Datasets related to both AKI and CKD were sourced from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). GSE139061 (GPL20301) consisted of 39 renal tissue samples from AKI patients and 9 normal renal tissue samples, while GSE30718 (GPL570) included samples included 28 from AKI patients and 11 from healthy kidney tissues, functioning respectively as the training set and validation set of AKI. Similarly, GSE66494 (GPL6480) consisted of 53 renal tissue samples from CKD patients and 8 normal renal tissue samples, while GSE104948 (GPL22945) included 50 CKD patients’ renal tissue samples and 18 normal renal tissue samples, serving respectively as the training set and validation set of CKD. Furthermore, GSE183277 (GPL24676) comprised single-cell RNA sequencing (scRNA-seq) data from kidney cortex tissue samples of 5 AKI patients, 2 CKD patients and 11 normal individuals.

Differentially expressed genes1 (DEGs1) between AKI and normal samples in the GSE139061 dataset were pinpointed by employing the DEseq2 (v 1.38.0) package (14).The dataset was normalized using the estimateSizeFactors function, and genes with counts ≤ 1 were filtered out. DEGs1 were selected with the thresholds of |log₂fold-change (FC)| > 1.5 and P < 0.05, and the false discovery rate (FDR) was applied to control for multiple comparisons. For the GSE66494 dataset, differential expression analysis between CKD and control samples was performed using the limma package (v3.44.3) (15). Genes with missing values were removed using the na.omit() function. DEGs2 were identified with the same thresholds of log₂FC > 1.5 and P < 0.05, and FDR correction was applied. Subsequently, DEGs1 and DEGs2 were visualized as volcano plots and heatmaps, displaying only the top 10 in descending order of log2FC for both up- and down-regulated genes. The visualizations were generated through the ggplot2 (v 3.3.2) package (16) and the pheatmap (v 0.7.7) package (17), respectively.

To identify candidate genes in AKI and CKD progression, the up-regulated genes from DEGs1 and DEGs2, as well as the down-regulated genes from DEGs1 and DEGs2 were separately overlapped using (v 1.7.3) ggvenn package (18). Subsequently, the clusterProfiler package (version 3.16.0) was used to perform Gene Ontology (GO) and Kyoto Encyclopedia of the Genome (KEGG) enrichment analyses on the candidate genes (19). The Benjamini-Hochberg (BH) method was applied to control the FDR, with a significance threshold of pvalueCutoff = 0.05. The top 10 most significantly enriched terms (ranked in ascending order of p-value) from the GO and KEGG analyses were visualized using the enrichplot package (v 1.14.2) (20). To explore protein interactions among the candidate genes, the protein-protein interaction (PPI) network (interaction score > 0.15) was constructed using the Searching for Interacting Genes (STRING, https://www.string-db.org) database and the results were visualized using Cytoscape (v 3.10.2) software (21).

The glmnet (v 4.1.4) package was used to apply the least absolute shrinkage and selection operator (LASSO) method to the candidate genes in the GSE139061 and GSE66494 datasets (22). The parameter family was set as binomial, and 10-fold cross-validation (nfolds = 10) was performed to determine the optimal lambda (λ) value. Potential feature genes were screened based on the lambda.min value for each dataset. Moreover, feature genes were obtained by overlapping the potential feature genes obtained from the GSE139061 and GSE66494 datasets, respectively. Immediately, to evaluate the potential of the feature genes to distinguish AKI samples from control samples, and CKD samples from control samples, these feature genes were subjected to ROC curve analysis using pROC (v 1.18.0) package (23) in the AKI training set and the AKI validation set, the CKD training set and the CKD validation set, respectively, and feature genes with area under the curve (AUC)>0.7 in all four datasets were named as candidate biomarkers. Simultaneously, the candidate biomarkers were subjected to gene expression analysis in the AKI training set and AKI validation set, the CKD training set and the CKD validation set, respectively, and the candidate biomarkers showing a notable difference (P<0.05) between the disease samples and the control samples in the four datasets and a consistent expression trend were selected as the biomarkers for the subsequent analyses.

To investigate the biological roles of biomarkers involved in AKI and CKD, GSEA was performed in the GSE139061 and GSE66494 datasets, respectively. For the analysis, the c2.cp.kegg.v2023.1.Hs.symbols.gmt gene set was acquired from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/gsea/msigdb/) to act as the background set. First, Spearman correlations between the biomarkers and other genes were calculated using the psych (v 2.2.5) package (24) in the GSE139061 dataset. Subsequently, GSEA for each biomarker was constructed using clusterProfiler (v 3.16.0) package, with significance determined at P <0.05 and |normalized enrichment score (NES)| > 1. The top five pathways in descending order of P-value were visualized using the enrichplot package (v 1.14.2). Similarly, GSEA of the biomarkers was carried out using the same methods and thresholds in the GSE66494 dataset.

