Single-cell and high-resolution spatial profiling of podocytopathies reveals core mechanisms of podocyte injury

Abstract

Podocytes (PODs) play a critical role in maintaining glomerular filtration function, and their injury is a key driver of progressive podocytopathies (PCPs). Despite their clinical importance, the molecular and spatial mechanisms underlying POD injury remain poorly understood. We generated a single-nucleus RNA sequencing (snRNA-seq) and high-resolution spatial transcriptomic (ST) dataset from kidney tissues of patients with five major PCP subtypes. Our analysis revealed distinct molecular signatures among PCP subtypes, injury-related cytoskeletal alterations, and POD-centered microenvironmental features associated with clinical outcomes. Through integrative analysis, we found PDE4DIP as a previously unrecognized regulator of the POD cytoskeleton. Functional experiments through in vitro and in vivo gene editing demonstrated that PDE4DIP maintains cytoskeletal integrity by scaffolding with AKAP9 and activating RAS-ERK and AKT signaling. Its expression correlated with renal function and prognosis. This study provides the comprehensive snRNA-seq and high-resolution ST atlas of PCPs, offering a valuable resource for understanding POD injury and identifying potential therapeutic targets.

PDE4DIP emerges as a potential target for protecting podocytes and preserving kidney filtration.

INTRODUCTION

Kidney disease remains a global health burden with rising prevalence (1). The integrity of the glomerular filtration barrier is essential for normal renal function (2), and podocytes (PODs), with their complex foot processes, constitute a critical component of this barrier (3). Structural or functional abnormalities in PODs compromise the filtration barrier, leading to massive proteinuria and nephrotic syndrome (4). Without timely and effective intervention, this can progress to chronic kidney disease (CKD) even end-stage renal disease (5).

POD injury is observed in nearly all glomerular diseases, but only when this injury serves as the initiating or dominant pathological event is it defined as a podocytopathy (PCP) (6). The underlying mechanisms are multifactorial, involving genetic (7), immune (8), and metabolic factors (9). Canonical PCPs include minimal change disease (MCD) and primary focal segmental glomerulosclerosis (FSGS), while other glomerulopathies such as membranous nephropathy (MN), diabetic nephropathy (DN), and obesity-related nephropathy (OBN) also exhibit marked POD damage (10). Given the limited regenerative capacity of PODs and the incomplete understanding of the molecular underpinnings of PCPs, current treatments rely heavily on immunosuppression and supportive care, with limited success in reversing injury (11).

Emerging single-cell/nucleus RNA sequencing (sc/snRNA-seq) and spatial transcriptomic (ST) technologies have provided transformative tools for dissecting renal pathophysiology. However, while these tools have been applied to models of acute kidney injury (AKI) and CKD, the patients with CKD have been primarily caused by hypertension or diabetes (12, 13). A comprehensive sc/snRNA-seq and ST atlas specific to PCPs has been lacking. This limitation hinders the mechanistic understanding of POD injury, as well as the development of targeted therapeutic strategies for POD repair and protection.

Here, we performed snRNA-seq and ST on kidney tissue from both human and murine models of PCP, generating the integrated transcriptomic atlas of POD injury (Fig. 1A) . This atlas enabled us to characterize conserved molecular and functional features of POD damage, map POD-centered microenvironments, and identify potential diagnostic and therapeutic targets.

Fig. 1. snRNA-seq atlas of PCPs in human and mouse kidneys.

(A) Study design overview. Q. Jianbo, Created in BioRender (2026); https://BioRender.com/c60b094. (B) Uniform Manifold Approximation and Projection (UMAP) plot of 155,922 nuclei from human kidney tissues across five PCP types, classified into 21 cell types. ART, artery; CNT, connecting tubule; DCT, distal convoluted tubule; FIB, fibroblast; GEC, glomerular capillary EC; ICA/ICB, intercalated cells A/B; LYM, lymphatic cell; MD, macula densa; PC, principal cell; PEC, parietal epithelial cell; PT-Inj/R, injured/repairing PT; PTC, peritubular capillary EC; TAL, thick ascending limb; TL, thin limb. (C) Bubble plot of marker gene expression in human POD subtypes (cyan: canonical; red: dedifferentiation; blue: endothelial related; pink: tubule related). (D) Stream plot of human POD fate trajectory. (E) T-distributed stochastic neighbor embedding (T-SNE) plot of human POD differentiation potential. (F) Proportion of POD subtypes in control and PCP groups. (G) UMAP of 141,834 nuclei from mouse kidney PCP models, classified into 24 cell types. (H) Bubble plot of marker gene expression in mouse POD subtypes. (I) Stream plot of mouse POD pseudotime trajectory. (J) T-SNE plot of mouse POD differentiation scores. (K) Jaccard similarity heatmap between human and mouse POD subtypes. (L) Proportions of POD subtypes in four mouse groups. (M) Correlation between POD subtype proportions and Scr/UACR levels. Cor, correlation. (N) Simulated fluorescence-activated cell sorting (FACS) plot of POD gene expression in control, patients with AKI, and patients with CKD. (O) Multiplex immunofluorescence images of kidney tissues from various sources. (P) Pathway activity scores in human POD subtypes. (Q) Expression of HIF1A and EPAS1 in human and mouse PODs. (R) mRNA levels of HIF1A, HIF2A, PECAM1, and LRP2 in cultured human PODs treated with hypoxia/HIF inhibitors. Data are mean ± SD (n = 5), analyzed by two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons. *P < 0.05, **P < 0.01, and ***P < 0.001 versus the hypoxia group; #P < 0.05 and ##P < 0.01 versus the inhibitor group.

RESULTS

Single-nucleus map of PCP and POD subtypes

We collected kidney biopsy samples from 24 patients diagnosed with five representative forms of PCPs (MCD, FSGS, MN, DN, and OBN) and adjacent healthy kidney tissue from four patients with renal clear cell carcinoma for snRNA-seq (n = 18) and 10× Visium HD ST (n = 15). All patients had an estimated glomerular filtration rate (eGFR) >90 ml/min per 1.73m² at tissue collection and were followed for 80 to 896 days to monitor eGFR decline (table S1). In addition, we established adriamycin-induced nephropathy (ADRN) and OBN mouse models and sequenced them to observe POD changes at the animal level.

Human snRNA-seq data were integrated using Harmony, demonstrating excellent performance (fig. S1A). After quality control, the Leiden algorithm was applied to cluster 155,922 nuclei into 45 clusters (fig. S1, B and C). These clusters were further categorized into 21 cell types, including epithelial cells, endothelial cells (ECs), immune cells, and stromal cells (SCs), based on classical markers (Fig. 1B, fig. S1D, and tables S2 and S3). A total of 2003 POD nuclei were captured, and except for MN, POD proportions in PCPs lower than in controls (fig. S1, E and F).

