The Mechanism of γ-Secretase Activities through High Molecular Weight Complex Formation of Presenilins Is Conserved inDrosophila melanogaster and Mammals*

Mutations in presenilin 1 (PS1) and PS2 genes contribute to the pathogenesis of early onset familial Alzheimer's disease by increasing secretion of the pathologically relevant Aβ42 polypeptides. PS genes are also implicated in Notch signaling through proteolytic processing of the Notch receptor in Caenorhabditis elegans, Drosophila melanogaster, and mammals. Here we show thatDrosophila PS (Psn) protein undergoes endoproteolytic cleavage and forms a stable high molecular weight (HMW) complex inDrosophila S2 or mouse neuro2a (N2a) cells in a similar manner to mammalian PS. The loss-of-function recessive point mutations located in the C-terminal region of Psn, that cause an early pupal-lethal phenotype resembling Notch mutant in vivo, disrupted the HMW complex formation, and abolished γ-secretase activities in cultured cells. The overexpression of Psn in mouse embryonic fibroblasts lacking PS1 andPS2 genes rescued the Notch processing. Moreover, disruption of the expression of Psn by double-stranded RNA-mediated interference completely abolished the γ-secretase activity in S2 cells. Surprisingly, γ-secretase activity dependent on wild-type Psn was associated with a drastic overproduction of Aβ1–42 from human βAPP in N2a cells, but not in S2 cells. Our data suggest that the mechanism of γ-secretase activities through formation of HMW PS complex, as well as its abolition by loss-of-function mutations located in the C terminus, are highly conserved features inDrosophila and mammals.

Mutations in presenilin (PS1) 1 or PS2 genes account for the majority of early-onset familial Alzheimer's disease (FAD), and these mutations cause an increase in the ratio or levels of production of amyloid ␤ peptides ending at position 42 (A␤42), that most readily form amyloid deposits (1). Presenilins are polytopic integral membrane proteins that span the membrane eight times and undergo endoproteolysis (2). The endoproteolytic fragments of PS are incorporated into a high molecular weight (HMW) complex (3,4) and are highly stabilized (t1 ⁄2 ϭ ϳ20 h), whereas holoprotein is rapidly degraded (t1 ⁄2 ϭ ϳ2 h) (5).
PS is implicated in ␥-cleavage of ␤APP, the final step in the generation of A␤ peptides, as well as in the ␥-cleavage-like intramembranous proteolysis of various transmembrane proteins (e.g. Notch, ErbB4, E-cadherin, and LRP) (reviewed in Ref. 6). Although the precise role of PS in the intramembranous proteolysis still remains unknown, following lines of evidence suggest that PS is a catalytic component of ␥-secretase. First, the ablation of PS genes in mice inactivated the total ␥-secretase activities (7,8). Second, mutating either of the two conserved aspartate residues within the transmembrane domains (TMD) 6 and 7 of PS inhibited the ␥-secretase activities (9). Third, the ␥-secretase activity solubilized by a mild detergent, CHAPSO, was immunoprecipitated by antibodies against PS1 in HMW fractions (10). Lastly, the transition-state analogue ␥-secretase inhibitors that are conjugated with photoaffinity labeling and/or biotin tags directly labeled the PS fragments (11)(12)(13). Recently, functional ␥-secretase complex containing PS fragments was partially purified by an immobilized ␥-secretase inhibitor (14). Taken together, it is strongly suggested that the stabilized HMW complex of PS represents the functional form of ␥-secretase and that the PS fragments harbor the catalytic center of ␥-secretase.
