γ-Secretase is responsible for proteolytic maturation of signaling and cell surface proteins, including amyloid precursor protein (APP). Abnormal processing of APP by γ-secretase produces a fragment, Aβ42, that may be responsible for Alzheimer's disease (AD). The biogenesis and trafficking of this important enzyme in relation to aberrant Aβ processing is not well defined. Using a cell-free reaction to monitor the exit of cargo proteins from the endoplasmic reticulum (ER), we have isolated a transient intermediate of γ-secretase. Here, we provide direct evidence that the γ-secretase complex is formed in an inactive complex at or before the assembly of an ER transport vesicle dependent on the COPII sorting subunit, Sec24A. Maturation of the holoenzyme is achieved in a subsequent compartment. Two familial AD (FAD)–linked PS1 variants are inefficiently packaged into transport vesicles generated from the ER. Our results suggest that aberrant trafficking of PS1 may contribute to disease pathology.
Alzheimer's disease (AD) is a neurodegenerative disorder characterized by the presence of amyloid plaques and neurofibrillary tangles in hippocampal regions of the brain. A major constituent of the amyloid plaque is a small Aβ peptide fragment derived from the β-amyloid precursor protein (APP). APP is cleaved along the secretory pathway by several proteases, two of which, the β- and γ-secretases, normally generate a 40-amino acid long Aβ40 peptide. Abnormal cleavage by γ-secretase yields an aggregation-prone 42-amino acid long Aβ42 peptide that builds up over time to form neuritic plaques. Mutations in APP and presenilin 1 and 2 (PS1 and PS2) account for most of the familial forms of AD (FAD), which lead to early onset of the disease. These mutations either enhance the ratio of Aβ42/Aβ40 or increase the level of both Aβ42 and Aβ40 without changing the ratio (for review see Walsh and Selkoe, 2004).
Presenilin (PS) is believed to be the catalytic subunit of γ-secretase. γ-Secretase is a multi-subunit protein complex whose minimal components are nicastrin (NCT), anterior pharynx-1 (APH-1), presenilin enhancer-2 (PEN-2), and PS (Yu et al., 2000; Francis et al., 2002; Goutte et al., 2002; LaVoie et al., 2003; Luo et al., 2003; Takasugi et al., 2003). Other possible regulatory components of γ-secretase, such as CD147 and TMP21, have been reported (Zhou et al., 2005; Chen et al., 2006). Although not essential, these regulator proteins affect Aβ production. Therefore, the assembly and activation of γ-secretase may be linked to processing and misprocessing of APP.
Early secretory compartments such as endoplasmic reticulum (ER), ER-Golgi intermediate compartment (ERGIC), and cis-Golgi play a critical role in the synthesis, assembly, and quality control inspection of multi-subunit membrane protein complexes (Ellgaard et al., 1999). In the case of γ-secretase, retrieval of unassembled Pen-2 and NCT, subunits of γ-secretase, is regulated by Rer1 at the cis-Golgi (Kaether et al., 2007; Spasic et al., 2007). Down-regulation of Rer1 causes an increase of γ-secretase at cell surface. Thus, abnormal assembly and traffic of γ-secretase in the early secretory pathway may impact γ-secretase function in later compartments.
Proteins traverse the secretory pathway in defined transport carriers, the first of which, COPII vesicles, conveys newly synthesized and assembled proteins from the ER to the ERGIC (Schekman and Orci, 1996). However, because these COPII vesicles are transient, it is difficult to analyze the molecular characteristics of protein intermediates in cells. Such intermediates may be evaluated by generating transport vesicles in vitro. We exploited a biochemical assay to monitor the export of γ-secretase components from the ER (Kim et al., 2005).
COPII proteins (coat protein complex II; Sar1, Sec13/31, and Sec23/24) assemble on the surface of the ER and capture proteins into vesicles at ER exit sites. The selection of cargo proteins at ER export sites is driven by the direct or indirect interaction of cargo proteins with the Sec24 subunit of the coat (Lee et al., 2004). In this paper, we describe the characteristics of the γ-secretase complex as it assembles in COPII vesicles in membranes harboring wild-type (WT) or FAD mutant forms of PS1.
The γ-secretase complex in the COPII vesicle
COPII vesicles budded from microsomal membranes incubated with nucleotide and cytosol as a source of COPII proteins. ER membranes and transport vesicles were separated by differential centrifugation (Kim et al., 2005). As a control, an ER resident protein, ribophorin I, was not released into the slowly sedimenting transport vesicle fraction (Fig. 1 A). In contrast, an ER-to-Golgi recycling protein, p58/ERGIC53/LMAN1, was released into the vesicle fraction in a cytosol- and nucleotide-dependent manner (Fig. 1 A). Furthermore, this cargo release was inhibited by a dominant-negative form (T39N) of Sar1 but not by Brefeldin A (Fig. 1 A; Fig. S1), suggesting that COPII vesicles are responsible for cargo capture.
