14 Dec 2010

Like squirmy toddlers dodging Grandma’s repeated attempts to photograph them, Aβ oligomers have proven evasive to the doting biophysicist. Indeed, progress has been slow for researchers toiling to discern the physical makeup of these neurotoxic clumps, widely seen as the form of amyloid-β most dangerous in Alzheimer’s disease. At the Society for Neuroscience (SfN) annual meeting held 13-17 November 2010 in San Diego, several labs presented new efforts to tease out the structure of Aβ oligomers. Surprisingly, a computational approach fingers the N-terminus as especially important for mediating Aβ42 neurotoxicity. Compared to the identical region in Aβ40, a less pathogenic Aβ peptide, the N-terminal end of Aβ42 flops around and is more exposed to solvent—making it potentially more susceptible to enzymatic cleavage and other modifications. The differences are captured in brief video clips (see below). Some scientists wonder how the results fit with prior work suggesting that the N-terminus is not involved in Aβ aggregation. Others seem captivated by the new data’s synchrony with growing evidence that N-terminally truncated forms may be the deadliest of Aβ oligomers. Further clarity should come with future experiments addressing whether amino acid substitutions that change solvent exposure correspondingly influence toxicity of the mutant Aβ peptides.

Having just two extra amino acids at its C-terminal end, Aβ42 seems to give neurons far more trouble than does Aβ40, but the reasons for this remain unclear. It would help if scientists could see the structures of these proteins. However, Aβ oligomers are constantly on the move, coalescing into larger assemblies and falling back apart—making them terribly unwieldy for high-resolution methods such as x-ray crystallography and solution-state nuclear magnetic resonance (NMR). While some scientists have used molecular tricks such as photochemical crosslinking (see Bitan et al., 2003) and disulfide bridges (see Sandberg et al., 2010; O’Nuallain et al., 2010) to tame Aβ oligomers for structural analysis, others are trying to manhandle the misfits from the comfort of their own laptops.

But even in the realm of computational modeling, Aβ oligomers have put up a challenge. The proteins are too complex for all-atom molecular dynamics—the computational models giving the most detailed structural information. “There are too many atoms, and the analysis would take way too long,” Brigita Urbanc, Drexel University, Philadelphia, Pennsylvania, told ARF after presenting her work at an SfN nanosymposium. To tackle the problem, Urbanc and colleagues went with a simplified approach called discrete molecular dynamics (DMD). This technique models atoms in motion as if they were colliding billiard balls. Proteins are reduced into strings of amino acids, each represented by a maximum of four beads, where one bead denotes, for example, a chemical group.

Of course, peptides do not exist in a vacuum, but are surrounded by solvent under normal physiological conditions, which makes molecular modeling even more complex. The analysis involved “no explicit solvent,” but instead used “a force field describing hydrophobic and hydrophilic effects,” Urbanc said. “With these simplifications, we can capture the full process of oligomerization, starting from non-interacting Aβ peptides and ending in a steady state where monomers and oligomers of various sizes are present.” As oligomers “form” on the computer screen, the scientists can separate them by size and do more detailed structural analysis on specific species such as dimers or trimers.

In earlier simulations of Aβ40 and Aβ42 oligomer formation using the four-bead protein method (Urbanc et al., 2004), “…the intriguing thing is that Aβ42 seemed to have more exposed N-termini than Aβ40 even though the structural differences are at the C-terminus,” Urbanc told attendees. Folding and assembly of even short proteins like Aβ are complex processes, she noted, and a sequence change at one position can cause structural changes anywhere in the sequence. This was also exemplified in her recent study (Urbanc et al., 2010), which used the same four-bead simplified approach to characterize oligomers of Aβ40 and Aβ42 and their Arctic mutants. With this mutation, “the amino acid substitution (E22G) is at the center of the peptide, and yet the [structural] differences are in the N-terminus,” Urbanc said. Specifically, her team found that the N-terminal end of Aβ40 has a short β strand that does not appear in Aβ42, or in the Arctic mutants of either full-length Aβ40 or Aβ42. These findings hinted that the N-terminus “presents the free energy barrier…that may be the reason Aβ40 doesn’t so readily form larger oligomers,” Urbanc said.

To test this prediction, Bogdan Barz, a postdoctoral fellow in Urbanc’s lab, selected 10 Aβ40 and 10 Aβ42 dimers formed in silico with the four-bead protein modeling, and examined their stability in a shorter (50-nanosecond) simulation using the more detailed all-atom approach. Aβ oligomers are typically too dynamic and complex for this comprehensive method, but in this case the researchers chose limited numbers of a specific Aβ species (i.e., dimers), making the analysis more manageable. Consistent with their structural data from the simplified model, the N-terminus “is doing some kind of dance, moving around and returning to its original position. This is typical for Aβ42, but not Aβ40,” Urbanc said as she showed video clips of each dimer “immersed” in water (see movies below).

Click on the images below to view each movie. (Please be patient; the movies may take several seconds to load.)

Aβ42 N-Termini—Movin’ and Groovin'
In computational models of Aβ oligomer formation, N-terminal region (with position 1 aspartic acid shown as red spheres) of Aβ42 dimer (right panel) is more flexible and less structured than the same region of Aβ40 dimers (left panel). Aβ peptides within each dimer appear in blue and green. Image credit: Bogdan Barz and Brigita Urbanc, Drexel University

In another round of analyses, researchers applied the four-bead protein method to study Aβ42 oligomer formation in the presence of three Aβ-derived C-terminal fragments (Aβ30-40, Aβ31-42, and Aβ39-42) previously shown to block Aβ42 neurotoxicity in cell culture studies done in the lab of coauthor Gal Bitan at the University of California, Los Angeles (see ARF related conference story and Fradinger et al., 2008).

