Exercise enhances learning and memory in aged mice
Physical exercise has garnered extensive empirical support for its efficacy in augmenting learning memory and attenuating cognition [6, 7]. To investigate the effects of exercise on learning and memory in aged mice, a natural aging mouse model was established, incorporating an exercise intervention. Eighteen 15-month-old C57BL/6J mice were selected and divided into an aging group (n = 9, O-Ctrl) and an aging exercise group (n = 9, O-Ex). Additionally, 3-month-old young mice were included as a control group (n = 9, Y-Ctrl). The Y-Ctrl and O-Ctrl groups were kept under standard conditions for 8 weeks. Meanwhile, the O-Ex exercise group was subject to an 8-week exercise protocol, as outlined in Fig. 1A.
Following the exercise regimen, a Barnes maze test was administered to all three mouse cohorts to evaluate alterations in spatial learning and memory due to aging and exercise [19]. Data analysis revealed a marked elevation in mean escape latency for the O-Ctrl group compared to Y-Ctrl (P < 0.001, see Fig. 1C). The low-intensity exercise regimen resulted in a significant reduction in mean escape latency in the O-Ex group compared to their O-Ctrl counterparts (P < 0.01, Fig. 1C). Noticeably, the low-intensity exercise regimen did not significantly alter the body weight of the aging mice (Fig. 1B). These outcomes suggest that exercise may mitigate the cognitive declines associated with aging.
Single-cell sequencing identifies ten cell types in mouse hippocampus
The hippocampus, a pivotal neural structure integral to learning and memory consolidation, exhibits various age-related alterations [20]. Research has established that aerobic exercise can augment synaptic plasticity within the hippocampus, thereby yielding enhancements in cognitive functions, memory, and mental health [21–24].To further elucidate the mechanisms, especially cell type dynamics, by which exercise may decelerate aging-related changes, hippocampal tissues were harvested from distinct mouse cohorts. Single-cell transcriptomic sequencing was then performed on these tissue samples to investigate gene expression profiles across different groups, as detailed in Fig. 2A. This analytical endeavor resulted in the acquisition of 21,537 cells, with the following distribution across the experimental groups: 6,257 cells were derived from the Y-Ctrl group, 7,906 cells from the O-Ctrl group, and 7374 cells from the O-Ex group.
Considering the premise that cells sharing comparable gene expression profiles demonstrate similar functional characteristics, we classified the cells into clusters based on their gene expression patterns. This was followed by rigorous data quality control and preprocessing. For dimensionality reduction, principal component analysis (PCA) was utilized, and the resultant data were visualized via uniform manifold approximation and projection (UMAP) for effective single-cell cluster categorization. This approach led to the identification of 20 distinct optimal cell subclusters, as delineated in Supplementary Figure S1E.
To further narrow down the cell types comprehensively, we utilized the lmmgen reference dataset and literature-derived marker genes [25], the correlation between the expression profiles of the analyzed cells and the reference dataset was computed. Ten cell types with different gene marker profiles were identified (Figs. 2B with violin plot and 2C with heatmap form. Supplementary Figure S1F). In addition, t-SNE plot, based on maximal correlation coefficients with the reference dataset, was used to identify and visualize cell types in the hippocampus across different groups [26–30]. The t-SNE enabled the identification of the same ten distinct cell types in hippocampus (Fig. 2D): arachnoid barrier cells (ABCs) marked by Cdh1, astrocytes/neural stem cells (AST/NSCs) identified by Gfap/Aldoc expression, choroid plexus cells (expressing SIc4a5), endothelial cells (characterized by Flt1), ependymal cells (AK7), fibroblasts (FCs) identified by Slc6a13, microglia (Cx3cr1), neurons (Grin2b), oligodendrocytes(Hapln2), and oligodendrocyte progenitor cells (Cspg4).
Exercise reverses the dynamics of cell populations induced by aging
To investigate the aging-related dynamics of cell types induced by exercise. We conducted the comparative analysis of cellular composition among different groups, the consistency of hippocampus cell types among Y-Ctrl, O-Ctrl, and O-Ex mice was noted, with minimal variations in their distribution (Fig. 3A). The study further delved into the impact of exercise on hippocampal cell type proportions by examining the cellular makeup in the hippocampal tissues of the three mouse cohorts.
