The search for the toxic species responsible for the neurodegeneration observed in Alzheimer disease has become this field’s Holy Grail. And much like that mythical search, the search for the toxic species has been full of false leads, dead ends, and even a couple of conspiracy theories. One thing that most of the field will agree on is that Aβ aggregation is a central element to the generation of the toxic species, with most of the recent focus being on the formation of smaller oligomeric forms. However, due to limitations of many methods, studying aggregating proteins and peptides has proved to be an inexact science. For this reason the work by Bernstein et al. using mass spectrometry coupled with ion mobility to characterize the early aggregation pathway of both Aβ40 and 42 is a technical tour de force. The approach is very elegant. It elucidates many of the intermediates on the aggregation pathway and clearly shows that Aβ40 and 42 behave differently. The major difference is that Aβ42 forms a meta-stable dodecamer structure, a species that has previously been identified as a...  Read more

The search for the toxic species responsible for the neurodegeneration observed in Alzheimer disease has become this field’s Holy Grail. And much like that mythical search, the search for the toxic species has been full of false leads, dead ends, and even a couple of conspiracy theories. One thing that most of the field will agree on is that Aβ aggregation is a central element to the generation of the toxic species, with most of the recent focus being on the formation of smaller oligomeric forms. However, due to limitations of many methods, studying aggregating proteins and peptides has proved to be an inexact science. For this reason the work by Bernstein et al. using mass spectrometry coupled with ion mobility to characterize the early aggregation pathway of both Aβ40 and 42 is a technical tour de force. The approach is very elegant. It elucidates many of the intermediates on the aggregation pathway and clearly shows that Aβ40 and 42 behave differently. The major difference is that Aβ42 forms a meta-stable dodecamer structure, a species that has previously been identified as a candidate for the toxic species (Lesne et al., 2006).

Is this dodecamer the toxic species? We don’t know. While the dodecamer may be the most stable species in a “test tube,” we also know that Aβ is capable of interacting with a range of biological substrates, all of which could stabilize other structural forms of Aβ. These binding partners include lipid membranes (e.g., Kayed et al., 2004; Tickler et al., 2005; Hung et al., 2008; Martins et al., 2008), protein receptors (e.g., Deshpande et al., 2009) and metal ions (Bush, 2003). All of these factors have been implicated in the generation of the toxic species.

The existence of the meta-stable dodecamer structure raises an interesting conundrum with respect to the many efforts that are currently ongoing to isolate the toxic Aβ species from various tissues. Function (toxic or otherwise) is related to structure; Aβ is a pleiomorphic molecule and its structure depends on its context—change the context and you change the structure. The findings by Bernstein et al. suggest that if you have isolated Aβ42, then the default structure that this peptide will most likely adopt is the dodecamer. Whether this is the structure the peptide originally had in the brain is an unanswered question.

The work by Harmeier et al. reminds us that there is more to Aβ toxicity than peptide aggregation. We have previously shown that mutations to the GxxxG motif can alter the rate of Aβ aggregation affecting the formation of smaller oligomers of Aβ and their interactions with lipid membranes (Hung et al., 2008). Harmeier et al. have extended this work into an in-vivo fly model and show that G33 is a key residue in regulating Aβ aggregation and toxicity. Substitutions that increase the hydrophobicity at this residue increase aggregation and in the process decrease the toxicity.

Similarly, we have previously published that a single substitution to Y10 of Aβ inhibited toxicity, even though this peptide formed dimers, trimers, and other oligomeric forms including a 56 KDa species (Barnham et al., 2004). These data show that Aβ aggregation is not the only requirement for toxicity. Other factors do indeed play a role, and it is likely to be a quite specific conformation of Aβ oligomer that is toxic. The requirement for a specific conformation for Aβ toxicity would be consistent with the concept that toxicity is mediated through a specific receptor interaction, e.g., such as with the NMDA receptor (Deshpande et al., 2009).

The work by Bernstein et al. and Harmeier et al. sheds more light on the search for the Holy Grail, but the quest will continue.

