γ-Secretase Complex Assembly within the Early Secretory Pathway*

γ-Secretase is an aspartyl protease complex composed of the four core components APH-1, nicastrin (NCT), presenilin (PS), and PEN-2. It catalyzes the final intramembranous cleavage of the β-secretase-processed β-amyloid precursor protein to liberate the neurotoxic amyloid β-peptide. Whereas unassembled complex components appear to be unstable and/or to be retained within the endoplasmic reticulum (ER), the fully assembled complex is known to exert its biological function in late secretory compartments, including the plasma membrane. We thus hypothesized that the γ-secretase complex undergoes a stepwise assembly within the ER. We demonstrate that γ-secretase-associated NCT can be actively retained within the ER by the addition of a retention signal. Under these conditions, complex assembly occurred in the absence of maturation of NCT, and ER-retained immature NCT associated with APH-1, PEN-2, and PS fragments. Moreover, a biotinylated transition state γ-secretase inhibitor allowed the preferential isolation of the fully assembled complex containing immature NCT. Furthermore, we observed a conformational change in immature NCT, which is known to be selectively associated with complete γ-secretase complex assembly. This was also observed for a small amount of immature endogenous NCT. ER-retained NCT also rescued the biochemical phenotype observed upon RNA interference-mediated NCT knockdown, viz. reduced amyloid β-peptide production; instability of PS, PEN-2, and APH-1; and accumulation of β-amyloid precursor protein C-terminal fragments. Finally, we demonstrate that dimeric (NCT/APH-1) and trimeric (NCT/APH-1/PS) intermediates of γ-secretase complex assembly containing endogenous NCT are retained within the ER and that the incorporation of the fourth and last binding partner (PEN-2) also occurs on immature NCT, suggesting a complete assembly of the γ-secretase complex within the ER.

␥-Secretase is an aspartyl protease complex composed of the four core components APH-1, nicastrin (NCT), presenilin (PS), and PEN-2. It catalyzes the final intramembranous cleavage of the ␤-secretase-processed ␤-amyloid precursor protein to liberate the neurotoxic amyloid ␤-peptide. Whereas unassembled complex components appear to be unstable and/or to be retained within the endoplasmic reticulum (ER), the fully assembled complex is known to exert its biological function in late secretory compartments, including the plasma membrane. We thus hypothesized that the ␥-secretase complex undergoes a stepwise assembly within the ER. We demonstrate that ␥-secretase-associated NCT can be actively retained within the ER by the addition of a retention signal. Under these conditions, complex assembly occurred in the absence of maturation of NCT, and ER-retained immature NCT associated with APH-1, PEN-2, and PS fragments. Moreover, a biotinylated transition state ␥-secretase inhibitor allowed the preferential isolation of the fully assembled complex containing immature NCT. Furthermore, we observed a conformational change in immature NCT, which is known to be selectively associated with complete ␥-secretase complex assembly. This was also observed for a small amount of immature endogenous NCT. ER-retained NCT also rescued the biochemical phenotype observed upon RNA interference-mediated NCT knockdown, viz. reduced amyloid ␤-peptide production; instability of PS, PEN-2, and APH-1; and accumulation of ␤-amyloid precursor protein C-terminal fragments. Finally, we demonstrate that dimeric (NCT/APH-1) and trimeric (NCT/ APH-1/PS) intermediates of ␥-secretase complex assembly containing endogenous NCT are retained within the ER and that the incorporation of the fourth and last binding partner (PEN-2) also occurs on immature NCT, suggesting a complete assembly of the ␥-secretase complex within the ER.
