Spatial transcriptomics reveals injury-responsive compartments and coordinated immune–fibrotic signaling in ANCA-associated renal vasculitis

1 Spatial transcriptomics reveals injury-responsive compartments and coordinated1 immune–fibrotic signaling in ANCA-associated renal vasculitis2 3 Authors4 Yucheng Tang1, Chunhua Zhu2, Shuang Chen1,2,3, Huimei Chen1,4,5, Hai Qian1,6, Aihua Zhang2,*, and5 Enrico Petretto1,4,5,*6 Affiliations7 1 Institute for Big Data and Artificial Intelligence in Medicine, School of Science, China Pharmaceutical8 University, Nanjing, Jiangsu, 211198, China;9 2 Department of Nephrology, Children’s Hospital of Nanjing Medical University, Nanjing 210008,10 China;11 3 Jiangsu Key Laboratory of Pediatrics, Nanjing Medical University, Nanjing, China;12 4 Programme in Cardiovascular and Metabolic Disorders, Duke-NUS Medical School, Singapore,13 Singapore;14 5 Centre for Computational Biology, Duke-NUS Medical School, Singapore, Singapore.15 6 Centre of Drug Discovery, State Key Laboratory of Natural Medicines, China Pharmaceutical16 University, Nanjing, Jiangsu, 211198, China;17 Yucheng Tang and Chunhua Zhu contributed equally to this work.18 Address for correspondence19 Enrico Petretto, Institute for Big Data and Artificial Intelligence in Medicine, School of Science, China20 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 2 Pharmaceutical University, Nanjing 210009, China; Programme in Cardiovascular and Metabolic21 Disorders, Duke-NUS Medical School, Singapore, Singapore; Centre for Computational Biology,22 Duke-NUS Medical School, Singapore, Singapore23 Email: [email protected] Aihua Zhang, Department of Nephrology, Children’s Hospital of Nanjing Medical University, Nanjing25 210008, China26 Email: [email protected] 28 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 3 Abstract29 ANCA-associated vasculitis (AAV) often presents with rapidly progressive glomerulonephritis, yet30 how immune and stromal programmes are organised within kidney tissue remains unclear. We31 applied spatial transcriptomics to renal cortex biopsies from patients with AAV of varying disease32 severity and normal-histology controls, generating a spatial map that localizes disease programmes33 in situ. Four disease-responsive compartments—immune/interstitial fibroblasts (IM/Fib), glomeruli,34 myofibroblasts, and vascular compartments—showed distinct compartment-specific signatures that35 correlated with histopathology at the patient level. We also identify coordinated immune–fibrotic36 signaling linked to severity. Within IM/Fib, the CXCR4–CD74 receptor complex co-localised with IgM⁺37 cells, and lumican (LUM) co-localised with collagen I/III; both were validated by immunofluorescence38 and associated with fibrotic injury. These tissue-anchored signatures constitute candidate39 diagnostic/prognostic biomarkers, and show the utility of this foundational spatial transcriptomics40 map to generate testable hypotheses which will guide validation in larger, stratified,41 treatment-annotated cohorts.42 43 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 4 Introduction44 Anti-neutrophil cytoplasmic antibodies (ANCA)-associated vasculitis (AAV) represents a group of rare45 systemic autoimmune diseases, characterised by inflammation of the small-to-medium-sized blood46 vessels1. Renal involvement occurs in >75% of patients, predominantly manifesting as rapidly47 progressive glomerulonephritis (RPGN)2. In pediatric AAV, renal involvement is more severe, with48 approximately 20%-35% of children progressing to renal failure, while another 20%-30% develop49 chronic kidney disease (CKD)3. However, children exhibit a higher capacity for renal function recovery50 compared to adults, which is potentially attributable to the greater plasticity of the immune system51 and the potential for earlier therapeutic intervention4.52 A central immunological mechanism of AAV involves the breakdown of immune tolerance to53 neutrophil granule proteins, primarily proteinase 3 (PR3) or myeloperoxidase (MPO), resulting in the54 production of ANCA5. These ANCAs activate neutrophils, promoting vascular injury through the55 formation of neutrophil extracellular traps (NETs), oxidative burst, and degranulation6. Activated56 neutrophils also release autoantigens into the extracellular space, where they are captured by57 dendritic cells (DCs) and presented to naïve T cells7. This leads to the expansion of effector T cells,58 which infiltrate renal tissue and contribute to crescent formation, interstitial fibrosis, and tubular59 atrophy8.60 Transcriptomic (single-cell and spatial) studies in AAV have identified key immune cell populations61 and provided insights into disease mechanisms. Single-cell analyses have highlighted62 disease-associated monocyte subsets (FCGR3A⁺ and FCGR1A⁺)9, SPP1⁺ lipid-associated macrophages63 that are associated with inflammation and fibrosis10, and cytotoxic T cells exhibiting pathogenic64 profiles in renal tissue11,12. Activation of the alternative complement pathway has been65 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 5 demonstrated in AAV13. Collectively, prior studies primarily focused on circulating immune cells,66 leaving unresolved how immune and stromal programmes are organised within kidney tissue and67 how this spatial organisation relates to injury severity and repair in AAV .68 Current AAV treatments rely on immunosuppressive therapy, typically cyclophosphamide with69 high-dose glucocorticoids, which improves survival but increases infection and malignancy risk14,15.70 The use of the anti-CD20 monoclonal antibody, rituximab (RTX), yields comparable renal outcomes,71 with more effective B-cell depletion and better ANCA seroconversion16. Avacopan is an oral C5a72 receptor antagonist that effectively blocks the activation of neutrophils and the release of NETs,73 thereby reducing inflammation and tissue damage in AAV17. It was authorised for adjunct use with74 cyclophosphamide or rituximab, combined with a glucocorticoid dosage lower than standard75 protocols. Yet 10–30% of patients fail to achieve remission, remain glucocorticoid dependent, or76 progress despite therapy18, and multiple immune-targeted agents are still in development12,19,20.77 Since current therapies improve survival yet leave a substantial fraction of patients with persistent78 activity, glucocorticoid dependence, and progression to CKD, there is a critical need for tissue-level79 biomarkers that quantify severity and resolve immune–fibrotic signaling in situ. We therefore80 hypothesized that spatially resolved transcriptomic signatures within the renal parenchyma would81 reveal physiologically relevant biomarkers and pathogenic circuits not detectable in blood. To test82 this hypothesis, we profiled rare, biopsy-limited renal cortex specimens from patients with AAV using83 10x Genomics Visium, defined AAV-responsive regions and constituent cell types, and linked regional84 programmes to histopathological severity. We identified cell-type-specific markers linked to disease85 pathology, and prioritized ligand–receptor interactions that serve as biomarkers of disease severity in86 renal biopsies and as potential therapeutic targets in ANCA-associated vasculitis.87 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 6 Results88 Spatial transcriptomic profiling of kidney biopsies in ANCA-associated vasculitis89 We categorized AAV patients into mild (n = 2) and severe (n = 3) groups and analysed kidney cortex90 biopsies alongside those of controls (n = 4) (Supplementary Table 1). The controls, obtained from91 patients with hematuria, showed normal renal histology without sclerosis or atrophy. In contrast,92 mild AAV biopsies displayed segmental sclerosis and tubular atrophy, while severe AAV consistently93 exhibited glomerular crescents, reflecting severity-dependent tissue damage (Fig. 1a). None of the94 patients had received treatment at the time of biopsy collection.95 We generated spatial transcriptomic (ST) profiles from kidney biopsies of control and AAV patients96 using the Visium platform, analysing 2-4 tissue sections per group (Fig. 1b). After standardizing97 quality control, we obtained 3,850 high-quality spots (1,413 in controls; 890 in mild AAV; 1,277 in98 severe AAV). Severe AAV samples exhibited increased transcript and gene counts per spot,99 accompanied by reduced mitochondrial content, consistent with altered composition and activation100 (Extended Data Fig. 1a).101 Using unsupervised clustering and marker gene expression, we identified 10 distinct cell types from102 the integrated dataset (Extended Data Fig. 1e, Extended Data Fig. 2 and Supplementary Table 2),103 which were visualized in a UMAP plot and spatially mapped across kidney sections (Fig. 1b,e). These104 included proximal tubule (PT), thick ascending limb (TAL), distal convoluted tubule (DCT), connecting105 tubule/principal cells (CT/PC), connecting tubule/intercalated cells (CT/IC), glomerulus (Glo), immune106 cells/interstitial fibroblasts (IM/Fib), vascular endothelial/smooth muscle cells (vEC/VSMC),107 myofibroblasts (Myo), and mixed tubules. Cell type distributions were consistent across patients, and,108 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 7 as expected, controls showed fewer IM/Fib cells and almost no Myo cells (Extended Data Fig. 1b).109 The IM/Fib and Myo clusters showed the most pronounced changes among cell types in both mild110 and severe AAV patients compared to controls (Fig. 1c), with progressive expansion correlating with111 increasing disease severity and associated pathological features such as immune infiltration and112 fibrosis. The IM/Fib cluster was most abundant in the severe AAV group. In contrast, glomerular (Glo)113 spots were reduced in AAV samples, especially in the mild group, suggesting a potential link to early114 glomerular sclerosis. While TAL, CT/PC, and mixed tubule populations showed some variability across115 conditions, their changes were less prominent. Notably, the mixed tubule cluster co-expressed116 markers from multiple cell types (e.g., MALAT1 for PC, SLC5A12 for PT, S100A9 for monocytes) (Fig.117 1d), had the lowest transcript and gene counts, and exhibited high mitochondrial content (Extended118 Data Fig. 1d), were flagged as low-quality by pre-specified QC thresholds (genes/UMIs/% mito) and119 excluded a priori from downstream analyses.120 We identified 2,865 upregulated differentially expressed genes (DEGs) across the 10 annotated121 kidney cell types (Supplementary Table 3). The top 20 DEGs showed distinct, cell-type-specific122 expression patterns (Fig. 1d, with strong concordance between our ST and scRNA-seq profiles from123 the Kidney Precision Medicine Project (KPMP) atlas21 (Supplementary Table 3). The IM/Fib and Myo124 clusters exhibited unique molecular signatures, including enrichment in complement pathway125 components (C7, C1QA, C1S, C1R, C3) and extracellular matrix organization (LUM, MMP7) (Extended126 Data Fig. 1f). Myo cells also showed upregulation of genes related to immune response (CXCR6),127 wound healing (TPM1), and cell junction assembly (CDH6). Our cell type annotations were further128 validated using scRNA-seq references from DISCO62 and the Kidney Cell Atlas (KCA)129 (https://www.kidneycellatlas.org/), confirming strong agreement with established marker gene130 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 8 expression (Extended Data Fig. 1c).131 This ST profiling of AAV kidney biopsies revealed the severity-dependent expansion of132 immune-fibroblast and myofibroblast populations, as well as reduced glomerular signatures and133 distinct cell-type-specific gene expression programmes enriched for complement activation and134 fibrosis-related pathways.135 Delineating pathogenic AAV-responsive regions during disease progression136 To characterise AAV-associated cell types in situ, we defined “disease-responsive” regions integrating137 changes in cell type proportions (Fig. 1c) with the expression of canonical AAV-related pathological138 marker genes (Supplementary Table 4 and Extended Data Fig. 3b). The AAV-responsive regions139 identify four key cell types – IM/Fib, Myo, Glo, and vEC/VSMC – as AAV-responsive (see Methods),140 which were consistently present across AAV patients, mirroring the distribution of injury markers141 (Extended Data Fig. 3b-d). These cell populations were spatially mapped across control, mild, and142 severe AAV groups (Fig. 2a), showing proximity