Inhibition of the γ-secretase enzyme that snips amyloid precursor protein (APP) to form Aβ has long been seen as a therapeutic option for Alzheimer disease, but finding a safe, effective inhibitor has proved frustrating. The risk of side effects from γ-secretase inhibition is high, in part because this secretase also cuts Notch, an essential protein for numerous biological functions. The recent cancellation of a high-profile γ-secretase inhibitor clinical trial (see ARF related news story) is the latest disappointment for approaches targeting this secretase. A paper in yesterday’s Nature online offers a new tack for this field by reporting the discovery of a γ-secretase activating protein (GSAP) that acts specifically to promote the binding of γ-secretase to APP, but not to Notch. Researchers led by Paul Greengard at the Rockefeller University in New York show that inhibition of GSAP reduces the production of Aβ by 40 to 50 percent both in vitro and in AD model mice, while having no effect on Notch processing. This suggests that GSAP inhibition could be a promising therapeutic approach to lowering amyloid levels while avoiding toxic side effects. Although the results still need to be replicated by other labs, the news has generated excitement in the field.

“This is a tour de force work that goes from identifying a protein, to making a mouse model where [the authors] actually show that this is a potential target for therapeutics,” said Gopal Thinakaran, of the University of Chicago in Illinois. “This is a very elegant set of studies, and the data are compelling.”

Several other proteins that modulate γ-secretase activity have been identified in recent years, including transmembrane protein TMP21 (see ARF related news story on Chen et al., 2006), G-coupled protein receptor 3 (see ARF related news story on Thathiah et al., 2009), and the γ-secretase complex component Aph1 (see ARF related news story on Serneels et al., 2009). What distinguishes GSAP, said Thinakaran, is that the authors have described a detailed mechanism of action for it, making it an attractive drug target.

Previous work by Greengard and colleagues had shown that the anti-cancer drug imatinib, also known as Gleevec, acted as a γ-secretase modulator, lowering production of Aβ38, 40, and 42 without affecting Notch processing (see ARF related news story on Netzer et al., 2003). Imatinib itself is not a good AD drug candidate because it does not enter the brain. Therefore Greengard and colleagues set out to understand how it blocked the secretase complex so they could develop a better drug. First author Gen He added a radiolabeled, photoactivatable nitrogen group to imatinib, then incubated this derivative with a membrane preparation containing the γ-secretase complex. When He and colleagues illuminated this preparation with ultraviolet light, the nitrogen bonds broke, creating reactive free radicals that bound to any proteins in the vicinity and radiolabeled them. Surprisingly, none of the known γ-secretase components bound the radiolabel, but a novel 16 kilodalton protein did. After purification, mass spectrometry identified this small protein as the C-terminal region of an uncharacterized protein known as pigeon homologue protein, which the authors renamed γ-secretase activating protein (GSAP).

He and colleagues characterized GSAP by using co-immunoprecipitations, showing that it binds both the γ-secretase complex and the membrane-bound, β-secretase-cleaved C-terminal fragment of APP. An experiment using truncated forms of APP showed that GSAP binds a region of about 10 amino acids that lies just next to the membrane on the cytoplasmic side of the precursor. Other experiments demonstrated that imatinib inhibits γ-secretase by interfering with the binding of GSAP to APP. In contrast, He and colleagues showed that GSAP does not bind Notch or affect Notch cleavage, explaining how imatinib spares Notch processing.

The crucial question, however, is what effect GSAP inhibition has on Aβ production. The authors addressed this first in cell cultures, where they used short interfering RNA to knock down GSAP expression by about three-quarters. This lowered production of Aβ38, 40, and 42 by about 50 percent each. Imatinib had no additional effect on Aβ production, confirming that the drug acts through GSAP. The authors then looked at the effects of blocking the activating protein in vivo by making a transgenic mouse that expressed short hairpin RNA for GSAP under a tetracycline inducible promoter. He and colleagues crossed these mice with double-transgenic animals carrying both APP with the Swedish mutation and a presenilin mutation (APPswe/PS1ΔE9). Induction of the interfering RNA for six months reduced GSAP RNA levels by 85 percent. Levels of Aβ40 and 42 dropped some 40 percent, as did the number of amyloid plaques (see figure below). These effects on amyloid were similar to those seen in AD mice treated with a global γ-secretase inhibitor, dibenzazepine, but without the toxic side effects created by faulty Notch processing.



