Presenilins Mutated at Asp-257 or Asp-385 Restore Pen-2 Expression and Nicastrin Glycosylation but Remain Catalytically Inactive in the Absence of Wild Type Presenilin*

The Presenilins are part of the γ-secretase complex that is involved in the regulated intramembrane proteolysis of amyloid precursor protein and other type I integral membrane proteins. Nicastrin, Pen-2, and Aph1 are the other proteins of this complex. The Presenilins probably contribute the catalytic activity to the protease complex. However, several investigators reported normal Aβ-peptide generation in cells expressing Presenilins mutated at the putative catalytic site residue Asp-257, contradicting this hypothesis. Because endogenously expressed wild type Presenilin could contribute to residual γ-secretase activity in these experiments, we have reinvestigated the problem by expressing mutated Presenilins in a Presenilin-negative cell line. We confirm that Presenilins with mutated Asp residues are catalytically inactive. Unexpectedly, these mutated Presenilins are still partially processed into amino- and carboxyl-terminal fragments by a “Presenilinase”-like activity. They are also able to rescue Pen-2 expression and Nicastrin glycosylation in Presenilin-negative cells and become incorporated into large ∼440-kDa complexes as assessed by blue native gel electrophoresis. Our study demonstrates that the catalytic activity of Presenilin and its other functions in the generation, stabilization, and transport of the γ-secretase complex can be separated and extends the concept that Presenilins are multifunctional proteins.

The ␥-secretase cleavage of the amyloid precursor protein (APP) 1 is the final step in the release of the amyloid peptide A␤, the main constituent of the amyloid plaques in the brain of patients suffering from Alzheimer's Disease (1). This proteolytic event, thought to occur in the hydrophobic environment of the cell membrane, is mediated by a multimolecular protein complex consisting of Presenilin (PS), Aph-1, Pen-2, and Nicastrin (2-7) reviewed by De Strooper (8). How these different proteins precisely interact with each other, how this complex can cleave substrates in their transmembrane domain, and whether additional proteins or lipids are needed for its full activity are questions that remain to be fully addressed.
The Presenilins have been identified some years ago by positional cloning (9). Missense mutations of Presenilin 1 (PS1) and its homologue Presenilin 2 (PS2) are the major cause of dominant inherited early onset Alzheimer's Disease and have been shown invariably to affect ␥-secretase processing of APP, increasing the production of the longer more amyloidogenic A␤42-species (10 -13). Genetic inactivation of PS in mice (14 -17), Caenorhabditis elegans (18), and Drosophila melanogaster (19,20) abrogates ␥-secretase processing of APP (21,22) and of other substrates like Notch (19,20,23), Erb-B4 (24), LRP (25), Nectin 1-␣ (26), CD44 (27,28), E-cadherin (29), Delta and Jagged (30), and others (31). The lethal phenotype of the PS knock-out animals can almost entirely be explained by Notch deficiencies, indicating that the major in vivo task of PS during embryogenesis is the regulation of Notch signaling (14 -17). Likewise, inactivation of the genes encoding the other members of the ␥-secretase complex, at least in Drosophila and C. elegans, also results in deficiencies in ␥-secretase processing of APP and a Notch-like phenotype (2,3,32,33). Although the genetic evidence implying PS in ␥-secretase processing is therefore strong, it remains an issue of debate whether PS contains the active catalytic site of the ␥-secretase complex (34) or whether PS has a more general role in trafficking, maturation, and/or stabilization of the different components of the ␥-secretase complex like Nicastrin (35)(36)(37)(38)(39) and Pen-2 (40) as well as some other proteins including APP (22), telencephalin (41), and the nicotinic acetylcholine receptor (42). Cell biological and pharmacological evidence (43)(44)(45) is consistent with the hypothesis that two aspartate residues, Asp-257 and Asp-385, located in putative transmembrane domains 6 and 7, respectively, constitute the core of an aspartyl-type catalytic site in PS1 and PS2 (34, 46 -48). In further support to this hypothesis, a minimal motif around the aspartyl residues is conserved in all of the PSs in evolution (49) and in a series of other bacterial (50) and mammalian candidate proteases (51) including the recently discovered signal peptide peptidase (52).
