Is protein folding hierarchic ? I . Local structure and peptide folding

ALTHOUGH SYNTHETIC ANALOGS of proteins might be developed ultimately, Flory’s succinct statement accurately summarizes the remarkable behavior of globular proteins. Today, an intense search is being made for the principles that guide the folding process. Although the work began with experimental studies (see Box 1), it has become increasingly clear that these results cannot be interpreted successfully without a satisfactory theory and simulations based on an adequate physical model. Accordingly, both experimentalists and theorists are now racing to learn what constitutes an adequate physical model. In pursuit of this goal, we examine a curious paradox. The folding kinetics of small proteins reveal the existence of two classes of molecule (which we call class I and class II) that appear to fold by quite different mechanisms. Class I proteins, typified by a-lactalbumin (a LA), apomyoglobin (apoMb), RNase H, barnase and cytochrome c (cyt c), fold by a hierarchic process in which native-like secondary structure forms rapidly and is stabilized in molten-globule intermediates. Class II proteins, typified by chymotrypsin inhibitor 2 (CI2)2 and cold-shock protein B (Csp B)3, fold rapidly, in a kinetically two-state manner that lacks detectable intermediates. Some workers believe that these conflicting results imply that two different folding mechanisms exist: (1) a hierarchic mechanism that involves proteins with populated intermediates; and (2) a tertiary nucleation mechanism in which intermediates are not detectable. We turn to this question in Part II of this article (which will appear in the February issue of TiBS ) by considering the structures and properties of observable intermediates in class I proteins, and data that describe the transition states of some class II proteins. Throughout this review, we are asking a fundamental question: is protein folding hierarchic? We define hierarchic folding as a process in which folding begins with structures that are local in sequence and marginal in stability; these local structures interact to produce intermediates of ever-increasing complexity and grow, ultimately, into the native conformation (see Box 2). Non-hierarchic folding is a process in which tertiary interactions not only stabilize local structures but actually determine them. It follows that protein secondary structure is determined largely by local sequence information if folding is hierarchic, but not if folding is non-hierarchic. We use this difference to distinguish between the two models. Hierarchic folding is an attractive model because it is both conceptually simple and computationally tractable. We consider several approaches for testing and evaluating hierarchic folding. For the model to be plausible, secondary structures (helices, turns and individual strands of sheet) must have at least borderline stability in peptides in the absence of tertiary interactions. Moreover, the local interactions inferred from inspection of helices, turns and strands in proteins of known structure should be reproducible in peptides, in which they can be measured quantitatively. The stop signals responsible for terminating protein helices should be found in the residue sequences that bracket the helix, not in tertiary interactions, and these local termination signals should operate in suitable peptide helices as well. We cover these topics in Part I of this article, which represents the straightforward, almost classical, part of this article. The results strongly support a hierarchic mechanism, but the evidence extends only to the initial stages of folding. For further evidence, we turn next to folding intermediates and transition states, the topics reviewed in Part II of this article. The interpretation of such work is far more controversial but, in our opinion, it is here that remaining conflicts between alternative folding models will be resolved. We attempt to show that present evidence can be reconciled with hierarchic folding and to argue that intermediates in class I folding reactions strongly support this view. A novel approach to the problem is provided by the LINUS program4, which can perform simple folding simulations in which structure is allowed to develop solely on the basis of local interactions. Unlike our definition, the term hierarchic folding is sometimes used to describe a unique, sequential pathway that progresses to the native state in successive steps through a strict series of everlarger, native-structure intermediates. By contrast, we suppose that alternative pathways of self-assembly are viable – as in the diffusion–collision model of Karplus and Weaver5, which is illustrated by the folding kinetics of the l repressor6. None of the five helices of the l repressor is entirely stable alone, but random collisions between helices are mutually stabilizing and produce higher-order intermediates. Many possible folding routes exist, and the energy landscape (i.e. the differing stabilities of the helices and their combinatorial intermediates) determines the dominant populations.