Cholesterol-dependent Generation of a Seeding Amyloid β-Protein in Cell Culture*

Deposition of aggregated amyloid β-protein (Aβ), a proteolytic cleavage product of the amyloid precursor protein (1), is a critical step in the development of Alzheimer’s disease (2). However, we are far from understanding the molecular mechanisms underlying the initiation of Aβ polymerization in vivo. Here, we report that a seeding Aβ, which catalyzes the fibrillogenesis of soluble Aβ, is generated from the apically missorted amyloid precursor protein in cultured epithelial cells. Furthermore, the generation of this Aβ depends exclusively on the presence of cholesterol in the cells. Taken together with mass spectrometric analysis of this novel Aβ and our recent study (3), it is suggested that a conformationally altered form of Aβ, which acts as a “seed” for amyloid fibril formation, is generated in intracellular cholesterol-rich microdomains.

A␤ 1 is physiologically secreted into the extracellular space; however, why and how soluble A␤ aggregates and forms amyloid fibrils remains to be elucidated. A great deal of effort has been made to clarify this issue, using mainly in vitro systems. In most such experiments, it has been found that A␤ at much higher concentrations than those prevailing in biological fluids is needed for A␤ aggregation. Thus, it has been hypothesized that aggregation of soluble A␤ involves seeded polymerization (4,5), although this assumption has not yet been proved in vivo.
We have recently reported the detection of a novel A␤ in the apical compartment of cultures of MDCK cells that had been stably transfected with APP cDNA (⌬C MDCK cell) with a truncated cytoplasmic domain (⌬C APP) (3). This A␤ species possesses unique molecular characteristics including its appearance as a smear on immunoblots and altered immunoreactivity. Significantly, these molecular characteristics disappeared dramatically following treatment of the cells with compactin or filipin, an inhibitor of de novo cholesterol synthesis and a cholesterol-binding drug, respectively. Based on pre-viously reported evidence for ⌬C APP being missorted to the apical surface (6) and the cholesterol concentrations of the apical plasma membrane and apical transport vesicles being higher than those in other cellular membranes (7), we concluded that a novel A␤ is generated from apically missorted APP in a cholesterol-dependent manner.
Regarding the involvement of cellular cholesterol in the generation of the pathogenic protein, it must be noted that cholesterol-rich lipid microdomains within cells, called caveolae-like domains, have been reported to be the likely sites of the conversion of the normal cellular form of prion protein (PrP c ) to its pathogenic form (PrP sc ) (8 -11). Thus, it would be of great interest to investigate whether the novel A␤ detected in our recent study (3), the generation of which is exclusively dependent on the presence of cholesterol in the cell, has the potential to act as a seed for fibrillogenesis of soluble A␤.

EXPERIMENTAL PROCEDURES
MDCK Cell Culture-Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum was used as the culture medium. We plated 2.5-3.0 ϫ 10 5 MDCK cells transfected with human APP695 cDNA, with full-length APP (wild-type MDCK cell) or APP with a truncated cytoplasmic domain (⌬C MDCK cell) (12) onto 24-mm Transwell filters (Costar) and cultured these cells on the filters for 3 days. To determine the integrity of the cell monolayers that grew on the filters, measurement of the electric resistance between the apical and basolateral compartments of the MDCK cell culture was performed by immersing electrodes into each of the compartments. The MDCK cells culture media were changed 3 days after plating, and the cells were cultured 24 h longer.
Thioflavin T Assay and Congo Red Assay of Fibril Formation of Synthetic A␤-Thioflavin T assay was performed as described elsewhere (13), on a spectrofluorophotometer (RF-5300PC, Shimadzu, Kyoto, Japan). Optimum fluorescence measurements of amyloid fibrils were obtained at the excitation and emission wavelength of 446 nm and 490 nm, respectively, with the reaction mixture (1.0 ml) containing 5 M thioflavin T (Nakalai tesque, Inc., Kyoto, Japan) and 50 mM of glycine-NaOH buffer, pH 8.5. Fluorescence was measured immediately after making the mixture and averaged for an initial 5 s. Synthetic A␤ (A␤1-40, Bachem Switzerland) was initially dissolved in ice-cold distilled water at a concentration of 100, 200, and 300 M, and then diluted with 9 volumes of PBS. Aliquots of the A␤ solutions were incubated in Eppendorf tubes at 37°C. Every hour after the start of incubation, 10 l of the solution of synthetic A␤ was taken and mixed with 990 l of the reaction mixture. The lot number of the A␤ used in the experiment of Fig. 1 was 518765 and that of the A␤ used in the other experiments was 510313. Peak fluorescence was dependent on the concentration of A␤. We used 20 M A␤ peptide for this study. Congo red assay was performed as described elsewhere (14).