The observation that site-directed mutagenesis of either Asp-257 or Asp-385 toward alanine (and also other amino acid) residues resulted in a dominant negative version of PS was pivotal for this hypothesis (34). These investigators found that strong overexpression of such "aspartate mutants" could suppress the expression of wild type PS, resulting in the complete inactivation of ␥-secretase activity. This "replacement phenomenon" of endogenously expressed wild type PS by overexpressed mutant PS is not well understood. Probably, the competition of endogenously and exogenously expressed PS for "limiting cellular factors" responsible for the stabilization of newly synthesized PS could explain this phenomenon (53). It is possible that the recently discovered Aph1, Pen-2, and Nicastrin are these limiting factors (5)(6)(7)(8). Precisely because of this replacement phenomenon, experiments studying overexpressed PS containing Asp mutations (PSD/A) in a wild type background are difficult to interpret. Effects can indeed be the result of a loss of catalytic function of the mutated PS itself, but it is as likely that these aspartyl mutations distort the three-dimensional structure of PS and therefore hamper the assembly of a "normal" active ␥-secretase complex (4). Indeed and in general, dominant negative effects of proteins are difficult to interpret mechanistically, opening opportunities for alternative interpretations (54). Using similar approaches for the generation of stable PSD/A-overexpressing cell lines, some investigators found that all A␤ generation failed in their cell lines (indicating the total absence of ␥-secretase-processing activity) (34,47), whereas others saw no significant effects on A␤ production, in particular, with the single Asp-257 mutant (55)(56)(57). Both groups of investigators claimed the total replacement of endogenously synthesized PS by the mutant PS, making it difficult to explain this discrepancy. Although the second series of observations are in essence not compatible with the hypothesis that Asp-257 is part of the catalytic site, these results are not conclusive because it is extremely difficult to rule out the possibility that variable small amounts of endogenously expressed wild type PS confuse the interpretation of the experiments (58). Therefore, we decided to reevaluate the effects of the Asp mutations on PS function in a PS-deficient cell line (39,59). We also analyzed the role of these aspartyl residues in the proteolytic cleavage of PS itself, the so-called "Presenilinase" activity that previously was proposed to be autocatalytic (34), and the incorporation of these mutant PSs into the ␥-secretase complex.
Confluent stable MEF cell lines were transduced with human APP-695 by infection at 50 multiplicity of infection for 2 h at 37°C with recombinant adenovirus Ad5/CMV-APP (see below) using a minimal volume of Dulbecco's modified Eagle's medium. After 2 h, cells were transferred to Dulbecco's modified Eagle's medium containing 1% fetal bovine serum and maintained overnight at 37°C. Conditioned medium then was collected and immunoprecipitated with mAb 4G8 (Senetek) recognizing A␤ and 30 l of protein G-Sepharose (Amersham Biosciences). The bound material was eluted from the beads using Nu-PAGE sample buffer and boiled for 10 min, and proteins were separated on a 12% Nu-PAGE precasted gel (Invitrogen). After the transfer, Western blot analysis was performed using mAb WO2 (1 g/ml) (60) followed by horseradish peroxidase-conjugated secondary antibodies and ECL detection.
Recombinant Adenovirus-Replication-deficient recombinant adenovirus Ad5/CMV-APP that expresses wild type human APP-695 was prepared by direct ligation as described previously (61). The APP-695 expression cassette, which includes a 5Ј-CMV immediate-early promoter and a 3Ј-polyadenylation sequence, was inserted into the E1a genomic region of human type 5 adenovirus (Ad5), and the resulting construct was propagated in permissive (human embryonic kidney 293) cells. Notch⌬E and Notch intracellular domain (NICD) cDNA was cloned in the pIPspAdapt6 plasmid and transfected with the appropriate cosmid into PerC6/E2A cells, yielding the Ad5/dE1dE2A/CMV Notch⌬E virus (62).
