Single-cell sequencing uncovers disrupted stromal-macrophage communication as a driver of intrauterine adhesion progression.

This study complied with all relevant ethical regulations and received approval from the Ethics Committee of First People’s Hospital of Yunnan Province, affiliated with Kunming University of Science and Technology (ethics number: KHLL2021-KY049) and has been performed in accordance with the principles of the Declaration of Helsinki. All patients were fully informed about the use of their samples in this study and provided written informed consent. All ethical regulations relevant to human research participants were followed. The endometrium undergoes cyclic changes influenced by estrogen and progesterone, including the proliferative, secretory, and menstrual phases. Since the secretory phase is predominantly regulated by progesterone, IUA may alter progesterone secretion levels, leading to variable endometrial changes. This variability could obscure the inherent molecular alterations associated with IUA when compared to normal endometrium. Therefore, we selected samples from the proliferative phase, which is primarily driven by estrogen and is expected to exhibit relatively stable molecular characteristics, ensuring a more accurate comparison of disease-specific changes in IUA. The information of all endometrial samples is shown in supplementary Table  1 . Fresh human endometrial samples were immediately immersed in ice-cold Dulbecco’s Phosphate-Buffered Saline (DPBS; Gibco, 14190144) supplemented with 1% fetal bovine serum (FBS; Gibco 10099-141) and kept on ice during transportation to preserve cell viability. A two-step enzymatic digestion protocol was employed to isolate single cells enriched for either stromal fibroblasts or epithelial components. Upon arrival, tissues were rinsed 2–3 times with PBS, dissected in a sterile 3.5 cm Petri dish to remove blood and mucus, and minced into small fragments. These were transferred to 10 mL of digestion buffer comprising 1 mg/mL Collagenase Type IV (Gibco, 17104-019), 2.5 IU/mL Dispase (Corning, 354235) and 100 ug/mL DNase I (Sigma-Aldrich, 11284932001) dissolved in DMEM. Samples were incubated at 37 °C under gentle agitation (50 r.p.m) for 20–30 min. This step preferentially released stromal fibroblasts into suspension, while epithelial glands remained largely intact. The resulting cell mixture was passed through a 40-μm nylon cell strainer (BD Falcon, 352340) with stromal cells collected in the filtrate and epithelial structures retained on the mesh. The retained epithelial fragments were recovered by backwashing with DMEM and subjected to further enzymatic digestion using 400 μLTrypLE Select (Thermo Fisher Scientific) for 20 min at 37 °C, with occasional pipetting to to promote dissociation into single cells. Stromal suspensions were treated with red blood cell (RBC) lysis buffer (Invitrogen, 00-4333-57), and all resulting cell suspensions were filtered once more through a 40-μm strainer and centrifuged at 1,000 r.p.m. for 5 min. Cells were then washed with PBS containing 0.04% Bovine Serum Albumin (BSA; Sigma-Aldrich, B2064), pelleted by centrifugation at 500 × g for 5 min, and resuspended in PBS with 0.04% BSA at a final concentration of ~1 × 10⁶ cells/mL. Viability was assessed using Trypan blue staining (Invitrogen, T10282 ), and viable single-cell suspensions were maintained on ice for subsequent single-cell RNA sequencing. Single-cell suspensions were cprocessed using the Chromium Single Cell 3’ Library & Gel Bead Kit v3 (10x Genomics) following the manufacturer’s protocol. This included encapsulation of single cells into Gel Beads-in-Emulsion (GEMs), cell barcoding, reverse transcription, amplification of full-length cDNA, and subsequent library construction. All steps-from GEM generation to indexing and amplification-were performed according to the standard procedures provided by 10x Genomics. Final libraries were subjected to high-throughput paired-end sequencing with a read length of 150 base pairs, using the Illumina HiSeq X Ten or NovaSeq 6000 platforms (PE150 mode), depending on availability. Raw sequencing data were aligned to the GRCh38 human reference genome using Cell Ranger (v5.0.1),generating gene-cell count matrices. Subsequent data processing and analysis were performed using the Seurat R package (v4.3.0.1). Unless otherwise specified, default parameters were applied for all functions.To correct for potential batch effects across different samples, integration was performed using FindIntegrationAnchors() followed by IntegrateData(). After integration, data normalization and scaling were conducted through NormalizeData() and ScaleData() functions, respectively. Quality control criteria were applied to exclude cells with fewer than 200 or more than 7500 detected genes, or those with mitochondrial gene expression exceeding 20%, as these were indicative of low-quality or dying cells. To identify and eliminate potential doublets, the DoubletFinder package (v2.0.4) was employed. The curated and normalized dataset was then subjected to dimensionality reduction and clustering analyses to delineate distinct cell populations. Dimensionality reduction was initially performed using principal component analysis (PCA), applying default settings. The top 30 principal components were selected for downstream clustering based on a shared nearest neighbor (SNN) modularity optimization algorithm via the FindClusters function (resolution = 0.6). To visualize the distribution of cell populations in reduced dimensional space, uniform manifold approximation and projection (UMAP) was employed. Differentially expressed genes (DEGs) for each cluster were identified using the FindAllMarkers function based on the Wilcoxon rank-sum test, using the following thresholds: |avg_log2FC | > 0.25, adjusted P -value  0.25. Cell type identities were subsequently assigned by integrating canonical marker gene expression and DEG profiles, visualized through feature and violin plots. The CellChat analysis was conducted with reference to previously published studies 34 ; specifically, the analysis was performed as follows:To comprehensively examine intercellular communication between macrophage subsets and distinct stromal cell populations, we applied the CellChat R package ( https://github.com/sqjin/CellChat ), which leverages a curated database of ligand-receptor interactions and associated cofactors. Separate analyses were first performed on control, adjacency, and adhesion samples to infer group-specific communication networks.Subsequently, these networks were integrated using joint manifold learning to evaluate the functional similarity and divergence of signaling patterns across conditions. For each signaling pathway, the overall information flow-quantified as the sum of communication probabilities between all interacting cell pairs-was computed and systematically compared between groups. Differentially expressed genes (DEGs) were analyzed using the Seurat FindMarkers function based on the Wilcoxon rank sum test. Gens with an average∣log2FC∣> 0.25 and adjusted p   <  0.05 were considered DEGs. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were performed with the R package “clusterProfiler(v3.18.1)”. Significantly enriched terms were visualized using bubble plots generated with the ggplot2 package. For scRNA-seq data, we inferred PROGENy pathway activity using the human version of PROGENy (version 1.12.0), a comprehensive resource that includes a curated collection of pathways and their core target genes with weights for each interaction. For our analysis, we utilized the human weights and selected the top 500 responsive genes, ranked by p-value and original pathway scores scaled to their respective null distribution, to calculate normalized pathway activity scores. To infer transcription factor (TF) activity from our scRNA-seq data, we utilized the human version of DoRothEA (version 1.2.2), a comprehensive gene sets resource of TF’s targets (regulons), in conjunction with VIPER [version 1.24.0]. This integration enhances the accuracy of TF activity predictions by incorporating the regulatory modes of each TF-target interaction, as highlighted by a recent benchmark study 48 , 49 . Results were visualized via heatmaps generated with the pheatmap package (version 1.0.12). Based on 48 M2 genes reported by Liu et al. 50 the “AddModuleScore” (Seurat, version 4.3.0) function was employed to calculate the M2 signature score of all macrophage clusters. By estimating M2 meta-signature scores between different groups based on the nonparametric