This study for scRNA-seq analysis was reviewed and approved by the Institutional Review Board of the Children’s Hospital of Chongqing Medical University (File Number: 2022 Research 124). The research for flow cytometric analysis was reviewed and approved by the Institutional Review Board of the Children’s Hospital of Chongqing Medical University. The File Number is 2022 Research 35. Written informed consent was obtained from all participants and their guardians.

In this study, we collected PBMCs from the blood and kidney tissues from three pediatric patients diagnosed with IgAN by renal biopsy, all of whom had IgAN with nephrotic-level proteinuria (NS-IgAN). The PBMCs were harvested from samples that remained after routine tests had been performed. Patients were classified as NGC, SGC, or LGC according to their glucocorticoid treatment at the time of sampling. We also included two kidney tissues as control kidney samples: one sample was Wilms’ tumor paracancerous tissue from a child (CTRL-1) while the other sample was from a resected duplex kidney (CTRL-2). All participants were enrolled from the Children’s Hospital of Chongqing Medical University, and each sample was collected for scRNA-seq individually. For comparative analysis, we downloaded scRNA-seq datasets from the Gene Expression Omnibus database. The kidney tissue datasets included one healthy adult (CTRL-3), four adult patients with IgAN (merged as HU-IgAN) from GSE171314, and PBMC datasets from three healthy children (Con-1, Con-2, Con-3) from GSE206295. Detailed information relating to the participants is shown in Supplementary Table 1 . To further investigate the changes in monocyte subsets in NS patients, we recruited a cohort of 13 healthy children, 26 NS patients with proteinuria, and 6 NS patients who were in remission after treatment. We collected blood from samples that remained after routine testing for flow cytometry.

To prepare PBMCs, we collected 2 mL of venous blood in EDTA collection vessels, which were then taken to the laboratory on ice. PBMCs were isolated using Ficoll medium (TBD, Tianjin, China) and cryopreserved according to the 10X genomics recommended protocol (CG00039). To dissociate the kidney tissue into single-cell suspensions, each fresh kidney sample was washed three times with Hanks’ balanced salt solution and immediately stored in GEXSCOPE tissue preservation solution (Singleron Biotechnologies, Nanjing, China) at 2–8°C. Then, the tissue was cut into small pieces and digested in 1 mL of Tissue Dissociation Mix (Singleron Biotechnologies) at 37°C for 15 minutes before being passed through a 40 μm filter. After centrifugation at 3500 g for 5 minutes, cell pellets were resuspended in 1 mL of cold PBS. To remove red blood cells, 2 mL of GEXSCOPE Red Blood Cell Lysis Buffer (Singleron) was added into the cell suspension and incubated at 25°C for 10 minutes. Cells were then centrifuged at 300 g for 5 min and resuspended in cold PBS. Next, cells were stained with trypan blue (Beyotime, Shanghai, China) and counted with a TC20 automated cell counter (Bio-Rad, California, USA). Sample processing and analysis were permitted once cell viability exceeded 85%.

Several different sequencing methods were used for the samples of PBMCs and kidneys. For PBMCs, each sample was diluted to a final concentration of 700–1200 cells µl⁻¹ and loaded onto a Chromium Single Cell Controller (10X Genomics, San Francisco, USA). The libraries for scRNA-seq were constructed using a Chromium Next GEM Single Cell 3´ GEM, Library and Gel Bead Kit v3.1 (10X Genomics) and then sequenced using an Illumina NovaSeq 6000 system. For each kidney sample, the single-cell suspension was adjusted to a concentration of approximately 300 cells µl⁻¹. A GEXSCOPE Single Cell RNA Library Kit (Singleron Biotechnologies) was then used to construct a single-cell RNA-seq library for kidney samples. The libraries were then sequenced with an Illumina HiSeq X 10 system. Each sample of PBMCs and kidney tissue was processed independently.

