A Loss of Function Mutation of Presenilin-2 Interferes with Amyloid β-Peptide Production and Notch Signaling*

Presenilin-1 (PS1) facilitates γ-secretase cleavage of the β-amyloid precursor protein and the intramembraneous cleavage of Notch1. Although Alzheimer’s disease-associated mutations in the homologous presenilin (PS2) gene elevate amyloid β-peptide (Aβ42) production like PS1 mutations, here we demonstrate that a gene ablation of PS2 (unlike that of PS1) in mice does not result in a severe phenotype resembling that of Notch-ablated animals. To investigate the amyloidogenic function of PS2 more directly, we mutagenized a conserved aspartate at position 366 to alanine, because the corresponding residue of PS1 is known to be required for its amyloidogenic function. Cells expressing the PS2 D366A mutation exhibit significant deficits in proteolytic processing of β-amyloid precursor protein indicating a defect in γ-secretase activity. The reduced γ-secretase activity results in the almost complete inhibition of Aβ and p3 production in cells stably expressing PS2 D366A, whereas cells overexpressing the wild-type PS2 cDNA produce robust levels of Aβ and p3. Using highly sensitive in vivo assays, we demonstrate that the PS2 D366A mutation not only blocks γ-secretase activity but also inactivates PS2 activity in Notch signaling by inhibiting the proteolytic release of the cytoplasmic Notch1 domain. These data suggest that PS2 is functionally involved in Aβ production and Notch signaling by facilitating similar proteolytic cleavages.

Presenilin-1 (PS1) facilitates ␥-secretase cleavage of the ␤-amyloid precursor protein and the intramembraneous cleavage of Notch1. Although Alzheimer's diseaseassociated mutations in the homologous presenilin (PS2) gene elevate amyloid ␤-peptide (A␤42) production like PS1 mutations, here we demonstrate that a gene ablation of PS2 (unlike that of PS1) in mice does not result in a severe phenotype resembling that of Notchablated animals. To investigate the amyloidogenic function of PS2 more directly, we mutagenized a conserved aspartate at position 366 to alanine, because the corresponding residue of PS1 is known to be required for its amyloidogenic function. Cells expressing the PS2 D366A mutation exhibit significant deficits in proteolytic processing of ␤-amyloid precursor protein indicating a defect in ␥-secretase activity. The reduced ␥-secretase activity results in the almost complete inhibition of A␤ and p3 production in cells stably expressing PS2 D366A, whereas cells overexpressing the wild-type PS2 cDNA produce robust levels of A␤ and p3. Using highly sensitive in vivo assays, we demonstrate that the PS2 D366A mutation not only blocks ␥-secretase activity but also inactivates PS2 activity in Notch signaling by inhibiting the proteolytic release of the cytoplasmic Notch1 domain. These data suggest that PS2 is functionally involved in A␤ production and Notch signaling by facilitating similar proteolytic cleavages. PS1 1 is required for A␤ generation (1) as well as Notch signaling (2). Recent evidence demonstrated a deficiency in ␥-secretase activity in mice lacking the PS1 gene (1). Neurons isolated from these animals secreted reduced levels of A␤ and accumulated the C-terminal fragments of ␤APP, which are normally proteolyzed by ␥-secretase (1). Moreover, mice lacking PS1 exhibit a phenotype, which resembles that caused by the deletion of the Notch1 gene (3,4). A direct involvement of PS1 in Notch signaling has now been demonstrated by the finding that cells lacking PS1 show reduced levels of the proteolytically generated cytoplasmic domain of Notch1 (Notch intracellular domain (NICD)) (5) as well as by genetic evidence derived from multiple model systems including Caenorhabditis elegans, Drosophila melanogaster, and mice (2, 6 -9). Work by Wolfe et al. (10) recently suggested that PS1 might be an unusual aspartyl protease, which exhibits the ␥-secretase activity required for the proteolytic release of A␤. Mutagenizing either one of the two critical Asp residues in transmembrane domain 6 or transmembrane domain 7 of PS1 inhibits A␤ production and results in the accumulation of C-terminal fragments of ␤APP (10). The biochemical phenotype caused by this mutation therefore closely resembles the effects of the PS1 ablation on A␤ production (1).
