The Nonconserved Hydrophilic Loop Domain of Presenilin (PS) Is Not Required for PS Endoproteolysis or Enhanced Aβ42 Production Mediated by Familial Early Onset Alzheimer's Disease-linked PS Variants

Presenilin 1 (PS1) and presenilin 2 (PS2) are polytopic membrane proteins that are mutated in the majority of early onset familial Alzheimer's disease (FAD) cases. Two lines of evidence establish a critical role for PS in the production of beta-amyloid peptides (Abeta). FAD-linked PS mutations elevate the levels of highly amyloidogenic Abeta ending at residue 42 (Abeta42), and cells with ablated PS1 alleles secrete low levels of Abeta. Several recent reports have shown that the hydrophilic loop (HL) domain, located between transmembrane domains 6 and 7, contains sites for phosphorylation, caspase cleavage, and sequences that bind several PS-interacting proteins. In the present report, we examined the metabolism of PS polypeptides lacking the HL domain and the influence of these molecules on Abeta production. We report that the deletion of the HL domain does not have a deleterious effect on the regulated endoproteolysis of PS, saturable accumulation of PS fragments, or the self-association of PS fragments. Abeta production was not significantly altered in cells expressing HL-deleted PS polypeptides compared with cells expressing full-length PS. Importantly, deletion of the HL domain did not affect FAD mutation-mediated elevation in the production of Abeta42. Furthermore, the deletion of the HL domain did not impair the role of PS1 or PS2 in facilitating Notch processing. Thus, our results argue against a biologically or pathologically relevant role for the HL domain phosphorylation and caspase cleavage and the association of PS HL domain-interacting proteins, in amyloid precursor protein metabolism and Abeta production or Notch cleavage.

Autosomal dominant mutations in genes encoding presenilin 1 (PS1) 1 and presenilin 2 (PS2) predispose individuals to familial early onset Alzheimer's disease (FAD) (1)(2)(3). Alzheimer's disease is pathologically characterized by the cerebral deposition of 40 -42 amino acid ␤-amyloid (A␤) peptides, which are generated by the proteolytic processing of amyloid precursor protein (APP) (4,5). PSs play an important role in the generation of A␤, as evidenced by the lack A␤ production in cells established from PS1 Ϫ/Ϫ mice (6,7). Furthermore, FAD-linked mutations in PS increase the production of highly fibrillogenic A␤42 (8 -11). PS are polytopic membrane proteins that undergo regulated endoproteolysis to generate saturable of stable NH 2 -(NTF) and COOH-terminal fragments (CTF), which are the preponderant PS-related species that accumulate in vivo (12,13). Although the precise steps involved in the maturation of synthetic PS polypeptides are not clearly defined, properly folded PS polypeptides are stabilized, undergo endoproteolysis, and form high molecular weight complexes (13)(14)(15)(16)(17)(18). The vast majority of overexpressed full-length PS and transgene-derived polypeptides that correspond to the NTF fail to form stable complexes and are rapidly degraded (15,18,19). Whereas endoproteolytic cleavage is not required for biological activity of PS polypeptides (12,20,21), several lines of evidence indicate that the NTF and CTF together form the functional PS unit in vivo (11). Little is known about the proteolytic cleavage of PS, the mechanism that regulates PS fragment accumulation, and the domains of PS that are essential for the pathogenic function of mutant PS.
In this study, we focused on the role of the cytoplasmic loop (Cys 263 -Leu 381 ) between transmembrane (TM) 6 and TM 7, termed the "loop" domain, in PS endoproteolysis and mutant PS-mediated A␤42 production. The NH 2 -terminal one-third of the loop domain is rich in hydrophobic residues, highly conserved between PS1 homologues, and harbors several FADlinked mutations; the site of regulated endoproteolytic cleavage (Met 292 ) is located within the hydrophobic stretch (21,22). The remaining two-thirds of the loop domain is hydrophilic and only weakly conserved between PS1 homologues. Nevertheless, sev-eral findings have suggested that the HL domain may be important for the biological functions of PS. For example, PS1 and PS2 are cleaved by caspases within this nonconserved region (23)(24)(25)(26). In addition, serine residues within the HL domain of PS1 and PS2 are modified by phosphorylation (14,27,28); it has been shown that phosphorylation of PS2 within the HL domain can regulate caspase cleavage and modulate antiapoptotic properties of the PS2 CTF (29,30). Finally, several laboratories have independently identified interactions between several members of the armadillo family of proteins and PS (31;reviewed in Ref. 32); the binding of p0071 and ␤-catenin to PS has been mapped to the HL domain (33; this study). To address the importance of the HL domain for the biology of PS, we chose to characterize PS1 and PS2 polypeptides that lack this region. Our studies show that the HL domain is dispensable for endoproteolysis of PS and stabilization of resultant derivatives. Furthermore, we document that the deletion of the PS HL domain does not affect FAD-linked mutant PS1-or PS2-mediated production of A␤42 peptides.