After determining the optimal dimensions and resolution, the 2003 PODs were divided into three subgroups (fig. S1, G to I), two of which exhibited reduced expression of classic POD markers (e.g., WT1 and NPHS1) and elevated expression of dedifferentiation markers (e.g., B2M and S100A6), classified as dedifferentiated PODs (dPODs). The remaining subgroup, with normal POD markers, represented healthy PODs (hPODs). Compared with hPODs, dPODs expressed high levels of endothelial (CD34, FLT1, and PECAM1) and tubular (LRP2, CUBN, and SLC13A1) markers (Fig. 1C and table S4) (13).

Thus, two dPOD subgroups were separately termed PODs expressing endothelial-related genes (ePODs) and tubular-related genes (tPODs). Enrichment analysis showed hPOD enrichment in kidney development, ePOD in endothelial differentiation, and tPOD in substance catabolism (fig. S1J). Notably, endothelial- or tubular-related genes were notably up-regulated in dPOD only when compared to hPODs, but their expression remained low relative to normal proximal tubule (PT) cells and ECs (fig. S1K). Pseudotime analysis further indicated that hPOD represents the starting point of the POD fate trajectory, while tPOD represents the terminal stage (Fig. 1, D and E, and fig. S1L). Notably, among all disease groups, dPODs were more prevalent in DN and MN (Fig. 1F).

We also performed snRNA-seq on kidney tissues from ADRN and OBN mice and integrated these with Humphreys’ DN mouse data (fig. S2, A and B) (14). This resulted in the classification of 141,834 mouse kidney cell nuclei into 24 types (Fig. 1G, fig. S2C, and table S5), with POD proportions lower in disease groups (fig. S2D). Similarly, 1006 POD nuclei were divided into three subgroups (fig. S2, E to G), with gene expression and biological process patterns consistent with human POD subtypes (Fig. 1H, fig. S2H, and table S6). Pseudotime analysis showed similar fate trajectories in both species (Fig. 1, I and J), with strong cross-species similarity in POD subtypes (Fig. 1K). In all mouse PCP models, the proportion of ePOD and tPOD was higher (Fig. 1L and fig. S2I). Correlation analysis revealed that hPOD proportion was negatively correlated with the urinary albumin–to–creatinine ratio (UACR) and serum creatinine (Scr), while tPOD proportion was positively correlated (Fig. 1M).

To validate our findings, we used kidney sc- and snRNA-seq data from the Kidney Precision Medicine Project (KPMP) database (13), including patients with AKI and patients with CKD, and extracted PODs (fig. S1M). In both AKI and CKD PODs, tPOD markers were up-regulated, especially in AKI. DPOD markers were mainly expressed in AKI PODs, while ePOD-related genes showed less difference (Fig. 1N and fig. S1N). These results indicate that POD undergoes dedifferentiation in both AKI and CKD.

Hypoxia-HIF axis drives POD dedifferentiation

We performed multiplex immunofluorescence staining on adjacent healthy kidney tissues from patients with renal clear cell carcinoma, kidney biopsy tissues from patients with MCD, and kidney tissues from ADRN mice and their healthy controls. Given the high spatial overlap between POD and EC, we focused on observing the presence of tPOD. In both patients with PCP and ADRN mice, we observed mild expression of LRP2 in POD regions (Fig. 1O).

To elucidate the mechanisms underlying POD dedifferentiation, we conducted gene set enrichment scoring on human and mouse PODs, incorporating six major pathways related to POD dedifferentiation: epithelial–mesenchymal transition (EMT) (15), hypoxia (16), NOTCH (17), phosphatidylinositol 3-kinase (PI3K) (18), transforming growth factor–β (TGF-β) (19), and WNT (20). In both human and mouse PODs, dedifferentiation pathways were significantly activated in ePOD, particularly the TGF-β signaling pathway, which is well documented for its key role in POD and kidney injury (21). However, it is important to note that the activity of pathways in tPOD, excluding hypoxia, was generally low, even lower than in hPOD. In contrast, the hypoxia pathway in tPOD was significantly up-regulated in both human and mouse models compared to other POD subtypes, with ePOD and tPOD in mice showing significantly higher activation of the hypoxia pathway than hPOD (Fig. 1P and fig. S2J). Various conditions leading to kidney injury can create hypoxic environments, triggering the activation of hypoxia-inducible factor (HIF) pathways, which are crucial for cellular adaptation and tissue repair during disease. However, prolonged activation of HIF signaling may exacerbate kidney injury and promote fibrosis (22, 23).

We observed distinct expression patterns of two key HIF pathway transcription factors, HIF1A (HIF-1α) and EPAS1 (HIF-2α), in PODs. In both human and mouse models, EPAS1 was primarily expressed in ePOD but was not prominent in AKI- or CKD-derived PODs. Conversely, HIF1A was up-regulated in tPOD across species and also showed significant expression in mouse ePOD. Moreover, HIF1A expression was markedly higher in PODs from patients with AKI and patients with CKD compared to controls (Fig. 1Q and fig. S1N).

Subsequently, we cultured human PODs in vitro and exposed them to hypoxia and HIF pathway inhibitors. After 48 hours of hypoxia treatment, we observed significant up-regulation of HIF1A, EPAS1, LRP2, and PECAM1 in PODs. Notably, both HIF1-α and HIF2-α inhibitors significantly suppressed the up-regulation of LRP2 and PECAM1, confirming that hypoxia is a key mechanism driving POD dedifferentiation (Fig. 1R).

PDE4DIP expression links POD cytoskeleton to proteinuria and disease progression

We calculated the differential genes between ePOD and tPOD relative to hPOD in both human and mouse samples. In human and mouse ePODs, 304 genes were up-regulated and 178 genes down-regulated in common, while tPODs showed 477 up-regulated and 370 down-regulated common genes. The differential genes between human and mouse PODs were highly consistent, primarily related to kidney development, and significantly enriched in cellular components associated with the cytoskeleton, such as the cell leading edge and cell cortex (fig. S3, A to D). We identified 85 genes that were significantly up-regulated and 131 genes that were significantly down-regulated in both ePOD and tPOD (Fig. 2, A and B), all of which are closely associated with POD development and cytoskeleton formation (Fig. 2C). POD foot process fusion and disappearance are key features of POD dedifferentiation and PCP, with the core mechanism involving cytoskeletal disruption (24, 25). We also compiled 158 genes related to the POD cytoskeleton, categorized into five groups: actin, POD-specific actin components, intermediate filaments, microtubules, and slit diaphragm (table S7). Apart from POD-specific actin components and slit diaphragm-related genes, only PDE4DIP and PALLD showed significant down-regulation in dPODs (Fig. 2D).

