The toxic aggregates of Aβ are no longer thought to be amyloid fibrils, the components of amyloid plaques, but smaller soluble oligomeric aggregates of Aβ, or larger, but still soluble, protofibrils, which contain smaller aggregates as building blocks. A major problem in research on Aβ aggregation is that there appears to be a multitude of soluble oligomeric species of which some might be toxic and others perhaps not. These are difficult to isolate from Aβ aggregation mixtures made in vitro. In particular, they are difficult to keep stable from further aggregation into (less toxic) amyloid fibrils, which are the endpoint of Aβ aggregation and which assemble into amyloid plaques. Hence, soluble Aβ aggregates are difficult to characterize in detail. As a result, several groups try different tricks to stabilize the correct (most toxic) aggregate to enable detailed studies of its properties (see, e.g., [1,2]).

A major contester for the role as building block of the penultimate toxic aggregate is an Aβ "dimer" which has been identified in CSF samples from AD patients (3). In their...  Read more

The toxic aggregates of Aβ are no longer thought to be amyloid fibrils, the components of amyloid plaques, but smaller soluble oligomeric aggregates of Aβ, or larger, but still soluble, protofibrils, which contain smaller aggregates as building blocks. A major problem in research on Aβ aggregation is that there appears to be a multitude of soluble oligomeric species of which some might be toxic and others perhaps not. These are difficult to isolate from Aβ aggregation mixtures made in vitro. In particular, they are difficult to keep stable from further aggregation into (less toxic) amyloid fibrils, which are the endpoint of Aβ aggregation and which assemble into amyloid plaques. Hence, soluble Aβ aggregates are difficult to characterize in detail. As a result, several groups try different tricks to stabilize the correct (most toxic) aggregate to enable detailed studies of its properties (see, e.g., [1,2]).

A major contester for the role as building block of the penultimate toxic aggregate is an Aβ "dimer" which has been identified in CSF samples from AD patients (3). In their elegant studies, Shankar et al. (3) also produced a synthetic dimer in which substitution of serine 26 in the Aβ peptide with cysteine 26 (the S26C mutant) results in a covalent disulfide bridge linking two monomeric units of Aβ40. This dimer was shown to be considerably more potent than wild-type Aβ in inhibiting long-term potentiation of memory (LTP) assays in mouse hippocampus. These results placed the S26C dimer as a candidate for a model of the neurotoxic dimer found in AD patients.

In the present article, O’Nuallain et al. in Dominic Walsh’s lab characterize the S26C dimer further. They find that it rapidly aggregates into large protofibril-like structures. These protofibrils are “synaptotoxic,” since they inhibit LTP, whereas the isolated S26C dimer or (reduced) S26C monomer are not. The protofibrils formed by S26C are morphologically similar to those formed by wild-type Aβ, and they appear to contain a significant amount of β-sheet secondary structure. Furthermore, the protofibrils formed by S26C are relatively stable towards further aggregation to amyloid fibrils. The authors conclude that dimer formation enables aggregation into stable toxic Aβ protofibrils.

Recently, Yamagushi et al. (4) employed another type of cysteine-linked dimer to stabilize protofibrils: They made C-terminal extensions of Aβ to link monomers into dimers at their C-termini. They argued that the dimer stabilized protofibrils by increasing the kinetic barrier for fibril formation. In our own work, we stabilized the protofibrillar state by using a disulfide bridge (double cysteine mutation) to lock the Aβ monomer into a hairpin conformation (the AβCC mutant). We showed that such Aβ hairpins readily form toxic oligomers and/or protofibrils, but not amyloid fibrils (2).

The present paper and those by Yamaguchi et al. (4) and Sandberg et al. (2) are all encouraging in similar ways and for a number of reasons. First, they all constitute new methods to “engineer” protofibrils while avoiding fibril formation. Second, they reveal similar and presumably significant features of the Aβ protofibrils: their overall dimensions and a large fraction of secondary β-sheet structure content. Third, they hint at very specific structural features of Aβ units in the protofibrils. For instance, the present work shows that linkage of residues 26 in two monomeric units is compatible with the protofibril structure, and Sandberg et al. show that Aβ in neurotoxic oligomers/protofibrils adopt a β-hairpin conformation. Fourth, the present study and that of Sandberg et al. both confirm the toxicity of Aβ protofibrils.

However, in my opinion, the structure and role of the Aβ dimer found in AD patients remains obscure. This is because basic physical chemistry predicts that the rate of protofibril formation must be enhanced every time monomer units are linked into dimers in a way that is compatible with the arrangement of Aβ monomer units within the protofibrils. This could be the explanation for why both the present S26E as well as C-terminally linked monomers result in rapid protofibril formation. Similar arguments may explain why the protofibrils formed by both are stable: the disulfides provide kinetic barriers to fibril formation.

