EFFECTS OF WEIGHT LOSS AND DIET MODIFICATION ON GUT MICROBIOME AND TOTAL ENERGY EXPENDITURE IN A MULTI-CENTER, RANDOMIZED TRIAL

Low-carbohydrate (carb) diets may have beneficial effects on total energy expenditure (TEE) and body weight. Dietary carbohydrate also has been associated with changes in the microbiome, which may mediate the physiologic effects of macronutrients. But the technical difficulty of controlling or recording dietary intake in human studies is a major limiting factor for understanding how diet influences the microbiome. Our goal was to evaluate microbial changes after weight loss and microbial differences after randomization to high, moderate, and low carb diets. We also aimed to identify microbes associated with TEE.

Participants in the trial achieved 10-14% weight loss on a run-in diet and then were randomized to 3 test diets for 20 weeks of weight-loss maintenance. Percentages of total energy from carb-fat were 20-60 (low-carb), 40-40 (moderate-carb), and 60-20 (high-carb). Food was provided to maintain controlled macronutrient profiles and minimize differences between diets other than carb-fat percentages. About one-third of participants were enrolled in our ancillary microbiome study (n=54). Stool samples were collected at baseline, after weight loss, and at 2 months on the randomized diets. We performed 16S sequencing and analyzed data using Qiime2 and custom Python scripts. We assessed taxonomic changes with weight loss and differences by test diet, and used machine learning to classify dietary interventions with microbiome data. Pooling data across test diets, we assessed correlations of microbial abundance with TEE.

A random forest classifier predicted the pre- and post-weight loss time points using relative abundances at the genus level (Accuracy: 0.86, AUC: 0.96). Enrichment in Akkermansia and depletion of Bifidobacterium were among the top microbiome features after weight loss. Diet randomization did not confer major differences in the stool microbiome. A similar classifier could not distinguish microbiomes from the 3 diets (Accuracy: 0.39, AUC: 0.56). On the taxonomic level, participants on the low-carb diet maintained higher relative abundances of Faecalibacterium and Erysipelotrichaceae compared to those on the moderate- or high-carb diet (Kruskal-Wallis, FDR<0.1). We found six species associated with a change in TEE including Bacteroides uniformis, Bacteroides ovatus, and species belonging to Phascolarctobacterium, Christensenellaceae and Coriobacteriaceae (Spearman, FDR<0.1).

In a controlled feeding study, weight loss resulted in microbiome changes, consistent with previous studies. The 3 diets resulted in limited microbiome differences, demonstrating that carb-fat ratio within the studied range may have only a minor effect on gut microbial communities when controlling for potential dietary confounders. The mechanistic basis for associations between microbial abundance and energy expenditure requires further study.