Reduction of Brain β-Amyloid (Aβ) by Fluvastatin, a Hydroxymethylglutaryl-CoA Reductase Inhibitor, through Increase in Degradation of Amyloid Precursor Protein C-terminal Fragments (APP-CTFs) and Aβ Clearance*

Epidemiological studies suggest that statins (hydroxymethylglutaryl-CoA reductase inhibitors) could reduce the risk of Alzheimer disease. Although one possible explanation is through an effect on β-amyloid (Aβ) metabolism, its effect remains to be elucidated. Here, we explored the molecular mechanisms of how statins influence Aβ metabolism. Fluvastatin at clinical doses significantly reduced Aβ and amyloid precursor protein C-terminal fragment (APP-CTF) levels among APP metabolites in the brain of C57BL/6 mice. Chronic intracerebroventricular infusion of lysosomal inhibitors blocked these effects, indicating that up-regulation of the lysosomal degradation of endogenous APP-CTFs is involved in reduced Aβ production. Biochemical analysis suggested that this was mediated by enhanced trafficking of APP-CTFs from endosomes to lysosomes, associated with marked changes of Rab proteins, which regulate endosomal function. In primary neurons, fluvastatin enhanced the degradation of APP-CTFs through an isoprenoid-dependent mechanism. Because our previous study suggests additive effects of fluvastatin on Aβ metabolism, we examined Aβ clearance rates by using the brain efflux index method and found its increased rates at high Aβ levels from brain. As LRP1 in brain microvessels was increased, up-regulation of LRP1-mediated Aβ clearance at the blood-brain barrier might be involved. In cultured brain microvessel endothelial cells, fluvastatin increased LRP1 and the uptake of Aβ, which was blocked by LRP1 antagonists, through an isoprenoid-dependent mechanism. Overall, the present study demonstrated that fluvastatin reduced Aβ level by an isoprenoid-dependent mechanism. These results have important implications for the development of disease-modifying therapy for Alzheimer disease as well as understanding of Aβ metabolism.

Alzheimer disease (AD) 2 is a progressive neurodegenerative disease, being the most prevalent disorder among dementia. The discoveries that the genes of familial AD-linked mutation up-regulate A␤ production and the increased rate of A␤42 production is associated with the age of onset provide conclusive evidence for the amyloid hypothesis in the pathogenesis of AD (1). Several therapies based on the amyloid hypothesis are being examined, including ␥-secretase inhibitors and A␤ vaccine therapy, as disease-modifying therapy. However, there are still many unresolved issues with their clinical application (2,3). Furthermore, recent failures of clinical trials of these therapies raise questions on delayed timing of intervention and the efficacy of targeting only one pathway of A␤ metabolism (4,5). More efficient treatment with higher safety is needed to treat AD.
On the other hand, from basic and clinical reports, statins (hydroxymethylglutaryl-CoA reductase inhibitors), which are widely used for the treatment of hypercholesterolemic patients, might be beneficial in AD. Clinically, many case control studies support the protective effect of statins (6 -8), whereas the results of prospective studies and randomized clinical trials are still controversial (9 -11). Larger randomized clinical trials considering intervention timing, dose, duration, kind of statins, and main end point (i.e. prevention or cognitive decline) are required to clarify the efficacy of statins (12).
Although statins affect A␤ metabolism, their proposed mechanism of action on A␤ production is quite diverse as follows: up-regulation of ␣-secretase processing, down-regulation of ␤-secretase processing, down-regulation of ␥-secretase processing, modulation of APP trafficking, and up-regulation of APP-CTF degradation (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). However, it should be noted that these various mechanisms were demonstrated mostly by in vitro studies. In considering the in vivo effects on A␤ metabolism, several points should be clarified. Firstly, which of the two effects (cholesterol-dependent effect and isoprenoid-dependent effect; Ref. 21) affects A␤ metabolism in vivo more strongly? Secondly, because the concentration of statin might be important (17), what are the physiological levels of statins at clinically relevant doses and how do statins affect A␤ metabolism at those levels? Thirdly, we previously demonstrated that the protective effect of fluvastatin in an A␤-induced memory impairment mouse model was associated with reduced A␤ accumulation, suggesting additional effects on A␤ metabolism other than A␤ production (25). Here, the present study demonstrated that fluvastatin affected A␤ metabolism in the brain through a reduction of A␤ production and an increase in A␤ clearance via up-regulation of lysosomal degradation of APP-CTFs and an increase in LRP1 at the BBB, respectively.

EXPERIMENTAL PROCEDURES
Animals-C57BL/6 mice as well as APP23 transgenic mice were used in this study. APP23 transgenic mice overexpress human APP with Swedish double mutation (KM670/671NL) under the control of Thy-1 promoter (26). All hemizygous (ϩ/Ϫ) transgenic animals were crossed with nontransgenic background strain animals (C57BL/6) to obtain transgenic (ϩ/Ϫ) animals. Animals were housed in specific pathogen-free facilities under a standard 12/12-h light/dark cycle with free access to both food and water. All experiments were carried out in accordance with the Guidelines for the Care and Use of Laboratory Animals of Osaka University School of Medicine.