To assess the infiltration of 64 immune cells in disease samples and control samples in the GSE139061 and GSE66494 datasets, respectively. In the case of the GSE139061 dataset, relative abundance was calculated using the xCell (v 1.1.0) package (25), and the proportionate distribution of the 64 immune cells of the AKI samples versus the control samples was visualized using the ggplot2 (v 3.3.2) package. Differences in infiltration scores between AKI samples and control samples in the GSE139061 dataset were then assessed using Wilcoxon test to screen for immune cell types with a significant difference in infiltration (P<0.05), which were named differential immune cells. Subsequently, Spearman correlation analysis was performed using corrplot (v 0.92) package (26) to explore the relationship between differential immune cells and the association between diverse immune cells and biomarkers(|cor| > 0.30, P < 0.05), and correlation heatmaps were plotted to show the results. In addition, immune infiltration and correlation analyses were carried out in the GSE66494 dataset with the same methods and thresholds.

Biomarkers targeted by miRNAs were forecasted using the TargetScan (http://www.targetscan.org/) and miRDB (http://mirdb.org/) databases. The transcription factors (TFs) that regulate biomarkers were predicted through the ChEA3 (https://maayanlab.cloud/chea3/) database. Then, the lncRNAs targeting the aforementioned miRNAs were predicted by means of the LncBase (http://carolina.imis.athena-innovation.gr/diana_tools/web/index.php?r=lncba) database. The miRNA-mRNA, TF-mRNA and TF-mRNA-miRNA networks were visualized by using the Cytoscape (v 3.10.2) software, and using the ggplot2 package (v 3.3.2), the lncRNA-miRNA-mRNA network was visualized. Additionally, the Comparative Toxicogenomics Database (CTD, http://ctdbase.org/) was employed to predict drugs targeting biomarkers and Cytoscape version 3.10.2 was employed to plot the biomarker-drug network.

Firstly, 5 AKI samples and 11 control samples were selected from the GSE183277 dataset as the AKI single-cell dataset. The AKI single-cell dataset was the “Seurat” package (v 4.1.0) was utilized for quality control (QC) to filter out cells with exceeded 20% of mitochondrial genes, cells with nCount_RNA under 200 and surpassed 30,000 genes, and cells with nFeature_RNA > 200 (27). Then, in light of the GSE183277 dataset, data were normalized by the “NormalizeData” function in the “Seurat” package (v 4.1.0), and highly variable genes (HVGs) were selected by the “FindVariableFeatures” function. Next, the “ScaleData” function in the “Seurat” package (v 4.1.0) was applied to scale data before principal components analysis (PCA). Subsequently, the “JackStraw” function within the “Seurat” package (v 5.0.1) was applied to execute PCA on HVGs. The “ElbowPlot” function within the “Seurat” package (v 4.1.0) was thereafter applied to draw a scree plot of the top 30 principal components (PCs), aiming to identify PCs that notably contributed to variation for subsequent analysis (p < 0.05). Afterward, cell cluster analysis was conducted on cells after dimensionality reduction utilizing “FindNeighbors” and “FindClusters” functions (resolution = 0.2, dimension = 30). Finally, the Seurat package’s FindNeighbors and FindClusters functions were employed to categorize all high-quality cells into various cell clusters using the uniform manifold approximation and projection (UMAP) clustering technique.The FindAllMarkers function was used to identify key marker genes for various populations, and the classical marker genes of relevant cells in the CellMarker (http://xteam.xbio.top/CellMarker/) database were used as the reference gene set to annotate each cell cluster ( Supplementary Table 1 ). Additionally, 2 CKD samples and 11 control samples were selected from the GSE183277 dataset as the CKD single-cell dataset and analyzed by scRNA-seq in the same way, with marker genes shown in Supplementary Table 2 .

Cellular communication networks between cell types of AKI samples and control samples as well as those between cell types of CKD samples and control samples were analyzed respectively using the CellChat (v 1.6.1) package (28) based on the AKI single-cell dataset and the CKD single-cell dataset. And visualization was carried out by using the patchwork (v 1.3.0) package (29). In addition, key cells were screened and obtained based on the expression situation of biomarkers in cell types within the 2 single-cell datasets.