Alzheimer's disease is a serious public health problem, and a dramatic increase in Alzheimer's disease patients is apparent in our rapidly aging society. Pathologically, Alzheimer's disease is characterized by the accumulation of tangles and senile plaques (1). Senile plaques are composed of amyloid ␤-peptide (A␤), 1 which is generated by proteolytic processing from the ␤-amyloid precursor protein (APP) (2). Secretases (named after the secretion of some of their cleavage products) cut APP within and around the A␤ domain. ␤-Secretase mediates the N-terminal amyloidogenic cut and competes with the non-amyloidogenic ␣-secretase, which prevents A␤ production by cleaving in the middle of the A␤ domain (2). The cleavage products of ␤and ␣-secretases (99-and 83-amino acid C-terminal fragments) are the direct substrates for the ␥-secretase enzyme. This cleavage occurs within the transmembrane domain (TMD) and is exerted by an aspartyl protease complex composed of the four core components presenilin (PS)-1 or -2, nicastrin (NCT), APH-1a/b (anterior pharynx-defective-1; in this work, we refer to the long splice variant APH-1aL (3)), and PEN-2 (presenilin enhancer-2) (reviewed in Ref. 2). PSs are most likely the catalytic subunit of this complex and provide the two critical aspartyl residues. This is now strongly supported by a number of independent observations. In the absence of PS1 and PS2, no authentic A␤ is produced (4,5). Moreover, numerous familial Alzheimer's disease-linked mutations in the PS genes shift the ratio from the predominant 40-amino acid A␤ species to the more aggregation-prone 42-amino acid variant (2). Furthermore, active-site ␥-secretase inhibitors can be cross-linked to PSs (6,7), and mutagenesis of the two critical aspartyl residues inhibits ␥-secretase activity (8). PSs belong to the novel family of polytopic aspartyl proteases of the GXGD type (9, 10). These proteases have a highly conserved C-terminal active-site motif composed of a GXGD motif containing the catalytically critical aspartate. This family of proteases, which includes the signal peptide peptidases and their homologs (11) and the type 4 prepilin peptidases (10,12), may have evolved independently to cleave hydrophobic sequences within or close to TMDs (13).
In contrast to all known putative GXGD proteases, which are active either on their own or as homodimers (11,12,14), PSs must associate with three other components to gain their proteolytic activity (15)(16)(17)(18). Coordinated regulation of the expression level of the four complex components severely complicates functional analysis of individual components (reviewed in Ref. 2). Down-regulation of PS1, NCT, APH-1, or PEN-2 decreases the levels of the other components and prevents maturation of NCT (2). Once assembled, the complex is, however, extraordinary stable (19). It appears that a dimeric complex composed of APH-1 and immature NCT (15, 20 -22) provides the scaffold for the addition of the remaining components. The PS holoprotein may be added to this putative precomplex to from a trimeric intermediate (15,23). Finally, PEN-2 apparently associates with the precomplex (21)(22)(23)(24)(25), resulting in a conformational change in NCT (20,26) and the induction of PS endoproteolysis (Refs. 15,18,23,and 25;reviewed in Ref. 27).
At the moment, it is unclear where within the secretory pathway ␥-secretase complex assembly occurs. We hypothesized that ␥-secretase complex components may be retained within the endoplasmic reticulum (ER) as long as the complex is not fully assembled. To identify the cellular compartment where ␥-secretase complex assembly occurs, we attached an ER retention signal to the cytoplasmic domain of NCT. By forcing the assembly of the ␥-secretase complex on immature NCT, we demonstrate that a fully functional ␥-secretase complex can be generated within the ER. Moreover, endogenously assembled complex intermediates containing only two or three components were retained within the ER, but were released after the missing ␥-secretase subunits were incorporated.

EXPERIMENTAL PROCEDURES
Cell Culture and Cell Lines-Human embryonic kidney (HEK) 293 cells stably expressing Swedish mutant APP (swAPP) and the NCT knockdown cell line were described previously (26). The swAPP-expressing cell line and the NCT knockdown cell line were stably transfected with the wild-type NCT (NCTwt/-V5) (28) or NCT-ER-V5 constructs using FuGENE 6 (Roche Diagnostics). For NCT expression, cells were selected with blasticidin (50 g/ml). The PEN-2 knockdown cell line and cell lines expressing PEN-2 derivatives have been described previously (25).
cDNA Constructs of NCT-All NCT constructs expressed in the NCT knockdown cell line contained a cluster of silent mutations conferring RNA interference (RNAi) resistance (26). The NCT-ER-V5 construct, containing the sequence motif for ER retention and a stop codon after the V5 tag, was generated by PCR-mediated mutagenesis. For the CD4-NCT-CT-V5 and CD4-NCT-CT-V5-ER fusion constructs, the C termini of NCT were amplified by PCR using the appropriate NCT construct as template and subcloned via an artificial NotI site after the last CD4 codon and the XhoI site of the vector pcDNA3 (Invitrogen). The CD4-NCT-TM-V5 construct was generated by PCR. All cDNA constructs were sequenced to verify successful mutagenesis.
