Transcriptome and single-cell profiling of the mechanism of diabetic kidney disease

Abstract

BACKGROUND

The NOD-like receptor thermal protein domain associated protein 3 (NLRP3) inflammasome may play an important role in diabetic kidney disease (DKD). However, the exact link remains unclear.

AIM

To investigate the role of the NLRP3 inflammasome in DKD.

METHODS

Using datasets from the Gene Expression Omnibus database, 30 NLRP3 inflammasome-related genes were identified. Differentially expressed genes were selected using differential expression analysis, whereas intersecting genes were selected based on overlapping differentially expressed genes and NLRP3 inflammasome-related genes. Subsequently, three machine learning algorithms were used to screen genes, and biomarkers were identified by overlapping the genes from the three algorithms. Potential biomarkers were validated by western blotting in a db/db mouse model of diabetes.

RESULTS

Two biomarkers, sirtuin 2 (SIRT2) and caspase 1 (CASP1), involved in the Leishmania infection pathway were identified. Both biomarkers were expressed in endothelial cells. Pseudo-temporal analysis based on endothelial cells showed that DKD mostly occurs during the mid-differentiation stage. Western blotting results showed that CASP1 expression was higher in the DKD group than in the control group (P < 0.05), and SIRT2 content decreased (P < 0.05).

CONCLUSION

SIRT2 and CASP1 provide a potential theoretical basis for DKD treatment.

Key Words: Diabetic kidney disease; Single-cell RNA sequencing analysis; NOD-like receptor thermal protein domain associated protein 3; Sirtuin 2; Caspase 1

Core Tip: The NOD-like receptor thermal protein domain associated protein 3 inflammasome is implicated in diabetic kidney disease. Our study makes a significant contribution to the literature by identifying sirtuin 2 and caspase 1 as potential biomarkers for diabetic kidney disease, providing new insights into the molecular mechanisms involving the NOD-like receptor thermal protein domain associated protein 3 inflammasome, and suggesting potential targets for diagnosis and treatment, which are crucial for clinical and policy implications.

INTRODUCTION

Diabetic kidney disease (DKD), a severe microvascular complication affecting approximately 40% of individuals with diabetes, is associated with various conditions, such as elevated levels of advanced glycation end products, oxidative stress, hyperglycemia, and dysregulated lipid metabolism[1-3]. The incidence of diabetes-induced end-stage renal disease has increased, with rates escalating from 22.1% in 2000 to 31.3% in 2015[4]. The current therapeutic strategy for DKD focuses on controlling blood pressure and sugar levels to slow disease progression. Moderate proteinuria is widely recognized as an important clinical indicator of DKD, but it is not sensitive to early detection of the disease[5]. However, due to the individual heterogeneity of DKD, not all patients achieve effective treatment outcomes[6]. Therefore, identifying highly sensitive and efficient biomarkers could aid in developing new strategies for the diagnosis and treatment of DKD.

NOD-like receptor thermal protein domain associated protein 3 (NLRP3) belongs to the NLR family, which is divided into four subfamilies: NLRA, NLRB, NLRC, and NLRP, based on the nature of their N-terminal structural domains. Activation of the NLR family triggers various downstream signals that promote inflammasome assembly and inflammatory responses[7]. Among these, NLRP3 and its associated inflammasomes have received the most attention in recent years. As an important pattern recognition receptor in the cytoplasm, the activation of NLRP3 leads to the formation of oligomeric complexes, including apoptosis-associated speck-like protein with a caspase-recruitment domain and caspase 1 (CASP1). This protein complex, known as the NLRP3 inflammasome, plays a crucial role in inflammatory and antiviral processes[8-10]. In a physiologically normal organism, the activation of the NLRP3 inflammasome is strictly regulated by host cells. When appropriately activated, it can bolster the host cellular defense against pathogenic microbial invasion[11]. However, irregular activation of the NLRP3 inflammasome can cause considerable pathological damage. Experimental evidence has highlighted the role of NLRP3 in triggering detrimental inflammation associated with various diseases, including neuroinflammation induced by human immunodeficiency virus infection[12], chronic hepatitis, liver damage caused by hepatitis C virus infection[13], and Alzheimer’s disease[14]. Recent studies indicate that hyperglycemia can activate the NLRP3 inflammasome, which mediates the onset and progression of DKD through multiple mechanisms[15-17]. Isoliquiritigenin shows effective potential in the treatment of DKD by significantly inhibiting the toll-like receptor 4/nuclear factor kappa-B/NLRP3 inflammasome pathway[18]. Additionally, Huajuxiaoji may also alleviate DKD damage and exert anti-DKD effects by inhibiting NLRP3-mediated inflammasome activation and pyroptosis in vitro and in vivo experiments[19]. Given the evidence obtained thus far, further research on the role of NLRP3 inflammasome-related genes in DKD is warranted. Currently, there is a lack of efficient diagnostic and treatment methods for this condition, and targeting NLRP3 could be a promising therapeutic approach.

