Fiendishly complicated, γ-secretase has proved a formidable foe to the structurally minded scientist. But goaded by éminence grise T.I., this Japanese bunch of rebels is deploying the force of SCAM and other stratagems to conquer it yet. Following our initial Keystone Treatment News series, read our next series of stories from the March 2008 Alzheimer Disease conference in Keystone, Colorado. Learn how some rebels explore the bends and recesses of the enzyme’s deep space, while others take on elusive science mysteries of Alzheimerology. May the Force be with them.
Keystone: γ Slowly Relinquishes Its Secrets
Fiendishly complicated, γ-secretase inspires both awe and weariness. Biochemists with an appetite for structure-function studies marvel at its intricate design; indeed, they thrive in the hip new subfield of RIP/I-CliP research that this protein complex of 19 transmembrane domains has helped inspire. (Spelled out, the acronyms, less elegantly, mean regulated intramembrane proteolysis/intramembrane-cleaving protease). At the same time, colleagues of pathological, genetic, or clinical persuasion silently wonder how much hardcore enzymology they have yet to endure to stay abreast of this part of their field. No matter where they are coming from, though, scientists agree that (along with β-secretase) γ-secretase remains a pivotal subject of AD research, and the Keystone conference held 24-29 March in Keystone, Colorado, offered new insights on just how it might work. Presentations addressed its structure, its sensitivity to lipids, and the way in which mutations causing familial Alzheimer disease corrupt its activity. Here is a summary of new data on these three issues, presented by three groups in the field.
Takeshi Iwatsubo of the University of Tokyo, Japan, presented the latest data from his group’s ongoing effort to glean indirect peeks at the structure of γ-secretase. A structure-based understanding of the enzyme would offer a starting point for drug development, much like it did for HIV protease. Ideally, a definitive view might come from atomic-level crystal structures of γ-secretase bound with substrate or inhibitor molecules; alas, given the size and complexity of this membrane protein complex, no one expects such structures anytime soon, even though numerous labs are working on the problem. Electron microscopy images of γ-secretase have given some hints about where water might enter its hollow inner cave, but otherwise reveal nowhere near enough detail (Lazarov et al., 2006; Ogura et al., 2006).
In the meantime, the challenge is to gather biochemical clues against which any future crystal structure can be validated. Using cysteine scanning mutagenesis, cross-linking, and inhibitor studies, Chihiro Sato, Shizuka Takagi (see image below), and Taisuke Tomita performed an arduous project toward this goal that drew praise from conference attendees. Such indirect methods of structure-function not only advance the field in the absence of X-ray crystallography, they also build an important body of data to assess later on whether a future structure has indeed captured the protein in its true, active state, said Michael Wolfe of Brigham and Women’s Hospital in Boston, Massachusetts.
These scientists have joined forces to illuminate the shadowy recesses of γ-secretase with cysteine scanning, cross-linking, and inhibitor studies. Image credit: Ikuo Hayashi
Put simply, the abiding question that animates these studies is this: how can an enzyme hydrolyze, i.e., break a molecular bond using water, right in the middle of a lipid bilayer—a space that is thoroughly water-repellant? To explore this question, Sato and colleagues used cysteine scanning mutagenesis, also called substituted cysteine accessibility method (SCAM). This technique probes the enzyme for the presence of water-accessible residues in the nine different transmembrane domains (TMD) that make up presenilin (the protein-cutting part of γ-secretase). Residue by residue, a cysteine-scanning study evaluates the residue’s respective role in γ-secretase function. In previous work, the Japanese scientists, and also Bart de Strooper’s group in Leuven, Belgium, had proposed the presence of a water-filled catalytic pore lined with hydrophilic sites, and had begun to define its composition made up of particular residues of TMDs 6 and 7 (Sato et al., 2006 with some attendant lightsaber rattling, all in the search for truth; Tolia et al., 2006). This pore is where the cleavage happens. Iwatsubo’s group believes the pore might be shaped like a funnel, with the two catalytic aspartates near its waist.
