COL1A1 ⁺ Epithelial Cells Orchestrate VEGFA‐VEGFR Signaling in Endometritis Revealed by Single‐Cell Analysis

ABSTRACT

Background

Endometritis is linked to adverse reproductive outcomes, but epithelial programs in disease initiation and persistence remain unclear. We aimed to systematically define inflammation‐associated epithelial states and regulation in endometritis using single‐cell analysis.

Method

The single‐cell RNA sequencing (scRNA‐seq) data from Gene Expression Omnibus (GEO) were processed using Seurat. After quality control, data were normalized with SCTransform and batch‐corrected using the Harmony package. Cell types were annotated based on canonical markers. Differential expression analysis was performed to identify genes altered in endometritis. Epithelial cells were subsetted for reclustering and trajectory inference using Monocle2. Cell–cell communication was inferred with CellChat, and transcriptional regulon activity was assessed using SCENIC and AUCell methods.

Results

A total of 153,877 cells formed 14 clusters across seven lineages, with lower epithelial proportion in endometritis. Epithelial cells included four subpopulations (SPDEF⁺, MT1H⁺, Ciliated, COL1A1⁺), with SPDEF⁺ and COL1A1⁺ expanded in disease. Upregulated epithelial genes enriched in ribosome, antigen processing/presentation, and estrogen‐related pathways. Pseudotime showed a continuous trajectory splitting into two fates: one ferroptosis‐related with glutathione metabolism and mineral absorption, the other ribosome and antigen presentation. Communication networks were denser in endometritis, with COL1A1⁺ epithelium as a hub and enhanced VEGFA–VEGFR signaling. SCENIC revealed elevated regulon activity in disease, especially AP‐1 (JUN/FOS) modules.

Conclusion

This epithelial‐centric single‐cell atlas delineates disease‐associated states, fate decisions, signaling axes, and regulatory programs in endometritis. The data support a model of epithelial fate remodeling coupled to angiogenic signaling and AP‐1–driven transcription, nominating testable targets for mechanistic validation and potential translation.

Single‐cell RNA‐seq of normal and endometritis endometrium shows reduced epithelial cells and fate divergence toward ferroptosis or ribosome/antigen‐presentation programs. COL1A1⁺ epithelial cells act as an AP‐1–driven hub that enhances VEGFA–VEGFR signaling and remodels the vascular microenvironment.

Abbreviations

ANOVA

analysis of variance

AP‐1

activator protein‐1

AUCell

area under the recovery curve‐based regulon activity scoring (SCENIC module)

BH

Benjamini–Hochberg

CD138

plasma‐cell marker used for CE pathology

CellChat

cell–cell communication inference toolkit

DDRTree

discriminative dimensionality reduction with trees (Monocle2 method)