It is, therefore, not certain that the present S26C or the C-terminally linked Aβ dimer actually mimics the dimer found in AD patients. The S26C dimer may, of course, still be a significant building block, and it would be important to find out what it looks like. However, given what is found in the present study by O’Nuallain et al., and my arguments above, I think that the major focus in research on structure of toxic Aβ aggregates now shifts from dimers to the protofibril.

References:
1. Ono, K., M.M. Condron, and D.B. Teplow, Structure-neurotoxicity relationships of amyloid beta-protein oligomers. Proc. Natl. Acad. Sci. USA, 2009. 106: p. 14745-14750. Abstract

2. Sandberg, A., et al., Stabilization of neurotoxic Alzheimer amyloid-beta oligomers by protein engineering. Proc. Natl. Acad. Sci. USA, 2010. 107: p. 15595-15600. Abstract

3. Shankar, G.M., et al., Amyloid-beta protein dimers isolated directly from Alzheimer's brains impair synaptic plasticity and memory. Nat. Med., 2008. 14(8): p. 837-42. Abstract

4. Yamaguchi, T., et al., A disulfide-linked amyloid-beta peptide dimer forms a protofibril-like oligomer through a distinct pathway from amyloid fibril formation. Biochemistry, 2010. 49: p. 7100-7107. Abstract

View all comments by Torleif Hard

Considerable evidence has accumulated to suggest that water-soluble oligomeric forms of Aβ are the principal causes of the neurotoxicity in AD. Studying the structures and the formation pathways of these oligomers is thus of great interest. Since the oligomers and the intermediates along the aggregation pathways have finite lifetimes, there is an interest in designing mutants that can form more stable mimics of the wild-type intermediates.

The Aβ amyloid peptide follows different oligomerization pathways, depending on the length of the peptide. Aβ42 tends to forms fibrils faster and larger oligomers than Aβ40 (Bitan et al., 2003). Also, different types of Aβ peptides (Arctic [E22G], Dutch [E22Q], and Flemish [A21G]) exhibit different rates (usually higher than wild-type in several cases) of secondary structure formation. In fact, they are more stable than the wild-type peptide (see, e.g., Urbanc et al., 2010 and Fawzi et al., 2007). These...  Read more

Considerable evidence has accumulated to suggest that water-soluble oligomeric forms of Aβ are the principal causes of the neurotoxicity in AD. Studying the structures and the formation pathways of these oligomers is thus of great interest. Since the oligomers and the intermediates along the aggregation pathways have finite lifetimes, there is an interest in designing mutants that can form more stable mimics of the wild-type intermediates.

The Aβ amyloid peptide follows different oligomerization pathways, depending on the length of the peptide. Aβ42 tends to forms fibrils faster and larger oligomers than Aβ40 (Bitan et al., 2003). Also, different types of Aβ peptides (Arctic [E22G], Dutch [E22Q], and Flemish [A21G]) exhibit different rates (usually higher than wild-type in several cases) of secondary structure formation. In fact, they are more stable than the wild-type peptide (see, e.g., Urbanc et al., 2010 and Fawzi et al., 2007). These differences are already observed at the early stage of oligomerization, i.e., at the monomer level. The interest in studying more complex structures such as protofibrils and oligomers had increased rapidly since they became associated with neurotoxicity.

The works of Walsh and Hard present two complementary strategies to address this issue. Hard reported an Aβ peptide variant with two mutations (A21C and A30C) for both Aβ40 and Aβ42 monomers. This creates an intramolecular disulphide bridge that enhances a β-hairpin conformation (accessible in monomer wild-type) in both cases. By engineering a mutant Aβ peptide variant that has a β-hairpin structure and can associate into protofibrils that are morphologically identical to wild-type protofibrils but much more stable and toxic, Hard and colleagues proposed that the toxic protofibrils have β-sheet structures and suggested to use the mutant protofibrils as model systems to study the structure and formation kinetics of the wild-type protofibrils.

Walsh, on the other hand, made synthetic stable dimers that can form protofibrils but not fibrils. This, according to him, supports the idea that Aβ dimers may stabilize the formation of fibril intermediates by a process distinct from that available to Aβ monomers, and that higher molecular weight prefibrillar assemblies are the proximate mediators of Aβ toxicity. These studies suggest that these Aβ dimers form protofibrils more rapidly than Aβ monomers, and the authors linked the presence of these dimers with the abundance of protofibrils.

