High-resolution spatial transcriptomics reveals fibroblast and neuroimmune microenvironments in endometriosis lesions

Spatial niches in ovarian and peritoneal lesions We performed whole transcriptome spatial analysis using Visium v2 (10x Genomics) on three tissue types: endometrium biopsy (EMB), peritoneal superficial lesion, and ovarian lesion. The Visium assay is a low-resolution spatial assay that employs 55 µm spots to capture protein-coding genes. These spots are spread out across the tissue with a 100 µm distance between the center of each spot, leaving sub- stantial tissue space uncaptured (Fig.1a). To identify epithelial glands– one of the key components of both EMB and lesions – we performed nuclei segmentation on H&E-counterstained tissues using QuPath and trained a learning model for gland detection. This model was inte- grated with spatial expression data for the epithelial markerEPCAM to define the spatial Epi cluster (Fig. 1a ) .I na d d i t i o nt ot h ek n o w np r e - sence of CD10+ (encoded by MME) fibroblasts, our previous endo- metriosis single-cell study led us to define differentfibroblast subtypes in lesions. We speci fically observed an increased osteoglycin ( OGN) expressing fibroblast subpopulation in endometriosis tissues (EMB and lesions) 10. Therefore, we investigated the spatial distribution of MME and OGN expression. Unsupervised clustering of the low- resolution spatial gene signatures failed to resolve the cellular com- plexity surrounding the glands, likely due to the heterogenous cell type composition within each spot, where we captured ~ 25 cells per spot (Fig. 1b). We therefore manually annotated two stromal regions based on spatial proximity to epithelial glands and gene expression: the ‘ Close Stroma’ , an immediately adjacent region, marked by ele- vated MME expression, and the ‘ Far Stroma’ ; and a more distal region enriched for OGN expression (Fig. 1c, d). This spatial strati fication revealed a clear, layered pattern of MME and OGN expression in both ovarian and peritoneal lesions (Fig.1c, d) but not in EMB tissue (Sup- plementary Fig. 1a– f). These findings suggest the presence of lesion- specific stromal organization around epithelial structures. To evaluate the molecular differences between stromal niches, we performed pseudo-bulking of the measured gene expression in Close and Far Stroma. We then conducted differential gene expression analysis (DEGs) within the peritoneal (Fig. 1e) and ovarian (Supple- mentary Fig. 1c) lesions. In peritoneal lesion, the Far Stroma niche was enriched for genes associated with WNT signaling (such asSFRP4)a n d Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press vascular processes signaling (CNN1 and THBS1) while the Close Stroma displays elevated expression of immune cell, particularly NK cell, related genes, includingGZMA, GZMB, GNLY,a n dNKG7 (Fig. 1e, f). We observed enrichment of nervous system-related pathways in Far Stroma of the peritoneal lesion (Fig. 1f). Though these findings were inconsistent in the ovarian lesion (Supplementary Fig. 1d), it became clear on further visual inspection of neuronal related genes that the platform lacked the sensitivity to reliably detect fine neuronal pro- cesses such as small axons, as reported by Liu et al. 12.T oo v e r c o m et h i s

and enhance spatial precision, we transitioned to high- resolution spatial transcriptomics using the Visium HD (10x Genomics) platform. a bc d EPCAM MME OGN NES EPCAM MME OGN Epi Close Far Gene Expression Level / Spot Epi Close Far Epi Close Far Peritoneal -1 0 1 2 log/g3676 FC -log/g3675/g3674 p.adj e f Fig. 1 | Low-resolution spatial transcriptomics of peritoneal endometriosis. a Schematic of the low-resolution spatial transcriptomics (Visium V2) workflow performed on peritoneal superficial lesion (n = 1), ovarian lesion (n = 1) and endo- metrium (n = 1). The H&E was used for nuclei segmentation and gland identifica- tion, the spatial transcriptomics data was used to identify EPCAM +, MME + , and OGN + spots. The working model is as pictured with Close Stroma enriched in MME + fibroblasts directly surrounding the epithelial glands and Far Stroma enri- ched in OGN + fibroblasts outside of the Close Stroma cluster. Created in BioR- ender. Haney, C.63. https://BioRender.com/e64ttk0is licensed under CC BY 4.0. b Peritoneal lesion unsupervised clustering (sample P5).c Zoom of identified gland area showing spots with detectableEPCAM, MME,a n dOGN expression.d Violin plot of EPCAM, MME,a n dOGN expression levels in peritoneal Epithelial (Epi), Close Stroma (Close) and Far Stroma (Far) clusters. e Volcano plot of the differentially expressed genes (DEGs) between the Close Stroma (blue, upregulated above zero) and Far Stroma (purple, downregulated below zero). The top 15 genes for each group are highlighted. DEGs were calculated using DEseq2 which uses a Wald Test. The horizontal gray dotted line represents the significance threshold of Benjamini- Hochberg-adjustedp-value (FDR) < 0.05, and the two vertical lines represent the Log2 fold change (FC) 0.5.MME (Log 2 FC = 0.611, FDR = 0.604) andOGN (Log2 FC = − 2.064, FDR = 0.048) are labeled in red. f Bar plot of significant GO: Biological Process pathways summarized by representative pathways in the Close Stroma [blue, positive normalized enrichment score (NES)] compared to the Far Stroma (purple, negative NES). A complete list of significant pathways and what the summarized processes represent can be found in Supplementary Data 5, 6. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press High-resolution spatial mapping of lesion microenvironment We performed high-resolution spatial transcriptomics using Visium HD on eight endometriosis lesions: four superficial peritoneal and four ovarian lesions (Fig.2a and Supplementary Fig. 2). All lesions used were selected based on the presence of epithelial glands surrounded by fibroblasts. We focused on these lesion types as they represent the two most prevalent clinical subtypes – peritoneal lesions are observed across all stages of disease, while ovarian lesions affect ~50% of patients undergoing treatment for fertility 25.V i s i u mH De n a b l e sh i g h - resolution spatial mapping by capturing transcriptomic data across a ab c d f Ovarian Lesion Peritoneal Lesion Ovarian Lesions NES Peritoneal Lesions NES EPCAM MME OGNH&E CD10 OGN Pan-CK/EpCAM 100 μm Spatial Cluster Epi Close Far P13 P12 P11 O9 Z-score Sample Spatial Cluster Sample g h i Spatial1 Spatial 2 Spatial1 Spatial 2 Spatial1 Spatial 2 Spatial1 Spatial 2 EPCAM MME OGN Epi Close Far Z-score Sample Spatial Cluster j e Ovarian LesionPeritoneal Lesion Ovarian LesionPeritoneal Lesion ASPN CPXM2 KCNMB1 JPH2 FILIP1L SYNM AOC3 SGCA MME MMP11 DIO2 COL7A1 WNT2 HOXD10 CD276 NTN1 CALB2 UPK3B MST1 HPSE PANX2 EPCAM CDH1 100 μm Distance from PanCK boundary (μm) Z-score across 8μm bins Fig. 2 | High-resolution spatial transcriptomics of ovarian and peritoneal endometriosis. a Schematic of high-resolution spatial transcriptomics and antibody-based platform validating spatial clusters on three independent cohorts of samples. Visium HD was performed on peritoneal superficial (n = 4) and ovarian (n = 4) lesions. Xenium spatial transcriptomic on peritoneal superficial (n =3 )a n d ovarian (n = 1) lesions. Imaging Mass Cyometry (IMC) on peritoneal superficial (n =4 )a n do v a r i a n(n = 4) lesions. Created in BioRender. Courtois, E. (2026)https:// BioRender.com/2ufvq8cis licensed under CC BY 4.0. b Zoom of identified gland area showing bins with > 0 expression ofEPCAM, MME,a n dOGN (n = 4 ovarian and 4 peritoneal lesions). c (left) Representative image of ovarian lesion (sample O1a) and peritoneal lesion (sample P1) spatial clusters labeled as Epi, Close Stroma, Far Stroma or Other. The other six tissues can be found in Supplementary Fig. 2. Scale bar 1 mm. d Heatmap of z-scoredEPCAM, MME and OGN expression per VisiumHD sample defining the Epi, Close and Far spatial clusters. e, f Bar plot of selected significantly enriched GO: Biological Process pathways in the Close Stroma [blue, positive normalized enrichment score (NES)] compared to the Far Stroma (purple, negative NES) for all ovarian lesions (e) and all peritoneal lesions (f). A full list of pathways is in Supplementary Data 9 and 10. g Representative images of Xenium spatial transcriptomics data from (top) ovarian lesion (sample O9) and (bottom) peritoneal lesion (sample P11), colored the same as in (c). Scale bars 1 mm. h Heatmap of Xenium spatial transcriptomic samples showing z-score expression of genes used to define spatial clusters based on VisiumHD differential gene ana- lysis (Supplementary Fig. 3 and Supplementary Data 11 and 12).OGN is not present in the predesigned human 5000 gene panel. i IMC protein validation of CD10 (encoded byMME) surrounding epithelial gland (Pan-cytokeratin/EpCAM collapsed into one color) as Close Stroma and OGN farther away as Far Stroma. Scale bar is 100 µm. Ovarian lesion (sample O7) and peritoneal lesion (sample P8). Additional images in Supplementary Fig. 3c (n = 4 ovarian and 4 peritoneal lesions). j Kernel- smoothed z-scored profiles of CD10 and OGN antibody expression plotted against the distance from the Pan-CK expressing boundary (epithelial gland), highlighting relative enrichment patterns. Source data are provided as a Source Data file. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press continuous tissue area using 2 × 2 µm bins, leaving no tissue unsam- pled. We aggregated 4 × 4 squares of these 2 × 2µmb i n si n t o8×8µm bins. This approach yielded a dataset comprising 2.62 million 8 × 8µm bins, with an average of 327,000 bins per tissue section, providing an unprecedented subcellular spatial map of lesions (Supplemen- tary Data 2). We identified epithelial glands using glandular feature detection on H&E images combined with spatial EPCAM expression across the bins. Unlike the low-resolution spatial analysis, unsupervised cluster- ing of the transcriptomic pro files of 8 × 8 µm bins outlined distinct spatial niches. For each lesion, we labeled the unsupervised clusters b a s e do nm R N Ae x p r e s s i o no fMME – for Close Stroma, OGN – for Far Stroma and EPCAM – for Epi (Fig. 2b– d). The Far Stroma included spatial clusters enriched in fibroblast-associated markers ( PDGFRA, LUM, SFRP4) in addition to OGN. Clusters labeled ‘ Other’ lacked fibro- blast markers or represented low-quality bins. This allowed for com- parison across multiple samples and lesion types. The Close and Far stroma gene expression pro files were subse- quently pseudo-bulked for DEG and pathway analysis. Pathway analysis revealed distinct, cell type speci fic biological processes associated with each spatial cluster, consistent across lesion types. Pathways related to lymphatic vessels in peritoneal lesions and B cells in ovarian lesions were significantly enriched in the Far Stroma (Fig. 2e, f). Strik- ingly, sensory perception signaling pathways were enriched in the Close Stroma compartment across both ovarian and peritoneal lesions (Fig. 2e, f). Although classically associated with olfactory tissues, these pathways encompass a family of ectopically expressing G-protein- coupled receptors and chemosensory receptors that are known to regulate inflammatory and neuroimmune signaling in peripheral tis- sues. Their enrichment in endometriosis lesions likely reflects altered chemosensory and neuroin flammatory activity within the Close Stroma microenvironment. This is in sharp contrast withfindings from low-resolution spatial results (Fig.1f), underscoring the need for high- resolution spatial profiling to resolve neuronal signatures accurately. To further validate thesefindings and increase the sample size, we performed single-cell spatial transcriptomics on an additional subset of peritoneal lesions ( n = 3) and ovarian lesions ( n =1 ) u s i n g t h e Xenium (10x Genomics) platform and predesigned human 5000 gene panel. To define the spatial clusters, we used unsupervised clustering of the transcripts (Points2Regions26), then evaluated the top differen- tially expressed genes from the VisiumHD data (Supplementary Fig. 3). For the Close Stroma, this includedMME, MMP11, DIO2, COL7A1, WNT2, HOXD10, CD276, NTN1, CALB2, UPK3B, MST1 ,a n d HPSE.F o rt h eF a r Stroma, this included A S P N ,C P X M 2 ,K C N M B 1 ,J P H 2 ,F I L I P 1 L ,S Y N M , AOC3, SGCA,a n d CCN5 – note that OGN is not included in the prede- signed Xenium 5k probe set. The results show well defined Close and Far clustering across all tissues, further validating our initial observa- tions using an independent high-de finition spatial transcriptomic platform and biological replicates (Fig. 2g, h). Consistent with these findings, quantitative protein expression relative to epithelial glands was con firmed by imaging mass cytometry (IMC) in independent peritoneal lesions (n = 4) and ovarian lesions ( n =4 ) ( F i g .2ia n dS u p - plementary Fig. 3c). The pixel intensity analysis averaged across all tissues shows that CD10 expression is highest in proximity to epithelial glands, whereas OGN increases as CD10 signal declines when moving further away from epithelial glands (Fig.2j). Together, across two spatial transcriptomic platforms and one protein expression platform, all ovarian and peritoneal lesions formed spatial clusters surrounding epithelial glands, with MME, MMP11 and WNT2 highly expressed in the Close Stroma andOGN, ASPN and FILIP1L highly expressed in the Far Stroma. Stromal compartments are conserved across lesions To assess the full cellular complexity of the lesion microenvironment, we predicted the most likely cell type within each bin using marker genes from in-house and published single-cell datasets10– 12 (Fig. 3a, b,