The identity of the residues that border this pore—creating, in essence, a watery microenvironment inside the otherwise hydrophobic membrane—is the topic of ongoing study. At Keystone, Iwatsubo presented new data extending beyond the published work on TMDs 6 and 7 to TMDs 8 and 9. SCAM analysis of these two domains indicated that TMD 8 is truly buried in the lipid bilayer without contributing any sites to the watery pore. TMD 9 is a different story. It contains many hydrophilic residues, Iwatsubo reported. They cluster around the so-called PAL motif, a three-amino acid stretch on one end of TMD 9 that has previously been implicated in the γ-secretase active site (e.g., Wang et al., 2006). Additional highly water-accessible residues were found further up on TMD 9 and also toward the C-terminus of presenilin. This tail section of the protein is amphipathic, i.e., both hydro- and lipophilic, Iwatsubo reported, and it might serve as the previously proposed “lateral gate” that lets in the γ-secretase substrate from nearby in the membrane, Iwatsubo suggested. (The enzyme does not accept its substrate from the watery milieu on either side of the membrane, like a channel lets in ions. Instead, it somehow grabs its substrate from within the lipid bilayer; how it does so and guides it into its watery core is at question.)
Structural clues about membrane proteins can also be deduced from disulfide cross-linking, a method that allows the scientist to estimate angstrom distances between individual amino acids. Using MTS cross-linkers, Iwatsubo and colleagues constructed a structural model of TMD and presenilin’s C-terminus. It places the PAL motif very close to the catalytic site that had been previously defined by the catalytic aspartates in TMDs 6 and 7 plus a so-called GxGD motif in TMD 7, and it envisions the luminal side of TMD 9 as forming one tilted funnel wall facing the inner pore. (The GxGD motif defines the active site of the I-CliP family of proteases; see more below.) Further structural hints came from an analysis of how the binding of three different types of γ-secretase inhibitor competed with SCAM, Iwatsubo reported. By this method, too, the proposed substrate binding sites comprise the catalytic aspartates in TMDs 6 and 7, the GxGD motif, the PAL motif, and residues on the luminal side of TMD 9.
The details of this study will follow in a paper. In essence, Iwatsubo suggested at Keystone that the group’s data support a mechanism whereby the substrate first binds to a distal piece of TMD 9 that constitutes the lateral gate, then passes through the C-terminal side of the PAL motif, and from there reaches the catalytic site in the middle of the lipid bilayer, which is made up of the luminal side of TMD 6, the GxGD motif, the PAL motif, and separate portions of TMD 9.—Gabrielle Strobel.
This is Part 1 of a three-part series. See also Part 2 and Part 3.
At the Alzheimer disease conference held 24-29 March in Keystone, Colorado, both Dennis Selkoe and Michael Wolfe of Brigham and Women’s Hospital, Boston, described in their respective talks joint new research on how the lipid composition of the membrane influences the enzyme activity of γ-secretase. γ-secretase modulation has become a buzzword in the field, and until recently, scientists thought primarily of small molecules (such as flurizan) and protein partners (such as TMP21) when studying γ-secretase modulation. But the enzyme is embedded in the lipid bilayer, after all, so lipid changes in the surrounding membrane environment would seem to be an obvious way of affecting its activity, as well. This may well be relevant to Alzheimer disease, as numerous groups have reported lipid differences between brain samples of AD patients and controls over the past dozen years. Even within a given cell, γ-secretase activity varies from compartment to compartment along with changes in their respective membrane lipid composition.
To study this question, postdoc Pamela Osenkowski, working with both Wolfe and Selkoe, modified a γ-secretase activity assay. Importantly, she removed the detergents that previously were routinely present in in-vitro activity assays as a leftover from the extraction protocol used to obtain γ-secretase in solution. (Besides being artificial and absent from γ-secretase in brain, detergents alter the enzyme’s activity in vitro, Wolfe said.) Specifically, Osenkowski added defined lipids and lipid mixtures to purified human γ-secretase, then removed the detergent. This allowed the lipids and γ-secretase to form proteolipid vesicles, or liposomes. This method creates a more natural environment of lipid bilayers to study the activity of purified γ-secretase, while at the same time giving Osenkowski control over precisely which lipids are present in those bilayers, Wolfe said. It allowed her to systematically characterize γ-secretase activity in different lipids and lipid mixtures.
First off, getting detergents out of the picture doubled the enzyme’s baseline activity, the scientists found. Further, the type of lipid present mattered greatly. Cerebrosides or gangliosides boosted γ-secretase activity over a baseline established with phosphatidylcholine only (this lipid is the most common kind in cell membranes and thought to provide their structural bulk). By contrast, phosphatidylinositol slowed the enzyme down. But a given lipid did not merely boost or crimp γ-secretase activity wholesale; some had quite specific effects. For example, adding cholesterol to phospholipids changed cleavage of APP, APLP1, or Notch-like substrates differentially and dose-dependently.