within tissue sections and adjacent positioning in143 UMAP space (Extended Data Fig. 3a), which suggests shared transcriptional features. Consistent with144 our previous observations (Fig. 1c), IM/Fib and Myo populations expanded with increasing disease145 severity, while Glo proportions declined. Although vEC/VSMC cells were relatively sparse, they146 expressed immune-related genes (C1S, C1R, C1QA–C1QC, IGHA1, JCHAIN), injury markers (IGFBP6,147 CST3), and profibrotic genes (COL1A2, COL3A1), overlapping transcriptionally with both148 immune-fibrotic and glomerular compartments (Extended Data Fig. 3b). These four cell types, most149 strongly associated with AAV progression, were selected for focused downstream analyses.150 To assess whether AAV-responsive cell types are linked to disease pathology, we correlated cell type151 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 9 proportions from ST with histopathological indicators from diagnostic kidney biopsies (see Methods).152 Among all cell types, IM/Fib and Myo showed the strongest positive associations with key153 pathological features, including focal/global glomerulosclerosis, glomerular injury, and interstitial154 fibrosis (Extended Data Fig. 3e). IM/Fib proportions correlated strongly with both focal and global155 sclerosis, while Myo showed the highest correlation with global sclerosis (R = 0.95, p = 0.015) (Fig.156 2b). In contrast, non-responsive regions exhibited negative correlations with these pathological157 measures. While PT cells also showed positive associations with injury, they were not prioritized due158 to minimal expression of AAV-relevant markers (Extended Data Fig. 3b). Glo and vEC/VSMC159 populations displayed distinct correlation patterns: Glo was positively associated with crescentic160 glomeruli, whereas vEC/VSMC showed inverse correlations. Together, these findings highlight161 dynamic shifts in IM/Fib and Myo populations, which may reflect their association with glomerular162 injury and tubulointerstitial fibrosis in AAV .163 Markers from AAV-responsive regions correlate with pathological indicators164 We identified differentially expressed marker genes in AAV-responsive cell types (Glo, IM/Fib, Myo,165 and vEC/VSMC) that significantly correlate with pathological indicators (R² > 0.7, P < 0.05) (Fig. 2c,d166 and Supplementary Table 6). Tubular (e.g., UMOD, SLC12A3) and glomerular (e.g., PODXL, HTRA1)167 markers were markedly reduced in severe AAV, reflecting extensive loss of renal architecture. Notably,168 VEGFA and PLA2R1, podocyte-enriched genes22, showed strong inverse correlations with glomerular169 injury, highlighting podocyte damage in AAV . In IM/Fib populations, immunoglobulin genes and170 classical complement components (C1QA, C1R, C1S) were upregulated. ECM-related genes (LUM,171 DCN, SERPINF1, ADH1B) were also elevated, with ADH1B and LUM showing the strongest172 associations (R² > 0.97) with fibrosis and focal glomerulosclerosis, respectively. Markers in Myo and173 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 10 vEC/VSMC populations, including VCAN, MGP, COL4A1, TPM1 and C3; NNMT and SOD2; as well as174 SPARCL1 and TAGLN, were significantly upregulated in severe AAV and positively correlated with175 fibrotic and glomerular injury indices. C3 had the highest correlation with fibrosis, while MGP, NNMT,176 and SOD2 were linked with glomerular injury.177 These cell types and marker associations with renal injury were independently confirmed in a178 separate AAV ST dataset12 (see Extended Data Fig. 3f–j), supporting their potential as diagnostic179 biomarkers for AAV in renal biopsies.180 Immune activation and extracellular matrix remodeling in AAV-responsive regions181 We performed pairwise differential expression analysis across control, mild, and severe AAV groups182 in ten different kidney cell types. Disease-associated transcriptional changes were quantified by183 overlapping DEGs with cell-type-specific markers (Supplementary Table 2 and Fig. 3a).184 AAV-responsive regions showed a progressive increase in upregulated marker genes with disease185 severity, particularly in the IM/Fib compartment (severe vs control: 65%; severe vs mild: 48%). While186 most Glo markers were downregulated, a small subset was upregulated in severe AAV .Tubular187 markers were broadly downregulated, except for modest increases in PT and TAL-specific genes.188 To explore functional implications of these transcriptional changes, we performed Reactome-based23189 gene set enrichment analysis (GSEA) on the DEGs from AAV-responsive regions (severe vs control; Fig.190 3b and Extended Data Fig. 5a). We identified immune activation and ECM remodeling as the most191 significantly enriched pathways across all compartments. Core immune pathways, including192 complement activation, cytokine signaling, and interferon responses, were upregulated across the193 IM/Fib, Glo, Myo, and vEC/VSMC regions. However, the specific DEGs in these enriched pathways194 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 11 varied by cell type. Specifically, the IM/Fib region showed signatures of T cell activation and195 neutrophil degranulation, consistent with known immune infiltration in AAV24,25. In the Glo region,196 IL-4 and IL-13 signaling was enriched, suggesting a STAT6-mediatedimmune regulation and fibrosis,197 as previously reported26. ECM remodeling pathways, including collagen synthesis, proteoglycan198 deposition, and matrix degradation, were also broadly upregulated in all AAV-responsive regions (Fig.199 3b and Extended Data Fig. 5a). A conserved set of ECM-related genes (TIMP1, FN1, COL3A1, VCAN,200 COL6A3, COL4A1, TNC, MMP7, FBN1, COL1A2, and SERPINE1) was identified across compartments,201 while others were region-specific (e.g., LUM, DCN in IM/Fib; COL8A1, CD44 in Glo). Furthermore, the202 downregulated DEGs were primarily involved in development, transporter function, and protein203 metabolism, indicating impaired renal function. Correlating DEGs with pathology revealed that genes204 most strongly associated with clinical indicators (e.g., glomerular injury, fibrosis) were concentrated205 in IM/Fib, followed by vEC/VSMC, Glo, and Myo (Fig. 3c and Extended Data Fig. 5b). Notably, LUM206 (IM/Fib) correlated with focal glomerulosclerosis, while FBN1 and SLC12A3 (IM/Fib), SOD2 (Myo),207 C1R (vEC/VSMC), and COL8A1 (Glo) were most associated with global sclerosis and fibrosis.208 Together, these findings suggest that immune activation and ECM remodeling are central to AAV209 pathogenesis, with both shared and region-specific molecular programmes across kidney210 compartments.211 Network analysis reveals coordinated immune and fibrotic pathways in AAV-responsive212 compartments213 To identify coordinated transcriptional programmes in AAV-responsive regions, we used hdWGCNA214 on our ST data from severe AAV samples, which yielded 13 distinct gene co-expression modules215 (Extended Data Fig. 6a,b). Four of these (M2, M3, M8, M11) were enriched in AAV-responsive cell216 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 12 types: M2 and M11 in IM/Fib, M3 in Myo, and M8 in Glo (Fig. 4a). Among these, M11 was highly217 specific to IM/Fib, while M2 was also moderately active in Myo and Glo (Fig. 4b). M3 was mainly218 expressed in Myo, and M8 showed exclusive upregulation in Glo.219 We then assessed module activity using eigengenes and AddModuleScores derived from the top 100220 hub-ranked genes (kME) (see Methods for details), and used this metric to evaluate each module's221 association with AAV severity. Notably, M2, M3, and M11 activities increased with disease severity in222 their respective cell types, while M8 in Glo showed no such trend (Fig. 4c,d). In all IM/Fib spots, M2223 features increased linearly with disease severity, while M11 features were induced in disease states224 (Fig. 4d).225 Pathway enrichment and hub gene analysis highlighted M2 and M11 as key networks in IM/Fib (Fig.226 4e–h). Compared to M3 and M8 (Extended Data Fig. 6c–f), M2 was enriched for ECM organization,227 featuring hub genes such as COL1A1, COL1A2, COL3A1, LUM, PDGFRA, and LRP1. These genes are228 involved in collagen biosynthesis and TGF-β signaling, both of which are known drivers of fibrosis in229 AAV nephropathy27-29. Additionally, the identified response to the corticosteroid pathway may be230 associated with corticosteroid treatment in AAV patients30. M11 was related to immune activation,231 including antigen presentation (CD74), leukocyte chemotaxis (CXCR4), cytotoxicity, and apoptosis (Fig.232 4h). Cellular communication analysis suggests that CD74 and CXCR4 might cooperate in directing233 immune cell infiltration (Extended Data Fig. 7a–c).234 In summary, although the spatial resolution did not allow for the complete separation of immune235 and fibrotic cells, M2 and M11 showed overlapping spatial features (Fig. 4d and i–j), suggesting that236 immune cell infiltration and fibrotic remodeling occur in adjacent or colocalized regions within AAV237 kidneys.238 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 13 CXCR4-CD74 crosstalk in the IM/Fib region mediates leukocyte chemotaxis in AAV239 Integration of ligand–receptor and hdWGCNA analyses identified CXCR4 and CD74 as hub genes in240 module M11, associated with immune activation in the IM/Fib region (Fig. 4f). CD74 and CXCR4 form241 a receptor complex, and using NICHES for intercellular interaction analyses, we found that cell–cell242 communication in IM/Fib increased with AAV severity (Extended Data Fig. 7a), with the CD74–CXCR4243 axis showing one of the most pronounced activations in severe AAV (Extended Data Fig. 7b), and244 enriched for leukocyte chemotaxis, the top pathway in M11 (Fig. 4h and Extended Data Fig. 7c).245 While CD74 was consistently upregulated across all AAV groups, CXCR4 was selectively elevated in246 severe AAV (Fig. 5c). Spatial transcriptomics also revealed the co-expression or adjacency of CXCR4247 and CD74 in AAV kidneys, particularly in severe disease (Fig. 5a,b and Extended Data Fig. 8a,b). In248 severe AAV, the co-expression of CXCR4 and CD74 was strongly positively correlated (Fig. 5d),249 suggesting disease-phase-dependent induction of this signaling axis, which may mediate leukocyte250 chemotaxis and subsequent immune activation.251 Immunofluorescence (IF) analysis corroborated these transcriptomic findings. CD74 and CXCR4252 protein expression increased with disease severity, and colocalized near IgM-positive immune cells253 (Fig. 5e). IGHM, a marker of plasma/memory B cells exclusively expressed in AAV patients, was254 increased and positively correlated with CXCR4 in severe AAV (Extended Data Fig. 8c–f). Spatial255 proximity analysis revealed that CD74/CXCR4 cells were 10–24 μm from IgM+ cells in the renal256 tubulointerstitium (Extended Data Fig. 8g,h), suggesting that CD74/CXCR4 may signal locally to257 attract nearby immune cells (a process consistent with paracrine chemotactic recruitment). We also258 reported AAV severity-dependent increases in CXCR4 protein expression and, to a lesser extent, in259 CD74 (Fig. 5f). The co-localization of CXCR4 and CD74, assessed by Pearson’s coefficients, was260 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 14 significantly higher in severe AAV (P = 0.0002; Fig. 5g), confirming a severity-linked interaction.261 In summary, ST and IF analyses suggest that CXCR4–CD74 co-expression in IM/Fib regions is spatially262 linked to immune infiltrates and correlates with AAV severity, supporting their role in leukocyte263 chemotaxis and immune activation during disease progression.264 LUM–Collagen I/III co-localization correlates with AAV-associated renal fibrosis265 In gene co-expression module M2, we identified LUM (lumican) as a central hub gene, with pathway266 and ligand-receptor analyses suggesting its interaction with COL1A2 and COL3A1 via ITGB1,267 implicating a role in ECM remodeling (Fig. 4e,g and Extended Data Fig. 7d). Spatial transcriptomics268 revealed strong spatial co-expression between LUM and both COL1A2 and COL3A1 within the IM/Fib269 region, either in the same or adjacent tissue spots (Fig. 5h, I and Extended Data Fig. 8i–l). Expression270 levels of LUM, COL1A2, and COL3A1 increased progressively with AAV disease severity (Fig. 5j and271 Extended Data Fig. 8m), and all genes were significantly upregulated in AAV compared to controls272 (Fig. 5k and Extended Data Fig. 8n). These data suggest that LUM may contribute