To study incorporation of PS1 into the γ-secretase complex, we first analyzed γ-secretase from the microsomal membranes. For γ-secretase complex analysis, digitonin (DIG) or n-dodecyl β-D-maltoside (DDM) was used to solubilize microsomes from N2a cells stably expressing PS1 WT or PS1 ΔE9, a FAD-linked mutation (Crook et al., 1998). The resultant mixtures were subject to BN-PAGE (Fig. 1 B). A summary of γ-secretase subcomplexes is described in Fig. 1 C (Fraering et al., 2004). The γ-secretase complexes (band VI) from DIG-solubilized microsomes migrated more slowly than complexes from DDM solubilized material, but in each case all four γ-secretase subunits were detected at roughly the same electrophoretic mobility (Fig. 1 B, band VI). This differential pattern of migration may be influenced by variable detergent binding. Alternatively, DDM may be harsher than DIG and change subunit stoichiometry. It was reported that DDM disrupts the γ-secretase complex into subspecies (Fig. 1 B) (Fraering et al., 2004). Two principal PS1 WT complexes of ∼440 kD (band VI) and ∼100–120 kD (band II) in DDM-solubilized microsomes represent a full complex (band VI) and a complex (band II) that contains PEN-2 and either PS1-NTF (N-terminal fragment) or PS1 FL (full length), respectively (Fig. 1 B, two left panels, lanes WT DDM) (Fraering et al., 2004). Band II was also observed in DDM-solubilized PS1 ΔE9 microsomes (Fig. 1 B two left panels, lanes ΔE9 DDM). Interestingly, the band II from PS1 WT and ΔE9 microsomes showed similar electrophoretic mobility (Fig. 1 B, two left panels, DDM lanes), whereas the band VI from PS1 WT and ΔE9 microsomes showed a mobility difference of 20–30 kD (Fig. 1 B). It is likely that the difference in electrophoretic migration of the apparently intact PS1 WT and PS1 ΔE9 complexes is a reflection of conformational alterations (Berezovska et al., 2005) or changes in subunit stoichiometry induced by the deletion of exon 9 within the PS1ΔE9 holoprotein.
To discover the biogenic origin of intact γ-secretase complex, we solubilized membranes from microsomes and transport vesicles using 1% DIG and applied cleared soluble samples to blue-native gels. From the transport vesicle fraction, PS1 was detected in a band migrating at a position (440 kD) characteristic of the intact γ-secretase complex (Fig. 1 D) (Edbauer et al., 2002; Kimberly et al., 2003). In addition, some PS1-containing subcomplexes were also found in the vesicle fraction (Fig. 1 D).
PS1 in the vesicle fraction and in the γ-secretase complex was mainly the full-length (FL) precursor form (Fig. 1, A and D). Apparently, association of minimal γ-secretase subunits (PS1, NCT, APH-1, and PEN-2) was not sufficient for efficient endoproteolysis of PS1. Perhaps another influence(s) in a later compartment governs PS1 endoproteolysis (see the PS1 is sorted by Sec24A section). From our results, it is clear that γ-secretase assembly occurs before or within COPII vesicles.
Packaging of the endogenous γ-secretase complex
Overexpression of PS1 may bias formation of the γ-secretase complex in the ER or COPII vesicles. To test whether the endogenous level of PS1 leads to formation of γ-secretase, we used CHO-K1 cells where PS1 FL is relatively easily detectable (Kim et al., 2005). The vesicle fraction contained almost exclusively PS1 FL and was enriched in a higher mobility form of NCT (Fig. 2 A). This higher mobility species of NCT predominated in PS1/2 null ES cells, thus it likely corresponds to the immature form (Fig. 2 B). When the vesicle fraction was analyzed by BN-PAGE to monitor γ-secretase, PS1 was found in band V and VI (Fig. 2 C), reinforcing the idea that normal levels of γ-secretase subunits assemble in an intact complex.