Wilcoxon rank sum test and determining the average expression level of the 48 genes for each cell, we obtained the M2 signature score.M2 scores were visualized using violin plots generated with the ggplot2 package (version 3.5.1). Trajectory analysis was performed via the R package Monocle (version 2.18.0) to explore the dynamic process of monocyte-to-macrophage differentiation. The top highly variable genes (HVGs) were selected as differential gene intersections from single-cell sequencing. The “estimateSizeFactors” and “estimateDispersions” functions were used to build statistical models to characterize the data. Additionally, the “reduceDimension” function was applied to reduce dimensions, and the “orderCells” function was used to place cells onto a pseudotime trajectory. Based on the pseudo-time analysis, branch expression analysis modeling (BEAM) was conducted to analyze genes associated with branch fate determination. To assess the most likely trajectories of cell progression of stromal cells, seven stromal subpopulations were then subjected to PAGA and RNA velocity analysis to ascertain the most likely intersubcluster trajectories. The edge confidences between each subcluster node for all edges is visualized using the python package scvelo(version 0.2.4). The expression data was subsetted into different cell types and stages, and the mean expression of cell type markers was calculated to represent the expression level of its corresponding cell type. Finally, the Spearman correlation test was carried out between Control, Adjacency and Adhesion groups.Results were visualized using heatmaps produced by the pheatmap package (version 1.0.12). To spatially map cell types defined by scRNA-seq analysis within the publicly Visium spatial transcriptomics data, we used cell2location 51 (version 0.1.4), a Bayesian framework for mapping single-cell transcriptomic data to spatial transcriptomics data. All analyses were conducted using default parameters in Python version 3.12. To further investigate potential batch effects across samples, we performed data integration and visualization using scvi-tools (version 1.3.0). Specifically, we applied the scVI integration method to re-integrate our single-cell transcriptomic data, grouping the data by sample to account for potential batch effects. The resulting low-dimensional coordinates were extracted and visualized using the ggplot2 (version 3.5.1) package in R. All analyses were conducted with default parameters in Python (version 3.12). This apprach allowed us to assess and mitigate batch effects while preserving the biological variability within the dataset. Endometrium tissue samples were fixed in 4% paraformaldehyde and embedded in paraffin. After antigen retrieval to activate endogenous peroxidase, the sections were incubated in 3% hydrogen peroxide for 30 min. The samples were then blocked with 10% normal goat serum for 1 h. The expression of COL1A1(1:400,ab138492, abcam),α-SMA(1:400,ab124964,Abcam),CD68(1:200,ab955,Abcam),S100A8(1:200,DF6556,Affinity),CCL2(1:200,10134-T26,Sino-biological),CCL5(1:200,DF7427, Affinity) were determined by incubating the sections with the primary antibody at 4 °C overnight. The slides were subsequently incubated with the secondary antibody dilution and then treated with DAB solution for 1 h. Hematoxylin solution was briefly applied for 15 seconds to stain the nuclei. Finally, the slides were examined under a microscope. The endometrium tissues were fixed in 4% paraformaldehyde for 24 h and subsequently processed for paraffin embedding. The paraffin-embedded sections were sliced into 4-μm sections and stained with Masson’s trichrome using standard protocols provided by Servicebio. The images of the Masson-stained tissues were analyzed via ImageJ software to calculate the area of fibrosis. Total RNA was extracted from the endometrium stomal cells stimulating by SPP1(120-35-100UG, peprotech) or CCL5(C062), novoprotein) for 48 h via the MiniBEST Universal RNA Extraction Kit (TaKaRa, Cat No. 9767) following the manufacturer’s instructions. The PrimeScript™ RT reagent kit, gDNA Eraser (Perfect Real Time) (TaKaRa, Cat No. RR047), was used to synthesize cDNA, which was then subjected to amplification via real-time fluorescence PCR. The fold change in the mRNA levels was determined via the 2 −ΔΔCt method. COL1A1 primer was: F: GAGGGCCAAGACGAAGACATC; R: CAGATCACGTCATCGCACAAC. ACTA2 gene primer was: F: GTGTTGCCCCTGAAGAGCAT; R: GCTGGGACATTGAAAGTCTCA. To further investigate the correspondence between our defined cell types and the HECA labels, we performed label transfer and comparative analysis using the Seurat package (v4.4.0). Specifically, we utilized the FindTransferAnchors() and MapQuery() functions to map cell type annotations between the reference (HECA Mareckov et al. 30 ). To visually validate the similarity, we generated an interactive Sankey diagram using the networkD3 package (v0.4.1), which quantitatively illustrates the robust overlap between our cell type definitions and the HECA classification. Statistical analysis was conducted using GraphPad Prism 9.0 software. Data are presented as means ± SD, with error bars indicating SD values. Various statistical tests, including One-way ANOVA or two-tailed Student’s t test. The significance levels were denoted as follows: p   >  0.05 indicates no significance (n.s.), * p   <  0.05, ** p   <  0.01, and *** p   <  0.001. Representative data were obtained from at least three different samples. Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

To characterize endometrial fibrosis resulting from IUA, we performed scRNA-seq on endometrial biopsies from nine individuals (Fig.  1a ). All samples were in the proliferative phase of the menstrual cycle without exogenous hormonal treatment (Fig.  1b , Supplementary Fig.  1a ). Healthy endometrium (Ctrl) refers to samples from individuals without IUA. Samples from individuals with mild IUA (MiIUA), moderate IUA (MoIUA), and tissue adjacent to adhesion (Mi/oIUA-a) were collected from seven individuals, with adhesion score varied from 5 to 17, based on the diagnostic criteria of the Chinese Medical Society of Obstetrics and Gynecology (Fig.  1c , Supplementary Table  1 ). Fig. 1 An overview of the single cell landscape of healthy controls and IUA patients. a Flowchart illustrating the origins of the scRNA-seq samples and the analysis workflow. b The bubble diagram shows the expression of classical stromal marker genes defining each subphase of menstrual cycle in the collected endometrial samples ( n  = 7). c Diagram showing scRNA-seq metrics per patient (left) and tissue type (right) after QC. These metrics indicate unique molecular identifier (UMIs) and total genes per cells across tissue types. The cord diagram (center) indicates the representation of each patient (miIUA: mild intrauterine adhesion, moIUA: moderate intrauterine adhesion, Ctrl control. a adjacent, l lesions) in each tissue type. d UMAP plot showing the 108,497 single-cells from control and IUA tissues. Four major cell types are identified. e UMAP plot in Fig. 1d is separated into five groups: controls ( n   =  2, providing 22936 cells), mi-IUA-a cases ( n   =  2, providing 32041 cells), mi-IUA-l cases ( n   =  2, providing 12147 cells), mo-IUA-a cases ( n   =  2, providing 27348 cells), mo-IUA-l cases ( n   =  2, providing 2037 cells). f Violin plot showing the expression of classical marker genes for the four major cell types-stromal, immune, epithelial, and endothelial identified in the scRNA-seq dataset from 9 endometrial samples. g Diagram showing the proportion of cell types in each tissue, grouped by ctrl, adjacency and adhesion. A one-way ANOVA was used to compare cell type proportion across the three groups and the P value between ctrl group and adjacency group is 0.032. * p  < 0.05. h Correlation matrix plot showing Pearson’s correlation coefficient between different endometrial pathology stages (adjacency and adhesion) and normal sample. a Flowchart illustrating the origins of the scRNA-seq samples and the analysis workflow. b The bubble diagram shows the expression of classical stromal marker genes defining each subphase of menstrual cycle in the collected endometrial samples ( n  = 7). c Diagram showing scRNA-seq metrics per patient (left) and tissue type (right) after QC. These metrics indicate unique molecular identifier (UMIs) and total genes per cells across tissue types. The cord diagram (center) indicates the representation of each patient (miIUA: mild intrauterine adhesion, moIUA: moderate intrauterine adhesion, Ctrl control. a adjacent, l