GSAP Promotes Aβ Production
In double-transgenic APP/PS1 mice, GSAP knockdown (right) attenuates Aβ production and plaque deposition in comparison to controls (left). Image credit: Nature Publishing Group

This demonstration of in vivo relevance is significant, said Michael Wolfe of Harvard Medical School. “Assuming that this is validated, it’s quite an important advance. You always increase your chances of finding better [therapeutic] agents when you know what the target is.”

Nonetheless, numerous questions remain about GSAP’s mode of action. Thinakaran said it would be important to know whether GSAP is bound to γ-secretase throughout the neuron, or whether there are specific subcellular compartments that favor the interaction, particularly in light of data showing that presynaptically as well as post-synaptically released Aβ may harm synapses. Another question is whether GSAP binds γ-secretase and APP in a one-to-one fashion. Greengard said his team believes the three proteins act as a trimer. To prove this, however, would require more detailed molecular modeling, Thinakaran said.

One notoriously controversial question in the field concerns the γ-secretase cleavage sites on APP. The transmembrane cleavage at the γ site releases toxic Aβ species, while a few amino acids farther toward the C-terminus, the ε site cleavage is the classic way to produce the APP intracellular domain (AICD), which may have important signaling roles. Some work suggests that these cleavages happen sequentially, first at the ε site and then at γ (see Gu et al., 2001; Lefranc-Jullien et al., 2006; Takami et al., 2009). This scenario was supported by a recent study showing that autoproteolysis of presenilin, the catalytic subunit of γ-secretase, occurs in a stepwise fashion (see ARF related news story). Other papers suggest, however, that the two types of cleavage can occur independently of each other (see Wiley et al., 2005; Kume et al., 2006; Bentahir et al., 2006). He and colleagues uncovered evidence in favor of independent cleavage. The Rockefeller scientists showed, in an in vitro assay, that adding recombinant GSAP increased the levels of Aβ produced by γ cleavage, while decreasing the amount of AICD produced by ε cleavage. Greengard said the cleavage sites were distinguished by the use of antibodies specific for each cleavage fragment.

If independent cleavage is occurring, as the paper suggests, some scientists wonder what the downstream consequences of this would be, and what becomes of the unusual fragments produced. Wolfe points out that, “Whenever you produce an Aβ [fragment], you produce AICD. It’s a one-to-one correspondence.” Wolfe would like to see what happened to the corresponding Aβ and AICD fragments that were not detected in this assay, i.e., long Aβ49 produced by ε cleavage, and long AICD produced by γ cleavage. Bart De Strooper, of the University of Leuven, Belgium, also finds this issue puzzling. He noted, “The question of what happens with Aβ49 has surprisingly not been addressed in the paper” (see full comment below). In an accompanying News & Views article, Peter St George-Hyslop and Gerold Schmitt-Ulms at the University of Toronto ask, “Are these amino-terminal fragments left in the membrane as Aβ49, or are they degraded by some other mechanism that does not generate disease-associated amyloid-β peptides?” Thinakaran speculates that “maybe the γ-secretase complex is capable of both sequential and independent cleavage, and this difference between the two modes of cleavage is dictated by the presence or absence of GSAP.”

For the moment, Greengard and colleagues are focused on the therapeutic potential of GSAP. Greengard said they are looking for the protein that cleaves GSAP into its mature form, with the hope that this protein itself might make an appealing therapeutic target. They are also looking for more proteins that bind to GSAP, Greengard explained, to see if GSAP has other functions in the cell that might be affected by knockdown of the modulator. Finally, he stated that they are also collaborating with molecular geneticists to see if GSAP mutations are associated with AD. It is not clear if the researchers have immediate plans to screen for small-molecule GSAP inhibitors. Paul Fraser of the University of Toronto, Canada, who led the discovery of TMP21, commented that this new paper “validates the notion that regulators [of γ-secretase] are going to be important. There has to be some way of keeping a hold on this protease so it doesn’t randomly traipse around and chew up things it shouldn’t.” He said the fact that GSAP “appears to be a positive regulator of γ-secretase opens a door that nobody knew was there before. It’s exciting stuff.”—Madolyn Bowman Rogers.

