The series of genetic codes that preceded our own

Abstract

We produce a working hypothesis for the steps by which the canonical genetic code might have been constructed. We assume amino acids were added in the order inferred from their enrichment or depletion in ancestrally reconstructed sequences dating back to the Last Universal Common Ancestor (LUCA) relative to that in ancient but post-LUCA controls. To convert this ordering into a time series of 64-codon genetic codes, we followed the “2-1-3 rule” that information appeared first at the middle codon position, then mostly at the first before the third, with purine vs pyrimidine distinctions tending to precede distinctions between purines or between pyrimidines. Two notable deviations from the 2-1-3 rule suggest an ancestral promiscuity between valine/isoleucine/methionine, and between alanine/threonine, to produce “statistical proteins.” The same promiscuities enable amino acids to be added in order of appearance at successive nodes in the cladograms of Class I and Class II aminoacyl-tRNA synthetase catalytic domains. Trifonov's previous amino acid ordering did not produce this consilience, with both a worse fit to the 2-1-3 rule, and discordance of synthetase trees with the order. More support for late resolution of hydrophobic promiscuity comes from post-LUCA differentiation between the synthetases of valine and isoleucine, in a manner dependent on 2 conserved tryptophans. The observed promiscuous activities of synthetases, and deep mutational scanning data on which amino acid substitutions are best tolerated, also align well with our reconstructed time series of genetic codes. As a working hypothesis, this series can guide research on random peptides, urzymes, and substitution models.