Raw sequencing reads from PBMCs were processed using Cell Ranger (version 6.0.0), including demultiplexing, genome alignment (GRCh38), barcode counting, and unique molecular identifier (UMI) processing. Similarly, raw data from the kidney tissues were processed by Celescope (version 1.10.0). We then used the Seurat (version 4.1.0) package to perform downstream analysis. To exclude low-quality cells, the cells were filtered by gene counts and UMI counts; cells with a high mitochondrial content were removed. Detailed information relating to the quality control (QC) threshold settings is given in Supplementary Table 2 . After cell filtering, 53,571 PBMCs and 47,602 renal cells were captured for downstream analysis. Next, we used DoubletFinder (version 2.0.3) to identify doublets and removed clusters with a high proportion of doublets. To remove batch-effects, we integrated sample datasets via the “integrate” function in Seurat. Principal component analysis (PCA) was then performed on the top 2000 highly variable genes; the top 20 PCs were used for subsequent analysis. We then used a graph-based clustering algorithm to identify clusters, thus allowing us to construct a K-Nearest Neighbor (KNN) graph by Euclidean distance. The Louvain algorithm was used to group cells and optimize modules. To display the distribution of cells by status, we used the uniform manifold approximation and projection (UMAP) algorithms to visualize clustered cells. First, cells were identified by SingleR (v1.6.1). This software compared the transcriptome of each cell cluster to various reference datasets (e.g., human primary cell atlas, Blueprint/ENCODE, Database of Immune Cell Expression, Novershtern hematopoietic data, and Monaco immune data). In order to exclude the influence of automatic assignment, we also manually adjusted the identities of clusters by combining data with the expression levels of canonical marker genes and existing annotated scRNA-Seq data ( Supplementary Tables 3–7 , 9 ).

Differentially expressed genes (DEGs) were computed using the FindMarkers function of Seurat. DEGs were defined as genes detected in at least 10% of cells, with a threshold of 0.25 log fold change using the Two-tailed Wilcoxon Rank-Sum Test with P < 0.01 following Bonferroni correction. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed with the Metascape web tool (www.metascape.org) and ClusterProfiler (version 4.2.2).

We used cell scores to evaluate the degree to which individual cells expressed a certain predefined expression gene set. This allowed us to define meaningful functions and states. The cell scores were calculated using the Seurat function “AddModuleScore”, which calculated the average expression of genes from the predefined gene set in the respective cell. The control gene sets were randomly selected based on aggregate expression level bins. The final gene set score was obtained by subtracting the control score from the predefined gene set score. We then used several well-defined EMT markers ( Supplementary Table 8 ) to define the EMT score.

To assess the statistical significance of scores, for each NS-IgAN patient or HU-IgAN group, the EMT scores were compared with that of the control group (CTRL) using the two-tailed Mann-Whitney U-test. Differences with a P value <0.05 were considered significant, *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001.

CellphoneDB was used with default parameters to reveal changes in interactions between different cell types.

R package Monocle2 (version 2.18.0) was used to perform pseudo-time analysis. To clarify the role of genes in cell fate decisions, branched expression analysis modeling (BEAM) from Monocle2 was applied.

The defined IgAN susceptibility genes were obtained based on previous genome-wide association studies (GWASs) in IgAN combined with NephQTL and eQTL (cf) online analysis tools, and were provided in Supplementary Table 10 . Cell-type specific expression of IgAN susceptibility genes was calculated by the average relative logarithmic expression values.

Human immortalized podocytes were provided from the lab of Zhihong Liu, and the cells were cultured as previously described (13). Cells were grown at the permissive temperature of 33°C (in 5% CO₂) and differentiated at 37°C (in 5% CO₂). After differentiating for 7-14 days, podocytes were treated with 0.25 µg mL^-1 doxorubicin (Sigma-Aldrich, Saint Louis, USA) and 0.5 µg mL^-1 doxorubicin for 24 hours.

The kidney tissue was dewaxed and then heat-mediated antigen repair was performed in sodium citrate solution (pH=6.0) for 15min, and blocked with goat serum. Cultured podocytes growing on a glass slide were fixed in 4% paraformaldehyde for 15 min. The sections and cells were incubated with the following primary antibodies: Prostaglandin D Synthase (PGDS) (ABclonal, Wuhan, China), anti-nephrin (PROGEN, Darmstadt, Germany) at 4°C overnight. Then incubated with the appropriate secondary antibody for 45 or 60 min at room temperature: anti-guinea pig IgG antibody conjugated with Alexa Fluor 568 (Invitrogen, California, USA), anti-rabbit IgG antibody conjugated with Alexa Fluor 633 (Invitrogen). Nuclei were counterstained with Hoechst (Thermo Fisher, Boston, USA). Fluorescence signals were viewed under a fluorescence microscope (Nikon A1R, Tokyo, Japan). NIS-Element (version 5.5) was used to quantify PGDS and nephrin staining intensity.

