Neuron, Vol. 38, 9–12, April 10, 2003, Copyright 2003 by Cell Press Aph-1, Pen-2, and Nicastrin Minireview with Generate an Active ␥-Secretase Complex

Bart De Strooper* tivity is a major step in the direction of answering this Neuronal Cell Biology and Transfer Laboratory question. Center for Human Genetics The (PS) appear to provide the active core K.U. Leuven and Flanders Interuniversitary Institute of the . Two mammalian homologs, PS1 and -kDa) span the cellular mem 50ف) for Biotechnology PS2, exist. The PS VIB4 Leuven branes several times (Table 2). Two aspartate residues Belgium (Asp257 and Asp385) located in transmembrane do- mains 6 and 7, respectively (Table 2), are essential for the catalytic activity of the protease. Although the working mechanism needs further scrutiny, ␥-Secretase may ␥-Secretase cleaves the Amyloid Precursor therefore indeed be considered an aspartyl protease (APP) in its transmembrane domain, releasing the am- (Wolfe et al., 1999). PS are synthesized as precursor yloid peptide A␤.A␤ is the main constituent of the that must become incorporated into a larger amyloid plaques in the brains of patients suffering from complex for stabilization. The pool that is not incorpo- Alzheimer’s disease. Several other type I integral mem- rated into these complexes is rapidly degraded by the brane proteins are also cleaved by this protease. Re- proteasome. The stabilization of PS is accompanied by cent work indicates that ␥-Secretase is a multiprotein a proteolytic “maturation” cleavage performed by an complex consisting of Presenilin, Nicastrin, Aph-1, and unknown “presenilinase” (Thinakaran et al., 1996). The Pen-2 and that all four proteins are necessary for full kDa) and 30ف ,resulting amino-terminal fragment (NTF -kDa) each contrib 20ف ,proteolytic activity. Since several paralogs and alter- carboxy-terminal fragment (CTF natively spliced variants of at least Presenilin and ute separately one aspartyl residue to the catalytic site. Aph-1 have been identified as well, it is anticipated Both fragments are part of a larger complex. The exact ␥ that -Secretase is not a homogeneous activity. molecular weight of this complex is an issue of debate ␥ -Secretase is an interesting but complex drug target and varies according to the techniques used. The mini- that challenges classical thinking about proteolytic mal estimate is 200–250 kDa (Kimberly et al., 2003), kDa (Edbauer et al., 2002) and even larger 440ف processing and intracellular signaling. but complexes have been described. Obviously this issue needs to be resolved, since it is a decisive piece of ␥-Secretase is the proteolytic activity responsible for information for estimating the minimal number of sub- the cleavage of a series of integral membrane proteins units that are needed to constitute the ␥-Secretase (Table 1), most notoriously the Amyloid Precursor Pro- complex. tein (APP) and Notch. Generally, ␥-Secretase cleaves Using antibodies against the PS fragments, a second the hydrophobic integral membrane domain of its sub- member of the complex, called Nicastrin (Nct), was puri- kDa 130ف strates (except for N-cadherin), resulting in the release fied (Yu et al., 2000). Nct is a glycosylated of protein fragments at the luminal (extracellular) and at integral membrane protein that binds relatively well to the cytoplasmic side of the membrane (Annaert and De both the NTF and the CTF of PS. Nct is synthesized as ف Strooper, 2002). In the case of Notch and some other a 110 kDa precursor protein that needs PS to leave substrates, the released cytoplasmic domains interact the endoplasmic reticulum and to reach the cell surface. with DNA binding proteins and regulate gene transcrip- In PS-deficient cells, the Nct precursor accumulates in tion, linking ␥-Secretase function to a series of signaling the endoplasmic reticulum. Conversely, suppressing processes (Table 1). Whether the small fragments that Nct expression using siRNA results in decreased are shed concomitantly at the luminal side also have steady-state levels of PS, indicating that Nct is one of biological functions (e.g., A␤ in case of APP or N␤ in the stabilizing factors of the PS fragments (Table 3; Edbauer et al., 2002). However, both proteins expressed case of Notch [Okochi et al., 2002]) remains uncertain. together do not suffice to increase ␥-Secretase activity In any event, the A␤ fragment is pathologically important in most cell lines. Therefore, other proteins are rate lim- since it is the main constituent of the amyloid plaques iting for the assembly of the active . Geneticists, in the brain of Alzheimer’s disease patients. For this studying C. elegans, overcame the final hurdles. Goutte reason, inhibiting ␥-Secretase is a therapeutic aim, even and colleagues used a screen for that cause an when doubts are raised concerning the potential side “anterior pharynx-defective phenotype,” reflecting defi- effects associated with long-term suppression of this cient glp-1 signaling (glp-1 and lin-12 are the two Notch activity (Table 1). It is still unknown, at the molecular receptors in C. elegans). They identified two such genes level, whether ␥-Secretase is a homogeneous or a het- called Aph-1 and Aph-2. Aph-2 is the homolog of mam- -kDa multimem 30ف erogeneous activity. If the latter holds true, it is possible malian Nct, while Aph-1 is a novel that different complexes preferentially cleave certain brane spanning protein that, similar to PS, is needed for substrates. This opens the door to novel, more specific the correct subcellular transport of Aph2/Nct to the cell ␥ -Secretase inhibitors. The recent identification of the surface (Goutte et al., 2002). Aph-1 (Pen-1) was also different molecular components that constitute this ac- identified independently in a screen for Presenilin en- hancers that cause a glp-1 sterility in a C. elegans strain *Correspondence: [email protected] partially deficient in PS (Francis et al., 2002). This screen Neuron 10