References:
Barnham KJ, Haeffner F, Ciccotosto GD, Curtain CC, Tew D, Mavros C, Beyreuther K, Carrington D, Masters CL, Cherny RA, Cappai R, Bush AI. Tyrosine gated electron transfer is key to the toxic mechanism of Alzheimer's disease beta-amyloid. FASEB J. 2004 Sep;18(12):1427-9. Abstract

Bush AI. The metallobiology of Alzheimer's disease. Trends Neurosci. 2003 Apr;26(4):207-14. Abstract

Deshpande A, Kawai H, Metherate R, Glabe CG, Busciglio J. A role for synaptic zinc in activity-dependent Abeta oligomer formation and accumulation at excitatory synapses. J Neurosci. 2009 Apr 1;29(13):4004-15. Abstract

Hung LW, Ciccotosto GD, Giannakis E, Tew DJ, Perez K, Masters CL, Cappai R, Wade JD, Barnham KJ. Amyloid-beta peptide (Abeta) neurotoxicity is modulated by the rate of peptide aggregation: Abeta dimers and trimers correlate with neurotoxicity. J Neurosci. 2008 Nov 12;28(46):11950-8. Abstract

Kayed R, Sokolov Y, Edmonds B, McIntire TM, Milton SC, Hall JE, Glabe CG. Permeabilization of lipid bilayers is a common conformation-dependent activity of soluble amyloid oligomers in protein misfolding diseases. J Biol Chem. 2004 Nov 5;279(45):46363-6. Abstract

Lesné S, Koh MT, Kotilinek L, Kayed R, Glabe CG, Yang A, Gallagher M, Ashe KH. A specific amyloid-beta protein assembly in the brain impairs memory. Nature. 2006 Mar 16;440(7082):352-7. Abstract

Martins IC, Kuperstein I, Wilkinson H, Maes E, Vanbrabant M, Jonckheere W, Van Gelder P, Hartmann D, D'Hooge R, De Strooper B, Schymkowitz J, Rousseau F. Lipids revert inert Abeta amyloid fibrils to neurotoxic protofibrils that affect learning in mice. EMBO J. 2008 Jan 9;27(1):224-33. Abstract

Tickler AK, Smith DG, Ciccotosto GD, Tew DJ, Curtain CC, Carrington D, Masters CL, Bush AI, Cherny RA, Cappai R, Wade JD, Barnham KJ. Methylation of the imidazole side chains of the Alzheimer disease amyloid-beta peptide results in abolition of superoxide dismutase-like structures and inhibition of neurotoxicity. J Biol Chem. 2005 Apr 8;280(14):13355-63. Abstract

View all comments by Kevin Barnham

We readily agree with some of the data and interpretations given in the interesting paper of Bernstein et al. Moreover, the study shows ESI-MS to be a useful method to analyze non-covalently linked oligomers in the gas phase. In my respectful opinion, some parts of the paper seem to focus too much on the Aβ*56 (12-mer).

There is no doubt that the dimer and tetramer of Aβ42 are important. Quite a while ago, we were able to show why, since engineered dimers have a twofold increased β-sheet content (Schmechel et al., 2003). This was the first report to show that covalently linked dimers of Aβ can serve as a nidus to start fibril growth and that homodimers of Aβ are a risk factor for the formation of higher oligomers.

Our data published last week in the Journal of Neuroscience (Harmeier et al., 2009) show that toxicity also requires a specific conformation of Aβ42 variants. The G33I substitution and mutant in Drosophila shows that it might be important to compare the conformations of Aβ42 toxic and non-toxic variants to...  Read more

We readily agree with some of the data and interpretations given in the interesting paper of Bernstein et al. Moreover, the study shows ESI-MS to be a useful method to analyze non-covalently linked oligomers in the gas phase. In my respectful opinion, some parts of the paper seem to focus too much on the Aβ*56 (12-mer).

There is no doubt that the dimer and tetramer of Aβ42 are important. Quite a while ago, we were able to show why, since engineered dimers have a twofold increased β-sheet content (Schmechel et al., 2003). This was the first report to show that covalently linked dimers of Aβ can serve as a nidus to start fibril growth and that homodimers of Aβ are a risk factor for the formation of higher oligomers.