He G, Luo W, Li P, Remmers C, Netzer WJ, Hendrick J, Bettayeb K, Flajolet M, Gorelick F, Wennogle LP, Greengard P. Gamma-secretase activating protein is a therapeutic target for Alzheimer’s disease. Nature 2010 Sep 2. Abstract

George-Hyslop PS, Schmitt-Ulms G. Selectively tuning gamma-secretase. Nature 2010 Sep 2. Abstract


Make a Comment

To make a comment you must login or register.

Comments on this content

  1. The identification of GSAP as a specific stimulator of APP processing is important in understanding how γ-secretase is regulated and which mechanism enables selection among the various substrates of the γ-secretase activity. Although more than 30 single transmembrane proteins have been reported as γ-secretase substrates, including APP, Notch, and E- and N-cadherins (see Marambaud et al., 2002, and 2003 for cadherins), no consensus sequence is recognized among these substrates. Accordingly, substrate recognition and broad specificity for γ-secretase remain an enigma. It is also noted that the γ-secretase complex detected by glycerol velocity gradient, or blue native-PAGE (Georgakopoulos et al., 1999; Gu et al., 2004; Evin et al., 2005; Kiss et al., 2008), shows an apparent molecular weight larger than 400 kDa, which exceeds the simple sum of the molecular weights of the γ-secretase core components, presenilin (PS)/N- and C-terminal fragments (28 and 18 kDa, respectively), Nicastrin (120 kDa), APH-1 (24 kDa) and PEN-2 (12 kDa). One interesting concept is that adapter proteins act as recruiting factors specific for one, or a subgroup of, substrate(s) and regulate γ-secretase cleavages. We proposed this model in our paper published two years ago (Kouchi et al., 2009).

    We have reported p120 catenin, an isoform of δ-catenin, as such a recruiting protein specific for N- and E-cadherins since it has binding affinity to PS1, as well as to these cadherins, and also mediates PS1-dependent cleavage of these substrates (Kouchi et al., 2009; also see Kiss et al., 2008, for investigations on γ-secretase catenin supercomplex). In this model, various recruiting factors could be indispensible for substrate recognition and explain the high-molecular-weight γ-secretase supercomplex(es) with broad substrate specificity.

    GSAP reported by He et al. seems to be another example of this kind of protein, but specific for the APP substrate. Interestingly, the interaction between p120 catenin and the cadherins involves the juxtamembrane region of the substrate, i.e., cadherin, just as in the case of GSAP and APP-CTFs. And, just like GSAP, p120 catenin plays the role as a bridge between the γ-secretase and the substrate. We have further identified a p120 binding site in PS1, amino acids 330-360 (Kouchi et al., 2009), although the GSAP binding site has yet to be identified. It is possibly in the PS1 CTF.

    Whereas p120 binds to cadherins, but not to APP, expression of p120 catenin not only promoted E-cadherin cleavage but also partially suppressed Aβ and AICD production (Kouchi et al., 2009). We explained this by a possible competition between substrates, i.e., E-cadherin and APP, for limited availability of γ-secretase, but now we can give an alternative interpretation—that p120 catenin and GSAP may competitively bind to the cytoplasmic loop region of PS1CTF. Since cadherins and p120 are implicated in dendritic spine formation and synaptic transmission, it seems to be a crucial issue whether GSAP affects cadherin/p120-catenin/PS interaction in order to validate GSAP as an attractive target for treatment of AD.


    . Transition-state analogue gamma-secretase inhibitors stabilize a 900 kDa presenilin/nicastrin complex. Biochemistry. 2005 Mar 22;44(11):4332-41. PubMed.

    . Presenilin-1 forms complexes with the cadherin/catenin cell-cell adhesion system and is recruited to intercellular and synaptic contacts. Mol Cell. 1999 Dec;4(6):893-902. PubMed.

    . The presenilin proteins are components of multiple membrane-bound complexes that have different biological activities. J Biol Chem. 2004 Jul 23;279(30):31329-36. PubMed.

