Evidence That Intramolecular Associations between Presenilin Domains Are Obligatory for Endoproteolytic Processing*

Mutations in genes encoding presenilins (PS1 and PS2) cosegregate with the majority of early onset cases of familial Alzheimer’s disease. PS1 and PS2 are polytopic membrane proteins that undergo endoproteolytic cleavage to generate stable NH2- and COOH-terminal derivatives (NTF and CTF, respectively). Several lines of evidence suggest that the endoproteolytic derivatives are likely the functional units of PSin vivo. In the present report, we examine the disposition of PS NTF and CTF assemblies in stable mouse N2a neuroblastoma cell lines expressing human PS polypeptides. We show that exogenous expression of PS1 NTFs neither assemble with endogenous CTF nor exhibit dominant negative inhibitory effects on the endogenous PS1 cleavage and the accumulation of derivatives. In cells co-expressing PS1 and PS2, PS1- and PS2-derived fragments do not form mixed assemblies. In contrast, cells expressing a chimeric PS1/PS2 polypeptide form stable PS1 NTF-PS2 CTF assemblies. Moreover, expression of chimeric PS1/PS2 polypeptides harboring a familial early onset AD-linked mutation (M146L) elevates the production of Aβ42 peptides. Our results provide evidence that assembly of structural domains contained within NH2- and COOH-terminal regions of PS occur prior to endoproteolytic cleavage.

reviewed in Ref. 3). Approximately 10% of all cases of AD, classified as familial early onset AD (FAD) (age of onset Ͻ60 years), is caused by autosomal dominant inheritance of mutations in genes encoding APP (for a review, see Ref. 4), presenilin 1 (PS1) (5) and presenilin 2 (PS2) (6 -8). Expression of FAD-linked APP mutations leads to increased production of highly fibrillogenic A␤ species ending at residue 42 (or 43) (A␤42) (reviewed in Ref. 3). In addition, expression of FADlinked PS1 and PS2 variants alters APP processing in a manner that leads to elevated production of A␤42 (9 -14). Thus, genetic mutations in all the three genes that cosegregate with FAD increase the production of the pathogenic A␤42 peptides. The role that PS plays in facilitating proteolytic cleavage of APP at the ␥-secretase site is not known, but PS1-deficient neurons exhibit selective defects in the production of A␤ (15,16).
PS1 and PS2 (PS) are homologous polytopic membrane proteins (17) that are subject to endoproteolytic cleavage (12, 18 -20). In previous efforts, we documented that PS1-derived 28-kDa NH 2 -terminal (NTF) and 17-kDa COOH-terminal fragments (CTF) are the preponderant PS1-related species that accumulate in vivo (18). The PS NTF and CTF accumulate to 1:1 stoichiometry, and the absolute levels of PS-derived fragments are established by a highly regulated and saturable mechanism such that overproduction of human PS fragments led to diminution of mouse PS fragments (18). In transfected cells overexpressing PS, only a small fraction of full-length PS is converted into stable fragments (t1 ⁄2 Ͼ 12 h); excess fulllength PS polypeptides are rapidly degraded (t1 ⁄2 ϳ 1 h) (18,(21)(22)(23)(24)(25). Moreover, accumulation of PS1 and PS2 fragments are coordinately regulated, providing strong evidence that association with limiting cellular components regulates the abundance of PS fragments (21). More recently, using in situ chemical cross-linking and co-immunoprecipitation approaches, we demonstrated that the NTF and CTF of either PS1 or PS2 can be coisolated (26). The accumulated data support our view that the more stable endoproteolytic derivatives are the biologically functional forms of PS. Arguing against this notion, Citron et al. (27) reported that expression of a truncated PS1 species corresponding to the human PS1 NTF with a FAD mutation failed to increase A␤42 production; the authors suggested that the pathogenic effect of mutant PS is elicited by the full-length molecule and not the endoproteolytic derivatives.