200μL whole blood was incubated with CD14-PE (BioLegend, California, USA) and CD16-APC (BioLegend, California, USA) antibodies. After staining for 20 minutes at room temperature in the dark, erythrocytes (BD Pharmingen, New Jersey, USA) in the samples were lysed by incubation with lysing solution for 5 minutes. Following centrifugation (300g/5 minutes, 4°C) and washing with PBS, cells were then examined using BD FACSCanto™. The datasets were analyzed using FlowJo (version 10.4.2).

Data are expressed as means ± standard deviation (SD). The EMT scores and gene expression levels were analyzed by the two-tailed Mann-Whitney U-test in SPSS 26.0. Differences in staining intensity between the two groups were analyzed by the two-tailed Student’s t test. All other analyses involved the two-tailed Wilcoxon Rank-Sum Test. Significance was defined as *P<0.05, **P<0.01, ***P<0.001. No specific indication is given if data were not significant. Graphs were generated by ggplot2 (version 3.3.5) and pheatmap (version 1.0.12) in R.

scRNA-seq was performed on PBMCs and kidney cells from three pediatric patients, all of whom had IgAN with nephrotic-level proteinuria, hypoalbuminemia, and hyperlipidemia (NS-IgAN) ( Figure 1A ; Supplementary Table 1 ). Kidney biopsy revealed the mesangial deposition of IgA and extensive or partial foot process effacement ( Supplementary Figures 1A–C ). It is worth noting that all three patients showed the pathological characteristics of mesangial hypercellularity. These patients had accepted different treatments at the time point of sampling, one patient was not on glucocorticoid (GC) therapy (NGC), one patient was on short-term GC therapy (SGC), and one patient was on long-term GC therapy (LGC) ( Supplementary Figures 1D–F ). Following the scRNA-seq of PBMCs, the raw data of three NS-IgAN patients and three healthy controls (Cons) from GSE206295 were merged and 53,571 PBMCs were captured following QC ( Figure 1B ; Supplementary Table 2 ). We annotated 30 cell types of PBMCs based on the expression of canonical markers, including CD4⁺ T cells (CD3D, CD4), CD8⁺ T cells (CD3D, CD8A), natural killer (NK) cells (KLRB1, GNLY), B cells (CD19, CD79A), myeloid cells (CD68, LST1), γδT cells (CD3D, TRDC), and megakaryocytes (PPBP, PF4) ( Supplementary Figure 2A ; Supplementary Table 3 ). For the kidney dataset, we downloaded data from GSE171314 as the HU-IgAN group, including four adult IgAN patients with hematuria and no nephrotic-level proteinuria ( Supplementary Table 1 ). In addition, we performed scRNA-seq on the kidney tissues from two pediatric patients and included a single-cell public database (GSE171314) from one adult kidney as a control group (CTRL) ( Figure 1A ). The data from a total of ten kidney samples from three groups (CTRL, NS-IgAN, HU-IgAN) were integrated and 47,602 kidney cells were captured after QC ( Supplementary Table 2 ). There were 26 clusters after dimension reduction ( Supplementary Figure 2B ). Then, 16 major cell types were annotated with the expression of canonical marker genes ( Figure 1C ; Supplementary Table 4 ). The expression levels of canonical marker genes for all cell types in the PBMCs and kidney cells are shown in Figures 1D, E .

Our investigation of the composition ratio of each sample in the same cell type found that the composition of SGC and LGC samples in myeloid cell types was prominent ( Figure 1B ). In response, we re-clustered the myeloid cell types into 14 cell clusters ( Figure 2A ). According to the expression of canonical marker genes for each cluster, we defined the cell types of clusters ( Figure 2B , Supplementary Table 5 ). According to the expression levels of CD14 and FCER3A (encoding CD16 protein), we respectively defined classical (CMs, CD14 ^high, FCER3A ^neg), intermediate (IMs, CD14 ^high, FCER3A ^low), and non-classical monocytes (NCMs, CD14 ^low, FCER3A ^high) ( Figure 2B ) (14). Of these, the CMs were composed of four cell clusters. Cluster 5 expressed interferon-related genes (ISG15, MX1) and was defined as CD14⁺ IFN-related monocytes (IFN-Mon) ( Figure 2B ). Notably, we found that cluster 9 expressed CCL4, IL1B, ICAM1, and CXCL2. This cluster was defined as inflammatory monocytes (INF-M) ( Figure 2B ). Although the proportion of INF-M in myeloid cells did not differ significantly between disease and healthy samples ( Figure 2C ), we found that this group of cells was dominant in the disease group ( Figure 2A ).