PS is an evolutionarily conserved protein that is present in every multicellular organism including vertebrates and invertebrates as well as plants, and the primary amino acid sequences of the C-terminal region of PS are highly conserved (15). We have previously shown that stabilization and formation of the HMW PS complex that are dependent on the integrity of the PS C terminus is required for the ␥-secretase activity (16). Missense mutations that replace the 1st proline of the C-terminal PALP motif, which is completely conserved in all PS family members, with leucine, lead to a loss-of-function of PS in Drosophila melanogaster presenilin (Psn) as well as in Caenorhabditis elegans Spe-4 (17,18). Moreover, Psn B3 allele, another loss-of-function mutant of Psn, that results in an amino acid substitution (G516E) in the C terminus of Psn has been reported (19). This glycine residue also is conserved in almost all known PS family members except for the C. elegans Hop-1 protein. We have previously shown that the 1st proline of the PALP motif is required for the stabilization, complex formation, and ␥-secretase activities of human PS in mamma-lian cells (15). However, the effects of the loss-of-function mutations (P507L or G516E) on the metabolism and complex formation of Psn polypeptides still remain unknown. In this study, we examined the modes of processing, complex formation, and function of Psn protein in Drosophila S2 cells as well as in mammalian cells and compared them with those of human PS.
Cell Culture and Transfection-Mouse neuro2a (N2a) neuroblastoma cells and SV40-transformed mouse embryonic fibroblasts (MEF) derived from PS1 Ϫ/Ϫ PS2 Ϫ/Ϫ littermates (provided by Dr. B. De Strooper) were maintained as described (15). Generation of stable N2a cell lines co-expressing ␤APP NL and Notch⌬E (NL/N) were described previously (25). Stable N2a NL/N cell lines expressing Psn or PS2 derivatives were generated by transfecting cDNAs using LipofectAMINE and selected in Dulbecco's modified Eagle's medium containing both hygromycin (Wako) at 250 g/ml and G418 (Calbiochem, San Diego, CA) at 500 g/ml. Transient transfection of cDNAs into MEF cells were performed using LipofectAMINE 2000 (Invitrogen) according to the manufacturer's instructions. After 24 h of transfection, 10 mM butyric acid was added for 24 h to drive protein expression.
Drosophila Schneider (S2) cells were maintained in Schneider's insect medium (Sigma) supplemented with 10% fetal bovine serum, 5% peptone, and penicillin/streptomycin (S2 medium) at 24°C (26). Transient transfection of cDNAs into S2 cells was performed using Cellfectin (Invitrogen) according to the manufacturer's instructions, and samples were collected after 48 h of transfection. Stable S2 cell lines were generated by transfection of cDNAs in pAc5.1/V5-His A vector together with those in pCoHygro (Invitrogen) vector (ratio of transfected cDNAs; 2:0.1 g) using Cellfectin and selection in S2 medium containing hygromycin at 250 g/ml.
Double-stranded RNA-mediated Interference (RNAi)-For the production of the double-stranded RNA (dsRNA), transcription templates that contained T7 RNA promoter sequences on each end were generated by PCR using the following oligonucleotides containing the T7 RNA polymerase binding site as primer pairs: 5Ј-TTAATACGACTCACTAT-AGGGAGAATGGCTGCTGTCAAT-3Ј for Psn, 5Ј-TTAATACGACTCAC-TATAGGGAGAATGGTGAGCAAGGGC-3Ј for EGFP as sense primers, 5Ј-TTAATACGACTCACTATAGGGAGAGACATCATTCCGACC-3Ј for Psn, 5Ј-TTAATACGACTCACTATAGGGAGATTACTTGTACAGCTC-3Ј for EGFP as reverse primers, respectively. dsRNAs were prepared from transcription templates by using MEGAscript T7 KIT (Ambion, Austin, TX) and transfected into S2 cells using Cellfectin. Cell lysates and conditioned media were harvested after incubation for indicated times.
Quantitation of A␤ by Two Site ELISAs-Two site ELISAs that specifically detect the C terminus of A␤ were used as described. BAN50 is a monoclonal antibody raised against a synthetic peptide of human A␤1-16; it preferentially reacts with the N-terminal portion of human A␤ starting at Asp-1, but does not cross-react with N-terminally truncated A␤ nor with rodent-type A␤ (20,28). BA27 and BC05 that specifically recognize the C terminus of A␤40 and A␤42, respectively, were conjugated with horseradish peroxidase and used as detector antibodies. Culture media were collected after an appropriate incubation period and subjected to BAN50/BA27 or BAN50/BC05 ELISAs as described (20,29).

