Influence of whole-wheat consumption on fecal microbial community structure of obese diabetic mice

Study design

The Institutional Animal Care Use Committee from Washington State University approved all experimental procedures (animal protocol approval number: 04436-001). Two strains of male mice were used in this study, BKS.Cg- + Lepr^db/+Lepr^db/OlaHsd obese diabetic (db/db), and lean BKS.Cg-Dock7m +/+ Lepr^db/OlaHsd (Harlan Laboratories, Kent, WA). Animals were purchased at 5–6 weeks of age and maintained in ventilated rack system with food and water provided ad libitum throughout the study. We received 11 mice for the lean group and 10 mice from all other groups. After 7 days of acclimatization, obese mice were randomly divided into two groups (n = 10 each) namely obese (AIN-93 G Purified Rodent Diet) and WW (whole-wheat supplemented diet). The wild type mice group (n = 11) was named lean (AIN-93 Diet). Diets were made by Dyets Inc. (Bethlehem, PA) (Table 1). Four or five mice per cage were housed in an environment-controlled room (23 °C, 12 hours dark-light cycle). All mice were visually inspected every day and body weight was recorded from all animals once a week.

Fecal collection and DNA extraction

Fresh distal colon contents (see qPCR analysis below) and fecal samples were obtained from all mice at the end of the study (8 weeks) and stored at −80 °C prior to DNA and 16S rRNA gene profiling analysis. Total DNA was extracted from at least two different fecal pellets weighting approximately 200 mg. Following bead-beating, the QIAamp DNA Stool Mini Kit (Qiagen Inc., Valencia, CA, USA) was used for DNA extraction following the manufacturer’s instructions. DNA concentration and purity was determined using a NanoDrop Spectrophotometer (Thermo Scientific, Wilmington, DE, USA) and diluted to a working concentration of 5 ng/μL.

High-throughput sequencing of 16S rRNA genes

Amplification and sequencing were performed as described elsewhere (Bokulich et al., 2014). Briefly, the V4 semi-conserved region of bacterial 16S rRNA genes was amplified using primers F515 (5′-GTGCCAGCMGCCGCGGTAA-3′) and R806 (5′-GGACTACHVGGGTWTCTAAT-3′), with the forward primer modified to contain a unique 8-nt barcode and a 2-nt linker sequence at the 5′ terminus. Amplicons were combined into one pooled sample and submitted to the University of California Davis Genome Center DNA Technologies Core for Illumina paired-end library preparation, cluster generation, and 250-bp paired-end sequencing on an Illumina MiSeq instrument in one runs. For data analysis, raw Illumina fastq files were demultiplexed, quality filtered, and analyzed using the freely available Quantitative Insights into Microbial Ecology (QIIME) Virtual Box v.1.8.0 (Caporaso et al., 2010). Operational Taxonomic Units (OTUs) were assigned using two different approaches: first, using UCLUST v.1.2.22 (Edgar, 2010) as implemented in QIIME using the open-reference clustering algorithm described in (Rideout et al., 2014) for alpha and beta diversity analyses; and second, using the pick_closed_reference_otus.py QIIME script for further analysis using PICRUSt (see Predicted metabolic profiles below). The Greengenes 13_5 97% OTU representative 16S rRNA gene sequences was used as the reference sequence collection (DeSantis et al., 2006). Alfa and beta diversity analyses were performed using 3000 random sequences per sample (lowest number of sequences in a sample after demultiplexing, filtering and OTU picking). Raw sequences were uploaded into the Sequence Read Archive at NCBI (accession number: PRJNA281761). The trim.seqs command in MOTHUR (Schloss et al., 2009) was used for splitting original fastq files per sample for uploading to SRA.

Predicted metabolic profiles

OTUs from the closed_reference script were normalized and used to predict metagenome functional content using the online Galaxy version of Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (Langille et al., 2013). PICRUSt uses existing annotations of gene content as well as 16S copy numbers from reference microbial genomes in the IMG database (Markowitz et al., 2012) and a functional classification scheme to catalogue the predicted metagenome content. The current galaxy version supports three types of functional predictions; this current study used the popular KEGG Orthologs (Kanehisa et al., 2012).