To explore the expression changes of biomarkers during the process of cell state transformation, key cells were first extracted respectively based on the AKI single-cell dataset and the CKD single-cell dataset for secondary dimensionality reduction and clustering, and the key cells were reclustered and divided into different cell subpopulations. Following this, the Monocle (v 2.30.0) package was used to conduct cell pseudo-time trajectory analysis on both the AKI and CKD single-cell datasets (30).

In this study, male C57BL/6J mice, approximately 8 weeks of age, were utilized. The strain was sourced from the University Model Animal Research Center at Guangxi Medical University. Ethical approval for the use of animals in this research was obtained in compliance with the Guidelines for the Management of Laboratory Animals as stipulated by the Ministry of Science and Technology of the People’s Republic of China, as well as the Guidelines for Ethical Review of Laboratory Animals according to the National Standard GB/T35892–2018 of the People’s Republic of China, and the protocols of the Animal Care and Welfare Committee at Guangxi Medical University (No:202506002). The mice were provided with food and water ad libitum, and the housing environment was maintained at a temperature of 25 ± 2°C with a 12-hour light/dark cycle. The experimental design included three groups of mice, with the model being established through renal artery ischemia-reperfusion surgery. For the intervention study, the C57BL/6J mice were divided into three groups (n = 5 or 6 per group): (1) normal control group; (2) AKI group; and (3) CKD group. Ischemic AKI was experimentally induced using a bilateral ischemia-reperfusion injury (BIRI) model. In this model, mice were anesthetized, and bilateral dorsal incisions were performed to access the kidneys. Both kidneys were then clamped to occlude blood flow for a duration of 30 minutes. CKD was simulated through a unilateral ischemia-reperfusion procedure combined with a contralateral total nephrectomy. Following anesthesia, a left dorsal incision was made to clamp the left kidney, obstructing blood flow for 30 minutes. Subsequently, 14 days post-procedure, a right dorsal incision was executed to facilitate the complete removal of the right kidney (31).

Kidney tissues were paraffin-embedded and sectioned into 4 μm slices. After deparaffinization and rehydration, antigen retrieval was conducted with EDTA buffer at pH 9.0 for 25 minutes. A 15-minute treatment with 10% hydrogen peroxide was used to block endogenous peroxidase activity, and secondary antibodies were blocked with 5% serum for 30 minutes at room temperature.The kidney tissues underwent overnight incubation at 4°C with primary antibodies (PLEKHA4,BD-PB3919, 1:300, Biodragon, Jiangsu, China; KLKI,YP-AB-02871, 1:200, UpingBio, Zhejiang, China; CLCNKB, DF9376, 1:150, Biodragon, Jiangsu, China) targeting the candidate biomarkers. Horseradish peroxidase (HRP)-conjugated antibodies were applied to the sections on the subsequent day. 3,3’-diaminobenzidine (DAB) (G1212-200T, Servicebio, Wuhan, China), a substrate specific to HRP, was used to highlight the stained areas in kidney tissue. Subsequently, counterstaining is performed using hematoxylin (G1004-100ML, Servicebio, Wuhan, China). Representative images were captured using an Olympus microscope, and ImageJ (NIH, USA) was employed to quantify the average optical density of the images to assess the expression levels of candidate biomarkers.

Immunofluorescence staining was conducted on 5 μm-thick paraffin-embedded sections of mice kidney tissue. Following deparaffinization and antigen retrieval using EDTA (pH 9.0), the sections were blocked with goat serum and incubated overnight at 4°C with primary antibodies targeting SLC4A1 (A17391, 1:150, ABclonal, Wuhan, China) and CA II (EM1801-08, 1:150, HuaAn, Zhejiang, China). Subsequently, the sections were treated with iFluor™ 647-conjugated goat anti-rabbit IgG and iFluor™ 488-conjugated goat anti-mouse IgG (HA1125 and HA1123, 1:300, HuaAn) for one hour at room temperature. Nuclei were counterstained with DAPI, and imaging was performed using a Zeiss Axio-Imager A2 confocal microscope (Carl Zeiss, Jena, Germany).

