The workflow of the study is presented in Fig. 1.

Fig. 1

The flowchart of the study

Cell types, distribution, interactions, and ferroptosis-related genes in AKI mice by single-cell analysis

Single cell analysis of IRI mice model leveraging the dataset from GES139506 identified eight cell types in kidneys of AKI mice, including proximal tubule cells, myeloid cells, epithelium collecting duct cells, macrophages, endothelial cells, fibroblasts, mesenchymal cells, and podocytes. The distribution of these cell types was depicted in t-distributed stochastic neighbor embedding (tSNE) plots and uniform manifold approximation and projection (UMAP) plots by Dimplot function (Fig. 2a). Figure 2b demonstrated the number of cell-to-cell interactions pathways between these eight identified cell types, including 13 interactions between proximal tubule cells and fibroblasts, and 8 interactions between proximal tubule cells and macrophages, and 8 interactions between macrophages and fibroblasts. The intensity of these cell-to cell interactions was illustrated in Fig. 2c.

Fig. 2

Cell types, distribution, interactions, and ferroptosis-related genes in AKI mice by single-cell analysis based on GSE139506. a Annotated tSNE and UMAP plots of all cell types in GSE139506, containing 8 categories. Only samples of AKI were contained in the analysis. b, c, respectively, show the numbers and weights of the crosstalk pathways among all types of cells from GSE139506. d Heatmap of expression levels of 205 ferroptosis-related genes in all types of cells in GSE139506

The heatmap in Fig. 2d presented the gene expression patterns of 205 ferroptosis-related genes across different cell types in AKI mouse kidneys. Genes such as Egr1, Hspa5, and Mtdh were expressed in all 8 types of cells highlighting their potential broad involvement in the ferroptosis process within the kidney. Specifically, proximal tubule cells showed a higher expression of genes such as Lgmn, Miox, Fth1, and Gpx4 suggesting a unique role of these cells in ferroptosis-driven processes. On the other hand, genes such as Ctsb and Cxcl2 had the highest expression in macrophages. Fibroblasts were characterized by enhanced expression of genes such as Pgrmc1 and Epas1. Interestingly, genes including Cxcl2 and Alc38a1 showed exclusive expression in only one or two cell type, pointing out the highly specialized functions of these genes in the context of AKI and ferroptosis.

Heterogeneity of ferroptosis-related genes in proximal tubule cells

Through pseudo-time analysis, we discovered substantial heterogeneity in the expression of ferroptosis-related genes at different stages of proximal tubule cell. This suggested the necessity of conducting more refined clustering based on ferroptosis-related genes. Genes such as Wipi1, Hspa5 and Jun reached peak expression abundance in proximal tubule at early stage. Conversely, genes like Dhodh and Il1b were found to show significantly higher expression abundance at later stages (Fig. 3a). The heatmap generated from the pseudo-time analysis revealed that various ferroptosis-related genes were present at different developmental stages of proximal tubule cells. The outcome of this study emphasized the importance of performing a more in-depth clustering approach to better understand the complex heterogeneity of ferroptosis-related gene expression in proximal tubule cells, which can provide insights into their roles across cellular stages. Moreover, kidney injury markers based on previous studies further verified the authenticity of the pseudo-time analysis. Havcr1 (also known as Kidney Injury Molecule 1, Kim1) is commonly regarded as a sensitive and specific biomarker for kidney injury, exhibiting significantly increasing expression during the damage of renal tubules. It was found that the expression of Havcr1 did progressively increase over time (Supplementary 1a, c) [37]. Similarly, Neutrophil Gelatinase-Associated Lipocalin (lipocalin 2, Lcn2), also known as Ngal, a widely recognized biomarker for kidney injury, demonstrated a parallel increasing trend with Havcr1 (Supplementary 1b, c) [38]. Building on this foundation, we utilized NMF analysis and defined 4 clusters of proximal tubule cells. As detailed in the Methods section, 4 cell groups were distinguished by NMF, where cluster 1 was characterized by Egr1, cluster 2 was characterized by Tfrc, cluster 3 was characterized by Jun and cluster 4 has no characterized gene. Therefore, Egr1 + PTC-C1, Tfrc + PTC-C2, Jun + PTC-C3 and non-ferroptosis PTC-C4, were defined (Fig. 3b). This nuanced classification through NMF analysis allowed for a more refined understanding of the diversity within proximal tubule cells, especially in the context of ferroptosis.