This study used bioinformatic approaches to explore the potential molecular mechanisms of NLRP3 inflammasome-associated genes (NIRGs) affecting DKD. First, we identified 10 intersecting genes by examining the intersection of differentially expressed genes (DEGs) and NIRGs. Subsequently, we screened for genes using 3 machine-learning methods. We further identified 2 biomarkers by taking the intersection of genes screened by these methods and validated them through gene expression and receiver operating characteristic (ROC) analyses. Based on these two biomarkers, we performed a series of analyses, including gene set enrichment analysis (GSEA), chromosome distribution analysis, immune infiltration analysis, subcellular localization, regulatory network construction, single-cell analysis, and pseudo-temporal analysis. These findings offer theoretical support for the clinical research and treatment of DKD related to NIRGs.

MATERIALS AND METHODS

Data sources and searches

The flowchart of the analysis of NLRP3 inflammatory vesicles in diabetic nephropathy is shown (Supplementary Figure 1). DKD datasets (GSE30529, GSE142025, and GSE131882) were acquired from the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/). The training set GSE30529 included 10 DKD and 12 control tubulointerstitial tissue samples based on the GPL571 platform, while validation set GSE142025 included 27 DKD and 9 control kidney biopsy tissue samples based on the GPL20301 platform. The data in the dataset was not further personalized as they had already been standardized and quality controlled when uploaded to the Gene Expression Omnibus database. The single-cell transcriptome sequencing dataset GSE131882 included 3 DKD and 3 control kidney tissue samples based on the GPL24676 platform. Additionally, 30 NIRGs were mined from the literature[20].

Differential expression analysis

First, the limma R package (version 3.52.4) was used to calculate the differences in gene expression levels between the control groups and DKD groups and to screen for significant DEGs using |log₂ fold change (FC)| ≥ 0.5 and P value < 0.05[21]. The results were then visualized using ggplot2 (version 3.3.6)[22] and ComplexHeatmap (version 2.14.0)[23] R packages, respectively.

Identification and function analysis of intersection genes

Intersection genes were identified by overlapping the DEGs and NIRGs using UpSetR (version 1.4.0)[24]. To explore the biological functions of the intersection genes, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were executed using the clusterProfiler R package (version 4.4.4)[25] (P < 0.05), and the results were visualized with ggplot2. Additionally, to further understand the protein interactions of the intersection genes, a protein-protein interaction network was constructed using the STRING database (https://string-db.org/) (confidence > 0.40), and the results were presented using Cytoscape (version 3.9.1)[26].

Machine learning, ROC, and gene expression analyses

Three machine algorithms were used to identify biomarkers based on the intersection genes. The least absolute shrinkage and selection operator (LASSO) analysis was performed using glmnet (version 4.1-4)[27], and lambda.min was used as the standard for the LASSO analysis. Support vector machine recursive feature elimination was conducted using e1071 (version 1.7-11)[28], selecting the model with the smallest error. Boruta (version 8.0.0) was used to rank and screen genes[29]. The biomarkers were identified by overlapping the genes from three machine learning algorithms using UpSetR. To better analyze biomarkers, we performed ROC [area under the curve (AUC) > 0.6] analysis using pROC (version 1.18.0)[30] and gene expression (P < 0.05) analysis of biomarkers in GSE30529 and GSE142025 datasets.

GSEA

Spearman’s correlation coefficients were calculated and ranked for each biomarker and all other genes in the GSE30529 dataset to explore the biological functions and pathways associated with the biomarkers. The background gene set of the KEGG gene sets was downloaded from msigdbr (version 7.5.1)[31]. Later, GSEA was performed and displayed using enrichplot (version 1.16.2)[32] (adjusted P value < 0.05).

Immune infiltration analysis

First, single sample GSEA was used to calculate enrichment scores for 28 immune cells using gene set variation analysis (version 1.44.5)[33], with the background gene set mined from the literature[34]. Subsequently, the Wilcoxon test was used to identify differential immune cells with differences in the enrichment scores of 28 immune cells between the DKD and control samples in the GSE30529 dataset (P < 0.05). The correlation between biomarkers and differential immune cells was analyzed using Spearman’s rank correlation coefficient.

Subcellular localization and chromosomal distribution analyses

To further understand the biomarkers, their subcellular localization was determined using Cell-PLoc 3.0. Furthermore, OmicCircos (version 1.38.0)[35] was used to visualize the distribution of biomarkers on the chromosome.

