Highly Abundant Proteins Are Highly Thermostable

Abstract

Highly abundant proteins tend to evolve slowly (a trend called E-R anticorrelation), and a number of hypotheses have been proposed to explain this phenomenon. The misfolding avoidance hypothesis attributes the E-R anticorrelation to the abundance-dependent toxic effects of protein misfolding. To avoid these toxic effects, protein sequences (particularly those of highly expressed proteins) would be under selection to fold properly. One prediction of the misfolding avoidance hypothesis is that highly abundant proteins should exhibit high thermostability (i.e., a highly negative free energy of folding, ΔG). Thus far, only a handful of analyses have tested for a relationship between protein abundance and thermostability, producing contradictory results. These analyses have been limited by 1) the scarcity of ΔG data, 2) the fact that these data have been obtained by different laboratories and under different experimental conditions, 3) the problems associated with using proteins' melting energy (Tm) as a proxy for ΔG, and 4) the difficulty of controlling for potentially confounding variables. Here, we use computational methods to compare the free energy of folding of pairs of human-mouse orthologous proteins with different expression levels. Even though the effect size is limited, the most highly expressed ortholog is often the one with a more negative ΔG of folding, indicating that highly expressed proteins are often more thermostable.

Keywords: expression levels; misfolding avoidance hypothesis; protein thermostability; rates of evolution; translational robustness hypothesis.

Fig. 1.

Changes in free energy of folding due to mutations (ΔΔG of mutations) used in the study. For each term, the direction of the arrow indicates the initial sequence and the final sequence. Positive ΔΔG values indicate that the final sequence is less stable than the initial sequence. Negative ΔΔG values indicate that the final sequence is more stable than the initial sequence.

Fig. 2.

Correlation between the difference in the free energy of folding of human and mouse proteins and the difference in the abundance of human and mouse. Each dot corresponds to a pair of human–mouse orthologs (n = 47). The shaded area represents the 95% confidence interval. Spearman's rank correlation test significance level: *, P < 0.05.

Fig. 3.

Correlation between changes in free energy of folding (ΔΔG) and changes in protein abundance at different branches of the primate and rodent phylogeny. Each dot corresponds to a human or mouse protein. The shaded areas represent the 95% confidence intervals. Spearman's rank correlation test significance level: *, P < 0.05.

Fig. 4.

Correlation between the differences in the free energy of folding due to all possible translation errors (ΔΔG_t) and protein abundance. Each dot corresponds to one human or mouse protein structure (n = 51). Average and median ΔΔG_t values were estimated by substituting every amino acid position with the other 19 standard amino acids. The percent of destabilizing translation errors was computed as the fraction of translation errors with ΔΔG_t > 1. The shaded areas represent the 95% confidence intervals.