Immunocytochemistry-COS-7 cells were cultured on poly-L-lysinecoated coverslips; transiently transfected with the CD4-NCT constructs using FuGENE 6; and 48 h after transfection, fixed with 4% paraformaldehyde and 4% sucrose in phosphate-buffered saline for 20 min. For intracellular staining, cells were treated with 0.2% Triton X-100. A standard immunofluorescence procedure was carried out using the anti-CD4 antibody (1:500 dilution) as the primary antibody and an Alexa 488-coupled anti-mouse secondary antibody (Molecular Probes, Inc.) for detection.
Protein Analysis-Cell lysates were prepared using 1% Nonidet P-40, 50 mM Tris (pH 7.6), 150 mM NaCl, and 2 mM EDTA. For analysis of secreted A␤, the medium was collected after 5 h and subjected to combined immunoprecipitation/immunoblotting using antibodies 3926 and 6E10 (35). For co-immunoprecipitations, CHAPS lysis buffer (2% CHAPS, 20 mM HEPES (pH 7.2), 100 mM KCl, 2 mM EDTA, and 2 mM EGTA) was used. Co-immunoprecipitations and glycerol velocity gradient centrifugation were carried out as described (36). Briefly, for velocity gradients, 100,000 ϫ g membranes were solubilized in CHAPS lysis buffer; insoluble material was pelleted by an additional 100,000 ϫ g centrifugation step; and 2 mg of the soluble material was loaded onto a 5-25% (v/v) glycerol gradient and separated by centrifugation at 200,000 ϫ g for 16 h. 12 fractions of 1-ml volume were collected. The trypsin resistance assay was carried out as described (20,26). Briefly, 10 l of CHAPS lysates (protein concentration of 3 mg/ml) were incubated with trypsin at a final concentration of 0.1 mg/ml, and the immunoprecipitated material was treated with trypsin at a final concentration of 0.02 mg/ml for 30 min at 30°C. The reaction was stopped by the addition of 1 l of soybean trypsin inhibitor (10 mg/ml). For the glycerol gradient, 100 l of each fraction was incubated with trypsin (final concentration of 0.01 mg/ml) for 30 min at 30°C, stopped with 1 l of soybean trypsin inhibitor, and precipitated with an equal volume of 20% trichloroacetic acid. Controls were treated similarly, but were incubated without trypsin. For stabilization of PEN-2⌬C, cells were treated with 5 M MG132 for 12 h. Cell lysis and co-immunoprecipitations were carried out as described above.
In Vitro ␥-Secretase Activity Assay-The assays were carried out as described previously (28).
Inhibitor Affinity Precipitations of the Active ␥-Secretase Complex-Confluent cell monolayers were washed twice with ice-cold phosphatebuffered saline, and cells were harvested in 20 mM EDTA in phosphatebuffered saline and sedimented by centrifugation at 1000 ϫ g for 10 min. Cells were resuspended in hypotonic buffer containing 10 mM HEPES (pH 7.4), 1 mM EDTA, and 0.25 M sucrose; incubated for 15 min on ice; and homogenized. Nuclei and cell debris were removed by centrifugation at 2500 ϫ g for 10 min. Membranes of the post-nuclear supernatant were precipitated by centrifugation at 100,000 ϫ g for 1 h and solubilized in lysis buffer containing 1% CHAPSO and 150 mM sodium citrate (pH 6.4). Insoluble material was pelleted by centrifugation at 100,000 ϫ g for 1 h, and the supernatant was subjected to affinity precipitation. For each sample, 800 l (protein concentration of 1 g/l) adjusted to a CHAPSO concentration of 0.5% was precleared for 30 min using 25 l of streptavidin-Sepharose (Amersham Biosciences). For specific capture, the precleared solubilized membrane preparation was incubated for 2 h in the presence of the biotinylated aspartyl transition state analog Merck C inhibitor (0.1 M) (37), followed by incubation with 25 l of streptavidin-Sepharose for 30 min. ␥-Secretase inhibitor complexes were precipitated by centrifugation, and the pellets were washed three times with 0.5% CHAPSO. To prove that the inhibitor concentration was sufficient, the supernatant after the first precipitation was subjected to a second precipitation with Merck C inhibitor and streptavidin-Sepharose. Specific binding was proven by the addition of a 100-fold excess of the non-biotinylated Merck A inhibitor (38). Samples were separated by SDS-PAGE and analyzed by Western blotting.