Relationship between biomarkers and DKD

To explore the relationship between biomarkers and DKD, the inference score between biomarkers and DKD was searched and visualized using the Comparative Toxicogenomics Database (https://ctdbase.org/), with higher scores indicating stronger interaction relationships.

Constructing networks

The transcription factors (TFs) targeting biomarkers were predicted using miRNet2.0 (https://www.mirnet.ca/) based on the Council for Higher Education Accreditation database (https://www.chea.org/) (degree ≥ 1), and a TFs-microRNAs (miRNAs) network was constructed. Additionally, the starBase database (http://starbase.sysu.edu.cn/) was used to predict miRNAs targeting biomarkers, and a competitive endogenous RNA (ceRNA) network was constructed using long non-coding RNAs (lncRNAs) targeting miRNAs (ExpNum ≥ 20). Furthermore, based on the predicted miRNAs from the starBase database, single nucleotide polymorphism (SNP) variants in their seed regions were searched using miRNASNP, and a miRNA-SNP-gene interaction network was built.

Drug predicted and related analysis

First, the Drug-Gene Interaction database was used to obtain biomarker-related drugs, and a gene-drug network was constructed. Furthermore, based on the gene-drug network, the protein molecular crystal structures of the biomarkers were downloaded from the Protein Data Bank database, the molecular structures of the drugs were downloaded from PubChem, and molecular docking of the target components was performed using AutoDock.

Single-cell RNA sequencing analysis

The samples in the GSE131882 dataset were integrated using the Seurat R package (version 4.3.0)[21]. The cells were retained as min.cells = 3 and min.features = 200. Then, low-quality cells were also filtered according to quality control criteria (nFeature_RNA ≤ 4000, nCount_RNA ≤ 50000, percent.mt < 20%). After data standardization, the top 2000 highly variable genes were selected, and the top 10 genes were labeled. Next, based on these 2000 highly variable genes, principal component analysis was performed. Principal components were selected for subsequent analysis according to the inflection point plot and principal component analysis displacement test. Additionally, an unsupervised clustering analysis was performed on the filtered cells using the FindNeighbors and the FindClusters functions from Seurat (resolution = 0.8). The cells were clustered using uniform manifold approximation and projection. Subsequently, based on marker genes from literature mining[36], the cells were annotated and visualized. Furthermore, the expression of biomarkers in annotated cells and in DKD and control samples was analyzed.

Pseudo-temporal analysis

To explore cell differentiation states and direction, pseudo-temporal analysis was performed on cells expressing all biomarkers using the Monocle R package (version 2.26.0)[37]. Additionally, cell differentiation in DKD and control samples was analyzed.

Experimental animals

Male diabetic mice (BKS^- db/db, 40-55 g, 8-week-old) and male control mice (control, 25-30 g, 8-week-old) were purchased from GemPharmatech (Nanjing, Jiangsu, China) and housed at the Experimental Animal Center of Fujian Medical University (Fujian, China). Before the experiment, all animals were provided with a normal diet and allowed to acclimatize for 1 week under a 12:12 light-dark cycle at room temperature (23 ± 1 °C) and approximately 60% humidity. All experimental protocols involving animals were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and approved by the Fujian Animal Research Ethics Commission (license No. IACUC FJMU 2022-0587).

Western blot

Renal cortex protein extracts were prepared using the Total Protein Extraction Kit (Epizyme Biomedical Technology, Shanghai, China). Protein concentrations of the cell extracts were measured using a Bicinchoninic acid protein assay kit (Epizyme Biomedical Technology, Shanghai, China). Cell lysates were electrophoresed on 7.5% or 15% sodium-dodecyl sulfate gel electrophoresis gels and transferred onto polyvinylidene fluoride membranes. The membranes were blocked with 5% skim milk for 1 hour at room temperature and probed with primary antibodies against CASP1 (catalog No. AF5418; 1:1000; Affinity, Jiangsu, China), sirtuin 2 (SIRT2, catalog No. AF5256, 1:1000, Affinity, Jiangsu, China), and α-tubulin (catalog No. ab176560, 1:2000, Abcam) at 4 °C overnight. After washing, the membranes were incubated with secondary antibodies for 1 hour at room temperature, followed by detection using an enhanced chemiluminescent substrate solution (catalog No. TG196601-01, TG-SCIENCE). The ChemiDoc imaging system (Bio-Rad) was used for blot imaging. Proteins were quantified through the signal ratio using α-tubulin as the loading control.

Statistical analysis

Statistical analysis was performed using R software (version 4.2.2) (https://www.r-project.org/). Differences between the groups were analyzed using the Wilcoxon signed-rank test (P < 0.05). The statistical methods of this study were reviewed by corresponding author Pei-Wen Wu again, and we confirmed it’s appropriate for the research.

