Introduction
Comparative studies of the folding of homologous domains have proved to be valuable in understanding protein folding.1,2 Two principal questions are asked in such studies: the first concerns the pathway; which elements of structure fold early and which fold late? The second is a question of mechanism; it has been suggested that there is a continuum of mechanisms,3 from strictly framework, diffusion collision mechanisms characterised by polarised transition states,4,5 to pure nucleation condensation mechanisms, characterised by diffuse transition states with strictly concomitant formation of secondary and tertiary structure.6 Although in a few protein families, the folding pathway does not appear to be conserved (e.g. protein G and protein L7) in many protein families the transition states of homologous proteins are remarkably similar; even where the transition state of one protein is much more structured, the general structural features of the transition state are maintained. This has been observed, for instance, in the immunity proteins Im7 and Im9, which have 60% sequence identity,8 and in the immunoglobulin-like (Ig-like) proteins including the Ig domain from human titin, TI I27, and a fibronectin type III domain from human tenascin, TNfn3, which have no significant sequence identity, although they share a common fold.9–11 Folding mechanisms within a fold can vary significantly between homologous proteins, even where the transition states have similarly structured regions. In the homeodomain superfamily, for instance, folding mechanisms range from a strict framework/diffusion collision mechanism (engrailed homeodomain) to pure nucleation condensation (hTRF1).3 In this case, the mechanism is apparently determined by secondary structure (helical) propensity. Thus, studies of the folding of related proteins suggest that both topology and the nature of the secondary structure content have a role in determining how a protein folds.12
We have studied the folding of two protein families extensively. The first is the complex, all-beta Greek key Ig-like fold.9–11,13–15 Topology is the dominant factor influencing the folding of Ig-like domains. All these proteins fold via a nucleation condensation process, where key long-range interactions, between residues in the central BCEF strands, form early to nucleate folding. The transition state of these proteins is an expanded form of the native state and local interactions are not involved in determining the folding pathway.
In contrast to the complex Greek key topology of the Ig-like domains, spectrin domains are simple three-helix coiled-coil structures with up-down-up connectivity. Three spectrin domains have been studied in our laboratory (R15, R16 and R17 from chicken brain α-spectrin).16–18 In all three of these proteins, two elements of secondary structure (helices A and C) form and interact early and the third (helix B) forms and packs after the rate-determining transition state. There is some suggestion of a change in mechanism, as is seen in the homeodomain family; R16 folds by a framework-like mechanism, with secondary structure (especially in helix C) preceding tertiary structure formation,17 whereas R15 folds by a nucleation condensation mechanism with secondary and tertiary structure forming concomitantly (our unpublished results). R17 shows a mechanism similar to that of R16, but with higher Φ-values and more helical structure in helix A.18
The principal aim of this study was to compare the folding of a death domain (DD) from human FADD (FADD DD), an all-helical protein with a complex Greek key topology, with the two other classes of protein studied in depth in this laboratory: spectrin domains, simple all-alpha three-helix bundles, and Ig-like domains, all-beta proteins with complex Greek key topology.19,20 The death domains provide an opportunity to study the interplay between formation of secondary structure and topology in protein folding. Death domains have six anti-parallel alpha helices, arranged in a Greek key topology, with helices 1, 5 and 6 grouped in an approximately orthogonal position above helices 2, 3 and 4 (Fig. 1).21 Human FADD comprises two domains from the DD superfamily, each approximately 100 amino acids; the C-terminal DD is the one studied here.
In the Ig-like domains, two elements of structure, twisted beta sheets, pack together to form the hydrophobic core. Central to this structure is a four-strand motif consisting of two pairs of anti-parallel beta strands, one from each sheet (B-E and C-F, Fig. 1). Similarly, death domains can be thought of as two three-helix bundles, packed together via a central four-helix motif: bundle 1 (B1) is made from the packing of H1, H5 and H6, and forms the B1 core. Bundle 2 is made of contiguous helices H2, H3 and H4, which pack to form the B2 core. The central four-helix motif comprises two pairs of parallel helices (H1-H5 and H2-H4) packed together orthogonally, to form the central core. As is shown in Fig. 1, the three cores are clearly separated, so that different faces of H1, H2, H4 and H5 contribute to different cores. The peripheral helices H3 and H6 contribute only to the packing of the bundle cores (B2 and B1, respectively). In contrast, the Ig-like domains have a single hydrophobic core formed by packing of the twisted beta sheets. Thus, FADD DD can be thought of as two spectrin-like, three-helix bundles packed together to form a Greek key structure with central elements forming the central core in a manner that is reminiscent of Ig-like domains.