Amino‑acid composition is a selection target for coding‑sequence retention: evidence from out‑of‑frame translation comparisons in Escherichia coli

Abstract

Protein evolution is a major engine of biological diversification, but the selective pressures that determine which sequences are retained as genes remain debated. In previous work, we showed that, across the three domains of life, per‑protein amino‑acid residue fractions in reference proteomes—defined here as the set of proteins encoded by the annotated coding sequences (CDS) of a reference genome—consistently form species‑specific, bell‑shaped distributions well approximated by binomial expectations for each of the 20 amino acids. If a genome can, in principle, encode a wide range of amino‑acid compositions yet the realized set of coding sequences occupies only a narrow region of that space, this would imply that amino‑acid composition itself is a target of selection during gene retention. Here we test this idea in Escherichia coli. Using this definition, we compared the distributions of per‑CDS residue fractions for the reference proteome’s native (+1) translations with those for genome‑encoded out‑of‑frame translations obtained by re‑parsing the unaltered CDS in non‑native frames (+2, +3 on the plus strand; −1, −2, −3 on the reverse complement). We then located the reference proteome within the composition space spanned by these alternatives. The reference proteome was concentrated within a markedly narrower, shifted region of composition space than that spanned by the out‑of‑frame translations, with especially strong separations for cysteine, aspartate, glutamate, and arginine. Thus, despite the genome’s capacity to encode diverse compositions, the reference proteome lies within a restricted subset—consistent with amino‑acid residue composition being an important target of selection in coding‑sequence retention.