Multi-modal Single-Cell Analysis Reveals Brain Immune Landscape Plasticity during Aging and Gut Microbiota Dysbiosis

Abstract

Phenotypic and functional plasticity of brain immune cells contribute to brain tissue homeostasis and disease. Immune cell plasticity is profoundly influenced by tissue microenvironment cues and systemic factors. Aging and gut microbiota dysbiosis that reshape brain immune cell plasticity and homeostasis has not been fully delineated. Using Cellular Indexing of Transcriptomes and Epitopes by sequencing (CITE-seq), we analyze compositional and transcriptional changes of the brain immune landscape in response to aging and gut dysbiosis. Discordance between canonical surface-marker-defined immune cell types and their transcriptomes suggest transcriptional plasticity among immune cells. Ly6C⁺ monocytes predominate a pro-inflammatory signature in the aged brain, while innate lymphoid cells (ILCs) shift toward an ILC2-like profile. Aging increases ILC-like cells expressing a T memory stemness (T_scm) signature, which is reduced through antibiotics-induced gut dysbiosis. Systemic changes due to aging and gut dysbiosis increase propensity for neuroinflammation, providing insights into gut dysbiosis in age-related neurological diseases.

Keywords: CITE-seq; CNS; aging; brain; brain immunity; dysbiosis; gut microbiota; single-cell sequencing.

Figure 1.. CITE-Seq Delineates the Global Immune Cell Diversity in the Brain

(A) Schematic of experimental approach. (B) Single cells projected onto UMAP with cells color coded by transcriptional cluster ID (left). UMAP clustering as in left plot with cells color coded by canonical Cell-ID (right). Dashed circle (a) on peripheral lymphoid cell populations and dashed circle (b) on non-microglia myeloid cells are shown. (C) Heatmap of top differentially expressed genes (DEGs) per transcriptional cluster. (D) Stacked bar charts of peripheral lymphoid cells (top) and peripheral myeloid cells (bottom) frequencies (n = 3 biological replicates per group). (E) Volcano plot of differentially enriched gene pathways (DEGPs) in CD8⁺ T cells between aged and young. (F) Violin plots of Ly6a and Dusp2 expression in CD8⁺ T cells in young and aged (n = 3 biological replicates per group). All plots were derived from pooling three biological replicates per experimental condition. Data in (E) and (F) were analyzed by Wilcoxon rank-sum test. Also see Tables S1, S2, S3 and S4.

Figure 2.. Aged Brain Is Enriched for Inflammation-Prone Brain-Resident Myeloid Cells

(A) UMAP clustering as in Figure 1B with Cell-ID microglia colored in pink (top) and with Cell-ID microglia colored by transcriptional clusters (bottom). (B) Stacked bar charts of Cell-ID microglia frequencies within transcriptional clusters (n = 3 biological replicates per group) (top). Volcano plot of DEGPs between RNA cluster 1 and RNA cluster 0 (bottom) is shown. (C) Heatmap with top 40 DEGs in Cell-ID microglia. (D) UMAP clustering as in Figure 1B with Cell-ID BAMs in blue (top) and with Cell-ID BAMs colored by transcriptional cluster ID (bottom). (E) Stacked bar charts of Cell-ID BAMs frequencies within transcriptional clusters (n = 3 biological replicates per group) (top). Volcano plot of DEGPs between RNA cluster 3 and RNA cluster 8 (bottom) is shown. (F) Heatmap with top 26 DEGs in Cell-ID BAMs. All plots were derived from pooling three biological replicates per experimental condition. Data in (B) and (E) were analyzed by Wilcoxon rank-sum test. Also see Tables S2, S3, and S4.

