One sequence plus one mutation equals two folds.

I n this issue of PNAS, Alexander et al. (1) report an observation that will be a topic of heated discussion among biochemists for a long time to come. A single amino acid substitution can, in the right context, completely change the fold of a protein. And the structural change produced by this one mutation cannot be dismissed as a semantic issue over what constitutes a different fold. As shown in Fig. 1, one conformation consists of a three-helix bundle, whereas the alternate form has a four-strand -sheet with a single -helix. Eighty-five percent of residues change their secondary structure, with only eight residues in the central -helix plus one or two turn residues retaining the same conformation in both forms. When I first heard this result in a public seminar, my mind literally began to reel, leaving me dizzy and slightly nauseated as all hope of understanding how sequence encodes structure seemed to suddenly vanish. This paper represents the dramatic culmination of a project begun more than 5 years ago (2) involving the systematic modification of two small protein domains taken from a large extracellular protein found in streptococci known as protein G. The fragment referred to as GA is a serum-binding domain, whereas the GB fragment is an IgG-binding domain. Both proteins are soluble, monomeric, and between 45 and 56 aa in length. And both have been extensively studied as model systems by the protein folding community for more than 15 years. Alexander et al. (2) set out to identify the mutational pathway that would convert one protein fold into the other with the fewest unstructured intermediates. The result was a series of sequence pairs, each encoding two stable functional folds, with sequence similarity increasing from essentially zero to differences in amino acid sequence at 20 positions (2), then at 11, 5, 3, and then finally at only 1 position (1). A tyrosine at position 45 specifies the four-strand/onehelix structure, whereas a leucine at this position yields the three-helix bundle. As described in their paper, many sequence variants were screened by phage display. In a stroke of good fortune, the key functional residues for both proteins could be preserved, allowing the authors to score for the GA conformation by albumin binding and the GB conformation by IgG binding. At each stage in this methodical convergence to a minimal number of sequence differences, the stabilities of the mutant proteins to thermal unfolding were quantified and the ‘‘nativeness’’ of several variants ascertained by analysis of the 15N–1H correlation spectrum, perhaps the best experimental method for detecting molten globule states. In addition, full NMR structures were determined on select variants to remove any doubt about the structural states of sequence pairs. Because we now live in a world where one amino acid sequence plus one mutation actually can give rise to two very different protein folds, what are the implications? How can we make sense of this observation by putting it into the context of our current understanding of the physical chemistry used by protein sequences to encode structure? And what are its implications for the evolution of protein structure over biological time? Should we expect more examples of dramatic protein fold switching, or is there something exceptional about these two protein domains that diminishes the generality of conclusions we might wish to draw? First of all, it comes as no surprise that the sequence of a protein can be extensively modified with little or no appreciable effect on structure. Using selection schemes to recover stably folded proteins from mutagenized libraries, it is often possible to substitute 50% or more of residues without changing the fold or greatly lowering the stability (3). Even the kinetics of folding may be little changed from the wild-type values. As one might expect, in such cases most mutant sites are located on the protein surface, where the strength and specificity of interactions are often low. The expected result on going to levels of substitution much higher than 50% would have been destabilization of the folded states, causing much of the sequence pathway connecting the two conformations to be fully unfolded/denatured. But this is not what happened. As pointed out by Alexander et al. (1), there are precedents in the literature for proteins undergoing changes in their ‘‘fold’’ as result of a few changes in amino acid sequence or cleavage of a