EXPERIMENTAL PROCEDURES
Construction of Expression Plasmids-A cDNA encoding PS1 lacking the HL domain (amino acids 304 -371) was generated as follows. A 345-base pair DNA fragment that encodes PS1 amino acids 195-303 and contains PflM1 and BbsI sites at the 5Ј-and 3Ј-end, respectively, was amplified by PCR using the primers 5Ј-CGCTACATTACTGTTG-CACTCC-3Ј and 5Ј-GGCCTCTGGGTCTTCCGGGTCTCCTTCTGC-3Ј. The PCR product was digested with PflM1 and BbsI and ligated to plasmid pBSPS1 (12) digested with PflM1 and BbsI; this strategy results in an in frame deletion of the region encoding residues 304 -371. The deletion was verified by sequencing, and the PS1⌬HL insert was cloned into a pCDNA3.1 expression vector. FAD-linked M146L, H163R, and C410Y mutations were introduced into PS1⌬HL cDNA by exchanging appropriate mutant DNA fragments from mutant PS1 cDNAs (9). A DNA fragment that encodes PS2 amino acids 269 -309 and 353-448 was generated by overlap extension PCR as follows. Two PCRs were performed, one with the primer pair 5Ј-CCGGATCCTGTCCCAAAGG-GCCT-3Ј (PS2Floop) and 5Ј-CCTTTCCTCCTCTTCCTCGGGGTCCAG-CTTCGCC-3Ј and another with the primer pair 5Ј-GGCGAAGCTGGA-CCCCGAGGAAGAGGAGGAAAGG-3Ј and 5Ј-CCGTCTAGACCTCAG-ATGTAGAGCTGATG-3Ј (PS2SR). Subsequently, the two PCR products from the earlier step were mixed and amplified using primers PS2Floop and PS2SR. The final PCR product was digested with BamHI and XbaI and ligated to pCDNA3.1 vector, and the insert sequence was verified. Subsequently, Asp718-PflMI fragments isolated from wild-type (Wt) or mutant PS2 cDNAs (13,34) were inserted to generate PS2⌬HL expression plasmids. Full-length PS2 and PS2⌬HL cDNAs were also cloned into the bicistronic vector, pIRES1hyg (CLONTECH). Expression plasmid encoding truncated Notch1, mNotch⌬E, has been described previously (35) (kindly provided by Dr. Jeffrey S. Nye, Northwestern University Medical School).
Cell Culture and Stable Cell Lines-Monkey COS-1 cells were cultured in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum. Mouse N2a neuroblastoma cells were cultured in 1:1 Dulbecco's minimal essential medium/OptiMEM (Life Technologies, Inc.) supplemented with 10% fetal calf serum. N2a cells were transfected with PS1⌬HL plasmid, and independent stable clones were selected in medium containing G418. To examine alternate cleavage of PS1, cells were incubated with 500 nM staurosporine for 6 or 12 h. Stable N2a cell line, Swe.10, stably transfected with APP695 harboring a "Swedish" double mutation (APPswe) was described previously (36). Double stable (APPswe and PS2) cell lines were generated by transfecting Swe.10 cells with PS2 or PS2⌬HL cDNAs cloned in pIRES1hyg and selecting stable pools in medium containing 0.4 mg/ml hygromycin. Because the selection marker is translated from a bicistronic transcript using an internal ribosomal entry site located 3Ј to the PS2 cDNA, all hygromycin-resistant clones express PS2 polypeptides albeit at varying levels. Primary mouse fibroblasts (PS1 ϩ/ϩ and PS1 Ϫ/Ϫ ) were cultured as described (43). Immortalized PS1 Ϫ/Ϫ fibroblast cells (provided by Dr. Jeffrey S. Nye, Northwestern University Medical School) were transfected with pIRES1hyg expression vectors containing PS1, PS1⌬HL, PS2, or PS2⌬HL cDNA, and stable pools were generated as described above.
