Identification of a Novel ECM Remodeling Macrophage Subset in AKI to CKD Transition by Integrative Spatial and Single-Cell Analysis

Abstract

The transition from acute kidney injury (AKI) to chronic kidney disease (CKD) is a critical clinical issue. Although previous studies have suggested macrophages as a key player in promoting inflammation and fibrosis during this transition, the heterogeneity and dynamic characterization of macrophages are still poorly understood. Here, we used integrated single-cell RNA sequencing and spatial transcriptomic to characterize the spatiotemporal heterogeneity of macrophages in murine AKI-to-CKD model of unilateral ischemia-reperfusion injury. A marked increase in macrophage infiltration at day 1 was followed by a second peak at day 14 post AKI. Spatiotemporal profiling revealed that injured tubules and macrophages co-localized early after AKI, whereas in late chronic stages had spatial proximity to fibroblasts. Further pseudotime analysis revealed two distinct lineages of macrophages in this transition: renal resident macrophages differentiated into the pro-repair subsets, whereas infiltrating monocyte-derived macrophages contributed to chronic inflammation and fibrosis. A novel macrophage subset, extracellular matrix remodeling-associated macrophages (EAMs) originating from monocytes, linked to renal fibrogenesis and communicated with fibroblasts via insulin-like growth factors (IGF) signalling. In sum, our study identified the spatiotemporal dynamics of macrophage heterogeneity with a unique subset of EAMs in AKI-to-CKD transition, which could be a potential therapeutic target for preventing CKD development.

Keywords: AKI; CKD; macrophage; single‐cell RNA‐seq; spatial transcriptomic.

Figure 1

Major cell dynamics after UIR. a) The experimental workflow. Experimental AKI‐to‐CKD transition was induced in mice by UIR. Samples were collected from the injured kidneys at days 0, 1, 3, 14, and 28 post‐UIR for 10 × chromium single‐cell and visium spatial transcriptomic procedures. b) After unsupervised clustering, 2D uniform manifold approximation and projection (UMAP) visualization of the 60010 cells identified 13 major cell types. PT, proximal tubule; M/DC, monocyte, macrophage and DC; T/NK, T/NK cell; B, B cell; Neu, neutrophil; Fib, fibroblast; Pro, proliferation cell; EC, endothelial cell; Epi, epithelial cell; CD, collecting duct; DCT, Distal convoluted tubule; Pod, podocyte; LH, loop of Henle. c) Connected bar plots showing the proportionate abundance of each cell clusters in each samples. Immune cells are enlarged to facilitate data visualization.

Figure 2

Spatiotemporal dynamics of main cell types after UIR. a) H&E staining of Visium Spatial Gene Expression samples. b) Representative Masson images at each timepoint. c,d) UMAP of spatial transcriptomics spots based on cell‐type compositions and the injury score in spatial transcriptomics. e)Injury scores in each time points. The arrows point to the areas with the highest injury score. f,g,h) The proportions of macrophages, neutrophils and multiple cells were deconvoluted from the scRNA‐seq data using the cell2location algorithm. Max, maximum; min, minimum. i) The proportion of macrophages and neutrophils infiltrated into each area according to the time‐point after UIR. j) Median relevance of cell‐type abundance in predicting other cell‐type abundances within a location.

Figure 3

Sub‐clustering of monocyte/macrophage. a) Heatmap of top 20 marker genes in each sub‐clusters. The color scheme is based on the distribution of z‐score. b) UMAP plot of all monocyte/macrophages. c) Dot plot showing expression of marker genes. d) The score comparison of typical functions among different monocyte/macrophage clusters (all time points). e) Connected bar plots displaying the proportional abundance in each time points.

Figure 4

Characteristics of ECM remodeling macrophages. a) Expression of representative ECM‐related genes in EAMs after UIR. b) Gene expression of Tgfbi and Mmp9 in spatial transcriptomics dataset. c) Expression of representative lipid metabolism‐related genes. d) Gene expression of Fabp5 and Pla2g7 in spatial transcriptomics dataset. e) Gene Ontology terms enriched from the differentially expressed genes of EAMs compared to all other Mac clusters in each time points. f) Inflammation, fibrosis and ECM scores of EAMs in each time points. g) Gene expression of Igf1 and Mmp12 in spatial transcriptomics dataset.

Figure 5

Macrophage trajectories during repair and fibrosis. a) RNA velocity analysis in each sub‐clusters. b) Inflammation, fibrosis, ECM, and wound healing scores along trajectory. c) Gene expression dynamics and Gene Ontology pathway enrichment analysis along pseudotime (using Monocle2). d) Pseudotime‐dependent gene expression along the lineage 1 and corresponding feature plots. e) Flow cytometric analysis of the expression of Itgam/Igf1, Itgam/Trem2, and Mertk/Vegfa in the injured kidneys at days 0, 1, and 14 post‐UIR.

