The single-cell RNA sequencing data used in this study was downloaded from the Gene Expression Omnibus (GEO) database, specifically from dataset GSE146912. Data processing was performed using Seurat version 4.4.0, which is widely used for single-cell RNA-seq analysis. Quality control (QC) measures were implemented to ensure the reliability of the data. The QC criteria included the requirement that each gene must be expressed in at least 1% of the cells, and the number of genes detected in each cell was constrained to a range of 400 to 7,000. Additionally, cells with a mitochondrial gene percentage greater than 5% and those with a high percentage of red blood cell-related genes (>1%) were excluded to minimize potential sources of contamination or low-quality data.

To further improve data quality, doublets (i.e., events where two or more cells are incorrectly captured as a single cell) were identified and removed using the DoubletFinder (version 2.0.3) algorithm. Subsequent cell clustering was based on these identified variable genes. The identification of differentially expressed genes (DEGs) was performed using the Seurat function FindAllMarkers, which helps in the detection of genes that are uniquely expressed in each cluster of cells. Cluster annotation was carried out by referencing typical cell-type-specific markers in the literature.

To address potential batch effects, which are common in large-scale single-cell datasets, the Harmony algorithm (version 1.2.0) was employed to integrate the data and mitigate these batch effects, ensuring more accurate and robust downstream analysis. This approach ensures high-quality and reliable results for subsequent analyses and interpretations.

In this study, we conducted cell communication analysis using CellChat V1.6.1, with the mouse database selected for the analysis. CellChat is a powerful tool that integrates prior knowledge of ligand-receptor interactions and their cofactors to model the probability of cell communication from single-cell RNA-seq data. The analysis process involves several key steps. First, we created a CellChat object by inputting the normalized data and cell metadata. Then, we set the mouse database in the CellChat object. Next, we preprocessed the expression data to identify overexpressed genes and interactions, which are crucial for subsequent communication inference. After that, we applied the projectData function to smooth the data using protein-protein interaction (PPI) data, enhancing the reliability of communication predictions. Subsequently, we computed the communication probabilities and filtered the interactions based on a minimum cell threshold. Finally, we aggregated the communication network at the pathway level. To assess intergroup differences, we performed differential interaction analysis using the netVisual_diffInteraction function. All visualizations were generated using built-in CellChat functions.

We conducted transcription factor analysis using pySCENIC V0.12.1, a powerful Python-based tool for inferring gene regulatory networks from single-cell RNA-seq data. The analysis process involves three main steps. First, we constructed a co-expression network of transcription factors and potential target genes using the grn command with the GRNBoost2 algorithm, which identifies regulatory relationships based on gene expression correlations. Next, we refined the network by integrating motif information with the ctx command, using the mm10_10kbp_up_10kbp_down_full_tx_v10_clust.genes_vs_motifs.rankings.feather file as input. This step filters out non-significant interactions and retains only those supported by motif evidence. Finally, we assessed the activity of each regulon across all cells using the aucell command, which provides a quantitative measure of regulon activity in each cell. The visualization process followed the SCENIC (R) tutorial, ensuring a comprehensive presentation of the results.

We performed pseudotime analysis using Monocle2 V2.26.0. The analysis workflow involved several key steps. First, we identified differentially expressed genes across clusters using the differentialGeneTest function with a full model formula that included cluster information. The top 1000 genes identified from this analysis were used to filter the dataset with the setOrderingFilter function. Next, we applied dimensionality reduction using the reduceDimension function with the DDRTree method, which is the default in Monocle2 and helps visualize the data in a low-dimensional space. Finally, cells were ordered along the inferred trajectory using the orderCells function. Additionally, genes that varied along the pseudotime trajectory were also identified using differentialGeneTest, and their expression patterns were visualized with the corresponding Monocle2 functions.

In this study, we employed the clusterProfiler package to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. For each cluster, the top 50 differentially expressed genes were selected as input for enrichment analysis. The enrichGO function was used to categorize genes into biological process (BP), molecular function (MF), and cellular component (CC), providing insights into their functional roles. Meanwhile, the enrichKEGG function identified significantly enriched signaling pathways associated with each cluster. This approach allowed us to systematically explore the biological significance of each cluster, uncovering potential functional pathways and molecular mechanisms involved in the dataset.

For subcluster identification, relevant cell types were extracted, followed by processing similar to the standard single-cell analysis workflow, except that no additional quality control (QC) was performed. The analysis was conducted using Seurat, following its standard pipeline. Specifically, NormalizeData(normalization.method = “LogNormalize”) was applied to normalize gene expression, and FindVariableFeatures(selection.method = “vst”, nfeatures = 2000) was used to identify highly variable genes. The ScaleData function was then applied to the highly variable genes identified previously, to remove unwanted sources of variation.

