Endurance Exercise Induces Molecular and Cellular Changes in Adult Mouse Spinal Cord.

(A) Experimental design: Adult male C57BL/6 mice underwent six weeks of voluntary wheel running (TRAIN) or remained sedentary (SED), followed by performance testing and spinal cord dissection (n=15 per condition). Then a subset of the cohort (n=4 per condition) was used for bulk proteomics and single-nucleus RNA sequencing (snRNA-seq). (B) Average daily running distances of TRAIN mice over the intervention period. Each data point reflects an individual mouse; error bars denote mean ± SEM (n = 16 mice). (C) Treadmill-based endurance performance measured pre- and post-intervention in SED and TRAIN groups. Bars depict mean ± SEM; individual data points shown. ****p < 0.0001, two- way ANOVA with multiple comparisons (Fisher’s LSD test). (D) Volcano plot: Differentially expressed proteins in lumbar spinal cord (TRAIN vs. SED); proteins with p < 0.05 and |log₂FC| >= 0.25 are highlighted in blue and top differentially regulated (DR) proteins in red. (E) Top Gene Ontology (GO) terms for DR proteins: upregulated (yellow), downregulated (blue). (F) Heatmap of all significant (p < 0.05 and |log₂FC| >= 0.5) differentially expressed proteins (scaled by z-score). (G) UMAP visualization of major spinal cord cell populations from snRNA-seq, colored by cell identity and split by condition (SED, TRAIN). (H) Quantification of cell-type abundance (percent of total nuclei) by group. Bars depict mean ± SEM; n=4 per group, individual data points shown. No significant changes observed; two-way ANOVA with Šídák’s multiple comparisons test. (I) Bar plot showing number of differentially expressed genes (DEGs, up/down) in each main cell-type after training. Genes were filtered for |log₂FC| > 0.5 and q < 0.01. Abbreviations: SED, sedentary; TRAIN, voluntary exercise group; UMAP, Uniform Manifold Approximation and Projection; FC, fold-change; SEM, standard error of the mean. Panel A created with BioRender.com.

Glial Populations Mediate Cell-Type–Specific Transcriptional Remodeling and Intercellular Network Reorganization After Endurance Exercise.

(A–C) Dot-plot of top gene set enrichment analysis (GSEA) terms associated with the differentially expressed genes up-regulated in trained spinal cords for oligodendrocytes (A), oligodendrocyte precursor cells (OPCs; B), and Astrocyte-1 (C) clusters. (D) Heatmap summarizes the overlap of key GSEA terms across selected glial cell-types after exercise training, highlighting shared pathways of neural adaptation. (E–F) Heatmaps illustrate expression of representative genes upregulated in oligodendrocytes, grouped by synaptic signaling (E) and axon development (F). Expression levels are indicated by color (red = upregulated, blue = downregulated) and grouped by sedentary (SED) or trained (TRAIN) state. All genes displayed were filtered for |log₂FC| > 0.5 and q < 0.01. (G–H) Heatmaps show top upregulated genes in Astrocyte-1 (G) and OPCs (H) after training, with relative expression indicated by color scale across groups. (I–J) Dot plots depict communication probabilities for selected ligand–receptor interactions: WNT, Pleiotrophin (PTN), and Nectin signaling from cholinergic neurons to various spinal cord cell-types (I), and JAM and FGF signaling from microglia-2 to all spinal cord populations (J), under sedentary (SED) and trained (TRAIN) conditions. Dot color represents communication probability all dots reflect significant interactions (p < 0.01).

Acute Exhaustive Exercise Induces Distinct Temporal and Training-Dependent Proteomic and Cellular Remodeling in Spinal Cord.