Fig. 2. Alterations in POD cytoskeleton genes in human and mouse PCPs.

(A) Volcano plot of differentially expressed genes (DEGs) between human and mouse hPOD and ePOD and between hPOD and tPOD [P < 0.05, |log₂ fold change (FC)| ≥ 0.25 considered significant]. (B) Quadrant plot comparing DEGs in dPODs versus hPODs in human and mouse datasets. (C) Gene Ontology enrichment analysis of genes down-regulated in dPODs. BP, biological process; CC, cellular component. (D) Circular heatmap of cytoskeleton gene expression in human POD subtypes, with Venn diagram of human and mouse dPOD DEGs. (E) Bubble plot of gene expression in POD subtypes for cytoskeleton regulation modules. (F) UMAP of coexpression network structure for cytoskeleton genes and WT1. (G) PDE4DIP expression across kidney cell types in human and mouse snRNA-seq atlases. (H) Boxplots of PDE4DIP expression in human and mouse snRNA-seq atlases and the KPMP dataset. (I and J) Correlation between PDE4DIP mRNA (bulk glomerular RNA-seq) and eGFR (I) glomerulosclerosis (J). FPKM, fragments per kilobase of transcript per million mapped reads. (K and L) Correlation of PDE4DIP protein (glomerular proteomic profiling) with eGFR (K) and glomerulosclerosis (L). (M) PDE4DIP expression in 18 mouse kidney injury models. (N) PDE4DIP and eGFR correlation in nephrotic syndrome patients (NEPTUNE). (O) Survival analysis based on PDE4DIP expression. Progression to end-stage renal disease was defined as the clinical outcome. (P) Immunohistochemical quantification of PDE4DIP in glomeruli from MCD (11) and FSGS (7) patients versus controls (4). Data are mean ± SD, one-way ANOVA followed by Tukey’s multiple comparisons test. **P < 0.01 and ****P < 0.0001. (Q) Correlation of PDE4DIP protein with eGFR and proteinuria. (R) Integrative analysis of snRNA-seq and GWAS using seismicGWAS. (S) Venn diagram of shared genes associated with UACR/microalbuminuria across GWAS–snRNA-seq datasets. SNP, single-nucleotide polymorphism. (T) Multiplex immunofluorescence images of glomeruli from five PCP types and controls. (U) Quantification of SYNPO and PDE4DIP intensity. Data are mean ± SD (n = 5), one-way ANOVA followed by Tukey’s multiple comparisons test. ***P < 0.001.

We then constructed gene coexpression networks for human PODs, identifying seven modules (M1 to M7; fig. S3, E to G, and table S8). The expression of genes in modules M1, M2, M6, and M7 was significantly reduced in dPODs (Fig. 2E). Notably, the core POD transcription factor WT1 (26), as well as slit diaphragm components NPHS1/2 and PDE4DIP, were coexpressed within the same module (Fig. 2F), which exhibited the lowest gene expression levels in both DN and MN (fig. S3H). PDE4DIP was predominantly expressed in PODs in both humans and mice (Fig. 2G), with significantly reduced expression in the disease groups across our human and mouse datasets, as well as in KPMP data (Fig. 2H). In addition, bulk RNA-seq and proteomic data from 485 CKD patient glomeruli revealed a significant positive correlation between PDE4DIP mRNA and protein levels with eGFR and a significant negative correlation with glomerular sclerosis (Fig. 2, I to L). In various nephropathy mouse models, including those induced by folate, unilateral nephrectomy, and lipopolysaccharide (LPS), Pde4dip expression in PODs was markedly decreased compared to healthy controls (Fig. 2M). The Nephrotic Syndrome Study Network (NEPTUNE) database follow-up data also demonstrated a significant positive correlation between PDE4DIP expression in glomeruli and eGFR, with patients exhibiting lower expression levels being more prone to progress to end-stage renal disease (Fig. 2, N and O).

Subsequently, we performed immunohistochemical staining on renal tissues from patients with FSGS (n = 7) and MCD (n = 11), along with four adjacent normal controls. The results confirmed a significant positive correlation between glomerular PDE4DIP expression and eGFR; conversely, a negative but nonsignificant correlation was observed with 24-hour urinary protein (Fig. 2, P and Q, and fig. S2I). Notably, while these immunohistochemical results provide preliminary evidence suggesting an association between glomerular PDE4DIP expression and PCP, they do not definitively establish a causal relationship between POD-specific PDE4DIP expression and PCP. We also integrated genome-wide association study (GWAS) data on UACR from European American and African American populations, along with cross-ethnic data on microalbuminuria, and performed joint analysis with human snRNA-seq data using the seismicGWAS package (Fig. 2R) (27). We inferred a potential genetic association between PDE4DIP and proteinuria (Fig. 2S and table S9). Furthermore, multiplex immunofluorescence staining further confirmed a significant reduction in PDE4DIP expression in PODs across five typical PCPs compared to controls (Fig. 2, T and U).

ST profiling of PCPs

We performed ST sequencing on kidney tissues from 14 patients with PCP (MCD, FSGS, MN, DN, and OBN) and one healthy control. Analysis was conducted in 16-μm mode, and after quality control (fig. S4A), a total of 298,596 bins (cells) were obtained. Using deconvolution, 23 cell types from the human snRNA data were annotated in the ST data (fig. S4B). Figure 3A shows partial regions of 15 samples, with complete ST images and representative glomeruli hematoxylin and eosin (H&E) images in fig. S4C. In the ST data, the proportion of hPOD was significantly lower in all five PCPs compared to controls, whereas ePOD was significantly increased. In addition, we observed a significant increase in the proportion of fibroblast (FIB), macrophage (MAC), injured PT (PT-inj), and T cells in PCPs, particularly in DN and MN (Fig. 3, B and C, and fig. S4, D and E). In addition, glomeruli in both datasets were classified by sclerosis status. Compared to nonsclerotic samples, hPOD proportions in sclerotic glomeruli dropped from 0.92 to 0.60 in humans and from 0.98 to 0.89 in mice (fig. S4F).

Fig. 3. High-resolution ST landscape of kidneys from human and mouse PCPs.