Our data published last week in the Journal of Neuroscience (Harmeier et al., 2009) show that toxicity also requires a specific conformation of Aβ42 variants. The G33I substitution and mutant in Drosophila shows that it might be important to compare the conformations of Aβ42 toxic and non-toxic variants to learn which receptors might mediate the toxicity. This view is supported by Bernstein and colleagues, who show that tetramers of Aβ42 matter but not of Aβ40. In this regard, the S26C-linked dimer published last year by Tom Kukar and colleagues shows that engineered Aβ40 dimers can inhibit LTP and might adopt a toxic conformation, although it should be added that data of this paper are being intensely discussed in the field (Kukar et al., 2008).

We agree that intrinsic factors matter, independent of the length of the peptide (i.e., 42 or 40 residues) and that the specific conformation is important. One also has to be mindful that LTP and cell toxicity are relevant assays, but their link to the AD pathology is still missing; in other words, are LTP inhibition or cell toxicity really the cause of degrading synapses and neuronal loss leading to memory loss in the end?

In our view, another important point is that a therapeutic approach that aims to arrest aggregation at the stage of tetramers might be dangerous since tetramer and dimers were described as the most toxic species. It is possible that if a therapeutic approach attempts to arrest oligomer growth, the level of (toxic) forms might be unwittingly increased from the reservoir of non-toxic oligomers which might be converted into toxic forms by changing the conformation, e.g., by binding to a specific receptor? Thus, one has to make sure that a conversion of non-toxic into toxic oligomers will not happen.

Fibril growth might not be a problem as long as those fibrils (non-toxic per se) are not a source of dimers or tetramers that are brought back into solution. It may sound crazy to some, but I believe it might be better to quickly deposit Aβ oligomers as fibrils that are then not harmful any longer, or at least less harmful. Finally, it could turn out that the longer dimers and tetramers are available over time, i.e., the longer their half-life, the higher the likelihood that damage accrues.

View all comments by Gerd Multhaup

Elusive oligomerization-mediated amyloid-β-protein toxicity: Where have all the trimers gone?
Evidence that the pathogenesis of Alzheimer disease (AD) is strongly associated with occurrence of oligomeric assemblies of Aβ is strongly challenging researchers who wish to identify suitable therapeutic strategies for prevention and cure of this debilitating illness. It is well known that of the two dominant alloforms of Aβ—Aβ40, and Aβ42—AD is more correlated with the latter, longer alloform. How does a small difference in the primary structure so critically affect the pathology? Within this past week, two inspiring papers addressed Aβ40 and Aβ42 oligomer formation and toxicity from unique angles.

Application of ion-mobility mass spectroscopy to resolve the oligomer sizes of Aβ40 and Aβ42 performed in Michael Bowers’ group, in collaboration with experimental labs of Gal Bitan and Dave Teplow and the computational group of Joan-Emma Shea, yielded both expected and unexpected results (Bernstein et al., 2009). Expected was the fact that Aβ40 and Aβ42 oligomerized through...  Read more

Elusive oligomerization-mediated amyloid-β-protein toxicity: Where have all the trimers gone?
Evidence that the pathogenesis of Alzheimer disease (AD) is strongly associated with occurrence of oligomeric assemblies of Aβ is strongly challenging researchers who wish to identify suitable therapeutic strategies for prevention and cure of this debilitating illness. It is well known that of the two dominant alloforms of Aβ—Aβ40, and Aβ42—AD is more correlated with the latter, longer alloform. How does a small difference in the primary structure so critically affect the pathology? Within this past week, two inspiring papers addressed Aβ40 and Aβ42 oligomer formation and toxicity from unique angles.

Application of ion-mobility mass spectroscopy to resolve the oligomer sizes of Aβ40 and Aβ42 performed in Michael Bowers’ group, in collaboration with experimental labs of Gal Bitan and Dave Teplow and the computational group of Joan-Emma Shea, yielded both expected and unexpected results (Bernstein et al., 2009). Expected was the fact that Aβ40 and Aβ42 oligomerized through distinct pathways, as has been experimentally demonstrated in 2003 by using photo-induced cross-linking of unmodified proteins (PICUP) (3). The present study by Bernstein and collaborators provides an independent confirmation of oligomerization differences between Aβ40 and Aβ42 and concurs that their differences are not induced by a particular chemistry of PICUP (5).

Unexpected was the lack of trimers of both Aβ40 and Aβ42. Trimers represented a significant oligomeric species of Aβ40 oligomer population in the original study by Bitan and collaborators (3). Bernstein and collaborators found trimers to be present in oligomeric states of the two Aβ40 mutants, one with F19P substitution and the other with oxidized M35, demonstrating that the lack of trimers was not due to limitations of the experimental setup.

