The first proline of PALP motif at the C terminus of presenilins is obligatory for stabilization, complex formation, and gamma-secretase activities of presenilins.

Mutations in presenilin (PS) genes cause early-onset familial Alzheimer's disease by increasing production of the amyloidogenic form of amyloid beta peptides ending at residue 42 (Abeta42). PS is an evolutionarily conserved multipass transmembrane protein, and all known PS proteins contain a proline-alanine-leucine-proline (PALP) motif starting at proline (P) 414 (amino acid numbering based on human PS2) at the C terminus. Furthermore, missense mutations that replace the first proline of PALP with leucine (P414L) lead to a loss-of-function of PS in Drosophila melanogaster and Caenorhabditis elegans. To elucidate the roles of the PALP motif in PS structure and function, we analyzed neuro2a as well as PS1/2 null fibroblast cell lines transfected with human PS harboring mutations at the PALP motif. P414L mutation in PS2 (and its equivalent in PS1) abrogated stabilization, high molecular weight complex formation, and entry to Golgi/trans-Golgi network of PS proteins, resulting in failure of Abeta42 overproduction on familial Alzheimer's disease mutant basis as well as of site-3 cleavage of Notch. These data suggest that the first proline of the PALP motif plays a crucial role in the stabilization and formation of the high molecular weight complex of PS, the latter being the active form with intramembrane proteolytic activities.

Alzheimer's disease (AD) 1 is a progressive dementing neurodegenerative disorder in the elderly characterized pathologically by the presence of senile plaques and neurofibrillary changes in the brains of affected individuals (reviewed in Ref. 1, and references therein). Senile plaques are composed of amyloid ␤ peptides (A␤) comprising ϳ40 amino acids that are proteolytically produced from ␤-amyloid precursor protein (␤APP). ␤APP is initially cleaved by ␤-secretase to generate a 99-residue C-terminal fragment (C99) that then is cleaved by ␥-secretase to generate A␤. A subset of AD is inherited as an autosomal dominant trait (familial AD: FAD). Genetic mutations in ␤APP genes that cosegregate with the clinical manifestations of FAD increase production of the amyloidogenic A␤42 species ending at Ala 42 (2); A␤42, which normally comprises only ϳ10% of total secreted A␤, aggregates much faster than the predominant A␤40 species (3), and A␤42 is the initially and predominantly deposited A␤ species in AD brains (4,5). These data implicated a seminal role of A␤42 in the pathogenesis of AD.
Mutations in presenilin (PS) 1 and PS2 genes are linked to the majority of early onset FAD. FAD-linked PS mutations affect ␥-cleavage of ␤APP leading to an increased production of A␤42 (1). In contrast, ablation of PS1 and PS2 genes in mice completely inhibited production of both A␤40 and A␤42, accompanied by accumulation of the ␤APP C-terminal stubs (i.e. C99 and C83) that are the direct substrates for ␥-secretase (6 -8). Furthermore, studies in Caenorhabditis elegans and Drosophila melanogaster, as well as in knockout mice, suggested that PS facilitates Notch signaling by activating the ligand-induced intramembranous proteolysis of Notch receptors at site-3 to release their cytoplasmic domains (NICD) (7)(8)(9)(10)(11). This cleavage appears to be very similar to ␥-cleavage of ␤APP because it occurs close to or within the membrane, inhibited by inactivation of PS genes, and can be blocked by peptidomimetic ␥-secretase inhibitors (reviewed in Ref. 12). The physiological role(s) of PS in intramembrane proteolysis has been unclear, but it has been shown that two aspartates within the sixth and seventh transmembrane (TM) domains are required for its ␥-secretase activities (13,14), and the amino acid sequences near these aspartates are homologous to those of the active sites of the polytopic aspartyl proteases of bacterial origin (15). Recently, transition state analogue ␥-secretase inhibitors that inhibit aspartic protease(s) have been shown to directly bind fragment forms of PS, strongly suggesting that PS harbors the catalytic center of ␥-secretase (16 -18).
PS1 and PS2 are polytopic integral membrane proteins that span membrane six to eight times and undergo endoproteolysis to generate N-and C-terminal fragments (NTF and CTF, respectively) (19,20). These fragments form heterodimers and are incorporated into the high molecular weight (HMW) protein complexes that are highly stabilized and acquire long half-life (t1 ⁄2 ϭ ϳ20 h), whereas holoproteins are rapidly degraded (t1 ⁄2 ϭ ϳ2 h) (21)(22)(23)(24). Mutagenesis studies showed that the endoproteolysis of PS is not required for its stabilization, complex formation, or function (25)(26)(27). In contrast, recent findings indicate that the formation of stabilized HMW complex of PS fragments is tightly related to its function. CHAPSO-solubilized membrane fractions containing PS1 fragments exhibit ␥-secretase activity (28); nicastrin, a novel single-pass transmembrane protein, has been identified as one of the components of the HMW PS complex that may regulate ␥-secretase activity (29). Moreover, the facts that the amount of stabilized NTF and CTF of PS are tightly regulated and that overexpression of exogenous PS replaces endogenous PS suggest that some cellular factor(s) of limited amount are required for the stabilization and complex formation of PS (19,30).
