Cholesterol Modulates -Secretase Cleavage of Amyloid Precursor Protein

Amyloid precursor protein (APP) and cholesterol metabolism are genetically linked to Alzheimer's disease, the latter through apolipoprotein E, a lipid and cholesterol transport protein. We have examined the hypothesis that the processing of APP is disrupted by elevated cholesterol, which is known to modulate the activity of several transmembrane proteins. In the current study, cholesterol, solubilized by methyl-β-cyclodextrin or ethanol, was added to the culture media of APP 751 stably transfected HEK 293 cells. Radiolabeled APP and APP (the soluble N-terminal derivative following α-secretase cleavage) were precipitated from lysates and conditioned media of stably transfected HEK 293 cells; the relative levels were determined by quantitative densitometry following separation by SDS-polyacrylamide gel electrophoresis. The data show that cholesterol, solubilized by methyl-β-cyclodextrin, greatly reduced the levels of APP. Low doses of ethanol-solubilized cholesterol similarly caused a dramatic reduction of APP. By contrast, levels of APP holoprotein remained the same or increased. The large decrease seen in APP production was not due to nonspecific inhibition of secretion because several secreted proteins increased in level. Cholesterol, which impedes membrane fluidity, may lower APP production by impeding the interaction of the substrate with its protease(s). If APP were to function trophically, as suggested by other studies, the current conclusion suggests that changes in cellular cholesterol levels in Alzheimer's disease could contribute to neuronal degeneration by decreasing the production of APP.

N-terminal derivative that is found in human cerebrospinal fluid (Seubert et al., 1993). Another pathway, termed the ␣-secretase pathway, cleaves within the A␤ segment to release APP sol , a nonamyloidogenic soluble N-terminal derivative, also found in human cerebrospinal fluid (Esch et al., 1990;Palmert et al., 1992). The exact cleavage site is between residues 16 and 17 of A␤ (Anderson et al., 1991;Wang et al., 1991). The soluble APP derivative found in the conditioned media of the cell line used in this study, HEK 293 cells, is almost exclusively the result of ␣-secretase cleavage (Wang et al., 1991).
Modulation of APP sol levels may be of physiological consequence. APP sol induces a 2-fold increase in the phosphorylation of tau, a microtubule-associated protein that is hyperphosphorylated in AD (Greenburg et al., 1994). In addition, APP sol has a trophic effect on cerebral neurons in culture (Araki et al., 1991) and is mitogenic for Swiss 3T3 cells (Schubert et al., 1989). APP sol must be added to the growth medium of two different cell lines with reduced APP production to restore normal cell proliferation (LeBlanc et al., 1992;Ninomiya et al., 1993). APP sol also protects neurons against hypoglycemic damage and glutamate toxicity, causing a rapid and prolonged reduction in intracellular Ca 2ϩ concentration (Mattson et al., 1993).
In this study, we have examined the possible disruptive effects of elevated cholesterol on APP processing. Increases in cholesterol previously have been shown to modify the function of certain membrane proteins. Function has decreased, as in the case of the Meta I-Meta II transition of rhodopsin (Mitchell et al., 1990), or increased, as in the cases of the Na ϩ -K ϩ -ATPase, carrier-mediated lactate transport, and the acetylcholine receptor (Yeagle, 1991;Grunze et al., 1980;Craido et al., 1982;Fong and McNamee, 1986). Several lines of evidence have linked alterations in cholesterol metabolism and transport to AD. The E4 allele of ApoE is associated with higher plasma cholesterol levels (Sing and Davignon, 1985), and ApoE4 is present with increased frequency in patients with sporadic and late onset familial AD Saunders et al., 1993;Corder et al., 1993). In addition, hypercholesterolemia is one of the major risk factors for critical coronary artery disease (cCAD), a condition that results in A␤ deposition 3-10 times more frequently than in non-heart disease controls (Sparks et al., 1990(Sparks et al., , 1993. We report that methyl-␤-cyclodextrin-solubilized cholesterol increases the levels of both mature and immature APP holoproteins in a dose-dependent fashion while dramatically reducing the production of APP sol .

