Exercise rejuvenates microglia and reverses T cell accumulation in the aged female mouse brain

Abstract

Slowing and/or reversing brain ageing may alleviate cognitive impairments. Previous studies have found that exercise may mitigate cognitive decline, but the mechanisms underlying this remain largely unclear. Here we provide unbiased analyses of single-cell RNA sequencing data, showing the impacts of exercise and ageing on specific cell types in the mouse hippocampus. We demonstrate that exercise has a profound and selective effect on aged microglia, reverting their gene expression signature to that of young microglia. Pharmacologic depletion of microglia further demonstrated that these cells are required for the stimulatory effects of exercise on hippocampal neurogenesis but not cognition. Strikingly, allowing 18-month-old mice access to a running wheel did by and large also prevent and/or revert T cell presence in the ageing hippocampus. Taken together, our data highlight the profound impact of exercise in rejuvenating aged microglia, associated pro-neurogenic effects and on peripheral immune cell presence in the ageing female mouse brain.

Keywords: T cells; active place avoidance; border‐associated macrophages; brain ageing; cognition; disease‐associated microglia; neuroinflammation.

FIGURE 1

Single cell RNA‐sequencing of cell populations isolated from the young or aged brain following exercise. (a) Overview of experimental design. Cells were isolated from the hippocampi of young sedentary mice (3 months old, “Young SED”), aged sedentary mice (18 months old; “Aged SED”) or aged mice given 21 days of voluntary access to a running wheel (“Aged RUN”). In Aged RUN mice, the exercise period was followed by 14 days of rest prior to sacrifice. Single, live cells from dissociated hippocampi were sorted by FACS and processed to create single‐cell (sc)RNA‐seq libraries (n = 3 female littermates per condition). (b) Uniform Manifold Approximation and Projection (UMAP) plot of 9336 cells from the three experimental groups (Young SED, Aged SED, Aged RUN) that passed quality control thresholds (Figure S1). Cells clustered into nine distinct cell types, which were annotated based on their expression of canonical markers (as indicated). (c) Proportions of identified nine cell types in Young SED, Aged SED and Aged RUN mice. (d) Forest plot showing Log₂FC of cell numbers between experimental conditions. The mean and 95% confidence interval (error bar; obtained from permutation) is shown for each cell comparison. Significance was obtained using a Monte Carlo permutation test with an FDR <0.05 and an absolute log change >1 (corresponding to a doubling of abundance compared to the reference group and represented as dashed lines). Note the age‐related increase in T cell numbers. (e) Independent validation of scRNA‐seq data showing increased T cell abundance in the ageing hippocampus. Representative confocal image (left) and quantification (right) of parenchymal CD3^pos T cells in the mouse hippocampus. Note that the number of CD3^pos T cells is increased in Aged SED mice compared to Young SED mice, and that access to a running wheel (i.e Aged RUN) significantly attenuates this change. Data points represent individual mice (n = 5–7 per group). Data are represented as mean ± SEM. Statistics: one‐way ANOVA, followed by Bonferroni post‐hoc comparison with Geisser–Greenhouse correction. ***p < 0.001, ****p < 0.0001.

FIGURE 2

Exercise reverses the effect of ageing in microglia. (a) Bar plot showing the number of differentially expressed genes (DEGs) identified between Young SED and Aged SED mice (grey; gene markers of ageing), and between Young SED and Aged RUN (purple) in oligodendrocytes, microglia, astrocytes and endothelial cells. (b) Bar Plot showing the Log2FC in gene expression and their related pathways in Aged SED microglia (light grey; gene markers of ageing) and Aged RUN microglia (purple) compared to Young SED microglia. (c) Log fold change comparisons of gene expression in microglia. Each dot represents the log fold change of one of the DEGs identified between Young SED and Aged SED mice. The red line represents the equation x = y. The blue line corresponds to the best fit of the linear regression following the equation: log fold change (Young SED/Aged RUN) ~ log fold change (Young SED/Aged SED). Shaded area shows the 95% confidence interval of the fitted values. Coefficients, standard error and R² for the linear regression are the following: 0.437x ± 0.012, R² = 0.69. (d) Heatmap and diagram showing the number of significant ligand‐receptor (LR) pairs found between astrocytes, endothelial cells, microglia and oligodendrocytes from Aged SED and Aged RUN mice. (e) Schematic diagram showing specific LR pairs regulated with exercise in aged hippocampus, as predicted from cell–cell interaction analysis. Blue arrows show unique LR interactions that were significantly enriched in Aged SED but not Aged RUN mice; red arrows indicate significant LR pairs that were regulated by exercise in Aged RUN mice and not present in their Aged SED counterparts. (f) Validation of predicted changes in identified ligand‐receptor pair, APP‐CD74, in response to ageing and exercise. Note that CD74 expression in IBA1^pos cells is increased with ageing, but attenuated by exercise. CD74^pos cells were often found closely associated with APP^pos (i.e. 6E10‐stained) vascular structures and/or cells. Staining control (primary antibody omission) shows the absence of non‐specific secondary antibody binding. Scale bar = 50 μm. Statistics: one‐way ANOVA, followed by Bonferroni post‐hoc comparison with Geisser–Greenhouse correction. ***p < 0.001, ****p < 0.0001.