PS1 is sorted by Sec24A
We have previously reported that recombinant mammalian COPII proteins, though functional in budding vesicles from synthetic liposomes, appear to be less active in budding transport vesicles from ER membranes (Kim et al., 2005). A limiting cytosolic factor restores the activity of COPII proteins on ER membranes. This limitation was overcome in incubations containing two- to threefold more than the normal level of ER membranes (semi-intact cells, SICs) (Fig. 3 A). To explore the COPII subunit preference for various cargo proteins, we purified all four human Sec24 proteins in complex with Sec23A (23A/24A, 23A/24B, 23A/24C, and 23A/24D) and used them in reactions containing an elevated level of ER membrane material. As previously seen, the recombinant COPII proteins showed lower cargo packaging activity than cytosol even though ∼10-fold more COPII protein was added to the reactions (Fig. 3 A). p58/ERGIC53/LMAN1 was captured preferentially by Sec24A and Sec24D. APP was recognized by Sec24A, Sec24C, and Sec24D. In contrast, PS1 was favored by Sec24A, although Sec24B and Sec24D were marginally active in PS1 packaging (Fig. 3 A)
PS1 packaging in authentic COPII vesicles should depend on all of the COPII components. Individual COPII components or combinations of two components did not show significant cargo packaging activity (Fig. 3 B). γ-Secretase packaging into vesicles generated with purified COPII (including 23A/24A) proteins was evaluated in 1% DDM samples resolved on BN-PAGE (Fig. 3 C). PS1 in the vesicle fraction was found in the γ-secretase complex (band VI) which also co-migrated with the NCT subunit (Fig. 3 C, left and right panels). Because Sec24 appears to confer cargo specificity to the COPII coat (Miller et al., 2003; Mossessova et al., 2003), Sec24A may contain a recognition site for one or more of the γ-secretase subunits.
We showed that even though γ-secretase formation occurs in the ER or in COPII vesicles, PS1 is found predominantly as FL and suggested that PS1 endoproteolysis occurs at post ER compartments. This model suggests that if we block export of PS1 from the ER, PS1 should remain intact. To test this possibility, we used shRNA transfection to lower the levels of individual Sec24 paralogues (Fig. 4, A and B). A 90% reduction in Sec24A caused a 75% increase in the ratio of PS1 FL/PS1-NTF (Fig. 4, C and D). Sec24C knockdown caused a reproducible but smaller (25%) reduction in PS1 maturation. Our results are consistent with the idea that PS1 endoproteolysis requires traffic to a downstream compartment(s).
In vitro maturation of γ-secretase
Having established that γ-secretase assembly occurs before or within a COPII vesicle, we explored PS1 endoproteolysis as a surrogate marker of γ-secretase activity because only processed PS1 fragments bind γ-secretase inhibitors (Esler et al., 2000; Li et al., 2000; Seiffert et al., 2000; Beher et al., 2003). We solubilized a total membrane fraction in 0.5% CHAPSO lysis buffer, incubated cleared lysate fractions at 37°C for 0–3 h, and analyzed PS1 maturation by SDS-PAGE. PS1-NTF was generated within 1 h of incubation at pH 6.5, suggesting that a slightly acidic environment is optimal for endoproteolysis of PS1 (Fig. 5 A). A concurrent increase of newly generated PS1-CTF was not observed (Fig. 5 A). Perhaps PS1-CTF was generated quickly and then degraded such that it did not accumulate during the incubation.
We then examined PS1 maturation in enriched COPII vesicles (Fig. 5, B and C). We prepared CHAPSO lysates using COPII vesicles generated from N2a cells overexpressing PS1 (Fig. 5 B; Vesicle), from CHO-K1 cells containing endogenous level of γ-secretase subunits (Fig. 5 C; K-1) or from γ-30 cells overexpressing PS1, NCT, PEN-2, and APH-1a (Fig. 5 C; γ-30) (Kimberly et al., 2003). Little if any PS1 proteolysis was observed in these samples. Our results suggest that the apparently intact γ-secretase complex formed in ER is not enzymatically competent. Traffic of γ-secretase to the Golgi may provide an environment (pH, ionic) conducive to endoproteolysis or expose the complex to another protein that promotes endoproteolysis.
Quality control at the ER exit site
Over 100 FAD-linked PS mutations have been found throughout the PS molecule, especially in PS1 (Alzheimer Disease and Frontotemporal Dementia Mutation database; http://www.molgen.ua.ac.be/ADMutations). Many of these mutations affect the folding of PS1 (Berezovska et al., 2005), and yet the mutant proteins promote aberrant APP processing in the secretory stations subsequent to the ER. It is reasonable to assume that these mutant proteins fold sufficiently well to pass ER quality control inspection. Because ER export is a benchmark of underlying ER quality controls, we directly monitored exit of these mutant proteins from the ER using in vitro COPII vesicle formation assay. We used microsomes derived from N2a cell lines stably expressing human APP695swe, a FAD-linked APP variant, and human PS1 variants (Fig. 6, A and B). To account for differences in cargo packaging between experiments, we normalized the packaging efficiency of each cargo molecule to that of p58 in the same reaction (Fig. 6, B and D). As we showed earlier using microsomes from CHO-K1 cells (Fig. 2 A), WT PS1 was found mainly as a full-length form in the transport vesicle fractions (Fig. 6). PS1 ΔE9, a deletion of exon 9 isolated from a family with early onset AD (Crook et al., 1998), showed reduced packaging into COPII vesicles. PS1 D385A, a putative active site mutant, showed a partial defect in the packaging PS1. In all cases, p58 packaging was normal, thus the PS1 ΔE9 packaging defect was cargo selective.