section and Supplementary Data 3 and 4). Shifts in predicted cell type proportions between Epi, Close and Far Stromal spatial clusters showed similarities and differences between ovarian and peritoneal lesions (Supplementary Fig. 4b, linear mixed effect model). As a good control, epithelial cells were sig- nificantly enriched in Epi spatial clusters. Across both lesion types, vascular cells (which comprises endothelial and perivascular markers) are signi ficantly enriched in the Far Stroma, and there is a trend t o w a r d si n c r e a s e dm o n o c y t e sa n dm a c r o p h a g e si nt h eC l o s eS t r o m a . Differences across lesion types were manifest in the signi ficant enrichment of neurons within the Close Stroma of peritoneal lesions, while neurons were evenly distributed across Close and Far Stroma in ovarian lesions. NK cells are significantly enriched in the Close Stroma cluster of ovarian lesions (Fig. 3a, b and Supplementary Fig. 4a, b). B cells have been associated with endometriosis and autoimmune disorders involving B cells have also been linked to the disease 27.Bc e l l counts are increased in the peritoneal fluid of patients 28– 31,a n dw e previously observed tertiary-like structure involving B cells in lesions10. However, their role remains unclear. Interestingly, we noted that B cells form aggregates throughout the lesion, and more abundantly in the Far Stroma in both lesion subtypes, across all tissues analyzed (Fig. 3a– d and Supplementary Fig. 5a– f). To evaluate globally the similarities across lesion types, we inte- grated all tissues into one dataset and performed unsupervised hier- archal clustering on the pairwise similarities between all bins. The

indicate similar transcriptomic signatures across lesion types, mostly aligning with the spatial clusters initially defined independently within each sample by the stromal compartments (Fig. 3e, black boxes). UMAP projection of spatial-clustered bins shows discrete Close, Far and Epi clusters, with a little overlap, while lesion and sample U M A Pp l o t si n d i c a t en os a m p l eb i a s( F i g .3f). This further confirms our strategy to define spatial compartments and that the identified spatial features apply across both ovarian and peritoneal lesions. Neurons and immune cells spatially organize within lesion Neuro-immune signaling has been hypothesized to play an important role in the cell-to-cell interactions that occur in endometriosis lesions due to the association between increased in flammation and pain signaling4,19. Thus, we evaluated the enrichment of immune cells in the immediate neuron periphery (within 16µm distance) and compared it to all other bins (Fig. 4a). Results indicated increased NK- and T-cell enrichment around neurons for the Epi cluster, while B cells appear to be less present within the immediate neuron periphery in different spatial clusters. We also noted an enrichment in mast cells in proximity to neurons within both the Close and Far Stromal clusters and across peritoneal and ovarian lesions. This observation is consistent with prior studies reporting mast cells in direct apposition to nervefibers in lesions, where they are believed to contribute to nociceptive signaling through degranulation and release of pro-inflammatory mediators 32,33 (Fig. 4a). Interestingly, macrophages appear to be mostly enriched within the immediate neuron periphery in the Epi cluster. Our analysis is based on a grid of 8 × 8 µm bins rather than true single-cell resolu- tion; therefore, individual bins may contain transcripts from multiple cell types. To assess classi fication robustness, we examined the ‘ sec- ond-best’ cell type assignment for bins classified as neuron, across all tissues. Although fibroblasts consistently emerged as the next most likely identity, the confidence score for neuronal assignment was sig- nificantly higher than that of any alternative cell type in these bins across all tissues, supporting the robustness of the neuron bin assignments (Supplementary Fig. 6). Crosstalk between macrophages and neurons is critical in endo- metriosis, as we and others have previously shown based on functional studies and in single-cell analysis 10,17,19. As such, we compared the dis- tribution of distances between macrophages and neurons to that of Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press distances between macrophages and non-neurons. Within spatial clusters, we showed a signi ficant correlation between macrophage – neuron distance, normalized to all other cell types, indicating that macrophages and neurons position closer to each other, while not establishing immediate contact. This was particularly emphasized within the Epi and Far Stroma for ovarian and peritoneal lesions, respectively (Fig. 4b). These results were further con firmed through IMC as we saw CD68+ macrophages located close to βIIITub+ neur- ites (Fig. 4c). To further evaluate the communication between neurons and macrophages, we performed ligand-receptor analysis on the neuron and macrophage bins across the Close Stroma and Far Stroma. Spatial ligand– receptor analysis revealed pronounced regional differences in neuroimmune signaling within endometriosis lesions. In the Close Stroma, we observed extensive bidirectional communication between macrophages and neurons. Both macrophage to neuron and neuron to macrophage signaling were represented by a broad set of ligand– receptor families, including PDGF, SEMA –NRP/PLXNA, MDK- LRP1 and multiple COL–ITG extracellular matrix interactions ac e Spatial Cluster Samples Lesion Type Spatial Cluster Samples Lesion Type Spatial Cell Type Assignment Ovarian Lesion f Peritoneal Lesion UMAP 1 UMAP 2 UMAP 1 UMAP 2 UMAP 1 UMAP 2 Infered Cell Type Proportion Neighborhood Analysis Lesions bd Spatial Cluster B cells DCs Epithelial Cells Fibroblasts Mast cells Monocytes NK cells Neurons Other T cells Vascular cells B cells DCs Epithelial Cells Fibroblasts Mast cells Monocytes NK cells Neurons Other T cells Vascular cells B cells DCs Epithelial Cells Fibroblasts Mast cells Monocytes NK cells Neurons Other T cells Vascular cells B cells DCs Epithelial Cells Fibroblasts Mast cells Monocytes NK cells Neurons Other T cells Vascular cells Samples MME UMAP 1 UMAP 2 OGN UMAP 1 UMAP 2 Spatial1 Spatial 2 Spatial1 Spatial 2 Epi Epi Close Close Far Far Other Other Proportion of Predicted Cell TypesProportion of Predicted Cell Types Fig. 3 | Global cellular map of endometriosis lesions. a, b (left) Spatial plot of representative ovarian (sample O2) (a) and peritoneal lesion (sample P2) (b) showing bin-level cell type assignment for Epithelial Cells, Fibroblasts, Vascular Cells, Neurons, Monocytes/Macrophages, T Cells, NK Cells, B Cells, DCs and Mast Cells (Supplementary Data 4). Arrows indicate B-cell clusters and the scale bar represents 1 mm in (a, b). (right) Bar plot represents the average proportion for each cell type in the Epi, Close Stroma, Far Stroma and Other spatially defined clusters for ovarian (a) and peritoneal lesions (b). Proportions for each tissue individually are in Supplementary Fig. 4a, and larger images are in Supplementary Fig. 11. A linear mixed-effect model was used to compare cell type proportion differences across the stroma clusters in each lesion type; significant comparisons are shown in Supplementary Fig. 4b.c, d Neighborhood analysis heatmap of global cell types for lesions represented in panel (a)a n d(b). Per tissue analysis is found in Supplementary Fig. 4. e Heatmap of unsupervised clustering across all merged samples comparing spatial cell cluster, individual samples, and lesion type. Unsu- pervised clustering comparing peritoneal lesions and, separately, ovarian lesions is in Supplementary Fig. 4c. Black boxes represent similar transcriptomic signatures in the spatially defined clusters. f Uniform manifold approximation and projection (UMAP) of bins across all merged samples, colored by lesion type (top left), spatial cell cluster (top right), and samples (bottom left) and MME and OGN expression (bottom right). Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press (Supplementary Fig. 7c). In contrast, within the Far Stromal regions, the communication network was markedly reduced (Supplementary Fig. 7d), with weaker and scarce signals, suggesting that Far Stroma maintains only a baseline level of neuroimmune activity rather than the amplified, lesion associated signaling observed near glands. Taken together, these results demonstrate that endometriosis lesions harbor a specialized stromal neuroimmune microenvironment characterized by intensified macrophage– neuron cross talk within the Close Stroma microenvironment. Next, we examined the spatial localization of macrophage sub- types (Fig. 4d, e), which we previously identified in our endometriosis single-cell atlas 10. We observed a trend in increased presence of acti- vated macrophages within the Epi and Close Stroma clusters for both lesions (Fig. 4f and Supplementary Fig. 7b). Ovarian lesions show a ab d f c Ovarian Lesion Peritoneal Lesion Ovarian Lesion Peritoneal Lesion Spatial Monocytes/Macrophages Cell Type Assignment T cells NK cells Monocyte/ Macrophages Mast cells DCs B cells Cell Type Enrichment in Neuron periphery vs Non-Neuron periphery Bins (Log/g3676 FC) Lesion Type Ovarian Peritoneal Ovarian Lesion Peritoneal Lesion Neuron Label Neuron bin Neuron periphery Non-neuron periphery Proportion of Predicted Macrophage Cell Types Nearest Macrophage Distance from Neuron vs Non-Neuron (Log2 FC) Lesion Type Ovarian Peritoneal CD10 OGN Pan-CK/EpCAM BIIItub CD68 Epi Close Stroma Far Stroma Epi EpiClose CloseFar FarOther Other g h Ovarian Lesion Peritoneal Lesion Visium HDXenium Epi Close Far Xenium Macrophage Transcript Counts Epi Close Far Other Macrophage Epi Close Far Spatial1 Spatial 2 Spatial1 Spatial 2 Spatial1 Spatial 2 Spatial1 Spatial 2 e Fig. 4 | Spatial association of neurons to immune cells and macrophage sub- populations. a Representation of bins annotated to neurons (orange) and two layers of bins (16 µm) surrounding a neuron bin labeled as neuron periphery (left). Per sample distribution of Log2 fold changes (FC) of immune cell types in neuron periphery compared to non-neuron periphery. Log2 FC > 0 indicates enrichment in neuron periphery <0 in non-neuron periphery.b Box plot of distance calculation between a neuron and macrophage bin compared to a neuron and all other cell type bins. The box represents the span of the interquartile range (IQR), the middle line in the box is the median, the whiskers extend to the smallest and largest value of 1.5 x IQR, the points represent the outliers. Log 2 FC is plotted, normalized to non- neuron bin distances to macrophages. Log2 FC < 0 (orange dotted line) have shorter distances between neurons and macrophages.N = 4 ovarian and 4 perito- neal lesions. Two-sided Wilcoxon test with Benjamini-Hochberg adjustedp-values (FDR) are plotted in order from left to right FDR = 1.41 × 10−2,1 . 7 9×1 0−10,6 . 1 7×1 0−5, 1.25 × 10−10,1 . 9 5×1 0−10,<2 . 2×1 0 −16. c Antibody staining (IMC) indicates that neurons (βIII-tubulin + ) are found located close to macrophages (CD68 + ) in proximity to glands, indicated by white arrows. Peritoneal lesion (sample P9,n =4 ) and ovarian lesion (sample O8,n = 4). Scale bar in (c) represents 100µm. d, e Spatial plots of monocyte/macrophage subpopulations and epithelial bins. Representative image of an ovarian (sample O2) (d) and a peritoneal lesion (sample P2) (e). f Bar plot of average proportion of monocyte/macrophage subpopulations in ovarian (n = 1) and peritoneal lesions (n = 3). Per sample proportions and linear mixed-effect model heatmap comparing average proportions in each lesion type are in Sup- plementary Fig. 7.g Spatial scatter plots of representative ovarian (sample O9) and peritoneal (sample P12) lesions colored by Epi, Close and Far stroma with an overlay of macrophage transcripts in dark purple (CD68, CD163, MRC1, CD14,a n dFCGR3A). h Box plot (same as b) of Xenium macrophage transcript counts in each spatial region (n = 4 samples). Comparisons between regions performed using two-sided Wilcoxon rank-sum test with Benjamini– Hochberg false discovery rate correction (* FDR = 0.029). Source data are provided as a Source Data file. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press significant increase in LYVE1+ and peritoneal macrophages within the Close Stroma cluster (Fig. 4f and Supplementary Fig. 7b). Altogether these results suggest a specific pattern of neurons and macrophages co-localizing in the endometriosis lesion microenvironment, where activated macrophages, expressingSPP1, FN1 and SPARC, are speci fi- cally positioned near epithelial glands. Notably,SPP1-expressing mac- rophages were reported to contribute to extracellular matrix (ECM) and tissue remodeling in endometriosis and other fibrotic diseases 34. These results were further con firmed on independent samples pro- cessed on an independent spatial transcriptomic platform, showing increased macrophage transcripts ( C D 6 8 ,C D 1 6 3 ,M R C 1 ,C D 1 4,a n d FCGR3A) within the Close Stroma (Fig. 4g, h). While this spatial profiling does not provide mechanistic insights, it does confirm that macrophages and neurons are key partners in the endometriosis microenvironment. Our data further supports the concept of speci fic immune cell-neuron interactions that are not incidental but represent a conserv ed feature across endometriosis lesion types. Nociceptors sensory neurons spatially distribute close to glands We next examined the spatial distribution of neuronal subtypes, by subclassifying neuron bins (Fig. 3), expressing pan-neuronal markers (Fig. 5a), based on subtype-specific markers (Fig. 5b). We verified the specificity of the selected neuronal markers by analyzing their expression in a dissociative single-cell endometriosis atlas dataset, which excludes the neuronal population 10. As expected, we found minimal or undetectable expression of neuronal markers in major canonical cell types found in lesions (Fig. 5c). Prior studies have reported increase of sensory neurons near endometriosis lesions 4,5,23,35– 38, prompting us to evaluate speci fic sensory neurons subtypes: nociceptor (TRPV1, NTRK1) which respond to in flammation and transmit pain signals17, proprioceptor (RUNX3, NTRK3)i n v o l v e di n body position sensing39, and mechanoreceptor (PIEZO2, NTRK2)w h i c h respond to mechanical stimuli40. We also included sympathetic neu- rons (TH, SCL6A2), down regulators of stress response41,a n dn e u r o n a l bins increased in immune-signaling related genes ( IL6R, TNFRSF1A), labeled neuroimmune (Fig.5a, b, d, e). 17% of the neuronal bins could not be subclassi fied, though they expressed pan-neuronal mar- kers (Fig. 5a). Among neuronal subpopulations, we identi fied a signi ficant enrichment of nociceptor neurons, marked by expression of pain receptors such asSCN9A and TRPV1, that localize within the Epi cluster in both ovarian and peritoneal lesions (Fig. 5d, e and Supplementary Fig. 8b). On average, nociceptor neurons account for 18% of the neu- ronal bins across lesions. Ovarian lesions show a trend towards increased neuroimmune neurons – which express inflammatory che- mokines such as TNFRSF1A and RAMP1 – within the Close Stroma. Mechanoreceptor neurons, marked by NTRK2 expression, show a trend toward enrichment in the Close and Far clusters across both lesion types (Fig.5d– f, Supplementary Fig. 8b). Notably, they represent the predominant neuronal subset accounting for an average of 40% of the neuronal bins. In contrast, s ympathetic neurons are the least represented, comprising only ~ 0.1% of the neuronal compartment. While we report average expression across lesions, individual tissues exhibit patient-specific heterogeneity (Supplementary Fig. 8a). Con- sistent with the gene expression data showing that TUBB3, the gene encoding βIII‑tubulin, is expressed at very low levels in all other cell types (Fig. 5c), representative IHC images demonstrate minimal colo- calization between βIII‑tubulin and CD10 (Fig. 5f). These findings confirm that βIII‑tubulin reliably marks neuronal cells. Applying this marker to an independent IMC sample cohort, we observed increased neuronal βIII-tubulin expression the Close Stroma of ovarian and peritoneal lesions (Fig.5g). In addition, nociceptor transcripts (SCN9A, TRPV1, TRPA1, NTRK1, CALCA ,a n d TAC1)a r ei n c r e a s e di nt h eC l o s e Stroma in the Xenium cohort of samples (Fig. 5h, i). Beyond nociceptive neurons, other sensory neurons, such as proprioceptors and mechanoreceptors, remain poorly characterized in endometriosis. Our data establishes a comprehensive overview of neuronal diversity within endometriosis lesions. Collectively, these

