Intercellular Communication Network of CellChat Uncovers Mechanisms of Kidney Fibrosis Based on Single-Cell RNA Sequencing

The incidence of chronic kidney disease (CKD) is on the rise globally, largely due to the aging population and the growing prevalence of diabetes, obesity, and hypertension [1, 2]. End-stage CKD, characterized by high mortality and morbidity, often requires renal replacement therapy [3]. In recent decades, blood pressure regulation and renin-angiotensin system inhibition have been the cornerstones of CKD treatment. Consequently, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers have become widely used due to their renal protective effects, which include vasodilation, reduction of intra-glomerular pressure, and decreased proteinuria [4]. Despite these advancements, renal fibrosis remains a chronic and progressive process that is the most common pathological feature of CKD. Currently, there are limited treatment options available for renal fibrosis and for delaying the loss of kidney function.

Sodium-glucose co-transporter-2 (SGLT2) inhibitors are a new class of antihyperglycemic drugs, blocking the activity of SGLT2 in the proximal renal tubules, thereby increasing urinary glucose excretion and reducing hyperglycemia. These drugs treat diabetes by increasing urinary glucose excretion and decreasing hyperglycemia by inhibiting glucose reabsorption through the kidneys [5]. Large, multicenter randomized trials have demonstrated that SGLT2 inhibitors can decrease the incidence of worsening nephropathy and lower the risk of cardiovascular mortality, heart failure hospitalizations, and all-cause death in patients with diabetic kidney disease (DKD) [6, 7]. These drugs have been shown to provide nephroprotection for DKD by modulating the renin-angiotensin-aldosterone system, altering energy utilization, and promoting anti-inflammatory, antioxidative, and immunomodulatory responses [8‒12]. Studies have also shown that SGLT2 inhibitors, such as empagliflozin (Empa), can effectively mitigate renal fibrosis and enhance renal function in DKD mice by regulating renal metabolic reprogramming [13] and alternatively activating macrophages in rats with 5/6 nephrectomy (5/6 Nx) [14]. Gene regulation analysis can be performed unbiasedly and reliably using single-cell RNA sequencing (scRNA-seq) across numerous cells at a single-cell resolution [15]. scRNA-seq can help identify new kidney cell types [16], elucidate the biological functions of the kidney [17], investigate cellular differentiation pathways, and shed light on intercellular communications [18] by profiling the transcriptome of a single kidney cell. This technology enables researchers to investigate the underlying mechanisms and pathogenesis of kidney disease, paving the way for more effective treatments [19, 20].

While the etiology of fibrosis in nondiabetic CKD has been thoroughly investigated, few studies have explored the intercellular communication networks involved in renal fibrosis in nondiabetic CKD [21]. In the present study, a nondiabetic CKD model with 5/6 Nx was developed, and scRNA-seq was used to elucidate the intercellular communication network of ligand-receptor pairs and signaling pathways involved in the fibrosis process in CKD. The study also investigated the benefits of SGLT2 inhibitors in treating nondiabetic CKD.

A total of 22 male Wistar rats, aged 6–7 weeks and weighing 200–250 g, were obtained from the Nanjing University Experimental Animal Center. Following a 2-week adaptive feeding period, the rats (n = 22) were divided into three groups: placebo-treated sham surgery rats (sham group, n = 7), placebo-treated 5/6 Nx rats (5/6 Nx group, n = 6), and Empa-treated 5/6 Nx rats (Empa group, n = 9). The rats were subjected to the following 5/6 Nx procedure [22]: we excised the superior and inferior poles of the left kidney in week 2, and the right kidney was removed in week 4. This model replicated CKD conditions by removing approximately 5/6 of the renal mass while preserving a minute portion of functional parenchyma. During the same time, sham laparotomies were performed. While producing 5/6 Nx, all rats were anesthetized with an intraperitoneal injection of 3, 3, 3-tribromoethanol at 300–400 mg/kg. The rats in the Empa group were given 15 mg/kg of Empa orally through gavage from week 10 to week 16 [14]. Similarly, rats in the sham and 5/6 Nx groups were given 15 mg/kg of sterile water. All rats were housed in a controlled SPF-grade environment with temperatures ranging from 20°C to 24°C and humidity levels of 45% to 65%. A 12-h light/dark cycle was maintained throughout the study. There were no restrictions on their diet or water intake.

