Abstract
Natural selection is believed to shape amino acid usage of the proteome by minimizing the energy cost of protein biosynthesis. Although this hypothesis explains well the amino acid frequency (AA(frequency)) difference among the 20 common amino acids within a given genome (species), whether it is applicable to cross-species difference remains to be inspected. Here, we proposed and tested a "metabolic rate hypothesis," which suggests that metabolic rate impacts genome-wide AA(frequency), considering that the energy allocated to protein biosynthesis is under selection pressure due to metabolic rate constraint. We performed integrated phylogenetic comparative analyses on proteomic sequence and metabolic rate data of 166 species covering 130 eumetazoan orders. We showed that resting metabolic rate (RMR) was significantly linked to AA(frequency) variation across animal lineages, with a contribution comparable to or greater than genomic traits such as GC content and codon usage bias. Consistent with the metabolic rate hypothesis, low-energy-cost amino acids are observed to be more likely at higher frequency in animal species with high (residual) metabolic rate. Correlated evolution of RMR and AA(frequency) was further inferred being driven by adaptation. The relationship between RMR and AA(frequency) varied greatly among amino acids, most likely reflecting a trade-off among various interacting factors. Overall, there exists no "one-size-fits-all" predictor for AA(frequency), and integrated investigation of multilevel traits is indispensable for a fuller understanding of AA(frequency) variation and evolution in animal.