Comparative metabolomics in primates reveals the effects of diet and gene regulatory variation on metabolic divergence

Abstract

Human diets differ from those of non-human primates. Among few obvious differences, humans consume more meat than most non-human primates and regularly cook their food. It is hypothesized that a dietary shift during human evolution has been accompanied by molecular adaptations in metabolic pathways. Consistent with this notion, comparative studies of gene expression levels in primates have found that the regulation of genes with metabolic functions tend to evolve rapidly in the human lineage. The metabolic consequences of these regulatory differences, however, remained unknown. To address this gap, we performed a comparative study using a combination of gene expression and metabolomic profiling in livers from humans, chimpanzees, and rhesus macaques. We show that dietary differences between species have a strong effect on metabolic concentrations. In addition, we found that differences in metabolic concentration across species are correlated with inter-species differences in the expression of the corresponding enzymes, which control the same metabolic reaction. We identified a number of metabolic compounds with lineage-specific profiles, including examples of human-species metabolic differences that may be directly related to dietary differences.

Figure 1. Known metabolites with human-specific levels.

Metabolite levels (y-axis, log-scale) in human (red), chimpanzee (green), and rhesus macaque (black). Error bars represent the ±SE within each species.

Figure 2. Interaction network enriched with human-specific metabolites.

Shaded nodes indicate metabolites showing a human-specific concentration, with undashed lines connecting compounds involved in the same metabolic reaction. Orange lines are connecting human-specific metabolites. The interaction network was generated using the Ingenuity Pathway Analysis (IPA) tool.

Figure 3. Metabolites involved in the same reactions change their levels in concert within and between species.

Cumulative distribution (y-axis) of correlation coefficients across individuals (x-axis) for metabolites involved in the same reaction (orange) and different reactions (grey), within humans (A), chimpanzees (B), and rhesus macaques (C). Below each distribution we show the median pairwise correlation with error bars representing a 95% confidence interval calculated using bootstrapping (1000 resamplings); the difference in medians is significant in all species (human P = 6 × 10⁻³, chimpanzee P = 2 × 10⁻³, and rhesus macaque P < 10⁻³, using a permutation test; see SI Materials and Methods). (D) Proportion of metabolite pairs that change their levels in the same direction in human and chimpanzee, considering pairs where both metabolites are differentially concentrated between the species. The proportion is significantly different between metabolite pairs involved in the same reaction (orange) and different reactions (grey, P = 0.019 using a one-sided Fisher's exact test). See Supplementary Fig. S8 for a similar plot considering pairs where at least one of the metabolites is differentially concentrated between the species. Error bars represent 95% confidence intervals calculated using bootstrap resampling.

Figure 4. Metabolites that are differentially concentrated between species are more likely to be involved in reactions catalyzed by enzymes that are differentially expressed between species.

(A) Cumulative distribution of likelihood ratios (LR) for differential expression (DE, x-axis) for enzymes associated with differentially concentrated (DC, blue) and non-DC (grey) metabolites. The panel within the plot illustrates the medians of the two distributions with 95% confidence intervals calculated using bootstrapping (see SI Materials and Methods). (B,C) The LR of enzymes associated with DC and non-DC metabolites (y-axis) is plotted for multiple cutoffs defining DC (x-axis) in human compared to chimpanzee (B) and rhesus macaque (C).

Figure 5. Higher enzyme differential expression between species predicts a higher level of differential concentration for metabolites involved in the same reaction.

(A) Mean LR (y-axis) for DC between human and chimpanzee in metabolites associated with enzymes that are DE (green) and non-DE (grey). Error bars correspond to 95% confidence intervals calculated using bootstrapping (see SI Materials and Methods). (B) The difference between the means described in A (y-axis) for multiple cutoffs to define enzyme DE (x-axis). The dashed line represents a linear regression of the plotted values (correlation significance is P = 1.37 × 10⁻⁵).