Single-cell and spatial transcriptomic profiling revealed niche interactions sustaining growth of endometriotic lesions

J. Leng and X.W. conceived and designed the study. S.L. and Z.G. performed the single-cell experiments. S.L. analyzed and interpreted the data. X.L., S.J., Y.W., and H.Y. performed the biopsies. Z.G., J.W., J.Z., and L.L. performed immunofluorescence and immunohistochemical experiments. Y.Y. and X.X. confirmed the pathological diagnosis of the biospecimen. S.L., X.W., S.J., and J. Leng wrote the manuscript. J.S., Y.D., and J. Lang discussed the manuscript.

To fully characterize the molecular profile of ectopic lesions, we performed single-cell sequencing on 21 specimens, including 8 endometriomas and 13 eutopic endometria, collected from 17 patients. Among the eutopic endometria, 8 were from patients with endometriosis and 5 from patients without endometriosis. All participants exhibited a regular menstrual cycle ranging from 27 to 33 days and had not received hormonal therapy at least 3 months prior to surgery. For patients with bilateral ovarian endometriosis, only one specimen per patient was included. The detailed clinical information for these specimens is listed in Table 1 . The hematoxylin and eosin (H&E) staining of 8 endometriomas is shown in Figure S1 . Table 1 Sample clinical information Specimen Patient ID Age Diagnosis Tissue Days after menstruation Menstrual phase Cell count Sample 1 p1 23 teratoma endometrium 11 proliferative 7,424 Sample 2 p2 35 infertility endometrium 12 proliferative 6,802 Sample 3 p3 28 OEM endometrium 13 proliferative 3,836 Sample 4 p4 34 OEM endometrium 14 proliferative 9,552 Sample 5 p5 39 OEM endometrium 9 proliferative 3,237 Sample 6 p6 34 OEM & DIE endometrium 13 proliferative 4,421 Sample 7 p7 26 teratoma endometrium 24 secretory 5,857 Sample 8 p8 34 Teratoma endometrium 27 secretory 4,691 Sample 9 p9 42 cervical cancer endometrium 23 secretory 6,333 Sample 10 p10 43 OEM endometrium 23 secretory 5,381 Sample 11 p11 32 OEM & DIE endometrium 20 secretory 7,260 Sample 12 p12 39 OEM & DIE endometrium 23 secretory 4,117 Sample 13 p13 29 OEM endometrium 22 secretory 4,392 Sample 14 p5 39 OEM ovarian endometrioma 9 proliferative 3,123 Sample 15 p6 34 OEM & DIE ovarian endometrioma 13 proliferative 6,489 Sample 16 p14 26 OEM ovarian endometrioma 18 secretory 7,352 Sample 17 p15 28 OEM ovarian endometrioma 23 secretory 4,544 Sample 18 p16 19 OEM ovarian endometrioma 22 secretory 4,255 Sample 19 p13 29 OEM ovarian endometrioma 22 secretory 3,153 Sample 20 p3 28 OEM ovarian endometrioma 13 proliferative 2,130 Sample 21 p17 27 OEM ovarian endometrioma 19 secretory 7,132 OEM, ovarian endometriosis; DIE, deep infiltrating endometriosis. Sample clinical information OEM, ovarian endometriosis; DIE, deep infiltrating endometriosis. From the 13 eutopic endometrium specimens, we obtained transcriptomes for 73,303 cells with a median of 2,344 genes per cell. Based on their gene expression profiles, these cells were categorized into 10 major types ( Figures 1 B and 1D): epithelial cells, ciliated epithelial cells, endometrial stromal cells, endothelial cells, lymphatic endothelial cells (LECs), pericytes, and four groups of immune cell types. Endometrial stromal (EnS) cells, marked by specific expression of MME ( Figure 1 D), were the most abundant in eutopic endometrium ( Figure 1 C). This is consistent with the use of CD10, the protein encoded by MME , as the immunohistochemical marker for identifying endometrial stroma. 28 We also noticed that both the epithelial and stromal cells in the endometrium showed phase-specific clustering, consistent with changes in cellular dynamics during the menstrual cycle ( Figures S2 A and S2B). When compared with a well-annotated single-cell transcriptome dataset from 11 endometrium samples (ArrayExpress: E-MTAB-10287 24 ), we identified similar cell type compositions and subtypes ( Figures S2 C–S2E). Figure 1 scRNA-seq identified ectopic EnS cells and lesion-specific stromal cells in endometriosis (A) Diagram of specimen collection for single-cell transcriptome profiling. (B and E) Uniform manifold approximation and projection (UMAP) plots of cell types identified from the 13 endometria (B) and 8 ovarian endometriomas (E). Colors indicate different cell types. See also Figures S2–S4 . (C and F) Bar plots of cell compositions for the endometria (C) and endometriomas (F). Colors indicate different cell types. Error bars represent standard deviations. (D and G) Violin plots displaying the marker gene expression for each identified cell type. EnS cell, endometrial stromal cell; LEC, lymphatic endothelial cell; OSC, ovarian stromal cell. (H and I) Immunofluorescence staining of endometrium (H) and ovarian endometrioma (I). DAPI (blue) indicates the cell nucleus, Ecad (E-cadherin; green) indicates epithelial cells, TCF21 (red) indicates OSCs, and CD10 (purple) indicates EnS s. EM, endometrium; OEM, ovarian endometrioma. Scale bars represent 1 mm (upper) or 50 μm (lower). scRNA-seq identified ectopic EnS cells and lesion-specific stromal cells in endometriosis (A) Diagram of specimen collection for single-cell transcriptome profiling. (B and E) Uniform manifold approximation and projection (UMAP) plots of cell types identified from the 13 endometria (B) and 8 ovarian endometriomas (E). Colors indicate different cell types. See also Figures S2–S4 . (C and F) Bar plots of cell compositions for the endometria (C) and endometriomas (F). Colors indicate different cell types. Error bars represent standard deviations. (D and G) Violin plots displaying the marker gene expression for each identified cell type. EnS cell, endometrial stromal cell; LEC, lymphatic endothelial cell; OSC, ovarian stromal cell. (H and I) Immunofluorescence staining of endometrium (H) and ovarian endometrioma (I). DAPI (blue) indicates the cell nucleus, Ecad (E-cadherin; green) indicates epithelial cells, TCF21 (red) indicates OSCs, and CD10 (purple) indicates EnS s. EM, endometrium; OEM, ovarian endometrioma. Scale bars represent 1 mm (upper) or 50 μm (lower). From the 8 ovarian endometriomas (OEMs), we generated expression data on 38,178 cells with a median of 1,647 genes per cell ( Figure 1 E). Unlike eutopic endometrium, no distinct clusters of epithelial or ciliated epithelial cells were observed. However, a small set of cells co-expressed EPCAM and WFDC2 ( Figure S3 A), indicative of the presence of endometrial epithelial cells. Meanwhile, ectopic EnS cells expressing MME formed a distinct cluster, accounting for 1.5% of all cells ( Figure 1 F). When we integrated the data from eutopic endometria and endometriomas, ectopic EnS cells from OEMs clustered together with eutopic EnS cells ( Figures S3 B and S3C), confirming their endometrium identity. Interestingly, another group of LUM -expressing stromal cells, characterized by specific expression of TCF21 and STAR ( Figures 1 E and 1G), was exclusively identified within endometriomas as the predominant cell type ( Figure 1 F). To investigate the potential ovarian origin of these stromal cells, we integrated our endometrioma with two published normal-ovary single-cell datasets (GEO: GSE118127 and GSE260685 ). 29 , 30 We found that these stromal cells clustered together with the theca/stromal cells from the normal ovary datasets ( Figures S4 A and S4B), which also specifically expressed TCF21 and STAR at high levels ( Figure S4 C). Gene expression correlation analysis showed that, while these cells share common fibroblast markers with EnS cells, they form distinct cell populations with moderate correlation ( Figure S4 D). Considering the hypothesis that endometriomas may develop through the inversion and progressive invagination of the ovarian cortex following the formation of superficial endometriotic implants on the ovarian surface, 31 we designated these stromal cells as ovarian stromal cells (OSCs). A group of STAR -expressing fibroblasts specifically in endometriomas has also been reported by Ma et al., though they did not specify the origin of these cells. 14 Immunofluorescence staining validated the lining of ectopic endometrial epithelial and stromal cells along OSCs in patient samples independent of the sequencing cohort. In the eutopic endometrium, CD10 signal marked EnS cells without any detectable TCF21 expression ( Figure 1 H). In ovarian endometriomas, we observed a layered structure. E-cadherin (Ecad) staining delineated a layer of ectopic endometrial epithelial cells. Adjacent to this, there was a layer of CD10 positive EnS cells, which were surrounded by a separate layer of TCF21-positive OSCs ( Figures 1 I and S3 D). We also compared our results with an independent dataset (GEO: GSE213216 16 ), which included scRNA-seq data of 54 samples from endometria, peritoneal endometriosis, ovarian endometriomas, and unaffected ovaries ( Figures S4 E–S4G). We observed similar cell compositions in eutopic endometria and ovarian endometriomas ( Figure S4 G). As expected, no MME -expressing EnS cells were found in unaffected ovaries, while OSCs with specific expression of TCF21 and STAR were present in both unaffected ovaries and endometriomas ( Figure S4 G). In peritoneal endometriosis, ectopic EnS cells formed a distinct cluster, separated from the tissue-specific stromal cells marked by high SFRP2 expression ( Figures S4 F and S4G). All of these stromal cells—EnS cells, OSCs, and peritoneal stromal cells—show specific expression of LUM and PDGFRA , which were widely expressed on fibroblasts; however, each of them has tissue-specific markers ( Figure S4 E). Together, the accurate identification of ectopic EnS cells and tissue-specific stromal cells in scRNA-seq data allows us to explore the unique characteristics of ectopic EnS cells and to investigate the cellular interactions between the implanted endometrium and the surrounding cells. Human endometrium undergoes periodic shedding and regeneration during the menstrual cycle. Previous studies have characterized the transcriptional variation of endometrial epithelial and stromal cells during proliferative and secretory phases. 