Differences in Gut Metabolites and Microbial Composition and Functions between Egyptian and U.S. Children Are Consistent with Their Diets

Abstract

Previous studies indicated that populations consuming a Mediterranean diet rich in fiber, vegetables, and fruits have a significantly lower risk of cardiovascular and metabolic diseases than populations of industrialized societies consuming diets enriched in processed carbohydrates, animal proteins, and fats. To explore the potential contributions of gut microbiota to the observed diet-related metabolic effects, we conducted an integrative analysis of distal gut microbiota composition and functions and intestinal metabolites in Egyptian and U.S. teenagers. All Egyptian gut microbial communities belonged to the Prevotella enterotype, whereas all but one of the U.S. samples were of the Bacteroides enterotype. The intestinal environment of Egyptians was characterized by higher levels of short-chain fatty acids, a higher prevalence of microbial polysaccharide degradation-encoding genes, and a higher proportion of several polysaccharide-degrading genera. Egyptian gut microbiota also appeared to be under heavier bacteriophage pressure. In contrast, the gut environment of U.S. children was rich in amino acids and lipid metabolism-associated compounds; contained more microbial genes encoding protein degradation, vitamin biosynthesis, and iron acquisition pathways; and was enriched in several protein- and starch-degrading genera. Levels of 1-methylhistamine, a biomarker of allergic response, were elevated in U.S. guts, as were the abundances of members of Faecalibacterium and Akkermansia, two genera with recognized anti-inflammatory effects. The revealed corroborating differences in fecal microbiota structure and functions and metabolite profiles between Egyptian and U.S. teenagers are consistent with the nutrient variation between Mediterranean and Western diets. IMPORTANCE The human gastrointestinal microbiota functions as an important mediator of diet for host metabolism. To evaluate how consumed diets influence the gut environment, we carried out simultaneous interrogations of distal gut microbiota and metabolites in samples from healthy children in Egypt and the United States. While Egyptian children consumed a Mediterranean diet rich in plant foods, U.S. children consumed a Western diet high in animal protein, fats, and highly processed carbohydrates. Consistent with the consumed diets, Egyptian gut samples were enriched in polysaccharide-degrading microbes and end products of polysaccharide fermentation and U.S. gut samples were enriched in proteolytic microbes and end products of protein and fat metabolism. Thus, the intestinal microbiota might be selected on the basis of the diets that we consume, which can open opportunities to affect gut health through modulation of gut microbiota with dietary supplementations.

Keywords: children; diet; metabolomics; metagenomics; microbiota; nutrition.

FIG 1

Comparison of distal gut microbiota composition between Egyptian and U.S. groups. (A and B) Sample similarity was assessed by unconstrained weighted UniFrac-based principal-coordinate analysis (A) and constrained weighted UniFrac-based redundancy analysis (B) run on the chord-transformed genus abundance data set. The statistical significance of group separation in PCoA is based on the Davies-Bouldin index. Group clouds represent areas of three standard errors around the group centroid (diamond), and dot sizes in PCoA are proportional to Shannon’s H′ alpha diversity values for that sample. Distance-based RDA used three explanatory variables (group, age, and body mass index). The age and the BMI of the subjects are represented by color gradients and the size of each dot, respectively. Arrows in the db-RDA biplot denote the magnitudes and directions of the effects of explanatory variables. (C) A variation partitioning diagram depicts the relative contributions of explanatory variables to the overall variability in the data set. (D) Results of random forest discriminant analysis of chord-transformed genus abundances were visualized through multidimensional scaling of the sample proximity matrix. The statistical significance of group separation is based on the Davies-Bouldin index. Group clouds represent areas of three standard errors around the group centroid (diamond). MDS, multidimensional scaling. (E) The relative abundances of the top 12 RF discriminatory genera are depicted on a violin plot. Each violin shows the density distribution of genus abundances among all samples in the group. (F) Community structure at the class level. Classes are ordered according to the phylum. Where shown, single asterisks (*) and double asterisks (**) indicate statistical significance of taxon abundance differences between two groups (p_u < 0.05 and p_u < 0.01, respectively) based on the FDR-adjusted Mann-Whitney U test.

FIG 2

Enterotypes of the distal gut microbial profiles. (A) Relative abundances of the main enterotype drivers, Prevotella and Bacteroides, among the two identified clusters. Each violin shows the density distribution of genus abundance among all samples within the cluster (thickness of the violin), median value (black dot), and 25% to 75% range of values (black bar). (B) Sample clustering into different enterotypes based on the between-cluster ordination analysis. (C) Comparison of Prevotella and Bacteroides relative abundances in four samples based on sequencing (Seq) and fluorescent in situ hybridization (FISH) results. Representative fluorescent images are shown on the right; pink coloring corresponds to genus-specific fluorescent probes, and blue coloring represents DAPI DNA staining. (D) Results of weighted UniFrac distance-based PCoA performed using OTU abundances showing separation of fecal samples between industrialized and nonindustrialized countries. The statistical significance of group separation is based on the Davies-Bouldin index. Panels E and F display the same PCoA space, with relative abundances of Prevotella (E) and Bacteroides (F) overlaid as dot color gradients.

FIG 3

Comparison of distal gut metabolite profiles between Egyptian and U.S. groups. (A) Overall distribution of metabolite categories among egkHLT and uskHLT samples shown on a stacked-column graph. The relative abundances of individual measured metabolites are represented as violin plots. Where shown, single asterisks (*) and double asterisks (**) indicate statistical significance of taxon abundance differences between two groups (p_u < 0.05 and p_u < 0.01, respectively) based on the FDR-adjusted Mann-Whitney U test. LDRMs, lipid degradation-related metabolites. (B) Statistically significant SparCC-based correlations among individual metabolites and metabolite groups. The associations between metabolite categories represent the median of pairwise correlations among individual metabolites from different categories. (C and D) Sample distribution in ordination space based on the centered log-ratio transformed metabolite relative abundances from principal-component analysis (C) and random forest discriminant analysis (D). P values indicate statistical significance of separation of sample groups based on the Davies-Bouldin index. Group clouds represent three standard errors around the group centroids (diamond). MDS, multidimensional scaling. (E) Coinertia analysis showing congruency of sample dispersal in ordination space based on metabolite and genus abundance profiles. The distance between the positions of each sample on two ordination plots is indicated by a connecting line. Shorter lines represent similar sample positions in the plots. Statistical significance and the relative fit of the ordinations were assessed by P value and RV coefficient, respectively.

FIG 4

Functional analysis of the distal gut microbiota metagenomes. (A) Differences in specific functional categories between Egyptian and U.S. gut microbiota. Each category comprises several SEED level 3 entries, each represented as individual bars with results calculated as log₂-transformed ratios of reads between egkHLT and uskHLT metagenomes. All displayed entries are statistically significant, with FDR-corrected P of <0.01. (B) Relative abundances of specific metabolites and the mean log₂ ratios of reads of metabolic enzymes that produce and/or utilize the metabolites are compared. CoA/P, coenzyme A/phosphate.