DEGs

differentially expressed genes

ECM

extracellular matrix

EpC

epithelial cells

EVs

extracellular vesicles

FDR

false discovery rate

GEO

gene expression omnibus

GO

gene ontology

IFN‐γ

interferon‐gamma

JUN/FOS

AP‐1 family components

KEGG

kyoto encyclopedia of genes and genomes

k‐NN

k‐nearest neighbors

MHC‐II

major histocompatibility complex class II

Monocle2

single‐cell trajectory inference toolkit

NK/T

natural killer/T

PCA

principal component analysis

QC

quality control

RIF

recurrent implantation failure

SCENIC

single‐cell regulatory network inference and clustering

scRNA‐seq

single‐cell RNA sequencing

Seurat

single‐cell analysis framework

TFs

transcription factors

t‐SNE

t‐distributed stochastic neighbor embedding

UMAP

uniform manifold approximation and projection

UMI

unique molecular identifier

VEGF

vascular endothelial growth factor

VEGFA

VEGF‐A

VEGFR

vascular endothelial growth factor receptor

1. Introduction

Endometritis is a common yet underrecognized inflammatory disorder of the female reproductive system that may present acutely or pursue a chronic, indolent course [1]; chronic endometritis (CE) is closely associated with infertility, recurrent implantation failure (RIF), and recurrent miscarriage, and it adversely affects embryo implantation and the maintenance of pregnancy in assisted reproductive cycles [2]. The putative mechanisms involve multifactorial dysregulation, including impaired immune tolerance at the embryo–maternal interface and aberrant remodeling of the mucosal milieu and vascular supply [3, 4]. Clinically, the diagnosis of CE integrates symptoms, imaging, hysteroscopic findings, and histopathology, with stromal plasma cells confirmed by CD138 immunohistochemistry widely used for pathological diagnosis [5, 6]; however, intercenter variation in sampling, threshold definition, and reader concordance leads to inconsistent sensitivity and specificity [7]. Antibiotic therapy is commonly employed in practice, yet treatment effectiveness remains uncertain due to a complex pathogen spectrum, antimicrobial resistance, and recurrence; recent clinical studies also suggest that improvements in pregnancy outcomes among patients with “mild” CE are not uniform, underscoring the importance of test‐of‐cure re‐evaluation to document therapeutic success [8, 9]. Collectively, this clinical heterogeneity and diagnostic–therapeutic uncertainty underscore the need to reconstruct disease mechanisms at the cellular and molecular levels, with particular emphasis on epithelial cells—the barrier, secretory, and interface‐regulating compartment—to explain disease initiation, persistence, and relapse and to inform precision stratification and targeted intervention [10, 11, 12].

Across the healthy menstrual cycle, the endometrial epithelium undergoes tightly programmed structural and functional remodeling governed by systemic hormones and local cues: dynamic shifts in the proportions of ciliated and secretory cells shape luminal flow and secretory profiles; the glycocalyx and mucus form the first line of defense; ion channels and transporters maintain the physicochemical milieu; and tight junctions, adherens junctions, and desmosomes, coupled to the cytoskeleton, preserve barrier integrity and selective permeability [13, 14, 15]. Under inflammatory or infectious conditions, this circuitry can be perturbed at multiple nodes: barrier disruption and increased permeability facilitate the translocation of danger‐ and pathogen‐associated signals; epithelial receptor and secretory profiles are reset, altering response thresholds to sterile and pathogen‐derived stimuli and rewiring costimulatory/coinhibitory circuits with immune cells; and changes in adhesion‐signaling coupling with the stroma and extracellular matrix can reprogram epithelial morphology and motility and perturb vascular permeability and angiogenic signaling, thereby amplifying or prolonging inflammation [16]. Accordingly, an epithelium‐centric systems perspective that integrates cell states, intercellular communication, and upstream regulatory programs not only helps explain symptom persistence and variable treatment responses but also delineates actionable nodes and translatable biomarkers.

Conventional histology, immunostaining, and bulk transcriptomics have provided important clues yet fall short of resolving rare or transient epithelial states and of systematically quantifying cross‐cell–type interactomes and upstream regulatory architectures [17]. Recent single‐cell studies of healthy or physiologic endometrium have mapped cellular composition and dynamics and established high‐resolution human endometrial reference atlases and visualization resources [18]; however, systematic epithelial‐focused single‐cell analyses in endometritis remain limited, particularly those integrating cell states with ligand–receptor communication and transcriptional regulation into a unified view [19]. To address this gap, we advocate a unified analytical pipeline that combines single‐cell RNA sequencing (scRNA‐seq) to define cell types and states, trajectory/pseudotime modeling to chart state transitions, and inference of cell–cell communication together with reconstruction of gene‐regulatory networks; in parallel, stringent quality control, batch correction, and cross‐sample integration are required to enhance robustness and reproducibility [20, 21, 22].

