Changes in intestinal microbiota composition and metabolism coincide with increased intestinal permeability in young adults under prolonged physiological stress
Abstract
The magnitude, temporal dynamics, and physiological effects of intestinal microbiome responses to physiological stress are poorly characterized. This study used a systems biology approach and a multiple-stressor military training environment to determine the effects of physiological stress on intestinal microbiota composition and metabolic activity, as well as intestinal permeability (IP). Soldiers (n = 73) were provided three rations per day with or without protein- or carbohydrate-based supplements during a 4-day cross-country ski-march (STRESS). IP was measured before and during STRESS. Blood and stool samples were collected before and after STRESS to measure inflammation, stool microbiota, and stool and plasma global metabolite profiles. IP increased 62 ± 57% (mean ± SD, P < 0.001) during STRESS independent of diet group and was associated with increased inflammation. Intestinal microbiota responses were characterized by increased α-diversity and changes in the relative abundance of >50% of identified genera, including increased abundance of less dominant taxa at the expense of more dominant taxa such as Bacteroides. Changes in intestinal microbiota composition were linked to 23% of metabolites that were significantly altered in stool after STRESS. Together, pre-STRESS Actinobacteria relative abundance and changes in serum IL-6 and stool cysteine concentrations accounted for 84% of the variability in the change in IP. Findings demonstrate that a multiple-stressor military training environment induced increases in IP that were associated with alterations in markers of inflammation and with intestinal microbiota composition and metabolism. Associations between IP, the pre-STRESS microbiota, and microbiota metabolites suggest that targeting the intestinal microbiota could provide novel strategies for preserving IP during physiological stress.
NEW & NOTEWORTHY Military training, a unique model for studying temporal dynamics of intestinal barrier and intestinal microbiota responses to stress, resulted in increased intestinal permeability concomitant with changes in intestinal microbiota composition and metabolism. Prestress intestinal microbiota composition and changes in fecal concentrations of metabolites linked to the microbiota were associated with increased intestinal permeability. Findings suggest that targeting the intestinal microbiota could provide novel strategies for mitigating increases in intestinal permeability during stress.
the intestinal barrier is a selective physical and immunological barrier that facilitates fluid and nutrient absorption while deterring translocation of potentially harmful luminal antigens into circulation (3). Disruption or dysfunction of the intestinal barrier increases intestinal permeability (IP), initiating a cycle in which translocation of luminal compounds (e.g., bacterial cell wall LPS) can induce immune and inflammatory responses that exacerbate intestinal barrier damage and further increase IP (3, 15, 54). Sequelae of increased IP and subsequent inflammation can include gastrointestinal distress (54), impaired nutrient absorption and metabolism (35), increased susceptibility to illness and infection (53), decrements in cognitive function and physical performance (12), and, if chronic, increased disease risk (19, 53).
The intestinal microbiota and its metabolites are integral mediators of intestinal barrier function and IP, capable of both perturbing and enhancing intestinal barrier integrity by modulating immune responses, oxidative stress, inflammation, vagal signaling, and nutrient availability (40). Intestinal microbiota composition and activity are malleable, influenced by the availability of undigested dietary components (13, 40) and the intestinal environment (e.g., pH, motility, inflammation, and immune activity) (48). Dietary ratios of fiber, carbohydrate, protein, and fat are also important, as low-fiber, high-protein, and high-fat diets have been reported to increase intestinal inflammation and IP by altering ratios of microbes and metabolites that modulate inflammation (13, 14, 41). Severe physical stress (12, 15, 54), psychological stress (34), sleep deprivation and circadian disruption (17, 50), and environmental stressors (9, 24) have also been independently associated with altered intestinal microbiota composition and increased IP. However, current understanding of the role of the intestinal microbiota in mediating effects of physical, psychological, and environmental stressors on the intestinal barrier is largely limited to information derived from animal models, which may not fully represent the human condition (12, 34).
Military training environments offer the opportunity for novel insights into the magnitude, temporal dynamics, and health effects of stress responses within the human intestinal microbiome, as military personnel commonly endure combinations of prolonged physical exertion, psychological stress, sleep deprivation, and environmental extremes during training and combat (31, 51). Transient and chronic gastrointestinal distress (46), suboptimal micronutrient status (21, 36), and cognitive decrements (31) have been reported in military personnel during training and combat. Although underlying etiologies are multifactorial, all are possible sequelae of increased IP, suggesting that intestinal barrier dysfunction and the intestinal microbiota may play a role. Recently, Li et al. reported that gastrointestinal distress during combat training was linked to stress, anxiety, inflammation, and increased intestinal and blood-brain barrier permeability (29, 30). Phua et al. observed changes in urinary concentrations of several metabolites potentially derived from the intestinal microbiota and an association of these changes with gastrointestinal symptomology and IP (44). Although they speculated that changes in intestinal microbiota composition may have contributed to these findings, the authors did not assess microbiota composition.
The present study used a physically demanding military training exercise as a model for elucidating the effects of physiological and metabolic stress on IP and intestinal microbiota composition and activity and to identify associations between dietary intake, IP, inflammation, and the intestinal microbiota. The data were collected during a trial designed to determine the extent to which dietary carbohydrate and protein supplementation spare whole body protein and attenuate decrements in physiological status during military training (32, 43). We hypothesized that the multiple-stressor environment, which was expected to induce negative energy balance and body weight loss, would adversely affect intestinal microbiota composition (e.g., decrease diversity, increase abundance of proinflammatory taxa, and decrease abundance of putatively beneficial taxa) and increase IP. We further hypothesized that supplemental protein would exacerbate these decrements by promoting the generation of potentially harmful bacterially derived metabolites, whereas carbohydrate supplementation would attenuate these decrements by reducing the magnitude of negative energy balance.
METHODS
Participants and experimental design.
Seventy-three Norwegian Army soldiers (71 men and 2 women) participating in a 4-day arctic military training exercise consented to participate in this randomized, controlled trial in January 2015 (32, 43). All soldiers >18 yr of age participating in the training were eligible for the study, which was approved by the Institutional Review Board at the US Army Research Institute of Environmental Medicine and the Regional Committees for Medical and Health Research Ethics (REK sør-øst, Oslo, Norway). Investigators adhered to the policies for protection of human subjects as prescribed in 32 CFR Part 219, US Department of Defense Instruction 3216.02 (Protection of Human Subjects and Adherence to Ethical Standards in DoD-Supported Research) and Army Regulation 70-25. The trial was registered on www.clinicaltrials.gov as NCT02327208.
Study staff block-randomized volunteers by body weight to a control (CNTRL, n = 18), protein-supplemented (PRO, n = 28), or carbohydrate-supplemented (CHO, n = 27) group in a 1:3 (control-to-intervention) ratio. All volunteers were provided with three Norwegian arctic rations per day to consume during the 4-day training exercise. The PRO group was also provided with four whey protein-based snack bars per day, while the CHO group was provided with four carbohydrate-based snack bars per day. Bars were similar in appearance, taste, and texture, enabling investigators, study staff, and volunteers to remain blind to the macronutrient composition. The training consisted of a 51-km cross-country ski-march, during which volunteers skied in 50:10-min work-to-rest ratios while carrying a ~45-kg pack (STRESS). Stool samples were collected over the 2 days before STRESS and on the night of or day after completion of STRESS in a self-selected subset of volunteers. Twenty-four-hour urine collections were completed on the day before STRESS and on day 3 of STRESS. Blood samples were collected on the morning before and the morning after STRESS. Primary study objectives were to determine the effects of macronutrient supplementation on whole body protein balance, body mass, and physiological status during military training and are reported elsewhere (32, 43). This report details secondary study objectives of determining the impact of a multiple-stressor military training environment on IP and intestinal microbiota composition and activity.