We considered the possibility that the APP695swe mutation may contribute to the abnormal trafficking PS1. To test this idea, we prepared microsomes from N2a cell lines stably expressing human wild-type APP695 and one of three human PS1 variants. Packaging defects for PS1 ΔE9 and PS1 M146L, another FAD-linked PS1 mutation, were obvious although PS1 ΔE9 displayed a more pronounced lesion (Fig. 6, C and D). One other PS1 allele, C410Y, was not detectably different from WT (unpublished data). We conclude that two FAD-linked PS1 mutations, ΔE9 and M146L, are sufficiently defective to impose a quality control delay in ER export of the misfolded forms of PS1.
Incorporation of newly synthesized PS1 into the transport vesicle
To explore the earliest events in the synthesis, folding and packaging of wild-type and mutant PS1, we adapted our budding reaction to measure the incorporation of membrane proteins synthesized and assembled in vitro. A proof of principle is shown in Fig. 7 A (left) in which both p58 and ribophorin I proteins were synthesized and translocated into permeabilized cells. Packaging of radiolabeled p58 into COPII vesicles was prevented by incubation with the dominant-negative Sar1a H79G. A low level of newly synthesized ribophorin I was released in the vesicle fraction in an incubation with or without Sar1a H79G. This was not observed for steady-state ribophorin I levels analyzed by immunoblot on the same PVDF membrane (Fig. 7 A, top panels). Importantly, radiolabeled wild-type PS1 was packaged into COPII vesicles, although less efficiently than p58 (Fig. 7 A, right).
We quantified the efficiency of wild-type and mutant PS1 protein export comparing radiolabel signal in ER membrane pellet and vesicle fractions (Fig. 7 B, middle and bottom). Quantification showed that PS1 ΔE9 export was reduced almost to background levels, whereas PS1 M146L displayed an intermediate defect (Fig. 7 C). The export of FAD mutants PS1 E280A and C410Y was comparable to wild-type levels. From these results we concluded that the export defects of newly synthesized PS1 ΔE9 and PS1 M146L are comparable to the export defects observed in N2a cells stably expressing these FAD-linked mutant proteins as shown in Fig. 6. Thus, the defect is not the result of overexpression of a mutant protein.
A chemical chaperone, TMAO, partially restores packaging of PS1 ΔE9
Protein misfolding accounts for aberrant export of mutant cystic fibrosis transmembrane conductance regulator (CFTR) molecules from the ER (Cheng et al., 1990; Gelman and Kopito, 2002; Sanders and Myers, 2004). Impaired trafficking of CFTR ΔF508 is partially rescued by incubating cells with media containing chemical chaperones such as glycerol and trimethyamine N-oxide (TMAO) (Brown et al., 1996; Sato et al., 1996; Qu et al., 1997; Bennion and Daggett, 2004). To test the idea that the export defect of PS1 ΔE9 is due to misfolding, we performed a transport vesicle formation assay in the presence of increasing concentrations of TMAO (Fig. 8). In addition, to be certain that our reactions recapitulate COPII vesicle formation, we performed control reactions supplemented with a purified dominant-negative Sar1a H79G, to specifically inhibit the generation of COPII vesicles. Packaging of PS1 ΔE9 was increased significantly by addition of TMAO in a concentration-dependent manner (Fig. 8, B and C). The packaging of WT PS1 was only marginally enhanced by TMAO (Fig. 8, A and C). These results are consistent with the notion that the underlying cause of the trafficking defect is at least partly due to protein misfolding. Because TMAO did not fully repair the packaging defect of PS1 ΔE9, it is possible that the exon 9 region may contain export information.
Next, we tested the effect of TMAO on the packaging of the γ-secretase complex bearing PS1 ΔE9. Because addition of 62.5 mM TMAO in the reaction enhanced the overall budding efficiency, we compared 62.5 mM TMAO and 250 mM TMAO conditions where only packaging of PS1 differed (Fig. 8 B, compare 0 and 62.5 mM TMAO conditions). Little or no complex was detected in transport vesicles produced from ΔE9 microsomes at 62.5 mM TMAO condition as expected. Packaging of apparently intact γ-secretase complex was enhanced by 250 mM TMAO from ΔE9 microsomes, whereas packaging of the WT complex was essentially unaffected (Fig. 9, A and B). This result suggests that rescued PS1 ΔE9 is incorporated into the complex.
We report that an apparently immature γ-secretase complex containing intact PS1 is captured in ER-derived COPII vesicles through the intervention of the sorting subunit Sec24A. In vitro PS1 maturation assays suggest that endoproteolysis occurs later in an acidic compartment(s) after a conformational change. Two FAD-linked PS1 alleles, ΔE9 and M146L, retard the packaging of PS1 and the γ-secretase complex. A chemical chaperone, TMAO, restores packaging of PS1 ΔE9, suggesting that this defect is partly due to protein misfolding. Because ER-retained γ-secretase is enzymatically incompetent, ER retention of the complex, per se, cannot directly influence cleavage preference either at the Aβ40 or Aβ42 site.