Although detailed knowledge on the function of PS1 is accumulating, we do not know the biological function of the homologous PS2 protein. Interestingly, PS2 mutations elevate A␤42 generation like PS1 mutations (11). However, Alzheimer's disease-associated mutations in PS2 are very rare, whereas numerous mutations have been reported in the PS1 gene. Moreover, PS2 mutations appear to be less aggressive and in contrast to PS1 the age of onset caused by PS2 mutations is apparently modified by the apoE phenotype (12). To understand the biological function of PS2, we analyzed the phenotype of PS2 deleted mice. We also generated a loss function mutation and investigated its influence on A␤ generation, NICD production, and Notch signaling in cultured cells and in a sensitive functional rescuing assay in C. elegans. From our experiments we conclude that loss of function mutations of PS2 interfere with A␤ production as well as Notch signaling, suggesting that PS2 is functionally involved in the proteolytic release of A␤ and NICD.
Generation of PS2 KO Mice-Mice carrying an ablated PS2 gene were created by targeting exon 5 of the mouse gene in ES cells. A small deletion in exon 5 and insertion of the neomycin resistance gene was sufficient to disrupt translation. Northern blot analysis was carried out according to standard procedures. For Western blot analysis, brain extracts were prepared according to standard procedures. Brain extracts from PS2 ϩ/ϩ , PS2 ϩ/Ϫ , and PS2 Ϫ/Ϫ mice were analyzed for PS1, PS2, and ␤APP expression using a combined immunoprecipitation/ immunoblotting protocol (see below).
Metabolic Labeling and Immunoprecipitation of PS2-To analyze expression of PS2, 293 cells were starved for 1 h in methionine-and serum-free minimum Eagle's medium and subsequently metabolically labeled with 700 Ci [ 35 S]methionine (Promix; Amersham Pharmacia Biotech) in methionine-and serum-free minimum Eagle's medium for 30 min. The PS2 holoprotein was immunoprecipitated from cell extracts as described (17,20) Combined Immunoprecipitation/Western Blotting-Extracts from brains or stably transfected 293 cell lines were prepared and subjected to immunoprecipitation using the polyclonal antibody 3027 to PS1 (18), 3711 to PS2 (19), or C7 to ␤APP (22). Following gel electrophoresis, immunoprecipitated PS proteins were identified by immunoblotting using the monoclonal antibody BI.3D7 (PS1; Ref. 20) or BI.HF5C (PS2; Ref. 20). ␤APP-CTFs were identified using the polyclonal antibody 5818. Bound antibodies were detected by enhanced chemiluminescence (Amersham Pharmacia Biotech).
Analysis of ␤APP Metabolites-Stably transfected 293 cell lines were grown to confluence. For the analysis of A␤ in conditioned media, cells were metabolically labeled with 450 Ci [ 35 S]methionine (Promix, Amersham Pharmacia Biotech) for 2 h, and chased for 2 h in medium containing excess amounts of unlabeled methionine. A␤ and p3 were immunoprecipitated from conditioned media with antibody 3926 (21) and separated on 10 -20% Tris-Tricine gels (Novex), and analyzed by fluorography. To analyze ␤APP-CTFs, cell lysates were subjected to immunoprecipitation with antibody C7 (22), separated on 10 -20% Tris-Tricine gels (Novex), and analyzed by fluorography. Quantitation of ␤APP-CTFs, A␤, and p3 was done by phosphoimager analysis. APP s was immunoprecipitated from conditioned media using antibody 5313.
Analysis of A␤40 and A␤42-Conditioned media (2 ml) were collected from confluent 293 cells grown in 6-well dishes for 24 h. The media were assayed for A␤40 and A␤42 using a previously described enzyme-linked immunosorbent assay (17,20).
Analysis of Notch1 Cleavage-cDNAs encoding Myc-tagged Notch1 derivatives (23) were cloned into the pcDNA3.1/Hygro(ϩ) expression vector (Invitrogen) and stably transfected into 293 cells. To analyze cleavage of Notch1, cells were starved for 1 h in methionine-and serum-free minimum Eagle's medium, subsequently metabolically labeled with 300 Ci [ 35 S]methionine (Promix, Amersham Pharmacia Biotech) for 20 min, and chased for 1 h in medium containing excess amounts of unlabeled methionine. Cell extracts were prepared and subjected to immunoprecipitation of Notch1 derivatives using the anti-Myc antibody 9E10 as described previously (23).