Based on clinical needs and methodological advances, this study constructs a single‐cell epithelial atlas of endometritis to dissect disease‐associated epithelial remodeling from multiple complementary angles. The four layers in which the project has been designed include: the characterization of inflammation‐associated epithelial states at single cell resolution provides the basis for the rest of the analysis; characterization of the communication axes through which these epithelial subsets communicate with immune, vascular, and stromal compartments—and how these sustain or amplify inflammation; discovery of upstream transcriptional regulators with tools like SCENIC that drive this state and communication change will turn descriptive cell states into causative candidates; and open analytical workflows and data for the community to test and translate the hypotheses generated. The four layers—states, signals, regulators, and resources form a coherent framework which moves from observation to intervention. This can facilitate mechanistic dissection and precision therapeutics for endometritis.

2. Materials and Methods

2.1. Data Source

Data were retrieved from the Gene Expression Omnibus (GEO) [23], accession GSE223639, comprising seven normal endometrium and seven endometritis samples for downstream analyses.

2.2. Data Preprocessing

For the scRNA‐seq dataset under GEO accession GSE223639, we implemented a standard workflow for quality control and integration. Genes were retained if detected in ≥ 3 cells, and cells were required to express ≥ 200 genes. At the cell level, we kept high‐quality profiles with nCount_RNA ≤ 100,000 and percent.mt < 15%. Data were normalized and variance‐stabilized using SCTransform, after which principal components were computed with RunPCA. Harmony was applied to remove cross‐sample batch effects, where sample origin was designated as the batch variable [24]. Harmony was applied with a lambda value of 0.5 and 50 maximum iterations. We constructed the k‐NN graph and performed community detection using FindNeighbors/FindClusters to define subpopulations; UMAP was used for visualization employing the top 20 principal components. Finally, cell identities were assigned by matching canonical markers curated in the CellMarker 2.0 database.

2.3. Differential Expression Analysis

Differential expression between Endometritis and Normal groups was performed using FindMarkers in the Seurat package on the normalized/integrated dataset [25]. Unless otherwise specified, we applied a two‐sided Wilcoxon rank‐sum test with Benjamini–Hochberg correction for multiple testing (adjusted p‐values reported as p_val_adj). Genes were considered differentially expressed if |avg_logFC| > 0.25 and p_val_adj < 0.05.

2.4. Construction of Single‐Cell Pseudotemporal Trajectories

Pseudotime analysis was performed on the epithelial subset using Monocle2 [20]. Raw UMI counts and associated phenotype metadata (group: Endometritis vs. Normal) were extracted from the normalized/integrated object to initialize a CellDataSet with newCellDataSet (negative binomial family for UMI data), followed by estimateSizeFactors and estimateDispersions. To reduce noise, we retained genes detected in ≥ 10 cells. Group labels were stored in the phenoData, and differentialGeneTest was applied to compare Endometritis versus Normal; significantly associated genes (based on multiple‐testing–adjusted q‐values) were designated as ordering genes. Dimensionality reduction was performed with reduceDimension (method = “DDRTree,” max_components = 2), and cells were arranged along pseudotime using orderCells. To orient trajectory direction, the branch enriched for Normal cells was assigned as the root state, from which pseudotime was computed. Visualization employed plot_cell_trajectory.

2.5. SCENIC Analysis

We applied the SCENIC workflow to infer transcriptional regulatory networks and quantify regulon activity in epithelial cells [26]. Briefly, transcription factors (TFs) and their coexpressed genes were first derived from the expression matrix, and GENIE3 was used to infer weighted TF‐target associations. Candidate edges were then pruned with RcisTarget by motif enrichment and genomic proximity using human cis‐regulatory motif databases, yielding high‐confidence regulons. Subsequently, AUCell computed per‐cell regulon activity scores (area under the recovery curve based on expression rank), which were normalized for visualization and group‐wise comparisons; when appropriate, AUCell‐derived thresholds were used to binarize regulon activity into “on/off” states.

2.6. Statistical Analysis

All tests were two‐sided with Benjamini–Hochberg FDR control (α = 0.05). Continuous data: Shapiro–Wilk for normality, then Welch's t‐test or Wilcoxon; multigroup: one‐way ANOVA (Welch if needed) or Kruskal–Wallis with FDR‐controlled post hoc. Categorical data: χ ² or Fisher's exact; correlations: Spearman's ρ with FDR adjustment. Enrichment (where applicable): hypergeometric/Fisher with BH correction. Analyses were conducted in R with fixed random seeds.