Figure 6

Ligand–receptor interactome of communication between macrophages and other cells. a) Heatmap of the number of relevant ligand‐receptor interaction pairs predicted by CellChatDB between main kidney cell types in cell‐cell interaction study. Scale = number of interactions. b) Dot plot showing expression of ligand‐receptor pairs in EAMs and all other clusters in each time points. c) UMAP scRNA‐seq and heatmap of select ligands expressed by Mac sub‐clusters and cognate receptor expression by fibroblasts on days 14 and 28 post injury. d) Co‐expression pattern of Igf1 and Igf1r in spatial transcriptomics dataset. Spatial feature plots showing the expression pattern of ligand gene Igf1 (red spots), receptor gene Igf1r (blue spots), and co‐expression pattern (purple spots). e) Multicolour immunohistochemistry staining of Igf1⁺Cd68⁺ cells (Igf1 and Cd68) and Igf1r⁺α‐SMA⁺ cells (Igf1r and α‐SMA) in each injury time points. Scale bars, 10 µm. The heatmaps show the density of Igf1⁺ Cd68⁺ cells and Igf1r⁺ α‐SMA⁺ cells. Results of nearest neighbor distance used to compute areas in which Igf1⁺Cd68⁺ cells (blue) and Igf1r⁺α‐SMA⁺ cells (yellow) lie within ≈1 µm of each other. The box plots show the quantification results. n = 15. Ligand–receptor interactome of communication between macrophages and other cells. f) The representative images of IGF1 and CD68 immunofluorescence costaining in normal, AKI, and CKD kidneys and g) its correlation with fibrosis (Masson staining). h) The urinary sediment, supernatant, and serum IGF1 concentrations of healthy control (n = 15), patients with AKI (n = 15) and CKD including CKD 2–3 stage (n = 15) and CKD 3–5 stage (n= 15) and its correlation with GFR‐EPI. *P < 0.05. i) Western blots of Col1a1, a‐SMA, Vim, and PCNA after stimulation with IGF1 (IGF1 10 ng ml⁻¹ 24 h and IGF1 10 ng/ml 48 h) in NRK49F cells. j) The representative images of IGF1 and CD68 immunofluorescence costaining in NRK49F cells after stimulation with IGF1 (IGF1 10 ng ml⁻¹ 24 h and IGF1 10 ng ml⁻¹ 48 h). k) Working model for spatiotemporal dynamics of macrophage heterogeneity in the process of AKI‐to‐CKD transition.

Figure 6

Ligand–receptor interactome of communication between macrophages and other cells. a) Heatmap of the number of relevant ligand‐receptor interaction pairs predicted by CellChatDB between main kidney cell types in cell‐cell interaction study. Scale = number of interactions. b) Dot plot showing expression of ligand‐receptor pairs in EAMs and all other clusters in each time points. c) UMAP scRNA‐seq and heatmap of select ligands expressed by Mac sub‐clusters and cognate receptor expression by fibroblasts on days 14 and 28 post injury. d) Co‐expression pattern of Igf1 and Igf1r in spatial transcriptomics dataset. Spatial feature plots showing the expression pattern of ligand gene Igf1 (red spots), receptor gene Igf1r (blue spots), and co‐expression pattern (purple spots). e) Multicolour immunohistochemistry staining of Igf1⁺Cd68⁺ cells (Igf1 and Cd68) and Igf1r⁺α‐SMA⁺ cells (Igf1r and α‐SMA) in each injury time points. Scale bars, 10 µm. The heatmaps show the density of Igf1⁺ Cd68⁺ cells and Igf1r⁺ α‐SMA⁺ cells. Results of nearest neighbor distance used to compute areas in which Igf1⁺Cd68⁺ cells (blue) and Igf1r⁺α‐SMA⁺ cells (yellow) lie within ≈1 µm of each other. The box plots show the quantification results. n = 15. Ligand–receptor interactome of communication between macrophages and other cells. f) The representative images of IGF1 and CD68 immunofluorescence costaining in normal, AKI, and CKD kidneys and g) its correlation with fibrosis (Masson staining). h) The urinary sediment, supernatant, and serum IGF1 concentrations of healthy control (n = 15), patients with AKI (n = 15) and CKD including CKD 2–3 stage (n = 15) and CKD 3–5 stage (n= 15) and its correlation with GFR‐EPI. *P < 0.05. i) Western blots of Col1a1, a‐SMA, Vim, and PCNA after stimulation with IGF1 (IGF1 10 ng ml⁻¹ 24 h and IGF1 10 ng/ml 48 h) in NRK49F cells. j) The representative images of IGF1 and CD68 immunofluorescence costaining in NRK49F cells after stimulation with IGF1 (IGF1 10 ng ml⁻¹ 24 h and IGF1 10 ng ml⁻¹ 48 h). k) Working model for spatiotemporal dynamics of macrophage heterogeneity in the process of AKI‐to‐CKD transition.