Association of active γ-secretase complex with lipid rafts

Cholesterol has been implicated in the pathogenesis of Alzheimer's disease (AD). Although the underlying mechanisms are not yet clear, several studies have provided evidence for the involvement of cholesterol-rich lipid rafts in the production of amyloid β peptide (Aβ), the major component of amyloid deposits in AD. In this regard, the γ-secretase complex is responsible for the final cleavage event in the processing of β-amyloid precursor protein (βAPP), resulting in Aβ generation. The γ-secretase complex is a multiprotein complex composed of presenilin, nicastrin (NCT), APH-1, and PEN-2. Recent reports have suggested that γ-secretase activity is predominantly localized in lipid rafts, and presenilin and NCT have been reported to be localized in lipid rafts. In this study, various biochemical methods, including coimmunoprecipitation, in vitro γ-secretase assay, and methyl-β-cyclodextrin (MβCD) treatment, are employed to demonstrate that all four components of the active endogenous γ-secretase complex, including APH-1 and PEN-2, are associated with lipid rafts in human neuroblastoma cells (SH-SY5Y). Treatment with statins, 3-hydroxy-3-methylglutaryl-CoA-reductase inhibitors, significantly decreased the association of the γ-secretase complex with lipid rafts without affecting the distribution of flotillin-1. This effect was partially abrogated by the addition of geranylgeraniol. These results suggest that both cholesterol and protein isoprenylation influence the active γ-secretase complex association with lipid rafts.

It is well established that the abnormal generation and deposition of amyloid ␤ peptides (A ␤ s) is a pathologic hallmark of Alzheimer's disease (AD). A ␤ is generated by two sequential proteolytic cleavage steps from the ␤ -amy-loid precursor protein ( ␤ APP) (reviewed in refs. [1][2][3]. ␤ APP is initially cleaved by ␤ -secretase, which has been identified as a ␤ -site APP-cleaving enzyme (4), followed by the subsequent intramembrane proteolysis of the membrane-bound C-terminal fragment ( ␤ CTF, C99) catalyzed by ␥ -secretase. In an alternative pathway, ␤ APP is cleaved by ␣ -secretase within the A ␤ domain and the remaining CTF (C83) is also cleaved by ␥ -secretase to release the nonamyloidogenic p3 peptide.
Recent studies suggest that the activation of ␥ -secretase requires the formation of a stable high-molecular-weight multiprotein complex, which includes an endoproteolysed and fragmented form of presenilin, nicastrin (NCT), APH-1, and PEN-2 (5). These four transmembrane proteins are presumed to be indispensable for ␥ -secretase activity, because their coexpression enables reconstitution of the ␥ -secretase activity (5)(6)(7)(8)(9), whereas the absence of even one results in the absence of ␥ -secretase activity and defects in the expression and/or maturation of the remaining partners (5,6). ␥ -Secretase is known to be responsible for the intramembranous cleavage of other type I transmembrane proteins, such as Notch, CD44, and E-cadherin (5).
There is growing evidence that cholesterol is linked to the development of AD. For example, the 4 allele of apolipoprotein E, which is a major apolipoprotein in the brain, has been identified as an important risk factor for AD (10). Several epidemiological studies have suggested that the use of 3-hydroxy-3-methylglutaryl-CoA-reductase inhibitors (statins) to treat hypercholesterolemia may reduce the risk of dementia (11)(12)(13). In vivo studies have demonstrated that high-cholesterol diets can increase A ␤ levels in rabbits (14) and in an AD mouse model (15). Treat-ment with cholesterol synthesis inhibitors can lower the A ␤ levels in guinea pig (16) and in the AD mouse (17). Cholesterol depletion also inhibits A ␤ generation in hippocampal neurons (16,18) and increases ␣ -secretase activity in cultured cells (19). Thus, it has been reported that cholesterol depletion by statins and/or cholesterolextracting agents appears to inhibit the amyloidogenic pathway while inducing the alternative nonamyloidogenic pathway (20).
Several lines of evidence suggest the involvement of lipid rafts in ␤ -and ␥ -cleavage of ␤ APP (21)(22)(23)(24)(25). Lipid rafts are membrane microdomains enriched in cholesterol and sphingolipids, and proteins can be selectively included or excluded from these microdomains (26). Lipid rafts have been shown to function as a concentrating platform for a variety of signal transduction molecules (27,28). It has been reported that the proteins relevant to A ␤ generation, including presenilin, NCT, and a small portion of ␤ APP, localize in lipid rafts (21)(22)(23)(24)(25)(29)(30)(31)(32). It has also been reported that ␤ -secretase localizes in lipid rafts and that cholesterol depletion abrogates this localization (22). Ehehalt et al. (21) have reported that ␤ APP exists in two pools, one associated with lipid rafts, in which ␤ -cleavage occurs, and another outside of lipid rafts, where ␣ -cleavage occurs. Recent reports have suggested that ␥ -secretase activity is predominantly localized in lipid rafts and that cholesterol can directly regulate the ␥ -secretase activity in isolated lipid rafts (24). Thus, lipid rafts offer a structural platform for examining the effect of cholesterol on A ␤ generation. Here, we report that the four endogenous components of the ␥ -secretase complex, including APH-1 and PEN-2 in SH-SY5Y cells, are associated with detergent-resistant membranes (DRMs). We also show that statin treatment affects the association of the active form of the ␥ -secretase complex with lipid rafts.