In this report, we examine the assembly of PS1 and PS2 derivatives in stable mouse N2a neuroblastoma cell lines expressing full-length PS, recombinant PS1 NTF, or chimeric PS1/PS2 polypeptides. We demonstrate that transgene-derived COOH-terminally truncated human PS polypeptides that correspond to PS1 NTF (residues 1-298) fail to co-assemble with endogenous PS1 CTF and are rapidly degraded. In cells co-expressing PS1 and PS2, processed fragments derived from full-length PS polypeptides form assemblies consisting entirely of either PS1 NTF⅐CTF or PS2 NTF⅐CTF. Based on these results, we conclude that the association of NH 2 and COOH domains of the precursor PS polypeptide precedes endoproteolytic cleavage and that PS1 and PS2 do not form heteromultimeric assemblies. Interestingly, fragments derived from proteolytic processing of a chimeric PS1/PS2 precursor polypeptide are capable of forming stable heteromeric PS1 NTF⅐PS2 CTF assemblies. Furthermore, expression of a FAD-linked M146L mutant PS1/PS2 chimera elevated the production of A␤42 peptides, indicating that the gain of function properties associated with the FAD-linked mutation is preserved in a PS1/PS2 chimera. Our results are consistent with a model where stabilization of intramolecular interaction(s) between domains within the NH 2 -and COOH-terminal regions of PS1 is a prerequisite for endoproteolytic processing.
Cell Culture and Generation of Stable Cell Lines-Mouse N2a neuroblastoma cells were cultured in 1:1 Dulbecco's minimal essential medium/OptiMEM (Life Technologies, Inc.) supplemented with 10% fetal calf serum. Monkey COS-1 cells were cultured in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum. Stable N2a cell lines expressing human PS1 and PS2 were described (18,21). Stable N2a lines expressing PS1 NTF or PS1/PS2 were generated by transfecting 0.5 g of the respective expression plasmids (see above). Expression of transgene-derived PS polypeptides in G418-resistant lines was determined by Western blot analysis with PS1 NT and ␣PS2Loop antisera (26). Cycloheximide was added to the culture medium at a final concentration of 30 g/ml to block protein synthesis in some experiments.
Co-immunoprecipitation and Western Blot Analyses-Detergent lysates were prepared by lysing cells in cold PBS containing 0.25% ndodecyl ␣-D-maltoside (Sigma) as described previously (26). We obtained similar results when lysates were prepared using buffers containing 1% digitonin (29). Aliquots of lysates were incubated with 3 l of antisera at 4°C for 2 h. PS1 NTF and CTF were immunoprecipitated using antisera PS1 NT and ␣PS1Loop, respectively (18,26). PS2 CTF was immunoprecipitated using ␣PS2Loop antiserum (21). A rat monoclonal antibody (mAb) specific for human PS1 (30) (provided by Dr. Allan I. Levey, Emory University School of Medicine, Atlanta) was used to immunoprecipitate transgene-derived human PS1 NTF. To generate antibodies against the PS2 NH 2 -terminal region, New Zealand White rabbits were immunized with a glutathione S-transferase fusion protein containing residues 1-75 of human PS2. The specificity of the resultant antiserum, termed PS2 NT , for PS2 was established by Western blot analysis of lysates prepared from stable N2a cell lines expressing PS1 or PS2 (data not shown). As negative controls, we used SOD1, an antiserum raised against human erythrocyte Cu/Zn superoxide dismutase 1 (31), and normal rabbit serum (Sigma). Immunoprecipitates were collected using protein A-agarose, washed twice in lysis buffer, and resuspended in Laemmli buffer. Proteins were eluted by incubation at 37°C for 15 min and analyzed by immunoblotting.
Quantification of A␤ by Two-site ELISAs-Monkey COS-1 cells were transfected by the DEAE-dextran method with cDNAs encoding PS1, PS2, or PS1/PS2 polypeptides and cDNAs encoding either full-length APP or the COOH-terminal 100 amino acids of APP as described (12,32). Conditioned media were collected 48 h after transfection, and secreted A␤ species ending at 40 or 42/43 were quantified using BNT77/ BA27 or BC05 two-site ELISAs (33-35) as described previously (12,32). Data obtained from full-length APP or APP C100 cDNA transfections were combined for analysis. To facilitate comparison between media from cells transfected with different cDNA, the ratio of A␤42/total A␤ was calculated for each sample and statistically examined using analysis of variance followed by Fisher LSD test and expressed as mean Ϯ S.E.

