Identifying Crucial Genes Associated with Pyroptosis in Lupus Nephritis

Abstract

Lupus nephritis (LN), a severe manifestation of systemic lupus erythematosus, involves immune complex deposition, inflammation, and kidney damage. Recent studies indicate that pyroptosis, a pro-inflammatory cell death process, drives renal injury in LN. This study intended to identify key pyroptosis-related genes in LN using datasets from the GEO database, encompassing glomerular, tubulointerstitial, and whole kidney tissues from LN patients. Identified differentially expressed genes related to pyroptosis and created a predictive model using univariate and LASSO regression analysis. LN patients were classified into subtypes through consensus clustering. Immune microenvironment characteristics and hallmark pathways were further analyzed. Using the WGCNA, key gene modules and hub genes were recognized, followed by an analysis of their clinical relevance and distribution patterns using the Nephroseq database and scRNA-seq data. Cellular experiments were conducted to validate the findings. We identified 26 differentially expressed pyroptosis-related genes in LN glomeruli and created a 10-gene model with high diagnostic accuracy (AUC: 0.968 for tubulointerstitium, 0.990 for whole kidney). Consensus clustering divided LN into two subtypes: subtype1, characterized by inflammation and immune activation, and subtype2, characterized by cellular metabolism. WGCNA highlighted the grey60 module linked to subtype1, and identified GBP2 and EIF2AK2 as hub genes. Cellular experiments showed that GBP2 and EIF2AK2 were upregulated in LPS-stimulated macrophages and glomerular endothelial cells, and their siRNA-mediated knockdown triggered a decline in pyroptosis-related marker expression, implying their possible role as therapeutic targets for modulating pyroptosis in LN. In conclusion, GBP2 and EIF2AK2 show potential as candidate molecules for targeted therapy in LN.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10753-025-02402-5.

Introduction

Lupus nephritis (LN), a severe renal complication of systemic lupus erythematosus (SLE), is driven by immune dysregulation, chronic inflammation, and irreversible kidney damage [1, 2]. Despite improvements in the knowledge and management of LN, it still plays a substantial role in the elevated morbidity and mortality rates among patients with SLE [3]. Abnormal renal cell death and heightened proinflammatory responses are commonly observed in LN patients [4]. Pyroptosis, a lytic and pro-inflammatory form of programmed cell death, has recently emerged as a critical contributor to renal injury in LN [5, 6]. In LN kidney tissue, gasdermin-induced pore formation and cytokine release (IL-1β, IL-18) drive renal cell death and inflammation [7, 8]. Given the central role of inflammation in LN pathogenesis, targeting pyroptosis-associated inflammatory signaling pathways has emerged as a promising therapeutic strategy to mitigate renal injury and preserve function. However, the specific genes regulating pyroptosis in LN remain poorly defined, hindering the development of precision therapies.

This study aims to systematically characterize pyroptosis-related gene networks in LN and identify potential therapeutic targets. Using glomerular, tubulointerstitial, and whole kidney tissue data from LN patients, we identified pyroptosis-associated hub genes and developed a pyroptosis gene signature. LN patients were stratified into distinct subtypes using an integrated analysis framework incorporating consensus clustering, immune profiling, and weighted gene co-expression network analysis (WGCNA). Employing Nephroseq for data analysis to explore key gene-clinical feature associations. Single-cell sequencing (scRNA-seq) data helped us pinpoint the cell types with changes in key pyroptosis-related genes, which we subsequently validated through cellular experiments. Our results may provide insights for developing new therapeutic strategies targeting LN-associated inflammatory pathways, which could potentially improve renal function and patients’ quality of life.

Materials and Methods

Obtainment and Preprocessing of LN Datasets

Gene expression data of LN patients were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). The detailed information of these datasets is listed in Supplementary Table S1. Four glomeruli datasets (GSE127797, GSE32591, GSE99339 and GSE104948, including 135 from LN patients and 51 control samples), three tubulointerstitium datasets (GSE69438, GSE127797, GSE32591, including 95 from LN patients and 15 control samples) and one whole kidney tissue dataset (GSE112943, including 14 from LN patients and 7 control samples) were chose for our next analysis. Using the GEOquery package, we obtain gene matrix data along with the probe annotation file. The expression level of a gene is determined by averaging the values from all probes that map to it.

Since the glomeruli and tubulointerstitium samples were derived from different GSE datasets, batch effect correction was necessary. To address batch effects across different platforms, we first visualized the data using principal component analysis (PCA) plots and boxplots of all samples (Supplementary Fig. S1a-b and S2a-b). These plots revealed significant batch effects among data from different sequencing platforms. We regarded the different platforms as the source of batch effects and applied the ComBat function from the sva package to remove them [9]. As shown in Supplementary Fig. S1c-f, after processing glomeruli samples with the ComBat function, abnormal samples were still present. Therefore, we removed 9 outlier samples identified through hierarchical clustering analysis using the “hclust” R package, based on a clustering height threshold of 140. Ultimately, we retained 128 LN samples and 49 normal control samples for the glomeruli datasets. For the tubulointerstitium samples, the PCA plot, boxplot, and hierarchical clustering plot after batch effect removal are shown in Supplementary Fig. S2c–e. These plots demonstrate that the batch effects in the tubulointerstitial data were successfully removed using this approach. The data quality control chart for the whole kidney is presented in Supplementary Fig. S3.