Fig. 3

Proximal tubule cells show heterogeneity in the process and outcome of ferroptosis. a Pseudo-time analysis of the ferroptosis-related genes in proximal tubule cells. Genes present within the pink module are predominantly expressed during the early stages of proximal tubule cell development, whereas genes within the earthy yellow module show higher expression levels during the later stages of proximal tubule cell development. b NMF analysis based on ferroptosis-related genes distinguished proximal renal tubular cells into 4 clusters. c, d Show the weights of the crosstalk pathways among all clusters of proximal tubule cells and fibroblasts, as well as macrophages. e Significant pathways obtained by GO enrichment analysis of the highly expressed genes in the proximal tubule cells of each cluster. f The heatmap of the transcription factors activity of each proximal tubule cell cluster obtained by SENIC analysis

Using the analysis of the Cellchat package, we found out that Egr1 + PTC-C1 interacted significantly more pronounced with fibroblasts and macrophages than all the remaining 3 proximal tubule cell types (Fig. 3c, d). Further exploration through GO enrichment analysis showed different pathway activities within the 4 clusters of proximal tubule cells. Egr1 + PTC-C1 cells displayed increased activities in the structural constituent of the ribosome, ubiquitin ligase inhibitor activity as well as rRNA binding pathways, which might indicate that protein synthesis and metabolism were at a higher level in this cluster of cells, while ATP-dependent chromatin remodeler activity, R-SMAD binding, and SMAD binding pathways were increased in Jun + PTC-C3 cells, reflecting the active histone modification in this cluster of cells. Sodium independent organic anion transmembrane transporter activity, organic anion transmembrane transporter activity and lipid transporter activity in Tfrc + PTC-C2 were notably high. Meanwhile, organic acid transmembrane transporter activity, carboxylic acid transmembrane transporter activity and C-cyltransferase activity showed significant enrichment in non-ferroptosis PTC-C4 (Fig. 3e).

The heatmap of SENIC analysis reveals that the transcription factors in the Egr1 + PTC-C1 and Jun + PTC-C3 clusters were more active than in the other 2 clusters, including Tfrc + PTC-C2 and non-ferroptosis-PTC-C4. Egr1, Irf7, Stat2 and Stat1 were significantly more active in Egr1 + PTC-C1, while the Jun and Jund transcription factor binding sites were significantly more active in Jun + PTC-C3 (Fig. 3f).

Recent scholarly articles have identified Egr1 + PTC-C1, Jun + PTC-C3, and non-ferroptosis PTC-C4 as novel subgroups, implying that these profiles provide latest information about the kidney's cellular landscape (Supplementary 1d). The Tfrc + PTC-C2 cluster, on the other hand, might be like healthy proximal tubule cells previously characterized in the literature, indicating a baseline or reference state for comparison [39].

Different metabolic and chemotactic orientations in macrophages during ferroptosis

The pseudo-time analysis heatmap revealed that different ferroptosis-related genes were expressed at different cellular stages of macrophages, suggesting the need for a deeper clustering of the heterogeneity of ferroptosis-related genes in macrophages (Fig. 4a). Expression of genes, including Ctsb, Hmox1, Park7, Epas1 and so on, were highest in early status, while expression of genes such as Il1b, Kdm6b, Bach1 and Jun peaked in late ones.

Fig. 4

Macrophages exhibit different metabolic and chemotactic orientations during ferroptosis. a Pseudo-time analysis of the ferroptosis-related genes in macrophages. The genes within the red module are more expressed in the early stage of macrophage development, while the genes of the earthy yellow and blue modules are more expressed in the late stage of macrophage development. b NMF analysis based on ferroptosis-related genes distinguished macrophages into 4 clusters. c, d show the weights of the crosstalk pathways among all clusters of macrophages and proximal tubule cells, as well as fibroblasts. e Significant pathways obtained by GO enrichment analysis of the highly expressed genes in the macrophages of each cluster. f Violin plots of inflammatory and immune-related genes expressed brightly in each cluster of macrophages

We used NMF clustering analysis to categorize AKI mice kidney macrophages, we successfully differentiated six distinct clusters. These clusters were Cxcl2 + Mac-C1, Ctsb + Mac-C2, Egr1 + Mac-C3, Il1b + Mac-C4, Miox + Mac-C5, and non-ferroptosis-Mac-C6 (Fig. 4b). The NMF analysis facilitated the segregation of six macrophage clusters in AKI mice kidneys, each of them defined by distinct gene expression patterns. Cluster 1 exhibited high expression of Cxcl2, cluster 2 was characterized by elevated levels of Ctsb, cluster 3 demonstrated significant expression of Egr1, cluster 4 displayed notable expression of Il1b, cluster 5 showed elevated expression of Miox, whereas cluster 6 did not exhibit any specific gene characterization. Cellchat analysis showed that Cxcl2 + Mac-C1 and Ctsb + Mac-C2 were the most communicative, showing the highest levels of interactions with each of the other clusters of macrophages (Fig. 4c, d). Specifically, we have found out that Ctsb + Mac-C2 had the highest Interaction weight with proximal tubule cells compared to the other 5 clusters of macrophages (Fig. 4c). Interactions between Cxcl2 + Mac-C1 and Egr1 + Mac-C3 were the most potent with fibroblasts (Fig. 4d).