Figure 2C shows the composition ratio of each monocyte subset to CD14⁺ cells in the controls and three cases. In NS-IgAN, the proportions of CM 1-4 and IMs tended to increase, and the proportion of NCMs tended to decrease; however, these trends were not statistically significant ( Figure 2C , Supplementary Figure 3A ). Previous evidence proved that monocyte subsets undergo changes in IgAN (15); our data suggested that similar changes may also exist in NS-IgAN. Since NS is the main clinical manifestation of NS-IgAN, we collected the peripheral blood of 26 NS patients with proteinuria, 6 NS patients with remission after treatment, and 13 healthy children to investigate changes in monocyte subsets in NS by flow cytometry. Monocytes were divided into classical monocytes (CMs, CD14⁺⁺CD16^-), intermediate monocytes (IMs, CD14⁺⁺CD16⁺), and non-classical monocytes (NCMs, CD14⁺CD16⁺⁺) through flow cytometry by the expression levels of CD14 and CD16 ( Supplementary Figure 3B ). The proportion of CMs in NS patients with proteinuria was significantly lower than that in healthy children, and the proportion of IMs and NCMs was significantly increased ( Figure 2D ). Following the glucocorticoid-induced remission of proteinuria, the proportion of CMs and NCMs returned to normal levels; the proportion of IMs was even lower than healthy controls ( Figure 2D ). Although the results of scRNA-seq and flow cytometry were different when compared between CMs and NCMs, the results consistently indicated a trend for an increased proportion of IMs. We found that IMs were characterized by high expression levels of VSIG4, HLA-DPA1, HLA-DPB1, and other MHC class II molecules ( Supplementary Figure 3C ). KEGG and GO enrichment analysis of the IMs showed that their function was mainly related to cytokine-cytokine receptor interaction, the chemokine signaling pathway, antigen processing and presentation, and MHC class II protein complex assembly ( Supplementary Figures 4A, B ). DEG analysis of IMs in the Cons and NS-IgAN groups showed that the expression of VSIG4, LYZ, HLA-C, FN1, and FCGRT was significantly increased in NS-IgAN ( Supplementary Figure 4C ). The significant up-regulation of DEGs in the IMs of NS-IgAN patients were related to oxidative phosphorylation by KEGG enrichment analysis ( Supplementary Figure 4D ). These results suggest that increased oxidative phosphorylation may be an important feature of IMs in NS-IgAN.

DEG analysis of monocytes in the Cons and NS-IgAN groups showed that the expression levels of CCR2 were significantly higher in the CMs (CM1-4), IMs, and NCMs of NS-IgAN patients ( Figures 2E, F ). CCR2 mediates monocyte chemoattractant recruitment to inflammatory regions and promotes the production of inflammatory cytokines (16). It has been reported that Ccr2-deficient mice with adriamycin-induced nephropathy showed reduced levels of injury, along with reduced macrophage and fibrocyte infiltration and inflammation in the kidney (17). To investigate the effect of high expression levels of CCR2 on monocytes in NS-IgAN, we divided monocytes into CCR2 ⁺ monocytes and CCR2 ^- monocytes for further analysis. We found that the proportion of CCR2 ⁺ monocytes was slightly increased in NS-IgAN, although this was not statistically significant ( Supplementary Figure 4E ). We also found that compared with CCR2 ^- monocytes, CCR2 ⁺ monocytes expressed high levels of LYZ and HLA-DQA2 ( Supplementary Figure 4F ). HLA-DQA2 encodes HLA class II alpha-chain proteins to constitute MHC class II molecules. A recent GWAS study of a Chinese IgAN cohort revealed significant associations between three HLA polymorphisms, thus indicating the extensive involvement of HLA-mediated immunity in IgAN development (18). Collectively these results suggested that the proportion of CCR2 ⁺ monocytes is increased in NS-IgAN and that this may participate in the pathogenetic process.