Identification of Differentially Expressed Pyroptosis‑Related Genes

We identified four biological process pathways pertinent to pyroptosis from the Molecular Signatures Database (MSigDB) [12]. The detailed information is presented in the Supplementary Table S2. We took the union of the genes from these 4 gene sets to obtain 93 pyroptosis-related genes for this study. Among these genes, 61 were expressed in glomeruli, 66 in tubulointerstitial tissues, and 86 in whole kidney tissues. By identifying the intersection of these renal-expressed genes, we defined a panel of LN-associated pyroptosis genes for subsequent analysis. To identify genes that were differentially expressed within the datasets, we utilized the “limma” R package, applying a significance threshold of P < 0.05. Subsequently, considering the varying sample sizes across different datasets, we conducted normality tests using either the Kolmogorov-Smirnov test or the Shapiro-Wilk test. Based on the results of these normality tests, we selected either the t-test or the Wilcoxon test to validate the differentially expressed pyroptosis-related genes that had been identified through the limma analysis. Finally, we visualized the findings using box plots.

Correlation Analysis

After identifying differentially expressed pyroptosis-related genes, we performed Pearson correlation analysis using the rcorr function from the “Hmisc” R package to investigate interactions and potential regulatory relationships among these genes in LN samples. Visualize the resulting correlations using the “corrplot” R package and create visualizations for the correlation between individual gene pairs using the “ggstatsplot” R package.

Establishment and Validation of a Pyroptosis Gene Signature

To identify key pyroptosis-related genes with predictive value from the differentially expressed genes(DEGs) and construct a robust predictive model, we initially screened variables using univariate logistic regression, and optimal predictive features were selected via least absolute shrinkage and selection operator (LASSO) regression. A generalized linear model was fitted to avoid model overfitting. For the pyroptosis-related DEGs with non-zero LASSO regression coefficients, they were selected and their corresponding regression coefficients were obtained to construct a model. Subsequently, the risk scores of this predictive model were calculated based on glomerular expression data. The “forestploter” R package is utilized to create a forest plot that illustrates the results of the logistic regression. The optimism of the model’s predictive probabilities is quantified by the average receiver operating characteristic (ROC) and area under the curve (AUC). The risk score for each sample is calculated as follows:

Consensus Clustering

To identify different pyroptosis-related molecular subtypes of LN, we did an unsupervised clustering analysis based on pyroptosis-related DEGs. We employed the R package “ConsensuClusterPlus”, selected the k-means method, and performed 1,000 iterations, choosing Euclidean distance as the clustering metric. The expression status of pyroptosis-related DEGs in the two subtypes was evaluated through t-tests, with a P-value < 0.05 considered statistically significant.

Immune Characteristics

To characterize the immune regulatory mechanisms underlying LN and their association with pyroptosis, we analyzed the immune microenvironment using a multi-dimensional approach. We quantified immune cell infiltration via CIBERSORTx with 1,000 permutations, a deconvolution algorithm applied to bulk RNA-seq data [10], and assessed immune response activity using Single Sample Gene Set Enrichment Analysis (ssGSEA) with ImmPort gene sets. Given the critical role of chemokines in leukocyte trafficking to inflamed kidneys, we curated LN-relevant chemokines and receptors from the literature [11]. We performed Pearson correlation analysis to link pyroptosis-related genes to these immune characteristics, including patterns of immune cell infiltration, immune pathway activity, and chemokine/chemokine receptor expression. Additionally, we explored correlations between molecular subtypes of LN (derived from pyroptosis-driven clustering) and immune features, encompassing immune cell fractions, pathway activity scores, and chemokine system dynamics.

Functional Enrichment Analysis

For the purpose of investigating the biological functions associated with pyroptosis-related molecular subtypes, we retrieved the gene set “h.all.v7.4.symbols” from the MSigDB database [12]. Pathway activation scores were calculated using the Gene Set Variation Analysis (GSVA) algorithm. The R package “limma” was applied to compare pathway activation score discrepancies between the two subtypes. To identify important modules associated with different modification patterns, WGCNA was conducted.

Correlation and Distribution Patterns of Hub Genes in Lupus Nephritis

The association between hub genes and various factors (including pathological classification, chronic kidney disease (CKD) stages, proteinuria, and renal function indicators) in LN patients was analyzed using the Nephroseq database (https://nephroseq.org/) [13, 14]. Additionally, the distribution and expression characteristics of these hub genes within the kidney were examined using single-cell RNA sequencing (scRNA-seq) data (https://singlecell.broadinstitute.org/single_cell/study/SCP279/amp-phase-1). Since one scRNA-seq dataset included both skin and kidney tissues [15, 16], skin-derived cells were excluded, and only kidney tissue cells were retained for subsequent analysis. Another dataset comprised scRNA-seq data of immune cells in the kidneys of LN patients [17]. T-Distributed Stochastic Neighbor Embedding (t-SNE) was used to visualize cell distribution, and the “plot1cell” R package was employed to visualize the expression of GBP2 and EIF2AK2 across different experimental groups.