Transgenic Lines of C. elegans and Rescue Assays-Constructs for transgenic expression of PS2 and PS2 D366A in C. elegans were generated as described previously (7). Transgenic lines were established by microinjection of plasmid DNA mixtures into the C. elegans germ line to create extrachromosomal arrays (7). Four independent lines from the progeny of F2 generation animals were established. As the sel-12(ar171) animals never lay eggs (7), rescue of the sel-12 defect can be quantified by scoring egg-laying behavior in transgenic animals (7). 50 transgenic animals of each line were analyzed for their ability to lay eggs. The numbers of egg laid by individual transgenic animals were counted and placed into four categories.: Eglϩϩϩ, robust egg laying, more than 30 eggs laid; Eglϩϩ, 15-30 eggs laid; Eglϩ, 5-15 eggs laid; EglϪ, no eggs laid.

RESULTS
To understand the function of PS2 we generated mice in which the PS2 gene is ablated. Northern and Western blot analysis confirmed the lack of PS2 expression (Fig. 1, A and B). In contrast, no significant changes were observed in the levels of PS1 (Fig. 1B). Surprisingly, in contrast to the very severe effects of the PS1 gene ablation on mouse embryonic development (3, 4) the lack of the PS2 gene in mice does not result in an obvious phenotype. Gross external examination of neonates (Fig. 1C) and adult null, heterozygote, and wt mice showed that the PS2-ablated mice develop normally up to the oldest age studied (1 year). Gross analysis of brain cytoarchitecture by Nissl stain (Fig. 1, D-G) showed that ablated and wt mice have similar neuroanatomy. In addition, adult-ablated mice do not show gross skeletal abnormalities reminiscent of those seen in PS1 null mutant embryos (3,4) when examined by x-ray analysis (data not shown). Although it is possible that subtle cellular abnormalities exist in the PS2-ablated mice, it is clear that the PS2 null phenotype is grossly different from that seen in the PS1 null mouse (3,4). This suggests that PS2 is not obligatorily required for normal embryonic development. Moreover, in contrast to the PS1 deletion (1, 24), analysis of ␤APP processing revealed no accumulation of APP CTFs (Fig. 1B) and no change in A␤ production (data not shown) in the PS2 Ϫ/Ϫ mice. However, the lack of a phenotype in the gene-ablated animals could be due to the compensation of PS2 activity by the significantly (5-fold) higher expression levels of PS1 during mouse development (25) and might therefore not exclude an essential function of PS2.
To allow a direct assessment of the functional role of PS2 in A␤ production and Notch signaling, we mutagenized a conserved aspartate residue (PS2 D366A) (15,16), which was previously shown to be required for the ␥-secretase promoting activity of PS1 (10), and generated stably transfected cell lines. As a control, cells were stably transfected with the wt PS2 cDNA. Overexpression of the mutation inhibits PS2 endoproteolysis, whereas the wt PS2 protein is proteolytically cleaved to generate the characteristic C-terminal fragments observed for presenilins (26 -29) (Fig. 2A). The lack of endogenous PS2 fragments is due to the previously observed replacement upon ectopic overexpression (17,20,27,28). Cells expressing the PS2 aspartate mutation accumulate C-terminal fragments of ␤APP, which are known to be turned over by the ␥-secretase activity to produce p3 and A␤ (10, 11) (Figs. 2B and 3A). Analysis of A␤/p3 generation revealed that cells stably expressing PS2 D366A secrete significantly less A␤ and p3 than cells overexpressing wt PS2 (Fig. 2B). The reduced A␤ and p3 production is not due to a decrease in protein secretion, because no difference in the secretion of APP s was found in the two cell lines (Fig.  2B). Quantitation demonstrates that A␤/p3 production is inhibited by ϳ90% as compared with wt PS2 expressing cells (Fig. 3, B and C). Moreover, inhibition of A␤ production affects both A␤ species, A␤40 (Fig. 3D) and A␤42 (Fig. 3E). As we (Ref. 30; see also Figs. 2B and 3B) and others (14,26) have shown earlier, A␤/p3 production occurs in the presence of overexpressed PS2, which also displaces both endogenous presenilins (Refs. 28 and 30; and data not shown). We therefore conclude that the PS2 D366A mutation interferes with A␤/p3 production. Because PS2, like PS1, appears to be involved in A␤ production, we also investigated whether the PS2 aspartate mutation interferes with Notch signaling. To investigate if functional PS2 is required for cell fate decisions mediated by Notch signaling, we expressed PS2 D366A in a mutant strain of C. elegans, which lacks a functional PS homologue (sel-12 (ar171), Ref. 2). The sel-12 (ar171) allele is known to cause a phenotype, which is due to reduced Notch signaling (2). Consistent with previous results (6), transgenic expression of PS2 in the mutant worm fully rescued the egg-laying phenotype (Table I). However, expression of PS D366A failed to exhibit any rescuing activity (Table I). Therefore, the same mutation, which interferes with the amyloidogenic function blocks the activity of PS2 in Notch signaling in vivo.
