Thermophilic Adaptation in Prokaryotes Is Constrained by Metabolic Costs of Proteostasis

Abstract

Prokaryotes evolved to thrive in an extremely diverse set of habitats, and their proteomes bear signatures of environmental conditions. Although correlations between amino acid usage and environmental temperature are well-documented, understanding of the mechanisms of thermal adaptation remains incomplete. Here, we couple the energetic costs of protein folding and protein homeostasis to build a microscopic model explaining both the overall amino acid composition and its temperature trends. Low biosynthesis costs lead to low diversity of physical interactions between amino acid residues, which in turn makes proteins less stable and drives up chaperone activity to maintain appropriate levels of folded, functional proteins. Assuming that the cost of chaperone activity is proportional to the fraction of unfolded client proteins, we simulated thermal adaptation of model proteins subject to minimization of the total cost of amino acid synthesis and chaperone activity. For the first time, we predicted both the proteome-average amino acid abundances and their temperature trends simultaneously, and found strong correlations between model predictions and 402 genomes of bacteria and archaea. The energetic constraint on protein evolution is more apparent in highly expressed proteins, selected by codon adaptation index. We found that in bacteria, highly expressed proteins are similar in composition to thermophilic ones, whereas in archaea no correlation between predicted expression level and thermostability was observed. At the same time, thermal adaptations of highly expressed proteins in bacteria and archaea are nearly identical, suggesting that universal energetic constraints prevail over the phylogenetic differences between these domains of life.

Keywords: chaperones; evolution; homeostasis; protein folding; thermophiles.

Fig. 1.

Material and energy flux in proteostasis. Amino acid biosynthesis, translation and polypeptide synthesis, and chaperone assisted protein folding consume a significant fraction of energy $\mathcal{E}$ available to a prokaryote. Maintenance of steady state concentrations of every amino acid bears a known energy cost, with cheaper amino acids preferred in highly expressed proteins (Akashi and Gojobori 2002). We propose that energy cost of chaperone activity depends on amino acid composition of client proteins, as protein foldability is affected by amino acid composition (Dill 1985; Berezovsky et al. 2007; Venev and Zeldovich 2015). Therefore, amino acid composition evolves under the energetic constraint from two distinct processes, amino acid biosynthesis costs and chaperone activity.

Fig. 2.

Convergence of the archaeal and bacterial trends of thermal adaptation. Slopes of the amino acid frequencies versus OGT regressions are compared between archaea and bacteria. (A) Proteome-wide, the temperature trends of amino acid usage in bacteria and archaea are weakly correlated. (B) Ribosomal proteins of archaea and bacteria have similar patterns of thermal adaptation. (C) Predicted highly expressed proteins (top 10% CAI) in the organisms with CUS show identical patterns of thermal adaptation between bacteria and archaea. (D) Correlation of trends of thermal adaptation in complete proteomes of organisms with CUS is statistically insignificant in the organism bootstrap test (see text for details).

Fig. 3.

Simulated frequencies of amino acids compared with the naturally evolved ones for bacteria. (A, B) Jensen–Shannon divergence between amino acid frequencies in simulated data and thermophilic (A) and mesophilic (B) proteomes exhibits clear minima with respect to the temperature T and shaperone-adjusted synthesis costs w. The best match between the model and mesophilic proteomes is achieved at a lower temperature than the best match to the thermophilic ones. (C) Pearson correlation coefficient R_A between amino acid frequencies in simulated data and all bacterial proteomes reaches $R\sim 0.9$ for $0.7<T<0.9$. (D) In the same temperature range, the temperature trends of amino acid in simulated data are strongly correlated with those in bacterial proteomes, R = 0.64.

Fig. 4.

Temperature trends of the amino acid synthesis and maintenance costs in prokaryotes. Proteome average cost of amino acid synthesis and amino acid maintenance for archaeal species (A, B) and bacterial species (C, D). Marker color represents the genome-wide GC content of each specie; as it is well-established, genome-wide GC content is not correlated with OGT, see also supplementary figure S1A and B, Supplementary Material online.

Fig. 5.

Similarity between proteins with high CAI and thermophilic proteins by comparing amino acid composition. Proteins in each organism are grouped into 20 bins according to CAI, and amino acid composition within each groups is compared with the average thermophilic composition using Jensen–Shannon divergence, separately for archaea (A), and bacteria (B). In bacteria, the higher is protein expression, the more similar is amino acid composition to a thermophilic one (decreasing JSD_T, red line, statistically significant). No such trend was observed for archaea. As a control, codon reshuffling was used to destroy the relation between CAI and amino acid composition of proteins. For both archaea and bacteria, the correlation between JSD_T and CAI for reshuffled codons was not significant. Error bars represent the 30% and 70% percentiles of the underlying distributions.