RESULTS

Ten intersection genes enriched in the NLR signaling pathway

In total, 2514 DEGs were identified (1603 upregulated and 911 downregulated) (Figure 1A and B). 10 intersection genes (SIRT2, ATAT1, PANX1, NLRP3, TXNIP, CARD8, HSP90AB1, CD36, PYCARD, and CASP1) were screened by overlapping 2514 DEGs and 30 NIRGs (Figure 1C). GO and KEGG enrichment analyses showed that these intersecting genes were enriched in 597 GO terms, including NLRP3 inflammasome complex assembly, inflammasome complex, and cysteine-type endopeptidase activator activity involved in the apoptotic process (Figure 1D-F). They were also enriched for 16 KEGG pathways, including the NLR signaling pathway, lipids, and atherosclerosis (Figure 1G). The protein-protein interaction network included 10 intersecting genes and 19 interactions, including CASP1, NLRP3, and CARD8 (Figure 1H). Based on these results, it is suggested that the identified intersecting genes were not only closely related to the assembly and function of NLRP3 inflammatory vesicles but also played an important role in the pathogenesis and progression of DKD.

Figure 1 The 10 intersection genes were enriched in the NOD-like receptor signaling pathway. A: Volcano plot of differentially expressed genes (DEGs); B: Heatmap of DEGs; C: Venn diagram of the intersection of DEGs and NLRP3 inflammasome-associated genes; D-F: Gene Ontology analysis of the intersection genes, including molecular function, cellular components, and biological processes; G: Significantly enriched Kyoto Encyclopedia of Genes and Genomes pathway analysis of the intersecting genes; H: Protein-protein interaction network of DEGs. DEGs: Differentially expressed genes; NLRP3: NOD-like receptor thermal protein domain associated protein 3; NIRGs: NLRP3 inflammasome-associated genes; GO: Gene Ontology; BP: Biological process; CC: Cellular component; MF: Molecular function; KEGG: Kyoto Encyclopedia of Genes and Genomes; SIRT2: Sirtuin 2; CASP1: Caspase 1.

SIRT2 and CASP1 were considered as biomarkers

Based on the 10 intersection genes (SIRT2, ATAT1, PANX1, NLRP3, TXNIP, CARD8, HSP90AB1, CD36, PYCARD, and CASP1), five genes were identified by LASSO (lambda.min = 3 × 10^-4): SIRT2, PANX1, NLRP3, CD36, CASP1; four by support vector machine recursive feature elimination: CASP1, SIRT2, NLRP3, CARD8RD; and eight by Boruta’s algorithm: SART2, ATAT1, PANX1, TXNIP, CARD8, HSP90AB1, PYCARD, CASP1 (Figure 2A-E). Two biomarkers (SIRT2 and CASP1) were subsequently selected by overlapping these genes using the three algorithms (Figure 2F). The expression of these biomarkers significantly differed between DKD and control samples (P < 0.05) in both GSE30529 and GSE142025 (Figure 2G and H). Specifically, SIRT2 expression was low in DKD samples, whereas the expression of CASP1 was high. The ROC curves in GSE30529 (AUC_SIRT2 = 1.000, AUC_CASP1 = 1.000) and GSE142025 (AUC_SIRT2 = 0.650, AUC_CASP1 = 0.794) showed that the biomarkers could distinguish between DKD and control samples (Figure 2I and J).

Figure 2 Sirtuin 2 and caspase 1 were considered as biomarkers. A and B: Feature genes were selected using least absolute shrinkage and selection operator to obtain five genes; C: The process of feature gene selection using support vector machine recursive feature elimination four feature genes were identified among the differentially expressed low-risk groups; D: Plot of variation in Z-score; E: Boruta selection of nine feature genes with importance rankings; F: Overlapping genes were identified using three machine algorithms; G: Expression of sirtuin 2 (SIRT2) and caspase 1 (CASP1) in GSE30529; H: Expression of SIRT2 and CASP1 in GSE142025; I: Receiver operating characteristic curves for SIRT2 and CASP1 in GSE30529, with an area under the curve value of 1; J: Receiver operating characteristic curves of SIRT2 and CASP1 in GSE142025 with an area under the curve value of 0.65 (SIRT2) or 0794 (CASP1). ^aP < 0.05; ^cP < 0.001; ^dP < 0.0001. LASSO: Least absolute shrinkage and selection operator; SVM-RFE: Support vector machine recursive feature elimination; SIRT2: Sirtuin 2; CASP1: Caspase 1; AUC: Area under the curve.