RunPCA(features = VariableFeatures(object)) was performed for dimensionality reduction, followed by batch effect correction using the Harmony algorithm. FindNeighbors(dims = 1:30, reduction = “harmony”) was then used to construct the nearest-neighbor graph. For clustering, FindClusters was run iteratively with the resolution parameter ranging from 0.1 to 1 to explore different clustering granularities. The optimal number of clusters was determined using the clustree package, which visualizes clustering results across different resolutions to aid in selecting an appropriate resolution value.

RunUMAP and RunTSNE were used for visualization, and marker genes were identified using FindAllMarkers with default parameters to annotate subclusters based on known cell type markers.

The paraffin-embedded sections were prepared for immunofluorescence. Antigen retrieval for the paraffin-embedded sections was conducted using high-temperature pressure in Tris-EDTA buffer (pH 9.0). Tissue sections were blocked with goat serum (Biosharp, BL1097A) and incubated overnight with primary antibodies (S100a8, R&D, MAB3059; Wilms Tumor 1, Immunoway, YM6533). Fluorescent secondary antibodies (Goat Anti-Rat IgG H&L, Abcam, Cat# ab150159; Goat anti-Mouse IgG (H+L), Invitrogen, Cat#A11001) were incubated for 1 hour. Nuclear staining was performed using DAPI (Biosharp, BL120A). Images were captured using a laser scanning confocal microscope.

To explore the changes in specific gene expression in mouse glomerular cells under differentpathological and physiological backgrounds, we intentionally included mice of different ages (Supplementary Table 1), with different disease backgrounds, and at different disease durations for a comprehensive and integrated analysis of the mouse glomerulus (Figure 1A). We integrated the enriched single-cell transcriptomic sequencing data (GSE146912) of glomeruli and conducted strict quality control and batch effect correction (Method). Through unsupervised clustering, we obtained 29 subpopulations (Figure 1B). Utilizing classical kidney cell markers (Supplementary Table 2), we identified and characterized the kidney cell subpopulations predominantly composed of glomerular cells, including podocytes (PODs), mesangial cells (MCs), endothelial cells (gECs), and parietal epithelial cells (PECs) (Figure 1C). Different from other kidney single-cell omics data (11–14), this dataset had captured a certain number of glomerular-related cell subsets (15, 16), which had provided the possibility for subsequent analysis.

We systematically profiled the expression of susceptibility genes identified through genome-wide association studies (GWAS) for major kidney diseases (focal segmental glomerulosclerosis-FSGS, diabetic nephropathy-DN, lupus nephritis-LN, immunoglobulin A nephropathy-IgAN) across diverse renal cell subtypes, and delineated their cell type–specific enrichment patterns (Figure 1D). Genes linked to FSGS were highly expressed primarily within glomerular cells, notably PODs, PECs, and MCs, but exhibited minimal influence on gECs, extraglomerular endothelial cells, or various immune cells. Conversely, DN-associated genes were broadly expressed across diverse renal cell subtypes. For LN, the pathogenic genes were not prominently expressed in PODs but were predominantly enriched in MCs and MAC-related subsets. In contrast, IgAN-associated genes were widely active in gECs, extraglomerular endothelial cells, and immune cells, yet showed limited expression in other intrinsic glomerular cells.

Notably, we identified a high expression of multiple disease-associated molecules in Cluster 17 (pro_EC). This subpopulation expresses gEC markers (Sost, Ehd3) along with characteristic proliferation-related molecules (Top2a, Stmn1). These results suggest the existence of a distinct proliferative gEC population within the glomerulus, which may play a pivotal role in disease pathogenesis.

Given the critical role of POD and PEC injury in proteinuria pathogenesis (4, 17, 18), we performed an integrated re-clustering analysis of these populations, which identified 13 distinct subclusters (Figure 2A). Clusters 8 and 10 both expressed established PEC markers (Igfbp6, Ptgis, Pla2g7 and Dkk3) (Figure 2B). Cluster 8 was designated as PEC_1 based on its high expression of AP-1-related molecules (Junb, Jun, Fos and Fosb). In contrast, Cluster 10, characterized by elevated expression of Cp, Pcp4, Cdh16, Fxyd1, C3, Pax8, Vcam1, and Cd9, was designated PEC_2. Notably, while PEC_1 contained cells derived from both healthy and diseased mice, the PEC_2 population was predominantly contributed by the disease group (Figure 2C).

Distinct injured podocyte subpopulations emerge across disease models (Figures 2A, B). These clusters exhibit unique molecular signatures (Figure 2B), disease origins (Figure 2C), transcription factor activities (Figure 2D), and functional enrichments (Figures 2E–G). POD_Injury_1, present across multiple disease models (Figure 2C), highly expresses Tagln, Cxcl1, Crct1, Angptl7, Tgfb2, and Flna (Figure 2B), suggesting a common injury response signature shared among different nephropathies. In contrast, Cluster 2 exhibits high expression of Endou, Cd200, Lgmn, Il18, Dmpk, and Spon2 (Figure 2B) and originates almost exclusively from ob/ob mice (Figure 2C), prompting its annotation as POD_ob. Similarly, Cluster 11, predominantly originating from ADR-treated mice, was designated POD_ADR based on elevated expression of Selenbp1, Lpar1, S100a8, S100a9, and Sult1a1. Transcription factor analysis revealed that POD_ob specifically expresses Churc1, Sp2, Dbp, Foxc2, Mafb, and E2f1, while POD_Injury_1 and POD_Injury_2 both upregulate Bcl3, Rela, Fosl1 and Nfkb2. POD_ADR is characterized by high expression of Cebpd (Figure 2D). GO analysis indicated that POD_Injury_1 (Figure 2E) and POD_ob (Figure 2F) are both enriched in terms related to cytoskeleton and actin filament organization. POD_ADR, however, is primarily involved in intracellular signaling, protein transport and secretion, smooth muscle cell migration, and apoptotic signaling pathways (Figure 2G).

To identify common injury mechanisms, we compared the transcriptional profiles of podocyte subclusters between disease and control groups, quantifying differentially expressed genes (DEGs). A substantial number of DEGs were identified in POD_AP1 across different age groups and at the 5-day nephritis time point. Additionally, POD_1, POD_2, and POD_3 also exhibited numerous DEGs under conditions of aging, diabetes, and nephritis (Figure 3A). Notably, POD_1—a major constituent of the podocyte population—significantly downregulated AP-1-related molecules (Fos, Junb, Jun, Fosb, Atf3) with advancing age. In the nephritic state, Spp1 and Cebpb was markedly upregulated (Figure 3B). Pseudotime analysis of podocyte subclusters distributed them across five distinct states, with a majority of injury-associated subclusters coalescing in State 5 (Figures 3C–E). GO analysis revealed that State 5 is primarily associated with cell–matrix interactions, actin filament assembly and organization, wound healing, morphogenesis of cardiac and muscle tissue, positive regulation of T-cell activation and positive regulation of leukocyte–cell adhesion (Figure 3F). Key injury-related genes, including Tagln, Cald1, Tagln2, Gpx3, Press23, and Lgals1, were also predominantly highly expressed in this state (Figure 3G). These characteristics suggest that State 5 likely represents a characteristic injury or repair state, involved in podocyte response, remodeling, and immunomodulatory processes following damage. Notably, the immunofluorescence staining of renal tissues confirmed the co-expression of Wt1 and S100a8 in ADR mice, demonstrating that some podocytes highly express S100a8 in this disease model (Figure 3H).

Mesangial cells (MCs) are crucial for maintaining glomerular structural integrity, regulating filtration rate, synthesizing or degrading extracellular matrix (ECM), and producing various cytokines and bioactive molecules (19). We performed reclustering analysis on MCs, identifying 13 distinct subclusters (Figure 4A), which were annotated based on their characteristic gene expression profiles (Figure 4B). Among these, we focused on cluster 1, annotated as Mes_Injury, which was predominantly derived from multiple disease groups (Figure 4C). This subpopulation highly expressed Cxcl1, Egr1, Hspa1b, Socs3, and Dnaib1 (Figure 4D). Notably, we identified several disease-specific MC subclusters—clusters 3, 11, and 12—that were markedly expanded in ob/ob mice (Figure 4C). Specifically, Mes_ob_1 (cluster 3) exhibited a transcriptome distinct from other MCs and appeared as an outlier in UMAP visualization (Figure 4A). It was characterized by high expression of Aldh1a2, Thbs1, Fbln5, Cdh11, and Loxl2 (Figure 4B). Transcription factor analysis revealed that Mes_Injury highly expressed Fos, Junb, Xbp1, and Cebpb, while all ob/ob-derived MCs shared high expression of Dbp and Prrx2 (Figure 4D). Mes_ob_1 uniquely expressed Mbd2, Spi1, and Pura (Figure 4E). Of translational relevance, Prrx2 is involved in development, tissue repair, and epithelial–mesenchymal transition (EMT), regulating ECM remodeling and cell migration (20, 21). Spi1, primarily expressed in the immune system, is a key regulator of hematopoiesis and immune responses (22, 23). Functionally, GO enrichment analysis indicated that Mes_Injury is primarily associated with intracellular signaling regulation and stress responses (Figure 4F), whereas Mes_ob_1 is more enriched in processes related to cell migration, tissue repair, and cell–matrix interactions (Figure 4G). Pseudotime trajectory analysis of all MC subclusters (Figure 4H) showed that ob/ob-derived MCs were concentrated in State 3 (Figures 4I, J), where enriched genes were functionally linked to antigen presentation, immune response, regulation of epithelial cell proliferation, and negative regulation of cell projection organization (Figure 4K).

Glomerular endothelial cells (gECs) are essential for maintaining the integrity of the filtration barrier, regulating molecular exchange, balancing coagulation and anticoagulation, and providing structural support (9). Their injury contributes to increased permeability (leading to hematuria and proteinuria) and microthrombosis, thereby accelerating kidney disease progression (9). We performed reclustering of gECs, identifying 18 subclusters that included 9 distinct subpopulations with unique transcriptional profiles (Figures 5A, B). Notably, gEC heterogeneity was not primarily driven by specific disease contexts (Figure 5A). However, cluster 9 emerged as an exception, being predominantly derived from multiple disease groups (Figure 5C). We annotated cluster 9 as gEC_Injury, identifying it as a potential common injury-specific subpopulation (Figure 5B). The gEC_Injury subpopulation was characterized by high expression of Apln, Mir147, Pgf, Actn1, Ercc and Procr (Figure 5B). Key transcription factors defining this cluster included Fosl, Tead4, and Relb (Figure 5D), which are known regulators of cell proliferation, migration, immune response, and tissue repair (24–26). Consistent with this molecular signature, GO analysis indicated that gEC_Injury is functionally enriched in processes related to cell migration, cytoskeletal regulation, immune response, as well as angiogenesis and vascular development (Figure 5E).

Transcriptomic analysis revealed substantial differential gene expression in gEC_1, gEC_2, gEC_3, gEC_4, and gEC_6 under nephritis and ADR conditions, with a predominant pattern of transcriptional upregulation (Figure 6A). Among these, gEC_1 and gEC_2 exhibited the most pronounced changes (Figure 6A). Disease-specific responses were observed for these two subpopulations: gEC_1 significantly upregulated Spp1, Sparc, and Mgp in the 5-day nephritis model (Figure 6B), whereas gEC_2 was characterized by elevated expression of Lrg1, S100a8, Fabp4, Serpine2, and Ch25h in the ADR model (Figure 6C). Pseudotime trajectory analysis (Figures 6D–J) further delineated injury-associated states. State 9 was predominantly populated by cells from disease groups and served as the primary distribution zone for the gEC_Injury subpopulation (Figures 6E–G). State 10 was also largely composed of disease-derived cells, with the ADR model being the major contributor (Figures 6E–G). Functional characterization showed that State 9 was significantly enriched for biological processes related to collagen synthesis and metabolism, inflammatory responses, angiogenesis, and responses to bacterium-derived molecules (Figure 6H). In contrast, State 10 was primarily associated with cellular responses to interleukin-17 and chemokines, chemokine-mediated signaling pathways, detoxification of inorganic compounds, and responses to metal ions and inorganic substances (Figure 6I). Furthermore, key genes involved in cellular stress, immune response, proliferation, and differentiation—including Socs3, Nfkbia, Atf3, and Icam1—were highly expressed in both State 9 and State 10 (Figure 6J).

We analyzed the interactions between glomerular cell subsets and other cell types under disease conditions (Figure 7A) and found that mesangial cells exhibited particularly prominent ligand–receptor relationships with other cell subpopulations (Figure 7B). Notably, we observed that VISFATIN-mediated ligand–receptor interactions were primarily identified between Mes_injury and Pod_injury_1 subsets (Figure 7C). This suggested that extracellular NAMPT derived from mesangial cells may directly or indirectly bind to and activate Toll-like receptor 4 (TLR4) on podocytes. Elevated Visfatin levels are known to promote chronic inflammation via the TLR4/NF-κB pathway (27). Consistent with our findings (Figure 7D), prior studies have reported that mesangial cells may upregulate Cxcl1 expression upon injury (16). Furthermore, we detected high Cxcl1 expression in other injured podocyte subsets. This may represent a shared mechanism by which mesangial cells and podocytes collectively facilitate the infiltration of CXCR2-high neutrophils in glomerular diseases. Injured podocytes may also recruit macrophages and neutrophils via CADM, CXCL, and MIF pathways (Figure 7E). Among glomerular cell types, Mif was predominantly upregulated in Pod_injury_1 under injury conditions, while its receptors—Cd74, Cd44, and Cxcr4—were highly expressed in macrophages, neutrophils, and Pod_Und subpopulations (Figure 7F). Existing evidence supports a key role for MIF signaling in inflammatory recruitment (28). Recent studies suggest that MIF–CD44 interactions may contribute to parietal epithelial cell injury and the development of focal segmental glomerulosclerosis (1, 15, 29). Our results indicated that this may represent a broad mechanism underlying glomerular inflammatory recruitment following podocyte injury across multiple disease contexts.