Volunteers began consuming provided rations 2 days before training and the intervention snack bars on day 1 of STRESS. Three Norwegian field rations provide 14.6 MJ of energy, 141 g of protein, 435 g of carbohydrate, and 126 g of fat. The four protein-based snack bars provided an additional 4.4 MJ of energy, 85 g of whey protein, 102 g of carbohydrate, 35 g of fat, and <1 g of fiber, while the four carbohydrate-based snack bars provided an additional 4.4 MJ of energy, 11 g of whey protein, 189 g of carbohydrate, 29 g of fat, and 1 g of fiber. All snack bars were manufactured by a third party that did not participate in data collection (Combat Feeding Directorate, Natick Soldier Systems Center, Natick, MA). Investigators, study staff, and volunteers were blind to the macronutrient composition of the bars. Volunteers were asked to consume the rations and bars as they normally would during training and to consume only foods and caloric beverages provided by the study team. All volunteers were provided with ration-specific food logs, which were collected and reviewed daily by study staff and used to calculate actual intakes (Table 1).
| CNTRL (n = 18) | CHO (n = 27) | PRO (n = 28) | |
|---|---|---|---|
| Age, yr | 19 ± 2 | 20 ± 1 | 20 ± 1 |
| BMI, kg/m2 | 23.6 ± 1.8 | 24.1 ± 2.3 | 23.3 ± 2.1 |
| Energy expenditure, MJ/day | 25.5 ± 1.7 | 25.8 ± 2.1 | 25.8 ± 2.5 |
| Energy intake, MJ/day | 10.5 ± 1.7 | 13.1 ± 2.6* | 11.8 ± 2.5 |
| (6.5–13.0) | (7.7–16.4) | (7.1–16.8) | |
| Carbohydrate, g/day | 312 ± 47 | 434 ± 86* | 321 ± 77† |
| (193–385) | (253–543) | (171–490) | |
| Protein, g/day | 100 ± 15 | 98 ± 22 | 148 ± 25*† |
| (65–124) | (58–130) | (96–191) | |
| Fat, g/day | 91 ± 20 | 107 ± 24 | 102 ± 23 |
| (57–117) | (56–146) | (59–141) | |
| Fiber, g/day | 25 ± 4 | 25 ± 6 | 22 ± 5† |
| (16–32) | (13–33) | (12–33) |
IP assay.
IP was assessed by quantifying the urinary excretion of orally ingested sugar substitutes (29, 38). Fasted volunteers consumed a solution of 2 g of sucralose and 4 g of mannitol dissolved in ~180 ml of water and then collected all urine produced over the subsequent 24 h. Sucralose is not degraded by the colonic microbiota, is excreted in proportion to paracellular permeability, and is a common marker for whole-gut IP (38). In contrast, mannitol is used for small bowel permeability measurements (3) but is degraded by the colonic microbiota, which prevents its use for IP measurements >5 h. Mannitol results are presented solely for comparison with a previous study conducted in a military training environment (29). Sucralose and mannitol concentrations were measured by HPLC (model 1100, Agilent, Santa Clara, CA) as previously described (1, 33). For calculation of fractional excretion, the measured concentration of each probe was multiplied by the total volume of urine collected, and the product was divided by the dose administered. Logistical constraints and adverse weather precluded more frequent urine collections and prevented procurement of complete post-STRESS urine collections from 24 volunteers.
Blood biochemistries.
After an overnight fast, blood was collected by antecubital venipuncture, separated into serum or plasma, and immediately frozen. Samples were then shipped on dry ice to the US Army Research Institute of Environmental Medicine, where they were stored at −80°C until they were shipped to Pennington Biomedical Research Center (Baton Rouge, LA) or Metabolon (Durham, NC) for analysis. Plasma LPS was measured by ELISA (Cusabio, College Park, MD), serum IL-6 by multianalyte profiling (MILLIPLEX, Millipore, Billerica, MA), serum high-sensitivity C-reactive protein (CRP) by a chemiluminescent immunometric assay (Immulite 2000, Siemens, Malvern, PA), and serum creatine kinase (a marker of muscle damage) by an automated chemistry analyzer (model DXC 600 Pro, Beckman Coulter, Brea, CA).
Stool microbiota composition.
Stool sample collection was optional to encourage maximal participation for primary study outcomes. A self-selected subset of 38 volunteers provided stool samples; 26 of these volunteers provided both pre- and post-STRESS samples.
Stool samples were collected into provided collection containers, immediately placed on ice, and frozen in ~500-mg aliquots within 12 h of collection. Samples were shipped on dry ice to the US Army Research Institute of Environmental Medicine, where they were stored at −80°C. Samples were then shipped to Metabolon for metabolomics analysis and to the US Army Center for Health and Environmental Research for intestinal microbiota composition analysis.
Samples were selected for DNA extraction in random order, and DNA was extracted using the PowerFecal DNA Isolation kit (MO BIO Laboratories, Qiagen, Carlsbad, CA). Primers designed to amplify the V3–V4 region of the 16S rRNA gene were employed for PCR amplification (22) according to the Illumina 16S Metagenomic Sequencing Library Preparation manual (catalog no. 15044223 Rev B, Illumina, San Diego, CA). A limited-cycle PCR generated a single amplicon of ~460 bp to which Illumina sequencing adapters and dual-index barcodes were added. Paired 300-bp reads and MiSeq v.3 reagents were used to generate full-length reads of the V3 and V4 regions in a single run on the Illumina MiSeq platform.
Sequencing data were processed using Quantitative Insights Into Microbial Ecology (QIIME) v.1.9.1 (8). Read quality assessment, filtering, barcode trimming, and chimera detection were performed on demultiplexed sequences using USEARCH (16). Operational taxonomic units (OTUs) were assigned by clustering sequence reads at 97% similarity. The most abundant sequences with a minimum sequence length of 150 bp were aligned against the Greengenes database core set v.gg_13_15 (37) using PyNAST (7). Taxonomic assignment was completed using the RDP classifier v.2.2 (55).
Stool and plasma metabolomics.
Stool and plasma aliquots from soldiers providing both pre- and post-STRESS stool samples were submitted for global metabolite profiling (Metabolon). Samples were analyzed using two separate reverse-phase (RP)/ultra-high-performance liquid chromatography (UPLC)-tandem mass spectrometry (MS/MS) methods with positive ion mode electrospray ionization (ESI), RP/UPLC-MS/MS with negative ion mode ESI, and hydrophilic interaction chromatography (HILIC)/UPLC-MS/MS with negative ion mode ESI.
Several recovery standards were added before the first step in the extraction process and were analyzed with the experimental samples for quality control. All analysis methods utilized an ACQUITY UPLC (Waters, Milford, MA) and a Thermo Scientific Q-Exactive high-resolution/accurate MS interfaced with a heated ESI-II source and Orbitrap mass analyzer operated at 35,000 mass resolution. Sample extracts were dried and reconstituted in solvents compatible with each of the four methods. Each reconstitution solvent also contained a series of standards at fixed concentrations to ensure injection and chromatographic consistency. One aliquot was analyzed using acidic positive ion conditions chromatographically optimized for more hydrophilic compounds. In this method, the extract was gradient-eluted from a C-18 column (Waters UPLC BEH C-18, 2.1 mm × 100 mm, 1.7 µm) using water and methanol containing 0.05% perfluoropentanoic acid and 0.1% formic acid. Another aliquot was also analyzed using acidic positive ion conditions chromatographically optimized for more hydrophobic compounds. In this method, the extract was gradient-eluted from the same C-18 column using methanol, acetonitrile, water, 0.05% perfluoropentanoic acid, and 0.01% formic acid and operated at an overall higher organic content. Another aliquot was analyzed using basic negative ion optimized conditions and a separate dedicated C-18 column. The basic extracts were gradient-eluted from the column using methanol and water, but with 6.5 mM ammonium bicarbonate at pH 8. The fourth aliquot was analyzed via negative ionization following elution from a HILIC column (Waters UPLC BEH Amide 2.1 mm × 150 mm, 1.7 µm) using a gradient consisting of water and acetonitrile with 10 mM ammonium formate at pH 10.8. The MS analysis alternated between MS and data-dependent MSn scans using dynamic exclusion. The scan range varied slightly between methods but covered 70–1,000 m/z.
