Molecular Dissection of Domains in Mutant Presenilin 2 That Mediate Overproduction of Amyloidogenic Forms of Amyloid b Peptides INABILITY OF TRUNCATED FORMS OF PS2 WITH FAMILIAL ALZHEIMER’S DISEASE MUTATION TO INCREASE SECRETION OF Ab42*

Mutations in presenilin (PS) 1 or PS2 genes account for the majority of early-onset familial Alzheimer’s disease, and these mutations have been shown to increase production of species of amyloid b peptide (Ab) ending at residue 42, i.e. the most amyloidogenic form of Ab. To gain insight into the molecular mechanisms whereby mutant PS induces overproduction of Ab42, we constructed cDNAs encoding mutant and/or truncated forms of PS2 and examined the secretion of Ab42 from COS or neuro2a cells transfected with these genes. Cells expressing full-length PS2 harboring both N141I and M239V mutations in the same polypeptide induced overproduction of Ab42, although the levels of Ab42 were comparable with those in cells engineered to express PS2 with one or the other of these PS2 mutations. In contrast, cells engineered to express partially truncated PS2 (eliminating the COOH-terminal third of PS2 while retaining the endoproteolytic NH2-terminal fragment) and harboring a N141I mutation, as well as cells expressing COOH-terminal fragments of PS2, did not overproduce Ab42, and the levels of Ab42 were comparable with those in cells that expressed fulllength, wild-type PS2 or fragments thereof. These data indicate that: (i) the Ab42-promoting effects of mutant PS2 proteins reach the maximum level with a given single amino acid substitution (i.e. N141I or M239V); and (ii) the expression of full-length mutant PS2 is required for the overproduction of Ab42. Hence, cooperative interactions of NH2and COOH-terminal fragments generated from full-length mutant PS2 may be important for the overproduction of Ab42 that may underlie familial Alzheimer’s disease.

rological disorder characterized pathologically by an extensive neuronal loss in the cerebral cortex as well as a massive deposition of amyloid ␤ peptides (A␤) as senile plaques and in the walls of blood vessels (1). A subset of early-onset AD is inherited as an autosomal dominant trait, and presenilin (PS) genes were identified as the major causative genes for these earlyonset familial AD (FAD). PS1 gene, which is linked to the majority of early-onset FAD located on chromosome 14 (2), and PS2 gene (3), which is responsible for a subtype of FAD linked to chromosome 1, encode homologous polytopic membrane proteins that predominantly localize to endoplasmic reticulum (4 -6) and span the membrane 8 times (7). More than 40 missense mutations (8), as well as an exon 10 deletion (9) in PS1 and two missense mutations of PS2 (3,10), thus far have been identified in pedigrees of FAD.
The physiological function of PS proteins is unknown, although recent data from studies in Caenorhabditis elegans (11,12) and PS1 gene knock-out mice (13,14) indicate that PS1 may play some role in Notch signaling. The mechanisms whereby mutations in PS1 or PS2 genes cause AD also remain elusive, but several lines of evidence suggest that they may lead to AD by promoting ␤-amyloid deposition. Amino acid substitutions, as well as an exon 10 deletion, of PS1 (15)(16)(17) and PS2 (17,18) have been shown to increase the secretion of a species of A␤ ending at residue 42 (A␤42), i.e. the most amyloidogenic form of A␤ (19 -21). Recent findings that the secretion of A␤ from primary neurons cultured from brains of mice that lack PS1 is decreased, despite the normal level of full-length ␤-amyloid precursor protein (␤APP) or the amyloidogenic COOH-terminal fragment thereof, argue for the notion that PS is an important co-factor for the proteolytic processing of ␤APP at the COOH terminus of A␤ termed ␥-cleavage (22). However, the mechanisms whereby mutant PS proteins affect ␥-cleavage and lead to the increased production of A␤42 is unknown. Some investigators have shown the direct association of PS and ␤APP in cultured cells (23,24), whereas others have not (25). Thus, one may speculate a direct or indirect "chaperone"-like effect of mutant PS or effects on intracellular vesicular trafficking to increase the susceptibility of ␤APP to be cleaved at position 42.
