The experimental strain of L. brevis GKEX provided by Grape King Bio Ltd. (Taoyuan, Taiwan) was originally isolated from douchi (fermented black soybeans) and preserved at the National Institute of Technology and Evaluation, Biological Resource Center (NBRC, Chiba, Japan) under accession number NITE BP-03696. A single colony was selected from an MRS agar plate and cultured in MRS broth (Difco, BD, United States) at 32 °C for 16 h with shaking at 100 rpm. The activated culture was then scaled up in 1 L of MRS medium with 0.1% inoculation and incubated under the same conditions. Afterward, the fermented broth was centrifuged at 25 °C at 5,000 rpm (Heraeus Megafuge 40R, Thermo Fisher Scientific Inc., United States) for 10 min to collect the bacterial pellet. The pellet was washed with reverse osmosis (RO) water, mixed with 20%(w/v) skim milk powder, and freeze-dried (FD24, Kingmech Scientific, Taoyuan, Taiwan) to obtain a live bacterial powder containing approximately 4 × 10¹1 colony-forming units (CFU) per gram, as determined by the plate counting method. For the preparation of the heat-killed GKEX sample (HK-GKEX), the washed bacterial pellet was resuspended in an equal volume of RO water, autoclaved at 121 °C for 15 min, and then freeze-dried. To prepare the freeze-killed GKEX sample (FK-GKEX), the washed bacterial pellet was resuspended in RO water, frozen at −20 °C for 3 days, and thawed at room temperature. This freeze–thaw cycle was repeated three times before processing into the powder by freeze-drying. All the bacterial powders were stored at 4 °C until use in subsequent animal experiments.

Eight-week-old male ICR mice with an average weight of approximately 30 g were sourced from BioLASCO Taiwan (Yi-Lan Breeding Center, Yi-Lan, Taiwan). The Institutional Animal Care and Use Committee of the National Taiwan Sport University (IACUC of NTSU) approved (IACUC-11108) all animal experiments conducted. All animals were provided LabDiet^® 5,001 (PMI Nutrition International, Purina Mills, MO, United States) and distilled water ad libitum. The room temperature was maintained at 24 ± 2 °C with a humidity of 65 ± 5% and a 12-h light–dark cycle. A 2-week acclimation period was provided during which the mice were subjected to adaptive training to minimize stress and familiarize them with treadmill use. The adaptive training protocol consisted of placing the mice on a stationary treadmill on days 1–3, followed by light running at a speed of 5–10 m/min for 10 min/day from days 4 to 7, with gradual increases in duration and intensity over the second week to prepare them for later endurance testing. A total of 48 mice were randomly divided into 6 groups, with each group containing 8 mice: vehicle (water only) and branched-chain amino acids (BCAA, positive control, 1,500 mg/day (14)). Feeding doses of different preparations of L. brevis GKEX groups in mice were based on the human-to-mouse conversion factor of 12.3 (15): low-dosage L. brevis GKEX (GKEX-L) (0.0615 mg/day, equivalent to 10 mg/day in a human with 60 kg), high-dosage L. brevis GKEX (GKEX-H) (0.615 mg/day, equivalent to 100 mg/day in a human with 60 kg), HK-GKEX (0.615 mg/day, equivalent to 100 mg/day in a human with 60 kg), and FK-GKEX (0.615 mg/day, equivalent to 100 mg/day in a human with 60 kg). During the animal experiment, L. brevis GKEX was administered daily at 9 a.m. via oral gavage for four consecutive weeks. Throughout this period, the body weight, food intake, and water consumption were recorded. The intensity of exercise was gradually increased from low to high in consideration of animal welfare. Endurance fatigue analysis was conducted on days 29 to 37 of the experiment. This included analysis of forelimb grip strength, an exhaustive running test, post-exercise blood lactate and blood urea nitrogen (BUN) concentrations, as well as liver and muscle glycogen content. The experimental procedure is shown in Figure 1.

The forelimbs of the mice were placed on a tension rod (diameter 2 mm, length 7.5 cm) using a grip strength meter (Model-RX-5, Aikoh Engineering, Nagoya, Japan). During this process, the tail of the mouse was repeatedly pulled back to record the maximum grip force (peak power) as previously described (16).

The mouse was placed on a mouse treadmill (MK680C, Muromachi Kikai, Chuo, Tokyo, Japan), set at a 15-degree incline with an initial speed of 12 m/min, increasing by 3 m/min every 2 min until the mouse falls into the shock zone (electric current of 2 Hz, 1.22 mA) for 5 s, showing no willingness to continue running forward, thereby indicating exhaustion. The time from the start to exhaustion was recorded as a measure of endurance performance (17).

Unloaded swimming tests were conducted according to a previous study (18, 19). Mice swam for 10 min and rested for 20 min for lactate analysis, whereas mice swam for 90 min and rested for 60 min for BUN analysis. The swimming test was conducted in water at 28 ± 1 °C, and blood samples were immediately collected after swimming. The blood samples were centrifuged at 4 °C, 1,500 × g for 10 min to collect blood for analysis. Post-exercise fatigue markers, including lactate and BUN, were measured using an automatic blood biochemical analyzer (Hitachi 7,060, Hitachi, Tokyo).