Cell Experiment

During this cellular experiment, RAW 264.7 and mouse glomerular endothelial cells (MGECs) were transfected with siRNA targeting GBP2 and EIF2AK2 (sequences detailed in Supplementary Table S3) using Lipomaster 3000 Transfection Reagent (catalog No. TL301-01-AA, Vazyme Biotech Co., Ltd) for 24 h. Following siRNA intervention, cells were stimulated with lipopolysaccharide (LPS; catalog number HY-D1056, MedChemExpress) at a final concentration of 10 ng/mL for an additional 24 h. The stimulated RAW 264.7 cells and MGECs were extracted according to the manufacturer’s instructions (phosphorylation protease inhibitor, PMSF, and protease inhibitor cocktail, catalog No. G2007-1ML, G2008-1ML, and G2006-250UL, Servicebio) and the protein concentration were quantified using the BCA Protein Assay (catalog No. G2026-1000T, Servicebio). Proteins were separated by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Molecular weight markers (catalog No. MP102, Vazyme Biotech Co., Ltd.) were used to indicate the protein sizes. Subsequently, the separated proteins were transferred onto polyvinylidene difluoride (PVDF) membranes (catalog No. WGPVDF45, Servicebio). The membranes were then blocked with 5% non-fat dried milk (catalog No. GC310001-100 g, Servicebio). Primary antibodies (GAPDH, 1:10000, catalog No.60004-1-Ig; Proteintech; EIF2AK2,1:1000, catalog No. A4047, Abconal; GSDMD (Full length + N terminal),1:1000, catalog No. A18281, Abconal; GBP2, 1:1000, catalog No. 11854-1-AP, Proteintech; NLRP3, 1:1000, catalog No. 15101, CST) were treated with the membranes at 4 °C overnight. The membranes were incubated for 1 h at room temperature with an HRP-conjugated secondary antibody (anti-mouse IgG, catalog No.AS003, 1:5000, Abconal; anti-Rabbit IgG, catalog No.AS014, 1:5000, Abconal). The gray values of the target protein bands were analyzed using Image J software (NIH, Bethesda).

Statistical Analysis

Statistical analyses were performed utilizing R software (version 4.3.3) and GraphPad Prism (version 9.5). Independent sample Wilcoxon rank-sum tests or t-tests were applied for conducting statistical comparisons. In cases involving multiple group comparisons, one-way analysis of variance (ANOVA) coupled with multiple comparison tests was employed. The flow chart depicting the study’s process is illustrated in Fig. 1.

Fig. 1.

Flow chart of the research study

Results

Pyroptosis-Related Gene Expression and Differential Analysis in Lupus Nephritis

From the MSigDB, we curated a list of 93 pyroptosis-associated genes. By integrating transcriptomic data from glomeruli, tubulointerstitial, and whole kidney tissues retrieved from the GEO database, we identified 58 pyroptosis-related genes commonly expressed across all three compartments in LN patients (Fig. 2a). A protein-protein interaction (PPI) network was constructed to visualize functional associations among these 58 genes (Fig. 2b). Differential expression analysis using the “limma” R package revealed 26 significantly dysregulated pyroptosis-related genes in glomeruli (Fig. 2c). Box plots (Fig. 2d-f) validated the limma-identified DEGs using t-tests for glomerular/tubulointerstitial datasets and Wilcoxon tests for whole-kidney datasets. Figure 2e and f validated the differentially expressed glomerular pyroptosis-related genes using datasets from tubulointerstitial tissue and whole kidney tissue, respectively. Notably, 8 genes (IFI27, GBP2, PYCARD, CASP1, GBP1, TRIM21, EIF2AK2, CARD8) were consistently upregulated across all three datasets. No overlapping downregulated pyroptosis-related genes were identified across datasets; however, MAPK8 exhibited reduced expression specifically in the glomerular and tubulointerstitium datasets. Overall, these findings highlight specific pyroptosis-related genes that are dysregulated in LN across different kidney compartments, providing insights into the potential mechanisms underlying the disease.

Fig. 2.

Pyroptosis-Related Gene Expression and Differential Analysis in Lupus Nephritis. a Venn diagram of co-expressed pyroptosis genes in different datasets of LN. b PPI network of co-expressed pyroptosis genes. c Volcano plot of the 26 differentially expressed pyroptosis-related genes in glomeruli. d Box plot of 26 pyroptosis-related DEGs in glomeruli. e Box plot of expression of 26 differentially expressed glomerular pyroptosis genes in the tubulointerstitial datasets. f Box plot of expression of 26 differentially expressed glomerular pyroptosis genes in the whole kidney dataset. P values were shown as: *P < 0.05; **P < 0.01; ***P < 0.001; **** P < 0.0001, ns: no significant difference

Correlation Analysis of Pyroptosis-Related Genes in Lupus Nephritis

To elucidate the functional interplay and regulatory networks underlying the dysregulated pyroptosis-related genes identified in differential expression analysis, we investigated their expression correlations across kidney compartments in LN samples. Figure 3a depicts the expression correlation of 26 pyroptosis-related differential genes in glomerular samples. Notably, the expression of CASP1 exhibits a significant positive correlation with GBP1 (R = 0.85, P = 3.76e-36), and similarly, CASP1 also demonstrates a significant positive correlation with GBP2 (R = 0.81, P = 1.47e-30). Figure 3b, on the other hand, illustrates the correlation of pyroptosis-related differential gene expression in tubulointerstitium from LN patients, where CASP1 again shows significant positive correlations with both GBP1 and GBP2. In whole kidney tissues, while a relatively weak and non-significant positive correlation is observed between CASP1 and GBP2, a notable significant positive correlation emerges between IFI27 and EIF2AK2 (R = 0.85, P = 3.76e-36, Fig. 3c).