Drug Administration to Animals-Administration of fluvastatin was started at 8 weeks of age and continued for 4 weeks in all experiments, except for that of co-administration with lysosomal inhibitors. In experiments with lysosomal inhibitors, fluvastatin treatment was continued for 5 weeks. Mice received fluvastatin at 5 mg mg/kg/day added as a diet admixture (0.008%) or vehicle. This dose of fluvastatin is equivalent to the dose in clinical usage (20 mg/day) and did not affect plasma cholesterol level or markers of hepatic toxicity (data not shown).
One week of chronic administration of leupeptin or E64 (Peptide Institute Inc., Osaka Japan) into the cerebral ventricle was performed as described previously (27,28). An osmotic minipump (model 2002; ALZET, Cupertino, CA) was loaded with ACSF buffer (148 mM NaCl, 3 mM KCl, 1.4 mM CaCl 2 , 0.75 mM MgCl 2 , 0.8 mM Na 2 HPO 4 , 0.2 mM NaH 2 PO 4 ), leupeptin (20 mg/ml in ACSF), or E64 (20 mg/ml in ACSF) connected to the brain infusion assembly (brain infusion kit 3; ALZET) and incubated with sterile saline at 37°C for 48 h. Anesthetized mice were placed in a stereotaxic apparatus (Narishige, Tokyo, Japan), and a midline incision was made to expose an area of the skull. A catheter was inserted into the lateral ventricle of the brain, and the connected osmotic minipump was implanted subcutaneously in the midscapular area of the back of each animal. The coordinates for cannula placement were: anteroposterior, 0.2 mm to bregma; mediolateral, 0.8 mm to bregma; and dorsoventral, 2.5 mm to cranium. A hole was drilled in the skull, the cannula was glued to the cleaned and scraped skull with Aron Alpha (jelly type; Toagosei, Tokyo, Japan), and the incision was closed over the assembly. During the intracerebroventricular administration of leupeptin or E64, there were no significant changes in body weight, food intake, or general appearance (data not shown).
In Vitro Cell Culture-Primary cultures of cortical neurons were prepared from day 16 -18 C57BL/6 mouse embryos as described previously (29). In brief, cerebral cortices devoid of meninges were digested with trypsin/DNase I and dispersed by pipetting. Dissociated neurons were seeded onto polyethyleneimine (Sigma-Aldrich, Tokyo, Japan)-coated culture dishes at a density of 10 5 cells/cm 2 . Neurons were cultivated in Dulbecco's modified Eagle's medium with high glucose (Invitrogen, Tokyo, Japan) plus 10% fetal bovine serum and 100 g/ml penicillin/ streptomycin. The medium was changed every 2 days. Before experiments, 5 M cytosine ␤-D-arabinofuranoside (Sigma-Aldrich) was added from the 4th day for 2 days to block the increase in glia cells. Primary human brain microvascular endothelial cells (HBMEC; ScienCell, San Diego, CA) were plated on collagen type I-coated dishes (Iwaki, Tokyo, Japan) and cultured in endothelial cell basal medium-2 (EBM-2) supplemented with EGM-2 containing 2% fetal bovine serum (Lonza, Walkersville, MD) and 100 g/ml penicillin/streptomycin according to the manufacturer's instructions.
Sample Preparation-Following drug administration, anesthetized mice were transcardially perfused with saline, and the brain was removed and cut sagittally into the left and right hemispheres. After removing the olfactory lobe and cerebellum, the hemispheres were snap-frozen in liquid nitrogen and stored at Ϫ80°C. For biochemical analysis, these brains or the harvested cells were homogenized in radioimmune precipitation lysis buffer (Millipore, Billerica, CA) containing a protease inhibitor mixture (PIM; Roche Diagnostics K.K., Tokyo, Japan) with a Teflon-glass homogenizer and then sonicated on ice. The resulting supernatants (total fraction) were used for Western blot analysis. In a separate study, the microsomal fraction was prepared as described previously (25). In brief, brain samples were homogenized in tissue homogenization buffer (THB; 0.25 M sucrose, 20 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM EGTA) containing PIM. After centrifugation at 1600 ϫ g for 5 min at 4°C, the postnuclear supernatant was further centrifuged at 100,000 ϫ g for 60 min at 4°C to collect the pellet. After washing with ice-cold 100 mM Na 2 CO 3 (pH 11.3), the precipitant was resuspended in THB containing PIM, solubilized by sonication, and used as the microsomal fraction, and the supernatant was also collected as the "soluble fraction." Protein concentration was determined by the Lowry method (Bio-Rad, Tokyo, Japan) to load an equal quantity of sample in Western blot analysis, which was performed as described previously (25).