23 Given that ectopic endometriotic lesions also exhibit cyclic bleeding, we aimed to explore whether ectopic EnS cells maintain responsiveness to hormonal changes during the menstrual cycle by examining their transcriptome alterations. We first performed pseudotime analysis of EnS cells from 13 eutopic endometria. Eutopic EnS cells from both endometriosis patients and controls followed a similar differentiation trajectory ( Figures S5 A and S5B), with distinct separation between EnS cells from the proliferative phase and secretory phase ( Figure 2 A), recapitulating known distinctions in proliferative versus secretory EnS cells. 23 To quantitatively define menstrual phase signatures, we derived a proliferative EnS cell score (PES) based on expression of the 89 proliferative phase-specific genes and a secretory EnS cell score (SES) from the 113 overexpressed genes in the secretory phase ( Table S1 ). These menstrual phase-specific genes were also confirmed by comparison with the data from Garcia-Alonso et al. 24 We found that 85 of our 89 proliferation-specific genes were upregulated in their endometrial stroma (eS), and 102 of our 113 secretory phase-specific genes were upregulated in their decidualized stroma (dS) ( Figures S5 C and S5D). In eutopic EnS cells, PES decreased while SES increased along the pseudotime axis of the trajectory ( Figures 2 B and 2C). We validated these scores using single-cell data from 19 healthy ovum donors spanning different cycle days (Wang et al. 23 ) ( Figures 2 D and 2E), as well in the classified eS or dS ( Figure S5 E), both demonstrated that the changes in PES and SES correlated well with menstrual stages. Figure 2 The menstrual dynamics of ectopic EnS cells (A) Heatmap of the gene expression dynamics for eutopic EnS cells along the pseudotime axis of the menstrual cycle. (B and C) Dot plot of PES (B) and SES (C) for each EnS cell identified in our data along the menstrual pseudotime. PES, proliferative phase EnS score; SES, secretory phase EnS score. (D and E) Violin plots of PES (D) and SES (E) for EnS cells using the GSE111976 dataset of healthy women. Pro, proliferative phase; Sec_early, early secretory phase; Sec_early-mid, early to mid-secretory phase; Sec_mid, mid-secretory phase; Sec_late, late secretory phase. (F and G) Comparison of proliferative and secretory gene expression scores of eutopic and ecotpic EnS cells between our data and an independent endometriosis dataset ( GSE213216 ). EM_Pro, eutopic endometrium from the proliferative phase; EM_Sec, eutopic endometrium from the secretory phase; OEM_Pro, ovarian endometriomas from the proliferative phase; OEM_Sec, ovarian endometriomas within the secretory phase. PEM_Pro and PEM_Sec denote peritoneal endometriosis within the proliferative phase and secretory phases, respectively. Student’s t test was used to compare the gene expression score of EnS cells between the proliferative phase and secretory phase. Δ = mean of PES – mean of SES. (H and I) Dot plots of representative menstrual-phase gene expression in eutopic and ectopic EnS cells for our data (H) and GSE213216 . (J) Heatmap of overexpressed genes in eutopic and ectopic EnS cells across different menstrual phases. The differentially expressed genes were identified by Wilcoxon rank-sum test with log 2 fold change > 1 and adjusted p  < 0.05. (K) Venn diagram displaying the number of overexpressed genes in ectopic EnS cells for the proliferative phase and secretory phase. (L) GO enrichment for the 42 overexpressed genes along the menstrual cycle. See also Figure S5 . The menstrual dynamics of ectopic EnS cells (A) Heatmap of the gene expression dynamics for eutopic EnS cells along the pseudotime axis of the menstrual cycle. (B and C) Dot plot of PES (B) and SES (C) for each EnS cell identified in our data along the menstrual pseudotime. PES, proliferative phase EnS score; SES, secretory phase EnS score. (D and E) Violin plots of PES (D) and SES (E) for EnS cells using the GSE111976 dataset of healthy women. Pro, proliferative phase; Sec_early, early secretory phase; Sec_early-mid, early to mid-secretory phase; Sec_mid, mid-secretory phase; Sec_late, late secretory phase. (F and G) Comparison of proliferative and secretory gene expression scores of eutopic and ecotpic EnS cells between our data and an independent endometriosis dataset ( GSE213216 ). EM_Pro, eutopic endometrium from the proliferative phase; EM_Sec, eutopic endometrium from the secretory phase; OEM_Pro, ovarian endometriomas from the proliferative phase; OEM_Sec, ovarian endometriomas within the secretory phase. PEM_Pro and PEM_Sec denote peritoneal endometriosis within the proliferative phase and secretory phases, respectively. Student’s t test was used to compare the gene expression score of EnS cells between the proliferative phase and secretory phase. Δ = mean of PES – mean of SES. (H and I) Dot plots of representative menstrual-phase gene expression in eutopic and ectopic EnS cells for our data (H) and GSE213216 . (J) Heatmap of overexpressed genes in eutopic and ectopic EnS cells across different menstrual phases. The differentially expressed genes were identified by Wilcoxon rank-sum test with log 2 fold change > 1 and adjusted p  < 0.05. (K) Venn diagram displaying the number of overexpressed genes in ectopic EnS cells for the proliferative phase and secretory phase. (L) GO enrichment for the 42 overexpressed genes along the menstrual cycle. See also Figure S5 . Intriguingly, ectopic EnS cells showed significant variations in PES and SES between proliferative and secretory phases ( Figure 2 F), although the dynamic range was compressed compared to eutopic EnS cells. Further analysis of 8 proliferative and 8 secretory EnS cell genes revealed similar expression trends for each gene matching the cycle phases in both ectopic and eutopic EnS cells ( Figure 2 H). These variations in PES and SES and the representative gene expression were confirmed in an independent endometriosis dataset ( Figures 2 G and 2I). In the independent dataset, peritoneal EnS cells exhibited PES/SES shift variations with a dynamic range comparable to eutopic endometrium ( Figure 2 G). In contrast, endometrioma showed a compressed range, indicating a tempered response of EnS cells to hormonal changes in the cyst environment. These results suggested that ectopic EnS cells retain intrinsic programming to undergo cyclic gene expression changes reminiscent of eutopic endometrium. We did not observe dynamic changes in the expression of menstrual cycle-related genes in OSCs as seen in EnS cells ( Figure S5 F). However, when comparing OSCs between proliferative and secretory phases, we identified 25 differentially expressed genes ( Figure S5 G). We next directly compared the gene expression of ectopic and eutopic EnS cells at proliferative and secretory phases. While dynamically expressed menstrual cycle genes (e.g., MMP11 ) were expressed consistently ( Figure 2 H), a group of genes was aberrantly overexpressed in ectopic EnS cells, such as CXCL8 , CXCL2 , and WNT5A ( Figure 2 J). For the proliferative phase, 106 genes were overexpressed, while 204 genes were overexpressed during the secretory phase. Among them, 42 genes were upregulated continuously throughout the menstrual cycle ( Figure 2 K; Table S2 ). The Gene Ontology (GO) terms of these genes were enriched in biological processes of wound healing, regulation of angiogenesis, epithelial cell proliferation, tissue migration, regulation of cell-cell adhesion, and inflammatory response ( Figure 2 L). These results suggest that ectopic EnS cells retain cyclical gene expression patterns similar to eutopic EnS cells but also exhibit distinct expression that may be related to the pathogenesis of endometriosis. To investigate the roles of OSCs in the pathogenesis of endometriosis, we first characterized their gene expression profiles in both our data and the Fonseca et al. dataset. 