(A) Spatial maps of kidney tissues from control and five patients with PCP, with 23 cell types annotated using robust cell type decomposition (RTCD) at 16-μm resolution; this image is also used in part in Fig. 1A. (B) Heatmap of Ro/E (observed/expected) indices for 23 cell types across control and patients with PCP. (C) Line plot showing changes in the proportions of POD subtypes, FIB, PT-Inj, MAC, and T cells in control and PCP kidneys. (D) Heatmap of Pearson correlations between clinical parameters [eGFR, eGFR slope, 24-hour urinary albumin (AU), 24-hour urinary total protein (TPU), age, and Scr] and 23 cell types. *P < 0.05 and **P < 0.01 [false discovery rate (FDR)]. (E) Simulated FACS plot of gene expression in PODs from patients with PCP. (F) Spatial maps of tPODs in patients with PCP, with corresponding H&E-stained regions. (G) Bubble plot of marker gene expression in three human POD subtypes in ST data (cyan: healthy markers; blue: endothelial related; pink: tubule related). FOC, fraction of cells. (H) Stacked bar plot showing relative proportions of POD subtypes in control and PCP kidneys. (I) Spatial maps of control and ADRN mouse kidneys, with 26 annotated cell types (8-μm resolution). (J) Bubble plot of marker gene expression in three mouse POD subtypes in ST data. (K) Stacked bar plot showing POD subtype proportions in control and ADRN mouse kidneys. (L) Line plot showing changes in POD subtypes, FIB, PT-Inj, MAC, and T cells in control and ADRN mouse kidneys. (M) Spatial feature plots of Pde4dip expression in control and ADRN mouse kidneys. (N) Bubble plot of Wt1, Nphs1/2, and Pde4dip expression in PODs from control and ADRN mouse kidneys. Color indicates the expression level of the corresponding gene in bins annotated as PODs.

The proportions of PT-inj, T cell, and MAC were significantly positively correlated with patient age and negatively correlated with eGFR slope, while the proportion of normal PT cells showed a significant positive correlation with the eGFR slope (Fig. 3D). In the ST data, the identified dPODs were predominantly categorized as ePODs, with five tPODs detected, which may be related to the limited sensitivity of ST detection for relatively low-abundance transcripts in PODs (Fig. 3E and fig. S4G) (28). Because Visium HD partitions tissue into spatial bins rather than performing cell segmentation, signals from ECs and PODs may overlap spatially. In contrast, tubular epithelial cells are absent from the glomerulus, allowing the spatial localization of tubule-associated gene signals to indicate the presence of tPOD populations (Fig. 3F).

The gene expression patterns of POD subtypes in the ST data were largely consistent with those from snRNA-seq and in the PCP group (Fig. 3G), and the proportion of ePOD was more than twice that of the control (Fig. 3H). The proportions were calculated from all PODs detected in each Visium HD dataset, as the analysis was performed on whole tissue sections without manual glomerular segmentation, thus including PODs from both nonsclerotic and sclerotic glomeruli.

We also performed ST sequencing on kidney tissues from one ADRN mouse and one healthy control mouse, analyzing the data in 8-μm mode, with quality control and deconvolution (fig. S3H). A total of 527,831 cells were annotated into 26 cell types, and we identified hPODs and dPODs in both ADRN and control mice (Fig. 3I and fig. S4I). The gene expression patterns of mouse POD subtypes mirrored those observed in humans (Fig. 3J). Notably, the proportion of dPODs was significantly higher in ADRN mice compared to controls (Fig. 3K and fig. S4J). In line with the human data, ADRN mice also exhibited marked increases in the proportions of FIB, MAC, T, and PT-inj cells, with particularly pronounced expansion of FIB and PT-inj populations (Fig. 3L and fig. S4K).

Notably, while Pde4dip expression was not detected in the human ST data, it was reliably identified in mouse samples, albeit in a small subset of cells. Its expression exhibited a graded decrease across hPOD, ePOD, and tPOD subtypes, respectively (Fig. 3J). At the disease level, although the expression of canonical POD markers (Nphs2, Podxl, and Wt1) was significantly reduced in ADRN PODs compared to control, Pde4dip expression was relatively higher in the ADRN group. This may be attributed to its restricted expression in a limited number of cells, and consistent with this, the expression differences of Pde4dip between groups were less pronounced (Fig. 3, M and N).

Spatial mapping of the POD-centered microenvironments

We defined the POD-centered microenvironment as the composition of cell types surrounding PODs (Fig. 4A). Although PODs typically neighbor glomerular cells like ECs, PECs, and SCs, spatial binning revealed additional cell types within POD-proximal regions. Given the relevance of FIBs, PT-inj cells, MACs, and T cells in PCPs and their clinical associations, we analyzed their abundance in the POD-centered microenvironment.

Fig. 4. Identification of POD-centered microenvironments.

(A) Schematic diagram illustrating the approach for defining the POD-centered microenvironments in human ST data, based on counting the number and types of cells directly adjacent to each POD. (B) Heatmap showing mean numbers of neighboring cell types within POD-centered microenvironments in control and PCP kidneys [log₁₀(count + 1^e−4) + 4]. (C) Bubble plots of normalized abundance of neighboring cell types across control and PCP groups. (D) Heatmap of Pearson correlations between proportions of 23 cell types within POD-centered microenvironments and clinical parameters (eGFR, eGFR slope, SU, TPU, age, and Scr); FDR corrected (*P < 0.05). (E) Bar plot showing the number of identified POD per ST sample and the proportion of POD whose POD-centered microenvironment lacked any of the following four cell types: FIBs, PT-Inj cells, MACs, or T cells. (F) Representative spatial network plots showing FIBs, PT-Inj cells, MACs, and T cells around PODs. Each point represents a single cell. (G and H) Heatmaps of number (G) and interaction strength (H) of inferred cell-cell communications between PODs and neighboring cell types (FIBs, PT-Inj cells, MACs, and T cells). (I) Bubble plot showing ligand-receptor pairs with significantly increased communication probabilities between PODs and the four neighboring cell types (FIBs, PT-Inj cells, MACs, and T cells) in PCP kidneys compared to controls. (J) Bubble plots displaying the average normalized abundance of neighboring cell types in the POD-centered microenvironments of control and ADRN mouse kidneys. Values were normalized as in (B). (K) Bubble plot showing ligand-receptor pairs with significantly increased communication probabilities between PODs and the four neighboring cell types (FIBs, PT-Inj cells, MACs, and T cells) in ADRN mouse kidneys compared to control.

Across five human PCPs, PT-inj cells and FIBs (except in OBN) were significantly increased around PODs compared to controls. DN showed the highest abundance of FIBs, PT-inj cells, MACs, and T cells (Fig. 4, B and C). The microenvironment of ePODs contained more PT-inj cells, FIBs, and T cells than hPODs but fewer MACs (fig. S5, A and B). Increased POD-centered microenvironment abundance of these cells correlated negatively with eGFR slope, suggesting a role in disease progression (Fig. 4D).

We further quantified the abundance and proximity of these cells to PODs and constructed spatial networks. DN exhibited the highest proportion of PODs adjacent to these cells, while OBN had the lowest (Fig. 4, E and F, and fig. S5C). Cell communication analysis revealed prominent interactions between PODs-MACs and PODs–PT-inj cells, particularly in MCD (Fig. 4, G and H). Key ligand-receptor pairs included APP-CD74 (PODs-FIBs and PODs–PT-inj cells) and SELL-PODXL (MACs-PODs and T cells-PODs). APP-TNFRSF21 also mediated interactions between PODs and all four cell types (Fig. 4I).