Table 1. Substrates of ␥-Secretase (Mammalian)

Substrates Priming Cleavage Fragments and Function APP, APLP-1, APLP-2 Bace and ADAM A␤ and AICD (gene regulation?) (constitutively) Notch 1, 2, 3, and 4 ADAM 17, ADAM 10 (?) N␤ and NICD (gene regulation) (ligand-induced) ErbB-4 Metalloproteases B4-ICD (function?) E-cadherin ? ICD (disassembly of E-cadherin/catenin complex) LRP ? ICD (function?) Nectin-1-␣ ? ICD (function?) Delta ADAM? ICD (function?) Jagged ADAM? ICD (function?) CD44 Metalloproteases CD44␤, ICD (function?) yielded in addition the fourth ␥-Secretase partner, and cell-free assays (Kimberly et al., 2003; Takasugi et Pen-2. Pen-2 is a small, hairpin-like membrane protein al., 2003; Edbauer et al., 2003). Thus, it seems that the -kDa (Table 2). minimal number of components needed for the proteo 12ف with Mr Aph-1 and Pen-2, like PS and Nct, are essential for lytic activity of the complex have been identified, Pen-2 normal Notch (glp-1 and lin-12) signaling in C. elegans and Aph-1 being apparently the long-sought “limiting (Table 3). What is, however, the evidence that they are cellular factors” controlling PS expression (Thinakaran also directly involved in the ␥-Secretase activity itself? et al., 1996). Also, the genetic screens in C. elegans Genetics provide a partial answer. Using constructs that indicate that the core of the complex is now probably directly drive expression of the Notch intracellular do- identified: only mutants of these four genes and their main, Francis et al. (2002) demonstrated that Aph-1 and homologs (Table 2) but no other new ones were identi- Pen-2 act at or upstream of the release of the Notch fied in the extensive screens by Francis et al. (2002). intracellular domain, like Presenilin does. Downregula- In mammalian species the situation is, as usual, more tion of one of the two new proteins in cells in culture complicated, as several paralogs of the individual proto- using siRNA leads to a decline in ␥-Secretase activity type proteins and a series of alternative spliced forms (Francis et al., 2002; Takasugi et al., 2003; Lee et al., of Aph-1a (Gu et al., 2003) have been identified (Table 2002), comparable to what was demonstrated before 2). It should be noticed that, in addition, other proteins with Nct (Edbauer et al., 2002) and Presenilin (De or lipids could be rate-limiting for the overall activity of Strooper et al., 1998) (Table 3). Thus, all four proteins the complex. are needed for cleavage of Notch and APP substrates. From the loss-of-function and overexpression experi- Conversely, overexpression of any combination of three ments performed in different species (Table 3), it is obvi- proteins does not increase processing of APP, while ous that the four basic components of the ␥-Secretase overexpressing the four proteins together results con- activity influence each other’s stability and maturation comitantly in the processing and stabilization of PS, the (Gu et al., 2003; Kimberly et al., 2003; Lee et al., 2002; increased expression of fully glycosylated Nct, and a Luo et al., 2003; Steiner et al., 2002; Takasugi et al., clear enhancement of ␥-Secretase activity in cell-based 2003; Edbauer et al., 2003). The effects can be explained

Table 2. The ␥-Secretase Complex Members Name (Generic) C. elegans Mammalian Functions