MATERIALS AND METHODS
Transfections and Cell Lines-All experiments used APP 751 stably transfected HEK 293 cells. The APP 751 construct was made with pRC-CMV (Invitrogen), which contains the cytomegalovirus promoter and neo R gene. Cells were maintained in minimal essential media (Life Technologies, Inc.) with 10% fetal calf serum (Life Technologies, Inc.), 1% penicillin/streptomycin/Fungizone (Life Technologies, Inc.), and 550 mg/ml G418 (Life Technologies, Inc.). Cells used in experiments with ethanol-solubilized cholesterol were maintained in 15% delipidated serum (Cocalico Biologicals).
Antibodies and Radiolabeled Immunoprecipitation Assay-APP holoproteins were immunoprecipitated with C-terminal polyclonal antisera 8256 (Abbott Laboratories) (described previously in Bodovitz et al. (1995)). APP sol , the soluble ␣-secretase derivative, was immunoprecipitated with N-terminal monoclonal antibody 22C11 (Boehringer Mannheim). A␤ was immunoprecipitated with 4G8 (Anderson et al., 1992). For the radiolabeled immunoprecipitation assay, approximately one million cells/6-cm 2 plate were preincubated for 24 h with methyl-␤cyclodextrin-solubilized cholesterol (Sigma), methyl-␤-cyclodextrin alone (Sigma), ethanol-solubilized cholesterol (Sigma), ethanol alone, or no treatment. The cells were washed once in Hanks' balanced buffer (Life Technologies, Inc.) and incubated with 2 ml of Met/Cys-free Dulbecco's modified Eagle's media (Life Technologies, Inc.), including the same treatment as the preincubation for 30 min. 50 Ci of 35 S-labeled Met/Cys (DuPont NEN) were added to each plate, and the cells were labeled for 8 h.

RESULTS AND DISCUSSION
Cholesterol, water solubilized by methyl-␤-cyclodextrin at a ratio of 46 mg of cholesterol/g of solid, was delivered to human embryonic kidney (HEK) 293 cells stably transfected with APP 751. The cholesterol exchange between cyclodextrin and membranes reaches equilibrium in less than 1 min and is proportional to the concentration of cyclodextrin used (Irie et al., 1992a). Cyclodextrin-solubilized cholesterol, at the doses used here, has been shown to be an effective delivery system for cell culture studies as well as parenteral administration to animals (De Caprio et al., 1992;Irie et al., 1992aIrie et al., , 1992b. We used this delivery system to determine if cholesterol modulation affected the stability of APP, a transmembrane protein with a half-life of only 20 -30 min (Kang et al., 1987;Weidemann et al., 1989). Following a 24-h incubation to allow for intracellular distribution and membrane turnover, the APP 751 stably transfected HEK 293 cells were labeled for 8 h with 35 S-labeled methionine/cysteine in the presence of cholesterol to obtain steady-state measurements. During the entire proce-dure, at all concentrations of cholesterol used, the cells remained attached to the plates; at 240 mg/dl they were slightly rounded, but at 160, 80, and 40 mg/dl cholesterol they looked indistinguishable from controls (data not shown). Cell lysates were used for immunoprecipitation of APP holoproteins with C-terminal polyclonal antiserum 8256 , previously characterized in Bodovitz et al. (1995). Antiserum 8256 immunoprecipitated two bands of ϳ110 and 130 kDa, which correspond, respectively, to immature and mature APP holoprotein (Fig.  1a). Both protein bands increased significantly following the addition of cyclodextrin-solubilized cholesterol ( Fig. 1a and Fig.  2, a and b). Mature APP holoprotein increased 50 Ϯ 10% (p Ͻ 0.001; n ϭ 6) at 40 mg of cholesterol/dl of media. (All values are normalized to no drug control and are given as ϮS.E.; p values were determined by a Student's t test.) The peak increase of 160 Ϯ 37% (p Ͻ 0.05; n ϭ 4) occurred at 160 mg/dl cholesterol, followed by a smaller increase of 70 Ϯ 14% (p Ͻ 0.05; n ϭ 6) at 240 mg/dl cholesterol. Immature APP holoprotein showed a similar rate of increase, e.g. 46 Ϯ 10% at 40 mg/dl cholesterol and 140 Ϯ 64% at 160 mg/dl cholesterol (p Ͻ 0.001; n ϭ 4), but was still increasing 660 Ϯ 190% (p Ͻ 0.001; n ϭ 6) at 240 mg/dl cholesterol.
