Long-term dietary intervention reveals resilience of the gut microbiota despite changes in diet and weight

Abstract

Background: With the rising rates of obesity and associated metabolic disorders, there is a growing need for effective long-term weight-loss strategies, coupled with an understanding of how they interface with human physiology. Interest is growing in the potential role of gut microbes as they pertain to responses to different weight-loss diets; however, the ways that diet, the gut microbiota, and long-term weight loss influence one another is not well understood.

Objectives: Our primary objective was to determine if baseline microbiota composition or diversity was associated with weight-loss success. A secondary objective was to track the longitudinal associations of changes to lower-carbohydrate or lower-fat diets and concomitant weight loss with the composition and diversity of the gut microbiota.

Methods: We used 16S ribosomal RNA gene amplicon sequencing to profile microbiota composition over a 12-mo period in 49 participants as part of a larger randomized dietary intervention study of participants consuming either a healthy low-carbohydrate or a healthy low-fat diet.

Results: While baseline microbiota composition was not predictive of weight loss, each diet resulted in substantial changes in the microbiota 3-mo after the start of the intervention; some of these changes were diet specific (14 taxonomic changes specific to the healthy low-carbohydrate diet, 12 taxonomic changes specific to the healthy low-fat diet) and others tracked with weight loss (7 taxonomic changes in both diets). After these initial shifts, the microbiota returned near its original baseline state for the remainder of the intervention, despite participants maintaining their diet and weight loss for the entire study.

Conclusions: These results suggest a resilience to perturbation of the microbiota's starting profile. When considering the established contribution of obesity-associated microbiotas to weight gain in animal models, microbiota resilience may need to be overcome for long-term alterations to human physiology. This trial was registered at clinicaltrials.gov as NCT01826591.

Keywords: diet; low-carbohydrate; low-fat; microbiome; obesity; weight loss.

Copyright © The Author(s) on behalf of the American Society for Nutrition 2020.

Figures

FIGURE 1

One year of a low-carbohydrate or low-fat diet results in specific dietary alterations and weight loss. Twelve-month study design for 2 diet groups: healthy low-carbohydrate (white) and healthy low-fat (gray).  A: Weight, food-intake assessment, and stool for microbiome analysis were collected at the indicated sampling times. B–E: Levels of carbohydrate intake (B; grams), fat intake (C; grams), calorie intake (D; kilocalories), and weight (E; kilograms) over time for participants, separated by diet group. White: low-carbohydrate; gray: low-fat. Significant differences were assessed using a Wilcoxon paired test, adjusted P value <0.05 (Benjamini-Hochberg). Low-fat: n = 24; low-carbohydrate: n = 25. Significant differences in panel B: low-fat BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo; low-carbohydrate BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo, 3-mo vs. 12-mo; in panel C: low-fat BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo, 3-mo vs. 12-mo; in panel D: low-fat BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo, 3-mo vs. 12-mo; low-carbohydrate BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo; in panel E: low-fat BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo; low-carbohydrate BL vs. 3-mo, BL vs. 6-mo, BL vs. 12-mo, 3-mo vs. 12-mo, 6-mo vs. 12-mo. P values are listed in Supplemental Table 4. BL, baseline.

FIGURE 2

Evidence that gut microbiota composition is perturbed by diet but exhibits resilience over time. Bray-Curtis distance between all samples was calculated, and principal coordinate analysis was used to find new axes that captured the most variance across sample distance. Values for principal coordinate 1 (PC1) of Bray-Curtis distance was plotted, grouped by participant and by diet. Filled black circles, baseline sample; open triangles, 3-mo sample; +, 6-mo sample; x, 9-mo sample; open diamond, 12-mo sample. Low-fat: n = 24; low-carbohydrate: n = 25. Carb, carbohydrate.

FIGURE 3

Each diet results in distinct changes in the gut microbiota composition after 3-mo. Diet-specific compositional changes at 3-mo on low-carbohydrate (A) or low-fat (B) diets. Fractional abundances of taxa that significantly changed between the baseline sample and 3-mo sample exclusively in either diet group are shown. Significance calculated as adjusted q value <0.05 (SAM 2-class paired). Plots colored by phylum—in panel A: white, Proteobacteria; light gray, Bacteroidetes; dark gray, Firmicutes; in panel B: light gray, Actinobacteria; dark gray, Firmicutes. Gray boxes denote shared lineage. Significant changes found in both groups are shown in Figure 4A. “x_” indicates phylogenetic level where x = p, phylum; c, class; o, order; f, family; g, genus. Number of samples used was based on the lowest number of samples available in low-fat or low-carbohydrate per time point: BL vs. 3-mo: n = 22/diet; BL vs. 6-mo: n = 18/diet; BL vs. 9-mo: n = 23/diet; BL vs. 12-mo: n = 21/diet. BL, baseline; SAM, significance analysis of microarrays.

FIGURE 4

Changes observed in both diets correlate with weight. A: Compositional changes at 3-mo shared between the low-carbohydrate diet and low-fat diet. Abundances of taxa that significantly changed between the baseline sample and 3-mo sample on both diets are shown (all participants plotted). Significance calculated as adjusted q value <0.05 (SAM 2-class paired). White, healthy low-carbohydrate; gray, healthy low-fat. All taxa shown share the same phylogeny (Bacteroidetes phylum). Diet-specific changes are shown in Figure 3. “x_” indicates phylogenetic level where x = p, phylum; c, class; o, order; f, family; g, genus. Number of samples used was based on the lowest number of samples available in low-fat or low-carbohydrate per time point: BL vs. 3-mo: n = 22/diet; BL vs. 6-mo: n = 18/diet; BL vs. 9-mo: n = 23/diet; BL vs. 12-mo: n = 21/diet. B: Modeling results of taxonomic abundance and weight. Linear mixed-effects models were optimized per taxa, with taxa significantly associated with weight listed in the table (adjusted P value <0.05, Benjamini-Hochberg). Taxa were tested whose abundance was >1% in ≥5% of the samples. *Denotes taxa that were identified in Figure 4A. Low-fat: n = 24, low-carbohydrate: n = 25. BL, baseline; SAM, significance analysis of microarrays.