Metabolomes evolve faster than metabolic network structures

Metabolism plays a key role in cell growth and functions, and the evolution of metabolism represents a crucial aspect in the origin and evolution of cells. To investigate the evolution of metabolism, efforts have focused on the structural comparison of metabolic networks, which can be readily reconstructed from genomes in the genomic era (1). However, the structure of the metabolic network alone, i.e., the metabolic enzyme repertoire, is not able to reflect the functional state of metabolism, which is under the control of multifaceted regulatory events, e.g., on enzyme expression and activity (2). Therefore, there might be a weak association between metabolic networks and functional states, e.g., intracellular metabolite levels. With advanced technologies in metabolomics, the collection of the metabolites present in cells, i.e., metabolome, serves as a measurable functional readout of metabolism. By leveraging metabolomics, Tengölics et al. in PNAS (3) explore the evolution of metabolomes across yeast species and populations, leading to several interesting findings.

First, metabolomes vary considerably within and among yeast species. They measure levels of intracellular metabolites in nine species of the Saccharomycetaceae family and 17 genetically distinct populations of Saccharomyces cerevisiae. Comparisons between yeast species show substantial divergence in the levels of metabolites of primary metabolism, especially amino acids, and this trend also holds when comparing among populations of S. cerevisiae. Thus, this uncovers a diversity of metabolic states across yeast species and populations, complementing the understanding of the yeast evolution (4, 5).

By leveraging metabolomics, Tengölics et al. in PNAS explore the evolution of metabolomes across yeast species and populations, leading to several interesting findings.

Second, the metabolome variations cannot be explained by the metabolic network structures. Intuitively, variations in intracellular metabolite levels might be caused by structural alterations in the metabolic network that result in changed biosynthetic capacities. To test it, they use genome-scale metabolic models (GEMs) of yeast species (6) and S. cerevisiae populations (7) to calculate the metabolite production capacity, i.e., maximum theoretical yield. Note that the calculation depends exclusively on the reaction stoichiometry in the metabolic network and thus can be seen as a proxy of the metabolic network structure. However, comparisons both within and between species show no significant association between divergence in metabolite levels and production capacity. Interestingly, while most S. cerevisiae populations display identical metabolite production capacities, they differ significantly in the levels of intracellular metabolites such as amino acids. This indicates faster evolution rates of metabolite levels than the metabolic network structures (Fig. 1) and thus highlights the limitation of the research that exclusively compares metabolic network structures across evolutionary timescales. The lack of association between the evolution of metabolomes and metabolic network structures is probably due to differences in multiple layers of metabolic regulation, which could be systematically investigated by more multi-omics experiments (8).

Fig. 1.

Metabolomes evolve faster than metabolic network structures.

Third, the metabolome variations can be explained by both phylogenetic distance and domestication. Based on correlation analysis, they find that phylogenetic distance alone cannot explain the divergence of metabolite levels across S. cerevisiae populations and thus hypothesize that lifestyle-changing events, typically domestication, might noticeably affect the metabolomes. This hypothesis is supported by the observation that metabolomes are substantially different between wild and domesticated populations of S. cerevisiae. In addition, when separately analyzing wild and domesticated populations, there are significant correlations between phylogenetic distance and metabolome divergence. Therefore, both domestication and phylogenetic distance contribute to the observed divergence of metabolite levels within S. cerevisiae species. Furthermore, they find that the metabolome divergence among domesticated populations is significantly larger than that among wild populations, indicating that metabolomes evolve faster in domesticated than in wild populations. Taken together, they identify a profound impact of human practices on shaping the yeast metabolomes.

Last, domesticated populations exhibit a metabolic syndrome. Despite the metabolome divergence among domesticated populations, Tengölics et al. identify domestication signatures in a set of individual metabolites, which have significantly changed levels in the domesticated compared to wild populations of S. cerevisiae. This suggests convergent evolution of specific metabolites across independently domesticated S. cerevisiae populations that are from distinct human-made niches. The domestication signatures are observed in central metabolites such as amino acids and TCA cycle intermediates, which could be explained by human-driven selection on the aroma production and fermentation.

In conclusion, Tengölics et al. propose the use of metabolomes, rather than the metabolic network structures that are encoded in genomes, to interpret the evolution of metabolism, and demonstrate that metabolomes indeed provide previously unrecognized insights. Therefore, metabolomics would advance the field of the evolution of metabolism. Furthermore, they identify the diversity of metabolic states among S. cerevisiae populations with identical metabolic networks, which evidences the substantial variability of GEM simulations using constraint-based approaches (9) and thus emphasizes the inclusion of additional constraints, e.g., based on evolutionary information, in metabolic modeling.