The modulation of APP holoproteins suggested that downstream catabolites would also be modulated. To test this possibility we immunoprecipitated A␤ and the N-terminal ␣-secretase derivative, APP sol (see below for characterization), from the conditioned media of APP 751 stably transfected HEK 293 cells using, respectively, A␤ monoclonal antibody 4G8 (as characterized by Buxbaum et al. (1992)) and N-terminal monoclonal antibody 22C11 (Boehringer Mannheim). Although A␤ levels showed no consistent change (data not shown), APP sol significantly and reproducibly decreased in the presence of augmented cholesterol (Figs. 1c and 2c). The decrease was 16 Ϯ 8% at 40 mg/dl cholesterol, indicating that APP sol levels were not as sensitive as holoprotein levels to low doses of cholesterol. The decrease was 65 Ϯ 9% (p Ͻ 0.05; n ϭ 4) at 160 mg/dl FIG. 1. Cholesterol modulation of mature and immature APP holoproteins and APP sol . HEK 293 cells stably transfected with APP 751 were preincubated with no drug or the indicated concentration (mg/dl) of methyl-␤-cyclodextrin-solubilized cholesterol or methyl-␤-cyclodextrin alone. The cells were changed to methionine/cysteine-free media containing the same concentration of cholesterol or cyclodextrin and labeled for 8 h with [ 35 S]methionine/cysteine. Cellular lysate was used to immunoprecipitate APP holoproteins with C-terminal antiserum 8256; conditioned media were used to immunoprecipitate the N-terminal ␣-secretase derivative, APP sol , with monoclonal antibody 22C11. Immunoprecipitates were separated by 10% Tris/Tricine SDS-PAGE electrophoresis and visualized by autoradiography. Representative autoradiograms are shown. a, both mature and immature APP holoproteins are increased at 160 mg/dl cholesterol. Mature APP decreases at 240 mg/dl cholesterol, but immature APP continues to increase. c, APP sol decreases dramatically at 160 and 240 mg/dl cholesterol. The control band had to be overexposed in order for the autoradiogram to exhibit any APP sol at 240 mg/dl cholesterol. b, methyl-␤-cyclodextrin was added at 0 or 5200 mg/dl; the latter corresponds to the amount used to solubilize 240 mg/dl cholesterol. This solubilization agent did not affect levels of immature or mature APP holoprotein. d, even at 5200 mg/dl, methyl-␤-cyclodextrin did not affect levels of APP sol . These data demonstrate that the modulation of APP processing seen on the left was the result of cholesterol and not its solubilization agent. cholesterol and 93 Ϯ 1% (p Ͻ 0.001; n ϭ 6) at 240 mg/dl cholesterol, indicating a strong response to higher doses of cholesterol.
The N-terminal soluble APP derivative can be released into conditioned media by either ␣or ␤-secretase cleavage. In order to determine the relative amounts of each cleavage in our system, we immunoprecipitated the total N-terminal derivative from conditioned media, split the precipitates in half, and resolved them on two separate SDS-PAGE gels. The proteins were transferred to membrane and probed with either 22C11, a monoclonal antibody against the N terminus of APP, or 6E10, a monoclonal antibody against residues 1-16 of A␤ (Pirtilia et al., 1995). 22C11 recognizes total N-terminal derivative whereas 6E10 recognizes that resulting from ␣but not ␤-secretase cleavage. Both antibodies recognized bands of similar intensity (Fig. 3). This finding, in conjunction with the sequence analysis of Wang et al. (1991), demonstrates that the soluble N-terminal APP derivative released into the conditioned media of HEK 293 cells is almost exclusively APP sol , the ␣-secretase cleavage derivative.