16 Interestingly, this analysis identified two distinct subpopulations of OSCs, which we termed OSCa and OSCb. Among them, OSCa showed high expression of expressed TCF21 and STAR . However, OSCb cells exhibited significantly elevated levels of FAP and COL10A1 ( Figures 3 A and 3B). Figure 3 Spatial profiling reveals the zonation of two subtypes of OSCs in endometriomas (A and B) UMAP plots of the two subgroups of OSCs, OSCa and OSCb, in our dataset (A) and  GSE213216 (B), displaying their respective marker gene expression. (C) H&E staining of formalin-fixed, paraffin-embedded sections from a patient with ovarian endometriosis. The scale bar represents 6.5 mm. (D) Unsupervised clustering of spatial transcriptomic spots from the sections in (C). A dominant cell type is assigned to each spot following the identification of marker genes within the clusters. OSCm, spots expressing markers of both OSCa and OSCb; OUn, unclassified cells. The scale bar represents 6.5 mm. See also Figure S6 . (E) Dot plot illustrating the overexpressed genes corresponding to each identified cell type from the spatial transcriptomic data. See also Figure S6 . (F) Marker gene expression in the spatial transcriptomic data for EnS cells, OSCb, and OSCa. MMP11 represents ectopic EnS cells, COL10A1 and FAP mark OSCb, and TCF21 marks OSCa. Color intensity indicates gene expression levels after sctransform in Seurat. See also Figure S6 . (G and H) Immunofluorescence staining of ectopic EnS cells and the two groups of OSCs in tissue sections of endometriomas. DAPI stains the cell nucleus, TCF21 marks OSCa, CD10 identifies ectopic EnS cells, and COL10A1 and FAP mark OSCb. Scale bars represent 500 μm. See also Figure S7 . (I) Dot plot illustrating the overexpressed genes in three groups of OSCs derived from GSE213216 , displaying the percentage of cells expressing each gene and the average expression level. (J) GO enrichment analysis of the overexpressed genes in each group of OSCs. Hypergeometric test was used to assess the gene enrichment, and false discovery rate was applied to account for multiple comparisons. Spatial profiling reveals the zonation of two subtypes of OSCs in endometriomas (A and B) UMAP plots of the two subgroups of OSCs, OSCa and OSCb, in our dataset (A) and  GSE213216 (B), displaying their respective marker gene expression. (C) H&E staining of formalin-fixed, paraffin-embedded sections from a patient with ovarian endometriosis. The scale bar represents 6.5 mm. (D) Unsupervised clustering of spatial transcriptomic spots from the sections in (C). A dominant cell type is assigned to each spot following the identification of marker genes within the clusters. OSCm, spots expressing markers of both OSCa and OSCb; OUn, unclassified cells. The scale bar represents 6.5 mm. See also Figure S6 . (E) Dot plot illustrating the overexpressed genes corresponding to each identified cell type from the spatial transcriptomic data. See also Figure S6 . (F) Marker gene expression in the spatial transcriptomic data for EnS cells, OSCb, and OSCa. MMP11 represents ectopic EnS cells, COL10A1 and FAP mark OSCb, and TCF21 marks OSCa. Color intensity indicates gene expression levels after sctransform in Seurat. See also Figure S6 . (G and H) Immunofluorescence staining of ectopic EnS cells and the two groups of OSCs in tissue sections of endometriomas. DAPI stains the cell nucleus, TCF21 marks OSCa, CD10 identifies ectopic EnS cells, and COL10A1 and FAP mark OSCb. Scale bars represent 500 μm. See also Figure S7 . (I) Dot plot illustrating the overexpressed genes in three groups of OSCs derived from GSE213216 , displaying the percentage of cells expressing each gene and the average expression level. (J) GO enrichment analysis of the overexpressed genes in each group of OSCs. Hypergeometric test was used to assess the gene enrichment, and false discovery rate was applied to account for multiple comparisons. To further explore the relationship between these two subtypes of OSCs and the ectopic EnS cells, we performed spatial transcriptome sequencing on three ovarian endometriomas, sequencing a total of 6,412 tissue spots with a diameter of 55 μm using the 10× Visium platform ( Figures 3 C, 3D, and S6 A–S6D). Unsupervised clustering categorized spots into known cell types, including ectopic EnS cells ( MME and MMP11 ), OSCa ( TCF21, C3 ), OSCb ( COL10A1, FAP ), pericytes ( MCAM and RGS5 ) and endothelial cells ( VWF and CD34 ) ( Figures 3 D and 3E). We also identified an intermediate group co-expressing OSCa and OSCb markers, possibly reflecting multi-cellular spots ( Figures 3 D, S6 B, and S6D). Notably, the expression of marker genes revealed an organized structure within the lesion: ectopic EnS cells formed discrete patches that were closely surrounded by OSCb cells, with OSCa cells predominantly located in the outer regions ( Figures 3 F and S6 E). To validate the spatial zonation of ectopic EnS cells and subtypes of OSCs, we performed immunofluorescence staining on four additional samples that were not included in spatial transcriptomics or scRNA-seq ( Figures 3 G and S7 A). We used CD10 to mark the ectopic EnS cells. Adjacent to the CD10-positive EnS cell zone, COL10A1-positive cells with low TCF21 expression were identified, corresponding to the OSCb subgroup. Surrounding this area were cells that highly expressed TCF21 but lacked COL10A1 expression, indicative of the OSCa subgroup ( Figures 3 G and S7 A). We also performed staining using FAP, which is expressed in both OSCb and EnS cells ( Figures 3 H and S7 B). These results consistently demonstrated the same spatial organization across all samples, with EnS cells surrounded by OSCb and OSCa in the outer regions. To explore the functions of different OSC subgroups in endometriomas, we compared the OSCs from endometriosis and the unaffected ovaries using the Fonseca et al. dataset. 16 Our analysis revealed distinct gene expression profiles in OSCa and OSCb subgroups. While both OSCa and OSCb from endometriomas showed elevated expression of inflammatory mediators and complement activation regulators, including IL6ST , C3 , and C7 , compared to OSCs from unaffected ovaries, these genes appeared to be more strongly expressed in OSCa ( Figures 3 I and 3J). This suggests that OSCa may play a more prominent role in regulating immune responses within the endometriotic lesions. In contrast, OSCb exhibited significantly higher levels of extracellular matrix and remodeling genes, including FAP , COL10A1 , COL3A1 , CTHRC1 , and FBLN1 ( Figure 3 I), potentially associated with pathways of connective tissue development, cell adhesion, and fibroblast migration ( Figure 3 J). To better understand the microenvironment of endometriotic lesions, we next investigated the spatial distribution of immune cells. Although we were unable to identify distinct immune cell clusters in the spatial transcriptomic data due to the limited resolution (each 55-μm-diameter spot may contain multiple cell types), we were able to analyze the distribution of immune cell markers within the tissue ( Figures 4 A, S8 A, and S8B). Figure 4 Spatial profiling reveals immune cell enrichment in distinct zones of endometriotic lesions (A) Marker gene expression in the spatial transcriptomic data for EnS cells, T cells, macrophages, and B cells. MMP11 represents ectopic EnS cells, while CD3D identifies T cells. CD68 and CD79A mark macrophages and B cells, respectively. Color intensity indicates gene expression levels after sctransform in Seurat. See also Figure S8 . (B) Dot plot illustrating immune cell marker expression across different clusters of spots in spatial transcriptomic data. The size of the dots indicates the percentage of spots expressing the marker within each zone, and the color intensity indicates the mean expression level. Data from three samples are indicated by 1, 2, or 3 following the cluster name. These results are quantified based on the spatial transcriptomic data presented in (A) and Figures S8 A and S8B. (C) Bar plots showing the percentage of CD68-positive (macrophage) and CD3D-positive spots across different zones of endometriomas. These results are linked to the dot plots in (B). Paired Student’s t test was used to compare two groups. ns, not significant. Error bars represent standard deviations. (D) Immunofluorescence staining of immune cells in different zones of endometrioma. CD10 (yellow) indicates ectopic EnS cells, CD68 (magenta) marks macrophages, CD3D (cyan) indicates T cells, TCF21 (red) marks OSCa, and FAP (orange) marks EnS cells and OSCb. The dashed lines delineate the EnS region and OSCa and OSCb regions separately. Scale bars represent 500 μm. See also Figure S8 . (E) The mean fluorescence intensity of CD68 and CD3D in the EnS cell, OSCb, and OSCa regions. Paired Student’s t test was used for statistical analysis. Spatial profiling reveals immune cell enrichment in distinct zones of endometriotic lesions (A) Marker gene expression in the spatial transcriptomic data for EnS cells, T cells, macrophages, and B cells. MMP11 represents ectopic EnS cells, while CD3D identifies T cells. CD68 and CD79A mark macrophages and B cells, respectively. Color intensity indicates gene expression levels after sctransform in Seurat. See also Figure S8 . (B) Dot plot illustrating immune cell marker expression across different clusters of spots in spatial transcriptomic data. The size of the dots indicates the percentage of spots expressing the marker within each zone, and the color intensity indicates the mean expression level. Data from three samples are indicated by 1, 2, or 3 following the cluster name. These results are quantified based on the spatial transcriptomic data presented in (A) and Figures S8 A and S8B. (C) Bar plots showing the percentage of CD68-positive (macrophage) and CD3D-positive spots across different zones of endometriomas. These results are linked