Similarly, we analyzed the POD-centered microenvironment in mice. The proportion of FIBs, PT-inj cells, and MACs was increased in the ADRN POD-centered microenvironment (Fig. 4J), although these changes were not significant across different POD subtypes (fig. S5, D and E). Similar to humans, the mouse T cells communicated with PODs through SELL-PODXL, but in mice, PODs communicated with PT-inj cells via APP-CD74 (Fig. 4K).

Furthermore, multiplex immunofluorescence staining was performed on kidney sections from control subjects and patients with PCP to validate the interaction between PODs and MACs. Representative images revealed that, in control glomeruli, PODs and MACs showed minimal direct contact. In contrast, in glomeruli from patients with PCP, marked MAC infiltration was observed, with MACs closely adjacent to PODs (fig. S5F).

The core role of PDE4DIP in maintaining POD cytoskeletal stability

Through correlation analysis of human glomerular bulk RNA-seq data, we identified 1258 genes significantly positively correlated with PDE4DIP expression (Fig. 5A and table S10), and 114 of which were the same as genes in the coexpression network containing PDE4DIP (Fig. 5B). These 114 genes are mainly involved in pathways closely related to the cytoskeleton, such as renin–angiotensin system (RAS) and PI3K-AKT signaling (Fig. 5C). PDE4DIP interacts with AKAP9 to promote protein kinase C (PKC) phosphorylation, which, in turn, activates downstream RAS signaling, leading to the activation of ERK and AKT pathways, thereby maintaining cytoskeletal integrity and stability (Fig. 5D) (29). To simulate POD injury under various conditions, we treated human PODs with four different stimuli: ADR, angiotensin II (ANG II), hypoxia, and LPS, each of which significantly reduced the expression of PDE4DIP and AKAP9, as well as the phosphorylation of PKCε, ERK, and AKT (Fig. 5E and fig. S6, A to C). Quantitative polymerase chain reaction (qPCR) analysis confirmed that these factors caused a significant reduction in PDE4DIP and AKAP9 mRNA levels in PODs (fig. S6, D and E), along with decreased RAS activity (Fig. 5F). Silencing PDE4DIP expression in PODs using lentivirus produced similar results (Fig. 5G and fig. S6F), and phalloidin staining further confirmed that PDE4DIP knockdown led to loss of cytoskeletal integrity and reduced actin bundle formation (Fig. 5H and fig. S6G).

Fig. 5. PDE4DIP regulates POD injury and cytoskeletal stability.

(A) Volcano plot showing genes correlated with PDE4DIP in human glomerular bulk RNA-seq (P < 0.05). (B) Venn diagram of PDE4DIP-correlated genes and coexpression module 1. (C) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of overlapping genes. (D) Schematic of PDE4DIP-AKAP9-PKC signaling regulating cytoskeleton via RAS/ERK/AKT pathways. (E) WB analysis of PDE4DIP, AKAP9, PKCε, P-PKCε, AKT, P-AKT, ERK, and P-ERK in PODs treated with ADR, ANG II, hypoxia, or LPS. Bottom: quantification of AKT and ERK phosphorylation. Data are mean ± SD (n = 3), two-tailed Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. (F) RAS activity under indicated conditions. Data are mean ± SD (n = 5), one-way ANOVA followed by Tukey’s multiple comparisons test. **P < 0.01, ***P < 0.001, and ****P < 0.0001. (G) WB analysis of the same proteins in PDE4DIP knockdown (KD) and control PODs. Data are mean ± SD (n = 3), unpaired two-tailed Student’s t test. *P < 0.05 and **P < 0.01. (H) Phalloidin staining of control and KD PODs [F-actin, green; nuclei, 4′,6-diamidino-2-phenylindole (DAPI)]. (I) Heatmap of top transcription factors by single-cell regulatory network inference and clustering (SCENIC) in human and mouse PODs. (J) Expression of transcription factors in PODs. (K) SCENIC-predicted regulatory network of PDE4DIP. (L) ATAC-seq signals around the PDE4DIP TSS (±2.5 kb) in control and hypoxia-treated PODs. (M) KEGG enrichment of genes from down-regulated peaks. cAMP, cyclic adenosine 3′,5′-monophosphate; mitogen-activated protein kinase. (N) Motif analysis showing WT1 binding sites in the PDE4DIP promoter. (O) WT1 enrichment profiles around peak centers. (P) WB analysis of PDE4DIP signaling axis in PODs with ADR treatment or WT1 overexpression. (Q) Quantification of (P). Data are mean ± SD (n = 3), two-way ANOVA with Tukey’s test, *P < 0.05, **P < 0.01, and ***P < 0.001 versus the ADR treatment group; #P < 0.05 and ##P < 0.01 versus the WT1 overexpression group. (R) Phalloidin staining after ADR treatment or WT1 overexpression.

Transcription factor activity analysis revealed that in both human and mouse hPODs, WT1 and ZBTB7C exhibited the highest activity, whereas in ePODs, PBX1 and TAL1 were the most active, and in tPODs, PPAPRA showed the highest activity (Fig. 5I and tables S11 and S12). The activity of these transcription factors was consistent with their expression patterns in POD subtypes (Fig. 5J). Furthermore, we found that PDE4DIP is a target gene of WT1 in both humans and mice (Fig. 5K). We subsequently performed assay for transposase-accessible chromatin using sequencing (ATAC-seq) on hypoxia-treated and control PODs. After quality control, peak calling, and annotation (fig. S6, H to K), we observed a significant reduction in peak signal within the PDE4DIP promoter region under hypoxic conditions (Fig. 5L and table S13). The down-regulated peaks were significantly enriched in pathways such as PI3K-AKT and cyclic adenosine 3′,5′-monophosphate (Fig. 5M and fig. S6L). Motif analysis revealed that WT1 significantly binds to the peaks in the PDE4DIP promoter region (Fig. 5N and table S14), and WT1 enrichment around peak centers was also markedly decreased following hypoxia treatment (Fig. 5O). Thus, we transfected human PODs with WT1 (fig. S6M). The results showed that WT1 overexpression effectively restored the ADR-induced reduction in PDE4DIP and AKAP9 expression, as well as the decreased phosphorylation of PKCε, ERK, and AKT (Fig. 5, P and Q). Furthermore, WT1 overexpression significantly alleviated the cytoskeletal disruption induced by ADR treatment in PODs (Fig. 5R and fig. S6N), and PDE4DIP overexpression produced a similar protective effect (fig. S7, A to D).

