In their hunt for compounds that divert γ-secretase away from generating Aβ42 without affecting its ability to cleave Notch, several groups independently are scrutinizing a small area of presenilin-substrate interaction that might be amenable to allosteric modulation. By studying APP dimerization, Gerd Multhaup of the Free University, Berlin, arrived at this very spot, as well. With graduate student Lisa Munter, Multhaup discovered a potentially critical dimerization motif on APP that influences the sequential cleavages of the γ-secretase substrate. By doing so, it helps determine what forms of Aβ come out of this process.

APP dimerization itself, by two different points of contact between monomers, had been published before (see Beher et al., 1996; Scheuermann et al., 2001). In her talk, Munter first noted that she and her colleagues had discovered a third, new site of contact between two APP molecules that mediates dimerization. This site includes three consecutive GxxxG motifs extending into the transmembrane sequence. Such motifs are known to promote interaction of protein helices, and a recent Alzforum Discussion raised the question of whether they might mediate Aβ toxicity. On APP, glycines 29 and 33 within the Aβ sequence represent particularly important residues of this GxxxG series, Munter and Multhaup observed. The scientists found that when they mutated the glycine 33 of APP, the mutant form not only failed to dimerize but also shifted Aβ generation away from Aβ42 and toward more of the smaller peptides Aβ37 and 38. Intriguingly, some NSAIDs have exactly the same effect on Aβ peptide distribution. This raises the question of whether they might do so by weakening the dimerization of the γ-secretase’s APP substrate C99, Munter and Multhaup speculated.

Here is how the German scientists explain their finding. The γ-secretase complex has been shown to process its C99 substrate by a series of sequential ε-, ζ-, and γ-cleavages. This is well-documented but has not yet been widely incorporated into people’s thinking about how this unusual protein machine works. Proceeding from the carboxyl terminus toward the N-terminus, the cleavages generate successively shorter Aβ peptides, from 49/48 to 46/45, to 43/40 and 42 (Qi-Takahara et al., 2005; Zhao et al., 2005; Chandu and Kopan, 2006). Quite possibly, ε and ζ cleavage proceed regardless of whether the substrate is a monomer or a dimer, yet after that, the dimer poses a steric obstacle that prevents the substrate from moving farther through the catalytic site. In essence, the idea is that dimers would stall successive cleavages at the Aβ42 site. Successive degradation of C99 would represent a physiological function of γ-secretase, and mutations that stabilize dimerization would increase production of an intermediate product of this normal degradation, namely Aβ42. By contrast, G29/G33 mutations—or perhaps some future therapeutic compounds—preventing dimerization would allow the substrate to be further processed into shorter peptides before it can leave the membrane.

γ-secretase inhibitors—alive and well?
When early classes of γ-secretase inhibitors began showing toxic side effects, especially severe ones in the intestinal tract and the maturation of lymphocytes, it looked like the approach might be dead in the water. Yet a more apt analogy might have been to not throw the baby out with the bathwater. It is true that companies are focusing on finding compounds that modulate γ-secretase rather than inhibit it outright, but they have quietly continued studying inhibitors, as well. “We really still would like to use these drugs. Some of them have great properties, if only we can get rid of their mechanism-based side effects,” said Christian Czech of Hoffmann-La Roche’s CNS Research group in Basel, Switzerland. With this comment, he echoed the sentiment of colleagues in other firms, and some in academia, as well.

One ray of hope in this area was implicit from one of a series of posters presented by Elan Pharmaceuticals in South San Francisco. In one, Guriqbal Basi and colleagues described an analysis of the structural determinants of different inhibitor classes. The upshot of this characterization of exactly which inhibitor binds to which amino acid residues on a given presenilin was that certain kinds of inhibitor clearly distinguish between presenilin-1 and presenilin-2. In this case, sulfonamides, as represented by a compound called BMS299897, acted more potently on the former than the latter, whereas DAPT and some other inhibitors acted comparably on both.

One can suspect that none of the inhibitors shown on the poster will be tomorrow’s drug, but the data is nonetheless important because it means that drugs can be found that will selectively block only certain kinds of γ-secretase complex. This might point a way toward finding a drug that acts in the brain but not in the gut, for example. Several labs have shown that different tissues of the body assemble the presenilin and Aph-1 isoforms in the human genome into γ-secretase complexes of distinct composition, and perhaps also distinct function.

Elan scientists showed more data on some new γ-secretase inhibitor series, and they were not alone. Nearly a dozen ICAD presentations dealt with γ-secretase inhibitors, some from academic labs, others from company labs such as Wyeth and Bristol-Myers Squibb. For his part, Mark Shearman, of Merck Research Laboratories, Boston, Massachusetts, described how scientists at his company used rational drug discovery to make new γ-secretase inhibitors. These inhibitors showed their ability to lower brain Aβ40 and 42 consistently when tested in Tg2576 mice, mice expressing full-length human APP off a yeast artificial chromosome, in guinea pigs, and then in a primate model.

To bridge the step from small animal to human, and to help predict an effective dose range in humans, the Merck scientists developed a rhesus monkey model that allowed them to monitor how a given inhibitor regimen affected Aβ levels in several fluid compartments over the course of days, weeks, even months. The partial nature of prior data sets—plasma time courses in this study, a spinal tap in that study—had made it difficult to compare between studies and especially to track how a given drug changed the separate pools of Aβ in the body over time (see ARF related news story on biomarker validation). Merck’s primate model overcomes this limitation. It is simple: a surgeon implants a catheter into the cisterna magna, a reservoir of cerebrospinal fluid at the lower back of the head. The catheter connects to a sterile port that allows the scientists to drain a few drops of CSF for analysis non-invasively whenever needed without having to sedate the animal. A cohort of about 30 monkeys has been living with this catheter for more than a year without ill effects, Merck scientists said. These animals have enabled the scientists to compare the effects of their γ-secretase inhibitors to those of γ-secretase modulators on Aβ40 and 42 levels in plasma and CSF. On a poster, Maria Michener, working with Lynn Cook and colleagues, showed data for 6-day studies, but internally the scientists have followed the effects of their compounds for longer.

Moreover, in an initial human safety and pharmacokinetics/ pharmacodynamics study presented by Laura Rosen, a lumbar catheter was implanted into 27 human volunteers at two clinical study sites in order to test a single oral dose of the MK-0752 γ-secretase inhibitor. CSF was then collected for up to 30 hours afterwards, as well as blood samples for up to 4 days. Besides being well-tolerated in this small study, MK-0752 made it into the CSF, where it replicated a dose-dependent Aβ40 reduction as seen in the primate model, Rosen reported. “We believe the basic biology of APP metabolism, and its monitoring in plasma and CSF in response to our inhibitors, is transferable from mouse through guinea pig through primates to the human,” Shearman said in an evening lecture detailing some of the company’s efforts on γ-secretase inhibitors, modulators, and BACE inhibitors. The MK-0752 γ-secretase inhibitor is currently in phase 1 trials as a candidate disease-modifying agent, and other compounds have recently been published (see Jelley et al., 2006; Shaw et al., 2006; Best et al., 2006).—Gabrielle Strobel.