to the dot plots in (B). Paired Student’s t test was used to compare two groups. ns, not significant. Error bars represent standard deviations. (D) Immunofluorescence staining of immune cells in different zones of endometrioma. CD10 (yellow) indicates ectopic EnS cells, CD68 (magenta) marks macrophages, CD3D (cyan) indicates T cells, TCF21 (red) marks OSCa, and FAP (orange) marks EnS cells and OSCb. The dashed lines delineate the EnS region and OSCa and OSCb regions separately. Scale bars represent 500 μm. See also Figure S8 . (E) The mean fluorescence intensity of CD68 and CD3D in the EnS cell, OSCb, and OSCa regions. Paired Student’s t test was used for statistical analysis. We quantified immune cell presence across lesion regions by counting the percentage of spots expressing specific markers and their average expression levels. Across the three samples, we observed CD68 (a macrophage marker) expression in an average of 67.5% OSCa spots, indicating significant macrophage enrichment within the OSCa region. CD3D expression (a T cell marker) was detected in 43.5% of spots in the ectopic EnS region ( Figure 4 B), suggesting active immune cells infiltration near EnS cells. Statistical analysis revealed significantly higher CD68 expression in the OSCa region compared to other areas ( Figure 4 C). While EnS spots exhibited a higher CD3D positivity rate, this difference was not statistically significant, probably due to variation among samples ( Figure 4 C). To validate these results at the protein level, we performed multiple rounds of immunofluorescence staining ( Figures 4 D, 4E, S8 C, and S8D). We observed a distinct zonation of CD10-postive EnS cells, FAP-positive and CD10-negative OSCb cells, and TCF21-positive OSCa cells, with CD3D-positive T cells and CD68-postive macrophages distributed throughout the tissue ( Figure 4 D). By quantifying the mean intensity of fluorescence signals in different areas, we found that CD3D intensity was highest in the EnS region, lower in OSCb, and intermediate in OSCa. CD68 intensity was lowest in the OSCb region and significantly higher in OSCa ( Figure 4 E). These results confirmed the enrichment of immune cells in the ectopic EnS regions, consistent with a role of ectopic EnS cells in initiating and sustaining inflammatory responses. It also suggests a role of OSCa in mediating macrophage interactions within the lesion. To explore functionally relevant signaling pathways underlying the spatial interactions between lesions and the environment, we applied CellChat to the spatial scRNA-seq data, taking into consideration both expression and spatial information. We found that the interaction strength between EnS cells and OSCb was the strongest ( Figure 5 A). The ectopic EnS cell-derived signaling mainly acted on OSCb and pericytes ( Figure 5 B). OSCb sent signals to EnS cells, pericytes, OSCa, and endothelial cells, while OSCa interacted with all cells except EnS cells, likely due to spatial constraints ( Figures 5 C and 5D). Figure 5 Cellular interaction analysis identified disease-specific signaling pathways associated with endometriomas (A) Heatmap illustrating the interaction strengths derived from spatial transcriptomic data of ovarian endometriomas. (B–D) Outgoing cellular interaction strengths for EnS cells (B), OSCb (C), and OSCa (D). The numbers indicate the interaction strengths between the cell types. (E) Heatmap depicting the secretory signaling pathways identified among cell types in the spatial transcriptomic data. The color represents the interaction strengths. (F) Heatmap depicting the secretory signaling pathways identified among cell types in the scRNA data of endometriomas and control ovaries. The color represents interaction strength. Green dashed lines indicate the endometriosis-specific signaling. (G) Dot plot illustrating the outgoing (x axis) and incoming signal strengths (y axis) for each cell type in endometriomas and control ovaries. Dot size represents the number of cells. (H–J) Circle plot illustrating cellular interaction strength for shared signaling pathways in endometriomas and control ovaries. Cellular interaction analysis identified disease-specific signaling pathways associated with endometriomas (A) Heatmap illustrating the interaction strengths derived from spatial transcriptomic data of ovarian endometriomas. (B–D) Outgoing cellular interaction strengths for EnS cells (B), OSCb (C), and OSCa (D). The numbers indicate the interaction strengths between the cell types. (E) Heatmap depicting the secretory signaling pathways identified among cell types in the spatial transcriptomic data. The color represents the interaction strengths. (F) Heatmap depicting the secretory signaling pathways identified among cell types in the scRNA data of endometriomas and control ovaries. The color represents interaction strength. Green dashed lines indicate the endometriosis-specific signaling. (G) Dot plot illustrating the outgoing (x axis) and incoming signal strengths (y axis) for each cell type in endometriomas and control ovaries. Dot size represents the number of cells. (H–J) Circle plot illustrating cellular interaction strength for shared signaling pathways in endometriomas and control ovaries. In total, 38 secretory signaling pathways were identified as significantly present among these cells ( Figure 5 E). Among the pathways originating from ectopic EnS cells, non-canonical WNT (ncWNT) and BMP pathways exhibited the highest interaction strength with OSCb. We also identified immune-regulating signaling, such as colony-stimulating factor (CSF), proteinase-activated receptors (PARs), ANNEXIN, interleukin-16 (IL-16), IL-10, and IL-2, activated within OSCa regions ( Figure 5 E). To further explore these spatial interactions and extend our analysis to immune cell populations that are not available in the spatial transcriptomic data, we compared the cellular interactions in our single-cell data of eight endometriomas and the control ovary dataset (GEO: GSE260685 30 ). This analysis identified 25 endometrioma-specific secretory signaling pathways ( Figure 5 F). In endometriomas, ectopic EnS cells, OSCb, and OSCa dominated the outgoing signals, while immune cells primarily acted as the signaling receivers in the endometriomas ( Figure 5 G). In contrast, in the control ovary, OSCs were the major signal senders, with endothelial cells and pericytes as primary receivers ( Figure 5 G). Immune cells in control ovaries exhibited weak cellular interactions, further validating strong immune cell activation in the endometriomas. Among the disease-related signaling pathways identified from our spatial data ( Figure 5 E), ncWNT and BMP signaling were confirmed in the single-cell data ( Figure 5 F). The immune-regulatory signaling specifically activated in endometriomas included ANNEXIN, 32 , 33 oncostatin M (OSM), 34 , 35 CSF, 36 , 37 and IL-10 38 , 39 signaling mediated by macrophages ( Figure 5 F), which primarily promote the M2 polarized phenotype and function in an anti-inflammatory capacity. Moreover, T cell-mediated PARs and IL-2 signaling ( Figure 5 F) contributed to inflammatory responses and immune cell migration, 40 while IL-2 ensures proper T cell activation and suppresses excessive immune responses. 41 Besides the disease uniquely activated signaling, we also observed immune-regulatory signaling pathways shared with a control ovary, such as macrophage migration inhitbitory factor (MIF), 42 GALECTIN, 43 and COMPLEMENT 44 signaling. These pathways were strengthened by ectopic EnS cells and modulated the environment through interactions with immune cells ( Figures 5 H–5J). Given the importance of ncWNT signaling in EnS-OSCb interaction in our analysis, we further examined the specific ligand-receptor pairs involved in the signaling. We identified WNT5A-FZD1 as having the strongest communication probability ( Figure 6 A), suggesting that WNT5A is secreted from ectopic EnS cells and acts on OSCb through its receptor FZD1. Spatial transcriptomic data also support the interactions, with the observation that WNT5A is highly expressed in ectopic EnS cells, and FZD1 is highly expressed in OSCb ( Figure 6 B). Using multiplex immunofluorescence, we confirmed high expression of WNT5A in ectopic EnS cells as well as the colocalization of WNT5A and FZD1 in the OSCb region ( Figures 6 C and S9 ). Figure 6 Ectopic EnS cells interact with OSCb through WNT5A-FZD1 signaling in endometriomas (A) Dot plot illustrating the communication probabilities for each ligand-receptor pair involved in ncWNT and BMP signaling among the cell types. The p value reflects the significance of the interactions as determined by permutation testing and is represented by the dot size, with color denoting the probability of communication. (B) Violin plots illustrating the expression levels of ncWNT signaling genes across various cell types. (C) Immunofluorescence staining reveals the interaction between ectopic EnS cells and OSCb through WNT5A-FZD1 pairings. DAPI (blue) stains cell nuclei, CD10 (white) marks ectopic EnS cells, and COL10A1 (cyan) labels OSCb cells. WNT5A (yellow), secreted by EnS cells, co-localizes with FZD1 (red) in OSCb regions. Scale bars represent 500 μm. See also Figure S9 . (D and E) Dot plots illustrating WNT5A expression in EnS cells from our data (D) and the GSE213216 dataset (E). Labels denote eutopic endometrium (EM) and ovarian endometriomas (OEMs) during both proliferative (Pro) and secretory (Sec) phases as well as peritoneal endometriosis (PEM) in corresponding phases. (F) Dot plot depicting WNT5A expression in EnS cells derived from the menstrual effluent dataset GSE203191 . (G and H) Immunohistochemical staining of WNT5A in ovarian endometriomas (G) and eutopic endometrium (H), accompanied by H&E sections. Blue arrows and the dashed line in (G) delineate the ectopic endometrium, and black arrows in (H) indicate WNT5A positivity. Scale bars represent 100 μm. Ectopic EnS cells interact with OSCb through WNT5A-FZD1 signaling in endometriomas (A) Dot plot illustrating the communication probabilities for each ligand-receptor pair involved in ncWNT and BMP signaling among the cell types. The p value reflects the significance of the interactions as determined by permutation testing and is represented by the dot size, with color denoting the probability of communication. (B) Violin plots illustrating the expression levels of ncWNT signaling genes across various cell types. (C) Immunofluorescence staining reveals the interaction between ectopic EnS cells and OSCb through WNT5A-FZD1 pairings. DAPI (blue) stains cell nuclei, CD10 (white) marks ectopic EnS cells, and COL10A1 (cyan) labels OSCb cells. WNT5A (yellow), secreted by EnS cells, co-localizes with FZD1 (red) in OSCb regions. Scale bars represent 500 μm. See also Figure S9 . (D and E) Dot plots illustrating WNT5A expression in EnS cells from our data (D) and the GSE213216 dataset (E). Labels denote eutopic endometrium (EM) and ovarian endometriomas (OEMs) during both proliferative (Pro) and secretory (Sec) phases as well as peritoneal endometriosis (PEM) in corresponding phases. (F) Dot plot depicting WNT5A expression in EnS cells derived from the menstrual effluent dataset GSE203191 . (G and H) Immunohistochemical staining of WNT5A in ovarian endometriomas (G) and eutopic endometrium (H), accompanied by H&E sections. Blue arrows and the dashed line in (G) delineate the ectopic endometrium, and black arrows in (H) indicate WNT5A positivity. Scale bars represent 100 μm. Since WNT5A act as the dominate ligand secreted by the EnS cells in endometriomas, we further evaluated its expression changes in eutopic and ectopic lesions across menstrual cycle phases from the single-cell data. In eutopic EnS cells, WNT5A expression was detected in 40.7% of cells during the proliferative phase and 25.4% of cells in the secretory phase, although at low levels. However, it was significantly overexpressed in ectopic EnS cells derived from endometriosis patients in both the proliferative and secretory phases ( Figure 6 D). The Fonseca et al. dataset confirmed significantly elevated WNT5A levels in EnS cells from endometriomas compared to the corresponding phases of eutopic controls ( Figure 6 E). For peritoneal lesions, WNT5A expression in ectopic EnS cells was also higher during the secretory phase compared to the proliferative phase ( Figure 6 E). The upregulation of WNT5A in ectopic EnS cells was further validated. In the endometriomas, although the ectopic endometrium constituted a small portion of the lesion, the ectopic EnS cells were almost entirely positive for WNT5A ( Figure 6 G). However, in the endometrium, only a small number of endometrial cells exhibited positive staining for WNT5A ( Figure 6 H). Interestingly, in an scRNA-seq dataset (GEO: GSE203191 ) of endometrial cells shed in menstrual effluent, 45 we found that patients with endometriosis or reported symptoms of endometriosis exhibited significantly elevated WNT5A expression in menstrual stromal cells ( Figure 6 F). Collectively, these findings strongly indicate that WNT5A-positive cells accumulate in ectopic lesions and induce an alteration through WNT5A-FZD1 signaling, facilitating ectopic implantation in patients with endometriosis.

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiaoyue Wang ( [email protected] ). This study did not generate new reagents. • The raw sequence data reported in this paper have been deposited at the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA006275) and are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human as of the date of publication. • The integrated gene expression matrix files for the 21 scRNA-seq datasets and the 3 spatial transcriptomic datasets of endometriomas were deposited at the OMIX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences ( https://ngdc.cncb.ac.cn/omix , accessions OMIX007675 and OMIX007676). They are available upon request. • The code for reproducing main analyses of the paper is available at https://github.com/Wangxiaoyue-lab/EMS and https://doi.org/10.5281/zenodo.14220067 . • Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. The raw sequence data reported in this paper have been deposited at the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA006275) and are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human as of the date of publication. The integrated gene expression matrix files for the 21 scRNA-seq datasets and the 3 spatial transcriptomic datasets of endometriomas were deposited at the OMIX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences ( https://ngdc.cncb.ac.cn/omix , accessions OMIX007675 and OMIX007676). They are available upon request. The code for reproducing main analyses of the paper is available at https://github.com/Wangxiaoyue-lab/EMS and https://doi.org/10.5281/zenodo.14220067 . Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Endometriosis, characterized by the displacement of endometrial tissue to various organs, involves heterogeneous cell types at lesions that are challenging to study in isolation. Here, using single cell RNA sequencing as well as spatially resolved approaches, we identify the small population of ectopic EnS cells from the predominant tissue-specific stromal cells within ovarian endometriotic cysts. Our analysis provides new insights into the aberrant functionality of displaced endometrium and its interaction with niche stroma in pathological growth. We uncovered that ectopic EnS cells retain intrinsic programming to undergo cyclic gene expression changes across menstrual phases. Despite decades of observations regarding menstruation-related symptoms in ectopic lesions, such as increased pain and recurrent hemorrhage during menstrual cycle, 5 the molecular evidence explaining these phenomena has remained obscure. 46 , 47 In this study, we found that the gene dynamics of ectopic stromal cells resemble those of eutopic EnS cells in both peritoneal and ovarian endometriotic lesions, albeit with attenuated gene expression dynamics, providing concrete molecular evidence of menstrual cycle effects on ectopic lesions. The attenuated gene expression in ectopic EnS cells indicates an aberrant response to hormones. In particular, decidualization gene markers, including DKK1 48 and IL15 , 49 , 50 were elevated in ectopic EnS cells but did not reach levels seen in eutopic EnS s, suggesting progesterone resistance. 11 , 51 This aligns with observations from other single-cell studies of endometriosis 18 and explains the limited efficacy of progestin therapies. 52 , 53 Furthermore, we found upregulation of pro-inflammatory cytokines such as CXCL8 and CXCL2 in ectopic EnS cells, which may contribute to a local chronic inflammatory environment, as reported previously. 54 , 55 , 56 , 57 This inflammation may disrupt normal progesterone signaling by inducing oxidative stress and promoting the release of pro-inflammatory mediators 58 , 59 such as tumor necrosis factor alpha and IL-6, 60 creating a microenvironment that sustains hormone resistance of ectopic EnS cells. Using spatial transcriptomics, our study illustrates the unique spatial organization of ectopic lesions. We identified two distinct OSC subpopulations within endometriotic lesions, OSCb and OSCa, each localized in different zones. OSCb cells, which closely surround ectopic EnS cells, highly express genes involved in wound healing, fibroblast migration, and connective tissue development—factors potentially contributing to fibrosis in endometriosis. 61 Conversely, OSCa cells, situated around OSCb, were associated with the inflammation process and overlapped with macrophage-rich areas. Notably, we observed a significant accumulation of inflammatory cells, particularly T cells, in the EnS region. The spatial arrangement of these cell types suggests a model for lesion establishment and growth. Once attached, ectopic EnS cells likely initiate the process by sending signals to OSCs for lesion establishment. Simultaneously, EnS cells secrete pro-inflammatory cytokines, leading to the recruitment and activation of immune cells. In response to EnS cells and the altered microenvironment, OSCs transitioned into two distinct states. OSCb cells, surrounding the EnS cells, express genes involved in wound healing, fibroblast migration, and connective tissue development, potentially contributing to fibrosis. 62 The outer layer of OSCa cells appears to mediate the interaction with macrophages. Our analysis revealed that macrophages in the lesion microenvironment are skewed toward an M2 phenotype, promoting tissue repair, angiogenesis, and immune tolerance. 