We generated POD-conditional Pde4dip knockout mice (Fig. 6A), specifically, POD-conditional Pde4dip knockout mice were maintained under normal housing conditions without any pharmacological intervention, and their phenotypes were monitored during natural aging. Multiplex immunofluorescence on renal tissue sections confirmed efficient gene knockout (Fig. 6, B and C). Kidneys of Pde4dip^−/− mice were markedly enlarged at 28 weeks (Fig. 6D), with survival reduced to 40% (Fig. 6E). UACR increased significantly by 20 weeks, accompanied by elevated Scr and blood urea nitrogen (BUN), and these differences became more pronounced by 28 weeks (Fig. 6, F to H). In isolated glomeruli, expression of Pde4dip, Wt1, Nphs1, and Nphs2 was significantly reduced, particularly at 28 weeks (Fig. 6, I to L). POD quantification revealed that POD counts in Pde4dip^−/− mice were comparable to controls at 8 weeks; however, they significantly declined starting at 20 weeks (fig. S7, E and F).

Fig. 6. POD-conditional deletion of Pde4dip establishes a mouse model of PCP.

(A) Schematic of POD-specific Pde4dip conditional knockout (CKO) strategy and workflow. Created in BioRender. Qing, J. (2026) https://BioRender.com/7cfjwlq. (B) Multiplex immunofluorescence staining of glomeruli showing PDE4DIP (red), SYNPO (green), and nuclei (DAPI). (C) Quantification of PDE4DIP fluorescence in PODs. Data are mean ± SD (n = 5), two-tailed Student’s t test. ***P < 0.001. (D) Gross kidney morphology at 8, 20, and 28 weeks (W). (E) Survival curves of control and CKO mice. (F to H) UACR, Scr, and BUN levels of control and CKO mice at 8, 20, and 28 weeks. (I to L) Relative mRNA expression of Pde4dip, Wt1, Nphs1, and Nphs2 in isolated glomeruli. (M) Representative kidney histology [H&E and periodic acid–Schiff (PAS)] and transmission electron microscopy (TEM) images from control and CKO mice at 8, 20, and 28 weeks. (N) Glomerulosclerosis index (GSI) in control and CKO mice at 8, 20, and 28 weeks. (O) Correlation between glomerular Pde4dip mRNA levels and GSI in control and CKO mice. R² and P values reflect linear regression fitting. (P) POD foot process effacement scores in control and CKO mice at 8, 20, and 28 weeks. (Q) Correlation between glomerular Pde4dip mRNA levels and foot process effacement scores. Data in (F) to (L), (N), and (P) are all mean ± SD (n = 5). Two-way ANOVA with Tukey’s test [(F) to (L), (N), and (P)]; linear regression [(O) and (Q)]. **P < 0.01, ***P < 0.001, and ****P < 0.0001 versus 8 weeks; #P < 0.05, ##P < 0.01, ###P < 0.001, and ####P < 0.0001 versus CKO.

Histology showed no overt abnormalities at 8 weeks, FSGS at 20 weeks, and progression by 28 weeks. Electron microscopy revealed intact POD foot processes at 8 weeks, partial effacement at 20 weeks, and extensive effacement by 28 weeks; notably, glomerular Pde4dip mRNA levels correlated positively with both glomerulosclerosis and foot process effacement (Fig. 6, M to Q). Collectively, these results demonstrate that Pde4dip deficiency in PODs leads to progressive glomerular injury and functional decline in mice.

DISCUSSION

Here, we present the snRNA-seq and high-resolution ST atlas of PCPs, encompassing five major disease types across both human and mouse kidney tissues. This integrated resource allowed us to define conserved molecular signatures of POD injury and define POD subtypes, addressing a major gap in previous studies. Notably, we identified PDE4DIP as a core gene shared across PCPs, with strong association with disease onset and progression, highlighting its translational potential as a diagnostic and therapeutic target.

Previous sc and ST studies in kidney disease have primarily focused on AKI and CKD (12, 13), with limited coverage of PCPs and relatively low capture of PODs. In this study, we integrated snRNA-seq data from multiple human and mouse models of PCPs and identified three conserved POD subtypes: hPODs and two injured subtypes, ePODs and tPODs. Pseudotime analysis positioned hPODs as the origin and tPODs as a terminal, dysregulated state within the injury trajectory. In contrast to earlier classifications that often-dichotomized PODs simply as “healthy” or “injured” based on marker expression, our work reveals distinct molecular pathways underlying POD injury.

Although PODs are considered terminally differentiated, severe or prolonged injury can induce dedifferentiation, loss of POD-specific markers, and cytoskeletal disruption (24, 30), ultimately leading to detachment and cell loss (31, 32). Previous studies, including work by Lake et al. (13) classified PODs into hPOD and dPOD at the sc level. The dPODs identified in their study share similar molecular markers, such as increased B2M expression. In addition, we further subcategorized dPODs into ePODs, which express endothelial markers, and tPODs, which express tubular markers. It is important to note that these expression differences are relative to hPODs. Although stringent doublet filtering was performed during quality control, the presence of binucleated cells cannot be completely ruled out. Nevertheless, we validated similar dPOD gene expression patterns in independent data from the KPMP consortium. Notably, dPOD can also be detected in normal human and mouse kidneys, as a small fraction of PODs may undergo transient stress or injury even under physiological conditions; however, their proportions are lower than those observed in disease states.

Previous studies have shown that endothelial markers, such as CD31 and alpha–smooth muscle actin (α-SMA), emerge during POD EMT (33, 34). As PODs are specialized epithelial cells with inherent endocytic potential, the endocytic system, consisting of LRP2 and CUBN, has also been detected (35, 36). Hypoxia emerged as a key driver of POD dedifferentiation in our analysis, with HIF1A and HIF2A being prominently activated in injured tPODs across species. Consistent with previous reports, chronic hypoxia disrupts cytoskeletal organization and promotes POD injury (37–39). Our in vitro experiments further confirmed that hypoxia induces dedifferentiation-associated gene expression, which can be partially reversed by HIF inhibition. This aligns with established literature linking renal hypoxia to both glomerular and tubular fibrosis (40–42). The observed reduction in chromatin accessibility at the PDE4DIP promoter under hypoxia, enriched for WT1 binding motifs, provides a plausible mechanistic link between the hypoxic microenvironment, loss of master transcription factors, and POD dysfunction.

POD cytoskeletal integrity is essential for maintaining the glomerular filtration barrier, and its disruption is a key driver of POD injury and loss (43). Therefore, stabilizing the cytoskeleton in injured PODs is essential for preserving POD integrity and restoring glomerular function (44, 45). For instance, in MCD, glucocorticoid therapy can promote foot process reextension and functional recovery (46). Central to our study is the discovery of PDE4DIP as a critical guardian of POD cytoskeletal integrity. Its expression was predominantly POD-specific, markedly down-regulated across human and murine PCPs, and strongly correlated with renal function and prognosis.