Figure 3.. Innate Ly6C^High Monocytes Show Microenvironment-Dependent Plasticity in Aged Brain

(A) UMAP clustering as in Figure 1B with RNA cluster 6 in turquoise (left) and with Cell-ID Ly6C^High darkened (right, split by conditions). (B) Re-clustered young and aged Ly6C^High cells on UMAP (top left) and stacked bar charts of cluster frequencies (n = 3 biological replicates per group) (top right) and heatmap of Ly6C^High subcluster DEGs (bottom). (C) Trajectory of Ly6C^High subclusters overlaid with indicated genes (top). Heatmap of trajectory DEGs in cluster ordering (bottom). Ordering is denoted by arrows at the bottom. (D) Volcano plot of DEGPs in Ly6C^High subcluster 2 versus all other Ly6C^High subclusters. (E) PCA triplot of sparse canonical correlation analysis (CCA) of microbiota families and Ly6C^High DEGs in young and aged. Dots represent cells from aged and triangles represent cells from young. Ly6C^High subclusters are represented by different colors. Gene names are in green and OTUs are in pink. (F) Gene-microbial family correlation plot. Red indicates positive and blue indicates negative correlation. All correlations have corr.test p < 0.01. All plots were derived from pooling three biological replicates per experimental condition. Data in (D) were analyzed by Wilcoxon rank-sum test. Also see Tables S2, S3, and S4.

Figure 4.. Aged Brain Increases Ly6C^Low Patrolling Monocyte Plasticity

(A) UMAP clustering as in Figure 1B with RNA cluster 3 in green (left) and with Cell-ID Ly6C^Low darkened (right, split by conditions). (B) Re-clustered young and aged Ly6C^Low cells on UMAP (top left) and stacked bar charts of cluster frequencies (n = 3 biological replicates per group) (top right). Heatmap of Ly6C^Low subcluster DEGs (bottom) is shown. (C) Trajectory of Ly6C^Low subclusters overlaid with indicated genes (top). Heatmap of trajectory DEGs in cluster ordering (bottom) is shown. Ordering is denoted by arrows at the bottom. (D) Volcano plot of DEGPs in Ly6C^Low subcluster 1 versus all other Ly6C^Low subclusters. (E) PCA triplot of sparse CCA of microbiota families and Ly6C^Low DEGs in young and aged. Dots represent cells from aged and triangles represent cells from young. Ly6C^Low subclusters are represented by different colors. Gene names are in green and OTUs are in pink. (F) Gene-microbial family correlation plot. Red indicates positive and blue indicates negative correlation. All correlations have corr.test p < 0.01. All plots were derived from pooling three biological replicates per experimental condition. Data in (D) were analyzed by Wilcoxon rank-sum test. Also see Tables S2, S3, and S4.

Figure 5.. CNS Innate Lymphoid Cell Plasticity Reflects Chronic Neuroinflammation in the Aged Brain

(A) UMAP clustering as in Figure 1B highlighting RNA-based clusters 7 (blue) and 9 (purple) (left) and with Cell-ID ILCs darkened (right, split by conditions). (B) Re-clustered young and aged ILCs on UMAP. Plots split by young and aged (top left) and stacked bar charts of cluster frequencies (n = 3 biological replicates per group) (top right). Dot plot of ILC subcluster expression level (color scale) and percentage (size of dot) for indicated cell types and associated marker genes (bottom). (C) Heatmap of ILC subcluster DEGs. (D) Volcano plot of DEGPs in ILC subcluster 1 versus all other ILC subclusters. (E) Trajectory of ILC subclusters overlaid with indicated genes. (F) Heatmap of trajectory DEGs in cluster ordering. Ordering is denoted by arrows at the bottom. All plots were derived from pooling three biological replicates per experimental condition. Data in (D) were analyzed by Wilcoxon rank-sum test. Also see Tables S2 and S3.

Figure 6.. ABX Treatment Alters ILC Plasticity Pattern in the Aged Mice

(A) Re-clustered young control and young ABX ILCs on UMAP (top left) and stacked bar charts of cluster frequencies (n = 3 biological replicates per group) (top right). Re-clustered aged control and aged ABX ILCs on UMAP (bottom left) and stacked bar charts of cluster frequencies (n = 3 biological replicates per group) (bottom right). (B) Heatmap of aged control and aged ABX ILC subcluster DEGs. (C) Trajectory of aged control and aged ABX ILC subclusters overlaid with indicated genes. (D) Heatmap of trajectory DEGs in cluster ordering. Ordering is denoted by arrows at the bottom. All plots were derived from pooling three biological replicates per experimental condition. Also see Tables S2, S3, and S4.