Exercise Alters Gut Microbiota Composition and Function in Lean and Obese Humans : Medicine & Science in Sports & Exercise

Journal Logo


Exercise Alters Gut Microbiota Composition and Function in Lean and Obese Humans


Author Information
Medicine & Science in Sports & Exercise 50(4):p 747-757, April 2018. | DOI: 10.1249/MSS.0000000000001495


Environmental stimuli and behavioral practices can modulate the composition and functionality of the gut microbiota. For instance, diet (1), obesity status (2), and mode of birth delivery (3) have been shown to significantly contribute to the metabolic and immune-regulating capacity of the gut microbiome throughout the lifespan. Physical activity status in humans correlates with microbiota composition; however, the effects of endurance-based exercise training on the human gut microbiota have not been explored.

Despite the lack of research in humans, data from our laboratory and others indicate that exercise training can significantly modulate the gut microbiota in animal models (4–6). Moreover, exercise-induced changes in the gut microbiota have been correlated with changes in host physiology, including alterations in metabolism (7), immunity (5), and behavior (6). Exercise training also increases microbiota-derived short chain fatty acids (SCFA) within the mouse gut (8). Short-chain fatty acids are two- to six-carbon-long fatty acids which serve as an energy source for a variety of tissues and have been shown to reduce inflammation (9), improve insulin sensitivity (10), and alter central nervous system morphology (11).

Among humans, Estaki et al. (12) showed that cardiorespiratory fitness (V˙O2max) was positively correlated with increased bacterial diversity and butyrate-producing bacteria, whereas Barton et al. (13) described that the gut microbiome in athletes support higher turnover of macromolecules (carbohydrates and protein) and higher gut SCFA concentrations compared to sedentary controls. Despite offering intriguing associations between exercise and the gut microbiota, these studies are limited by their cross-sectional design because other factors that transiently alter the gut microbiota are not controlled. Thus, the present study was designed to assess gut microbial changes in a longitudinal design that controls for diet and antibiotic use.

The effects of exercise on the human gut microbiota have also not been investigated in the context of obesity. An obesity-associated gut microbiota can significantly contribute to host weight gain through enhanced energy harvest (14) and drive inflammation through reduced gut barrier integrity (15). Exercise training can contribute to improvements in obesity status and associated comorbidities by increasing insulin sensitivity, reducing systemic inflammation (16), and improving V˙O2max (17). Exercise training in mice has also been shown to limit gut barrier integrity disruptions and tissue inflammatory responses induced by obesity, offering evidence of a possible gut microbiota-mediated mechanism by which exercise improves outcomes associated with obesity (18). However, considering the wide-ranging effectiveness of exercise interventions in combating weight gain and obesity-associated comorbidities in humans (19), a more comprehensive investigation into the biological systems that are responsive to both obesity and exercise training, including the gut microbiota, is needed.

In this study, we investigated whether 6 wk of aerobic exercise training alters gut microbial communities and fecal SCFA in previously sedentary lean and obese adults with controls in place to ensure consistent dietary patterns. We hypothesized that exercise training would stimulate orthogonal shifts in gut microbial communities dependent on body mass index (BMI) status and lead to increased fecal SCFA levels as well as functional metabolic capacity for the gut microbiota to produce SCFA. Lastly, we also predicted that composition, functional capacity, and metabolic output of the gut microbiota would return to baseline levels after an extended (6 wk) reversion to sedentary activity.


Participants and Ethical Approval

Thirty-two, previously sedentary women (n = 20) and men (n = 12) were recruited based on a lean or obese BMI; lean (BMI < 25 kg·m−2; n = 18; 9 women, 9 men) and obese (BMI > 30 kg·m−2; n = 14; 11 women, 3 men). Baseline participant characteristics are shown in Table 1. This study was approved by the University of Illinois Institution Review Board, written informed consent was obtained for all participants, and all procedures and protocols conformed to the standards of use of human participants in research as outlined in the Sixth Declaration of Helsinki.

Subject characteristics at baseline.

Inclusion and Exclusion Criteria

Participants were age 20 to 45 yr, had a BMI < 25 kg·m−2 (lean) or a BMI > 30 kg·m−2 (obese) and were sedentary (≤30 min of moderate- or high-intensity exercise per week and ≤10 aggregate Godin-Shepard Leisure Time Physical Activity Questionnaire (GSLTQ; Godin-Leisure) score. Subject medical history and medication use was assessed through questionnaire. Subjects that qualified for the study were free of metabolic and gastrointestinal disease, not pregnant or lactating, not taking medications that would impact bowel function, and not taking antibiotics for at least 3 months before or during the study.

Study Design

The study was conducted as a longitudinal design lasting 14 wk. After 2 wk of baseline testing, all subjects completed a 6-wk endurance-based exercise intervention. The exercise training period was followed by a 6-wk “washout” period, during which participants were instructed to refrain from exercising. Fecal and blood samples were collected throughout the study, with 3-d dietary controls in place before each collection (see Figure, Supplemental Digital Content 1, Experimental Design, https://links.lww.com/MSS/B104).

Screening and Diet Control

At baseline, subjects came to the laboratory for four visits. During these visits, GSTLQ were administered to ensure participants were sedentary over a period of at least 3 months. Participants were asked to complete a 7-d dietary record, which included detailed descriptions of the types and amounts of foods and beverages consumed. Then, in consultation with a registered dietitian, participants designed a 3-d food menu that consisted entirely of foods and drinks from the 7-d dietary recall. Participants were asked to follow this 3-d food menu before each fecal collection. Beyond the acute dietary control, subjects were also instructed to maintain overall dietary patterns, which included maintenance of alcoholic/caffeinated beverage consumption and continuation of any dietary supplement usage that occurred before the beginning of the study. Three-day menus were analyzed for macronutrient and micronutrient composition, food groups, and individual food item consumption (see Table, Supplemental Digital Content 2, Baseline dietary consumption, https://links.lww.com/MSS/B105) using Nutrition Data System for Research (NDSR 2014 version, University of Minnesota, Minneapolis, MN).