Electron Microscopy-Samples were spread on carbon-coated grids, negatively stained with 1% phosphotungstic acid, pH 7.0, then examined under a Hitachi H-7000 electron microscope with an acceleration voltage of 75 kV.
Enzyme Immunoassay-The enzyme immunoassay (EIA) was per-* This study was supported by a research grant for Longevity Sciences (8A-1) and Brain Research Science from the Ministry of Health and Welfare and by Core Research for Evolutional Science and Technology, Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ ‡ To whom correspondence should be addressed. Tel.: 81-562-44-5651 (ext. 834); Fax: 81-562-44-6594; E-mail: katuhiko@nils.go.jp. 1 The abbreviations used are: A␤, amyloid ␤-protein; APP, amyloid formed essentially as described previously (3,15). Briefly, 100 l of the media of the MDCK cell cultures were diluted with 400 l of buffer C (20 mM phosphate buffer (pH 7.0), 0.4 M NaCl, 2 mM EDTA, 10% Block Ace (Dai-nippon, Tokyo, Japan), 0.2% bovine serum albumin, and 0.05% NaN 3 ), and 100 l of the mixture were subjected to the multiwell plates coated with 4G8, a monoclonal antibody specific for A␤17-24 (16). Appropriate amounts of synthetic A␤40 (A␤1-40, Bachem Switzerland) were applied to the multiwell plates for the construction of a standard curve. The plates were incubated at 4°C overnight. After rinsing with PBS, loaded wells were reacted with appropriately diluted horseradish peroxidase-conjugated BA27, a monoclonal antibody specific for A␤40, at 4°C overnight. Bound enzyme activities were measured using the TMB Microwell peroxidase substrate system (Kirkegaard and Perry Laboratories, Gaithersburg, MD). Effect of Addition of the A␤ Immunoprecipitated from the Medium of the MDCK Cells on the Fibril Formation of Synthetic A␤-Aliquots (1.5 ml) of the medium from the transfected MDCK cell cultures were incubated with 4G8 (1.5 g) at 4°C overnight for A␤ immunoprecipitation 3 days after changing the medium. The mixtures were then incubated with protein G-Sepharose at 4°C for 3 h and centrifuged. The pellets were washed thoroughly in RIPA buffer once, and then in Tris-saline buffer four times. The A␤ in the immunoprecipitates was extracted in 25 l of a buffer containing 2% SDS by boiling. Following dilution with 975 l of Tris-saline buffer, 5 l of the solution (containing 1 pmol of the immunoprecipitated A␤) was mixed with 10 l of the synthetic A␤ solution (containing 2 nmol of A␤1-40) and 85 l of PBS buffer. The level of the immunoprecipitated A␤ was determined by enzyme immunoassay (data not shown), and the molecular ratio between synthetic A␤ and the immunoprecipitated A␤ was approximately 2,000:1. Amyloid fibril formation of synthetic A␤ was quantitatively determined by measuring fluorescence intensity at 2 and 6 h after starting the incubation as described before. The background fluorescence intensity was determined using an extract of protein G-Sepharose that was incubated with 4G8 in fresh medium.
Inhibition of de Novo Cholesterol Synthesis-To inhibit de novo cholesterol synthesis, MDCK cells were incubated for 90 min with compactin (Sigma), a potent competitive inhibitor of 3-hydroxy-3-methylglutaryl-coenzyme A reductase activity, at concentrations of 0.5 and 1.0 M, before changing the media to fresh ones. The cells were further cultured for 3 days with compactin, and then 1.5-ml aliquots of the medium from the apical and basolateral compartments of the culture were used for immunoprecipitation.
