Modular Evolution of the Drosophila Metabolome

Abstract

Comparative phylogenetic studies offer a powerful approach to study the evolution of complex traits. Although much effort has been devoted to the evolution of the genome and to organismal phenotypes, until now relatively little work has been done on the evolution of the metabolome, despite the fact that it is composed of the basic structural and functional building blocks of all organisms. Here we explore variation in metabolite levels across 50 My of evolution in the genus Drosophila, employing a common garden design to measure the metabolome within and among 11 species of Drosophila. We find that both sex and age have dramatic and evolutionarily conserved effects on the metabolome. We also find substantial evidence that many metabolite pairs covary after phylogenetic correction, and that such metabolome coevolution is modular. Some of these modules are enriched for specific biochemical pathways and show different evolutionary trajectories, with some showing signs of stabilizing selection. Both observations suggest that functional relationships may ultimately cause such modularity. These coevolutionary patterns also differ between sexes and are affected by age. We explore the relevance of modular evolution to fitness by associating modules with lifespan variation measured in the same common garden. We find several modules associated with lifespan, particularly in the metabolome of older flies. Oxaloacetate levels in older females appear to coevolve with lifespan, and a lifespan-associated module in older females suggests that metabolic associations could underlie 50 My of lifespan evolution.

Keywords: Drosophila; coevolution; evolution; metabolome; modular.

Fig. 1.

Evolution of the Drosophila Metabolome. (A) Samples from each strain at each age (young 5 days, or old 31 days) and sex are plotted along principal components 1 (PC1) and 2 (PC2) of the mean-centered and scaled log-abundance of 97 targeted metabolites. Each sample is mapped by lines to the centroid of its respective sex and age group. The colored legend at bottom right refers to the sex and age groups in each figure. (B) The mean of PC1 (left) or PC2 (right) of each species is plotted along the phylogeny (Kumar et al., 2017). (C) The pairwise metabolome distance between all samples within each age and sex is plotted over the divergence time between species pairs. Intraspecies sample distances are plotted over time = 0. The metabolome distance was calculated as the average squared difference of the log-metabolite abundance for the 97 targeted metabolites between all pairs of samples. The lines in (C) represent the fit to an ordinary least squares linear model: metabolome divergence ∼ divergence time within each group and are equivalent to the BM model of trait evolution.

Fig. 2.

Modular Evolution of the Metabolome. Heatmaps showing the pairwise absolute correlation between PICs for the 97 metabolites in each sex and age group (young 5 days, or old 31 days). The modules identified by WGCNA are shown on the left axis and partitions are included in the heatmap corresponding to these modules. Metabolites not fitting the criteria for modularity are in the “gray” category. Hierarchical clustering was done in WGCNA (Materials and Methods) and is shown in supplementary figure S4, Supplementary Material online.

Fig. 3.

Metabolite Variation by Sex and Age. (A) The effects (β) of sex, age, or their interaction on metabolite levels in Drosophila estimated in a phylogenetic mixed model. For sex, metabolites that are more abundant in males have a positive β. The effects are clustered by metabolite (row). (B) The number of metabolites effected by sex, age, or their interaction (n = 97, FDR ≤ 0.05).

Fig. 4.

Lifespan and Metabolite Coevolution. Mean lifespans of females (A) and males (B) from 11 Drosophila species are mapped on the phylogeny and a color scale is added based on estimated rates of continuous trait evolution. Means are from 16 to 378 individuals from each of 1–3 strains per species per sex. At each sex and age, PICs of mean lifespan for each species were regressed on PICs for 97 targeted metabolites using major axis regression forced through the origin. Plots of the regression for the three lowest P values from females (C) and males (D) of each age group, and the FDR of each association is show inside the plots. The regression data for all metabolites are shown in supplementary table S6, Supplementary Material online.