed by and holds Alithea Genomics company stock. MB is currentl\nemployed by Lonza AG. PCS is currently employed by and holds F. Hoffmann-La Roche Ltd.\ncompany stock.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 18 of 26\nReferences\nDonoso MB, Serra R, Rice GE, Gana MT, Rojas C, Khoury M, et al. (2019) Normality Ranges of\nMenstrual Fluid Volume During Reproductive Life Using Direct Quantification of Menses\nwith Vaginal Cups Gynecol Obstet Invest 84:390–5\nIvarsson MA, Stiglund N, Marquardt N, Westgren M, Gidlöf S, Björkström NK (2017)\nComposition and dynamics of the uterine NK cell KIR repertoire in menstrual blood\nMucosal Immunology 10:322–31\nAlcayaga-Miranda F, Cuenca J, Luz-Crawford P, Aguila-Díaz C, Fernandez A, Figueroa FE, et al.\n(2015) Characterization of menstrual stem cells: angiogenic effect, migration and\nhematopoietic stem cell support in comparison with bone marrow mesenchymal stem\ncells Stem Cell Res Ther 6\nRossignoli F, Caselli A, Grisendi G, Piccinno S, Burns JS, Murgia A, et al. (2013) Isolation,\nCharacterization, and Transduction of Endometrial Decidual Tissue Multipotent\nMesenchymal Stromal/Stem Cells from Menstrual Blood BioMed Research International\n2013:1–14\nFaramarzi H, Mehrabani D, Fard M, Akhavan M, Zare S, Bakhshalizadeh S, et al. (2016) The\nPotential of Menstrual Blood-Derived Stem Cells in Differentiation to Epidermal Lineage:\nA Preliminary Report World J Plast Surg. 5:26–31\nLiu Y, Niu R, Yang F, Yan Y, Liang S, Sun Y, et al. (2018) Biological characteristics of human\nmenstrual blood-derived endometrial stem cells J Cell Mol Med 22:1627–39\nPatel AN, Park E, Kuzman M, Benetti F, Silva FJ, Allickson JG (2008) Multipotent Menstrual\nBlood Stromal Stem Cells: Isolation, Characterization, and Differentiation Cell Transplant\n17:303–11\nSabbaj S, Hel Z, Richter HE, Mestecky J, Goepfert PA (2011) Menstrual Blood as a Potential\nSource of Endometrial Derived CD3+ T Cells PLoS ONE 6\nCindrova-Davies T, Zhao X, Elder K, Jones CJP, Moffett A, Burton GJ, et al. (2021) Menstrual flow\nas a non-invasive source of endometrial organoids Commun Biol 4\nFilby CE, Wyatt KA, Mortlock S, Cousins FL, McKinnon B, Tyson KE, et al. (2021) Comparison of\nOrganoids from Menstrual Fluid and Hormone-Treated Endometrium: Novel Tools for\nGynecological Research JPM 11\nWarren LA, Shih A, Renteira SM, Seckin T, Blau B, Simpfendorfer K, et al. (2018) Analysis of\nmenstrual effluent: diagnostic potential for endometriosis Mol Med 24\nSchmitz T, Hoffmann V, Olliges E, Bobinger A, Popovici R, Nößner E, et al. (2021) Reduced\nfrequency of perforin-positive CD8+ T cells in menstrual effluent of endometriosis\npatients Journal of Reproductive Immunology 148\n1.\n2.\n3.\n4.\n5.\n6.\n7.\n8.\n9.\n10.\n11.\n12.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 19 of 26\nShih AJ, Adelson RP, Vashistha H, Khalili H, Nayyar A, Puran R, et al. (2022) Single-cell analysis\nof menstrual endometrial tissues defines phenotypes associated with endometriosis BMC\nMed 20\nNayyar A, Saleem MI, Yilmaz M, DeFranco M, Klein G, Elmaliki KM, et al. (2020) Menstrual\nEffluent Provides a Novel Diagnostic Window on the Pathogenesis of Endometriosis Front\nReprod Health 2\nVan Der Molen RG, Schutten JHF, Van Cranenbroek B, Ter Meer M, Donckers J, Scholten RR, et\nal. (2014) Menstrual blood closely resembles the uterine immune micro-environment and\nis clearly distinct from peripheral blood Human Reproduction 29:303–14\nFraser IS, McCarron G, Markham R, Resta T (1985) Blood and total fluid content of menstrual\ndischarge Obstetrics and gynecology 65:194–8\nHosseini S, Shokri F, Tokhmechy R, Savadi-Shiraz E, Jeddi-Tehrani M, Rahbari M, et al. (2015)\nMenstrual blood contains immune cells with inflammatory and anti-inflammatory\nproperties J of Obstet and Gynaecol 41:1803–12\nBrulport A, Bourdon M, Vaiman D, Drouet C, Pocate-Cheriet K, Bouzid K, et al. (2024) An\nintegrated multi-tissue approach for endometriosis