63 This polarization allows the ectopic tissue to evade immune clearance while, paradoxically, supporting lesion growth and survival. 2 , 11 , 20 , 64 , 65 Our findings on macrophage distribution in endometriomas are consistent with recent work by Tan et al. 15 They found that tissue-resident macrophages, which express genes associated with immune tolerance and tissue remodeling, were enriched in endometriotic lesions. This model is consistent with the notion that tissue fibrosis is a consequence of chronic inflammation induced by EnS cells. 62 Meanwhile, the accompanying inflammation in the OSCa region may result from tissue damage induced by OSCb fibrosis, which further affects the development of oocytes. This model also correlates with the widespread observations of pain and infertility in endometriosis patients. Chronic inflammation can sensitize nerve endings, leading to pain, 66 while fibrotic changes in ovarian tissue may impair folliculogenesis and oocyte development, contributing to infertility. 67 With the resolved hierarchical structure of endometriomas, we were able to identify the essential signaling driven by the ectopic EnS cells. Our investigation of the interactions between ectopic EnS cells and OSCs revealed aberrant activation of the non-canonical WNT pathway, notably through upregulation of the WNT5A ligand in ectopic endometrium. This new finding suggests a role of WNT5A in ectopic lesion formation. WNT5A can activate several ncWNT signaling pathways by binding to different members of the Frizzled- and Ror-family receptors. Aberrant activation of WNT5A signaling has been reported to participate in regulating angiogenesis, 68 tumor metastasis, 69 , 70 and inflammation. 71 , 72 WNT5A has been reported to regulate collective cell migration during sprouting angiogenesis by promoting the formation of strong adherens junctions and coordinating collective cell polarity. 73 The impact of WNT5A on cell migration and adherens junction formation may contribute to the establishment of ectopic lesions by facilitating cell attachment to the ovarian stroma. We hypothesize that WNT5A-expressing cells may be more likely to successfully attach and grow ectopically, especially in the ovary, when refluxed during menstruation. This could potentially explain the accumulation of WNT5A-positive cells in ectopic lesions and the elevated expression in menstrual effluent from endometriosis patients. Importantly, the observed upregulation of WNT5A in endometriosis patients and its enriched presence at the junctions of engrafted human endometrium underscore its potential as a therapeutic target. However, WNT5A also plays significant roles in other physiological and pathological processes, 74 , 75 , 76 necessitating careful consideration of potential off-target effects in any therapeutic approach targeting this pathway. While our findings suggest new avenues for non-invasive therapies, it is important to note that surgical intervention currently remains a crucial approach in the management of endometriosis, particularly for ovarian endometriomas. Surgery plays a vital role in both pain management and fertility preservation. However, small endometriotic lesions are often difficult to excise completely, and there is a high risk of recurrence post surgery. Therefore, the development of targeted, non-hormonal treatments based on molecular insights, such as the WNT5A pathway, could potentially complement surgical approaches and improve long-term outcomes for endometriosis patients. Future research should focus on developing targeted therapies that can modulate WNT5A signaling specifically in endometriotic lesions while minimizing systemic effects. One notably limitation of our study is the scarcity of epithelial cells identified from ovarian endometriomas, which restricted our ability to explore the cellular interactions between ectopic EnS cells and ectopic endometrial epithelial cells. The endometrial epithelium, comprising gland, luminal, and ciliated epithelium, undergoes differentiation along the menstrual cycle while maintaining the capacity for self-renewal. 24 Previous studies have identified distinct epithelial progenitor cell populations in the endometrium, 77 and evidence suggests that epithelial cells exhibit resistance to apoptosis in ovarian endometriosis. 17 Moreover, cancer driver gene mutations have been detected in ectopic endometrial epithelial cells but not in stromal cells, 16 , 78 highlighting their roles in the onset of disease. The scarcity of epithelial cells in our ectopic lesion samples is consistent with previous observations and may be partly attributed to disruption of the surface epithelium during surgical removal of endometriomas. Given the importance of endometrial epithelial cells in endometriosis pathogenesis, further investigations are needed to comprehensively explore the combined effect of both the epithelial and stromal cell populations in the development of endometrioma. Another consideration when interpreting our results is the potential heterogeneity across different types of endometriomas. Endometriotic cysts can be categorized as cortical invagination, surface inclusion related, physiological cyst related, or unclassified. Our spatial transcriptomic analysis was conducted on samples that were not classifiable into specific endometrioma subtypes due to lack of definitive evidence at histological examination. This challenge of classification is common in the field. A previous study found that 66% of endometriomas were unclassifiable due to insufficient evidence. 79 It is not known whether the spatial pathological features of other types of endometriotic cysts follow the same hierarchical structure. In particular, surface inclusion cyst-related endometriotic cysts, which arise from further budding of an existing cyst, are considered highly invasive and might reflect a different pathogenesis. 79 Future studies with larger sample sizes could further elucidate subtype-specific features and provide insights into the origin and development of endometriomas.

Endometriosis, present in 10% of reproductive-age women, is characterized by growth of functional endometrial tissue outside the uterine cavity. 1 , 2 This condition can be classified into three major forms depending on lesion location: peritoneal, ovarian, and deep infiltrating. Among them, ovarian endometriomas occur in 17%–44% of women with endometriosis 3 and are significantly associated with pelvic deep infiltrating endometriosis, ovarian adhesions, and pouch of Douglas obliteration. 4 A notable characteristic of ectopic lesions is their recurrent hemorrhage during menstruation, 5 indicating preservation of endometrial functionality and hormonal responsiveness. Despite decades of research, the pathophysiology of the disease remains only partially understood. Current treatments are limited to surgical resection and hormonal suppressive therapy with many adverse effects that rarely provide long-term relief. 6 , 7 Elucidating the mechanisms that enable the ectopic growth of endometrium is essential for developing novel targeted therapies. The retrograde menstruation theory, first proposed by Sampson in 1927, 8 hypothesizes that endometrial fragments reflux through the fallopian tubes and implant in the peritoneal cavity, potentially resulting in significant clinical symptoms. 9 , 10 , 11 , 12 However, despite 80%–90% of women of childbearing age exhibiting menstrual reflux, only 10%–15% develop endometriosis. 13 This discordance suggests that additional factors, possibly involving a complex interplay of cellular and molecular signals, are required for promoting the attachment and growth of these ectopic endometrial cells within the lesions. Single-cell transcriptome sequencing has been used recently to profile cell heterogeneity within and across endometriosis sites 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 as well as the dynamic endometrial remodeling throughout the menstrual cycle. 23 , 24 , 25 , 26 , 27 These studies have revealed distinct transcriptomic signatures distinguishing epithelial and stromal cells from ectopic lesions compared to eutopic endometrium. However, a critical knowledge gap remains regarding the responses of ectopic endometrial cells to hormones during the menstrual cycle. It is also unclear what interactions occur between these ectopic cells and the surrounding niche cells that enable sustained growth of the ectopic lesion. Here, we performed single-cell RNA sequencing (scRNA-seq) on 13 eutopic endometria and 8 ovarian endometriomas and integrated these data with available datasets for comprehensive profiling of ectopic endometriotic lesions. We demonstrated that ectopic endometrial cells retained responsiveness to menstrual-cycle hormones. Our analysis revealed that stromal cells in the ectopic niche are specialized upon lesion formation, forming distinct pro-fibrotic and pro-inflammatory zones, as observed in spatial transcriptomic profiling. Further examination uncovered a strong interaction mediated by non-canonical WNT signaling between ectopic endometrial and ovarian stromal cells. This signaling pathway could be critical in driving ectopic growth of endometrial cells and offers a potential target for more effective endometriosis treatments.