FIGURE 3

Microglial subtypes are differentially impacted by ageing and exercise. (a) UMAP plots showing the five identified sub‐clusters of brain myeloid cells, namely homeostatic microglia (green), inflammatory microglia (brown), disease‐associated microglia (DAM; red), an unknown microglia subtype (light grey), and border‐associated macrophages (blue) for the different experimental groups (Young SED, Aged SED and Aged RUN). (b) Heatmap of the top gene markers for each sub‐cluster. A full list of genes and publications of origin can be found within Table S6. (c) Stacked bar chart showing the proportion of each microglial/myeloid cell subtype for each of the experimental groups. (d) Dot plot representing coefficients of linear regressions fitted to each cluster using the 310 DEGs identified between the Young SED and Aged SED condition. The orange dashed line represents the coefficient identified when all subtypes were combined (Coefficients and standard error: 0.437 ± 0.012). Dot colours reflect microglia subtypes and/or states and BAMs, as specified in Figure 3a. Coefficients and standard errors are: 0.419 ± 0.036 (homeostatic microglia; green), 0.607 ± 0.042 (inflammatory microglia; brown), 0.278 ± 0.033 (disease‐associated microglia; red) and 0.932 ± 0.049 (BAMs, blue).

FIGURE 4

Benefits of exercise on cognitive abilities of aged mice remain with microglial depletion. (a) Overview of experimental timeline. Young adult (3‐month‐old) and aged (18‐month‐old) mice underwent active place avoidance (APA) testing over a course of 5 days (APA1). Aged mice were then placed on PLX5622‐containing chow (to deplete microglia), or control chow and re‐tested for acquisition of a new APA task (APA2) 21 days later. (b) Diagram of the visual cues used during APA testing prior to (APA1) and after (APA2) PLX5622 (or control chow) administration. The maroon triangle indicates the shock zone location. Note that different visual cues were used in APA2 versus APA1 to assess spatial learning (as opposed to task recall). (c) Shock zone entries during APA1 testing (10‐min trials/day). Aged mice had significantly more entries on day 5 compared to young adult mice (198% increase, t(115) = 2.79, p = 0.031, n = 12–13). (d) Percentage improvement in APA1 performance of individual mice (t = 2.50, df = 23, p = 0.012; minimum, 25% percentile, median, 75% percentile, maximum, Young SED: −16.67, 59.70, 86.34, 91.96, 100.00; Aged SED: −60.00, 24.16, 45.45, 51.67, 78.95, 138.90). (e) Schematic overview showing the split of aged sedentary mice used in APA1 into two groups, receiving either control chow (yellow), or PLX5622‐containing chow (to deplete microglia, pink). (f) Confocal images of IBA1 immunostaining showing effective depletion of microglia in the hippocampus of aged (18‐month‐old) mice that were fed PLX5622 (PLX) or control (CON) chow. Scale bar: 50 μm. (g) Entries into the shock zone during APA2 (10‐min trial/day) for aged mice given either control of PLX5622 chow (F[1,9] = 0.075, p = 0.79). (h) Percentage improvement in APA2 testing for aged mice with and without microglia (t = 0.31, df = 10, p = 0.77; minimum, 25% percentile, median, 75% percentile, maximum, Aged SED/control chow: −57.14, −33.83, 26.67, 70.38, 78.26; Aged SED/PLX5622: −169.20, −25.00, 11.11, 73.33, 83.33). (i) Overview of experimental layout and timeline. Aged (18‐month‐old) mice received either