Fig. 3.

Correlation Analysis of Pyroptosis-Related Genes in Lupus Nephritis Samples. a Correlation analysis of pyroptosis-related genes in glomerular samples from LN. b Correlation analysis of pyroptosis-related genes in tubulointerstitial samples from LN. c Correlation analysis of pyroptosis-related genes in the whole kidney samples from LN

Establishment and Verification of a Pyroptosis-Associated Gene Signature for Lupus Nephritis

Building upon the dysregulated pyroptosis-related genes identified through differential expression analysis, we further employed univariate logistic regression and LASSO regression to distill a minimal gene signature with optimal diagnostic performance for LN. We initially utilized univariate logistic regression analysis to screen out genes linked to LN. A total of 23 genes meeting the significance threshold (P < 0.05) were retained for downstream analysis (Fig. 4a). The glomeruli dataset was partitioned into a training set (70%) and a validation set (30%) using the caret R package. These 23 candidate genes were then subjected to LASSO regression, which selected 10 genes based on the optimal lambda (λ) value to form the final predictive model (Fig. 4b, c). The risk score for each sample was calculated as follows: Risk Score = (0.6023 × IFI27 expression) + (0.4547 × MYD88 expression) + (0.2765 × GBP2 expression) + (0.3541 × PYCARD expression) + (0.2670 × TRIM21 expression) + (1.3060 × EIF2AK2 expression) + (0.2417 × CASP4 expression) + (−0.7105 × MAPK8 expression) + (0.8358 × NLRP3 expression) + (−1.8923 × BTK expression). Notably, LN patients exhibited significantly higher risk scores than controls in both the training and validation cohorts (Fig. 4d). The validation set of the predictive model demonstrated an AUC of 0.968, indicating its excellent performance in distinguishing between healthy and LN samples (Fig. 4e). For the independent datasets comprising tubulointerstitium and whole kidney tissues, the AUCs were 0.968 and 0.990, respectively, further underscoring the model’s proficiency in classifying samples (Fig. 4f, g). The distribution based on the 10 pyroptosis gene risk scores and gene profiles is presented in Fig. 4h. Notably, we observed elevated expressions of IFI27, MYD88, GBP2, PYCARD, TRIM21, EIF2AK2, CASP4, NLRP3, and BTK in the high-risk group, whereas MAPK8 exhibited decreased expression in this group. These results suggest that the model demonstrates reliable performance in distinguishing LN cases from controls across diverse datasets, indicating it may serve as a candidate diagnostic or prognostic biomarker.

Fig. 4.

Establishment and verification of a pyroptosis-associated gene signature for lupus nephritis. a Univariate logistic regression analysis of the 26 pyroptosis-related DEGs in glomeruli. b LASSO regression of the 23 pyroptosis-related DEGs. c Cross-validation for tuning the parameter selection in the LASSO regression. d Risk score of the Lasso model of the normal and LN group. e - g The predictive capacity of the pyroptosis-related gene signature was evaluated in validation sets of glomeruli (e), tubulointerstitial tissues (f), and whole kidney datasets (g) by calculating their combined AUCs. h The risk score distribution is based on the pyroptosis-related gene signature and gene expression profiles from the study group

Pyroptosis-Driven Immune Heterogeneity in Lupus Nephritis

To elucidate the immunopathological mechanisms of LN, we conducted a comprehensive analysis integrating pyroptosis-related gene signatures expression profiles with immune microenvironment characteristics. We characterized immune cell dynamics via CIBERSORTx, revealing distinct correlations between pyroptosis gene expression and immune subsets: activated mast cells, monocytes, and macrophages (M0/M1/M2) showed positive associations, while regulatory T cells and resting mast cells were negatively correlated (Supplementary Fig. S4a). ssGSEA analysis linked most pyroptosis genes to enhanced immune receptor signaling (chemokine/cytokine receptors, TCR/BCR pathways) but reduced interferon activity (Supplementary Fig. S4b). We incorporated literature-curated LN-related chemokine/chemokine receptors (e.g., CCR1, CCR5, CCL2, CXCL10) and found positive correlations for 9/10 pyroptosis genes (IFI27, MYD88, NLRP3, etc.) with these factors, contrasting with MAPK8’s inverse associations (Supplementary Fig. S4c).