COPII cargo recognition and sorting is achieved in large measure by the Sec24 subunit (Miller et al., 2003; Mossessova et al., 2003). Mammalian cells have four Sec24 variants (Sec24A, B, C, and D), which show partially overlapping cargo preferences. Despite this redundancy, PS1 showed preference for Sec24A in vitro and in cultured cells (Fig. 3 and Fig. 4). This cargo preference suggests that there will be other cargo proteins whose export from the ER depends solely on a particular Sec24 protein. In yeast, Pma1 export depends on Lst1, a Sec24 paralogue (Roberg et al., 1999; Shimoni et al., 2000). The expression pattern of Sec24 proteins may vary in cell types and during developmental stages. Therefore, spatiotemporal regulation of expression of Sec24 paralogues may be coupled with expression of PS1 to facilitate development.
Membrane protein mutations are a common cause of human disease. In the case of cystic fibrosis conductance regulator (CFTR), over 1,200 cystic fibrosis (CF)–causing mutations have been classified (Gelman and Kopito, 2002): class I mutations abrogate synthesis of the CFTR protein, class II mutations cause defective protein trafficking, class III mutations lead to unstable or nonfunctional protein at the cell surface, and class IV mutations interfere with channel activation and regulation by physiological agonists. The CFTR ΔF508 mutation, which causes misfolding and arrests traffic in the ER, is the most common human allele (more than 90%) (Kerem et al., 1989; Riordan et al., 1989). The CFTR ΔF508 mutation was initially believed to affect trafficking alone (for review see Gelman and Kopito, 2002). However, the folding defect also produces an abnormal response to β-adrenergic agonists and instability at the plasma membrane. Incubation of cells expressing CFTR ΔF508 in a medium containing osmolytes such as glycerol and TMAO increases the steady-state localization of functional CFTR ΔF508 in the plasma membrane (Brown et al., 1997). Unfavorable interaction between osmolytes and the peptide backbone favors the folded state in normal and misfolded mutant proteins (Bolen and Baskakov, 2001). The effect of TMAO on traffic of the PS1 ΔE9 mutant protein resembles the behavior of CFTR ΔF508 mutant protein, thus PS1 misfolding causes pleiotropic effects such as a trafficking defect and abnormal processing of APP.
Misfolded proteins in the ER are subjected to quality control and in severe cases to ER associated degradation (ERAD) (Römisch, 2005). Although the deletion mutant PS1 ΔE9 is profoundly defective in ER export, the mutant protein is as stable as PS1 WT (Ratovitski et al., 1997). In fact, the steady-state distribution of PS1 ΔE9 is only mildly different from that of PS1 WT (Kim et al., 2000). Although the ΔE9 mutant protein is substantially defective in sorting into COPII vesicles in vitro, this delay is overcome with time and the mutant protein is eventually targeted to later compartments. Delays at later stations may have a direct bearing on the misprocessing of APP in cells harboring ΔE9 or other FAD mutant proteins.
Previous studies suggest that formation of the γ-secretase complex can occur in the ER (Kim et al., 2004; Capell et al., 2005). These studies used a recombinant NCT construct with an ER-retrieval motif (KKXX) inserted into the cytosolic portion of NCT in order to retain NCT in the ER. In cells in which the normal NCT was down-regulated, the ER-retained NCT was incorporated into the γ-secretase complex and generated Aβ at a normal level, suggesting that the active γ-secretase complex formation can happen in the ER. However, because the ER-retrieval signal exerts its effect through retrieval rather than inhibition of ER export, it is not clear where activation of the γ-secretase complex occurs. For example, it is possible that γ-secretase is assembled and activated in the ERGIC or in the cis-Golgi and retrieved back to ER. In our study, we monitored the content of ER derived vesicles directly. In addition, normal levels of endogenous PS1 were detected in the γ-secretase complex packaged in ER-derived vesicles.
We discovered that complex formation and activation of this proteolytic enzyme are discrete processes. A conformational change and an acidic environment may be critical requirements for maturation of γ-secretase under normal conditions. Our data suggest that this process occurs at a post-ER compartment(s). Consistent with this model, it is reported that the ER form of NCT is sensitive for trypsin treatment whereas complex glycosylated NCT is resistant (Shirotani et al., 2003). Our results indicate that an acidic environment is not sufficient for γ-secretase activation.
Retention of mutant γ-secretase in the ER may have direct or indirect consequences that influence the development of AD. For example, PS1 may serve as a Ca2+ leak channel in the ER. Disturbance of Ca2+ homeostasis in cells expressing the two FAD-linked PS1 mutations (ΔE9 and M146V) may be the consequence of ER retention (Tu et al., 2006; Nelson et al., 2007). Alternatively, in light of the recent finding that PS has a γ-secretase–independent function in cytoskeletal organization, it is also possible that an ER export delay of γ-secretase or PS1 may disturb cytoskeleton-mediated traffic at later compartments (Khandelwal et al., 2007).