Antibodies, Protein Analysis, and Quantification of A␤-PS1 antibodies PS1 NT and ␣PS1loop, PS2 antibody PS2 NT , and methods for Western blot and coimmunoprecipitation analyses were described previously (12,13,18,37). To generate an antibody that reacted with the COOH terminus of PS1, a rabbit was immunized with a keyhole limpet hemocyanin-coupled peptide, which corresponds to PS1 amino acids 453-467 of PS1. The resulting antiserum, PS-C3, was affinity purified and characterized as described previously (38). COOH-terminal epitope-tagged mNotch⌬E polypeptides were detected by monoclonal Myc antibody 9E10. Monoclonal ␤-catenin antibody was purchased from Transduction Laboratories.
To quantify APP synthesis, cells were pulse labeled with [ 35 S]methionine for 10 min, and APP was immunoprecipitated using a COOHterminal antibody, CT15, as described (36). Secreted A␤ species ending at 40 or 42/43 residues were quantified from media conditioned by double-transfected COS cells or stable N2a cell lines using BNT77/ BA27 or BC05 two-site ELISA as described previously (10,18). Aliquots of conditioned media (20 l/lane) were fractionated on bicine/urea gels (39) and immunoblotted with 26D6, a monoclonal antibody that reacts with A␤ residues 1-12 (a gift from Dr. Steve Wagner, SIBIA Neurosciences, Inc.). Measurements of A␤x-40, A␤x-42 (43), or A␤x-42(43)/ total A␤ ratio were statistically examined using analysis of variance (ANOVA) followed by Scheffe's test and expressed as mean Ϯ S.E.

Metabolism of PS1
Lacking the HL Domain-To investigate whether the nonconserved HL region of PS plays an essential role in PS1 metabolism, we constructed a cDNA that encodes a human PS1 polypeptide without the HL region (Glu 304 -Gly 371 ), termed PS1⌬HL (Fig. 1A). The site of regulated endoproteolytic cleavage of PS1 (Met 292 ) (21,22) is present in PS1⌬HL, but the site of caspase cleavage (Asp 345 ) has been deleted. Endoproteolysis of PS1⌬HL was examined by generating independent stable N2a cell lines expressing PS1⌬HL. Antibody PS1 NT , generated against PS1 NH 2 terminus (37), reacted with the "full-length" PS1⌬HL ( Fig. 1B, top panel, lanes 2-4). In addition, PS1 NT reacted with a ϳ30-kDa PS1⌬HL-derived NTF, which exhibited slightly accelerated migration on gels compared with endogenous mouse PS1-derived NTF (lane 1) and co-migrated with full-length human PS1-derived NTF (Fig. 1C, compare lanes 1 and 2). Thus, the 30-kDa fragment is a derivative of the endoproteolytic processing of human PS1⌬HL. We previously demonstrated that PS1 NTF and CTF accumulation is a saturable process and that overexpression of human PS polypeptides in N2a cells interfered with the accumulation of endogenous murine PS derivatives (12,13). To determine if the expression of PS1⌬HL also resulted in the diminished accumulation of endogenous murine PS1 derivatives, we performed Western blot analysis using ␣PS1Loop. Because the epitopes recognized by ␣PS1Loop lie within the HL domain (12), this reagent does not react with PS1⌬HL or the CTF derived from PS1⌬HL (CTF ⌬HL ). In each of the stable PS1⌬HL clones, ␣PS1Loop only detected a weak signal for the mouse PS1derived CTFs as compared with cells transfected with empty vector (Fig. 1B, bottom).