To investigate the connection between glomerular expression of pyroptosis-related gene signatures and LN subtypes, we performed consensus clustering on all LN patients. As we increased the clustering variable (k) from 2 to 10, it became evident that at k = 2, the intra-cluster correlation was maximized while the inter-cluster correlation was minimized. This suggests that the 128 LN patients could be effectively classified into two clusters based on the 10 DEGs mentioned earlier. (Fig. 5a-c). 51 cases were included in pyroptosis-related subtype1, and 77 cases were included in pyroptosis-related subtype2. Subsequent differential analysis of the 10 pyroptosis-related genes revealed that subtype1 exhibited higher expression levels of IFI27, MYD88, GBP2, PYCARD, TRIM21, EIF2AK2, CASP4, NLRP3, and BTK, while subtype2 was characterized by enhanced MAPK8 expression (Fig. 5d). Additionally, we performed comparative analysis of chemokine/receptor expression profiles across LN subtypes. As illustrated in Fig. 5e, Subtype1 exhibited significantly higher expression of CCL2, CCL5, CCR1, CCR5, CX3CR1, CXCL10, and CXCR4 compared to subtype2. And we employed the CIBERSORT algorithm to figure out the immune infiltration scores of LN samples between the two clusters. Comparing the scores between subtype1 and subtype2, significant differences were found in two types of T cells. Subtype2 had higher infiltration levels of naive CD4⁺ T cells and follicular helper T cells (Fig. 5f). Regarding immune responses, Antigen Processing and Presentation, Antimicrobials, and BCR Signaling Pathway were more active in subtype1, whereas TCR Signaling Pathway and Natural Killer Cell Cytotoxicity were more active in subtype2 (Fig. 5g). Pyroptosis genes may play a crucial role in regulating the immune microenvironment of LN. To delve into the biological functional pathways that may be impacted by pyroptosis genes, we conducted GSVA to evaluate the enrichment status of these biological pathways. Figure 5h displays the enrichment differences of hallmark pathways between the two subtypes, indicating that inflammatory pathways such as α-interferon response, γ-interferon response, inflammatory response, IL6/JAK/STAT3 signal, as well as damage response pathways like apoptosis and DNA repair, were more enriched in subtype1. Conversely, KRAS signaling and xenobiotic metabolism were more enriched in subtype2. Furthermore, we assessed the International Society of Nephrology/Renal Pathology Society (ISN/RPS) LN class distribution across the two pyroptosis-related subtypes using GSE127797 data. Proliferative classes (III/IV/III + V/IV + V) predominated over non-proliferative classes (I/II/V) in both subtypes (Supplementary Fig. S5). This proliferative dominance suggests these molecular subtypes harbor heightened proliferative capacity, potentially correlating with aggressive disease progression and adverse prognosis. Collectively, our findings demonstrate that pyroptosis-related gene signatures stratify LN into two immunologically distinct subtypes, with subtype1 exhibiting a pro-inflammatory phenotype marked by elevated chemokine/chemokine receptors activity, interferon responses, and IL6/JAK/STAT3 pathway activation, suggesting its role as an inflammation-driven LN entity with therapeutic implications.

Fig. 5.

Differential Analysis of Immune Microenvironment, and Biological Functional Pathways in LN Subtypes. a Consensus clustering of cumulative distribution function (CDF) for k = 2–10. b Elbow plot shows the relative change in area under the CDF curve. c Consensus clustering matrix for k = 2. d The two subtypes exhibit distinct expression statuses of a boxplot. e Chemokine/receptor expression profiles differentiate LN subtypes. f The immune cell infiltration scores in two distinct pyroptosis-associated subtypes. g The immune response gene sets in two distinct pyroptosis-associated subtypes. h Differences in hallmark pathway enrichment pyroptosis-associated subtypes. *P < 0.05, *P < 0.01, ***P < 0.001, ****P < 0.0001, ns not significant

WGCNA Analysis Reveals Key Gene Modules and Hub Genes Related To Pyroptosis

To elucidate the regulatory networks underlying pyroptosis-driven immune microenvironment remodeling in LN, we performed WGCNA to identify co-expressed gene modules and hub genes that orchestrate these processes. By clustering genes with similar expression profiles, we delineated 16 distinct modules (Fig. 6c). Among these, the MEgrey60 module was enriched in pathways such as “response to type I interferon” and “regulation of viral life cycle”—pathways critically linked to inflammatory and antiviral responses. Other modules included MEmidnightblue (“immune response signaling pathway”), MEgreenyellow (“interleukin-1 production”), and MEpurple/MEdarkgreen (“fatty acid catabolic process”) (Fig. 6d). Notably, MEgrey60 exhibited the strongest positive correlation with inflammation-driven subtype1 (Fig. 6e), suggesting its pivotal role in shaping the pro-inflammatory microenvironment. A gene co-expression network was subsequently constructed using MEgrey60 genes (Fig. 6f). Genes overlapping between the pyroptosis-related signature and MEgrey60 were prioritized as candidate hub genes, with GBP2 and EIF2AK2 identified as key regulators. MEgrey60 was selected as the master module based on its pathway relevance to pyroptosis-mediated inflammation, alignment with subtype1’s pro-inflammatory features, and validation of core regulatory genes. Collectively, WGCNA advanced our understanding of pyroptosis-driven immunopathology in LN by pinpointing MEgrey60 as a central regulatory module and GBP2/EIF2AK2 as critical hub genes within this network.

Fig. 6.

WGCNA Analysis Reveals Key Gene Modules and Hub Genes Related to Pyroptosis. a Scale-free fitting index analysis and mean connectivity of soft threshold power from 1 to 20. b Clustering dendrograms for pyroptosis genes: hierarchical clustering of pyroptosis-related genes, with modules merged if their correlation exceeds 0.75 based on dynamic tree cutting. c Module subtype correlation heatmap: visualizing the correlation between different module subtypes. d The dot plot of the top GO enrichments among various modules. e Scatterplot of subtype1 in the grey60 module. f A gene co-expression network for the grey60 module with a threshold of 0.05. g Venn Diagram of Pyroptosis-related gene signature and module grey60 genes