Biomarkers involved in Leishmania infection pathway

Exploring the relationship between biomarkers and DKD, we found that CASP1 (inference score = 71.81) had a stronger effect on DKD than SIRT2 (Figure 3A). GSEA revealed that SIRT2 was mainly involved in olfactory transduction, spliceosome pathways, and Leishmania infection (Figure 3B). CASP1 was mainly enriched in allograft rejection, cell adhesion molecules, chemokine signaling pathways, and Leishmania infection (Figure 3C). These functional pathways may be involved in DKD development.

Figure 3 The relationship between biomarkers and diabetic kidney disease. A: The relationship between biomarkers and diabetic kidney disease; B and C: Representative pathways enriched in the identified genes, as determined by gene set enrichment analysis (normal P < 0.05); D: Enrichment scores of differentially expressed immune cells between the two groups; E: Corrgrams showing the correlations between biomarkers and differential immune cells based on Pearson’s r values; F: Subcellular localization analysis of biomarkers; G: Chromosomal distribution of biomarkers. ^aP < 0.05; ^bP < 0.01; ^cP < 0.001; ^dP < 0.0001. KEGG: Kyoto Encyclopedia of Genes and Genomes; SIRT2: Sirtuin 2; CASP1: Caspase 1.

High correlation between biomarkers and effector memory CD4⁺ T cell

The Wilcoxon test identified 19 differential immune cells, including natural killer (NK) cells, activated B cells, eosinophils, mast cells, and effector memory CD4⁺ T cells. Except for eosinophils and immature dendritic cells, the enrichment scores of differential immune cells were higher in the DKD samples (Figure 3D). A high correlation was observed between biomarkers and differential immune cells. Effector memory CD4⁺ T cells had the highest negative correlation with SIRT2 (r = -0.87), while NK cells had the highest positive correlation with CASP1 (r = 0.90) (Figure 3E). These findings suggested that the pathophysiological process of DKD may involve complex immune cell interactions, in which the low expression of SIRT2 may be associated with the abnormal function of effector memory CD4⁺ T cells, whereas the high expression of CASP1 may be closely related to the NK cells.

CASP1 and SIRT2 might be expressed in cytoplasm

Subcellular localization analysis revealed that CASP1 and SIRT2 were expressed in the cytoplasm (Figure 3F). Furthermore, the chromosomal distribution indicated that CASP1 and SIRT2 are located on chromosome 11 and chromosome 19 respectively (Figure 3G).

HELLP associated lncRNA/hsa-miR-128-3p/CASP1 was found in the ceRNA network

Thirty TFs were predicted using biomarkers, with early growth response factor 1 (EGR1) jointly predicted by CASP1 and SIRT2 (Figure 4A). 16 miRNAs and 28 lncRNAs were separately predicted, constructing a ceRNA network, such as HELLP associated lncRNA/hsa-miR-128-3p/CASP1 and solute carrier family 9 member A3/hsa-miR-324-5p/SIRT2 (Figure 4B). 23 SNPs were predicted by miRNAs. The miRNA-SNP-biomarkers network (2 biomarkers, 13 miRNAs, 23 SNPs) was constructed after integrating the relationships, such as hsa-miR-128-3p/rs752296690/CASP1 and hsa-miR-302c-3p/rs1391170578/SIRT2 (Figure 4C).

Figure 4 Prediction of drugs through biomarkers and construction of competitive endogenous RNA networks. A: A total of 30 transcription factors were predicted by biomarkers; B: MicroRNAs-long non-coding RNAs co-regulated the competitive endogenous RNA network; C: The microRNA-single nucleotide polymorphism-biomarker network; D: A total of 35 drugs were predicted using biomarkers; E: Docking energy between the drugs and biomarkers. SIRT2: Sirtuin 2; CASP1: Caspase 1.

Vermistatin and paullone might contribute to the treatment of DKD

A total of 35 drugs were predicted by biomarkers (9 drugs by SIRT2 and 26 drugs by CASP1), including vermistatin-CASP1 and paullone-SIRT2 (Figure 4D). Molecular docking showed the docking energy between vermistatin and SIRT2 was -6.63 kcal/mol, and between paullone and CASP1 was -5.46 kcal/mol (Figure 4E and Table 1), confirming their potential as molecular drugs for treating patients with DKD.

Table 1 Names, IDs, and energy of the two molecules.

Molecule_name	CID	Gene_name	PDB_ID	Energy (kcal/mol)
Vermistatin	5467588	SIRT2	1bmq	-6.63
Paullone	369401	CASP1	4rmh	-5.46

CID: Compound ID; PDB: Protein Data Bank; SIRT2: Sirtuin 2; CASP1: Caspase 1.

