Abstract
Objectives
Previous studies have defined dietary patterns for comparison with microbiome features using factor analysis from food frequency questionnaires. In this study, we applied a new tree-based method to directly define dietary patterns from 24-hour food records. We aimed to determine if these patterns corresponded with microbial features.
Methods
Daily fecal samples and daily 24-hour food records (ASA24–2016) were collected from 32 healthy adults over 17 days. Dietary patterns were derived using all reported foods for each subject. Foods were arranged into a tree structure using USDA food groups. Tree-based weighted Unifrac food distances (QIIME 1.9.1) were used for principal coordinate analysis to define five dietary patterns. Each pattern was named after its most influential food groups. Average microbiome composition was determined from metagenomic sequencing. Dietary patterns were compared with subjects’ average microbiome composition using correlation analysis. Spearman correlations were corrected for multiple comparisons within each taxonomy level. Constrained redundancy analysis (RDA) was used to determine the explanatory power of dietary patterns.
Results
Four of the five most discriminatory dietary patterns (DPs) were associated with microbial taxa (A). DP1 was positively correlated with an unclassified family in the order Burkholderiales and negatively correlated with the species Lachnospiraceae bacterium TF01–11. DP3, DP4, DP5 were most representative of a western diet. DP3 was negatively correlated with family Pasteurellaceae. DP4 was positively correlated with family Erysipelotrichaceae and negatively correlated with family Sutterellaceae. DP5 was positively correlated with members of class Bacteroidia including two specific Bacteroides speciesHMSC073E02 and HMSC067B03. Constrained RDA using the five dietary patterns revealed a gradient of Phylum Bacteroidetes along an axis driven by DP3, DP4, and DP5 (B).
Conclusions
The dietary patterns derived using our tree-based method reveal relationships between diet microbial taxa. In agreement with previous studies, our tree-based patterns show that the western diet corresponds to increased Bacteroidetes, demonstrating the utility of this method.
Funding Sources
Funding for this study was provided by General Mills.
Supporting Tables, Images and/or Graphs