Clinical and Pathological Associations of GBP2 and EIF2AK2 in Lupus Nephritis

Having identified GBP2 and EIF2AK2 as key hub genes within the MEgrey60 module, we investigated their clinical relevance in LN pathogenesis using the Nephroseq database to assess translational potential. GBP2 showed significant upregulation in the World Health Organization (WHO) class II/III LN patients compared to healthy controls (Fig. 7a). Its expression progressively increased across CKD stages 1–3, peaking at stage 2 (Fig. 7b). While elevated in both subnephrotic and nephrotic proteinuria groups, subnephrotic patients exhibited notably higher GBP2 levels (Fig. 7c). In the ISN/RPS pathological classifications, GBP2 expression was numerically greater in combined III + V and IV + V subtypes compared to other categories (Fig. 7d), though differences lacked statistical significance due to limited sample sizes. EIF2AK2 demonstrated parallel trends to GBP2. It was significantly upregulated in the WHO class II/III LN versus controls (Fig. 7e), increased across CKD stages 1–2 (peaking at stage 2, Fig. 7f), and showed higher expression in subnephrotic than nephrotic proteinuria patients (Fig. 7g). Unlike GBP2, EIF2AK2 expression exhibited no significant variation across ISN/RPS LN pathological classifications (Fig. 7h). Neither GBP2 nor EIF2AK2 expression correlated significantly with eGFR or serum creatinine levels (Fig. 7i–l). This suggests their expression changes may not directly influence conventional renal function markers, or that their pathophysiological roles in LN progression involve mechanisms beyond direct impacts on glomerular filtration or creatinine clearance. These findings highlight their potential as biomarkers for specific disease manifestations while emphasizing the complexity of their molecular contributions.

Fig. 7.

Expression Analysis of GBP2 and EIF2AK2 in LN and Their Association with Pyroptosis-Related Gene Signatures in Specific Immune Cell Types. a-c The boxplot illustrates the relationship between GBP2 expression and both the control group and the WHO lupus nephritis pathological classification, in addition to its association with the stage of CKD and proteinuria. d The boxplot illustrates the relationship between GBP2 expression and the ISN/RPS lupus nephritis pathological classification. e-g The boxplot illustrates the relationship between EIF2AK2 expression and both the control group and the lupus nephritis pathological classification, in addition to its association with the stage of CKD and proteinuria. h The boxplot illustrates the relationship between GBP2 expression and the ISN/RPS lupus nephritis pathological classification. i-l The correlation between the expression levels of GBP2 and EIF2AK2 and eGFR, as well as serum creatinine. *P < 0.05; **P < 0.01; ***P < 0.001

Single-Cell Sequencing and Functional Validation Reveal Cell-Type-Specific Expression and Pyroptosis-Regulating Roles of GBP2 and EIF2AK2 in Lupus Nephritis

To investigate the expression profiles of GBP2 and EIF2AK2 in LN, we analyzed two scRNA-seq datasets focusing on renal tissue and immune cell compartments. t-SNE visualization of kidney tissue scRNA-seq data (Fig. 8a) revealed distinct cellular clusters. Across these clusters, GBP2 showed elevated expression in multiple LN-associated cell types compared to controls, while EIF2AK2 exhibited increased expression in most LN cell types except for a notable decrease in stromal cells (Fig. 8b-c). Notably, neither gene demonstrated significant expression changes in renal tubules between the LN and control groups. In kidney immune cells (excluding CE0 epithelial cells, Fig. 8j), both GBP2 and EIF2AK2 were generally upregulated in LN immune subsets compared to controls (Fig. 8k-l). scRNA-seq data analysis across the ISN/RPS LN pathological classifications revealed differential expression patterns: GBP2 was relatively enriched in class III + V and IV LN, whereas EIF2AK2 showed peak expression in class IV LN (Fig. 8d, g). These findings suggest distinct pathological roles for the two genes. Further exploration of interstitial fibrosis and inflammation (Fig. 8e-f, h-i) revealed dynamic expression changes: GBP2 expression was highest under minimal/no inflammation and peaked at the lowest fibrosis stage, indicating its potential involvement in early disease activity rather than progressive fibrosis. EIF2AK2 expression reached maximum levels under moderate inflammation and severe fibrosis, suggesting a later-stage pathogenic role. These data highlight tissue- and cell-type-specific expression patterns of GBP2 and EIF2AK2 in LN, with differential associations to pathological classifications and disease activity stages. Their opposing dynamic expression profiles in fibrosis/inflammation contexts suggest complementary roles in LN pathogenesis.

Fig. 8.

Single-Cell Analysis of GBP2 and EIF2AK2 Expression Dynamics in Lupus Nephritis. a t-SNE plot of single-cell sequencing data from kidney tissues. b-c Expression patterns of GBP2 and EIF2AK2 in different cell types between the control group and LN patients. d-f Expression patterns of GBP2 in the different LN ISN/RPS pathological classifications, Interstitial Inflammation, and Tubular Interstitial Fibrosis. g-i Expression patterns of EIF2AK2 in the different LN ISN/RPS pathological classifications, Interstitial Inflammation, and Tubular Interstitial Fibrosis. j t-SNE plot of single-cell sequencing data from kidney immune cells. k-l Expression patterns of GBP2 and EIF2AK2 in different immune cell types between the control group and LN patients. CE0: Epithelial cells, CD0: Dividing cells, CM0: Inflammatory CD16⁺ macrophages, CM2: Tissue-resident macrophages, CM3: Conventional dendritic cells, CT5b: CD56bright CD16⁻ NK cells, CT0a: Effector memory CD4⁺ T cells, CT3b: TFH-like cells, CT0b: Central memory CD4⁺ T cells, CT5a: Resident memory CD8⁺ T cells, CB2a: Naïve B cells, CT4: GZMK⁺ CD8⁺ T cells, CT1: CD56dim CD16⁺ NK cells, CT3a: Treg cells, CT2: Cytotoxic T Lymphocytes, CM1: Phagocytic CD16⁺ macrophages, CB0: Activated B cells, CB2b: Plasmacytoid dendritic cells, CM4: M2-like CD16⁺ macrophages, CB1: Plasma cells and plasmablasts, CT6: ISG-high CD4⁺ T cells, CB3: ISG-high B cells