In order to determine that the modulation of APP processing by cholesterol was not affected by methyl-␤-cyclodextrin, the solubilization agent, we added this agent to our stably transfected cell system and immunoprecipitated APP holoproteins and APP sol . The concentration used, 5200 mg/dl, corresponded to the solubilization of 240 mg/dl cholesterol. Even at this high concentration, methyl-␤-cyclodextrin did not affect levels of APP holoprotein (Fig. 1b) or levels of APP sol (Fig. 1d).
As a further control for the methyl-␤-cyclodextrin-solubilized cholesterol delivery system, we immunoprecipitated APP holoproteins and APP sol from cells maintained in delipidated serum and treated with ethanol-solubilized cholesterol. This delivery system, in the same concentration range used here, has been shown to induce a significant linear increase in cellular free cholesterol (Lasa et al., 1991). Ethanol-solubilized cholesterol, in contrast to cyclodextrin-solubilized cholesterol (Fig. 1a), did not modulate APP holoprotein levels (Fig. 4a). The reason for the difference between the two delivery systems is unknown but could be a result of variation in the intracellular distribution of cholesterol; differential distribution might be the result of cyclodextrin, but not ethanol, mediating both cholesterol influx and efflux at the plasma membrane (Irie et al., 1992a). However, the impact of ethanol-delivered cholesterol on APP sol was dramatic, with reductions even more pronounced than seen with cyclodextrin-solubilized cholesterol (Fig. 4b). Levels of APP sol were greatly reduced with as little as 40 mg/dl cholesterol and all but gone at 80 mg/dl and higher (Fig. 4b). Both delivery systems thus support our major finding that increased cellular cholesterol dramatically inhibits ␣-secretase cleavage.
As a final control for our cyclodextrin delivery system, we examined the effects of methyl-␤-cyclodextrin-solubilized cholesterol on general production of cellular and secreted proteins. We analyzed [ 35 S]methionine/cysteine-labeled proteins from cell lysate and conditioned media by 10% Tris/Tricine SDS-PAGE gel electrophoresis (Fig. 5). A small percentage of the cellular proteins changed in level, but there were no significant global differences in distribution or intensity between the spectra of labeled proteins, even at 240 mg/dl cholesterol (Fig. 5a). This lack of change indicates that cellular protein metabolism was largely unaffected by cholesterol modulation. By contrast, there were several changes in unimmunoprecipitated [ 35 S]methionine/cysteine-labeled secreted proteins (Fig. 5b). There was a decrease in a ϳ120-kDa band with increasing cholesterol concentration; this band was APP sol . Its elevated levels in conditioned media were the result of the overexpression of the APP 751 gene in stably transfected HEK 293 cells. There were no other significant differences in the spectra of labeled secreted proteins between 0 and 160 mg/dl methyl-␤-cyclodextrin-solubilized cholesterol. At 240 mg/dl cholesterol, however, one band at ϳ65 kDa appeared, and several bands in the 40-kDa range increased in intensity. The increase is in opposition to the decrease in APP sol levels, demonstrating that the latter change is not due to nonspecific cholesterol modulation of secreted proteins.
The increase in APP stability with methyl-␤-cyclodextrinsolubilized cholesterol is consistent with several experimental observations. APP is a transmembrane protein (Kang et al., 1987) that is tyrosine-sulfated and O-and N-glycosylated (Wei-

FIG. 2. Densitometric analysis of cholesterol modulation of APP holoproteins and APP sol .