to pathological273 collagen deposition in AAV .274 Immunofluorescence staining confirmed increased expression and co-localization of LUM with275 collagen I and III in AAV kidneys. Lumican and collagen I protein levels were elevated in both mild and276 severe AAV, with significantly stronger co-localization in severe cases (P = 0.0037, Fig. 5l-n). Similar277 patterns were observed for collagen III, which also showed enhanced co-localization with LUM in278 severe AAV (P < 0.0001; Extended Data Fig. 8o–q).279 Together,these findings suggest that LUM is associated with fibrotic progression in AAV, potentially280 through spatial and functional interactions with fibrillar collagens involved in ECM accumulation.281 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 15 Discussion282 Prior AAV studies have largely profiled circulating and infiltrating immune cells, with less attention to283 how disease programmes are organised within the renal microenvironment. Extensive immune284 infiltration in glomeruli has been reported31, and scRNA-seq analysis has revealed prominent285 populations of infiltrating B cells and plasma cells implicated in local immune responses within the286 kidney32. Further integration of ST and scRNA-seq has characterised inflammatory niches and287 highlighted pathogenic T cells in the kidneys of patients with ANCA-associated glomerulonephritis288 (GN)12. Building on these studies that identified inflammatory niches, we provide a pilot spatial map289 of AAV kidneys that localises immune–fibrotic activity in situ, and links it to histopathological severity290 at the patient level.291 We defined four AAV-responsive renal regions, IM/Fib, Glo, Myo, and vEC/VSMC, and identified292 compartment-specific molecular signatures correlating with disease severity at the patient level.293 Among these, IM/Fib showed the most pronounced transcriptomic changes, including the highest294 number of differentially expressed genes, cell type markers, and strongest correlations with295 pathological indicators. The latter were evaluated at the patient level (to avoid pseudoreplication),296 i.e., for each cell type/region, we generated pseudo-bulk expression per patient and related these to297 histopathological indices using Spearman rank correlations with FDR control. Complement pathway298 activation was evident, with increased expression of classical pathway components C1S and C1R in299 IM/Fib and C3 in Myo. While prior work emphasises alternative-pathway activation in AAV33, our300 increased classical components within IM/Fib are consistent with complement crosstalk in301 immune-complex-rich niches, potentially contributing to glomerular injury and thromboembolic risk302 in AAV33-35. We also identified ADH1B as a potential fibrosis-associated marker in IM/Fib, with303 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 16 minimal expression in controls and strong patient-level correlation with fibrosis scores. Given304 ADH1B’s roles in aldehyde/retinoid metabolism, we speculate that altered aldehyde handling could305 contribute to carbonyl stress and fibroblast activation and ECM deposition36,37 in AAV . This306 observation remains hypothesis-generating and warrants validation at the protein/activity level and307 dedicated functional analyses.308 The immune response-associated module (M11) also includes immunoglobulins, HLA family309 members, and complement components that are enriched among cell type-specific DEGs in the310 IM/Fib region of the severe AAV group. In M11, the most prominent pathways play a central role in311 the development of kidney inflammation and renal injury in AAV .During leukocyte activation, ANCA312 binds to PR3 and MPO, leading to excessive neutrophil activation. These cells release large amounts313 of pro-inflammatory cytokines, damaging the vascular endothelium7. Activated monocytes and314 dendritic cells present antigens via MHC II (CD74, HLA-DBP1, HLA-DRB1, HLA-DRA, HLA-DQA1),315 activating T cells and establishing a persistent autoimmune response6. Vascular endothelial cells and316 immune cells also secrete chemokines (CCL19, CCL5, CCL21), attracting T cells and dendritic cells to317 accumulate at the sites of inflammation38,39. These processes collectively sustain tissue injury and318 loss of renal function. In addition, negative regulation of neutrophil and lymphocyte apoptosis may319 lead to sustained tissue damage (Supplementary Table 7).320 CXCR4 and CD74 were co-expressed and co-localised within IM/Fib neighbourhoods. CXCR4 is a vital321 chemokine receptor implicated in leukocyte trafficking and fibrosis40,41. While prior studies support322 its involvement in kidney fibrosis, the potential interaction between CXCR4 and CD74 in AAV has not323 been reported. Rather than a classical ligand–receptor pair, CD74 and CXCR4 form a receptor324 complex that can be engaged by macrophage migration inhibitory factor (MIF)42 and tuned by325 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 17 CXCL12, promoting chemotaxis and inflammatory signaling43. Although MIF did not increase326 uniformly in our dataset, the spatial adjacency of CXCR4/CD74 with IgM⁺ cells and the severity-linked327 co-localisation are consistent with chemotactic recruitment. These features position the328 CXCR4–CD74 complex as a candidate tissue biomarker and a putative therapeutic axis in AAV,329 pending functional validation. We will quantify MIF/CXCL12 protein and test perturbations (e.g.,330 blocking antibodies) to establish causality. However, further investigation is needed to identify the331 immune cell types involved and clarify whether these pathways also contribute to fibrogenesis.332 In the ECM remodeling module (M2), key pathways including ECM organization, collagen metabolic333 processes, and TGF-  signaling were identified as drivers of fibrotic tissue injury27,29,44, which is334 consistent with findings from DEGs in the IM/Fib region of the severe AAV group. Beyond these335 major fibrosis-related pathways, we also identified pathways associated with the humoral immune336 response mediated by circulating immunoglobulin (C1R, C1S, C7, SERPING1, CD81, SVEP1) and the337 response to viruses (IFITM3, IFI27, TPT1, ISG15, DDIT4) within the M2 network (Supplementary Table338 8). Interferon-stimulated genes (ISGs) were highly expressed in AAV, indicating abnormal activation339 of innate immunity and the maintenance of a pro-inflammatory and pro-fibrotic environment45,46.340 Hence, these findings suggest a crosstalk between fibrosis, innate immunity, and chronic341 inflammation within the AAV-responsive region.342 A second key finding was the identification of LUM (lumican), a small leucine-rich proteoglycan and343 structural ECM component, as the hub gene in an ECM remodeling module (M2). LUM was strongly344 co-expressed with collagen genes COL1A2 and COL3A1 and spatially co-localized in fibrotic IM/Fib345 regions. These collagens are early markers of fibrosis47, and integrin-mediated interactions between346 LUM and collagen may contribute to excessive ECM deposition. Immunofluorescence confirmed347 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 18 LUM-collagen I/III co-localization and disease severity-linked expression. Interestingly, both LUM and348 collagen were upregulated even in mild AAV, suggesting early involvement in pathogenesis. Although349 LUM is not specific to AAV (being implicated in other kidney fibrotic diseases)48-50, its early and350 progressive upregulation in AAV suggests it as a potential biomarker for incipient fibrosis, when used351 in combination with other molecular indicators.352 Moreover, these cell type-specific markers, including CXCR4, CD74, and LUM, may offer added value353 in enhancing diagnostic granularity and enabling patient stratification in renal biopsies from AAV354 patients. Unlike conventional histopathological assessments, these transcriptomic signatures showed355 correlations with disease severity and pinpointed specific pathological features, including immune356 infiltration and fibrosis. For instance, LUM and its co-localization with COL1A2/COL3A1 may reflect357 fibrotic remodeling, while CXCR4–CD74 co-expression could be indicative of localized immune358 activation. While preliminary, these findings suggest that integrating such molecular markers with359 histopathology may refine staging and guide therapies, pending validation in larger independent360 cohorts. We postulate that these markers constitute candidate tissue biomarkers with361 complementary roles: diagnostic (refining lesion classification in biopsies) and prognostic (risk362 stratification by immune activation and fibrotic burden). As the cost of spatial transcriptomics363 (ST)-based assays continues to decline, we envisage a composite spatial severity score (e.g., IM/Fib364 M11 and M2 module activity combined with marker gene load) to complement conventional365 histopathology. In addition to the IM/Fib region, we identified additional disease-associated markers366 in the Myo, Glo, and vEC/VSMC compartments, many of which correlated with pathological features,367 expanding the landscape of potential diagnostic markers.368 Nevertheless, we acknowledge several limitations in our study. First, AAV rarity and the challenge of369 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 19 obtaining high-quality diagnostic biopsies resulted in a small cohort; accordingly, suggesting CXCR4,370 CD74, and LUM as candidate biomarkers remains preliminary. Validation in larger, independent371 cohorts will be necessary to establish their diagnostic and prognostic utility. Second, 10x Visium372 captures transcripts from multiple cells per spot, which limits single-cell resolution and introduces373 potential ambient RNA contamination and spot-mixing. We mitigated these by conducting374 patient-level analyses (i.e., pseudo-bulk per cell type/region with Spearman correlations, and375 implementing Benjamini–Hochberg FDR control), applying reference-guided cell-type enrichment376 analyses using several datasets (e.g., KPMP21, DISCO62 and KCA, and Giotto hypergeometric test), and377 confirming key findings by conducting blinded immunofluorescence quantification. Third, our cohort378 may not equally represent PR3- and MPO-ANCA subtypes or adult AAV cases, limiting generalizability.379 None of the patients had received treatment at the time of biopsy collection (hence minimizing this380 confounder), future studies should expand to larger, stratified cohorts based on serotype and age to381 replicate and extend these findings. Finally, finer cellular resolution will benefit from next-generation382 platforms such as Xenium, MERFISH, or seqFISH+51-53 and validation of main findings in larger,383 treatment-annotated cohorts is warranted.384 In conclusion, we provide a foundational spatial map of AAV kidneys, which extends the previous385 immune cell subsets analysis12,32. These findings allowed us to generate testable hypotheses and a386 framework to refine biopsy classification and can help advance our understanding of AAV387 nephropathy. Future work integrating higher-resolution spatial transcriptomics with multi-omics388 approaches could further delineate cell-type-specific responses, enabling precise mapping of389 immune, fibrotic, and parenchymal cell dynamics in AAV nephropathy.390 391 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 20 Methods392 Ethical compliance393 We have complied with all ethical regulations related to this study. The use of human kidney biopsy394 specimens in this study was approved by the Committee on Research Ethics of the Children’s Hospital395 of Nanjing Medical University. All experiments involving human samples were conducted in396 accordance with all relevant guidelines and regulations. All participants (or their legal guardians)397 provided written informed consent, and participation was entirely voluntary.398 Human kidney cortex biopsy samples used in spatial technologies and immunostaining399 This study involved renal cortex biopsies from patients with ANCA-associated vasculitis (AAV),400 approved by the Independent Ethics Committee (IEC) of Children’s Hospital of Nanjing Medical401 University (Approval #202008089-1). Kidney samples from patients with AAV and healthy controls402 were analysed using spatial technologies and immunostaining (clinical details are provided in403 Supplementary Table 1). Healthy control tissues were obtained from non-diseased regions of patients404 with hematuria. In