Raw data were extracted, peaks were identified, and data were quality control-processed using Metabolon’s proprietary hardware and software. Compounds were identified by comparison with a library that is maintained by Metabolon and contains entries of purified standards or recurrent unknown entities. Biochemical identifications were based on three criteria: retention index within a narrow retention index window of the proposed identification, accurate mass match to the library ±10 ppm, and the MS/MS forward and reverse scores between the experimental data and authentic standards. The MS/MS scores were based on a comparison of the ions present in the experimental spectrum with the ions present in the library spectrum. Peaks were quantified using area under the curve.
Bioinformatics.
Analyses were completed using R v.3.3.1, Multiexperiment Viewer v.4.9.0, SPSS v.21, and XLSTAT v.2015. We obtained an average of 140,762 ± 103,480 16S rDNA sequences per stool sample, which clustered into 2,015 OTUs at 97% sequence identity. OTUs could be assigned to 12 phyla and 83 genera. α-Diversity (Shannon and Chao1 indexes and observed OTUs) was calculated using the phyloseq R bioconductor package, and β-diversity was calculated using Bray-Curtis distances. Prior to statistical analysis of sequencing data, phylum-, genus-, and OTU-level relative abundances were calculated by dividing the number of reads for each taxon by the total number of reads in the sample. Ordination and cluster analyses were conducted on OTU-level relative abundances, whereas differential analyses were conducted on phylum- and genus-level relative abundances. For differential analyses, any OTUs that could not be assigned to the genus level were grouped at the next-lowest level of classification possible (e.g., family or order). Relative abundances were arcsine square-root-transformed before differential analysis to stabilize variance and better approximate normality. Prior to analysis of stool and plasma metabolites, any missing values were imputed using the minimum observed value for each compound, normalized to set the median equal to 1, and log10-transformed.
Ordinations were conducted by principal coordinates analysis (PCoA) of the OTU Bray-Curtis dissimilarity matrix, principal components analysis (PCA) of metabolite data, and hierarchical complete-linkage clustering of Euclidean distances (OTU and metabolite data). Supervised classification of pre- and post-STRESS samples was conducted using random forest analysis, and the mean decrease accuracy was used to identify taxa driving classification. To examine associations between stool microbiota composition and global metabolite profiles, metabolite PCA ordinations were compared with OTU PCoA ordinations using Procrustes analysis implemented in the R package vegan.
A knowledge-based approach was used to better identify microbially derived metabolites by predicting changes in stool metabolite profiles based on changes in stool microbiota composition. For these analyses, PICRUSt v.1.0.0 was first used to predict metagenome functional content from 16S rDNA data (26). Final metagenome functional predictions were performed by multiplying normalized OTU abundance by each predicted functional profile. Differences in predicted metagenomic profiles were examined by comparing KEGG orthologs between pre- and post-STRESS samples and PCA. Changes in metagenome functional counts over time were examined following Trimmed Mean of M component normalization by fitting linear models using moderated standard errors and the empirical Bayes model. Metabolites predicted to derive from significantly altered KEGG orthologs (P ≤ 0.05) were annotated using HMDB v.2.5, KEGG v.80.0 (compounds, pathways, orthologs, and reactions), SMPDB v.2.0, and FOODB v.1.0. These metabolites were then compared with the list of metabolites in stool that increased or decreased over time (P < 0.10). Overlapping metabolites were considered indicative of functional relationships between changes in the microbiome and the metabolome.
Statistical analysis.
Sample size calculations were based on primary study outcomes, which are reported elsewhere (32, 43). Statistical analyses were completed using SPSS v.21 and R v.3.3.1. Data were assessed for normality before analysis and transformed if necessary to meet model assumptions. When transformation was not successful, nonparametric tests were used. Repeated-measures ANOVA was used to test effects of STRESS and diet and their interaction on study outcomes. Pair-wise comparisons of pre- and post-STRESS genus relative abundances were conducted using Wilcoxon’s signed rank test, and between-group comparisons of changes in genus relative abundances were conducted using the Kruskal-Wallis test. Spearman’s rank correlation (ρ), Pearson’s correlation (r), multiple linear regression, and linear mixed models were used to examine associations among variables. Relationships between surcalose excretion and LPS, IL-6, and CRP concentrations and ordinations of stool microbiota composition and stool/plasma metabolites were also assessed using linear mixed models. All mixed models included subject as a random factor and time as a continuous covariate. Sucralose excretion or LPS, IL-6, or CRP concentration was entered as dependent variable, and scores for the first three principal components of the ordinations were included as independent variables. Finally, backward stepwise regression was used to identify the strongest predictors of changes in IP. Independent variables included in the regression model were those that were significantly correlated with changes in sucralose excretion and included dietary parameters (protein intake), change scores for inflammation markers (IL-6 and CRP), pre-STRESS stool microbiota characteristics (Shannon diversity and Actinobacteria and Proteobacteria relative abundances), and change scores for stool metabolites linked to changes in microbiota composition changes (cysteine and arginine). Changes in Shannon diversity and pre-STRESS Sutterella relative abundance were also considered in place of pre-STRESS Shannon diversity and Proteobacteria relative abundance, respectively.
The false discovery rate for all tests including taxa or metabolite data was controlled by adjusting P values using the Benjamini-Hochberg procedure. Adjusted P values are presented as Q values. Values are means ± SD unless otherwise noted. Statistical significance was set at P ≤ 0.05 or Q ≤ 0.10.
RESULTS
Macronutrient intakes varied across study groups as planned (Table 1). Specifically, mean protein intake was higher in PRO than CNTRL and CHO (P < 0.05), mean carbohydrate intake was higher in CHO than CNTRL and PRO (P < 0.05), and fat intake did not differ between groups. Energy intake was higher in CHO than CNTRL and PRO (P < 0.05; Table 1). Energy expenditure was high, averaging 25.7 ± 2.2 MJ/day and did not differ between groups (32). The high energy expenditure resulted in a 55% energy deficit and 2.7 ± 1.2 kg loss of body mass, which also did not differ between groups (32, 43). Serum creatine kinase, IL-6, and CRP concentrations are reported elsewhere (43). All increased during STRESS independent of diet group, indicating that muscle damage and inflammation were induced during STRESS.
The volunteers who chose to provide stool samples were men and did not differ in age (P = 0.59), body mass index (P = 0.47), or body mass loss (P = 0.98) or change in intestinal permeability (P = 0.42), energy intake (P = 0.51), macronutrient intake (P ≥ 0.11), or energy expenditure (P = 0.94) during STRESS relative to volunteers who chose not to provide stool samples.
IP, plasma LPS, and inflammation.
Sucralose excretion increased 62 ± 57% during STRESS independent of diet (main effect of time, P < 0.001; Fig. 1A), suggesting increased IP, and was correlated with changes in creatine kinase (r = 0.34, P = 0.02), CRP (ρ = 0.36, P = 0.01), IL-6 (Fig. 1B), and protein intake (ρ = −0.31, P = 0.03). Mannitol excretion also increased during STRESS independent of diet [28 ± 8% (pre-STRESS) vs. 33 ± 13% (post-STRESS), main effect of time, P = 0.01]. Plasma LPS concentrations did not differ from pre- to post-STRESS (P = 0.79; Fig. 1C). However, soldiers with increased LPS concentrations demonstrated a trend to greater increases in IL-6 concentration than those with no change or a decrease in LPS concentration (Fig. 1D).