Mechanistically, we demonstrate that PDE4DIP functions by scaffolding with AKAP9 to facilitate PKC phosphorylation, thereby activating prosurvival RAS-ERK and AKT signaling pathways. This mechanism was disrupted by various injurious stimuli and was essential for maintaining actin cytoskeleton architecture, as PDE4DIP knockdown led to profound cytoskeletal disassembly. Furthermore, we established PDE4DIP as a transcriptional target of WT1, whose activity is diminished in dPODs. The severe phenotype of POD-specific Pde4dip knockout mice, characterized by early proteinuria, progressive FSGS, foot process effacement, and reduced survival provides compelling in vivo evidence of its nonredundant role in POD homeostasis.

Our findings have important clinical implications. The strong correlation between PDE4DIP expression and renal function suggests its potential utility as a prognostic biomarker. Therapeutically, restoring PDE4DIP expression or function represents a promising strategy for POD-protection. This could be achieved by targeting its upstream regulators (e.g., mitigating hypoxia and enhancing WT1 activity) or directly modulating its downstream pathways (RAS-ERK/AKT).

Because of their rarity and sensitivity, PODs are difficult to capture using either scRNA-seq or snRNA-seq technology, which limits the comprehensive analysis of PCPs (47). In contrast, ST, particularly high-resolution platforms, enables more accurate identification and localization of PODs within intact kidney tissue. Our ST data unveiled profoundly altered POD-centered microenvironments in PCPs. We observed a significant expansion of PT-inj cells, FIBs, MACs, and T cells around PODs, particularly in DN. The number of these adjacent cells negatively correlated with eGFR slope, indicating that POD-centered microenvironment alteration accelerates disease progression. Cell-cell communication analysis predicted active cross-talk between PODs and these infiltrating cells via ligand-receptor pairs such as APP-CD74 and SELL-PODXL, suggesting potential pathways through which the microenvironment may directly influence POD health.

Following injury, PODs could express major histocompatibility complex class II and costimulatory molecules or secrete inflammatory cytokines, which contribute to the recruitment and activation of immune cells (48, 49). The accumulation of immune cells can then exacerbate POD damage, forming a self-amplifying cycle. In addition, injured PODs release profibrotic factors such as TGF-β, which activate FIBs and promote collagen deposition (50, 51). As proteinuria increases due to POD injury, the reabsorption burden on PT also rises. When this burden exceeds their capacity, it leads to PT injury. Together, injured PODs and PT-Inj cells become key sources of proinflammatory and profibrotic signaling (52, 53). This cascade ultimately contributes to irreversible kidney damage. Therefore, timely protection of PODs and disruption of their interactions with immune cells and FIBs may offer an effective therapeutic strategy to prevent disease progression.

Despite these insights, our study has several limitations. The sample sizes of some rare PCP subtypes remain relatively small. Moreover, the multiplex immunofluorescence and transcriptomic data cannot definitively establish the existence of ePOD and tPOD populations; instead, they suggest the presence of potential POD states that require further validation in larger datasets and through additional experimental approaches. In addition, PDE4DIP expression was not captured in the ST datasets, and the precise functional contributions of the predicted cell-cell interactions will require further experimental validation, such as coculture systems or conditional knockout models.

In conclusion, our integrated multiomic atlas of PCPs offers a comprehensive resource that refines the molecular taxonomy of POD injury and its microenvironment, serving as a valuable tool for the broader research community to investigate disease mechanisms and discover potential therapeutic targets in PCPs. We delineate a pathogenic axis wherein microenvironmental hypoxiadrives POD dedifferentiation and down-regulates PDE4DIP, a key cytoskeletal regulator, via altered WT1 activity. The subsequent loss of PDE4DIP-mediated RAS-ERK/AKT signaling culminates in cytoskeletal collapse and proteinuria. This work not only advances our mechanistic understanding of PCPs but also nominates PDE4DIP and its associated pathways as compelling targets for future therapeutic development.

MATERIALS AND METHODS

Ethical approval

This study was conducted in accordance with the principles of the Declaration of Helsinki. The collection of clinical samples and all animal experiments were approved by the Institutional Review Board (IRB) of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (IRB no. 2025-0487) and performed in compliance with institutional guidelines. Written informed consent was obtained from all participants after full disclosure.

Human samples

Kidney tissues were obtained from 24 patients diagnosed with PCPs at Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, based on clinical presentation, laboratory tests, kidney biopsy, and electron microscopy. All patients had an eGFR above 90 ml/min per 1.73 m² at diagnosis and had not received glucocorticoids or immunosuppressive therapy. Control samples were collected from the tumor-adjacent normal kidney tissues of four patients undergoing unilateral nephrectomy for renal clear cell carcinoma at the same hospital. These controls also had preoperative eGFR above 90 ml/min per 1.73 m², and their adjacent tissues were confirmed to be histologically and ultrastructurally normal. Clinical and pathological data were collected for all participants, excluding any personally identifiable information. This study were approved by the Institutional Review Board of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (IRB no. 2025-0487).

Animal models

All animal experiments were performed using male mice purchased from Cygen Biotechnology Co. Ltd. To establish the ADRN model, 6-week-old BALB/c mice received a single tail vein injection of ADR (10 mg/kg; Sigma-Aldrich), while littermate controls received saline. Mice were euthanized 8 weeks after injection. For the OBN model, 6-week-old C57 mice were fed a high-fat diet (Research Diets; D12492) for 16 weeks, with control mice maintained on a standard diet (54). POD-specific Pde4dip conditional knockout mice (C57BL/6JGpt) were generated by crossing Pde4dip^flox/flox mice with Nphs2-Cre⁺ transgenic mice, followed by intercrossing of offspring. All mice were euthanized under approved protocols, and serum, urine, and kidney tissues were collected for further analysis. In addition, glomeruli were isolated from mouse kidneys for qPCR (Supplementary Methods) (55).

Cell culture and in vitro injury model

The conditionally immortalized human POD line (CIHP-1, F-biology; FBCL2160) was cultured at 33°C in RPMI 1640 (Thermo Fisher Scientific, 11875085) medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and insulin-transferrin-selenium (ITS; MCE, HY-150287) until 70 to 80% confluence was reached. To induce differentiation, cells were shifted to 37°C for 10 days in medium containing 2% FBS. To establish in vitro POD injury models, differentiated cells were treated with ADR (0.8 μg/ml; MCE, HY-15142), ANG II (10⁻⁷ M; MCE, HY-13948), and LPS (10 μg/ml; MCE, HY-D1056) (56) or cultured under hypoxic conditions (1% O₂) for 24 hours in low-serum RPMI 1640 medium (2% FBS, 1% penicillin/streptomycin, 1× ITS). Lentiviral vectors for PDE4DIP knockdown, PDE4DIP overexpression, and WT1 overexpression (Obio Technology, Shanghai, China) were transduced into PODs according to the manufacturer’s instructions.