13 cells annotated by single-cell analysis

Ineligible cells were filtered after quality control, and the remaining cells and genes were used for subsequent analyses (Figure 5A). Two thousand highly variable genes were screened, and the top 10 (PLA2R1, SLC12A1, and PTPRQ) were displayed (Figure 5B). Principal component analysis was performed with no apparent outliers (Figure 5C). The top 30 principal components were used for subsequent analyses (P < 0.05) (Figure 5D). Using uniform manifold approximation and projection cluster analysis, 19 cell clusters were identified (Figure 5E). However, only 13 cells were annotated, namely distal convoluted tubule, proximal convoluted tubule, principal cells, thick ascending limb, connecting tubule, type A intercalated cells, subpopulation of proximal tubule with vascular cell adhesion molecule-1 expression, endothelial cells (ENDO), parietal epithelial cells, podocyte, type B intercalated cells, leukocytes, mesenchymal fibroblast (Figure 5F and G). SIRT2 was mainly expressed in type B intercalated cells and ENDO, while CASP1 was mainly expressed in leukocytes and ENDO (Figure 6A). SIRT2 was highly expressed in control samples, while CASP1 was highly expressed in DKD samples (Figure 6B). This further confirmed the reliability of the previous results.

Figure 5 The 13 cells were annotated by single-cell analysis. A: After quality control, the remaining cells and genes were identified; B: Variance diagram showing variation in gene expression in all cells. Blue dots represent highly variable genes, and black dots represent non-variable genes; C: Principal component analysis exhibited no outliers; D: Principal component analysis identified the top 30 principal components at P < 0.05; E: The uniform manifold approximation and projection algorithm was applied to the top 30 principal components for dimensionality reduction, and 19 cell clusters were successfully classified; F: Expression levels of marker genes in each cell cluster; G: All 13 cell clusters were annotated using singleR and CellMarker based on the composition of marker genes. UMAP: Uniform manifold approximation and projection; MES_FIB: Mesenchymal fibroblast; LEUK: Leukocytes; ICB: Type B intercalated cells; PODO: Podocyte; PEC: Parietal epithelial cells; ENDO: Endothelial cells; PT_VACM1: Proximal tubule with vascular cell adhesion molecule-1 expression; ICA: Type A intercalated cells; CNT: Connecting tubule; TAL: Thick ascending limb; PC: Principal cells; PCT: Proximal convoluted tubule; DCT: Distal convoluted tubule.

Figure 6 Diabetic kidney disease mostly occurred during the mid-differentiation stage of endothelial cells. A: Expression levels of caspase 1 (CASP1) and sirtuin 2 (SIRT2) in each cell cluster; B: Expression of CASP1 and SIRT2 in each group; C: Pseudo-temporal analysis reveals 11 states in the differentiation process of endothelial cells; D: Pseudo-temporal analysis reveals that diabetic kidney disease mostly occurs during the mid-differentiation stage of endothelial cells. DKD: Diabetic kidney disease; SIRT2: Sirtuin 2; CASP1: Caspase 1; ENDO: Endothelial cells; TAL: Thick ascending limb; PT_VACM1: Proximal tubule with vascular cell adhesion molecule-1 expression; PODO: Podocyte; PEC: Parietal epithelial cells; PCT: Proximal convoluted tubule; PC: Principal cells; MES_FIB: Mesenchymal fibroblast; LEUK: Leukocytes; ICB: Type B intercalated cells; ICA: Type A intercalated cells; DCT: Distal convoluted tubule; CNT: Connecting tubule.

DKD mostly occurred during the mid-differentiation stage of ENDO

ENDO was selected for pseudo-temporal analysis because both biomarkers were expressed in ENDO. ENDO differentiated from left to right over time, with blue indicating the early stages of differentiation, and there were 11 states in the differentiation process (Figure 6C). DKD samples were fewer in the early and later stages of differentiation, indicating that DKD mostly occurred during the mid-differentiation stage of ENDO (Figure 6D).

Western blot

Western blotting was performed to evaluate SIRT2 and CASP1 expression. CASP1 expression in the renal cortex increased in the DKD group compared to the control group (P < 0.05, Figure 7A and B). Conversely, SIRT2 levels decreased in the DKD group compared to the control group (P < 0.05, Figure 7C and D). These findings not only validated previous findings, but also revealed the important roles of CASP1 and SIRT2 in the pathological mechanisms of DKD.

Figure 7 Validation of high caspase 1 and low sirtuin 2 expressions in diabetic kidney disease models. A and B: Western blot staining of caspase 1 between the diabetic kidney disease and control groups; C and D: Western blot staining of sirtuin 2 between the diabetic kidney disease and control groups (n = 4). CASP1: Caspase 1; SIRT2: Sirtuin 2.