also reveal distinct spatial localization of neuronal subtypes, with sensory neurons – particularly nociceptors – enriched in the epithelial compartment. In vitro validation of neuron-lesion dynamic interactions Given the preferential localization of the sensory neurons near the epithelial glands and adjacent to the Close Stroma compartment, we sought to validate the dynamic interaction between epithelial, stroma cells – particularly MME expressing fibroblasts – and neurons. We generated 3D in vitro models recapitulating each of these components using: (1) peripheral sensory brain organoids (PSBOs) derived from two female iPSC lines; (2) primary epithelial organoids derived from EMB and peritoneal and ovarian lesions; and 3) CD10+ fibroblasts derived from EMB and lesions, co-cultured in Matrigel domes (see Methods) (Fig. 6a). PSBOs were differentiated for 86 days. Differ- entiation was validated by observing a decrease in proliferation mar- ker, Ki67, and the increase in βIII-tubulin and TRKA protein detection over the course of day 35 to day 95 (Fig.6b). It takes considerable time for neurons to reach maturity in in vitro cultures, as others have reported 42,43. We evaluated the gene expression of mature neuronal markers (MAP2, PENK,a n d RBFOX3) and nociceptor markers ( CALCA, CALCB, NTRK1, NTRK2, NTRK3 ,a n d TAC1) in day 95 brain organoids (Supplementary Fig. 9a). All markers showed clear expression in both PSBO iPSC lines, supporting the emergence of mature neuronal and nociceptor‑like identities. Co-cultures were maintained for 10 days, and RNA was extracted for bulk RNA-sequencing to evaluate the effect of epithelial and stro- mal endometriosis cells on neuronal signaling pathways (Supplemen- tary Movie 1). Canonical makers for epithelial organoids ( KRT8, EPCAM), fibroblasts ( PDGFRA, MME ) and mature neurons ( MAP2, RBFOX3) were used to validate the experimental design and support that these co-cultures mimic the major cell types recapitulated in Epi/ Close Stroma environments (Fig. 6c and Supplementary Fig. 9a). We next performed DEG analysis between the different conditions. PSBOs cultured alone differentiated towards TH + sympathetic and PRPH + sensory neurons with a distribution of 55: 45 percent. However, when the PSBOs were co-cultured with epithelial organoids (PSBO+Org) or with both organoids and fibroblasts (PSBO + Org + Fib), this propor- tion shifted to 40% for TH+s y m p a t h e t i ca n d6 0 %PRPH+s e n s o r y neurons (ratio in Fig. 6d, normalized expression Supplementary Fig. 9b). Fibroblasts alone were not able to drive this neuronal shift in PSBOs. These data suggest that the presence of e pithelial organoids, endometrial and lesion-derived, drive neurons toward a sensory neuron fate. We performed pathway analysis between ovarian and peritoneal co-cultures for the PSBO + Org + Fib conditions. We extracted the pathways involved in neuronal signaling. Genes associated with sen- sory neuron and neuroin flammation signaling pathways were increased in both lesion types, which further validates our spatial DEG analysis (Fig. 6e). We also noted that motor neuron and central ner- vous system signaling-related genes were downregulated in both lesion type-derived co-cultures. These data suggest that epithelial organoids in combination with MME+fibroblasts push neurons toward a sensory fate and increase the neuroinflammation response signaling pathway alongside pain receptor expression, validating spatial tran- scriptomic findings. Finally, we next compared the PSBO neuronal gene signatures to gene sets used to de fine the neuronal subpopulations in the spatial transcriptomics (Fig.6e). Inflammatory genes IL6R and TNFRSF1A are both increased in the presence of organoids, while RAMP1 and CSF1R seem to be driven by the presence of fibroblasts (Supplementary Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press Fig. 9c). We also observe increased gene expression of inflammation- related genes, such as NINJ1, IL1B, TNF, IL18 in PSBO+Org+Fib co- cultures (Supplementary Fig. 9d). Mechanoreceptor genes, such as PVALB and NTRK2, are downregulated in the Org+Fib conditions for all of the tissue-derived cells. Expression of the nociceptor geneCALCA is specifically upregulated in the peritoneal and ovarian lesion-derived organoid and fibroblast co-cultures compared to EMB co-cultures. However, some background signal is still present in the epithelial organoid– only and fibroblast‑only conditions, suggesting additional sources of baseline activity. Future in depth work is needed to delineate the speci fic signaling pathways that drive the neuroin- flammatory crosstalk between cell types. ab eOvarian Lesion Peritoneal Lesiond Pan-Neuronal Markers Ovarian Lesion Peritoneal Lesion f OvarianPeritoneal Proportion of Predicted Neuron Types Epi EpiClose CloseFar FarOther Other Spatial Neuronal Cell Type Assignment CD10Merge βIIITub TRKAMerge TRKB c gh Nociceptor Transcipt Counts Epi Close FarOvarian Peritoneal Spatial1 Spatial 2 Spatial1 Spatial 2 i TRPV1 CALCA SCN9A NTRK1 TRPA1 EPCAM Fig. 5 | Spatial distribution of neuronal subtypes. a Dot plot of pan-neuronal genes used to identify neuron bins.b Dot plot of the neuron only bins showing gene markers used to define the subtypes of neurons.c D o tp l o to fp r e v i o u s l yp u b l i s h e d single-cell dataset10 from endometriosis lesion tissue showing specificity of neuron subtype genes to neurons. d Spatial plots of representative ovarian (sample O2; left) and peritoneal lesions (sample P2; right) showing the distribution of neuronal bins across the tissues, and the dot size was increased to visualize distribution patterns. Epithelial cells are plotted to localize the lesion area. Scale bar in (d) represents 1 mm.e Bar plot of average neuronal subtypes in ovarian and peritoneal lesions across the spatial clusters.f IHC images of a representative peritoneal lesion (sample P10) showing CD10+fibroblasts andβIII-tubulin+ neurons surrounding the epithelial gland (top row) and TRKA + and TRKB + neurons close to the epithelial gland (bottom row). Scale bar represents 100µm. g Boxplots show the distribution of neuron fraction (βIII-tubulin+ cells in region/ total βIII-tubulin+ cells in tissue) across spatial regions (Epi, Close, and Far) within ovarian (n = 4) and peritoneal lesions (n = 4) from Imaging mass cytometry (IMC). Each point represents an independent tissue sample, overlaid as jittered dots. Boxes indicate the median and interquartile range, and whiskers extend to 1.5 × the interquartile range. Statistical differences between regions were assessed separately within each lesion type using two-sided Wilcoxon rank-sum tests with Benjamini– Hochberg false discovery rate (FDR) correction across comparisons (Close vs Epi, Close vs Far, and Epi vs Far) (* FDR = 0.00119).h Box-and-whisker plots with overlaid jittered points show noci- ceptor transcript counts in Epi, Close, and Far stromal regions across independent tissues (n = 4) from Xenium. Boxes denote the median and interquartile range, and whiskers extend to 1.5 × IQR; individual points indicate biological replicates. Comparisons between regions were performed using a two-sided Wilcoxon rank- sum test with Benjamini– Hochberg false discovery rate (FDR) correction, resulting in no significant comparisons. i Representative image of Xenium nociceptor tran- scripts in a peritoneal lesion (sample P11). Source data are provided as a Source Data file. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press Follicle AMH levels correlate with lesion epithelial distance The link between infertility and endometriosis remains unclear, parti- cularly how the presence of endometriosis lesions impacts the quality of ovarian follicles. We unexpectedly observed the presence of follicles close to the endometriosis lesion in one of the ovarian samples ( s a m p l eO 3 ) .W ei d e n t ified follicles based on H&E staining and the expression of speci fic follicular markers ( ZP3, GDF9,OOSP2 and FIGLA) 44(Fig.7a). Anti-Mullerian Hormone (AMH), which is specifically expressed by small, growing follicles, is widely used as a marker for ovarian reserve and ovarian stimulation capacity 45. Patients with higher AMH levels have been associated with better fertility outcomes46.A l l identified follicles were located in the Far Stroma (Fig. 7b), with a variable range ofAMH levels per follicle (Fig.7c). We noted a significant positive correlation between a follicle’ s AMH gene expression level and its distance from the epithelial gland (Fig.7d), suggesting that follicles may be impacted by the lesion. While these data are exploratory and a similar analysis needs to be replicated in a larger cohort, they do demonstrate the remarkable capacity of continuous high-resolution spatial transcriptomics to infer potential mechanisms associated with endometriosis infertility. ab e Cell type Tissue type EPCAM KRT8 PDGFRA MME c Neuronal Pathway comparison in Lesion Org+Fib Cell type Tissue type d βIIITub/ KI67 /DAPI TRKA / DAPI Day 35Day 56Day 95 Proliferation Maturation Log2FC(PRPH/TH) PSBO PSBO +Fib PSBO +Org PSBO +Org +Fib 1.0 0.5 0.0 - 0.5 Tissue type Peritoneal Ovarian Fig. 6 | In vitro cell model of neuronal-epithelial-fibroblasts interactions in endometriosis. aSchematic of experimental design. Primary cell lines for epithelial organoids (Org) and fibroblasts (Fib) were derived from endometriosis patient endometrium, peritoneal or ovarian lesion tissues. Peripheral sensory brain orga- noids (PSBO) were derived from a female iPSC-line following the growth factor protocol outlined in the Methods section and Supplementary Fig. 10. Cells were co- cultured for 10 days in different combinations for each tissue type: PSBO, PSBO + org, PSBO +fib, PSBO + org +fib. Bulk RNA-sequencing was performed on all the co- culture conditions (n = 2). The resulting transcriptomics signatures were compared for DEGs and pathway analysis. Supplementary Movie 1 of co-cultures. Created in BioRender. Haney, C. (2026) https://BioRender.com/as0z5ynis licensed under CC BY 4.0. b IHC staining was performed on day 35, 56 and 95 PSBOs, confirming the successful differentiation of iPSC to mature neurons (βIII-tubulin), sensory noci- ceptor (TRKA + ) and a decrease in proliferation (KI67). Scale bar represents 25µm. c Heatmap showing the expression of epithelial markers (EPCAM, KRT8)a n d fibroblast markers (PDGFRA, MME) in analyzed co-culture conditions.d Bar plot of the proprotion of PRPH+ sensory neurons over TH + sympathetic neurons in all co- culture conditions, normalized to the PSBO alone condition.n =3P S B O ,2 PSBO + fib, 2 PSBO + org, 5-8 PSBO + org + fib per tissue type. Error bars repre- sent +/− SE. e Dot plot of neuronal pathways significantly upregulated (NES > 0) or down regulated (NES < 0) in the ovarian and peritoneal co-culture PSBO + org +fib conditions. NES is calculated by normalizing raw enrichment scores against the mean of permutaiton-based enrichement scores for each gene set. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press

