Phylogenetic analysis of mutational robustness based on codon usage supports that the standard genetic code does not prefer extreme environments

Abstract

The mutational robustness of the genetic code is rarely discussed in the context of biological diversity, such as codon usage and related factors, often considered as independent of the actual organism’s proteome. Here we put the living beings back to picture and use distortion as a metric of mutational robustness. Distortion estimates the expected severities of non-synonymous mutations measuring it by amino acid physicochemical properties and weighting for codon usage. Using the biological variance of codon frequencies, we interpret the mutational robustness of the standard genetic code with regards to their corresponding environments and genomic compositions (GC-content). Employing phylogenetic analyses, we show that coding fidelity in physicochemical properties can deteriorate with codon usages adapted to extreme environments and these putative effects are not the artefacts of phylogenetic bias. High temperature environments select for codon usages with decreased mutational robustness of hydrophobic, volumetric, and isoelectric properties. Selection at high saline concentrations also leads to reduced fidelity in polar and isoelectric patterns. These show that the genetic code performs best with mesophilic codon usages, strengthening the view that LUCA or its ancestors preferred lower temperature environments. Taxonomic implications, such as rooting the tree of life, are also discussed.

The Mutational Robustness of the Genetic Code and Codon Usage in Environmental Context: A Non-Extremophilic Preference?

“…Unfortunately, the exact effect on hydrophobicity is difficult to entangle in this study, albeit a slightly pronounced AT-bias is the most likely to minimize distortion. This could be in accord with an earlier analysis accounting for phylogenetic bias in a small sample of prokaryotes [ 69 ], and other observations attesting for a general AT-biased substitution pattern [ 57 , 58 , 59 ]. If that is the case, an A/T-mutation should be more likely to fix due to its lower risk of perturbing hydrophobic patterns in proteins.…”

“…This is demonstrated on two datasets: (i) one dataset documents optimal temperature, salt concentration, and pH conditions for Bacteria and Archaea [ 67 ], whereas the (ii) second compiles optimal growth temperature conditions for a more comprehensive sample including members of all three domains [ 68 ]. Although Bacteria are overrepresented in both, our earlier analysis on a smaller but phylogenetically controlled dataset showed similar effects that are robust to taxon sampling [ 69 ]. We can also confirm that the inclusion of Eukaryotes is unlikely to change this picture.…”

“…Distortion (Equation (1)) is used to estimate the average effect of mutations [ 16 , 17 , 18 , 69 ]. It is calculated given a source distribution of codons, P(c i ) , and the uncertainty of the code resulting from noise, P (Y = c j |X = c i ) , i.e., the probability of codon c i mutating into c j (see next section about background mutation model).…”

“…We must estimate the raw, a priori performance of the genetic code without natural selection introducing additional bias. To this effect, we employ a simple model of random mutations [ 69 ] (Equations (2)–(4)), which is reminiscent of Kimura’s two parameter model [ 73 ]. Here, κ denotes the transition/transversion rate ratio, otherwise known as the ti/tv-ratio, and µ is the mutation rate.…”

Robust genetic codes enhance protein evolvability

“…To overcome this limitation, we additionally considered definitions of code robustness based on a large number of diverse physicochemical properties, one at a time (Section 2 in S1 Text ), and on a custom amino acid similarity score specific to each of the 6 data sets (Section 3 in S1 Text ), with qualitatively the same results. Fourth, the error tolerance of a genetic code, standard or nonstandard, is influenced by mutation bias and codon usage [ 16 , 83 ] as they make some mutations more likely than others. While mutation bias and codon usage may influence peak accessibility in adaptive landscapes [ 84 ], they do not affect landscape topography, which is why we have not considered these effects here.…”