control or PLX5622‐containing chow (to deplete microglia) for 61 days. Mice were allowed to run for 21 days, allowed a 2‐week rest period, and then tested in APA 14 days after completion of the exercise paradigm. (j) Distance travelled during the habituation trial of APA testing (shock zone off; minimum, 25% percentile, median, 75% percentile, maximum, Aged SED/control chow: 85.21, 92.01, 100.5, 106.40, 125.30; Aged RUN/control chow: 86.10, 92.72, 103.00, 109.20, 127.40; Aged RUN/PLX5622 run: 68.33, 82.44, 98.38, 111.0, 114.0). (k) Total number of entries into the shock zone during APA testing (20‐min trial/day; F(2,44) = 3.81, p = 0.0297). (l) Percentage improvement in APA performance for individual mice, assessed by the change in entries on testing day 5 versus day 1 for individual mice (F[2,44] = 8.64, p = 0.0007; minimum, 25% percentile, median, 75% percentile, maximum, Aged SED/control chow: 38.46, −20.83, 0, 18.75, 50.0; Aged RUN/control chow: −44.44, 18.63, 39.57, 59.24, 86.96; Aged RUN/PLX5622: 7.69, 22.70, 47.70, 58.17, 96.30). (m) Schematic of experimental timeline, as detailed above, used to examine neuroplasticity and neurogenic effects of exercise. Aged (18‐month‐old) mice received either control or PLX5622‐containing chow (to deplete microglia) for 56 days. Mice were allowed 14 days of rest after 21 days of voluntary wheel running to allow newborn cells to differentiate into DCX^pos immature neurons. (n) Exercise (RUN) increased the number of synaptophysin (SYN) puncta in the hippocampus of 18‐month‐old mice compared to sedentary (SED) aged‐matched controls (F(2,12) = 12.88, p = 0.0010). PLX5622 did not significantly alter the increase in SYN observed in Aged RUN mice (Aged SED/Control chow vs Aged RUN/PLX5622, p > 0.99; n = 5/group; minimum, 25% percentile, median, 75% percentile, maximum: Aged SED/Control chow: 0.0016, 0.0018, 0.0496, 0.058, 0.065; Aged RUN/Control chow, 0.079, 0.086, 0.106, 0.110, 0.112; Aged RUN/PLX5622: 0.061, 0.073, 0.087, 0.105, 0.112). (o) Representative confocal images for synaptophysin (SYN) staining in the hippocampus of Aged SED/Control chow, Aged Run/Control chow and Aged RUN/PLX5622 mice. Scale bar = 50 μm. (p) Exercise (RUN) increased the number of immature DCX^pos neurons in 18‐month‐old mice treated with a control diet compared to sedentary age‐matched controls (SED; 2.64‐fold increase, t(26) = 2.84, p = 0.017, n = 8–11). PLX5622 significantly decreased the number of DCX^pos immature neurons in mice that underwent exercise (2.50‐fold decrease, t(26) = 2.81, p = 0.019, n = 6–11; minimum, 25% percentile, median, 75% percentile, maximum, Aged SED/control chow: 1.69 2.43, 3.07, 3.44, 4.05; Aged RUN/control chow, 2.25, 3.68, 6.23, 9.71, 19.69; Aged RUN/PLX5622: 0.57, 1.72, 3.16, 4.74, 5.13). (q) Representative confocal images of immature DCX^pos neurons in hippocampus of aged mice fed either control or PLX5622‐containing chow. Scale bar = 50 μm. Note the exercise‐induced increase in DCX^pos cells in the RUN condition, and the absence of this when microglia were depleted (PLX5622). Data are represented as mean ± SEM unless specified otherwise. Statistics: unpaired Student's t‐test (d, h), repeated two‐way ANOVA (c, g, k), or one‐way ANOVA (j, l, n, p) both followed by Bonferroni post‐hoc comparison with Geisser–Greenhouse correction. *p < 0.05, **p < 0.01, ***p < 0.001. Data points represent individual mice.