Impact of diet and individual variation on intestinal microbiota composition and fermentation products in obese men

Abstract

There is growing interest in understanding how diet affects the intestinal microbiota, including its possible associations with systemic diseases such as metabolic syndrome. Here we report a comprehensive and deep microbiota analysis of 14 obese males consuming fully controlled diets supplemented with resistant starch (RS) or non-starch polysaccharides (NSPs) and a weight-loss (WL) diet. We analyzed the composition, diversity and dynamics of the fecal microbiota on each dietary regime by phylogenetic microarray and quantitative PCR (qPCR) analysis. In addition, we analyzed fecal short chain fatty acids (SCFAs) as a proxy of colonic fermentation, and indices of insulin sensitivity from blood samples. The diet explained around 10% of the total variance in microbiota composition, which was substantially less than the inter-individual variance. Yet, each of the study diets induced clear and distinct changes in the microbiota. Multiple Ruminococcaceae phylotypes increased on the RS diet, whereas mostly Lachnospiraceae phylotypes increased on the NSP diet. Bifidobacteria decreased significantly on the WL diet. The RS diet decreased the diversity of the microbiota significantly. The total 16S ribosomal RNA gene signal estimated by qPCR correlated positively with the three major SCFAs, while the amount of propionate specifically correlated with the Bacteroidetes. The dietary responsiveness of the individual's microbiota varied substantially and associated inversely with its diversity, suggesting that individuals can be stratified into responders and non-responders based on the features of their intestinal microbiota.

Figure 1

Hierarchical clustering of the bacterial fingerprints of fecal samples collected from 14 subjects during four different diets (M diet, square; RS diet, circle; NSP diet, triangle; WL diet, diamond). The subject-wise clustering is highlighted with boxes. Vertical line is drawn at Pearson correlation of 0.95, which represents a reference value for the temporal stability of the microbiota in subjects without dietary intervention (for details, see Discussion section).

Figure 2

Dynamics of the microbiota diversity per volunteer during dietary shifts. Diets with statistically significant (P<0.05) difference are indicated with asterisks.

Figure 3

Heatmaps of changes in bacterial abundance between the M diet and the three test diets (N=non-starch polysaccharides, R=resistant starch, W=weight loss). Each row represents genus-like phylogenetic groups of bacteria whose mean abundance differed significantly between at least one pairwise comparison of the diets. The leftmost three columns show the mean change of those taxa that reached statistical significance (P<0.05) compared with the M diet, red denoting increase and blue decrease. The large heatmap represents intra-individual changes. The logarithmic fold-change compared with the M diet is indicated by the colors ranging from dark blue (logFC –1 or lower) to dark red (logFC 1 and higher).

Figure 4

Changes in abundant bacteria during dietary shifts as revealed by pairwise comparisons. The x axis shows logarithmic fold change of bacterial groups that differed significantly (P<0.05) and represented >0.5% of the total hybridization signal. The direction of bars indicates which of the two diets had higher abundance of the taxa listed on left side of each plot. The full list of significantly altered taxa is given in Supplementary Table S4.

Figure 5

Impact of diet on the abundance of Ruminococcus species in four volunteers, as assessed by qPCR. Primer pairs were designed that are specific for R. bromii, R. albus plus R. bicirculans, R. callidus plus R. flavefaciens plus R. champanellensis, and R. champanellensis together with R. champanellensis. Also shown are results obtained with two genus-level primer sets, the one used by Walker et al. (2011) and a newly designed set that gives better amplification of non-R.bromii species. Results are shown as % of the signal obtained with the universal 16S ribosomal RNA gene primer set. Primer sequences and amplification conditions are summarized in Supplementary Table S2. All of the available samples were analyzed from these four volunteers, providing a complete time course.

Figure 6

Relationship between proportion of propionate and of Bacteroides and Prevotella 16S ribosomal RNA sequences, as detected by qPCR. Data are means for all samples obtained from the last 2 weeks of each dietary period (Walker et al., 2011).

Figure 7

Percentage of total variation in the data that is explained by diet and by individual (ID). (a) Proportion of variance explained in the microbiota (as analyzed with the HITChip), SCFAs and the markers of insulin sensitivity. (b) Variance explained by diet when looking at subsets of data collected during M diet and one of NSP, RS or WL diets. (c) Proportion of variance explained on individual metabolites. Data refer to relative proportion (%) of each metabolite. Graphs (a, b) are based on multivariate analysis of variance (ANOVA) using distance matrices allowing for effects of volunteer and diet, (c) is based on univariate ANOVA allowing for effects of volunteer and diet. Asterisks indicate variables where the explained proportion reached statistical significance (P<0.05).

Figure 8

Relationship between plasma insulin and the proportion of bifidobacteria in the fecal microbiota. (a) HITChip data, based on one sample per diet. (b) qPCR data, based on means of all available samples taken in the last 2 weeks of the NSP, RS and WL diets and in the single week of the M diet.

Figure 9

Dietary responsiveness of the individual's microbiota. (a) Biplot of the study subjects' baseline microbiota and its net response to diets indicated by arrow length. Subjects with stabile microbiota, that is, non-responders, are circled. Methanogen carriage status assessed by qPCR (based on an average of 20 samples per volunteer, Walker et al., 2011) is indicated as follows: – absence of detectable methanogens; + present in <50% of samples; ++ present in >50% of samples. (b) Diversity of the microbiota on the M (run-in) diet in the non-responders and responders. Numbers refer to specific study subjects. Asterisk indicates statistically significant difference (P<0.05).