We and others have shown that the ectopically expressed NTFs of PS1 or PS2 are not stabilized and do not overproduce A␤42 on FAD mutant basis (31)(32)(33). Furthermore, we have found that the integrity of the C terminus of PS is required for its stabilization and abnormal ␥-cleavage of ␤APP (i.e. A␤42 overproduction) (34). PS are evolutionarily conserved proteins that are present in almost all multicellular organisms including vertebrates and invertebrates as well as plants, and the C-terminal region of PS is highly conserved. Notably, we found that all known PS proteins contain an amino acid motif comprised of four consecutive amino acids, proline-alanine-leucineproline, at the C terminus, including C. elegans SPE-4 that harbors a shorter C-terminal region, which we designated PALP motif (Fig. 1A). In C. elegans and Drosophila, many of the loss-of-function mutants of PS defective in Notch signaling have been reported, and the first proline of PALP motif is replaced with leucine in recessive loss-of-function mutants of C. elegans spe-4 (35) and Drosophila Psn (36). In C. elegans spe-4 (eb-12) allele, which results in P440L (corresponding to the first proline of PALP) amino acid substitution of SPE-4 protein, morphogenesis of fibrous body-membranous organelle complexes is defective and spermatogenesis arrests at an unusual cellular stage in a similar manner to a null phenotype, despite robust synthesis of mutant proteins (35). Drosophila Psn (Dps) 46 mutant allele causing P507L amino acid substitution also resulted in reduction of Notch signaling similar to that in a Psn null mutant allele (36). These results suggested that the first proline of PALP motif plays an important role in the ␥-secretase functions of PS. In contrast, it has been documented that the second proline of PALP motif is replaced with serine (37) or glutamine (38) in some PS1 mutant FAD pedigrees. To elucidate the structural and functional roles PALP motif in the C terminus of PS, we analyzed neuro2a and PS1/2 null fibroblast cell lines transfected with PALP-mutated PS and examined their effects on the metabolism of PS proteins as well as on ␥-cleavage of ␤APP and site-3 cleavage of Notch.
Cell Culture and Transfection-Mouse neuro2a (N2a) neuroblastoma cells were maintained as described previously (20). Generation of SV40transformed fibroblasts derived from PS1 Ϫ/Ϫ PS2 Ϫ/Ϫ littermates (40) by B. D. S. will be described elsewhere. Stable N2a cell lines were generated by transfecting cDNAs using LipofectAMINE (Life Technologies, Inc.) and selection in DMEM containing G418 (Life Technologies, Inc., or Calbiochem, San Diego, CA) at 500 g/ml as described (32). N2a cells stably coexpressing Notch⌬E and PS2 were generated by transfecting a cDNA encoding Notch⌬E in pcDNA3.1/hyg(ϩ) in N2a cells stably expressing PS2 and selection in DMEM containing hygromycin at 160 g/ml. Transient transfection was performed by LipofectAMINE or LipofectAMINE 2000 (Life Technologies, Inc.) according to the manufacturer's instructions, and expression of transgenes was driven by addition of 10 mM butyric acid for 12-24 h.
For immunoblot analysis of cell lysates, cells were lysed in 2% SDS sample buffer and briefly sonicated. The samples were separated by SDS-PAGE without previous heating, transferred to polyvinylidene difluoride membrane (Millipore, Bedford, MA), and probed with each antibody as described (20,32,34). The immunoblots were developed using an ECL system (Amersham Pharmacia Biotech) or Immunostar reagents (Wako), and detected on Hyperfilm ECL (Amersham Pharmacia Biotech) or using LAS-1000plus (Fuji). Scanned images were quantitated with Image Gauge (Fuji) or Scion Image (Scion Corp.) software.
To evaluate the half-lives of transfected PS proteins and fragments thereof by blocking total cellular protein synthesis, cultured cells were treated with cycloheximide (30 g/ml) for 12 or 24 h and then analyzed by immunoblotting with appropriate PS antibodies.
