24 December 2003. Scientists prospecting the amyloid hypothesis for new treatment strategies have set their sights on the idea of boosting the enzymatic destruction of the Aβ peptide. A flurry of recent papers and presentations at the 33rd Annual Meeting of the Society for Neuroscience, held last month in New Orleans, Louisiana, indicate that a growing number of research labs are investigating the proteases that can chew up Aβ. They are trying to gain an understanding of which ones are relevant in vivo, whether tweaking them affects amyloid deposition and relevant disease markers, and whether this could be done safely. While Aβ degradation remains poorly understood, results from several different approaches are beginning to converge.
Traditionally, Aβ production has received a much larger share of attention than has Aβ degradation (and indeed, on the production side, the current hot topic of investigation has shifted toward other APP cleavage products and away from Aβ itself). What’s more, while therapeutic attempts to curtail production—chiefly γ- and β-secretase inhibitors and vaccination protocols—are wending their way through the (mostly pre-) clinical pipeline, it has become as plain as day that these approaches are challenging and risky. So, at least theoretically, enzymatic breakdown of monomeric Aβ shines as the next-best hope of cutting off the cascade of detrimental consequences that occur when this peptide begins to accumulate, aggregate, and deposit. Below is a news account of recent developments.
Aβ-Degrading Enzymes: Lifesaver for Mice?
The most recent paper on this topic appeared in the December 18 Neuron. In it, Malcolm Leissring, Wesley Farris, and Dennis Selkoe of Brigham and Women’s Hospital, and colleagues, report that chronic elevation of the two most extensively studied Aβ proteases in neurons did not overtly harm mice. Instead, it reduced the mice’s brain Aβ levels and pathology, and even prevented premature death. The scientists created two strains of transgenic mice, one that overproduced insulin-degrading enzyme (IDE—see ARF related news story, ARF news story), and one that overproduced neprilysin, a neural endopeptidase studied intensely by Takaomi Saido’s lab (see below). The former mouse strain expressed twice the normal level of IDE and its enzymatic capacity was doubled; the latter strain expressed eight times the normal neprilysin level and also had a corresponding increase in activity. Leissring and colleagues then bred each of these two strains with APPSwInd mice (Mucke et al., 2000; also see APP Research Models).
Strikingly, this appeared to rescue a premature death phenotype that the authors observed when raising APPSwInd mice. Mouse models of AD amyloidosis generally have neither massive neuronal death nor full-blown AD. Even so, some lines of APP transgenic mice have been reported to die prematurely, prior to overt plaque deposition Hsiao et al., 1995), and Leissring and colleagues observed a similar effect in the APPSwInd mice, with one in five perishing by six months of age. This rate dropped to about one in 20 in the IDE+APP double-transgenics, and to nil in the neprilysin+APP cross. When quantifying brain Aβ levels in these crosses, the scientists found them cut in half in the IDE+APP transgenics, and down by 90 percent in the neprilysin+APP mice. Further experiments suggest that the proteases degraded chiefly Aβ monomers, not its aggregated forms. The extent of this reduction is surprising, the authors write, given that this particular model produces a 10,000-fold increase in Aβ over endogenous mouse brain levels, and it leads them to suggest that a even a small enhancement of Aβ proteolysis might over time have a therapeutic benefit. In addition, the plaque load of the IDE+APP mice was halved and the neprilysin+APP mice had barely any plaques at all. Antibody markers for astrocytosis, microgliosis, and dystrophic neurites were also down in the crosses.
This October, a different in-vivo study yielded broadly similar findings. Scientists led by Thomas Beach at Sun Health Research Institute in Sun City, Arizona, infused the neprilysin inhibitor thiorphan into the cerebral ventricles of rabbits for five days, and then found increases in cerebral and CSF Aβ levels Newell et al., 2003).
Both IDE and neprilysin are known to cleave many other substrates besides Aβ, for example, insulin. Could a drug raising their activity ever be safe? This question has no answer yet, but Leissring et al. note that the IDE or neprilysin single-transgenic mice are healthy, fertile, and without neuropathological abnormalities. This is but a rough first-pass check for possible mechanism-based side effects. For one, the transgenic mice had IDE and neprilysin upregulated chronically only in the brain, as the transgenes were driven by the neural CamKII promoter. (For this reason, the Neuron study also does not address links between insulin metabolism and AD; see ARF Live Discussion). For another, unexpected side effects in humans sometimes crop up which were not seen in extensive mouse studies. To pick a recent example from AD research, encephalitis was never reported in mouse studies of AD immunotherapy, but later developed in humans. That said, upregulating Aβ-degrading enzymes is unlikely to provoke adverse immune reaction. Finally, genetically manipulated mice frequently appear fine until a specific challenge (e.g., oxidative, neurotoxic, or metabolic) reveals more subtle vulnerabilities, so the IDE- or neprilysin-overexpressing mice might yet turn up other phenotypes, once challenged.