lesions) in each tissue type. d UMAP plot showing the 108,497 single-cells from control and IUA tissues. Four major cell types are identified. e UMAP plot in Fig. 1d is separated into five groups: controls ( n   =  2, providing 22936 cells), mi-IUA-a cases ( n   =  2, providing 32041 cells), mi-IUA-l cases ( n   =  2, providing 12147 cells), mo-IUA-a cases ( n   =  2, providing 27348 cells), mo-IUA-l cases ( n   =  2, providing 2037 cells). f Violin plot showing the expression of classical marker genes for the four major cell types-stromal, immune, epithelial, and endothelial identified in the scRNA-seq dataset from 9 endometrial samples. g Diagram showing the proportion of cell types in each tissue, grouped by ctrl, adjacency and adhesion. A one-way ANOVA was used to compare cell type proportion across the three groups and the P value between ctrl group and adjacency group is 0.032. * p  < 0.05. h Correlation matrix plot showing Pearson’s correlation coefficient between different endometrial pathology stages (adjacency and adhesion) and normal sample. A total of 139,395 single cells were obtained, of which 96509 passed stringent quality control filters for subsequent analysis (Supplementary Fig.  1b ). To further mitigate batch effects introduced by sampling, we applied dimensionality reduction on each sample using scVI 27 and Integrate 28 , both yielding well-defined cell clusters with minimal batch bias (Supplementary Fig.  1c ). In these tissues, a median of 1864 gene per cell could be detected (Fig.  1c ). Unsupervised graph clustering partitions the cells into clusters and visualized the clusters via uniform manifold approximation and projection (UMAP). These clusters were assigned to four major cell types: epithelial, stromal, endothelial, and immune cells (Fig.  1d, e ). Epithelial cells were identified by EPCAM expression, stromal cells by COL1A1 and PDGFRA , endothelial cells by PECAM1 and VWF , and immune cells by PTPRC and CD3G (Fig.  1f ). The frequency of immune cell subsets appears to increase with IUA severity (Fig.  1g , Supplementary Fig.  1d ), suggesting a potential role in disease progression, though further study is needed. Meanwhile, the decline in stromal cells may reflect shifts in cellular composition influenced by sampling depth or endometrial region. Further investigation of cell population similarities revealed significant differences in epithelial cells from adjacent and adhesion sites compared to those in normal tissues (Fig.  1h ). Together, these findings underscore regional heterogeneity in the endometrium and offer a basis for further investigation into IUA pathophysiology. Given that stromal cells are the most abundant cell type in the endometrial microenvironment and play a crucial role in fibrosis pathogenesis, we first explored the stromal cell subsets and their functions in IUA progression. Initially, we attempted to map our data to stromal subpopulations defined in previously published studies by Tan et al. 22 , Aloson et al. 29 , Santamaria et al. 20 , Marečková et al. 30 , the distribution of cells among many clusters confirms the inherent heterogeneity of the sampled tissue. Nevertheless, smaller subclusters of projected cells generally aligned with the manual classifications of major cell groups (i.e., predominantly stromal cells), (Supplementary Fig.  2a–d ). Then, the HECA datasets were incorporated into our analysis. The Sankey plot further confirmed the consistency between our manual annotation and major cell lineages from the reference dataset (Supplementary Fig.  3a, b ). Based on these results, we performed unsupervised graph-based clustering, which identified eight distinct clusters visualized using the UMAP algorithm (Supplementary Fig.  4a ). To simplify our message and enhance the relevance of our comparisons, we focused on clusters 0-6, which represented the majority of the mesenchymal population (Fig.  2a , Supplementary Fig.  4b ), while clusters 7 contained fewer than 80 cells (Supplementary Fig.  4a ). Each stromal subpopulation exhibited a distinct gene expression profile, as shown in the heatmap (Supplementary Fig.  4c ). Furthermore, gene ontology (GO) enrichment analysis revealed functional heterogeneity among the stromal subsets (Supplementary Fig.  4d ). Finally, we identified seven subsets of stromal cells, all of which were present in both Ctrl and IUA clusters (Fig.  2b, c , Supplementary Fig.  5a, b ). We confirmed that these stromal clusters were differentially distributed at various stages of disease progression (Fig.  2d ). Nearly all cells expressed VIM (vimentin) and PDGFRB , confirming their mesenchymal origin (Fig.  2e ). Specific pericyte markers, RGS5 (Regulator of G protein signaling 5) and NOTCH3 were predominantly detected in clusters 5. The pan-fibroblast marker genes PDGFRA and REV3L were expressed across all clusters, with the highest expression observed in cluster 1 (Fig.  2e ). Compared with healthy controls, IUA tissues exhibited a higher proportion of cluster 3 (Fig.  2d ). Fig. 2 Identification of distinct mesenchymal clusters and changes in transcriptional profile during IUA. a UMAP of scRNA-seq data from 24,453 mesenchymal cells across 8 patients (2 controls and 6 IUA samples) showing the first 7 clusters (clusters 0 to 6). Colors indicate clusters defined by a graph-based clustering method. b Same UMAP as in ( a ) showing cell distribution across control samples (C1, C2) and IUA patient samples (P1 to P6). c Same UMAP as in ( a ) showing cell distribution according to disease status (Ctrl in pink for patients without IUA; adjacent tissues in green and adhesion tissues in blue from patients with IUA). d Bar plot showing the percentages of different stromal clusters according to disease status: Ctrl, Adjacency and Adhesion ( N   =  2, 4 and 2, respectively). A one-way ANOVA was used to compare cell type proportion across the three groups and the P -value between ctrl group and adjacency group is 0.033. * p  < 0.05. e UMAP (top) and Violin plot (bottom) showing expression of marker genes according to the different clusters for mesenchymal cells ( PDGFRB , VIM ), pericytes ( RGS5 , NOTCH3 ) and fibroblasts ( REV3L , PDGFRA ). f UpSet Plot showing the integrated comparative analysis of upregulated DEGs in major stromal cell types between Ctrl and Adhesion groups. Upregulated DEGs: upregulated in Adhesion, downregulated in Ctrl group. Distribution and comparison of the fibrosis score in stromal cell populations across three types of endometrial samples-Ctrl, adjacency, and Adhesion. g Representative GO terms and pathways enriched in upregulated DEGs based on functional enrichment analysis in stromal cells. P value was derived by a hypergeometric test. h A heatmap showing the Q values of 18 metabolic pathways across 7 cell types of adhesion compared to ctrl. The lower Q value is depicted in blue, the higher Q value in red. i Z scores of ligand-receptor products averaged across groups. Ligand expression is averaged across all cell types; receptors are in fibroblasts only. The receptor-ligand interactions shown were deemed significantly different between the three groups by a permutation test. a UMAP of scRNA-seq data from 24,453 mesenchymal cells across 8 patients (2 controls and 6 IUA samples) showing the first 7 clusters (clusters 0 to 6). Colors indicate clusters defined by a graph-based clustering method. b Same UMAP as in ( a ) showing cell distribution across control samples (C1, C2) and IUA patient samples (P1 to P6). c Same UMAP as in ( a ) showing cell distribution according to disease status (Ctrl in pink for patients without IUA; adjacent tissues in green and adhesion tissues in blue from patients with IUA). d Bar plot showing the percentages of different stromal clusters according to disease status: Ctrl, Adjacency and Adhesion ( N   =  2, 4 and 2, respectively). A one-way ANOVA was used to compare cell type proportion across the three groups and the P -value between ctrl group and adjacency group is 0.033. * p  < 0.05. e UMAP (top) and Violin plot (bottom) showing expression of marker genes according to the different clusters for mesenchymal cells ( PDGFRB , VIM ), pericytes ( RGS5 , NOTCH3 ) and fibroblasts ( REV3L , PDGFRA ). f UpSet Plot showing the integrated comparative analysis of upregulated DEGs in major stromal cell types between Ctrl and Adhesion groups. Upregulated DEGs: upregulated in Adhesion, downregulated in Ctrl group. Distribution and comparison of the fibrosis score in stromal cell populations across three types of endometrial samples-Ctrl, adjacency, and Adhesion. g