Metabolic Labeling and Immunoprecipitation-N2a cells stably coexpressing PS2 and Notch⌬E were starved for 2 h in methionine-and serum-free DMEM (Life Technologies, Inc.) containing 10 mM butyric acid. Cells were subsequently metabolically labeled for 15 min with 6 MBq/ml [ 35 S]methionine (Expre 35 S 35 S, PerkinElmer Life Sciences) and chased for 60 min in complete DMEM containing 10% fetal bovine serum with or without addition of 10 M clasto-lactacystin ␤-lactone (Sigma) in culture media. Cells were then washed three times in cold Tris-buffered saline and lysed in RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) containing Complete protease inhibitor mixture (Roche Diagnostics). Before precipitating, the extracts were incubated with 10 l of protein G-agarose beads (Life Technologies, Inc.) and spun down to remove debris and proteins nonspecifically bound to agarose beads. The supernatants were mixed overnight with 50 l of protein G-agarose beads and a monoclonal antibody anti-c-Myc (9E10) at 4°C. Beads were washed twice with RIPA buffer, and then twice with TS (50 mM Tris-HCl, pH 7.5, 150 mM NaCl), and extracted in SDS sample buffer. Immunoprecipitates were fractionated by SDS-PAGE, and autoradiograms were analyzed using BAS-1800II (Fuji).
Glycerol Velocity Gradient Centrifugation-Glycerol velocity gradient centrifugation was performed as described previously with some modifications (23,41). N2a cells stably expressing PS2 cDNAs were grown to confluence on a 15-cm dish. All of the following steps were carried out at 4°C. Scraped cell pellets were resuspended in 4 ml of homogenization buffer A (10 mM HEPES, pH 7.4, 150 mM NaCl, 10% glycerol, Complete protease inhibitor mixture). Cells were disrupted by a Polytron homogenizer (Hitachi) at power level 4 for 30 s, and nuclei and large cell debris were pelleted by centrifugation at 1,500 ϫ g for 10 min. The postnuclear supernatants were centrifuged for 1 h at 100,000 ϫ g. The vesicle pellets were extracted with 0.5 ml of homogenization buffer A containing 1% CHAPSO for 1 h. The resulting membrane extracts were cleared by centrifugation for 1 h at 100,000 ϫ g. The supernatants were loaded on a 11-ml linear 15-30% (v/v) glycerol gradient in the gradient buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 0.5% CHAPSO). Molecular mass marker proteins (thyroglobulin, 669 kDa; ferritin, 440 kDa; catalase, 232 kDa; lactate dehydrogenase, 140 kDa; bovine serum albumin, 67 kDa; Amersham Pharmacia Biotech) were fractionated in parallel to allow estimation of the molecular mass of the fractions. Sedimentation was carried out for 16 h at 38,000 rpm in a Beckman SW41 rotor, and the gradients were fractionated from the bottom to give 18 -20 fractions of 600 l. 20 l of each fraction was separated by SDS-PAGE for immunoblot analysis.
Subcellular Fractionation by Iodixanol Gradients-Subcellular fractionation was performed using iodixanol as medium according to the previously described method with some modifications (42,43). Cell pellets from a 15-cm dish were resuspended in 4 ml of homogenization buffer B (10 mM HEPES (pH 7.4), 1 mM EDTA, 0.25 M sucrose, Complete protease inhibitor mixture). All of the following steps were carried out at 4°C. Cells were disrupted by a Polytron homogenizer and centrifuged at 1,500 ϫ g for 10 min as in glycerol velocity gradient centrifugation. The postnuclear supernatants were centrifuged for 1 h at 65,000 ϫ g. The vesicle pellets were resuspended in 0.8 ml of homogenization buffer B. Gradients were set up in 13-ml Beckman SW41 centrifuge tubes by diluting iodixanol (Optiprep, 60% w/v, Life Technologies, Inc.) with homogenization buffer B (2.5% iodixanol, 1 ml; 5%, 2 ml; 7.5%, 2 ml; 10%, 2 ml; 12.5%, 0.5 ml; 15%, 2 ml; 17.5%, 0.5 ml; 20%, 0.5 ml; 30%, 0.3 ml). The resuspended vesicle fractions were loaded on the top of the gradients and centrifuged in a SW41 rotor at 40,000 ϫ rpm for 2.5 h. The resulting gradients were collected in 1-ml fractions. 20 l of each fraction was separated by SDS-PAGE for immunoblot analysis.
Quantitation of A␤ by Two-site ELISAs-Two-site ELISAs that specifically detect the C terminus of A␤ were used as described (20,32,34). BNT77, which was raised against human A␤11-28 and recognizes full-length as well as N-terminally truncated A␤, was used as a capture antibody; BNT77 binds human as well as rodent-type A␤, but does not react with the 3-kDa fragment (p3) (44). BA27 and BC05 that specifically recognize the C terminus of A␤40 and A␤42, respectively, were conjugated with horseradish peroxidase and used as detector antibodies (2). Culture media were collected after an appropriate incubation period of 24 h and subjected to BNT77/BA27 or BNT77/BC05 ELISAs as described (20,32,34).