scRNA-seq revealed cell-type-specific expression patterns of GBP2 and EIF2AK2 in LN kidneys. Both genes were significantly upregulated in glomerular endothelial cells and most immune cell subsets in the LN group, while no marked changes were observed in renal tubular epithelial cells. Based on these findings, we focused on validating their functional roles in glomerular endothelial cells and macrophages. Using siRNA, we effectively reduced GBP2 and EIF2AK2 expression in both cell types. LPS stimulation of siRNA-treated cells demonstrated attenuated expression of pyroptosis markers NLRP3 and GSDMD-N compared to controls (Fig. 9a, b). This suggests that GBP2 and EIF2AK2 may modulate pyroptosis pathways in these cells. Notably, siRNA-mediated knockdown of either gene in macrophages reciprocally influenced the expression of the other, indicating potential regulatory crosstalk. These findings collectively underscore the complementary roles of GBP2 and EIF2AK2 in LN pathogenesis, highlighting their potential as both biomarkers of disease activity and therapeutic targets for modulating inflammatory cell death pathways.

Fig. 9.

Verifying GBP2 and EIF2AK2 roles in pyroptosis via cell experiments. a In glomerular endothelial cells (via Western Blotting). b In Raw264.7 cells (via Western Blotting). *P < 0.05; **P < 0.01; ***P < 0.001; **** P < 0.0001, ns: no significant difference

Discussion

The development and advancement of LN are notably influenced by pyroptosis [18]. Inflammasomes trigger inflammatory cascades, activating caspase-1 and promoting the release of IL-1β and IL-18 [19]. Under pathological stress, this leads to cell death and caspase-1-dependent pyroptosis [20]. The NLRP3 inflammasome is linked to LN pathogenesis, causing both pyroptosis and apoptosis [21–23]. Studies show that caspase-1/gasdermin D-induced pyroptosis worsens podocyte injury in LN, while its inhibition can slow disease progression [24]. Ursolic acid exhibits nephroprotective effects in MRL/lpr mice by inhibiting SUMO1-mediated NLRP3 sumoylation [25]. Collectively, these findings emphasize the critical role of pyroptosis in LN progression and underscore the therapeutic potential of targeting pyroptotic pathways to alleviate disease severity. In this study, we systematically explored the role of pyroptosis-related genes in LN using an integrative bioinformatics approach. Our results not only provided new insights into the molecular mechanisms underlying LN pathogenesis but also identified potential therapeutic targets and diagnostic biomarkers. The identification of 26 differentially expressed pyroptosis-related genes in glomerular tissues is consistent with recent research on NLRP3 inflammasome activation in LN progression. The development of a 10-gene diagnostic model (AUC > 0.96) demonstrates strong potential for clinical translation. Notably, its robust performance across glomerular, tubulointerstitial, and whole-kidney tissues highlights its broad applicability in heterogeneous LN manifestations.

Our study identified two molecular subtypes of LN through consensus clustering, shedding new light on pyroptosis-related pathogenesis and highlighting heterogeneity in LN mechanisms. Subtype1, characterized by inflammatory and interferon responses, may link inflammation to pyroptosis-related gene activation. Subtype2, driven by KRAS-mediated metabolic pathways, suggests a compensatory response to oxidative stress, reflecting chronic injury processes. This subtyping could inform personalized treatments, such as immune suppression for subtype1 and metabolic regulation for subtype2. Future research should explore associations between these subtypes.

Through WGCNA, hub genes related to pyroptosis in the gray60 module associated with subtype1 were identified, namely GBP2 and EIF2AK2. Guanylate binding proteins (GBPs) belong to the GTPase protein family and are thought to play a crucial role in the host’s innate immune response by mediating resistance to invading pathogens [26]. Previous studies have demonstrated that GBPs can modulate inflammasome activity [27]. For instance, GBP1 is known to activate caspase 4, resulting in the release of the pro-inflammatory cytokine IL18, which subsequently triggers pyroptosis and cell death [28]. Similarly, GBP3 has been shown to exacerbate the progression of LN by suppressing cell proliferation, promoting inflammation, and inducing apoptosis in mouse models of LN [29]. In this study, GBP2 is a marker of macrophage activation. This finding aligns with prior studies indicating that NLRP3 activation in resident macrophages leads to the upregulation of IL-33, which subsequently exacerbates inflammatory responses in renal tubular epithelial cells and plays a pivotal role in the pathogenesis of LN [30]. In LN, GBP2 may be involved in the abnormal activation of the immune system and inflammatory processes, albeit the precise mechanisms underlying its actions remain to be elucidated. EIF2AK, also known as protein kinase R, is a stress-related kinase that serves as a pivotal regulator in governing cellular stress responses, inflammatory reactions, and immune responses [31]. EIF2AK2, which is also an IFN-stimulated gene, demonstrates increased expression in T cells of patients with SLE, selectively modulating immune responses and the transcription of histone genes associated with SLE [32]. Our analysis revealed that GBP2 was significantly upregulated in the WHO class II/III LN and exhibited a non-significant trend toward increased expression in the ISN/RPS class III + V/IV + V subtypes, suggesting its potential role in early immune activation and complex pathological progression. EIF2AK2 showed similar upregulation in the WHO II/III LN but no significant variation across the ISN/RPS subtypes, indicating a broader immune stress response. scRNA-seq data further supported GBP2 enrichment in III + V/IV subtypes and peak EIF2AK2 expression in class IV LN, implying distinct roles in LN immunopathology. Future studies should expand sample sizes and incorporate functional assays to validate these genes as therapeutic targets or biomarkers for LN subtyping and prognosis.