Transgene-derived Human PS1 NTF Fails to Assemble with
Endogenous PS1 CTF-In previous studies, we and others reported that the NTF and CTF derived from endoproteolysis of PS1 (or PS2) are noncovalently associated to each other in vivo (26,36,37). To further characterize the NTF and CTF assemblies, we generated stable mouse N2a neuroblastoma cell lines transfected with a plasmid that encodes human PS1 residues 1-298 (analogous to NTF). Expression of transgene-derived human PS1 NTF was assessed by Western blotting using PS1 NT , a polyclonal PS1 antiserum that recognizes epitopes in the NH 2 terminus of PS1 (26). Transgene-derived human PS1 NTF was readily detected in several independent stable N2a cell lines (Fig. 1A, top, lanes [3][4][5]. Consistent with previous reports (10,11,18,21), the recombinant human PS1 NTF exhibited accelerated migration relative to the mouse PS1 NTF on SDS-polyacrylamide gel electrophoresis (lane 1) and comigrated with the human NTF derived from full-length human PS1 (lane 2). Previously, we reported that the expression of human full-length PS1 in mouse N2a cells resulted in a decrease in the levels of murine PS1-derived NTF and CTF (21). To examine whether the expression of recombinant human NTF abrogated accumulation of endogenous murine PS1 derivatives, we performed Western blots using ␣PS1Loop, an antiserum that reacts with epitopes in the hydrophilic loop domain of PS1 (18). The steady-state levels of mouse PS1 CTF in three independent N2a lines examined were essentially identical to N2a cells transfected with empty vector (21) (Fig. 1A, bottom), indicating that human NTF did not interfere with the accumulation of mouse CTF.
Recently, Steiner and colleagues (25) reported that recombinant human PS1 NTF expressed in stably transfected human 293 cells undergoes rapid degradation mediated by the proteasome. While endogenous NTF and CTF were detected in a stable human 293 cell line expressing recombinant human NTF, it could not be determined whether high level expression of recombinant human NTF affected endogenous NTF because of the comigration of endogenous and exogenous NTFs (25,27). We addressed this issue in our stable N2a cell lines expressing human PS1 NTF. A PS1 NT -reactive species co-migrating with the mouse NTF was visible in lysates of stable N2a cells expressing low levels of recombinant human NTF (Fig. 1A, top,  compare lanes 1 and 3). However, the mouse NTF and the recombinant human NTF were not sufficiently resolved on SDS-polyacrylamide gel electrophoresis to convincingly demonstrate whether or not the expression of human NTF influenced the accumulation of mouse NTF. Therefore, we used an alternate strategy that takes advantage of the findings that while the recombinant human NTF is short lived (25), endogenous PS-derived NTF has a half-life of several hours (22,23). We incubated stable N2a cell line PS1 298 .6, which expresses high levels of human NTF (Fig. 1B, lane 1), with cycloheximide to arrest protein synthesis and examined the decay of human NTF over a 6-h interval. Immunoblot of cell lysates collected at various time points revealed that the faster migrating PS1 NTreactive species, which corresponds to recombinant human NTF, degraded rapidly (t 1/2 ϳ 1.5 h) after the addition of cycloheximide. The slower migrating PS1 NT -reactive species, which corresponds to mouse PS1 NTF was readily visible after 1 h of cycloheximide treatment and was the only remaining PS1 NTF after 4 h of cycloheximide treatment (Fig. 1B, top). Reprobing of the blots using ␣PS1Loop antiserum revealed that the levels of mouse CTF remained unchanged during the 6-h cycloheximide treatment (Fig. 1B, bottom). Similar results were obtained in cell lines expressing low or high levels of human PS1 NTF (data not shown). These results provide evidence for the presence of stable pools of endogenous NTF in N2a cell lines expressing high levels of recombinant human NTF. Furthermore, the short half-life of transgene-derived human NTF in our stable N2a cell lines is entirely consistent with previous studies (25). From these results, we conclude that the expression of transgene-derived human NTF failed to interfere with the endoproteolysis of endogenous murine PS1, or the accumulation of mouse PS1 derivatives.