candidate biomarkers: a systematic\nreview Reprod Biol Endocrinol 22\nKoks CAM, Dunselman GAJ, De Goeij AFPM, Arends JW, Evers JLH (1997) Evaluation of a\nmenstrual cup to collect shed endometrium for in vitro studies Fertility and Sterility 68:560–\n4\nWang W, Vilella F, Alama P, Moreno I, Mignardi M, Isakova A, et al. (2020) Single-cell\ntranscriptomic atlas of the human endometrium during the menstrual cycle Nat Med\n26:1644–53\nChevronnay HP Gaide, Selvais C, Emonard H, Galant C, Marbaix E, Henriet P (2012) Regulation\nof matrix metalloproteinases activity studied in human endometrium as a paradigm of\ncyclic tissue breakdown and regeneration Biochimica et Biophysica Acta (BBA) - Proteins and\nProteomics 1824:146–56\nMutlu L, Hufnagel D, Taylor HS (2015) The Endometrium as a Source of Mesenchymal Stem\nCells for Regenerative Medicine1 Biology of Reproduction [Internet 92\nBozorgmehr M, Gurung S, Darzi S, Nikoo S, Kazemnejad S, Zarnani AH, et al. (2020)\nEndometrial and Menstrual Blood Mesenchymal Stem/Stromal Cells: Biological\nProperties and Clinical Application Front Cell Dev Biol 8\nWang X, Park J, Susztak K, Zhang NR, Li M (2019) Bulk tissue cell type deconvolution with\nmulti-subject single-cell expression reference Nat Commun 10\nChen ST, Shi WW, Lin YQ, Yang ZS, Wang Y, Li MY, et al. (2023) Embryo-derive TNF promotes\ndecidualization via fibroblast activation eLife 12\nLiu L, Dong H, Guan Y, Fan T, Sun W, Bagchi IC, et al. (2023) Regulation of HAND2 Expression\nby LncRNA HAND2-AS1 in Ovarian Endometriosis Involving DNA Methylation Journal of the\nEndocrine Society 7\n13.\n14.\n15.\n16.\n17.\n18.\n19.\n20.\n21.\n22.\n23.\n24.\n25.\n26.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 20 of 26\nHuyen DV, Bany BM (2011) Evidence for a conserved function of heart and neural crest\nderivatives expressed transcript 2 in mouse and human decidualization REPRODUCTION\n142:353–68\nKlemmt PAB, Carver JG, Kennedy SH, Koninckx PR, Mardon HJ (2006) Stromal cells from\nendometriotic lesions and endometrium from women with endometriosis have reduced\ndecidualization capacity Fertility and Sterility 85:564–72\nBarragan F, Irwin JC, Balayan S, Erikson DW, Chen JC, Houshdaran S, et al. (2016) Human\nEndometrial Fibroblasts Derived from Mesenchymal Progenitors Inherit Progesterone\nResistance and Acquire an Inflammatory Phenotype in the Endometrial Niche in\nEndometriosis1 Biology of Reproduction 94\nBulun S, Monsavais D, Pavone M, Dyson M, Xue Q, Attar E, et al. (2012) Role of Estrogen\nReceptor-β in Endometriosis Semin Reprod Med 30:39–45\nBurney RO, Talbi S, Hamilton AE, Vo KC, Nyegaard M, Nezhat CR, et al. (2007) Gene Expression\nAnalysis of Endometrium Reveals Progesterone Resistance and Candidate Susceptibility\nGenes in Women with Endometriosis Endocrinology 148:3814–26\nChen S, Chai X, Wu X (2022) Bioinformatical analysis of the key differentially expressed\ngenes and associations with immune cell infiltration in development of endometriosis\nBMC Genom Data 23\nLavogina D, Samuel K, Lavrits A, Meltsov A, Sõritsa D, Kadastik Ü, et al. (2019)\nChemosensitivity and chemoresistance in endometriosis – differences for ectopic versus\neutopic cells Reproductive BioMedicine Online 39:556–68\nVelho RV, Taube E, Sehouli J, Mechsner S (2021) Neurogenic Inflammation in the Context of\nEndometriosis—What Do We Know? IJMS 22\nProestling K, Husslein H, Hudson QJ, Witzmann-Stern M, Widmar B, Bagó-Horváth Z, et al.