This study presents a high-resolution spatial transcriptomic profiling of human ovarian and peritoneal endometriosis lesions. Previously, we had established a comprehensive single-cell atlas, based on dis- sociative approaches, of similar lesions. Yet this atlas lacked the spatial information needed to understand dynamic interactions and ecosys- tems driven by the spatial distributions of cells. In addition, due to technical limitations of dissociative approaches, neurons, a known component of lesions and a key driver of endometriosis symptoms, were not detected. Now we have established a comprehensive spatial atlas of endometriosis lesions, at sub-cellular resolution, and provide an unbiased survey of the neuronal landscape within these structures. Although endometriosis is a high ly heterogeneous condition, we employed spatially defined regions to identify shared pathways across two of the most prominent lesion types (Fig. 8). This approach emphasizes the importance of understanding the disease’ sc o m p l e x i t y and spatial patterning to identify common features and mechanisms among different lesion types. While spatial transcriptomics provides a powerful approach for inferring ecosystems that play a critical func- tion in lesion pathobiology, it provides only a snapshot view of the dynamic cellular interactions underlying a complex microenviron- ment. Thus, we developed a 3D co-culture model, using patient- derived primary lines, which recapitulates some of the critical cell types and their functional interactions to begin to understand dynamics between neuron-epithelial partners. From our spatial transcriptomic analysis, we defin e ds h a r e ds p a - tial niches across lesion types, labeled Epi, Close and Far Stroma, based on mutual distance from glandular structures and shared expression signatures. Surprisingly, we found compelling similarities between spatial clusters of super ficial peritoneal and ovarian lesions, two morphologically distinct lesions, with variable clinical stages and treatments. Our data support findings of a recent endometrioma spatial analysis from Liu and colleagues, where spatial niches were also observed, with fibroblasts proximal to epithelial glands implicated in wound healing, whereas distal fibroblasts were de fined as pro- inflammatory 12. The authors also determined that low-resolution spa- tial transcriptomics, while informative, does not provide suf ficient clarity for comprehensive analysis of immune cells and neuronal organization. Indeed, we also demonstrated that low-resolution spatial transcriptomics did not reliably depict the positioning of immune and neuronal cells and only provided a discontinuous map of lesions. Our spatial transcriptomic analysis showed the presence of B cell aggregates in the Far cluster consistently across all ovarian and peri- toneal lesions. Previous reports have described an increased presence ab c d Ovarian Lesion AMH Gene Expression in Follicules Fig. 7 | Ovarian follicleAMH level is correlated to distance from epithelial gland. a Spatial plot of ovarian lesion (sample O3) with Epi, Close Stroma, Far Stroma and Follicles labeled.b AMH expression in Follicle bins.c Boxplot of average AMH levels per follicle in the Far Stroma. We did not identify follicles in the Close Stroma. Each point represents a follicle. Box indicates the median and interquartile range, and whiskers extend to 1.5 × the interquartile range.d Linear model comparing the follicle distance from the centroid of the epithelial gland to the averageAMH level of the follicle. Blue line is the bestfit linear regression model (y =− 0.3 + 5.8 × 10 −5x). Shaded gray is 95% confidence interval. The two-sided Pearson’ s correlation was calculated to be R = 0.53 and the p = 0.00099. n = 1 ovarian lesion, 35 follicles. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press of peripheral and peritonealfluid B cells in endometriosis patients, and showed that plasma cells produce autoantibodies against the endo- metrium and autoantibodies commonly found in autoimmune diseases 29,47. However, most of the B-cell biology in female reproduc- tive pathologies remain unknown 48. Our work further suggests the contribution of B cell aggregates to the Far Stroma niche likely involved in promoting a chronic inflammatory environment. Parallels between endometriosis and autoimmune diseases, or immune-related disorders 27, have been suggested in previous studies; however, our study further explores this perspective and identifies these aggregates as a common immunological mark across lesion subtypes, targeting of which could represent a promising therapeutic approach. An unresolved question in the field is whether neurogenesis