Representative GO terms and pathways enriched in upregulated DEGs based on functional enrichment analysis in stromal cells. P value was derived by a hypergeometric test. h A heatmap showing the Q values of 18 metabolic pathways across 7 cell types of adhesion compared to ctrl. The lower Q value is depicted in blue, the higher Q value in red. i Z scores of ligand-receptor products averaged across groups. Ligand expression is averaged across all cell types; receptors are in fibroblasts only. The receptor-ligand interactions shown were deemed significantly different between the three groups by a permutation test. Functional enrichment analysis revealed that cluster 3 was mainly related to chromosome segregation suggesting a profibrotic cluster may originated from this proliferating subpopulation (Supplementary Fig.  4d ). We then conducted trajectory analysis using RNA velocity and PAGA to investigate the differentiation trajectories of proliferating stromal cells in IUA. This analysis indicated a branched structure, starting with proliferating cell and bifurcating into multiple stromal states (Supplementary Fig.  5c, d ). The proportion of cluster 1 also increased with the progression of IUA, although this increase was not statistically significant (Fig.  2d ). To identify the genes involved fibrosis, we focused on the upregulated genes in stromal cell subpopulations between control and adhesion groups. The Upset plot indicated that cluster 0, 2, 3, 4 and 6 were strongly associated with fibrosis, as they exhibited the highest number of upregulated genes. Notably, 40 up-regulated genes were shared by all stromal subpopulations (Fig.  2f ). The degree of endometrial fibrosis at different sites was evaluated using the selected 40 genes, revealing that fibrosis was more pronounced at the adhesion site compared to adjacent areas (Fig.  2f ). To further elucidate the biological functions associated with the upregulated DEGs, we performed GO analysis (Fig.  2g ). The results indicated significant enrichment in various biological processes, including wound healing, TGF-beta signaling pathway, Hippo signaling pathway, focal adhesion, extracellular matrix organization, and collagen fibril organization. These processes are crucial for tissue repair and fibrosis development, suggesting that the upregulated DEGs play a pivotal role in the fibrosis pathogenesis of IUA. To explore metabolic activity of stromal cells in ctrl and adhesion sites, we analyzed the gene expression of 18 metabolic pathways. Among the seven stromal subtypes, cluster 3 displayed the largest proportions of statistically significant differentially expressed genes (DEGs) calculated to the total number of genes in each pathway in the comparison of ctrl vs. adhesion (Supplementary Fig.  6a ). In cluster 3, Oxidative phosphorylation (53/158,33.54%), fatty acid elongation (8/27,29.63%), glutathione metabolism (12/53,22.64%), cysteine and methionine metabolism (8/37, 21.62%), were among the top altered pathways. We then applied single-cell pathway analysis (SCPA) across all stromal cell types, identifying oxidative phosphorylation, HIF-1, and insulin signaling as the top three pathways with Q values of 7.0, 6.2, and 5.7, respectively (Supplementary Fig.  6b ). Cluster 0 showed the highest Q values for most of the pathways compared to other cell types (Fig.  2h ), indicating high pathway differences between ctrl and adhesion site. Additionally, we performed the analysis of transcription factors (TF) associated with metabolic genes. Among the top 25 TFs (Supplementary Fig.  6c ), the transcription factor FOXO1 plays a role in endometrial remodeling during the menstrual cycle, regulating epithelial cell polarity and progesterone receptor expression 31 , 32 ; however, its expression was downregulated at adhesion sites. Similarly, HIF-1α, essential for normal endometrial repair 33 , was reduced following IUA formation. Cell-to-cell communication analysis revealed that adhesion sites exhibited the highest communication scores for both pro-inflammatory and anti-inflammatory signals, as well as elevated extracellular matrix (ECM) communication scores. Notably, TGF-β signaling, a master regulator of fibrosis, exhibited the highest activity at adhesion sites (Fig.  2i ). To evaluate the contribution of immune cells to the progression of endometrial fibrosis, we sub-cluster 20,827 immune cells into 16 subpopulations using the harmony algorithm (Supplementary Fig.  7a, b ). We filtered out 3298 cells, including cluster 4 and 5, which expressed stromal cell markers, and cluster 12 that could not be accurately defined (Supplementary Fig.  7c ). Ultimately, we defined six major immune cells: macrophages ( CD14 , CSF1R ), mast cells ( MS4A2 , HPGD ), T cells ( CD3E , CD3D ), NK cells ( GZMB , NCAM1 ), B cells ( CD79A , MS4A1 ), and proliferating T cells ( MKI67 ) (Supplementary Fig.  7d, e ). Expression heatmaps of the top 20 signature genes for each defined cell type reflected the specificity of distinct cell populations (Supplementary Fig.  7f ). Numerous studies have reported that macrophages are closely associated with fibrosis progression. Therefore, we further analyzed macrophages heterogeneity and developmental trajectories during endometrial fibrosis progression. Based on previously published literature, we categorized 15 macrophage subpopulations into 6 types according to the expression of MRC1 , CD163 , CLEC10A , S100A8 , GATA6 , CD24 , and MKI67 (Supplementary Fig.  8a, b ). We then identified six main macrophage subsets: CD163 + , CD301 + , CD24 + , Monocyte, proliferating, and an undefined population (unknown) (Fig.  3a, b ). Interestingly, we found that the percentage of the 6 subsets of human endometrial macrophage cells showed different changes with the different degree of fibrosis (Fig.  3c) . Among these, CD163 + was the predominant population in the endometrium during the proliferative phases. Although we observed a decreasing trend in CD163⁺cells and an increase in monocytes at adhesion sites, these changes were not statistically significant. In contrast, proliferating macrophages showed a significant increase in the adjacency group compared to the ctrl group, suggesting a localized activation and expansion of macrophage populations in para-fibrotic tissues (Fig.  3c ). Fig. 3 Heterogeneity of macrophages and developmental changes in macrophages during disease progression. a Bubble diagram showing the expression of marker genes for each subpopulation of macrophages ( n  = 9 samples). b The UMAP plot showing subtypes of macrophages from controls and patients with IUA ( n   =  9 samples). c The percentage of each macrophage subtype among the three groups: Ctrl Adjacency, and Adhesion ( n   =  2, 4 and 3, respectively). A one-way ANOVA was used to compare cell type proportion across the three groups and the P -value between ctrl group and adjacency group is 0.044. * p  < 0.05. d Pseudo-time analysis showing the distribution of each macrophage subset along the developmental trajectory. Each point corresponds to a single cell, and each color represents a macrophage population as indicated. e Potential polarization trajectory of different macrophage subsets inferred by Monocle 2, showing two branches of monocyte polarizing to CD24 + or CD163 + subsets. f Pseudotemporal analysis showed the distribution of macrophages in different disease states. g Differentially expressed genes along the pseudotime were hierarchically clustered into four profiles, with gene expression in a branch-dependent manner. Each row indicated the standardized kinetic curves of a gene(left). The kinetic curves for the four clustered genes are displayed in the middle. Representative enriched GO terms and pathways are shown(right). h Expression patterns of representative genes along the reprogramming trajectory. a Bubble diagram showing the expression of marker genes for each subpopulation of macrophages ( n  = 9 samples). b The UMAP plot showing subtypes of macrophages from controls and patients with IUA ( n   =  9 samples). c The percentage of each macrophage subtype among the three groups: Ctrl Adjacency, and Adhesion ( n   =  2, 4 and 3, respectively). A one-way ANOVA was used to compare cell type proportion across the three groups and the P -value between ctrl group and adjacency group is 0.044. * p  < 0.05. d Pseudo-time analysis showing the distribution of each macrophage subset along the developmental trajectory. Each point corresponds to a single cell, and each color represents a macrophage population as indicated. e Potential polarization trajectory of different