The proportion of Tregs in the peripheral blood of IgAN was significantly lower than that in controls, thus an increase in the proportion of Treg cells could improve clinical prognosis (19). We performed a re-clustering analysis of CD4⁺ T cells and identified 13 clusters ( Supplementary Figure 5A ); these were defined as eight cell types by the expression of canonical marker genes ( Figures 3A, B ; Supplementary Table 6 ). Of these, two clusters expressed FOXP3 and IL2RA; we defined these as Treg1 and Treg2 cells ( Figure 3C ). According to the canonical marker genes expressed by different subsets of Tregs (20, 21), we found that Treg1 cells expressed high levels of CCR7, SELL, and TCF7; which may indicate that the cells are in a naive state; Treg2 cells expressed high levels of CCR4 and HLA class II molecules, which may indicate that the cells are in an effector state ( Figure 3D ). The two groups of Treg cells consistently expressed some classical genes, including TIGIT, IKZF2H, and RTKN2, but also had molecular expression characteristics ( Supplementary Figures 5B, C ). We found that the proportion of Treg2 cells in NS-IgAN tended to be higher than that in the Cons group, while the proportion of Treg1 cells tended to be lower, although these differences were not statistically significant ( Figure 3E ). The results of GO enrichment analysis suggested that the functions of Treg2 cells were mainly related to peptide antigen binding, MHC protein complex, and antigen processing and presentation ( Figure 3F ). In Treg2 cells from NS-IgAN patients, we found that the expression levels of genes such as FOS, JUN, and JUNB were decreased, while those of CCR4 were upregulated ( Figure 3G ). Recent research reported that a significant increase in the number of GATA3⁺ Tregs in the kidney was closely related to disease remission (22). CCR4 is known as an important chemokine receptor that promotes the infiltration of GATA3⁺ Tregs in the kidney during the later phases of injury (22). Our results suggest that there may be an increased proportion of CCR4⁺ GATA3⁺ Tregs in the circulation of NS-IgAN patients, thus indicating that dynamic changes of this special subset of Tregs may be involved in recovering from kidney injury.

Considering the important role of B cells in IgAN, we also performed a re-clustering analysis of B cells ( Supplementary Figure 6A ). We defined 12 clusters as 6 cell types by the expression of marker genes ( Supplementary Figures 6B, C , Supplementary Table 7 ). Consistent with expectations, the IGHA1 encoding IgA was mainly expressed in plasma cells. Pseudo-time analysis of B cells suggested that plasma cells at the end of differentiation were increasingly dominant in the disease, and expressed IGHA1 and IGHA2 ( Supplementary Figures 6D–G ). We compared the expression of key genes related to IgAN in diseased and healthy B cells; analysis suggested that C1GALT1 expression was downregulated in IgAN; this may be related to the formation of Gd-IgA1 ( Supplementary Figure 6H ). In addition, the expression of IgG and IgA-related genes in the B cells of SGC and LGC patients was upregulated ( Supplementary Figure 6H ). Findings related to Treg cells and plasma cells are restrictive and limited by the number of captured cells, meaning that further research needs to be undertaken to validate these findings.

Understanding alterations in the molecular characteristics of kidney cells may help us to understand the downstream mechanisms of kidney injury in NS-IgAN. Considering the clinical characteristics of patients with NS-IgAN who have “nephrotic-range” proteinuria, we focused on the podocyte cluster that significantly expressed NPHS2 and FGF1. Podocytes specifically expressed NPHS2 and FGF1 at both protein and transcriptional levels ( Figures 1E , 4A, B ). Unfortunately, only two podocytes were captured in the SGC patient, and we did not include this group in the subsequent comparative analysis ( Supplementary Table 4 ). We found that some genes showed a downward trend in terms of their expression by podocytes in the NS-IgAN group, including the canonical marker genes (NPHS1 and CLIC5), the genes encoding membranous nephropathy-associated autoantigens (PLA2R1 and THSD7A), and genes that have not been extensively investigated in podocytes, such as PCOLCE2 and PTGDS ( Figure 4C ). Our data indicated that CCL2 was highly expressed in the podocytes of NS-IgAN patients ( Figure 4C ). It has been reported that the expression of CCL2 in podocytes is closely related to podocyte injury and proteinuria (24). In the adriamycin-induced mouse model of nephropathy, researchers found that CCL2 in the kidney may recruit the infiltration of inflammatory and pro-fibrotic bone marrow-derived cell populations through its receptor CCR2; furthermore, a deficiency of Ccr2 in mice can ameliorate renal injury (17). We detected high expression levels of CCL2 in NS-IgAN glomerular podocytes and significantly increased expression levels of CCR2 in circulating monocytes ( Figure 2F ), thus indicating that crosstalk may occur in patients between podocytes and circulating monocytes through the CCL2-CCR2 ligand receptor. Furthermore, CFH was highly expressed in the podocytes of LGC ( Figure 4C ). Complement factor H (CFH) is one of the important circulating regulators of the alternative pathway, serves as an essential cofactor for complement factor I (CFI)-mediated C3b cleavage (25). Podocytes produce functionally active complement components, such as CFH; these could influence the local glomerular complement activation and regulation (26).