At the end of the experiment, the mice were euthanized by 95% CO2 asphyxiation. Their blood was collected and then centrifuged at 4 °C, 1,500 × g for 15 min. The serum was aliquoted and analyzed for general clinical blood biochemistry, including glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), creatine kinase (CK), BUN, creatinine (CREA), uric acid (UA), total cholesterol (TC), triglycerides (TG), glucose (GLU), and albumin (ALB), using a blood biochemical automatic analyzer (Hitachi 7,080, Hitachi, Tokyo, Japan).

This analytical method directly quantifies glycogen according to Chamberland and Rioux (20). Mice were sacrificed 30 min after the last feeding, and the liver and hind limb muscles were collected. After washing with physiological saline and drying, the tissues were homogenized with 5 times (w/v) tissue homogenization buffer using a Bullet Blender tissue homogenizer (Next Advance, Cambridge, MA, United States). The homogenates were then centrifuged at 4 °C, 12,000 × g for 15 min, and the supernatant was used for hepatic glycogen content analysis (21). Additionally, a calibration curve was prepared using a commercially available glycogen standard (Glycogen Sigma) to calculate changes in glycogen storage in the liver and muscle tissues among the different groups.

Fresh samples from the cecum of mice were collected after sacrifice and stored at −80 °C for DNA extraction using the Qiagen DNA Mini Kit (Qiagen, MD, United States). DNA was quantified using a NanoDrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States) and maintained at −80 °C. Furthermore, the V3–V4 region of the 16S rRNA gene was amplified using the forward primer 341F (5’-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′) and reverse primer 805R (5’-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′). Subsequently, the PCR product was subjected to a second PCR amplification using the Nextera XT Index Kit (Illumina, United States) and quantified by real-time quantitative polymerase using a KAPA library quantification kit (KAPA Biosystems, United States). The Illumina HiSeq platform was applied to the V3 and V4 regions of 16S rRNA genes for taxonomic profiling of microbiota in cecums from kingdom to species, according to the Greengenes database¹. Linear discriminant analysis effect size (LEfSe) analysis was performed to identify the differences between the abundance of taxonomic biomarkers of the gut microbiota in each group, and the results were visualized using a heatmap. Alpha diversity was assessed using the Shannon diversity index to evaluate the species richness and evenness within each group. The results were presented as boxplots to visualize differences in microbial diversity. Beta diversity using Bray–Curtis dissimilarity, unweighted UniFrac, and weighted UniFrac distances, which were visualized via constrained principal coordinates analysis (CPCoA).

Values are presented as mean ± SD, with each group consisting of eight mice. To assess group differences, a one-way analysis of variance (ANOVA) using Duncan’s post hoc test was conducted. The analysis was performed using SAS 9.0 (SAS Institute, Cary, NC, United States). Statistical differences in alpha diversity among groups were analyzed using the Kruskal–Wallis test, and for pairwise comparisons between groups, post hoc Dunn’s tests were performed. PERMANOVA accompanies CPCoA to determine whether the observed separation is statistically significant. A p-value of < 0.05 was considered statistically significant.

During the experimental period, the body weight of mice in all six groups showed a stable increase over time, indicating no adverse effects associated with L. brevis GKEX supplementation. Additionally, there were no significant changes in the average body weight among the groups during the 4-week experimental period (Figure 2). As shown in Supplementary Table S1, no significant differences were observed in the average daily water and chow intake among the groups. Therefore, supplementation of different treatment methods of L. brevis GKEX did not significantly affect the drinking or feeding behavior of the mice. Furthermore, there were no significant differences observed in the absolute or relative weight of the tissues and organs among the groups (Supplementary Table S4), suggesting that no abnormal damage and changes were observed.

In order to determine whether L. brevis GKEX supplementation can enhance mouse muscle strength, the mice were fed GKEX for 4 weeks, and the forelimb grip strength of the L. brevis GKEX groups was compared with that of the control group. Forelimb grip strength performance may be affected by individual body weight. To adjust for this, the relative grip strength normalized to body weight (%) was calculated, as shown in Figure 3. The relative forelimb grip strength for the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups was 347.02 ± 17.68, 385.02 ± 25.23, 415.55 ± 28.74, 426.50 ± 15.13, 407.79 ± 28.11, and 418.07 ± 20.45, respectively. Compared to the Vehicle group, the L. brevis GKEX group significantly increased the forelimb grip strength. Furthermore, compared to the BCAA group, the GKEX-L, GKEX-H, and FK-GKEX groups significantly increased by 1.08-fold (p = 0.0117), 1.11-fold (p = 0.0009), and 1.09-fold (p = 0.0069), respectively. The absolute values of grip strength are presented in Supplementary Table S2.

The running exhaustion times for the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups were 7.59 ± 1.52, 10.23 ± 0.97, 13.45 ± 1.11, 16.28 ± 0.72, 13.00 ± 1.06, and 14.96 ± 0.80 min, respectively (Figure 4). Compared to the Vehicle group, the BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups showed significantly improved running exhaustion performance. Additionally, compared to the BCAA group, the GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups showed significantly enhanced running exhaustion performance by 1.31-fold, 1.59-fold, 1.27-fold, and 1.46-fold, respectively (p < 0.0001). Thus, continuous supplementation with L. brevis GKEX effectively increased the running exhaustion performance.

