The UM-HET3 model was created by crossbreeding four inbred strains: BALB/cJ and C57BL/6J females with C3H/HeJ and DBA/2J males. Because each animal inherits a different mix of genes, the strain mirrors human genetic variability far better than the single-strain mice that dominate aging research, and it has become a workhorse of longevity studies. Yet, until now, its physiological responses to exercise were almost unexplored. Only one prior study had examined voluntary wheel running in these animals, and voluntary activity varies widely across strains and does not necessarily reproduce the effects of imposed, progressive training.

In the new experiment, 40-week-old male UM-HET3 mice were randomly assigned to an exercise-trained group of 17 animals or a sedentary group of 16. After a one-week acclimation period and treadmill familiarization, the trained mice ran five days per week for four weeks, completing 60-minute sessions at roughly 60 to 65 percent of their maximal work output, an intensity chosen because it elicits physiological responses comparable to human endurance exercise. Maximal exercise capacity was tested before and after the intervention using graded treadmill tests in which speed increased by 2.5 meters per minute every three minutes and incline rose by 5 degrees every nine minutes until exhaustion. Body weight, fat mass and lean mass were measured by quantitative magnetic resonance imaging, and fecal samples were collected at the end of the study for 16S rRNA gene sequencing in 13 mice per group.

The clearest physiological signal was in exercise performance. Both groups improved their maximal work between the pre- and post-intervention tests, but the magnitude of improvement was dramatically different: exercise-trained mice gained an average of 483.3 plus or minus 48.8 joules, compared with 194.7 plus or minus 52.0 joules in sedentary mice, a difference the authors report as statistically significant at P = 0.0003. The researchers attribute the unexpected improvement in sedentary animals, at least in part, to the heterogeneous genetic background of the UM-HET3 strain, which can influence how individual mice respond and adapt to repeated testing.

Body composition told a subtler story. Within both groups, total body weight, fat mass, lean mass and percentage body fat changed significantly from pre- to post-intervention, but exercise training did not significantly modify most of these measures between groups. The one exception was body fat percentage: trained mice showed a significant reduction within their own group, and after training they carried a lower body fat percentage than sedentary mice (P = 0.02). The authors note that this pattern echoes an earlier UM-HET3 study by Willow and colleagues, in which aging exerted a stronger influence on body composition than voluntary wheel running, and they suggest that age-related changes in body composition may outweigh the effects of exercise in aging mice.

The microbiome results were the study’s central focus. Sequencing yielded 407 amplicon sequence variants, and while the dominant phyla in both groups were Firmicutes and Bacteroidota, consistent with typical murine gut profiles, the community composition clearly diverged between trained and sedentary animals. Principal coordinate analysis based on Bray-Curtis dissimilarity showed significant separation between the groups along the first axis (P = 0.01), and PERMANOVA and ANOSIM confirmed the difference (adjusted P = 0.001). A PERMDISP test found no significant difference in within-group dispersion, indicating the separation was not an artifact of unequal variability. A hierarchical clustering dendrogram likewise split the animals into two major clusters corresponding to their training status.

Alpha diversity, by contrast, shifted only modestly. Exercise-trained mice showed a higher median Chao1 richness of 219.5 and a slightly higher Shannon diversity index (4.10 versus 3.948), but neither difference reached statistical significance. At the phylum level, trained mice had modestly higher relative abundances of Firmicutes (53.61 percent versus 52.68 percent) and Verrucomicrobiota (2.94 percent versus essentially zero in sedentary mice), and lower abundances of Bacteroidota (41.95 percent versus 45.13 percent) and Desulfobacterota (0.57 percent versus 1.56 percent). At the family level, Lachnospiraceae was more abundant in trained mice (16.05 percent versus 12.61 percent), and genus-level plots tracked changes in Lachnospiraceae NK4A136, Bacteroides, Lactobacillus, Muribaculaceae and an unclassified group.

Several of these taxa carry intriguing health associations, though the authors are careful to frame them as links rather than proven mechanisms. Verrucomicrobiota includes Akkermansia species, which have been associated with improved gut barrier integrity and metabolic health, whereas Desulfobacterota comprises sulfate-producing bacteria that may contribute to intestinal inflammation when present in excess. Lachnospiraceae are important producers of short-chain fatty acids, particularly butyrate, which supports intestinal barrier function, immune regulation and host metabolism. Trained mice also showed slightly higher Lactobacillus abundance, a genus with recognized probiotic properties, and some Lactobacillus strains have displayed ergogenic effects on fatigue resistance, recovery and inflammatory responses in prior work, feeding the idea of a gut-muscle axis.

Crucially, however, the correlation analysis came up empty. Spearman’s correlations between phylum-level relative abundances and the pre-to-post changes in body composition and exercise capacity revealed no statistically significant associations, although trends hinted at positive relationships with Verrucomicrobiota and negative ones with Desulfobacterota. The authors caution that relying on compositional changes alone is limited by the substantial functional redundancy of the gut microbiome: different bacterial communities can perform similar metabolic roles, so shifts in taxonomy do not necessarily translate into shifts in function.

The study also carries important limitations that shape how its findings should be read. The mice were 40 weeks old in a strain with a lifespan of 96 to 104 weeks, placing them in early-to-mid adulthood rather than old age, so the results may not extend to older populations. Only males were studied, despite a known female survival advantage in UM-HET3 mice, leaving open the possibility of sex-specific responses. Because fecal samples were collected only at the end of the intervention, baseline microbiome composition was unavailable, meaning pre-existing group differences cannot be excluded even though animals were randomized. In addition, mice were co-housed by treatment group, and microbial exchange between cage-mates can itself alter community composition, so some of the observed differences may reflect cage-level environmental effects. The authors call for future work incorporating baseline sampling, metabolite profiling of short-chain fatty acids and bile acids, exercise biomarkers such as lactate, and metagenomic approaches, along with larger samples and longer interventions, to determine whether these exercise-associated microbial shifts genuinely contribute to healthy aging.

Subject of Research: Effects of aerobic exercise training on gut microbiota composition and exercise capacity in genetically heterogeneous UM-HET3 mice

Article Title: Aerobic exercise training alters gastrointestinal microbiota and enhances exercise capacity in UM-HET3 mice

Article References: Smith, C., Garza, K., Azari, H., Vellers, H., Abdelrahman, M., Dhanasekara, C. S., Salmon, A. B., Liang, H., Fernandez, E., Musi, N., Massett, M. P., & Albracht-Schulte, K. (2026). Aerobic exercise training alters gastrointestinal microbiota and enhances exercise capacity in UM-HET3 mice. GeroScience. https://doi.org/10.1007/s11357-026-02582-4

Image Credits: AI Generated

DOI: 10.1007/s11357-026-02582-4

Keywords: gut microbiome, aerobic exercise, UM-HET3 mice, aging, exercise capacity, 16S rRNA sequencing, body composition, Firmicutes, Verrucomicrobiota, Lachnospiraceae, GeroScience, mouse model