3. Results

3.1. Single‐Cell Atlas of Endometritis

We integrated scRNA‐seq data from seven normal endometrium and seven endometritis samples in GSE223639. After stringent quality control, normalization, dimensionality reduction, and clustering, 153,877 high‐quality cells remained, resolving 14 clusters (Figure 1A). Cluster annotation using canonical marker panels identified seven major lineages: NK/T cells, mesenchymal cells, epithelial cells, myeloid cells, mast cells, endothelial cells, and erythrocytes (Figure 1B–C). Per‐sample cellular compositions are summarized in Figure 1D. Comparative composition analysis showed a lower epithelial fraction in endometritis relative to controls, consistent with epithelial reduction/remodeling under inflammatory conditions (Figure 1D).

FIGURE 1.

Overview of the single‐cell atlas. (A) UMAP of unsupervised Seurat clusters. (B) UMAP of annotated cell types. (C) Dot plot of canonical marker expression across cell types. (D) Stacked bar plots of per‐sample cell‐type composition.

3.2. Identification of Epithelium‐Associated Differentially Expressed Genes (DEGs) in Endometritis

Differential expression between Endometritis and Normal was performed on the normalized/integrated object using Seurat/FindMarkers, with significance defined as p_val_adj < 0.05 (Benjamini–Hochberg) and |avg_logFC| > 0.25 (Figure 2A–B). Epithelial and Mast cells exhibited the largest numbers of DEGs, suggesting heightened transcriptional remodeling in these compartments. Functional enrichment of upregulated epithelial genes revealed significant over‐representation of Ribosome, Antigen processing and presentation, and Estrogen signaling pathway terms, among others (Figure 2C–D). Collectively, these findings highlight epithelial transcriptional reprogramming in endometritis and nominate testable pathways for subsequent mechanistic interrogation.

FIGURE 2.

Differential expression and enrichment overview. (A) Rose plot showing the number of DEGs by cell types. (B) Scatter plot of epithelial DEGs: Upper, log₂ fold‐change distribution with top five upregulated genes; lower, log₂ fold‐change distribution with top five downregulated genes. (C) KEGG enrichment dot plot for upregulated epithelial DEGs. (D) GO–Biological Process enrichment dot plot for upregulated epithelial DEGs.

3.3. Epithelial Heterogeneity

To resolve epithelial‐level changes, epithelial cells from the endometritis and normal groups were subsetted for dimensional reduction and reclustering (t‐SNE; Figure 3A). Using cluster‐specific marker signatures, we annotated four epithelial subpopulations: SPDEF⁺ EpC, MT1H⁺ EpC, Ciliated EpC, and COL1A1⁺ EpC (Figure 3B). Compositional comparison revealed increased proportions of SPDEF⁺ EpC and COL1A1⁺ EpC in endometritis (Figure 3C), suggesting disease‐associated redistribution and activation of secretory/differentiation‐linked and matrix‐associated epithelial states. Gene Ontology—Biological Process enrichment on subcluster‐specific markers (Figure 3D–G) showed that COL1A1⁺ EpC markers were significantly enriched for “extracellular matrix organization,” “extracellular structure organization,” and “epithelial cell proliferation.” These patterns indicate epithelial ECM remodeling and proliferative reprogramming in the inflammatory milieu, providing testable mechanistic hypotheses for tissue repair or pathological remodeling in endometritis.

FIGURE 3.

Epithelial heterogeneity. (A) t‐SNE of annotated epithelial subclusters. (B) Violin plots of subcluster‐specific marker genes (SPDEF, MT1H, FOXJ1, COL1A1). (C) Proportional composition of epithelial subclusters in Endometritis vs. Normal. (D) GO–Biological Process enrichment for SPDEF⁺ EpC markers. (E) GO–Biological Process enrichment for MT1H⁺ EpC markers. (F) GO–Biological Process enrichment for Ciliated EpC markers. (G) GO–Biological Process enrichment for COL1A1⁺ EpC markers.