(A) Experimental design: Sedentary (SED) and endurance-trained (TRAIN) mice subjected to a single session of treadmill-based exhaustion exercise, with lumbar spinal cords collected at 6- and 24-hours post-exercise; n = 4 mice per group/timepoint. (B–E) Volcano plots from bulk proteomics showing differentially expressed proteins (DEPs) following exercise. (B) SED 6h vs. SED baseline, (C) SED 24h vs. SED baseline, (D) TRAIN 6h vs. TRAIN baseline, (E) TRAIN 24h vs. TRAIN baseline. Proteins with p < 0.05 and |log₂FC| >= 0.25 are highlighted in blue and top differentially regulated proteins in red. Numbers reflect total significantly up- or downregulated proteins per comparison (p < 0.05, |log₂FC| > 0. 5). (F) UMAP plots of integrated single-nucleus RNA sequencing data illustrate clustering and abundance of major spinal cord cell-types split across all six experimental conditions. (G) Spatial overlays of nuclei from each condition on the integrated UMAP, with green representing SED and blue representing TRAIN groups, to visualize cluster distribution and identify cell populations that diverge in UMAP space across conditions. (H–I) Quantification of non-neuronal cell-type proportions across conditions: (H) SED baseline, 6h, 24h; (I) TRAIN baseline, 6h, 24h. Proportions plotted as percent of total nuclei (mean ± SEM, n = 4/group). * p < 0.05, ns = not significant, 0.05<p<0.1 (two-way ANOVA with Tukey’s multiple comparisons test). Panel A created with BioRender.com.

Training Status Tunes Glial Transcriptional Dynamics after Acute Exercise.

(A–C) Bar plots show the number of significantly upregulated (dark blue) and downregulated (light blue) differentially expressed genes (DEGs) in oligodendrocytes (A), astrocyte-1 (B), and oligodendrocyte precursor cells (OPCs; C) at 6h and 24h post-acute exercise, comparing sedentary (SED) and trained (TRAIN) mice relative to respective baselines. Genes were considered significant when |log₂FC| > 0.5 and q < 0.01. (D) Gene set enrichment analysis (GSEA) for oligodendrocyte DEGs highlights top enriched functional categories—metabolic, synaptic, structural, signaling, and transport—across timepoints and training states. Circle size reflects statistical significance (–log10p), color indicates up- (yellow) or downregulation (blue). (E–F) Line plots depict normalized temporal expression profiles of representative oligodendrocyte DEGs: (E) metabolic genes and (F) synaptic signaling genes across baseline (PRE), 6h, and 24h. Shaded areas represent ± SEM. (G) GSEA for astrocyte-1 DEGs showing dynamically enriched metabolic, synaptic, and signaling pathways by training state and timepoint. (H–I) Line plots depicting expression trajectories for selected metabolic (H) and synaptic signaling (I) genes in astrocyte-1 across conditions. All gene expression data are log-normalized average expression. n = 4 per timepoint and condition.

Endurance Training Modulates Dynamic Intercellular Communication Networks in Spinal Cord Following Acute Exercise.

(A and B) Bar graphs quantifying the total number (left) and aggregate strength (right) of predicted intercellular communications among spinal cord cell-types inferred by CellChat, across baseline, 6h, and 24h after acute exercise for sedentary (SED, green, A) and trained (TRAIN, blue, B) mice. (C and D) Heatmaps illustrating overall strength of communication for all major signaling pathways for each sender cell-type, highlighting dynamic reorganization of intercellular signaling at baseline, 6h, and 24h in SED (C) and TRAIN (D) conditions. Selected pathways are highlighted with dashed lines. (E–G) Dot plots showing top ligand–receptor interactions mediating Semaphorin (E), Slit (F), and Acetylcholine (G) signaling from cholinergic neurons to glial, vascular, and neuronal targets across conditions. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01), dark grey dots are not significant (p > 0.01).

Glial Populations Fine-Tune Intercellular Communication Networks after Exercise in a Training-Dependent Manner.

(A) Dot plots display the probability of JAM family ligand–receptor interactions from astrocyte-1 and astrocyte-2 to multiple cell-types by group and timepoint. Top: SED, Bottom: TRAIN. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01), dark grey dots are not significant (p > 0.01). (B–D) Line plots showing mean (±SEM) expression of key JAM (Jam2, Jam3, F11r), PDGF (Pdgfd, Pdgfrb), and FGF (Fgfr2) pathway genes in astrocyte-1 (B), astrocyte-2 (C), and microglia-2 (D), comparing SED and TRAIN animals across PRE, 6h, and 24h post-exercise. (E) Dot plots ligand–receptor probabilities for PDGF signaling from astrocyte-1, microglia-2, and oligodendrocytes to selected target populations across timepoints and condition. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01), dark grey dots are not significant (p > 0.01). (F) Line plots showing mean (±SEM) expression profiles of key FGF (Fgfr1, Fgfr2, Fgfr3) and PDGF (Pdgfd, Pdgfrb) pathway genes in astrocyte-2 in SED and TRAIN conditions. (G) Dot plots of key semaphorin, JAM, and FGF ligand–receptor interactions from microglia-2 to various targets by group and timepoint. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01).