Expression of PS2 Harboring Point Mutation in PALP Motif-Amino acid sequence alignments of PS proteins in different
organisms showed that all known PS proteins contain a highly conserved amino acid motif in the proximal portion of the C terminus, which comprises ϩϩ(A/T)(A/L)PALPX(S/P)(I/L/ V)XXX(G/A)XX(F/C)(Y/C)F ("ϩ" indicates charged residues), and especially, the proline-alanine-leucine-proline sequence was completely conserved, which we designated the PALP motif (Fig. 1A). Moreover, the first and fourth proline residues of PALP motif have been shown to be linked to loss-of-function (35,36) or FAD (37, 38) mutations of PS, respectively. To elucidate the structural and functional roles of each amino acid in the PALP motif, we constructed cDNAs encoding wild-type (wt) or N141I FAD mutant (mt) PS2 harboring P414L (PS2/ P414L) or P417S (PS2/P417S) mutations and stably transfected them in N2a cells. In addition, we established N2a cells stably expressing two other types of cDNAs encoding modified PS2, i.e. PS2/D366A, which replaces the Asp residue in the seventh TM domain and works as a dominant negative mutant on ␥-cleavage of ␤APP and site-3 cleavage of Notch (14); and PS2/⌬loopN, which lacks amino acids 296 -325 in the sixth hydrophilic loop of PS2 including the standard cleavage site (25)(26)(27). Structures of the PS derivatives used in this study are schematically shown in Fig. 1B. Immunoblot analysis of the cell lysates showed that neither PS2/D366A nor PS2/⌬loopN underwent endoproteolysis to give rise to a 35-kDa NTF and a 23-kDa CTF that are normally produced from full-length (FL) PS2 as reported previously ( Fig. 2A, upper panel, arrow) (13,14,(25)(26)(27). wt and mt PS2/P414L also were expressed as holoproteins and did not yield endoproteolytic fragments, whereas PS2/P417S were cleaved to form fragments ( Fig. 2A, upper  panel, arrowheads). Overexpression of exogenous wt FL PS2 or PS2/P417S as well as of PS2/D366A or PS2/⌬loopN resulted in the replacement of endogenous murine PS1 fragments ( Fig. 2A,  lower panel, arrowhead). These data were consistent with the notion that replacement of endogenous PS is not coupled to endoproteolysis (24,25,30). In contrast, replacement of endogenous PS1 did not occur in N2a cells expressing PS2/P414L, suggesting that P414L mutation abolishes the interaction of PS with the cofactor(s) that stabilizes intracellular PS.
To evaluate the half-lives of transfected PS proteins and fragments thereof, we performed a chase experiment upon treatment with cycloheximide that blocks total cellular protein synthesis (24,34). In cycloheximide-treated N2a cells, PS2 holoproteins were rapidly degraded, whereas fragments were stabilized at 24 h of chase. NTF and CTF derived from PS2/ P417S also were highly stabilized (Fig. 2B, upper panel, arrowhead). A portion of PS2/D366A or PS2/⌬loopN was stabilized as holoproteins, suggesting that the abolition of endoproteolysis does not affect stabilization of PS (Fig. 2B, lower panel, arrow). In contrast, PS2/P414L holoproteins were degraded in a simi- lar manner to holoproteins of wt PS2 (Fig. 2B, upper panel,  arrow). The stabilization of PS2/D366A or PS2/⌬loopN holoproteins compared with those of wt PS2, PS2/P414L, or PS2/P417S was confirmed by densitometric analysis (Fig. 2C). These results suggest that the first proline of PALP motif plays an important role in the stabilization and replacement capacities of PS, whereas substitution of the second proline does not affect PS metabolism.
Formation of High Molecular Weight Complex and Subcellular Distribution of PS2 Proteins-PS proteins are known to form a HMW complex with other components (21)(22)(23). To examine the capacity of modified PS2 proteins to form HMW complex, membrane fractions of N2a cells were solubilized in 1% CHAPSO, and extracted proteins were separated on a linear glycerol velocity gradient. Endoproteolytic fragments derived from wt FL PS2 were predominantly present in the 232-443-kDa HMW range, whereas PS2 holoproteins were fractionated in the 140 -232-kDa low molecular weight (LMW) range (Fig. 3A). PS2/D366A and PS2/⌬loopN, which were stabilized but not cleaved, were present as holoproteins broadly within LMW and HMW ranges (Fig. 3, B and C). In contrast, unstable PS2/P414L holoproteins were fractionated exclusively in the LMW range (Fig. 3D). Fragments of PS2/P417S, which were stabilized in a similar fashion to FL PS2, were also re-covered in HMW fractions, whereas holoproteins were present in the LMW range (Fig. 3E).
