Amino acid coevolution induces an evolutionary Stokes shift

Abstract

The process of amino acid replacement in proteins is contextdependent, with substitution rates influenced by local structure, functional role, and amino acids at other locations. Predicting how these differences affect replacement processes is difficult. To make such inference easier, it is often assumed that the acceptabilities of different amino acids at a position are constant. However, evolutionary interactions among residue positions will tend to invalidate this assumption. Here, we use simulations of purple acid phosphatase evolution to show that amino acid propensities at a position undergo predictable change after an amino acid replacement at that position. After a replacement, the new amino acid and similar amino acids tend to become gradually more acceptable over time at that position. In other words, proteins tend to equilibrate to the presence of an amino acid at a position through replacements at other positions. Such a shift is reminiscent of the spectroscopy effect known as the Stokes shift, where molecules receiving a quantum of energy and moving to a higher electronic state will adjust to the new state and emit a smaller quantum of energy whenever they shift back down to the original ground state. Predictions of changes in stability in real proteins show that mutation reversals become less favorable over time, and thus, broadly support our results. The observation of an evolutionary Stokes shift has profound implications for the study of protein evolution and the modeling of evolutionary processes.A major focus of modern evolutionary studies is to understand how structural and functional contexts determine the patterns of evolutionary change at different positions in a biological macromolecule. Such an understanding is important to phylogenetics, partially because the position-specific processes of evolution are known to determine our ability to reconstruct deep nodes in the tree of life (1) but also because features of the evolutionary process such as convergence can deterministically mislead phylogenetic reconstruction (2). Understanding patterns of evolution and how they respond to details of structure and function can also potentially help us to better decode the evolutionary record, allowing us to distinguish between structural and functional constraints and identify the signatures of positive selection. This finding can lead to improved understanding of a biomolecule's structure, dynamics, thermodynamics, functionality, and physiological context. Key questions concern how evolutionary processes vary among sites and over time and particularly, how the evolution at different locations influences each other. For example, coevolution between different locations in a protein can slow the amino acid replacement process, allowing phylogenetic inference at deep nodes that would otherwise have been swamped by recurrent neutral changes. Furthermore, coevolution tends to depend on proximity in the 3D structure of proteins, leading to the hope that, if properly understood, it could improve our ability...

Biophysics of protein evolution and evolutionary protein biophysics

“…DDG prediction is widely used to screen large numbers of mutations, often in combination with laboratory experiments [53][54][55][56][57]. The approach has also served as fitness estimators in simulation studies of protein evolution [58,59].…”

“…Rather it should depend on the preceding mutations at other sites. This phenomenon is illustrated by a recent simulation study of the evolution of purple acid phosphatase by generating random mutations in the structure [58]. Based on stability calculations in the model, the propensities for all amino acids at selected solvent-exposed and buried sites were determined.…”

“…The results indicated considerable variations of these propensities because stability effects of amino acid substitutions at a given site change with time as the evolutionary process progresses. The simulation showed that an enforced destabilizing mutation (which could arise in real proteins owing to functional constraints such as the need to preserve an active site) can be compensated by subsequent mutations, thus increasing the future viability of the already-mutated residue at that same site and rendering the reverse mutation detrimental and therefore less probable [58]. Although this model probably overestimated native stability as well as stability effects of mutations and underestimated the probability for misfolding because of its simplified treatment of the unfolded state (see discussion above on marginal stability; §2.4.2), it offers an excellent elucidation of the 'holistic' nature of intra-protein interactions and the biophysical forces that govern the mutational effect on stability and how it may depend on the temporal order in which the mutations occur [58].…”

Extensively Parameterized Mutation–Selection Models Reliably Capture Site-Specific Selective Constraint

“…A growing body of research has demonstrated that these constraints lead individual protein sites to have distinct tolerances to different amino acids (Porto et al 2004;Ramsey et al 2011;Pollack et al 2012;Ashenberg et al 2013;Bloom 2014aBloom , 2014bRisso et al 2014;Abriata et al 2015;Doud et al 2015;Echave et al 2016). Recent experimental studies have further demonstrated that, for at least several proteins, site-wise amino-acid preferences are broadly conserved over evolutionary time (Ashenberg et al 2013;Risso et al 2014;Doud et al 2015).…”

“…All rights reserved. For permissions, please e-mail: journals.permissions@oup.com time, e.g., due to epistatic interactions (Weinreich et al 2006;Pollack et al 2012;Ashenberg et al 2013;Draghi and Plotkin 2013;Gong et al 2013;McCandlish and Stoltzfus 2014;Shah et al 2015). As such, our use of the word fitness throughout this study should be interpreted primarily as a mutationselection model parameter indicating conserved site-specific properties.…”