Several questions arise from these observations. The mechanistic effects of a given single amino acid substitution in mutant PS protein leading to AD are not yet understood. Previous data showing that either of the two known PS2 mutations, i.e. the Volga German mutation that substitutes Asn-141 for Ile (N141I) or the Italian mutation causing a Met-239 to Val (M239V) mutation, considerably increase the percentage of secreted A␤42 (17,18,26), suggesting that a single amino acid substitution on PS2 may lead to a significant change in the protein folding and/or interaction with other proteins of PS2 compared with those with wild-type PS2. To gain insights into the nature of pathogenic effects caused by PS2 mutations, we first examined whether a mutant PS2 molecule harboring both the N141I and M239V mutations would increase the overproduction of A␤42, compared with singly mutated PS2 with one or the other of these mutations. PS1 and PS2 have been shown to undergo endoproteolytic cleavage that yields a long NH 2 -terminal fragment (NTF) and a short COOH-terminal fragment (CTF) spanning the membrane 6 and 2 times, respectively (18,27). These fragments are the predominant forms of PS1 or PS2 in cultured cells or brain tissues that do not overexpress PS (27). However, the relationship between cleavage and function of PS is not well understood. Next we sought to examine if the NTF or CTF forms of mutant PS2 alone are capable of promoting the secretion of A␤42. To this end, we expressed partially truncated forms of PS2 (eliminating the COOH-terminal third of PS2 while retaining the endoproteolytic NTF) harboring a N141I mutation, as well as CTFs of PS2 in cultured cells and examined the COOHterminal properties of A␤ secreted from these cells.

EXPERIMENTAL PROCEDURES
Construction of Expression Plasmids-A full-length cDNA encoding wild-type (WT) human PS2 was obtained by PCR from a normal human cDNA library, and the N141I or M239V PS2 mutations were introduced by the dU-template method as described previously (18). WT as well as mutant (mt) PS2 cDNAs were subcloned into pBluescript, and the coding region was then subcloned into a mammalian expression vector pcDNA3. The N141I/M239V double mutation was introduced by digesting the N141I mt PS2 cDNA in pBluescript with BamHI and BstXI and then inserting the resultant ϳ0.5-kilobase pair product between the BamHI-BstXI sites of M239V mt PS2 in pcDNA3.
To express derivative polypeptides of PS2 fused to glutathione Stransferase (GST), cDNAs encoding the NH 2 -terminal (2-84) or the loop (301-361) portions of PS2 were amplified by PCR and subcloned into an Escherichia coli expression vector (pGEX-6P-1, Amersham Pharmacia Biotech). These ligations resulted in the fusion of PS2 sequences COOH-terminal and in-frame with GST.
Cell Culture and Transfection-Monkey COS-1 cells or mouse neuro2a (N2a) cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and penicillin/streptomycin at 37°C in 5% CO 2 atmosphere as described (18). Transient expression in COS-1 cells was performed by the DEAE-dextran method, and stable N2a cell lines were generated by transfecting the cDNAs in pcDNA3 vector using a mammalian transfection kit (Stratagene) or LipofectAMINE (Life Technologies, Inc.) and selection in Dulbecco's modified Eagle's medium containing G418 (Life Technologies, Inc.) at 400 or 500 g/ml. Double transfection of COS-1 cells with cDNAs encoding PS2 or its derivatives in pcDNA3 and those encoding the COOH-terminal 100 amino acids (C100) of ␤APP in p91023(E) was performed as described previously (18). Co-expression of NTF and CTF of PS2 was performed by transiently transfecting a cDNA encoding PS2/304ctf into N2a cells stably expressing WT or N141I mt PS2/ 303stop using LipofectAMINE according to the manufacturer's instructions.