from chronic in flammation or is driven by the presence of lesion-specific epithelial cells and fibroblasts in human tissues. With our data, we show an increase in nociceptor neurons expressing pain receptors close to epithelial glands across both peritoneal and ovarian lesions. Our in vitro 3D cellular model utilizing sensory brain organoids provides functional evidence of epithelial cells andfibroblasts in pro- moting sensory pain neurons. Further spatial analysis supports the increased presence of activated macrophages in areas where noci- ceptor neurons and epithelial glands colocalize. These data suggest a potential synergistic effect whereby epithelial cells influence neurons to express pain receptors. In turn, these neurons likely increase the production of inflammatory molecules (such asNINJ1), promoting the recruitment and activation of immune cells, particularly macrophages, and contributing to the sustained in flammatory microenvironment. Indeed, NINJ1 was shown to mediate tissue remodeling and in flam- mation by regulating leukocyte entry, adhesion, activation, and migration 49, and represents a potential candidate factor involved in endometriosis neuroimmune crosstalk. Further functional and mechanistic studies with complex 3D systems, recapitulating all these central cellular components, will help elucidate the role of macro- phages or mast cells in this process. Endometriosis is an estrogen-dependent disease, and steroid hormone cyclic variation influences the immune landscape in repro- ductive female organs and related-diseases. Technical and cost lim- itations of high-resolution spatial transcriptomics hinder the profiling of large clinical datasets. In this study, we chose to focus on lesions from patients under hormonal contraception (HC) to minimize the impact of the menstrual cycle variation on our dataset. HC is also a standard treatment for endometriosis patients not seeking pregnancy. Interestingly, this study also suggests that HC may not be effective against the neurogenic and pain-modulating processes driven by the presence of epithelial glands and close stroma. Additional studies have shown that estrogens play an important role in the mast and macrophage-neuron communication in endometriosis 33,50. Therefore, further research is necessary to evaluate the impact of HC on lesions in the context of neuroimmune signaling and to investigate how this may be altered in non-hormonally treated patients, and across menstrual cycle phases. Finally, one limitation of spatial transcriptomics is that it encom- passes only a single spatial modality. Future integration of spatial proteomics, ECM organization and other modalities such as lipidomics will provide additional insights, further enhancing our understanding of the complex microenvironment associated with endometriosis lesions. Spatial transcriptomics infers co-localization of cells but can- not prove direct cell-cell contact. Further in vitro cellular modeling is needed to truly address cell-cell contact and functional impact. Our initial spatial profiling focused on the two most common subtypes -- superficial peritoneal and ovarian lesions. However, considering the significant heterogeneity of endometriosis, it is essential to investigate all lesion subtypes, including deep in filtrating endometriosis and extra-pelvic lesions, in order to identify shared pathways that could contribute to the development of comprehensive treatment strategies for all lesion variants. By integrating whole transcriptome data with tissue morphology, spatial transcriptomics has revolutionized our understanding of dis- eases such as cancer and irritable bowel disease 51,52. Here, we fill a Fig. 8 | Representation model of common cellular features across lesions. Schematic of proposed model for spatial cluster cell type distributions in ovarian and peritoneal lesions, highlighting common features across lesions. Epi Stroma cluster showed increased expression in Nociceptor neurons. The Close Stroma cluster showed increased expression of CD10+fibroblasts, Nociceptor neurons, activated macrophages in both lesion types and NK cells in ovarian lesions only (indicated with *). The Far Stroma cluster showed increased vascular structures, B cells and lymphatic vessels in both lesion types. Globally, sympathetic neurons were sparse, and there was a large increase in sensory neurons in the lesion microenvironment. Lastly, the data showed mast cells are enriched close to neu- rons in both the Close and Far Stroma clusters. Created in BioRender. Courtois, E. (2026) https://BioRender.com/husfc4nis licensed under CC BY 4.0. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press critical gap in endometriosis research by revealing precise spatial cellular interactions within lesions and establishing parallels between two of the most common endometriosis lesions. We found that ovarian and peritoneal lesions share spatial features. Our findings highlight a prominent role for immune cells, uncover fibroblast compartments surrounding glands, and define the spatial distribution of neuronal and macrophage subsets. Understanding the cellular and molecular drivers of pain— typically among the earliest symptoms— may offer a path to preventing or mitigating the disease, particularly its most debilitating manifestations, and informing the development of therapeutic strategies.

Human endometriosis tissue collection This study complies with all relevant ethical regulations and was approved by The Jackson Laboratory and the University of Connecticut Health Center Institutional Review Board (IRB). Patients ’ enrollment was conducted under either The Jackson Laboratory IRB 19-145J-2 and the Endometriosis Research Innovation Support & Education Bior- epository (EndoRISE) CT biorepository IRB 24-111JSF-1, with all parti- cipants providing written informed consent, including agreement to share demographic information. Individuals assigned female at birth, ages 18 and above, were consented and enrolled in the CT EndoRISE biorepository for longitudinal endometriosis research, agreeing to share biospecimens (tissues, whole blood, urine, peritonealfluid) and phenotypic data. Informed consent included secondary usage of tis- sues for endometriosis research, such as organoids and cell line gen- eration, and subsequent research applications. Study participants donating tissue samples for this study did not receive compensation. Tissues and phenotypic data (including quality of life, clinical and medical history data) were collected by endometriosis specialists at UCHC following the World Endometriosis Research Foundation (WERF) Endometriosis Phenome and Biobanking Harmonization Pro- ject (EPHect) guidelines and standard operating procedures. Speci- mens were immediately transported on ice to the biorepository at The Jackson Laboratory for processing and long-term storage (within 1 h of collection). The EndoRISE Sample Access Review Committee (com- prised of staff from UCHC and The Jackson Laboratory) reviewed the request for the data and biospecimen distribution for this project and approved the request unanimously. Patient information and use of specimens in this project can be found in Supplementary Data 1. Tissues were collected at the University of Connecticut Health Center and transported to The Jackson Laboratory on ice in MACS tissue storage solution (Miltenyi, Cat:130-100-008). Tissues werefixed in 10% Formalin in Phosphate Buffered Saline (PBS) for 24 h at 4 oC. Following fixation, tissues were washed three times in PBS and stored in 70% Ethanol at 4 oC until sent to embedded in paraf fin blocks. Formalin-fixed Paraffin-embedded (FFPE) tissue blocks were sectioned at 5 µm and stained with Hematoxylin and Eosin (H&E) to visualize histology or Picrosirius Red (PS) to identify areas of dense collagen staining in lesions. For tissues where epithelial glands surrounded by fibroblasts were not obvious with the histological staining, we per- formed immunohistochemistry using anti-CD10 (R&D Systems, Cat:AF1182) to identifyfibroblasts and anti-Pan-cytokeratin (Biolegend, Cat:914204) to identify epithelial glands as previously described 10 (see below IMC, IHC and confocal imaging section). Tissue processing for spatial transcriptomics Once a lesion area was identi fied with epithelial glands and CD10+ fibroblasts, a 6.5 × 6.5 mm region of interest (ROI) was selected for spatial transcriptomics. RNA quality control was performed on the selected FFPE blocks with the Qiagen RNAeasy FFPE kit (Cat: 72504), followed by Agilent tapestation analysis of RNA. Only FFPE blocks with a DV200 above 30% were used for all spatial transcriptomics platforms. FFPE blocks that passed all quality control metrics were sectioned at 5 µm and put on glass slides. The sectioned tissues were either processed following the 10X Visium V2 protocol, Visium HD V1 pro- tocol, or Xenium Human 5 K panel by the Single Cell Biology Core at The Jackson Laboratory. Visium V2 was run on three samples, eutopic endometrium (EMB), peritoneal lesion and ovarian lesion, from two patients. Visium HD was run on eight samples, four peritoneal lesions and four ovarian lesions, from seven patients. Supplementary Data 2 includes tissue information and spatial metrics. Xenium was run on three peritoneal lesions and one ovarian lesion. Spatial transcriptomics pre-processing and cluster identification The Illumina base call libraries were converted to FASTQfiles. The 10x Genomics Space Ranger pipeline (v3.0.0) was used to align the sequencing reads to the human reference genome GRCh38-2020-A. For tissues processed with the Visium V2, we kept the spots that express more than 100 genes and have less than 10% of mitochondrial gene counts. QuPath was used to segment the nuclei in the H&E image. The transcriptional data was used to manually label the spots con- taining EPCAM + epithelial glands, MME + closefibroblasts and OGN + far fibroblasts. For tissues processed with Visium HD V1, count matrices of aggregated 8 × 8µm bins resulting from SpaceRanger (10x Genomics) were processed with custom R and Python scripts. Bins with low UMI counts were excluded using sample-specific UMI cutoffs. Each cutoff was established by first de fining a tissue mask of contiguous bins intersecting at least one nucleus. The UMI cutoff was then taken as that UMI count that best separated bins within the tissue mask from those outside the tissue mask, as quantified by F1 score [i.e., 2 x precision x recall / (precision + recall)]. Nuclei were counted in Python by applying the ‘ 2D_versatile_he’ model of stardist to an image normalized to the 5th and 95th percentiles. Nuclei overlapping each bin were counted by performing spatial joins between the nuclei coordinates and bin coordinates using geopandas in Python. The tissue mask was then defined by applying morphological smoothing operations to bins intersecting at least one nucleus using OpenCV in Python. Specifically, islands on the periphery of the initial mask of non-empty bins were connected using morphological closing (via morphologyEx and a square kernel, each of whose sides were 80 µms — i.e., ten times the size of an aggregated bin). Small, isolated groups of bins were then removed via morphological opening (via the same function and ker- nel). This resulted in a refined tissue mask. For each tissue individually, 8 µm bins were removed outside of the refined tissue mask, and 8 µm bins were removed with a UMI count less than the threshold that separates the foreground and the background of the re fined tissue mask. The H&E was used to identify the epithelial glands, but unsu- pervised clustering was used to identify thefibroblast compartments. The clusters closest to the glands and highest inMME expression were labeled as ‘ Close Stroma’ . The clusters outside of the MME+ areas and higher in OGN were labeled as ‘ Far Stroma’ . The left-over clusters that did not fitt h i sd efinition were labeled as ‘ Other’ . For tissues processed with Xenium 5 K human panel, each tissue was run through the Points2Region pipeline 26,53 using a subsample of 750k transcripts. This created a spatial unsupervised clustering of the transcripts was projected back to the entire tissue. The clusters were labeled as Epi, Close or Far based on the expression of the top DEGs from the VisiumHD data (Fig. 2h and Supplementary Fig. 3). Visium HD sample integration and clustering Due to large sample size (Supplementary Data 2) sample integration was performed on a representative set of cells from each sample. This sampling approach is based on a geometric sketching method, and it ensures that rare populations are also included in the subset 54.T h e integration of eight samples was done with Seurat using the Harmony Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press