this study, cryopreserved kidney tissue sections were utilised for spatial405 transcriptomic analysis, whereas conventional FFPE (formalin-fixed paraffin-embedded) sections406 were prepared for fluorescence immunohistochemical staining.407 Histopathological Analysis of Renal Tissues408 Renal tissue samples were fixed in 4% paraformaldehyde (PFA) at room temperature for 48 hours and409 then processed for histological examination. Serial sections were prepared at a thickness of 2 μm for410 Periodic Acid-Schiff (PAS) and Masson's trichrome staining, all performed according to established411 protocols54. These kidney samples were systematically evaluated for structural abnormalities in the412 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 21 glomerular and tubular basement membranes and fibrotic changes, with subsequent Berden413 classification to stratify histological severity in ANCA-associated glomerulonephritis (details are414 provided in Supplementary Table 1).415 Construction and sequencing of spatial gene expression libraries416 Spatial transcriptomic analysis was performed using the Visium HD Spatial Gene Expression platform417 (10x Genomics). Human kidney biopsy samples were prepared according to the Visium Spatial418 Protocols – Tissue Preparation Guide (10x Genomics, CG000240). Subsequent processing was419 conducted using the Visium Spatial Tissue Optimization Reagents Kit (10x Genomics, CG000238) and420 the Gene Expression Reagent Kit (10x Genomics, CG000239). Three Visium slides were sequenced,421 comprising control (n = 4, two sections each), mild (n = 2, with three and four sections), and severe422 (n = 3, two sections each). OCT-embedded tissue sections (10 µm) were methanol-fixed, hematoxylin423 and eosin (H&E)-stained, and imaged with a Leica DMi8 microscope (10x magnification) following the424 10x Genomics protocols (CG000241). Following permeabilization for 12 minutes, mRNA was captured425 within the fiducial capture areas of the Visium slides. cDNA libraries were generated via426 second-strand synthesis, and sequencing was performed on a Novaseq X Plus system (Illumina).427 Spatial transcriptomics data processing, filtering, and gene annotation428 Spatial transcriptomics data were processed using the Space Ranger pipeline (10x Genomics, v2.0.0)429 with the GRCh38 human reference genome (refdata-gex-GRCh38-2020-A) to generate raw unique430 molecular identifier (UMI) count matrices, mapped to 55 μm barcoded spatial spots.431 Gene-spot matrices from Visium samples underwent individual quality assessment and432 preprocessing in R Studio (v4.3.2) with STutility (v1.0.0)55 and Seurat (v5.0.1)56. For each group433 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 22 (control, mild, and severe), raw counts, histological images, spatial coordinates, and scaling factors434 were loaded into STutility. Metadata (group, patient identifiers, and section numbers) were435 annotated with the “ManualAnnotation” function. Low-quality tissue spots were removed using the436 “subsetSTData” function. Metadata were transferred to Seurat objects, removing genes detected in437 <10% of spots. Mitochondrial content was calculated but not used for filtering due to biopsy size and438 possible mitochondrial expression changes in AAV lesions.439 For each Seurat object, the gene-spot matrix was normalized using the “NormalizeData” function,440 and 2,000 highly variable features were identified via the “FindVariableFeatures” function.441 Integration anchors were then established using canonical correlation analysis (CCA) through the442 “SelectIntegrationFeatures” and “FindIntegrationAnchors” functions, enabling the integration of443 datasets for control, mild, and severe groups with the “IntegrateData” function. Then, we performed444 clustering analysis following Seurat's standard workflow. The integrated data was scaled using the445 “ScaleData” function. Then, a combined principal component analysis (PCA) was performed using the446 “RunPCA” function, selecting the first nine principal components (PCs) based on variance inflection447 points identified through the visualization from the “ElbowPlot” function. These PCs served as input448 for subsequent UMAP dimensional reduction by the “RunUMAP” function and neighbourhood graph449 construction by the “FindNeighbors” function. Finally, we applied cluster detection at a resolution of450 0.4 by the “FindClusters” function, which identified ten distinct transcriptional clusters within our451 dataset. The "DimPlot" function was used to display UMAP plots, utilising the "umap" reduction452 method and organising the data by either "cell types" or "groups," or splitting them as required.453 Cell type annotation454 Cell types were assigned by identifying differentially expressed genes (DEGs) for each cluster versus455 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 23 all others using Seurat’s “FindAllMarkers” function (MAST method) with a non-parametric Wilcoxon456 rank sum test: log2 fold change (FC) ≥ 0.25, and adjusted P ≤ 0.0557. Clusters were annotated based457 on marker genes, cross-referenced with literature markers (Supplementary Table 2) and the Kidney458 Tissue Atlas21. Ten cell types were identified: proximal tubule (PT: MIOX, ALDOB), thick ascending459 limb (TAL: SLC12A1, CLDN10), distal convoluted tubule (DCT: SLC12A3), connecting tubule/principal460 cell (CT/PC: AQP2, AQP3), connecting tubule/intercalated cell (CT/IC: FOXI1, ATP6V1B1), glomerulus461 (Glo: NPHS2, PTPRO), immune cell types and interstitial fibroblasts (IM/Fib: CD74, IGHG1), vascular462 endothelial cells/smooth muscle cells (vEC/VSMC: MYH11, ACTA2), myofibroblasts (Myo: COL8A1),463 and mixed tubules. Cellular composition was calculated as the proportion of spots per cell type464 relative to total spots for inter-group comparison, and per patient relative to total spots for465 patient-level analysis.466 Gene co-expression network analysis467 High-dimensional Weighted Gene Co-expression Network Analysis (hdWGCNA) was employed to find468 gene co-expression modules specific to the AAV-responsive cell types and associated with disease469 progression. Using the hdWGCNA package (v0.3.3) with a soft power threshold of 958, we performed470 WGCNA on the AAV severe group (Extended Data Fig. 6a). The optimal soft power threshold is471 selected when the scale-free topology model-fit achieves its maximum value while the median472 connectivity attains its minimum to identify clusters with the strongest intra-group connections and473 the weakest inter-group connections. Module eigengenes (MEs) represent the first principal474 component (PC1) of the gene expression profiles within a co-expression module, serving as a475 summary of the module's overall expression pattern. The average expression levels of the top 100476 co-expressed genes (based on kME value) in each module were calculated using the477 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 24 “AddModuleScore” function in the Seurat package. The kME value (module membership measure) is478 a key metric that quantifies the strength of a gene’s connection to its assigned module. Based on cell479 type annotation, four modules (M2, M3, M8, and M11) were enriched in AAV-responsive cell types480 and exhibited gene expression changes correlated with disease severity. The topological overlap481 matrix (TOM) of these modules was computed using the “GetTOM” function in hdWGCNA. Hub gene482 networks for each module were visualized using Cytoscape (v3.9.1), and ligand-receptor genes within483 each module were annotated based on human ligand-receptor pairs from the FANTOM5 database59.484 Differential expression analysis485 Differential expression analysis was performed in Seurat using “FindMarkers” (MAST method) for486 each cell type between mild vs. control, severe vs. control, and severe vs. mild groups57. MAST487 (Model-based Analysis of Single-cell Transcriptomics) is a statistical framework that utilises488 generalized linear models (GLMs) to analyse two gene expression features: a binomial component for489 gene detection probability and a continuous component for transcript abundance. In the490 “FindMarkers” function, MAST improves differential expression analysis by integrating covariates491 with both gene detection rates and expression levels, hence increasing biomarker identification by492 accounting for technical uncertainty and cellular heterogeneity. DEGs were defined as log2FC ≥  0.25,493 and adjusted P  ≤ 0.05 (Wilcoxon rank sum test).494 Validation of cell type annotations495 To validate cell type annotations, we performed hypergeometric test-based enrichment analysis496 using the Giotto package (v3.3.0)60. For each group (control, mild, and severe), Giotto objects were497 generated with Seurat metadata. References (Supplementary Table 9) were restricted to scRNA-seq498 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 25 data from healthy kidneys and cortex-derived cell types to avoid disease-related bias.499 Cell-type-specific marker genes were identified using Seurat’s “FindAllMarkers” function (MAST500 method) from the Kidney Cell Atlas mature kidney dataset61 and healthy kidney data in the Deeply501 Integrated Single-Cell Omics (DISCO) database62. Marker genes cutoff: log2FC ≥ 0.25 and adjusted P  ≤502  0.05 (Wilcoxon rank sum test). We performed hypergeometric testing using Giotto's503 "runHyperGeometricEnrich" function to generate enrichment scores, defined as -log10(P) per spatial504 spot. Mean scores for each annotated cell type were visualized as a heatmap.505 Identification of AAV-responsive regions506 AAV-responsive regions were defined by two criteria: (1) cell type proportions differed significantly (P507 ≤ 0.05) between mild vs. control groups or severe vs. control groups; (2) AAV marker genes508 (Supplementary Table 4) were expressed in disease but not controls, with scaled average509 expression >0 in both mild and severe groups (Supplementary Table 5 and Extended Data Fig. 3b).510 Although vEC/VSMC showed no significant proportion changes and had few spots, they were511 included in the AAV-responsive region because these cell types are associated with vasculitis512 according to the disease characteristics. Besides, mixed tubules were excluded because the spatial513 transcriptomics resolution/quality was insufficient to distinguish tubular subtypes. In summary, we514 defined IM/Fib, Myo, Glo, and vEC/VSMC as AAV-responsive regions, which meet the criteria above.515 Association between cell type and pathological indicators516 The association between cell type proportions from integrated spatial transcriptomics and517 pathological indicators from patient biopsies was assessed using Pearson correlation using the ‘cor’518 function from the stats package (v4.3.2), with significance determined via P-values within 95%519 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 26 confidence intervals from the ‘cor .mtest’ function in the corrplot package (v0.95). The pathological520 indicators were derived from quantitative metrics obtained from diagnostic biopsies (Supplementary521 Table 1). Results were visualized as a clustered heatmap by the ‘pheatmap’ function, with rows as522 pathological indicators, columns as cell types, and colour intensity indicating the Pearson correlation523 coefficient (Extended Data Fig. 3e). Statistically significances were annotated with asterisks (*P ≤ 0.1,524 ** P ≤ 0.05) and hierarchical clustering to identify cell types potentially linked to clinical features.525 Correlation between cell type markers/DEGs and pathological indicators526 To evaluate correlations between cell-type-specific markers/DEGs derived from integrated spatial527 transcriptomics and pathological indicators from patient biopsies, we employed linear regression528 analyses using the 'lm' function from the stats package implemented in a computational loop. For529 each gene-pathology indicator pair, separate regression models were fit, with R² and P-values.530 Correlations meeting dual thresholds (P 0.2) were considered biologically meaningful.531 Functional enrichment analysis532 To determine the key biological pathways associated with each cell type, we performed533 over-representation analysis (ORA) on the commonly upregulated DEGs from each cell type using the534 clusterProfiler package (v4.8.3)63. Enrichment was performed against Gene Ontology Biological535 Process (GOBP) terms64, with redundant GO entries removed using the simplify