DISCUSSION

Fibrosis is the overall pathological progression of DKD, with inflammatory cell infiltration as a key mechanism of kidney fibrosis in DKD[38]. Recent studies suggested that NLRP3 overexpression in the kidney is associated with macrophage infiltration and kidney fibrosis during the transition from renal inflammation to fibrosis[39]. The NLRP3 inflammasome is also considered a key modulator of inflammation in DKD. NLRP3 expression is significantly increased in the glomeruli and renal tubular epithelial cells of patients and mice with DKD. Inhibition of the NLRP3 inflammasome can reduce inflammation and fibrosis in DKD renal tissue[40]. For example, the deletion of NLRP3 can reduce inflammation and renal fibrosis in diabetic mice. Li et al[41] demonstrated that Astragaloside IV alleviates renal tubular interstitial inflammation and renal tubular epithelial cell apoptosis in DKD rats by inhibiting NLRP3 inflammasome activation. Another experimental research has shown that the NLRP3 inflammasome, as an initiating signal, positively regulates interleukin-1β secretion in a myeloid differentiation primary response protein 88-dependent manner, promoting the development of renal inflammation and fibrosis in mice[42]. Therefore, the NLRP3 inflammasome plays a significant role in the onset and progression of DKD. Further studies are imperative to deepen our understanding of DKD and identify potential therapeutic targets.

Using a range of analyses, we identified two biomarkers, SIRT2 and CASP1, that were significantly correlated with the NLRP3 inflammasome. SIRT2, initially found in yeast, is an NAD⁺-dependent deacetylase with histone deacetylase activity[43]. Recent studies have highlighted the pivotal role of SIRT2 in various diseases, including hypertensive nephropathy[44]. Jennings et al[45] reported that the deacylase SIRT2 acts as the first known “eraser” of lactylation, expanding the knowledge of enzymatic regulation of non-enzymatic derived protein post-translational modifications. Overexpression of SIRT2 in renal tubular epithelial cells of mice with unilateral ureteral obstruction and unilateral ischemia-reperfusion injury mitigated the degree of renal fibrosis[46]. This suggests that SIRT2 could potentially serve as an effective therapeutic approach for the prevention and treatment of renal fibrosis in chronic kidney disease. Although research on the role of SIRT2 in DKD remains limited, a recent experimental study revealed the efficacy of tenovin-1, a SIRT inhibitor, in mitigating the fibrotic phenotype in DKD[47]. CASP1, an essential enzyme in the cellular pyroptosis pathway and a key inflammasome component, is significantly associated with various diseases, including cancer and coronary heart disease[48-50]. Evidence has also revealed potential connections between the classical pyroptosis pathway involving inflammasomes and CASP1[51]. Activated CASP1 contributes to the mediation of inflammation. CASP1 can cleave gasdermin D and plays a role in promoting pyroptosis, an inflammatory form of cell death[52]. The activation of CASP1, which induces inflammation, is associated with the occurrence of acute kidney injury, fibrosis, and DKD. This indicates that CASP1 inhibitors could potentially provide a clinically translatable treatment strategy for DKD. However, no CASP1 inhibitors have officially entered the clinic yet, as there is still a need to overcome issues such as insufficient efficacy, poor targeting, or adverse side effects, indicating that further optimization is required. Therefore, we speculate that SIRT2 and CASP1 may promote the progression of DKD through inflammation and pyroptosis.

Using GSEA, we identified that the two biomarkers under investigation were mainly enriched in pathways associated with Leishmania infection, cell adhesion molecules, and chemokine signaling. Notably, the Leishmania infection pathway was recently demonstrated to be an upstream activation pathway of the NLRP3 inflammasome[53]. Numerous studies indicate that cell adhesion molecules play a crucial role in diabetic vascular complications by contributing to cellular proliferation, differentiation, formation of cell connections, and stimulation of white blood cells at inflammation sites[54]. Similarly, chemokines significantly contribute to the inflammatory phenotype of DKD, as corroborated by numerous preclinical studies[55,56].