Our results suggest a model in which γ-secretase is assembled in the ER and activated in a downstream acidic compartment. A delay in ER export of the γ-secretase complex may play a role in the disruption of cell physiology.
Materials And Methods
Antibodies were used as described earlier (Kim et al., 2005). Antibodies were diluted as follows: anti-Sec24A antibody, 1:1,000; anti-Sec24B antibody, 1:1,000; anti-Sec24C antibody, 1:1,000; anti-Sec24D antibody, 1:1,000; anti-Sec23 antibody, 1:1,000; anti-actin C4 antibody (Mab), 1:20,000 (MP Biomedicals); anti-PS1 NTF (Ab14), 1:10,000 (a gift from Sam Gandy, Farber's Institute of Neuroscience, Philadelphia, PA); anti-NCT antibody (Guinea pig), 1:2,000 (Chemicon); mouse monoclonal anti-NCT antibody (9C3), 1: 7,000; mouse monoclonal anti-PEN-2 antibody (7D3), 1:9,000; polyclonal anti-APH1a antibody (B80.2), 1:1,000. Anti-NCT antibody, anti-PEN-2 antibody, and anti-APH1a antibodies were gifts from W. Annaert (Center for Human Genetics, Leuven, Belgium).
The human Sec24A coding region was amplified from MGC-12985 (ATCC) using synthetic oligonucleotides JK108 (5′-CCTTCCAACCCCATGGGAATGTCCCAGCCGGGAATACCGGCC-3′) and JK109 (5′-TTTTTTTTGTGCGGCCGCAAATTTTATTTCACATTAAGTCAGTAT-3′). The amplified fragments were digested with NcoI and NotI and inserted into corresponding sites of pHis-hSec24C to replace Sec24C with Sec24A. The resultant plasmid, pJK13s, has a His6 tag at the N terminus, and the C-terminal region is truncated. To construct the full-length Sec24A, we digested pJK13s and pBS-hSec24A (J.-P. Paccaud) with StuI and XhoI. The StuI- and XhoI-digested fragment from pBS-hSec24A containing an intact Sec24A C-terminal region was inserted into the corresponding site of pJK13s to generate a full-length human Sec24A (pJK13). The sequence of this resultant Sec24A is identical to AAH19341.
The human Sec24B coding region was amplified from pCiNeo-Flag-human Sec24B using synthetic oligonucleotides JK110 (5′-TACAAAGACGCCATGGGAATGTCGGCCCCCGCCGGGTCCTCTCAC-3′) and JK111 (5′-GTCTGCTCGAAGCATTAACCCTCA-3′). The amplified fragments were digested with NcoI and NotI and inserted into corresponding sites of pHis-hSec24C to replace Sec24C with Sec24B. This plasmid, pJK14, has a His6 tag at the N terminus.
The human Sec24D coding region was amplified from MGC-46092 (ATCC) using synthetic oligonucleotides JK114 (5′-TGGAATGATAGAATTCAAATGAGTCAACAAGGTTACGTGGCTACA-3′) and JK113 (5′-TTTTTAAGAGCGGCCGCAGTGATTTATTGATACTGGTTAACA-3′). The amplified fragments were digested with EcoRI and NotI and inserted into corresponding sites of pHis-hSec24C to replace Sec24C with Sec24D. This plasmid, pJK15, has a His6 tag at the N terminus.
Plasmids pHis-hSec24C, pJK13, pJK14, and pJK15 are derivatives of pFastBac HTb (Invitrogen) suitable for baculoviral manipulation.
Expression and purification of proteins
Rat liver cytosol was prepared as described previously (Kim et al., 2005). Hamster SarA2 (previously known as Sar1a) protein, Sec23A/24, and Sec13/31A complexes were also purified as described previously (Kim et al., 2005). Sf9 cells infected with either Sec23A/His6Sec24C or Sec13/His6Sec31A viruses were lysed by sonication in a lysis buffer [20 mM Hepes (pH 8.0), 500 mM KOAc, 250 mM sorbitol, 10% glycerol, 10 mM imidazole, 0.1 mM EGTA, and 5 mM β-mercaptoethanol] supplemented with a protease inhibitor cocktail tablet (1×, Roche). The lysate was cleared by successive centrifugation in a Sorvall SS34 rotor for 30 min at 15,000 rpm and in a Beckman 45 Ti rotor for 1 h at 40,000 rpm. The cleared lysate was incubated with prewashed Ni-agarose resin for 1 h at 4°C and poured into a column. The column was washed first with lysis buffer and washing buffer (lysis buffer with 50 mM imidazole) and then with elution buffer (lysis buffer with 250 mM imidazole). Eluates were pooled, distributed into small aliquots, snap frozen in liquid nitrogen, and kept at −80°C.