Treatment with Filipin-MDCK cells were cultured in medium containing filipin (Sigma), a polyene antibiotic that specifically binds to cholesterol, at concentrations of 0.1 and 0.3 g/ml for 90 min before changing the media to fresh ones. The cells were further cultured for 3 days with filipin, and then 1.5-ml aliquots of the medium from the apical and basolateral compartments of the culture were used for immunoprecipitation.
Addition of Exogenous Cholesterol-To confirm that the seeding ability of the apical A␤ depends on the presence of cholesterol, we investigated whether the effect of compactin and filipin was reversed by the addition of exogenous free cholesterol.
Determination of the Total Cellular Level of Cholesterol and the Level of de Novo Cholesterol Synthesis-To determine the total level of cellular cholesterol in MDCK cells, cultures treated with compactin (1 M) or filipin (0.3 g/ml) for 2 days were washed 3 times in PBS and dried at room temperature. The samples were extracted with hexane/isopropanol (3:2 v/v) and dried with nitrogen. The total cholesterol levels in the samples were determined using a cholesterol determination kit (Kyowa Medical Co. Ltd., Tokyo, Japan). Protein concentration was determined using the BCA protein assay kit (Pierce), with bovine serum albumin as the standard. To determine the level of de novo cholesterol synthesis, MDCK cells were pretreated with compactin (1 M) and filipin (0.3 g/ml) in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum for 2 h. 2 Ci/ml [ 14 C]acetate was added into the cultures treated with the compounds. After 3 h of incubation, the cultures were washed three times in PBS and dried at room temperature. The samples were then extracted with hexane/isopropanol (3:2 v/v) and dried with nitrogen. The samples were quantitatively spotted on thin-layer chromatography plates and developed in a solvent system of hexane/ ethyl ether/acetic acid (80:30:1). The radioactivities of the spots were detected and quantified by the Bio-imaging Analyzer System-2500 Mac (Fuji Film Co. Ltd., Tokyo, Japan).
Mass Spectrometry-Mass spectrometric analysis was performed essentially as described elsewhere (17). Aliquots of medium (5 ml) from the apical and basolateral compartments were incubated with 4G8 (5 g) at 4°C overnight for immunoprecipitation of A␤. The mixtures were then incubated with protein G-Sepharose at 4°C for 3 h and centrifuged. The pellets were washed thoroughly with RIPA buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% SDS, 0.5% deoxycholic acid, 1.0% Nonidet P-40) once and then with distilled water four times. The immunoprecipitated A␤ was extracted with 10 l of trifluoroacetic acid. 1 l of the extracted solution was mixed with 10 l of UV-laser desorption matrix (trifluoroacetic acid/water/acetonitrile (1:20:20, v/v/v), containing saturated ␣-cyano-4-hydroxycinnamic acid; 0.5 l of this mixture was loaded onto the mass spectrometer sample probe and dried at room temperature. Mass spectra were measured using a UV-laser desorption/ ionization time-of-flight mass spectrometer (Voyager Elite, PerSeptive Biosystem).

RESULTS AND DISCUSSION
MDCK cells were cultured as described previously (6). Stable transfection of these cells with APP cDNA was performed as described elsewhere (12). To investigate whether the novel A␤ in the medium of the apical compartment of the ⌬C MDCK cell cultures, referred to as apical A␤, potentially accelerates amyloid fibril formation of synthetic A␤, we performed a thioflavin T assay as described previously (13), using the A␤ peptide immunoprecipitated from the conditioned media. As shown in Fig. 1a, when synthetic A␤ (A␤1-40) was incubated with the apical A␤ derived from the ⌬C MDCK cells, the fluorescence increased without a lag phase and proceeded to equilibrium hyperbolically. This time-course curve suggests that the apical A␤ acts as a seed in this experiment. A perfect linear semilogarithmic plot (r ϭ 0.998) shown in Fig. 1b, indicates that F(t) satisfies a differential equation: FЈ(t) ϭ B Ϫ CF(t), where B and C are constants (18). Based on this differential equation, amyloid fibril formation from synthetic A␤ incubated with the apical A␤ can be explained by a first-order kinetic model; i.e. the extention of amyloid fibrils may proceed via the consecutive association of synthetic A␤ first onto the apical A␤ then onto the ends of growing fibrils (18).