Urine samples were collected at 24 h, while blood samples were taken at weeks 1 (baseline), 4, 10, and 16. At the end of the 16th week, the rats were anesthetized with an intraperitoneal injection of 2, 2, 2-tribromoethanol, and blood samples were collected from the inferior vena cava to induce hemorrhagic shock and fatality. Subsequently, the blood was centrifuged for 10 min at 3,000 rpm and at 4°C. Upper serum was collected and quickly frozen in liquid nitrogen before being stored at −80°C. The kidneys were weighed and cut in half longitudinally; one half was preserved in 4% paraformaldehyde, embedded in paraffin for histology, and the other half for scRNA-seq. We centrifuged the urine for 10 min at 2,000 rpm and stored the plasma and supernatant at −80°C for later experiments.

This study utilized the GEXSCOPE^® Single Cell RNA Library Kit TissueV2 (Cat# 5180011) purchased from Singleron Biotechnologies to perform single-cell isolation and sequencing library preparation. The integrated workflow system comprises the following key components: tissue preservation solution for sample stabilization, enzymatic tissue dissociation buffer, a microfluidic microwell chip for single-cell capture, molecularly barcoded magnetic beads for single-cell identification, along with optimized amplification and library construction reagents. This fully contained platform supports end-to-end processing from tissue specimen preservation through complete single-cell transcriptomic library generation, ensuring experimental consistency while minimizing technical variability across samples.

Fresh kidney tissue samples were gently washed with phosphate-buffered saline and minced into approximately 1 mm³ pieces. The minced tissue was transferred into centrifuge tubes containing tissue dissociation buffer. The tubes were then placed in a 37°C constant temperature shaker set at 180 rpm. The samples were monitored every 15 min until complete digestion of the tissue was achieved, with a total dissociation time ranging from 30 to 60 min. The resulting cell suspension was filtered through a 70-µm cell strainer to remove undigested tissue fragments and cell clumps. The filtered cell suspension was centrifuged at 300 × g for 5 min, and the supernatant was discarded. The cell pellet was resuspended in an appropriate volume of phosphate-buffered saline to prepare a single-cell suspension. Cell viability was assessed using trypan blue staining, ensuring that viability exceeded 70%. The cell concentration was adjusted to 1 × 10⁶–5 × 10⁶ cells/mL using a hemocytometer.

The SCOPE-chip was prepared by washing with the appropriate buffers and loading barcode beads into the chip, ensuring that each microwell contained a single barcode bead. The prepared single-cell suspension was then loaded into the sample inlet of the SCOPE-chip, and single cells were captured into the microwells. The capture process lasted for 10–15 min. Microscopic observation of the chip was performed to confirm that each microwell contained a single cell, with a capture efficiency of over 70%.

The single-cell suspension containing captured cells was subjected to multiple brief centrifugations to ensure cell integrity. Subsequently, lysis buffer and reverse transcriptase were added to lyse the cells and initiate the reverse transcription reaction, converting RNA into cDNA. The reaction conditions were set at 42°C with a centrifugation speed of 1,300 rpm for 90 min. The resulting cDNA was amplified using PCR with amplification master mix and amplification enzyme. The amplified cDNA was fragmented using fragmentation buffer and fragmentation enzyme mix. The fragmented cDNA was processed using end repair and adapter ligation reagents, including ligation mix, ligation booster, and adaptors. The ligated cDNA was further amplified using library amplification mix (Library Amp Mix V2) and indexing primer mix. The quality of the sequencing library was assessed using Agilent 4150 TapeStation System to evaluate the fragment size distribution and concentration of the library.

Each library was diluted to 4 nm and combined with 150 bp paired-end reads for Illumina HiSeq × sequencing. featureCounts was used to quantify the genes after the Rawraw fastq files were aligned to the Ensembl Rnor_6.0 reference genome. In addition, the Seurat R package (version 3.1.2) performs the data quality assessment, initial processing, and dimensionality reduction. We screened the cells expressing <200 or >3,000 genes, with Unique Molecular Identifiers <30,000, and mitochondrial gene expression >50% to obtain high-quality matrices. After normalizing the data with robust point matching, 2,000 genes with high variability were selected for principal component analysis. The top 20 dimensions were then visualized and clustered. A total of 14 distinct cell clusters were identified in the cell clustering process using the FindClusters function with a resolution of 0.8. The heterogeneity of T-cell clusters and myeloid cells was determined by sub-clustering T cells and myeloid cells with a resolution of 1.2. We identified genes with differential expression between cell clusters using the FindMarkers function and the Wilcoxon rank-sum test with adjusted Benjamini-Hochberg p value < 0.05 as default parameters.

We used the CellChat 1.0.0 algorithm to analyze and quantify cell-cell communication at the single-cell level [23]. CellChat calculated the probability of ligand-receptor interaction to assess the communication incidence of the signaling pathway. p value <0.05 indicates a significant interaction. The number of cell-cell interactions and communication strength were then depicted using chord and circle diagrams.

