In yesterday’s Neuron, researchers presented fresh in-vivo support for the amyloid hypothesis, and they bolster the status of the APP protease BACE as the current favorite target for AD therapy development. Led by John Disterhoft and Robert Vassar at Northwestern University Feinberg School of Medicine in Chicago, with collaborators elsewhere, the study suggests that the cognitive deficits described in APP-transgenic mice are due to the Aβ peptide, not its parent protein APP, because depleting Aβ in transgenic mice reverses their memory deficits. The work further suggests that excess Aβ may wreak havoc on memory function by somehow interfering with cholinergic excitation of hippocampal neurons. The study focuses on soluble forms of Aβ. The paper also hints at a physiological function of this peptide in spatial memory. “This is an excellent and important study,” comments Mark Mattson.
The paper is the first in what will be a wave of studies examining crosses of BACE knockout mice with older transgenic strains that overproduce mutant human APP, and interpreting it fully will become easier once these other studies are available, as well. In 2001, three teams of scientists—one at Amgen including some of the present authors (Luo et al., 2001), one at Johns Hopkins’ University (Cai et al., 2001), and one at Elan Pharmaceuticals (Roberds et al., 2001)—each had created BACE knockout strains. All three strains appeared normal, raising hope that inhibiting BACE might have fewer side effects than inhibiting γ-secretase (see related New Orleans news story). Since then, these groups have bred their BACE-less mice to APP-transgenics, reasoning that such mice would have high levels of APP (and presumably the products of α-secretase cleavage) but no Aβ. This might allow scientists to finally answer the lingering question of whether the behavioral deficits described in APP-transgenic mice result from elevated Aβ levels or indeed from overexpression of the APP transgene itself. Despite abundant evidence implicating Aβ in Alzheimer's pathogenesis, direct proof of its role in cognitive decline, and any mechanism thereof, remains scant (see ARF related news story). Initial data of three studies similar to the Disterhoft/Vassar study were presented at the annual meeting of the Society for Neuroscience in New Orleans last November; brief summaries follow below.
In the present study, first author Masuo Ohno and colleagues crossed their BACE knockout mice with Tg2576 mice. ELISA analysis showed that, as expected, the bigenic mice indeed had almost no Aβ in their brains despite massive overproduction of APP, though it did detect residual Aβ40 and 42 levels that the authors suspect may be Aβ variants generated by other proteases. Ohno et al. compared the performance of their bigenic mice to that of Tg2576, BACE1 knockouts, and wild-type mice in two tests of hippocampus-based learning. First, they used a social recognition task (Kogan et al., 2000), which measures how well a mouse remembers the smell of a previously encountered littermate (a bit like the sniffing of one dog meeting another). Second, they used spontaneous alteration in the Y maze (Lalonde, 2002), which the authors believe is sensitive to early onset cognitive decline. Reflecting a recent shift in attention to soluble species of Aβ, the authors performed all their memory tests in mice at four to six months of age, well before plaques form in their brains. In both tests, the bigenic mice performed as well as wild-type mice, but Tg2576 mice had deficits, suggesting that accumulation of soluble Aβ caused these deficits, the scientists report.
Intriguingly, the BACE knockout mice, which had much less brain Aβ than did wild-type mice, also showed a mild deficit in the Y maze. This implies that Aβ may play a physiological role in learning and memory, a widely recognized but sparsely studied question (see Kamenetz et al., 2003; see also ARF related conference story). In terms of drug treatment, this result cautions that any future drug would require careful dosing to avoid depleting Aβ. Current research into additional targets of BACE besides APP also suggests as much (see ARF Live Discussion on BACE). “Nevertheless, this potential issue does not diminish BACE1 as a therapeutic target,…” the authors write.
Soluble Aβ and Cholinergic Function in Vivo
To get a sense of the mechanism by which high Aβ concentrations might impair social recognition and spatial memory, Ohno and colleagues performed electrophysiology in hippocampal slices of their bigenic mice. They picked up previous research by Rene Quirion, Mark Mattson, and others who had shown that Aβ applied to cultured neurons or slices inhibits cholinergic signal transduction. Further inspiration came from more recent work suggesting that two-month-old Tg2576 mice already have defects in protein kinase C activation downstream of muscarinic acetylcholine receptors, and that this in turn impairs the normal cholinergic modulation of GABAergic transmission in prefrontal cortex (Zhong et al., 2003). Cholinergic deficits in AD have been described as early as the 1970s, and most current AD drugs elevate acetylcholine temporarily (see Live Discussion). To test for cholinergic dysfunction in their mice, Ohno et al. measured changes in an acetylcholine-dependent consequence of learning called post-burst after hyperpolarization (AHP). The scientists report that in hippocampal slices from Tg2576 mice, they recorded a weakening of the normal stimulation of neuronal excitability by the acetylcholine receptor agonist carbachol, but that this stimulation reverted to the level of wild-type controls in the bigenic mice. Moreover, hippocampal slices from the BACE knockout mice responded like those of wild-type mice, suggesting that a different mechanism must account for the memory deficit these mice show in the Y maze, the authors add.