\n(2023) MLLT11 Regulates Endometrial Stroma Cell Adhesion, Proliferation and Survival in\nEctopic Lesions of Women with Advanced Endometriosis IJMS 25\nYang F, Wu Y, Hockey R, Doust J, Mishra GD, Montgomery GW, et al. (2023) Evidence of shared\ngenetic factors in the etiology of gastrointestinal disorders and endometriosis and\nclinical implications for disease management Cell Reports Medicine 4\nSapkota Y, Steinthorsdottir V, Morris AP, Fassbender A, Rahmioglu N, De Vivo I, et al. (2017)\nMeta-analysis identifies five novel loci associated with endometriosis highlighting key\ngenes involved in hormone metabolism Nat Commun 8\nMatalliotaki C, Matalliotakis M, Rahmioglu N, Mavromatidis G, Matalliotakis I, Koumantakis G,\net al. (2019) Role of FN1 and GREB1 gene polymorphisms in endometriosis Mol Med Report\nTrachsel B, Imobersteg S, Valpreda G, Singer G, Grabherr R, Ormos M, et al. (2024) Relaxed\nfibronectin: a potential novel target for imaging endometriotic lesions EJNMMI Res 14\nLis-Kuberka J, Kubik P, Chrobak A, Pająk J, Chełmońska-Soyta A, Orczyk-Pawiłowicz M (2021)\nFibronectin Molecular Status in Plasma of Women with Endometriosis and Fertility\nDisorders IJMS 22\n27.\n28.\n29.\n30.\n31.\n32.\n33.\n34.\n35.\n36.\n37.\n38.\n39.\n40.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 21 of 26\nWarzecha D, Załęcka J, Mańka G, Kiecka M, Lipa M, Spaczyński R, et al. (2022) Plasma and\nPeritoneal Fluid Fibronectin and Collagen IV Levels as Potential Biomarkers of\nEndometriosis IJMS 23\nTseng JF, Ryan IP, Milam TD, Murai JT, Schriock ED, Landers DV, et al. (1996) Interleukin-6\nsecretion in vitro is up-regulated in ectopic and eutopic endometrial stromal cells from\nwomen with endometriosis The Journal of Clinical Endocrinology & Metabolism 81:1118–22\nAibar S, González-Blas CB, Moerman T, Huynh-Thu VA, Imrichova H, Hulselmans G, et al. (2017)\nSCENIC: single-cell regulatory network inference and clustering Nat Methods 14:1083–6\nMeola J, Hidalgo GDS, Silva JCRE, Silva LECM, Paz CCP, Ferriani RA (2013) Caldesmon: New\nInsights for Diagnosing Endometriosis1 Biology of Reproduction 88\nLiang Y, Wang W, Huang J, Tan H, Liu T, Shang C, et al. (2015) Potential Role of Semaphorin 3A\nand Its Receptors in Regulating Aberrant Sympathetic Innervation in Peritoneal and\nDeep Infiltrating Endometriosis PLoS ONE 10\nWu J, Xie H, Yao S, Liang Y (2017) Macrophage and nerve interaction in endometriosis J\nNeuroinflammation 14\nAghajanova L, Giudice LC (2011) Molecular Evidence for Differences in Endometrium in\nSevere Versus Mild Endometriosis Reprod Sci 18:229–51\nJohnson MC, Torres M, Alves A, Bacallao K, Fuentes A, Vega M, et al. (2005) Augmented cell\nsurvival in eutopic endometrium from women with endometriosis: Expression of c-myc,\nTGF-beta1 and bax genes Reprod Biol Endocrinol. 3\nKisovar A, Becker CM, Granne I, Southcombe JH (2023) The role of CD8+ T cells in\nendometriosis: a systematic review Front Immunol 14\nUhlén M, Fagerberg L, Hallström BM, Lindskog C, Oksvold P, Mardinoglu A, et al. (2015) Tissue-\nbased map of the human proteome Science 347\nJuárez-Barber E, Corachán A, Carbajo-García MC, Faus A, Vidal C, Giles J, et al. (2024)\nTranscriptome analysis of adenomyosis eutopic endometrium reveals molecular\nmechanisms involved in adenomyosis-related implantation failure and pregnancy\ndisorders Reprod Biol Endocrinol 22\nGirling JE, Lockhart MG, Olshansky M, Paiva P, Woodrow N, Marino JL, et al. (2017) Differential\nGene Expression in Menstrual Endometrium From Women With Self-Reported Heavy\nMenstrual Bleeding Reprod Sci 24:28–46\nKobayashi H, Yamashita Y, Iwase A, Yoshikawa Y, Yasui H, Kawai Y, et al. (2012) The\nferroimmunomodulatory role of ectopic endometriotic stromal cells in ovarian\nendometriosis Fertility and Sterility 98:415–422\nFranzén O, Gan LM, Björkegren JLM (2019) PanglaoDB: a web server for exploration of\nmouse and human single-cell RNA sequencing data Database\nZheng GXY, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, et al. (2017) Massively parallel\ndigital transcriptional profiling of single cells Nat Commun 8\n41.\n42.\n43.\n44.\n45.\n46.\n47.\n48.\n49.\n50.\n51.\n52.\n53.\n54.\n55.