Quantitative real-time PCR (qPCR) analysis

DNA was extracted from distal colon content using the ZR Fecal DNA MiniPrep™ kit following the manufacturer’s protocol (Zymo Research, Irvine, CA, USA). qPCR was used to detect specific bacterial groups as described elsewhere (Noratto et al., 2014). Table 2 shows the primers sequences used for all qPCR analyses.

Measurement of Short Chain Fatty Acids (SCFAs) in caecal contents

SCFAs were quantified as reported elsewhere (Campos et al., 2012). Briefly, samples were analyzed by an HPLC-PDA system using an Aminex HPX-87H strong cation-exchange resin column (300 × 7.8 mm) and fitted with an ion exchange microguard refill cartridge (Bio-Rad, Hercules, CA, USA). The HPLC-PDA system consisted of a Water 2695 Separation Module (Waters, Milford, MA), which was equipped with a Water 2996 Photodiode Array detector (PDA). Samples (20 μL) were eluted isocratically with 5 mM sulfuric acid at 0.6 mL/min, and the column temperature was held at 50 °C. Sodium butyrate, acetic acid, oxalic acid, and succinic acid were identified and quantified by comparing retention time and UV-Visible spectral data to standards.

Statistical analysis

ANOVA and the non-parametric alternative Kruskal-Wallis test were used to analyze final body weights and SCFAs concentrations, respectively. Multiple-comparisons were performed using Tukey and Mann-Whitney tests. The Bonferroni (for Mann-Whitney tests) and False Discovery Rate (for Tukey’s tests) corrections were used to adjust for multiple comparisons. Analysis of Similarities (ANOSIM) was used to test for clustering of microbial communities using weighted and unweighted UniFrac distance matrices. QIIME v.1.8.0, R v.3.0.3 (R Core Team, 2014), PAST (Hammer, Harper & Ryan, 2001) and Excel were used for statistics and graphics. The Linear Discriminant Analysis (LDA) Effect Size (LEfSe) method was used to assess differences in microbial communities using a LDA score threshold of 3 (Segata et al., 2011). STAMP (Parks & Beiko, 2010) was used to visualize and analyze the PICRUSt data with ANOVA and False Discovery Rate. Unless otherwise noted, an alpha of 0.05 was considered to reject null hypotheses.

Fecal microbiota composition

A total of 8686 different OTUs were detected using the open reference algorithm described by Rideout et al. (2014). On the other hand, the closed_reference method used to generate data for PICRUSt (see PICRUSt below) only yielded 1302 OTUs. Fecal microbial composition of all mice was mostly comprised by Firmicutes (average: 58.7% across all samples) and Bacteroidetes (average: 32.8%) (Fig. 1). Other less abundant Phyla were Actinobacteria (∼4%), Proteobacteria (∼3%), Verrucomicrobia (∼0.8%) and others (Fig. 1). There was no statistical difference in relative abundance of the two most abundant phyla (Firmicutes and Bacteroidetes), partly because of the high variability among individual mice. The ratio Bacteroidetes/Firmicutes was lower in the lean (median: 38.7%) compared to the obese group (median: 85.3%) and the WW group (median: 75.4%) but this difference did not reach significance (p = 0.12, Kruskal-Wallis). Several statistical differences were found in low abundant phyla. Actinobacteria and Verrucomicrobia were higher in lean, Cyanobacteria, TM7 and Tenericutes were higher in the WW group, and Deferribacteres was higher in obese-control (p < 0.01, Kruskal-Wallis, Fig. 1).