Exercise Training Protocol

The exercise intervention consisted of three supervised 30 to 60 min, moderate-to-vigorous intensity (60%–75% of HR reserve [HRR]) aerobic exercise sessions per week. Subjects chose from a cycle ergometer or treadmill during each exercise session. Percent of HRR was calculated as:

HRmax was determined during the baseline V˙O2max test. During week 1 of training, sessions were 30 min at 60% HRR. Training sessions for week 2 lasted 45 min at a similar intensity. Training sessions at week 3 were increased to 60 min at 60% HRR. During weeks 4 to 6 of training, there was an increase in intensity of 5% HRR per week, progressing up to 75% HRR for 60 min during week 6. All participants were 100% compliant in completing necessary requirements for the exercise portion of the study.

Washout Period

After posttraining assessments, all participants were instructed to refrain from exercise for the next 6 wk. To assess washout compliance, GSLTQ were again administered. Participants that scored below 10 were deemed to have returned to sedentary activity and were included in the washout analysis. A total of seven subjects (n = 4 lean and n = 3 obese) remained physically active during the washout (GSLTQ >10) and were not included in the washout analysis.

Fecal collections

Fecal samples were collected at baseline (E0), after 6 wk of exercise (E6), and after 6 wk of return to sedentary activity (W6). All collections were preceded by the 3-d food menu. Subjects were instructed to bring samples into the laboratory within 30 min of defecation to ensure minimal degradation of volatile SCFA. Researchers asked participants the time defecation to ensure sample was received within 30 min. Once received, a small portion of the sample (~0.5 g) was aliquoted for SCFA analysis, and the rest was stored at −80°C until future analysis.

Body composition assessments

Body composition was assessed using dual-energy x-ray absorptiometry (Hologic QDR 4500A, Bedford, MA) at E0, E6, and W6.

Maximal oxygen consumption (V˙O2max)

At baseline (E0) and after the exercise intervention (E6), participants completed a maximal oxygen uptake test to assess cardiorespiratory fitness (V˙O2max) (Parvo Medics True Max 2400, Sandy, UT). This test involved walking or running to maximal exertion on a treadmill using the Bruce testing protocol as described previously (20). HR was continuously recorded via wireless Polar Monitor (Polar Electro, Lake Success, NY) and the test ended upon volitional fatigue. V˙O2max was verified if two or more of the following criteria were met: 1) RPE > 18, 2) HR within 10 bpm of age-predicted max HR, 3) a plateau in HR (<3 bpm change) over the last two intensity stages, and/or 4) RER of > 1.10.

Bacterial DNA isolation and 16S rRNA sequencing

Fecal bacterial DNA was extracted using the PowerLyzer PowerSoil DNA Isolation Kit (MOBIO Laboratories, Inc.) and quality was assessed via gel electrophoresis. The DNA library was constructed using a Fluidigm Access Array system in the Functional Genomics Unit of the Roy J. Carver Biotechnology Center at the University of Illinois at Urbana (UIUC). After library construction, a 300-bp region of the 16S rRNA gene (V4) was amplified using a two-step PCR amplification method modified from Muturi et al. (21). Briefly, DNA samples were diluted to 2 ng·μL−1 and amplified with Roche High Fidelity Fast Start Kit and 20× Access Array loading reagent before PCR. Samples were amplified using the following Access Array cycling program: 50°C for 2 min (1 cycle), 70°C for 20 min (1 cycle), 95°C for 10 min (1 cycle), followed by 10 cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 1 min, 2 cycles at 95°C for 15 s, 80°C for 30 s, 60°C for 30 s, and 72°C for 1 min, 8 cycles at 95 °C for 15 s, 60°C for 30 s, and 72°C for 1 min, 2 cycles at 95°C for 15 s, 80°C for 30 s, 60°C for 30 s, and 72°C for 1 min, 8 cycles at 95°C for 15 s, 60°C for 30 s, and 72°C for 1 min, and 5 cycles at 95°C for 15 s, 80°C for 30 s, 60°C for 30 s, and 72°C for 1 min. The PCR was quantified on a Qubit fluorimeter (Thermo-Fisher). Evenly diluted product was used for a second round of amplification with Illumina linkers and barcodes with the following PCR conditions: 1 cycle 95°C for 10 min, 1 cycle 95°C for 15 s, 1 cycle of 60°C for 30 s, 14 cycles of 72°C for 1 min followed by a 72°C extension for 3 min. Products were quantified on a Qubit fluorimeter and stored at −20°C. All samples were run on a Fragment Analyzer (Advanced Analytics, Ames, IA), and amplicon regions and expected sizes were confirmed. Samples were then pooled in equal amounts according to product concentration. The pooled products were then size selected on a 2% agarose E-gel (Life Technologies) and extracted from the isolated gel slice with Qiagen gel extraction kit (Qiagen). Cleaned size selected products were run on an Agilent Bioanalyzer to confirm appropriate profile and determination of average size. The final library pool was spiked with 10% nonindexed PhiX control library (Illumina®) and sequenced using Illumina® MiSeq® V3 Bulk system. The libraries were sequenced from both ends of the molecules to a total read length of 300 nt from each end. High-quality (>25) sequence data were analyzed with QIIME 1.9.054. Quality control consisted of depleting or removing barcodes, primers, and short sequences (<237 bp), sequences with ambiguous base calls, and sequences with homopolymer runs exceeding 6 bp. After removal of singletons, operational taxonomic units (OTUs) were classified using closed reference picking with the Ribosomal Database Project at 97% similarity. β-diversity (weighted and unweighted UniFrac distances) were computed at an even sampling depth of 14,201 sequences per sample based on α-diversity (Chao1) rarefaction plots (see Figure, Supplemental Digital Content 3, Rarefaction α- diversity plot, https://links.lww.com/MSS/B106).