Quantitative analysis indicated an increase in the percentage of astrocytes/neural stem cells (AST/NSCs) and a concurrent decrease in neuronal populations with aging(O-Ctrl) compared with the Y-Ctrl group (Fig. 3B). Strikingly, post-exercise intervention revealed a reverse of cell types: a decrease in AST/NSCs and an increase in neuronal proportions in the O-Ex group compared with the O-Ctrl group. Remarkably, the cellular proportions in the O-Ex group closely mirrored those observed in the O-Ctrl cohort. Dysfunctions in astrocytes triggered by genetic mutations, RNA, and protein level alterations, and or environmental stress changes, can lead to central nervous system (CNS) imbalances, fostering neurodegeneration and contributing to neurodegenerative disease pathogenesis [31]. Our single-cell sequencing data suggests the decline in the astrocyte (AST) cell ratio and increase of neuron proportions following exercise, might play a crucial role in the improvement of learning and memory of aging mice induced by exercise.
To validate these sequencing findings, immunofluorescence (lF) assays were employed to quantitatively assess astrocyte numbers in the hippocampal region across different murine cohorts. Astrocyte numbers in the hippocampal CA1 area have been reported to be strongly associated with aging. The statistical analysis indicated a marked decrease in glial fibrillary acidic protein(GFAP)-positive astrocytes in the hippocampal CA1 area in the O-Ex mice compared to the O-Ctrl group (P < 0.05). Conversely, the difference in GFAP-positive astrocyte counts in the hippocampal CA1 zone between the O-Ex and Y-Ctrl groups was not statistically significant. This supports the notion that exercise can counteract the age-related escalation in AST cell numbers, as detailed in Figs. 3C-E.
Beyond astrocytes (AST), the age-associated neuronal decline was shown in our data. Given the key role of neurons in inducing reduced dendritic density, primarily via the diminution of synaptic interconnections [32–34]. We embarked on an in-depth analysis of neuron subtypes under aging and exercise-induced conditions. We classified neurons into 17 distinct clusters (Figs. 3F and 3G), observing notable differential expression in clusters 6, 7, 9, and 11. Post-aging, clusters 6 and 7 demonstrated upregulated expression, which was reversed following exercise intervention. In contrast, clusters 9 and 11 exhibited an inverse expression pattern. Leveraging established literature [35], we identified clusters 5, 9, 13, and 15 as predominantly GABAergic neurons, clusters 1,2, 3, 4, 6, 7, 8, 10, 11, 12, and 16 as glutamatergic neurons (GLUT), and specifically, cluster 7 as calretinin-positive neurons (CR). These categorizations were pivotal for further detailed investigation and analysis.
Aging-related gene expression profiles in astrocytes and neurons induced by exercise
Focusing on astrocytes/neural stem cells (AST/NSCs) and neurons, we acknowledged their critical roles in nervous system functionality, particularly considering the pronounced alterations in cell type composition identified in our preliminary findings. Our objective was to delineate the impact of exercise on the relative proportions and intracellular dynamics of AST/NSCs and neurons. To achieve this, a differential gene expression analysis was executed.
This analysis illuminated that both aging and exercise exert a significant regulatory effect on the gene expression profiles within AST/NSCs (Fig. 4A). In light of the previously identified subpopulation dynamics in neuron numbers, characterized as GABAergic, glutamatergic, and calretinin-positive neurons (GABA, GLUT, and CR)(Figs. 3F and 3G), we proceeded with an exhaustive gene expression analysis. We evaluated the top 10 genes exhibiting differential expression in AST/NSCs, GABAergic neurons (GABA), and GLUT neurons in aged mice, both with and without exercise intervention. The results of these analyses were systematically presented as volcano plots (Fig. 4B. Supplementary Figure S2A, 3B), offering a comprehensive view of the genetic shifts influenced by aging and exercise.
We next delved into an examination of the alterations in cellular pathways and processes associated with aging and post-aging exercise using gene set enrichment analysis (GSEA). This analysis revealed that the gene expression modifications induced by aging and exercise across various cell types were primarily concentrated in areas such as developmental processes, immune system functions, metabolic pathways, translation regulator activity, and other aging-associated pathways (as shown in Fig. 4C, Supplementary Figure S2C, 3D).