RESULTS
NCT that is not associated with the ␥-secretase complex remains as an immature species within the ER (39 -42). This suggests that NCT contains an ER retention motif, which may be masked during complex formation. To investigate whether NCT contains such active retention motifs, we used CD4 as a reporter protein (43). First, the C-terminal domain of NCT including a V5 tag (note that the V5 tag does not interfere with the physiological function of NCT; see below) was fused to CD4, creating CD4-NCT-CT-V5 (Fig. 1A). Second, the TMD of CD4 was exchanged with the TMD of NCT to create the CD4 variant CD4-NCT-TM-V5 (Fig. 1A). Upon transfection of COS-7 cells with CD4-NCT-CT-V5 or CD4-NCT-TM-V5, intracellular and plasma membrane immunostaining was carried out (Fig. 1B). Both fusion proteins showed punctuate intracellular staining as well as strong labeling of the plasma membrane (Fig. 1B). A similar pattern was found when CD4 alone (data not shown) or a cDNA construct encoding CD4 fused to 36 amino acids of the C terminus of a K ATP channel with a mutagenized inactive ER retention signal (CD4-AAA surface ) (43) was investigated (Fig.  1B). This indicates that the cytoplasmic tail and TMD of NCT do not contain an active ER retention signal. To further address the question of whether ␥-secretase complex assembly occurs within the ER, NCT was forced to be retained within the ER by the addition of an ER retention signal. We first confirmed the function of the well established KKXX motif for ER retention (44) fused to the C terminus of a V5 tag variant of NCT using the CD4 reporter system (Fig. 1A). CD4-NCT-CT-V5-ER was retained within the ER, and only very little surface staining could be detected (Fig. 1B). Similarly, the CD4-RKR ER variant (Fig. 1A), in which the last 36 amino acids of the C terminus of a K ATP channel with an active ER retention signal (RKR) (43) are fused to CD4, was also efficiently retained within the ER (Fig. 1B).
To analyze whether NCT that is actively retained within the ER influences ␥-secretase complex formation and function, we transfected an NCT variant with a functionally active ER retention signal (for ER retention signal, see Fig. 1) in an NCT knockdown HEK 293 cell line stably expressing swAPP (26,28,39). Consistent with our previous results (26,28,39), NCT expression was strongly decreased by RNAi ( Fig. 2A). As expected, this was accompanied by reduced PS1, APH-1aL, and PEN-2 expression and a significant increase in the APP Cterminal fragments (CTFs), followed by reduced A␤ generation (Fig. 2B). When this cell line was stably transfected with the RNAi-insensitive NCTwt-V5 construct (26,28,39), the biochemical phenotype of the NCT knockdown was rescued (Fig.  2B). The APP CTFs did not accumulate anymore, and A␤ production was fully restored (Fig. 2B). Maturation of NCTwt-V5 was observed ( Fig. 2A), which is consistent with the functional restoration of ␥-secretase activity (39,41). In contrast, NCT maturation was reduced upon expression of NCT-ER-V5 ( Fig. 2A). Although maturation of NCT was significantly reduced by the addition of the ER retention signal, NCT-ER-V5 rescued the loss of ␥-secretase function caused by the RNAimediated knockdown of endogenous NCT. Like NCTwt, ERretained NCT rescued PS1, APH-1aL, and PEN-2 expression; reduced APP CTF levels; and restored A␤ generation (Fig. 2B). No significant change in the A␤ 40 /A␤ 42 ratio was observed upon ER retention of NCT (Fig. 2B). To exclude that remaining small amounts of endogenous NCT were responsible for the full rescue, we performed in vitro ␥-secretase assays using ␥-secretase complexes selectively immunoisolated via the V5 tag of ectopically expressed NCT. This revealed that ␥-secretase complexes containing NCT-ER-V5 efficiently produced A␤ and APP intracellular domain-like complexes composed of NCTwt (Fig. 2C).