SCFA Analysis

Short-chain fatty acids were analyzed as described previously by Panasevich et al. (22). Briefly, fresh fecal samples were acidified in 6.25% meta-phosphoric acid solution and stored at −20°C. The SCFA concentrations were determined by gas chromatography (Hewlett-Packard 5890A Series II) and a glass column (180 cm × 4 mm i.d.), packed with 10% SP-1200/1% H3PO4 on 80/100 + mesh Chromosorb WAW (Supelco, Inc.). Nitrogen was the carrier gas with a flow rate of 75 mL·min−1. Oven, detector, and injector temperatures were 125°C, 175°C, and 180°C, respectively. Acetic, n-butyric, and propionic acid solutions (Sigma-Aldrich) were used as standards.

Functional Gene Quantification

Isolated bacterial DNA was assessed for the relative abundance of BCoAT gene and methylmalonyl-CoA decarboxylase (mmdA) content by qPCR as described by Louis and Flint (2007) and Reichardt (2014), repectively (23,24). Briefly, real-time PCR experiments were performed with SYBR Green Master Mix (Applied Biosystems) in a total volume of 20 μL. The BCoAT and mmdA genes were quantified in parallel with a universal 16S rRNA gene probe (see Table, Supplemental Digital Content 4, qPCR target genes and primer sequences, https://links.lww.com/MSS/B107).

Statistical Analysis

Gut microbiota diversity and taxonomy evaluation

β-diversity was calculated using QIIME version 1.9.0 (25) at an even sample depth of 14,038 sequences per sample, visualized using EMPeror (26), and analyzed by permutational multivariate analysis of variance (PERMANOVA). Genera level taxonomy assignment was generated using Ribosomal Database Project database through QIIME and was used as the primary outcome variable in subsequent factor analysis analyzing changes in gut microbiome. Preexercise (E0) to postexercise (E6) changes in SCFA concentrations, SCFA-producing genes, body composition, and V˙O2max were analyzed using a 2 × 2 repeated-measures ANOVA with BMI (lean, obese) and exercise [time] (E0–E6) as independent factors. Because of nonuniversal adherence to washout period guidelines, postexercise (E6) to washout (W6) changes in SCFA concentrations, relative SCFA-producing genes, body composition, and V˙O2max were completed using a separate 2 × 2 repeated-measures ANOVA with BMI and washout [time] (E6–W6) as independent variables.

Random Forest analysis and Boruta feature selection

To determine whether exercise training induced orthogonal shifts in the gut microbiota based on obesity status, Random Forest (RF) analysis was performed. The training set used preexercise to postexercise change in bacterial genera (Δ E0–E6 expressed as percent of total bacteria) with BMI status as the grouping variable. Random Forest was also performed with the change in bacterial genera during the washout (Δ E6–W6 percent of total bacteria) as the independent training set. The cross-validation error rates of correctly predicted groupings were averaged across BMI groups compared with the baseline error rates that would be achieved by randomly assigning to either grouping variable by Matthews Correlation Coefficient (see Figure, Supplemental Digital Content 5, RF classification, https://links.lww.com/MSS/B108). Next, Boruta feature selection was implemented to determine which shifts in bacterial taxa best discriminated between lean and obese groups. Boruta selection uses RF to iteratively compare importance of independent variables with that of pseudo-random (shadow) attributes (27). Variables “confirmed” by Boruta feature selection had significantly higher RF importance scores than “shadow” attributes and could therefore delineate between lean and obese groups. Analysis was performed on R random forest package with 1000 trees and all other default values.

Principle component factor reduction

To determine if exercise caused physiologically relevant changes in gut microbiota composition, principle component factor analysis (PCA) was completed on the change scores of bacterial genera (ΔE0–E6) that paralleled similar changes in SCFA (fecal butyrate or BCoAT; or cardiorespiratory fitness (V˙O2max) within each BMI group. First, the analyses revealed that the change in seven bacterial genera paralleled change in butyrate or BCoAT production in the lean participants and ten genera paralleling change in V˙O2max in the obese participants via Spearman Rho correlations. Next, retracted components explaining the overall change in these bacteria were calculated by PCA. This factor equation was then applied (unbiased) to the washout change scores of the same genera (ΔE6–W6). Resulting factors were then compared to each other and to outcome variables through Pearson r correlation analysis. Loading variables with small loading coefficients (<0.4) were suppressed for all PCA analysis. Significance was set a priori at α ≤ 0.05 for analyses.


Exercise training improved body composition and V˙O2max, which were reversed after returning to sedentary activity

Exercise training increased total lean body mass (P < 0.01, Fig. 1A), decreased body fat percentage (P < 0.01, Fig. 1B), and increased bone mineral density (P < 0.05, Fig. 1C) in both lean and obese subjects. Exercise also led to a marked improvement in cardiorespiratory fitness as measured by relative V˙O2max in both lean and obese groups (P < 0.05, Fig. 1D). After a return to sedentary activity for 6 wk (W6), lean mass and body fat changes that occurred in response to training were reversed toward baseline values in both lean and obese participants (P < 0.05; Fig. 1A–B), whereas BMD remained higher versus baseline (Fig. 1C). V˙O2max was not measured at W6.