Signaling pathways were classified based on structural and functional similarities to investigate their characteristics further. The Euclidean distance was used to measure the distance between signaling pathways to develop an impression of signaling pathway adaptation among groups. The information flow value in each signaling pathway was compared across groups, and the synergistic mode of cell populations and signaling pathways was illustrated. Hierarchical diagrams were used to examine the network structure formed by the cellular signal senders and receptors in target signaling pathways. The specific roles of cell populations in signaling pathways, such as senders, receivers, mediators, and regulators, were illustrated by heatmaps. Moreover, we used the online database the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING, http://string-db.org) to predict the function of the target gene.

Statistical analysis and visualization of data were conducted using GraphPad Prism 9.5 and SPSS 26.0 software. One-way ANOVA was employed for the analysis of laboratory indicators with normally distributed data, while the Kruskal-Wallis test was utilized for data that did not follow a normal distribution. Differential gene expression analysis was performed using the Wilcoxon rank-sum test, and the false discovery rate method was applied to correct for multiple testing errors.

The study flowchart is depicted in Figure 1a. 5/6 Nx significantly elevated blood urea nitrogen, serum creatinine, and urinary albumin-to-creatinine ratio (p < 0.01) in rats, which was mitigated by Empa treatment (p < 0.05) (Fig. 1b–d). The 5/6 Nx group exhibited increased renal interstitial fibrosis, and this was also improved by Empa treatment (p < 0.05) (Fig. 1e, f).

Three kidney samples were randomly selected from each group for scRNA-seq atlas generation. After filtering, normalizing, and reducing the dimensionality of the data, 14 cell clusters were grouped from 89,420 cells. Stacked bar graphs depict the differences in the proportion of various cell subsets between groups (Fig. 2a; online suppl. Table S1; for all online suppl. material, see https://doi.org/10.1159/000545209). We classified the cells into clusters based on their marker genes: endothelial cells (ECs; markers: Pecam1, Cdh5, Kdr), mesangial cells (MCs; markers: Fh12, Itga8), pericytes (markers: Rgs5, Abcc9, Kcnj8), proximal tubule cells (PTCs; markers: Lrp2, Slc4a1, Cubn, Havcr1), loop of Henle (LOH) cells (markers: Umod, Slc12a1, Cldn16, Bst1, Aqp1, Spp1, Id1, Cryab), distal tubule cells (markers: Slc8a1, Slc12a3, Trpm6, Calb1), collecting duct principal cells (PCs; markers: Aqp2, Aqp3), collecting duct intercalated cells (markers: Slc4a1, Atp6v0d2, Foxi1), ureter transitional epithelial cells (UT_epithelial; markers: Upk3a, Upk1a, Upk1b), smooth muscle cells (markers: Acta2, Tagln, Mylk, Myh11), fibrocytes (Fib; markers: Dcn, Colla2, Colla1), proliferating cells (markers: Mki67, Top2a, Epcam), myeloid cells (markers: Lyz2, Cd68, Mrc1, Cd209a, Clec9a, Irf8, Xcr1), and T cells (markers: Trbc1, Cd3g, Cd3d, Cd2). We also subclustered the T cells into four subpopulations, including regulatory T cells (markers: Cd3d, Cd4, Tnfrsf9, Tnfrsf4), IL-17-producing helper T cells (markers: Il17a, Ccr6, Cxcr6), follicular helper T cells (markers: Cd40lg, Icos, Cxcr6), and CD8⁺ effector T cells (markers: Cd3d, Cd8a/b, Nkg7, Gzma, Gnly). Myeloid cells were further divided into 6 subpopulations: monocytes (markers: Lyz2, Sell, Mnda), M1 macrophages (markers: Lyz2, Il1b, Tnf), M2 macrophages (M2 macro; markers: Lyz2, Mrc1, C1qc, C1qb, Cd68, Ccl22), and 3 dendritic cell subsets – conventional type 1 dendritic cells (markers: Lyz2, Xcr1, Clec9a, Cadm1), plasmacytoid dendritic cells (markers: Siglech, Ccr9, Tlr7, Gzmb), and migratory dendritic cells (markers: Lyz2, Cd83, Ccr7). Further details are provided in our previous study [14].

CellChat 1.0.0 was then used to infer intercellular communication networks systematically. Intercellular analysis revealed that the incoming and outgoing interaction strength was significantly higher in the 5/6 Nx rats than in the sham group (Fig. 2b). Our findings also revealed a significant increase in ligand-receptor pairs involved in cell-cell communication in 5/6 Nx rats. Moreover, we observed that Empa inhibited intercellular communication (Fig. 2c). Based on the above findings, we further investigated the incoming communication patterns of target cells and the outgoing communication patterns of secreting cells in 5/6 Nx rats. There are 5 distinct patterns of outgoing signaling and 5 distinct patterns of incoming signaling. Notably, endothelial and MC communication patterns are significantly similar. Particularly, pattern 4 is shared by both types of cells for incoming and outgoing signaling (Fig. 2d).