3.4. Pseudotime Analysis

To delineate dynamic epithelial transitions in endometritis, we inferred trajectories with Monocle2 and performed pseudotime analysis. Epithelial cells progressed along a continuous trajectory and bifurcated into two prominent branches at the terminal stage (Figure 4A–C). The branch enriched for Normal cells was assigned as the root, enabling a temporal characterization from putative healthy states toward disease‐associated states. Branch and fate analysis indicated two distinct terminal outcomes (Figure 4D). Branch 1 showed pathway‐level enrichment for Ferroptosis, accompanied by activation of Glutathione metabolism and Mineral absorption, consistent with iron‐dependent lipid peroxidation stress and metabolic reprogramming. Branch 2 was enriched for Ribosome and Antigen processing and presentation, suggesting enhanced protein synthesis and immune presentation functions. Together, these results depict a continuous‐to‐branching epithelial transition under inflammatory conditions and nominate testable pathways for mechanistic follow‐up.

FIGURE 4.

Epithelial pseudotime trajectory. (A–C) Single‐cell trajectory of epithelial cells colored by group (Endometritis/Normal), pseudotime, and epithelial subclusters. (D) Heatmap of branch‐specific DEGs with corresponding KEGG pathway enrichment analysis.

3.5. Cell–Cell Communication

To delineate cross‐lineage signaling, we used CellChat to compare ligand–receptor networks between Endometritis and Normal. At the global level, the number of interactions and overall network strength were higher in endometritis, indicating augmented intercellular signaling under inflammatory conditions. Ranking information flow across pathways revealed group‐biased signaling: pathways marked in red were enriched in Endometritis, whereas those in green were enriched in Normal (Figure 5A). At cell‐type resolution, COL1A1⁺ EpC exhibited broadened incoming/outgoing communication and occupied a hub‐like position across multiple pathways (Figure 5B). Ligand–receptor comparisons further showed selective enhancement of VEGFA–VEGFR1/2 and VEGFA–VEGFR2 axes in the disease network (Figure 5C–D), implicating angiogenic signaling in remodeling of the pathological microenvironment. Collectively, these findings indicate a denser and selectively activated communication landscape in endometritis, nominating testable pathways and candidate intervention points for subsequent studies.

FIGURE 5.

Cell–cell communication. (A) Comparison of interaction counts and pathway information flow between Endometritis and Normal. (B) Communication network centered on COL1A1⁺ EpC. (C) Ligand–receptor dot plot: Interactions between COL1A1⁺ EpC and other cell types. (D) Ligand–receptor dot plot: Interactions between other cell types and COL1A1⁺ EpC.

3.6. TFs Regulating COL1A1 ⁺ EpC

To identify key regulators within COL1A1⁺ epithelial cells (EpC), we applied the SCENIC pipeline to infer regulons and quantify per‐cell activity using AUCell. Groupwise comparisons revealed higher regulon activity in endometritis than in controls (boxplots, Figure 6A), indicating broad transcriptional upregulation under disease conditions. We then performed pathway enrichment on targets from the most discriminative regulons highlighted in Figure 6A; notably, targets of JUN‐ and FOS‐associated regulons showed significant enrichment in KEGG pathways (Figure 6B–C), spanning inflammatory signaling, epithelial remodeling, and related processes. Collectively, these findings implicate AP‐1 family factors (JUN/FOS) as putative regulators of disease‐associated programs in COL1A1⁺ EpC, thereby nominating candidates for mechanistic validation and therapeutic exploration.

FIGURE 6.

Regulon activity and pathway enrichment. (A) Box plots of transcription factor AUCell scores in COL1A1⁺ EpC for Endometritis vs. Normal. (B) KEGG enrichment dot plot for targets of the JUN‐associated regulon. (C) KEGG enrichment dot plot for targets of the FOS‐associated regulon. Statistical significance is indicated as follows: ^**** p < 0.0001.