NCT with and without an ER retention signal co-immunoprecipitated with both PS1 fragments (Fig. 2D), suggesting insertion into a complete ␥-secretase complex. In contrast to NCTwt-V5, where mainly mature NCT associated with PS fragments, significant levels of immature NCT-ER-V5 co-immunoprecipitated with PS (Fig. 2, D and E). Both variants FIG. 1. The C terminus and TMD of NCT do not contain an ER retention signal. A, schematic representation of NCT and the fusion constructs of the reporter protein CD4. The NCT C terminus without (CD4-NCT-CT-V5) or with (CD4-NCT-CT-V5-ER) an ER retention signal was fused to the C terminus of CD4, or the TMD of CD4 was exchanged with the TMD of NCT (CD4-NCT-TM-V5). As controls, two CD4 constructs containing either an ER retention signal (CD4-RKR ER ) or a mutated retention signal (CD4-AAA surface ) were used. TM, the transmembrane region; hatched boxes, the NCT TMD; striped boxes, the NCT C terminus; black bars, V5 epitope. B, subcellular localization of the CD4-NCT-CT-V5 and CD4-NCT-TM-V5 fusion constructs and the ER-retained CD4-NCT-CT-V5-ER construct, including the surface-located control (CD4-AAA surface ) and the ER-retained control (CD4-RKR ER ). COS-7 cells were transiently transfected with the indicated cDNAs. 48 h after transfection, intracellular staining (left panels) or surface staining (right panels) was carried out using the anti-CD4 monoclonal antibody.
(NCTwt and ER-retained NCT) efficiently assembled into a ␥-secretase complex. This suggests that ␥-secretase complex formation occurs within an early secretory compartment and that ␥-secretase complexes containing immature NCT can be biologically active. Indeed, quantification of co-immunoprecipitation experiments showed a decreased interaction of mature NCT with PS1 in the case of the ER-retained variant (Fig. 2E). Although significantly less mature NCT was observed, A␤ generation was unchanged (Fig. 2E), indicating proteolytic activity of an ER-retained fully assembled ␥-secretase complex.
To further investigate whether ER-retained immature NCT indeed assembles into a functional ␥-secretase complex, we investigated the conformational switch of NCT known to be associated with ␥-secretase activity (20,26). As expected from our previous findings (20,26), mature NCTwt-V5 was selectively trypsin-resistant, whereas the immature variant was degraded under the same conditions (Fig. 3A). In contrast, substantial amounts of immature NCT-ER-V5 were found to be resistant to trypsin (Fig. 3A). This suggests that a significant portion of immature NCT-ER-V5 is incorporated into a func-

FIG. 2. NCT containing an ER retention signal entirely rescues the NCT knockdown-induced phenotype and assembles with complex components.
A, NCT knockdown HEK 293 cells expressing swAPP (sw NCT RNAi ) were stably transfected with NCTwt-V5 (wt-V5) and NCT-ER-V5 (ER-V5) cDNA constructs harboring a cluster of silent mutations to escape RNAi-mediated destruction. Cell lysates were analyzed for NCT expression by Western blotting with the anti-NCT-CT, anti-NCT-NT, or anti-V5 antibody. Note that in the swAPP-expressing NCT knockdown HEK 293 cells, low levels of endogenous NCT could be detected with the antibodies against the N and C termini of NCT. B, cell lysates were analyzed for levels of PS1 CTFs, APH-1aL, PEN-2, and APP CTFs by Western blotting using anti-PS1 loop antibody O2C2 (for APH-1 detection) and anti-PEN-2 and anti-APP-CT antibodies. A␤ was analyzed in conditioned medium by a combined immunoprecipitation/Western blot protocol with anti-A␤ antibodies 3926 and 6E10. A␤ was separated in a Tris/Tricine gel system and additionally in a Tris/Bicine/urea gel system to discriminate A␤ 40 and A␤ 42 . NCTwt-V5 and NCT-ER-V5 restored the expression of the components of the ␥-secretase complex, prevented APP CTF accumulation (no reproducible difference was observed upon expression of NCTwt-V5 or NCT-ER-V5; data not shown), and rescued A␤ 40 and A␤ 42 production. C, CHAPSO cell lysates were prepared, and NCT-V5-containing complexes were selectively immunoprecipitated (IP) with the anti-V5 antibody. As a control, the endogenous ␥-secretase complex was precipitated from CHAPSO lysates of HEK 293 cells expressing swAPP (sw) using the anti-PS1 loop antibody. The precipitated material was incubated with purified recombinant APP C100 (16) and analyzed for APP intracellular domain (AICD) and A␤ generation by Western blotting. D, CHAPS lysates were analyzed by immunoprecipitation with the anti-PS1 loop antibody, followed by Western blotting with the anti-V5, anti-NCT-CT, anti-PS1-NT, or anti-PS1 loop antibody (first to fourth panels, respectively). The arrow indicates the position of the PS1 holoprotein. NTF, N-terminal fragment. E, A␤ production and NCT maturation of HEK 293 cells expressing swAPP, NCTwt-V5, and NCT-ER-V5 were quantified. In vitro generation of A␤ was analyzed as described for C and is shown as a percentage of the average value obtained in swAPP-expressing cells. Maturation of NCT incorporated into the ␥-secretase complex was analyzed by immunoprecipitation of CHAPS lysates with the anti-PS1 loop antibody, followed by Western blotting with the anti-NCT-CT or anti-V5 antibody. Shown is the ratio of mature (NCT m ) versus immature (NCT im ) NCT. For all quantifications, the data shown are means Ϯ S.D. of three independent experiments. Note that although significantly less NCT maturated upon expression of NCT-ER-V5, A␤ production was not affected.
