Codon frequency is modulated by proteic selection, resulting in a coding profile in Archaea and Yeast

Abstract

Codons as fragments of the genetic code articulate both nucleotidic and proteic constraints. If codon usage bias is now admitted to be mainly influenced by GC content, codon frequencies in general may display a more subtle compromise between base composition and selection at proteic level. In order to investigate the existing non-GC content factors of codon frequencies, we compared coding sequences (CDS) of 280 Archaea plus S. cerevisiae genomes to their randomized version (same base-composition and same length). Through dedicated counts we identified several CDS vs random patterns in Archaea some of which reflecting probable or evident proteic constraint : in particular, the systematic enrichment of CDS in negatively charged amino acids, and the strong constraint existing on codons having a T in second position, which, on the basis of hydrophobic cluster analysis attests a folding constraint. The sum of these patterns constitutes a coding profile that enables to accurately classify about 99% of individual archaea sequences between CDS and randomized CDS. In S. cerevisiae, whose coding profile shares similarities with Archeae of close GC content, phylostratigraphic methods allowed to investigate the coding profile of CDS based on their relative age. This analysis reveals that contrary to other genes, the youngest genes (only found in S. cerevisiae) as a whole do not have a strong coding profile. This can be explained by their relative shortness in comparison with other genes. But even when taking length into account, a clear enrichment of misclassified sequences appears in the youngest S. cerevisiae genes. This enrichment may reflect an insufficient proteic optimization operated by selection.