Crystal structure of human prion protein fragment reveals a motif for oligomer formation

Abstract

The structural transition of the prion protein from α-helical to β-sheet rich underlies its conversion into infectious and disease-associated isoforms. Here we describe the crystal structure of a fragment from human prion protein consisting of the disulfide bond linked portions of helices 2 and 3. Instead of forming a pair-of-sheets steric zipper structure characteristic of amyloid fibers, this fragment crystallized into an β-sheet rich assembly of hexameric oligomers. This study reveals a never before observed structural motif for ordered protein aggregates, and suggests a possible mechanism for self-propagation of misfolded conformations by such non-amyloid oligomers.

Amyloid fibers and non-amyloid oligomers have been shown to be associated with a number of ‘conformational diseases’ including Alzheimer’s, Parkinson’s, and prion. Current evidence suggests that, at least in some cases, non-amyloid oligomers rather then amyloids may be the primary toxic species.¹ Prion diseases are particularly intriguing because aggregates of the prion protein (PrP) can be both toxic and infectious, with emerging evidence suggesting that these two traits are carried out by different types of aggregated species.² Furthermore, there are indications that small oligomers of PrP may be more infectious then larger fibrillar aggregates.³

Elucidating high-resolution structures of polypeptides in amyloid or oligomer conformations presents a formidable challenge for current structural biology techniques. Recently, an approach which reduces the complexity of aggregation prone regions of proteins by dissecting them into smaller parts has illustrated that investigations into their atomic structures can be amenable to X-ray crystallography. While structural studies with relatively short segments would be of limited value to learn about the overall structure and function of globular proteins, structures of segments from the amyloidogenic regions of proteins have shed important new light on the molecular mechanisms of aggregation into amyloid and oligomers. Using this approach, atomic insight into the assembly of amyloid has been gained with numerous structures.^4–11 These have defined a common ‘steric zipper’ motif characterized by a pair of β-sheets that extend indefinitely along the length of a fiber and a self-complementary interface between them composed of tightly interacting sidechains.^4,5 Although the consensus properties of non-amyloid oligomers inferred from biophysical studies include a multimeric assembly typically described as β-sheet rich,¹² until recently their structural organization has been elusive. The atomic-resolution structure of ‘cylindrin’, formed from a segment of αB-crystallin, has provided an example of one possible oligomeric architecture.¹³

In the present study we sought to gain a structural insight into the role that the conserved intramolecular disulfide bond in PrP may play during its conversion into misfolded aggregates. In the normal form of the prion protein, PrP^C, the disulfide bond bridges two long alpha helices, α2 and α3 (Figure S1a). However, during the formation of both amyloid fibrils^14,15 and oligomers from recombinant PrP in vitro,¹⁶ these helices have been shown to undergo a vast structural rearrangement into β-sheet structure. Recent evidence suggests a similar conformational change in this region of the protein upon conversion to the infectious PrP^Sc isoform,¹⁷ although alternative structural models have also been proposed.^18,19 Here we describe the 1.4 Å crystal structure (Table S1) of a fragment, referred to as DBPrP, from portions of α2 and α3 of human PrP corresponding to two discontinuous segments ¹⁷⁷HDCVNI¹⁸² and ²¹¹EQMCIT²¹⁶ that are covalently linked by a disulfide bond between Cys179 and Cys214 (Figure S1). Instead of forming a pair of sheets steric zipper reflective of amyloid, the DBPrP fragment assembled into distinct β-sheet oligomers.

The crystal structure of DBPrP reveals a structural motif never observed in any previously crystallized amyloidgenic fragments. Six DBPrP fragments were found in the asymmetric unit of the crystals, each one making up a subunit of a hexameric oligomer. The overall structural organization of this hexamer is that of three four-stranded, anti-parallel β-sheets arranged like the faces of a triangular prism (Figure 1a). The subunits form similar contacts between each other, and their assembly can be described as a trimer of dimers (D₃ point group symmetry), in which each four-stranded β-sheet is formed by an association of two subunits along a twofold axis, and the hexamer results from an association of dimeric β-sheets along a threefold axis (Figure 1).

Figure 1.