Expression and Metabolism of Psn in Drosophila S2 or Mouse
N2a Cell Lines-Drosophila presenilin (Psn) gene encodes 508 -541 amino acid proteins with ϳ50% identity to its vertebrate counterparts (21). The occurrence of endoproteolytic cleavage of Psn protein in vivo and in the Drosophila S2 cell line has also been documented, although a detailed analysis on the metabolism of Psn polypeptides is yet to be performed (18,30). To examine the expression and metabolism of endogenous and transfected Psn proteins in S2 or mouse N2a cell lines, we stably transfected these cells with Psn and analyzed by immu- noblotting with antibodies against the N terminus or hydrophilic 6th loop of Psn (i.e. anti-GDN1 and anti-GDL1, respectively). Immunoblot analysis of lysates of untransfected S2 cells revealed a ϳ27-kDa N-terminal fragment (NTF) as well as a ϳ32-kDa C-terminal fragment (CTF) (Fig. 2A). These bands disappeared when the blots were probed by antibodies preadsorbed with immunogen proteins (data not shown). A faint band of ϳ55-60 kDa, corresponding to the full-length Psn protein, also was detectable. These results confirmed the previous reports on the endoproteolysis of Psn as well as the predominance of fragment forms as endogenous Psn, which was similar to those seen with mammalian PS (18,30).
We next analyzed the lysates of S2 cells stably transfected with Psn ( Fig. 2A). A ϳ55-kDa band corresponding to a FL Psn polypeptide was detected by the N-and C-terminal antibodies, whereas the levels of NTF and CTF did not increase, suggesting that the levels of Psn fragments also are regulated by a "limiting co-factor" in a similar manner to mammalian PS (31). To further characterize the metabolism and function of Psn, we stably transfected the Psn cDNA into a mouse N2a cell line stably expressing both ␤APP NL and Notch⌬E (N2a NL/N cell line) ( Fig. 2A). Immunoblot analysis revealed that Psn polypeptides expressed in N2a NL/N cells underwent endoproteolysis to give rise to NTF and CTF of the same molecular weights as the endogenous ones in S2 cells. Moreover, the overexpression of Psn in N2a NL/N cells compromised the accumulation of endogenous murine PS fragments, suggesting that Psn retains the capacity to replace the endogenous PS by competing for limiting cofactor(s) in a similar fashion to that observed with mammalian PS (Fig. 2B).
Fragments of mammalian PS are highly stabilized and incorporated into HMW protein complexes of ϳ200 -600 kDa that are distributed in the ER as well as in Golgi/TGN, whereas holoproteins are rapidly degraded, fractionated in the low molecular weight (LMW) range of ϳ100 -200 kDa, and exclusively distributed in ER (15). To examine the stability of Psn protein, we treated S2 or N2a NL/N cells stably expressing Psn with cycloheximide (CHX) (Fig. 3A). The levels of endoproteolytic fragments of Psn did not decrease during CHX treatment of 10 -12 h, whereas the Psn holoproteins were rapidly degraded similarly to mammalian PS holoproteins. To examine the capacity of Psn proteins to form HMW complexes, we solubilized the membrane fractions of S2 or N2a NL/N cells in 1% CHAPSO, and separated the extracted proteins on a linear glycerol velocity gradient (Fig. 3B). Endoproteolytic fragments derived from Psn were predominantly distributed in the HMW range of 232-443 kDa, whereas Psn holoproteins were fraction- ated in the LMW range of 140 -232 kDa. Moreover, subcellular fractionation studies using discontinuous Iodixanol gradients showed that endoproteolytic Psn fragments were recovered in fractions containing ER vesicles as well as Golgi membranes, whereas holoproteins were detected in ER fractions in N2a NL/N cells (Fig. 3C and data not shown). These data suggest that Psn proteins are metabolized in Drosophila S2 cells by a similar cellular machinery to that working in mammalian cells, and appropriately metabolized by a mammalian PS-metabolic pathway (i.e. properly folded, assembled with binding partners, stabilized, and forming HMW complex) in mouse N2a cells.
