Individuality and ethnicity eclipse a short-term dietary intervention in shaping microbiomes and viromes

Abstract

Many diseases linked with ethnic health disparities associate with changes in microbial communities in the United States, but the causes and persistence of ethnicity-associated microbiome variation are not understood. For instance, microbiome studies that strictly control for diet across ethnically diverse populations are lacking. Here, we performed multiomic profiling over a 9-day period that included a 4-day controlled vegetarian diet intervention in a defined geographic location across 36 healthy Black and White females of similar age, weight, habitual diets, and health status. We demonstrate that individuality and ethnicity account for roughly 70% to 88% and 2% to 10% of taxonomic variation, respectively, eclipsing the effects a short-term diet intervention in shaping gut and oral microbiomes and gut viromes. Persistent variation between ethnicities occurs for microbial and viral taxa and various metagenomic functions, including several gut KEGG orthologs, oral carbohydrate active enzyme categories, cluster of orthologous groups of proteins, and antibiotic-resistant gene categories. In contrast to the gut and oral microbiome data, the urine and plasma metabolites tend to decouple from ethnicity and more strongly associate with diet. These longitudinal, multiomic profiles paired with a dietary intervention illuminate previously unrecognized associations of ethnicity with metagenomic and viromic features across body sites and cohorts within a single geographic location, highlighting the importance of accounting for human microbiome variation in research, health determinants, and eventual therapies. Trial Registration: ClinicalTrials.gov ClinicalTrials.gov Identifier: NCT03314194.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Experimental design and diet profiles.

(A) The VMIC study examined short-term longitudinal, multiomic data from multiple body sites over 9 days with 3 broad categories of evaluation (before: habitual diet, days −1 and 0; during: vegetarian diet, days 1 through 4; after: habitual diet, days 5 through 7). (B, C) Nutrient profiles inclusive of macro- and micronutrients (normalized by kcals) of the habitual diet and daily, vegetarian, study diets (Data underlying plot B and C can be found at S1 Data) are shown by NMDS ordination analysis based on Euclidean distance and statistically compared for significant differences by PERMANOVA. Human cliparts in plot A were created with BioRender.com. NMDS, nonmetric multidimensional scaling; PERMANOVA, permutational multivariate analysis of variance; VMIC, Vanderbilt Microbiome Innovation Center.

Fig 2. Dendrograms summarizing relationships of microbial community compositions before, during, and after the diet.

Participants are individually colored in the branches, and ethnicities (light blue and pink) are denoted at the tips of the branches along with the before (green), during (red), and after (blue) stages of the dietary intervention. Unweighted pair group method with arithmetic mean (UPGMA) clustering trees are shown for the (A, D) gut and (B, E) oral microbial community compositions and (C, F) gut viral community composition in (A-C) cohort 1 and (D-F) cohort 2 based on Bray–Curtis distances (permutations = 999). Data underlying this figure can be found at S1 Data.

Fig 3. Ethnicity-associated variation in metagenomic communities and functions.

The gradient of colors denotes variation explained by the variables, and asterisk symbol indicates p-value of PERMANOVA with Bray–Curtis distance that was performed at all combined dietary stages for taxonomy at the strain level, COGs, ARGs, and CAZymes [peptidoglycanase for virome]. adonis2(data ~ Ethnicity + Antibiotic Use + Hormonal Contraceptive, permutations = perm, method = “bray”, by = “margin”), where perm = with (data, how(nperm = 999, blocks = Day)). See S2 Fig for details of functional categories used for assembly-based PERMANOVA. Data underlying this figure can be found at S1 Data. ARG, antibiotic-resistant gene; CAZymes, carbohydrate active enzymes; COG, cluster of orthologous groups; PERMANOVA, permutational multivariate analysis of variance.

Fig 4. Ethnicity-associated variation in abundant microbial taxa and phage.

The 219 taxa in cohort 1 (A) and 182 taxa in cohort 2 (B) included in the phylogeny are present in all participants and have relative abundances of >1% in at least one of the gut or oral microbial metagenomes. The inner circle indicates differential taxa in the gut between the 2 ethnicities; the outer circle indicates differential taxa in the saliva between the 2 ethnicities. Pink or blue color gradients indicate FDR-adjusted p-value of significance (LinDA function in MicrobiomeStat package fitting linear mixed-effects models ‘~ Ethnicity + Antibiotic Use + Hormonal Contraceptive + (1|Day)’) for centered log-ratio transformed abundance between the 2 ethnicities. Data underlying this figure can be found at S1 Data. Pink indicates more abundant in Black participants, and blue indicates more abundant in White participants. Star indicates differential abundant taxa in the gut between the 2 ethnicities in both cohorts. Taxon name with a “G” before the name indicates the taxon was classified at the genus level.

Fig 5. Dietary intervention alters urine and plasma metabolomes.

Univariate analysis on metabolomic samples before and after a short-term diet identified 180 significantly changed total metabolites in cohort 1 and 152 metabolites in cohort 2 (PFDR < 0.05, Wilcoxon signed-rank test). Total urine metabolome compositions were not significantly different between diet stages or ethnicities in either cohort (A, E) (PFDR < 0.05, PERMANOVA). In cohort 1, 97/842 urine metabolites (D) significantly changed over the diet period, while in cohort 2, 80/835 urine metabolites (H) significantly changed. Total plasma metabolome compositions were significantly different between diet phases (I, M) in both years (PFDR < 0.05, PERMANOVA). In cohort 1, 83/796 plasma metabolites (L) significantly changed between diet phases, while in cohort 2, 72/736 plasma metabolites (P) were significantly different. Pairwise Euclidean distance between all participants and interethnic groups significantly changed in urine (B, C, F, G) and plasma (J, K, N, O) across cohorts (P < 0.05, Wilcoxon rank-sum test). Statistics in parentheses denote the removal of outlier participants (see Materials and methods, Metabolomics analysis). Individual scaled abundances are shown above by stage and by ethnicity and ordered by super pathway. Heatmaps reflect all significantly changed metabolites within cohort-matched, color-coded dots and ordered by decreasing log 10-fold change within each super pathway. Circles denote metabolites significant in both cohorts; squares represent metabolites significantly different in both cohorts and in both urine and plasma. Abundance corresponds to scaled and transformed intensity value of each metabolite. A key of ordered metabolites, fold changes, and p-values can be found in S1 Data.