LEfSe showed statistical significant differences for several microbial groups at lower taxonomic levels (Fig. 2). Among the bacterial groups that showed differences indicating an effect of WW-supplementation include the genus Lactobacillus, the class Gammaproteobacteria and the controversial S24-7 family http://groups.google.com/forum/#!topic/qiime-forum/Ds75aZoVrFY (Fig. 2). Other differences in bacterial abundances suggested that WW-supplementation did not generate a lean-like microbiome. For example, the genera Bifidobacterium, Allobaculum and Akkermansia were higher in lean compared to both obese and WW group (Fig. 2). Also, the family Ruminococcacea was more similar between obese (median: 6.9%) and WW-supplemented mice (5.6%) compared to lean (2.7%). Despite these differences, overall the fecal microbial composition of obese-control and WW mice was more similar to each other compared to lean although WW-supplementation yielded a unique pattern of bacterial abundances that did not necessarily cluster together with all obese samples (Fig. 3).

Alpha diversity

There was no significant difference in number of OTUs and Chao1 diversity index. Interestingly, samples from the obese group showed a more disperse distribution of OTUs (Fig. 4). Rarefied plots of number of OTUs showed that more than the 3000 sequences per sample used in this study are needed to fully describe the fecal microbiota of all mice.

Beta-diversity

Principal Coordinate Analysis (PCoA) of weighted and unweighted UniFrac metrics showed different clustering of microbial communities. Weighted (which takes phylogenetic information as well as sequence abundance into account) metrics clearly showed a different microbial structure in lean individuals compared to obese and WW groups (ANOSIM, p = 0.001) (Fig. 5). This was expected based on the clustering of lean subjects using relative abundance of sequence reads (Fig. 3). On the other hand, the qualitative (does not take sequence abundance into account) unweighted UniFrac analysis shows that the microbiota of the WW group clustered separately from the lean and obese groups (ANOSIM, p = 0.001) (Fig. 5).

Predicted metabolic profile

The taxa predicted by 16S RNA marker gene sequencing was used to predict the functional profile of the fecal microbiome in all three experimental groups. Using a p < 0.01 for ANOVA tests in STAMP, several statistical differences were found (Table 3). Overall, obese and WW groups tended to share more similar abundances of gene families compared to lean mice, an observation that supports the differences in bacterial abundances.

qPCR assessment of microbiota in distal colon contents

We performed qPCR analysis for bacterial groups of interest to health in distal colon contents. Similarly to the sequencing results from fecal samples, qPCR results revealed several differences in relative abundance for different bacterial groups (Fig. 6).

SCFAs caecal concentrations

There was a statistically significant difference among the treatment groups for several SCFAs in caecal contents (Table 4). Butyrate concentrations were higher in the WW group compared to both the lean and the obese group (p < 0.001, Kruskal-Wallis). Also, WW consumption was associated with lower succinic acid concentrations (p = 0.009, Kruskal-Wallis).

Caveats

This study was designed to obtain preliminary information about the influence of WW consumption on gut microbial ecology of obese diabetic mice; therefore, we did not aim to determine the exact compound(s) behind the observed effects. Wheat is a fiber-rich grain and consumption of fiber alone is associated with changes in the gut microbiome and the immune system of the host (Bermudez-Brito et al., 2015). Aside from fiber, WW also contains other bioactive compounds (e.g. polyphenols) that may be responsible for specific effects on host metabolism, physiology and immune system. For instance, it has been recently shown that wheat-derived alkylresorcinols were capable of showing beneficial effects on diet-induced obese mice (Oishi et al., 2015). Interestingly, our group showed that carbohydrate-free polyphenol-rich juice from plum is capable of impeding body weight gain in obese Zucker rats (Noratto et al., 2014), a finding that was not observed with WW consumption in mice in this current study. More research is necessary to investigate the separate effect of the different nutrients in WW.

Future directions

In humans, obesity is a multifactorial disease that can be partly controlled with dietary modifications. This paper adds valuable information to the current literature with regard to the potential influence of WW consumption on the gut microbiota of obese diabetic mice. However, research is needed to investigate the effect of WW on obese human individuals.