As described above, endoproteolysis of PS1⌬HL at or near Met 292 generated a 30-kDa NTF similar in size to the NTF generated from full-length PS1. Hence, we expected that a shorter ϳ9-kDa CTF ⌬HL will accumulate in PS1⌬HL cells. To confirm that the diminution in murine PS1 CTF was accompanied by the accumulation of CTF ⌬HL , we performed Western blots using PS-C3, an antibody raised against the COOH terminus of PS1. Antibody PS-C3 reacted with the 18-kDa CTF derived from full-length PS1, and ϳ9-kDa CTF ⌬HL derived from PS1⌬HL (Fig. 1C, lanes 3 and 4, respectively). As was the case of CTF derived from full-length PS1, CTF ⌬HL accumulated to saturable levels in independent clones that expressed variable levels of PS1⌬HL (data not shown). Next, we performed coimmunoprecipitation studies to assess whether the NTF and CTF ⌬HL derived from PS1⌬HL could be coisolated. Using previously described methods (37), we show that PS1 NT antiserum could coimmunoprecipitate NTF and CTF ⌬HL from detergent extracts of stable N2a cells expressing PS1⌬HL (Fig. 1C, lanes 5 and 6). PS1-or PS1⌬HL-derived CTFs were not coprecipitated by a control antibody, superoxide dismutase 1, generated against superoxide dismutase (lanes 7 and 8). These results demonstrate that the deletion of the HL domain of PS1 does not affect the regulated endoproteolytic processing of PS1, saturable accumulation of PS1 fragments, or the association of NTF and CTF.
Lack of Caspase Cleavage of PS1⌬HL-Previously, it was demonstrated that in cells treated with staurosporine, an agent that induces apoptosis, full-length PS1 undergoes cleavage by caspases to generate "alternate" CTF (aCTF) (23,25,26). Caspase cleavage of PS1 has been mapped to a single site (Asp 345 ) located within the HL domain (25, 26) ( Fig. 2A). Because this cleavage site has been deleted in the PS1⌬HL polypeptide, we anticipated that PS1⌬HL could not be cleaved by caspase-type proteases during cellular apoptosis. To assess whether PS1⌬HL polypeptides can be cleaved by caspases at alternate site(s), we treated stable N2a cells with staurosporine as described previously (23,25,26) and analyzed alternate cleavage of PS1 by Western blotting. Consistent with the earlier findings, we show that in staurosporine-treated cells fulllength PS1 undergoes caspase cleavage to generate an ϳ14-kDa aCTF, with a concomitant decrease in the levels of ϳ18-kDa CTF (Fig. 2B, top panel). In contrast, PS1⌬HL polypeptide fails to generate an alternate CTF ⌬HL , and the levels of CTF ⌬HL remained unchanged in staurosporine-treated cells (Fig.  2B, bottom panel). These results confirm our prediction that PS1⌬HL polypeptide, which lacks the previously identified caspase cleavage site fails to undergo alternate caspase-mediated cleavage.
Mapping of ␤-Catenin-PS1 Interaction within the HL Domain-Previous studies have shown that the connecting loop between TMs 6 and 7 of PS1 is sufficient for its interaction with ␤-catenin (31,40). To assess whether the ␤-catenin binding site is localized within the HL domain that has been deleted in PS1⌬HL, we mapped the sequence required for the interaction between both proteins. Because PS1 and ␤-catenin could not be reproducibly coimmunoprecipitated from N2a cells, 2,3 we used an affinity precipitation method previously used to document PS1-␤-catenin interaction (33,40). GST-PS1 and GST-PS2 fusion proteins were purified, and their binding capacity to ␤-catenin from mouse brain homogenates was analyzed as described under "Experimental Procedures." As shown in Fig. 3, GST-PS1 fusion proteins containing amino acids 331-351, 331-360, 319 -360, and 263-407, but not those containing 319 -340 and 351-375, interacted with ␤-catenin. Recombinant GST-PS2 containing amino acids 306 -366 within the PS2 loop did not bind ␤-catenin under the same conditions. Lack of interaction between PS2 and ␤-catenin is not surprising, in view of the lack of similarity between PS1 and PS2 within the HL region. Expression of PS1 mutants in which the minimal binding site (331-351) was deleted confirmed the lack of association between the PS1 deletion mutant and ␤-catenin in stable 293 cells. 4 From these results we conclude that the region spanning amino acids 331 to 351 within the HL of PS1 is sufficient for binding to ␤-catenin in vitro, and PS1⌬HL lacks the ␤-catenin binding site. analysis of PS1⌬HL derivatives. Detergent lysates prepared from stable cells expressing PS1 or PS1⌬HL were analyzed by immunoblotting with PS1 NT (lanes 1 and 2) or COOH-terminal PS1 antibody, PS-C3 (lanes 3 and 4). Lysates prepared under nondenaturing conditions were used for coimmunoprecipitation using PS1 NT (lanes 5 and 6) or control superoxide dismutase antibodies (lanes 7 and 8) and analyzed by immunoblotting with PS-C3. PS1 CTF and CTF ⌬HL are indicated.