Electron microscopic analysis showed the formation of typical amyloid fibrils with a diameter of approximately 10 nm and helical structure (Fig. 1c, left panel). Similar helical filament structure was observed when 400 M of synthetic A␤ (A␤1-40) was incubated at pH 7.5, 37°C for 3 days (18). When synthetic A␤ was incubated with other types of immunoprecipitated A␤, the increase in the fluorescence was small and no fibrillar structures were observed with electron microscopic analysis (Fig. 1, a and c, right panel).
Accelerated fibrillogenesis upon addition of the apical A␤ was further confirmed using Congo red assay (data not shown). The extent of enhancement of fibrillogenesis by the apical A␤ was statistically significant (Fig. 2a). We excluded the possibility that these results were caused by alteration in the amount of A␤ secreted from the cells by determining the level of A␤ by enzyme immunoassay (Fig. 2b).
To investigate the molecular mechanism of generation of the apical A␤ with seeding ability, we first asked whether the generation is dependent on the cellular cholesterol because the level of cholesterol in the apical plasma membrane is higher than that of basolateral plasma membrane (7). When we incubated the cells with compactin or filipin, acceleration of fibrillogenesis upon addition of the apical A␤ was substantially inhibited in a dose-dependent manner as shown in Fig. 3a. Again, we excluded the possibility that these results were caused by alteration in the amount of A␤ secreted from the cells by determining the level of A␤ by enzyme immunoassay (Fig.  3b). To further confirm that the seeding ability of the apical A␤ depended on the presence of cholesterol, we performed an experiment to see if the effect of compactin and filipin was reversed by the addition of exogenous cholesterol. As shown in Fig. 3c, the inhibition of the apical A␤-induced acceleration of fibrillogenesis of synthetic A␤ following treatment with compactin and filipin, was dramatically reduced by the addition of exogenous cholesterol. Total cellular level of cholesterol was not dramatically decreased in cultures treated with compactin or altered at all in those treated with filipin, whereas the de novo cholesterol synthesis was markedly suppressed in cultures treated with compactin (Fig. 3d). These results suggest that generation of the seeding A␤ requires the presence of cholesterol in specific microdomains and does not depend on the total cellular level of cholesterol. The altered A␤ species was also generated in cultures that were grown in media not Note that acceleration of the amyloid fibril formation was observed in the mixture containing A␤ immunoprecipitated from the medium of the apical compartment of the ⌬C MDCK cells. *p Ͻ 0.01 (Student's t test). Panel b, determination of the level of A␤40 secreted into the media of the MDCK cells by EIA. EIA was performed as described previously (15). Note that the levels of A␤40 in the media were not correlated with the fluorescence intensities (panel a). containing fetal bovine serum or supplemented with lipoprotein-deprived serum (data not shown), indicating that its generation does not depend on the presence of lipoprotein(s) or other factors in the serum.
We then attempted to characterize the novel SDS-stable A␤ species. It was previously reported that A␤ can form SDSstable oligomers (19); thus, we attempted to determine its molecular mass using matrix-assisted laser desorption/ionization mass spectrometry. The molecular mass of the major peak for the apical A␤ was identical to that of human A␤6 -40 peptide (Fig. 4a), whereas that of the A␤ immunoprecipitated from the media of the basolateral compartment was identical to the molecular mass of A␤5-40 (data not shown), which is consistent with a previous report (12). Notably, the molecular mass of the apical A␤ did not change following treatment of the cells with compactin (Fig. 4b), whereas its smearing on the immunoblot was lost following compactin treatment (data not shown) as described previously (3). To exclude the possibility of a different A␤ species being extracted in the experiment of mass spectrometry because of a difference in the used extracting buffer, we performed Western blotting analysis of the immunoprecipitates extracted with trifluoroacetic acid, which was used for the mass spectrometric study, but the smear appearance persisted (data not shown). This result indicates that the acquisition of these unique molecular characteristics by the apical A␤ is not due to the association of the A␤ with other molecules, but rather because of some conformational alteration. This assumption is also supported by our previous finding that the altered immunoreactivity of this novel A␤ was restored by treatment of the cells with compactin (3). Furthermore, formic acid treatment of the apical A␤ abolished its ability to act as a seed (data not shown).