Antibodies and Immunoblot Analysis-Polyclonal antibodies were raised in rabbits using synthetic peptides conjugated to keyhole limpet hemocyanin or GST fusion proteins corresponding to the following predicted amino acid sequences of human PS2 as immunogens: anti-G2N2 against GST fused to amino acids 2-84 of PS2, anti-G2L against GST fused to amino acids 301-361 of PS2, and anti-PS2C2 against peptides corresponding to amino acids 443-448 of PS2. Anti-G2N2 and anti-PS2C2 antisera were further affinity-purified to their immunogens as described (28). Anti-PS2loop antiserum raised against GST fused to the loop domain of PS2 (29) was kindly provided by Dr. G. Thinakaran. The locations of immunogen peptides/protein fragments within PS2 are shown in Fig. 1.
Cells were lysed in 2% SDS sample buffer and briefly sonicated. The samples were separated by SDS-PAGE without prior heating, transferred to polyvinylidene difluoride membrane (Millipore), and probed with each of the anti-PS antibodies as described (18). The immunoblots were developed using an ECL system (Amersham Pharmacia Biotech).
Immunofluorescence Microscopy-Transiently transfected COS-1 cells were cultured on glass coverslips. Cells were fixed by incubation in phosphate-buffered saline (PBS) containing 4% paraformaldehyde at room temperature for 30 min, permeabilized, and blocked with PBS-TB (10 mM phosphate buffer, pH 7.4, containing 150 mM NaCl, 0.1% Triton X-100 and 3% bovine serum albumin) for 30 min at room temperature. Coverslips were then incubated with primary antibodies against PS2 for 2 h and fluorescein isothiocyanate-or Texas Red-conjugated secondary antibodies in PBS-TB for 1 h, mounted in PermaFlour Aqueous Mounting Medium (IMMUNON), and viewed with a confocal laser scanning microscope (Fluoview, Olympus, Tokyo) as described (30).
Quantitation of A␤ by Two-site ELISAs-Two-site ELISAs that specifically detect the COOH terminus of A␤ were used. BNT77 raised against human A␤11-28, which recognizes full-length as well as NH 2terminally 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) beginning at the Leu-17 residue of A␤. BA27 and BC05, monoclonal antibodies that specifically recognize the COOH termini of A␤40 and A␤42, respectively, were conjugated with horseradish peroxidase and used as detector antibodies. The specificity and sensitivity of these ELISAs have been characterized previously (18,31). Culture media were collected after appropriate incubation periods (60 h in COS-1 and 12 or 24 h in N2a cells) and subjected to BNT77/BA27 or BC05 ELISAs as described (18).

Expression of Full-length PS2 Harboring N141I, M239V, or Both of the Mutations in Cultured Cells and Effects on A␤
Secretion-We transfected cDNAs encoding WT or N141I, M239V single mt as well as N141I/M239V double mt PS2 cDNAs transiently into COS-1 cells or stably into N2a cells and analyzed the cell lysates by Western blots using anti-PS2 antibodies. As we have previously observed (18), when we transfected cells with cDNAs encoding WT or N141I mt PS2, 50 -55-kDa polypeptides corresponding to full-length (fl) PS2, 35-40-kDa NTFs migrating as a doublet as well as a ϳ19-kDa CTF were detected in COS-1 cells (data not shown, but see Fig. 4 showing identical patterns of processing of PS2 in COS-1 cells); this ϳ19-kDa CTF was presumed to be produced by caspase-3-like proteolytic activities (32) because its generation was inhibited by a caspase-3 inhibitor, DEVD-CHO. 2 In addition, a small amount of a ϳ23-kDa CTF was detected. In N2a cells, 50 -55-kDa fl PS2 (Fig. 2, arrow), 35-kDa NTF (Fig. 2, closed arrowhead), as well as a 23-kDa CTF (Fig. 2, open arrowhead) were detected. However, the amounts or ratios of NH 2 -and COOH-terminal fragments were similar in M239V or N141I/ M239V double mt PS2 compared with N141I mt or WT PS2 both in COS-1 and N2a cells (Fig. 2).