The HL Domain Is Not Required for FAD-linked PS-mediated Elevation of A␤42 Production-We next sought to assess whether the HL domain of PS1 is essential for the FAD mutation-mediated increase in the production of A␤42 (8,9). For these studies, we generated cDNAs that encode PS1⌬HL polypeptides harboring the FAD-linked missense mutation M146L, H163R, or C410Y. To measure A␤ production, we cotransfected Wt or mutant PS1⌬HL cDNAs into COS cells along with a cDNA that encodes APPswe. In parallel, we transfected COS cells with cDNAs that encode full-length Wt and mutant PS1. As expected from previous studies (9), ELISA quantification revealed a 2-fold increase in A␤42/total A␤ ratio in conditioned media collected from cells transfected with mutant fulllength PS1 (% of A␤42/total A␤ for PS1 Wt was 8.04 Ϯ 0.38 versus PS1 mutants, 17.28 Ϯ 0.73; p Ͻ 0.0001). The A␤42 ratio was not significantly different between cells transfected with Wt PS1 and PS1⌬HL (PS1 was 8.04 Ϯ 0.38 versus PS1⌬HL, 9.03 Ϯ 0.35; p ϭ 0.959), indicating that deletion of the HL domain, by itself, did not influence the production of A␤42 (Fig.  4). Notably, A␤42 ratio was elevated 1.7-fold by the expression PS1⌬HL harboring FAD-linked mutations M146L, H163R, and C410Y (PS1⌬HL Wt was 9.03 Ϯ 0.35 versus PS1⌬HL mutants, 15.7 Ϯ 0.91; p ϭ 0.0026).
Having demonstrated that the deletion of the HL domain of PS1 did not affect A␤ production, we asked whether the same is true for PS2. For these studies, we generated cDNAs that encode Wt and mutant PS2 polypeptides lacking the HL domain (PS2⌬HL). As expected, PS2⌬HL polypeptides were also endoproteolytically processed into stable fragments (see Fig. 5). Consistent with the previous studies (10), expression of mutant PS2 significantly altered A␤42 ratio in transfected COS cells (PS2 Wt was 8.99 Ϯ 0.59 versus PS2 mutants, 14.46 Ϯ 0.58; p ϭ 0.0055) (Fig. 4). Similar to our PS1⌬HL findings, deletion of the PS2 HL domain did not have a significant influence on A␤42 ratio (PS2 was 8.99 Ϯ 0.59 versus PS2⌬HL, 10.57 Ϯ 0.84; p ϭ 0.813). Furthermore, expression of PS2⌬HL polypeptides harboring FAD-linked N141I or M239V mutations significantly

FIG. 2. PS1⌬HL is not cleaved at alternate site(s) during staurosporine treatment.
A, schematic representation of the region from TM 6 to the COOH terminus of PS1 and PS1⌬HL. The sites of regulated endoproteolysis (arrow) and caspase cleavage (arrowhead), the transmembrane domains (boxes), and the recognized by ␣PS1Loop and PS-C3 are indicated. Thick lines represent the 18-kDa CTF and 9-kDa CTF ⌬HL generated by regulated endoproteolysis of PS1 and PS1⌬HL, respectively, and 10-kDa aCTF generated by caspase cleavage of PS1. B, analysis of CTF generated in N2a cell lines treated with staurosporine. Stable PS1 and PS1⌬HL cell lines were incubated with 500 nM staurosporine for the indicated time and lysed in a buffer containing SDS (12). Aliquots of lysates were fractionated by Tris/Tricine gels and analyzed by Western blotting with ␣PS1Loop (top) or PS-C3 (bottom). Note the near complete loss of CTF and the appearance of aCTF in PS1 cells during staurosporine treatment. PS1⌬HL cells continued to accumulate CTF ⌬HL , and no aCTF-like species was generated during staurosporine treatment. UT, untreated; STS, staurosporine-treated.

FIG. 3. ␤-Catenin binding site is contained within the PS1 HL domain.