Here we report, for the first time, that a seeding A␤, which catalyzes the fibrillogenesis of soluble A␤, is endogenously generated in cell culture. A noteworthy finding in this study is that generation of the seeding A␤ depends exclusively on the presence of cholesterol as shown in Fig. 3, a and c. Determination of the intracellular site of the generation of the seeding A␤ remains to be determined; however, lipid microdomains, called rafts (20), sharing a high content of cholesterol and glycosphingolipid with caveolae or caveolae-like domains are likely to be the best candidate for the following reasons: first, the axonally sorted APP, analogous to apically sorted APP in epithelial cells, is conveyed via caveolae-like domains in neurons (21); second, localization of APP in caveolae has recently been reported (22); third, further evidence to support the generation of A␤ in the cholesterol-rich microdomains is accumulating (23,24).
At this point, it is extremely difficult to elucidate the molecular mechanism of acquisition by the apical A␤ of its unique molecular characteristics, including its seeding ability; however, it may be reasonable to assume that the apical A␤ adopts an altered conformation based on the following experimental results obtained from this and previous studies (3): first, the immunoreactivity of the apical A␤ to BAN50, in addition to its smearing behavior on gel electrophoresis, changed following treatment of the cells with compactin without any alteration in its mass number (Ref. 3 and Fig. 4); second, the BAN50 immunoreactivity for the apical A␤ recovered following treatment of the A␤ with formic acid (3). In the putative conformational alteration of the apical A␤, auxiliary factors localized in the lipid microdomains may be involved, as is suggested in the conversion of the normal cellular form of prion protein (PrP c ) to its pathogenic form (PrP sc ) (9). Among the candidates for such factors, we prefer to consider GM1 ganglioside for the following reasons: first, GM1 ganglioside is one of the main resident molecules in the microdomains (25); second, we have previously found GM1 ganglioside-bound A␤ in human brains in the early stages of Alzheimer's disease (26); third, A␤ undergoes alteration of its secondary structure via interaction with GM1 ganglioside (27,28); and fourth, it has recently been reported that amyloid fibril formation of A␤ is drastically accelerated in the presence of GM1 ganglioside (14). Thus, it is intriguing to speculate that the A␤ generated from the apically missorted APP undergoes conformational alteration via association with GM1 ganglioside in the lipid microdomains and then acts as a template for the consecutive conversion of a nascent soluble A␤ into a seeding A␤.
Recently, much attention has been focused on the conformational alteration of constitutive proteins in the brain in various neurodegenerative diseases (4,5,29). In such processes, a constitutive protein in the brain undergoes minor perturbations of structure, leading to an increase in ␤-sheet content; it has been proposed that these disease processes be grouped into one new category, the conformational diseases (29). Conformational conversion of the prion protein with resultant aggregation of its pathogenic form is a well-known example belonging to this category (30). Although Alzheimer's disease could also be included in the conformational diseases group (4,5,29), to date, no study has ever shown the generation of a conformationally altered isoform of A␤ with seeding ability. In this regard, our results present, for the first time, evidence for the generation of a seeding A␤ in cell culture, and furthermore, may be used to explain the molecular mechanism underlying initiation of amyloid fibril formation in vivo.
Finally, from the results of this and other (31) studies, one can consider the possibility that missorting of APP or altered intracellular trafficking of A␤ plays a role in the pathogenesis of Alzheimer's disease. Further studies, using polarized differentiated neurons should be carried out to investigate the consequences of generation of a seeding A␤ from axonal or presynaptic membranes. FIG. 4. , Mass spectrometry of the apical A␤. Matrix-assisted laser desorption/ionization mass spectrometry spectra of the A␤ immunoprecipitated from the medium of the apical compartment of ⌬C MDCK cell cultures without (panel a) and with (panel b) compactin treatment. Peaks observed in the spectra were labeled with protonated molecular masses (MϩH ϩ ), which corresponded to the calculated protonated molecular mass of 3712.2 Da for human A␤6 -40. We did not detect significant peaks corresponding to A␤42, probably due to its presence in low amounts. Note that the molecular mass of the A␤ obtained from the compactin-treated cells was identical to that obtained from nontreated cells.