Presenilin HOP-1, SEL-12 PS1, PS2 Catalytic subunit and assembly of (PSEN in Drosophila) ␥-secretase (glycosylation of Nct, Pen-2 expression); ␤-catenin phosphorylation

Nicastrin APH-2 Nct Assembly of ␥-secretase (stabilization of PS, Pen-2)

Aph-1 APH-1 (PEN-1) Aph1aL and Aph1aS; Assembly of ␥-secretase (stabilization Aph1b (Aph1c in of PS/Nct) rodents)

Pen-2 PEN-2 Pen-2 Assembly of ␥-secretase (proteolytic processing PS) Minireview 11

Table 3. Loss- and Gain-of-Function Effects of the Different ␥-Secretase Subunits

Gain of Function (Overexpression in Cells) Loss of Function (in Cells) Loss of Function (Animals)

PS PS full-length accumulation, Pen-2↓, Nct glycosylation↓, and Notch phenotype (C. elegans, no stabilization cell surface transport Nct↓ D. melanogaster, M. musculus) Nct Unglycosylated Nct accumulation PS↓; Pen-2↓ Notch phenotype (C. elegans, D. melanogaster, M. musculus) Aph-1 PS full-length accumulation, PS↓, cell surface transport Nct↓ Notch phenotype (C. elegans) stabilization Nct? Pen-2 Processing of stabilized PS Accumulation of full-length PS Notch phenotype (C. elegans, M. musculus) in different ways: each component could be essential Overall, Aph-1 seems to be least sensitive to changes for one step in the activation of the protease, in which in expression of the other proteins, and therefore could case none of them needs to be a partner in the final be considered as the scaffold of the complex that binds active complex. However, the available evidence, al- PS and Nct. Downregulation of PS affects the stability though still somewhat shaky, suggests the alternative of Pen-2 and the glycosylation of Nct and, vice versa, possibility that the four proteins are subunits of a larger, downregulation of Nct affects PS stability. This suggests relatively stable active complex. Thus, in material pre- that PS and Nct need each other to leave the endoplas- cipitated from cell extracts using antibodies against one mic reticulum, probably in concert with Aph-1. Finally, of the subunits, all other subunits are reproducibly de- Pen-2 downregulation results in the accumulation of full- tected. Unfortunately, all such experiments made use length PS, but in a stable conformation (Takasugi et of Western blotting to demonstrate the presence of the al., 2003). Thus, Pen-2 seems to be required for the associated proteins, making it hard to know how specific “Presenilinase” cleavage of PS once PS becomes incor- and quantitative the observed coprecipitations are. This porated in the complex with Aph-1 and Nct. Further assay also does not exclude the possibility that a mix of work is needed to know how Pen-2 controls this “Pre- trimeric complexes is precipitated. Moreover, relatively senilinase” and to understand the kinetics of complex mild detergents were used, and the studies were per- formation. Also, the subcellular compartments in which formed with cells that strongly overexpress the proteins. the different subunits bind each other remain to be in- This raises concerns regarding their correct subcellular vestigated. It seems unlikely that the three proteins as- localization before solubilization of the cells and also sociated to PS only function in the global maturation of increases the possibility of nonspecific interactions. the PS protease. Maybe they are also important for the Kimberly et al. (2003) provided an interesting alternative regulation of the activity of the complex and the recogni- for the coimmuneprecipitation experiments using an af- tion of the different ␥-Secretase substrates. finity column that retains only the active form of ␥-Secre- Now that several pieces of the puzzle are coming tase. Apparently all four components were retained and together, it is important to tackle the question of the coeluted, in support of the hypothesis that they are all molecular mechanism behind the clinical mutations in part of one enzymatic complex. However, here again, PS causing Alzheimer’s disease. As amply shown, the cells strongly overexpressing proteins were used as more than one hundred different missense mutations in starting material. Quantitative analysis of the degree of PS1 and PS2 that cause familial Alzheimer’s disease purification of the activity and of the coenrichment of (http://molgen-www.uia.ac.be/ADMutations/) all cause the putative ␥-Secretase components was also lacking. a similar defect, i.e., an increase in the ratio of the A␤42/