Fig. 1.Intestinal permeability (IP), plasma LPS, and inflammation during military training. A and C: intestinal permeability (IP) measured by 24-h urine collection following ingestion of 2 g of sucralose (n = 49) and plasma LPS concentrations (n = 67) before (PRE) and after (POST) military training. Boxes, median and interquartile range; whiskers, 1.5 times the interquartile range, or minimum and maximum if no observations within that range; circles, data points >1.5 times the interquartile range. *P < 0.001 (repeated-measures ANOVA, main effect of time). B: correlation of changes in IP with changes in serum IL-6 concentrations (Pearson’s correlation, n = 46). D: a trend for larger increases in serum IL-6 in soldiers experiencing increases in plasma LPS during training than in those experiencing a decrease or no change in plasma LPS (P = 0.07, repeated-measures ANOVA, time × ΔLPS interaction). #P < 0.05 vs. PRE. CHO, carbohydrate-supplement group; CNTRL, control group (rations only); PRO, protein-supplement group.
Stool microbiota composition.
The Shannon α-diversity index increased during STRESS independent of diet (main effect of time, P = 0.04), whereas the Chao1 index (main effect of time, P = 0.42) and total observed OTUs (main effect of time, P = 0.45) were not affected by STRESS or diet, indicating an increase in the evenness, but not the richness, of the stool microbiota (Fig. 2A). PCoA (Fig. 2B) and cluster analysis (Fig. 2C) demonstrated an effect of STRESS on the microbiota independent of diet. Random forest analysis differentiated pre- and post-STRESS samples with 100% accuracy. The top 10 taxa contributing to the high prediction accuracy were Peptostreptococcus, Christensenella, Faecalibacterium, Staphylococcus, unassigned taxa within the Mogiobacteriaceae, Christensenellaceae, and Planococcaceae families, and unassigned taxa within the CW040 and RF39 orders (see Supplemental Table S1 in Supplemental Material for this article available online at the Journal website). At the phylum level, decreases in Bacteroidetes and increases in Firmicutes and several other phyla were observed (Q < 0.10; Fig. 2D). At the genus level, changes in the relative abundance of 48 of 83 identified genera were observed (Q < 0.10; see Supplemental Table S1). Changes in genus relative abundances did not differ by diet group (Q > 0.75 for all).

Fig. 2.Military training elicits changes in intestinal microbiota composition. A: α-diversity before (PRE) and after (POST) military training. Boxes, median and interquartile range; whiskers, 1.5 times the interquartile range, or minimum and maximum if no observations within that range; circles, data points >1.5 times the interquartile range. *P = 0.04 (repeated-measures ANOVA, main effect of time). B: principal coordinates (PC) analysis of Bray-Curtis dissimilarity matrix indicates that composition of the stool microbiota community was more strongly influenced by training environment than by individual variability or diet group. Data points represent the stool microbiota community of a single individual. Points closer together are more similar. C: hierarchical complete-linkage clustering of Euclidean distances of operational taxonomic unit (OTU) relative abundances measured in stool collected before and after military training (n = 38). Colored bars are data points representing the stool microbiota composition of an individual. Branches (lines) within the same node (points where branches split) reflect similarity in composition of the stool microbiota community. Clustering of branches by time point indicates that composition of the stool microbiota community was more strongly influenced by training environment than by individual variability or diet group. D: phylum-level shifts in gut microbiota composition. Bars, mean relative abundances. Arrows indicate direction of change in relative abundance from PRE to POST. *P < 0.05 (repeated-measures ANOVA, main effect of time). CHO, carbohydrate-supplement group (n = 9); CNTRL, control group (rations only, n = 5); PRO, protein-supplement group (n = 12).
Stool and plasma metabolites.
A total of 694 compounds were identified in stool. PCA (Fig. 3A) and cluster analysis (Fig. 3B) of these compounds did not suggest an effect of time point or diet. However, random forest analysis correctly differentiated pre- and post-STRESS stool samples with 84% accuracy (Fig. 3C), and 274 compounds demonstrated statistically significant changes (Q < 0.10). Of these, 81%, including several metabolites of amino acid, fatty acid, carbohydrate, and energy metabolism, decreased during STRESS (see Supplemental Table S2). Secondary bile acids and amino acid metabolites (Fig. 4) known to be solely or partially derived from microbial metabolism were generally decreased as well or unchanged, with the notable exception of p-cresol, a microbial metabolite of tyrosine fermentation, which was increased in stool post-STRESS.

Fig. 3.Stool metabolomics before (PRE) and after (POST) military training. A–C: principal components (PC), hierarchical complete-linkage clustering of Euclidean distances, and random forest analyses of stool metabolites (n = 25). A: individual data points represent metabolite composition within a single individual. Points closer together are more similar. B: columns are individuals and rows are metabolites shaded by abundance within a sample. Branches (lines) within the same node (points where branches split) reflect similarity in metabolite composition. Stool metabolites did not demonstrate a distinct clustering pattern. C: top-30 metabolites with the strongest influence on prediction accuracy of the random forest analysis presented in order of importance (top to bottom). Random forest analysis used individual metabolite profiles to predict whether the samples were from pre- or posttraining. Mean decrease in prediction accuracy is the mean decrease in the percentage of observations classified correctly when that metabolite is assigned a random value. Arrows indicate direction of metabolite change from pre- to posttraining. CHO, carbohydrate-supplement group; CNTRL, control group (rations only); PRO, protein-supplement group.

Fig. 4.Qualitative changes in phenylalanine and tyrosine (A) and tryptophan (B) metabolites in stool and plasma during military training. Arrows indicate direction of change in stool (black) and plasma (gray) from pre- to posttraining (repeated-measures ANOVA, main effect of time, Q < 0.10). Metabolites circled by dashed line are compounds known to be wholly or partially derived from microbial metabolism. Compounds without arrows were either unchanged (Q > 0.10) or not detected.
A total of 737 compounds were identified in plasma; of these, 478 demonstrated statistically significant changes during STRESS (Q < 0.10). Changes primarily reflected increases in host energy metabolism, lipolysis, fatty acid oxidation, branched-chain amino acid catabolism, and steroid metabolism (data not shown). However, changes in plasma concentrations of several metabolites known to be partially or fully derived from microbial metabolism were also observed. Specifically, mean concentrations of phenylalanine and tyrosine metabolites, including p-cresol sulfate (+48%), p-cresol glucuronide (+79%), phenylacetate (+44%), phenyllactate (+42%), phenylacetylglutamine (+24%), and 3-(4-hydroxyphenyl)lactate (+40%), were increased (Fig. 4). In contrast, mean concentrations of the benzoate metabolites 2-hydroxyhippurate (−22%), 3-hyroxyhippurate (−61%), and 4-hyroxyhippurate (−35%) were decreased (Q < 0.10). Mean concentrations of secondary bile acids in plasma demonstrated more variable responses: glycolithocolate sulfate (+21%), glycohyocholate (+6%), taurolithocholate 3-sulfate (+89%), and taurocholenate sulfate (+56%) concentrations increased, while deoxycholate (−66%), ursodeoxycholate (−63%), and isoursodeoxycholate (−51%) concentrations decreased (Q < 0.10).
Associations between stool microbiota composition, stool and plasma metabolites, IP, and inflammation.
