CD36 restricts lipid-associated macrophages accumulation in white adipose tissues during atherogenesis

Abstract

Visceral white adipose tissues (WAT) regulate systemic lipid metabolism and inflammation. Dysfunctional WAT drive chronic inflammation and facilitate atherosclerosis. Adipose tissue-associated macrophages (ATM) are the predominant immune cells in WAT, but their heterogeneity and phenotypes are poorly defined during atherogenesis. The scavenger receptor CD36 mediates ATM crosstalk with other adipose tissue cells, driving chronic inflammation. Here, we combined the single-cell RNA sequencing technique with cell metabolic and functional assays on major WAT ATM subpopulations using a diet-induced atherosclerosis mouse model (Apoe-null). We also examined the role of CD36 using Apoe/Cd36 double-null mice. Based on transcriptomics data and differential gene expression analysis, we identified a previously undefined group of ATM displaying low viability and high lipid metabolism and labeled them as "unhealthy macrophages". Their phenotypes suggest a subpopulation of ATM under lipid stress. We also identified lipid-associated macrophages (LAM), which were previously described in obesity. Interestingly, LAM increased 8.4-fold in Apoe/Cd36 double-null mice on an atherogenic diet, but not in Apoe-null mice. The increase in LAM was accompanied by more ATM lipid uptake, reduced adipocyte hypertrophy, and less inflammation. In conclusion, CD36 mediates a delicate balance between lipid metabolism and inflammation in visceral adipose tissues. Under atherogenic conditions, CD36 deficiency reduces inflammation and increases lipid metabolism in WAT by promoting LAM accumulation.

Keywords: atherosclerosis; inflammation; lipid; macrophage; scRNA-Seq; visceral adipose tissue.

Figure 1

ATM represent the major immune cell type in the white adipose tissue during atherogenesis. (A) The Apoe-null or Apoe/Cd36 double-null mice fed on chow diet or HFD for 10 weeks were sacrificed (n = 2 per group) and CD45⁺ cells were isolated from pgWAT and subjected to scRNA-Seq analysis. Cells were unbiasedly clustered based on the FindNeighbors method from Seurat package. Distinct cell clusters are shown through uniform manifold approximation and projection (UMAP). Immune cell types were determined using the SingleR method. (B) Violin plots of relative expression of macrophage marker genes among CD45⁺ adipose tissue cells. (C) UMAP plots of 4 different conditions were shown. (D) Percentages of immune cell populations as defined by SingleR within total CD45⁺ cells in each condition were shown.

Figure 2

Identification of subpopulations of pgWAT ATM, which are highly dynamic during atherogenesis. (A) ATM from all 4 conditions were selected and combined. They were divided into 10 clusters (0–9) using the same method as in Figure 1 and shown in the UMAP. (B) Violin plots of relative expression of macrophage marker genes among ATM clusters. (C) Heatmap showing the 10 most upregulated genes in each cluster as defined in (A) Selected enriched genes used for biological identification of each cluster are shown on the right of the heatmap. (D) The ATM names corresponding to different cluster numbers in the texts are shown. (E) Violin plots of relative expression of enriched genes in the cluster 0, 1, 2, 3, 5, and 6. (F) Gene expression patterns projected onto the UMAP plots showing the enrichment of Trem2, Fabp5, Gpnmb, Cd9, Lagls3, Atp6v0d2, Ctsl, Lipa, and Lpl in LAM (cluster 3). The graphical panel in Figure 2D was created with BioRender.com.

Figure 3

LAM relative content is upregulated by HFD specifically in the apoe/Cd36 double-null mice. (A) UMAP plots of 4 different conditions demonstrate distinct ATM constitutions. (B) Bar chart of the relative frequency of ATM subpopulations among 4 conditions. (C) Flow cytometry analysis of the percentage of F4/80 ⁺Trem2⁺ ATM among all CD11b ⁺F4/80⁺ ATM. Data are shown as mean ± S.E. in the bar graphs (n = 8-9 individuals per group) and Two-Way ANOVA was performed. (D) Representative images of H&E-stained adipose tissues. Scale bar: 150 μm. (E) Areas of the adipocytes shown in (D) were quantified and shown in the dot plot chart. The medium value in each condition was denoted with a horizontal line. The n value means the number of individuals. One-Way ANOVA was performed. ns: not significant. (F) Plasma insulin levels and (G) pgWAT IL-6 levels were shown in the bar graphs (n = 6-11 individuals per group). *P < 0.05, **P < 0.01, ****P < 0.0001.

Figure 4

LAM are more active in lipid metabolism and less inflammatory, compared to other ATM. (A) Gene ontology enrichment analysis of biological pathways showing upregulated pathways (left panel) and downregulated pathways (right panel) in the LAM. (B) GSEA on LAM upregulated pathways including cholesterol metabolism, lysosome, PPAR signaling, and synaptic vesicle cycle. (C) GSEA on LAM downregulated pathways including IL-17 signaling, NF-κB signaling, TNF signaling, and viral protein signaling. (D) Isolated ATM were incubated with Bodipy-labeled palmitate for 15 min. Flow cytometry analysis of palmitate uptake by CD11b ⁺F4/80⁺ ATM. Mean fluorescence intensity (MFI) of Bodipy was quantified. Data are shown as mean ± S.E. in the bar graphs (n = 7 individuals per group) and Two-Way ANOVA was performed. ns: not significant. **P < 0.01.

Figure 5

Transcription factor regulons of pgWAT ATM. (A) SCENIC results showing the binary regulon activity matrix. The regulons that are specific to ATM subpopulations are highlighted in rectangle boxes (A–E). The highly enriched transcription factor regulons in each box are listed underneath the matrix. (B) Violin plots of relative expression of LAM-active regulons among ATM clusters.

Figure 6

Cd36 deficiency reverses the pro-inflammatory and metabolic responses to HFD. (A) Upper panel: Volcano plot showing Apoe-null chow vs. Apoe-null HFD gene expression fold change (x-axis, log2 scale) and their adjusted p-value (y-axis, -log10 scale). Highly significant genes are indicated by a green dot. Bottom panel: GSEA showing significantly different pathways compared between Apoe-null chow and Apoe-null HFD. (B) Same analysis as in (A) comparing Apoe/Cd36 double-null chow and Apoe/Cd36 double-null HFD. (C) Same analysis as in (A) comparing Apoe-null chow and Apoe/Cd36 double-null chow. (D) Same analysis as in (A) comparing Apoe-null HFD and Apoe/Cd36 double-null HFD. (E) The schematic diagram shows the strategy for identifying genes that are most impacted by CD36 deficiency. (F) Fold change (y-axis, Log2 scale) of individual genes (x-axis) as induced by HFD in Apoe-null or Apoe/Cd36 double-null pgWAT ATM respectively. The top 17 differentially regulated genes are shown.