By using this approach in combination with additional A␤40-species. The A␤42 peptide is, like A␤40, generated purification steps, it should be possible in principle to by ␥-Secretase cleavage of APP and contains two amino purify ␥-Secretase to homogeneity from cells that do acids more at its carboxyl terminus. This peptide has not overexpress proteins. This will allow researchers to increased aggregation properties and is believed to trig- really evaluate whether the different proteins are stably ger the pathogenic cascade leading to neurodegenera- associated and to settle definitively whether any other tion in Alzheimer’s disease. Since a loss of PS function factor is needed to generate the active complex. inhibits the generation of both types of A␤ (De Strooper Awaiting this further work, we propose that the working et al., 1998), it is tempting to consider the mutations hypothesis of four proteins acting together in one stable in PS as “gain-” or “change-” of-function mutations. complex is reasonable, although not entirely proven. However, it is unclear how the many different mutations Further work is also needed to better understand the (indicated by dots in Table 2) scattered over the protein molecular tasks of the different subunits. The loss- and could all cause such a similar “change” of function. It gain-of-function experiments that were published over is not unlikely that the PS mutations cause subtle de- the last months (summarized in Table 3) overall show fects in the quaternary structure of the ␥-Secretase com- that any change in the expression level of one of the plex, which could even result in loss of function toward individual components affects the stability, the proteo- other substrates than APP. The fact that human PS1 lytic maturation, the glycosylation, or the cell surface is able to rescue PS deficiency in C. elegans and in transport of the other subunits (Gu et al., 2003; Kimberly mammalian cultured cells, while some PS mutants can- et al., 2003; Lee et al., 2002; Luo et al., 2003; Steiner et not, would argue for this possibility, at least in regard al., 2002; Takasugi et al., 2003; Edbauer et al., 2003). to Notch processing (Baumeister et al., 1997; Levitan et Apparently all four proteins closely regulate each other. al., 1996; Song et al., 1999). Neuron 12