macrophage subsets inferred by Monocle 2, showing two branches of monocyte polarizing to CD24 + or CD163 + subsets. f Pseudotemporal analysis showed the distribution of macrophages in different disease states. g Differentially expressed genes along the pseudotime were hierarchically clustered into four profiles, with gene expression in a branch-dependent manner. Each row indicated the standardized kinetic curves of a gene(left). The kinetic curves for the four clustered genes are displayed in the middle. Representative enriched GO terms and pathways are shown(right). h Expression patterns of representative genes along the reprogramming trajectory. To understand the role of monocyte-derived macrophages in the endometrial fibrosis progression, we performed trajectory analysis using monocle to investigate macrophages polarization in IUA. This analysis indicated a branched structure originating from monocytes and CD301 + macrophages and diverging into multiple polarization states (Fig.  3d ). As illustrated in Fig.  3d, e , monocyte, CD301 + and proliferating macrophage were primarily distributed along the early to intermediate stages of the trajectory, while CD24 + and CD163 + macrophages were predominantly located at the terminal branches. This distribution suggests that monocytes progressively polarize into distinct macrophage subtypes during fibrosis. Furthermore, group-specific pseudotime mapping confirmed this trend, showing increased polarization as endometrial fibrosis advanced (Fig.  3f ). We further examined transcriptional and functional changes across pseudotime. The expression of genes associated with the cell cycle, organelle fission and DNA replication genes (e.g., CENPF , and MKI67 ) peaked during proliferating macrophage differentiation, but exhibited a decreasing trend toward terminal polarization. Conversely, genes (e.g, ITGA4 and BTG1 ) involved in integrin-mediated signaling, IFN-γ mediated signaling pathway and positive regulation of vasculature development were markedly downregulated from the monocyte stage to other macrophage branches in the trajectory. Notably, gene involved in in extracellular matrix organization (e.g, IGF1 , COL1A1 ), TGF-β receptor signaling pathway, and collagen fibril organization were downregulated in the early stage but upregulated in the late stage (Fig.  3g, h ). In contrast, pathways related to negative regulation of protein polymerization, actin filament organization and neutrophil activation involved in immune response were highly enriched during the middle and late stage of macrophage polarization (Fig.  3g, h ). Together, these data reveal that monocytes are specific cell subsets that emerge during the progression of endometrial fibrosis. Notably, these monocytes exhibit high expression of S100A8 , a factor implicated in driving fibrosis in endometrial stromal cells (EnSCs) through the RAGE-JAK2-STAT3 pathway 15 . Therefore, targeting S100A8 -expressing monocytes could offer a potential therapeutic approach for treating endometrial fibrosis. Building on our previous pseudo-temporal analysis of macrophages in endometrial fibrosis, we hypothesized that specific macrophage subpopulations contribute to the acceleration of ECM deposition. We first conducted PROGENy pathway analysis, which showed that monocyte subsets exhibited high expression of pro-inflammatory signaling pathways, including NF-κB, JAK-STAT, and TNF-α. In contrast, pro-fibrotic signaling pathways, such as TGF and WNT, were most active in both monocytes and CD24 + subsets, aligning with the pseudo-time trajectory analysis results (Fig.  4a ). To further characterize the functional attributes of endometrial macrophage subpopulations, we calculated M2 polarization scores using macrophage polarization-related gene sets and found that the CD163 + macrophages, the most abundant in the endometrium, closely resembled the M2 macrophage phenotype (Fig.  4b ). The role of M2 macrophages in tissue fibrosis remains controversial. Therefore, we specifically analyzed the upregulated genes in CD163 + macrophages and identified 283 differentially expressed genes in macrophages in adjacency group and 1029 genes specifically upregulated in adhesion group (Fig.  4c ). KEGG pathway analysis revealed that these upregulated genes were predominantly enriched in pro-inflammatory responses, including TNF-α, IL-17 and NF-κB signaling (Fig.  4d ). Fig. 4 Molecular profiling and intercellular communication of profibrotic macrophages in intrauterine adhesion. a Heatmap of the PROGENy-predicted pathway activity scores of each macrophage clusters. b Violin plots indicating relative expression levels of M2 signatures scores across the macrophage clusters. c Venn diagram showing the number of shared and specific upregulated genes between adjacency versus control and adhesion versus control in CD163 + macrophage. d Enriched KEGG pathways of shared and specific upregulated genes in ( c ) are shown. KEGG, Kyoto Encyclopedia of Genes and Genomes. e Violin plots revealing the expression of two pro-fibrosis representative genes (CCL2, CCL5) across the three groups: control, adjacency, and adhesion. f Expression of cell-type-specific ligand-receptor interactions inferred by cell-chat. Shown are predicted interactions between macrophages, epithelium, endometrium stromal cells, endothelium, NK cells and T cells at different disease stages. Circle size indicates the proportion of ligand-receptor expression in interacting cells and circle colour indicates mean expression of ligand and receptor genes for interacting pairs. a Heatmap of the PROGENy-predicted pathway activity scores of each macrophage clusters. b Violin plots indicating relative expression levels of M2 signatures scores across the macrophage clusters. c Venn diagram showing the number of shared and specific upregulated genes between adjacency versus control and adhesion versus control in CD163 + macrophage. d Enriched KEGG pathways of shared and specific upregulated genes in ( c ) are shown. KEGG, Kyoto Encyclopedia of Genes and Genomes. e Violin plots revealing the expression of two pro-fibrosis representative genes (CCL2, CCL5) across the three groups: control, adjacency, and adhesion. f Expression of cell-type-specific ligand-receptor interactions inferred by cell-chat. Shown are predicted interactions between macrophages, epithelium, endometrium stromal cells, endothelium, NK cells and T cells at different disease stages. Circle size indicates the proportion of ligand-receptor expression in interacting cells and circle colour indicates mean expression of ligand and receptor genes for interacting pairs. In conclusion, these results suggest that although CD163 + macrophages in the endometrium largely exhibit an M2-like type, their function gradually shifts towards a pro-inflammatory state during fibrosis. This transition is accompanied by a reduction in their number, a finding corroborated by previous studies 34 . Although macrophages can differentiate into myofibroblasts and contribute to tissue fibrosis through ECM secretion, our study specifically examined the cytokines they produce. Our findings revealed that as fibrosis progression, the endometrium secretes high amount of CCL2 and CCL5 —two factors known to promote fibrosis in other tissues (Fig.  4e ). To further explore the functional interactions between macrophages and other cell types, we classified macrophages as ligand-expressing cells and other cell types as receptor-expressing cells. The analysis indicated that macrophages at adhesion sites secreted significantly higher levels of the pro-fibrotic factor SPP1 compared to those in normal and para-adhesion regions. These factors were found to bind to the receptors ITGAV , ITGA9 , ITGA8 , ITGA5 , ITGA4 , and CD44 on epithelial, stromal, vascular, macrophage, NK, and T cells, respectively, aligning with previous findings (Fig.  4f ). Significant alterations were also observed in other cytokine and chemokine interactions, including CXCL8 , CCL20 , and IL1B , which were all expressed by macrophages. Additionally, the secretion of EREG and NAMPT by macrophages was notably elevated at adhesion sites, suggesting their potential role in promoting endometrial fibrosis (Fig.  4f ). Collectively, these data highlight the pivotal regulatory role of macrophages in modulating other cell types and suggest that CCL2 , CCL5 , CXCL8 , and IL1B could serve as potential therapeutic targets for endometrial fibrosis. Ligand-receptor-mediated cell-cell interactions play a pivotal role in regulating cellular functions. The progression of IUA is closely linked to