In fact, age-related oxidative stress does appear to affect these enzymes, though not in a way that would contradict the general gist of this therapeutic approach. Instead, neprilysin and IDE may be means through which oxidative byproducts fuel amyloid pathology. This October, scientists led by Dennis Dickson at the Mayo Clinic in Jacksonville, Florida, reported that 4-hydroxynonenal (HNE) adducts form on NEP proteins in AD brains in greater proportion than in control brains (Wang et al., 2003). And at the Neuroscience conference in New Orleans, Frank LaFerla’s group at the University of California, Irvine, showed preliminary data suggesting that IDE also is a substrate for HNE and is oxidized disproportionately in the hippocampus of AD brain (202.11).
More on Neprilysin
Also in New Orleans, Nobuhisa Iwata, working with Takaomi Saido at RIKEN Brain Sciences Institute in Saitama, Japan, and colleagues elsewhere, presented the latest findings of that group’s longstanding research on neprilysin. Previously, this group had shown that this enzyme localizes to presynaptic sites and axons (Fukami et al., 2002). They also generated neprilysin knockout mice that have elevated brain Aβ levels Iwata et al., 2001). At the conference, Iwata presented data on an experimental gene therapy approach to test the notion that increasing neprilysin activity can stem the amyloid component of AD and the synaptic damage associated with it. Using adeno-associated virus to ferry the neprilysin gene into cells, Iwata et al. showed that neprilysin transfer decreased Aβ levels in the brains of both neprilysin knockout mice and APP-transgenic mice (525.6). (Last March, Eliezer Masliah’s group at University of California, San Diego, reported that overexpressing neprilysin from a lentiviral vector lowered amyloid deposition; see Marr et al., 2003; also see ARF related news story and scroll to Masliah).
What’s Age Got to Do with It?
Last year, Saido’s group had shown that neprilysin expression wanes with age in the perforant pathway connecting the entorhinal cortex to the hippocampus and on mossy fibers, suggesting local reductions of neprilysin in areas known for synaptic damage in AD Iwata et al., 2002), while Antonella Caccamo and colleagues in LaFerla’s lab picked up the question whether age-related changes in IDE and NEP could account for amyloid buildup in sporadic AD. Last month at the Neuroscience conference in New Orleans, Caccamo presented initial data comparing steady-state levels of IDE and neprilysin in brain extracts of normal mice at two and 18 months of age. She found that both proteases fell with age in the hippocampus; in the cortex, neprilysin was unchanged and IDE appeared to increase slightly. In cerebellum, a brain region largely spared by AD, neprilysin increased with age and IDE stayed unchanged. When compared directly, cerebellum was the brain region with the highest steady-state levels in old mice. Likewise, human brain extracts had higher IDE levels in cerebellum than in hippocampus or cortex (202.11). In a separate study, Zhongmin Xiang, working with Giulio Pasinetti and colleagues at Mt. Sinai School of Medicine in New York, assayed IDE activity in postmortem brain samples of people with AD and correlated it with clinical progression. With advancing dementia, these investigators reported in New Orleans, IDE activity appeared to decrease in the hippocampus but not in the visual cortex, an area unaffected by AD (202.10).
What Have Genes Got to Do with It?
As yet, the jury is still out on the question of whether IDE is an Alzheimer’s risk gene. As occurs frequently in AD genetics, an initial report (Bertram et al., 2000) was thrown into doubt by failures to repeat (Abraham et al., 2001; Boussaha et al., 2002). However, more recent case-control studies (Edland et al., 2003) and haplotype analysis (Prince et al., 2003) keep finding hints that it might contribute significantly, after all. For a news summary on the insulin connection, scroll down in ARF ISOA meeting report. Lars Bertram, Rudy Tanzi, and colleagues of Massachusetts General Hospital presented data in New Orleans that appear to strengthen the candidacy of IDE.
No neprilysin mutations or independently confirmed polymorphisms predisposing to AD have been found to date (but see Clarimon et al., 2003). Last January, however, Saido’s group implicated neprilysin functionally, if not genetically, in some familial forms of AD. The Dutch, Flemish, Italian, and Arctic mutations of APP are unusual in that they lie within the Aβ peptide sequence of APP (see ARF related news story). They also, it turns out, make the mutant Aβ more resistant to proteolysis by neprilysin, the scientists reported (Tsubuki et al., 2003).
While this report focused on IDE and neprilysin, there are clearly more enzymes capable of chewing up the Aβ peptide, chief among them endothelin-converting enzyme and plasmin. And new candidates are cropping up, for example, a yet-unidentified serine protease reported by Carmela Abraham of Boston University last month in New Orleans (524.15). To date, scientists do not have a firm understanding of the relative roles of neprilysin, IDE, or the other Aβ-degrading proteases in the normal regulation of brain Aβ levels. What’s more, the field has not shown whether these Aβ-degrading proteases actually play a significant role in causing the disease. In other words, except in some forms of familial AD, it is unclear whether the spigot is turned up or the drain is clogged, or both. However, many researchers believe that, to the extent that Aβ accumulation contributes to Alzheimer’s, opening the drain makes intuitive sense, and perhaps by now this idea has garnered enough experimental support to warrant attention from drug developers.—Gabrielle Strobel.