We then examined whether the transgene-derived human NTF associated with endogenous mouse PS1 CTF. Previous studies demonstrated that after repeated immunodepletion of PS1 NTF⅐CTF complexes, only recombinant human NTF re-mained in the lysates of human 293 cells expressing human PS1 NTF (25). We were interested in directly testing whether we could isolate complexes containing human PS1 NTF and mouse PS1 CTF using a human PS1-specific antibody. To distinguish between the transgene-derived human NTF and endogenous mouse NTF, we used a rat mAb that reacts specifically with the human PS1 NH 2 terminus (21,30). Stable N2a cells expressing human PS1 NTF were lysed under nondenaturing conditions, and aliquots of the lysates were incubated with either the anti-human PS1 mAb or polyclonal PS1 NT antiserum (which reacts with both human and murine PS1). The immunoprecipitates were first probed using PS1 NT to confirm successful immunoprecipitation of PS1 NTF by both antibodies (Fig. 1C, top panels, lanes 1 and 3). As expected, control immunoprecipitations performed using monoclonal anti-Myc antibody, 9E10, and polyclonal anti-superoxide dismutase antibody, SOD1, failed to isolate PS1 NTF (lanes 2 and 4). Reprobing of the blots using ␣PS1Loop antiserum revealed the presence of co-purified mouse PS1 CTF in immunoprecipitates prepared using PS1 NT antiserum (Fig. 1C, bottom panels, lane  3), but not by using anti-human PS1 mAb (lane 1). These results demonstrate that transgene-derived recombinant human NTF could not assemble into complexes containing the endogenous murine PS1 CTF.
Cells Co-expressing PS1 and PS2 Do Not Form heteromeric PS1⅐PS2 Assemblies-Both PS1 and PS2 have been reported to form heteromeric assemblies and migrate as high molecular weight complexes on size exclusion columns and velocity density gradients (29,36,37). However, it is not clear whether PS1 derivatives form heteromeric complexes with PS2 derivatives. To address this issue, we used a stable N2a cell line transfected with human full-length PS2 cDNA. As we reported previously (21), high level expression of human PS2 in mouse N2a cells results in the diminution of mouse PS1 derivatives (Fig. 2A,  lane 6). Therefore, we chose a cell line that expresses low levels of human PS2 (lane 2) for the present studies; PS1 derivatives are only partially replaced in this cell line (compare lanes 4 and 5). We performed co-immunoprecipitation analyses using PS1 and PS2 antibodies: PS1 derivatives were immunoprecipitated . C, stable N2a cells expressing human PS1 NTF were lysed under nondenaturing conditions, and lysates were used to co-immunoprecipitate PS1 NTF using polyclonal PS1 NT antiserum, a PS1 mAb (which reacts specifically with human PS1 NH 2 -terminal epitopes), or control Myc and SOD1 antibodies. The resulting immunoprecipitates and total lysates (corresponding to 1 ⁄20 of the volume used for immunoprecipitations) were probed with PS1 NT (top panels) or ␣PS1Loop (bottom panels). PS1 NTF and CTF are indicated by arrowheads. The slower migrating ␣PS1Loop-reactive band probably represents the phosphorylated form of PS1 CTF (36,43). The high molecular weight PS1NT-reactive species (lane 5) represent nonspecific aggregation of overexpressed human NTF. The small decrease in the CTF signal in lane 5 is not reproducible and is seen only in samples with aggregated human NTF.
using antibodies PS1 NT and ␣PS1Loop; PS2 NTF and CTF were immunoprecipitated using PS2 NT and ␣PS2Loop, respectively. In parallel, polyclonal SOD1 antiserum and normal rabbit serum were used in control immunoprecipitations. The immunoprecipitates were analyzed by sequential immunoblotting using PS1 and PS2 antisera. As expected, PS1 NT and ␣PS1Loop antisera co-immunoprecipitated PS1 CTF and NTF, respectively (Fig. 2B, top panels, lanes 1 and 5), but failed to co-purify PS2 derivatives (Fig. 2B, bottom panels, lanes 1 and  4). Furthermore, PS2 NT and ␣PS2Loop antisera co-immunoprecipitated PS2 CTF and NTF, respectively (Fig. 2B, bottom  panels, lanes 2 and 6), but failed to co-purify PS1 derivatives (Fig. 2B, top panels, lanes 2 and 6). These results demonstrate that the derivatives resulting from endoproteolysis of PS1 and PS2 form either PS1 NTF⅐PS1 CTF or PS2 NTF⅐PS2 CTF assemblies but do not form mixed complexes containing the derivatives from both of the precursor polypeptides.