N2a cells stably expressing human APP695 were transfected with pAG3-PS1WT, or pAG3-PS1M146L or pAG3-PS1ΔE9 plasmid and selected in a medium containing 50% DME, 50% OptiMEM, and 5% FBS supplemented with 200 μg/ml G418 and 400 μg/ml zeocin. N2a cells were transfected with pAG3-PS1WT or pAG3-PS1ΔE9 plasmid and selected in a medium containing 50% DME, 50% OptiMEM, and 5% FBS supplemented with 400 μg/ml zeocin. Colonies derived from single cells were screened for similar levels of PS1 expression (∼15-fold more than endogenous mouse PS1, probably an overestimation because anti–human PS1 NTF antibody was used to quantify signals). Human embryonic kidney (HEK 293) cells stably expressing human APP695 (HEK.APP) were maintained in DME supplemented with 1× NEAA, 10% FBS, and 400 μg/ml G418. HEK 293 cells stably expressing both human APP695 and PS1WT (HEK.APP.PS1 WT) were maintained in DME supplemented with 1× NEAA, 10% FBS, 550 μg/ml G418, and 400 μg/ml hygromcyin. CHO-K1 cells were maintained as described previously (Kim et al., 2005). γ-30 cells were a gift from Dr. Michael Wolfe (Harvard Medical School, Boston, MA).
In vitro vesicle-formation assay
Microsomes, rat liver cytosol, and recombinant COPII proteins were prepared as described previously (Kim et al., 2005). A microsome suspension (30 μl) was mixed in 500 μl reaction buffer for the measurement of absorbance (600 nm). Normally, at an A600 = 0.025, 30 μl of the original microsome suspension was used for a 100-μl budding reaction. When semi-intact cells (SICs) were used instead of microsomes, a suspension (5 μl) was mixed in 500 μl reaction buffer for the measurement of absorbance (600 nm). Normally, at an A600 = 0.02, 5 μl of the original SIC suspension was used for a 100-μl budding reaction. For most samples, 20% of total membranes and 75% of a transport vesicle fraction were loaded into a well. Proteins were resolved by SDS-PAGE and transferred to PVDF membranes. Membrane strips were probed with appropriate primary antibodies and 35S-labeled secondary antibody, then quantified and visualized using a Phosphorimager. Radioactive signals were measured to calculate the budding/packaging efficiency of each cargo molecule based on total input and the efficiency of a cargo protein was normalized to the budding efficiency of p58. Because PS1ΔE9 and PS1D385A are not processed into N-terminal and C-terminal fragments (NTF and CTF), we summed the signals of full-length (FL) form and NTF when calculating the total input of PS1WT and PS1M146L for comparison.
In vitro PS1 endoproteolysis assay
The procedure to make SICs has been described (Wilson et al., 1995). 100 μl of standard SICs (see in vitro vesicle-formation assay section) were centrifuged and the pellet was solubilized in a 100 μl MNM (50 mM MES, pH 6.5, 150 mM NaCl, and 5 mM MgOAc) buffer containing 0.5% CHAPSO. The lysate was cleared by centrifugation at 100,000 g for 30 min. The supernatant was sampled into three tubes (15 μl each) and incubated at 37°C as indicated. After the reaction, samples were analyzed by SDS-PAGE and immunoblotting.
In vitro tranlation, translocation, and budding reaction
The procedure to make SICs and the subsequent in vitro translation reactions have been described (Wilson et al., 1995). After translating the proteins for 45 min at 30°C, 0.75 ml ice-cold KHM buffer (110 mM KOAc, 20 mM Hepes, pH 7.2, and 2 mM Mg(OAC)2) was added to stop the translation reaction and the SICs were harvested at 3,000 g for 3 min at 4°C. The SICs were resuspended in 20 μl KHM and aliquoted before assembling the budding reaction as described above. To calculate the percentage of budding, we correlated the radiolabeled signal in the vesicle fractions (both −rATP and +rATP) to the radiolabeled signal in the ER pellet fractions of each translated protein. Nonspecific leakage found in the vesicle fractions from budding reactions without ATP regeneration mix (−rATP) was subtracted from radiolabeled signal found in vesicle fractions from budding reactions with ATP regeneration mix (+rATP).
Membranes were solubilized with HN buffer containing 50 mM Hepes (7.0), 50 mM NaCl, and appropriate detergents for 30 min at 4°C. Lysates were centrifuged for 30 min at 65,000 rpm in a Beckman TLA100.3 rotor and separated on 4–12.5% polyacrylamide gradient gels (18 cm × 16 cm × 1.5 mm) with high molecular weight markers from Amersham Biosciences (Schägger and von Jagow, 1991).