Changes from preexercise (E0) to postexercise (E6) and washout (W6) for; (A) lean mass % [upper right inset: absolute lean mass], (B) body fat % (BF%) [upper right inset: absolute fat mass], (C) bone mineral density (BMD) g·cm−2 and (D) relative maximal oxygen consumption (V˙O2max) in mL·kg−1·min−1. *P < 0.05 versus E0 in respective group. N/A indicates that V˙O2max tests were not administered at the washout time point.

Exercise induces orthogonal shifts in the gut microbiome dependent on BMI status

β-diversity analysis revealed that gut microbiota composition was different between lean and obese individuals at baseline (E0 PERMANOVA, P = 0.034; Fig. 2A). After exercise training, there was no difference between the two groups (E6, PERMANOVA, P = 0.31; Fig. 2B) and remained not significantly different after the washout period (W6, PERMANOVA, P = 0.27; Fig. 2C). To further investigate whether exercise training initiated orthogonal shifts in gut microbiota communities’ dependent on obesity status, RF analysis using preexercise to postexercise change scores of bacterial genera was implemented. Change scores of genera could be adequately trained to predict obesity status (cross-validation error rate, 19.35 ± 3.32; P < 0.05 versus baseline entry; see Figure A, Supplemental Digital Content 5, RF classification, https://links.lww.com/MSS/B108). Next, RF was applied to change scores of bacterial genera from postexercise (E6) to washout (W6) and again, could be used to predict obesity status (cross-validation error rate, 18.21 ± 13.35, P < 0.05 versus baseline entry; see Figure B, Supplemental Digital Content 5, RF classification, https://links.lww.com/MSS/B108). Bacterial taxa that shifted differentially in response to exercise as a result of obesity status (confirmed by Boruta feature selection) are shown in Figure 2D (Exercise) and Figure 2E (Washout). Notably, changes in three bacterial genera (Collinsella spp., Faecalibacterium spp., and Lachnospira spp.) were strong predictors of BMI status across both RF analyses and responded in a contrasting manner as a result of exercise training (E0–E6; Fig. 2D) compared with washout (E6–W6; Fig. 2E). No differences in α-diversity of the microbiota were observed between lean and obese participants at baseline (see Figure, Supplemental Digital Content 3, Rarefaction α-diversity plot, https://links.lww.com/MSS/B106); P > 0.05) or at any of remaining collection time points (P > 0.05; data not shown).

(A–C) Principle coordinates analysis (PCoA) of gut microbiota community composition (Weighted Unifrac) comparing Lean (LN) versus Obese (OB) groups at (A) preexercise (E0), (B) after exercise training for E6 and, (C) after a return to sedentary activity for 6 wk (W6). PERMANOVA P < 0.05 deems significant difference between LN and OB microbiota communities, ns, not significantly different. (D–E) Comparison of bacterial genera that changed orthogonally based on obesity status as a result of (D) Exercise training (E0–E6) and E) Returning to sedentary activity washout (E6–W6); data are represented as Δ E0–E6 change (± SEM). Genera represented were “confirmed” by Boruta feature selection as satisfactory in describing obesity category.

Exercise training increased fecal SCFA concentrations and gut microbiome SCFA-producing capacity depending on BMI status

We investigated the effects of exercise training on fecal SCFA concentrations and genes within the gut microbiota important for SCFA production and observed that aerobic exercise increased fecal concentrations of the three most abundant SCFA: acetate [C2], propionate [C3] and butyrate [C4]. This effect appeared to be dependent on BMI status, because the changes in acetate and butyrate were observed primarily in the lean participants (Exercise × BMI P < 0.05; Fig. 3Ai and Fig. 3Aiii). Changes in SCFA were observed concomitant to an exercise-induced increase in the relative abundance of the butyrate-regulating gene BCoAT and the propionate-regulating gene mmdA (P < 0.05, respectively; Fig. 3Bi-ii). At W6, fecal acetate concentrations remained elevated (P > 0.05), whereas propionate and butyrate declined toward baseline levels (P < 0.05). In the obese group, concentrations of all three SCFA did not change from E6 to W6 (P > 0.05; Fig. 3Ai-iii). However, the washout period was accompanied by a reduction in SCFA-producing capacity of the gut microbiome, as depicted by E6 to W6 reduction in relative abundance of BCoAT and mmdA (P < 0.05; Fig. 3Bi-ii).

(A) Changes in fecal SCFA concentrations: (i( acetate [C2], (ii) propionate [C3], and (iii) butyrate [C4] as a result of 6 wk of exercise training (E0–E6) and returning to sedentary activity for 6 wk (E6–W6, upper right graphs), (B) changes in relative abundance of (i) BCoAT and (ii) mmdA as a result of 6 wk of exercise training (E0–E6) and return to sedentary activity for 6 wk (E6–W6, upper right graphs). *Significant time [exercise or washout] × group [BMI] interaction effect. #Significant Exercise or Washout main effect; &Significant BMI main effect at P < 0.05.

In obese individuals, we observed greater relative BCoAT compared with lean individuals (P < 0.01; Fig. 3Bi). The differences in baseline BCoAT levels, however, were not corroborated by differences in fecal SCFA concentrations between BMI categories at baseline (Fig. 3Ai-iii). Interestingly, and contrary to the notion that the SCFA-producing capacity of the gut microbiota may be detrimental to obese individuals, relative concentrations of SCFA and BCoAT were associated with lower body fat (%), higher lean mass (%) and higher V˙O2max within the obese group at baseline (see Figures A–D, Supplemental Digital Content 6, SCFA and functional genes versus baseline measures, https://links.lww.com/MSS/B109; P < 0.05, respectively).