We then quantitated the levels of A␤40 and A␤42 secreted from cells transfected with these mt PS2 cDNAs. The percentages of A␤42 secreted from COS-1 cells doubly transfected with ␤APP C100 and N141I or M239V mt PS2 were elevated to almost similar levels (ϳ30%) by 2.2-fold compared with those from cells with WT PS2 and ␤APP C100 (average of A␤42, 13.8%), and the absolute levels of secreted A␤42 were increased by 2.3 (N141I) and 1.7 (M239V) times, respectively, compared with WT PS2. When COS-1 cells were transfected with N141I/ M239V double mutated PS2 and ␤APP C100, the A␤42 was 31.4%, and the A␤42 level also was similar to those in cells with singly mutated PS2 (Fig. 3A). Similar results were obtained in COS-1 cells doubly transfected with full-length WT or Swedishtype mutant (i.e. 595/596 KM-NL) ␤APP together with mt PS2 (data not shown).
We then examined the secretion of A␤ from stably transfected N2a cell lines expressing WT, N141I, M239V, or N141I/ M239V mt PS2. As described previously (18), N2a cells expressing N141I mt PS2 secreted considerably increased amounts or percentages of A␤42 (71.6% of total A␤ and 4.2 times compared with those with WT PS2: 17.0%). N2a cells expressing M239V mt PS2 also secreted significantly increased levels (3.4 times compared with those with WT PS2) or percentage (mean, 58.3%) of A␤42. However, the secretion of A␤42 from cells expressing N141I/M239V double mutant PS2 were again similar to those with either of the single PS2 mutations in terms of the absolute levels (4.1 times compared with those with WT PS2) or percentages (mean, 59.3%) (Fig. 3B)  of endoproteolytic processing of PS2 and especially to examine whether the NTF or CTF of PS2 is biologically active, we then expressed truncated forms of WT or N141I mt PS2 in cultured cells and characterized their metabolism and subcellular localization. Two categories of cDNAs encoding truncated PS2 were used (see Fig. 1): (i) NTF constructs (WT or N141I mt) ending at residues 270, 303, or 388 (designated PS2/270stop, PS2/ 303stop, and PS2/388stop), retaining the NH 2 -terminal 6 (or 7) transmembrane (TM) domains and shorter than, close to, or longer than the predicted size of PS2 NTF, respectively (33,34); accordingly, PS2/388stop retains the entire length of the loop region as well as the 7th TM domain of PS2; (ii) CTF constructs starting at residues 271, 304, or 344 (designated PS2/271ctf, PS2/304ctf, and PS2/344ctf), longer than, close to, or shorter than the predicted size of native CTF, respectively. PS2/344ctf is close to the size of the "ALG-3" fragment, which was found to inhibit cellular apoptosis (35).
In stable N2a cells, the expression patterns of PS2 derivatives were essentially similar to those in COS-1 cells with some differences (Fig. 5). Notably, 45-50-kDa polypeptides corresponding to PS2/388stop were barely processed to form 35-kDa NTF (Fig. 5A) that was present in cells with fl PS2 (Fig. 5A, arrowhead). PS2/271ctf (Fig. 5B) and PS2/304ctf (Fig. 5B) also did not produce proteolytic fragments of smaller sizes. The patterns of expression of endogenous ␤APP were almost similar between these cell lines (data not shown).
Next we examined the subcellular localization of the PS2 NTFs or CTFs in COS-1 cells by immunofluorescence microscopy. Remarkably, all constructs encoding NTFs and CTFs of WT or N141I mt types of PS2 showed similar distribution in a fine meshlike pattern throughout the cytoplasm as well as dense immunostaining in the perikaryal areas, which corresponded to those with BiP, a marker for endoplasmic reticulum (Fig. 6). N2a stable cells also showed similar patterns of ER localization of PS2 derivatives (data not shown).
Characterization of A␤ Secreted from Cells Expressing Truncated Forms of WT or N141I mt PS2-We then quantitated the levels and percentages of A␤40 and A␤42 secreted from cells expressing truncated forms of N141I mt or WT PS2. In COS-1 cells doubly transfected with ␤APP C100 and each of the three types of truncated mt PS2, the A␤42 in total A␤ was ϳ10% in all, and they were similar to those in cells with corresponding forms of truncated WT PS2, whereas A␤42 comprised 22.3% of total A␤ in cells expressing fl mt PS2, which was 1.7 times relative to that in cells expressing fl WT PS2 (13.5%). However, the total levels of A␤ were increased by ϳ2-fold in cells with truncated mt PS2 compared with those with truncated WT PS2. When CTFs of PS2 were transfected together with ␤APP C100, the A␤42 ranged between 11.9 and 14.6%, which also was similar to cells with fl WT PS2 (Fig. 7A). Similar results were obtained in COS-1 cells doubly transfected with WT or Swedish-type mutant (i.e. 595/596 KM-NL) ␤APP together with these PS2 derivatives (data not shown).