Microsomal membranes were solubilized in buffer containing 0.5% dodecylmaltoside, 20% glycerol, and 25 mM Bis-Tris/HCl, pH 7.0, for 30 min on ice. After ultracentrifugation (100,000 ϫ g for 30 min), the protein concentration of the supernatant was measured (protein assay of Bio-Rad). Finally, extracts were aliquoted and stored at Ϫ80°C until use. For Blue native-PAGE, 30-l extract was mixed with 3 l of the following buffer: 5% Coomassie Blue G-250, 200 mM Bis-Tris/HCl, 1 M 6-aminocaproic-acid, pH 7.0, and 30% sucrose. 70 g of total protein extract was loaded in each lane. Blue native-PAGE was performed at 4 -6°C. After an initial run at 80 V to allow the protein sample to run into the stacking gel, electrophoresis was continued at 200 V for 1 h. The Coomassie Blue G-250 containing cathode buffer was replaced by a buffer without the dye and run for another 1.5 h. The gel was incubated in 0.1% SDS, 25 mM Tris, 192 mM glycine, 20% methanol for 10 min at room temperature and transferred onto polyvinylidene difluoride membranes. The blot was de-stained for 1 h in distilled water/methanol/ acetic acid (60/30/10) and reacted with the appropriate antibodies.
Antibodies-Polyclonal antibodies directed against human PS1 (B14.5 and SB129) and Nicastrin (B59.4) have been described previously (39,64,65). For the detection of the additional ␥-secretase components, polyclonal antibodies were generated in New Zealand White rabbits. MNLERVSNEEKLNLC in the COOH terminus of Pen-2 and CSRVMVYSALRIPPED in the COOH terminus of the long splice variant of Aph1a were coupled to keyhole limpet hemocyanin (Pierce Imject). All of the polyclonal antibodies were affinity-purified against the peptide antigen immobilized on a NHS-Sepharose matrix according to the manufacturer's instructions (Amersham Biosciences). Antibody SZ 909 was raised against PALIYSST in PS1 and kindly provided by Dr. Helmut Jacobsen (Hoffmann-La Roche). Polyclonal 4627 recognizing the amino acids 457-467 of PS1 was kindly provided by Dr. Dennis Selkoe (Harvard, Boston, MA). "Cleaved Notch1 (Val-1744) antibody," specific for the ␥-secretase-cleaved NICD fragment was purchased from Cell Signaling Technology.
Cell Free Assay-For the cell free assay, MEFs were harvested and centrifuged. The cell pellet was resuspended in 250 mM sucrose, 5 mM Tris-HCl, pH 7.4, and 1 mM EGTA supplemented with protease inhibitors and homogenized using a ball-bearing cell cracker (10 passages, clearance 10 m). After low speed centrifugation (800 ϫ g, 10 min), the postnuclear supernatant was ultracentrifuged (100,000 ϫ g, 1 h). The resulting microsomal pellet was washed twice in 0.02% saponin, resuspended in 5 mM Tris, 1 mM EDTA, pH 7, containing 0.5% CHAPS, and incubated for 1 h at 4°C. Cleared extracts (100,000 ϫ g, 1 h) were incubated next overnight (37°C) with recombinant FLAG-tagged APP C100 (66). Finally, de novo formed A␤ was analyzed by SDS-PAGE on 10% Bis-Tris Nu-PAGE gels (Invitrogen) in MES running buffer followed by Western blotting and ECL detection.