Exercise training increases butyrate-regulating bacterial taxa that strongly associate with parallel shifts in body composition in lean individuals

Despite observing clear overall shifts in community composition, few individual taxa were consistently altered by exercise training across all individuals. In light of this apparent individualized response, we focused on whether functional groups of bacteria (rather than individual taxa), changed as a result of physical activity. First, we examined whether there existed a “SCFA-regulating bacterial group” that was affected by exercise training in lean participants only, as SCFA concentrations in obese participants were altered to a lesser degree by exercise training compared to lean participants. Using Spearman rank correlations, we observed that the change from E0 to E6 of five bacterial genera, Roseburia spp., Lachnospira spp., Clostridiales spp., Faecalibacterium spp., and f Lachnospiraceae unclass., positively correlated with changes in butyrate and/or BCoAT, whereas changes in Bacteroides spp. and Rikenella spp. negatively correlated with changes in butyrate and/or BCoAT (P < 0.05 for both analyses; Fig. 4A). Using principle components analysis, the relative changes in these selected taxa were then modeled into one factor, which explained 61.2% of the total variance in how these genera changed as a result of exercise training (EXERCISELN_Factor1; Fig. 4B). Next, we correlated this resulting factor against changes in body composition and V˙O2max and report that exercise-induced changes in “butyrate-regulating bacterial group,” along with changes in SCFA concentrations and BCoAT levels, were significantly related to changes in body composition (e.g., increases in lean mass and reductions in fat mass) (Fig. 4C). Moreover, the five bacteria genera that were positively correlated to butyrate production were modeled into one factor termed “butyrate producers.” These bacterial genera (Clostridiales spp, Lachnospira spp, Roseburia spp. f_Lachnospiraceae unclass, and Faecalibacterium spp.), which together consisted of over 30% of the total representative genera, increased in abundance with exercise training irrespective of BMI (P < 0.05), and then decreased after a return to sedentary activity (P < 0.05; Fig. 4D). Within the obese group, we observed fewer relationships between genera changes and parallel shifts in SCFA. Nevertheless, we observed a functional group of genera (10) that paralleled changes in cardiorespiratory fitness (see Figure, Supplemental Digital Content 7, Obese microbiota changes versus V˙O2max, https://links.lww.com/MSS/B110).

(A) Spearman correlations relating exercise-induced changes in seven bacterial genera to parallel changes in fecal concentrations of butyrate and relative abundance of BCoAT within the LN group only. EXERCISELN_Factor1 represents the component factor that most thoroughly explained the variance (61.2% of total) of the change in the highlighted bacterial genera. Spearman ρ, *P < 0.05, **P < 0.01. (B) Loading coefficients for each selected taxa used in primary component factor explaining largest overall variance. EXERCISELN_FACTOR1 is representative of the factor component that most thoroughly explained the variance in change of these select bacteria genera (61.2%). (C) Exercise-induced changes in butyrate-regulating taxa (ExerciseLN_factor1), butyrate and BCoAT associate with parallel changes in body composition but not cardiorespiratory fitness within the LN participants (E0-E6), Pearson r, *P < 0.05, **P < 0.01. (D) Indicates the effects of exercise training on the overall change in the five bacterial taxa (Roseburia spp + Lachnospira spp. + f_Lachnospriaceae + Clostridiales spp. + Faecalibacterium spp.) that positively associated with changes in butyrate or BCoAT concentrations, termed “Butyrate Producers.” #Main effect of Exercise (E0–E6) or Washout (E6–W6) at P < 0.05.

A return to sedentary activity for 6 wk led to a BMI-dependent reversion in gut microbiome composition

Lastly, we investigated whether a return to sedentary activity (E6–W6) would reverse the gut microbiota changes observed after exercise training. To accomplish this, we analyzed the effects of exercise training (ΔE0–E6) and return to sedentary activity (ΔE6–W6) on the bacterial genera within the BMI-dependent bacterial groups depicted in Figure 4 (LN) and Figure, Supplemental Digital Content 7, Obese microbiota changes versus V˙O2max, https://links.lww.com/MSS/B110 (OB). First, our results reveal that the relationship between genera changes that occurred from ΔE0–E6 were inversely related to the genera changes that occurred from ΔE6–W6. In fact, five of the seven bacterial genera in the lean bacterial clade and eight of the ten genera in the obese bacterial clade were inversely regulated by PA status (Figs. 5Ai. and Bi.). We then used the component factor equation that most thoroughly explained the change in bacterial genera from E0–E6 (EXERCISEfactor1) to calculate a factor explaining genera changes that occurred during E6–W6 (WASHfactor1). Using this unbiased predictive model, we found that the calculated factor component from the washout period WASHfactor1 was inversely correlated to that of the factor component from exercise period EXERCISEfactor1 within both BMI categories (Lean r = −0.831, P < 0.01, Fig. 5Aii. Obese r = −0.630, P < 0.05; Fig. 5Bii.), thus further validating that gut microbiota composition is dependent on PA status.

Spearman correlations describing the relationship between the gut microbiota response to exercise-training (E0–E6) versus the gut microbiota responses to washout from exercise (E6–W6) regarding individual bacterial genera within: (Ai) lean and (Bi) obese “functional clades.” (Aii–Bii) Factor equations explaining most variance in bacterial change during exercise period among both BMI categories (EXERCISELN_factor1 or EXERCISEOB_Factor1) were applied directly to those genera changes that occurred during the washout. The resulting factor component, termed WASHLN_factor1 or WASHOB_FACTOR1, was then compared with Pearson r correlations to the corresponding EXERCISE factor component.


In this study, we provide evidence that 6 wk of aerobic exercise training alters the gut microbiota and microbial-derived SCFA in previously sedentary lean and obese adults without any changes to dietary patterns. At baseline, obese individuals had different gut microbiota compositions from lean individuals, and the microbiotas of the two groups responded differently to exercise training. Additionally, we demonstrate that exercise-induced modulations of the gut microbiota and SCFA were strongly associated with changes in body composition in lean participants and V˙O2max in obese participants, highlighting the role of the gut microbiota in regulating physiological adaptations to exercise training. Lastly, we report that gut microbiome changes that occurred after a return to sedentary activity were inversely related to the modifications that occurred in response to exercise training, providing further evidence that physical activity status contributes significantly to the composition of the human gut microbiota.