on Principal Component space. Then, the integration was expanded to the full dataset. An aggregated pseudo-bulk-based workflow was used, where each cell is an aggregated sample of one cluster in each sample. Then clustering was done using the complete method and Euclidean dis- tance (Fig. 3e). After cell aggregation, cell type- speci fic differential expression was performed using the DESeq2 method and the Wilcoxon rank sum test. Pathway analysis for the biological process gene ontology terms (GO:BP) was run. Visium HD cell type assignment and distance analysis To predict the major cell type within each bin, Seurat ’ s AddModule- Score function was used with known markers for the following cell types: Epithelial Cells, Fibroblast Cells, Vascular Cells, Neurons, Mac- rophages, NK cells, T Cells, B Cells, DCs and Mast Cells (Supplementary Data 3). The resulting scores were scaled and filtered for a z-score above 1 to be labeled as a cell type. The spatial location of all the cell types was plotted using SpatialDimPlot, and increased the point size to visualize the patterns. The proportion of each cell type per sample in each defined fibroblast compartment (Epi, Close or Far) was calculated by taking the number of bins for that cell type in the fibroblast com- partment divided by the total number of bins in that fibroblast com- partment and graphed it in a proportion plot. To assess which cell type proportions were signi ficantly different between the stroma com- partments across lesion types, post-hoc analysis was performed of linear mixed-effect models fit independently for each cell type and lesion type, with compartment as the fix e de f f e c ta n ds a m p l ea st h e random effect (Supplementary Fig. 3b) To identify subpopulations of neurons, all the bins labeled as neurons from the previous global cell typing were subsetted and run through AddModuleScore with another set of genes speci fict ot h e subpopulations (Supplementary Data 3). Again, the scores were scaled, but a z-score above 0.01 was used to identify the subpopulations of neurons, as these genes are often low in abundance. The same process was performed to identify macrophage subpopulations using the gene markers from a published single-cell dataset of endometriosis tissue 10 (Supplementary Data 3). The same linear mixed-effect model was used to calculate the differences between the subtypes of neurons and macrophages in the fibroblast compartments (Supplementary Fig. 5b and Fig. 5g). To understand the spatial connectivity of cells, Squidpy neighbors enrichment analysis was performed on each VisiumHD sample indivi- dually across the global cell types. This resulted in a heatmap for each sample showing which cell types are spatially clustered close to each other (Fig. 3c, d and Supplementary Fig. 5a– f). Enrichment analysis between neurons and immune cells To check whether immune cells are enriched around neuron bins, the proportion of each immune cell type in the bins directly surrounding the neuron bin, de fined by two layers of 8 ×8 µmb i n s( n e u r o np e r - iphery in Fig. 4a), was calculated. The proportion of each immune cell type around all the non-neuron bins was calculated (non-neuron per- iphery in Fig. 4a). These values were normalized to show their fold change. To evaluate the enrichment of macrophages in the vicinity of neurons within each compartment (Epi, Close Stroma, and Far Stroma), the distance was calculated between each neuron bin in that compartment and its closest macrophage, noting that the neuron and macrophage bins may not fall in the same compartment. Similarly, the distance between each non-neuron, non-macrophage bin in each compartment was calculated relative to its closest macrophage bin. The adjusted p-value was graphed from the result of a Wilcoxon test corrected using the Benjamini-Hochberg procedure. Within each sample, all distances were log normalized to the median of the non- neuron/non-macrophage to macrophage distances within that sample (Fig. 4b). As such, a log-normalized distance less than zero indicates that a macrophage is closer to a neuron than the median distance between a macrophage and a non-neuron, non-macrophage. Ligand-receptor analysis We inferred spatially resolved ligand – receptor– mediated cell – cell communication using COMMOT55, a framework that reconstructs sig- naling interactions in spatial transcriptomics data through a collective optimal transport model. This method incorporates spatial distances between cells and competition among ligand and receptor species, enabling biologically realistic estimation of communication strength and directionality. Spatial transcriptomics data were preprocessed following stan- dard COMMOT guidelines, including log‑transformation and normal- ization of total transcript counts. Ligand– receptor pairs were obtained from curated databases of CellChat 56 and CellPhoneDB57 available through COMMOT, and were formatted into a structured data frame for downstream analysis. COMMOT then computed pairwise ligand- receptor communication matrices using spatial constraints (distance threshold of 50µm) and support for heteromeric receptor complexes. Downstream analyses included quantifying sender and receiver activities for interactions between macrophages and neurons within Close Stroma and Far Stroma. IMC, IHC and confocal imaging We used previously published Imaging Mass Cytometry (IMC) data reanalyzed for markers specific to this project (Fig. 2 & 4)8.I m m u n o - histochemistry (IHC) was adapted from a previously described protocol 8 for the localization of TRKA and TRKB neurons (Fig. 5). Briefly, FFPE tissue sections were melted for 10 minutes at 55 oC, deparaffinized in 2 washes of Histoclear (National Diagnostics, Cat: HS- 200). The tissues were rehydrated in a gradient of ethanol to water washes. Antigen retrieval was performed in EDTA pH 9 (Vector Labs, Cat: H-3301-250) for 15 min at 95 oCin a pressure humidity chamber (BioSB TintoRetriever). Sections were permeabilized in 0.1% triton for 10 min, then blocked in 10% donkey serum and 3% BSA solution for 1 hour at room temperature. Primary antibodies were incubated in 3% BSA in PBS overnight at 4 oC. Supplementary Table 1 includes primary antibody information. Unbound primary antibody was washed away, then sections were incubated in secondary antibodies for 2 h at room temperature. Secondary antibodies were used at 1:1000 dilution, Donkey Anti-Goat IgG-Alexa fluor 647 (Invitrogen, Cat: A32849) and Donkey anti-Mouse IgG- Alexa Fluor 555 (Jackson Immunological, Cat: AB_3095480). A Dapi counter stain was used. Slides were coverslipped with Flouramount-G (Life Tech, Cat:00-4958-02) (Supplementary Table 1). IHC images were acquired using a Leica TCS SP8 laser scanning confocal microscope (Leica Microsystems, Wetzlar, Germany) equip- ped with a DM6000 upright microscope stand. The system was con- trolled by Leica Application Suite X (LAS X) software (version 3.5.7.23225). Images were obtained using the HC PL APO CS2 40x/1.30 Oil objective (Cat #: 11506358) and HC PL APO CS2 63x/1.30 Glyc

(Cat #: 11506353). Leica Microsystems’ Type F immersion oil (Cat #: 11513859) was used with the 40x objective, while Leica Micro- systems’ Type G Immersion Liquid (Cat #: 11513910) was used with the 63x objective. Excitation was provided by a combination of a 405 nm diode laser and a White Light Laser (WLL). Laser power was controlled via the software and was typically set between 1 – 10% to minimize photobleaching. Fluorescence emission was detected by a combina- tion of photomultiplier tubes (PMTs) and high-sensitivity hybrid detectors (HyDs). The spectral detection bands were tuned to opti- mally capture the emission of each fluorophore while minimizing crosstalk. The spectral detection bands were tuned as follows: • DAPI: Excitation at 405 nm, emission collected at 410– 494 nm. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press • Alexa Fluor 488: Excitation at 499 nm, emission collected at 509– 572 nm. • Alexa Fluor 568: Excitation at 577 nm, emission collected at 587– 648 nm. • Alexa Fluor 647: Excitation at 653 nm, emission collected at 658– 779 nm. Images were acquired as 8-bit files with a frame size of 512 × 512 pixels. A conventional galvanometric scanner was used. For imaging, a scan speed of 400 Hz was used with a line averaging of 1. The pinhole was set to 1 Airy unit (AU) to ensure confocality. For Z-stacks, images were acquired using system-optimized Nyquist values for the step size, resulting in a step size of 0.333 µm for the 63x objective (total axial distance of 7.66 µm) and 0.346 µm for the 40 x objective (total axial distance of 13.85 µm). Sequential scanning was employed for multi- channel imaging to prevent bleed-through between channels. The scan sequence was set to “between frames”. Raw image files (.lif) were processed and analyzed using LAS X software and Fiji/ImageJ. The captured signals were max projected. IMC quantification Per‑cell marker intensities and centroid coordinates were extracted from an AnnData object and processed independently for each ima- ging field‑of‑view (FOV). Raw marker values (PanCK, CD10, OGN) were arcsinh‑transformed (cofactor = 5) and robustly scaled to [0,1] using the 1st– 99th percentiles per FOV. PanCK‑positive cells were defined as those at or above the 90th percentile of normalized PanCK within each FOV. Distances from every cell to the nearest PanCK+ cell in the same FOV were computed using a cKDTree nearest ‑neighbor search, con- verted to microns, and constrained to outside ‑only values by setting negative distances to zero. To quantify spatial gradients, normalized marker intensities were binned by distance (8 µm bins), averaged per FOV, and then aggregated across images, followed by light Gaussian smoothing ( σ =1 . 5 b i n s ) . A k e r n e l‑smoothed variant was also com- puted by evaluating Gaussian ‑weighted means on a 0 – 200 µmg r i d (8 µms p a c i n g ;σ = 1.5 × bin width). The pipeline yielded per‑cell anno- tations (*_norm, PanCK + mask, distance values) and cohort‑level dis- tance‑dependent profiles for CD10 and OGN. AMH-follicle distance analysis We received one lesion that came from an ovarian cortex and con- tained follicles near an endometriosis lesion (Fig.7). The Seurat func- tion SpatialFeaturePlot was used to plot the AMH level for the follicle cluster on the tissue. K-nearest-neighbor was used to spatially localize the follicles into the closest stroma cluster. To analyze the quality of those follicles, the centroid of the epithelial gland and each follicle were calculated. DBscan was used to detect the bins for each follicle, resulting in 35 follicles and the average AMH for each follicle was cal- culated. The Euclidean distance from the epithelial gland centroid to the follicle centroid was compared to the mean AMH level for each follicle using a linear regression model. Endometrial epithelial organoid andfibroblast culture Tissues were dissociated into single-cell suspension as previously described10. The single-cell suspension was resuspended in cold Matrigel (Corning, Cat:356231) in 50 µL domes in a 24-well plate. Epi- thelial organoid media, previously described58, was added to cover the domes. For thefirst passage, to separate the epithelial organoids from the fibroblasts, the fibroblasts were allowed to grow to the bottom of the dish, then the Matrigel dome was carefully removed to leave the fibroblasts in place. The fibroblasts were washed with DPBS, and fibroblast media was added. Fibroblast media contained DMEM/F12 (Gibco, Cat:11320082) supplemented with 10% dextran-coated char- coal stripped FBS (Sigma, Cat:C6241, Gibco, Cat:A5256801), 1% peni- cillin-streptomycin, 2 mM L-glutamine, 1 nM β-Estrodiol (Sigma, Cat:E4389), 2 ug/mL insulin (Sigma, Cat:91077 C). Epithelial organoids were passaged every 7 days according to previously described protocol 58. Fibroblasts were passaged 1:4 every 7 – 10 days once 80% confluent with a five-minute digestion of TrypLE (Gibco, Cat:1260510) to lift the cells from the plate. Peripheral sensory brain organoid culture Human induced pluripotent stem cells (hiPSCs) were cultured and induced to brain organoids, adapted from the following protocols43,59. All material and reagents can be found in the protocols.io (dx.doi.org/ 10.17504/protocols.io.kxygxw1ddv8j/v1). The hiPSCs were derived from two healthy females of African and Puerto Rican ancestry at the Jackson Laboratory (GeDiX697.3 J and GeDiX653.1 J). This hiPSC lines were fully validated for trilineage differentiation potential and whole genome sequencing. Brie fly, to induce neuronal progenitor cells in organoid formation, the dual-SMAD method was adapted as pre- viously described 60 on a shaking platform at 95 rpm inside a 37 oC incubator. Briefly, cells were cultured in 100% KSR medium supple- mented with 10 μMS B 4 3 1 5 4 2a n d1μML D N 1 9 3 1 8 9f o rd a y s4– 8p o s t - seeding. To promote the differentiation to sensory neurons, a protocol was adapted 61,2 . 5μMS U 5 4 0 2 ,1μMD A P Ta n d1 . 5μM CHIR99021 was added on days 5– 13 of differentiation. The cells were transitioned from 100% KSR media to 100% Neural Induction Media (NIM) over a 7 day gradient from day 8– 14. Once in 100% NIM on day 14, the supplements were switched to 10 ng/mL FGF2 and 10 ng/mL EGF2 for 6 days, washing every other day. The cells were finally put into Terminal Dif- ferentiation (TD) media supplemented with 20 ng/mL BDNF, 20 ng/mL GDNF, 25 ng/mL NGF, 10 ng/mL NT-3, 200 µMA s c o r b i cA c i da n d 0.5 mM dbcAMP on day 20 post feeding. The cells were washed every 3 days with fresh media and supplements (Supplementary Fig. 10). To verify the neural induction to peripheral sensory organoids, the brain organoids were were fixed in 4% PFA for 4 hours at r4 oC, washed with PBS 3 times for 5 min, then put in 30% sucrose for 48 h at 4 oC. Once the organoids sunk in the sucrose solution, they were embedded in OCT and sectioned at 14 µm. The sections were then stained for anti-β-III-tubulin (Promega, Cat: G7121), anti-KI67 (Invitro- gen, Cat:14-5698-82), TRKA (R&D Systems, Cat:AF175) to detect the decrease in proliferation and increase in maturation. The immune fluorescence protocol was as follows: cells were washed 3 times with PBS, permeabilized with 0.25% Triton-X100 in PBS for 15 minutes, washed 3 time with PBS, blocked with 10% donkey serum, 1% BSA in PBS for 1 h at room temperature, and incubated in primary antibodies at 1:100 dilution in 0.1% Triton-X100, 1% BSA in PBS overnight at 4oC. On the second day, cells were washed 3 times for 5 min in PBS, incu- bated in secondary antibodies 1:1000 for 2 h room temperature, washed 3 times for 5 min with PBS, incubated with DAPI for 5 min, washed 3 times for 5 min with PBS. PSBO IHCs were imaged on Leica TCS SP8 laser scanning confocal microscope using a 63x objective as described above. Images were optimized for brightness and contrast across the entire field of view with FIJI. Co-culture and bulk RNA-seq of brain organoids, epithelial organoids and fibroblasts To assess the molecular changes that occur when endometriosis epi- thelial cells and fibroblasts are in contact with sensory neurons, co- culture experiments were performed. Eutopic endometrial, peritoneal lesion and ovarian lesion epithelial organoids and fibroblasts were cultured in different combinations (denoted in Fig. 6a, patient line information in Supplementary Data 1) with 86 day sensory peripheral brain organoids. There were 6 conditions: PSBO alone (n =4 ) ,e p i t h e l i a l organoids alone ( n =4 ) , fibroblasts alone ( n =3 ) , P S B O+fib( n =2 ) , PSBO + epi (n =2 ) ,P S B O+fib+e p i( n = 20, 4 EMB, 8 ovarian, 8 perito- neal). After 10 days of co-culturing, the Matrigel domes were melted on ice for 30 min with Cell Recovery Solution (Corning, Cat: 354253), the cells were spun down 520 g for 5 min at 4 oC, cells were washed with Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press PBS once and spun down again. The cells went right into RNA extrac- tion following the Qiagen RNeasy Mini Kit protocol (Cat: 74106). RNA quality was assessed with Agilent Tapestation, and only samples with a RIN above 6 were used for bulk RNA sequencing. Samples were prepped for whole-genome library prep. Then sequenced on Novaseq XP l u s1 0 Bw i t h3 0 0c y c l e sw i t ha na v e r a g eo f2 . 9m i l l i o nr e a d sp e r sample. Analysis of bulk RNA-seq Reads were aligned with the human genome GRCh38-2020-A,filtered with nf-core/rnaseq (v.3.19) utilizing the STAR aligner 62,a n dc o u n t s were normalized to counts per million. DeSeq2 was used to subtract the brain organoid alone conditions from all the rest of the conditions to get DEGs. The VST scaled matrix was used to assess gene expression differences between groups. Then, the different co-culture conditions were compared using the FGSEA pathway analysis. The neuronal- related pathways were selected to compare between groups. Statistics and reproducibility All statistical tests were run as described in the above methods sec- tions. Adjustedp-values or FDR were used when possible. No statistical

