Identification of functional heterogeneity of immune cells and tubular-immune cellular interplay action in diabetic kidney disease

Single-cell transcriptomic data preparation

First, we downloaded scRNA-seq data from the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo) dataset GSE181382, which contained the single-nucleus transcriptomic data of db/mc control group, db/db group and db/db SGLT2i groups produced by 10× genomics.

Determination of cell type

We followed the methods of Zhang et al. 2022. The clusters were annotated using cell type-specific signatures and marker genes, for example, macrophages of C1qa, C1qb, Cd74, Cd79a and Cd79b, and T cell of CD4, CD8, Ltb, Nkg7 Gzma.

Pathway analysis and intercellular crosstalk analysis

We followed the methods of Zhang et al. 2022.^[15]

Single cell RNA sequence data preparation

The raw gene expression matrix was obtained and processed to align reads with the reference genome (Homo_sapiens_GRCh38_96) using Cell Ranger (version 5.0.0). Data filtration and normalization were performed using the R package Seurat (version 4.0.0) according to the manufacturer’s manual. We used Seurat (4.0.0) to filter low quality cells, cells with the number of detected genes < 200 or > 5000, and cells with > 25% of the UMI counts belonging to mitochondrial genes were also omitted. Only scRNA seq data that met the quality control criteria were analyzed in this study.

Canonical correlation analysis, dimensionality reduction, and clustering

After quality control and filtering, the clustering and visualization was finished by Seurat (V4.0.0), with the following steps: First, library-size normalization to each cell was performed by NormalizeData. The variable genes were calculated by FindVariableGenes. Then, all libraries were combined using FindIntegrationAnchors and IntegrateData with default parameters, and using ScaleData to regress out the variability of the number of UMIs. Then the RunPCA and RunUMAP were used to reduce dimensions. FindClusters was used to cluster cells using 20 dims at a resolution of 0.8 (total 27 clusters). FindAllMarkers was used to compare each cluster to all others to identify cluster-specific marker genes. Each retained marker gene was expressed in a minimum of 25% of cells and at a minimum log-fold change threshold of 0.25. The clustering differential expressed genes were considered significant if the adjusted P-value was less than 0.05, and the avg_log₂FC was ≥ 0.25.

Deferentially expressed genes in specified clusters

After cell annotation, differentially expressed genes in specified cell types were analyzed using the FindMarkers function based on the bimod algorithm of the R package Seurat (version 4.0.0). Fold changes 1.25 and P < 0.05 were considered significantly modulated.

Cell-cell interactions

CellPhoneDB is a public repository of curated receptors, ligands, and their interactions. In this study, cell crosstalk interaction was performed using CellPhoneDB (version 2.1.1) according to the manufacturer’s manual (https://www.cellphonedb.org/). The mean value represents the average ligand and receptor expression in a specific cell type, which is calculated based on the percentage of cells expressing the specific gene and the gene expression mean. The P-value is calculated based on the proportion of means that are as high as or higher than the actualmean, which represents the likelihood of a specific cell type of a given receptor–ligand complex.

Spontaneous type 2 diabetic db/db mice model and measurements of urinary albumin excretionWe purchased the db/db and db/mc mice and measured the UAE based on our previous research.^[16] The method of treatment to db/db and db/mc mice were followed by the methods of Wu et al. 2021.^[13] The db/db mice were defined as DKD by albuminuria at 10 weeks of age and gavaged for 8 weeks, vehicle control (phosphate-buffered saline [PBS]) (n = 4), 3 mg/kg/d of dapagliflozin (n = 4), and sacrificed at the age of 18 weeks. All animal studies were performed according to the protocol approved by the Animal Care Committee at the First Medical Center of Chinese PLA General Hospital. The urine albumin concentration was measured using the ELISA kit (Bethyl Laboratories, Montgomery, TX, USA).