Changes in sucralose excretion were inversely associated with pre-STRESS Shannon diversity (ρ = −0.43, P = 0.05) and Actinobacteria relative abundance (ρ = −0.53, Q = 0.09) and positively correlated with pre-STRESS Proteobacteria (ρ = 0.64, Q = 0.02) and Sutterella (ρ = 0.68, Q = 0.09) relative abundance (Fig. 5; see Supplemental Table S1) and changes in Shannon diversity (ρ = 0.58, P = 0.02). No statistically significant correlations between the pre-STRESS relative abundance of any taxa or the change in relative abundance of any taxa and changes in LPS, IL-6, or CRP concentration were detected. Additionally, no association between these variables and scores extracted from the first three principal components of the stool microbiota PCoA was detected.

Fig. 5.Factors associated with increased IP during military training (A–D). IP was measured by 24-h urine collection following ingestion of 2 g of sucralose [Spearman’s correlation (ρ), n = 21]. P values for correlations with taxa were adjusted using the Benjamini-Hochberg correction (Q).
Procrustes analysis demonstrated a significant association between the ordinations of stool metabolites and stool microbiota composition (M2 = 0.76, Monte Carlo P = 0.001; Fig. 6A), indicating an association between stool metabolites and the stool microbiota. Additionally, prediction models linked changes in stool microbiota composition to 69 of the metabolites found to be altered in stool (see Supplemental Table S3). These models were supported by Procrustes analysis on ordinations of the significantly altered taxa and these metabolites (M2 = 0.72, Monte Carlo P = 0.001). Of the 69 metabolites, amino acid and nucleotide metabolites comprised the majority and were generally lower post-STRESS than pre-STRESS (Q < 0.10). Changes in two of these metabolites, arginine and cysteine, were correlated with changes in sucralose excretion during STRESS (Table 2). Changes in the concentrations of another 14 metabolites were also inversely correlated with changes in sucralose excretion (Table 2). Scores on the first principal component from the ordination of stool metabolite data were associated with sucralose excretion (β ± SE = −0.05 ± 0.01, P = 0.01), indicating that the effect of STRESS on stool microbiota was associated with IP.

Fig. 6.Stool microbiota composition is associated with stool metabolite and plasma metabolite concentrations. Procrustes analysis of stool microbiota data ordinated using principal coordinates analysis of Bray-Curtis distances and stool (A) and plasma (B) metabolite profiles ordinated using principal components analysis. The first 3 components of each ordination were extracted and analyzed using Procrustes rotation, which attempts to rotate ordinations to maximal similarity. Circles, stool microbiota community of a single individual before or after military training; arrowheads, stool or plasma metabolite profile of a single individual before or after military training. Vectors connect microbiota composition with metabolite profiles of the same individual for each time point. Longer vectors indicate greater intraindividual dissimilarity. The fit of each Procrustes rotation over the first 3 dimensions is reported as the M2 value. P values were calculated after 1,000 permutations. Results indicate similar clustering patterns between stool microbiota composition and stool metabolites and between stool microbiota composition and plasma metabolites. C: Venn diagram of stool and plasma metabolites that were significantly altered during military training (Q ≤ 0.10). Prediction models linked changes in stool microbiota composition to 69 of the metabolites found to be altered in stool, 30 of which were also significantly altered in plasma.
| Super Pathway | Sub Pathway | Biochemical Name | ρ | P Value | Q Value |
|---|---|---|---|---|---|
| Amino acid | Leucine, isoleucine and valine metabolism | 3-Methylglutaconate | −0.75 | 0.001 | 0.05 |
| Methionine, cysteine, S-adenosylmethionine, and taurine metabolism | N-acetyltaurine | −0.73 | 0.001 | 0.06 | |
| l-Cysteine | −0.70 | 0.003 | 0.07 | ||
| Taurine | −0.68 | 0.004 | 0.08 | ||
| N-acetylmethionine sulfoxide | −0.67 | 0.005 | 0.09 | ||
| Polyamine metabolism | N-acetylputrescine* | −0.78 | <0.001 | 0.05 | |
| Urea cycle; arginine and proline metabolism | l-Arginine | −0.70 | 0.002 | 0.06 | |
| Carbohydrate | Amino sugar metabolism | Glucuronate | −0.68 | 0.004 | 0.08 |
| Cofactors and vitamins | Nicotinate and nicotinamide metabolism | Nicotinate ribonucleoside | −0.69 | 0.003 | 0.07 |
| Lipid | Endocannabinoid | Linoleoyl ethanolamide | −0.75 | 0.001 | 0.05 |
| Oleoyl ethanolamide | −0.71 | 0.002 | 0.06 | ||
| Mevalonate metabolism | Mevalonate | −0.71 | 0.002 | 0.06 | |
| Phospholipid metabolism | Trimethylamine N-oxide | −0.71 | 0.002 | 0.06 | |
| Secondary bile acid metabolism | 7-Ketodeoxycholate | −0.86 | <0.001 | 0.01 | |
| 12-Dehydrocholate | −0.71 | 0.002 | 0.06 | ||
| Xenobiotics | Xanthine metabolism | 1-Methylxanthine | −0.76 | 0.001 | 0.05 |
Procrustes analysis also demonstrated a significant association between the ordinations of plasma metabolites and stool microbiota composition (M2 = 0.49, Monte Carlo P = 0.001; Fig. 6B), indicating an association between plasma metabolites and the stool microbiota. Furthermore, plasma concentrations of 30 of the 69 metabolites that linked the stool microbiota to the stool metabolome in prediction models were altered (Fig. 6C; see Supplemental Table S3). However, plasma metabolite changes were not correlated with changes in sucralose excretion or IL-6 or CRP concentration.
Backward stepwise regression was used to identify the strongest predictors of changes in IP. The final model comprising pre-STRESS Actinobacteria relative abundance, change in serum IL-6 concentrations, and changes in stool cysteine concentrations explained 84% of the variability in the change in sucralose excretion (Table 3). Collectively, these findings demonstrate an association between intestinal microbiota composition, stool metabolite concentrations, and changes in IP.
| β ± SE | Standardized β | P Value | |
|---|---|---|---|
| Actinobacteria relative abundance (pre-STRESS) | −45.0 ± 8.5 | −0.59 | <0.001 |
| Δlog10 IL-6 (pg/ml) | 0.4 ± 0.6 | 0.43 | 0.003 |
| Δlog10 stool cysteine | −2.4 ± 0.6 | −0.43 | <0.001 |
| Intercept | 1.4 ± 0.3 | <0.001 | |
| Adjusted R2 = 0.84 | <0.001 |
DISCUSSION
The magnitude, temporal dynamics, and physiological effects of intestinal microbiome responses to stress are poorly characterized. Our findings demonstrate that a multiple-stressor environment characterized by high physical exertion, suboptimal energy intake, muscle damage, and inflammation adversely affects intestinal barrier integrity concomitant with alterations in intestinal microbiota composition and metabolism. Associations between increased IP, the pre-STRESS microbiota, and stool metabolites associated with the microbiota suggest that targeting the intestinal microbiota could provide novel strategies for maintaining intestinal barrier integrity during physiological stress.