This study revealed that compositional and functional microbiome changes due to exercise participation (or a return to sedentary activity) were dependent on BMI. This was indicated by obesity-dependent differences in baseline gut microbiota composition that were not maintained after 6 wk of exercise. Further highlighting the BMI-dependent response to training, exercise-induced shifts in SCFA-producing taxa (Faecalibacterium spp. and Lachnospira spp.) and genetic machinery (BCoAT) were more substantial in lean versus obese participants.

Exercise-induced increases in fecal SCFA concentrations observed in this study corroborate data presented in rodent models (8), as well as a previous cross-sectional study in humans relating fecal butyrate concentrations to V˙O2max and muscle turnover (12). The mechanism(s) responsible for exercise-induced increases in fecal SCFA concentration is not yet clear, but could involve increased production of SCFA through endogenous metabolic input (e.g., lactate), increased mixing of intestinal contents and bacterial fermentation of dietary fibers, increased anaerobic fermentation due to colonic oxygen saturation or pH changes, or reduced intestinal utilization and uptake of SCFA (28). Exercise-induced increases in fecal SCFA may also be augmented by shifts in the gut microbiota’s functional capacity to produce SCFA. Of the genes within the gut microbiota involved in SCFA production, BCoAT is one of two known butyrate-producing genes that contributes significantly to butyrate levels (29). Functionally, BCoAT transfers a co-enzyme A group from butyrl CoA to acetate, which results in the formation of acetyl CoA and butyrate, two substrates used by tissues for macromolecule synthesis or energy production (30). Methylmalonyl-CoA decarboxylase, meanwhile, is an integral gene in the succinate pathway, one of three propionate regulating pathways highly conserved within the mammalian gut microbiota (24). In light of the prominent roles of these genes in SCFA production and their strong association with SCFA-producing bacteria, we hypothesize that exercise-induced increases in these functional genes may partially explain the parallel rise in fecal SCFA. Moreover, it is possible that these shifts in SCFA genetic machinery may augment responses to other environmental cues within the gut, including increased potential for dietary fiber fermentation, ultimately leading to increased SCFA production (31).

Exercise-induced increases in fecal BCoAT and SCFA levels also strongly paralleled improvements in body composition, including most notably, increases in lean mass. The mechanisms behind such an association are unclear, but many possible explanations exist. SCFA can improve skeletal muscle insulin sensitivity (32), reduce inflammation (33), and regulate satiety (34), all of which may contribute to the improvements in body composition observed in this study. Additionally, SCFA are also energy substrates for numerous tissue types, including colon (35), adipose (36), and muscle (32), indicating that SCFA can contribute to enhanced energy harvest from the diet, ultimately leading to enhanced tissue growth, including skeletal muscle. Considering the findings presented here and previous cross-sectional studies (12,13), we postulate that an exercise-induced increase towards a greater SCFA-producing capacity may represent a beneficial adaptation, ultimately contributing to enhanced replenishment of expended calories and improvements in body composition. The SCFA-producing capacity of the gut microbiota also returned to pretraining levels after the washout period, as indicated by a return of fecal BCoAT and mmdA levels (within the lean and obese groups) and butyrate-producing bacteria taxa (within the lean group) toward baseline values. In light of these findings, we postulate that exercise-induced shifts in the metabolic capacity of the gut microbiome may be transient and likely dependent on repeated exercise stimuli.

This study also identified differences in the baseline microbiome characteristics and SCFA producing capabilities between lean and obese individuals. Most notably, obese participants displayed nearly double the baseline levels of BCoAT compared to lean counterparts. The possible role of BCoAT and SCFA in energy harvest during obesity was first described in a seminal study by Turnbaugh et al. (14), who showed that a fecal transplant from obese individuals into germ-free mice led to increased gut SCFA concentrations and an obese phenotype in the recipient mice. In humans, successful weight loss through gastric bypass surgery was accompanied by reduced levels of BCoAT (37). However, contrary to previous studies, we observed no differences in the concentrations of any SCFA between lean and obese individuals. Moreover, and conflicting with the notion that SCFA lead to aberrant body composition phenotypes during obesity, this study indicates that within the obese participants, baseline fecal butyrate concentrations were strongly associated with higher lean mass, reduced body fat, and higher cardiorespiratory fitness (V˙O2max), whereas BCoAT levels were associated with higher lean mass. Therefore, these data challenge the assumption that the gut microbiome’s potential to produce SCFA is a biomarker for a maladaptive obese phenotype. Rather, it appears that these factors are indicative of healthier body composition and enhanced cardiorespiratory fitness during obesity.

Exercise-induced shifts in the gut microbiota and SCFA may have implications for diseases of the gut. Physical activity has been shown to reduce risk of colorectal cancer and irritable bowel disease (38). Butyrate is a primary fuel source for colonocytes (39), is required for maintaining gut barrier function (40), and is a critical metabolite regulating the anti-inflammatory/regulatory phenotypes of gut-resident immune cells (41). Interestingly, genera that positively associated with exercise-induced increases in butyrate concentrations (Faecalibacterium spp., Roseburia spp., Lachnospira spp., f_Lachnospiraceae, and Clostridiales spp.), are well documented butyrate producers (31,42), and have been associated with notable physiological outcomes in vivo. In particular, Faecalibacterium spp. has been shown to induce local and systemic anti-inflammatory effects and may be protective against bowel diseases, including ulcerative colitis (43,44). Thus, we postulate that exercise-induced shifts in SCFA, butyrate-producing taxa and metabolic capacity to produce SCFA may have the potential to prevent the occurrence or offset the symptomology of irritable bowel disease. Future studies are needed to address these hypotheses more closely.