the activities of ECM-secreting mesenchymal cells and their interactions with other cell types. We initially visualized the incoming and outgoing interactions of all cell types at both para-adhesion and adhesion sites. The direction and length of each arrow represent the predominance and strength of IUA-driven incoming or outgoing interactions, respectively (Fig.  5a ). Endometrial macrophages at both the para-adhesion and adhesion site exhibited the most pronounced increases in incoming and outgoing interactions, indicating their critical role in IUA-associated signaling. In contrast, EnSCs and NK cells displayed reduced interaction levels (Fig.  5a ). Fig. 5 Altered intercellular signaling landscape reveals fibrotic pathways in intrauterine adhesion. a Arrow plot showing changes in outgoing and incoming interaction strength between adjacency (point of the arrow) and control conditions (base of the arrow) for specific cell types(left) and between adhesion (point of the arrow) and control conditions (base of the arrow) in the endometrium (right). b Heatmap showing the differential interaction strength among stromal and macrophage cell types in the endometrium under adjacency (left) and adhesion (right) conditions. Red and blue shading indicate increased or decreased signaling, respectively, compared to control. c Circle plots showing the top 25% of increased (red) or decreased (blue) signaling interactions in the endometrium for specific pathways in adjacency (left) and adhesion (right) compared to controls. d Bar plot illustrating significant signaling pathways in Adhesion, Adjacency, and control groups. The overall flow of information in a signaling network is accomplished by aggregating all potential communication probabilities within the network. e Circle plots representing the top 25% of endometrium cell-cell communications inferred for the TGF-β pathways in the control, adjacency and adhesion groups. f Circle plots showing the top 25% of cell-cell communications inferred for the SPP1 and complement pathways among stromal subtypes and macrophage subtypes in the endometrium specific for adhesion groups. Each node represents a cell type, and the interaction is shown by color-coded lines. a Arrow plot showing changes in outgoing and incoming interaction strength between adjacency (point of the arrow) and control conditions (base of the arrow) for specific cell types(left) and between adhesion (point of the arrow) and control conditions (base of the arrow) in the endometrium (right). b Heatmap showing the differential interaction strength among stromal and macrophage cell types in the endometrium under adjacency (left) and adhesion (right) conditions. Red and blue shading indicate increased or decreased signaling, respectively, compared to control. c Circle plots showing the top 25% of increased (red) or decreased (blue) signaling interactions in the endometrium for specific pathways in adjacency (left) and adhesion (right) compared to controls. d Bar plot illustrating significant signaling pathways in Adhesion, Adjacency, and control groups. The overall flow of information in a signaling network is accomplished by aggregating all potential communication probabilities within the network. e Circle plots representing the top 25% of endometrium cell-cell communications inferred for the TGF-β pathways in the control, adjacency and adhesion groups. f Circle plots showing the top 25% of cell-cell communications inferred for the SPP1 and complement pathways among stromal subtypes and macrophage subtypes in the endometrium specific for adhesion groups. Each node represents a cell type, and the interaction is shown by color-coded lines. Given that stromal cells and macrophages are the most affected by IUA, we conducted a detailed analysis of communication dynamics between stromal and macrophage subsets at para-adhesion and adhesion sites. We assessed the interaction strength between cell type pairs, visualized by a heatmap (Fig.  5b ) and a network plot (Fig.  5c ). At adhesion site, interactions between macrophage subsets and stromal subsets showed that S0 and S1 subsets had the strongest interactions with CD163 + , CD301 + , and monocyte (Fig.  5b, c ). Interestingly, communication involving the stromal cell subset S0 was significantly reduced at both para-adhesion and adhesion sites (Fig.  5b, c ). In summary, fibrotic endometrium is characterized by reduced communication between stromal cell subsets, particularly within S0 subsets, and increased communication among macrophage subsets and between macrophage subsets and stromal subsets. Given the substantial changes in macrophage and stromal cell signaling in IUA, we then examined specific signaling pathways contribution to these cell interactions to identify those highly upregulated in IUA (Fig.  5d ). The TGF-β signaling pathway was substantially involved in interactions between CD301 + and proliferating macrophage subsets and stromal subsets in both control and adjacent groups (Fig.  5e ), potentially indicating its role in ECM accumulation in fibrosis-associated endometrium. In the intrauterine adhesion group, interactions between macrophages and stromal cells were largely lost, except for the S6 subset, while TGF-β signaling between monocytes and CD163 + macrophages, as well as within CD163 + macrophages, was significantly enhanced (Fig.  5e ). The SPP1 signaling pathway was predominantly driven by CD163 + and proliferating macrophages, with numerous interactions directed toward stromal subset only in the adhesion group (Fig.  5f ). The complement signaling pathway was predominantly mediated by CD163 + macrophages, which exhibited outgoing signals to CD301 + , monocyte, and proliferating macrophage subsets (Fig.  5f ). Together, these analyses reveal that endometrial fibrosis is characterized by macrophage-stromal signaling involving pro-fibrotic and inflammatory pathways, orchestrated through cell-cell crosstalk between macrophages and non-immune cell types. To validate the expression of macrophages derived cytokines identified through single-cell sequencing during the pathological progression of IUA in humans, we compared the expression levels of S100A8, CCL2, CCL5 and CD68 in normal endometrium and IUA samples using immumohistochemical staining. The results demonstrated that the expressions of these cytokines were significantly higher in the endometrium of patients with IUA than in normal endometrial tissue (Fig.  6a–d , Supplementary Fig.  9a, b ). Furthermore, this increased cytokine expression was accompanied by elevated levels of COL1A1 and α-SMA, key markers of fibrosis in IUA (Fig.  7a–c , Supplementary Fig.  9a, b ). To investigate the interaction between macrophage-derived factors and stromal cells, in vitro stimulation of endometrial stromal cells with SPP1 and CCL5 significantly increased COL1A1 expression, but not ACTA2 , suggesting a partial induction of myofibroblast differentiation (Fig.  7d, e ). Taken together, our results suggest that macrophage-derived cytokines, particularly CCL5 and SPP1, contribute to myofibroblast differentiation and thereby facilitate the progression of IUA. These findings underscore the critical role of macrophage-stromal cell interactions in the pathogenesis of endometrial fibrosis. Fig. 6 Pathological features of macrophages in the IUA tissue. a –d Immunohistochemical staining showing positive expression of CD68 ( a ), CCL2 ( b ), CCL5 ( c ) and S100A8 ( d ) in normal and IUA endometrium ( n  = 4 per group). (low magnification, scale bar, 500 μm; high magnification, scale bar, 100 μm). The values are the means ± SDs. * p   <  0.05, ** p   <  0.01, *** p   <  0.001, **** p   <  0.0001.Statistical significance was determined using unpaired two-tailed t-tests. Fig. 7 Macrophage-derived factors promote endometrial fibrosis. a Masson’s trichrome staining was used to analyze changes in fibrosis in the different groups. b , c Immunohistochemical staining of COL1A1( b ) and-SMA ( c ) in normal and IUA endometrium ( n  = 4 per group). (low magnification, scale bar, 500 μm; high magnification, scale bar, 100 μm). d , e Myofibroblast differentiation on endometrium fibroblasts following SPP1, CCL5 stimulation. Primary endometrium fibroblasts were isolated, plated, and subsequently stimulated with SPP1 (100 ng/ml) ( d ) and CCL5 (100 ng/ml) ( e ), 48 h later, RNA was isolated, and transcripts were determined for COL1A1, ACTA2. The values are the means ± SDs. * p   <  0.05, ** p   <  0.01, *** p   <  0.001, **** p   <  0.0001.Statistical significance was determined using unpaired two-tailed t-tests. a –d