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 22 of 26\nFleming SJ, Chaffin MD, Arduini A, Akkad AD, Banks E, Marioni JC, et al. (2023) Unsupervised\nremoval of systematic background noise from droplet-based single-cell experiments\nusing CellBender Nat Methods\nHao Y, Hao S, Andersen-Nissen E, Mauck WM, Zheng S, Butler A, et al. (2021) Integrated\nanalysis of multimodal single-cell data Cell 184:3573–3587\nPliner HA, Shendure J, Trapnell C (2019) Supervised classification enables rapid annotation\nof cell atlases Nat Methods 16:983–6\nSubramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. (2005) Gene\nset enrichment analysis: A knowledge-based approach for interpreting genome-wide\nexpression profiles Proc Natl Acad Sci USA 102:15545–50\nAlpern D, Gardeux V, Russeil J, Mangeat B, Meireles-Filho ACA, Breysse R, et al. (2019) BRB-seq:\nultra-affordable high-throughput transcriptomics enabled by bulk RNA barcoding and\nsequencing Genome Biol 20\nChen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, et al. (2013) Enrichr: interactive and\ncollaborative HTML5 gene list enrichment analysis tool BMC Bioinformatics 14\nLove MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for\nRNA-seq data with DESeq2 Genome Biol 15\nLiberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P (2015) The Molecular\nSignatures Database Hallmark Gene Set Collection Cell Systems 1:417–25\nAuthor information\nPetra C Schwalie\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, Roche\nInnovation Center Basel, Pharma Research and Early Development, Basel, Switzerland\nORCID iD: 0000-0002-6004-8095\nCemsel Bafligil\nWellcome Trust Sanger Institute, Wellcome Genome Campus, Cambridge, UK\nJulie Russeil\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland\nMagda Zachara\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland\nMarjan Biocanin\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland\n56.\n57.\n58.\n59.\n60.\n61.\n62.\n63.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 23 of 26\nDaniel Alpern\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland\nEvelin Aasna\nGerman Cancer Research Center (DKFZ), Somatic Evolution and Early Detection Heidelberg,\nGermany\nBart Deplancke\nLaboratory of Systems Biology and Genetics, Institute of Bioengineering, School of Life\nSciences, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland\nGeraldine Canny\nUniversity College Dublin, Research Office, Dublin, Ireland\nAngela Goncalves\nWellcome Trust Sanger Institute, Wellcome Genome Campus, Cambridge, UK, German Cancer\nResearch Center (DKFZ), Somatic Evolution and Early Detection Heidelberg, Germany\nORCID iD: 0000-0002-3890-7658\nFor correspondence: a.goncalves@dkfz.de\nEditors\nReviewing Editor\nAxelle Brulport\nSenior Editor\nDiane Harper\nUniversity of Michigan-Ann Arbor, Ann Arbor, United States of America\nReviewer #1 (Public review):\nSummary:\nThe characteristics of endometrium health are an increasing topic in women's health issues,\nespecially in the context of endometriosis. In this respect, having access to information is\nhampered by the inaccessibility of the uterine tissue. The authors propose here using the\nmenstrual fluid (easily accessible by non-invasive methods) as an access door towards getting\nrelevant information.\nOverall, the paper is divided into two parts:\n(1) The comparison between menstrual fluid samples and biopsies of the endometrium.\n1. As a proof of concept, the authors then compared 11 controls and 7 endometriosis\ncases in this way, from different severity stages.\nStrengths:\nIn Figure 1, general features of the 15 samples are presented (volume/number of\ncells/hematopoietic cells - cd45 labeling). The authors then used single-cell RNA-seq to\ncharacterize the different samples. Through having access to endometrium biopsies, they\nwere able to compare the profiles obtained.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 24 of 26\nIn the MF samples from the second part of the paper - aiming at comparing endometriosis\nand controls - one question is raised about the effect of culture. The authors compared freshly\nisolated and cultured tissues (ex vivo vs in vitro) by bulk RNA seq. Biases induced by the\nculture procedure were identified. Deconvolution was applied to strengthen this observation,\nwith an important increase of seemingly stromal and unknown cells, especially in the\nunsorted cells and the CD45+ cells.