Similar levels of expression were obtained with HuPS1 containing the different Asp mutations (PS1D/A, Fig 1, three last lanes). All three Asp mutants displayed also partial proteolytic processing into NTF and CTF fragments, albeit less efficiently and to a much more variable extent than wild type HuPS1 (Fig.  2, B and quantitated in C). However, the fact that they can be cleaved in a PS-negative background is surprising because it is believed that Presenilins are cleaved by an autocatalytic process depending on these aspartate residues (7,34). One possibility is that the PS1D/A proteins are cleaved by caspases (67,68). However, the caspase-cleaved CTF fragments of both HuPS1 and PS1D/A are much smaller than the fragments observed here (result not shown). This was further ruled out because antibody Pep35.2 against amino acids 310 -331 that is amino-terminal to the caspase cleavage site in the PS1 loop reacts with the PS1-CTFs (Fig. 2, panel Bc). This experiment also confirmed that the observed cleavage occurs closely to the region where "Presenilinase" cleaves PS1 (69). The "Presenilinase"-like activity therefore identified in the current experiments is apparently not autocatalytic since cleaving the PS1D/A proteins. It should be noticed that the "Presenilinase" cleavage site displays heterogeneity and that it is not proven that all of the cleavages are performed by one and the same protease (69). One of the major cleavage sites identified occurs carboxyl-terminal to the 283 PALIYSST 291 sequence. We obtained antibody SZ909 that reacts with full-length PS1 and PS1-NTFs that are specifically cleaved at this cleavage site (Fig. 2Bb). In our cell culture conditions, only a marginal fraction of the total pool of PS1-NTF is reacting with this antibody, suggesting that the major "Presenilinase"-like processing occurs at the other identified positions (Met-298 and Ala-299) (69).
PS deficiency also causes impaired glycosylation of Nicastrin ( Fig. 1) (see also Refs. 35,36,38,and 39) and destabilizes Pen-2 expression (Fig. 1) (70). The expression of another ␥-secretase component, Aph1a, is not affected by the absence of PS expression (Fig. 1). Interestingly, HuPS1/D257A, HuPS1/D385A, and HuPS1/D257A/D385A were all able to reconstitute Nicastrin maturation and Pen-2 expression to a similar extent as wild type HuPS1, indicating that these Asp mutations do not interfere with the roles of PS in the transport of Nicastrin through the endoplasmic reticulum and the Golgi complex and the stabilization of Pen-2.
As discussed above, some authors have suggested that the Asp mutations affect the three-dimensional structure of PS1, thereby interfering with the normal incorporation of PS1D/A into the ␥-secretase complex. We investigated this problem using Blue native gel electrophoresis. In these non-denaturing gels, PS1 migrates at ϳ440 kDa because it is non-covalently bound to other proteins like Nicastrin and Pen-2 (Fig. 3, lane  2). Interestingly, similar complexes are observed in PSϪ/Ϫ fibroblasts expressing PS1/D385A or PS1/D257A/D385A, indicating that the PS1D/A mutants become incorporated into a complex together with Nicastrin and Pen-2 (Fig. 3). In PSϪ/Ϫ fibroblasts, no ϳ440-kDa complex is observed and Nicastrin and Pen-2 migrate at a lower apparent M r . Thus, the incorporation of both proteins into the ϳ440-kDa complex depends on PS. Interestingly, Pen-2 (that is expressed at significantly lower levels in PSϪ/Ϫ cells, see also Fig. 1) is distributed over two M r fractions (Fig. 3).
Overall, it is clear from these experiments that the Asp/Ala mutations do not affect the incorporation of PS1 into the fully mature complex. Previously, Yu et al. (4) investigated the incorporation of PS1D/A mutants in high M r complexes using glycerol gradients. They found that PS1D/A is not incorporated into such complexes in contrast to their wild type counterpart. However, it is very well possible that under the overexpression conditions used in those experiments, the other protein partners of the ␥-secretase complex became rate-limiting. This increases the chance that only partial complexes are generated, providing an explanation for the discrepancy between these previous findings and those presented here. In any event, it is clear from our data that the Asp mutations do not necessarily affect the interaction of PS1 with the other members of the ␥-secretase complex, in line also with the results displayed in Fig. 1.