function (similarity536 cutoff=0.6). Pathways with q-value ≤ 0.05 were considered significant. WGCNA modules and537 ligand–receptor functions in AAV-responsive regions were analysed with the same method. To538 investigate the functions of DEGs in the AAV-responsive regions, we performed gene set enrichment539 analysis (GSEA) on the DEGs63,65, which were ranked by their average Log2FC values from each540 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 27 AAV-responsive cell type. This analysis was performed using the “gesPathway” function from the541 ReactomePA package (v1.44.0)66. The normalized enrichment scores (NES) were weighted based on542 the overlap between the input gene sets and Reactome Pathway's canonical functional gene sets.543 Multiple testing correction was applied using the Benjamini-Hochberg (BH) method, with significance544 thresholds set at an FDR ≤ 20% and a minimum gene set size (minGSSize) of 10. The most significant545 DEGs and associated functions were visualized using the circlize package (v0.4.16).546 Human AAV spatial transcriptomics reference dataset used to validate key features of547 AAV-responsive regions548 A recently available human AAV spatial transcriptomics reference confirmed the characteristics of549 AAV-responsive regions in our dataset12. The reference datasets comprised 10 predefined normal550 renal cell types (including CT/PC, CT/PC/IC, DCT/CT ,Glo, LOH, PT, PT/DCT, PT/TAL, tubulointerstitium,551 and vasculature) and 2 AAV-responsive cell populations (inflamed Glo and tubulointerstitium),552 derived from 19 ANCA-crescentic glomerulonephritis (GN) patients and 8 controls (Supplementary553 Table 9). We evaluated the similarity between our spatial transcriptomics data and the reference’s554 gene expression profiles using the fgsea package (v1.26.0). The DEGs identified by the “FindMarkers”555 function in the reference dataset (between controls and ANCA-GN patients, with log2FC ≥1 and556 adjusted p ≤ 0.05) and the up-regulated DEGs from our AAV severe group (ranked by log2FC) were557 used as input for analysis. The NES was then calculated to show the enrichment of cell-type features558 from the reference dataset within our cell types. The key cell type markers identified in our559 AAV-responsive regions were further validated within the spatial reference dataset.560 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 28 Cell-cell communication analysis using NICHES561 Ligand-receptor (L-R) cell-cell communication analyses were performed using the NICHES package562 (v1.0.0)67. NICHES analyses the gene expression matrix with spot metadata from spatial563 transcriptomics as input and generates interaction matrices as output using the “RunNICHES”564 function. In the matrices, each row represents a known ligand-receptor pair from OmniPath565 databases68, while each column shows potential cell-cell communication events. The interaction566 strength for each pair is calculated by multiplying the ligand's expression level in the sending cell by567 the receptor's expression level in the receiving cell. The ligand-receptor interaction matrices can be568 analysed in Seurat using the “FindMarkers” function, allowing for a comparison of cell-type-specific569 ligand-receptor interactions between the AAV and control groups. The significance thresholds were570 set using the non-parametric Wilcoxon rank sum test with the following cutoffs: log2FC ≥ 0.25 and571 adjusted P ≤ 0.05.572 We used the InterCellar package (v2.6.0) to construct interaction networks, visualizing cell-cell573 communication patterns69. In these networks, nodes represent distinct cell types from each574 experimental group, while edges indicate either autocrine (within the same cell type) or paracrine575 (between different cell types) interaction frequencies.576 Immunofluorescence staining577 Immunofluorescence (IF) analysis was conducted on 3-μm paraffin-embedded kidney sections.578 Tissue samples were derived from three groups: controls, mild, and severe AAV patients (n = 3 in579 each experimental group). The IF staining for triple-labeling tissue sections using a TSA-based580 four-colour kit system (Abclonal, #RK05903) begins with deparaffinization by baking slides at 65C for581 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 29 30 minutes, followed by sequential immersion in xylene (37°C), absolute ethanol, and an ethanol582 gradient (95%, 85%, 70%) for rehydration, then rinsed in PBS. Antigen retrieval is performed in583 Tris-EDTAbuffer (pH 9.0, Beyotime, #P0084) using an antigen retrieval cooker (Aptum, 2100 Retriever)584 for 20 minutes, cooled to room temperature (RT), and washed with PBS. Endogenous peroxidase585 activity was quenched with 3% H2O2 (ORIGENE, #PV-9000) for 20 minutes at RT, followed by PBS586 rinses and blocking with 5% BSA (Beyotime, #P0102) for 1 hour at RT. Primary antibodies are applied587 sequentially over three days: Day 1, CXCR4 antibody (Abcam, #ab124824, 1:100); Day 2, CD74588 antibody (Invitrogen, #14-0747-82, 1:100); Day 3, IgM antibody (ORIGENE, #24052901, 1:30, direct589 green fluorescence without secondary antibody). First antibodies (without IgM) are incubated590 overnight at 4°C, then washed, and treated with HRP-conjugated secondary antibody (Abclonal,591 #RK05903-5) for 50 minutes at RT. IF staining used stains TYR-570 for CD74 antibody (red, Abclonal,592 #RK05903-2); TYR-690 for CXCR4 antibody (violet, Abclonal, #RK05903-3), amplified with593 TSA+enhancer (1:200, Abclonal, #RK05903-4). After each round, the antibodies are stripped using a594 95°C EDTA buffer for 40 minutes, and the cycle (peroxidase blocking to TSA) is repeated for595 subsequent targets. Finally, these sections are mounted using an antifade mounting medium with596 DAPI (Vector Laboratories, #H-1200). Slide images were acquired using a Panoramic MIDI slide597 scanner (3DHISTECH, #PMIDI23C3001) fitted with a Grasshopper3 USB 3.0 camera (FLIR,598 #GS3-U3-51S5M-C) and a Zeiss 20X Plan-Aprochromat objective. Image capture parameters included599 an 8-bit depth and variable exposure times ranging from 8 to 60 ms. Post-acquisition processing was600 performed using CaseViewer software (3DHISTECH).601 The IF staining for double-labeling tissue sections using a TSA-based three-colour kit system602 (Abclonal, #RK05902). The procedure for IF staining was performed following the same protocol as603 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 30 described above. Primary antibodies are applied sequentially over two days: Day 1, the lumican604 antibody (Proteintech, #10677-1-AP ,1:400) or the CXCR4 antibody; Day 2, the collagen I and collagen605 III antibodies (Bioss, #BS-10423R and #BS-0549, 1:400) or the CD74 antibody. IF staining used stains606 TYR-520 for CXCR4 and lumican antibodies (green, Abclonal, #RK05903-1); TYR-570 for CD74,607 collagen I, and collagen III antibodies (red, Abclonal, #RK05903-2). Fluorescence was detected using608 an X-cite Series 120 Q unit (Excelitas, USA) attached to a light microscope Imager A2 (Zeiss, Germany)609 with an Axiocam 503 colour digital camera (Zeiss, Germany). Five randomly selected fields per610 section (200X magnification) were imaged, with all acquisitions performed under standardized611 optical and exposure conditions. Quantitative analysis was conducted using ImageJ (v1.54k, National612 Institutes of Health, USA)70.613 Statistical analysis614 Statistical analyses were performed using R Studio (v4.3.2) or GraphPad Prism software (v10.4.1;615 GraphPad Software, LLC, www.graphpad.com). Continuous variables are expressed as mean ±616 standard deviation (SD), with between-group differences evaluated by two-tailed, unpaired Student’s617 t-tests. Statistical significance was defined as a P < 0.05, unless otherwise indicated. Differences in618 cell type proportions were assessed via Student's t-test, with significance levels denoted as * (P ≤619 0.05) and ** (P ≤ 0.01).620 Packages used to generate spatial transcriptomics figures621 The top DEGs for each cell type were plotted as a heatmap using the "DoHeatmap" function in Seurat,622 representing the integrated datasets. Other heatmaps were generated using the pheatmap package623 (v1.0.12), employing the “pheatmap” function with optimized parameters.624 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 31 The spatial distribution of cell types or AAV-responsive regions in control, mild, and severe datasets625 was visualized using the “SpatialDimPlot” function in Seurat. The “SpatialFeaturePlot” function was626 used to reveal the scaled expression and spatial patterns of key cell type markers in AAV-responsive627 regions. Violin plots were created using the “VlnPlot” function in Seurat. All dot plots were generated628 using the “DotPlot” function in Seurat to display feature expression changes across different cell629 types.630 Figures not mentioned before (such as bar plots, combined violin-box plots, and scatter plots) were631 generated using the ggplot2 package (v3.5.1) with optimized parameters.632 633 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 32 Data availability634 The spatial transcriptomic data supporting the findings of this study have been deposited in the635 CNSA database and will be made publicly available upon publication.636 Code availability637 All analyses were conducted using open-source software and packages, as cited in the Methods638 section. Additional information can be provided upon request.639 Competing interests640 All the authors declared no competing interests.641 642 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 33 Figure legends643 Fig.1: Overview of study design and clinical information in control (ctrl), mild and severe644 ANCA-associated vasculitis (AAV) datasets. (a) Histological staining of patient samples from control,645 mild AAV, and severe AAV groups. There were three groups of renal cortex tissues and646 corresponding histochemical staining images extracted from 4 (control), 2 (mild AAV) and 3 (severe647 AAV) patients respectively. Control patients are labeled as ‘ctrl’. The aav1 patient (mild group)648 exhibiting segmental glomerular sclerosis (Light microscopy, Masson trichrome-stained section). The649 aav2 patient (mild group) demonstrating glomerular global sclerosis with renal tubular atrophy (Light650 microscopy, PAS-stained section). The aav3-5 patients (severe group) showing crescentic formations651 in glomeruli (Light microscopy, aav3 & aav4: Masson trichrome; aav5: PAS). Yellow circles highlight652 pathological regions in all panels. Scale bars represent 50 μm. (b) Schematic workflow showing the653 study design and the subsequent spatial transcriptome analysis. The tissue sections were overlaid on654 Visium slides and sequenced. The data underwent quality control and subsequent downstream655 integrated analysis. UMAP plot showing the unsupervised clusters with cell type annotations656 assigned in integrated datasets. PT, proximal tubule; TAL, thick ascending limb; DCT, distal657 convoluted tubule; CT/PC, connecting tubule/principal cell; CT/IC, connecting tubule/intercalated658 cell; Glo, glomerulus; IM/Fib, immune cell and fibroblast; vEC/VSMC, vascular endothelial659 cell/vascular smooth muscle cell; Myo, myofibroblast. (c) Ratios of cell types in each group (left). The660 cell proportions in all patient samples were compared using Student's t-test between the control,661 mild, and severe groups. The –log₁₀(P) are shown in the heatmap. * P ≤ 0.05, ** P ≤ 0.01. (d)662 Heatmap showing the top 20 most differentially expressed genes (DEGs) for each cell type in663 integrated datasets. For each cell type, five representative markers were highlighted. (e) H&E images664 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 34 overlayed by the annotated cell types in control, mild and severe AAV datasets.665 Fig.2: Identify and characterise AAV-responsive regions involved in disease progression. (a)666 Highlighted AAV-responsive regions, the matched histology and related cell types in AAV samples.667 Control samples are provided as a comparison. (b) The scatter plot illustrates the correlation668 between the pathological indicators of the kidney sections from AAV patients (x-axis) and the ratios669 of representative cell types/regions (y-axis) in five AAV patient samples. Non-responsive regions670 encompass cell types lacking AAV-responsive regions. (c) The heatmap displays the expression of cell671 type markers across control, mild, and severe AAV groups. Representative