These data suggest that stabilized PS2 proteins (i.e. endoproteolytic fragments and a portion of holoproteins in uncleavable mutants) participate in the formation of HMW PS2 complexes, which are present in Golgi/TGN. In contrast, unstabilized PS2 proteins (i.e. holoproteins of wt FL PS2, PS2/ P414L, and nascent form of PS2/D366A and PS2/⌬loopN) form LMW PS2 complexes that remain in ER. These observations support the idea that stabilization and transport to Golgi/TGN of PS proteins require an interaction of PS with unknown cellular cofactor(s), and that the first proline of the PALP motif plays an important role in the formation of PS HMW complex.
Effects of PALP Mutations on Site-3 Cleavage of Notch-PS is known to serve as a critical component of Notch signaling by executing the site-3 cleavage of Notch (7)(8)(9)(10)(11). It has been reported that PS2/D366A abrogates the proteolytic cleavage of mouse Notch-1 in mammalian cells and interferes with the Notch signaling in C. elegans in vivo (14). Thus, it is tempting to speculate that the loss-of-function mutations in invertebrate PS may suppress Notch signaling by inhibiting the proteolytic release of NICD. To examine whether PS2/P414L or PS2/ P417S mutations affect the ␥-secretase-like site-3 cleavage activity for Notch, we analyzed the proteolytic release of NICD in N2a cells transiently transfected with a cDNA encoding Notch⌬E that lacks the extracellular domain but retains the TM domain and flanking cytoplasmic region including site-3 ( Fig. 5A) (39). Proteolytic generation of NICD from Notch⌬E was inhibited in N2a cells overexpressing PS2/D366A as reported previously (Fig. 5A, arrowhead) (14). However, PS2/ P414L mutation did not affect the site-3 cleavage of Notch and normal levels of NICD were generated either on wt or mt basis. In contrast, FAD-linked mt FL PS2 (i.e. N141I), PS2/M239V, and PS2/P417S, the latter harboring an amino acid substitution of the second proline in PALP motif at an analogous position to a FAD-linked PS1 P436S mutation (37), suppressed the proteolytic release of NICD. Upon longer exposure, a very small amount of NICD was observed in cells expressing PS2/ M239V or PS2/P417S (lower panel). To further corroborate our observation that the FAD-linked and/or A␤42-overproducing mutations of PS2 affect the proteolytic release of NICD, we generated N2a cell lines stably coexpressing PS2 and Notch⌬E and analyzed the NICD production (Fig. 5B). Overexpression of PS2/D366A completely abolished the production of NICD in stable N2a cells, whereas stable coexpression of mt FL PS2 or PS2/P417S significantly, but not completely, suppressed NICD formation. Pulse-chase analysis using N2a cell lines stably coexpressing wt FL PS2 and Notch⌬E showed that the proteolytic release of NICD occurred at 30 min of chase, whereas the Notch⌬E holoprotein was rapidly degraded (Fig. 5C, upper left  panel). It has been suggested that the proteasomal activity regulates the metabolism and degradation of Notch⌬E and NICD, and that treatment with lactacystin, a specific proteasome inhibitor, stabilizes Notch⌬E and NICD without inhibiting ␥-secretase activity (10). clasto-Lactacystin ␤-lactone is a derivative of lactacystin known as a cell-permeable, irreversible proteasome inhibitor. Incubation with clasto-lactacystin ␤-lactone during chase period stabilized the Notch⌬E polypeptides and did not inhibit the proteolytic generation of NICD (Fig. 5C, upper center panel, arrow and arrowhead). These data suggested that the proteasomal activity is responsible for the degradation of Notch⌬E in N2a cell lines, which progresses very rapidly compared with that in primary neurons or HEK293 cell lines (see Refs. 10 and 14). In contrast, overexpression of mt FL PS2, PS2/D366A, or PS2/P417S markedly reduced the proteolytic generation of NICD after 30 -60 min of  GRP94 (F, arrowheads), which carry the KDEL sequence at the C terminus, in fractions 9 -12, and that Golgi/TGN marker adaptin-␥ was distributed mainly in fractions 5-9 (G, arrowhead). chase even under treatment with clasto-lactacystin ␤-lactone, whereas overexpression of PS2/P414L did not affect NICD formation (Fig. 5C, upper right and lower panel, arrowhead). However, very small amounts of NICD were detected at 60 min of chase in cells expressing mt FL PS2 or PS2/P417S, suggesting that these mutations partially abolish the site-3 cleavage activity of PS2.