tional ␥-secretase complex as early as within the ER. To investigate whether trypsin-resistant immature endogenous NCT can be observed as well, we performed immunoprecipitations of cell lysates derived from untransfected cells with antibodies directed against endogenous NCT or PS. When lysates were immunoprecipitated with the anti-NCT antibody, we observed large amounts of immature NCT, which were almost completely trypsin-sensitive (Fig. 3B). In contrast, when ␥-secretase complexes were isolated with the anti-PS1 loop antibody, we observed only small amounts of immature NCT, which were, however, fully trypsin-resistant (Fig. 3B). This suggests that excess immature NCT fails to become incorporated into the ␥-secretase complex, whereas immature NCT associated with the ␥-secretase complex undergoes the conformational change characteristic of biologically active ␥-secretase complexes. Thus, immature endogenous NCT is apparently incorporated into the ␥-secretase complex within the early secretory pathway before its maturation.
To obtain further evidence that immature NCT-ER-V5 is incorporated into the high molecular weight ␥-secretase complex, we performed glycerol velocity gradient centrifugation (36). Under these conditions, components of the ␥-secretase complex, including mature NCT, accumulated in fractions 9 -12 ( Fig. 4A) (31). These studies revealed significant amounts of immature NCT-ER-V5 co-fractionating with endogenous PS1 CTFs (Fig. 4A), suggesting incorporation of ER-located NCT into a high molecular weight complex. Importantly, a substantial amount of immature NCT co-fractionating with the PS1 CTFs became trypsin-resistant (Fig. 4B, fractions 9 -12), whereas immature NCT in low molecular weight fractions was trypsin-sensitive (Fig. 4, A and B, compare fractions 4 -8). In contrast, in cells expressing NCTwt-V5, immature NCT was almost completely digested in all fractions (Fig. 4B), whereas mature NCT comigrating with PS1 CTFs was selectively stable (fractions 9 -12). This again supports the observation that ERlocated immature NCT becomes part of a fully assembled high molecular weight ␥-secretase complex.
To analyze whether fully assembled ␥-secretase complexes containing immature NCT are proteolytically active, we selectively isolated fully functional ␥-secretase using a biotinylated affinity ligand. The biotinylated Merck C inhibitor selectively binds biologically active ␥-secretase as described previously (6,37,45). Membranes of NCT-ER-V5-or NCTwt-V5-expressing cells were solubilized in CHAPSO and incubated with the biotinylated Merck C affinity ligand, followed by precipitation of enzyme-inhibitor complexes with streptavidin-Sepharose beads. Upon expression of NCTwt-V5, the inhibitor precipitated almost exclusively mature NCT, although a robust amount of immature NCT was present in the membrane lysate (Fig. 5). In contrast, when NCT-ER-V5 was expressed, immature NCT was also precipitated by the biotinylated Merck C inhibitor (Fig. 5). Capture specificity was demonstrated since the binding could fully be abolished in the presence of a 100fold excess of the Merck A inhibitor (Fig. 5). Furthermore, no ␥-secretase activity could be precipitated after initial clearance of the lysate with the biotinylated Merck C affinity ligand (data not shown).

FIG. 4. ER-retained immature NCT is incorporated into a high molecular weight complex and becomes trypsin-resistant.
