Exercise Attenuates PCB-Induced Changes in the Mouse Gut Microbiome

Animals and voluntary exercise. All animal protocols were approved by the Institutional Committee on Animal Care of the University of Miami School of Medicine and followed National Institutes of Health guidelines (National Research Council 2011). The animals were treated humanely and with regard for alleviation of suffering. Male C57BL/6 mice (11–13 months of age; Charles River Laboratories, Wilmington, MA) were housed under 12:12-hr light/dark conditions with access to food and water ad libitum. Mice were randomly assigned to either the exercised or sedentary group (n = 6/group). Mice were housed individually in plastic cages equipped with a running wheel (Coulbourn Instruments, Whitehall, PA); however, wheels were locked in cages of sedentary mice. Running activity was monitored 24 hr/day, 7 days/week and analyzed using Clocklab (Actimetrics, Wilmette, IL) and Matlab (Mathworks, Natick, MA) software. Mice were allowed to exercise for 5 weeks, including 1 week to adapt to running and solitary housing.

Mice averaged 10–12 km/24 hr. The average speed was 1.12 km/hr, and the mean ± SE time spent running was 10.3 ± 0.33 hr/day. Exercised mice had 30% lower body weight on average than sedentary mice. At the end of the exercise period, we observed no changes in gut transit or stool frequency between the two groups.

PCB exposure. We selected the type of PCB congeners and their relative ratio based on PCB content in contaminated food (Ashley et al. 2000). The three environmentally relevant PCB congeners, PCB138, PCB153, and PCB180 (AccuStandard, New Haven, CT), were mixed at a molar ratio of 1.7:3.2:1 in tocopherol-stripped safflower oil (vehicle; Dyets Inc., Bethlehem, PA). The total PCB dose of 150 μmol/kg administered to mice results in a PCB plasma level of 5 μM (Choi YJ et al. 2010), which is comparable to plasma levels of PCBs in an acutely exposed human population (Jensen 1989; Wassermann et al. 1979). All treatments were performed via oral gavage using an 18-G gavage needle (3 cm length, curved, 2.25 mm ball diameter; Popper and Sons, New Hyde Park, NY).

A diagram of the study design is shown in Figure 1. At the end of the exercise period, mice were administered tocopherol-stripped safflower oil (5 mL/kg) to allow them to adjust to the vehicle. Feces were collected for 2 days (allowing collection of a sufficient amount of sample), and samples were analyzed for control/background microbiome data. Mice were then treated with the PCB mixture, and feces were again collected for 2 days. To minimize the influence of individual variations on the gut microbiome, each mouse served as its own control for PCB treatment.

Analysis of the gut microbiome. All analyses were performed using PhyloChip Arrays (Second Genome Inc., San Bruno, CA). The GenBank accession numbers were used as identification and classification of bacterial operational taxonomic units (OTUs) based on their 16S rRNA sequences. Bacterial 16S rRNAs were amplified with the primer sets 27F.1 (5´-AGRG​TTTG​ATCM​TGGC​TCAG-3´) and 1492R.jgi (5´-GGTT​ACCT​TGTT​ACGA​CTT-3´). Gene amplification was accomplished by polymerase chain reaction (PCR; 35 cycles) as follows: denature step at 95°C for 30 sec, annealing step at 50°C for 30 sec, and extension step at 72°C for 2 min using the TaKaRa Ex Taq system (Takara Bio Inc., Japan). For each sample, amplified products were concentrated using a centrifuge filtration method and quantified by electrophoresis using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). PhyloChip Control Mix (Second Genome Inc.) was added to each amplified product. The combined PCR products and controls were fragmented, biotin labeled, and hybridized to the PhyloChip Array (version G3). PhyloChip arrays were washed, stained, and scanned using a GeneArray scanner (Affymetrix, Cleveland, OH). Each scan was captured using the GeneChip Microarray Analysis Suite (Affymetrix). Hybridization scores and the fluorescence intensity for each taxon were calculated as a trimmed average, with maximum and minimum values removed before averaging.