Because of their structural and functional similarities, all signaling pathways were represented on 2-dimensional images, identifying four functional and four structural groups. Our findings revealed that different states of each signaling pathway were altered. For instance, MIF, LAMININ, and COLLAGEN had structural similarities in the sham group. They were assigned to separate groups in the 5/6 Nx group, indicating that their architecture of cell-cell communication changed in CKD (Fig. 3a). In addition, we compared the information flow for each signaling pathway that reflects the cumulative communication probabilities between cell populations in the inferred network. The 5/6 Nx group had significant enrichment of multiple signaling pathways compared to the control group, such as fibrotic signaling pathways COLLAGEN, FN1, THBS, and LAMININ (Fig. 3b). This indicates that CKD effectively triggers fibrotic responses. The diversity was determined by measuring the Euclidean distance between each pair of shared signaling pathways. The 5/6 Nx group demonstrated a more pronounced MIF pathway remodeling than the control group.

Moreover, Empa treatment resulted in GRN pathway reconstruction (Fig. 3c; online suppl. Table S2). For the functional network association analysis, we focused on the ligand-receptor genes that were important in the pathways above (Fig. 3d; online suppl. Fig. 1–4). Figure 3e illustrates the predicted interactions among a series of proteins potentially associated with renal fibrosis. Each node represents an individual protein, and the edges connecting the nodes signify predicted interactions, with different colors indicating the type of evidence supporting these interactions. We observed significant associations among the ligand-receptor genes presented in Figure 3d. For instance, SDC4 exhibits database and co-expression evidence links with ITGA8 and also shows co-expression evidence association with CXCR4. Both FN1 and ITGB1 have database evidence connections with ITGA8, while LAMB1 demonstrates experimental and database evidence links with both type IV alpha 1 collagen (COL4A1) and ITGA1. The STRING interaction network map reveals the intricate web of interactions among proteins related to renal fibrosis. In addition, we performed functional annotations of the genes using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (www.genome.jp/kegg). These genes were enriched in various functional categories, including extracellular matrix (ECM)-receptor interaction, cell-matrix proliferation, transforming growth factor-β (TGF-β) pathway, cell adhesion, and focal adhesion (Fig. 3f; online suppl. Table S3).

We first focused on the TGF-β pathway because of its important role in renal fibrosis [24]. Figure 4a presents chord diagram that illustrates the interaction strengths between macrophages and other cell types. Compared to the sham group, the 5/6 Nx group exhibited significantly higher interaction strengths. Specifically, M2 macro had the highest sender, receiver, and mediator scores within the 5/6 Nx group, highlighting their pivotal role in the communication network of the TGF-β signaling pathway in CKD. Furthermore, Figure 4a demonstrates that the Empa therapy significantly reduced these interactions. Moreover, we used a hierarchical diagram to illustrate the intercellular communication mediated by each cell type and their roles in the TGF-β pathway. Cross-referencing source and target signaling patterns revealed information about the autocrine and paracrine pathways of specific cell types. ECs were the primary sources of the TGF-β pathway in the sham group. In contrast, M2 macro were identified as the primary source and target of the TGF-β pathway in the 5/6 Nx group. Furthermore, M2 macro were autocrine and paracrine to ECs, MCs, and Fib. Figure 4b demonstrates that the Empa group has inhibited autocrine and paracrine interactions between M2 macro and other cells. To investigate the potential correlation between cell communication intensity and gene expression, alterations in the ligand and receptor genes in the TGF-β pathway were analyzed across 3 groups. The ligands of the TGF-β pathway detected in this study include transforming growth factor-β1 (TGFB1), transforming growth factor-β2 (TGFB2), and transforming growth factor-β3 (TGFB3). The TGF-β receptors include transforming growth factor-β receptor 1 (TGFBR1) and transforming growth factor-β receptor 2 (TGFBR2). Among all the ligand-receptor pairs, the TGFB1-(TGFBR1 + TGFBR2) pair contributed the most to intercellular communication, followed by the TGFB3-(TGFBR1 + TGFBR2) pair (Fig. 4c). In the 5/6 Nx group, the expression of TGFB1 in Fib, TGFB2 in Fib and MCs, and TGFBR2 in ECs was upregulated.

