Hippocampal microglia exhibit progressive age-related transcriptional inflammatory activation.

A, UMAP plot of microglia separated into transcriptional clusters. (n = 1 pool of 5 animals for each age) B, Superimposition of ages onto the UMAP plot with dashed lines signifying relative cluster demarcation. C, Percent composition of each cluster by age. D, Dot plot of expression of cluster markers sorted by age. Percent of cells expressing the gene and average normalized expression are represented. E, Violin plots of genes with dynamic age-related expression patterns. F, Representative images and quantification of C1q (yellow), C3 (red), and Iba1 (cyan) staining in the hilus and molecular layer (ML) of 6- and 24-month-old mice. (n=4-5 mice per group; mixed effects analysis followed by Dunnett’s multiple comparisons; *P<0.05, **P<0.01, ****P<0.0001). G, Number of differentially expressed genes from the 6-month timepoint at each age. Bars above the intersect represent increased expression and those below represent decreased expression. H, The average expression change at all ages for genes differentially regulated genes at individual ages represented by the color scheme in B. I, J, K, Volcano plots of differentially expressed genes in microglia between 6- and 12-months (I), 6- and 18-months (J), and 6- and 24-months (K) and corresponding gene ontology analysis of genes with significantly increased (brown) or decreased (green) expression for each comparison. Data are shown as mean±s.e.m.

Hippocampal microglia exhibit age-related heterogeneity during aging.

A, UMAP plot of all Cd11b+ cells. Several adjacent clusters of microglia were identified, as well as a cluster of proliferating microglia. Smaller populations of peripheral immune cell types – macrophages and neutrophils – were identified. Clusters of astrocytes and vascular cells were also found. Overall, greater than 82% of cells were microglia. (n = 1 pool of 5 animals for each age) B, Dot plot showing expression (average expression and percent of cells expressing) of top two markers for each cluster. C, UMAP plots with expression levels of microglia markers superimposed onto cells. Notice that peripheral immune cells express microglia markers; however, they are distinguished from microglia based on marker expression from (b). D-G, Volcano plots of differential gene expression for the clusters identified in Figure 1A compared to every other cluster for Homeostatic (D), Transition (E), Activation (F), and Interferon (G) microglia clusters. H, Cluster composition of non-proliferating microglia by age. I, Standardized variation of non-proliferating microglia for each age. J, Representative images of NFκB (p65) (yellow) and IBA1 (cyan) staining in the hippocampus of 3- and 24-month-old mice and quantification of staining in 3-, 6-, 12-, 18-, and 24-month-old mice. (n=5 mice per group; one-way ANOVA with Dunnett’s post hoc test; *P<0.05, ***P<0.001). K, Representative images of CD68 (red) and IBA1 (cyan) staining in the hippocampus of 3- and 24-month-old mice and quantification of staining in 3-, 6-, 12-, 18-, and 24-month-old mice. (n=4-5 mice per group; one-way ANOVA with Dunnett’s post hoc test; *P<0.05, **P<0.01). Data are shown as mean±s.e.m.

Hippocampal microglia aging advances through intermediate states that respond to systemic interventions or disease states.

A, Pseudotime trajectories of microglia from an anchor point located in 6-month microglia presented in a UMAP plot. (n = 1 pool of 5 animals for each age) B, Microglia ages superimposed over pseudotime trajectories. C, Gene expression modules representing sections of the inflammatory aging trajectory over the right half of the UMAP plot (left). Modules were discovered using Moran’s I autocorrelation test. Top Gene Ontology terms and representative genes in each module (right). D, Dotplot of pseudotime modules sorted by age. Percent of cells expressing the gene and average normalized expression are represented. E, F, G, Average gene expression changes for each aging module represented as log2 fold change of 12-months (E), 18-months (F), or 24-months (G) over 6-months. H, Diagram of the heterochronic parabiosis model with the comparisons made in scRNA-Seq. I, Average gene expression changes for each aging module represented as log2 fold change of heterochronic young (HY) over isochronic young (IY) adult parabionts. Data from 31. J, Diagram of microglia surrounding an Aβ plaque. K, Average gene expression changes for each aging module represented as log2 fold change of the AppNL-G-F genotype (AD) over wildtype (WT). Data from17. (one sample T-test with the expected value of 0 (no change); *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001).

Hippocampal microglia aging advances through intermediate states that respond to systemic interventions or disease states.