Schematic structure of the TM 6-TM 7 connecting loop domain of PS1 and PS1⌬HL and the recombinant PS1 proteins used to map ␤-catenin binding site within the HL of PS1 are shown. Aliquots of 1 g recombinant GST-PS1 or GST-PS2 proteins were incubated with 200 g of Nonidet P-40 soluble mouse brain homogenate, and ␤-catenin binding was assessed by immunoblotting. Results from a representative experiment are shown. The binding site of ␤-catenin to the HL of PS1 falls within amino acids 331 to 351. Note that this stretch is part of the region that has been deleted (residues 304 -371) in PS1⌬HL. Recombinant PS2 does not bind ␤-catenin in this assay, as predicted from the lack of similarity between PS1 and PS2 residues in the loop region.

FIG. 4. Influence of HL deletion on A␤ production in COS cells.
COS cells were transfected with PS1, PS1⌬HL, PS2, or PS2⌬HL cDNA (Wt or FAD mutants as indicated) along with cDNA encoding APPswe, and conditioned medium was collected 48 h after transfection. The amount of secreted A␤ x-40 and A␤ x-42 was quantified from using two-site ELISAs, and A␤ x-42/total A␤ ratio (mean Ϯ S. Deletion of the HL Domain Does Not Influence A␤ Production in Double Stable N2a Cell Lines-To provide conclusive evidence that deletion of the HL domain did not influence the total levels of secreted A␤40 and A␤42, we generated stable N2a cells that coexpressed APP and PS2. Swe.10, a stable N2a cell line that expresses human APP695 harboring the Swedish double mutation (36), was transfected with empty vector, fulllength PS2, or PS2⌬HL cDNA cloned in a bicistronic vector, and stable pools of hygromycin resistant cells were derived. As shown in Fig. 5 (top), PS2 NT , a PS2 NH 2 -terminal antibody reacted with a 30-kDa PS2 NTF in stable populations of cells transfected with full-length PS2 or PS2⌬HL cDNAs (lanes 2-7) but not in cells transfected with empty vector (lane 1). Furthermore, despite the differences in the levels of full-length 55-kDa PS2 (lanes 2-4) and 35-kDa PS2⌬HL polypeptides (lanes 5-7), PS2 NTF accumulated to comparable levels. As expected from clonal lines, each of the stable pools synthesized similar levels of APP, quantified by 10-min pulse labeling using [ 35 S]methionine followed by immunoprecipitation and phosphorimaging analysis (Fig. 5, middle).
Secreted A␤40 and A␤42 in media conditioned by the double stable pools were analyzed by fractionation on bicine/urea gels (39) followed by immunoblotting with monoclonal antibody 26D6. Expression of PS2 or PS2⌬HL harboring FAD mutations caused marked increases in the levels of A␤ in stable N2a cells 42 (Fig. 5, bottom). To quantify the levels of A␤40 and A␤42, we performed two-site ELISA as described (10). These analyses revealed no significant differences in A␤40 secreted by PS2⌬HL cells as compared with full-length PS2 cells or cells transfected with an empty vector ( Table I). The only significant difference in the amounts of secreted A␤40 was observed in PS1 N141I cells (p ϭ 0.035, N141I versus vector transfected cells). Although it appears that there is a small increase in the amounts of A␤42 secreted by PS2⌬HL cells as compared with PS1 Wt cells, this difference did not reach statistical significance by ANOVA (p ϭ 0.99) or by t test comparison (p ϭ 0.069). Furthermore, the percentage of A␤42/total A␤ was also not significantly different between PS2Wt and PS2⌬HL cells (11.68 Ϯ 1.32 and 15.04 Ϯ 0.82, respectively; p ϭ 0.97) (Fig. 6). As expected from previous studies (10), the amounts of secreted A␤42 were significantly higher in FAD-linked mutant cells (7.68-and 5.74-fold more in PS2 N141I and PS2 M239V, respectively; p Ͻ 0.0001), compared with PS2 Wt cells. Consistent with immunoblotting results, the amounts of secreted A␤42 were also significantly higher from mutant PS2⌬HL cells (7.72and 5.41-fold more in N141I and M239V, respectively; p Ͻ 0.0001), compared with PS2⌬HL cells (Fig. 6). Collectively, these results demonstrate that the deletion of the HL domain of PS1 or PS2 has little influence on the levels of secreted A␤. Furthermore, FAD-linked mutations elevated the levels of secreted A␤42 to similar levels in cells expressing the full-length or PS2⌬HL molecules.