In contrast, the expression of these genes was downregulated in the Empa group. Moreover, ligand and receptor gene expression levels in other cell types, such as M2 macro, were noticeably low in all three groups (Fig. 4d; online suppl. Tables S5–7). The Empa group had higher TGFB3 gene expression than the 5/6 Nx group, but this difference was not statistically significant (online suppl. Table S7). Although there is a strong correlation between cells in the TGF-β pathway in the 5/6 Nx group, it is important to note that the communication strength of each cell is not always positively correlated with gene expression levels. Given the preceding, we hypothesize that Empa may reduce intercellular communication by inhibiting gene expression. However, it is essential to recognize that other factors may influence the strength of intercellular communication. The findings shed light on the intricate intercellular communication mechanisms of the TGF-β pathway.

During our investigation of additional pathways involving M2 macro in the 5/6 Nx group, we observed that M2 macro communicate with tubular cells via the MIF and GRN pathways (Fig. 5a, b). The MIF pathway in the sham group primarily depends on PTCs and the LOH as signal transmitters, with collecting duct PCs serving as primary receptors. However, M2 macro emerged as the primary signal recipients in the 5/6 Nx group (Fig. 5a), receiving signals from tubular cells via paracrine signaling (online suppl. Fig. 5A). A significant feature of CKD progression was the observed transition in receptor contribution from ACKR3 to CD74 and CXCR4 (Fig. 5a). MIF expression was observed ubiquitously across multiple cell types, indicating its presence in the kidney.

In contrast, ACKR3 was primarily expressed in PCs, whereas CD74 was predominantly expressed in M2 macro. Lower expression of CXCR4 and CD44 supported ACKR3 and CD74 as key receptor genes in the MIF pathway (Fig. 5c). Despite the lack of significant changes in gene expression in the Empa treatment group, we identified a significant difference in intercellular communication strength. The chord and heatmap diagrams revealed decreased signal reception from M2 macro, with Empa PCs resuming their role as signal receivers. Notably, the most significant ligand-receptor pair returned to MIF-ACKR3 in the Empa group, indicating a return to normalcy (Fig. 5a). The GRN pathway is primarily involved in each of the three groups by the ligand-receptor pair GRN-SORT1. The chord diagram demonstrated that all cell types engaged in extensive intercellular communication.

Moreover, heatmap analysis revealed alterations in GRN signaling pathway communication in CKD. Under sham conditions, tubular cells were identified as the primary signal transmitters. However, there is a significant shift in CKD, where M2 macro assumed the role of primary signal transmitters (Fig. 5b). The hierarchy diagram reveals that M2 macro significantly increased the paracrine effects of collecting duct intercalated cells and PCs. However, Empa treatment significantly reduced this effect (online suppl. Fig. 5B). The violin plot depicted GRN expression in all cells, revealing significantly higher levels in M2 macro of the 5/6 Nx group (Fig. 5d). Empa effectively downregulated GRN gene expression (p < 0.05, online suppl. Table S4). These findings indicate that Empa has the potential to inhibit the interaction between M2 macro and other cells by weakening signal transmission in the MIF pathway and inhibiting gene expression in the GRN pathway.

The pro-fibrotic signaling pathways of COLLAGEN, FN1, LAMININ, and THBS were significantly altered in CKD, compared to the sham group. Fib and MC signal output in these pathways was significantly higher in the 5/6 Nx group, with Fib having the highest output. Figure 6a depicts EC-tubular cell communication in the COLLAGEN and LAMININ pathways. The heatmap analysis revealed that tubular cells were the primary targets of signals in all four pro-fibrotic pathways, demonstrating the significant impact of the fibrotic process on tubular cells (Fig. 6b).

Our findings revealed a significant upregulation of ligand genes, such as COL4A1 and COL4A2 in the COLLAGEN pathway, FN1 in the FN1 pathway, LAMB1 in the LAMININ pathway, and thrombospondin-1 (THBS1) and THBS2 in the THBS pathway. These upregulations were particularly evident in the Fib and MCs of the 5/6 Nx group (Fig. 6c). In ECs, the circle plot illustrated that the COL4A1 and COL4A2 expression levels in the COLLAGEN pathway, and LAMB2 in the LAMININ pathway, were associated with their communication strength (Fig. 6b). In addition, the primary receptor genes involved in the four pathways (ITGA1, ITGB1, SDC4, and CD47) had significant expression levels in tubular cells, supporting the findings of the heatmap that tubular cells are the primary signal recipients. Although the receptor gene expression was higher in Fib and MCs (Fig. 6b), the heatmap did not identify them as the primary signal recipients. These findings suggest that ligand genes facilitate cell communication between Fib, MCs, and tubular cells in CKD. The decreased ligand gene expression in Fib within the relevant pathways due to Empa treatment explained the observed attenuation in cellular communication strength in the Empa group.

