Exercise induction of gut microbiota modifications in obese, non-obese and hypertensive rats

Animals

Animals were obtained from the Federal University of São Paulo, Brazil (UNIFESP) and started the experiment at ~18 weeks of age. Three different strains from two different genotypes were used: an obese genotype, homozygous (fa/fa) obese Zucker rats [25] (Obese rats; n = 5, 389.4 ± 21 g) and a hypertensive genotype, spontaneous hypertensive rats (Hypertensive rats; n = 5, 227.4 ± 29.3 g), a strain obtained by the selective breeding of Wistar-Kyoto rats with high blood pressure [26]. A strain of Wistar rats (WR; n = 5, 223.2 ± 27.3 g) was used as normotensive control for SHR and as a non-obese phenotype [26] (Additional file 1). All animals were allocated to collective cages according to their lineages, being kept in a 12 h light–dark cycle environment, with food and water ad libitum. All experimental procedures and interventions in the present study, involving animal-welfare, were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the local ethics committee for standards in animal use, at the Institute of Biological Sciences, University of Brasilia, Brazil (UnBDOC no. 48695/2010); these were also in accordance with international standards. After experimental procedures, all animals were deeply anesthetized with 2% Xylazine (50 mg.kg⁻¹) and 10% Ketamine (80 mg.kg⁻¹) and euthanized by cervical dislocation. Through the entire experiment, all efforts were made to minimize animal suffering.

Exercise training

Before the training period, all animals underwent a familiarization period on a treadmill device (Li 870, Letica Scientific Instruments, Barcelona, Spain) for 2 weeks to reduce external stress. Duration and speed on treadmill were increased progressively (up to 12.5 m.min⁻¹ for Obese rats; 20 m.min⁻¹ for Hypertensive and Wistar rats) as previously described [27, 28]. In addition, blood pressure of Wistar and Hypertensive rats was measured by the tail-cuff method [29] at the beginning of the experiment to characterize the hypertensive phenotype from the SHR strain (171.4 ± 7.7 mm.Hg⁻¹ for Hypertensive rats and 128 ± 5.9 mmHg⁻¹ for Wistar rats) (Additional file 2). As regards the Obese rats group (obese (fa/fa) Zucker rat), these animals are homozygous for the fa allele and are one of the most common models of genetic obesity where rats normally exhibit hyperlipidemia, hyperinsulinemia, hyperphagia and a significant weight gain by the 3rd to the 5th week of age [25]. In this model, conflicting results were reported on whether these obese Zucker rats are hypertensive compared to the lean control rats, showing that blood pressure is not elevated in this model [30, 31]. However, the obesity phenotype may enhance arterial peripheral resistance and the animal may develop hypertension secondary to obesity and to secondary mechanisms [32, 33]. Considering that obese Zucker rats do not present a regular pattern of blood pressure as reported by different studies, the systolic blood pressure was not measured in this group.

After the adaptation period, all animals were trained for 30 min per day, 5 days per week for 4 weeks. Running intensity was set corresponding to maximal lactate steady state (MLSS), previously identified in obese Zucker rats [28] and SHR rats [27]. Therefore, for Obese rats, running velocity was set at 12.5 m.min⁻¹ and for Hypertensive and Wistar rats at 20 m.min⁻¹. A new MLSS identification was performed after the fourth week of exercise training to assess each animal’s cardiovascular adaptation.

Blood lactate analysis

Capillary blood samples (10 μL) were collected through a small incision in the distal portion of the tail of the animals at rest and every 5 min during the MLSS test. Capillary blood samples were placed in microtubes (0.6 mL) containing 20 μL of 1% sodium fluoride and stored at −20°C. Analyses were performed by electro-enzymatic method with YSI 1500 Sports (Yellow Springs, OH, USA). The MLSS was considered when there was no increase over 1 mmol.L⁻¹ of blood lactate from 10 to 25 min of exercise tests [34].

Fecal DNA extraction and barcoded pyrosequencing of the 16S rRNA gene

Fecal content from all animals was individually collected in three replicates before exercise training period and in vivo 24 h after the last session of exercise (Additional file 1). Fecal samples were stored in RNA later® (Life technologies) until samples were frozen at −80°C. Three out of five individual fecal samples from each animal lineage were randomly selected. Fecal microbial DNA was extracted from ~0.25 g using the PowerFecal DNA Isolation Kit (MoBio, Carlsbad, CA, USA) according to the manufacturer’s instructions. The triplicate DNA extractions were not pooled.