    . p120-catenin is a key component of the cadherin-gamma-secretase supercomplex. Mol Biol Cell. 2008 Oct;19(10):4042-50. PubMed.

    . p120 catenin recruits cadherins to gamma-secretase and inhibits production of Abeta peptide. J Biol Chem. 2009 Jan 23;284(4):1954-61. PubMed.

    . A CBP binding transcriptional repressor produced by the PS1/epsilon-cleavage of N-cadherin is inhibited by PS1 FAD mutations. Cell. 2003 Sep 5;114(5):635-45. PubMed.

    . A presenilin-1/gamma-secretase cleavage releases the E-cadherin intracellular domain and regulates disassembly of adherens junctions. EMBO J. 2002 Apr 15;21(8):1948-56. PubMed.

Comments on Primary Papers for this Article

  1. Regulated intramembrane proteolysis (RIP) is currently the hot topic in the field of proteases, and what this new paper is telling us is that there is more to it than just the four subunits of active γ-secretase, which has become famous for its role in amyloid-β production in Alzheimer disease (AD).

    He et al., from Paul Greengard’s laboratory, describe a novel, so-called γ-secretase activating protein (GSAP) in their current publication. Importantly, the authors show that when it is processed into a 16-kDa fragment, GSAP—the direct target of an anticancer drug that inhibits amyloid formation—directly interacts with the γ-secretase substrate APP. When cellular expression of GSAP was reduced by 72 percent with RNAi, a dramatic reduction of APP-derived amyloid peptides with 38, 40, and 42 residues was observed. Thus, this finding strengthens the idea of substrate targeting to lower amyloid production, which surfaced when NSAIDs were characterized as γ-secretase modulators (GSMs) and found to directly bind to the GxxxG interaction motif of the transmembrane sequence of APP (Kukar et al., 2008; Richter et al., 2010).

    GSAP could have been regarded erroneously as a novel γ-secretase inhibitor, since AICD is reduced. This occurs most likely because it binds to AICD and thereby inhibits its degradation. However, since it affects both product lines and enhances Aβ42, which was suggested to be the precursor of Aβ38, we can be confident that it is a real activator. This explanation is based on the current model of the γ-secretase mechanism. In a consecutive cleavage process, two product lines containing either Aβ40 or Aβ42 are generated in the degradation of remnants of membrane APP. Although we do not know much about GSAP and its 16-kDa active fragment, which makes the story even more complicated, the work clearly shows that an understanding of the fine-tuning of the γ-mediated cleavage by cofactors is very important. This means that results from in vitro assays of the γ-secretase should be considered with great care when they are obtained in the absence of important cofactors. The loss of cofactors (when purified away during the enrichment of the γ-secretase module) may explain why findings from in vitro assays do not always match to those obtained with living cells. Unfortunately, researchers sometimes assume that in vivo results will parallel in vitro findings.

    Given that regulatory factors such as TMP21 (Chen et al., 2006), GSAP, and others exist (for a review see De Strooper and Annaert, 2010), it is evident how dangerous it can be to extrapolate from the minimal composition of a protein module with four subunits to a fully active unit containing all cofactors (many of which might have been otherwise lost during the preparation of the module for the in vitro assay). In addition, other factors, such as the dimerization of the APP-derived substrate β-CTF, have been recognized as an important mechanism to regulate Aβ42 production. Production of Aβ42 is reduced in favor of an increase of Aβ38 when dimerization of the substrate is attenuated (Munter et al., 2007; Munter et al., 2010). This nicely fitted into findings from Yasuo Ihara’s group (see Qi-Takahara et al., 2005 and Takami et al., 2009), from which we have learned that γ-secretase is able to process Aβ49 or Aβ48 at every third residue, leading to the stepwise generation of shorter Aβ peptides representing intermediate products of the γ-secretase-driven degradation process.