Mutations of the Highly Conserved Amino Acid Residues at the C Terminus of Psn or Their Equivalents in Human PS2 Affect the Formation of Stable HMW PS Complex-The formation of the stabilized HMW complex of mammalian PS, that requires the integrity of the conserved PS C terminus, is essential to the acquisition of ␥-secretase activity, and an aspartate residue within 7th TMD (TMD7) is crucial to the ␥-secretase activity in mammalian PS (9,32). To verify the effects of missense mutations in Psn that cause Notch (i.e. loss-of-function) phenotype in Drosophila in vivo, on the metabolism of Psn polypeptides, we introduced the two types of amino acid substitutions (i.e. P507L or G516E) and stably expressed the mutant Psn in N2a NL/N cells. In addition, we established N2a NL/N cells stably coexpressing Psn carrying D461A mutation that replaces the highly conserved aspartate residue in the TMD7 with alanine, to see if it works as a dominant negative mutant on ␥-cleavage as in mammalian PS (9,15,32). Structures of the Psn derivatives used here are schematically shown in Fig. 1. Immunoblot analysis of cell lysates showed that neither Psn/P507L, Psn/G516E nor Psn/D461A underwent endoproteolysis to give rise to NTF and CTF that normally occurs with wild type Psn (Fig. 4A). The replacement of endogenous PS1 did not occur in N2a NL/N cells coexpressing Psn/P507L or Psn/G516E. Upon CHX treatment of the N2a cells, the Psn/ P507L or Psn/G516E holoproteins were rapidly degraded in a similar manner to wild type Psn holoprotein ( Fig. 4B and data not shown). In contrast, the overexpression of Psn/D461A resulted in a complete replacement of endogenous murine PS1 fragments, and a portion of Psn/D461A was stabilized as a holoprotein, as previously described in aspartate mutants of mammalian PS (i.e. PS1/D385A, PS2/D366A) (9,15,32). We next analyzed the HMW complex formation of Psn and its derivatives (Fig. 4C). The unstable Psn/P507L or Psn/G516E holoproteins were fractionated exclusively in the LMW range. In contrast, Psn/D461A, which was stabilized but not cleaved, was present as holoproteins broadly within LMW and HMW ranges in a similar manner to that of mammalian PS2/D366A (15).
To further elucidate the structural and functional roles of the conserved glycine residue at the C terminus of PS, we constructed cDNAs encoding wild type or N141I FAD mutant human PS2 harboring a G423E mutation, which is equivalent to G516E mutation of Psn, and stably transfected them in N2a NL/N cells. Western blot analysis revealed that PS2/G423E was expressed as holoproteins but neither underwent endoproteolytic cleavage nor replaced endogenous PS1 CTF (Fig. 5A). FAD-linked N141I mutation did not affect the metabolism of G423E mutant PS2 polypeptides. We next analyzed the halflife and HMW complex formation of PS2/G423E. CHX treatment showed that PS2/G423E holoproteins were unstable (Fig.  5B). Moreover, the PS2/G423E polypeptides were fractionated exclusively in LMW fractions by glycerol velocity centrifugation, in a similar manner to unstable PS2 proteins (e.g. PS2 holoprotein or PS2/P414L) (15), indicating that the G423E mutation abolished the HMW complex formation of PS2 pro-tein (Fig. 5C). These results suggest that the conserved glycine residue in the C terminus of PS plays an important role for the stabilization and formation of HMW complex of PS polypeptides in diverse organisms including Drosophila as well as mammals, as we have previously shown with the conserved proline residue at the PALP motif (15).
Thus, the two loss-of-function mutations of Psn at the conserved amino acid residues at the C terminus abolished the stabilization and HMW Psn protein complex formation, and stabilized Psn proteins participated in the formation of HMW Psn complexes, whereas unstable Psn proteins formed only LMW protein complexes. Taken together, these data strongly suggested that the molecular mechanism of PS metabolism is preserved beyond species from Drosophila to humans.