In the 5/6 Nx CKD model, TGF-β pathway activation has promoted the proliferation and activation of M2 macro (Fig. 2a). In addition, CD74 receptor gene expression has been upregulated in M2 macro (Fig. 5c). The interaction between tubular cell-expressed MIF and CD74 promotes the recruitment of inflammatory cells. Furthermore, upregulation of the ligand gene GRN on M2 macro (Fig. 5c; online suppl. Table S4) facilitates M2 macro polarization, aggravating renal damage. Moreover, M2 macro communicate with ECs, Fib, and MCs via the TGF-β pathway, promoting fibrosis progression. Increased expression of fibrotic genes (COL4A1 [25], FN1 [26], LAMB1 [25], and THB1 [27]) in Fib and MCs, combined with upregulation of receptor genes (ITGA1 [28], ITGB1 [29], SDC4 [30], and CD47 [31]) in tubular cells, promotes ECM deposition, improves cell adhesion, and accelerates the progression of renal fibrosis. The number of M2 macro was significantly reduced after Empa treatment (Fig. 2a). Empa inhibits the expression of the TGF-β1 gene in Fib and the GRN gene in M2 macro, reducing intercellular communication in the TGF-β pathway between M2 macro, ECs, Fib, and MCs. This attenuation is critical for modulating the cellular interactions in renal fibrosis progression. Moreover, there was a significant inhibition of communication between Fib, MCs, and tubular cells in the fibrotic pathway during Empa treatment. Therefore, we hypothesize that Empa slows renal fibrosis progression by reducing the number of M2 macro and inhibiting signal transduction in both pro-inflammatory and fibrotic pathways (Fig. 7).

The present study aimed to establish a 5/6 Nx model in nondiabetic CKD rats and employ scRNA-seq to elucidate the communication network of ligand-receptor pairs and signaling pathways associated with the fibrosis process in CKD. Additionally, we sought to investigate how Empa influences cell-cell communication in CKD. Our results demonstrated that Empa treatment significantly weakened intercellular interactions, which are more pronounced in CKD patients. Furthermore, we identified and analyzed key cell types and signaling pathways within the TGF-β signaling pathway and fibrosis pathway. We emphasized the critical roles of M2 macro and fibroblasts (Fib) in CKD and demonstrated that Empa can inhibit renal fibrosis progression by modulating intercellular signaling among these cells.

M2 macro are recognized for their role in tissue repair and immune regulation [32, 33]. However, their overactivation can lead to excessive deposition of ECM and recruitment of Fib, thereby contributing to tissue fibrosis [34]. In the context of renal fibrosis, numerous studies [35‒37] demonstrated that M2 macro play a significant role in both the development and progression of fibrosis. Our team recently investigated the role of macrophage subpopulations in CKD progression in 5/6 Nx rats [14]. The main pro-fibrotic process involved M2 macro polarization from CD206⁻CD68⁻ to CD206⁺CD68⁺ state, increasing pro-fibrotic genes and transforming into Fib [36]. Empa inhibited the expression of the pro-fibrotic genes IFG1 and TREM2, which promote the transformation of M2 macro into fibroblasts. It also inhibited the genes GPNMB, LGALS3, PRDX5, and CTSB that promoted polarization by regulating the mTOR pathway and mitophagy and attenuated the inflammatory signaling of CD8⁺ effector T cells. Therefore, the 5/6 Nx group in this study demonstrated increased interaction strength between macrophages and other cell types compared to the sham group. In CKD, M2 macro have been identified as key players in the communication flow of the TGF-β signaling network. The Empa treatment effectively reduced these interactions. The hierarchical diagram revealed the specific roles of different cell types and their autocrine and paracrine pathways in the TGF-β pathway. The ligand-receptor pairs TGFB1-(TGFBR1 + TGFBR2) and TGFB3-(TGFBR1 + TGFBR2) contributed the most. Gene expression analyses revealed altered TGFB1, TGFB2, and TGFBR2 expression levels in several cell types, which were suppressed in the Empa group. These findings provide valuable insights into the complex mechanisms of intercellular communication in the TGF-β pathway.