    Another point which I would like to stress is that major recent findings regarding the γ-secretase module have been made by classical biochemical approaches. These were based on thorough analyses of interactions with γ-secretase subunits (e.g., by Chen et al., who identified TMP21), or with compounds such as NSAIDs and Gleevec, elegantly used as photoactivatable derivatives (Kukar et al., 2008; He et al., 2010) to characterize binding partners of secretase substrates or cofactors. Together with complementing studies that were oriented to understand enzyme-substrate interactions (Munter et al., 2007; Munter et al., 2010; Richter et al., 2010), the knowledge of the substrate conformation (i.e., dimer or monomer) and an understanding of regulatory factors such as GSAP will help to find answers to questions regarding the physiology and the structure/ function relation within the γ-secretase module. All this tells us that there is an additional level of activity modulation that is more complex than previously thought and which has a high potential for the development of novel therapeutic strategies, which are urgently needed for AD.

    View all comments by Gerd Multhaup
  2. I am very enthusiastic about this paper. The authors identify a 16-kDa protein that apparently binds to APP-CTF at the juxtamembrane area—close to the ε cleavage site—and which is able to modulate the different cleavages of the APP-CTF. Blocking the interaction with APP-CTF (either using siRNA or imatinib) lowers Aβ generation and increases AICD formation. The protein does not seem to affect Notch signaling (in vitro or in vivo), although further work is needed to see whether it affects other substrates or whether the effect is entirely APP selective. In addition, if GSAP inhibition stimulates AICD formation and inhibits γ-cleavage, then long Aβ49 should be observed. The question of what happens with Aβ49 has surprisingly not been addressed in the paper.

    The fact that its activity can be modulated by imatinib provides proof of concept that the protein is a drug target (although it will not be easy to generate drugs that are specific enough and brain permeable to interfere with the proposed protein-protein interaction). Interestingly, several γ-secretase modulators have been proposed to bind also to APP-CTF and thus modulate γ-cleavages of this substrate. It is unclear whether they act via a similar mechanism to γ-secretase activating protein.

    Many other exciting questions remain to be answered in follow-up studies: How precisely does this protein interact with γ-secretase? How do mutations in presenilin affect this interaction? What is the role of the large precursor protein of GSAP, and are the proteases that process this precursor candidate drug targets themselves?

    View all comments by Bart De Strooper
  3. In this paper, He and colleagues work revealed that the γ-secretase processing mechanisms of Aβ generation from APP C-terminal fragments (CTFs) by γ-site cleavage were distinct from the mechanisms of ε-site cleavage. This result is consistent with our previous reports (Kume et al., 2004; Kume and Kametani, 2006). These show that APP CTFs are cleaved in two γ-secretase processing pathways and clarify the relationship between the processing sites and their products in these pathways.

    One is the so-called γ-secretase regulated pathway. γ-secretase is responsible for processing not only APP CTFs, but also various type 1 membrane proteins, including Notch and cadherins (Wolfe and Kopan, 2004). In general, this cleavage occurs near the cytoplasmic membrane boundary region and releases the intracellular cytoplasmic domain for intracellular signaling. In APP CTFs, γ-secretase cleaves at the ε-site near the cytoplasmic membrane boundary region and produces AICDε (C50) (Gu et al., 2001; Sastre et al., 2001; Weidemann et al., 2002). As previously reported, ε-site cleavage preferentially occurs on the α-secretase processing product C83, which is the major APP CTF (Kume et al., 2004; Kume and Kametani, 2006). Furthermore, it was recently reported that the β-secretase processing product C99, which is the minor APP CTF, is an inefficient substrate for proteolysis by γ-secretase (Funamoto et al., 2010). Therefore, in the γ-secretase processing pathway, trace amounts of Aβ49 may be produced from C99, while AICDε (C50) is mostly produced from C83. Long p3 is also produced in this pathway (Kametani, 2004).

    The other regulatory pathway is via GSAP regulation processing. As this paper describes, this processing involves GSAP binding to the cytoplasmic domain of APP CTFs. GSAP/γ-secretase/APP CTF ternary complex alters the structural relationship between γ-secretase and APP CTFs. In the presence of GSAP, γ-secretase may directly cleave at γ-site in the middle of the membrane domain of APP CTFs. Thus, γ-site processing may produce Aβ from C99, p3 from C83, and AICDγ (C57/59) from C99 and C83.