The increase in IP in association with increased inflammation (Fig. 1) is consistent with the only other study to our knowledge that has assessed IP in military personnel during training (29). In these environments, intense or prolonged exercise may reduce splanchnic perfusion, which can trigger intestinal hypoxia, inflammation, and oxidative stress, which collectively degrade intestinal barrier integrity and increase IP (15, 24, 54). Stress-induced muscle damage may also contribute to inflammation, potentiating increases in IP by inducing tight junction dysfunction (15). Ultimately, the increase in IP is thought to result in mild endotoxemia and inflammation and contribute to gastrointestinal distress in endurance athletes (4, 15, 20, 24) and possibly military personnel (29). Although gastrointestinal symptoms were not assessed in the present study, in the study of Li et al. (29), 70% of soldiers participating in a 6-wk combat training course reported gastrointestinal distress symptomology (i.e., abdominal pain, diarrhea, and constipation); these symptoms were more frequent in soldiers with the largest increases in IP and were associated with psychological decrements. Gastrointestinal distress, to include infectious diarrhea, is historically the leading nonbattle injury encountered in deployed military personnel, representing a significant burden to military health care and operational readiness (45–47). Identification of mediators of intestinal barrier responses to severe stress and development of strategies to target those mediators may therefore have substantial benefit for military personnel.
Our findings suggest that the intestinal microbiota may be one mediator of IP responses to severe physiological stress and that targeting the microbiota before stress exposure may be one strategy for maintaining IP. In particular, increasing microbiota diversity and Actinobacteria relative abundance and decreasing Proteobacteria and Sutterella relative abundances before stress exposure may be effective in lieu of the observed associations with changes in IP during stress (Fig. 5). Greater microbiota diversity is generally considered indicative of a healthy intestinal ecosystem, having been frequently associated with lower chronic disease risk (11, 19). Similarly, species within the Actinobacteria phylum, including those belonging to the Bifidobacterium and Collinsella genera, have favorable anti-inflammatory and immunomodulatory effects that may protect the intestinal barrier during stress (2, 42). Bifidobacterium strains are included in multistrain probiotics that have demonstrated some, although weak, efficacy for favorably impacting IP in athletes (25, 49). Use of prebiotics such as oligofructose to increase Bifidobacterium relative abundance has also been shown to promote intestinal barrier integrity in animal models (6). In contrast, Proteobacteria are endotoxin producers that have been linked to inflammatory bowel diseases and subclinical inflammation (19, 27). Sutterella, a genus within the Proteobacteria phylum, has been shown to promote inflammatory bowel disease by inhibiting immunoglobulin A secretion (39). As such, although findings are correlative and the study design precludes determination of causality, the observed associations between the pre-STRESS microbiota and changes in IP during STRESS are plausible and provide potential targets for further study.
To our knowledge, this study is the first to examine intestinal microbiota responses during military training and expands knowledge regarding the temporal effects of exercise and psychological stress on the microbiome, which is largely limited to animal studies at present (12, 34). Human studies have demonstrated that drastic changes in diet impact intestinal microbiota composition (13, 41) by altering the availability of metabolic substrates for intestinal microbes (23). In contrast to these earlier reports, our findings demonstrate alterations in microbiota composition that most likely were not solely attributable to diet and were more pronounced than those commonly reported in human diet studies (Fig. 2). Although potential mechanisms were not directly assessed, changes in immune activity, intestinal inflammation and oxidative stress, and altered hypothalamic-pituitary-adrenal axis and vagal signaling have been postulated as mechanisms through which physical and psychological stress modulate the microbiome (12, 34).
The increase in Shannon α-diversity and the numerous genus-level changes in relative abundance demonstrated that changes in microbiota composition were broadly characterized by an increase in abundance of less dominant taxa at the expense of more dominant taxa such as Bacteroides (Fig. 2). This included increased relative abundances of several potentially deleterious and infectious taxa (e.g., Peptostreptococcus, Staphylococcus, Peptoniphilus, Acidaminococcus, and Fusobacterium) and decreased relative abundances of several taxa thought to deter pathogen invasion, reduce inflammation, and promote immunity (e.g., Bacteroides, Faecalibacterium, Collinsella, and Roseburia). As such, an increase in the ratio of less-abundant, potentially harmful taxa to beneficial taxa may explain the unexpected observation that greater increases in diversity during training were correlated with larger increases in IP. However, several alternative explanations exist. Individuals with the lowest pre-STRESS Shannon diversity also demonstrated the largest increases in diversity during STRESS (r = −0.60, P = 0.001). Therefore, the association between increased diversity and increased IP may be attributable to lower pre-STRESS diversity. Alternately, higher stool microbiota diversity has been correlated with longer intestinal transit time and higher urinary concentrations of potentially harmful degradation products of bacterial protein metabolism (48). In this study, stool and plasma concentrations of protein degradation products did not uniformly change, although they were more commonly decreased in stool and increased in plasma (Fig. 4; see Supplemental Tables S2 and S3). Whether these observations reflect changes in transit time could not be determined from the collected data. Nonetheless, no protein degradation metabolite was independently associated with increased IP or inflammation. This observation contrasts with reports that bacterial protein metabolites induce intestinal barrier damage and inflammation in vitro (56) and suggests that the positive association between protein intake and increases in IP during training was not mediated by bacterial metabolism of diet-derived amino acids.
Decreased concentrations of several stool metabolites were associated with increased IP (Table 2). Metabolites included two amino acids, arginine and cysteine, which were predicted to be associated with changes in microbiota composition and are plausible modulators of IP on the basis of known physiological functions. Specifically, arginine is a precursor to polyamines required for intestinal mucosal growth and repair and for nitric oxide, a potent vasodilator that may protect intestinal barrier integrity by improving splanchnic perfusion, deterring pathogen invasion, and modulating inflammation (28, 54). It has been reported that arginine supplementation preserves intestinal barrier integrity in various animal stress and intestinal injury models (2), although the effects in humans are less clear (5). Cysteine is an essential component of glutathione, an antioxidant tripeptide critical to maintaining a favorable redox balance in the intestine (10). Phua et al. (44) recently reported that increases in urinary concentrations of a glutathione metabolite, possibly reflecting increased oxidative stress, were associated with gastrointestinal symptomology during military training. Our findings also suggest that interactions between the intestinal microbiota and dietary fat metabolism may impact IP (Table 2). 7-Ketodeoxycholate and 12-dehydrocholate are secondary bile acids derived from bacterial metabolism of bile acids secreted in response to dietary fat intake. Secondary bile acids are recognized as important signaling molecules with functions that are thought to include promotion of gut barrier integrity (52). Collectively, these findings suggest that changes in intestinal microbiota composition and metabolism may impact IP during physiological stress by modulating the availability of amino acid precursors critical to moderating inflammation and oxidative stress and of secondary bile acids.
Study strengths include the provision of diets of known composition providing a range of macronutrient intakes and the integration of physiological, stool microbiota composition, and metabolomics data. However, results should be interpreted in the context of the study design and several limitations. The physically demanding environment coupled with the physiological demands imposed by undereating may have masked some associations and limited generalizability of the findings but provides unique and novel insights into the temporal dynamics of host-microbiome interactions during prolonged physical stress. While psychological and sleep deprivation stresses were likely also present, we did not quantify those responses. Study participants were predominantly young men, and findings may not be generalizable to older populations or women. Limitations include the correlative nature of associations between outcomes from which causality cannot be determined, despite evidence of plausibility, and limited statistical power for some analyses, especially those including between-group comparisons resulting from only a subset of the full cohort participating in stool collections. The method for measuring plasma LPS concentrations is also a limitation, as it did not quantify endotoxin activity, which is known to vary between LPS forms (18). Nonetheless, the weak association between changes in plasma LPS and IL-6 concentrations is consistent with the well-established proinflammatory effects of the compound (18). Inclusion of metagenomic or transcriptomic analysis of stool samples would have strengthened findings and complemented the metabolomics analysis by allowing more accurate functional predictions of microbiota function. Reliance on stool for measurements of microbiota composition and metabolites is also a limitation, as the composition of the stool may be more reflective of the distal colon than the entirety of the gastrointestinal tract. However, plasma metabolite measurements were included to better capture bacterial metabolism along the full gastrointestinal tract. Finally, logistical constraints prevented more frequent measurements, which would have provided additional insight into temporal dynamics.