In summary, total RNA was extracted from renal tissues using the Trizol method (15596026, Invitrogen, USA). Equivalent amounts of mRNA were reverse transcribed into cDNA utilizing the HiScript RT SuperMix kit (R122-01; Vazyme, China). Quantitative real-time PCR (qRT-PCR) was conducted with the ChamQ Universal SYBR qPCR Master Mix (Q711-02; Vazyme, China) on a Viia 7 quantitative real-time PCR instrument (Thermo-Fisher Scientific, USA). The PCR amplification protocol consisted of 35 cycles at 95°C for 30 seconds, 58°C for 30 seconds, and 72°C for 30 seconds. The following primers were employed: Aqp6 forward: GCCGTCATTGTTGGGAAGTTC and reverse: GGCTCCAGGTCTACCACTTTC; Kit forward: GAACAGGACCTCGGCTAACAA and reverse: CCTTTGCTCTGCTCCTGTACA;Slc4a1 forward: CCTCGTCCAATACATCTCCCG and reverse: CGTCATGGCAAGTAGGAAGGT. RT-PCR products were separated on a 1.5% agarose gel and visualized under UV light. The quantification of qRT-PCR was performed using the 2−ΔΔCt method and expressed as relative fold changes.

To investigate the expression of candidate biomarkers in patients with AKI and CKD), we selected a cohort comprising five patients with AKI and five with CKD. Additionally, we included five patients diagnosed with renal malignancy, from whom normal renal tissue adjacent to the tumor was obtained during surgical procedures. Patients were identified as having AKI if they fulfilled any of these conditions: (a) a rise in serum creatinine (Scr) exceeding 26.5 μmol/L within 48 hours; (b) a 50% increase in Scr over the course of one week; or (c) urine output below 0.5 mL per kilogram per hour lasting over 6 hours. CKD patients were recognized by an estimated glomerular filtration rate (eGFR) under 60 mL/min/1.73 m².Approval for this study was granted by the ethics committee of The First Affiliated Hospital of Guangxi Medical University, with informed consent obtained from the patients (No. 2024-E0918).The data analysis design of this study was showed in Figure 1 .

Bioinformatic analysis were performed in the R (v 4.2.2). GraphPad Prism statistical software was used for experiment statistical analyses in the study. We employed the unpaired t test to compare continuous variables between two groups. Values are shown as mean ± SEM, with statistical significance set at P<0.05.

In the GSE139061 dataset, 1153 differentially expressed genes (DEGs1) were screened out, among which 913 were up-regulated and 240 were down-regulated. Similarly, in the GSE66494 dataset, 153 differentially expressed genes (DEGs2) were screened out, with 60 being up-regulated and 93 being down-regulated. The top 10 up- and down-regulated DEGs in both datasets and their expression profiles were labeled on the volcano plots and heatmaps respectively ( Figures 2A–D ). Subsequently, by overlapping the 913 up-regulated DEGs1 with the 60 up-regulated DEGs2, 6 common up-regulated genes were identified ( Figure 2E ). And by overlapping the 240 down-regulated DEGs1 with the 93 down-regulated DEGs2, 13 common down-regulated genes were obtained ( Figure 2F ). The 6 common up-regulated genes and the 13 common down-regulated genes were combined, and 19 candidate genes were determined. In conclusion, this analysis focused on the discovery of candidate genes that might play important roles in the progression of AKI and CKD.

Enrichment analyses of the 19 candidate genes showed that they were enriched in 22 GO entries, such as organic anion transport ( Figure 2G ; Supplementary Table 3 ), whereas KEGG analyses revealed that the candidate genes were significantly enriched in the Renin-angiotensin system ( Figure 2H ; Supplementary Table 4 ). In addition, in the constructed PPI network, genes such as ALB, SLC22A6 and SLC12A3 were highly associated with other genes ( Figure 2I ).

Based on the candidate genes, 7 potential feature genes in the AKI training set and 16 potential feature genes in the CKD training set were obtained respectively through the LASSO regression analysis ( Figures 3A, B ). Then, 5 feature genes were finally obtained by overlapping ( Figure 3C ). Subsequently, it was found in the AKI training set, validation set as well as the CKD training set and validation set that the AUC values of CLCNKB, KLK1 and PLEKHA4 were all greater than 0.7, and thus they could be regarded as the candidate biomarkers for this study ( Figures 3D, E ). Moreover, the expression analysis of the candidate biomarkers showed that the expression trends of CLCNKB, KLK1 and PLEKHA4 were consistent in the four datasets. Among them, CLCNKB and KLK1 were significantly down-regulated in AKI and CKD samples, while PLEKHA4 was significantly up-regulated ( Figures 3F, G ). Moreover, the expression trends of CLCNKB, KLK1 and PLEKHA4 in renal tissues of different groups of patients and different groups of mice models were consistent with our results ( Figure 4 ).