Antibodies and reagents

The antibodies and reagents used in this study included rabbit anti-CD206 (ab64693, 1:400, Abcam, Cambridge, UK), goat anti-S100a8 (AF3059, 1:100, R&D systems, MN, USA), goat anti-S100a9 (1:100, AF2065, R&D systems, MN, USA), DNA was stained with Hoechst 33342 (1:1000, H3570; Thermo Fisher Scientific, CA, USA), dapagliflozin (3 mg/kg/d, S1548, Selleck, TX, USA)

Immunofluorescence microscopy and kidney histology

For tissue immunofluorescence and histology analysis, we keep following the methods in our previous research.^[16] Images were acquired at room temperature with a 60×/1.42 oil objective on an Olympus FV1000. All acquisition settings were kept constant for the experimental and control groups in the same experiment. All raw images were analyzed with Velocity 6.0 software (Perkin Elmer).

Statistical analysis

Statistical analysis was performed with GraphPad Prism. The statistically analyzed data are expressed as the mean ± SD. To compare the means between the three groups, oneway ANOVA followed by Dunnett’s multiple comparison test was used, as noted in the figure legends. For all tests, differences were considered statistically significant if P values were < 0.05 (as indicated with ^***P < 0.001). Investigators were blinded during the assessment of all staining assays.

Immune cells atlas at single-cell resolution in the pathogenesis and treatment with SGLT2i in DKD

To decipher the comprehensive landscape of potential dynamic features of macrophages in the DKD progression and treatment with SGLT2i, a total of 58760 single-cell transcriptomes from GSE138182 dataset of db/mc, db/ db and SGLT2i-treated groups were analyzed. Using unsupervised clustering (uniform manifold approximation and projection [UMAP]), 27 separate cell clusters were identified after pooling all samples together (Figure 1A and 1B) and then categorized into ten distinct cell types, including proximal tubules (PT), neutrophil, distal convoluted tubule (DCT), loop of henle (LOH), T cells (T), macrophage, B cells (B), collecting duct (CD), endothelial cell (EC), proliferative PT (prolif-PT) (Supplementary Figure S1). Cluster 19 was annotated as macrophage since it is characterized by representative macrophage genes (C1qa, C1qb and CD74) (Figure 1C). Cluster 15 was annotated as T cell, characterized by its marker genes (Ltb, Nkg7, and Gzma) (Figure 1D).

Figure 1

Integrated scRNA-seq of db/mc, db/db and SGLT2i-treated samples. We performed UMAP analysis (A), identified gene markers (B, C, D), of the 58760 cells.

Dapagliflozin enhanced the portion S100a8/a9 high macrophage, possibly by glucogenesis to fuel anti-inflammatory function. These 1306 macrophages were further divided into 5 clusters after passing mitochondrial quality control. Through unsupervised clustering, the representative genes for each cluster were C1qa, CCl₃, H2-DMb2, Spp1, Pck1(Figure 2A). After dapagliflozin treatment, compared with the db/db and db/mc groups, the portion of Ccl3+ group (S100a8/a9 high) and Pck1+ group significantly increased (Figure 2B)

Figure 2

Functional heterogeneity of macrophage subcluster in the DKD. The distributions of the 5 subcluster macrophage in db/mc, db/db, and SGLT2i treated-db/db groups (A). The portion of subcluster in each group (B). The heatmap shows the top 9 significantly differentially expressed (SDE) genes in each subset of macrophages (C). GSVA analysis indicates enriched pathways of Ccl3+ macrophages (D).

We next focused on the potential function of Ccl3+ macrophages in the three groups, and found that it exhibited a high level of anti-inflammatory and antimicrobial signatures such as Lcn2, Hp, S100a8, S100a9 (Figure 2C), while the portion of Ccl3+ macrophage in db/db group decreased and reversed in the db/db_SGLT2i treated groups (Figure 2C), suggesting the host susceptibility to severe bacterial infection in the progression of DKD.