The Endoproteolytic Derivatives from a PS1/PS2 Chimeric Polypeptide Can Form Functional PS1/PS2 Assemblies-PS1 and PS2 have similar protein structures, share a high degree of amino acid sequence identity, and undergo endoproteolytic cleavage, and the resulting fragments derived from each precursor associate with each other. Despite these similarities, our results indicate that fragments derived from endoproteolytic cleavage of PS1 and PS2 fail to form mixed complexes. The absence of complexes containing PS1 NTF and PS2 CTF or PS2 NTF and PS1 NTF strongly suggests that the noncovalent association between the NH 2 -and COOH-terminal structural domains of the precursor polypeptide may precede endoproteolytic cleavage. If this prediction were correct, one would expect that endoproteolysis of a chimeric polypeptide containing PS1 NTF and PS2 CTF should result in the generation of processed PS1 NTF and PS2 CTF derivatives that can be co-immunoprecipitated. To test this hypothesis, we generated stable N2a cell lines transfected with an expression plasmid encoding a chimeric polypeptide corresponding to amino acids 1-280 of human PS1 fused to amino acids 287-448 of human PS2. Expression of the PS1/PS2 chimera was assessed in several cell lines by Western blotting using PS1 NT and ␣PS2Loop antisera (data not shown). The chimeric polypeptide was endoproteolytically processed to generate fragments corresponding to PS1 NTF and PS2 CTF, and in cell lines expressing high levels of the PS1/PS2 chimera, endogenous PS1 derivatives were replaced (Fig. 3A). Thus, the metabolism of the PS1/PS2 chimera was similar to that of PS1 or PS2, and the accumulation of the resulting derivatives (PS1 NTF and PS2 CTF) was similar to that of the fragments derived from full-length PS1 or PS2. Next, we performed co-immunoprecipitation studies to assess whether the fragments derived from endoproteolytic processing of the PS1/PS2 chimera could be co-immunoprecipitated. For these studies, we chose a stable cell line in which the endogenous mouse PS1 derivatives are incompletely replaced. Nondenaturing lysates were subject to immunoprecipitation with PS1 and PS2 antibodies. As expected, PS1 NT antiserum co-immunoprecipitated endogenous murine PS1-derived CTF (Fig. 3B,  top panels, lane 1). Interestingly, reprobing of the blot using ␣PS2Loop antiserum revealed that PS1 NT antiserum also copurified PS2 CTF (bottom panels, lane 1). Furthermore, ␣PS1Loop antiserum co-precipitated the slower migrating mouse PS1 NTF, whereas ␣PS2Loop antiserum co-precipitated the faster migrating human PS1 NTF (top panels, lanes 7 and 8, respectively). Because endogenous murine PS2 NTF is not detectable in total lysates prepared from the PS1/PS2 cell line used for this experiment (bottom panel, lane 12), we failed to detect any signal for PS2 CTF or NTF in co- precipitations  (bottom panels, lanes 2 and 8, respectively). These results dem-onstrate that the PS1 NTF and PS2 CTF are capable of forming mixed complexes when derived from cleavage of a chimeric polypeptide and strongly support our view that interaction between NH 2 -and COOH-terminal regions of a precursor PS polypeptide precedes endoproteolysis.
Next we examined whether the chimeric PS1 NTF⅐PS2 CTF assembly is stable. To address this issue, we incubated cells expressing PS1/PS2 chimera with cycloheximide to block protein synthesis. In parallel, we incubated cells expressing fulllength PS1 or PS2 with cycloheximide. Detergent lysates prepared at the end of the incubation period were analyzed by immunoblotting using PS1 and PS2 antisera. As expected from previous studies (21)(22)(23)(24), in stable N2a cells expressing PS1 the full-length PS1 molecules were rapidly degraded, while the NTF and CTF were stable (Fig. 4A). Similar results were obtained in stable cell lines expressing PS2 (data not shown). Analysis of lysates from stable N2a cells expressing the PS1/ PS2 chimera revealed that full-length PS1/PS2 polypeptides were also rapidly degraded, while the endoproteolytic fragments derived from the chimera remained stable over a 6-h incubation period in the presence of cycloheximide (Fig. 4A). From these results, we conclude that endoproteolytic fragments derived from the processing of PS1/PS2 chimera are long lived.