Schematic Summary of Exercise-Induced Transcriptional Programs and Their Temporal Dynamics in the Healthy Lumbar Spinal Cord.

(A) Summary of major transcriptional processes modulated by endurance exercise across spinal cord cell types. Arrows indicate upregulated or downregulated pathways identified by snRNA-seq, including metabolic coordination, synaptic support, lipid turnover, cytoskeletal remodeling, and inflammatory or growth-associated signaling. Each cell type is represented by its dominant gene programs and functional themes derived from differential expression and pathway analyses. Abbreviation: PNS, peripheral nervous system. (B) Temporal trajectories of representative gene programs illustrating exercise-induced dynamics in sedentary and trained mice. Curves depict the relative activation of three key functional categories across baseline, 6 h, and 24 h after an acute bout of exercise. Training accelerates the induction and resolution of transcriptional responses, whereas sedentary animals show delayed and more persistent activation. Together, these schematics illustrate the cell-type specificity and temporal logic of spinal cord adaptation to exercise and highlight the influence of prior training on glial and neuronal plasticity. Created with BioRender.com.

Endurance Exercise Selectively Preserves Lean Mass and Modifies Spinal Cord Cell-Type Landscape.

(A–C) Longitudinal assessment of whole-body composition (A, total mass; B, relative lean mass; C, relative fat mass) in SED and TRAIN groups pre- and post-intervention. Data represent mean ± SEM; individual biological replicates shown. **p < 0.01; ****p < 0.0001; ns, not significant; two-way ANOVA with multiple comparisons (Fisher’s LSD test). (D) Violin plots displaying snRNA-seq quality metrics across identified spinal cord cell types: number of genes (nFeature_RNA), unique molecular identifiers (nCount_RNA), and mitochondrial proportion (percent.mt). (E) Dot plot summarizing top marker gene expression of each identified celltype and subcluster in the spinal cord; size of dots indicates percent of cells expressing each marker, color gradient reflects average normalized expression.

Distinct and Overlapping Transcriptional Programs Across Spinal Cord Cell Types Following Endurance Training.

(A) UpSet plot demonstrating limited overlap among differentially expressed genes upregulated in main glial cell types following endurance training. All DEGs analyzed were filtered for |log₂FC| > 0.5 and q < 0.01. (B–C) Heatmaps show relative expression of upregulated genes involved in ion transport (B) and cellular signaling (C) in oligodendrocytes, comparing sedentary (SED) and trained (TRAIN) groups; upregulation is indicated in red and downregulation in blue. (D–G) Heatmaps display top DEG expression profiles in microglia-2 (D), Inhibitory Neurons-1 (E), Inhibitory Neurons-5 (F), and Excitatory Neurons-1 (G). (H–K) Dot plots present selected gene set enrichment analysis (GSEA) results for Microglia-2 (H), ExN-1 (I), InhN-1 (J), and InhN-5 (K), indicating significantly enriched pathways, gene counts, and statistical significance (represented by dot size and color). (L) Heatmaps summarizing overall outgoing signaling pathway patterns and major differences between SED and TRAIN spinal cords, with selected pathways highlighted (red arrows). (M) Interaction heatmaps illustrating differences in the number (left) and strength (right) of intercellular communications between SED and TRAIN, grouped by sending and receiving cell types.

CellChat Analysis Reveals Exercise-Driven Remodeling of Spinal Cord Cell–Cell Signaling Pathways.

(A–E) Dot plots display the top CellChat-inferred ligand–receptor interactions mediating signaling from cholinergic neurons (A), astrocytes (B–C), oligodendrocytes (D), and microglia-2 (E) to diverse recipient cell populations in sedentary (SED) and trained (TRAIN) spinal cords. Notable exercise-responsive signaling axes highlighted include Neurexin (Nrxn1/2), Laminin (Lama2, Lamc1), and Semaphorin (Sema4) pathways. Dot color represents communication probability, and dot size indicates statistical significance (all plotted interactions: p < 0.01)

Acute Exercise Modulates Molecular Pathways and Cell-Type Proportions in a Training- and Time-Dependent Manner.