S100a8/a9/a11, S100 calcium-binding protein a8/a9/a11 (S100a8/a9/a11) is a member of the S100 family, S100a8/ a9 is related to inflammation, obesity and diabetes.^[17,18,19] S100a8 attenuates the production of proinflammatory mediators^[20] and decreases mast cell degranulation and cytokines secretion.^[21] S100a8 forms a homodimer or a heterodimer with S100a9.

Pathway analysis of the unique gene sets in subcluster Ccl3 and Pck1 revealed that the pathways enriched in subcluster Ccl3 were absent in subcluster Pck1, while the pathways enriched in subcluster Pck1 were absent in subcluster Ccl3, indicating that these two subclusters likely have an exclusive function in DKD (Figure 2D). The GO term enrichment analysis of the marker revealed that subcluster Ccl3+ macrophage was enriched by response to a molecule of bacterial origin and lipopolysaccharide while the portion of Ccl3+ macrophage in total macrophage was decreased, suggesting that during the progression of diabetic kidney disease, the antiinflammatory activity of subcluster Ccl3+ macrophage was inhibited and then reversed by SGLT2i treatment.

To further validate these results, type 2 diabetic db/ db mice at 10 weeks of age were treated with vehicle, SGLT2i (dapagliflozin) for a total of 8 weeks (Figure 3A). Nondiabetic db/mc mice were used as their controls. Compared with db/m mice, db/db mice had increased 24-h urinary albumin excretion (UAE), a reduction in UAE was observed in SGLT2i-treated groups (Figure 3B). Histologically, both glomerular volume and mesangial matrix expansion were increased in db/db mice compared with db/mc mice, but these were reduced by dapagliflozin treatment (Figure 3C). The specific markers of Ccl3+ macrophage (S100a8/S100a9) were validated by immunostaining of mouse kidneys (Figure 3D) from db/ mc, db/db and SGLT2i-treated mice.

Figure 3

Characterization of mice and validation of macrophage after 8 weeks of treatments. (A) Schematics of the experimental design. UAE were monitored in 8-week-old db/db mice. Treatment (PBS, SGLT2i) was started after db/db mice developed DKD approximately 10 weeks of age. All mice were sacrificed 8 weeks after treatment, Four mice per group. (B) Urinary albumin excretion (UAE) in db/mc, db/db and db/db+SGLT2i group. (C) Kidney sections from each group were stained with PAS. Scale bars, 20 μm. (D) Kidney samples of each group were stained with the S100a8 (green), S100a9 (green) and M2 macrophage (CD206, red). Scale bars, 20 μm. The data are the means ± SD. One-way ANOVA was performed followed by Dunnett’s multiple comparisons. ^***P < 0.001.

SGLT2i treated group also showed the increased portion of Pck1+ macrophage, Pck1 is a main control point for the regulation of gluconeogenesis, showing the metabolism shifts toward gluconeogenesis, fueling the anti-inflammatory function of M2 macrophage.^[22] So, we hypothesized that dapagliflozin can promote its anti-inflammatory effect mainly by changing the glucose metabolism pathway of macrophages.

Characteristics and dynamics of T cells in DKD

T cells were categorized into 7 subclusters based on available markers (Ltb, Nkg7, and Gzma) by UMAP analysis (Supplementary Figure S2A), and were performed to functionally annotate the seven T cell subsets (Supplementary Figure S2B). CD8a+ Tcm (central memory) subcluster decreased in the db/db group and reversed by the SGLT2i treatment (Supplementary Figure S2C). CD8a+ Tcm is characterized by its signature genes, Ms4a4b, Ms4a6b, Nkg7, and Ccl5. Ms4a4b, Ms4a6b, and Nkg7, preoccupied with immune defense, use MS4A family to enhance sensitivity to extrinsic antigen stimulation, modulates T cell activation and cytotoxic function.^[23,24] Ccl5 is a superfamily of secreted proteins involved in immunoregulatory and inflammatory processes, leading to an innate immune response.^[25] The GO term enrichment analysis of the marker genes revealed that subcluster Cd8a+ Tcm was enriched by the phosphatidylinositol-3-kinase signaling pathway (Supplementary Figure S2D). It suggested that the central T cell memory activation and function was inhibited in the progression of diabetic kidney disease and reversed by dapagliflozin treatment.