As already mentioned, ␥-Secretase cleaves quite a Leuven, the European Union (DIADEM-QLK3-CT-2001-02362), and broad range of substrates with a relaxed specificity. In the Federal Office for Scientific, Technical, and Cultural Affairs, Bel- fact, it looks like ␥-Secretase cleaves almost by default gium (IUAP P5/19). I would like to thank Dr. Bassem Hassan and Dr. Dieter Hartmann for their critical reading and suggestions. any type I integral membrane protein whose ectodomain is shorter than a certain number of amino acid residues Selected Reading (Struhl and Adachi, 2000). Regulation of the process (in case of Notch) occurs at the level of a first “priming” Annaert, W., and De Strooper, B. (2002). Annu. Rev. Cell Dev. Biol. cleavage step performed by members of the metallopro- 18, 25–51. tease family ADAM (Table 1). From the novel data, a Baumeister, R., Leimer, U., Zweckbronner, I., Jakubek, C., Grunberg, much more complicated picture is, however, emerging. J., and Haass, C. (1997). Genes Funct. 1, 149–159. If the total molecular weight of the individual subunits De Strooper, B., Saftig, P., Craessaerts, K., Vanderstichele, H., Guhde, G., Annaert, W., Von Figura, K., and Van Leuven, F. (1998). is taken together, a close approximation of the estimate Nature 391, 387–390. for the minimal molecular weight of the intact complex, Donoviel, D.B., Hadjantonakis, A.K., Ikeda, M., Zheng, H., Hyslop, i.e., 200–250 kDa, is obtained (Kimberly et al., 2003). P.S., and Bernstein, A. (1999). Genes Dev. 13, 2801–2810. This implies a 1:1:1:1 stoichiometry. Therefore, taking Edbauer, D., Winkler, E., Haass, C., and Steiner, H. (2002). Proc. into account the two mammalian PS and the two mam- Natl. Acad. Sci. USA 99, 8666–8671. malian Aph-1 homologs, the existence of at least four Edbauer, D., Winkler, E., Regula, J.T., Pesold, B., Steiner, H., and different ␥-Secretase complexes in mammalian species Haass, C. (2003). Nat. Cell Biol., in press. can be inferred. This estimate is a conservative one, Francis, R., McGrath, G., Zhang, J., Ruddy, D.A., Sym, M., Apfeld, since it does not take into account the alternatively J., Nicoll, M., Maxwell, M., Hai, B., Ellis, M.C., et al. (2002). Dev. Cell spliced forms of PS and Aph-1a (for which a long and 3, 85–97. a short variant have been identified [Gu et al., 2003; Goutte, C., Tsunozaki, M., Hale, V.A., and Priess, J.R. (2002). Proc. Table 2]). The higher molecular weight estimates for the Natl. Acad. Sci. USA 99, 775–779. complex reported by other groups, would, obviously, Gu, Y., Chen, F., Sanjo, N., Kawarai, T., Hasegawa, H., Duthie, M., imply an even higher degree of molecular complexity. Li, W., Ruan, X., Luthra, A., Mount, H.T., et al. (2003). J. Biol. Chem. 9, 7374–7380. In any event, if different complexes can be generated, the question is whether different complexes cleave dif- Herreman, A., Hartmann, D., Annaert, W., Saftig, P., Craessaerts, K., Serneels, L., Umans, L., Schrijvers, V., Checler, F., Vandersti- ferent substrates, and what cellular and physiological chele, H., et al. (1999). Proc. Natl. Acad. Sci. USA 96, 11872–11877. parameters determine the assembly and disassembly Kimberly, W.T., LaVoie, M.J., Ostazewski, B.L., Ye, W., Wolfe, M.S., of the different ␥-Secretase complexes. At the biological and Selkoe, D.J. (2003). Proc. Natl. Acad. Sci. USA, in press. level, some indications for such complexity do indeed Lee, S.F., Shah, S., Li, H., Yu, C., Han, W., and Yu, G. (2002). J. Biol. exist. For instance, the gene inactivation studies of PS1 Chem. 277, 45013–45019. and PS2 in mice indicate that the two mammalian PS Levitan, D., Doyle, T.G., Brousseau, D., Lee, M.K., Thinakaran, G., have overlapping but different functions. Mice with inac- Slunt, H.H., Sisodia, S.S., and Greenwald, I. (1996). Proc. Natl. Acad. tivated PS1 are lethal, displaying a partial Notch signal- Sci. USA 93, 14940–14944. ing deficiency, while the mice with inactivated PS2 ap- Luo, W.J., Wang, H., Li, H., Kim, B.S., Shah, S., Lee, H.J., Thinakaran, pear largely normal. The double-deficient mouse is early G., Kim, T.W., Yu, G., and Xu, H. (2003). J. Biol. Chem. 278, 7850– embryonic lethal, displaying a severe Notch phenotype 7854. (Donoviel et al., 1999; Herreman et al., 1999). Interest- Okochi, M., Steiner, H., Fukumori, A., Tanii, H., Tomita, T., Tanaka, T., Iwatsubo, T., Kudo, T., Takeda, M., and Haass, C. (2002). EMBO ingly, the many genetic crosses performed in C. elegans J. 21, 5408–5416. also suggest that different combinations of the proteins Song, W., Nadeau, P., Yuan, M., Yang, X., Shen, J., and Yankner, could have different biochemical properties. For in- B.A. (1999). Proc. Natl. Acad. Sci. USA 96, 6959–6963. stance, loss-of-function mutations of Pen-2 and Aph-1 Steiner, H., Winkler, E., Edbauer, D., Prokop, S., Basset, G., Yama- in combination with partial loss-of-function mutations saki, A., Kostka, M., and Haass, C. (2002). J. Biol. Chem. 277, 39062– of Nct or the PS homologs in C. elegans (Table 2) cause 39065. not entirely overlapping glp-1- or lin-12-related pheno- Struhl, G., and Adachi, A. (2000). Mol. Cell 6, 625–636. types (Francis et al., 2002). Obviously it is also possible Takasugi, N., Tomita, T., Hayashi, I., Tsuruoka, M., Niimura, M., that this variability reflects different expression levels Takahashi, Y., Thinakaran, G., and Iwatsubo, T. (2003). Nature 422, of the proteins in diverse tissues. Further work is clearly 438–441. needed to address this important question in further Thinakaran, G., Borchelt, D.R., Lee, M.K., Slunt, H.H., Spitzer, L., detail, especially if drugs have to be designed that spe- Kim, G., Ratovitsky, T., Davenport, F., Nordstedt, C., Seeger, M., et cifically interfere with the ␥-Secretase cleavage of APP. al. (1996). Neuron 17, 181–190. In conclusion, a major step forward has been made Wolfe, M.S., Xia, W., Ostaszewski, B.L., Diehl, T.S., Kimberly, W.T., in the field by identifying the main molecular partners and Selkoe, D.J. (1999). Nature 398, 513–517. constituting the “minimal” core of the ␥-Secretase com- Yu, G., Nishimura, M., Arawaka, S., Levitan, D., Zhang, L., Tandon, A., Song, Y.Q., Rogaeva, E., Chen, F., Kawarai, T., et al. (2000). plex. In the years to come, we will see an increasing Nature 407, 48–54. understanding of the complexity of this enzymatic activ- ity. To this aim, studies should focus on the complexes expressed under physiologically relevant conditions.

Acknowledgments

Research in the laboratory was supported by a Pioneer Grant from the Alzheimer’s Association and by the FWO, the VIB, the K.U.