In this study a systems biology approach was used to confirm the hypothesis that a multiple-stressor environment can induce increases in IP that are associated with inflammation, as well as intestinal microbiota composition and metabolism. Furthermore, these findings extend the current evidence base by demonstrating that such environments can induce rapid and pronounced changes in the intestinal microbiota and suggest that the pre-STRESS intestinal microbiota and changes in microbial metabolism may be important for mediating intestinal barrier responses to stress. As such, targeting the intestinal microbiota could provide novel strategies for mitigating increases in IP and associated sequelae induced by physically and psychologically demanding environments.
GRANTS
This work was supported by the US Army Medical Research and Material Command, the US Defense Health Agency, and the Norwegian Defense Research Establishment under agreement W81XWH-12-0279.
DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the authors.
DISCLAIMERS
The opinions or assertions contained herein are the private views of the author(s) and are not to be construed as official or as reflecting the views of the Army or the Department of Defense. Citation of commercial organizations or trade names in this report does not constitute an official Department of the Army endorsement or approval of the products or services of these organizations. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the US Army.
AUTHOR CONTRIBUTIONS
J.P.K., L.M.M., E.H.M., J.W.C., Y.G., A.V.H., M.W.L., R.K., N.C., A.G., R.H., S.M., S.J.M., and S.M.P. conceived and designed research; J.P.K., L.M.M., E.H.M., N.E.M., J.W.C., Y.G., A.V.H., M.W.L., S.M., and S.J.M. performed experiments; J.P.K., A.V.H., M.W.L., and R.K. analyzed data; J.P.K., R.K., N.C., and A.G. interpreted results of experiments; J.P.K. prepared figures; J.P.K. drafted manuscript; J.P.K., L.M.M., E.H.M., N.E.M., J.W.C., Y.G., A.V.H., M.W.L., R.K., N.C., A.G., R.H., S.M., S.J.M., and S.M.P. edited and revised manuscript; J.P.K., L.M.M., E.H.M., N.E.M., J.W.C., Y.G., A.V.H., M.W.L., R.K., N.C., A.G., R.H., S.M., S.J.M., and S.M.P. approved final version of manuscript.
ACKNOWLEDGMENTS
We thank the study volunteers and command staff from the 2nd Battalion, Brigade North, Norway for their participation and support, Dr. Jennifer Rood and her staff at the Pennington Biomedical Research Center (Baton Rouge, LA) for assistance with biochemical assays, Paul Maguire and Danielle Anderson for their assistance developing the snack bars, and Ingjerd Thrane, Hilde Teien, Pål Stenberg, Christopher Carrigan, Albert Bohn, Anthony Karis, Jamie Templar, Myra Jones, and Andrei Loban for significant contributions to the study.
REFERENCES
- 1. . A simple method for the analysis of urinary sucralose for use in tests of intestinal permeability. Ann Clin Biochem 42: 224–226, 2005. doi:10.1258/0004563053857923.
Crossref | PubMed | ISI | Google Scholar - 2. . The role of immunomodulators on intestinal barrier homeostasis in experimental models. Clin Nutr 34: 1080–1087, 2015. doi:10.1016/j.clnu.2015.01.012.
Crossref | PubMed | ISI | Google Scholar - 3. . Intestinal permeability—a new target for disease prevention and therapy. BMC Gastroenterol 14: 189, 2014. doi:10.1186/s12876-014-0189-7.
Crossref | PubMed | ISI | Google Scholar - 4. . Strenuous exercise causes systemic endotoxemia. J Appl Physiol (1985) 65: 106–108, 1988.
Link | ISI | Google Scholar - 5. . The effect of arginine or glycine supplementation on gastrointestinal function, muscle injury, serum amino acid concentrations and performance during a marathon run. Int J Sports Med 20: 315–321, 1999. doi:10.1055/s-2007-971137.
Crossref | PubMed | ISI | Google Scholar - 6. . Selective increases of bifidobacteria in gut microflora improve high-fat-diet-induced diabetes in mice through a mechanism associated with endotoxaemia. Diabetologia 50: 2374–2383, 2007. doi:10.1007/s00125-007-0791-0.
Crossref | PubMed | ISI | Google Scholar - 7. . PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 26: 266–267, 2010. doi:10.1093/bioinformatics/btp636.
Crossref | PubMed | ISI | Google Scholar - 8. . QIIME allows analysis of high-throughput community sequencing data. Nat Methods 7: 335–336, 2010. doi:10.1038/nmeth.f.303.
Crossref | PubMed | ISI | Google Scholar - 9. . Gut microbiota orchestrates energy homeostasis during cold. Cell 163: 1360–1374, 2015. doi:10.1016/j.cell.2015.11.004.
Crossref | PubMed | ISI | Google Scholar - 10. . Intestinal redox biology and oxidative stress. Semin Cell Dev Biol 23: 729–737, 2012. doi:10.1016/j.semcdb.2012.03.014.
Crossref | PubMed | ISI | Google Scholar - 11. . Gut microbiota composition correlates with diet and health in the elderly. Nature 488: 178–184, 2012.
Crossref | PubMed | ISI | Google Scholar - 12. . Exercise-induced stress behavior, gut-microbiota-brain axis and diet: a systematic review for athletes. J Int Soc Sports Nutr 13: 43, 2016. doi:10.1186/s12970-016-0155-6.
Crossref | PubMed | ISI | Google Scholar - 13. . Diet rapidly and reproducibly alters the human gut microbiome. Nature 505: 559–563, 2014. doi:10.1038/nature12820.
Crossref | PubMed | ISI | Google Scholar - 14. . A dietary fiber-deprived gut microbiota degrades the colonic mucus barrier and enhances pathogen susceptibility. Cell 167: 1339–1353, 2016. doi:10.1016/j.cell.2016.10.043.
Crossref | PubMed | ISI | Google Scholar - 15. . Intestinal epithelial barrier function and tight junction proteins with heat and exercise. J Appl Physiol (1985) 120: 692–701, 2016. doi:10.1152/japplphysiol.00536.2015.
Link | ISI | Google Scholar - 16. . Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26: 2460–2461, 2010. doi:10.1093/bioinformatics/btq461.
Crossref | PubMed | ISI | Google Scholar - 17. . Systemic bacterial invasion induced by sleep deprivation. Am J Physiol Regul Integr Comp Physiol 278: R905–R916, 2000.
Link | ISI | Google Scholar - 18. . Endotoxemia: methods of detection and clinical correlates. Clin Microbiol Rev 8: 268–292, 1995.
Crossref | PubMed | ISI | Google Scholar - 19. . Inflammatory bowel disease as a model for translating the microbiome. Immunity 40: 843–854, 2014. doi:10.1016/j.immuni.2014.05.013.
Crossref | PubMed | ISI | Google Scholar - 20. . Relationship between gastro-intestinal complaints and endotoxaemia, cytokine release and the acute-phase reaction during and after a long-distance triathlon in highly trained men. Clin Sci (Lond) 98: 47–55, 2000. doi:10.1042/cs0980047.
Crossref | PubMed | ISI | Google Scholar - 21. . Randomized, double-blind, placebo-controlled trial of an iron-fortified food product in female soldiers during military training: relations between iron status, serum hepcidin, and inflammation. Am J Clin Nutr 92: 93–100, 2010. doi:10.3945/ajcn.2010.29185.
Crossref | PubMed | ISI | Google Scholar - 22. . Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Res 41: e1, 2013. doi:10.1093/nar/gks808.
Crossref | PubMed | ISI | Google Scholar - 23. . How glycan metabolism shapes the human gut microbiota. Nat Rev Microbiol 10: 323–335, 2012. doi:10.1038/nrmicro2746.