was used to predetermine sample size due to limited avail- ability to precious human samples. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability The raw FASTQ data for Visium, VisiumHD, and bulk-RNA sequencing is deposited on the dbGAP public data repository (Accession # phs004891.v1.p1 [ http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/ study.cgi?study_id=phs004891.v1.p1]). Access requests are submitted through dbGAP’ ss t a n d a r dD a t aA c c e s sR e q u e s tp r o c e s sa n dr e v i e w e d by the relevant Data Access Committee, with a typical response time of 2– 4 weeks. Xenium data and Visium Space Ranger outputs have been deposited on EGA (EGAS50000002140). Access requests are sub- mitted through EGA and reviewed by the Data Access Committee (EGAC50000001139). Requests will be required to complete a Data Access Agreement through JAX and the requesters institution. Source data are provided for Xenium and IMC data quanti fications in the Source Data file. Restrictions on use for all data include: (1) research must be for the sole purpose of advancing endometriosis research, diagnostic tools, therapeutic development and advancing patient care. (2) users are prohibited from posting or making available data through open- access public websites and/or databases (controlled access databases are acceptable). (3) Users shall appropriately acknowledge and refer- ence the EndoRISE Biorepository in all publications, presentations, and other scholarly or scienti fic communications that include data or findings derived, in whole or in part, from Materials obtained through the EndoRISE Biorepository. Source data are provided in this paper. Code availability All code developed and used for this study is available on GitHub: https://github.com/TheJacksonLaboratory/endometriosis-visium-hd. (https://doi.org/10.5281/zenodo.21826389) 63.

1. Chapron, C., Marcellin, L., Borghese, B. & Santulli, P. Rethinking mechanisms, diagnosis and management of endometriosis.Nat. Rev. Endocrinol.15,6 6 6–682 https://doi.org/10.1038/s41574-019- 0245-z (2019). 2. Missmer, S. A. et al. Impact of Endometriosis on Life-Course Potential: A Narrative Review.Int. J. Gen. Med. 14,9 –25 https://doi. org/10.2147/IJGM.S261139(2021). 3. Harder, C., Velho, R. V., Brandes, I., Sehouli, J. & Mechsner, S. Assessing the true prevalence of endometriosis: A narrative review of literature data.Int. J. Gynecol. Obstet.167, 883–900 https://doi. org/10.1002/ijgo.15756(2024). 4 . G o d i n ,S .K . ,W a g n e r ,J . ,H u a n g ,P .&B r e e ,D .T h er o l eo fp e r i p h e r a l nerve signaling in endometriosis.FASEB BioAdvances3,8 0 2–813 https://doi.org/10.1096/fba.2021-00063(2021). 5. Morotti, M., Vincent, K., Brawn, J., Zondervan, K. T. & Becker, C. M. Peripheral changes in endometriosis-associated pain.Hum. Reprod. Update 20,7 1 7–736 https://doi.org/10.1093/humupd/ dmu021 (2014). 6. Velho, R. V., Sehouli, J. & Mechsner, S. Mechanisms of peripheral sensitization in endometriosis patients with peritoneal lesions and acyclical pain. Arch. Gynecol. Obstet.308 1327–1340, https://doi. org/10.1007/s00404-023-07110-9(2023). 7 . A s - S a n i e ,S .e ta l .E n d o m e t r i o s i s :Ar e v i e w .Jama 334,6 4–78 https:// doi.org/10.1001/jama.2025.2975(2025). 8. Sasamoto, N. et al. Trends in pelvic pain symptoms over 2 years of follow-up among adolescents and young adults with and without endometriosis.Pain 164,6 1 3–624 https://doi.org/10.1097/j.pain. 0000000000002747(2023). 9. Ortega-Gutiérrez, M., Muñoz-Gamez, A. & Girón-Prieto, M. S. Pri- mary care approach to endometriosis: diagnostic challenges and management strategies—a narrative review. J. Clin. Med. 14, 4757 (2025). 1 0 . T a n ,Y .e ta l .S i n g l e - c e l la n a l y s is of endometriosis reveals a coor- dinated transcriptional programme driving immunotolerance and angiogenesis across eutopic and ectopic tissues.Nat. Cell Biol. 24, 1306–1318 https://doi.org/10.1038/s41556-022-00961-5(2022). 11. Fonseca, M. A. S. et al. Single-cell transcriptomic analysis of endometriosis.Nat. Genet. 55,2 5 5–267 https://doi.org/10.1038/ s41588-022-01254-1(2023). 12. Liu, S. et al. Single-cell and spatial transcriptomic profiling revealed niche interactions sustaining growth of endometriotic lesions.Cell Genomics 5,1 0 0 7 3 7https://doi.org/10.1016/j.xgen.2024. 100737 (2025). 13. Clower, L., Fleshman, T., Geldenhuys, W. J. & Santanam, N. Tar- geting oxidative stress involved in endometriosis and its pain.Bio- molecules 12, https://doi.org/10.3390/biom12081055 (2022). 14. Garcia Garcia, J. M. et al. Endometriosis: cellular and molecular mechanisms leading tofibrosis.Reprod. Sci.30,1 4 5 3–1461 https:// doi.org/10.1007/s43032-022-01083-x(2023). 15. Bacci, M. et al. Macrophages are alternatively activated in patients with endometriosis and requiredfor growth and vascularization of lesions in a mouse model of disease. Am. J. Pathol. 175,5 4 7–556 https://doi.org/10.2353/ajpath.2009.081011(2009). 16. Capobianco, A. & Rovere-Querini, P. Endometriosis, a disease of the macrophage.Front Immunol. 4,9 https://doi.org/10.3389/fimmu. 2013.00009(2013). 1 7 . F a t t o r i ,V .e ta l .N o c i c e p t o r-to-macrophage communication through CGRP/RAMP1 signaling drives endometriosis-associated p a i na n dl e s i o ng r o w t hi nm i c e .Sci. Transl. Med 16,e a d k 8 2 3 0 https://doi.org/10.1126/scitranslmed.adk8230(2024). 18. Forster, R. et al. Macrophage-derived insulin-like growth factor-1 is a key neurotrophic and nerve-sensitizing factor in pain associated with endometriosis.Faseb J. 33, 11210–11222 https://doi.org/10. 1096/fj.201900797R(2019). 19. Greaves, E. et al. Estradiol is a critical mediator of macrophage- nerve cross talk in peritoneal endometriosis.Am. J. Pathol. 185, 2286–2297 https://doi.org/10.1016/j.ajpath.2015.04.012(2015). Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press 20. Scheerer, C., Frangini, S., Chi antera, V. & Mechsner, S. Reduced Sympathetic Innervation in Endometriosis is Associated to Sema- phorin 3C and 3F Expression.Mol. Neurobiol.54,5 1 3 1–5141 https:// doi.org/10.1007/s12035-016-0058-1(2017). 21. Zhu, H., Wang, Y., He, Y. & Yu, W. In flammation-mediated macro- phage polarization induces TRPV1/TRPA1 heteromers in endome- triosis. Am. J. Transl. Res. 14,3 0 6 6–3078 (2022). 22. Barcena, M. L. et al. In fluence of Nerve Growth Factor in Endometriosis-Associated Symptoms.Reprod. Sci. 18,1 2 0 2–1210 https://doi.org/10.1177/1933719111410711(2011). 23. Yan, D., Liu, X. & Guo, S. W. Neuropeptides Substance P and Cal- citonin Gene Related Peptide Accelerate the Development and Fibrogenesis of Endometriosis.Sci. Rep.9, 2698 https://doi.org/10. 1038/s41598-019-39170-w(2019). 24. Vercellini, P. et al. Association between endometriosis stage, lesion type, patient characteristics andseverity of pelvic pain symptoms: a multivariate analysis of over 1000 patients.Hum. Reprod. 22, 266–271 https://doi.org/10.1093/humrep/del339(2006). 25. Imperiale, L., Nisolle, M., Noël, J. C. & Fastrez, M. Three types of endometriosis: pathogenesis, diagnosis and treatment. State of the Art. J. Clin. Med. 12, https://doi.org/10.3390/jcm12030994 (2023). 26. Andersson, A. et al. Points2Region s: Fast, interactive clustering of imaging-based spatial transcriptomics data.Cytom. A105,6 7 7–687 https://doi.org/10.1002/cyto.a.24884(2024). 27. Rahmioglu, N. et al. The genetic basis of endometriosis and comorbidity with other pain and inflammatory conditions.Nat. Genet. 55,4 2 3–436 https://doi.org/10.1038/s41588-023-01323- z (2023). 28. Chishima, F. et al. Peritoneal and p eripheral B-1-cell populations in patients with endometriosis.J. Obstet. Gynaecol. Res. 26,1 4 1–149 https://doi.org/10.1111/j.1447-0756.2000.tb01298.x(2000). 29. Riccio, L. G. C., Baracat, E. C., Chapron, C., Batteux, F. & Abrão, M. S. The role of the B lymphocytes in endometriosis: A systematic review. J. Reprod. Immunol.123,2 9–34 https://doi.org/10.1016/j.jri. 2017.09.001 (2017). 30. Riccio, L. G. C. et al. B lymphocyt es inactivation by Ibrutinib limits endometriosis progression in mice.Hum. Reprod. 34, 1225–1234 https://doi.org/10.1093/humrep/dez071(2019). 31. Gagné, D. et al. Blood leukocyte s ubsets are modulated in patients with endometriosis.Fertil. Steril.80,4 3–53 https://doi.org/10.1016/ s0015-0282(03)00552-1(2003). 32. Anaf, V. et al. Pain, mast cells, and nerves in peritoneal, ovarian, and deep infiltrating endometriosis.Fertil. Steril.86,1 3 3 6–1343 https:// doi.org/10.1016/j.fertnstert.2006.03.057(2006). 33. McCallion, A. et al. Estrogen mediates in flammatory role of mast cells in endometriosis pathophysiology.Front. Immunol.13, 961599 https://doi.org/10.3389/fimmu.2022.961599(2022). 34. Henlon, Y. et al. Single-cell analysis identi fies distinct macrophage phenotypes associated with prodisease and proresolving functions in the endometriotic niche.Proc. Natl. Acad. Sci. USA 121, e2405474121https://doi.org/10.1073/pnas.2405474121(2024). 35. Bohonyi, N. et al. Local upregulation of transient receptor potential ankyrin 1 and transient receptor potential vanilloid 1 ion channels in rectosigmoid deep infiltrating endometriosis.Mol. Pain. 13, 1744806917705564https://doi.org/10.1177/ 1744806917705564(2017). 36. Tokushige, N., Markham, R., Russell, P. & Fraser, I. S. Nerve fibres in peritoneal endometriosis.Hum. Reprod.21,3 0 0 1–3007 https://doi. org/10.1093/humrep/del260(2006). 37. Tokushige, N. et al. Nerve fibers in ovarian endometriomas.Fertil. Steril. 94,1 9 4 4–1947 https://doi.org/10.1016/j.fertnstert.2009.12. 074 (2010). 38. Yan, D., Liu, X. & Guo, S.-W. Nerve fibers and endometriotic lesions: partners in crime in inflicting pains in women with endometriosis. Eur. J. Obstet. Gynecol. Reprod. Biol.209,1 4– 24 https://doi.org/10. 1016/j.ejogrb.2016.06.017(2017). 39. Moon, K. M. et al. Proprioception, the regulator of motor function. BMB Rep. 54,3 9 3–402 https://doi.org/10.5483/BMBRep.2021.54. 8.052 (2021). 40. Morotti, M., Vincent, K. & Becker, C. M. Mechanisms of pain in endometriosis.E u r .J .O b s t e t .G y n e c o l .R e p r o d .B i o l .209,8 –13 https://doi.org/10.1016/j.ejogrb.2016.07.497(2017). 41. Arnold, J. et al. Imbalance between sympathetic and sensory innervation in peritoneal endometriosis.Brain, Behav., Immun.26, 132–141 https://doi.org/10.1016/j.bbi.2011.08.004(2012). 42. Lu, T. et al. Decoding transcrip tional identity in developing human sensory neurons and organoid modeling.Cell 187,7 3 7 4–7393 https://doi.org/10.1016/j.cell.2024.10.023(2024). 43. Rigamonti, A. et al. Large-Scale Production of Mature Neurons from Human Pluripotent Stem Cells in a Three-Dimensional Suspension Culture System. Stem Cell Rep. 6,9 9 3–1008 https://doi.org/10. 1016/j.stemcr.2016.05.010(2016). 44. Wagner, M. et al. Single-cell analysis of human ovarian cortex identifies distinct cell populations but no oogonial stem cells.Nat. Commun. 11, 1147 https://doi.org/10.1038/s41467-020-14936- 3 (2020). 45. Moolhuijsen, L. M. E. & Visser, J . A. Anti-Müllerian Hormone and Ovarian Reserve: Update on Assessing Ovarian Function.J. Clin. Endocrinol. Metab.105, 3361–3373 https://doi.org/10.1210/clinem/ dgaa513 (2020). 4 6 . T r e m e l l e n ,K .P . ,K o l o ,M . ,G i l m o r e ,A .&L e k a m g e ,D .N .A n t i - mullerian hormone as a marker of ovarian reserve.A u s t .N .Z .J . Obstet. Gynaecol.45,2 0–24 https://doi.org/10.1111/j.1479-828X. 2005.00332.x(2005). 47. Osuga, Y. et al. Lymphocytes in Endometriosis. Am. J. Reprod. Immunol.65,1 –10 https://doi.org/10.1111/j.1600-0897.2010.00887. x (2011). 48. Shen, M. et al. The role of endometrial B cells in normal endome- trium and benign female reproductive pathologies: a systematic review. Hum. Reprod. Open 2022,h o a b 0 4 3https://doi.org/10. 1093/hropen/hoab043(2022). 49. Lee, H.-J., Ahn, B. J., Shin, M. W., Choi, J.-H. & Kim, K.-W. Ninjurin1: a potential adhesion molecule and its role in inflammation and tissue remodeling.Mol. Cells 29, 223–228 https://doi.org/10.1007/ s10059-010-0043-x(2010). 50. Greaves, E. et al. Estrogen rece ptor (ER) agonists differentially regulate neuroangiogenesis in peritoneal endometriosis via the repellent factor SLIT3.Endocrinology155,4 0 1 5–4026 https://doi. org/10.1210/en.2014-1086(2014). 51. Garrido-Trigo, A. et al. Macrophage and neutrophil heterogeneity at single-cell spatial resolution in human inflammatory bowel disease. Nat. Commun. 14,4 5 0 6https://doi.org/10.1038/s41467-023- 40156-6 (2023). 52. Khaliq, A. M. et al. Spatial transcriptomic analysis of primary and metastatic pancreatic cancers highlights tumor microenviron- mental heterogeneity.Nat. Genet 56,2 4 5 5–2465 https://doi.org/ 10.1038/s41588-024-01914-4(2024). 53. Marco Salas, S. et al. Optimizing Xenium In Situ data utility by quality assessment and best-practice analysis workflows. Nat.