To overcome the caveat that site-3 cleavage activity associated with endogenous PS masks the activities of transfected PS, we transiently co-transfected PS2 derivatives and Notch⌬E in immortalized PS-null fibroblast cell line derived from PS1/ PS2 double-knockout mice (Fig. 5D) (7,40). NICD generation was completely absent in PS-null fibroblasts, which was restored upon transfection with wt FL PS2 whereas the overexpression of mt FL PS2 did not rescue (Fig. 5D, arrowhead). PS2/D366A did not rescue the site-3 cleavage of Notch⌬E as documented previously (14). PS2/P414L and PS2/P417S also did not restore the proteolytic release of NICD in PS-null fibroblasts. We therefore conclude that the amino acid substitution in the first proline of PALP motif interferes with the site-3 cleavage activity of PS2 by inhibiting the formation of HMW PS complex.

Effects of PALP Mutations on A␤ Generation in N2a
Stable Cells-To examine the effects of mutations in PALP motif on ␥-cleavage of ␤APP, we analyzed N2a cells stably expressing PS2/D366A, PS2/P414L, or PS2/P417S. Overexpression of PS2/ D366A in N2a cells inhibited ␥-cleavage of ␤APP, resulting in a marked accumulation of ␤APP C-terminal stubs (i.e. C83 and C99), which are the direct precursors for p3 and A␤, respectively (Fig. 6A, arrowheads) (14). In contrast, cells expressing PS2/P414L or PS2/P417S did not accumulate ␤APP CTFs, suggesting that mutations in either of the two prolines of PALP motif in PS2 does not have a dominant negative effect on ␥-cleavage of ␤APP unlike PS2/D366A. To further characterize the effects of PALP mutations of PS2 on A␤ production, the levels of secreted A␤40 and A␤42 in conditioned media were measured by A␤ C-terminal specific ELISAs (44). Secretion of A␤ from N2a cells expressing PS2/D366A was significantly inhibited compared with that in cells expressing wt FL PS2 as reported previously (see Fig. 8D) (14). In contrast, the total levels of A␤ secreted from cells expressing PS2/P414L or PS2/ P417S were comparable to those in cells with wt FL PS2 (Fig.  6B). The percentage of A␤42 as a fraction of total A␤ (ϭ A␤x-40 ϩ A␤x-42) (%A␤42) secreted by cells stably expressing mt PS2/ P414L was ϳ10%, and this was similar to the %A␤42 secreted from cells expressing wt FL PS2 or wt PS2/P414L, whereas the %A␤42 secreted from cells expressing mt FL PS2 was constantly elevated to ϳ75% as previously documented (Fig. 6C) (20,32,34,43,44). Moreover, the %A␤42 in cells expressing PS2/P417S also was elevated to ϳ30%. These results indicate that the P414L mutation abrogates the overproduction of A␤42 on a FAD-linked mutant basis, whereas P417S mutation harbors a pathogenic function like FAD mutant PS to increase A␤42.
Effects of PS1 Harboring P433L Mutation on Site-3 Cleavage of Notch and A␤ Generation-To determine whether our results regarding the role of PALP motif of PS2 in site-3 cleavage and A␤ generation are applicable to PS1, we constructed cDNAs encoding wt and P267S FAD-linked mutant (mt) PS1 with a proline to leucine amino acid substitution of the first proline in PALP motif (P433L). Consistent with the results obtained with PS2, PS1/P433L as well as PS1/D385A did not rescue the site-3 cleavage of Notch⌬E in PS-null fibroblasts (Fig. 7A). In stably transfected N2a cells, the total levels of secreted A␤ were not affected by expression of wt or mt PS1/ P433L (Fig. 7B). mt FL PS1 increased the %A␤42 by about 2-fold compared with that of wt FL PS1, whereas %A␤42 was not elevated in cells expressing mt PS1/P433L (Fig. 7C). These results suggested that the first proline of the PALP motif in PS1 C terminus also is important for its assembly and stabilization that are the prerequisites for ␥-secretase activities.