Of all peptides under study, only Aβ42 formed oligomers larger than tetramers and was thus uniquely characterized by hexamers and dodecamers.

Experimental data by Gerd Multhaup's group (Harmeier et al., 2009) show that a single-point mutation at G33 but not G29 can offset the oligomerization pathway of Aβ42. Both single-point mutations of Aβ42 under investigation, G33A and G33I, resulted in an enhanced oligomerization characterized by larger oligomers as compared to Aβ42 wild-type. Surprisingly, despite enhanced oligomerization, the resulting mutant oligomers were less toxic than Aβ42 wild-type oligomers! In the old days, the amyloid hypothesis of AD was associating neuronal loss and toxicity with amyloid plaques until a paradigm shift from amyloid fibrils to oligomers occurred. The present work by Harmeier and collaborators now demonstrates that toxicity is not necessarily correlated with the oligomerization propensity either (Harmeier et al., 2009), suggesting that specificity of the oligomer structure plays a crucial role in mediating toxicity.

Both studies identified as the low molecular mass Aβ42 oligomeric species (in addition to monomers) also dimers, tetramers, and hexamers. The common feature between the two studies was also the fact that Aβ42 oligomer size distribution was multi-modal in contrast to the Aβ40 oligomer size distribution, in agreement with the prior work by Bitan et al. (3) , with additional peak at dodecamers (Bernstein et al., 2009) or 16- to 20-mers (Harmeier et al., 2009). No trimers were detected in either of the two studies. The two studies differ in their interpretation of Aβ42 toxicity. Bernstein et al. associated Aβ42 toxicity with occurrence of Aβ42 dodecamers with a proposed annular structure resembling ion channels, which if inserted into the membrane bilayer would cause ion leakage and thereby induce cell death. Harmeier et al. identified the key amino acid responsible for Aβ42 toxicity, G33, which if substituted by alanine or isoleucine would result in decreased toxicity of Aβ42(G33A) or Aβ42(G33I) assemblies despite increased oligomerization.

The accompanying commentary on the work by Bernstein et al. was given by Clemmer and Valentine, who discussed whether the specific experimental set-up used in ion-mobility mass spectroscopy, involving peptides in solution to be sprayed into droplets, followed by solvent evaporation resulting in peptide assemblies immersed in a gas, alters the structure of resulting Aβ oligomers or not. If the answer was yes, then the resulting conformers might be less relevant to the etiology of AD, unless gaseous phase could be considered a model of a water-free membrane environment. If the answer is no, then one can assume that the resulting conformers had structure characteristic of an aqueous solution. The paper by Bernstein and collaborators discusses their experimental results using a modeling approach where individual spherical monomer conformers of Aβ42 form a planar hexamer ring, resulting in a close conformation that would not allow for a further addition of monomers and would, upon two-ring hexamer merge, form a potentially toxic dodecamer (Bernstein et al., 2009).

Considering the possible Aβ oligomer structure in an aqueous environment from the viewpoint of a computational biophysicist whose work revolves around Aβ structural predictions (2), I see an alternative explanation of the observed structural differences between Aβ40 and Aβ42 oligomers and resulting oligomer-mediated toxicity. The sequence of either Aβ can be viewed as ~1/3 hydrophilic in the N-terminal region and ~2/3 hydrophobic in the central and the C-terminal regions. In water, the hydrophobic regions will be adjusted to achieve the least contact with water. Because the majority of the peptide is hydrophobic, a folded monomer will have many hydrophobic residues exposed to solvent. However, as oligomers form, the central and C-terminal regions will be more efficiently screened from water molecules by the hydrophilic N-terminal regions. The simulations which captured the oligomer size distribution differences as reported by previous work of Gal Bitan and Dave Teplow (3) revealed globular structures of all oligomers with a most intriguing structural difference between Aβ40 and Aβ42 oligomers at the N-terminal region (2). Aβ40 oligomers were characterized by a β-strand structure involving A2-F4 region, rendering more ordered and restricted N-termini, compared to Aβ42 oligomers with random-coil-like and spatially more extended structure at the N-termini (2), suggesting an increased cross-section for Aβ42 oligomers as compared to Aβ40 oligomers and thereby providing an alternative model for explaining the differences in the cross-sections reported by Bernstein and collaborators (Bernstein et al., 2009).