Statistical analysis. Data processing and multivariate statistical analyses were performed using the PhyCA-Stats analysis software package (Second Genome Inc.). Probe intensities were background subtracted and scaled to values for PhyloChip Control Mix. Subsequently, hybridization scores were calculated as log₂(mean probe fluorescence intensity) × 1,000. OTUs were defined by high 16S rRNA gene sequence similarity, with most demonstrating > 99% intra-OTU concordance. We determined presence/absence of an OTU by a threshold based on perfect match and mismatch intensities of multiple probes per probe set (Hazen et al 2010). Before classification analysis, we performed data reduction via multiple filtering steps as described by Hazen et al. (2010). Taxa-sample intersections were then calculated using abundance (AT) and binary matrices (BT). Pairwise BT and AT dissimilarity scores were computed using the Unifrac distance metric (Lozupone et al. 2006) and weighted Unifrac distance metric, respectively; the weighted Unifrac metric considers OTU abundance in addition to phylogenetic distance between OTUs. We used hierarchical clustering via average-neighbor (HC-AN) and principal coordinate analysis (PCoA) to graphically summarize intersample relationships according to AT and BT dissimilarity scores. Unsupervised classification using the nearest shrunken centroid method (Tibshirani et al. 2002) as implemented in Prediction Analysis of Microarrays (PAM; http://www-stat.stanford.edu/~tibs/PAM/) was used to identify OTUs with the most significant differences in abundance between comparison groups. We used a randomization/Monte Carlo permutation–based test (Adonis test) to determine significant differences between discrete and continuous variables. Finally, we employed Welch’s test and Student’s t-test for statistical significance, with p < 0.05 considered significant.

PCB exposure decreases the abundance of the gut microbiota. We first analyzed the effects of PCB exposure on the gut microbiome in sedentary mice. The Welch’s test revealed that exposure to the PCB mixture significantly (p < 0.05 by Student’s t-test) altered the abundance of 1,223 bacterial taxa in these mice (1,133 taxa had decreased abundance and 90 taxa had increased abundance). As a result of these changes, the overall abundance of bacteria significantly diminished (by 2.2%) in PCB-exposed mice. Table 1 lists bacterial taxa with the greatest decrease in abundance (4.0- to 5.6-fold) after PCB treatment. These taxa belong primarily to phylum Proteobacteria, but the classes and families are diverse. The group of bacterial taxa with increased abundance after PCB treatment was relatively modest, and the changes did not exceed 2-fold; the bacterial taxa of these with the highest increases in abundance are listed in Table 2. They belong to several difference phyla, including Bacteriodetes, Actinobacteria, Verrucomicrobia, and Firmicutes.

Exposure to the PCB mixture did not alter biodiversity of the gut microbiome; all 11,229 taxa were detected in at least one sample from sedentary mice. After PCB exposure, the number of taxa dropped to 10,798, but the change was not statistically significant.

Exercise alters the composition of the gut microbiome. As analyzed by weighted Unifrac distance, the structure of the gut microbiome of the exercised mice was significantly different that of the sedentary mice (Adonis test, p < 0.05). PCoA of unweighted Unifrac distance with given presence/absence metrics also showed prominent categorization of the composition of the gut microbiome between the exercised and sedentary mice (Figure 2A).

Among detected bacterial taxa, 93 were present exclusively in either exercised or sedentary mice. Specifically, 67 taxa were detected only in the exercised mice [see Supplemental Material, Table S1 (https://doi.org/10.1289/ehp.1306534)], and 26 taxa were unique to sedentary mice (see Supplemental Material, Table S2).

A group of 2,510 taxa showed differences in abundance between the exercised and sedentary mice. These taxa were then analyzed by PCoA with weighted Unifrac distance, which indicated significant differences in the composition of the microbial communities between the exercised and sedentary mice (Figure 2B). HC-AN analysis based on weighted Unifrac distance confirmed a shift of the composition of the gut microbiome related to physical activity (exercised vs. sedentary mice) (Figure 2C).

Further examination using the PAM identified 10 taxa with substantially different abundance between the exercised and sedentary mice (Table 3). The taxa that were more abundant in the exercised group were in phylum Firmicutes, class Bacilli, and most of these were in order Lactobacillales. The taxa that were decreased in the exercised group belonged to phyla Tenericutes, Bacteroidetes, and Firmicutes. The most striking change in exercised mice was a decrease in Erysipelotrichaceae bacterium C11_K211 from phylum Tenericutes, which decreased dramatically in exercised compared with sedentary mice.

Exercise attenuates PCB-induced alterations of gut microbiome composition. Comparison of the gut microbiome among all experimental groups (sedentary and exercised mice with or without PCB exposure) identified 1,568 bacterial taxa that were differentially abundant in at least one of these groups. Analysis of these taxa for dissimilarity between the groups using PCoA with weighted Unifrac indicated significant differences between the composition of the gut microbiome before and after PCB exposure in sedentary mice. Importantly, exercise altered PCB-mediated effects on the gut microbiome, as indicated by a loss of bacterial clustering (Figure 3A). This phenomenon was subsequently confirmed by HC-AN analysis (Figure 3B).