Because under our experimental conditions, the PS1D/A mutants are incorporated into the fully mature ␥-secretase complex, we found it important to reinvestigate the effects on ␥-secretase activity and thus cleavage of APP, especially given the existing controversy in the literature concerning the catalytic activity of these mutants. As shown in Fig. 4, APP-CTF fragments generated from endogenously expressed mouse APP clearly accumulate in PSϪ/Ϫ cells, whereas expression of HuPS1 restores the level of APP-CTF to the ones observed in wild type fibroblasts. None of the PS1D/A mutants was active in this assay and APP-CTF fragments accumulated to a similar extent (between 2.3-and 4.0-fold compared with wild type cells) as in the PSϪ/Ϫ cells (3.3-fold compared with wild type cells). It should be noticed that the levels of endogenously produced murine A␤ are at the detection limits of our assays (59). Therefore, we also analyzed processing of human APP transduced in the cells using adenoviral vectors. A clear accumulation of (human) APP-CTF was again observed in the PSϪ/Ϫ fibroblasts that was reversed in the wild type HuPS1 but not in the HuPS1D/A-expressing cells (Fig. 5A). As shown in the same panel, wild type fibroblasts generated substantial amounts of A␤ under those conditions, whereas PSϪ/Ϫ cells did not produce any detectable A␤ (59). Expression of HuPS1, but not HuPS1D/A, in these cells restored completely A␤ generation. We also investigated the activity of the ␥-secretase complexes under cell free conditions. This assay allows us to measure not only newly generated A␤ but also the cytoplasmic fragment of PALIYSST 291 sequence just amino-terminal to one of the major "Presenilinase" cleavage sites of PS1. This antibody reacts with full-length PS1 and with stronger affinity with PS1-NTF containing a free PALIYSST sequence. The low amount of NTF fragments detected here indicates that "Presenilinase" processing in this experiment does not occur at this site; (c) pep3 5.2 monoclonal antibody against Human PS1 (Ser-310/Asp-331); and (d) 4627 recognizing amino acids Met-457/Ile-467 at the carboxyl terminus of PS1 (CTF-HuPS1). Panel C, NTF-HuPS1 and FL-HuPS1 were quantitated by densitometric scanning using a Kodak imager. The density of the NTF bands is given relative to the total density of NTF and FL-HuPS1. The mean Ϯ S.E. of seven independent experiments is indicated. aa, amino acid. Fig. 5B, none of the PS1D/A mutants was able to cleave the APP substrate, providing further evidence that the Asp/Ala mutations in PS1 inactivate the catalytic site in the ␥-secretase complex.

APP (APP intracellular domain). As demonstrated in
We finally analyzed processing of another ␥-secretase substrate, Notch⌬E (23). Again, wild type human PS1 was able to restore processing in PS-deficient cells, whereas the aspartate mutants were not active (Fig. 5C).