A 5 percent difference in the primary structure at the C-terminus (Aβ40 versus Aβ42) results in an alteration of oligomer pathway as well as structural differences in the N-terminal region. Moreover, structural differences between alloforms appear already at the stage of folding. Our computational study exploring folded structures of Aβ(1-40), Aβ(1-42), and their Arctic mutants demonstrated a lack of the β-strand structure at the N-terminal regions of both Arctic mutants, rendering the folded structure of the Arctic mutants proximate to the more toxic Aβ(1-42) (1). If sequence differences at the C-terminus (Aβ40 and Aβ42) or a single-point mutation in the central hydrophobic region effectively changes the structure at the N-terminal region, then generalizations of any experimental or theoretical results from one alloform to another should be questioned and applied with great caution. Considering such long-range structural effects of a single-point mutation, the conclusion of Harmeier et al. that the amino acid G33 is the key amino acid that mediates toxicity should be slightly rephrased. It is possible that the substitution of G33 by alanine or isoleucine changes not only the local but also the global structure of the resulting oligomers and thus the key structural change that directly impacts the toxicity might be remote from the position G33.

Substantial experimental and computational evidence on Aβ folding and oligomer formation demonstrates that we are dealing with a protein very sensitive to subtle changes in preparation and environment required for application of a specific experimental technique. Perhaps this very disordered nature of Aβ is key to understanding its oligomer formation and the highly selective toxicity of the resulting structures.

References:
1. Lam AR, Teplow DB, Stanley HE, Urbanc B. Effects of the Arctic (E22-->G) mutation on amyloid beta-protein folding: discrete molecular dynamics study. J. Am. Chem. Soc. 130, 17413-17422 (2008). Abstract

2. Urbanc B, Cruz L, Yun S, Buldyrev SUV, Bitan G, Teplow DB, Stanley HE. In silico study of amyloid beta-protein folding and oligomerization. Proc. Natl. Acad. Sci. USA 101, 17345-17350 (2004). Abstract

3. Bitan G, Kirkitadze MD, Lomakin A, Volles SS, Benedek GB, Teplow DB. Amyloid beta-protein (Abeta) assembly: Abeta 40 and Abeta 42 oligomerize through distinct pathways. Proc. Natl. Acad. Sci. USA 100, 330-335 (2003). Abstract

4. Bitan G, Volles SS, Teplow DB. Elucidation of primary structure elements controlling early amyloid beta-protein oligomerization. J. Biol. Chem. 278, 34882-34889 (2003). Abstract

5. Bitan G. Structural study of metastable amyloidogenic protein oligomers by photo-induced cross-linking of unmodified proteins. Methods Enzyme. 413, 217-236 (2006). Abstract

View all comments by Brigita Urbanc

I am concerned about drawing firm conclusions about what happens in the human brain from pure synthetic oligomers. My lab prefers to work with naturally produced oligomers, even though they have obvious experimental limitations, and biophysical measures unfortunately cannot be applied due to their small quantities. I do believe from my own work that there will not turn out to be one predominant synaptotoxic oligomer form in the human brain but several assembly forms that are in dynamic equilibrium in vivo.

View all comments by Dennis Selkoe

Two New Articles Use Synthetic Aβ to Study Oligomerization
The first article by Bernstein et al. uses ion mobility coupled with mass spectrometry to study how Aβ40 and Aβ42 oligomerize in vitro. The measurements of arrival time distributions show that Aβ40 oligomers are restricted to low-n species (i.e., dimers and tetramers) while Aβ42-derived oligomers self-assemble into two additional structures, hexamers and dodecamers. The collision cross-sections for each Aβ42 oligomer led them to propose that Aβ42 tetramers are folded in an open structure able to accept one additional dimer to form hexamers, and that Aβ42 hexamers form a planar hexagon which can then stack with one more hexamer to create the largest oligomeric assembly, Aβ42 dodecamers (whose estimated mass was 55.2 kDa).