Autoradiograms were scanned into the computer, and bands corresponding to mature APP, immature APP, and APP sol were densitometrically analyzed with the Metamorph imaging system. a and b, increased cellular cholesterol loading increased the steady-state levels of mature and immature APP holoproteins. The increase in mature protein reached a peak and then began to decline, although still remaining above control levels at the highest cholesterol dose used, 240 mg/dl. By contrast, the levels of immature protein were still dramatically increasing at 240 mg/dl cholesterol. c, increased cellular cholesterol loading decreased APP sol levels. The change in level lagged behind the changes seen with the holoproteins at low doses but reached a 93 Ϯ 1% (p Ͻ 0.001; n ϭ 6) decrease at 240 mg/dl cholesterol.  3. Characterization of the N-terminal APP secretase derivative released into the conditioned media. N-terminal secretase derivative, in total, was immunoprecipitated from conditioned media with monoclonal antibody 22C11. The precipitate was split in half and resolved on two separate SDS-PAGE gels. The proteins were transferred to membrane and probed with either 22C11, a monoclonal antibody against the N terminus of APP, or 6E10, a monoclonal antibody against residues 1-16 of A␤. 22C11 recognizes the total N-terminal derivative whereas 6E10 recognizes that resulting from ␣but not ␤-secretase cleavage. Both antibodies recognized bands of similar intensity. This finding, in conjunction with the sequence analysis of Wang et al. (1991), demonstrates that the soluble N-terminal APP derivative released into the conditioned media of HEK 293 cells is almost exclusively APP sol , the ␣-secretase cleavage derivative. demann et al., 1989). APP only has a half-life of 20 -30 min (Weidemann et al., 1989) but is found in many different intracellular membranes, including lysosomal and plasma (Haass et al., 1992), suggesting rapid intracellular movement. Increasing the cholesterol content of phospholipid bilayers increases their rigidity by ordering the acyl chain region and may slow down APP transport, in turn slowing down the transitions from immature to mature and mature to degraded. Even a small decrease in the rate of these transitions would have, given the short half-life of APP, a large effect on the net steady-state levels of mature and immature APP holoproteins.
Mature APP is cleaved by ␣-secretase to generate APP sol , demonstrating a precursor-product relationship (Weidemann et al., 1989), yet augmenting methyl-␤-cyclodextrin-solubilized cholesterol increases levels of mature APP while decreasing levels of APP sol . ␣-Secretase function thus is reduced, either directly or indirectly. Indirectly, function would be reduced if enzyme and substrate were compartmentalized and if cholesterol were to block APP access to the protease-containing compartment. Cholesterol also could have a direct effect on ␣-secretase activity, which appears to be membrane-associated. ␣-Secretase cleavage requires that APP be inserted into a membrane, and it cleaves APP at a fixed distance from the membrane instead of at a specific amino acid sequence (Sisodia, 1992). Furthermore, cleavage is blocked by APP site-directed mutagenesis of three adjacent lysines to glutamic acid residues, located just carboxyl to the proposed transmembrane domain (Usami et al., 1993); this substitution demonstrated that perturbation of the association of APP with the membrane inhibits ␣-secretase cleavage. Another perturbation may be the result of the stiffening of the membrane due to cholesterol loading, possibly inhibiting lateral movement and the required contact between enzyme and substrate (Fig. 6). Modulation of membrane fluidity also affects the external accessibility of membrane proteins, as evidenced by the complex changes in ligand binding of the serotonin receptor as a function of membrane fluidity (Heron et al., 1980). A change in the external accessibility of either ␣-secretase or APP may inhibit the en-zyme activity by disrupting a cleavage event that occurs at a fixed distance from the membrane.