\nInterestingly, since the authors got successive samples from the same donor, they could\nevaluate the consistency of the samples and reveal indeed an overall stability of the\nmolecular profile of the samples in a given patient.\nThe authors then attempted - quite originally - to characterize biomarkers in two major cell\ncompartments that they studied - CD45- (stromal-like) and CD45+ (immune cells).\nWeaknesses:\nA potential problem is the justification of the a priori mix of cell types of three different\nphenotypes (CD45+, CD45- EPCAM+, and CD45- EPCAM-) from each patient before moving to\nthe scRNAseq. It is not clear to me why this has been done, I guess that using directly the\nsamples would supposedly bias the result. But in this case, why is it supposed that three\ncategories are enough (immune cells, epithelial cells, and stromal cells)? I suppose that other\nmarkers could characterize other subtypes of the cells, and take into account the possibility of\nother cell types, for instance, connected to pain sensitivity, such as neuron precursor. Hence,\nthe justification of the organized mixes should be much more detailed in my opinion.\nIt is a bit unclear to me when the biopsies were collected in the cycle of the donor patients.\nThe description of these markers that are deregulated is presented as a list, and connected\nwith existing publications, which could rather be presented in discussion than in the results.\nThe authors do tend to demonstrate that the Menstrual Fluid is a good proxy to analyse the\nendometrium health status of the women affected with endometriosis.\nThe identification of MTRNR2L1 seems to be a major discovery of the paper, as well as in a\nlesser measure HBG2, and it is a bit strange why these putative markers were not emphasized\nin the abstract. HBG2 was certainly identified previously in endometriosis endothelial cells\nbut seems extremely variable from one sample to another - Geo profile (GDS3060, GDS3060 /\n213515_x_at (inist.fr)).\nOverall, the transcriptome analysis is a bit shallow, with no effort made to try to find\npotential transcription factors or miRNA that could activate/inhibit a series of modified\ngenes; it could be relevant to identify such master genes or master regulators through\nbioinformatics analyses and wet-lab validations, to understand better the cascade of events.\nAnother issue that was overlooked is the presence of 'stem-cells' in the MF obtained. Since\nendometriosis is supposed to occur from the implantation of uterine stem cells, this category\ncould be a major topic of scrutiny, in terms of quantity in the MF, as well as in terms of their\nspecific molecular properties.\nhttps://doi.org/10.7554/eLife.99558.1.sa1\nReviewer #2 (Public review):\nSummary:\nThe authors provided further evidence that menstrual fluid (MF) can be used as a non-\ninvasive source of endometrial tissue for studying its normal physiological state and when it\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 25 of 26\nis abnormal such as in endometriosis. Single-cell RNA sequencing confirmed the presence of\nthe major cell types -blood and tissue immune cells and endometrial stromal, epithelial, and\nvascular cells. The major new finding was that interindividual variation for the blood\nimmune cells was minimal between multiple MF samples from an individual. A comparison\nbetween the ex vivo MF gene profile and cultured MF showed the expected attachment and\nculture of stromal (and a small number of epithelial) cells, but the immune cells failed to\nattach. Several differentially expressed genes between controls and endometriosis were\nsuggested as potential biomarkers of the disease, however, these were a mitochondrial\npseudogene and a hemoglobin subunit, both very unlikely related to endometriosis\npathogenesis.