Materials
Dulbecco's modified Eagle's medium, fetal bovine serum, M ␤ CD, lipoprotein-deficient serum (LPDS), and geranylgeraniol were purchased from Sigma. L-685,458 and kifunensine were obtained from Calbiochem. Hybond-ECL nitrocellulose membrane and protein G Sepharose were purchased from Amersham. Complete protease inhibitor cocktail was from Roche Diagnotics. All other reagents were purchased from Wako, unless otherwise specified.

Cell lines and culture
Human SH-SY5Y neuroblastoma cells were obtained from The American Type Culture Collection. Wild-type and presenilin 1 Ϫ / Ϫ /presenilin 2 Ϫ / Ϫ mouse embryonic fibroblasts (33) were kindly provided by Drs. W. Annaert and B. De Strooper. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and penicillin/streptomycin (Gibco-BRL) at 37 Њ C in 5% CO 2 . For the inhibition of ␥ -secretase activity, cells were incubated with 2 M L-685,458 for 18 h. For cholesterol depletion experiments, cells were incubated with or without 2 mM M ␤ CD in serum-free media for 30 min at 37 Њ C.
In statin treatment experiments, cells were washed with PBS, and then cultured in Dulbecco's modified Eagle's medium supplemented with 5% LPDS and 250 M sodium mevalonate in the absence or presence of statins (50 M compactin from Dr. A. Endo or 5 M pitavastatin from Kowa Co., Ltd.) for 48 h. In statin and geranylgeraniol treatment experiments, cells were cultured with 5% LPDS and 10 M pitavastatin in the absence or presence of 10 M geranylgeraniol for 48 h. For the inhibition of mannosidase I, cells were treated in the absence or presence of 1 g/ml kifunensine for 48 h.

Isolation of DRMs
All steps were carried out at 4 Њ C. Confluent cells were washed and scraped into ice-cold PBS and resuspended in 200 l buffer R (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA) containing 2% 3-[(3-cholamidopropyl)dimethylammonio]-2-hydroxy-1-propanesulfonate (CHAPSO) and Complete protease inhibitor cocktail. After a 20 min incubation on ice, the cell lysate was adjusted to 45% sucrose and placed in the bottom of a 5 ml Beckman ultra-clear centrifuge tube and overlaid with 3 ml 35% sucrose in buffer R, followed by 1 ml 5% sucrose in buffer R. After centrifugation at 100,000 g for 16 h at 4 Њ C in a Beckman SW55 rotor, 10 fractions (500 l) were collected from the top of the gradient.