GO enrichment analysis showed that lipid metabolism-related pathways, including integrin binding, lipoprotein particle binding and protein lipid complex binding, were prominently more active in Ctsb + Mac-C2. This suggested a significant role in lipid processing and metabolism within this cluster. In contrast, the pathways of chemokine activity, chemokine receptor binding and cytokine activity were more active in Cxcl2 + Mac C1, hence indicating an elevated capacity for initiating and propagating inflammatory response and chemotaxis. Additionally, pathways related to peptide binding, MHC class II protein complex binding and MHC protein complex binding were enhanced in Egr1 + Mac-C3. This enhancement suggested a pivotal role in regulating T cell-mediated immune responses, potentially influencing the overall immune landscape within the AKI context. In Miox + Mac-C5, we found that the symporter activity and secondary active transmembrane transporter activity pathways were significantly more enriched, suggesting that the cluster of cells was exhibiting a higher level of activity in the transmembrane transport of substances, indicating an enhanced capacity for moving materials across cell membranes. Moreover, immune receptor activity was observed to be elevated, indicating an enhanced responsiveness of the immune system in Il1b + Mac − C4, while transferrin receptor binding, ferric iron binding and acetylcholine receptor inhibitor activity were enhanced in non-ferroptosis-Mac-C6 (Fig. 4e).

Violin plot analysis of the inflammatory and immune-related genes revealed that the expression of inflammatory response-related genes Il1b, Tnf, Cxcl2, Cxcl3 and Ccrl2 were highest in Cxcl2 + Mac-C1. An analysis that further supported the conclusion that this cluster of cells exhibited the most potent inflammatory response and chemotactic activity among the six identified macrophage clusters (Fig. 4f).

Ferroptosis-related genes distinguished the heterogeneity of fibroblasts

Pseudo-time analysis of ferroptosis-related gene expression in fibroblasts revealed that fibroblast also displayed time-dependent heterogeneity (Fig. 5a). Dpp4 had the highest expression in early fibroblast status, while Gpx4, Scp2 and Egr1 spiked in late fibroblast status. The pseudo-time analysis heatmap, revealed that different fibroblasts exhibit the presence of various ferroptosis-related genes at different stages. This observation highlighted the importance of conducting a more comprehensive clustering analysis paving the way for a better understanding of the heterogeneity of ferroptosis-related genes in fibroblasts.

Fig. 5

Ferroptosis-related genes distinguished the heterogeneity of fibroblasts. a Pseudo-time analysis of the ferroptosis-related genes in fibroblasts. The genes within the earth-yellow module are expressed more in the initial stages of fibroblast development, while the red genes are expressed more in the late stages of fibroblast development. b NMF analysis based on ferroptosis-related genes distinguished macrophages into 4 clusters. c, d show the weights of the crosstalk pathways among all clusters of fibroblasts and macrophages, as well as proximal tubule cells. e Assessment of the immune microenvironment of each cluster of fibroblasts. f Heatmap of CAF signature score for each cluster of fibroblasts

NMF analysis was employed to cluster fibroblasts into four distinct groups, namely Egr1 + Fib-C1, Scp2 + Fib-C2, Miox + Fib-C3, and Gpx4 + Fib-C4 (Fig. 5b). We used NMF clustering analysis to categorize AKI mice kidney macrophages into these four distinctive clusters: Egr1 + Fib-C1, Scp2 + Fib-C2, Miox + Fib-C3, and Gpx4 + Fib-C4. By analyzing specific gene expression profiles, the NMF analysis facilitated the differentiation and classification of these four cell groups, providing a clear delineation based on their unique molecular characteristics. Cluster 1 showed an elevated expression of Egr1, cluster 2 was characterized by increased levels of Scp2, cluster 3 demonstrated significant expression of Miox, and cluster 4 displayed notable expression of Gpx4. CellChat analysis showed that Egr1 + Fib-C1 had the strongest intercellular interactions with both proximal tubule cells and macrophages (Fig. 5c, d).