The serum lactate levels were measured before swimming, after 10 min of swimming, and after 20 min of rest, followed by swimming. The absolute values of lactate were recorded at different timepoints in Supplementary Table S3. Before swimming, there were no significant differences in serum lactate levels between the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups. After 10 min of swimming, the lactate increase ratio was calculated based on the lactate levels before and 10 min after the exercise. Compared to the Vehicle group, the GKEX-H group showed a significant decrease of 1.2-fold (p = 0.0213) in the lactate increase rate (Figure 5A).

The lactate clearance ratio, determined following a 20-min rest period after exercise, was used to assess lactate recovery. Compared to the Vehicle group, GKEX-H and FK-GKEX groups significantly increased the lactate clearance ratio. Furthermore, compared to the BCAA group, the supplementation of the GKEX-H significantly increased the lactate clearance ratio by 1.48-fold (p = 0.0145) (Figure 5B).

As shown in Figure 6, BUN concentrations following a 90-min swimming test were measured in each group of mice. The BUN concentrations in the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups showed no significant differences among the six groups.

Additionally, BUN concentrations after a 90-min swimming test followed by a 60-min rest were tested for each group of mice. Compared to the Vehicle group, supplementation with L. brevis GKEX significantly reduced BUN concentration. Furthermore, compared to the BCAA group, supplementation with GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX significantly decreased BUN concentrations by 1.09-fold (p = 0.0094), 1.14-fold (p < 0.0001), 1.07-fold (p = 0.0217), and 1.11-fold (p = 0.0011), respectively.

Hepatic and muscular glycogen contents are shown in Figure 7. The hepatic glycogen content in the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups was 1.45 ± 0.26, 1.66 ± 0.31, 2.05 ± 0.28, 2.54 ± 0.36, 1.99 ± 0.24, and 2.08 ± 0.19 (mg/g), respectively. Compared to the Vehicle group, the L. brevis GKEX group showed significantly increased hepatic glycogen content. In addition, compared to the BCAA group, the GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups significantly increased hepatic glycogen content by 1.23-fold (p = 0.0074), 1.53-fold (p < 0.0001), 1.20-fold (p = 0.0227), and 1.25-fold (p = 0.0044), respectively. On the other hand, the muscular glycogen content in the Vehicle, BCAA, GKEX-L, GKEX-H, HK-GKEX, and FK-GKEX groups was 0.57 ± 0.07, 0.79 ± 0.06, 1.13 ± 0.87, 1.37 ± 0.08, 1.08 ± 0.10, and 1.25 ± 0.08 (mg/g), respectively. Compared to the Vehicle group, the GKEX-H and HK-GKEX groups showed significantly increased muscular glycogen content. Compared to the BCAA group, the GKEX-H group significantly increased muscular glycogen content by 1.38-fold (p = 0.0444).

The heatmap resulting from LEfSe analysis shows the taxonomic groups in the intestines and their abundance distribution in each group. As shown in Figure 8A, the GKEX-L group showed higher proportions of Mollicutes (Class), Anaeroplasmatales (Order), Anaeroplasmataceae (Family), and Anaeroplasma (Genus). Furthermore, an LDA score above 2 is widely considered a threshold for biologically meaningful differences in microbiota analysis, assuming that it passes the statistical significance test (Supplementary Figure S1). The abundances of Roseburia, Blautia (Family Lachnospiraceae), and Eubacterium at the genus level and that of Christensenella minuta (Phylum Firmicutes) significantly increased in the GKEX-H group compared with the vehicle group. In addition, the genus Peptococcus in the HK-GKEX group was significantly increased, whereas the phylum Proteobacteria was significantly increased in the FK-GKEX group. Supplementation with different doses of live and dead bacteria, particularly high doses of live L. brevis GKEX, increased the taxa of the microbiome.

Alpha diversity refers to the diversity of the microbial species within the experimental groups. The α-diversity of the gut microbiota, as measured by the Shannon index, showed no significant differences among the groups (p = 0.38). Pairwise comparisons also revealed no statistically significant differences (all p > 0.05), although a trend of increased diversity was observed in the GKEX-H and HK-GKEX groups (Figure 8B). Beta diversity was visualized by CPCoA (Figure 8C). Beta diversity measures the differences in microbial community composition between each group. BCAA and GKEX-L groups showed a high degree of overlap, indicating that their microbiota effects were similar and not markedly distinct. The FK-GKEX group overlapped partly with the Vehicle group and partly with the BCAA and GKEX-L groups, suggesting that its microbiota composition shared similarities with all three. In contrast, the GKEX-H and HK-GKEX groups were clearly separated from the others, indicating that their microbiota profiles were substantially different from both the Vehicle and other treatment groups. A clear separation between groups was observed, and PERMANOVA indicated a significant difference (p = 0.008) in community structure between groups.