4. Discussion

Traditional views have regarded the endometrial epithelium as a relatively homogeneous barrier layer, yet single‐cell resolution offers an opportunity to reexamine this assumption [10]. CE is closely associated with infertility, RIF, and recurrent miscarriage, yet a unified model of its pathogenesis and molecular pathways remains unclear. In this study, we utilized single‐cell technology to identify a COL1A1⁺ epithelial subpopulation exhibiting characteristics reminiscent of epithelial‐mesenchymal transition. Additionally, we observed functional differentiation of epithelial cells at the terminal end of pseudotemporal trajectories. Within the intercellular communication network, COL1A1⁺ epithelial cells occupy a pivotal position, accompanied by enhanced signaling along the VEGFA–VEGFR axis. At the level of transcriptional regulation, AP‐1–related programs are activated. The chain of evidence indicates an imbalance coupling in the “epithelial fate–vascular pathway–regulatory network,” which provides a cellular‐level explanatory framework for the continuous mechanisms of CE [27, 28]. Thus, single‐cell resolution does not overturn existing theories but rather supplements the current pathological model with a previously underappreciated cellular state, thereby refining our understanding of tissue remodeling during inflammatory conditions.

Pseudotime analyses show that endometrial epithelial cells diverge toward two functional endpoints along a continuous trajectory. One endpoint is enriched for ferroptosis, glutathione metabolism, and mineral absorption pathways, indicating susceptibility to iron‐dependent lipid peroxidation and oxidative stress. The other endpoint is enriched for ribosome and antigen processing/presentation pathways, indicating heightened protein translation and increased immune visibility. For the first endpoint, multiple reviews and primary studies in female reproductive disorders have proposed a cascade of iron overload, glutathione depletion, lipid peroxidation, and ferroptosis, which can compromise epithelial barrier integrity and amplify mucosal inflammatory tone, thereby providing a plausible mechanistic substrate for the “stress‐biased” state [29, 30]. For the second endpoint, historical and contemporary evidence demonstrates that human endometrial epithelium possesses MHC‐II–dependent antigen‐presenting potential that can be upregulated by IFN‐γ; broader mucosal immunology places epithelia among “nonprofessional antigen‐presenting cells,” capable of lowering the threshold for immune crosstalk and sustaining chronicity under inflammatory conditions [31]. Concordance between these reports and the concomitant rise of antigen‐presentation and translation programs in the present data suggests that the “presentation‐biased” state may sustain inflammation by increasing the frequency of epithelial–immune interactions. Functionally, although the trajectories differ, both endpoints converge on reduced endometrial receptivity and persistent inflammation, in line with the risk signals observed for reproductive outcomes in CE [32].

Analyses of cell‐type redistribution and reconstructed communication networks suggest that COL1A1⁺ epithelial cells increase in proportion and activity, with reinforcement of extracellular matrix organization and epithelial proliferative programs. The overall number and strength of intercellular communications increase, and inbound/outbound information flow concentrates on the COL1A1⁺ epithelial compartment. At the signaling‐axis level, selective enhancement of ligand–receptor pairings between VEGFA and VEGFR1/2 is observed, implicating angiogenesis and permeability control in local microenvironmental remodeling [33]. Mechanistically, VEGFA acts through VEGFR2 on endothelial cells to activate downstream focal adhesion kinase (FAK) signaling that promotes endothelial cell migration, tube formation and vascular permeability [34]. In CE, there is an increase in the expression of endometrial VEGFA and VEGFR2, which also correlates with raised microvascular density and excessive vascularity of the peri‐implantation endometrium [27]. While such VEGF‐driven angiogenic and permeability changes have been implicated in other gynecologic conditions (such as endometriosis), where they facilitate inflammatory cell infiltration and lesion maintenance, their specific role in endometritis remains to be functionally validated [35, 36]. Clinical and translational experience with VEGF‐targeted agents further supports the feasibility of short‐course, low‐dose, individualized modulation of this pathway, although reproductive safety profiles and washout intervals require careful consideration [37, 38]. Methodologically, communication inferences are probabilistic, relying on differential expressions of ligands/receptors and curated complex databases; such analyses are statistical extrapolations. The principles and limitations of CellChat emphasize an interpretation strategy that privileges strong pathway priors and multievidence integration, supporting prioritization of the VEGF axis for experimental validation [39]. In addition, reconstructed regulatory networks indicate increased activity of AP‐1 (JUN/FOS). Cross‐tissue multiomics studies position AP‐1 as an integrator of mechanical and inflammatory cues capable of driving ECM‐stress and growth‐factor programs. On this basis, a positive‐feedback model—“COL1A1⁺ epithelium → AP‐1 → VEGF axis → endothelium/stroma”—is plausible and provides a pathway‐level explanation for exudation, bleeding, and periodic symptom fluctuations [40, 41].