    Under normal conditions, γ-secretase processing of APP CTFs occurs in these two pathways. An increased level of expression of GSAP induces an increase in Aβ production and a decrease in AICDε production, according to the authors. Therefore, GSAP binding to the cytoplasmic domain of APP CTF and GSAP regulated processing may act upstream of γ-secretase processing. Aβ may be mostly produced by the GSAP regulated processing, and GSAP regulated processing is a key event of AD.

    View all comments by Fuyuki Kametani


News Citations

  1. Lilly Halts IDENTITY Trials as Patients Worsen on Secretase Inhibitor
  2. First Look at the Secretase, New Kid on the Block
  3. Big Haul? A G Protein-coupled Receptor Regulates Aβ Production
  4. Double Paper Alert—Keystone Presentations Now in Press
  5. Gleevec for Alzheimer's?
  6. Research Brief: Presenilin Simplicity—Evidence for Autoproteolysis

Paper Citations

  1. . TMP21 is a presenilin complex component that modulates gamma-secretase but not epsilon-secretase activity. Nature. 2006 Apr 27;440(7088):1208-12. PubMed.
  2. . The orphan G protein-coupled receptor 3 modulates amyloid-beta peptide generation in neurons. Science. 2009 Feb 13;323(5916):946-51. PubMed.
  3. . gamma-Secretase heterogeneity in the Aph1 subunit: relevance for Alzheimer's disease. Science. 2009 May 1;324(5927):639-42. PubMed.
  4. . Gleevec inhibits beta-amyloid production but not Notch cleavage. Proc Natl Acad Sci U S A. 2003 Oct 14;100(21):12444-9. PubMed.
  5. . Distinct intramembrane cleavage of the beta-amyloid precursor protein family resembling gamma-secretase-like cleavage of Notch. J Biol Chem. 2001 Sep 21;276(38):35235-8. PubMed.
  6. . APPepsilon, the epsilon-secretase-derived N-terminal product of the beta-amyloid precursor protein, behaves as a type I protein and undergoes alpha-, beta-, and gamma-secretase cleavages. J Neurochem. 2006 May;97(3):807-17. PubMed.
  7. . gamma-Secretase: successive tripeptide and tetrapeptide release from the transmembrane domain of beta-carboxyl terminal fragment. J Neurosci. 2009 Oct 14;29(41):13042-52. PubMed.
  8. . Familial Alzheimer's disease mutations inhibit gamma-secretase-mediated liberation of beta-amyloid precursor protein carboxy-terminal fragment. J Neurochem. 2005 Sep;94(5):1189-201. PubMed.
  9. . Abeta 11-40/42 production without gamma-secretase epsilon-site cleavage. Biochem Biophys Res Commun. 2006 Nov 3;349(4):1356-60. PubMed.
  10. . Presenilin clinical mutations can affect gamma-secretase activity by different mechanisms. J Neurochem. 2006 Feb;96(3):732-42. PubMed.
  11. . Gamma-secretase activating protein is a therapeutic target for Alzheimer's disease. Nature. 2010 Sep 2;467(7311) PubMed.
  12. . Alzheimer's disease: Selectively tuning gamma-secretase. Nature. 2010 Sep 2;467(7311):36-7. PubMed.

External Citations

  1. APPswe/PS1ΔE9

Further Reading


  1. . Gamma-secretase activating protein is a therapeutic target for Alzheimer's disease. Nature. 2010 Sep 2;467(7311) PubMed.
  2. . Alzheimer's disease: Selectively tuning gamma-secretase. Nature. 2010 Sep 2;467(7311):36-7. PubMed.


  1. Lilly Halts IDENTITY Trials as Patients Worsen on Secretase Inhibitor
  2. First Look at the Secretase, New Kid on the Block
  3. Research Brief: Presenilin Simplicity—Evidence for Autoproteolysis
  4. Big Haul? A G Protein-coupled Receptor Regulates Aβ Production
  5. Double Paper Alert—Keystone Presentations Now in Press
  6. Gleevec for Alzheimer's?

Primary Papers

  1. . Gamma-secretase activating protein is a therapeutic target for Alzheimer's disease. Nature. 2010 Sep 2;467(7311) PubMed.
  2. . Alzheimer's disease: Selectively tuning gamma-secretase. Nature. 2010 Sep 2;467(7311):36-7. PubMed.