The bacterial community partial 16S rRNA gene was amplified with primer pair 787 F-1492R [35]. For pyrosequencing analysis, the primer set was modified, where the forward primer included the Roche 454-A pyrosequencing adapter and a 12-bp barcode (unique to each sample), while the reverse primer included only the Roche 454-B pyrosequencing adapter. The 20 μl reaction mixture for the PCR contained approximately 10 ng of metagenomic fecal DNA, 1X PCR buffer (Invitrogen), 3.0 mM MgCl2, 10 pmol of each primer, 0.25 mM dNTP, and 1.5 U Taq DNA polymerase (Invitrogen). The cycling protocol started with an initial denaturation step of 3 min at 95°C, followed by 25 cycles of denaturation for 30 s at 95°C, annealing for 30 s at 58°C, and extension for 1.40 min at 72°C, followed by a final extension for 7 min at 72°C and cooling to 10°C. Finally, the rRNA amplicons from bacterial communities were purified with the QIAquick PCR Purification Kit (Chatsworth, CA). The concentrations of the rRNA amplicons were measured by Qubit fluorometer (Invitrogen), and subsequently the massively parallel GS FLX Titanium technology was performed in the Roche 454 Life Sciences Corporation, Branford, CT, USA.

Analysis of 16S rRNA sequences

A total of 1,398,681 16S rRNA sequences were obtained by the 454 GS FLX Titanium sequencer. All the 16S amplicons were processed by the quantitative insights into microbial ecology (QIIME) pipeline version 1.6.0-dev [36]. Briefly, all the 16S rRNA amplicons were sorted by their barcodes, and subsequently the reads with a length less than 180 bp, ambiguous sequences, bases with Phred values of <30 and their primers, barcodes and adaptor sequences were removed. The remaining sequences were submitted to Denoiser algorithm [37] to remove pyrosequencing errors. Operational taxonomic units (OTUs) were clustered at 97% similarity using an ‘open-reference’ OTU picking protocol, where sequences are clustered against the Greengenes database [38] using Uclust [39]. One of the most-abundant reads from each OTU was aligned using the PyNAST algorithm [36]. The chimeric OTUs were detected with ChimeraSlayer [40], and taxonomic classifications were assigned with the naïve Bayesian classifier of the Ribosomal Database Project (RDP) classifier [41] applying 80% of confidence threshold. Shannon indices and observed richness were used to evaluate community richness, and the unweighted Unifrac algorithm was performed to generate principal coordinate plots (PCoA).

Statistical analysis

Statistical differences between the groups pre-exercise and after exercise were tested using the analysis of similarities (ANOSIM) by permutation of group membership with 999 replicates [42], and bivariate relationships were measured with Pearson correlations and regression analysis available through QIIME [36]. Statistical tests on the taxonomic differences between samples were calculated by Welch’s t-test combined with Welch’s inverted method for calculating confidence intervals (nominal coverage of 95%), using the Statistical Analysis of Metagenomic Profiles (STAMP) software version 2.0.0 (STAMP) [43].

Accession number

The 454 FLX Titanium flowgrams (sff files) have been submitted to the National Center for Biotechnology Information (NCBI) Sequence Read Archive database, project number: PRJNA246617.

Effect of exercise training on aerobic capacity

After four weeks of treadmill running exercise at moderate intensity, a novel MLSS assessment evidenced that all animals from each group had enhanced their aerobic capacity as demonstrated by improvement in the MLSS corresponding velocity (15 m.min⁻¹ for Obese rats and 30 m.min⁻¹ for Hypertensive and Wistar rats) (Figure 1A). Furthermore, when comparing the initial and final velocity of exercise, a significant reduction was evidenced in blood lactate concentration (BLC) of ~49% for Wistar rats ~39% for Hypertensive rats and ~33% for Obese rats. This reduction in BLC (before vs. after exercise training) demonstrates the effectiveness of the proposed exercise training intensity for all animals (p < 0.01) (Figure 1B).

Composition of fecal microbiota in rat lineages

After quality filtering, 889,124 out of 1,398,681 sequences were obtained from fecal samples collected pre-exercise and post-exercise, after 4 weeks of moderate exercise training (Additional file 1). An average of 49,951 denoised sequences per animal were obtained (average read length of 524.8) which composed an average of 583.1 distinct observed OTUs. Post-training samples presented a higher Shannon index compared to pre-training samples (6.4 ± 0.5 vs. 6.8 ± 0.2) (Additional file 3). Detailed sequencing information from all individual rats is presented in (Additional file 3).

