8 November 2004. The more scientists learn about γ-secretase, the crazier it looks to them, according to Todd Golde of the Mayo Clinic in Jacksonville, Florida. Golde is but one of many avid students of this eccentric enzyme, which slices the APP protein twice inside the membrane to release the much-studied Aβ peptide and the little-studied C-terminal AICD fragment. Setbacks and complications notwithstanding, many academic and company researchers still scrounge the APP cleavage process for chances to interfere therapeutically, and consequently, γ-secretase was the topic of several dozen presentations at the 34th annual meeting of the Society for Neuroscience, held last week in San Diego. Here we present selected highlights.
Christian Haass, from Munich’s Ludwig-Maximilians-University, Germany, reported on his group’s efforts to understand the steps by which the γ-secretase complex comes together and matures. Remember, in addition to presenilin, the complex comprises at least Pen-2, nicastrin, and Aph-1 (see report from the recent Swiss Society of Neuropathology meeting).
Haass and other researchers have reported that maturation of the complex takes place in the endoplasmic reticulum (see, for example, ARF related news story). Might specific ER retention signals hold the proteins there until their partners have arrived? Haass reported that presenilin 1 (PS1) indeed has an ER retention signal in its C-terminus (SfN abstract 264.4). A CD4 chimeric protein having the PS1 C-terminal end was retained in the ER, for example, while a chimera with the PS1 N-terminal end migrated rapidly to the cell membrane. Postdoctoral fellow Christoph Kaether mutated the C-terminal end of PS1 to locate this retention signal. It turned out to be quite long, but it overlaps with the Pro-Ala-Leu-Pro stretch of amino acids that is found in all other presenilins. When the Kaether and Haass mutated this motif, PS1 still formed a complex with its three partners, but it moved rapidly out of the ER and it lacked γ-secretase activity. In contrast, when they tacked the wild-type retention signal on nicastrin, the complex assembled and was active. The results suggest that the retention signal keeps the partners of the complex in the ER long enough for maturation to occur. Haass presented evidence that maturation is accompanied by a masking of the retention signal, triggering migration of the secretase to the cell membrane (part of this work is in press in the EMBO Journal, Kaether et al.).
Despite this complex process of maturation, several labs have succeeded in reconstituting γ-secretase in other systems, for example, yeast (Edbauer et al., 2003). In San Diego, Takeshi Iwatsubo’s group at the University of Tokyo in Japan offered another system for study by describing a new method to express human γ-secretase components in budded baculovirus particles (Hahashi et al., 2004 and SfN abstract 90.10).
By contrast, Patrick Fraering, working with Michael Wolfe and Dennis Selkoe at Harvard Medical School in Boston, took a conventional biochemistry approach to studying this enzyme complex. He described how developing a six-step protocol for purifying human γ-secretase (Fraering et al., 2004) has enabled him to analyze directly the role of γ-secretase interactors and modifiers.
Using APP’s C99 fragment as a substrate (i.e., a recombinant protein consisting of the &β-CTF portion of APP), Fraering first used his in-vitro assay to test how lipids and membrane-perturbing detergents influence γ-secretase cleavage. Sphingomyelin and phosphatidylcholine increased the enzyme’s activity but not the ratio of Aβ42/40 products, whereas the detergents CHAPSO and SDS did change that ratio. The results suggest that the precise cleavage specificity of the enzyme is sensitive to lipid conditions. Several other groups are studying how membrane lipids alter APP processing and, in turn, how various forms of Aβ itself might compromise the integrity of membranes (see ARF related news story).
Next, Fraering used this system to reproduce aspects of a study from Paul Greengard’s laboratory, which had suggested that the approved cancer drug Gleevec selectively inhibits γ-secretase cleavage of APP while leaving Notch alone (Netzer et al., 2003). This paper seemed to revive γ-secretase as a drug target, an idea that had taken a beating as the list of γ-secretase substrates grew and early inhibitors proved toxic (see ARF related news story). Tantalizing as the Greengard paper was, it also seemed puzzling. It purported that Gleevec, a card-carrying tyrosine kinase inhibitor, did not affect γ-secretase through this primary function, but instead worked by a different, unknown mechanism, possibly one having to do with the ATP requirement for Aβ production. How could such a mechanism enable Gleevec to shift specifically the cutting of APP but not Notch? The final answer is still not in, but Fraering used the purified human enzyme complex to confirm the Netzer et al. data. His data indicate that Gleevec acts directly on the γ-secretase. He showed that ATP itself can modulate the cleavage activity of the purified γ-secretase, and noted that a screen of 50 tyrosine kinase inhibitors brought up a few others that also block Aβ but not Notch cleavage, perhaps via an ATP binding site. (SfN abstract 264.14).
“This is the most convincing demonstration yet that you can distinguish between APP and Notch cleavage,” said Golde. This does not mean that a drug is at hand. The doses needed would be too toxic for chronic use in humans; besides, Gleevec does not cross the blood-brain barrier well. Moreover, APP cleavage modulated by Gleevec still increases APP C-terminal fragments, which themselves might be toxic, Golde added.