Effects of P414L Mutation on the Metabolism and the Dominant Negative Function against ␥-Secretase Activity of PS2/ D366A-It has been reported that aspartate mutations in PS affect not only ␥-secretase activity but also HMW complex formation in stably transfected MEF cells (45). However, a portion of PS2/D366A formed a HMW complex of a similar size to that derived from wt PS2, which was stabilized in a similar manner to fragments derived from FL PS2 (Fig. 3). To determine whether the HMW complex formation and stabilization of PS2/D366A is the prerequisite for its dominant negative function, we stably transfected PS2/D366A harboring P414L mutation (PS2/D366A/P414L) in N2a cells and examined its metabolism as well as its effect on ␥-secretase activity. Western blot analysis revealed that mutated PS2/D366A/P414L was no more capable of replacing endogenous PS1 CTF or accumulating ␤APP CTFs (Fig. 8, A and B). These data suggested that P414L mutation affected the metabolism as well as the dominant negative function of PS2/D366A. We next analyzed the half-life and HMW complex formation of PS2/D366A/P414L. Cycloheximide treatment showed that PS2/D366A/P414L protein was unstable (Fig. 8B). Moreover, the fractionation pattern of PS2/D366A/P414L polypeptides in glycerol velocity centrifugation was very similar to that of PS2/P414L, indicating that P414L mutation abolished the HMW complex formation of PS2/D366A protein. Finally, we analyzed the amount of A␤ secreted from N2a cells. Total levels of secreted A␤ from N2a cells expressing PS2/D366A/P414L were comparable to that in cells expressing wt FL PS2, whereas overexpression of PS2/ D366A inhibited the secretion of A␤ (Fig. 8D). These data suggest that PS2/D366A/P414L double mutant protein does not undergo stabilization and HMW complex formation, thereby losing its dominant negative effects on ␥-secretase activities. DISCUSSION Genetic and biochemical studies have shown that PS is not only AD-related pathological proteins, but also critical components of Notch signaling (1,9). Although the precise molecular function of PS remains elusive, accumulating evidence suggests that PS is involved in intramembranous cleavage of various transmembrane proteins (6 -8, 10, 46, 47). In this study, we demonstrated that (i) the first proline of the highly conserved PALP motif at the C terminus of PS plays a critical role in the stabilization, HMW complex formation and transportation out of ER to Golgi/TGN of PS, and (ii) lack of assembly and stabilization as a HMW complex of PS caused by the amino acid substitution in this proline deprives PS of the capacities to support intramembranous ␥-cleavage of ␤APP and site-3 cleavage of Notch. The results of our studies provide unequivocal evidence that stabilization and HMW complex formation of PS, which require the integrity of the C-terminal domain of PS, are the prerequisites for its physiological and pathological ␥-secretase functions.
The PALP motif is completely conserved in all PS species thus far identified (Fig. 1A). Here we showed that replacement of the first proline of PALP motif for leucine abolished the normal metabolism (i.e. endoproteolysis, stabilization, replacement, and HMW complex formation) of PS in mammalian cells. Although subcellular localization of site-3 cleavage activity of Notch remains unknown, Golgi vesicle-enriched fractions have been shown to contain ␥-secretase activities for ␤APP in vitro (42). Thus, the loss-of-function phenotypes of PS2/P414L for cleavages of ␤APP and Notch that we have shown in this study may be due to defects in stable HMW complex formation and proper transportation from ER to Golgi/TGN. Similar mechanisms may underlie the recessive PS-null phenotype exhibited by C. elegans spe-4 (eb-12) (35) and Drosophila Psn (Dps) 46 (36) alleles in vivo. Taken together, it is strongly suggested that the first proline of PALP motif plays an important role in the proper metabolism of PS family proteins to confer normal or abnormal functions.
In contrast, PS2/P417S underwent endoproteolysis to give rise to stabilized PS complexes and %A␤42 was increased as observed with other FAD-linked PS mutations. Furthermore, Notch site-3 cleavage was partially suppressed with both N141I and P417S PS2 mutants. It has been documented that the proteolytic production of NICD is impaired by some FADlinked PS1 mutations (48), although NICD has been shown to be produced normally with other PS1 mutations (49). In this regard, Kulic et al. (50) have recently reported an interesting relationship between A␤42 overproduction and Notch site-3 cleavage; they have introduced arbitrary mutations at position 286 of PS1 and shown that the strength of A␤42 promoting effects of PS1 mutations is correlated with the extent of reduction in NICD generation. The precise mechanism underlying these changes in ␥and site-3 cleavage remains unknown. However, one could argue that FAD-related PS2 mutations that are more potent in A␤42 promoting effects compared with those in PS1 (Ref. 51 and see Figs. 6C and 7C) might have behaved like some of the artificial PS1 mutations, leading to inhibition in Notch site-3 cleavage. It is tempting to speculate if this apparent "loss-of-function" in Notch cleavage caused by some of the FAD-linked PS mutations could be mechanistically related to the shift of ␥-cleavage from position 40 to 42 of A␤. However, the relevance of NICD reduction to the pathophysiology of AD remains unknown; if mutant PS1 allele were inactive for Notch cleavage, sufficient NICD could be produced by the activity of wt PS1 on the normal allele, considering the normal development of heterozygous PS1 KO mice in the absence of PS2 (40,52), and Notch phenotype has never been reported in FAD patients.
