Evolutionary systems biology of amino acid biosynthetic cost in yeast

Abstract

Every protein has a biosynthetic cost to the cell based on the synthesis of its constituent amino acids. In order to optimise growth and reproduction, natural selection is expected, where possible, to favour the use of proteins whose constituents are cheaper to produce, as reduced biosynthetic cost may confer a fitness advantage to the organism. Quantifying the cost of amino acid biosynthesis presents challenges, since energetic requirements may change across different cellular and environmental conditions. We developed a systems biology approach to estimate the cost of amino acid synthesis based on genome-scale metabolic models and investigated the effects of the cost of amino acid synthesis on Saccharomyces cerevisiae gene expression and protein evolution. First, we used our two new and six previously reported measures of amino acid cost in conjunction with codon usage bias, tRNA gene number and atomic composition to identify which of these factors best predict transcript and protein levels. Second, we compared amino acid cost with rates of amino acid substitution across four species in the genus Saccharomyces. Regardless of which cost measure is used, amino acid biosynthetic cost is weakly associated with transcript and protein levels. In contrast, we find that biosynthetic cost and amino acid substitution rates show a negative correlation, but for only a subset of cost measures. In the economy of the yeast cell, we find that the cost of amino acid synthesis plays a limited role in shaping transcript and protein expression levels compared to that of translational optimisation. Biosynthetic cost does, however, appear to affect rates of amino acid evolution in Saccharomyces, suggesting that expensive amino acids may only be used when they have specific structural or functional roles in protein sequences. However, as there appears to be no single currency to compute the cost of amino acid synthesis across all cellular and environmental conditions, we conclude that a systems approach is necessary to unravel the full effects of amino acid biosynthetic cost in complex biological systems.

Figure 1. Comparison of amino acid biosynthetic cost estimates.

Amino acid biosynthetic cost estimates are compared in the barcharts on the left hand side. Similarities among different cost types are visualised as a tree on the right hand side. The closer two cost types are in the tree the more similar the cost estimates. Each barchart axis shows the minimum and maximum value of each cost type, rounded to three significant figures. We note that absolute costs computed using our systems biology approach are unitless (see Materials and Methods for details). The cost comparison dendrogram was generated using complete agglomerative clustering of Spearman's Rank correlations between each cost type (see File S2).

Figure 2. Comparison of factors explaining observed transcript and protein levels in S. cerevisiae.

Each point is the contribution of the variable to explaining either protein or transcript levels. Points to the right have a greater contribution and vice versa for points to the left. Multiple points are shown for each variable in the figure, one for each cost type used in separate regression models fitted to explain transcript and protein levels using the following explanatory variables: average per residue protein carbon, nitrogen, sulphur content, average per residue protein biosynthetic cost, average per codon tRNA gene number and transcript CAI. The contribution of each variable to explaining transcript and protein data was then estimated by removing the variable from the regression and then estimating the size of the effect on explanatory power measured using Akaike's Information Criterion (AIC).

Figure 3. Comparison of amino acid substitution rate with biosynthetic cost.

Each point is the mean substitution rate for one of the twenty standard amino acids. Substitution rates were estimated from alignments of Saccharomyces genes by Wall et al. . Each amino acid substitution rate was normalised by tree length and then averaged across all alignment columns corresponding to the amino acid at that site in the ancestral protein sequence. Alignment columns containing gaps were excluded. The standard error of the mean for each amino acid substitution rate is shown as a bar in each point. Robust linear regression and 95% confidence intervals are used to indicated trend. Each plot indicates the Spearman's rank correlation between amino acid substitution rate and biosynthetic cost.