In summary, the authors show that genetically removing Aβ rescues two hippocampal learning and memory defects of Tg2576 mice, demonstrating that these deficits are due to Aβ, not APP. The authors also suggest that the rescue of these memory deficits occurs because purging the hippocampus of Aβ restores the normal regulation of neuronal excitability via cholinergic receptors.
In their discussion, Ohno and colleagues raise a caveat to the conclusion that Aβ is the only offending APP fragment. The bigenic mice lack not only Aβ, but also a product of BACE’s APP cleavage called β-CTF. This fragment might also contribute to memory dysfunction; prior work has indeed implicated it already (Nalbantoglu et al., 1997). Moreover, neuron-specific conditional knockouts of γ-secretase, which lead to lack of Aβ but presence of the β-CTF, do not rescue certain cognitive deficits (DeWachter et al., 2002; see Shen section of ARF New Orleans news story). Future experiments will need to sort out the contributions of various APP fragments to synaptic function and learning. This question grows even more complicated by the finding that BACE1 cleaves not only APP, but also its cousins APLP1 and APLP2 (Li and Sudhof, 2003).
Others Corroborate—or Do They?
Three other groups that are independently investigating what eliminating BACE does to APP-transgenic mice presented some of their unpublished data in New Orleans. Researchers led by Philip Wong at Johns Hopkins University presented a poster showing that breeding their BACE1 knockout mice with the APPswe/PS1 strain rescued the reference memory deficits observed in APPswe/PS1 transgenics (SfN abstract 524.11). A team led by John Morley at St. Louis University School of Medicine, Missouri, reported that reducing both Aβ and BACE levels by repeated injection of an antisense RNA construct into the brain ventricles of adult SAMP8 mice reversed the learning and memory deficits seen in these mice with the T-maze and lever press tests (525.2). (Harvard researchers are developing a parallel mRNAi approach; see Kao et al., 2003.) A third group appears to obtain somewhat different results. Researchers led by Mark Buttini and Lisa McConlogue of Elan Pharmaceuticals reported that removing BACE1 from PDAPP mice by breeding them to Elan’s BACE1 knockout strain indeed prevented progressive plaque pathology and hippocampal synaptic deficits described in PDAPP mice, but that their hippocampal atrophy remained (SfN abstract 524.7). What’s more, their colleagues Dione Kobayashi, Karen Chen, and others, tested spatial memory of these bigenic mice with the serial water maze paradigm, a variation of the original Morris water maze described in Chen et al., 2000. Like Ohno et al., this group found a spatial memory deficit in their homozygous BACE knockout. But unlike in the Ohno et al. study, this deficit was comparable to the deficit of the PDAPP mice. In addition, the Elan group saw in their BACE knockout/PDAPP mice no rescue of the PDAPP deficit, but instead the greatest memory impairment of all strains studied in this experiment (SfN abstract 524.4). These scientists also propose that Aβ may be required for normal working memory. Interestingly, this group also noted that calbindin levels changed in the mice, and that the degree of deviation from wild-type calbindin levels correlated with the memory impairment (see ARF related news story). You can view abstracts mentioned in this story at the SfN/ScholarOne website.
What Else Is New? More BACE Tidbits
BACE1 behavior, or rather, how BACE affects mouse behavior, has become an intriguing topic with a paper by researchers at GlaxoSmithKline in England, who reported last November in Molecular and Cellular Biology that they have made neuron-specific BACE1 transgenic mice that are especially bold and exploratory, whereas the corresponding BACE1 knockouts are notably timid (Harrison et al., 2003). In this month’s issue of the same journal, researchers led by Weihong Song of the University of British Columbia, Vancouver, characterize the BACE1 promoter. They fingered the transcription factor Sp1 in their attempt to understand the regulation of BACE expression (Christensen et al., 2004). On the genetics front, Chinese researchers report in the January 1 American Journal of Medical Genetics that they see a weak association of the single nucleotide polymorphism 1239G/C of the BACE1 gene in two cohorts of Han Chinese (Shi et al., 2004). And finally, a flurry of papers on experimental BACE inhibitors has just come online (Hom et al. 2004; Chen et al., 2004; Lamar et al., 2004; Park and Lee, 2003; Hu et al., 2003; Owens et al., 2003). Has your favorite experiment not made it into this snapshot of the BACE landscape? Send a comment and help complete the story.—Gabrielle Strobel
Q&A with Masuo Ohno, Robert Vassar, and John Disterhoft. Questions by Gabrielle Strobel.