There are limitations to this study. Seven participants did not adhere to washout guidelines and were excluded from analysis. We also did not include V˙O2max at the end of the study because of the predicted subject burden of an extended exercise training study. We did, however, assess washout body composition to verify that physiological parameters reverted to preexercise levels after a return to sedentary activity.

In summary, these data indicate for the first time that exercise training can significantly modulate the composition and metabolic capacity of the human gut microbiota. Exercise-induced changes in the gut microbiota are valuable prognostic features that are integral to understanding the overall physiological response to exercise training in humans.

The authors would like to acknowledge Mark Band, Chris Wright and Alvaro Hernandez at the Roy J. Carver Biotechnology Center at the UIUC for their contribution to sequencing preparation and analysis.

This work was partially funded by a doctoral student research grant from the American College of Sports Medicine (ACSM). Authors have no professional relationships with companies or manufacturers who will benefit from the results of the present study. Results of the present study do not constitute endorsement by ACSM. Results of the study are presented clearly, honestly and without fabrication, falsification, or inappropriate data manipulation.


1. David LA, Maurice CF, Carmody RN, Gootenberg DB, Button JE, Wolfe BE, et al. Diet rapidly and reproducibly alters the human gut microbiome. Nature. 2014;505(7484):559–563.
2. Cani PD. Gut microbiota and obesity: lessons from the microbiome. Brief Funct Genomics. 2013;12(4):381–7.
3. Dominguez-Bello MG, Costello EK, Contreras M, Magris M, Hidalgo G, Fierer N, et al. Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns. Proc Natl Acad Sci U S A. 2010;107(26):11971–5.
4. Allen JM, Berg Miller ME, Pence BD, Whitlock K, Nehra V, Gaskins HR, et al. Voluntary and forced exercise differentially alters the gut microbiome in C57BL/6J mice. J Appl Physiol (1985). 2015;118(8):1059–66.
5. Choi JJ, Eum SY, Rampersaud E, Daunert S, Abreu MT, Toborek M. Exercise attenuates PCB-induced changes in the mouse gut microbiome. Environ Health Perspect. 2013;121(6):725–30.
6. Kang SS, Jeraldo PR, Kurti A, Miller ME, Cook MD, Whitlock K, et al. Diet and exercise orthogonally alter the gut microbiome and reveal independent associations with anxiety and cognition. Mol Neurodegener. 2014;9:36.
7. Evans CC, LePard KJ, Kwak JW, Stancukas MC, Laskowski S, Dougherty J, et al. Exercise prevents weight gain and alters the gut microbiota in a mouse model of high fat diet-induced obesity. PLoS One. 2014;9(3):e92193.
8. Matsumoto M, Inoue R, Tsukahara T, Ushida K, Chiji H, Matsubara N, et al. Voluntary running exercise alters microbiota composition and increases n-butyrate concentration in the rat cecum. Biosci Biotechnol Biochem. 2008;72(2):572–6.
9. Kaczmarczyk MM, Miller MJ, Freund GG. The health benefits of dietary fiber: beyond the usual suspects of type 2 diabetes mellitus, cardiovascular disease and colon cancer. Metabolism. 2012;61(8):1058–66.
10. Watterson KR, Hudson BD, Ulven T, Milligan G. Treatment of type 2 diabetes by free fatty acid receptor agonists. Front Endocrinol. 2014;5:137.
11. De Vadder F, Kovatcheva-Datchary P, Goncalves D, Vinera J, Zitoun C, Duchampt A, et al. Microbiota-generated metabolites promote metabolic benefits via gut-brain neural circuits. Cell. 2014;156(1–2):84–96.
12. Estaki M, Pither J, Baumeister P, Little JP, Gill SK, Ghosh S, et al. Cardiorespiratory fitness as a predictor of intestinal microbial diversity and distinct metagenomic functions. Microbiome. 2016;4(1):42.
13. Barton W, Penney NC, Cronin O, Garcia-Perez I, Molloy MG, Holmes E, et al. The microbiome of professional athletes differs from that of more sedentary subjects in composition and particularly at the functional metabolic level. Gut. 2017. [Epub ahead of print]. doi: 10.1136/gutjnl-2016-313627.
14. Turnbaugh PJ, Ley RE, Mahowald MA, Magrini V, Mardis ER, Gordon JI. An obesity-associated gut microbiome with increased capacity for energy harvest. Nature. 2006;444(7122):1027–31.
15. McPhee JB, Schertzer JD. Immunometabolism of obesity and diabetes: microbiota link compartmentalized immunity in the gut to metabolic tissue inflammation. Clin Sci (Lond). 2015;129(12):1083–96.
16. Woods JA, Vieira VJ, Keylock KT. Exercise, inflammation, and innate immunity. Immunol Allergy Clin North Am. 2009;29(2):381–93.
17. Hammer RL, Barrier CA, Roundy ES, Bradford JM, Fisher AG. Calorie-restricted low-fat diet and exercise in obese women. Am J Clin Nutr. 1989;49(1):77–85.
18. Campbell SC, Wisniewski PJ, Noji M, McGuinness LR, Haggblom MM, Lightfoot SA, et al. The effect of diet and exercise on intestinal integrity and microbial diversity in mice. PLoS One. 2016;11(3):e0150502.
19. Goldberg Y, Boaz M, Matas Z, Goldberg I, Shargorodsky M. Weight loss induced by nutritional and exercise intervention decreases arterial stiffness in obese subjects. Clin Nutr. 2009;28(1):21–5.