Bacterial diversity was assessed by rarefaction measure of observed species against the number of sequences per sample, and the observed OTUs were identified with 97% of identity. Here, the rarefaction measure showed that species diversity (Additional file 4: A; Wistar rats; B; Hypertensive rats and C; Obese rats) reached a plateau tendency in all samples as the number of sequences increased, indicating that in the present study more sequences are unlikely to yield many additional species. As demonstrated in Additional file 4, the rarefaction curves revealed that OTU richness in post-exercise fecal samples is more species-rich than those found in the pre-exercise fecal samples. This was further evidenced in hypertensive rats and obese rats samples (Additional file 4B and C).

The relative abundance of the main dominant bacterial phylum from all fecal samples collected before and after exercise training is shown in Additional file 5A. Here, Firmicutes and Bacteroidetes are the most dominant phyla, followed by Proteobacteria. Firmicutes was shown to be enhanced after exercise training (1.1 fold change, p < 0.05) (Additional file 5B), thus being more evidenced in Obese rats (Obese rats; 0.69 ± 0.03 vs. Exercised Obese rats; 0.78 ± 0.04, p < 0.05). On the other hand, Proteobacteria were shown to be 1.8 fold reduced after exercise training (p < 0.05) (Additional file 5C). The Bacteroidetes phylum was shown to be 1.3 fold reduced after exercise only in Wistar rats (Wistar rats; 0.23 ± 0.04 vs. Exercised Wistar rats; 0.17 ± 0.03, p < 0.05).

Composition of bacterial communities before, during and after exercise training

The relative abundance at bacterial genus level for all animal lineages in response to exercise training was assigned only to those that presented a minimum variation at significant level (p < 0.05) (Figure 2). In Wistar rats (A), Streptococcus was the only genus that presented a significant alteration within its abundance, while untrained rats were more enriched with Streptococcus when compared to post-exercise (p < 0.05) (Figure 2A). In hypertensive rats, three genera (Allobaculum, Aggregatiobacter and Sutterella) were shown to be altered by exercise training. Despite minimal variation in the relative abundance of Allobaculum between pre-exercise and post-exercise samples, this genus was enriched by exercise training (p < 0.05) (Figure 2B). This was in contrast to Allobaculum, Aggregatibacter and Suturella where both were more abundant in pre-exercise samples (Figure 2B). Aggregatibacter presented a minimal variation in their relative abundance in fecal samples pre- and post-exercise training; however, exercise was shown to reduce the abundance of this genus (p < 0.05) (Figure 2B). The Suturella genus was also shown to be more enriched pre-exercise, with a greater relative proportion of all in this genus in hypertensive rats (p < 0.05). In Obese rats, Pseudomonas and Lactobacillus were both significantly altered after exercise training (Figure 2C). Minimal variation in Pseudomonas relative abundance was observed between samples (P < 0.05), while Lactobacillus presented the higher relative abundance after exercise from all identified genera (p < 0.05) (Figure 2C).

The proportion of sequences (%) of the main bacterial species from fecal samples collected before and after exercise training is shown in box plots (Figure 3). In pre-exercise fecal samples (Figure 3A and B), only two species (Bacteroides acidifaciens and Ruminococcus flavefaciens) presented a significant differential abundance, in contrast to fecal samples post-exercise training, where six species (Streptococcus alactolyticus, Bifidobacterium animalis, Ruminococcus gnavus, Aggregatibacter pnemotropica and Bifidobacterium pseudolongum) presented a differential abundance (Figure 3C-G). From all samples (pre and post-exercise) only one species (Ruminococcus flavefaciens) was less abundant in obese animals (Figure 3B), with all other species being significantly more enriched in the obese animals (Figure 3C-G).