Numerous studies have demonstrated that glomerular and tubular cells may contribute to renal fibrosis through mechanisms such as endothelial dysfunction [38‒41], MC apoptosis [42, 43], and tubular cell necrosis [44‒46]. Similarly, we used scRNA-seq to analyze intercellular communications between M2 macro and tubular epithelial cells to investigate further the potential contribution of M2 macro in the development of renal fibrosis via cellular cross talk. M2 macro communicate with tubular cells through the MIF and GRN pathways. The MIF pathway in the sham group primarily used PTCs and LOH as signal transmitters, with collecting duct PCs as the primary receivers. However, M2 macro emerged as the primary signal receivers in the 5/6 Nx group, receiving signals from tubular cells via paracrine signaling. The observed transition in receptor contribution from ACKR3 to CD74 + CXCR4 was a critical aspect of CKD progression. Empa treatment can potentially reduce interaction between M2 macro and other cells by inhibiting signal transmission in the MIF pathway and gene expression in the GRN pathway. MIF is a potent cytokine that recruits leukocytes and stimulates pro-inflammatory responses by binding to CXCR4 or CD74 [47, 48]. The increased MIF expression in endothelial and renal tubular epithelial cells is associated with macrophage accumulation and renal dysfunction [49]. Anti-MIF therapy has been shown to ameliorate IgA nephropathy and decrease TGF-β expression in glomerulonephritis [50]. The TGF-β pathway is a classic pathway in renal fibrosis, and numerous studies [44, 51, 52] demonstrated that it plays an important role in matrix protein synthesis by inhibiting matrix degradation. GRN is a multifaceted protein involved in cell growth, tissue remodeling, and inflammation. In addition, it has been associated with systemic lupus erythematosus [53, 54]. Moreover, GRN is involved in M2 macro polarization in lupus nephritis [54], which can exacerbate renal fibrosis [55]. Importantly, Empa treatment effectively downregulated the GRN gene expression in our study. Therefore, our findings indicate that Empa treatment has the potential to attenuate the interaction between M2 macro and other cells by weakening the signal transmission in the MIF pathway and inhibiting gene expression in the GRN pathway.

Excessive ECM deposition within the tubulointerstitium of the kidney is the defining feature of interstitial fibrosis. The ECM primarily comprises four types of proteins: collagen, laminin, fibronectin, and thrombospondin [56]. Under pathological conditions, these proteins can be excessively synthesized and deposited, causing renal fibrosis progression. The present study observed significant alterations in pro-fibrotic signaling pathways in CKD, specifically collagen synthesis (COL4A1), fibronectin synthesis (FN1), laminin synthesis (LAMB1), and thrombospondin synthesis (THBS1). The 5/6 Nx group revealed increased signal output from Fib and MCs in all four pathways, with Fib having the highest signal output. Communication between ECs and tubular cells was also observed in the collagen and laminin pathways. Collagens and fibronectin, essential ECM proteins, regulate various biological processes such as tissue repair, wound healing, and embryonic development [57, 58]. After tissue injury, collagens and fibronectin are the first proteins secreted by fibroblasts and other migratory cells at the injury site [59]. Collagens are also the most abundant proteins in the ECM, which provide mechanical support and structure to tissues [57].

In contrast, fibronectin acts as a scaffold for cell migration and tissue regeneration, as well as facilitating cell adhesion [60]. Collagens and fibronectin function as a fundamental framework, allowing ECM molecules such as proteoglycans and glycosaminoglycans to accumulate and increase the complexity and diversity of the ECM. Therefore, collagens and fibronectin deposition are critical in the wound-healing process. However, dysregulation in their expression or function can lead to pathological conditions such as fibrosis. Type IV collagen, with laminins, proteoglycans, and entactin/nidogen, is the primary structural component of the glomerular basement membrane, forming a mesh-like structure. Previous studies have demonstrated that COL4A1 plays an important role in the fibrosis progression [61]. TGF-β is thought to be responsible for increased COL4A1 and FN1 expression, which leads to renal fibrosis [62]. Ligand genes COL4A1, COL4A2, FN1, LAMB1, THBS1, and THBS2 were significantly upregulated in this study, particularly in the Fib and MCs of the 5/6 Nx group. The significant expression of receptor genes ITGA1, ITGB1, SDC4, and CD47 in tubular cells is consistent with their role as primary signal receivers. Our findings emphasize the importance of ligand genes in facilitating cell communication between Fib, MCs, and tubular cells in CKD. The observed decrease in ligand gene expression in Fib due to Empa therapy explains the weakened cellular communication in the Empa-treated group. LAMB1, an ECM protein [63], mediates cell adhesion and migration. Researchers have focused increasingly on its role in tissue fibrosis, a common aging-related organ dysfunction [64]. Recent research suggests that renal tubular epithelial cells can act as key mediators of renal fibrosis, a common age-related organ dysfunction [65, 66]. Multiple ECM molecules, including LAMB1, are produced by renal tubular epithelial cells, which can interact with various immune cells to promote fibrosis progression [67, 68]. Activated platelets secrete THBS proteins that activate the CD47 receptor on target cells, activating fibrosis-related downstream signaling pathways.