A, CHAPS-extracted membrane preparations of NCTwt-V5-or NCT-ER-V5-expressing cells were separated on 5-25% glycerol velocity gradients. 12 fractions were collected, and 100 l of each fraction was subjected to trichloroacetic acid precipitation and separation by SDS-PAGE. Overexpressed NCT and endogenous PS1 CTFs were analyzed using anti-NCT-NT and anti-PS1 loop antibodies. NCT m , mature NCT; NCT im , immature NCT. B, 100 l of each fraction was subjected to trypsin digestion, followed by trichloroacetic acid precipitation and Western blot analysis.
The findings described above suggest that ␥-secretase assembly occurs within the ER. However, most of these experiments were conducted with ER-retained NCT variants. To obtain further evidence for ER assembly of the ␥-secretase complex under native conditions, we investigated the stepwise assembly of the endogenous complex components. First, the dimeric precomplex consisting of NCT and APH-1 (15,20,23) was investigated. To do so, fibroblasts from PS1/2 double knockout mice (42) were investigated for NCT maturation. Cell lysates were immunoprecipitated with the anti-APH-1a antibody, and coimmunoprecipitated proteins were detected by Western blotting. As observed previously (15,20,23), immature NCT coimmunoprecipitated with APH-1, whereas no PEN-2 was detected (Fig. 6A). Thus, a dimeric NCT/APH-1 intermediate appears to be assembled and retained within the ER. To investigate the trimeric intermediate of the assembly (15,27), we used a cell line in which PEN-2 was stably knocked down by RNAi (25). Under these conditions, PS fails to be endoproteolyzed, but still associates with APH-1 and NCT (15,27). Coimmunoprecipitations using the anti-APH-1a antibody revealed that such a trimeric complex could indeed be isolated (Fig. 6B). Moreover, this complex also contained preferentially immature NCT (Fig. 6B), indicating that this intermediate complex was still located within the ER. Furthermore, immature NCT that co-immunoprecipitated with APH-1 or PS1 was trypsin-sensitive (Fig. 6C). However, upon expression of an RNAi-insensitive PEN-2 variant, the complex left the ER. Under these conditions, all four ␥-secretase complex components were co-isolated (Fig. 6B), and mature NCT was preferentially co-immunoprecipitated (Fig. 6B).
To search for a fully assembled ␥-secretase complex containing immature endogenous NCT, we made use of the observation that PEN-2 is required for endoproteolysis of PS holoprotein and for stabilization of PS fragments (25,46,47). We (25) and others (46,47) have previously shown that PEN-2 containing a C-terminal deletion (PEN-2⌬C) associates with the PS holoprotein and triggers its endoproteolysis, but that the resulting PS N-and C-terminal fragments, as well as PEN-2⌬C itself, are unstable and degraded by the proteasome. Thus, PEN-2⌬C inefficiently rescues PEN-2 knockdown (25,46,47). Therefore, we sought to stabilize the complex by blocking the proteasomal degradation of PS N-and C-terminal fragments and PEN-2⌬C using MG132. Under these conditions, co-immunoprecipitation with the anti-APH-1a antibody indeed resulted in the co-purification of the complete ␥-secretase quartet (Fig.  6D). Moreover, we selectively co-purified immature NCT (Fig.  6D), demonstrating the association of all four components before complex glycosylation in the late secretory pathway occurred. Note that even without stabilization by proteasomal inhibition, small amounts of PS1 CTFs and PEN-2⌬C associated with APH-1 and immature NCT.

DISCUSSION
The so-called spatial paradox claimed that PS is located mostly within the ER, whereas ␥-secretase activity is observed within the late secretory pathway, the plasma membrane, and endosomes (48,49). Indeed, only small amounts of PS associated with mature NCT are located on the plasma membrane (50). This suggested that unincorporated or excess amounts of ␥-secretase components may reside within the ER and are released only upon their assembly into the (complete) native ␥-secretase complex. This prediction was supported by the observation that down-regulation of PS results in the accumulation of immature NCT (39,41,42,50). These observations prompted us to investigate whether ␥-secretase complex assembly can occur within the ER. Such a mechanism would indicate that complex formation and correct folding of components of the complex could be controlled by ER-located chaperones. Misfolded complexes or unincorporated components would be degraded within the ER before reaching later compartments of ␥-secretase function (51,52).