We also found that genes related to epithelial-to-mesenchymal transition (EMT) were significantly upregulated in NS-IgAN, including TGFB1, CAV1, TAGLN, and COL1A1 ( Figure 4C ). We constructed an EMT gene set ( Supplementary Table 8 ) and compared the gene set scores of podocytes between different groups. The EMT scores of podocytes in the NGC and LGC were significantly increased, while the EMT scores of podocytes in HU-IgAN did not change significantly ( Figure 4D ). EMT is an important feature of podocyte injury (27). Recent studies have found that podocytes in the urine of patients with NS also have EMT characteristics (28). Our results suggest that EMT is an important molecular feature of NS-IgAN podocytes. Interestingly, the expression of PRSS23 was significantly elevated in NS-IgAN patients and to a greater degree than in HU-IgAN patients ( Figure 4C ). Bulk RNA-seq results also confirmed the reduced expression of PRSS23 in various glomerular diseases ( Figure 4E ). These results suggest that a novel serine protease encoded by PRSS23 may be related to podocyte injury.

By performing the scRNA-seq on kidney tissues, we found that PTGDS was only specifically expressed in the podocytes ( Figure 4F ). In addition, PTGDS was significantly downregulated in NS-IgAN podocytes (P < 0.0001) ( Figure 4G ). The glomerular transcriptome sequencing of various glomerular diseases, including membranous nephropathy (MN), focal segmental glomerulosclerosis (FSGS), and minimal change disease (MCD) confirmed that the transcriptional level of PTGDS was significantly decreased ( Figure 4H ). Hence, we reviewed the previously published scRNA-seq data of kidney tissues and found that PTGDS only expressed at high levels in human kidney podocytes ( Supplementary Figures 7A, B ). By using immunofluorescence and immunohistochemistry experiments, we were able to confirm the downregulation of PGDS (prostaglandin D2 synthase, encoded by PTGDS) in the glomeruli of patients with NS-IgAN ( Figures 4I, J ; Supplementary Figures 8A, B ). The induction of injury in human immortalized podocytes in vitro (by applying doxorubicin) also led to a significant reduction in the expression levels of PGDS ( Figures 4K, L ; Supplementary Figures 8C, D ). PGDS, also known as β-Trace protein (BTP), is an emerging novel marker for glomerular filtration rate (29). PGDS is produced de novo by both the glomeruli and LOHs in monkey kidneys (30), although the handling process of BTP in the kidney is uncertain. BTP is an eicosanoid that plays a role in a variety of important physiological processes, including vasodilation, inflammation, and adipogenesis (29). Our findings, and those of other researchers, indicate that PTGDS may represent a candidate marker gene for podocytes at the transcriptional level. PTGDS is expressed at high levels in normal podocytes and at far lower levels in injured podocytes.

We investigated cell-to-cell communication between kidney cells by applying CellphoneDB. In the CTRL group, the cell types present in the proximal nephrons and kidney immune cells exhibited obvious intercellular communication, whereas distal convoluted tubules, intercalated cells and principal cells located in the distal nephron did not ( Figure 5A ). Our results suggest that the intercellular communication of the cells associated with the proximal nephron may be stronger than that of the cells associated with the distal nephron. In the CTRL and HU-IgAN groups, PECs may have served as the center for intercellular crosstalk in kidney cells ( Figure 5A ). In the NGC and LGC patients, the center of intercellular crosstalk was replaced by podocytes and mesenchymal stromal cells (MSCs) ( Figure 5A ). In an injury state, podocytes were affected by collagen-related, extracellular matrix-related, and inflammation-related signals from PECs and MSCs, such as COL3A1, FN1, and CCL2 ( Figure 5B ). These findings were consistent with previous findings related to IgAN (7, 8).