markers for each cell type672 are indicated, with those correlated to pathological indicators from AAV-responsive regions673 highlighted in bold (Pearson’s correlation: P ≤ 0.05. (d) Scatter plots show the relationship between674 kidney pathology indicators in AAV patients (x-axis) and the expression levels of key cell type675 markers in AAV-responsive regions (y-axis). Violin plots compare the expression levels of these676 markers across control, mild, and severe groups in each AAV-responsive region. P-values were677 determined using a Student's t-test.678 Fig.3: Functions of differentially expressed genes (DEGs) in disease-responsive regions and their679 relationships with clinical indicators. (a) The bar plot shows the ratios of up/down-regulated DEGs680 to cell type markers among control, mild, and severe groups. (b) Circular plots showing the top DEGs681 in AAV-responsive regions (IM/Fib, Myo and Glo) and their associated key biological functions, as682 identified through Gene Set Enrichment Analysis (GSEA) in the Reactome database. All functions are683 colour-coded on the outer ring to distinguish higher-level Reactome pathway categories, and the684 average log2 fold change (log2FC) for each gene is also annotated. (c) Heatmap (left) displaying the685 z-score of the expression levels of the top DEGs from AAV-responsive regions (IM/Fib, Myo and Glo)686 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 35 in each patient, which are correlated with clinical indicators (R² > 0.2, P < 0.05). The right heatmap687 illustrates the R² values representing the correlation between the DEGs and these clinical indicators.688 ** P < 0.01. The DEGs with the most significant correlation with indicators are highlighted in bold689 font.690 Fig.4: High-dimensional weighted gene co-expression network analysis (hdWGCNA) further691 characterises the gene co-expression networks in AAV-responsive regions of spatial transcriptomic692 data. (a) hdWGCNA identified 13 co-expression modules in the severe AAV group, with module693 eigengenes (MEs) representing the dominant expression patterns of each cell type. (b) Violin plots694 display expression of module features (top 100 genes) in AAV-responsive cell types of the severe695 group, with M2 and M11 predominantly localized in the IM/Fib region, M3 in the Myo region, and696 M8 in the Glo region. Feature expression levels were computed using the AddModuleScore. P-values697 were determined using a Student‘s t-test. (c) Violin plots show the comparison of module feature698 expression levels in each corresponding region across the control, mild and severe groups. P-values699 were determined using a Student‘s t-test. (d) Group-labeled scatter plot visualizing the correlation700 between M2 and M11 AddModuleScores in the IM/Fib region. (e-f) The network shows the701 eigengenes in M2 and M11. The 50 hub genes are represented by coloured circles, while ligands and702 receptors are also highlighted. (g-h) Top 5 enriched gene ontology biological process (GOBP)703 pathway from M2 and M11. Ligands or receptors from hub genes involved in each pathway are704 labeled in the figure. (i-j) Module activity patterns were spatially mapped using SpatialFeaturePlots,705 displaying the tissue distributions of M2 and M11 features generated by AddModuleScore.706 Fig.5: Ligand–receptor interactions reveal leukocyte chemotaxis and ECM remodeling in IM/Fib707 regions. (a) Spatially overlapped and scaled expression of CXCR4 and CD74 in a severe AAV patient.708 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 36 Expressions of CD74 and CXCR4 are visualized as an outer hollow circle and an inner solid circle,709 respectively. Fewer than 10 co-expressing spots were detected in mild AAV, precluding correlation710 analysis. (b) A close-up view (inset in a) reveals the spatial co-expression of CXCR4 and CD74 at711 spatial transcriptomic resolution. (c) Violin plots compare the expression levels of CXCR4 and CD74712 across control, mild, and severe AAV groups in the IM/Fib. P-values were determined using a713 Student‘s t-test. (d) Scatter plots illustrate the correlation between CD74 (x-axis) and CXCR4714 expression (y-axis) in spots co-expressing both markers (excluding spots where either CD74 or CXCR4715 expression is zero) in the IM/Fib region of the severe AAV group. (e) Antibody-staining of CXCR4716 (magenta), CD74 (red), and IgM (green) across control, mild, and severe groups. White boxes717 highlighted regions containing IgM⁺ cells (green) adjacent to CXCR4–CD74 co-localized cells718 (magenta/red) in representative mild and severe AAV patient samples. Scale bars represent 20 μm.719 (f) Relative protein expression levels of CXCR4 and CD74 (% area fraction of each antibody) across720 control, mild, and severe AAV groups. P-values were determined using a Student’s t-test. (g)721 Heatmap showing Pearson’s colocalization coefficients of CXCR4 and CD74 protein expression in722 mild and severe AAV groups, with P-values determined using Student’s t-test between all723 immunofluorescence (IF) staining fields of the two groups and labeled on the top of the heatmap.724 The Y-axis represents measurements from five IF fields per patient, while the X-axis indicates725 patients from each group (n = 3 per group). (h-k) Same as in (a–d), measured in LUM and COL1A2. (l)726 Antibody-staining of lumican (green) and collagen I (red) across control, mild, and severe groups.727 Scale bars represent 50 μm. (m-n) Same as in (f-g), measured in lumican and collagen I.728 729 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 37 Reference730 1 Nakazawa, D., Masuda, S., Tomaru, U. & Ishizu, A. Pathogenesis and therapeutic731 interventions for ANCA-associated vasculitis. Nature Reviews Rheumatology 15, 91–101732 (2019).733 2 Geetha, D. & Jefferson, J. A. ANCA-associated vasculitis: core curriculum 2020. American734 Journal of Kidney Diseases 75, 124–137 (2020).735 3 Tseng, S.-T., Tseng, M.-H. & Huang, J.-L. Concurrent pulmonary hemorrhage and deep vein736 thrombosis in a child with ANCA-associated vasculitis: case report and review of literature.737 Pediatric Rheumatology 13, 1–7 (2015).738 4 Barnini, C., Oni, L. & Kronbichler,A. Course of paediatric ANCA-associated glomerulonephritis:739 advocating for an age-inclusive approach. RMD open 10, e004481 (2024).740 5 Gómez-Puerta, J. A. & Bosch, X. Anti-neutrophil cytoplasmic antibody pathogenesis in741 small-vessel vasculitis: an update. The American journal of pathology 175, 1790–1798742 (2009).743 6 Jarrot, P .-A. & Kaplanski, G. Pathogenesis of ANCA-associated vasculitis: an update.744 Autoimmunity reviews 15, 704–713 (2016).745 7 Wigerblad, G. & Kaplan, M. J. Neutrophil extracellular traps in systemic autoimmune and746 autoinflammatory diseases. Nature Reviews Immunology 23, 274–288 (2023).747 8 Linke, A., Tiegs, G. & Neumann, K. Pathogenic T-cell responses in immune-mediated748 glomerulonephritis. Cells 11, 1625 (2022).749 9 Tang, Y ., Cao, Q., Liu, J. & Zhuang, Q. Immune Landscape Variation in Antineutrophil750 Cytoplasmic Antibody‐Associated Vasculitis Circulation Before and After Plasmapheresis by751 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 38 Single‐Cell Transcriptome. Mediators of Inflammation 2025, 5531382 (2025).752 10 Vegting, Y . et al. Infiltrative classical monocyte-derived and SPP1 lipid-associated753 macrophages mediate inflammation and fibrosis in ANCA-associated glomerulonephritis.754 Nephrology Dialysis Transplantation, gfae292 (2024).755 11 Mueller, A. et al. Transcriptional and clonal characterization of cytotoxic T cells in crescentic756 glomerulonephritis. Journal of the American Society of Nephrology 34, 1003–1018 (2023).757 12 Engesser, J. et al. Immune profiling-based targeting of pathogenic T cells with ustekinumab in758 ANCA-associated glomerulonephritis. Nature Communications 15, 8220 (2024).759 13 Moiseev, S. et al. The alternative complement pathway in ANCA-associated vasculitis: further760 evidence and a meta-analysis. Clinical & Experimental Immunology 202, 394–402 (2020).761 14 O'Sullivan, K. M. & Holdsworth, S. R. Neutrophil extracellular traps: a potential therapeutic762 target in MPO-ANCA associated vasculitis? Frontiers in immunology 12, 635188 (2021).763 15 Koster, M. J. & Warrington, K. J. Recent advances in understanding and treating vasculitis.764 F1000Research 5, F1000 Faculty Rev–1436 (2016).765 16 Hartinger, J. M., Kratky, V., Hruskova, Z., Slanar, O. & Tesar, V. Implications of rituximab766 pharmacokinetic and pharmacodynamic alterations in various immune-mediated767 glomerulopathies and potential anti-CD20 therapy alternatives. Frontiers in immunology 13,768 1024068 (2022).769 17 Merkel, P .A., Jayne, D. R., Wang, C., Hillson, J. & Bekker, P .Evaluation of the safety and770 efficacy of avacopan, a C5a receptor inhibitor, in patients with antineutrophil cytoplasmic771 antibody–associated vasculitis treated concomitantly with rituximab or772 cyclophosphamide/azathioprine: protocol for a randomized, double-blind, active-controlled,773 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 39 phase 3 trial. JMIR research protocols 9, e16664 (2020).774 18 Flossmann, O., Scott, D. G. I., WITTER, J., YAZICI, H. & LUQMANI, R. A. EULAR775 recommendations for conducting clinical studies and/or clinical trials in systemic vasculitis:776 Focus on anti-neutrophil. Annals of the rheumatic diseases 66, 605–617 (2007).777 19 Amudala, N. A. et al. Obinutuzumab as treatment for ANCA-associated vasculitis.778 Rheumatology 61, 3814–3817 (2022).779 20 Minopoulou, I. et al. Anti-CD19 CAR T cell therapy induces antibody seroconversion and780 complete B cell depletion in the bone marrow of a therapy-refractory patient with781 ANCA-associated vasculitis. Annals of the Rheumatic Diseases 84, e4–e7 (2025).782 21 Hansen, J. et al. A reference tissue atlas for the human kidney. Science advances 8, eabn4965783 (2022).784 22 Lake, B. B. et al. An atlas of healthy and injured cell states and niches in the human kidney.785 Nature 619, 585–594 (2023).786 23 Croft, D. et al. Reactome: a database of reactions, pathways and biological processes. Nucleic787 acids research 39, D691–D697 (2010).788 24 Marinaki, S. et al. Abnormalities of CD4+ T cell subpopulations in ANCA-associated vasculitis.789 Clinical & Experimental Immunology 140, 181–191 (2005).790 25 Söderberg, D. & Segelmark, M. Neutrophil extracellular traps in ANCA-associated vasculitis.791 Frontiers in immunology 7, 256 (2016).792 26 Yan, J., Zhang, Z., Yang, J., Mitch, W. E. & Wang, Y . JAK3/STAT6 stimulates bone793 marrow–derived fibroblast activation in renal fibrosis. Journal of the American Society of794 Nephrology 26, 3060–3071 (2015).795 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 40 27 Csernok, E. et al. Transforming growth factor-beta (TGF-β) expression and interaction with796 proteinase 3 (PR3) in anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis.797 Clinical & Experimental Immunology 105, 104–111 (1996).798 28 Vegting, Y .et al. Monocytes and macrophages in ANCA-associated vasculitis. Autoimmunity799 Reviews 20, 102911 (2021).800 29 Abou Faycal, C. et al. An adapted passive model of anti-MPO dependent crescentic801 glomerulonephritis reveals matrix dysregulation and is amenable to modulation by CXCR4802 inhibition. Matrix Biology 106, 12–33 (2022).803 30 Walsh, M. et al. Plasma exchange and glucocorticoids in severe ANCA-associated vasculitis.804 New England Journal of Medicine 382, 622–631 (2020).805 31 Lin, W. et al. Glomerular immune deposition in MPO-ANCA associated glomerulonephritis is806 associated with poor renal survival. Frontiers in Immunology 12, 625672 (2021).807 32 Merino-Vico, A. et al. POS1006 PROFILING B AND PLASMA CELLS IN KIDNEY BIOPSIES FROM808 ANCA-ASSOCIATED VASCULITIS PATIENTS WITH GLOMERULONEPHRITIS AT SINGLE-CELL809 RESOLUTION. Annals of the Rheumatic Diseases 83, 1072 (2024).810 33 Chen, M., Jayne, D. R. & Zhao, M.