The immune infiltration analysis conducted in this study showed that, among the 19 significantly different immune cells, the expression of all except eosinophils and immature dendritic cells was higher in the disease group than in the control group, indicating a higher concentration of immune cells in patients with DKD. The principal biomarkers identified were strongly correlated with T cells, B cells, and NK cells. Contemporary studies have shown that immune cell infiltration, encompassing elements from both the innate and adaptive immune systems, contributed to renal injury triggered by hyperglycemia[57]. Studies suggested that SIRT2 negatively affected key enzymes in various metabolic pathways through deacetylation, including glycolysis, the tricarboxylic acid cycle, fatty acid oxidation, and glutaminolysis. These pathways are crucial for T cell effector functions. In the absence of SIRT2, T cell activation led to hyperacetylation and enhanced activity of multiple metabolic enzymes, resulting in increased aerobic glycolysis, oxidative phosphorylation, fatty acid oxidation, and glutaminolysis[58]. Consistent with our research findings, low expression of SIRT2 in DKD may be associated with abnormal functions of effector memory CD4⁺ T cells. During DKD, inflammatory cytokines secreted by T cells can induce the transformation of epithelial cells into mesenchymal cells and extracellular matrix accumulation, revealing the intrinsic link between T cells and the inflammatory phenotype in DKD[59]. Furthermore, a substantial body of research has verified the critical involvement of diverse T cell subsets, including T helper cells, regulatory T cells, CD8⁺ T cells, and mucosal-associated invariant T cells, in the pathophysiology of DKD[60]. Although an increase in eosinophil count is common in many kidney diseases[61], it rarely occurs in DKD and may be related to drug-induced hypersensitivity reactions[62] or interstitial nephritis[63]. Consistent with our findings, eosinophils in the control group showed higher immune enrichment scores. Additionally, studies have indicated that elevated levels of blood NK cells are a risk factor for DKD[64]. Our research found that the high expression of CASP1 may be closely related to NK cells, suggesting a possible interaction. This indicates that NK cells may serve as a marker for inflammasome activation in DKD. Therefore, we speculate that CASP1 may affect the occurrence and development of DKD by promoting the activated state and function of NK cells in the blood. Further basic experiments are still needed to verify this. Hence, reducing NK cell activation or inhibitors of CASP1 are expected to become therapeutic targets, providing a theoretical basis for the development of new diagnostic biological markers for DKD and further research into its molecular mechanisms.

Analysis of the TFs-miRNA regulatory network revealed that EGR1 functions as a shared TF modulating SIRT2 and CASP1. EGR1 is a TF that is rapidly activated by various stimuli, such as hypoxia and growth factors[65]. It encodes a protein with a zinc-finger motif and plays a pivotal role in regulating genes essential for cell differentiation, proliferation, and inflammation[66,67]. Relevant studies have also identified it as a characteristic gene of endoplasmic reticulum stress associated with DKD[68], demonstrating its potential role in DKD progression.

Single-cell analysis revealed that endothelial cells commonly expressed all key biomarkers identified in this study. As the first point of contact between the kidney and blood, endothelial cells are primary targets of hyperglycemic damage. Hyperglycemia activates multiple pathways, causing the release of pro-inflammatory cytokines and renal endothelial cell dysfunction, which may further accelerate DKD progression[69]. Endothelial cell dysfunction triggered by various mechanisms, including inflammation, plays a role in the onset of DKD[54]. The key biomarkers identified in this study may also be involved in this process.

Spatiotemporal analysis findings revealed that a significant proportion of endothelial cells in the DKD samples were in an intermediate stage of differentiation, potentially owing to the pathogenesis of DKD. Compared to existing literature[70,71], this study enhances the credibility of our findings by integrating bioinformatics analysis with laboratory research, offering robust evidence for our conclusions. Additionally, by employing single-cell analysis, this research provides deeper insights into cellular mechanisms, significantly contributing to the understanding of related studies in this field.

The key data obtained from this study lay the foundation and provide direction for further exploration of the pathological mechanisms and therapeutic targets of DKD. However, despite progress in revealing potential mechanisms, our study has some limitations. For example, the clinical manifestations of human DKD are highly heterogeneous and may not be fully captured by animal models. To address this, we plan to increase the sample size in subsequent studies and strive to collect more types of clinical samples. We will validate the performance of biomarkers in a larger and more diverse population to comprehensively verify and consolidate our research findings while avoiding statistical significance and generalizability issues. Additionally, further research is needed to enhance the reliability and validity of biomarkers. Finally, the role of biomarkers such as SIRT2 and CASP1 in different chronic kidney diseases remains to be explored.

CONCLUSION

This study identified two biomarkers, SIRT2 and CASP1, associated with the NLRP3 inflammasome in DKD. GSEA revealed that these biomarkers were predominantly enriched in pathways, such as Leishmania infection, cell adhesion molecules, and chemokine signaling. Immune infiltration analysis showed significant differences in 19 types of immune cells between the disease and control groups, correlating with these biomarkers. Subcellular localization analysis indicated that both biomarkers were expressed in the cytoplasm. Using single-cell analysis, we identified 13 annotated cell types and performed a pseudo-temporal analysis of endothelial cells expressing both biomarkers. We found that most endothelial cells in the disease samples were in the mid-stage of differentiation. Because of this study’s limitations, further experimental validation is required to confirm these results. Based on these findings, we will continue focusing on understanding the mechanisms underlying DKD for its diagnosis, treatment, and identification of biomarkers.

ACKNOWLEDGEMENTS

We sincerely thank the Gene Expression Omnibus database.