HL Domain Is Not Required for PS Facilitation of Notch Processing-Previously it was reported that PS1 facilitates the intramembranous proteolytic processing Notch, which releases the Notch intracellular domain (NICD) (41). To assess whether this function of PS is impaired by the deletion of the HL domain, we examined Notch cleavage in stable N2a cell lines and in PS1 Ϫ/Ϫ fibroblasts stably transfected with PS1, PS2, or the corresponding ⌬HL cDNAs. To examine the processing of Notch, we transiently expressed mNotch⌬E, which contains the signal sequence, the TM domain, and the intracellular domain of murine Notch-1 and COOH-terminal Myc epitope tags (35); mNotch⌬E is widely used to assess Notch processing in mammalian cells (35,(41)(42)(43)(44). Western blot analysis of transfected primary fibroblasts derived from wild type (PS1 ϩ/ϩ ) embryos revealed three bands staining with Myc antibody (Fig.  7A, lane 3). As described previously, the two upper bands correspond to mNotch⌬E and a cotranslational product (initiated from an internal AUG site Met 1726 ), and the lower band corresponds to the NICD (35). Consistent with previous findings (41), production of NICD from mNotch⌬E is impaired in primary fibroblasts derived from PS1 Ϫ/Ϫ embryos (Fig. 7A, lane  3). A comparison of stable N2a cell lines expressing full-length  -4), or PS2⌬HL (lanes 5-7) cDNAs in pIREShyg vector (Wt and FAD mutants as indicated). Stable pools of hygromycin-resistant clones were selected, and the expression of PS2 polypeptides was analyzed by immunoblotting with PS2 NT antiserum (top panel). Full-length PS2, PS2⌬HL, and PS2 NTF are indicated. APPswe expression was measured by 10-min pulse labeling with [ 35 S]methionine followed by immunoprecipitation using CT15 and phosphorimaging (middle panel). Aliquots of conditioned medium (20 l) from stable pools of cells were fractionated on bicine/urea gels and analyzed by immunoblotting with 26D6, a monoclonal antibody that reacts with A␤ residues 1-12 (bottom panel). As previously reported, A␤42 migrates faster than A␤40 in this gel system (39). Cells expressing PS2 or PS2⌬HL harboring FAD-linked mutations secreted more A␤42 compared with Wt controls. PS1 or PS1⌬HL revealed that Notch cleavage was not affected by the deletion of the HL domain (Fig. 7A, lanes 5 and 6). To confirm that PS lacking the HL domain will "rescue" the Notch processing deficiency in PS1 Ϫ/Ϫ fibroblasts, we stably transfected immortalized PS1 Ϫ/Ϫ fibroblasts with cDNAs encoding full-length PS1, full-length PS2, PS1⌬HL, or PS2⌬HL. Subsequently mNotch⌬E was introduced by transient transfection. Consistent with the results of primary PS1 Ϫ/Ϫ fibroblasts, mNotch⌬E was not efficiently processed to NICD in immortalized PS1 Ϫ/Ϫ fibroblasts transfected with an empty vector (Fig.   7B, lane 2). This deficiency in NICD production was restored in stable PS1 Ϫ/Ϫ fibroblasts expressing full-length PS1 or PS2, as well as PS1⌬HL or PS2⌬HL (Fig. 7B, lanes 3-6). Collectively, these results provide evidence that the HL domain of PS1 or PS2 is dispensable for Notch processing.

DISCUSSION
Protein topology studies predict that PS contains up to eight TM domains (45,46). Apart from the NH 2 terminus (81 residues) and the COOH terminus (39 residues), the loop between TM 6 and TM 7 is the only cytoplasmic domain that is greater than 10 residues in length. The hydrophilic stretch within the TM 6-TM 7 loop is only weakly conserved between human PS1 and PS2 and other PS homologues. Despite the weak sequence similarity, the HL domains of PS1 and PS2 share several characteristics. For example, both PS1 and PS2 undergo cleavage by caspases within the HL domain (23)(24)(25)(26). Serine residues within the HL domain of PS1 and PS2 are modified by phosphorylation (14,27,28). Furthermore, using the HL domain and neighboring sequences as the bait, yeast two-hybrid studies from independent laboratories have identified interactions between PS1 and PS2 with several members of the armadillo family of proteins (31; reviewed in Ref. 32). In this study, we characterized the metabolism and pathogenic activities of PS1 and PS2 polypeptides lacking the HL domain. Our studies show that the metabolism and stability of PS polypeptides lacking the HL domain is indistinguishable to that previously described for full-length PS1. Moreover, deletion of the HL domain did not impair the role of PS in facilitating Notch processing. Finally, FAD-linked mutant PS1 and PS2 lacking the HL domain enhanced the production of A␤42 peptides, consistent with the view that the pathogenic activity of PS polypeptides does not require the presence of the HL domain.