The authors declare no competing interests.

REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples Human endometrium and endometrioma samples Peking Union Medical College Hospital Ethics Committee of Perking Union Medical College Hospital (JS-1532) Antibodies mouse monoclonal anti-WNT5A Santa Cruz Biotechnology Cat#sc-365370; RRID:AB_10846090 Rabbit monoclonal anti-CD10 ABclonal Cat#A22179; Rabbit Recombinant anti-E-Cadherin Cell Signaling Technology Cat#3195s; Mouse monoclonal anti-POD-1 (TCF21) Santa Cruz Biotechnology Cat#sc-377225 Rabbit monoclonal anti-FAP Abcam Cat#ab207178; RRID:AB_2864720 Rabbit Polyclonal Collagen X antibody Abcam Cat#ab58632; RRID:AB_879742 Rabbit anti-Homo sapiens (Human) FZD1 Polyclonal antibody CUSABIO Cat# CSB-PA891570LA01HU Mouse monoclonal anti-CD68 antibody Abcam Cat# ab955; RRID:AB_307338 Rabbit monoclonal anti-CD3 epsilon antibody Abcam Cat# ab16669; RRID:AB_443425 Chemicals, peptides, and recombinant proteins Collagenase type II Sigma Cat# C6885-1G Collagenase type IV Sigma Cat#C5138-1G Red blood cell lysis buffer Invitrogen Cat#00-4333-57 Critical commercial assays Absin 4-Color IHC Kit Absin Cat#abs50012 Chromium Single Cell 3′ Gel Bead and Library Kit 10x Genomics Cat#120237 QuickBlock™ Blocking Buffer Beyotime Cat#P0260 Mounting medium, anti-fading (with DAPI) Solarbio Cat#S2110 Deposited data raw single cell RNA sequencing data of 21 samples This paper GSA: HRA006275 at  https://ngdc.cncb.ac.cn/gsa-human Integrated gene expression matrix of 21 scRNA datasets This paper OMIX: OMIX007675 at https://ngdc.cncb.ac.cn/omix Gene expression matrix and corresponding image data for 3 endometriomas This paper OMIX: OMIX007676 at https://ngdc.cncb.ac.cn/omix scRNA-seq data of endometriosis Fonseca et al. 16 GEO: GSE213216 ; Processed data downloaded from https://cedars.app.box.com/s/1ks3eyzlpnjbrseefw3j4k7nx6p2ut02 scRNA-seq data of menstrual effluent Shih et al. 45 GEO: GSE203191 scRNA-seq data of control endometrium Wang et al. 23 GEO: GSE111976 scRNA-seq data of 11 endometrium samples from 5 donors Garcia-Alonso et al. 24 ArraryExpress: E-MTAB-10287 in https://www.ebi.ac.uk/biostudies/arrayexpress scRNA-seq data of 31 tissue samples from 5 ovaries Fan et al. 29 GEO: GSE118127 scRNA-seq data of 3 adult human ovaries Jones et al. 30 GEO: GSE260685 Software and algorithms Codes and scripts This paper https://github.com/Wangxiaoyue-lab/EMS ; https://doi.org/10.5281/zenodo.14220067 Cellranger v6.0.0 10x genomics https://www.10xgenomics.com R v4.3.1 The R Project for Statistical Computing https://www.r-project.org/ Seurat V5 Butler et al. 80 and Stuart et al. 81 https://satijalab.org/seurat/ ClusterProfiler v3.12.0 Yu et al. 82 https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html scanpy Wolf et at. 83 https://scanpy.readthedocs.io/en/stable/ Monocle2 v.2.8.0 Trapnell et al. 84 FV1000 Cellchat v2.1.0 Jin et al. 85 https://github.com/jinworks/CellChat Space Ranger 10x genomics https://www.10xgenomics.com BBKNN Polanski et al. 86 https://github.com/Teichlab/bbknn Other Illumina NovaSeq 6000 Illumina N/A The Pannoramic Digital Slide Scanners 3DHISTECH Company Flash 250 Chromium Controller instrument 10x Genomics N/A 40-μm nylon cell strainer Corning Cat#352340 This study was approved by the Ethics Committee of Peking Union Medical College Hospital (JS-1532) and all the patients had provided their written informed consent. All specimens were obtained from the patients who experienced laparoscopy. These participants had regular menstrual cycles ranged from 27 to 33 days, and none of them received hormonal therapy three months prior to the surgery. Endometrial biopsies were collected using uterine biopsy curette, and ovarian endometriotic cysts were collected from patients by laparoscopy. In addition to the chocolate-like effusion observed during the operation, ovarian endometriosis was further confirmed by postoperative pathology. For the 8 endometriomas used in single-cell RNA sequencing, 2 were cortical invagination endometriotic cysts, 1 was surface inclusion cyst-related endometriotic cysts, the other 5 were unclassified endometriotic cysts ( Figure S1 ). For the other 3 paraffin-embedded ovarian endometrioma samples used in spatial transcriptomics, all of them were unclassified endometriotic cysts. Endometrial histology was performed to confirm the cycle stage for 13 patients from whom eutopic endometrial samples were collected. Histological evaluation was conducted by two experienced pathologist blinded to the last menstrual period (LMP) data. For 4 patients in the secretory phase from whom only endometrioma samples were collected, serum progesterone levels were tested, with levels above 5.16 ng/mL considered indicative of the secretory phase. Fresh tissues were collected after surgery and transported in ice-cold phosphate-buffered saline (PBS) (Corning, 21-040-CV). The tissues were washed 3 times with PBS to remove the blood and mucus followed by mincing into pieces as small as possible with surgical scissors. Then, small pieces of tissues were digested in DMEM medium with 1 mg/mL collagenase type II (Sigma, C6885-1G) and 1 mg/mL collagenase type IV (Sigma, C5138-1G). The tissue was enzymatically digested at 37°C with a shaking speed of 150 r.p.m. The digestion time was 45 min for endometrium tissue and 90 min for endometrioma. Cell suspensions were filtered through a 40-μm nylon cell strainer (Corning, 352340) and re-suspended in ice-cold PBS. The suspension was centrifuged at 1000 r.p.m for 3 min to collect the cell pallet. Dissociated cells were treated by red blood cell lysis buffer (Invitrogen, 00-4333-57) to remove red blood cells and re-suspended in ice-cold PBS. All the samples were processed within 1 h after dissection. For endometrium, single-cell suspensions with cell viability over 80%, cell mass ratio less than 10% and cell density over 1x 10 6 /mL were qualified for single cell library construction; For the endometrioma, single-cell suspensions with cell viability greater than 60%, cell mass ratio less than 10% and cell density over 1x 10 6 /mL were subjected for single cell library construction. Chromium single cell 3′ reagent kits (V2 and V3.1, 10x genomics) were used to construct scRNA-seq libraries. Briefly, about 6000 live cells were loaded on a Chromium Controller instrument (10x Genomics) for each sample to generate single-cell gel bead-in-emulsions (GEMs). Subsequent steps include reverse transcription, cDNA amplification, enzymatic fragmentation, end-repair, A-tailing, adaptor ligation and sample index PCR were carried out according to the Single Cell 3′ Reagent Kit User Guide of 10x genomics. Sequencing libraries were loaded on an Illumina NovaSeq 6000 platform to obtain a sequencing depth of approximately 100,000 reads per cell. The raw reads were aligned and quantified against the GRCh38 human reference genome through cell ranger software suite (version 6.0.0, 10x Genomics) to generate the gene-cell barcode expression matrix. The expression matrix for all samples were then aggregated using “cellranger aggr” with default parameters for downstream analysis. Scanpy 83 was used for cell clustering analysis. For our own data, the ribosome genes were removed from the matrix data, and genes identified in less than 10 cells were abandoned. Cells with less than 600 or more than 6000 detected genes, or the mitochondrial genes expression exceeding 20% were discarded. The genes expression for each cell were normalized and log transformed. The highly variable genes among the cells were identified and used for dimension reduction. The effects of total counts and the mitochondrial percent were regressed out for each cell, and the genes were scaled to unit variance. The principal component analysis (PCA) was applied to reduce the dimension, the BBKNN 86 method was used to remove the batch effect within different samples. Uniform manifold approximation and projection (UMAP) was applied for cell embedding, and Leiden algorithm were used for cell clustering. Cluster marker genes were identified by Wilcoxon rank-sum tests, and the cell type was identified based on the top highly expressed genes for each cluster. For the public endometriosis ( GSE213216 ) data, 16 the analyzed data were downloaded and transformed into an h5ad file format, then loaded into Scanpy for reanalysis. The ribosome genes were removed, and dataset were segmented according to tissue source. No further filtering of genes or cells were performed. Then, the same pipeline was applied to re-cluster the cells for each dataset. For the integration analysis of our endometriomas data with control ovary single cell transcriptomics data ( GSE118127 and GSE260685 ), each dataset was analyzed separately using the same pipeline by Scanpy, and get the cell annotations based on the corresponding cell type markers. The different datasets were combined and re-dimensionally reduced for cluster analysis, the batch effect was removed by BBKNN method. For the integration analysis of our endometrium data with endometrium donors single cell transcriptomics data (E-MTAB-10287). The public data was downloaded with well annotated cell types. And the cell type markers were calculated for each subgroup. Using the detailed subgroup cell markers, we carried out subgroup data analysis with our endometrium data. Then the two datasets were combined and dimension reduced to compare the cell subgroup clusters. All eutopic endometrial stromal cells from our data were extracted and loaded into the Seurat package, followed by normalization and data scaling. The cells were grouped by menstrual phases, and menstrual feature genes were calculated using the ‘FindAllMarkers’ function, with ‘min.pct = 0.5’ parameters. The overexpressed genes were selected based on the following criteria: (1) Avg_log 2 (Fold Change) > 1; (2) Adjusted p -value <0.05. 