What, then, is the mechanistic role of PALP motif in the stabilization of PS? One possibility would be that this motif serves as the binding site for the cellular factor required for stabilization. Because of its unique side-chain structure, proline is known as the critical amino acid essential to the proper structure of peptide backbone and function of proteins (53). In particular, a number of protein interaction domains harbor proline-rich sequences, that tend to adopt the polyproline II helix, i.e. an extended structure with three residues per turn and thus well placed to interact with the protein. Notably, SH3 domains recognize proline-rich sequences containing the core sequence of PXXP (53). The classification of SH3 ligand is dictated by the location of a positively charged residue that forms a salt bridge with an acidic residue in the SH3 domains, and peptides with a motif ϩXXPXXP or PXXPXϩ (where ϩ refers to a positively charged residue) correspond to class I and II motifs, respectively. Thus, the PALP motif of PS (i.e. KAL-PALP) fulfils the criteria for a class I SH3 ligand. Indeed, secondary structure prediction using Chou-Fasman method suggested that the C-terminal region of PS flanking the PALP motif forms an ␣-helical structure, whereas replacement of the first proline disrupted ␣-helix and converted it to a ␤-sheet. 2 Although PS-binding proteins carrying polyproline binding domain such as SH3 domain have not been identified, it is tempting to speculate that the ␣-helical structure around PALP motif serves as the binding site for the limiting cellular cofactor(s).
An alternative possibility would be that the PALP motif is required for maintaining the proper conformation of the Cterminal region of PS. We have shown previously that subtle changes in the C-terminal region of PS disrupt the stability of PS and its ␥-secretase functions (34). Similarly, the PALP motif may play a seminal role in the maintenance of proper membrane topology and assembly of the TM domains that are essential to the stabilization and complex formation of PS. It is also possible that the structural integrity of the entire C terminus is required for the association of PS with cofactor(s).
PS2/P414L proteins were fractionated exclusively in LMW ranges, whereas wt PS2 or PS2/P417S were separated in HMW fractions. Moreover, a portion of uncleavable PS2/D366A or PS2/⌬loopN, that were stabilized as holoproteins, were also fractionated in HMW ranges. These data strongly suggest that the stabilization of PS proteins coincides with the formation of HMW complex and that the endoproteolysis of PS proteins is not required for these processes. Thus, the difference in molecular size between unstable and stabilized PS complexes may be due to an association with the cofactor(s) including those required for stabilization (30).
Recently, it was reported that the formation of HMW complex is impaired with aspartate mutants of PS (45). However, a portion of PS2/D366A formed a HMW complex of a similar size to that derived from wt PS2 (Fig. 3). The reason for this discrepancy is unknown, but it may be due to differences in detergents and/or cell types. Nonetheless, the occurrence of stabilization and replacement with PS2/D366A that normally are observed with wt PS ( Figs. 2A and 9A) suggests that aspartate mutation does not impair the association of PS with the cofactor(s) required for stabilization (Fig. 9B). In addition, introducing P414L mutation in PS2/D366A completely abolished the HMW complex assembly and dominant negative effects on ␥-secretase activity (Fig. 8). Aspartate mutant thus acts as a dominant negative molecule because of its ability to replace the active endogenous PS, as well as its inactivity as a ␥-secretase. We have shown previously that PS2/I448R, in which the C- FIG. 9. Schematic diagram showing the hypothetical process of PS complex formation. A, nascent PS polypeptides initially form LMW complexes (left row), and then associate with the limiting cofactor(s) for stabilization and/or replacement (shaded circle) to form stabilized HMW complexes (middle row) that represent active ␥-secretases (right row). B, a portion of aspartate mutant PS binds these cofactor(s), forms HMW complexes (middle row), and replaces endogenous PS while it is devoid of the ␥-cleavage activities (right row). C, C-terminal modifications of PS at PALP motif or carboxyl tip (black box) impair association of PS with the limiting cofactor(s), thereby interfering with the formation of the functional HMW complex (middle row).
terminal isoleucine is replaced with arginine, is defective in stabilization, replacement, and ␥-cleavage activity in a similar manner to PS2/P414L (34). We have confirmed that PS2/I448R also does not form HMW complex. 2 Taken together, C-terminal region of PS, including PALP motif and carboxyl tip, may play important roles in the association with cofactor(s) required for stabilization and formation of HMW complex (Fig. 9C).
Mutational analysis of the highly conserved PALP motif provided additional evidence that the stabilized HMW complex of PS represents the active form of ␥-secretase. Further attempts to define the molecular mechanism of PS stabilization and to identify the components of PS complexes relevant to intramembranous proteolytic activity will facilitate understanding of the pathogenetic mechanisms of AD.