To prove whether the PS2 D366A mutation blocks the potentially intramembraneous cleavage of Notch1 (5, 23, 31), we  (19). Overexpression of both PS2 derivatives results in the accumulation of the PS2 holoprotein. As observed previously (26,27), no endogenous holoprotein could be observed. To detect the PS2 CTF, cell lysates from untransfected 293 cells, or 293 cells stably expressing wt PS2 or the PS2 D366A mutation were immunoprecipitated with antibody 3711 (19), and PS2 CTFs were detected with the monoclonal antibody BI.HF5C (20). Robust levels of PS2 CTFs are detected in untransfected cells as well as cells stably expressing wt PS2, whereas generation of the PS2 CTF is inhibited in cells stably expressing PS2 D366A. As observed before, overexpression of PS does not result in an increased production of the proteolytic fragments (28). B, expression of PS2 D366A results in the accumulation of C-terminal proteolytic fragments of ␤APP and a marked reduction of A␤/p3 production. 293 cells stably expressing wt PS2 or D366A were metabolically labeled with [ 35 S]methionine for 2 h followed by a cold chase for an additional 2 h. Cell lysates were immunoprecipitated with antibody C7 (22) to detect the C-terminal fragments of ␤APP, and conditioned media were immunoprecipitated with antibody 3926 to synthetic A␤ (21). Immunoprecipitation of conditioned media with antibody 5313 revealed no difference in the secretion of APP s . FIG. 3. Quantitation of the effect of the PS2 D366A mutation on A␤/p3 production as well as the accumulation of ␤APP C-terminal fragments. A, expression of PS2 D366A results in a marked accumulation of C-terminal fragments of ␤APP generated by ␣and ␤-secretase. B, expression of PS2 D366A inhibits total A␤ production. C, expression of PS2 D366A inhibits p3 production. D and E, the PS2 D366A mutation affects production of both A␤40 and A␤42. Bars represent the mean Ϯ S.E. of three independent experiments.
analyzed the proteolytic release of the NICD. Control cells or cells expressing the PS2 D366A mutation were stably transfected with the previously described Notch⌬E cDNA (N⌬E) construct (23). To monitor the generation of NICD from N⌬E, we also generated cell lines expressing the Notch ICV cDNA, which only encodes the derivative corresponding to NICD (23). In control cells, proteolytic release of NICD from N⌬E was observed (Fig. 4). In agreement with previous results (23), the proteolytically released NICD co-migrates with the Notch ICVencoded protein (Fig. 4). In contrast, overexpression of the PS2 D366A mutation interferes with the proteolytic release of NICD (Fig. 4). Because this domain is required for Notch signaling (23,31), the lack of processing may explain why this construct failed to rescue the sel-12 mutant phenotype (see Table I).

DISCUSSION
In this work, we have demonstrated that mutagenesis of a highly conserved aspartate residue of PS2 results in a functionally inactive PS2 variant. This variant interferes with the proteolytic release of A␤ and NICD and is totally inactive in an in vivo assay monitoring Notch signaling.