For cytoskeletal staining, cells grown on glass coverslips were fixed in 4% paraformaldehyde for 15 min and permeabilized using phosphate-buffered saline (PBS) containing 0.2% Triton X-100 for 10 min. After blocking with 1% bovine serum albumin (BSA) in PBS, cells were incubated with phalloidin (1:200; Proteintech, PF00001) for 30 min to visualize F-actin. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and imaged using a fluorescence microscope.

Single-nucleus RNA sequencing

Kidney tissues from both human and mouse samples were processed using the Shbio Nuclear Extraction Kit (Shbio, 52009-10) to prepare nuclear suspensions. These suspensions were then loaded into the individual chambers of a Chromium Chip K, along with barcoded gel beads and oil, using the 10X Genomics Chromium System to generate gel beads-in-emulsion (GEMs). Following generation of GEMs, reverse transcription was carried out in a PCR instrument. After reverse transcription, cDNA concentration and fragment size were determined using Qubit and the Agilent 2100 Bioanalyzer. cDNA amplification was followed by enzymatic fragmentation and size-based selection using magnetic beads to retain the optimal fragments. Library preparation included end-repair, A-tailing, adapter ligation with sequencing primer Read2 and PCR enrichment using P5 and P7 adapters. The libraries were purified using magnetic beads, and their concentration was assessed with Qubit, while fragment size was evaluated using the Agilent 4200 TapeStation System. Cluster generation and primer hybridization were performed according to the Illumina’s protocol. Clusters were loaded onto the sequencing flow cell, and paired-end sequencing was performed using the Illumina’s sequencing platform, with data collection monitored by Illumina’s software.

Visium HD

Four-micrometer FFPE kidney tissue sections were placed onto slides in the capture area for immunohistochemistry. Total RNA was extracted from the FFPE tissue blocks using the RNeasy FFPE Kit (QIAGEN, 73504). The quality of the extracted RNA was assessed by calculating the DV200. Tissue sections that passed quality control (DV200 > 30%) were subjected to ST analysis. FFPE sections were prepared according to the Visium HD manual (CG000684), including deparaffinization, H&E staining, and imaging. Subsequently, the sections were stained with H&E and imaged using a PANNORAMIC MIDIII digital scanner (3DHISTECH) at ×20 magnification. Probe hybridization, probe ligation, slide preparation, probe release, extension, library preparation, and sequencing followed the Visium HD Spatial Gene Expression Kit User Guide (CG000685). Sequencing was performed on an Illumina NovaSeq 6000 with paired-end reads. The tissue sections were processed using Space Ranger v3.0 to align the sequencing data with the microscope and CytAssist images, producing a gene barcode matrix for further analysis.

ATAC-seq

Human PODs cultured under normoxic and hypoxic conditions were lysed using lysis buffer (Novogene Kit). The nuclei were collected by centrifugation and resuspended in Tn5 transposase reaction buffer. Transposition was carried out at 37°C for 30 min, after which equal molar amounts of Adapter1 and Adapter2 were added, and the libraries were amplified by PCR using the ATAC-seq protocol (Novogene Kit). The libraries were then purified using AMPure beads. Library quality was assessed, and sequencing was performed on the Illumina NovaSeq platform, generating paired-end 150-bp reads.

Enzyme-linked immunosorbent assay

UACR and Scr levels in mice were measured using high-sensitivity albumin and creatinine enzyme-linked immunosorbent assay (ELISA) kits (EthosBiosciences, 1011; Nanjing Jiancheng, C011-2-1). BUN was determined using a urease-based assay kit (Elabscience, E-BC-K183-M). Ras activity was assessed using the Ras guanosine triphosphatase Chemi ELISA Kit (ActiveMotif, 52097).

Histology and immunohistochemistry

FFPE kidney tissues from humans and mice were sectioned at 4-μm thickness for histological evaluation, immunohistochemical, or multiplex immunofluorescence staining. For transmission electron microscopy, mouse kidney tissues were used to evaluate POD ultrastructure. Briefly, tissues were fixed in 2.5% glutaraldehyde, postfixed with 1% osmium tetroxide, dehydrated in graded ethanol, embedded in epoxy resin, and sectioned into ultrathin slices. Sections were stained with uranyl acetate and lead citrate and imaged using a transmission electron microscope. Detailed protocols for staining procedures, antibodies used, imaging conditions, and scoring or quantification criteria are provided in Supplementary Methods.

WB and qPCR

For WB, cells or tissues were lysed in radioimmunoprecipitation assay buffer supplemented with protease inhibitor phenylmethylsulfonyl fluoride and phosphatase inhibitors. Protein concentrations were determined using the BCA assay. Equal amounts of protein were separated by SDS–polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. Membranes were blocked with 5% nonfat dry milk or BSA in TBST for 1 hour at room temperature. Primary antibodies were incubated overnight at 4°C, followed by incubation with horseradish peroxidase–conjugated secondary antibodies for 1 hour at room temperature. The signal intensity was quantified using ImageJ software. For qPCR, RNA was extracted from tissue or cells using an RNA extraction kit, followed by cDNA synthesis using the HiScript III 1st Strand cDNA Synthesis Kit (Vazyme, R312) with 1 μg of RNA from each sample. SYBR Green–based qPCR was performed using SYBR Green (Vazyme, Q711-03) on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Relative gene expression levels were calculated using the ΔΔCt method. Detailed experimental procedures are provided in Supplementary Methods. Antibody information and primer sequences are provided in tables S15 and S16.

Bioinformatic analysis and visualization

We developed an R package, SCNT, for the analysis and visualization of sequencing data (57). More detailed bioinformatic procedures are described in Supplementary Methods.

Statistical analyses

All quantitative data are presented as mean ± SD unless otherwise indicated. Comparisons between two groups were performed using an unpaired two-tailed Student’s t test. For comparisons among multiple groups, one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test was applied. For experiments involving two independent variables, two-way ANOVA followed by Tukey’s multiple comparisons test was used. A P value < 0.05 was considered statistically significant. Survival analysis was performed using Kaplan-Meier survival curves with log-rank tests to evaluate the effect of Pde4dip knockout on mouse survival. All statistical analyses and data visualization were performed using GraphPad Prism (version 9.5).

Supplementary Materials

The PDF file includes:

Supplementary Methods

Figs. S1 to S7

Legend for data S1

Other Supplementary Material for this manuscript includes the following:

Data S1