Single-cell RNA sequencing analysis revealed a notable upregulation of GBP2 and EIF2AK2 across various immune cell subsets as well as renal parenchymal cells, including endothelial cells, within LN kidneys. It is well-established that both T and B lymphocytes, along with different subsets of innate immune cells, are key players in the pathogenesis of SLE [33]. The release of HMGB1 can provoke inflammation through the production of proinflammatory cytokines, the recruitment of immune cells via chemotaxis, and the subsequent induction of further pyroptosis in macrophages [34]. Their abnormal activation and functional dysregulation collectively contribute to the complex pathological processes underlying SLE. Experimental evidence supports the direct involvement of GBP2 and EIF2AK2 in pyroptosis, as demonstrated by the upregulation of these genes in RAW 264.7 and MGECs following LPS stimulation. Furthermore, our studies in RAW 264.7 cells and MGECs showed that knocking down GBP2 and EIF2AK2 with siRNA markedly reduced pyroptosis marker expression after LPS stimulation, indicating their involvement in regulating pyroptosis pathways. Taken together, these findings underscore the complementary roles of GBP2 and EIF2AK2 in the pathogenesis of LN. They not only highlight the potential of these genes as biomarkers for disease activity but also suggest their promise as therapeutic targets for regulating inflammatory cell death in SLE. While prior research has linked pyroptosis-related genes like NLRP3 to LN progression [5, 15–17], our research is the first to thoroughly analyze the pyroptosis gene network, unveiling its connections to LN subtypes and identifying new targets (GBP2 and EIF2AK2). This expands our understanding of pyroptosis regulation, including potential interactions with the classical NLRP3 pathway.

Although this research yields valuable insights, several limitations must be acknowledged. First, the heterogeneity of GEO data is a major challenge. Variations in sample collection times, treatment backgrounds, and experimental conditions across datasets can introduce bias, compromising the generalizability and reliability of our findings. This requires cautious result interpretation and consideration of these factors in future studies. Second, the small sample size in analyses integrating clinical-pathological factors (e.g., ISN/RPS LN classes) and gene-clinical indicator correlations (e.g., eGFR, serum creatinine) limited statistical power, potentially due to insufficient samples or complex disease mechanisms, thereby constraining the robustness of diagnostic model applications and clinical relevance validation. Subgroup analysis results should be viewed cautiously and validated in larger, more diverse cohorts to confirm clinical relevance. Third, the scarcity of publicly available scRNA-seq data from LN patients restricts our analysis’s depth and breadth. Larger-scale datasets are urgently needed to validate findings and better understand LN’s cellular and molecular mechanisms. Moreover, the study mainly used in vitro experiments, which may not fully replicate the complex in vivo LN environment. Interactions among cell types, the immune system, and the renal microenvironment in vivo can affect gene-related interventions, so in vitro results may not directly apply clinically, necessitating in vivo studies. Lastly, longitudinal studies are crucial to assess the long-term clinical impact of targeting identified genes. Well-designed studies with adequate follow-up are needed to evaluate their therapeutic potential and safety in real-world contexts.

Conclusion

This work identified GBP2 and EIF2AK2 as key regulators of pyroptosis in LN. A robust 10-gene diagnostic model exhibited high predictive accuracy, underscoring the clinical potential of pyroptosis-related genes as biomarkers. These findings enhance our comprehension of pyroptosis regulation in LN and lay the foundation for targeted interventions aimed at improving renal outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

All authors contributed to the study’s conception and design. Data collection and analysis were performed by M. S. S.M and Q. A performed cell experiments. The first draft of the manuscript was written by M.S and all authors commented on previous versions of the manuscript. R. Z.and Y.Y conceived the project and supervised and coordinated all the work. Y.Y and H.Z rendered financial support. All authors read and approved the final manuscript.

Funding

This work was financially supported by the National Natural Science Foundation of China (Grants 82170701, 82370700, 82370699, and 81974086) and the Young Scientists Fund of the National Natural Science Foundation of China (Grant 82200768).

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Ethics Approval

The current study exclusively utilized pre-existing data sourced from publicly available resources, thus no specific ethical considerations were necessary.

Clinical Trial Number

not applicable.

Consent for Publication

Not applicable.

Competing Interests

The authors declare no competing interests.