Structure of hexameric oligomer formed by DBPrP fragments reveals a symmetric, tightly packed, prismatic assembly of β-sheets. (a) The hexamer is shown in two orientations, looking along the side of the prismatic assembly on the left, and down its quasisymmetric threefold axis represented as a triangle on the right. Dashed grey lines are intended to draw attention to the three four-stranded β-sheets that define the hexamer. The main chain is represented as cartoon β-strands and side-chains as sticks. Each subunit is a different color, starting with chain A in orange with chains B through F arranged clockwise. This color scheme is consistent with panels c, d and e. (b) A side view of the oligomer with one segment removed reveals the tight complementary packing of hydrophobic sidechains as pink spheres at the interior and exclusion of hydrophilic sidechains as green spheres at its exterior. (c) A schematic representation of one dimeric β-sheets illustrates the arrangement of hydrogen bonds (dashed lines) between strands. Hydrogen bonds drawn in light blue are not observed in all six subunits within the hexamer due to conformational flexibility of the strand termini. Side chains which protrude towards the exterior of the hexamer are depicted as larger circles. Smaller pink circles highlight hydrophobic residues that form a hydrophobic cluster along one of its interior faces. A black oval represents a twofold symmetry within the association of two subunits. (d) A molecular stick representation with a transparent overlay of cartoon β-strands shows the same dimeric β-sheet diagramed in panel c. (e) A similar molecular representation of the interaction between edge strands of two adjacent β-sheets shows the Asp178 sidechain twisting away from the hydrophobic interior of the hexamer. In doing so it hydrogen bonds to the peptide backbone, and prevents the association of edge strands around the circumference of the hexamer.

The assembly of DBPrP fragments into the prismatic hexamer is driven by the maximization of hydrogen bonding and burial of hydrophobic sidechains (Figure 1b). Figure 1c and Figure 1d show the consensus pattern of hydrogen bonding between strands in each of the antiparallel sheets, illustrating that their arrangement allows for pairing of nearly all intra- and intermolecular backbone hydrogen bond donors and acceptors except those present on the outer edge strands. In this arrangement of strands, hydrophobic residues (Val180, Ile182, Met213, and Ile215) appear as a cluster on the interior face of the dimeric β-sheets (Figure 1c). The threefold assembly of these sheets along these clusters forms a dry interface, burying approximately 40% of the total surface area of each subunit. Furthermore, the shape complementarity (S_c) of the buried surfaces indicates tight interactions in the self-complementary association of the hydrophobic sidechains comparable to those found at the dry interface of steric zippers (Table S2). Hydrophilic residues face the outside of hexamer, making extensive interactions with solvent molecules (Figure S2). A closed topology of hydrogen bonding within the hexamer is prevented by residue Asp178 which faces the hydrophobic interior of the hexamer, but twists away in order to avoid burying its charged headgroup. In doing so, it makes hydrogen bonding interactions to the peptide backbone of edge strands at the interface of the sheets, and obstructs the formation of a continuous β-sheet around the hexamer (Figure 1e). As a result, the hexamer possesses open edge β-strands.

The hexameric structure of the DBPrP fragment is also found in solution. In size exclusion chromatography experiments performed under physiological pH the DBPrP fragment elutes as one distinct peak with a retention volume corresponding to a size larger than expected for a monomer of 1.4 kDa (Figure S3a). The elution profile is independent of fragment concentration, indicating that, at least at the concentration range tested (0.5–4 mg/ml), this multimeric species is not in equilibrium with the monomer. The peak has a similar elution volume to that of the aprotinin standard. Even though the molecular mass of aprotinin is ~6.5 kDa (i.e., smaller than that expected for the DBPrP hexamer), the radius of gyration calculated from the crystal structure of the latter protein (11.2 Å from pdb id 4pti) is nearly equal to that calculated for the crystal structure of DBPrP hexamer (10.8 Å). Thus, similar elution volumes of aprotinin and DBPrP suggest that, as in the crystal, also in solution DBPrP exists as a hexamer.. This is further corroborated by native electrospray mass spectrometry experiments that show only a single stable oligomer with a mass of 8.5 kDa (Figure S3b–c) consistent with a hexamer. Together, these experiments indicate that the DBPrP fragment forms a monodisperse population of hexamers in solution with the same compact shape as is observed in the crystal structure.

Although the hexameric assembly of the DBPrP fragments is the effective unit of crystal growth, a notable contact occurs between hexamers through their open edge strands. Within each DBPrP fragment, as Asp178 twists away from the hydrophobic core of the hexamer, it perturbs the N-terminal portion the edge β-strands it is located on, allowing the exposed edge strands to make hydrogen bonding contacts with identical strands on adjacent hexamers (Figure 2a). This interaction creates a small antiparallel β-sheet stabilized by intermolecular backbone hydrogen bonds between residues His177 and Cys179 (Figure 2b and Figure 2c). Although any of the six edge strands in the hexamer are capable of making such interactions, only two strands do so, linking hexamers together in a head-to-tail fashion. This head-to-tail assembly leads to a filament-like morphology throughout the crystal which, unlike in amyloid fibrils, does not have the alignment of β-strands perpendicular to the axis of elongation.

Figure 2.