␥-Secretase Activity of Psn in Mouse N2a Cells-To evaluate the ␥-secretase activity of Psn, we analyzed the levels of secreted A␤1-40 and A␤1-42 in conditioned media from N2a NL/N cells stably expressing wild type Psn or Psn/D461A by ELISAs (Fig. 6A). Surprisingly, overexpression of wild type Psn resulted in a ϳ5-7 fold increase in A␤42 secretion as compared with those secreted from cells expressing an empty vector or wild type human PS2. Whereas the percentage of A␤42 as a fraction of total A␤ (A␤1-40 ϩ A␤1-42) (%A␤42) secreted by untransfected N2a NL/N cells was ϳ10%, the %A␤42 secreted from N2a cells expressing wild type Psn was constantly elevated to ϳ50 -75%. Overexpression of Psn/D461A in N2a cells inhibited ␥-cleavage of ␤APP NL , resulting in a marked decrease in the secretion of both A␤1-40 and A␤1-42 accompanied by the accumulation of ␤APP C-terminal stubs (i.e. C83 and C99), that are the direct precursors of p3 and A␤, respectively (data not shown). We next analyzed the levels of secreted A␤ from N2a cells expressing Psn/P507L or Psn/G516E (Fig. 6A). In contrast to the expression of wild type Psn, the levels of A␤ or %A␤42 secreted from cells expressing Psn/P507L or Psn/G516E were comparable to those in cells with wild type FL PS2. These results indicated that the P507L or G516E mutation abrogated the FAD-linked mutant-like A␤42-promoting effect of Psn in N2a cells. We finally analyzed A␤ secreted from N2a NL/N cells expressing C-terminally modified PS2 (Fig. 6B). Total levels or %A␤42 of secreted A␤ from N2a cells expressing wild type or FAD mutant PS2/G423E was comparable to those in cells expressing wild type FL PS2. These data suggest that Psn/P507L, Psn/G516E, or PS2/G423E, that failed to undergo stabilization and HMW complex formation, lost the ␥-secretase activities, as we have previously observed with PS2/P414L, a PS2 equivalent of P507L mutant of Psn. These data further support our view that the stabilization and formation of HMW complex of PS mediated by the integrity of its C terminus is required for the ␥-secretase activity (15).
Psn is known to serve as a critical component for Notch signaling in vivo by executing the proteolytic release of Notch intracellular domain (NICD) at site-3 (1,6). To examine the activity of Psn in ␥-cleavage-like site-3 cleavage in mammalian cells, we transiently co-transfected wild type or mutant Psn, together with Notch⌬E, in an immortalized PS-null fibroblast cell line derived from PS1/PS2 double-knockout mice (7,15). Overexpression of wild type Psn restored the proteolytic generation of NICD, suggesting that Psn harbors a site-3 protease activity in mammalian cells. In sharp contrast, Psn/D461A, Psn/P507L, and Psn/G516E did not restore the proteolytic release of NICD in PS-null fibroblasts. We therefore conclude that Psn exhibits ␥-secretase activities that partially recapitu- late those of FAD-mutant PS (i.e. overproduction of A␤42) in mammalian cells, and that these activities are dependent on the formation of HMW PS complex as well as on the aspartate residue within the TMD7, in a similar manner to mammalian PS.
␥-Secretase Activity to Generate A␤ in Drosophila S2 cells-Psn-dependent ␥-secretase activity in Drosophila has been shown to cleave Notch and other transmembrane proteins in vivo (6,(33)(34)(35). The amino acid sequence of APPL, a Drosophila homologue of ␤APP, is not homologous to that of mammalian ␤APP especially within the TMD, and ␥-cleavage of APPL has not been documented (36). However, it has been shown that overexpression of the C-terminal 99 amino acid fragment of human ␤APP elicits the cleavage to generate A␤1-40 by a ␥-secretase-like activity in Drosophila SL-2 cells, although Drosophila cells lack ␤-secretase activity (37). To evaluate the ␥-secretase-like activity for proteolytic processing of the TMD sequence of human ␤APP in Drosophila S2 cells, we transiently transfected a cDNA encoding SC100, that corresponds to the C-terminal fragment of human ␤APP starting at the 1st residue of A␤ preceded by a signal peptide, and analyzed the conditioned media by ELISA (20,29). A␤ secretion was readily detectable in conditioned media of cells expressing SC100; surprisingly, however, %A␤42 was ϳ15%, which was in sharp contrast to the robust A␤1-42 overproduction in mouse N2a cells, that is mediated by the same PS species, i.e. wild type Psn (Fig. 7A). To exclude the possibility that ␥-secretase-like activity in S2 cells is incapable of producing excessive amounts of A␤1-42, we constructed a cDNA encoding SC100 harboring an isoleucine to phenylalanine substitution at residue 716 of ␤APP (SC100/I716F), that has been shown to cause robust increase in A␤1-42 secretion in COS cells (38). Transfection of SC100/I716F into S2 cells resulted in a dramatic increase in A␤1-42 secretion and simultaneous decrease in A␤40 secretion (Fig. 7B), suggesting that the endogenous ␥-secretase-like activity mediated by Psn normally cleaves the TMD sequence of human ␤APP predominantly at A␤40 position, but is capable of cleaving predominantly at position 42 under pathogenic conditions (e.g. ␤APP mutation) in S2 cells. Thus, Psn-dependent ␥-cleavage in S2 cells shows similar characteristics to those in mammalian cells, whereas it may be shifted to position 42 by some unknown mechanism in mouse N2a cells.