Because the Ala mutations at positions Asp-257 and Asp-385 do not affect the incorporation of PS1 into the mature ␥-secretase complex and because Pen-2 protein expression and Nicastrin glycosylation can be completely reconstituted with these mutants, we conclude that the Asp mutations do not alter significantly the overall structure of the PS protein. Therefore, these mutations affect very specifically the catalytic activity of the complex, strongly agreeing with the interpretation that these two aspartate residues indeed constitute the catalytic site of the ␥-secretase complex. The ongoing controversy in the field is based on the observation that the Asp-257 mutation in PS1 did not significantly affect A␤ generation in certain cell lines (see Introduction). However, in contrast to these previous experiments, potential incomplete replacement of endogenous PS does not complicate the interpretation of the current results. It is possible that in previous experiments using transfected PS1D/A, small amounts of endogenously expressed wild type PS1 and/or PS2 have escaped detection in the Western blot techniques used to confirm replacement. A low level of  Fig. 4 were transduced with adenovirus-driving expression of human APP-695. Membrane extracts were analyzed as in Fig. 4. APP fulllength (FL-APP) and carboxyl-terminal fragments (APP-CTFs) were detected using antibody B10.4 recognizing the carboxyl terminus of APP. A␤-peptide from the conditioned media was immune precipitated using mAb 4G8 and detected in Western blotting with mAb WO2. In the first and the third lane, no adenovirus was added and therefore only endogenously expressed mouse APP was detected. Panel B, membranes prepared from fibroblasts as indicated were solubilized in 0.5% CHAPS, and APP substrate was added. After overnight incubation, material was analyzed in a 12% Nu-PAGE and probed with mAb WO2. Carboxylterminal fragment was analyzed using mAb M2 against the FLAG tag incorporated in the synthetic APP substrate. Panel C, fibroblasts as indicated were transduced with adenovirus driving the expression of Notch⌬E or NICD. Cell extracts were analyzed by Nu-PAGE (7% Tris acetate), and Notch⌬E or NICD was detected, respectively, using mAb recognizing the Myc tag (upper panel) and Cleaved Notch1(Val-1744) antibody (lower panel). This antibody only recognizes the NICD fragment after proteolytic cleavage by ␥-secretase.
residual ␥-secretase activity in these cells could indeed lead to an accumulation of APP-CTF substrate as observed by these investigators. However, A␤ will still be generated at a lower rate and it is possible that in the time course of the experiment a steady state level is reached in the medium that is comparable with the one reached in the wild type cells. Although Notch processing was invariably affected in these experiments, the time course of the experiments analyzing Notch processing and APP processing is quite different. Therefore, it is plausible that clear deficits in Notch processing are observed, whereas the deficits in APP processing are much less obvious.
We conclude that the Asp-257 and Asp-385 are indeed part of the catalytic site in PS1 as initially suggested by Wolfe et al. (34). However, in disagreement with these authors, our experiments suggest that the proteolytic maturation of PS is not always dependent on the ␥-secretase activity of PS1. Indeed, the catalytically dead PS1D/A mutants, when expressed at physiological relevant levels in a PSϪ/Ϫ background, are still processed by an elusive "Presenilinase" (Figs. 1 and 2). Although several authors found that ␥-secretase inhibitors inhibit "Presenilinase" activity (71,72), their inhibitory profiles are not entirely overlapping, suggesting that "Presenilinase" and ␥-secretase are pharmacologically different activities. For instance, pepstatin A is a good inhibitor of "Presenilinase," whereas it is only in very high concentrations also active as a ␥-secretase inhibitor (72). Moreover, the cleavages at the "Presenilinase" site in PS1 are heterogeneous (69), implying that further work is needed to understand fully the exact molecular processes underlying the proteolytic maturation of Presenilin.
The most important implication from our work is probably that the assembly, activity, and biology of the Presenilins (and the other components of the ␥-secretase complex) have to be investigated at physiologically relevant expression levels. Not only the subcellular localization (73) but also the interaction with the different ␥-secretase components (this paper), the processing by "Presenilinase" (this paper), and the glycosylation of Nicastrin (39) are all heavily affected by the overexpression of Presenilin or one of the other proteins of the complex. Therefore, if one wants to make reliable conclusions concerning the physiology of ␥-secretase, it is needed to remain as close as possible to physiologically relevant levels of expression.
Finally, our experiments confirm that deficiency of PS has important repercussions for the stability of other components of the ␥-secretase complex. This finding suggests strongly an important role for PS in stabilization and trafficking of these proteins. Apparently, this function of PS1 can be dissociated from its role in the catalytic cleavage of integral membrane proteins, indicating that PSs are multifunctional proteins.