Immunohistochemical staining showing positive expression of CD68 ( a ), CCL2 ( b ), CCL5 ( c ) and S100A8 ( d ) in normal and IUA endometrium ( n  = 4 per group). (low magnification, scale bar, 500 μm; high magnification, scale bar, 100 μm). The values are the means ± SDs. * p   <  0.05, ** p   <  0.01, *** p   <  0.001, **** p   <  0.0001.Statistical significance was determined using unpaired two-tailed t-tests. a Masson’s trichrome staining was used to analyze changes in fibrosis in the different groups. b , c Immunohistochemical staining of COL1A1( b ) and-SMA ( c ) in normal and IUA endometrium ( n  = 4 per group). (low magnification, scale bar, 500 μm; high magnification, scale bar, 100 μm). d , e Myofibroblast differentiation on endometrium fibroblasts following SPP1, CCL5 stimulation. Primary endometrium fibroblasts were isolated, plated, and subsequently stimulated with SPP1 (100 ng/ml) ( d ) and CCL5 (100 ng/ml) ( e ), 48 h later, RNA was isolated, and transcripts were determined for COL1A1, ACTA2. The values are the means ± SDs. * p   <  0.05, ** p   <  0.01, *** p   <  0.001, **** p   <  0.0001.Statistical significance was determined using unpaired two-tailed t-tests. To comprehensively characterize the cellular composition and dynamics of IUA throughout the menstrual cycle, we analyzed approximately 240,000 high-quality cells from 42 individuals (Supplementary Fig.  10a ). First, we integrated three publicly available scRNA-seq datasets (Wang et al. 8 , Santamaria et al. 20 , Lv et al. 34 ) with our own dataset. Given the significant impact of hormonal fluctuations on gene expression, we separately examined the transcriptional profiles of IUA in the proliferative and secretory phases (Fig.  8a–d ). Following data integration, we identified nine major cell types in both phases: stromal cells, endothelial cells, non-ciliated epithelial cells, ciliated epithelial cells, B cells, mast cells, T cells, NK cells, and macrophages (Fig.  8a ). Fig. 8 Single-cell transcriptomic analysis reveals phase-specific epithelial and stromal dynamics across the menstrual cycle and in IUA. a UMAP projections of scRNA-seq data from a total of 42 individuals. b UMAP representations colored by menstrual phase. c UMAP projections of scRNA-seq data from a total of 17 individuals in the proliferative phase. d UMAP projections of scRNA-seq data from a total of 25 individuals in the secretory phase. e , f GO and KEGG enrichment analysis of up-regulated and down-regulated DEGs in epithelial cells ( e ), stromal cells ( f ). g Trajectory reconstruction of epithelial cells from both ctrl and IUA groups, along with the distribution of clusters corresponding to different phases, is presented. Additionally, the gene expression dynamics of representative implantation window markers are displayed. a UMAP projections of scRNA-seq data from a total of 42 individuals. b UMAP representations colored by menstrual phase. c UMAP projections of scRNA-seq data from a total of 17 individuals in the proliferative phase. d UMAP projections of scRNA-seq data from a total of 25 individuals in the secretory phase. e , f GO and KEGG enrichment analysis of up-regulated and down-regulated DEGs in epithelial cells ( e ), stromal cells ( f ). g Trajectory reconstruction of epithelial cells from both ctrl and IUA groups, along with the distribution of clusters corresponding to different phases, is presented. Additionally, the gene expression dynamics of representative implantation window markers are displayed. Differential gene expression analysis of epithelial cells in both the proliferative and secretory phase identified 997 and 1,732 DEGs, respectively. Among these, 1570 genes were upregulated, while 1159 were downregulated (Supplementary Fig.  10b ). The heatmap illustrates the overlapping upregulated and downregulated genes among the top 100 differentially expressed genes in the proliferative and pre-secretory phases. (Supplementary Fig.  10c ). GO and KEGG analysis revealed that the upregulated genes were primarily associated with the estrogen signaling pathway and interferon-gamma mediated signaling pathway, whereas the downregulated genes were enriched in the IL-17 and Hippo signaling pathway (Fig.  8e ). Additionally, GO and KEGG analyses of stromal cell upregulated genes in both the proliferative and secretory phase showed significant enrichment in pathways related to focal adhesion and the PI3K-Akt signaling pathway, while the downregulated genes were enriched in insulin-like growth factor II binding (Fig.  8f ). Notably, the transcriptional phenotypes of stromal cells in IUA varied considerably between the proliferative and secretory phases. To further investigate the developmental trajectories of IUA, we applied the Monocle 2 algorithm to normal and IUA samples across both menstrual phases (Supplementary Fig.  10d ). Pseudotime analysis revealed a typical progression of normal epithelial cells from the proliferative to the secretory phase. However, in IUA samples, the secretory phase lagged behind the expected proliferative trajectory (Fig.  8g ). This finding suggests that IUA disrupts the normal transition from the proliferative to the secretory phase, potentially hindering the opening of the epithelial implantation window. To investigate this further, we analyzed the dynamic expression of key implantation window markers along the pseudo-time trajectory. Our results revealed abnormal expression patterns of ESR1 , FOXO1 , PAEP , and SPP1 in IUA samples (Fig.  8g ). Furthermore, we identified genes differentially expressed genes associated with cell fate that exhibited progressive alterations in the two branches compared to the pre-branch in both the ctrl and IUA groups (Supplementary Fig.  10e, f ). Collectively, these results provide a comprehensive transcriptional landscape of IUA across different menstrual phases. The endometrial epithelium comprises a complex basal glandular network, functional glands extending into the uterine cavity, a layer of luminal epithelial cells, and various immune cells. To comprehensively understand the diversity and dynamics of macrophages and stromal cells during endometrial regeneration and differentiation throughout the natural menstrual cycle, we mapped the spatial distribution of the identified cell types in our scRNA-seq data using a published spatial transcriptomic dataset of normal endometrium from the proliferative and secretory phases of two individuals (Alonso et al.). We analyzed two sections of A30 tissue from the secretory phase (152811 and 1528070) and two sections of A13 tissue from the proliferative phase (152810 and 152806), each annotated for the endometrial and myometrial layers of the uterus (Fig.  9a ). Both macrophage and stromal cell subpopulations were localized within the endometrial layer (Fig.  9b, c ). During the transition from the proliferative to the secretory phase, stromal cell subpopulations 0, 1, 2, 4, and 6 migrated from the basal layer to the functional layer, whereas stromal cell subpopulations 3 and 5 remained evenly distributed throughout the endometrium (Fig.  9b ). Macrophage subpopulations exhibited distinct spatial distributions: CD24⁺, monocyte, and CD301⁺ macrophages were predominantly localized near the basal layer, CD163⁺ macrophages were concentrated near the functional layer, and proliferating macrophages were dispersed throughout the endometrium (Fig.  9c ). These findings suggest that macrophage-stromal interactions are spatially dynamic and highly overlapping, highlighting their potential role in endometrial remodeling. Fig. 9 Spatial transcriptomics reveals regional distribution of stromal and macrophage subpopulations in human endometrium. a Haematoxylin and eosin staining of the slides in the Visium arrays from Alonso et al. Two individuals were selected: proliferative phase A13 and secretory phase A30. b , c Estimated amount of mRNA (color intensity) contributed by each stromal cell population ( b ), macrophage subsets ( c ) to each spot (color) shown over the H&E image of proliferative (A13, 152810 slide and 152806 slide) and secretory (A30, 152811 slide and152807 slide) endometrium. a Haematoxylin and eosin staining of the slides in the Visium arrays from Alonso et al. Two individuals were selected: proliferative phase A13 and secretory phase A30. b , c Estimated amount of mRNA (color intensity) contributed by each stromal cell population ( b ), macrophage subsets ( c ) to each spot (color) shown over the H&E image of proliferative (A13, 152810 slide and 152806 slide) and secretory (A30, 152811 slide and152807 slide) endometrium.

