Dietary convergence induces individual responses in faecal microbiome composition

Abstract

Background: Dietary variation has been identified as a key contributor to microbiome diversification. However, assessing its true impact in a cross-sectional setting is complicated by biological confounders and methodological hurdles. We aimed to estimate the impact of a reduction of dietary variation (dietary convergence) on faecal microbiota composition among individuals consuming a Western-type diet.

Methods: 18 healthy volunteers recruited in the region of Flanders (Belgium) were followed up for 21 days. Participants were allowed to consume their habitual diet during a baseline and follow-up period (7 and 8 days, respectively), intersected by a 6-day intervention during which dietary options were restricted to oat flakes, whole milk and still water. Faecal samples were collected on a daily basis. Quantitative microbiome profiles were constructed, combining 16S rRNA gene amplicon sequencing with flow cytometry cell counting. Blood samples were taken at the beginning and end of each study week.

Results: While the intervention did not affect transit time (as assessed through the analysis of stool moisture), consumption of the restricted diet resulted in an increased prevalence of the Bacteroides2 microbiome community type. Microbial load and Faecalibacterium abundance decreased markedly. Despite dietary restrictions, no convergence of microbial communities (reduction of interindividual and intraindividual variation) was observed. The effect size (ES) of the intervention on genus-level microbiome community differentiation was estimated as 3.4%, but substantial interindividual variation was observed (1.67%-16.42%).

Conclusion: The impact of dietary variation on microbiome composition in a Western population is significant but limited in ES, with notable individual exceptions. Dietary convergence does not invariably translate into interindividual convergence of faecal microbial communities.

Keywords: Diet; Gastrointestinal microbiome; Intestinal microbiota; Microbiota; Nutrients; Prospective studies.

Figure 1. Design of The Oatmeal Study. The 21-day study period consisted of a baseline, an intervention and a follow-up phase following an A–B–A reversal design. During the intervention period, participants consumed only oatmeal, whole milk and still water. A food diary was kept throughout the entire study period, and a daily faecal sample was collected. An initial health assessment was performed by a medical practitioner and blood samples were taken on a weekly basis. BMI, body mass index.

Figure 2. Restriction of dietary variation as observed during the intervention period. (A) Principal component analysis of weight-based, food-group-level dietary variation over the study period (Euclidean distance; n=378). Each dot represents the dietary consumption of an individual participant during a day. Ellipses represent the 95% CI of the relevant group. (B) Boxplot presenting dietary variance observed before/after (baseline and follow-up combined) and during the intervention period. Variance significantly decreased during the intervention (n=378; dispersion test on Euclidean distances at food group level). The body of the boxplot represents the first and third quartiles of the distribution, with the median line, and the whiskers extend from the quartiles to the last data point within 1.5 times IQR, with outliers beyond. *Adjusted p value<0.001. (C) Distribution of weight-based average daily consumption of food groups (n=378). CI, confidence interval; IQR, interquartile range; PCA, Principal Component Analysis.

Figure 3. Microbiome variation associated with dietary convergence. (A) Boxplot comparing faecal microbial load variation over study phases. Load decrease during the intervention period (n=266; Kruskal-Wallis with post hoc Dunn test). (B) Weighted distribution of community-type prevalence over study phases. Bact2 prevalence increased significantly on dietary convergence (n=266; online supplemental figure 9). (C) Boxplot representing Faecalibacterium abundance distribution over study phases. A decrease in the abundance was observed during the intervention period (n=266; Kruskal-Wallis with post hoc Dunn test). (D) Boxplot presenting quantitative microbiome variance observed before/after (baseline and follow-up combined) and during the intervention period. Variance significantly increased during the intervention (n=266; dispersion test on Bray-Curtis dissimilarities at genus level). (A–D) Analyses performed excluding lag samples. (A,C,D) The body of the boxplot represents the first and third quartiles of the distribution, with the median line and the whiskers extend from the quartiles to the last data point within 1.5 times IQR, with outliers beyond. *adjP<0.05, **adjP<0.001. adjP, adjusted p value; Bact1, Bacteroides1; Bact2, Bacteroides2; IQR, interquartile range; Prev, Prevotella; Rum, Ruminococcaceae.

Figure 4. Impact of dietary convergence on microbiome diversification. (A) Principal coordinate analysis representing quantitative genus-level microbiome variation (n=381; Bray-Curtis dissimilarity). Arrows on the plot indicate direction and relative magnitude of shifts observed between baseline and intervention (dietary convergence) vs intervention and follow-up (dietary divergence). Microbiome community-type distribution is shown on the insert. Boxplots below axis 1 show the distribution of microbiomes belonging to the different study phases along the principal axis of variation. (B) Distance-based redundancy analysis of the association of potential covariates with quantitative genus-level microbiome variation (n=294). Transparent bars correspond with covariates that, while displaying a significant association in a univariable model, did not contribute significantly to microbiome variation in the stepwise variant. Bact1, Bacteroides1; Bact2, Bacteroides2; HbA1c, haemoglobin A1C; HDL, high-density lipoprotein; Prev, Prevotella; Rum, Ruminococcaceae.