Crossref | PubMed | ISI | Google Scholar - 24. . Intestinal barrier dysfunction, endotoxemia, and gastrointestinal symptoms: the “canary in the coal mine” during exercise-heat stress? Med Sport Sci 53: 61–73, 2008. doi:10.1159/000151550.
Crossref | PubMed | Google Scholar - 25. . Exercise, intestinal barrier dysfunction and probiotic supplementation. Med Sport Sci 59: 47–56, 2012. doi:10.1159/000342169.
Crossref | PubMed | Google Scholar - 26. . Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat Biotechnol 31: 814–821, 2013. doi:10.1038/nbt.2676.
Crossref | PubMed | ISI | Google Scholar - 27. . Richness of human gut microbiome correlates with metabolic markers. Nature 500: 541–546, 2013. doi:10.1038/nature12506.
Crossref | PubMed | ISI | Google Scholar - 28. . Amino acids and immune function. Br J Nutr 98: 237–252, 2007. doi:10.1017/S000711450769936X.
Crossref | PubMed | ISI | Google Scholar - 29. . Combat-training increases intestinal permeability, immune activation and gastrointestinal symptoms in soldiers. Aliment Pharmacol Ther 37: 799–809, 2013. doi:10.1111/apt.12269.
Crossref | PubMed | ISI | Google Scholar - 30. . Combat-training stress in soldiers increases S100B, a marker of increased blood-brain-barrier permeability, and induces immune activation. Neuro Endocrinol Lett 35: 58–63, 2014.
PubMed | ISI | Google Scholar - 31. . The fog of war: decrements in cognitive performance and mood associated with combat-like stress. Aviat Space Environ Med 76 Suppl: C7–C14, 2005.
PubMed | Google Scholar - 32. . Effects of supplemental energy on protein balance during 4-d arctic military training. Med Sci Sports Exerc 48: 1604–1612, 2016. doi:10.1249/MSS.0000000000000944.
Crossref | PubMed | ISI | Google Scholar - 33. . Simultaneous HPLC determination with light-scattering detection of lactulose and mannitol in studies of intestinal permeability in pediatrics. Clin Chem 44: 1685–1691, 1998.
Crossref | PubMed | ISI | Google Scholar - 34. . Gut/brain axis and the microbiota. J Clin Invest 125: 926–938, 2015. doi:10.1172/JCI76304.
Crossref | PubMed | ISI | Google Scholar - 35. . Iron deficiency and obesity: the contribution of inflammation and diminished iron absorption. Nutr Rev 67: 100–104, 2009. doi:10.1111/j.1753-4887.2008.00145.x.
Crossref | PubMed | ISI | Google Scholar - 36. . Randomized, double-blind, placebo-controlled trial of iron supplementation in female soldiers during military training: effects on iron status, physical performance, and mood. Am J Clin Nutr 90: 124–131, 2009. doi:10.3945/ajcn.2009.27774.
Crossref | PubMed | ISI | Google Scholar - 37. . An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea. ISME J 6: 610–618, 2012. doi:10.1038/ismej.2011.139.
Crossref | PubMed | ISI | Google Scholar - 38. . Effects of timing, sex, and age on site-specific gastrointestinal permeability testing in children and adults. J Pediatr Gastroenterol Nutr 50: 269–275, 2010. doi:10.1097/MPG.0b013e3181aa3aa9.
Crossref | PubMed | ISI | Google Scholar - 39. . Vertically transmitted faecal IgA levels determine extra-chromosomal phenotypic variation. Nature 521: 90–93, 2015. doi:10.1038/nature14139.
Crossref | PubMed | ISI | Google Scholar - 40. . Host-gut microbiota metabolic interactions. Science 336: 1262–1267, 2012. doi:10.1126/science.1223813.
Crossref | PubMed | ISI | Google Scholar - 41. . Fat, fibre and cancer risk in African Americans and rural Africans. Nat Commun 6: 6342, 2015. doi:10.1038/ncomms7342.
Crossref | PubMed | ISI | Google Scholar - 42. . Probiotic bacteria and intestinal epithelial barrier function. Am J Physiol Gastrointest Liver Physiol 298: G807–G819, 2010. doi:10.1152/ajpgi.00243.2009.
Link | ISI | Google Scholar - 43. . Effects of exercise mode, energy, and macronutrient interventions on inflammation during military training. Physiol Rep 4: e12820, 2016. doi:10.14814/phy2.12820.
Crossref | PubMed | ISI | Google Scholar - 44. . Gastrointestinal symptoms and altered intestinal permeability induced by combat training are associated with distinct metabotypic changes. J Proteome Res 14: 4734–4742, 2015. doi:10.1021/acs.jproteome.5b00603.
Crossref | PubMed | ISI | Google Scholar - 45. . The incidence and gastrointestinal infectious risk of functional gastrointestinal disorders in a healthy US adult population. Am J Gastroenterol 106: 130–138, 2011. doi:10.1038/ajg.2010.371.
Crossref | PubMed | ISI | Google Scholar - 46. . Gastrointestinal infections in deployed forces in the Middle East theater: an historical 60 year perspective. Am J Trop Med Hyg 93: 912–917, 2015. doi:10.4269/ajtmh.15-0200.
Crossref | PubMed | ISI | Google Scholar - 47. . Past trends and current status of self-reported incidence and impact of disease and nonbattle injury in military operations in Southwest Asia and the Middle East. Am J Public Health 98: 2199–2206, 2008. doi:10.2105/AJPH.2007.131680.
Crossref | PubMed | ISI | Google Scholar - 48. . Colonic transit time is related to bacterial metabolism and mucosal turnover in the gut. Nat Microbiol 1: 16093, 2016. doi:10.1038/nmicrobiol.2016.93.
Crossref | PubMed | ISI | Google Scholar - 49. . Effects of probiotics supplementation on gastrointestinal permeability, inflammation and exercise performance in the heat. Eur J Appl Physiol 114: 93–103, 2014. doi:10.1007/s00421-013-2748-y.
Crossref | PubMed | ISI | Google Scholar - 50. . Disruption of the circadian clock in mice increases intestinal permeability and promotes alcohol-induced hepatic pathology and inflammation. PLoS One 8: e67102, 2013. doi:10.1371/journal.pone.0067102.
Crossref | PubMed | ISI | Google Scholar - 51. . Energy requirements of military personnel. Appetite 44: 47–65, 2005. doi:10.1016/j.appet.2003.11.010.
Crossref | PubMed | ISI | Google Scholar - 52. . Gut permeability, its interaction with gut microflora and effects on metabolic health are mediated by the lymphatics system, liver and bile acid. Future Microbiol 10: 1339–1353, 2015. doi:10.2217/FMB.15.54.
Crossref | PubMed | ISI | Google Scholar - 53. . Intestinal mucosal barrier function in health and disease. Nat Rev Immunol 9: 799–809, 2009. doi:10.1038/nri2653.
Crossref | PubMed | ISI | Google Scholar - 54. . Physiology and pathophysiology of splanchnic hypoperfusion and intestinal injury during exercise: strategies for evaluation and prevention. Am J Physiol Gastrointest Liver Physiol 303: G155–G168, 2012. doi:10.1152/ajpgi.00066.2012.
Link | ISI | Google Scholar - 55. . Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol 73: 5261–5267, 2007. doi:10.1128/AEM.00062-07.
Crossref | PubMed | ISI | Google Scholar - 56. . Relevance of protein fermentation to gut health. Mol Nutr Food Res 56: 184–196, 2012. doi:10.1002/mnfr.201100542.
Crossref | PubMed | ISI | Google Scholar