These new findings complement the observations reported by our group in Tg2576 APP transgenic mice (Lesne et al., 2006). We identified and isolated a putative dodecameric Aβ assembly, Aβ*56, a ~56 kDa assembly that correlates with...  Read more

Two New Articles Use Synthetic Aβ to Study Oligomerization
The first article by Bernstein et al. uses ion mobility coupled with mass spectrometry to study how Aβ40 and Aβ42 oligomerize in vitro. The measurements of arrival time distributions show that Aβ40 oligomers are restricted to low-n species (i.e., dimers and tetramers) while Aβ42-derived oligomers self-assemble into two additional structures, hexamers and dodecamers. The collision cross-sections for each Aβ42 oligomer led them to propose that Aβ42 tetramers are folded in an open structure able to accept one additional dimer to form hexamers, and that Aβ42 hexamers form a planar hexagon which can then stack with one more hexamer to create the largest oligomeric assembly, Aβ42 dodecamers (whose estimated mass was 55.2 kDa).

These new findings complement the observations reported by our group in Tg2576 APP transgenic mice (Lesne et al., 2006). We identified and isolated a putative dodecameric Aβ assembly, Aβ*56, a ~56 kDa assembly that correlates with memory dysfunction in Tg2576 and J20 mice (Lesne et al., 2006; Cheng et al., 2007), and disrupts memory consolidation when applied to young, healthy rats. Although we found putative hexamers migrating at ~27 kDa on SDS-PAGE (also confirmed by non-denaturing SEC) preceding the age at which Aβ*56 appears, we found trimers but no dimers or tetramers in the brains of these “middle-aged” mice (6-11 months), and therefore proposed that trimers, not dimers or tetramers, are the basic building block of hexamers and Aβ*56. In vitro analyses performed on Tg2576 primary neurons supported trimers as the basic unit synthesized within neurons (Lesne et al., 2006; Supplementary data). Not until the mice were “old” (>13 months) and showed dense-core plaques did tetramers and dimers appear, suggesting an alternate pathway leading to Aβ fibrils that uses dimers as the principle building element.

Another element of similarity between our two studies is the observation that, in both cases, endogenous Aβ*56 and in vitro Aβ dodecamers are the terminal soluble species.

Bernstein et al. begin to address the quaternary structure of Aβ oligomers. However, until the quaternary structures of endogenous Aβ oligomers can be ascertained, we cannot know whether the pathways leading to Aβ oligomerization in vitro result in the same structures as those leading to oligomers in vivo. Hopefully, the same methodologies applied to synthetic Aβ oligomers can soon be utilized to examine purified endogenous Aβ oligomers, in order to answer this question.

The second article by Harmeier et al. reports the role of Aβ glycine 33 in oligomerization, toxicity and neuronal plasticity. These authors show that single residue substitutions at position 33 (G33A or G33I) drastically affect Aβ oligomerization, resulting in increased high-n oligomers (10-, 16- and 20-mers) at the expense of the low-n Aβ species, dimers and tetramers. Using MS/MS spectrometry analyses, the authors demonstrate that the subtle folding change induced by G33A/G33I alters the intermolecular peptide interactions which control the assembly of Aβ to form larger Aβ oligomers. With these artificially created peptides favoring high-ordered Aβ42 oligomers, the authors then performed toxicity assays of the respective oligomers separated by size exclusion chromatography using neuronal cell lines and 10-day-old primary neurons, and generated transgenic drosophila expressing the various glycine 33 variants. They also examined how well the different glycine 33 mutants inhibited LTP. All results point to tetramers inducing cell death and inhibiting LTP more effectively than high-n oligomers.

Based upon our earlier conclusion, that fibrils and plaques are derived from dimers, this work may be more relevant to the plaque / fibril pathway than to the Aβ*56 pathway. It is noteworthy that Tg2576 mice and J20 mice producing Aβ*56 for nearly a lifetime do not develop neuronal loss, indicating that Aβ*56 does not induce neurotoxicity leading to cell death, unlike the low-n species studied here.

In summary, the diversity of Aβ-induced effects suggest that multiple oligomeric Aβ species trigger specific mechanisms at different stages of AD. We believe that Aβ*56 initiates AD pathogenesis, perhaps during the asymptomatic or pre-dementia phase of disease. In contrast, low-n oligomers, associated with dense core plaques, may play a role in the progression of AD at later stages.

View all comments by Karen Hsiao Ashe
View all comments by Sylvain Lesne