A more realistic recapitulation of the natural environment had different effects still. Complex lipid mixtures reconstituted γ-secretase in structures resembling lipid rafts, and this environment afforded high γ-secretase activity. Furthermore, brain lipid extracts by far topped liver or heart lipid extracts in terms of revving up γ-secretase activity. Brain lipid extract led to the greatest Aβ40 production of all lipid mixtures tested, whereas E. coli or soybean lipid extract did not support γ-secretase activity at all.
It’s not clear yet whether the lipids affect γ-secretase because they change the general properties of the surrounding membrane, for example, its fluidity and packing, or because certain charged lipids interact directly with γ-secretase. This research is important in part because understanding and tweaking the lipid environment may aid efforts to grow γ-secretase crystals suitable for X-ray crystallography. It may also add small molecules that target lipids to the mostly protein-oriented drug development efforts that predominate today, the scientists note. Finally, interest in the interactions between membrane lipids and APP secretases is growing across the field. Just last Friday, a paper in Science reported that tethering a β-secretase inhibitor to membrane lipids, rather than having it float through soluble compartments, sharply increased its activity (Rajendran et al., 2008). The work of Osenkowski et al. is submitted.—Gabrielle Strobel.
This is Part 2 of a three-part series. See also Part 1 and Part 3.
Keystone: Loss Versus Gain—Mutations a Drag on γ-Secretase
A third area of γ-secretase research that drew attention at the Keystone conference held 24-29 March in Keystone, Colorado, is the hotly debated question of how presenilin mutations cause familial Alzheimer disease (FAD). Scientists wonder exactly how these mutations manage to shift the cleavage of APP in such a way that the ratio of Aβ40 and 42 changes enough for the more fibrillogenic Aβ42 to occur in excess over a period of years. The question has been framed in several ways. Some investigators asked whether FAD mutations cause a gain or a loss of presenilin function, some asked what that meant for drug development, and others questioned whether the distinction was real and important or merely semantic. The Alzforum has addressed the issue in two Live Discussions (see Davies/De Strooper discussion; see Shen/Kelleher discussion). At the Keystone conference, Christian Haass of Ludwig-Maximilians University in Munich, Germany, joined the debate with new data from his lab. In an opening plenary lecture, Haass first broadly summarized what’s known about amyloidogenic APP proteolysis and its players so far. Then he presented data suggesting that at least one FAD presenilin mutation his group studied in detail—the G384A mutation that sits smack in the crucial GxGD motif in the active site—slows down the processive cuts the enzyme usually performs on the substrate in such a way that relatively more Aβ42 gets made at the expense of the less harmful Aβ40, whose production plummets. In essence, then, the verdict is Solomonic: a relative loss of presenilin function (Aβ40 generation) amounts to a relative gain of toxic function (Aβ42 aggregation).
Haass arrived at this conclusion by first using SPP2b, a simpler member of the GxGD-type family of aspartyl proteases that previous research by his and other groups had defined. These enzymes make up the family of I-CLiPs. SPPs in particular come in handy for the study of the AD-relevant presenilin because they work alone, without requiring presenilin’s three protein partners Nicastrin, Aph-1, and Pen-2. Hence they are simpler to study but otherwise function almost identically to presenilin. They are also fascinating in their own right in that they operate as virtual mirror images of presenilin. The orientation by which their nine TMDs weave in and out of the lipid bilayer is flipped relative to that of presenilin; they cut substrates existing in type 2 orientation, while presenilin cuts substrates existing in the opposite type 1 orientation, and their cleavage products, too, are identical but opposite in orientation. “Their orientation across the membrane is opposite; everything else is in parallel,” Haass said.
SPP2b offered a handle to test the problem of the FAD mutations when Regina Fluhrer in Haass’ group found that this protease processes tumor necrosis factor α (TNFα, a type 2-oriented transmembrane substrate) in much the same way as presenilin cuts APP. The similarity even includes a few slightly different cleavage sites inside the membrane, much like the different cleavages known for presenilin. The major reason GxGD proteases perform multiple cuts starting from these different cleavage sites is to remove the hydrophobic sequence of the substrate as a part of its degradation, Haass speculated.