For a more nuanced understanding of the specific regulatory mechanisms at play, Gene Ontology (GO) enrichment analysis was applied to genes implicated in senescence-associated pathways post-enrichment. These were then juxtaposed with the top 5 and top 20 differentially expressed genes in AST/NSCs, GABAergic (GABA), and glutamatergic (GLUT) neurons. This rigorous analysis identified genes demonstrating both universal and cell type-specific expression changes after aging and exercise in these three neuronal categories(Figs. 4D, 4E). As depicted in Fig. 4D, our findings suggest that these genes are likely instrumental in modulating senescence across AST/NSCs, GABA, and GLUT cells, and in mediating the senescence-delaying impacts of physical exercise.
In our investigation, we extended our focus to the identification of key molecular players implicated in the regulation of aging across AST/NSCs, GABA, and GLUT cell types. Notably, genes such as Apoe, Cst3, and Rnpc3 emerged as significant, having a shared role in the pathways relevant to all three cell types, as delineated in Fig. 4E.
To elucidate the distinct genetic alterations linked to aging and subsequent exercise intervention in AST, we pinpointed 10 genes within AST that demonstrated a decrease in expression due to aging, but an increase following exercise. In contrast, 23 genes exhibited an inverse expression pattern (as shown in Fig. 4F). A comprehensive review of existing literature revealed that six of these genes-Apoe, Fkbp5, Ccnd3, AKT3, Zbtb16, and SIc24a4-are implicated in cerebral senescence. For instance, FKBP5 has been identified as a key gene associated with various neurological disorders, including Parkinson's Disease (PD), Alzheimer's Disease (AD), posttraumatic stress disorder, and schizophrenia [36–42]. In the context of PD, FKBP5 has been observed to promote neuronal cell death via interaction with PTEN-induced putative kinase43. Additionally, genes like APOE, AKT3, Ccnd3, and Zbtb16 have been linked to neuroinflammatory responses and apoptotic processes and exhibit modulated expression in AD and cognitive dysfunctions[10, 44–46].
Initially, to validate the mRNA expression levels of the six identified genes Apoe, Fkbp5, Ccnd3, AKT3, Zbtb16, and SIc24a4-in the mouse hippocampus, we employed qPCR and the statistical analysis indicated significant changes: in the exercised mice, mRNA levels of Fkbp5, Ccnd3, AKT3, Zbtb16, and SIc24a4 were reduced, whereas Apoe mRNA levels were elevated compared to the aging group (P < 0.01). This concordance with our sequencing data (Fig. 4G)substantiates our initial findings. Further, we extended our validation to the protein level, employing Western blot analysis to assess the protein expression of Fkbp5, Ccnd3, and Zbtb16 in the hippocampus. This analysis revealed an increase in the expression of these proteins following aging, with a subsequent decrease post-exercise (Figs. 4H-M). Based on these observations, we postulate that exercise may confer neuroprotective effects, enhancing learning and memory capabilities and decelerating brain aging, through the downregulation of these aging-associated genes in astrocytes.
Investigating aging-related transcriptional regulatory networks induced by exercise
To elucidate the complex transcriptional regulatory networks among aging-associated genes that are downregulated following exercise (O-Ex vs. O-Ctrl). Utilizing the String database, we probed into the collective and synergistic effects of these genes. Further, we employed the Cistrome database to identify six common transcription factors associated with these differentially expressed aging-related genes. These factors potentially act as upstream regulators, orchestrating the expression of the six identified aging-related genes.
Intrigued by the possibility that genes like Apoe, Ccnd3, and AKT3 may themselves be involved in transcriptional regulation, we extended our inquiry through additional analyses in the String database. This led to the discovery of a dynamic regulatory network comprising these six genes (Figs. 5A) and their shared transcriptional regulators (Figs. 5B).
This investigation posits that physical exercise could intricately and synergistically influence the aging trajectory of astrocytes. It does so by modulating key aging-related genes, namely Apoe, Fkbp5, Ccnd3, AKT3, Zbtb16, and SIc24a4, along with their co-regulatory transcription factors. Such modulation is hypothesized to impact the hippocampal aging process and potentially regulate cognitive functions, including learning and memory, in aged populations.