The coding regions of human Sec24A, Sec24B, Sec24C, and Sec24D were used to design short hairpin RNA (shRNA) constructs for specific gene silencing of the corresponding Sec24 isoforms using the siRNA Target Finder software from Ambion (http://www.ambion.com/techlib/misc/siRNA_finder.html). The shRNA target sequences for individual human Sec24 isoforms were as follows: 5′-AGGAGATGGAAGTGTAACT-3′ (Sec24A), 5′-GGCATTGGTACCGTGTTAT-3′ (Sec24B), 5′-TGTAGCTTCTCAGTTCACT-3′ (Sec24C), and 5′-CCCTTTACAAATACAACAA-3′ (Sec24D). Custom-made 62-nt DNA oligonucleotides were designed to contain a 19-nt sense strand target sequence that was linked to the 19-nt reverse complement antisense sequence by a short spacer region (5′-TTCAAGAGA-3′), followed by the RNA polymerase III termination sequence (5′-TTTTTT-3′), with the addition of BglII and BamHI restriction enzyme sites flanking the 5′ and 3′ ends, respectively. A pair of complementary oligonucleotides for each Sec24 target was synthesized (Integrated DNA Technologies), as follows: 5′-GATCAGGAGATGGAAGTGTAACTTTCAAGAGAAGTTACACTTCCATCTCCTTTTTTTGGAAA-3′ and 5′-AGCTTTTCCAAAAAAAGGAGATGGAAGTGTAACTTCTCTTGAAAGTTACACTTCCATCTCCT-3′ for Sec24A;5′-GATCGGCATTGGTACCGTGTTATTTCAAGAGAATAACACGGTACCAATGCCTTTTTTGGAAA-3′ and 5′-AGCTTTTCCAAAAAAGGCATTGGTACCGTGTTATTCTCTTGAAATAACACGGTACCAATGCC-3′ for Sec24B; 5′-GATCTGTAGCTTCTCAGTTCACTTTCAAGAGAAGTGAACTGAGAAGCTACATTTTTTGGAAA-3′ and 5′-AGCTTTTCCAAAAAATGTAGCTTCTCAGTTCACTTCTCTTGAAAGTGAACTGAGAAGCTACA-3′ for Sec24C; 5′-GATCCCCTTTACAAATACAACAATTCAAGAGATTGTTGTATTTGTAAAGGGTTTTTTGGAAA-3′ and 5′-AGCTTTTCCAAAAAACCCTTTACAAATACAACAATCTCTTGAATTGTTGTATTTGTAAAGGG-3′ for Sec24D.The complementary pair of DNA oligonucleotides was annealed and ligated into the linearized pSUPER.retro.puro shRNA expression vector (OligoEngine) in between the BglII and HindIII sites. This shRNA vector is expressed under the control of a human H1 RNA polymerase III promoter, to direct the synthesis of small RNA transcripts that can fold into short hairpin structure.
HEK.APP and HEK.APP.PS1 WT cells were transiently transfected with 3 μg of either pSUPER.retro.puro-Sec24 constructs or pSUPER.retro.puro empty vector as control using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's instructions. Cell lysates were harvested at 96 h post-transfection in cold lysis buffer [50 mM Tris (pH 7.5), 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, 0.1% SDS, and 1 mM PMSF] supplemented with protease inhibitor cocktail tablet (Roche). Equal amounts of total proteins were analyzed by SDS-PAGE and immunoblotting. Membranes were probed with the appropriate primary antibodies and IRDye labeled secondary antibodies, then visualized and quantified using the LI-COR Odyssey IR Imaging System (LI-COR Biosciences). The level of β-actin was used as a loading control for each membrane.
Online supplemental material
Figure S1 shows packaging of endogenous cargo proteins from CHO-K1 microsomes. Figure S2 shows that anti-PS1-NTF antibody specifically recognizes PS1 derivatives.
We are grateful to Dr. Michael Wolfe for providing the γ-30 cell line, to Dr. Sam Gandy for Ab14 antibody, and to Dr. Wim Annaert for NCT, APH, and Pen2 antibodies. We thank members of Schekman laboratory for critical discussion. We thank Ann Fisher in the cell culture facility for technical assistance.
B. Kleizen was supported by a European Molecular Biology Organization (EMBO) fellowship. This work was supported by funds provided by the Adler Foundation and the Howard Hughes Medical Institute (to R. Schekman).
J. Kim's present address is Section of Genetics, Department of Pediatrics, M.I.N.D. Institute, University of California-Davis Medical Center, Sacramento, CA 95817.
Abbreviations used in this paper: APP, amyloid precursor protein; COPII, coat protein complex II; CTF, C-terminal fragment; ERGIC, ER-to-Golgi intermediate compartments; FAD, familial Alzheimer's disease; FL, full length; NTF, N-terminal fragment; PS, presenilin; SIC, semi-intact cell; WT, wild type.