We detected 10,799 taxa in sedentary mice exposed to the PCB mixture, compared with 13,383 taxa in the exercised mice treated with PCBs. Although there appeared to be a tendency toward increased biodiversity in the exercise-plus-PCB group, these changes were not significant. In contrast, the abundance of bacterial species was elevated by 2.9% (Student’s t-test, p < 0.05) in PCB-treated exercised mice as compared PCB-treated sedentary mice, providing additional evidence that exercise can protect against PCB-mediated alterations in the gut microbiota. As shown in Figure 3C, exercise appeared to prevent a PCB-induced decrease in abundance of Proteobacteria in sedentary mice (Figure 3C).

In view of recent evidence indicating that gut bacteria can be involved in the preabsorptive metabolism of heavy metals and organic chemicals (Lapanje et al. 2007; Liebert et al. 1997; Pinyayev et al. 2011), the gut microbiome has been proposed to play a role in the assessment of health risks associated with environmental chemicals (Betts 2011; Snedeker and Hay 2012). Therefore, our demonstration that short-term exposure to an environmentally relevant PCB mixture resulted in profound changes in the gut microbiome in mice is highly significant. The most striking change in the intestinal microbial profiles was a decrease in the overall abundance of bacterial species. Although these results are the first to show effects of PCBs on the gut microbiome, the decrease in bacterial abundance we observed corresponds with the finding that PCB-contaminated soil is characterized by a shift in structure and abundance of bacterial community (Petric et al. 2011). Incubation of soil slurries with higher-chlorinated PCB congeners (e.g., PCB28, PCB77, Aroclor 1242) has been reported to result in lower bacterial numbers (Correa et al. 2010). Even though the gut and soil provide completely different bacterial environments, PCB exposure appears to elicit environmental stress on the structure and composition of bacterial communities that results in diminished bacterial abundance. In the present study, these changes selectively affected bacterial phylotypes in phylum Proteobacteria. Such changes may be most relevant to immune functions of the host, because the gut microbiome has been shown to play important roles in mucosal immunity and interactions with intestinal and colonic epithelial cells, dendritic cells, and T and B immune cells. Microbiota composition has functional effects on T-effector- and T-regulatory-cell balance, immune responsiveness, and homeostasis (Kelly and Mulder 2012). Thus, it is likely that alterations of the gut microbiota compromise a novel mechanism leading to immunological alterations, which develop in response to exposure to PCBs (Maule et al. 2005; Weisglas-Kuperus et al. 2004). In addition, alterations of the gut microbiome can affect PCB-induced disruption of the intestinal barrier and translocation of lipopolysaccharides into the blood stream (Choi JJ et al. 2012; Choi YJ et al. 2010). In the present study, we observed no statistical differences in bacterial community structure in the exercised mice before and after PCB treatment. Thus, exercise provided protection against PCB-induced changes in the gut microbiome. In particular, exercise prevented a PCB-induced decrease in abundance of Proteobacteria, which was observed in sedentary mice.

Although the data we reported here are novel, effects of physical activity on several other aspects of gut physiology (e.g., peristalsis; Song et al. 2012) and pathology have been reported. For example, exercise was demonstrated to decrease the risk of developing several intestinal diseases, including colon cancer (Friedenreich et al. 2006), inflammatory bowel disease and irritable bowel syndrome (Lustyk et al. 2001), and other disorders that are accompanied by changes in the gut microbiome (De Hertogh et al. 2012; Nelson et al. 2011; Walker et al. 2011).

The mechanisms of exercise-mediated changes in gut ecology are not known; however, they are likely to be mediated by altering the host factors that influence the intestinal microenvironment. For example, physical activity has been reported to increase excretion of primary bile acids to the gastrointestinal tract (Meissner et al. 2011) and to suppress the formation of secondary bile acids (Hagio et al. 2010). The primary bile acids, such as cholic, deoxycholic, or chenodeoxycholic acids, have established antimicrobial activity, which is mediated by the reduction in internal pH levels of bacteria, dissipation of their transmembrane electrical potential, and disturbances of membrane integrity, leading to leakage of ions and cell death (Kurdi et al. 2006). In support of this hypothesis, Islam et al. (2011) demonstrated that cholic acid induced substantial changes in the cecal microbiome composition by stimulating the growth of Firmicutes at the expense of Bacteroidetes and outgrowth of several bacteria in the classes Clostridia and Erysipelotrichi. Thus, the antimicrobial activity of the bile acids may elicit selective pressure on the bacterial communities in exercised mice, leading to a shift of the gut microbiome structure as observed in the present study.