Fluhrer made the FAD presenilin G384A mutation in SPP2b (where it is a G420A mutation). To her surprise, she found that it produces an elongated TNFα intracellular domain (ICD), which, because of the reaction’s flipped orientation, corresponds to Aβ in presenilin cleavage. This means that an FAD mutation in SPP2b created a change corresponding to an Aβ40-to-42 shift in APP processing, Haass said. Further analysis of TNFα cleavage with this FAD mutation showed that the normal, stepwise cleavage that degrades the substrate and finally liberates the ICD slows down in the mutant and becomes much less efficient. It takes the mutant protease much longer to produce the smaller peptides. Indeed, the effect of the mutant can be mimicked by simply chilling the wild-type enzyme down to 20 degrees Celcius from the 37 degrees at which it usually operates, Haass showed.
The finding held true in the same way for G384A presenilin and Aβ generation: Aβ40 production slowed to a crawl; Aβ42 production stayed unchanged. The reasons for this are not proven yet, but Haass interprets it this way at present: prior research by Yasuo Ihara at the University of Tokyo and others has shown that γ-secretase has different production lines, whereby the secretase latches on at a slightly different starting point and then cleaves processively. The production line that latches on at the so-called ε site releases Aβ40, and that is the one most hampered by the mutation. This leaves more substrate for the production line that produces Aβ42. (The exact details of how processive cleavage works, and how the ε, ζ, and γ cleavage sites relate to how much of each final product is made remains unclear. Audience members asked those questions, but no γ-secretase speaker was willing to speculate, which is as good a hint as any that several γ-secretase labs are trying hard to get a grip on the problem.)
In essence, aggressive FAD mutations slow down the pace and efficiency of γ-secretase in a selective loss of function, Haass suggested. His stance adds new data and a new voice to a position articulated recently by Michael Wolfe and Bart de Strooper (see also Winklhofer et al., 2008). Furthermore, it fits with a recent study that found a relatively protective role of Aβ40 in amyloid deposition (Kim et al., 2007).
Haass closed his talk with a cautionary note about γ-secretase modulation. This drug development approach is being actively pursued at several different biotech and pharmaceutical companies, and Phase 3 clinical trial results about one older such compound, Flurizan, are expected at the ICAD conference to be held this July in Chicago. The premise of γ-secretase modulation grew out of the twin realizations that outright γ-secretase inhibition could be toxic and that certain NSAIDs are able to subtly tweak γ-secretase so as to produce less Aβ42 and more of the presumably innocuous Aβ38. New science is now raising fresh questions for scientists to address while developing this approach. For one, a Keystone speaker claimed that under certain conditions, Aβ38 can aggregate, too (see ARF related news story). For another, Haass presented data from his own lab that support a recent, separate paper by Sascha Weggen and colleagues at Heinrich-Heine-University in Duesseldorf, Germany. Weggen’s data implied not only that certain FAD-based models were unsuitable for finding effective γ-secretase modulators, but also that people carrying severe, early onset FAD mutations were unlikely to respond to γ-secretase modulators even if those drugs do treat the majority of AD cases (Czirr et al., 2007).
For his part, Haass started out testing an underlying assumption of γ-secretase modulation. It is that Aβ38 and 42 generation are connected, i.e., that more of one means less of the other. This turned out to be wrong, Haass said. In data recently published (Page et al., 2008), the scientists found that the age of onset of FAD mutation indeed correlates with the amount of Aβ42, but not with the amount of Aβ38. Extending this observation to NSAIDs and to the γ-secretase inhibitors DAPT and Merck’s compound E, the scientists found that, unlike wild-type presenilin, certain FAD mutants fail to reduce their Aβ42 production in response to these drugs, yet they did respond with a robust increase in Aβ38, showing again how these two products are not connected. As do Weggen’s data on this issue (Czirr et al. 2008), Haass' results also imply that people with strong FAD presenilin mutations should not be treated with NSAIDs, Haass said. “I think the mutant γ-secretase is locked in a pathological conformation,” he added.
This emerging research reinforces a point of caution pharma researchers had raised in prior conversations with this reporter when pressed about why they resisted doing drug trials with families carrying eFAD mutations. Many obstacles, some of them surmountable, hold back such efforts. One scientific concern is that the genetic cause of AD may stand in the way of certain treatment mechanisms, and more research is needed to understand better which treatments stand the best chance of working in families with autosomal-dominant AD (for specific reference, see Drug Trials in EOAD essay; for general overview, see eFAD Research essay). The last word clearly is not spoken on this issue, but for now, it looks as if people with presenilin mutations are well advised to stay away from NSAIDs and look toward immunotherapies or small molecules such as scyllo-inositol instead.—Gabrielle Strobel.
This is Part 3 of a three-part series. See also Part 1 and Part 2.
Paul Coleman
Comments
No Available Comments
Make a Comment
To make a comment you must login or register.