In vitro ␥ -secretase assay and coimmunoprecipitation
In vitro ␥ -secretase assays using each gradient fraction and quantitation of A ␤ by two-site ELISA were performed as previously described, with some modifications (6,9,36). Purified recombinant C100-Flag/Myc/His (1 M) was incubated with each fraction in 1 ϫ ␥ buffer [HEPES buffer containing 5 mM EDTA, 5 mM 1,10-phenanthroline, 10 g/ml phosphoramidon, 0.1% phosphatidylcholine (Avanti), Complete protease inhibitor cocktail] adjusted to 0.25% CHAPSO at 4 or 37 Њ C for 6 h. The reaction was stopped by boiling for 2 min. The samples were centrifuged, and the supernatants were analyzed by two-site ELISA. For coimmunoprecipitation, fraction 3, representing the DRM, and fraction 9, representing the detergent-soluble fraction, were precleared with protein G Sepharose for 2 h at 4 Њ C, incubated with anti-G1L3 or control rabbit IgG overnight, followed by reaction with protein G Sepharose for 3 h at 4 Њ C. After washing 3 times with buffer R containing 1% CHAPSO, immunoprecipitates were eluted in 1 ϫ SDS-PAGE sample buffer, and were then analyzed by immunoblotting. and the resulting lower chloroform phase was dried under nitrogen gas. Lipids were redissolved in the eluent of an HPLC. Free cholesterol was measured with an HPLC equipped with a UV detector at 210 nm as previously described (37). An LC-8 column (4.6 ϫ 250 mm, 5 m particle size, Supelco, Tokyo) and isopropyl alcohol-acetonitrile (70:30, v/v) as an eluent were used at a flow rate of 1.0 ml/min. Protein concentrations were measured by BCA protein assay kit (Pierce).

Endogenous ␥ -secretase complex is associated with DRM in active forms
To confirm that endogenous ␥ -secretase complex is present in lipid rafts in cultured cells, we used SH-SY5Y cells, a human neuroblastoma cell line, to isolate DRM by a discontinuous sucrose density gradient method using 2% CHAPSO as the nonionic detergent. Under these experimental conditions, the use of Triton X-100 is inappropriate, because it results in the inactivation of ␥ -secretase activity (38). CHAPSO has been previously used with efficiency similar to that of Triton X-100 (39) for the preparation of DRM enriched in presenilin without affecting ␥ -secretase activity (24). SH-SY5Y cells were cultured in complete medium containing 10% fetal bovine serum, and the 2% CHAPSO lysate was subjected to sucrose density gradient centrifugation.
by guest, on July 19, 2018 www.jlr.org Downloaded from majority of NCT observed in DRM was in the mature form (Fig. 1A). A recent report (40) indicated that two APH-1a splice forms, APH-1aL and APH-1aS, and APH-1b exist in mammalian cells, and each APH-1 variant is a constituent of six distinct active ␥ -secretase complexes that contains either presenilin 1 or presenilin 2 as the catalytic subunit. We therefore examined the localization of APH-1aS and APH-1b in addition to APH-1aL. In SH-SY5Y cells, APH-1aS and APH-1b were also observed in DRM as well as APH-1aL (Fig. 1A). To examine the localization of the substrate of ␥ -secretase, ␤ CTF/C99, which is undetectable under normal conditions, a specific ␥ -secretase inhibitor (L-685,458) was used. L-685,458 treatment resulted in the accumulation of ␤ CTF/C99 in DRM as well as in the detergent-solu-  ble fraction (Fig. 1A). There was no change in the composition of the ␥ -secretase complex in DRM with this treatment (data not shown).
We next investigated whether these ␥ -secretase components retained ␥ -secretase activity in DRM. Coimmunoprecipitation analysis with an antibody against presenilin 1 CTF showed that mature NCT, APH-1aL, and PEN-2 were coimmunoprecipitated with presenilin 1 from DRM (fraction 3), but flotillin-1 was not (Fig. 1B). Using recombinant C100 as a substrate, we assessed ␥ -secretase activity in all fractions, and the activity was normalized to the protein content. As shown in Fig. 1C, a high specific activity of ␥ -secretase was detected in DRM, consistent with a previous report (24). Small amounts of residual ␥ -secretase activity were observed in the middle fractions 5-8 from our DRM preparation (Fig. 1C). This activity is probably at-tributable to the small amounts of the ␥ -secretase complex observed in fractions 5-8 after DRM preparation, as confirmed by immunoblotting (Fig.1A). These results indicate that the four essential components form the active ␥ -secretase complex in DRM.
Cholesterol is an essential component of lipid rafts, and depletion of cholesterol by M ␤ CD, a cholesterol-sequestering reagent, is commonly used to disrupt lipid rafts, resulting in lipid raft proteins becoming solubilized upon detergent treatment (27). Therefore, we treated SH-SY5Y cells with 2 mM M ␤ CD for 30 min. This treatment caused the reduction of cholesterol in DRM ( Fig. 2A ), and resulted in the partial solubilization of flotillin-1 (Fig. 2G). Under these conditions, presenilin 1 CTF, presenilin 2 CTF, mature NCT, APH-1aL, and PEN-2 shifted from DRM to fractions 9 and 10 ( Fig. 2B-F). These results demonstrate that The maturation of NCT is associated with ␥-secretase complex assembly (33,(41)(42)(43) and can be blocked by kifunensine, a mannosidase I inhibitor (33,43). In the presence of kifunensine, immature-like NCT was increased in SH-SY5Y cells, and was also observed in DRM (Fig. 3). In addition, the distributions of presenilin 1, presenilin 2, APH-1aL, and PEN-2 were unchanged upon kifunensine treatment.