In N2a cells stably expressing three types of truncated forms of N141I mt PS2, the percentage of A␤42 that comprised the total A␤ ranged between 14.2 and 18.0%, which was similar to those in cells with truncated (12.3-16.5%) or fl WT PS2 (20.5%), and the absolute amounts of secreted A␤ were at similar levels between cells expressing truncated WT or mt PS2 (Fig. 7B). This was in sharp contrast to the marked increase in the percentage or level of A␤42 from cells with fl mt PS2 (52.1%). The levels as well as percentages of A␤42 secreted from N2a cells expressing PS2 CTFs (10.1-14.2%) also were similar to those with fl WT PS2 or mock-transfected cells (Fig. 7B).
To examine if coexpression of mt PS2 NTF together with CTF reconstitutes overproduction of A␤42, we transiently transfected PS2/304ctf in N2a cells stably expressing WT or mt PS2/303stop. Upon co-transfection of PS2/304ctf, the total levels of secreted A␤ were decreased by ϳ50% both in WT and mt PS2/303stop stable cells relative to those in mock-transfected cells, whereas the A␤42 remained unchanged both in cells expressing WT (9.7% in double transfection versus 11.3% in mock transfection) and mt (11.1% in double transfection versus 10.3% in mock transfection) PS2/303stop (Fig. 7C).

DISCUSSION
In this study, we have clearly shown that (i) full-length PS2 harboring both N141I and M239V mutations in the same polypeptide induced overproduction of A␤42 at similar levels to those in cells expressing PS2 with one or the other of these PS2 mutations (i.e. N141I or M239V); (ii) NTFs or CTFs of PS2 expressed in cells predominantly localize in ER; and (iii) cells expressing partially truncated PS2 (eliminating the COOHterminal third of PS2 while retaining the endoproteolytic NH 2terminal fragment) and harboring a N141I mutation, as well as cells expressing COOH-terminal fragments of PS2, did not overproduce A␤42, and the levels or percentages of A␤42 were comparable with those in cells that expressed full-length and wild-type PS2 as well as fragments thereof.
The nature of the structural or functional changes of the polytopic membrane protein PS2 caused by the two known mutations is not fully understood at present. Regarding the N141I Volga German PS2 mutation, a PS1 mutation at the homologous site (N135D) was reported (36), and these homologous residues in PS1 and PS2 are located at the NH 2 -terminal flank (designated N-cap position) of the second transmembrane (TM2) domain, which is believed to be important in the accurate positioning of the transmembrane ␣-helix structure (37). Another PS2 mutation of the Italian type (M239V) is situated within the TM5 domain; a PS1 mutation linked to FAD at the homologous site (M233T) also was documented (38), and substitution of Met for Val was observed in multiple residues in the TM2 domain of PS1 (i.e. M139V and M146V) (8), suggesting that Met to Val substitution may cause some common structural changes in the TM domains of PS1 or PS2. Our observation that the N141I/M239V double mutation did not have additive effects on the increase in the levels or percentages of secreted A␤42 suggests that A␤42-promoting capacities of mt PS2 proteins reach the maximum level with a given single amino acid substitution (i.e. N141I or M239V). This contrasts with the recent observation that the A␤42-promoting effects of M146L/L286V double mutant PS1 were additive (39) and also with the clinical observation that FAD patients with PS1 mutations develop AD at a uniformly early age, whereas the age of onset in Volga German families with the N141I PS2 mutation is variable and relatively late (40). The reason for these discrepancies is not clear at present. However, one should con- sider the differences in the protein levels of endogenous PS1 versus PS2 in the brains of FAD patients (10,18). For example, it may be that the changes in A␤42-promoting effects of PS2 caused by a given single mutation per molecule is stronger than those with mt PS1, whereas the overall pathogenic effects of mt PS1 become more intense than those of mt PS2 because the total amount of PS1 proteins in neurons or brain tissues is higher than that of PS2.