-H. Complement in ANCA-associated vasculitis: mechanisms811 and implications for management. Nature Reviews Nephrology 13, 359–367 (2017).812 34 Amara, U. et al. Molecular intercommunication between the complement and coagulation813 systems. The Journal of Immunology 185, 5628–5636 (2010).814 35 Conway, E. Reincarnation of ancient links between coagulation and complement. Journal of815 Thrombosis and Haemostasis 13, S121–S132 (2015).816 36 Zhou, Y .et al. Identification of afatinib-associated ADH1B and potential small-molecule drugs817 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 41 targeting ADH1B for hepatocellular carcinoma. Frontiers in Pharmacology 14, 1166454818 (2023).819 37 Vilar-Gomez, E. et al. ADH1B∗ 2 is associated with reduced severity of nonalcoholic fatty liver820 disease in adults, independent of alcohol consumption. Gastroenterology 159, 929–943821 (2020).822 38 Bonasia, C. G. et al. Circulating immune profile in granulomatosis with polyangiitis reveals823 distinct patterns related to disease activity. Journal of Autoimmunity 146, 103236 (2024).824 39 Ma-Krupa, W. et al. Activation of arterial wall dendritic cells and breakdown of self-tolerance825 in giant cell arteritis. The Journal of experimental medicine 199, 173–183 (2004).826 40 Bianchi, M. E. & Mezzapelle, R. The chemokine receptor CXCR4 in cell proliferation and tissue827 regeneration. Frontiers in Immunology 11, 2109 (2020).828 41 Yuan, A., Lee, Y .,Choi, U., Moeckel, G. & Karihaloo, A. Chemokine receptor Cxcr4 contributes829 to kidney fibrosis via multiple effectors. American Journal of Physiology-Renal Physiology 308,830 F459–F472 (2015).831 42 Djudjaj, S. et al. Macrophage migration inhibitory factor mediates proliferative GN via CD74.832 Journal of the American Society of Nephrology 27, 1650–1664 (2016).833 43 Lue, H., Dewor, M., Leng, L., Bucala, R. & Bernhagen, J. Activation of the JNK signalling834 pathway by macrophage migration inhibitory factor (MIF) and dependence on CXCR4 and835 CD74. Cellular signalling 23, 135–144 (2011).836 44 Kronbichler, A. et al. Immunopathogenesis of ANCA-associated vasculitis. International837 journal of molecular sciences 21, 7319 (2020).838 45 Gül, A. et al. The pathogenesis, clinical presentations and treatment of monogenic systemic839 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 42 vasculitis. Nature Reviews Rheumatology, 1–12 (2025).840 46 Kessenbrock, K. et al. Netting neutrophils in autoimmune small-vessel vasculitis. Nature841 medicine 15, 623–625 (2009).842 47 Genovese, F., Manresa, A. A., Leeming, D. J., Karsdal, M. A. & Boor, P .The extracellular matrix843 in the kidney: a source of novel non-invasive biomarkers of kidney fibrosis? Fibrogenesis &844 tissue repair 7, 1–14 (2014).845 48 Feng, S. et al. Identification of lumican and fibromodulin as hub genes associated with846 accumulation of extracellular matrix in diabetic nephropathy. Kidney and Blood Pressure847 Research 46, 275–285 (2021).848 49 Sultana, Z. et al. Spatio-temporal interaction of immune and renal cells determines849 glomerular crescent formation in autoimmune kidney disease [Preprint]. bioRxiv850 2024-Dec-18 (accessed 2025-Jun-6), 629206 (2024).851 https://doi.org/https://doi.org/10.1101/2024.12.18.629206852 50 Danaher, P .et al. Childhood-onset lupus nephritis is characterized by complex interactions853 between kidney stroma and infiltrating immune cells. Science Translational Medicine 16,854 eadl1666 (2024).855 51 Marco Salas, S. et al. Optimizing Xenium In Situ data utility by quality assessment and856 best-practice analysis workflows. Nature Methods, 1–11 (2025).857 52 Choi, J. et al. Spatial organization of the mouse retina at single cell resolution by MERFISH.858 nature communications 14, 4929 (2023).859 53 Shah, S., Lubeck, E., Zhou, W. & Cai, L. seqFISH accurately detects transcripts in single cells860 and reveals robust spatial organization in the hippocampus. Neuron 94, 752–758. e751861 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 43 (2017).862 54 Peng, H. et al. Myokine mediated muscle-kidney crosstalk suppresses metabolic863 reprogramming and fibrosis in damaged kidneys. Nature communications 8, 1493 (2017).864 55 Bergenstråhle, J., Larsson, L. & Lundeberg, J. Seamless integration of image and molecular865 analysis for spatial transcriptomics workflows. BMC genomics 21, 1–7 (2020).866 56 Stuart, T. et al. Comprehensive integration of single-cell data. cell 177, 1888–1902. e1821867 (2019).868 57 Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes869 and characterizing heterogeneity in single-cell RNA sequencing data. Genome biology 16,870 1–13 (2015).871 58 Morabito, S., Reese, F., Rahimzadeh, N., Miyoshi, E. & Swarup, V. hdWGCNA identifies872 co-expression networks in high-dimensional transcriptomics data. Cell reports methods 3873 (2023).874 59 Ramilowski, J. A. et al. A draft network of ligand–receptor-mediated multicellular signalling in875 human. Nature communications 6, 7866 (2015).876 60 Dries, R. et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression877 data. Genome biology 22, 1–31 (2021).878 61 Stewart, B. J. et al. Spatiotemporal immune zonation of the human kidney. Science 365,879 1461–1466 (2019).880 62 Li, M. et al. DISCO: a database of Deeply Integrated human Single-Cell Omics data. Nucleic881 acids research 50, D596–D602 (2022).882 63 Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. The883 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint 44 innovation 2 (2021).884 64 Ashburner, M. et al. Gene ontology: tool for the unification of biology. Nature genetics 25,885 25–29 (2000).886 65 Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for887 interpreting genome-wide expression profiles. Proceedings of the National Academy of888 Sciences 102, 15545–15550 (2005).889 66 Yu, G. & He, Q.-Y .ReactomePA: an R/Bioconductor package for reactome pathway analysis890 and visualization. Molecular BioSystems 12, 477–479 (2016).891 67 Raredon, M. S. B. et al. Comprehensive visualization of cell–cell interactions in single-cell and892 spatial transcriptomics with NICHES. Bioinformatics 39, btac775 (2023).893 68 Türei, D., Korcsmáros, T. & Saez-Rodriguez, J. OmniPath: guidelines and gateway for894 literature-curated signaling pathway resources. Nature methods 13, 966–967 (2016).895 69 Interlandi, M., Kerl, K. & Dugas, M. InterCellar enables interactive analysis and exploration of896 cell− cell communication in single-cell transcriptomic data. Communications biology 5, 21897 (2022).898 70 Abràmoff, M. D., Magalhães, P .J. & Ram, S. J. Image processing with ImageJ. Biophotonics899 international 11, 36–42 (2004).900 901 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint Figure1 d b ctrl1 ctrl2 ctrl3 ctrl4 aav1 aav2 aav3 aav4 aav5 e Control Mild Severe c SLC12A3 TMEM52B DEFB1 FXYD2 SERPINA5 SLC12A1 UMOD CLDN16 ITGB6 PPP1R1A GPX3 NAT8 GATM UGT2B7 ALDOB AQP2 AQP3 SCNN1G FXYD4 CDH16 RHCG SLC4A1 KLK1 TMEM213 CALB1 PODXL HTRA1 SPOCK2 CCN2 PTPRO C7 C1QA C1S LUM C1R ADIRF PLN MYH11 MCAM ACTA2 CXCL6 TPM1 MMP7 C3 CDH6 MALAT1 (PC) MTRNR2L12 (TAL) SLC5A12 (PT) MTRNR2L1 S100A9 (monocyte) KRT17 -2 -1 0 1 2 Scaled expression PT TAL DCT CT/PC CT/IC Glo IM/Fib vEC/VSMC Myo Mixed tubules 0.00 0.25 0.50 0.75 1.00 Control Mild Severe ctrl1 ctrl2 ctrl3 ctrl4 aav1 aav2 aav3 aav4 aav5 Mixed tubules Ratios of cell types PT TAL DCT CT/PC CT/IC Glo IM/Fib vEC/VSMC Myo** ** * * * * * * 3 2.5 2 1.5 1 0.5 -log10(P) a 6.5 mm UMAP-2 Visium spatial gene expression slides ctrl1 ctrl2 ctrl3 ctrl4 aav1 aav2 aav3 aav4 aav5 Spatial transcriptomes (ST) 6.5 mmQuality control Integrated analysis Control Mild Severe UMAP-1 Mixed tubules PT TAL DCT CT/PC CT/IC Glo IM/Fib vEC/VSMC Myo Ctrl Mild Severe Fig.1: Overview of study design and clinical information in control (ctrl), mild and severe ANCA -associated vasculitis (AAV) datasets. (a) Histological staining of patient samples from control, mild AAV, and severe AAV groups. There were three groups of renal cortex tissues and corresponding histochemical staining images extracted from 4 (control), 2 (mild AAV) and 3 (severe AAV) patients respectively. Control patients are labeled as ‘ctrl’. The aav1 patient (mild group) exhibiting segmental glomerular sclerosis (Light microscopy, Masson trichrome-stained section). The aav2 patient (mild group) demonstrating glomerular global sclerosis with renal tubular atrophy (Light microscopy, PAS-stained section). The aav3-5 patients (severe group) showing crescentic formations in glomeruli (Light microscopy, aav3 & aav4: Masson trichrome; aav5: PAS). Yellow circles highlight pathological regions in all panels. Scale bars represent 50 μm. (b) Schematic workflow showing the study design and the subsequent spatial transcriptome analysis. The tissue sections were overlaid on Visium slides and sequenced. The data underwent quality control and subsequent downstream integrated analysis. UMAP plot showing theunsupervised clusters with cell type annotations assigned in integrated datasets. PT, proximal tubule; TAL, thick ascending limb; DCT, distal convoluted tubule; CT/PC, connecting tubule/principal cell; CT/IC, connecting tubule/intercalated cell; Glo, glomerulus; IM/Fib, immune cell and fibroblast; vEC/VSMC, vascular endothelial cell/vascular smooth muscle cell; Myo, myofibroblast. (c) Ratios of cell types in each group (left). The cell proportions in all patient samples were compared using Student's t-test between the control, mild, and severe groups. The –log₁₀(P) are shown in the heatmap. * P ≤ 0.05, ** P ≤ 0.01. (d) Heatmap showing the top 20 most differentially expressed genes (DEGs) for each cell type in integrated datasets. For each cell type, five representative markers were highlighted. (e) H&E images overlayed by th e annotated cell types in control, mild and severe AAV datasets. ctrl2 ctrl3 Renal puncture (cortex) ctrl1 n = 2 n = 3 Control n = 4 Mild Severe Sample pathology of patients aav1 aav2 aav3 aav4 aav5 ctrl4 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint Figure2 Control Mild Severe P = 0.29 5 0 1 2 3 4 5 0 1 2 3 4 Control Mild Severe P = 0.15 P = 1.2  10-7 P = 4.2  10-15 Control Mild Severe 5 0 1 2 3 4 P = 0.048 P =3.7  10-10 0 2 4 Control Mild Severe P = 3.9  10-5 P = 4.6  10-12 P = 0.000276 P =0.0084 Non-responsive regions AAV-responsive regions IM/Fib Myo vEC/VSMC Glo Histology (H&E) a Control Mild Severe Cell types b d 6 4 2Average gene expression0.00 0.25 0.50 0.75 1.00 R = 0.99, P = 0.00084 IM/Fib: ADH1B 15 10 5 0.10 0.15 0.20 0.25 Ratio of fibrosis Ratio of focal sclerotic glomeruli R = 0.97, P = 0.0049 IM/Fib: C1S 10 5 0.2 0.4 0.6 0.8 Ratio of global sclerotic glomeruli R = 0.95, P = 0.012 10 5 0.2 0.4 0.6 0.8 Ratio of global sclerotic glomeruli R = 0.93, P = 0.021 IM/Fib: LUM IM/Fib: C1R P =1.1  10-8 P =0.00077 P =2.3  10-10 P = 3.7  10-7 P < 2.22  10-16 P < 2.22  10-16 Scaled gene expression 0 1 2 3 4 Control Mild Severe Control Mild Severe 5 0 1 2 3 4 5 0 1 2 3 4 Control Mild Severe P = 4.6  10-7 P = 3.9  10-14 5 0 1 2 3 4 Control Mild Severe P = 8.2  10-14 P < 2.22  10-16 P = 9.8  10-14 Ratio of injured glomeruli 6 4 2Average gene expression 8 Glo: VEGFA 0 0.996 0.998 1.000 Ratio of injured glomeruli R = -0.99, P = 0.0021 Glo: PLA2R1 6 4 2 0 R = -0.95, P = 0.014 0.996 0.998 1.000 vEC/VSMC: TAGLN vEC/VSMC: SPARCL1 25 20 15 10 0.00 0.25 0.50 0.75 1.00 Ratio of fibrosis R = 0.97, P = 0.0056 10 5 0 0.10 0.15 0.20 0.25 Ratio of focal sclerotic glomeruli R = 0.9, P = 0.035 Scaled gene expression Scaled gene expression Control Mild Severe 0 1 2 3 4 P = 2.8  10-11 Control Mild Severe 0 1 2 3 P = 2.2  10-13 Control Mild Severe P = 4.8  10-7 0 1 2 3 4 0 2 4 6 Control Mild Severe 30 20 10 0 0.00 0.25 0.50 0.75 1.00 20 10 0 0.2 0.4 0.6 0.8 Ratio of global sclerotic glomeruli 30 20 10 0 40 0.2 0.4 0.6 0.8 Ratio of global sclerotic glomeruli R = 0.99, P = 0.0016R = 0.93, P = 0.022R = 0.98, P = 0.0045 60 40 20 0 0.996 0.998 1.000 Ratio of injured glomeruli R = 0.97, P = 0.0061 Average gene expression Ratio of fibrosis Myo: NNMT Myo: SOD2 Myo: MGPMyo: C3 0.8 R = -0.87, P = 0.058 Ratio of focal sclerotic glomeruli 0.150.10 0.20 0.25 1.0 0.6 R = 0.82, P = 0.0880.10 0.05 0.00 -0.05 0.150.10 0.20 0.25 R = -0.87, P = 0.057 Ratio of non- responsive regions Ratio of global sclerotic glomeruli 0.40.2 0.6 0.8 0.9 0.8 0.7 0.6 0.5 R = 0.86, P = 0.061 Ratio of IM/Fib Ratio of global sclerotic glomeruli 0.4 0.3 0.2 0.1 0.0 0.2 0.6 0.8 0.1 R = 0.86, P = 0.0630.4 0.3 0.2 0.0 -0.1 0.150.10 0.20 0.25 Ratio of global sclerotic glomeruli R = 0.95, P = 0.015 Ratio of Myo 0.075 0.050 0.025 0.000 0.40.2 0.6 0.8 Mild Severe 0.4 Ratio of non- responsive regions Ratio of IM/FibRatio of Myo Ratio of focal sclerotic glomeruli Ratio of focal sclerotic glomeruli c COL4A1 HLA-DPB1 AQP3 SPP1 PODXL TAGLN LUM NNMT MGP C1S NAT8 GPX3 IGHG3 IGHA1 IGLC1 IGKC IGLC2 IGHG1 IGLC3 IGHG4JCHAIN C1R IGHG2 SOD2 VCAN Scaled expression -4 -2 0 2 4 PDZK1IP1 UMODSLC12A1 SLPI SLC12A3 AQP2 RHCG SLC4A1 HTRA1 CCN2 POSTN PLAT C1QA DCN SPARCL1 C3 TPM1 VEGFA MME IL1RL1 PCOLCE2 PLA2R1 SERPINF1 