Finally, we examined whether the PS1/PS2 chimera could influence APP processing and elevate the production of A␤42 peptides. For these studies, we constructed expression plasmids encoding PS1/PS2 chimeric polypeptide that harbors the FAD-linked M146L substitution. COS cells were co-transfected with plasmids encoding wild-type or mutant PS1/PS2, and APP (or C100; see "Materials and Methods"). In parallel, we also transfected cells with plasmids encoding wild-type or mutant full-length PS1 or PS2. The levels of secreted A␤ x-40 and x-42 peptides in conditioned media of transfected cells were quantified using a two-site ELISA assay as described (33)(34)(35). To account for variations in the levels of A␤40 and A␤42 due to differences in transfection efficiencies of different expression plasmids, we calculated the ratio A␤ x-42/total A␤ for each expression construct and compared the relative ratios (32,27). Results of these analyses are presented in Fig. 4B. As reported previously (10, 12, 13), expression of mutant PS1 (A␤ x-42/total FIG. 3. PS1 NTF and PS2 CTF derived from a chimeric precursor can be co-precipitated. A, detergent lysates prepared from stable N2a cells transfected with an empty vector or cDNA encoding a chimeric PS1/PS2 polypeptide (PS1/2) were analyzed by immunoblotting using PS1 NT (lanes 1 and 2) and ␣PS2Loop antibodies (lanes 3 and 4). The positions of mouse (Mo) and human (Hu) PS1 NTF and PS2 CTF are indicated. B, stable N2a cells expressing PS1/PS2 chimera were lysed under nondenaturing conditions, and lysates were used for coimmunoprecipitation analyses as described in the legend to Fig. 2. Total lysate corresponding to 1 ⁄20 of the volume of lysate used for immunoprecipitation and an aliquot of stable PS2 cell line were also fractionated on SDS gels as markers for PS2 NTF and CTF. Note that the ϳ30-kDa band in the PS2 NT blot indicated by an asterisk (bottom panel, lane 8) is the PS1 NTF signal that remained due to the incomplete stripping of PS1 NT antibody.

DISCUSSION
Our understanding of the nature of the functional PS "unit" is still very incomplete. Knowledge about the subunit stoichiometry is essential for further studies on the structure and function of the PS complex. In the present study, we exclusively focused on the composition and stoichiometry of PS NTF and CTF complexes immunoprecipitated from stable mouse neuroblastoma cells that express recombinant human PS1 NTF, full-length PS2, or a chimeric PS1/PS2 polypeptide. These studies demonstrate that transgene-derived recombinant PS1 NTF neither influenced the metabolism of endogenous PS1 polypeptides nor associated with endogenous PS1-derived CTF. PS1 and PS2 fragments derived from the cleavage of co-expressed full-length PS1 and PS2 do not form mixed heteromeric assemblies. In contrast, endoproteolysis of a PS1/PS2 chimera generates stable PS1 NTF⅐PS2 CTF assemblies. The results of our studies provide an important insight: an intramolecular association(s) between domains of the PS NTF and CTF is established prior to endoproteolytic cleavage.
In the present study, by examining the metabolism of a PS1/PS2 chimera we provide evidence that intramolecular association(s) between structural domains located within NH 2and COOH-terminal halves of PS precedes endoproteolytic cleavage. Our conclusion is consistent with previous gel filtration chromatography and velocity gradient analyses, which revealed that full-length PS and the PS-derived NTF and CTF are components of high molecular weight complexes (29,36,37). Independently, two studies demonstrated that complexes containing PS1 NTF and CTF were larger than the complexes containing full-length PS1 (29,37). Interestingly, a FAD-linked PS1 deletion mutant that fails to undergo endoproteolytic cleavage (18,38) was found to be in a high molecular weight complex, similar in size to that of the cleaved fragments, suggesting that cleavage was not necessary for the assembly of stable PS complexes (37). Nevertheless, attempts to bypass the requirement for endoproteolysis by co-expressing truncated PS2 polypeptides, which correspond to FAD-linked mutant NTF and CTF, did not lead to elevated A␤42 production (32). Together with our present results, these studies indicate that the formation of a functional PS complex may be an ordered and stepwise process in which newly synthesized PS polypeptides establish intramolecular association(s) during the process of folding and subsequently bind to other proteins expressed at limiting levels that "stabilize" the PS complex; properly folded and "mature" PS molecules that have associated with limiting accessory molecules are then cleaved (or not in the case of PS1 lacking exon 9-encoded residues). The mechanisms involved in facilitating PS endoproteolysis are not known.