NTFs or CTFs of PS2 of various sizes predominantly localized to ER. Recently, it has been shown that the NH 2 -terminal 166 residues, but not 138 residues, of PS2 are sufficient for the ER targeting, suggesting that the initial two transmembrane domains are necessary for ER localization (41). Our findings confirmed these observations with respect to the NTFs and further extended these data by showing that CTFs of PS2, including those corresponding to the COOH-terminal 103 amino acids (ALG-3) also localize to ER. Although the precise orientation of the membrane insertion of these CTFs is yet to be determined, the occurrence of "caspase-type" cleavage (32) of these CTFs in similar patterns to those observed in cells expressing fl PS2 suggests that the NH 2 -terminal portions (i.e. loop region of PS2) of these CTFs are properly oriented to the cytoplasmic side. Moreover, these CTFs harbor two transmembrane domains (i.e. TM7 and -8). Taken together, the COOHterminal region may harbor other ER-targeting signal sequences besides those in the NH 2 -terminal region, or alternatively, the presence of multiple (i.e. more than two) TM domains, but not particular subregions, of PS2 may determine its ER localization.
The most unexpected, yet intriguing, finding in this study was that cells expressing COOH-terminally truncated N141I mt PS2 that are equivalent to or longer than the endoproteolytic NH 2 -terminal fragment did not overproduce A␤42. This was surprising because most of the PS proteins in native cells or tissues (including brains) exist as NTF and CTF forms, and the NTFs contain six of the eight TM domains of PS molecules. Recently, it has been suggested that the levels of PS within cells are strictly regulated by competition for limiting cellular factors (29). Moreover, it was shown that NTF and CTF of PS1 or PS2 remain noncovalently bound to each other after cleavage forming a very stable complex (25,42,43) and that they may form a 100 -150-kDa molecular mass complex (43). Our finding that NTF of mt PS2 or CTF alone does not promote A␤42 overproduction supports the notion that the stable complex forms of PS NTF and CTF constitute the functional units under biological as well as pathological conditions. An alternative possibility would be that the nascent, full-length form of PS is functional and the cleavage at loop domain is a switch-off phenomenon. However, the observations that full-length PS is short-lived (44) and not readily incorporated into the stable complex (29,43) and that full-length PS is rare in native cells (27) render this possibility rather unlikely. The COOH-terminally truncated forms of PS1 (29) or PS2 (i.e. PS2/388stop in this study), which contain the proteolytic cleavage sites, were not cleaved, and the truncated PS1 did not influence the stable complex formation in stably transfected N2a cells (29). Taken together with our results that mt PS2/388stop did not promote overproduction of A␤42, it is highly conceivable that the condition under which nascent PS proteins are stabilized and properly cleaved to produce NTF and CTF and form a stoichiometric stable complex is the prerequisite for the normal or pathological function of PS (29,44). Our data that co-expression of mt PS2/303stop together with PS2/304ctf did not reconstitute overproduction of A␤42 further support the notion that the heterodimeric complex of NTF and CTF of PS derived from a full-length PS molecule is required for the pathological effects of PS mutations.
What is the nature of the limiting factor(s) that determines the integrity of the functional complex of PS molecules? Our finding that mt PS2/388stop, which comprises the NH 2 -terminal 87% of the entire length of PS2 as well as 7 of the 8 putative TM domains, failed to overproduce A␤42 strongly suggests that some critical subdomain in the COOH-terminal portion of PS2, or an as yet unidentified binding protein(s) that interacts with this domain, should play key roles in the formation of functional complex of PS. An alternative possibility would be that the integrity of the whole PS molecule, not particular subdomains, is required for the formation of the complex. Further efforts to define this subdomain of PS2 or identify binding proteins that are essential for the function as well as stabilization of PS should facilitate our understanding of the mechanisms whereby mt PS proteins influence the ␥-cleavage of ␤APP to overproduce A␤42, thereby leading to AD.