GSEA was performed on the GSE139061 and GSE6649 datasets to investigate the biological roles of the biomarkers. In the GSE139061 dataset of AKI, CLCNKB, KLK1 and PLEKHA4 were significantly enriched in 50, 15 and 51 pathways respectively ( Supplementary Tables 5 – 7 ). It was worth noting that CLCNKB and PLEKHA4 were co-enriched in the valine leucine and isoleucine degradation pathways and oxidative phosphorylation, and KLK1 and PLEKHA4 were co-enriched in the neuroactive ligand receptor interaction pathway ( Figures 5A–C ). In the GSE6649 dataset of CKD, CLCNKB, KLK1 and PLEKHA4 were significantly enriched in 39, 60 and 44 pathways respectively ( Supplementary Tables 8 – 10 ). Among them, CLCNKB and KLK1 were jointly enriched in the oxidative phosphorylation and valine, leucine, and isoleucine degradation pathways, as well as in the cytokine-cytokine receptor interaction pathway ( Figures 5D–F ).

Figures 6A, B illustrated the infiltration levels of 64 immune cells in AKI versus control samples, and CKD versus control samples, respectively. Among them, the infiltration levels of 6 types of immune cells (differential immune cells 1) were significantly different in AKI and control samples ( Figure 6C ), and the infiltration levels of 26 types of immune cells (differential immune cells 2) were significantly different in CKD and control samples ( Figure 6D ), and the common differential immune cells included Astrocytes, Th2 cells. Furthermore, among the differential immune cells 1 in AKI, Fibroblasts had the most significant positive correlation with aDC (cor = 0.32), and Astrocytes had the most significant negative correlation with Fibroblasts (cor = -0.36) ( Figure 6E ; Supplementary Table 11 ). Whereas PLEKHA4 had the strongest positive relationship with pDC (cor = 0.58) and the strongest negative relationship with Astrocytes (cor = -0.38), CLCNKB had the strongest positive relationship with Astrocytes (cor = 0.30) and the strongest negative relationship with Th2 cells (cor = -0.49), but KLK1 was significantly correlated with Differential Immune Cells 1 ( Figure 6F ; Supplementary Table 12 ). Subsequently, among the differential immune cells 2 in CKD, cDC had the most significant positive correlation with DC (cor = 0.75) and the highest positive association with Macrophages M2 and Neurons (cor = -0.74) ( Figure 6G ; Supplementary Table 13 ). In contrast, PLEKHA4 had the greatest positive connection with Th2 cells (cor = 0.58) and the greatest negative linkage with MEP (cor = -0.40), CLCNKB had the most significant positive correlation with Th1 cells (cor = 0.79) and the strongest inverse relationship with NKT (cor = -0.67), and KLK1 had the greatest positive connection with MEP (cor = 0.62) and the most prominent negative connection with Th2 cells (cor = -0.43) ( Figure 6H ; Supplementary Table 14 ).

Initially, merely 2 miRNAs were predicted for PLEKHA4, whereas no miRNAs could be predicted for CLCNKB and KLK1 ( Figure 7A ). Subsequently, the TF-mRNA networks consisting of 133, 34 and 56 TFs corresponding to PLEKHA4, CLCNKB and KLK1 respectively were acquired from the ChEA3 database ( Figure 7B ). Then, a miRNA-mRNA-TF network was established by integrating the 2 miRNAs ( Figure 7C ). Eventually, 8 lncRNAs upstream of miRNAs were predicted and a lncRNA-miRNA-mRNA network was constructe, such as EBLN3P-hsa-miR-3187-3p-PLEKHA4 ( Figure 7D ). In a nutshell, this analysis centered around the finding that PLEKHA4 was likely to play a more crucial role in the progression of AKI and CKD.

Drugs were screened for activation of CLCNKB and KLK1, which are down-regulated in expression, and inhibition of PLEKHA4, which is up-regulated in expression, including 27 drugs targeting KLK1, 19 drugs targeting CLCNKB and 19 drugs targeting PLEKHA4 ( Supplementary Tables 15 – 17 ). A biomarker-drug network was constructed accordingly ( Figure 7E ). It was noteworthy that CLCNKB, PLEKHA4 and KLK1 were simultaneously targeted by Tetrachlorodibenzodioxin.