Ectopic expression of PS proteins is now well known to result in the efficient replacement of the endogenous presenilins (17,20,27,28). This is also the case when we overexpress PS2 variants (Ref. 30, Fig. 2A, and data not shown). One might therefore argue that we simply followed the loss of PS1 function due to the displacement by the PS2 D366A mutation. However, consistent with previous data, overexpression of wt PS2 (Refs. 14, 26, and 30; Fig. 2B) and familial Alzheimer's disease-associated PS2 mutations allows robust A␤ production (14,26,30). Therefore PS2 D366A is not only displacing PS1 but also appears to interfere directly with A␤ production and Notch cleavage. We can however not exclude that the effects observed by the overexpression of the PS2 D366A mutation are due to the inhibition of PS2 function (by the loss of function mutation) together with the inhibition of PS1 function (by replacement). To prove that PS2 is directly involved in Notch signaling we used an in vivo rescuing assay in C. elegans. In this system the PS homologue, the sel-12 gene, is nonfunctional, which leads to a prominent Notch phenotype. PS2 expression in the mutant worm leads to a full rescue (Table I) directly demonstrating PS2 activity in Notch signaling. In contrast, expressing PS2 D366A fails to exhibit any rescuing activity demonstrating that Asp-366 is critically required for PS2 function in Notch signaling. The very same mutation also blocks the proteolytic release of NICD and significantly reduces A␤ production. Based on these data, we conclude that PS2 promotes A␤ and NICD production by facilitating the endoproteolytic processing of their precursors. Therefore PS2 (data shown here) and PS1 functions (1, 2, 5, 6 -10, 32) appear to be similar. This may also explain our finding that a PS2 ablation causes no obvious phenotype, because the highly expressed PS1 could take over PS2 function (25). Based on these findings one would predict that a double knock-out of PS1 and PS2 would enhance the pathological phenotype. Moreover, reverse genetics in C. elegans revealed a redundant function of sel-12 and hop-1 in Notch signaling (32). A redundant function of PS1 and PS2 is also supported by the findings that wt PS2 can functionally replace sel-12 in C. elegans like PS1 (Table I). However, further work is required to support these predictions. Based on the work reported here it may be interesting to prove if co-expression of nonfunctional PS1 and PS2 derivatives could increase the effects on proteolytic processing of ␤APP. It may also be possible to rescue the reduced A␤ production in cells derived from PS1 Ϫ/Ϫ mice by overexpression of PS2. We also want to note that additional functions of PS2, which may differ from PS1 cannot be excluded from our work. Indeed, PS2 may be functionally involved in apoptosis (34,35). In that regard it is interesting to speculate that PS2 activity in apoptosis may also be required for intramembraneous processing of receptor proteins involved in signal transduction during programmed cell death.
Whether the aspartate mutations inactivate an intrinsic aspartyl protease activity as suggested by Wolfe et al. (10) or indirectly affect cellular transport of certain target proteins (24) remains to be determined. However, based on our data not only PS1 but also PS2 may be a potential target for A␤-lowering drugs. Because very minor amounts of the cytoplasmic Notch domain are required for signal transduction (23, 31), a partial inhibition of PS2/PS1 activity may be sufficient to reduce A␤ and amyloid plaque formation.
Note Added in Proof-While this manuscript was in press, De Strooper and colleagues (36) found that a double knock-out of PS1 and PS2 leads to a full Notch phenotype.

TABLE I
Rescue of the sel-12 egg-laying defect by human PS2 derivatives expressed from the sel-12 promoter For 50 transgenic animals each, the numbers of progeny were counted and grouped in the following categories: ϩϩϩ, over 30 progeny laid by individual animals; ϩϩ, 15-30 progeny laid; ϩ, 5-15 progeny laid; Ϫ, no progeny laid. Strains BR1336 -38, BR1390, and BR1391 also carry an additional unc-1(e538) mutation that does not affect egg laying (7).

Strain
Transgene Genotype Egg-laying behavior ]methionine for 20 min and chased for 1 h. Notch1 derivatives were immunoprecipitated with the antibody 9E10. In control cells the release of NICD from N⌬E is observed consistent with previous results (23). The cleaved derivative co-migrates with the Notch ICV -encoded protein as expected (23). In contrast, expression of PS2 D366A blocks the proteolytic generation of NICD.