The structure of fitness landscapes — the mapping of DNA or protein sequences to organismic fitness (or a proxy) — is a central problem in evolutionary biology. Even in this day and age of unprecedented molecular data and computational power, we are still asking rather basic questions: Are fitness landscapes mostly smooth, with one or just a few local maxima? Are they highly rugged, with lots of suboptimal local maxima? Or do they have large flat regions separated by sharp thresholds? Here are schematic illustrations of these different scenarios:
Large flat regions (sets of genotypes with nearly indistinguishable fitness) as shown in panel C are traditionally known as “neutral networks,” since genotype space has a complex network structure unlike the 2D Euclidean space shown in the figure. Neutral network models have long been part of the evolutionary theorist’s repertoire (some examples: van Nimwegen et al. 1999, Bornberg-Baur and Chan 1999, Wilke 2001, Draghi et al. 2010), but so far I would say experimental data on them has been very limited, if for no other reason than the fact that experimental data on fitness landscapes in general has been hard to come by.
That is why I thought this recent paper was so exciting:
Pervasive degeneracy and epistasis in a protein-protein interface Anna I. Podgornaia and Michael T. Laub http://dx.doi.org/10.1126/science.1257360
In this paper the authors consider four residues key to the binding interaction between PhoQ and PhoP in E. coli. (Binding of PhoQ controls phosphorylation of PhoP, which is an important step in a signaling pathway.) They systematically constructed all 204 = 160000 possible amino acid sequences for these positions and determine that 1659 maintain wild-type function. These 1659 sequences therefore form a neutral network in the whole sequence space. The neutral network appears to have very nontrivial structure, as shown in the following diagram:
In particular, there is widespread epistasis: certain mutations are tolerated on some genetic backgrounds but not on others. The diversity of amino acid combinations that are and are not functional is pretty striking. This epistasis leads to nontrivial evolutionary pathways between accessible sequences, since the most direct mutational pathways sometimes pass through nonfunctional sequences and are therefore inaccessible. I am definitely looking forward to seeing both how this neutral network compares with theoretical models, as well as further experimental studies of real neutral networks in simple systems.