89 overexpressed genes in proliferative phase were identified as proliferative feature genes, and 112 overexpressed genes in secretory phase were used as secretory feature genes ( Table S1 ). Gene expression scores were calculated by the ‘AddModuleScore’ function in Seurat. The monocle2 package (v2.8.0) was used for the pesudotime analysis. 84 The 201 menstrual feature genes were selected to order the cells, and the DDRTree method was applied for dimension reduction. We set “State 4” as the root state using the ‘orderCells’ function in Monocle 2 to orient the pseudotime trajectory. Gene expression changes along the pseudotime was calculated using the ‘differentialGeneTest’ function. For the analysis of menstrual feature gene expression scores using public data ( GSE111976 and E-MTAB-10287), endometrial stromal cells were subsetted, and the PES/SES scores were calculated using the “AddModuleScore” function in Seurat or the "sc.tl.score_genes" function in Scanpy. Ovarian stromal cells from ovarian endometriomas and unaffected ovaries in the public dataset ( GSE213216 ) were selected and reanalyzed using the Seurat package. The subgroups of stromal cells were labeled as OSCa, OSCb, and unaffected OSC, based on their original identities annotated by Scanpy. Overexpressed genes were identified using the 'FindAllMarker' function with default parameters. Differentially expressed genes were ranked by adjusted p -value, and the top 100 genes for each group were used for gene ontology (GO) analysis with the clusterProfiler (v3.12.0) package. 82 Significantly enriched biological process GO terms were selected based on the FDR adjusted p -value. Three formalin fixation and paraffin embedding samples were used for spatial transcriptomics. In order to detect the layered structure of the ovarian endometrioma cyst, longitudinal sections of the cyst were prepared when possible. Spatial transcriptomic sequencing was carried out using the 10x Visium CytAssist platform. Space Ranger was used to align the sequencing reads to the human reference genome. Subsequently, the Seurat package was used for clustering analysis and visualization of target gene expression. Spots meeting the criteria “nFeature_Spatial >1000 & nCount_Spatial >1000 & percent.mt < 15 & nCount_Spatial <100,000” were retained. To integrate multiple single-cell spatial transcriptomics datasets, we utilized the Seurat v5 package and its SCTransform-based integration workflow. First, highly variable features across the datasets were selected using the ‘SelectIntegrationFeatures’ function, where the top 2000 highly variable genes were chosen for subsequent analysis. Next, we performed SCTransform normalization to regress out technical noise. Using the selected features, we prepared the datasets for integration with the ‘PrepSCTIntegration’ function, which ensures consistent scaling across datasets. To identify shared cell states across the datasets, the ‘FindIntegrationAnchors’ function was applied, with the SCT normalization method and the selected variable features as input. The datasets were integrated using the ‘IntegrateData’ function with SCT-based normalization to correct for batch effects and generate a combined, batch-corrected dataset. Following integration, the data were scaled and dimensionality reduced with the first 15 principal components for downstream analyses. Uniform Manifold Approximation and Projection (UMAP) was then applied to visualize the data in two dimensions using the ‘RunUMAP’ function. The major cell type for each spot was annotated by the marker genes identified from our single cell data. To analyze immune cell enrichment in the spatial transcriptomics data, the positive rate of immune cell marker gene expression was calculated for all spots in each zonation. Paired Student’s t test was used to analyze immune cell enrichment in EnS and OSCa regions. CellChat (version 2.1.0) 85 package was used to analyze the cellular communications. All cell groups and normalized gene expression data were loaded into CellChat to create the analysis object. The secreted signaling database was selected and used. The standard workflow was carried out to predict the major signaling inputs and outputs for each cell type. The significant signaling interactions were identified using “computeCommunProbPathway” function with the threshold of the p value = 0.05. Significant interactions between two cell groups were identified using a permutation test by randomly permuting the group labels of cells, and then recalculating the communication probability of a given ligand-receptor pair between two cell groups. Network centrality analysis was then performed to understand the importance of different cell types in the signaling network. For the spatial transcriptomic data, the spatial image information was also imported to calculate the likelihood of intercellular communication by taking spatial constraints into account. The ovarian cyst and endometrial tissue sections were baked at 60°C for 3h, dewaxed with xylene, rehydrated with gradient alcohol and fixed with 10% neutral buffered formalin. The mIHC experiment was conducted using the Absin Company kit (Absin, #abs50012, China) according to the provided instruction manual. The mIHC Antigen retrieval was performed by boiling tissue sections in epitope retrieval buffer in a microwave oven for 5 min, followed by immediate cooling for 20min. After cooling to room temperature, tissue sections were washed three times in 1x Tris-Buffered Saline with Tween 20 (CST, #9997). Nonspecific binding was blocked in QuickBlock Blocking Buffer (TBST) (Beyotime, #P0260, China) at room temperature for 15min. The tissue sections were then incubated for 2 h with the primary antibody at room temperature. After washing three times with TBST, the slides were incubated in HRP secondary antibody at room temperature for 10 min and washed with TBST again. To amplify the fluorescent signal, 1X amplifying working solution was dropped onto the slides to fully cover the sample area for 10 min at room temperature. Subsequently, the working liquid was discarded, and the slides were washed with TBST. Following the single staining, antigen repair was performed once again. After blocking, subsequent staining was repeated. Lastly, nuclear staining and sealing were then conducted, with a drop of mounting medium, anti-fading (with DAPI) (Solarbio, #S2110, China) applied to the section and the coverslip subsequently sealed. Following staining, the slices were scanned using The Pannoramic Digital Slide Scanner Family (Flashh 250, 3DHISTECH Company). Endometrioma and endometrial tissue sections were baked at 60°C for 3 h, dewaxed with xylene and rehydrated with gradient alcohol. Antigen retrieval was performed by boiling tissue sections in epitope retrieval buffer (pH = 9) in a microwave oven for 5 min, followed by immediate cooling for 20min. After returning to room temperature, tissue sections were washed three times in PBS. Nonspecific binding was blocked in goat serum working fluid at room temperature for 1-2h. The tissue sections were then incubated overnight with the primary antibody at 4°C. The next day, after washing three times in PBS, the slide was incubated in secondary antibody at room temperature for 30 min. The quantitative and statistical analyses are described in the relevant sections of the method details or in the figure legends.

We thank Prof. Chengyu Jiang and Ran Zhang from the Institute of Basic Medical Science, Peking Union Medical College, for kind suggestions and technical expertise regarding tissue dissociation. We appreciate the support from the Clinical Biobank , 10.13039/501100008235 Peking Union Medical College Hospital , 10.13039/501100005150 Chinese Academy of Medical Sciences . This work was supported by the 10.13039/501100012166 National Key R&D Program of China ( 2022YFC2704000 and 2017YFC1001200 to J. Leng ), the 10.13039/501100001809 National Natural Science Foundation of China ( 32122023 to X.W., 82201822 to S.L., and 82071628 to J. Leng), the National High Level Hospital Clinical Research Funding ( 2022-PUMCH-A-201 to S.L., 2022-PUMCH-B-085 to J. Leng, and 2023-PUMCH-E-008 to X.W.), and the 10.13039/501100005150 Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences ( CIFMS 2017-I2M-4-003 ).