In the AKI single-cell dataset, a total of 78,791 cells were retained after quality control ( Supplementary Figure S1 ). Subsequently, the top 2,000 highly variable genes and the top 30 PCs were applied to UMAP clustering ( Figures 8A, B ). All high-quality cells were divided into 17 different cell clusters ( Figure 8C ). In addition, marker genes had high specificity in different cell clusters ( Figures 8D, E ). The cell clusters were annotated and 14 cell types were determined, such as Injured Proximal tubular cell and Loop of Henle cell ( Figure 8F ). Subsequently, in the CKD single-cell dataset, a total of 58,561 cells were retained after quality control ( Supplementary Figure S2 ). Next, the top 2,000 highly variable genes and the top 30 PCs were applied to UMAP clustering ( Figures 9A, B ). All high-quality cells were divided into 16 different cell clusters ( Figure 9C ). Moreover, marker genes also had high specificity in different cell clusters ( Figures 9D, E ). The cell clusters were annotated and 13 cell types were determined, such as Nephron epithelial cell and Loop of Henle cell ( Figure 9F ).

In the AKI single-cell dataset, the injured proximal tubular cells in AKI samples had a relatively large number of interactions and a relatively high intensity with other cells ( Figure 10A ), while in the control samples, B cells had a relatively large number of interactions and a relatively high intensity with other cells ( Figure 10B ). Interestingly, in the CKD single-cell dataset, Nephron epithelial cells and B cells had a relatively large number of interactions and a relatively high intensity with other cells both in CKD and control samples ( Figures 10C, D ). In addition, KLK1 and CLCNKB had relatively high expression levels in Type A intercalated cells and Collecting duct-principal cells in both single-cell datasets ( Figures 10E, F ). To evaluate the abundance of type A intercalated cells in kidney disease, immunofluorescence staining was performed on kidney sections of AKI/CKD mice. Consistent with the reduced expression of SLC4A1 (a specific marker for this cell type) in diseased kidneys ( Figure 11A ), the expressions of additional markers (Aqp6, Kit and Slc4a1) were also significantly downregulated ( Figures 11B–D ), confirming the loss of type A intercalated cells in AKI/CKD. Consequently, Type A intercalated cells could have been part of the disease’s development.

Secondary dimensionality reduction clustering analysis was performed on Type A intercalated cell and Collecting duct-principal cell. It was found that Type A intercalated cell and Collecting duct-principal cell were divided into 10 and 8 subgroups respectively in the AKI single-cell dataset ( Figures 12A, B ). Whereas in the CKD single-cell dataset, Type A intercalated cell was divided into 10 subgroups and Collecting duct-principal cell was divided into 9 subgroups ( Figures 12C, D ). Subsequently, the different subgroups within Type A intercalated cell and Collecting duct-principal cell were arranged on the developmental trajectory according to the differentiation time. A darker blue indicates earlier cell differentiation. In addition, after different cell subgroups were mapped to the pseudo-time trajectory plot, it was found that they exhibited different differentiation states. In the AKI single-cell dataset, Type A intercalated cell had 10 differentiation states, with State 4 being the earliest and most specific in differentiation. Collecting duct-principal cell had 8 differentiation states, and State 4 was also the earliest and most specific ( Figures 12E, F ). In the CKD single-cell dataset, Type A intercalated cell also had 10 differentiation states, with State 9 being the earliest in differentiation. Collecting duct-principal cell had 8 differentiation states, and State 0 was the earliest and most specific ( Figures 12G, H ). In the AKI single-cell dataset, with the differentiation of Type A intercalated cells, the expressions of KLK1 and PLEKHA4 had no significant changes. The expression of CLCNKB showed a trend of first decreasing, then increasing and finally decreasing again ( Figures 12I ). With the development of Collecting duct-principal cells, PLEKHA4 had no significant change. The expression of CLCNKB showed a trend of first increasing and then decreasing, and the expression of KLK1 showed a trend of first remaining unchanged, then increasing, then decreasing and finally remaining unchanged ( Figures 12J ) In the CKD single-cell dataset, with the development of Type A intercalated cells, PLEKHA4 had no significant change. The expression of CLCNKB showed a trend of first decreasing and then increasing, and the expression of KLK1 showed a trend of first decreasing, then remaining unchanged and finally increasing ( Figures 12K ). With the development of Collecting duct-principal cells, PLEKHA4 had no significant change. CLCNKB expression consistently declined, while KLK1 expression initially decreased and then stabilized ( Figures 12L ).