To examine whether Psn plays an essential role in A␤ generation by a ␥-secretase-like activity in S2 cells, we generated a S2 cell line stably expressing SC100 (S2-SC100) and suppressed the expression of endogenous Psn gene by doublestranded RNA (dsRNA)-mediated interference (RNAi). After a 48-h transfection of Psn dsRNA, the expression of Psn polypeptide in fragment forms was completely and specifically abolished in S2-SC100 cells, although the expression of other endogenous or exogenous genes (i.e. tubulin and EGFP) was not affected ( Fig. 7C and data not shown for EGFP co-transfection). After incubation in fresh media for additional 24 h, the cell lysates and conditioned media were analyzed. Immunoblot analysis revealed an accumulation of SC100 as well as of a ϳ10-kDa polypeptide comigrating with C83 of mammalian cells. The latter band presumably represents the SC100 derivative cleaved by an ␣-secretase-like activity, that has been reported in Drosophila and SL-2 cells (37). No A␤ secretion was observed in conditioned media, suggesting that the total suppression of the expression of Psn by RNAi resulted in a complete loss of ␥-secretase activity (Fig. 7D). Thus, Psn-dependent ␥-secretase activity is required for A␤ generation from a human ␤APP derivative (i.e. SC100) in Drosophila S2 cells.

DISCUSSION
In this study, we examined the metabolism and function of Psn protein in mammalian and Drosophila cell lines and showed the following. (i) Psn is metabolized in a manner similar to that of human PS. (ii) Loss-of-function mutations of Psn that result in an early pupal-lethal phenotype in Drosophila completely disrupt the stabilization and HMW complex formation of Psn polypeptides. (iii) Overexpression of wild type Psn in N2a cells increases the secretion of A␤1-42, whereas alanine substitution of the aspartate at position 461, that corresponds to one of the putative catalytic aspartates in mammalian PS, abolishes the ␥-secretase activity. (iv) Expression of Psn in PS-null murine fibroblasts restores the ␥-like site-3 cleavage Notch, and (v) the disruption of the expression of Psn by double-stranded RNAi completely abolish the ␥-secretase activity in S2 cells. These data suggest that the formation of HMW complex containing PS underlying the ␥-secretase activities is a highly conserved process that is common to Drosophila and mammals.