A, Volcano plot of differential gene expression of 12-month microglia versus all other ages with significant genes in teal. B, Gene ontology analysis of biological processes enriched in those genes with increased expression at 12-months of age. C, Representative images and quantification of CD68 (red) and IBA1 (cyan) staining in the hippocampus of isochronic young (IY) and heterochronic young (HY) along with a diagram of the comparisons. (n=5 mice per group; unpaired Student’s T-test; ***P<0.001). Data are shown as mean±s.e.m. D, Dotplot of pseudotime modules in young (Y), isochronic young (IY), and heterochronic young (HY) parabiont microglia. Data is from Palovics et al61. Percent of cells expressing the gene and average normalized expression are represented. E, Volcano plots of differential gene expression of AppNL-G-F genotype (AD) over wildtype (WT) microglia at 6-(left) and 12-months of age. Significant genes are in teal. Data is from Frigerio62. F, Dotplot of pseudotime modules across ages (3-, 6-, 12-, and 21-months old) and genotypes (AppNL-G-F genotype (AD) over C57Bl/6 (WT)). Data is from Frigerio17. Percent of cells expressing the gene and average normalized expression are represented.

Intermediate states of microglia aging act as checkpoints on inflammatory progression.

A, Representation of the microglia aging trajectory over the UMAP plot highlighting the region of peak Tgfb1 expression. B, Representative RNAscope images and quantification of Tgfb1 (red) expression in IBA1 (cyan) cells across ages. (n=5 per group; one-way ANOVA with Dunnett’s post-hoc test; *P<0.05, **P<0.01). C, Schematic of the heterochronic parabiosis model and quantification of hippocampal microglia expression of Tgfb1 from isochronic young (IY) and heterochronic young (HY) adult parabionts. Data derived from Pálovics et al31. D, Top gene ontology terms of genes with significantly decreased expression in bulk microglia RNA-Seq following TGFB1 treatment compared to control (DMSO) in LPS treated microglia. (n=5 per group) E, Heatmap of top 10 genes in each aging module following TGFB1 compared to DMSO in LPS treated microglia. F, Average gene expression changes for each aging module represented as log2 fold change of TGFB1 treatment over DMSO. (one sample T-test with the expected value of 0 (no change); *P<0.05, ***P<0.001, ****P<0.0001). G, Representation of the microglia aging trajectory over the UMAP plot highlighting the stage where CX-5461 modulates the trajectory. H, Representative images of S6 (magenta) and Iba1 (cyan) staining in the hippocampus of 6- and 24-month-old mice and quantification across aging. (n=3 mice per group; one-way ANOVA with Tukey’s post hoc test; *P<0.05, **P<0.005, ****P<0.0001). I, Schematic of the heterochronic parabiosis model and quantification of hippocampal microglia expression of translation module from isochronic young (IY) and heterochronic young (HY) adult parabionts. Data derived from Pálovics et al31. (one sample T-test with the expected value of 0 (no change); **P<0.01). J, Top gene ontology terms of genes with significantly decreased expression in bulk microglia RNA-Seq following CX-5461 treatment compared to control (DMSO) in LPS treated microglia. (n=3 per group) K, Heatmap of top 10 genes in each aging module following CX-5461 compared to DMSO in LPS treated microglia. L, Average gene expression changes for each aging module represented as log2 fold change of CX-5461 treatment over DMSO. (one sample T-test with the expected value of 0 (no change); *P<0.05, **P<0.01, ****P<0.0001).

Intermediate states of microglia aging act as checkpoints on inflammatory progression.

A, Quantification of percentage of Tgfb1 RNAscope signal within IBA1 cells. B, Quantification of IHC of pTGFBR1 signal in IBA1 cells across ages. (n=4-5 mice per group; one-way ANOVA; *P<0.05) C, Dotplot of the expression values of TGFB1 signaling components from scRNA-Seq of aging hippocampal microglia (6-, 12-, 18-, and 24-month-old). Percent of cells expressing the gene and average normalized expression are represented. D, Dotplot of the expression values of TGFB1 signaling components in young (Y), aged (A), isochronic young (IY), and heterochronic young (HY) parabiont microglia. Data is from Palovics et al61. Percent of cells expressing the gene and average normalized expression are represented. UMAP plot of Tgfb1 scRNA-Seq colored by genotype. E, Heatmap of gene expression changes in the top 10 genes in each aging module induced by LPS. F, Overlap between microglia gene expression changes induced by LPS and aging. (ξ2<2.2e-16). G, PCA plot of pharmacological manipulations with or without LPS treatment.