(A–D) Gene Ontology (GO) enrichment analysis of significantly up- and down-regulated proteins following acute exercise as determined by bulk proteomics: (A) SED at 6h post-exercise, (B) SED at 24h, (C) TRAIN at 6h, (D) TRAIN at 24h. (E) Violin plots depict key quality control metrics from integrated single-nucleus RNA-seq, including total detected genes (nFeature_RNA), UMI count (nCount_RNA), and percentage mitochondrial reads (percent.mt) across groups and conditions. (F–G) Quantification of neuronal cell-type proportions across conditions: (F) SED baseline, 6h, 24h; (G) TRAIN baseline, 6h, 24h. Proportions plotted as percent of total nuclei (mean ± SEM, n = 4/group). *p < 0.05, **p < 0.01, ****p < 0.0001, ns = not significant (two-way ANOVA with Tukey’s multiple comparisons test).

Training Modulates Temporal and Cell-Type–Specific Glial Transcriptional Programs after Acute Exercise.

(A–C) Bar plots displaying the number of significantly upregulated (dark blue) and downregulated (light blue) differentially expressed genes (DEGs) in microglia-2 (A), astrocyte-2 (B), and microglia-1 (C) at 6h and 24h post-acute exercise in sedentary (SED) and trained (TRAIN) groups. (D–G) UpSet plots showing overlap of DEGs between main glial subtypes for SED-6h vs SED (D), SED-24h vs SED (E), TRAIN-6h vs TRAIN (F), and TRAIN-24h vs TRAIN (G), highlighting generally limited overlap. (H) Gene set enrichment analysis (GSEA) for OPC DEGs highlights top enriched functional categories—metabolic, nuclear, synaptic, transport, signaling, and structural—across timepoints and training states. Circle size reflects statistical significance (–log10 p), color indicates up- (yellow) or downregulation (blue). (I) GSEA of microglia-2 DEGs for immune, phagocytic, and regulatory categories across timepoints and training states. Circle size reflects statistical significance (–log10 p), color indicates up- (yellow) or downregulation (blue). (J–K) Line plots showing normalized expression trajectories (mean ± SEM) for selected DEGs: representative metabolic and signaling genes in OPCs (J), and immune-related genes in microglia-2 (K), for SED vs. TRAIN animals across baseline, 6h, and 24h post-exercise. All gene expression data are log-normalized average expression. n = 4 per timepoint and condition.

Receptor-Level Regulation Underlies Rapid Remodeling of Glial and Neural Communication After Training.

(A and B) Line plots displaying mean (±SEM) expression of Semaphorin and Slit pathway receptors in Astrocyte-1 (A) and Astrocyte-2 (B) across experimental groups and timepoints. (C and D) Line plots displaying mean (±SEM) expression of Semaphorin, Slit, and cholinergic ligand and receptor genes in cholinergic neurons under SED and TRAIN conditions. (E and H) Dot plots indicating CellChat-inferred probabilities for FGF (E), and PTPR (H) ligand– receptor interactions from cholinergic neurons to all major neural and glial cell types by group and timepoint. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01), dark grey dots are not significant (p > 0.01). (F and G) Line plots for OPCs (F) and Astrocyte-1 (G) showing mean expression (±SEM) of key FGF receptors (Fgfr1, Fgfr3) and PTPR family members (Ptprd, Ptprf, Ptprs) over time in SED and TRAIN.

Training State Selectively Regulates Glial and Microglial Communication with the Spinal Cord Niche.

(A–B) Line plots showing mean (±SEM) expression profiles of select CellChat ligand/receptor genes in OPCs (A) and oligodendrocytes (B) under SED and TRAIN conditions, illustrating differences in signaling dynamics. (C–D) Dot plots of Semaphorin ligand–receptor interactions sent from astrocyte-1 (C) and FGF ligand–receptor interactions from oligodendrocytes (D) to selected cell types. Top: SED, Bottom: TRAIN. Dot color represents communication probability, all colored dots are statistically significant (p < 0.01), dark grey dots are not significant (p > 0.01).