In these simulations, the C-terminal toxicity inhibitors stuck themselves in between Aβ42 molecules, relaxing the β-strand structure of the full-length peptides and hindering their aggregation. Each toxicity inhibitor “acted like a glue that clamped the N-termini of Aβ42 and prevented them from moving around,” Urbanc told ARF. “Aβ40 is less toxic and has smaller solvent exposure at the N-terminus than Aβ42. But toxicity inhibitors seem to make Aβ42 have less solvent exposure and behave more like Aβ40.” The scientists also tested the effect of a control peptide, Aβ21-30, which did not affect Aβ42 toxicity in cell cultures, and showed that Aβ21-30 increased the solvent exposure of Aβ42 N-termini. “We conclude that perhaps it’s the exposure of the N-terminus to solvent that is in some way mediating the toxicity of Aβ42 oligomers.”

The new data are “extraordinarily compelling,” Thomas Bayer of the University of Goettingen, Germany, wrote in an e-mail to ARF. Urbanc and colleagues “demonstrated that the N-terminus of full-length Aβ oligomers is sticking out, and therefore exposed to potential peptidases for N-terminal truncation.” Given increasing evidence from Bayer’s and other labs that N-terminally truncated AβX-42 oligomers are highly toxic, aggregating more quickly than full-length Aβ (see ARF related conference story on pyroglutamate Aβ), Bayer noted that in Urbanc’s studies, “structural form and potential pathological function nicely fit together.”

Michael Nichols of the University of Missouri, St. Louis, was also intrigued by the new data, but more in a head-scratching sort of way. Whereas Urbanc’s results suggest that the N-terminus may affect Aβ oligomerization, “many studies, including our own (Touchette et al., 2010), indicate that the N-terminus is not involved in Aβ aggregation and remains available for modification after fibrils are formed (see Petkova et al., 2002 and ARF related news story; Whittemore et al., 2005; Kheterpal et al., 2001),” he wrote in an e-mail to ARF. In support of the SfN studies, researchers led by coauthor David Teplow at UCLA reported that two familial AD-linked mutations at the peptide’s N-terminus (English and Tottori) hastened Aβ assembly and increased toxicity (Ono et al., 2010).

As follow-up to the current work, Urbanc said her team plans to use both computational and experimental methods to test the predictions implicating the solvent exposure of Aβ42’s N-terminus in mediating toxicity. “Computationally, my group will explore single or double amino acid substitutions to predict how these changes affect solvent exposure of both Aβ40 and Aβ42, and then correlate our findings with available toxicity data on these mutated peptides,” Urbanc told ARF. “My group will also collaborate with the Teplow and Bitan labs to experimentally characterize structural changes induced by selected substitutions.”

At SfN, Leonid Breydo of Charles Glabe’s lab at the University of California, Irvine, described his latest efforts to characterize Aβ oligomers in “wet lab” structural studies. The Glabe lab has developed a number of conformation-specific Aβ antibodies (see Kayed et al., 2003 and Kayed et al., 2007), which they use to classify oligomeric Aβ as prefibrillar or fibrillar. The current study used site-specific denaturation and spectroscopy methods to compare the stability of Aβ40 fibrils, fibrillar oligomers, and prefibrillar oligomers. (All experiments were done with Aβ40 instead of Aβ42 because the latter aggregates so quickly that it is hard to work with, Breydo said during question time.) For the denaturation experiments, the researchers made 13 Aβ40 mutants, each with a different acrylodan-labeled cysteine substitution, and generated fibrils, fibrillar oligomers, and prefibrillar oligomers from each mutant peptide. Acrylodan is a polarity-sensitive fluorescent dye used here to measure local hydrophobicity of different regions in the tested proteins. The researchers measured hydrophobicity by using fluorescence spectroscopy to monitor the various Aβ species under various denaturing conditions. These data support the idea that fibrillar oligomers are structurally similar to Aβ fibrils, whereas prefibrillar oligomers look different. Their N-termini are more exposed, since “the acrylodan environment there is very polar—as polar as it is in water,” Breydo noted in an e-mail to ARF. “Prefibrillar Aβ oligomers may have a micelle-like structure,” Breydo said. “We think this because hydrophilic N-termini of prefibrillar oligomers are fully exposed, while hydrophobic regions of the structure are buried.”

Bitan mentioned a potential technical issue with this study—the possibility that the results could be influenced by experimental manipulations that compromise biological relevance. “The substitution by cysteine and introduction of the acrylodan fluorophore, which is quite bulky compared to most amino acid side chains, may alter the structure of the oligomers,” he wrote in an e-mail to ARF. He acknowledged that such concerns plague nearly every structural study of Aβ that uses molecular trickery to coax the peptide into forms amenable for analysis. And, perhaps more importantly, since the study only looked at Aβ40, he suggested, it is unclear whether the data “advance our understanding of the Aβ oligomer structures that are relevant to AD.” Breydo told ARF his team may examine Aβ42 oligomerization in future studies.—Esther Landhuis.