Two Is a Pair, Three Is a Network

One of the intriguing questions in molecular evolution is how new protein folds emerge. The creation of new structures by mutations of an existing sequence is a significant challenge. The article in this issue of the Biophysical Journal by Porter et al. (1) adds a new twist to our view of this process.

A large change in the fold of a protein (a structural flip) as a result of a point mutation of a protein sequence is an extremely rare event. Evidence that the process is indeed rare is the success of homology modeling in which structures of proteins are modeled according to templates with as low as 30% sequence identity. Hence significant variation in sequence does not imply a significant change in structure. Furthermore, between the options that the same fold will be retained upon a change of one amino acid, or that the protein will unfold or lose its function, the remaining chance of a structural flip is very small. Several computational models illustrated and estimated this probability (2,3).

Given the rarity of the switching sequences, the growing empirical evidence for such changes is truly remarkable (4–7). These experiments indicate that it is possible to find pairs of distinct protein folds that are related by the smallest evolutionary change, a change of one residue. This is supporting evidence to a model in which proteins structures evolved from one or a few basic folds and as time passed the protein flipped between the alternative structures that we see today. However, astonishing as these observations are, links between pairs of structures are insufficient to support full connectivity in the space of protein folds.

It is convenient to place the space of all folds on a graph in which every protein structure is a node on the graph. Mutation events that flip between protein structures (change the fold or the node of the sequence) are represented by graph edges. If it is possible to reach any other fold in the graph by point mutations, starting from any node on the graph, we say that the graph is connected. If, however, some nodes cannot be reached by point mutations from some other nodes we say that the graph is disconnected or disjoint. Connected graphs of folds were discussed in Meyerguz et al. (2) and Cao and Elber (3).

It is possible that the pairs of folds discovered experimentally are disjoint from other pairs and do not allow for extensive exploration of fold space by point mutations. Hence, to argue that a single or a few protein sequences could generate all folds requires a connected graph in which all protein shapes are accessible from a few core structures. It is a challenge to find even a few pairs of sequences that flip between folds (or edges in a graph of folds).

Is it possible to find more than one edge to a fold in order to establish a network of protein structures connected by point mutations?

The new article by Porter et al. (1) answers the above question in the affirmative. Building on their clever design of a protein flip between a three-helix bundle and an α/β protein (8) they added another edge to the small network originating at the three-helix bundle that is pointing now also to a β-sheet domain. Sequences of different folds were designed to be as similar as 80% sequence identity while retaining their original folds, in contrast to the widely accepted rule-of-thumb of homology. The new domains are supported by interdomain interactions. These interactions make the notion of foldable isolated protein domains more complex and require the existence of additional stabilizing contacts to support switches. It is an important new finding that opens the way for proteins to interact and exchange their folds on a larger multistep scale, forming a network. Computational models have suggested a number of edges per fold that is larger by an order of magnitude (3) compared to the number found here. Further experimental exploration of protein switches and the investigations of the feasibility of the network of sequence flow are highly desirable. This research may have significant impact on the way we think about evolution of protein structures, and the results reported in Porter et al. (1) are extremely encouraging.