These reports seem to indicate that the aggregation mechanism may involve some cooperative effects. It is an open issue how this depends on aggregate size, how this affects oligomer formation, and how to characterize their physical and chemical properties. The meta-stable mutant intermediates proposed in both cases of Hard and Walsh can be a good start for future studies. The current studies do not directly monitor the structures and pathways. Coherent multidimensional non-linear spectroscopies, which are optical analogues of NMR (Mukamel, 2000), can reveal most valuable additional information about the structure and kinetics of misfolded peptides. This can be done by either looking at vibrations in the infrared or at electronic excitations in the ultraviolet, making use of the remarkable sensitivity of these spectra to couplings among residues.

These applications were demonstrated in our recent papers (Jiang et al., 2010; Zhuang et al., 2010; see also “Two Dimensional Ultraviolet (2DUV) Spectroscopic Tools for Identifying Fibrillation Propensity of Protein Residue Sequences,” J. Jiang and S. Mukamel, Angewandte Chemie Int. Ed. 2010, in press). Experimental two-dimensional infrared studies of fibrils have been reported by the groups of Hochstrasser (Kim et al., 2009) and Zanni (Strasfield et al., 2009).

View all comments by Shaul Mukamel

I remain agnostic as to what studies of in vitro toxicity tell us about pathogenic mechanisms in Alzheimer's disease. I think the data is unequivocal that dumping protein aggregates on neurons in culture is harmful, but it has been challenging to prove that this drives neurodegeneration in AD.

With that disclaimer, I have always been a fan of the kinetic model of toxicity. This model was first proposed based on experimental data in the manuscript by Wogulis et al. published by Russ Rydel and colleagues in the Journal of Neuroscience in 2005. This manuscript demonstrated in a quite convincing fashion that Aβ fibril growth was closely linked to and necessary for in vitro toxicity. The current manuscript by O’Nuallain and colleagues extend these data to dimeric assemblies and effects on LTP, but they do not cite this earlier work.

Indeed, the Wogulis manuscript is rarely cited in much of the current literature on Aβ toxicity, but in my mind, the Wogulis study is one of the more insightful papers on this subject. I simply suggest that anyone who studies toxicity of protein...  Read more

I remain agnostic as to what studies of in vitro toxicity tell us about pathogenic mechanisms in Alzheimer's disease. I think the data is unequivocal that dumping protein aggregates on neurons in culture is harmful, but it has been challenging to prove that this drives neurodegeneration in AD.

With that disclaimer, I have always been a fan of the kinetic model of toxicity. This model was first proposed based on experimental data in the manuscript by Wogulis et al. published by Russ Rydel and colleagues in the Journal of Neuroscience in 2005. This manuscript demonstrated in a quite convincing fashion that Aβ fibril growth was closely linked to and necessary for in vitro toxicity. The current manuscript by O’Nuallain and colleagues extend these data to dimeric assemblies and effects on LTP, but they do not cite this earlier work.

Indeed, the Wogulis manuscript is rarely cited in much of the current literature on Aβ toxicity, but in my mind, the Wogulis study is one of the more insightful papers on this subject. I simply suggest that anyone who studies toxicity of protein aggregates, and who has not read this, should. It also points out that using MTT as an indicator of amyloid toxicity is highly problematic, as MTT interaction with amyloid results in needle-like formazan crystals that are toxic. This probably explains the initial reports that claimed neurotoxicity (induction of neuronal death) by low nanomolar levels of oligomers. Oligomers do appear to be slightly more potent with respect to inducing death, but still require low micromolar concentration to induce death. Again, if they induce fast growth into protofibrils, then this would be consistent with such kinetic models of toxicity.

A kinetic model for toxicity is interesting to consider with respect to inducing neurodegeneration in vivo, as it could help to explain two observations from human studies: the imperfect correlation between plaques and cognitive status, and higher levels of soluble Aβ being better associated with cognitive status. Of course, the imperfect correlation between plaques and cognition has many alternative explanations, but if one solely considers a kinetic model where toxicity is related to growth, then slow growth of plaques over many years might result in a lot of plaques but less neurodegeneration as opposed to rapid growth of fewer plaques that could result in more neurodegeneration. Also, one might think that higher levels of soluble Aβ (the vast majority of which is not oligomeric), which has been reported to correlate better with cognitive status, would drive faster growth once seeding has taken place, leading to more damage.

Finally, I think that we should not ignore plaques. They may be less toxic than oligomers in certain systems, but they may also promote damage in more insidious ways by activating the innate immune system, or through a more subtle type of toxicity. AD is not an acute degenerative disease; it is a slow, progressive neurodegeneration. Acute toxicity through depression of LTP or apoptotic death in culture may not be a good surrogate for what happens in vivo.

References:
Wogulis M, Wright S, Cunningham D, Chilcote T, Powell K, Rydel RE. Nucleation-dependent polymerization is an essential component of amyloid-mediated neuronal cell death. J Neurosci 2005;25:1071-80. Abstract

View all comments by Todd E. Golde