First, consider the shape issue. Writing in Nature, David Eisenberg and colleagues at the University of California, Los Angeles, and at Caltech, Pasadena, report that amyloid fibrils can form from proteins that retain substantial native conformation. In a set of elegant experiments, they showed that fibrils made from a modified ribonuclease (RNase) A still had nuclease activity. But the results are more profound than that. First author Shilpa Sambashivan and colleagues demonstrate that the reason these growing amyloid-like fibrils are so strong is because of a particularly powerful protein adhesive called domain swapping.

Domain swapping, where domain A of one polypeptide insinuates itself into domain B of another, and vice versa, yields a very tight protein-protein bond that gives certain protein dimers their extreme strength. This is well known for viral capsids, for example. Three-dimensional domain swapping, where each protein in a fibril shares one domain with the next in line, has previously been proposed by Eisenberg and others, including Andrew Miranker at Yale University, to account for the self-specificity of amyloid fibers.

However, the problem with domain swapping as an explanation for fiber formation was that it is energetically unfavorable. In an accompanying Nature News & Views, Miranker writes “There must be something about a swap that makes it more favorable to be joined up than to be free and single.” In the present study, that something turns out to be a hinge region.

The hinge in RNase A is a four-amino-acid loop connecting the swapped domain with the core domain of the protein. It is known that RNase A can form domain-swapped dimers under certain conditions. Researchers have speculated that if the hinge loop were expanded with an amyloidogenic segment, then the domain-swapping dimer might be turned into a domain-swapping polymer. This is exactly what Sambashivan and colleagues have achieved. By expanding the loop with the addition of ten glutamine residues, the RNase A forms aggregates that bind Congo red, a dye that binds to amyloids such as those formed from the amyloid-β (Aβ) found in AD.

Of course, just because these modified RNase A proteins form amyloid fibrils doesn’t mean that they do so by three-dimensional domain-swapping. An alternative explanation would be that the polyglutamine stretches stack together in a β-sheet conformation. This can happen when other proteins are interrupted by longer glutamine expansions, such as in Huntington’s disease, and it would leave the RNase part of the molecule untouched. Not so, Sambashivan and colleagues write. They report incontrovertible proof that domain-swapping is going on in these fibrils.

The evidence comes from the catalytic center. It depends on the presence of two histidine amino acids, at positions 12 and 119, supplied by different domains, one being the proposed swapped domain. Mutating either of these histidines to alanine causes the native molecule to lose its activity. When the authors mixed alanine 12 mutants with alanine 119 mutants and allowed them to form fibrils, lo and behold, activity returned. The most plausible explanation for this result is that domain-swapping reconstitutes the correct active site by juxtaposing two histidines. (A bit of Mendelian genetics here: Domain-swapping should result in four different types of “active” sites, H12:H119, H12:A119, A12:H119, and A12:A119, with only the first one being active. This means the fibrils should have a maximum activity about 25 percent that of normal. In fact, the activity of the fibrils was about half that. Given that the relatively large fibril must present itself in the correct orientation to its RNA substrate, the poorer performance could be due to stearic hindrance.)

The take-home message from these results is that “for the RNase A amyloid-like fibril there is no need to invoke a distinctly different, generic structural form of protein, other than for a small segment that forms the β-sheet spine,” as the authors write. Whether this holds true for other amyloids is sure to be questioned. Protofibrils of Aβ, for example, do not bind well to thioflavin T, which binds β amyloids similarly to Congo red. This suggests that for Aβ, at least, intermediate steps in fibril formation do exist (see ARF related news story). As Miranker sums up in his News & Views: “There is considerably more mystery to be solved in the formation of amyloid fibers—for example, the structural nature of transient intermediates and the basis of their cytotoxicity.”

Toxicity was the topic addressed by Byron Caughey and colleagues at the National Institute of Allergy and Infectious Diseases in Hamilton, Montana. First author Jay Silveira and coworkers focused on the issue of size. Over the last few years the field has come to realize that bigger does not necessarily mean more toxicity when one is dealing with fibrils. In the case of Aβ, for example, there is now substantial evidence that protofibrils are more toxic than the large aggregates (see ARF related news story and ARF news story), which may even be relatively protective (see ARF related news story). Now Silveira and colleagues report in last week’s Nature that the same is true for prion protein toxicity.