Heatmap analysis detailing the expression of extracellular matrix metabolism-related genes showed that Egr1 + Fib-C1 had the highest expression of Bgn, Mmp9 and Cxcl14. Additionally, Scp2 + Fib-C2 showed significantly higher expressions of Dcn, Ragln, Mmp3, and Pln, whereas Miox + Fib-C3 showed elevated levels of Pdgfa, Cxcl12, Ccl2, and Il7, indicating distinct functional emphases within these fibroblast subsets. While the expression abundance of Serpine1, Tgfb1, Myh11, Rasl12, Rasgrp2 and C3 was higher in Gpx4 + Fib-C4 as presented in Fig. 5e.

Correlation analysis with tumor-associated fibroblasts showed that Egr1 + FibC1 and Scp2 + Fib-C2 were highly correlated with pan-iCAF-2 and pan-pCAF, respectively (Fig. 5f). The analysis revealed that Egr1 + Fib-C1 showed a negative correlation with pan-pCAF and pan-iCAF. Among them, Pan-pCAF was associated with cell proliferation, while Pan-iCAF was thought to control the transcriptional program associated with inflammation [40]. Some studies have reported the association of pan-iCAF-2 with extracellular matrix remodeling, while pan-pCAF was associated with regulation of cell proliferation [40].

According to the most recent studies, the four identified fibroblast subgroups had characteristics like previously reported mesangial cells (Supplementary 1e). This finding implied that mesangial cells played a critical role in the onset of ferroptosis during acute kidney injury, emphasizing their importance in the pathological process [39].

Validation of heterogeneity of ferroptosis-related genes in AKI at the RNA-seq level

Analysis of GSE34351 demonstrated that the collective expression of 205 ferroptosis-related genes was notably elevated in the kidneys of the AKI mouse model in comparison with the control samples. We extracted the characteristic genes of each cluster of cells and analyzed the infiltration in the kidneys of IRI mice model and controls (Fig. 6b). A total of 8 clusters of cells, including Egr1 + PTC-C1, Tfrc + PTC-C2, Jun + PTC-C3, Cxcl2 + Mac-C1, Ctsb + Mac-C2, l1b + Mac-C4, Egr1 + Fib-C1, and Gpx4 + Fib-C4, had significant differences in the abundance of their signature genes infiltrating the kidneys between AKI and control mice.

Fig. 6

Transcriptomic validation of study results by GSE34351 and in vivo models. a Expression levels of ferroptosis-related genes in control (n = 3) and IR model mice (n = 3). b Infiltration abundance of characteristic genes of each NMF typing cell cluster in control and IRI mice model. c Volcano maps of DEG analysis between control and IRI mice model with annotation of cell cluster marker genes with significantly different abundance of characteristic gene infiltration. d PCR results of Egr1 transcript levels in an in vivo model (n = 3 vs n = 3). e PCR results of Jun transcript levels in an in vivo model (n = 3 vs n = 3). f PCR results of Cxcl2 transcript levels in an in vivo model (n = 3 vs n = 3). g Colocalization of AQP1 and Egr1 confirming the presence of Egr1 + PTC-C1. h Colocalization of AQP1 and Jun confirming the presence of Jun + PTC-C3. i Colocalization of CD68 and Cxcl2 confirming the presence of Jun + PTC-Cxcl2 + Mac-C1. j Colocalization of FAP and Egr1 confirming the presence of Egr1 + Fib-C1

The expression of the 7 signature genes across the 8 clusters of cells, namely Egr1, Tfrc, Jun, Cxcl2, Ctsb, l1b, and Gpx4, were validated in the kidneys of IRI mice and controls, and revealed that Egr1, Jun and Cxcl2 expression was significantly higher in AKI mice (Fig. 6c). RT-qPCR of Egr1, Jun and Cxcl2 from kidney revealed that all the 3 genes were significantly elevated in IRI mice compared to controls (Fig. 6d–f).

Multiplex immunofluorescence staining on renal pathology biopsy samples from human AKI patients was performed. This staining confirmed the presence of Egr1 + PTC-C1, Jun + PTC-C3, Cxcl2 + Mac-C1, and Egr1 + Fib-C1 cells during the occurrence of AKI. This technique demonstrated fluorescence co-localization, providing clear visual evidence of these specific cell types during AKI episodes. We used AQP1 immunofluorescence to localize proximal tubule cells and observed their co-staining with Egr1 and Jun, confirming the presence of Egr1 + PTC-C1 (Fig. 6g) and Jun + PTC-C3 (Fig. 6h). Similarly, we used CD68 to locate macrophages and observed their fluorescence colocalization with Cxcl2, confirming the presence of Cxcl2 + Mac-C1 (Fig. 6i). Likewise, we used FAP to locate fibroblasts and observed their co-staining with Egr1, confirming the presence of Egr1 + Fib-C1 (Fig. 6j).