It is likely that both protein stabilization and endoproteolytic cleavage may rely on proper folding of the synthetic PS polypeptides. The lack of an intact COOH terminus, for example, might prevent proper folding and/or association with PS binding proteins and thereby lead to rapid degradation. In this regard, previously we showed that COOH-terminally truncated PS polypeptides, which lack the last two transmembrane domains and the cytoplasmic tail, failed to undergo endoproteolysis and were rapidly degraded (21). Our present results confirm studies by Steiner et al. (25) that recombinant transgene-derived PS1 NTF is not incorporated into complexes containing endogenous PS1 CTF and is rapidly degraded. In recent studies, it was reported that recombinant PS1 NTF with a FAD mutation failed to increase A␤42 production, even after stabilization of recombinant NTF by incubation with proteasome inhibitors (25,27). Based on the recombinant NTF studies, it was suggested that full-length PS molecules, and not the endoproteolytic derivatives, were the functional PS units in vivo (27). It is known that proper assembly contributes to the structural maturation as well as the function of oligomeric membrane proteins. Thus, our demonstration that exogenously expressed human NTF failed to assemble with endogenous CTF provides direct evidence that the recombinant NTFs were not incorporated into PS complexes. We argue that the transgenederived NTF molecules are structurally quite different from NTF⅐CTF assemblies derived from full-length PS polypeptides; hence, their inability to increase A␤42 production cannot be interpreted in a way that provides any meaningful information regarding the functional form of PS.
The outcome of co-expression of full-length and truncated membrane receptors has been investigated for several membrane proteins. In several instances, expression of truncated polypeptides disrupted the functional assembly and activity of endogenous proteins by a dominant negative mechanism (39). For example, expression of fragments derived from the extracellular NH 2 -terminal domains of ␣-, ␦-, and ␥-subunits of the acetylcholine receptor blocked the assembly and cell surface expression of functional acetylcholine receptors (40). In analogous studies, co-expression of truncated and full-length gonadotropin-releasing hormone receptors resulted in the inhibition of gonadotropin-releasing hormone signaling via the full-length FIG. 4. Analysis of PS1/PS2 stability and the influence of mutant PS1/PS2 on A␤42 production. A, stable N2a cells expressing human PS1 or PS1/PS2 chimera were incubated in culture medium containing cycloheximide (30 g/ml) for the intervals indicated. Detergent lysates prepared from stable PS1 cells were analyzed by immunoblotting with PS1 NT and ␣PS1Loop (top three panels); lysates from PS1/PS2 cells were blotted with PS1 NT and ␣PS2Loop (bottom three panels). Only the regions of gels corresponding to the full-length polypeptide (FL) and endoproteolytic derivatives (NTF and CTF) are shown. B, COS cells transfected with PS1, PS2, or PS1/PS2 cDNA along with cDNA encoding full-length APP or APP C100 cDNA (see "Materials and Methods"). The levels of A␤ x-40 and A␤ x-42 secreted from doubly transfected cells were quantified by two-site ELISAs. Mean values Ϯ S.E. from two independent experiments are shown. *, p ϭ 0.0012; **, p Ͻ 0.001, relative to values from cells transfected with the corresponding wild-type cDNA.
receptor, probably because of specific interactions between the two receptor proteins resulting in impaired maturation of the full-length receptor (41). In contrast to these studies, our analysis of stable cell lines expressing transgene-derived human PS1 NTF shows that the expression of PS1 NTF did not have any dominant negative inhibitory effect on the cleavage of endogenous full-length PS1 or on the stable accumulation of endogenous PS1-derived fragments. Moreover, we and others showed that transgene-derived PS1 NTF cannot be co-isolated with endogenous PS1-derived CTF, excluding any direct physical interaction between truncated PS molecules and endogenous PS1. Together, these results strongly suggest that the COOH terminus of PS might provide information essential for protein folding and/or interaction with proteins that interact with PS. In this regard, a calcium-binding protein, termed calsenilin, has been shown to bind to the COOH-terminal 40 residues of PS (42).
As described above, PS1 and PS2 fragments derived from cleavage of co-expressed full-length PS1 and PS2 neither form mixed heteromeric assemblies nor exchange subunits to result in complexes containing PS1-NTF⅐PS2 CTF or PS2 NTF⅐PS1 CTF. Yet, overexpression of human PS1 resulted in diminution of endogenous PS2 derivatives (21). In studies presented here, we demonstrate that a chimeric polypeptide consisting of PS1 NTF and PS2 CTF undergoes endoproteolysis, and the resulting derivatives accumulate to saturable levels. Furthermore, expression of chimeric PS1/PS2 polypeptides harboring the FAD-linked M146L mutation resulted in the elevated production of A␤42. Together, these findings indicate that PS1 and PS2 share common pathways of processing and likely interact with similar factors that modulate ␥42 processing of APP. Future efforts to identity of individual polypeptide components of the oligomeric PS complex will be required to define the multiple steps involved in the regulation of assembly of the functional PS "unit".