␥-Secretase components shifted from DRM to nonraft by statin treatment
We next tested the effects of statin treatment, a 3-hydroxy-3-methylglutaryl-CoA-reductase inhibitor, on the distribution of the ␥-secretase complex. Cells were cultured in medium supplemented with LPDS to reduce intracellular cholesterol stores, and sodium mevalonate (250 M) in the absence or presence of statins. It has been previously reported that under these conditions, statin treatment prevents the production of A␤ in cultured cells (16,18,19,21). SH-SY5Y cells were treated with compactin, a widely used statin, or pitavastatin (47), a novel and robust statin, for 48 h, followed by DRM isolation. Under control conditions, i.e., in medium supplemented with LPDS and mevalonate in the absence of statins, a significant amount of cholesterol was found in DRM (Fig. 4A). In contrast, statin treatment reduced the cholesterol content in DRM (Fig. 4A). In the results of immunoblotting, the distribution of flotillin-1 did not dramatically change upon treatment with statins (Fig. 4G). In regard to the ␥-secretase complex, the shifts of presenilin 1 CTF, presenilin 2 CTF, mature NCT, APH-1aL, and PEN-2 from DRM to the detergent-soluble fraction were observed upon statin treat-ment ( Fig. 4B-F). Statin treatment did not affect the expression level of the ␥-secretase complex or the maturation of NCT. In addition to their cholesterol-lowering effects, statins have pleiotropic effects attributed to inhibition of protein isoprenylation (48). To define the effects of statins on the distributional change of the ␥-secretase complex resulting from the blockage of isoprenylation, cells were treated with or without geranylgeraniol. Geranylgeraniol can serve as the major substrate for isoprenylation in the presence of a statin, and is more cell-permeable than geranylgeranylpyrophosphate (49). Addition of geranylgeraniol partially abrogated a shift of presenilin 1 CTF from DRM to detergent-soluble fraction upon statin treatment (Fig. 5).