In this study, we investigated alterations in macrophages and stromal cells during endometrial fibrosis in IUA. Using single-cell sequencing, we compared cellular and molecular changes in IUA, adjacent endometrium, and normal endometrium. We identified a gene set that can assess fibrosis score in stromal cells and revealed several potential therapeutic targets for endometrial fibrosis, including chemokines CCL2 and CCL5, SPP1, and complement signaling pathway from a macrophage perspective. Fibrosis, marked by excessive ECM deposition driven by COL1A1 and α-SMA-expressing myofibroblasts, is the pathological hallmark of IUA 35 , 36 . Single-cell sequencing effectively resolves cellular heterogeneity. Both Lv, et al. 34 and Santamaria et al. 20 identified myofibroblasts in IUA using this approach. However, our stromal analysis did not acapture myofibroblasts, despite their confirmed presence in fibrotic tissue via immunohistochemistry. This likely reflects their low survival during single-cell suspensions preparation from fibrotic tissue, limiting exploration of their origin and features. Future studies on stromal cell heterogeneity, function, and spatial organization will be crucial for advancing endometrial research. Macrophages, as highly plastic immune cells, are central to organ fibrosis. Upon injury, they migrate to the damaged site to support healing or drive fibrosis 37 . Our data provide several key insights: first, monocytes accumulate at the adhesion site; second, under physiological conditions, macrophages and stromal cells maintain robust communication via TGF-β signaling, which diminishes with unresolved inflammation and fibrosis; third, at adhesion site, macrophages-derived TGF-β mainly acts in an autocrine manner. Notably, we observe macrophage-to-myofibroblast transition (MMT), as CD24 + macrophages co-express CD68 and ACTA2 , despite unchanged abundance compared to normal tissues. These findings highlight the importance of restoring macrophage-stromal communication to repair endometrial function. The endometrium is a highly dynamic tissue that undergoes cyclical shedding and regeneration during the menstrual cycle, with macrophages playing distinct roles in repair and remodeling. In murine models, monocytes rapidly infiltrate the endometrium following injury and differentiate into tissue-resident macrophages to promote repair 38 . Numerous studies have demonstrated that macrophage-stromal cell interactions are critical for myofibroblast formation. For example, a single-cell atlas of liver fibrosis identified TREM2 + macrophages as promoters of myofibroblast differentiation, while Chakarov et al. demonstrated that depletion of Lyve1 hi MHCII lo macrophages exacerbates lung injury 39 , 40 . Similarly, in models of abdominal adhesions, a common post-surgical complication, macrophages depletion exacerbated adhesion severity 41 . These findings highlight the functional diversity of macrophages across and within organs, and underscore the unique heterogeneity of endometrial macrophages. Lv et al. identified CD301 + macrophages as key regulators of fibroblast differentiation and survival, with GAS6 secreted by these cells activating AXL receptors on stromal cells to promote myofibroblast formation 34 . CD163 + macrophages, the most abundant subtype in the endometrium, exhibit an anti-inflammatory M2-like phenotype. However, their role in endometrial fibrosis-whether protective or pathogenic, remains unclear. Understanding their function could provide valuable insight into the regenerative capacity of the endometrium. Using pseudo-time analysis, we found that monocytes differentiate into two distinct macrophage lineages within both para-fibrotic and fibrotic regions of the endometrium: M2-like CD163 + macrophages and myofibroblast-like CD24 + macrophages.Platelets are the initial responders to tissue injury and form aggregates with monocytes, which shape macrophage polarization via the CXCL4-SPP1 axis to drive fibrosis 42 . Additionally, a stiffened ECM can mechanically activate monocytes, promoting their polarization into fibrosis-inducing macrophages 43 . These insights position macrophages as promising therapeutic targets for fibrosis 44 . Furthermore, based on our single-cell RNA-seq data, we identified ligand–receptor interactions between macrophages and stromal cells—particularly involving SPP1, CCL2, and CCL5—that could, in combination with traditional estrogen therapy, inform future clinical management strategies for IUA. This study has several limitations. First, it is challenging to digest the fibrotic endometrial tissue into single cells. The number of truly fibrotic cells, particularly myofibroblast obtained in this study was relatively limited and requires improvement. Additionally, the subpopulations of macrophages and stromal cells identified in this study were preliminarily validated in the published spatial transcriptome data of normal endometrium, and the factors derived from macrophages, such as SPP1 and CCL5, were preliminarily confirmed to promote the formation of myofibroblasts through in vitro experiments. More research efforts should focus on in vivo animal experiments to verify the interaction between stromal cells and macrophages in the future. Collectively, through unbiased single-cell RNA-seq analysis, we delineate the cellular landscape of IUA during the proliferation phase, highlighting key populations such as macrophages and stromal cells. We focused on molecular changes in these populations and revealed that their interaction are crucial to disease initiation and progression. Given that the endometrium is hormonally regulated by estrogen and progesterone, we integrated single-cell data from both the proliferative and secretory phases in normal and IUA samples,identifying signaling pathways consistently altered across both phases. Notably, We observed excessive activation of estrogen signaling in epithelial cells, suggesting a potential role in IUA pathogenesis. Pseudotime analysis of epithelial cells further indicated that IUA may disrupt the opening of the endometrial implantation window, potentially contributing to infertility. At the tissue level, we observed upregulation of CD68, S100A8, CCL2, and CCL5 in IUA samples compared to normal endometrium. In vitro experiments confirmed that macrophage-derived factors promoted stromal-to-myofibroblast transition. These results underscore the pivotal role of macrophage-stromal interactions in IUA pathogenesis. Furthermore, we identified key transcription factors-FOSL2, FOXO1, and HIF1A-involved in regulating endometrial fibrosis in stromal cells. Notably, FOSL2 has been implicated in renal and cardiac fibrosis in previous studies 45 , 46 . From the macrophages perspective, we identified key targets driving endometrial fibrosis, such as SPP1, CCL2, and CCL5. Importantly, our findings are in line with recent insights emphasizing that dysregulated immune-hormonal crosstalk underlies fibrotic endometrial pathologies, as reviewed by Murphy et al. 47 . These findings provide a framework for developing therapeutic strategies aimed at modulating macrophage-stromal cell interactions to mitigate IUA.

IUA is a common gynecological condition that significantly affects the physical and mental health of women of childbearing age worldwide 1 . It usually arises from infection or trauma to the endometrial basal layer, causing symptoms such as secondary amenorrhea, pain, and infertility 2 . Although first described by Joseph Asherman In 1948, its specific etiology and molecular drivers remain largely unknown 3 . Current treatments include hysteroscopic transcervical resection of adhesion (TCRA) and hormonal therapy. vasoactive agents and non-degradable stents are also used postoperatively 4 – 6 . However, recurrence and preserving fertility remain major challenges due to limited understanding of the mechanisms behind impaired endometrial repair. Unlike other tissues, the endometrium is highly dynamic and regenerative, undergoing repeated cycle of proliferation, differentiation, and shedding throughout reproductive life to support embryo implantation 7 , 8 . It consists of epithelial, stromal, vascular, and immune cells, which coordinate transcriptional programs to achieve scarless healing 9 , 10 . This repair process is tightly regulated by a balance of pro- and anti-fibrotic cytokines under physiological condition 11 . Endometrial fibrosis is a hallmark of IUA, with its resolution depending on progenitor cells migrating from the basal layer to damaged areas 12 – 15 . Stromal cells, the most abundant population, are central to ECM production via collagen secretion. Fibroblasts, highly injury-sensitive, initiate fibrosis through activation and excessive ECM accumulation 16 . Transforming growth factor (TGF)-β1 is a key mediator of fibroblast activation and collagen secretion through the canonical TGF-β/Smad signaling pathway 17 . Activated TGF-βR1 phosphorylates Smad2/3, which binds Smad4 and translocates into the nucleus to activate genes involved in fibroblast function and matrix deposition. This signaling is inhibited by Smad7 via negative feedback on TGF-βR1 18 . Despite advances in understanding TGF-β1 signaling in tissue fibrosis, the specific factors driving endometrial fibrosis remain unclear. Investigating the pathophysiological changes and molecular mechanisms underlying IUA is therefore urgently needed. Both fibroblasts and macrophages are crucial critical for ECM degradation and remodeling; however, the lack of cell-type-specific gene expression data limits understanding of IUA etiology. Recent studies have utilized scRNA-seq to investigate diseases microenvironments, including thin endometrium, endometriosis, and IUA, and to characterize cell heterogeneity and gene expression patterns in the human endometrium across the menstrual cycle 8 , 19 – 23 . In the present study, we profiled the transcriptomes of normal and IUA endometrial tissues during the proliferative phase using scRNA-seq. We identified eight distinct cell types with unique gene expression profiles at the single-cell resolution and uncovered the critical roles of stromal cells and macrophages in ECM remodeling. Macrophages exhibit high plasticity, with phenotypes shifting in response to environmental cues 24 . They differentiate into myofibroblasts, secrete ECM to promote fibrosis 25 , and regulate mesenchymal cell activation and ECM degradation via secreted factors 26 . Given this dual role, we focus subsequent bioinformatic analyses on macrophage heterogeneity and their interactions with other cell types, particularly stromal cells. By comparing IUA and control transcriptomes, we identified diverse stromal cell roles and screened a series of profibrotic factors, including cytokine (IL-6, IL-1β), and anti-inflammatory factors (TGF-β1, TGF-β2, TGF-β3). Ligands-receptor interaction analysis revealed a dynamic signaling network, highlighting the role of TGF-β signaling in macrophages at adhesion sites. Stromal cells were classified into seven distinct subgroups (S0–S6). S1 was enriched in adhesion sites and positioned at the terminal pseudotime stage, suggesting its involvement in fibrosis progression. S0 exhibited the highest metabolic activity, indicating a distinct role in the endometrial microenvironment. Notably, SPP1 secreted by macrophages promotes IUA formation by acting on S0 and S1, while CCL2 and CCL5 also contribute to fibrosis. Macrophage re-clustering reveals profibrotic populations with distinct trajectories. In vitro experiments suggested CCL5 and SPP1 promote fibroblast-to-myofibroblast transition. Finally, integration with public single-cell data from both proliferative and secretory phase showed that IUA alters estrogen signaling and impairs the opening of the endometrial implantation window. Together, our findings offer important insights into the roles of macrophages and stromal cells at different fibrotic sites, supporting future exploration of therapeutic strategies.

Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary