Patient selection and study design

To understand the molecular pathogenesis of RPGN, we conducted a comprehensive analysis of kidney samples from patients with lupus nephritis (SLE, n = 19), ANCA-GN (n = 32) and anti-GBM disease (n = 6) and from healthy kidney tissue (Fig. 1a). All patients with RPGN were included in the Hamburg Glomerulonephritis Registry (Table 1). Gene expression profiling of 480 genes at high resolution was performed on tissue sections using an in situ multiplexed mRNA assay (10x Xenium)9,11. Kidney biopsies were distributed across eight Xenium slides (Fig. 1a). Samples from different diseases were distributed across the slides to minimize potential systematic variations (Extended Data Fig. 1a).

Fig. 1: Major cell types in the human renal cortex in health and inflammation.

a, Overview of the patient cohort from the Hamburg Glomerulonephritis Registry and the analysis performed. b, Schematic showing the cross-section of a human kidney tissue with the localization of various cell types in different regions. Inset: cross-section of a glomerulus (top) and tubulointerstitial region (bottom) and the cell types therein. c, DAPI-stained images from kidney biopsies of control (left) and a patient with ANCA-GN (right). The images show glomeruli and surrounding regions overlaid with cell boundaries determined using a cell segmentation algorithm. Localization of specific marker genes is indicated, with glomerular boundaries highlighted for clarity. d, Uniform Manifold Approximation and Projection (UMAP) of ~3.2 million cells retained after QC filtering. Cells are colored based on their annotated cell type. e, Stacked bar plot showing the proportions of different cell types from the complete biopsy tissues across the four disease conditions. f, images from control (left four images) and ANCA-GN (right four images) kidney biopsies showing a glomerulus and tubulointerstitial region. For each region, the adjoining plots show the segmented cells color coded according to their cell-type annotation, illustrating cellular composition differences between conditions in different regions of the tissue. aGBM, anti-GBM antibodies; ATL, ascending thin limb of loop of Henle; bx, biopsies; cDC, conventional dendritic cell; CNT, connecting tubule; cycNKC/T, cycling NK cytotoxic T cell; DCT, distal convoluted tubule; DTL, descending thin limb of loop of Henle; FIB, fibroblast; Fib. MC, fibrotic mesangial cell; IC, intercalated cell of collecting duct; KEGG, Kyoto Encyclopedia of Genes and Genomes; MAC, macrophage; MC, mesangial cell; MDC, monocyte-derived cell; Mono, monocyte; N, neutrophil; NEU, neuronal cell; PpE, papillary tip epithelial cells abutting the calyx; pDC, plasmacytoid dendritic cell; PL, plasma cell; POD, podocyte; PT, proximal tubule; TAL, thick ascending limb of loop of Henle; VSM/P, vascular smooth muscle cells or pericytes.

Table 1 Baseline characteristicsPanel design to identify cells and disease progression

Our first step was to design a comprehensive gene panel for the identification of different kidney cells (Fig. 1b), immune cells and genes that may be important in glomerulonephritis at the molecular level. We started with a detailed literature review to identify cell-type-specific markers that would enable the identification of kidney and immune cells, including subtype populations12,13,14. Next, we included genes that are: (1) implicated in the pathogenesis of cGN; (2) essential for pathways involved in leukocyte infiltration and activation; (3) associated with cytokine and chemokine signaling and inflammation from the Kyoto Encyclopedia of Genes and Genomes15, Reactome16 and Gene Ontology17 databases; and (4) differentially expressed between normal and inflamed glomerular regions, based on our recently published spatial transcriptomics data from patients with ANCA-GN and controls18. These steps resulted in the selection of 480 genes that were used in this study (Supplementary Table 1).

Segmentation and classification of kidney and immune cells

To localize cells and determine their boundaries, we first segmented them using Baysor19, resulting in the identification of 3,218,210 filtered cells that passed our quality control (QC) metric, with between 303,732 and 501,421 cells in each slide (Extended Data Fig. 1b,c). The cells expressed a median of 24 genes (control: 14; SLE: 25; ANCA-GN: 26; and anti-GBM: 27) and 36 median transcripts (control: 21; SLE: 41; ANCA-GN: 39; anti-GBM: 41) (Extended Data Fig. 1d). The median segmented cell diameter was 7.7 μm2 (Extended Data Fig. 1e). We compared Baysor-based segmentation with competing approaches (Xenium’s default segmentation and Cellpose) and found that Baysor cell segmentation followed by Xenium’s default, expansion-based cell segmentation achieved the lowest signal of negative markers compared to Cellpose (Extended Data Fig. 2a). Transcript assignment confidence was comparable across disease conditions and was not substantially affected by tissue region density (Extended Data Fig. 2b–d). We detected transcripts of cell-type-specific marker genes in both control and diseased glomeruli (Fig. 1c). These included markers for podocytes (PODXL), mesangial cells (PDGFRB) and PECs (CLDN1). In the tubulointerstitial regions, we found transcripts specific to proximal tubule cells (GPX3) and collecting duct cells (AQP2). Glomerulonephritis samples additionally showed expression of collagen (COL1A1) and immune cell markers for monocytes or macrophages (LYZ and CD163) and T cells (CD3D, CD3E and CD3G) (Fig. 1c).

Subsequently, we used a logistic regression classifier to predict cell types with a median confidence >0.89 across cell types (Fig. 1d and Extended Data Fig. 2e). The cell types showed differential enrichment of cell-type-specific marker genes (Extended Data Fig. 2f) and exhibited some of the expected differential abundance across diseases, such as the loss of podocytes and the increase of innate and adaptive immune cells in RPGN versus healthy controls (Fig. 1e). Exemplary localization of cell types within control and ANCA-GN are presented in Fig. 1f, together with the corresponding hematoxylin and eosin (H&E)-stained images, highlighting that glomerular and periglomerular regions are affected by cellular changes in disease. In summary, segmentation and classification resulted in a very high-quality set of >3 million spatially resolved renal and immune cells.

Defining and analyzing the glomerular and periglomerular niches

As, in glomerulonephritis, the glomerulus and the surrounding tubulointerstitial tissue might be most informative for understanding the immunopathogenesis of the disease (Fig. 1f), we focused on these areas. To obtain reliable annotation of the glomerular niches, we performed an automatic spatial domain annotation by employing NichePCA20 (Fig. 2a). The boundaries of the glomerular niches were subsequently extended by a 100-µm perimeter defining the periglomerular niches (Fig. 2b). We termed individual glomeruli and their associated periglomerular niche regions of interest (ROIs). Examples of identified glomerular and periglomerular niches are illustrated in Fig. 2c. Areas outside of glomerular and periglomerular niches were annotated as the tubulointerstitial domain. Figure 2d illustrates examples of these areas for each clinical condition, with cells colored by their annotated types. Control and, to some extent, SLE samples showed healthy glomerular structures with a monolayer of PECs around the glomeruli and low numbers of immune cells. In the ANCA-GN glomerulus, an increased number of PECs formed a crescentic structure inside the glomerulus. ANCA-GN and anti-GBM samples also showed increased accumulation of immune cells. A systematic comparison of the spatial accumulation of different cell types is shown in Fig. 2e and Extended Data Fig. 3a,b. In the glomerulus, podocytes constituted a median of 30% of cells in controls, which progressively decreased in SLE (25.9%) and ANCA-GN (25%) and was lowest in anti-GBM (23.8%) (Fig. 2e and Extended Data Fig. 3b).

In contrast, the percentage of PECs in the glomerulus exhibited the opposite trend, showing an increase in the disease conditions (control: 6%; SLE: 8%; ANCA-GN; and anti-GBM: 15%) (Fig. 2e and Extended Data Fig. 3b). In addition, there was a decrease in the percentage of mesangial cells (control: 17.9%; SLE: 10.2%; ANCA-GN: 7.1%; and anti-GBM: 4.3%) and a concomitant increase in fibrotic mesangial cells (control: 1.6%; SLE: 2%; ANCA-GN: 6.4%; and anti-GBM: 8.8%) in the glomerulus. The fibrotic mesangial cells display a distinct fibrotic signature, characterized by increased production of extracellular matrix (ECM) components such as collagens and fibronectin (Extended Data Fig. 3c–e). These cells have gene expression similar to mesangial cells (Extended Data Fig. 3c), but also express fibrosis-related genes such as COL1A1 and FN1 (Extended Data Fig. 3d,e) and their abundance is inversely correlated to that of mesangial cells (Extended Data Fig. 3f).

Fig. 2: Spatial transcript analysis of renal compartments in glomerulonephritis.

a, Workflow of the NichePCA algorithm used for automated annotation and spatial mapping of glomeruli. b, Schematic illustration of a glomerulus and its periglomerular niche, defined by expanding the glomerular boundaries to include surrounding regions. c, Representative H&E image showing glomeruli (blue) as defined by NichePCA, their corresponding periglomerular regions (green) and the remaining tubulointerstitial area (orange). d, Representative images highlighting the spatial localization of various kidney and immune cell types. The top panels show the spatial localization of various kidney and immune cells within the glomerular and periglomerular regions, whereas the bottom panels depict the same in tubulointerstitial regions across the four disease conditions. e, Stacked bar plots showing the proportions of different cell types across the four disease conditions within three spatial niches: the glomerular niche (left), periglomerular niche (center) and tubulointerstitial areas (right).

A decline in the proportion of glomerular endothelial cells (ECs) was also observed. Among immune cell types, a marked increase in innate immune cells was seen. Macrophages accounted for <0.5% of all glomerular cells in control samples which increased to 2% in SLE and ANCA-GN and 7.8% in anti-GBM (Fig. 2e and Extended Data Fig. 3b). Other innate immune cells were also increased within the glomerular niches. Notably, there was negligible increase in adaptive immune cells such as T cells, natural killer (NK) cells, B cells or plasma cells in these regions (Fig. 2e and Extended Data Fig. 3b).

By contrast, the periglomerular niches showed an increased accumulation of both innate and adaptive immune cells. T cells, which in control samples constitute about 1.1%, increased to a median of 2.5%, 4.8% and 7.4% in SLE, ANCA-GN and anti-GBM, respectively (Fig. 2e and Extended Data Fig. 4a,b). There was also a rise in the numbers of NK cells, B cells, plasma cells and several innate immune cells in the periglomerular area. In parallel with the periglomerular niches, tubulointerstitial regions showed enhanced infiltration by both innate and adaptive immune cells. T cells, for instance, increased from 1.13% in control samples to median proportions of 2.28% in SLE, 3.82% in ANCA-GN and 4.4% in anti-GBM (Extended Data Fig. 4c,d). In addition, NK cells, B cells, plasma cells, fibroblasts and other innate immune populations also increased in proportions in the tubulointerstitial regions (Extended Data Fig. 4d). Notably, we were not able to detect major differences between the periglomerular and remaining tubulointerstitial area (Extended Data Fig. 4).

Taken together, we observed an increased accumulation of immune and fibrotic cell types in all three diseases, whereas the spatial accumulation of the immune cells showed distinct characteristics. The glomerular compartment in these diseases exhibited a selective accumulation of innate immune cells, with a notable absence of adaptive immune cell infiltration. However, the periglomerular and tubulointerstitial domains exhibited increased innate and adaptive immune cells.

Common trajectory of glomerular crescent formation

As outlined in the previous section, the different diseases showed qualitatively similar cellular changes across the three compartments, albeit with differing median quantities. These findings suggest that the path to glomerulonephritis might be common across ANCA-GN, SLE and anti-GBM, whereas temporal stages and the speed of progression might be distinct. To find common trajectories and identify the molecular and cellular components of progression, we analyzed the trajectory of crescent formation based on a principal component analysis (PCA) of 782 combined unique ROIs (glomeruli plus periglomerular area) (Extended Data Fig. 5). The first principal component (PC1) of the ROIs showed a cellular signature of crescent formation, including positive correlations with immune cells, PECs and fibroblasts, while negatively correlating with podocytes and ECs (Extended Data Fig. 5a,b). In line, ROIs from controls displayed the lowest PC1 values and, from anti-GBM disease, the highest PC1 values (Fig. 3a,b).

Fig. 3: Trajectory of the immune cell ecosystem in glomerulonephritis.

a, PCA plot showing 782 unique ROIs, each comprising a distinct glomerulus and its associated periglomerular area. ROIs are colored by clustering (left) and by disease condition (right). b, Distribution of PC1 values assigned to the ROIs across the four disease conditions. c, Correlation analysis of the median PC1 and kidney function (eGFR) of each sample (Spearman’s correlation, two-sided). d, Representative H&E images showing glomeruli from each of the four pseudotime quadrants, highlighting morphological variations: cluster 1: control (left); cluster 2: ANCA-GN (middle left); cluster 3: ANCA-GN (middle right); cluster 4: ANCA-GN (right). e, Change in the percentage distribution of different cell types within glomerular regions (left) and periglomerular regions (right) across the crescent formation trajectory (PC1).

To further understand the molecular and cellular characteristics of the spatiotemporal progression into crescents, we clustered the ROIs using PCs, revealing four clusters (C1–C4) containing 194, 220, 210 and 158 ROIs, respectively, each showing distinct distributions of conditions (Extended Data Fig. 5c). The first cluster was dominated by control ROIs (89.69%). The second cluster showed a mixed distribution, comprising control (27.73%), SLE (57.27%), ANCA-GN (14.55%) and anti-GBM (0.45%) ROIs. In contrast, the third and fourth clusters were primarily characterized by ANCA-GN and anti-GBM (Extended Data Fig. 5c). We compared the gene expression profiles of clusters to find marker genes. The ROI clusters showed different marker genes (C1: 75; C2: 21; C3: 19; and C4: 170) (Extended Data Fig. 5d,e). Among the top 10 genes enriched in clusters 3 and 4 were several immune-related cell type and fibrosis-associated genes (immune: MZB1, LYZ and S100A9; fibrosis: FN1, LUM and TIMP1)21,22. These two clusters also showed a higher enrichment of several terms related to fibrosis and ECM remodeling, with a higher enrichment in cluster 4 (Extended Data Fig. 5f). It is interesting that the results of the PCA closely resembled those of an additional diffusion pseudotime analysis, observing a significant correlation (Spearman’s r = 0.80) between PC1 and the pseudotime path (Extended Data Fig. 5g–l). This suggests that PC1 captures the common, progressive formation of crescents across the diseases, rather than the disease-specific progression time, which is partially captured in the other PCs.

Analysis of individual patient biopsies showed a large variation in the PC1 values of different ROIs from the same patient, several times spanning three clusters (Extended Data Fig. 6a). We computed the median PC1 value across all ROIs per patient and correlated that with other clinical data. The median PC1 value showed an inverse correlation with kidney function (estimated glomerular filtration rate (eGFR)) at the time of biopsy (Fig. 3c and Extended Data Fig. 6b). Among patients with ANCA-GN, the median PC1 values also positively correlated with the ANCA renal risk score, linking PC1 to clinical severity (Extended Data Fig. 6c). Differential gene expression analysis between ROIs in cluster 1 (preserved glomeruli) and those in clusters 2, 3 and 4 (more affected glomeruli) showed higher expression of inflammatory and fibrotic genes (for example, C3, LUM and FN1) in clusters 2–4. Conversely, genes associated with normal kidney function, such as SLC12A1, were expressed higher in cluster 1 ROIs (Extended Data Fig. 6d). The expression of these genes showed variation between ROIs within the same patient (Extended Data Fig. 6e,f). Examples of ROIs within each cluster are presented in Fig. 3d and indicate progressive glomerular pathology. The change in the percentage distribution of different cell types within the glomerular and periglomerular regions across the crescent formation trajectory (PC1) is shown in Fig. 3e.

These results indicate that SLE, ANCA-GN and anti-GBM share a common cellular and molecular trajectory of crescent formation, which might contain shared intercellular and intracellular signaling pathways. In this context, it needs to be pointed out that the results from anti-GBM samples (n = 6) have restricted power compared to SLE (n = 19) and ANCA-GN (n = 32). The common crescentic path dominated both the pseudotime and the PC-based analyses, but the data also underscore disease-specific temporal changes. However, these data are derived from a cross-sectional study and need to be distinguished from time course analysis, which could provide true information about actual temporal changes.

Intercellular signaling of glomerular crescent formation

To elucidate the intercellular signaling pathways that underlie the RPGN development of glomerulonephritis, we performed interactome analyses using CellChat v.223, which provides a framework for inferring cell interaction probabilities by integrating the ligand–receptor expression data. Importantly, CellChat v.2 uses the spatial proximity between cells and a maximum diffusion threshold of 250 µm (ref. 23). We focused on the regions of biggest change, the glomerular and periglomerular regions. In general, cGN samples were found to have a higher number of cell–cell interactions compared to control samples. A comparison of the cell-interaction networks between cGN and control samples revealed that PECs exhibited increased interactions with neighboring cell types in glomerulonephritis (Fig. 4a). In addition, analysis of affected glomeruli revealed a significantly higher proportion of PECs in the crescentic regions compared to noncrescentic regions (Extended Data Fig. 7a,b). Given the increased signaling activity originating from PECs, the high abundance of PECs in crescents and their potential central role in crescent formation, we investigated which pathways result in PEC activation. We filtered all significant cell–cell interactions in cGN and identified PDGF, TGFβ, FASL and SLIT pathways showing activity toward PECs (Fig. 4b). The primary source of PDGF ligands to PECs were ECs followed by mesangial cells, vascular smooth muscle cells or pericytes, PECs and macrophages. TGFβ was the other key signaling pathway targeting PECs, signaling from fibrotic mesangial cells, ECs, podocytes, PECs, fibroblasts and different immune cells (macrophages, monocytes, T cells, NK cytotoxic T cell (NKC/T) and B cells) (Fig. 4b). The interaction maps for the three cGN diseases separately showed similar results (Extended Data Fig. 7c,d). Cell interaction analysis of the tubulointerstitial regions revealed increased signaling from proximal tubules and fibroblasts to different immune cell types (Extended Data Fig. 7e).

Fig. 4: Mapping the immune–epithelial cell interactions in glomerular crescent formation.

a, Circle plot showing cell–cell interactions in cGN compared to control samples. The interaction edges are colored the same as the source cell type whereas edge weights are proportional to the interaction strength. b, Heatmap showing the signaling pathways that have increased activity directed toward PECs in cGN and the source cell types producing the corresponding ligands, colored by the interaction probability. c, Scores for PDGF and TGFβ pathway activation in PECs plotted along pseudotime (lines show quadratic regression fits with 95% confidence intervals). d, Schematic representation of the PDGF signaling pathway, highlighting the genes upregulated in PECs in an snRNA-seq dataset from samples from patients with ANCA-GN (n = 11) over controls (n = 4). e, Multiplexed protein staining on a Xenium slide showing overexpression of Ki-67 in cGN samples (representative images of eight samples: two control (top left), three ANCA-GN (bottom left), two SLE (top right) and one anti-GBM disease (bottom right); representative of two independent experiments). f, Expression of Ki-67 protein in PECs along PC1. The primary axis corresponds to fluorescence intensity values of Ki-67. The secondary axis curve shows the change in percentage of Ki-67-positive cells per glomerulus. g, Schematic representation of the TGFβ signaling pathway, highlighting the genes from this pathway upregulated in PECs in an snRNA-seq dataset from samples from patients with ANCA-GN (n = 11) over controls (n = 4). h, Scatter plot showing correlation of TGFβ pathway score with fibrosis score in PECs (top) and the patient median TGFβ pathway score to the individual kidney function (bottom) (Spearman’s correlation, two-sided). i, Schematic illustrating the role of PDGF-mediated and TGFβ-mediated activation of PECs in crescent formation and glomerulosclerosis in cGN.

PDGF induces PEC proliferation and crescent formation

We next investigated the dynamics and effects of early PDGF activation in PECs in glomerulonephritis. Therefore, we quantified the activation levels of the PDGF pathway in PECs by creating a pathway-specific score. We used the Reactome gene set for PDGFR signaling in disease (curated gene set collection (C2) from the Molecular Signatures Database, MSigDB)24. We plotted this PDGF receptor (PDGFR) score in PECs against the crescent trajectory (PC1) from our analysis (Fig. 4c). The PDGFR score exhibited an increase at early to medium progression stages and decreased at advanced stages.

To validate the activation of PDGF pathways observed in PECs within the spatial transcriptomics dataset, using an independent orthogonal dataset with higher transcript resolution per cell, we performed single nuclear transcriptome sequencing (single nuclear RNA sequencing (snRNA-seq)) from kidney biopsies of patients with ANCA-GN (n = 11) and controls (n = 4), resulting in 136,376 nuclei after QC filtering (Extended Data Fig. 8a). Cell types were predicted using a logistic regression classifier (Extended Data Fig. 8b,c). Genes associated with the PDGF pathway were found to be upregulated in PECs from ANCA-GN compared to controls, further confirming the activation of these pathways in ANCA-GN (Fig. 4d). Genes transcriptionally upregulated downstream of PDGF included those involved in regulating proliferation, such as NRG1 (ref. 25) and genes associated with fibrosis such as ADAMTS1 (ref. 26) (Fig. 4d).

To investigate whether high PDGF signaling results in proliferation of PECs, we investigated the presence of Ki-67 protein along the crescent trajectory (PC1). To this end, we combined Xenium transcript analysis with Akoya PhenoCycler-based spatial proteomics of the same slide for eight samples (two controls, three ANCA-GN, two SLE and one anti-GBM disease; Fig. 4e). To align spatial transcriptomic and proteomic images, we registered the Akoya PhenoCycler and Xenium DAPI signals using VALIS27. Ki-67 intensity levels were assigned to PECs within all ROIs based on the cell boundaries obtained by our cell segmentation. Thus, we were able to identify PECs based on their transcriptional profile, align glomeruli according to their PC1 trajectory and determine proliferation by Ki-67 quantification. These analyses confirmed that PEC proliferation followed similar dynamics (Fig. 4f) to PDGF signaling in PECs (Fig. 4c).

TGFβ signaling results in glomerulosclerosis

Next, we determined the functional effects of TGFβ signaling on PECs in advanced crescents. We used the human gene set of the TGFβ signaling signature (Hallmark gene set collection H from MSigDB24). The TGFβ score showed a continuous increase over the crescent trajectory (PC1) (Fig. 4c). For TGFβ, snRNA-seq of differentially expressed genes from ANCA-GN versus control PECs revealed upregulation of the TGFβ pathway genes that play crucial roles in fibrosis (COL4A1 and NOX4) (Fig. 4g). COL4A1 encodes type IV collagen, a key component of the ECM that contributes to ECM stiffening and tissue remodeling28. NOX4 is a major source of reactive oxygen species and promotes oxidative stress, driving fibrosis and ECM deposition29,30. The role of NOX4 in renal fibrosis is complex and has also been shown to have a protective role31. Other genes activated by the TGFβ pathway are BMPR1A and BMPR2, which code for receptors for BMP4, a ligand within the TGFβ superfamily, and NCOR2, coding for a transcriptional corepressor that modulates fibrosis32.

Finally, we aimed at understanding whether TGFβ signaling in PECs could indeed result in glomerulosclerosis in human glomerulonephritis. The TGFβ signaling score correlated with a fibrosis score (Fig. 4h, upper panel) and inversely correlated with kidney function (eGFR) (Fig. 4h, lower panel). These findings support the hypothesis that TGFβ signaling is activated in RPGN and contributes to glomerulosclerosis and kidney failure.

To further investigate the activation of PECs in RPGN, we performed immunofluorescence staining of human control samples and kidney samples of patients with ANCA-GN. PECs were defined as claudin-1+ and synaptopodin−. These analyses confirmed prominent MAPK signaling (pERK1/2) in PECs from patients with ANCA-GN and PEC activation (CD44) compared to control PECs (Fig. 5a). These findings indicate that PDGF signaling acts on PECs and induces cell activation and proliferation in patients with RPGN. Immunofluorescence staining confirmed activation of the TGFβ signaling pathway by prominent SMAD3 phosphorylation and nuclear expression in PECs of patients with ANCA-GN (Fig. 5b).

Fig. 5: Immunohistochemistry of PECs and intracellular signaling molecules.

a, Exemplified presentation of immunofluorescence staining showing prominent phospho-ERK1/2 and CD44 positivity in claudin-1+ or synaptopodin− PECs in human kidney biopsies of patients with ANCA-GN (bottom) compared to healthy control tissue (top). b, Exemplified presentation of immunofluorescence staining showing prominent phospho-SMAD3 staining in PAX8+ or synaptopodin− PECs in human kidney biopsies of patients with ANCA-GN (bottom) compared to healthy control tissue (top) in exemplified immunofluorescence staining. Representative staining of four controls and four patients with ANCA-GN is shown.

To assess the generalizability of our findings, we analyzed publicly available single-cell (sc)RNA-seq or snRNA-seq data from the Kidney Precision Medicine Project, which includes biopsies from patients with nonimmune-mediated kidney diseases13 (Extended Data Fig. 9a). We found that PDGF and TGFβ pathway activation in PECs from patients with ANCA-GN in our snRNA-seq dataset was substantially higher than that seen in acute kidney injury or chronic kidney disease compared to respective controls (Extended Data Fig. 9b–d).

Functional role of PDGF and TGFβ signaling in experimental GN

The analysis in human RPGN highlights that, first, PDGF induces cell proliferation of PECs and crescent formation and, later, immune and kidney cell-derived TGFβ results in PEC-mediated and fibrotic, mesangial cell-mediated glomerulosclerosis (Fig. 4i).

Next, we aimed to identify potential treatment options for targeting PDGF and TGFβ signaling. We therefore performed digital pharmacology18. Specifically, we calculated TGFβ and PDGF pathway scores and matched upregulated pathway genes against DrugBank33 to identify drugs targeting these pathways, prioritizing inhibitors. We identified nintedanib34,35 with the highest score for blocking PDGF signaling in PECs. In addition, fostamatinib was identified as a potential blocker of the TGFβ pathway. Although primarily targeting SYK, fostamatinib also exhibits off-target inhibition of TGFβ receptor 1 (ref. 36).

For functional validation of these findings, we used a well-established mouse model of cGN (also referred to as nephrotoxic nephritis)37. To block the PDGF pathway, after induction of cGN, mice were treated with nintedanib (oral (p.o.)) from day 5 to day 9 (Fig. 6a). At day 10, renal histology showed reduced crescents (Fig. 6b,c), accompanied by reduced ERK1/2 phosphorylation and decreased activation and proliferation of PECs (Fig. 6d,e).

Fig. 6: PDGF and TGFβ blockade in experimental crescentic glomerulonephritis.

a, Experimental setup of PDGF blockade with nintedanib in experimental cGN. b, Periodic acid–Schiff staining of kidney sections of the respective groups. c, Quantification showing reduction of crescents in mice treated with nintedanib (data pooled from two independent experiments): left, untouched; middle, cGN; right, cGN + nintedanib. d, Exemplified presentation of immunofluorescence staining showing prominent phospho-ERK1/2 and CD44 positivity in claudin-1+ or synaptopodin− PECs from cGN mice treated with nintedanib (bottom) in comparison to PBS (top). e, Low expression of the proliferation marker PCNA in claudin-1+ or synaptopodin− PECs in exemplified immunofluorescence staining of cGN mice treated with nintedanib (bottom) in comparison to phosphate-buffered saline (PBS; top). f, Experimental setup of TGFβ blockade with fostamatinib in experimental cGN. g, Periodic acid–Schiff staining of kidney sections of the respective groups: left, untouched control; middle, cGN; right, cGN + fostamatinib. h, Quantification showing reduction of glomerulosclerosis in mice treated with fostamatinib (data pooled from two independent experiments). i, Exemplified presentation of immunofluorescence staining showing less prominent phospho-SMAD3 positivity in PAX8+ or synaptopodin− PECs. j, Glomerular SMA staining in kidney sections of mice with fostamatinib-treated cGN (bottom) and c