ADH1B Fig.2: Identify and characterise AAV-responsive regions involved in disease progression. (a) Highlighted AAV-responsive regions, the matched histology and related cell types in AAV samples. Control samples are provided as a comparison. (b) The scatter plot illustrates the correlation between the pathological indicators of the kidney sections from AAV patients (x-axis) and the ratios of representative cell types/regions (y-axis) in five AAV patient samples. Non-responsive regions encompass cell types lacking AAV-responsive regions. (c) The heatmap displays the expression of cell type markers across control, mild, and severe AAV groups. Representative markers for each cell type are indicated, with those correlated to pathological indicators from AAV-responsive regions highlighted in bold (Pearson’s correlation: p-value ≤ 0.05. (d) Scatter plots show the relationship between kidney pathology indicators in AAV patients (x-axis) and the expression levels of key cell type markers in AAV-responsive regions (y-axis). Violin plots compare the expression levels of these markers across control, mild, and severe groups in each AAV-responsive region. P-values were determined using a Student's t-test. Control Mild Severe P < 2.22 10-16 P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint VCAN STAT1 CALM2 a b COL3A1 C1QC PI3KRT16 Detoxification of Reactive Oxygen Species Myo PPP2CB C3 C1SC1R KRT6BSLC26A7 IM/Fib Glo Immune system ECM organization Gene expression Vesicle-mediated transport Developmental Biology Disorders of transmembrane transporters Metabolism of proteins Signal transduction Cellular responses to stress Transport of small molecules Average Log2FC 2 0 -2 c Groups Control Mild Severe aav1ctrl4ctrl3IM/Fib ctrl1 ctrl2 aav2 aav3 aav4 aav5 C3 C1S C1R CFB PROS1 MT2A CEBPD VIM TIMP1 SAMHD1 HLA-DPB1 FOXO3 PSME1 FTH1 CRISPLD2 LUM COL3A1 VCAN DCN COL4A1 TNC FBN1 PPIB CD163 LRP1 SLC26A7 SLC12A3 SLC4A1 1.0 0.9 0.8 0.7 0.6 R2 z-score 2 1 0 -1 -2 ** ** ** ** ** ** ** ** Glo ctrl1 ctrl2 ctrl3 ctrl4 aav1 aav2 aav3 aav4 aav5 COL3A1 COL1A2 FTH1 COL8A1 COL16A1 FBN1 C3 VIM MT2A TNFRSF12A CRLF1 FOXO3 SLC12A1 JUP KRT16 ** ** ** ** ** ** VCAN SERPINE1 TNC COL4A1 SOD2 MCL1 PSMA5 CEBPB DDIT4 NPM1 aav1 aav2 aav3 aav4 aav5Myo ** ** ** ** Mixed tubules Myo vEC/VSMC IM/Fib Glo CT/IC CT/PC DCT TAL PT 0.0 0.1 UpDown Mild vs control Severe vs control Severe vs mild 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.00.7 Figure3 Fig.3: Functions of differentially expressed genes (DEGs) in disease-responsive regions and their relationships with clinical indicators. (a) The bar plot shows the ratios of up/down-regulated DEGs to cell type markers among control, mild, and severe groups. (b) Circular plots showing the top DEGs in AAV-responsive regions (IM/Fib, Myo and Glo) and their associated key biological functions, as identified through Gene Set Enrichment Analysis (GSEA) in the Reactome database. All functions are colour-coded on the outer ring to distinguish higher-level Reactome pathway categories, and the average log2 fold change (log2FC) for each gene is also annotated. (c) Heatmap (left) displaying the z-score of the expression levels of the top DEGs from AAV-responsive regions (IM/Fib, Myo and Glo) in each patient, which are correlated with clinical indicators (R²> 0.2, P < 0.05). The right heatmap illustrates the R²values representing the correlation between the DEGs and these clinical indicators. ** P < 0.01. The DEGs with the most significant correlation with indicators are highlighted in bold font. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint Figure4 Fig.4: High-dimensional weighted gene co-expression network analysis (hdWGCNA) further characterises the gene co-expression networks in AAV-responsive regions of spatial transcriptomic data. (a) hdWGCNA identified 13 co-expression modules in the severe AAV group, with module eigengenes (MEs) representing the dominant expression patterns of each cell type. (b) Violin plots display expression of module features (top 100 genes) in AAV-responsive cell types of the severe group, with M2 and M11 predominantly localised in the IM/Fib region, M3 in the Myo region, and M8 in the Glo region. Feature expression levels were computed using the AddModuleScore. P-values were determined using a Student‘s t-test. (c) Violin plots show the comparison of module feature expression levels in each corresponding region across the control, mild and severe groups. P-values were determined using a Student‘s t-test. (d) Group-labeled scatter plot visualizing the correlation between M2 and M11 AddModuleScores in the IM/Fib region. (e-f) The network shows the eigengenes in M2 and M11. The 50 hub genes are represented by coloured circles, while ligands and receptors are also highlighted. (g-h) Top 5 enriched gene ontology biological process (GOBP) pathway from M2 and M11. Ligands or receptors from hub genes involved in each pathway are labeled in the figure. (i-j) Module activity patterns were spatially mapped using SpatialFeaturePlots, displaying the tissue distributions of M2 and M11 features generated by AddModuleScore. HLA-DRA PTGDS IGHG3 HLA-DRB1 CXCR4 HLA-DPB1 IGKCIGHG1 HLA-DPA1 CD74 HLA-DQB1 IGLC2 HLA-DQA1 C1QB B2M IGLC3 C1QCHLA-B HLA-C JCHAIN MZB1 CCL5 SMAP2 CORO1A SSR4 CD53 HCLS1 IGHA1 TXNIP HCST C1QA ISG20 IGLC1 CCL19 LAPTM5 CD163 MS4A6A IGHG4 CD52 HLA-E ALOX15B ACAP1 NCF1 IL2RG ITGB2 MS4A4A HLA-DRB5 RNASE6 TYROBP LCP1HLA-F SRGN CSF1R CTSS TRBC2 CD37 IGHG2 HLA-DMA GMFG CD3D VSIR PTPRC AIF1 IL7R ARHGDIB PECAM1 SIGLEC1 CD79A MS4A7 SLCO2B1 IKZF1 CCL21 C1orf162 POU2AF1 TENT5C CELF2 ATP2A3 CRIP1 ARHGAP45 GZMKTRAC FXYD5 STK17B CD2 GPR34 VWF IL16 KLF2 GZMA ADA2 FCGR2B CD48 CD27 RNASET2 SESN1 MEI1 SLC1A3 CALHM6 RAC2 CXCL12 SLA ZFP36L2 TRBC1 LCP2 GPR183 IKZF3 CCL14 DERL3 SLAMF7 FCRL5 GIMAP5 MPEG1 H1FX PARM1 MGPLUM AEBP1 IGFBP7 TMSB4X VIM DCN C1RC1S SERPINF1 IFI27 COL3A1 MOXD1 MT2ACOL1A2 TSC22D3 CYBRD1TIMP2 A2M IFITM3FAU PDGFRA TPT1 C7 ISLR BGN ISG15 CTSK PLTP TMSB10 LY6E CYP1B1 LRP1 PRELP LHFPL6 IGFBP4 C12orf57 SELENOP FRZB CCDC80 DEPP1 FTH1 ADH1B VCAN MT1E ADGRE5 BTF3 TMEM204 RAPGEF5 IFNGR1 WASF2 IFI44 MIR100HG GGT5 FMOD IFI6 COL14A1 PALLD CST3 CD99 COL1A1 DDIT4 ZFAS1 TAGLN TNXB TPM2 PTEN PDGFRB KCTD12 HOPX MMP2CEBPBGLULCEBPD BST2 PHLDA1 SPARCL1 UBC ANXA1 CCDC3 FKBP5 MFAP4 GAS5 RACK1 MT1M NBL1 MYLK COL6A2 TIMP1 IL6ST MT1X SPARC TGFBR2 ENG MARCKS CD302 PROS1 THBS2IFI44L MATN2 FBLIM1 IFITM1 CCND2 FBN1 FLNA FSTL1 EMILIN1 CALD1 COL6A3 PPP1R14A DKK3 SERPING1 NR2F1 IGFBP5 NFKBIA TCF21RARRES2 SLIT3 LBH ELN CD81 COL16A1 IL1R1 PDK4 SNHG32 MX1 IFI16 NAP1L1 ACTA2 MDK ANTXR1 C11orf96 SVEP1 SNAI2 ABI3BP JUNB FHL1 ADGRA2 LAMA2 EDNRB LTBP4 TSC22D1 SYNPO2 PFDN5 P = 0.02P = 3.9  10-15 IM/Fib - M2 IM/Fib – M11 Myo – M3 Glo – M8 2.0 1.5 1.0 0.5 0.0 2.0 1.5 1.0 0.5 0.0 AddModuleScore 0.8 0.6 0.4 0.2 0.0 0.8 0.6 0.4 0.2 0.0 P = 1.6  10-10 P = 7  10-10 P = 0.00019 P = 0.016 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 M13 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Response to corticosteroid Negative regulation of growth Transforming growth factor beta stimulus Collagen metabolic process Extracellular matrix organization LUM, COL3A1, COL1A2, PDGFRA, LRP1, COL1A1 SERPINF1, A2M, COL1A1 COL3A1, COL1A2, COL1A1 COL1A2, COL1A1 M11 M11M2 Ligand Receptor Ligand/Receptor a f i j Scaled MEs 2.51.50.0-0.5 2.01.00.5 Percent expressed 2550 10075 c b P = 1.2  10-7 M2 M11 M3 M8 2.0 1.5 1.0 0.5 0.0 2.0 1.5 1.0 0.5 0.0 1.00 0.75 0.50 0.25 0.00 1.00 0.75 0.50 0.25 0.00 P = 1.9e  10-9 P < 2.22  10-16 P = 5.7  10-13 AddModuleScore d IM/Fib – M11 IM/Fib - M2 AddModuleScore 1.5 1.0 0.5 0.0 1.51.00.50.0 Control Mild Severe e g h -log10(q-value) Negative regulation of leukocyte apoptotic process 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 Antigen processing and presentation Positive regulation of leukocyte activation Cell killing Leukocyte chemotaxis CD74, HLA-DPB1, HLA-DRB1, HLA-DRA, B2M, HLA-DQA1, HLA-E, HLA-F CD74, HLA-DPB1, HLA-DRB1, HLA-DRA, B2M, HLA-DQA1, HLA-E, CCL5, ITGB2, HLA-F CD74, CXCR4, CCL5, ITGB2 CD74, CCL5 IGHG1, HLA-DRB1, HLA-DRA, B2M, HLA-C, HLA-B, HLA-E, HLA-F -log10(q-value) M2 -1 0 1 2 3 4 -1 0 1 2 3 4 AddModuleScoreAddModuleScore Control Mild Severe Control Mild Severe P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16 P < 2.22  10-16P < 2.22  10-16 P < 2.22  10-16 .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint control mild severe 0 1 2 3 4 5 CD74 % area fraction 1 2 3 CD74CXCR4 R = 0.53, P < 2.22  10-16 Scaled expression of CD74 Severe 2.5 3.0 3.52.0 4.0 2.5 2.0 1.5 1.0 0.5 0.0Scaled expression of CXCR4 P = 4  10-5 Control Mild Severe 2 4 6 0 CD74 P = 1.3  10-5 P = 0.66 Scaled gene expression Pearson’s coefficients of CXCR4-CD74 P < 0.0001 P < 0.0001 P < 0.0001 a b 0 1 2 3 4 CXCR4 CD74 0 1 2 3 4 d Severe c e g h j Severe 0 1 2 3 4 LUM COL1A2 0 1 2 3 4 l m n f i Control Mild Severe 1 2 3 0 CXCR4 Scaled gene expression P = 0.051 P = 6.9  10-5 P = 4.5  10-6 COL1A2 Control Mild Severe 2 4 6 0Scaled gene expression 1 2 3 0 Control Mild Severe 4 5 P = 1.1  10-8 P = 2.3  10-10 P = 0.00077 LUM Scaled gene expression CXCR4 CD74 IgM Control DAPI Mild CXCR4 CD74 IgM DAPI Severe CXCR4 CD74 IgM DAPI control mild severe 0 2 4 6 8 10 12 CXCR4 % area fraction 1 2 3 %Area fraction % Area fraction Control ControlMild MildSevere Severe P < 0.0001 P = 0.0259 P < 0.0001 1 2 3 0 4 5 2 4 6 0 8 10 12 Mild Severe 1 P = 0.0002 2 3 4 5 IF staining fields 0.1 0.2 0.3 Merged Merged Merged k R = 0.48, P < 2.22  10-16 R = 0.159, P = 0.121 R = -0.137, P = 0.627 1 2 3 0 4 1 2 3 0 4 1 2 3 0 4 Scaled expression of COL1A2 Scaled expression of LUM 1 2 30 4 1 2 30 4 1 2 30 4 Mild Severe Control Merged Mild Control Lumican Collagen I DAPI Merged Lumican Collagen I DAPI Severe Merged Lumican Collagen I DAPI control mild severe 0 1 2 3 4 5 LUM % area fraction 1 2 3 Lumican P < 0.0001 P = 0.4316 P < 0.0001 % Area fraction Control Severe 1 2 3 0 4 5 Mild Severe P = 0.0037 Pearson’s coefficients of lumican-collagen I 0.2 0.3 0.4 0.5 1 2 3 4 5 IF stained fields Mild control mild severe 0 2 4 6 8 collagen I % area fraction 1 2 3 Collagen I Control SevereMild P < 0.0001 P = 0.0413 P < 0.0001 2 4 0 6 8% Area fraction P < 2.22  10-16 Figure5 Fig.5: Ligand–receptor interactions reveal leukocyte chemotaxis and ECM remodeling in IM/Fib regions. (a) Spatially overlapped and scaled expression of CXCR4 and CD74 in a severe AAV patient. Expressions of CD74 and CXCR4 are visualized as an outer hollow circle and an inner solid circle, respectively. Fewer than 10 co-expressing spots were detected in mild AAV, precluding correlation analysis. (b) A close-up view (inset in a) reveals the spatial co-expression of CXCR4 and CD74 at spatial transcriptomic resolution. (c) Violin plots compare the expression levels of CXCR4 and CD74 across control, mild, and severe AAV groups in the IM/Fib. P-values were determined using a Student‘s t-test. (d) Scatter plots illustrate the correlation between CD74 (x-axis) and CXCR4 expression (y-axis) in spots co-expressing both markers (excluding spots where either CD74 or CXCR4 expression is zero) in the IM/Fib region of the severe AAV group. (e) Antibody-staining of CXCR4 (magenta), CD74 (red), and IgM (green) across control, mild, and severe groups. White boxes highlighted regions containing IgM⁺ cells (green) adjacent to CXCR4–CD74 co-localised cells (magenta/red) in representative mild and severe AAV patient samples. Scale bars represent 20 μm. (f) Relative protein expression levels of CXCR4 and CD74 (% area fraction of each antibody) across control, mild, and severe AAV groups. The three patients in each group are indicated by different point shapes. P-values were determined using a Student’s t-test. (g) Heatmap showing Pearson’s co-localisation coefficients of CXCR4 and CD74 protein expression in mild and severe AAV groups, with P-values determined using Student’s t-test between all immunofluorescence (IF) staining fields of the two groups and labeled on the top of the heatmap. The Y-axis represents measurements from five IF fields per patient, while the X-axis indicates patients from each group (n = 3 per group). (h-k) Same as in (a–d), measured in LUM and COL1A2. (l) Antibody-staining of lumican (green) and collagen I (red) across control, mild, and severe groups. Scale bars represent 50 μm. (m-n) Same as in (f-g), measured in lumican and collagen I. .CC-BY-NC 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted September 21, 2025. ; https://doi.org/10.1101/2025.09.19.677268doi: bioRxiv preprint