Phenotypic response of yeast metabolic network to availability of proteinogenic amino acids

Abstract

Genome-scale metabolism can best be described as a highly interconnected network of biochemical reactions and metabolites. The flow of metabolites, i.e., flux, throughout these networks can be predicted and analyzed using approaches such as flux balance analysis (FBA). By knowing the network topology and employing only a few simple assumptions, FBA can efficiently predict metabolic functions at the genome scale as well as microbial phenotypes. The network topology is represented in the form of genome-scale metabolic models (GEMs), which provide a direct mapping between network structure and function via the enzyme-coding genes and corresponding metabolic capacity. Recently, the role of protein limitations in shaping metabolic phenotypes have been extensively studied following the reconstruction of enzyme-constrained GEMs. This framework has been shown to significantly improve the accuracy of predicting microbial phenotypes, and it has demonstrated that a global limitation in protein availability can prompt the ubiquitous metabolic strategy of overflow metabolism. Being one of the most abundant and differentially expressed proteome sectors, metabolic proteins constitute a major cellular demand on proteinogenic amino acids. However, little is known about the impact and sensitivity of amino acid availability with regards to genome-scale metabolism. Here, we explore these aspects by extending on the enzyme-constrained GEM framework by also accounting for the usage of amino acids in expressing the metabolic proteome. Including amino acids in an enzyme-constrained GEM of Saccharomyces cerevisiae, we demonstrate that the expanded model is capable of accurately reproducing experimental amino acid levels. We further show that the metabolic proteome exerts variable demands on amino acid supplies in a condition-dependent manner, suggesting that S. cerevisiae must have evolved to efficiently fine-tune the synthesis of amino acids for expressing its metabolic proteins in response to changes in the external environment. Finally, our results demonstrate how the metabolic network of S. cerevisiae is robust towards perturbations of individual amino acids, while simultaneously being highly sensitive when the relative amino acid availability is set to mimic a priori distributions of both yeast and non-yeast origins.

Keywords: cellular resource allocation; enzyme-constrained model; flux balance analysis; genome-scale metabolic model; metabolic networks.

FIGURE 1

Correlation between in vivo and in silico amino acid mass fractions (g gDW⁻¹) during exponential growth on a defined, minimal glucose medium. Experimental mass fractions were estimated from the GECKO-implemented proteins only. Error bars denote the simulated variability of amino acid usage at 99% of the optimal objective value.

FIGURE 2

(A) Simulated mass fraction (g gDW⁻¹) of amino acids in the acidFBA-GEM at varying growth rates. For each simulation, the growth rate was constrained and the resulting flux phenotype was predicted my minimizing the overall enzyme usage. Shaded areas denote the region of protein-limited growth. (B) Rank-ordered, absolute relative amino acid mass-fraction deviations of the acidFBA-GEM of a fully fermentative versus a fully respiratory metabolism.

FIGURE 3

Distributions of the relative deviations from the mean of proteinogenic amino acids across N = 5, 000 randomly sampled nutrient conditions. Each condition was defined by a unique combination of viable nutrient sources belonging to each of the four elemental classes: carbon, nitrogen, phosphorus, and sulphur.

FIGURE 4

Violin plots of the distributions of mean-normalized flux ranges across N = 5, 000 sampled nutrient combinations for the 20 proteinogenic amino acids of the acidFBA model. The feasible flux ranges were simulated by performing a flux variability analysis (FVA) using an optimality threshold of 99%. Dotted line in red denote the selected deviation from growth optimality.

FIGURE 5

Absolute relative difference of each amino acid between the species-specific amino acid distributions and the simulated amino acid profiles of the acidFBA-GEM at increasing relative growth rates. The simulated profiles were obtained by minimizing the Euclidean distance to the species-specific distributions. Sce: S. cerevisiae, Eco: E. coli, Bsu: B. subtilis, Ppu: P. putida.