Intracellular immune-tubular cell communication in DKD

Since cross-talk between cell types might underline renal inflammation, promoting the progression of diabetic kidney disease. We characterized the intercellular receptor-ligand pairs and molecular interactions of the two cell types by the CellPhoneDB algorithm. Interestingly, we found that the frequency of receptor-ligand interactions in Figure 4A and 4B varies markedly when comparing the dbdb group, dbmc group versus dbdb_SGLT2i (SGLT2i) group, and summarized the key interactions between ligand and receptor from tubular-immune cell communication (Table 1). Notably, the GRN plays central roles in the most active receptor-ligand interactions.These results indicate that the crosstalk between macrophages via diverse receptor-ligand signals may exert a profound effect on diabetic kidney disease progression and treatment.

Progranulin (GRN) is a secreted glycoprotein that is widely expressed in many cell types including leukocytes and neurons, and is involved in various physiological and pathological progressions, including embryogenesis, wound healing, hostdefense, and tumorigenesis.^[26] GRN has been shown to inhibit the inflammatory response in chronic inflammatory conditions such as rheumatoid arthritis, osteoarthritis, and inflammatory bowel disease.^[27] Macrophage-derived progranulin associated with M2 phenotypes^.[28,29,30] Besides, Zhou et al. reported that PGRN prevents podocyte injury by facilitating mitophagy and mitochondrial biogenesis in DKD. Sortilin 1 (SORT1) endocytoses and delivers PGRN to lysosomes, sortilin-mediated PGRN endocytosis is likely to play a central role in DKD pathophysiology. Under diabetes conditions, insufficient PGRN accelerates DKD progression by disrupting mitochondrial homeostasis.^[31]

Figure 4

Intracellular immune-tubular cell communication in DKD. Overview of selected interactions of ligands and receptors (A). Detailed view of ligandreceptor connections between tubular cell and immune cell subclusters. Each color arrow line indicates the ligands to receptors from one cell population to another. The thickness of arrow line is proportional to the number of ligand-receptor pairs (B). The purple circle represents LOH, the yellow circle represents macrophage, the green circle represents PT, the blue circle represents T cell, The orange circle represents DCT.

Our data analysis showed that macrophage and PT clusters were revealed by enriched signaling pathways, including EGFR-GRN pair, suggesting that skewed macrophage plorization from M2 to M1, accelerating the immunologic injury in DKD. Macrophage-derived TNFRSF1A drives the progression to DKD.

In normal mouse and human kidneys, low CD74 expression limits inflammation to extracellular stimuli. The increased renal cell CD74 expression observed during kidney injury (e.g., abnormally high concentrations of specific metabolites, glucose, and lyso-Gb3), including kidney cancer may contribute, together with increased ligand availability, to elicit biological responses during kidney disease. Thus, overexpression of CD74 led to upregulation of NF-ϰB-dependent genes encoding cytokines in macrophages and to NFϰB activation and proliferation in human embryonic kidney cells. CD74, a chaperone molecule expressed in antigen-presenting cells, mediates the load of antigen peptides onto the MHC class-II molecule.^[32,33]

Notably, EGFR_GRN_SORT1 and TNFRSF1A_GRN complexes were likely the most active receptor-ligand interactions between macrophage and epithelial cells in dbdb group. The CD74 related signaling pathways were enriched in the db/mc and SGLT2i group than those in db/db group, suggesting that these cell-to-cell connections might be important for delay the progression to DKD.

Together, our unbiased dissection of the key ligand-receptor interactions between tubular and immune cells highlights EGFR-GRN-SORT1 and CD74 signaling axes as essential to delay the progression of the to DKD.