20. Pollock ML, Foster C, Schmidt D, Hellman C, Linnerud AC, Ward A. Comparative analysis of physiologic responses to three different maximal graded exercise test protocols in healthy women. Am Heart J. 1982;103(3):363–73.
21. Muturi EJ, Donthu RK, Fields CJ, Moise IK, Kim CH. Effect of pesticides on microbial communities in container aquatic habitats. Sci Rep. 2017;7:44565.
22. Panasevich MR, Allen JM, Wallig MA, Woods JA, Dilger RN. Moderately fermentable potato fiber attenuates signs and inflammation associated with experimental colitis in mice. J Nutr. 2015;145(12):2781–8.
23. Louis P, Flint HJ. Development of a semiquantitative degenerate real-time pcr-based assay for estimation of numbers of butyryl-coenzyme A (CoA) CoA transferase genes in complex bacterial samples. Appl Environ Microbiol. 2007;73(6):2009–12.
24. Reichardt N, Duncan SH, Young P, Belenguer A, McWilliam Leitch C, Scott KP, et al. Phylogenetic distribution of three pathways for propionate production within the human gut microbiota. ISME J. 2014;8(6):1323–35.
25. Kuczynski J, Stombaugh J, Walters WA, Gonzalez A, Caporaso JG, Knight R. Using QIIME to analyze 16S rRNA gene sequences from microbial communities. Curr Protoc Microbiol. 2012; 27:E:1E.5:1E.5.1–1E.5.20.
26. Vazquez-Baeza Y, Pirrung M, Gonzalez A, Knight R. EMPeror: a tool for visualizing high-throughput microbial community data. Gigascience. 2013;2(1):16.
27. Kursa MB. Robustness of Random Forest-based gene selection methods. BMC Bioinformatics. 2014;15:8.
28. Turnbaugh PJ, Gordon JI. The core gut microbiome, energy balance and obesity. J Physiol. 2009;587(Pt 17):4153–8.
29. Louis P, McCrae SI, Charrier C, Flint HJ. Organization of butyrate synthetic genes in human colonic bacteria: phylogenetic conservation and horizontal gene transfer. FEMS Microbiol Lett. 2007;269(2):240–7.
30. Charrier C, Duncan GJ, Reid MD, Rucklidge GJ, Henderson D, Young P, et al. A novel class of CoA-transferase involved in short-chain fatty acid metabolism in butyrate-producing human colonic bacteria. Microbiology. 2006;152(Pt 1):179–85.
31. Hippe B, Zwielehner J, Liszt K, Lassl C, Unger F, Haslberger AG. Quantification of butyryl CoA:acetate CoA-transferase genes reveals different butyrate production capacity in individuals according to diet and age. FEMS Microbiol Lett. 2011;316(2):130–5.
32. Canfora EE, Jocken JW, Blaak EE. Short-chain fatty acids in control of body weight and insulin sensitivity. Nat Rev Endocrinol. 2015;11(10):577–91.
33. Roelofsen H, Priebe MG, Vonk RJ. The interaction of short-chain fatty acids with adipose tissue: relevance for prevention of type 2 diabetes. Beneficial microbes. 2010;1(4):433–7.
34. Kaji I, Karaki S, Kuwahara A. Short-chain fatty acid receptor and its contribution to glucagon-like peptide-1 release. Digestion. 2014;89(1):31–6.
35. Astbury SM, Corfe BM. Uptake and metabolism of the short-chain fatty acid butyrate, a critical review of the literature. Curr Drug Metab. 2012;13(6):815–21.
36. den Besten G, Bleeker A, Gerding A, van Eunen K, Havinga R, van Dijk TH, et al. Short-chain fatty acids protect against high-fat diet-induced obesity via a PPARgamma-dependent switch from lipogenesis to fat oxidation. Diabetes. 2015;64(7): 2398–408.
37. Patrone V, Vajana E, Minuti A, Callegari ML, Federico A, Loguercio C, et al. Postoperative changes in fecal bacterial communities and fermentation products in obese patients undergoing bilio-intestinal bypass. Front Microbiol. 2016;7:200.
38. Wolin KY, Yan Y, Colditz GA, Lee IM. Physical activity and colon cancer prevention: a meta-analysis. Br J Cancer. 2009;100(4):611–6.
39. Donohoe DR, Garge N, Zhang X, Sun W, O’Connell TM, Bunger MK, et al. The microbiome and butyrate regulate energy metabolism and autophagy in the mammalian colon. Cell Metab. 2011;13(5):517–26.
40. Peng L, Li ZR, Green RS, Holzman IR, Lin J. Butyrate enhances the intestinal barrier by facilitating tight junction assembly via activation of AMP-activated protein kinase in Caco-2 cell monolayers. J Nutr. 2009;139(9):1619–25.
41. Furusawa Y, Obata Y, Fukuda S, Endo TA, Nakato G, Takahashi D, et al. Commensal microbe-derived butyrate induces the differentiation of colonic regulatory T cells. Nature. 2013;504(7480):446–50.
42. Louis P, Duncan SH, McCrae SI, Millar J, Jackson MS, Flint HJ. Restricted distribution of the butyrate kinase pathway among butyrate-producing bacteria from the human colon. J Bacteriol. 2004;186(7):2099–106.
43. Machiels K, Joossens M, Sabino J, De Preter V, Arijs I, Eeckhaut V, et al. A decrease of the butyrate-producing species Roseburia hominis and Faecalibacterium prausnitzii defines dysbiosis in patients with ulcerative colitis. Gut. 2014;63(8):1275–83.
44. Qiu X, Zhang M, Yang X, Hong N, Yu C. Faecalibacterium prausnitzii upregulates regulatory T cells and anti-inflammatory cytokines in treating TNBS-induced colitis. J Crohns Colitis. 2013;7(11):e558–68.


Supplemental Digital Content

© 2018 American College of Sports Medicine