Kidney immune cells were re-clustered and defined into eight cell subsets by the expression of canonical marker genes (31) ( Figure 5C , Supplementary Table 9 ). C1QA, C1QB, and C1QC are characteristically expressed in two subsets of tissue-resident macrophages (RTMs). Of these, RTM-1 expressed LYVE1, IGF1, and CD209, and RTM-2 expressed TREM2 and MARCO ( Figure 5D ). Our results suggest that FCN1 and VCAN may be specific markers of monocytes in the human kidney. Classical monocytes (CMs) in the kidney represented the main subset of monocytes and expressed CXCL3. In contrast, NCMs expressed CX3CR1 ( Figure 5D ). XCR1, CLEC9A, and IDO1 were expressed in Classical DC-1, and CD1C and CD1E were expressed in Classical DC-2. Plasmacytoid DC expressed IL3RA, CLEC4C, GZMB, and TCL1A. It is worth noting that cluster 12 simultaneously expressed the marker genes of B cells, including CD79A and MZB1, and also expressed the marker genes of T cells such as CD3D, CD3E, and TRAC ( Figure 5D ). These results indicated that this cluster may be composed of B lymphocytes and T lymphocytes; hence, we defined this cluster as lymphocytes. We performed a developmental analysis of eight groups of kidney immune cells and confirmed the difference between tissue-resident macrophages and monocytes ( Figure 5E ). Previous studies comprehensively demonstrated the molecular characteristics of myeloid cell subsets in the kidney (31–34); however, there was a lack of understanding of lymphocyte subsets in the kidney. We further performed pseudo-time analysis on lymphocytes and found that they were in three different states (states 1–3) in the kidneys ( Figure 5F ), with different molecular characteristics (clusters 1–3) ( Figure 5G ). Lymphocytes in state 3 significantly expressed genes related to cluster 1 ( Figure 5G ), which were related to B cell activation, phagocytosis, and recognition ( Figure 5H ). Lymphocytes in state 1 significantly expressed genes related to cluster 3 ( Figure 5G ). These results were related to the production of cytokines from T cells ( Figure 5H ). Surprisingly, state 2 lymphocytes significantly expressed genes related to cluster 2 ( Figure 5G ). These findings are related to T cell activation and actin filaments ( Figure 5H ). These results suggest that there may be a group of T cells with actin filament-related functions in the kidney. By applying a scRNA-seq technology platform, we were fortunate to capture immune cells in the kidney and provide a preliminary exploration of the molecular characteristics of lymphocytes. However, due to the limited number of immune cells captured from the kidney tissue involved in this study, we were not able to conduct further analysis.

Based on the SNPs detected in a recent GWAS meta-analysis of IgAN (18), we combined NephQTL and eQTL (cf.) to predict susceptibility genes that may be affected in different regions of the kidney tissue ( Supplementary Table 10 ) and mapped these genes to kidney cells ( Figure 6A ) and PBMCs ( Figure 6B ). In the kidney tissue, renal resident macrophages, macrophages, and DC cells expressed high levels of genes related to HLA ( Figure 6A ). In circulation, multiple B cell subsets (except plasma cells), cDCs, and pDCs expressed high levels of genes related to HLA, while monocytes did not ( Figure 6B ). Previous studies had identified HLA molecules as the main disease related susceptibility loci for IgAN (18). Research evidence also indicated that DCs play key roles in the pathogenesis of IgAN (35). Our present data showed that DCs express high levels of HLA-related genes in both the circulation and kidney tissue. This indicated that attention needs to be paid to the changes in DC cells in the circulation and kidney tissue of NS-IgAN patients in future research ( Figures 6A, B ). The CFH-related genes CFHR1, CFHR3, and CFHR4 were exclusively expressed in kidney MSCs ( Figure 6A ). We performed a re-clustering of the MSCs that significantly expressed POSTN and ACTA2 ( Supplementary Table 4 ; Supplementary Figure 9A ). This analysis suggested that MSCs were a mixed subset composed of fibroblasts, myofibroblasts, vascular smooth muscle cells, pericytes, and mesangial cells ( Supplementary Figure 9B ). Unfortunately, in our study, mesangial cells could not be defined by canonical marker genes such as PDGFRB, PIEZO2, ITGA8, and GATA3 ( Supplementary Figure 9B ).