DISCUSSION
Several epidemiological studies have suggested that cholesterol lowering by statin treatment is beneficial for some AD patients (11)(12)(13). In vitro and in vivo studies have further demonstrated that A␤ generation is decreased by treatment with these lipid-lowering drugs (16)(17)(18)(19). These observations led us to study the relationship between the ␥-secretase complex and lipid rafts. As has been demonstrated in several reports, the four essential components of the ␥-secretase complex affect each other's stability, localization, and maturation, and the relative abundance of each component is rate limiting for the assembly of the active enzyme (5). Therefore, exogenously expressed proteins may not adequately represent the physiological state of ␥-secretase in the membrane. In this study, we evaluated endogenously expressed proteins in the neuroblastoma cell line SH-SY5Y and in nonneuronal mouse embryonic fibroblasts. When the cells were lysed with CHAPSO and examined by sucrose density gradient ultracentrifugation, active presenilin forms (presenilin CTF and presenilin NTF) and mature NCT were recovered in DRM as well as in the CHAPSO-soluble fractions, consistent with previous reports (24,25,29,32). In addition, we now report that each of the endogenous APH-1 variants and endogenous PEN-2 are also located in the same DRM fractions as the presenilin fragments. In vitro ␥-secretase assay and coimmunoprecipitation results together showed that essential ␥-secretase components in lipid rafts formed the active ␥-secretase complex. M␤CD pretreatment resulted in the shifts of ␥-secretase components from DRM to the soluble fraction, confirming that the active ␥-secretase complex localizes in cholesterol-rich lipid rafts.
The majority of NCT observed in lipid rafts was found in the mature form (32). Inhibition of NCT maturation by kifunensine treatment resulted in the incorporation of immature NCT into the ␥-secretase complex, and did not influence the distribution of presenilin 1, presenilin 2, APH-1aL, or PEN-2 into lipid rafts. These data are consistent with the observation that kifunensine treatment does not affect A␤ production or the cell surface localization of NCT (33,43). On the other hand, presenilin 1/presenilin 2 ablation caused the dissociation of NCT from lipid rafts. It has been reported that in the absence of presenilin 1/ presenilin 2, NCT and PEN-2 are retained in the endoplasmic reticulum (33,(44)(45)(46). Taken together, these results suggest that the maturation of NCT is not essential for the targeting of the ␥-secretase complex to lipid rafts, but the localization of NCT and PEN-2 in lipid rafts is dependent on presenilin expression. Under the same conditions, a significant amount of APH-1aL was associated with lipid rafts in the absence of presenilin 1/presenilin 2. Therefore, there may be a presenilin-independent trafficking pathway for APH-1 to lipid rafts, suggesting that APH-1 may play an important role in the raft association of the ␥-secretase complex.
Inhibition of 3-hydroxy-3-methylglutaryl-CoA-reductase by statin treatment induced a shift in the location of ␥-secretase components from lipid rafts to the detergentsoluble fraction. However flotillin-1, a well-characterized raft-resident protein, did not shift upon statin treatment. In addition, flotillin-1 was not found to be coimmunoprecipitated by the anti-presenilin 1 antibody, suggesting that flotillin-1 and ␥-secretase may be associated with different rafts. Other possibilities include that they may behave differently in terms of their partitioning into cholesterol-rich membrane microdomains, or that ␥-secretase is more rapidly recycled between membranes than is flotillin-1 (27). Recently, it has been reported that flotillin-1 DRM association in Jurkat T cells and U937 cells was resistant to M␤CD treatment, suggesting that the different sensitivities to cholesterol are considered to be the most plausible mechanism (50).
In addition to their effects on cellular cholesterol levels, statins are known to exert pleiotropic effects through the blockage of isoprenylation (48). In the presence of pitavastatin in LPDS, addition of geranylgeraniol partially abrogated the shift of presenilin 1 from DRM to detergentsoluble fraction. This observation suggests that protein isoprenylation may also play an important role in the association of the ␥-secretase complex with lipid rafts. Wahrle et al. (24) have reported that cholesterol-dependent ␥-secretase activity localizes in the buoyant cholesterolrich membrane microdomain. In a conflicting report, Wada et al. (25) observed that ␥-secretase activity in lipid rafts does not depend on cholesterol. The findings presented here suggest that the ␥-secretase complex association with lipid rafts is influenced by cholesterol and protein isoprenylation, and may affect the efficiency of contact with its substrate in lipid rafts.
Statin treatment has been reported to reduce the generation of A␤ in vitro (16,18,19,21). In these reports, the cells were treated with medium containing the statin, LPDS as a serum, and a small amount of sodium mevalonate for 48-72 h. Ehehalt et al. (21) suggested that the clustering of the ␤APP and ␤-secretase contained in lipid rafts occurs during endocytosis and that ␤-cleavage of ␤APP is carried out in endosomes. It has been previously shown that cholesterol depletion abrogates the localization of ␤-secretase in lipid rafts (22). Therefore, the reduction of A␤ generation by statin treatment may depend on the inaccessibility of ␤-secretase in lipid rafts and/or a decrease in the clustering of the ␤APP and ␤-secretase rafts (16,18,(20)(21)(22)(23). The results presented here demonstrate that the ␥-secretase complex and C99 also colocalize in lipid rafts, and that the association of ␥-secretase with lipid rafts is affected by statin treatment. Our results suggest that the reduction in A␤ production upon statin treatment may be due in part to the dissociation of the ␥-secretase complex from lipid rafts. In the future, it will be important to determine whether statin treatment can induce a similar distributional change of the ␥-secretase complex in neuronal cells in vivo.
The association of the active ␥-secretase complex with lipid rafts and its relation to cholesterol metabolism shown here will provide new insights for understanding the mechanism of A␤ generation in AD.