Saido became interested in Aβ43 when he saw a poster presentation indicating that a mutation in presenilin-1, the catalytic component of the γ-secretase that extricates Aβ from its precursor, shifts Aβ cleavage toward the 43 position (see subsequent paper by Nakaya et al., 2005). Shortly afterward, researchers led by Martin Rossor at University College, London, reported language impairment in people carrying one of the same genetic variants, R278I (see Godbolt et al., 2004), suggesting the mutation, and possibly Aβ43, are pathogenic. One of those carriers has since gone on to develop AD-like learning problems.

To test this idea, joint first authors Takashi Saito, Takahiro Suemoto, and colleagues developed a knock-in model, replacing the endogenous mouse presenilin (PS1) gene with one containing the R278I mutation. Mice heterozygous for the mutant PS1 appeared normal, but homozygous animals died in the womb and had severe developmental defects. Though γ-secretase, a complex of presenilin and three other proteins, appeared to form normally in the homozygotes, it did not properly process substrates, including APP and Notch, which are essential for development.

To test if a single copy of the mutant PS1 is sufficient to alter Aβ levels in the heterozygotes, Saito and colleagues developed enzyme-linked immunosorbent assays (ELISAs) to specifically identify Aβ40, 42, and 43. The ELISAs detected more guanidine hydrochloride soluble Aβ43 in the cortex of 24-month-old heterozygous mice than controls. On the other hand, the mutant mice produced less Aβ40. The pattern is in keeping with coauthor Yasuo Ihara’s finding that γ-secretase sequentially cleaves APP in three-amino-acid steps (see Takami et al., 2009 and ARF related news story), and suggests that the mutation retards the cleavage of Aβ43 to Aβ40.

This could have important implications. The ratio of Aβ42 to Aβ40 is commonly used as a measure of toxicity, but Saido believes an increase in that ratio could equally reflect an increase in Aβ43 and a concomitant decrease in Aβ40. Indeed, in the PS1 mutant heterozygotes, the Aβ42/40 ratio is higher than in controls, while Aβ42 levels are unchanged.

The researchers found that the R278I PS1 caused rampant elevations of Aβ43 in cells overexpressing wild-type amyloid precursor protein (APP). To see what effect this might have in vivo, Saito and colleagues crossed the R278I PS1 mutant mice with transgenic cousins overexpressing human APP. Three- and nine-month-old crosses produced more Aβ42 and 40 than controls, but even more Aβ43, such that the ratio of solubilized Aβ42 to 40 appeared normal, but Aβ43/40 was elevated. In the R278I PSs x APP crosses, amyloid deposits began to appear by six months and extensively decorated the hippocampus by nine months. The plaques had more Aβ43 than Aβ42, in contrast to plaques in control mice that overexpressed wild-type human APP with a M146V mutant presenilin. In the latter, plaques were predominantly made up of Aβ42. In R278I PS1 x APP mice, pathological changes were preceded by behavioral dysfunction. At four months old, the mice performed poorly in a Y-maze test of learning and memory.

All told, the evidence suggests that the R278I presenilin mutation shifts APP processing toward Aβ43, which then drives plaque pathology and learning and memory deficits. Could that be relevant to human AD? Researchers led by Lars Tjernberg at the Karolinska Institute in Stockholm found Aβ43 in amyloid plaque cores taken from both sporadic and familial AD cases (see Welander et al., 2009). Quantification by mass spectrometry suggested that Aβ43 represents only about 5 percent of the total; however, the peptide is more difficult to ionize for mass spectrometry than Aβ40 or 42, suggested Saido. “We must be cautious interpreting mass spec data,” he told ARF. Even if Aβ43 makes up a small amount of the total Aβ, it could still be influential. Saito and colleagues found that in a test tube, it aggregates faster, seeds aggregation more efficiently than either Aβ40 or Aβ42, and is more toxic to cells. Clinical evidence points to a role as well. The researchers found a correlation between age of onset, PS mutations, and Aβ43. Those PS mutations that produce the most Aβ43 when expressed in cells cause the earliest onset of AD. The same is true for PS mutations and Aβ42, but the correlation for Aβ43 was tighter.

Though Wolfe noted evidence pointing to poor removal of Aβ as the major problem in Alzheimer’s (see ARF related news story and upcoming Webinar on the role of ApoE isoforms in Aβ clearance), he said that it is an oversimplification to talk about Aβ40 and 42. There is a whole range of different Aβ peptides that are understudied, he said. As Bart De Strooper and Iryna Benilova, KU Leuven, Belgium, point out in an upcoming Nature News & Views, there are two major paths for Aβ production, both yielding peptides that are successively shorter by three amino acids. The Aβ40 line starts with Aβ49, which is chopped down into Aβ46, and Aβ43. Aβ42 originates from Aβ45 and Aβ48. Shifting γ-secretase to favor the former pathway, and, in particular, to have it go to completion, could be a potential therapeutic strategy. Saido said that Ihara already identified some compounds that facilitate processing of Aβ43 to Aβ40. Saido is also launching a screen to search chemical libraries for compounds that can do the same. Aβ43 might also be a useful biomarker. “Aβ42 is reduced in the CSF because it is trapped in the brain. Since Aβ43 is more prone to aggregate than Aβ42, I would expect that the CSF Aβ43 would be a more sensitive marker—if we can measure it,” said Saido. Tjernberg agrees. “My prediction is you would find an increase in Aβ43 in familial cases that have high Aβ42/40 ratios,” he told ARF. The problem is that no one ever measures Aβ43, and, in fact, some of the antibodies, such as BCO5, that were used in earlier studies of Aβ42, also recognize the longer peptide. Both Tjernberg and Saido are working on ELISAs that could be used to measure Aβ43 in CSF and other biological samples. However, in an e-mail to ARF, Erik Portelius, University of Gothenburg, Sweden, added words of caution, noting that the amount of Aβ43 in the brain is very small compared to Aβ42. “Whether Aβ43 can be used as a biochemical marker in cerebrospinal fluid, if present, remains to be shown,” he wrote (see full comment below).

What is clear is that these new mouse models raise many questions about Aβ43 and its role in AD. Ironically, Adams’s followers know that questions are as important as answers. The biological supercomputer that arrived at the ultimate answer to the ultimate question was powered by none other than sagacious mice. Unfortunately, it took them 7.5 million years to arrive at the answer, by which time the original question had long been forgotten. Hopefully, it will take much less time for mouse models to help answer whether Aβ43 is important in Alzheimer’s disease.—Tom Fagan.

Reference:
Saito T, Suemoto T, Brouwers N, Sleegers K, Funamoto S, Mihira N, Matsuba Y, Yamada K, Nilsson P, Takano J, Nishimura M, Iwata N, Van Broeckhoven C, Ihara Y, Saido TC. Potent amyloidogenicity and pathogenicity of Abeta43. Nature Neuroscience. 2011. Abstract