Q: In your bigenic mice, what happens to the cleavage products of α-secretase? What are their levels? Might they influence the effects you see?
A: BACE1-/-/Tg2576+ mice have about twofold higher levels of brain APPs-α and C83 (the α-secretase cleavage products) as compared to BACE1+/+/Tg2576+ (Luo et al., 2001). Therefore, although unlikely, we cannot exclude the possibility that the effects we observed may involve α-secretase cleavage products to some extent. However, the vast majority of previous work by our group and others makes it very likely that reduced cerebral levels of Aβ were the reason for the rescue of the memory deficits that we observed in BACE1-/-/Tg2576+ mice.
Q: Why did you not use the water maze or the radial arm maze? That would have made it easier to compare the bigenic mice's behavioral phenotype with the results of others.
A: Recent work has focused on soluble Aβ assemblies (e.g., “ADDLs” and protofibrils) as potential neurotoxic species in AD. Therefore, we wanted to assay memory in young (4-6 months of age) Tg2576 mice before amyloid deposits developed so we would be able to analyze the effects of soluble, non-plaque-associated Aβ species on learning and memory. We chose social recognition and spontaneous alternation in the Y maze because these paradigms are very sensitive for measuring early, pre-deposit deficits in hippocampus-dependent memory in APP transgenic mice. Water maze or radial arm maze tests are possibly more standard assays of hippocampal learning. However, these tests appear less sensitive for measuring early memory deficits in young Tg2576 mice, although these paradigms become more robust for measuring memory deficits after Tg2576 mice begin to develop amyloid plaques at ~9 months of age. In addition, social recognition and spontaneous alternation in the Y maze are hippocampus-dependent tasks that rely on the natural social and exploratory behavior of mice.
Q: Why did you measure AHP? Again, would not commonly used measures of LTP have made it easier to interpret the results?
A: Emerging evidence suggests that cholinergic signaling is impaired by low concentrations of Aβ in younger Tg2576 mice, which might contribute to the early phase of cognitive impairments we observed in 4-6-month-old Tg2576 mice. Therefore, we tested the possible cholinergic mechanisms underlying the memory rescue in BACE1-/-/Tg2576+ bigenic mice. We investigated increased hippocampal neuronal excitability, as assessed by a reduction in the slow post-burst after hyperpolarization (AHP) in the bigenic mice. Our research group has done extensive investigation into AHP, which is modulated by acetylcholine, and we have shown it to be a cellular mechanism of learning. LTP, which is an electrophysiological model of synaptic plasticity and is mainly regulated by glutamatergic signals, is not a direct measure of the excitability change that we hypothesize underlies the learning rescue in the bigenic mice. Previous data showed that hippocampal LTP is normal in Tg2576 mice at 2-8 months of age, although LTP becomes impaired by the age of 15-17 months. Since we observed learning deficits during the time before plaques were formed and at ages when LTP was not impaired in the transgenic, this was further reason to concentrate on the AHP measure that we used.
Q: In New Orleans, scientists from Elan presented behavioral data on BACE ko/PDAPP bigenic mice, and these mice appear to have serious memory problems that progress with age when tested in the serial water maze. What might explain this apparent discrepancy?
A: We currently do not have an explanation for the discrepancy between Elan’s results and our own. We can only speculate that the discrepancy may possibly be related to differences in the behavioral paradigms or in the specific mice used by our two groups. For example, the Elan group used water maze while we used social recognition and Y maze tests; therefore, it is difficult to determine directly how their behavioral results relate to our own. In addition, Elan used a different APP transgenic (PDAPP) having the London APP FAD mutation (rather than the Swedish mutation in Tg2576), and the PDAPP transgene is driven by the PDGF promoter (rather than by the prion promoter in Tg2576). The mice used by our two groups also had different strain backgrounds, which together with the different APP transgenes may influence behavioral results. Clearly, more research will be needed to determine the cause of this discrepancy.
Q: There has been a flurry of papers on BACE1 inhibitors in the past couple of months. Are these all experimental tools, or do you consider any of them promising leads for drug discovery. If so, which ones?
A: The majority of BACE1 inhibitors that have been published to date are peptide-based, and therefore do not exhibit ideal drug-like properties. As such, they are mainly useful as research tools. Importantly, however, they also provide a starting point for rational drug design of small molecule BACE1 inhibitors with more favorable drug-like characteristics that may be developed into lead compounds.