22,8 1 3–823 https://doi.org/10.1038/s41592-025-02617- 2 (2025). 5 4 . H i e ,B . ,C h o ,H . ,D e M e o ,B . ,B r y s o n ,B .&B e r g e r ,B .G e o m e t r i c Sketching Compactly Summarizes the Single-Cell Transcriptomic Landscape.Cell Syst. 8,4 8 3–493 https://doi.org/10.1016/j.cels. 2019.05.003(2019). 55. Cang, Z. et al. Screening cell –cell communication in spatial tran- scriptomics via collective optimal transport.Nat. Methods 20, 218–228 https://doi.org/10.1038/s41592-022-01728-4(2023). Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press 56. Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12,1 0 8 8https://doi.org/10.1038/s41467- 021-21246-9(2021). 57. Efremova, M., Vento-Tormo, M., Teichmann, S. A. & Vento-Tormo, R. CellPhoneDB: inferring cell–cell communication from combined expression of multi-subunit ligand–receptor complexes.Nat. Pro- toc. 15,1 4 8 4–1506 https://doi.org/10.1038/s41596-020-0292- x (2020). 58. Boretto, M. et al. Patient-derived organoids from endometrial dis- ease capture clinical heterogeneity and are amenable to drug screening.Nat. Cell Biol. 21,1 0 4 1–1051 https://doi.org/10.1038/ s41556-019-0360-z(2019). 59. Bozal, S. B. et al. Development of an automated 3D high content cell screening platform for organoid phenotyping.SLAS Discov. 29, https://doi.org/10.1016/j.slasd.2024.100182 (2024). 60. Chambers, S. M. et al. Combined small-molecule inhibition accel- erates developmental timing and converts human pluripotent stem cells into nociceptors.Nat. Biotechnol.30,7 1 5–720 https://doi.org/ 10.1038/nbt.2249(2012). 61. Hiranuma, M., Okuda, Y., Fujii, Y., Richard, J.-P. & Watanabe, T. Characterization of human iPSC-derived sensory neurons and their functional assessment using multi electrode array.Sci. Rep. 14, 6011 https://doi.org/10.1038/s41598-024-55602-8(2024). 62. Ewels, P. A. et al. The nf-core framework for community-curated bioinformatics pipelines.Nat. Biotechnol.38,2 7 6–278 https://doi. org/10.1038/s41587-020-0439-x(2020). 63. Haney, C., Alizadeh, E., White, B.& Courtois, E. High-Resolution Spatial Transcriptomics Reveals Fibroblast and Neuroimmune Microenvironments in Endometriosis Lesions Code.Zenodohttps:// doi.org/10.5281/zenodo.21826389 (2026).

The authors would first like to thank all of the study participants that generously donated their tissues tohelp us better understand the biol- ogy of endometriosis. Without their participation, none of this would be possible. We heartfully thank the EndoRISE Biorepository for enabling part of this research through access to biospecimens, and in particular Kayceety Mullaj and Makayla Murphy for their support. Next, we would like to thank all of The Jackson Laboratory core facilities that supported this work and in particular: Genome Technologies, Histology, Micro- scopy and Cyberinfrastructure High-performance computing resources. Of particular importance, we gratefully acknowledge the advanced expertise of the Single Cell Biology Lab and their continuous support. These shared services are supported in part by the JAX Cancer Center (P30 CA034196). We thank the OB/Gyn department and UConn Health Surgery Center personnel for their assistance in consenting patients and collecting specimens. A special thank you to the Robson and Skarnes Labs at the Jackson Laboratory for the use of iPSC cells, especially Juliana Alcoforado Diniz, Enricca Pellegrino and Cole Loring for their guidance and Adrian Melo Carrilo for his help with the bulk-RNA-seq analysis. This study makes use of data and/or resources generated by The Jackson Laboratory’s Genetic Diversity (GeDi) hiPSC initiative and made available through internal data access and sharing mechanisms. Data from the following GeDi samples were used: GeDiX697.3 J and GeDiX653.1 J. Author contributions E.T.C conceived and designed the study. C.M.H. and E.T.C performed and supervised experiments. D.E.L collected consent from participants and clinical samples. C.M.H., J.L., J.K, and E.T.C. processed and biobanked samples for downstream experiments. J.K. provided EndoR- ISE biospecimen and de-identified patient information. C.M.H., J.L., and M.S. performed IHC experiments and imaging. C.M.H. conceived and performed in vitro experiments and bulk-RNA seq. C.M.H and W.F.F performed bulk-RNA-seq analysis. P.R. provided knowledge and resources for iPSC brain organoid cultures. C.M.H, E.A., and B.S.W. performed computational data analysis for visium HD. B.S.W, W.F.F., and E.T.C. provided crucial data interpretation. C.M.H. and E.T.C. wrote the manuscript and generatedfigures and schematics. All authors reviewed a n de d i t e dt h em a n u s c r i p t . Funding This study was supported by The Mayday Fund and the State of Con- necticut, through HB6672, which funds the CT Endometriosis Tissue and Data Biorepository, which is part of the EndoRISE program. The iPSC line used in this study was derived with support from The Jackson Labora- tory, Director’s Innovation Fund, award number 19000-22-12. Competing interests The authors declare no competing interests. Additional information Supplementary informationThe online version contains supplementary material available at https://doi.org/10.1038/s41467-026-78453-5 . Correspondenceand requests for materials should be addressed to Elise T. Courtois. Peer review informationNature Communicationsthanks Carlos Simon and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Reprints and permissions informationis available at http://www.nature.com/reprints Publisher’s note Springer Nature remains neutral with regard to jur- isdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted

derived from this article or parts of it. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicatedotherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by-nc-nd/4.0/ . © The Author(s) 2026 This Article in Press is shared early to give you faster access to new research. It is citable and carries a permanent DOI. The final edited version will replace it automatically. Article https://doi.org/10.1038/s41467-026-78453-5 Nature Communications Article in Press