If cholesterol were to increase in the AD brain, our data from the cyclodextrin delivery system suggest an increase in immature and mature APP holoprotein and a decrease in APP sol . The significance of the former change is, at present, unclear. APP may function as a receptor (Kang et al., 1987;Ferreira et al., 1993;Nishimoto et al., 1993) and may be associated with the heterotrimeric G-protein, G o (Nishimoto et al., 1993); its overproduction could lead to aberrant intracellular signaling. Alternatively, the main function of APP may be to serve as a precursor for APP sol and other cleavage derivatives. A decrease in APP sol would be the reduction of a mitogenic and trophic factor as well as a protective agent against hypoglycemic damage and glutamate toxicity (Araki et al., 1991;Schubet et al., FIG. 5. Cholesterol modulation of general cellular and secreted proteins. Aliquots of the cellular lysates and conditioned media collected for use in the immunoprecipitation studies of Fig. 1 were left unimmunoprecipitated, separated with 10% Tris/Tricine SDS-PAGE electrophoresis, and visualized by autoradiography. a, a small percentage of the [ 35 S]methionine/cysteine-labeled cellular proteins changed in level, but there were no significant global differences in distribution or intensity between the spectra of labeled proteins at cholesterol doses of 0, 160, or 240 mg/dl. This lack of change demonstrates that general protein metabolism was unaffected by cholesterol modulation. b, by contrast, there were several changes in unimmunoprecipitated [ 35 S]methionine/cysteine-labeled secreted proteins. There was a decrease in a ϳ120-kDa band with increasing cholesterol concentration; this band was APP sol . Its elevated levels in conditioned media were the result of the overexpression of the APP 751 gene in stably transfected HEK 293 cells. There were no other significant differences in the spectra of labeled secreted proteins between 0 and 160 mg/dl cholesterol. At 240 mg/dl cholesterol, however, one band at ϳ65 kDa appeared, and several bands in the 40-kDa range increased in intensity. The increase is in opposition to the decrease in APP sol levels, demonstrating the latter change is not due to nonspecific cholesterol modulation of secreted proteins. FIG. 6. Possible mechanism of the inhibition of ␣-secretase cleavage by cholesterol. The stiffening of the membrane due to cholesterol loading may decrease ␣-secretase cleavage of APP by inhibiting lateral movement (indicated by horizontal arrows) and the required contact between enzyme and substrate.
FIG. 4. Ethanol-solubilized cholesterol modulation of APP holoproteins and APP sol . As a control for the methyl-␤-cyclodextrinsolubilized cholesterol delivery system, we immunoprecipitated APP holoproteins and APP sol from cells maintained in delipidated serum and treated with ethanol-solubilized cholesterol. a, this delivery system, in contrast to cyclodextrin-solubilized cholesterol, did not modulate APP holoprotein levels. The reason for the difference between the two delivery systems is unknown but could be a result of variation in the intracellular distribution of cholesterol; differential distribution might be the result of cyclodextrin, but not ethanol, mediating both cholesterol influx and efflux at the plasma membrane. b, the two delivery systems did, however, yield similar results on the modulation of APP sol , with ethanol-solubilized cholesterol generating an even more dramatic reduction. Levels of APP sol were greatly reduced at as little as 40 mg/dl cholesterol and all but gone at 80 mg/dl and higher. Both delivery systems, despite the differences in modulation of APP holoproteins, support our major finding that increased cellular cholesterol dramatically inhibits ␣-secretase cleavage.
1989; Mattson et al., 1993). The loss of such a protein could exacerbate the cell death in AD.
Some evidence is available to suggest that cholesterol levels increase in AD. There is a predisposition for AD associated with the E4 allele of the cholesterol and lipid transport protein ApoE Saunders et al., 1993;Corder et al., 1993). Lipoproteins associated with ApoE4 are cleared more efficiently than those containing either of the other two alleles, E3 or E2 (Poirier et al., 1993), yet the E4 allele is linked to higher plasma cholesterol levels (Sing and Davignon, 1985) and is the most common of the three alleles in hypercholesterolemia (Utermann, 1984). The discrepancy stems from the efficient clearing of ApoE4 in the liver, possibly leading to a down-regulation of the low density lipoprotein receptor and, hence, an elevation of serum low density lipoprotein cholesterol (for review, see Davignon et al. (1988)).
Hypercholesterolemia is one of the major risk factors for cCAD, a condition that often results in A␤ deposition similar to that found in AD (Sparks et al., 1990(Sparks et al., , 1993. Cerebral A␤ plaques are 3-10 times more common in cCAD than in nonheart disease controls (Sparks et al., 1993). In addition, hypercholesterolemia induced in rabbits resulted in elevated immunoreactivity of A␤ and ALZ-50, an epitope only found in mature brain afflicted by AD (Sparks et al., 1994). All of these data suggest that increased cholesterol levels are a risk factor for AD. Our data provide a mechanism, namely the direct disruption of APP processing.