Regarding pre-exercise samples, the proportion of sequences from the Bacteroides acidifaciens species was significantly more abundant in Obese rats than in Wistar and Hypertensive rats (p < 0.05) (Figure 3A). However, an opposite profile is observed in sequences attributed to the Ruminococcus flavefaciens species, where greater abundance is seen in Wistar rats followed by Hypertensive rats, with no abundance seen in Obese rats (p < 0.05) (Figure 3B). After exercise training, the proportion of sequences indicated that all listed species (Streptococcus alactolyticus, Bifidobacterium animalis, Ruminococcus gnavus, Aggregatibacter pneumotropica and Bifidobacterium pseudolongum) were more abundant in Obese rats compared to Wistar and Hypertensive rats lineages (Figure 3C-G respectively). The relative abundance of Streptococcus alactolyticus in Obese rats diverged significantly from Hypertensive and Wistar rats (p < 0.05), while a diminished proportion of sequences was seen in both rat strains (Figure 3C). The Bifidobacterium animalis species was seen to be highly enriched in Obese rats (p < 0.05) and absent in Wistar and Hypertensive rats (Figure 3D). In relation to Ruminococcus gnavus, this species was poor in Wistar and almost absent in Hypertensive rats, but more abundant in Obese rats (Figure 3E). The Aggregatibacter pneumotropica species presented a similar profile to the previous species, being also more abundant in Obese rats compared to Wistar and Hypertensive rats (p < 0.05) (Figure 3F). Lastly, from the Actinobacteria phylum, Bifidobacterium pseudolongum abundance was almost exclusive to Obese rats (p < 0.05), being almost absent in Hypertensive rats and completely absent in Wistar rats group (Figure 3G).

Principal coordinates analysis (PCoA)

Principal coordinates analysis (PCoA) of unweighed UniFrac distances was calculated and compared between all fecal samples collected pre and post-exercise from the three rat lineages in order to observe similarity in microbiota composition and the effect of exercise training (Figure 4). All three biological replicates from each animal lineage (Wistar, Hypertensive and Obese rats) were shown to cluster with a high correlation between them (R = 0.79, P < 0.001). UniFrac (PCoA) analysis showed that Wistar and Hypertensive rats share a similar bacterial composition, thus clustering far from Obese rats, indicating a distinct bacterial community composition between these rat lineages (Figure 4). It also indicated that microbiota from Wistar, Hypertensive and Obese rats were significantly altered by exercise training, where pre-exercise samples clustered significantly far from fecal samples collected after four weeks of exercise training (Figure 4). However, even though exercise altered microbial community composition in every animal lineage, Wistar and Hypertensive rats still maintained a similar bacterial composition, still clustering far from Obese rats microbiota (Figure 4).

Correlation of bacterial abundance and lactate concentration

As shown in Figure 1A, when comparing the initial and final velocity of exercise training, a significant reduction in BLC was evidenced in all rat lineages, where lower BLC (means of 2.3 mmol.L⁻¹, Figure 1B) is associated to an improved aerobic capacity from trained status when compared to higher BLC (3.8 mmol.L⁻¹) from untrained rats (pre-exercise samples from all rat lineages). Therefore, fecal bacterial communities were plotted against BLC to establish a correlation between microbial abundance and training status (Figure 5). OTUs from two bacterial families (Clostridiaceae and Bacteroidaeae) and two genera (Oscillospira and Ruminococcus) were found to be significantly correlated with BLC. The OTU abundance from both bacterial families was negatively correlated to BLC (Clostridiaceae, R = −0.82 P < 0.01; Bacteroidaceae, R = −0.73 P < 0.01). In both cases, higher abundance in OTUs was observed to correlate with lower lactate concentrations, indicating that exercise training may be favorable to the proliferation of these OTUs from both bacterial families (Figure 5A and B). The relative abundance of OTUs from Bacteroidaceae family was shown to be close to zero when BLC reached ~4 mmol.L⁻¹ (Figure 5B), being associated with untrained status. Regarding the genera, the abundance of OTUs from Oscillospira and Ruminococcus genera presented divergent correlations with BLC. OTUs from Oscillospira were shown to be positively correlated to BLC (R = 0.78 P < 0.01) and Ruminococcus was negatively correlated (R = −0.75 P < 0.01) (Figure 5C and D). The OTUs from Oscillospira genus were shown to be almost absent in lower lactate concentrations, increasing their abundance with higher concentrations over 3.5 mmol.L⁻¹, which indicates that exercise training may be seen as an unfavorable factor for a specific OTU from Oscillospira genus and its proliferation in gut environment (Figure 5C). However, the OTUs from Ruminococcus microbial genus were shown to be more abundant at lower lactate concentration and with almost no abundance at higher BLC, indicating that exercise training may also influence proliferation in this genus.