\nStrengths:\nThe Spearman correlation analysis between the control MF gene profiles of multiple samples\nfrom the same individual and its graphic presentation provided strong evidence that there is\nlittle variation between MF samples. Together with another study which showed similar\nfindings for endometrial stem cells and a number of proteins in MF supernatant, this\nimportant data shows MF as a promising biofluid for pathology testing.\nThe bioinformatic analyses conducted by bioinformatic and computational experts are a\nmajor strength of the manuscript and in particular the comparison between MF and\nendometrial biopsy data obtained from published scRNAseq studies. This is an important\nfinding, particularly if comparisons included late secretory and early proliferative stage\nbiopsy tissue which would be most similar to shedding menstrual endometrium.\nThe inclusion of workflows in the Figures for the various studies and the use of symbols in\nthe various panels is very helpful for the reader.\nMF cell suspensions were enriched for stromal and epithelial cells to enable a detailed\nbioinformatic analysis of their respective gene profiles\nWeaknesses:\nTwo patient cohorts from different institutions were used in the study and somewhat\ndifferent methods were used to extract the cellular fraction from these cohorts for further\nstudy: (1) sample dilution and differential filtration to separate blood-derived immune cells\nfrom endometrial tissue then dissociated into single cells and separated into CD45+, CD45-\nEpCAM+ and CD45-EpCAM- cells, and (2) gradient density separation to generate unsorted,\nCD45+, CD45- and putative mesenchymal stem cells (MSC) CD45-CD105+ which were also\ncultured. In addition, questions on pelvic pain and proven fertility would have addressed the\n2 key symptoms of endometriosis.\nThe use of CD105 to purify MSC from MF rather than well-characterised markers of\nclonogenic, self-renewing, and mesodermal differentiating endometrial MSC such as\nCD146+PDGFRB+ or SUSD2 (both mentioned in references 22 and 23) is a weakness. The ISCT\nmarkers are not specific and are also found on stromal fibroblasts of many tissues (Phinney\nand Sensebe Cytotherapy 2013; Demu et al Acta Haematologica 2016).\nThe UMAPs generated from the scRNAseq were at low resolution and more individual\nimmune and endometrial cell types have previously been identified and reported in MF.\nMore comparisons with these studies would also have enhanced the Discussion.\nIt was not always possible to work out how the data was reported in the gene expression\ntables (Supplementary Tables 2, 4-10) as they were not in adjusted P value order and\nsometimes positive log2 fold change values appeared amongst the negative log2FC. In some\ncomparisons described, the adj P values were not significant but were described as up or\ndown-regulated in the text.\n\nPetra C Schwalie et al., 2024 eLife. https://doi.org/10.7554/eLife.99558.1 26 of 26\nThe 2 DEGs highlighted in the endometriosis and control arm of the study appear as poor\nchoices from many others that could have been chosen as MTRNR2L1 is a mitochondrial\npseudogene and HBG2 is a hemoglobin subunit. Neither are likely indicators of endometriosis\npathogenesis.\nThe manuscript format and organisation could be improved by reducing the discussion in the\nResults section and providing a more in-depth Discussion. More references need to be\nincluded in the Discussion and other work in the MF analysis field that supports - or not - the\nauthors' findings or at least puts them into context, and should be included and referenced.\nThe potential to use MF as a non-invasive source of endometrial tissue for potential diagnosis\nis a very important avenue of research that is currently in its infancy and could have a major\nimpact in the endometriosis research arena.\nhttps://doi.org/10.7554/eLife.99558.1.sa0","source_license":"CC0","license_restricted":false}