Targeting age-related changes in microglia-derived TGFB1 promotes microglia advancement along inflammatory trajectories in the hippocampus and impairs cognition.

A, Schematic of in vivo manipulation schema of Module 2. Adult (7-8 months) littermate Cx3cr1-Cre-ER+/wt; Tgfb1flox/flox (cKO), Cx3cr1-Cre-ER+/wt; Tgfb1flow/wt (Het), and Cx3cr1-Cre-ER+/wt; Tgfb1wt/wt (WT) mice were administered tamoxifen and subject to hippocampal microglia scRNA-Seq and behavior analysis two months later. B, UMAP plot of microglia separated into transcriptional clusters. (n = 2 pools of 3 animals per genotype) C, Stacked bar plot of the normalized relative percentage of cells of each genotype in the identified clusters. D, Dot plot of expression of aging module markers sorted by genotype. Percent of cells expressing the gene and average normalized expression are represented. E,F, Average gene expression changes for each aging module represented as log2 fold change of either Het over WT (E) or KO over WT (F). (one sample T-test with the expected value of 0 (no change); *P<0.05, ****P<0.0001). G, Novel object recognition task. (Top) Diagram of the training and testing phases of the novel object recognition paradigm. (Bottom) Quantification of the NOR testing phase represented as percentage of time spent with the novel object (over the total time spent interacting with the objects). (n=9-13 per genotype; one sample T-test with the expected value of 50 (equal time spent with each object); ***P<0.001)(differences between groups determined by one-way ANOVA; **P<0.01). H, Contextual fear conditioning. (Top) Diagram of the fear conditioning paradigm. (Bottom) Quantification of the percentage of time mice froze in the contextual fear conditioning testing phase. (n=14-16 per genotype; one-way ANOVA; *P<0.05). Data are shown as mean±s.e.m.

Targeting age-related changes in microglia-derived TGFB1 promotes microglia advancement along inflammatory trajectories in the hippocampus and impairs cognition.

A, Representative images and quantification of IBA1 (cyan)/CD68 (red)-positive microglia in wildtype, Tgfb1 cHet, and Tgfb1 cKO hippocampi. (n=3-5; one way ANOVA with Tukey post hoc test; *P<0.05) B, Dotplot of the expression values of TGFB1 signaling components from scRNA-Seq of Tgfb1 WT, cHet, and cKO microglia. (n = 2 pools of 3 animals per genotype) C, UMAP plot of Tgfb1 scRNA-Seq colored by genotype. D, FACS plots of gating strategy for microglia isolation for RNA-Seq in cHet and cKO hippocampi. Quantification of the flow cytometry analysis of CD11b and CD45. (n=3-5; mixed effects analysis; *P<0.05, ****P<0.0001). E, Example histogram and quantification of CD48 in control and Tgfb1 cKO FACS-sorted microglia. (n=3-5 per group; T-test; ****P<0.0001). F, Top gene ontology terms associated with genes significantly increased in microglia bulk RNA-Seq in Tgfb1 cKO samples (n=3-5 samples per genotype). G, Top gene ontology terms associated with genes significantly decreased in microglia bulk RNA-Seq in Tgfb1 cKO samples. H, Overlap between concordant microglia gene expression changes induced by Tgfb1 knockout and aging. (ξ2<2.2e-16). I, Heatmap of gene expression of the top 10 genes in each aging module in control and Tgfb1 cKO microglia. J, Average gene expression changes for each aging module represented as log2 fold change of Tgfb1 cKO over control. (one sample T-test with the expected value of 0 (no change); ****P<0.0001). K, Freezing during the cued fear conditioning phase of the fear conditioning test. The percent freezing was calculated for the last 30 seconds of the test following the conditioned stimulus. (n=13-16 per genotype) L, Diagram and results of Y-maze test. Results are presented as preference for the novel arm versus the trained armed (time in novel arm/ (time in novel arm + trained arm) x 100). (n=14-16 per genotype) M, Percent time in the center of the open field and distance traveled during the open field test. (n=14-16 per genotype) Data are shown as mean±s.e.m.