Indeed, we recently identified an ER retention signal within PS1 (53), which holds unassembled PS back in the early secretory pathway. For NCT, we could detect an ER retention motif neither in the TMD nor in the cytoplasmic domain. However, NCT that is not incorporated into the ␥-secretase complex resides within the ER; therefore, accumulation of immature NCT may be explained by unsuccessful folding. This is consistent with our previous observation that ␥-secretase complex formation/activity is closely associated with a major conformational change in NCT (20,26). Thus, forward transport of NCT apparently depends on its interaction with the other ␥-secretase complex components.
Consistent with previous reports (15,22,27), we have demonstrated a sequential assembly of the complete ␥-secretase complex. Immature NCT and APH-1 appear to be a dimeric ER-located scaffold for further complex assembly. Next, the PS holoprotein binds and forms a trimeric intermediate, which still contains immature NCT. Subsequently, PEN-2 may come into play (27). After PEN-2 association with the trimeric complex, PS is endoproteolyzed; NCT undergoes a conformational switch; and the complete complex is than apparently released from the ER as monitored by its trypsin resistance and maturation. A rapid release of ␥-secretase from the ER may make it difficult to investigate the cellular site of its assembly. We therefore used several independent approaches to study the assembly of the ␥-secretase complex. First, we forced NCT to be retained within the ER by the addition of a functionally active ER retention motif. Under these conditions, NCT maturation is significantly reduced, although small amounts of NCT do escape retention. Upon assembly into the complex, the retention signal may thus be partially covered, and consequently, ␥-secretase complexes containing NCT variants with the ER retention motif can escape the ER and undergo maturation. Additional evidence for formation of a high molecular weight ␥-secretase complex containing immature NCT was obtained by velocity gradient centrifugation and trypsin digestion. ER- FIG. 5. Affinity capture of immature NCT by the biotinylated Merck C inhibitor. The CHAPSO-solubilized ␥-secretase complex was incubated with the biotinylated Merck C inhibitor (0.1 M) and precipitated with streptavidin-Sepharose beads. In NCTwt-V5 (wt-V5)-transfected cells, the inhibitor specifically bound to mature NCT (NCT m ), whereas in NCT-ER-V5 (ER-V5)-transfected cells, substantial amounts of immature NCT (NCT im ) were precipitated as well. To investigate whether sufficient amounts of inhibitor were used, the supernatant of the first precipitation was subjected to a second affinity capture, which did not precipitate additional NCT. CHAPSOsolubilized membranes were incubated with the Merck C affinity ligand (0.1 M) in the presence of a 100-fold excess of the nonbiotinylated Merck A inhibitor. Precipitation with streptavidin-Sepharose followed by Western blot analysis revealed that the precipitation of NCT was abolished in the presence of a 100-fold excess of the Merck A inhibitor. As a loading control, 2.5% of the membrane lysate used for precipitation was directly analyzed by Western blotting. The anti-V5 antibody was used for detection. sw NCT RNAi , NCT knockdown HEK 293 cells expressing swAPP.
retained immature NCT co-purified with PS fragments in high molecular weight fractions. Moreover, upon digestion with trypsin, we detected substantial amounts of trypsin-resistant immature NCT in high molecular weight fractions. The appearance of trypsin-resistant NCT is associated with the formation of a complete ␥-secretase complex (20,26). Thus, the identification of trypsin-resistant immature NCT suggests ER assembly of the complete ␥-secretase complex. Furthermore, by selective trypsin resistance, we have demonstrated that immature endogenous NCT assembles into a ␥-secretase complex under physiological conditions. In addition, the biotinylated Merck C inhibitor, which selectively binds to biologically active ␥-secretase (37), also preferentially purified ␥-secretase complexes containing immature NCT from cells expressing NCT with the ER retention motif. Finally, we isolated the ␥-secretase complex containing all four components, including immature wild-type (endogenous) NCT, upon expression and stabilization of an unstable PEN-2 variant (25,46,47). This suggests that even PEN-2, which is the last component to enter the complex, associates with the trimeric intermediate already within the ER, before NCT maturation occurs. An immediate release of fully assembled complexes from the ER may have made this observation so far impossible, and consequently, it was expected that the final complex assembly, refolding, and activation occur in late compartments (54).
Taken together, our data indicate a sequential assembly of the ␥-secretase complex within the ER/early secretory compartments. Immediately after full assembly, PS is endoproteolyzed; NCT undergoes the conformational switch; and the complex may be released from the ER and targeted to its sites of biological activity in late secretory/endocytic compartments.