Several limitations should be acknowledged. First, this study primarily relies on single‐cell transcriptomic data sourced from public databases, thus being unable to directly link specific pathogens to molecular events such as AP‐1 activation. Future prospective cohort studies should concurrently collect endometrial samples for single‐cell sequencing and microbiome analysis to establish causal chains between pathogens and host responses. Second, pseudotime analysis, intercellular communication analysis, and transcriptional regulatory network reconstruction are all computational inferences based on expression data, influenced by algorithm parameters and knowledge base completeness. Therefore, their reliability requires further validation through spatial transcriptomics and multiplex immunofluorescence staining to clarify the colocalization relationship between COL1A1⁺ epithelial cells and VEGFA signaling and AP‐1 activity in the tissue in situ. Third, the driving role of COL1A1⁺ epithelial cells has yet to be functionally validated. Future studies should establish endometrial organoids or mouse models to verify the necessity and therapeutic potential of these cells and pathways in inflammation maintenance through approaches such as gene editing and drug intervention. Finally, the proposed therapeutic strategy targeting the COL1A1⁺–VEGFA–AP‐1 axis faces practical challenges in current clinical translation. Future efforts will focus on clinical translation, prioritizing the exploration of more feasible intervention strategies such as anti‐inflammatory drugs targeting the local endometrium or small‐molecule compounds targeting downstream effector molecules. These approaches will be validated within a rigorous preclinical reproductive safety assessment framework.

5. Conclusion

This study provides a single‐cell atlas of the endometrial epithelium in endometritis, identifying SPDEF⁺ and COL1A1⁺ subsets that expand during inflammation. The latter shows features of extracellular matrix remodeling and emerges as a signaling hub, with enhanced VEGFA‐VEGFR communication pointing to angiogenic involvement in the local microenvironment. Pseudotime analysis indicates two distinct epithelial trajectories: one linked to ferroptosis and metabolic stress, the other to antigen processing and presentation. SCENIC further implicates AP‐1 (JUN/FOS) as a potential upstream regulator of these disease‐associated programs. The collective outcomes of these studies provide a cellular framework that helps to understand the mechanisms through which epithelial states change during endometritis. Further, they highlight potential avenues for mechanistic or therapeutic exploration including VEGF signaling, AP‐1 activity, and ferroptosis‐related mechanisms.

Author Contributions

Chengzi Tian led conceptualization, methodology development, key investigations, and drafting of the original manuscript. Zaiyi Li supervised the project, conducted formal data analysis, revised the manuscript critically, and managed overall project administration. All authors reviewed and approved the final manuscript.

Funding

This work was supported by Central Government Guidance Fund for Local Science and Technology Development (202407AB110013).

Ethics Statement

Ethics approval was not required for this study because it is not involved in any human experiments.

Consent

The consent was not required for this study because it is not involved in any human experiments.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The datasets generated and/or analyzed during the current study are available in the (GSE223639) repository, (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE223639).