Short-chain fatty acids (SCFAs) may be another factor that regulates the gut microbiome in response to exercise. Indeed, Matsumoto et al. (2008) showed that rats that participated in voluntary running exercise had increased butyrate concentration in the cecum compared with sedentary rats. These authors directly linked this effect to the beneficial effect of exercise on the gut microbiota and the development of gastrointestinal disorders. SCFAs (e.g. butyrate and acetate) increase colonic epithelial cell proliferation and decrease the risk of colorectal cancer. Their influence on the composition of microbial environment has been linked to a decreased pH in the gut (Wong et al. 2006). Nevertheless, compared with the results of the present study, the effects of butyrate infusion on the rumen microbiome in cows was relatively minor because only 19 genera and 43 bacterial taxa were significantly affected in response to butyrate (Li et al. 2012); these data suggest that this SCFA may be only one of several factors involved in exercise-mediated changes in the gut microbiome. Treatment with SCFAs may also affect host-related intestinal factors because butyrate promotes cell differentiation and cell-cycle arrest, inhibits the enzyme histone deacetylase, and decreases the transformation of primary to secondary bile acids as a result of colonic acidification (Wong et al. 2006).

Finally, exercise may influence the composition of the gut microbiome by altering the intestinal immune system. Viloria et al. (2011) observed that physical activity increased expression of IgA and cytokines such as interleukin-6 and tumor necrosis factor-α. These changes in the intestinal immune system may lead to secondary alternations of the host–bacterial interaction and induce selective pressure on bacterial selection.

The development of chronic diseases related to the exposure to environmental toxicants are associated with age, and the benefits of exercise are also being emphasized in older individuals; therefore, we used aged mice (11–13 months of age) in the present study. Older mice tend to have higher body mass than younger animals. In fact, the average body weight of sedentary mice in the present study was 46.8 ± 1.4 g, and body weight of exercised mice was approximately 30% lower. Recent evidence from Ley et al. (2005) indicated a strong association of the intestinal microbiome with the development of obesity. Genetically obese ob/ob mice were characterized by a major decrease in the abundance of Bacteroidetes and an increase in Firmicutes compared with lean ob/+ wild-type littermates and lean ob/+ mothers fed the same diets. Similar changes were observed in wild-type mice fed a high fat/high polysaccharide diet (Turnbaugh et al. 2008) and in obese humans (Ley et al. 2005). An increase in Firmicutes (such as Lactobacilli) and a decrease in Bacteroidetes have been confirmed in obese humans (Armougom et al. 2009). In another study, overweight pregnant patients had reduced abundance of Bifidobacteria and Bacteroidetes and increased abundance of selected Firmicutes (e.g., Staphylococcus) and Proteobacteria (e.g., Enterobacteriaceae) (Santacruz et al. 2010). In line with these reports, it is relevant that we detected an increased abundance of several Firmicutes, primarily Enterococcaceae (e.g., Enterococcus faecium), in the exercised mice.

Although Enterococci are commensal bacteria, they are also important nosocomial pathogens that cause bacteremia, endocarditis, and other infections in humans. Some strains are resistant to multiple antibiotics and possess virulence factors, such as adhesins, invasins, pili, and haemolysin (Willems et al. 2011). Nevertheless, E. faecium isolates from clinical outbreaks are different strains than E. faecium from animals, food, and humans in the community (Franz et al. 2011). In fact, several enterococci, including E. faecium strains, are used as probiotics in the form of pharmaceutical preparations. They are administered to treat diarrhea, antibiotic-associated diarrhea, and irritable bowel syndrome; to lower cholesterol levels; or to improve host immunity (Franz et al. 2011).

The most striking change we found in the gut microbiome of exercised mice was a > 300-fold decrease in the abundance of Erysipelotrichaceae (Table 3). This family plays an important role in metabolic disorders and energy metabolism (Chen et al. 2012). Chen et al. (2012) reported that Erysipelotrichaceae were enriched in obese humans and mice, as well as in mice fed a high-fat diet, and noted that the abundance of Erysipelotrichaceae was also increased in patients with colorectal cancer. Thus, our data are consistent because we observed a decrease in the number of Erysipelotrichaceae in exercised mice that had lost a substantial amount of body weight; this is associated with the role of Erysipelotrichaceae in energy production and adiposity (Claus et al. 2011; Goodman et al. 2011; Zhang et al. 2009).

Here we demonstrate that oral exposure to a mixture of environmentally relevant PCB congeners significantly altered the abundance of the gut microbiome by decreasing the levels of Proteobacteria. These results suggest that the gut microbiome may be one of the primary targets of PCB-induced toxicity in subjects exposed orally to these environmental toxicants. We observed that PCB-induced alterations of the gut microbiome were attenuated by voluntary exercise.