Prion protein, responsible for spongiform encephalopathies in sheep and cows and Creutzfeldt-Jakob disease in humans, comes in normal and infectious forms. The infectious variety can convert normal forms into toxic ones (see below), and toxic forms gradually take over in the brain. Toxic prions form long aggregates much like amyloid fibrils, and prion researchers have begun asking, just like AD researchers have for years, whether size matters when it comes to toxicity.

Silveira and colleagues isolated prion amyloids, disaggregated them, separated them by size, and then measured the toxicity and converting activity of the various fractions. They found a tight correlation between specific infectivity and specific prion converting activity. Both peaked in the 300-600 kDa range, which equates to prion particles of 14-28 monomers. In this study, particles of fewer than five monomers were not infective.

So what is the nature of the most infectious prion? Previous work has suggested that prion conversion takes place by the simple addition of monomers to a growing chain (see ARF related news story). This would be incompatible with the present study, which uses transmission electron micrographs to suggest that the most infectious particles are, in fact, spherical or ellipsoidal in nature.

However, a subtle distinction is important here, the authors point out. On a per-particle basis, the authors found that the difference in infectivity between infectious particles of different sizes does not exceed twofold. This means that while the smaller polymers with 14-28 peptides may have by far the highest specific activity on a weight basis, on a molar basis they are not much more infectious than the largest particles. This distinction should not be overlooked. As the authors point out in a supplementary discussion, which is available online: “As long as infectious particles are above a minimum size, particle concentration is a key parameter in determining scrapie infectivity titer.” The operative term here is “particle.”

The finding reinforces doubts that have previously been raised about using aggregate busting as a potential therapy (see ARF related news story). In prion diseases, such a strategy might unintentionally increase infectivity, the authors caution.

Last week also brought an advance on another prion question that has puzzled researchers for some time. It is whether the conversion from normal to toxic prion can occur with existing proteins or whether it happens only as protein is being synthesized. In mammalian brain, for example, the first sign of conversion occurs in the Golgi apparatus, suggesting that it somehow accompanies synthesis (see ARF related news story). Contradicting this implication, Prasanna Satpute-Krishnan and Tricia Serio of Brown University in Providence, Rhode Island, reported in Nature last week that, in yeast, conversion of the prototypical prion protein Sup35 is immediate and occurs in fully synthesized protein.

The authors made their discovery by using a Sup35 chimera. They coupled Sup35 to the green fluorescent protein (GFP) and made it removable by targeted proteolysis. Sup35-GFP is resistant to conversion by the toxic version of the prion, Sup35[PSI+]. The reasoning behind this experiment was that if conversion cannot occur in fully synthesized protein, then Sup35 released by proteolysis from Sup35-GFP should be stable even in the presence of toxic Sup35[PSI+]. The authors found the opposite. Sup35[PSI+] readily converted fully synthesized Sup35 released by proteolysis. This resulted in an immediate phenotypic switch because Sup35 is a translational suppressor whose presence or absence can easily be assayed.

The authors note the parallels between prion conversion and formation of protein aggregates associated with AD, PD, HD, and other neurodegenerative diseases. “The remodeling of existing proteins from their normal configurations to either a pathogenic intermediate or a potentially protective amyloid may have profound effects on the appearance and progression of these diseases,” they write. They add that therapy development should proceed with an understanding of the in vivo protein species that can undergo such physical state transitions.—Tom Fagan.

References:
Satpute-Krishnan P, Serio TR. Prion protein remodeling confers an immediate phenotypic switch. Nature. September 2005;437:262-265. Abstract

Sambashivan S, Liu Y, Sawaya MR, Gingery M, Eisenberg D. Amyloid-like fibrils of ribonuclease A with three-dimensional domain-swapped and native-like structure. Nature. 2005 Sep 8;437(7056):266-9. Abstract

Miranker A. Structural biology: fibres hinge on swapped domains. Nature. 2005 Sep 8;437(7056):197-8. No abstract available. Abstract

Silveira JR, Raymond GJ, Hughson AD, Race RE, Sim VL, Hayes ST, Caughey B. The most infectious prion protein particles. Nature. September 2005;437:257-261. Abstract