Psn polypeptides underwent endoproteolysis to give rise to NTF and CTF in cultured cells as previously documented (18,30). These fragments were highly stabilized and formed a HMW complex in a similar manner to mammalian PS. Moreover, overexpression of wild type Psn resulted in a complete replacement of endogenous PS in mammalian cells. These results suggest that Drosophila Psn protein is metabolized in a similar manner to mammalian PS and competes for the "limiting cofactor" with mammalian PS (5, 31). We further studied the molecular mechanism of loss-of-function caused by Psn 46 and Psn B3 alleles (18,19), and found that these mutations (P507L and G516E, respectively) completely abolished the stabilization, HMW complex formation as well as the ability to replace endogenous PS, of Psn polypeptides. Thus, proper metabolism of Psn, that requires the integrity of its C-terminal region including a couple of highly conserved residues (i.e. Pro 507 or Gly 516 ), is essential to its ␥-secretase-like function in a similar manner to mammalian PS. Missense mutations leading to substitution of one or the other of these residues lead to Notch phenotype probably due to failure in ␥-secretase-like activities mediated by Psn, although the precise nature of alteration in the structure of Psn caused by these single amino acid substitutions has yet to be elucidated. Formation of NTF and CTF as well as replacement of endogenous PS have been shown to occur in HEK293 cells transfected with zebrafish (Danio rerio) PS (39), whereas the C. elegans PS, i.e. SEL-12, failed to recapitulate these features in mammalian cells (Ref. 40). 2 The amino acid sequences of the C-terminal ϳ11 residues of PS family proteins are highly homologous among mammals, zebrafish, and Drosophila, whereas they are relatively divergent in C. elegans Sel-12 and Spe-4. Taken together, it is strongly suggested that the integrity of the C terminus, as well as a couple of highly conserved amino acid residues flanking this region including the PALP motif (Ref. 15; the first proline corresponding to Pro 507 in Psn), play an important role in the common molecular mechanism underlying the ␥-secretase-like functions that are conserved from Drosophila to mammals.
We have generated a Drosophila S2 cell line stably expressing the C-terminal stub of human ␤APP (SC100), and found that endogenous Psn forms HMW protein complexes in a similar pattern to mammalian PS, and that ␥-secretase-like activity cleaves SC100 to secrete A␤. Moreover, RNAi-based "knockdown" technique confirmed that A␤-generating protease activities in S2 cells are dependent on Psn expression, as previously shown for the Notch site-3 activities (41). The present experiment also highlights the usefulness of RNAi in the molecular dissection analysis of the PS complex; indeed, Francis et al. (42) have recently identified two additional cofactors of Psn, i.e. APH-1 and PEN-2, using genetic screen in C. elegans, and demonstrated by RNAi that expression of these proteins are essential to the A␤-generating activities of Drosophila cells transfected with Notch or APP C100, using a cellular system similar to ours.
Another intriguing finding in this study was the difference in preponderant ␥-cleavage sites by wild type Psn in N2a and S2 cells: In N2a cells, overexpression of wild type Psn caused a robust A␤1-42 overproduction, which was dependent on the aspartate residue in TMD7. Similar overproduction of A␤1-42 by transfection of "wild type" PS has also been observed with zebrafish PS1 (39). We compared the deduced amino acid sequence of Psn for variations at positions with known mutations causing FAD in human PS, and found that ϳ8 amino acid residues in wild-type human PS1 (e.g. Met 84 , Met 139 , Cys 263 ) are different from the corresponding codon in Psn (Lys 106 , Leu 161 , Ser 285 , respectively), where FAD-linked mutations have been identified (although the substituted amino acids are not identical). One possibility is that the naturally occurring differences in amino acid sequences, which coincidentally behaved like human FAD mutations, caused the overproduction of A␤1-42 in mammalian cells. In contrast, Psn-dependent ␥-secretase activity in Drosophila S2 cells did not cause A␤1-42 overproduction and the %A␤42 was at normal level (ϳ15%). The molecular mechanism of overproduction of A␤1-42 caused by FAD-linked amino acid substitutions in human PS still remains unknown. However, our observation that overexpression of SC100/I716F mutant in S2 cells resulted in an enormous secretion of A␤1-42 like in mammalian cells indicates that Psn-dependent ␥-secretase activity in S2 cells retains the capacity to cleave the TMD sequence of ␤APP at A␤42 position. Another speculative idea is that the differences in the composition or structure of components, as well as in the three-dimensional structures, of PS complexes might have caused the differences in substrate recognition or cleavage sites. Alternatively, the difference in the composition and metabolism of membrane lipids between mammalian and Drosophila cells may underlie the distinct behaviors in ␥ 42 -secretase activities, related to the unusual enzymatic characteristics of ␥-secretase to take place within membranes. In fact, it has been shown that phosphatidylethanolamine is the predominant phospholipids in cellular membranes of Drosophila, whereas the major phospholipid in mammalian cells is phosphatidylcholine (43). Genetic, biochemical and proteomic approaches to determine the components of PS complex in mammalian and S2 cells, as well as the efforts to reconstitute ␥-cleavage in vitro, will clarify these problems.