We previously reported that a deletion of the COOH-terminal 106 or 64 amino acids of PS1 interfered with regulated endoproteolysis (13). FAD-linked mutations introduced into COOH-terminally truncated PS1 molecules failed to elevate the levels of A␤42. 5 Furthermore, it was recently reported that addition, deletion, or substitution of amino acids at the extreme COOH terminus of PS2 impairs proteolytic cleavage and pathogenic activity of PS2 (47). In contrast to the critical role played by the COOH terminus, deletion of the NH 2 -terminal 76 residues of PS2 had no influence on PS metabolism and enhanced A␤42 production by FAD-linked PS variants (47). In this report we provide evidence that the HL domain of PS1 or PS2 is not required for endoproteolysis, saturable accumulation of PSderived fragments, or the association between NTF and CTF. Collectively, these results suggest that the HL domain or the proteins that associate with this domain are unlikely to play a critical role in regulating the stability or accumulation of PS NTF and CTF.
Our findings are noteworthy considering the number of proteins that have recently been shown to bind to the loop domain of PS1 or PS2. Significantly, members of the armadillo family of proteins including ␤-catenin, ␦-catenin, p0071, and neuronal-specific plakophilin bind to PS1 or PS2 through a region within the connecting loop between TM 6 and TM 7 (31; reviewed in Ref. 32). Binding of p0071 and ␤-catenin has been mapped to the HL domain (33; this study). Other PS1 CTFbinding proteins reported thus far include Bcl-XL, filamin family of actin-binding proteins, and SEL10, a Cdc4p-related protein (reviewed in Ref. 32). It is known that PS NTF and CTF form heteromeric high molecular weight complexes in vivo (14,16,17), whereas excess full-length PS polypeptides fail to as-  1 and 4) or a cDNA encoding a COOH-terminally Myc-tagged mNotch⌬E polypeptide (mN⌬E) (35) (lanes 2, 3, 5, and 6). Lysates of cells were analyzed by Western blotting with a monoclonal Myc antibody, 9E10. B, stable PS1 Ϫ/Ϫ fibroblasts expressing full-length or ⌬HL PS were transiently transfected with mN⌬E and Notch processing analyzed by blotting with 9E10. Arrows indicate mN⌬E and NICD. An open circle indicates the previously described cotranslational product, initiated from an internal translation initiation site (Met 1726 ) (35). semble into stable complexes and are subject to rapid degradation (13,15,18,19). Based on our analysis of PS⌬HL metabolism, we predict that HL-interacting proteins may not play a role in the formation of high molecular weight PS complexes. Unlike PS polypeptides with substitutions of aspartate residues in TM 6 and TM 7 (42,44,48), overexpression of PS⌬HL neither influenced ␥-secretase processing of APP nor intramembranous cleavage of Notch 1. Furthermore, our analyses of A␤ secreted from cells expressing PS1⌬HL or PS2⌬HL harboring FAD-linked mutations strongly argues against a role for caspase cleavage within the HL domain in mediating the effects of FAD mutations on A␤42 generation. A similar conclusion regarding the connection between caspase cleavage and A␤42 production was reached in an earlier study (49). Although our studies provide clear evidence that under normal culture conditions caspase cleavage of PS does not contribute to enhanced production of A␤42 by mutant PS, we cannot rule out a role for caspase cleavage of PS in influencing A␤42 production under pathogenic conditions in vivo. We are currently generating transgenic mice that express PS⌬HL to characterize the in vivo role of the PS1 HL domain, and assess whether expression of HL-deleted PS polypeptides can rescue the developmental defects of PS1 Ϫ/Ϫ mice.