One possible reason why small molecules may influence APP cleavage but not that of Notch may be that they don’t bind to the active site of γ-secretase, but instead influence the enzyme-substrate interaction in other ways. In San Diego, Wolfe suggested APP processing may require two independent steps of docking followed by cleavage (see SfN abstract 264.1)
His evidence centers on the use of a classic tool in enzymology, i.e., transition-state analogues that bind to the active site of γ-secretase. His group also designed small, 10- to 13-residue peptides to mimic the 3D structure of the section of APP that protease cleaves. To keep these peptides in their correct α-helical shape, Wolfe and colleagues substitute α-methyl alanine (Aib), which favors the helical conformation, for selected APP amino acids (see Bihel et al., 2004). Using photoaffinity labeling, Wolfe showed that these Aib APP peptides bind to both the N- and C-terminal fragments of presenilin 1, though preferentially to the N-terminus. They also inhibit γ-secretase potently in cell-free systems.
But it was when Wolfe and his colleagues tried to use transition-state analogs to prevent these peptides from binding to presenilin that they could show the two types of compounds bound to distinct locations. Unexpectedly, one helical peptide, D10, bound to presenilin even in the presence of transition-state analogs (TSA), while another one that’s three amino acids longer did compete with the TSAs. The Wolfe lab suggested the following explanation: The shorter peptides bind only to the APP docking site on presenilin, which may be between two of the transmembrane helices that make up the barrel-shaped protease. The longer peptide not only binds to the docking site, but also protrudes sufficiently through the presenilin helices to access the active site that is on the inside of the barrel. In support of this theory, Wolfe noted that FAD mutations in presenilin substantially alter TSA-binding, but have little or no effect on D10 binding.
This research also ties in with Golde’s and Eddie Koo’s own collaborative work. They were the first to discover that certain NSAIDs tweak γ-secretase activity in a way that shifts APP cleavage away from Aβ42 production while not touching other substrates. In San Diego, Thomas Kukar from Golde’s lab told the audience that he has performed a broad search for other compounds that do this more potently than the initially discovered agents, ibuprofen and indomethacin (SfN abstract 264.13). This search turned up a surprising variety of compounds that raise Aβ42 production in cell-based assays, including some COX-2-inhibiting NSAIDs such as Celecoxib. Ironically, Vioxx, which was retracted this fall (see ARF related news story), does not raise Aβ42.
But no COX-2 inhibitor lowered Aβ42. In fact, when the scientists made chemical derivatives of a given Aβ42-lowering NSAID to turn it into a COX-2 selective compound, they found that the derivative now raised Aβ42 production. This prompted Kukar to claim that COX-2 inhibition and Aβ42-lowering are mutually exclusive functions on a given molecule. Moreover, in collaboration with Koo, Golde, and Christian Haass, Sarah Sagi from University of California, San Diego, reported that NSAIDs that can alter the cleavage site in APP, including sulindac sulfide and flurbiprofen, have no effect on the cleavage site for Notch or CD44, indicating that the NSAID effect is substrate-specific (see SfN abstract 264.10).
One of the most potent Aβ42-raising agents Kukar found is the PPARγ agonist and lipid regulator fenofibrate. Others are isoprenoid compounds that are dietary metabolites. In principle, this raises the question of whether some dietary isoprenoids, or compounds that somehow alter isoprenoid metabolism, could pose an environmental risk for AD, the scientists said. It is clearly too early to say, but this pathway should be studied, they add.
Finally, Kukar addressed the question of how the compounds act. To do so, they tried to confirm a recent paper (Zhou et al., 2003) that suggested that the effect is indirect and mediated through the Rho-Rock pathway of small GTPases. Kukar and colleagues were unable to reproduce these data. He instead suggested that these modulators act on γ-secretase directly, not via a second messenger pathway, as did Fraering with regard to Gleevec.
Another development γ-secretase aficionados follow closely is that of the mysterious epsilon cleavage. Two years ago, German researchers reported that they found yet another cleavage at play in the sequential digestion of APP, and termed it epsilon (Weidemann et al., 2002). It occurs closer to the cytoplasmic side of the membrane than the γ cleavage and gives rise to AICD. Since then additional groups have tried to understand the relationship of this new cleavage with the ones already known, particularly γ. In San Diego, two groups presented new data indicating that the epsilon cleavage can precede γ cleavage of APP (see SfN abstracts 90.14 and 146.4.) This is intriguing in part because it would open up an even more upstream opportunity of interfering with Aβ generation than is being targeted currently.
Clearly, this bewildering enzyme complex is only beginning to yield its secrets. Even so, many scientists at the Neuroscience meeting shared a sense that there may yet be a workable drug strategy in it, eventually.—Gabrielle Strobel and Tom Fagan.