An interaction between hexamers is mediated through backbone hydrogen bonding. (a) A cartoon representation of symmetry related hexamers in different colors illustrates their head-to-tail association along the c-axis of the unit cell (grey box). (b) A detailed molecular view of one of these interactions shows that they are mediated through hydrogen bonding of the peptide backbone between two edge strands of adjacent hexamers, forming a short anti-parallel β-sheet. (c) A schematic of this interaction shows the pattern of backbone hydrogen bonds between His177 and Cys179 from the two involved edge strands. The diagram is cut-off where interactions are the same as that described for individual hexamers (Fig. 1).

The DBPrP hexamer presented here represents a never before observed structural motif for non-amyloid oligomers. As previously postulated,²⁰ the organization of β-sheets into such oligomers may allow for many more degrees of freedom in terms of sheet-to-sheet packing arrangements compared to fibrillar (amyloid) aggregates of which there are eight classes.⁵ The recently described ‘cylindrin’ structures of toxic oligomers formed by an eleven residue segment of αB crystallin have revealed one such possible arrangement with characteristics of the β-barrel fold.¹³ The DBPrP hexamer shows yet another possible type of structural organization for β-sheet oligomers. This latter structure shows a striking similarity to the structural motif found in a handful of globular proteins in which a prism-like association of three four-stranded β-sheets has been classified as a β-prism I fold.²¹ The cylindrin and DBPrP hexamer represent two possible structural classes for β-sheet oligomers identified to date. These distinct structural arrangements may contribute to the great polymorphism described for these types of protein aggregates²².

Although structurally distinct, the oligomeric assemblies of cylindrins and DBPrP share some structural commonalities. Both are symmetric hexamers, the cylindrin with 6 strands and the DBPrP hexamer with 12 strands per oligomer. Furthermore, both have features in common with steric zippers, including a self-complementary and tight association of β-sheets at a dry interface. This suggests that aggregation through self-complementary interactions may be an innate property of polypeptides in both amyloid and non-amyloid oligomers. However, a major difference between these two structural motifs is that cylindrin has a closed β-sheet topology where the first strand hydrogen bonds to the last one, whereas the DBPrP hexamer has exposed edge strands. These exposed edge strands promote the association of hexamers into larger assemblies, the implications of which are described below.

Globular proteins have evolved molecular features to prevent aggregation through the exposed edges of β-sheets.²³ In aberrant misfolded forms of proteins where evolution has not intervened, such as those represented by steric zippers and the DBPrP oligomer described herein, exposed β-sheet edges provide a nucleation element for further association. The assembly of DBPrP hexamers illustrates how the specific hydrogen bonding interactions between exposed β-strand edges facilitates the formation of large molecular weight aggregates composed of identical oligomeric building blocks. This type of interaction could potentially explain the great heterogeneity of sizes and various morphologies reported for non-amyloid oligomers,²² including that of infectious PrP^Sc.³ Interestingly, studies on a partially disaggregated form of infectious PrP^Sc have shown oligomers with a three-fold symmetry¹⁸ similar to that found in the DBPrP hexamer.

The crystal structure of the DBPrP hexamer shows one possible conformation of this fragment of PrP in a β-sheet rich conformation and provides insight into how it can scaffold the assembly of the protein into a similar structure. To illustrate the latter point, we created a model which shows how the entire C-terminal amyloidogenic region of the prion protein^14,24 can be accommodated into a three-fold symmetric hexameric oligomer scaffolded on the DBPrP fragment structure (Figure S4). Since there is evidence that oligomers share common structural features²⁵ it can be imagined that other amyloidogenic proteins where oligomer formation has been observed may also adopt similar conformations.

The structural organization of the DBPrP fragment described here allows us to speculate on a general mechanism by which misfolded protein conformations could self-propagate within the context of non-amyloid oligomers. This issue is of particular importance to the understanding of prion diseases. The infectious PrP^Sc isoform in prion disease replicates by binding to the cellular PrP^C isoform and templating its conversion to the PrP^Sc structure in a process that is often modeled by an analogy to seeded polymerization of amyloid fibrils.^26,27 This templating takes place at the open β-strands at the ends of amyloid fibers, forming very stable non-native interactions between misfolded proteins. Models based on this analogy are, however, complicated by the uncertainty regarding the molecular nature of infectious prion particles: while some strains of mammalian prions have characteristics of amyloid, in others these characteristics are not detected.²⁸ Furthermore, it has been shown that highest infectivity per mass unit is associated with prion particles substantially smaller than long fibrils.³ These observations, together with very limited infectivity of amyloid fibrils formed from the recombinant PrP in vitro,²⁹ implicate non-amyloid oligomers in the process of infectivity and underscore the need for alternative (i.e., non-amyloidbased) models for the self-propagation of mammalian prions. The DBPrP structure presented here shows how oligomers, just like amyloid, can posses open β-strands that can template the growth of larger assemblies of misfolded protein. Hence, this novel structural motif for non-amyloid aggregation provides a paradigm for an alternative basis for the self-propagation of misfolded proteins through such non-amyloid aggregates.