Q&A with Malcolm Leissring—Posted 24 December 2003.
Q: The premature mortality of the APPSwInd mice was news to me. Has it been described somewhere?
A: As mentioned in the paper, other APP transgenics, specifically Karen Hsiao's (1995 Neuron paper), have been reported to die early, prior to plaque deposition. The speculation at the time was that mechanisms unrelated to amyloid deposition might be involved.
Q: Do the dying mice have neurodegeneration as in AD?
A: I don't know, but based on other APP transgenics, I doubt it. Lennart Mucke, in collaboration with Roger Nicoll and Robert Malenka, reported that there were electrophysiological changes in these mice at an early age (Hsia et al., 1999), as well as sharp reductions in synaptophysin (Mucke et al., 2000). The synaptophysin changes might count as a form of "neurodegeneration," but I know of no mouse model that shows degeneration of the type seen in AD.
Q: If not, what kills them?
A: Soluble oligomer-induced desynchronization of respiratory neuronal circuits during sleep??? Your guess is as good as mine. But this is an equally interesting question to ask of clinicians: What ultimately does kill AD patients? Neurodegeneration, per se? Probably not. Whatever does kill them, the results from our study suggest that it is likely a soluble species of Aβ, rather than aggregates, since the effects are seen prior to plaque deposition.
Q: Your neprilysin or IDE transgenic mice seem normal. But Roberto Malinow and other scientists are suggesting a physiological role for Aβ in synaptic function (Kamenetz et al., 2003). With Aβ presumably decreased in your mice right from the get-go, were you able to detect an LTP or a learning/memory phenotype in them?
A: We did not specifically address this question in our mice, but there is a strong inference that the LTP deficits present in Mucke's APP transgenics would be reversed by coexpression of either Aβ-degrading protease. We are currently crossing these mice to Frank LaFerla's triple-transgenic AD mice (Oddo et al., 2003), and they might have the resources to do this. It would certainly be interesting to find out.
Q: Did you assess behavior in the IDE+PDAPP and neprilysin+APP double-transgenics? If the described deficits improved only partially, this would be a way to distinguish between the roles of Aβ and some other APP cleavage products in learning and memory.
A: I agree that the double-transgenics are a nifty model in this respect: no Aβ, but lots of every other APP metabolite (except AICD in the IDE transgenics). I am actually hoping someone will use our mice to look at just this, as this would be too much work for our small IDE group (three people, counting Dennis).
Q: Therapeutically, how would one boost IDE or neprilysin?
A: One promising way would be disinhibition, that is, finding a small molecule that displaces an endogenous inhibitor. There are a few known endogenous inhibitors of IDE (e.g., ATP and fatty acids) and others for NEP (spinorphin, sialorphin) so this might not be impossible. Wes Farris in our lab discovered that IDE naturally exists as a homodimer, and possibly a trimer or tetramer, and Lou Hersh recently published the same result ( Song et al., 2003). Intriguingly, Lou found that low (nanomolar) levels of other substrates could increase the hydrolysis of Aβ, and we have seen similar effects resembling cooperativity with different substrates. Lou's idea is that the dimer is the more active form, and substrate binding might favor the dimer form, thus activating the protease. In this regard, a small molecule that would disrupt tetramer formation, or stabilize dimer formation, could conceivably activate the molecule.
Another way is to just screen as many compounds as possible, and hope to get lucky and find some molecule that lodges itself into the enzyme and in some way alters the Km or Vmax. Note that activators have been found for other enzymes, including the red wine constituent resveratrol activating the sirtuins Howitz et al., 2003).
There are other ways, including identifying compounds that upregulate the expression of the proteases. I'm sure you have heard Saido talk about somatostatin upregulating neprilysin, and estrogens have been reported to do this, as well. Interestingly, in our paper we found that IDE was actually downregulated in the NEP transgenics. This suggests that some common substrate might regulate the expression of IDE, and if you can identify what is regulating it, you can conceivably exploit that to activate expression.
For our own part, we are actively pursuing compounds that might enhance the activity or expression of IDE and other Aβ-degrading proteases. Over the past few years, I have used a newly devised high-throughput Aβ-degradation assay (Leissring et al., 2003) to screen over 130,000 compounds against naked recombinant IDE (i.e., with no endogenous inhibitor). While we have found a few activators, none are particularly impressive, showing only subtle effects at rather high concentrations. But there is much more to do, such as repeating the screens with endogenous inhibitors present. The nice thing about our assay is that it is universal: it can be used with any Aβ-degrading protease. So we've got a lot of screening lined up for this assay in the future. Please keep your hopes high but your expectations low!