Furthermore, the THBS/CD47 signaling pathway has been identified as a key regulator of renal fibrosis [69]. Importantly, this pathway inhibits the fibrosis progression in experimental models of various diseases, including diabetes, liver cirrhosis, and pulmonary fibrosis [70]. THBS1 is a glycoprotein found in numerous tissues, including the kidney. THBS1 is upregulated in glomerular MCs and tubular epithelial cells in DKD, contributing to the development of renal fibrosis [71]. TGF-β signaling activation is one way by which THBS1 promotes fibrosis. THBS1 can bind to and activate latent TGF-β, causing phosphorylation and nuclear translocation of Smad proteins, which regulate the ECM gene expression. In addition, THBS1 can induce the TGF-β receptor expression and improve TGF-β signaling sensitivity in cells. These effects of THBS1 on TGF-β signaling contribute to renal fibrosis in diabetic nephropathy [71].

In conclusion, THBS1 is an important endogenous TGF-β activator in diabetic nephropathy and promotes the development of renal fibrosis. THBS1 and its downstream signaling pathways are potential therapeutic targets to treat diabetic nephropathy. These findings highlight the importance of LAMB1 and the THBS/CD47 signaling pathway in the development of fibrosis in organ dysfunction associated with aging and suggest that targeting these pathways may have therapeutic potential for reducing tissue fibrosis.

It is important to acknowledge the limitations of this study. First, the use of animal models may not fully capture the complexity of human physiology. Second, the scRNA-seq technique has inherent limitations, such as the potential for batch effects and the possibility of false positives in differential gene expression analyses. Further research is needed to address these limitations and to validate the findings in a larger human population. For instance, long-term studies with larger sample sizes could provide more robust evidence of the efficacy of Empa in inhibiting renal fibrosis progression. Additionally, the study of different genetic backgrounds and pathological conditions in animal models could help further elucidate the mechanisms of action of Empa in CKD.

The 5/6 Nx CKD model revealed that the TGF-β pathway stimulates M2 macro. The upregulation of the CD74 receptor gene in M2 macro and the interaction between MIF expressed in tubular cells and CD74 facilitate the recruitment of inflammatory cells. Furthermore, the GRN gene upregulation on M2 macro promotes polarization, thereby exacerbating renal injury. In addition, increased fibrotic gene expression in Fib and MCs, in conjunction with the receptor gene upregulation in tubular cells, promotes increased ECM deposition, cell adhesion, and renal fibrosis progression. Empa treatment significantly reduced the number of M2 macro and downregulated the expression of TGF-β1 in Fib and the GRN gene in M2 macro. This effectively suppressed the intercellular communication within the TGF-β pathway between M2 macro, ECs, Fib, and MCs, modulating the cellular interactions involved in renal fibrosis progression. Therefore, we believe Empa slows the progression of renal fibrosis by reducing the number of M2 macro and inhibiting signal transduction in both pro-inflammatory and fibrotic pathways.

We are thankful to Singleron Biotechnologies for their technical support in single-cell sequencing and data analysis. Their assistance was instrumental in the success of this research project.

The animal experiment was approved by the Experimental Animal Ethics Committee of Jinan University (Approval No. 00245787). We strictly followed the National Law for Laboratory Animal Experimentation guidelines. Animal welfare and well-being were prioritized throughout the study. The animal experiments conducted in our study adhered to ARRIVE guidelines.

No potential conflict of interest relevant to this article was reported.

This work was supported by the National Natural Science Foundation of China (Grant No. 82100716, 82170690), Guangdong Basic and Applied Basic Research Foundation (Grant No. 2022A1515110267), and the Shenzhen Science and Technology Innovation Committee of Guangdong Province of China (Grant No. JCYJ20210324123200003).

T.Z.: writing – review and editing, supervision, methodology, funding acquisition, and conceptualization. B.H.: writing – review and editing, supervision, methodology, investigation, and conceptualization. Y.-P.L.: writing – review and editing, supervision, investigation, and funding acquisition. L.L.: writing – original draft, data curation, methodology, and investigation. Y.-X.X., M.-L.L., and Y.L.: writing – original draft, data curation, and investigation. Z.-Y.Z., H.-W.W., T.-J.C., and X.-M.Z.: writing – original draft and data curation. C.T.: writing – original draft and investigation. X.-H.W., D.D., T.K., and B.K.K.: writing – review and editing, and methodology. Z.-H.Z.: writing – review and editing, supervision, methodology, funding acquisition, and conceptualization. All the authors have read and approved the paper and declared no potential conflicts of interest in the paper. All the authors agreed to publish this paper.

Lei Lei, Yun-Xiu Xiang, and Mao-Lin Luo contributed equally to this work.

All data generated or analyzed during this study are included in this published article (in the online suppl. tables). Further datasets are available from the corresponding author on reasonable request.