Assessment of A␤ and APP Metabolites-The microsomal fraction as described above was used to detect full-length APP (APP-FL) and APP C-terminal fragment (APP-CTFs: APP-CTF␤ and APP-CTF␣), and the soluble fraction was used to detect secreted APP (sAPPs: sAPP␣ and sAPP␤) in Western blot analysis and ELISA analysis (IBL, Gumma, Tokyo, Japan). For measurement of endogenous A␤ in the brain, the diethylamine extraction method was used (30). Briefly, the above described brain homogenate in THB was further homogenized with an equal volume of cold 0.4% diethylamine and 100 mM NaCl on ice. After centrifugation at 100,000 ϫ g for 1 h at 4°C, the supernatant was neutralized with a 1/10 volume of 0.5 M Tris base (pH 6.8) and applied to the A␤ ELISA system (Wako, Osaka, Japan).
To check our detection methods, APP metabolites in the brain of C57BL/6 mice (male, 8 weeks old) were measured after treatment with ␥and ␤-secretase inhibitors, using APP23 mice as a positive control (supplemental Fig. 1). Firstly, we measured their levels in the brain 4 h after subcutaneous administration of DAPT (Peptide Institute Inc.) at 100 mg/kg as reported previously (31). DAPT administration reduced the A␤ levels by ϳ60% (data not shown) and increased the APP-CTF levels (supplemental Fig. 1A). Secondly, we measured their levels in the brain at 3 h after intracerebroventricular administration of BACE inhibitor-IV (Calbiochem) according to previously described methods (32). BACE inhibitor-IV reduced the sAPP␤ level and increased the sAPP␣ level (supplemental Fig. 1B). These preliminary studies confirmed that the changes in APP metabolites in the brain could be tracked by the methods employed in our study. The brain microsomal fraction was also subjected to dephosphorylation with protein phosphatase (Millipore) for 4 h in a supplied buffer containing PIM and 10 M DAPT to detect dephosphorylated APP-CTFs as described previously (33).
Biochemical Fractionation-The hemispheres of the mouse forebrain were homogenized in THB buffer containing PIM. After collecting the postnuclear supernatant, the samples were centrifuged at 100,000 ϫ g for 60 min at 4°C and resuspended in suspension buffer (500 mM sucrose, 40 mM Tris-HCl, pH 7.4, 2 mM EDTA, 2 mM EGTA) followed by an additional 20 strokes in 1-ml syringes fitted with a 24-, 27-, and 30-gauge needle, sequentially. Then, the samples were mixed with 2 volumes of 60% iodixanol (Axis-Shield Plc, Dundee, Scotland, UK) and applied to the bottom of a preformed 2-30% iodixanol continuous gradient in THB containing PIM and DAPT, which was made by Gradient Master (BIOCOMP, Fredericton, NB, Canada). The gradient was centrifuged for 20 h at 100,000 ϫ g in a rotor (model SW41; Beckman Coulter, Fullerton, CA) at 4°C. Twenty fractions were collected from the top of the tubes by a piston gradient fractionator (BIOCOMP). The samples were trichloroacetic acid-precipitated and resolubilized in 2 volumes of sample buffer for Western blot analysis.
Brain Efflux Index (BEI) Study-The in vivo brain elimination experiments were performed using the intracerebral microinjection technique reported previously (34 -36). Fluvastatin-treated or untreated C57BL/6 mice were anesthetized with intraperitoneal xylazine and ketamine (20 and 100 mg/kg, respectively) and placed in a stereotaxic apparatus. Core body temperature was maintained at 37°C using a heating pad with a thermoprobe (ATC-101B; Unique Medical, Tokyo, Japan). pump (NE-1000; Neuroscience, Osaka, Japan). The intactness and quality of 125 I-A␤ were confirmed by HPLC fractionation, SDS-PAGE fluorography, trichloroacetic acid precipitation, and ELISA analysis, as reported previously (34). Synthetic A␤-(1-40) peptides (Peptide Institute Inc.) were solubilized in 1,1,1,3,3,3-hexafluoro-2-propanol (Kanto Chemical, Tokyo, Japan), dried, and resolubilized in ECF buffer. To minimize any backflow of injectate, the needle was left in place for 3 min after administration. The dose-dependent effect with or without fluvastatin treatment was determined at 60 min. At the designated time, the ipsilateral cerebrum was excised, snap-frozen in liquid nitrogen, and homogenized in THB with PIM. About half of this homogenate was solubilized in tissue solubilizer (Soluene-350; PerkinElmer Life Sciences) at 65°C and mixed with liquid scintillation mixture (LumaSafe Plus; PerkinElmer Life Sciences). The remains of the homogenate were trichloroacetic acid-precipitated to calculate the rate of intact A␤ (34). 125 I radioactivity of the samples was measured with a ␥-counter (ARC-2000; Aloka, Tokyo, Japan), and 14 C radioactivity was measured with a liquid scintillation counter (WALLAC 1409; PerkinElmer Life Sciences) using a double-channel system. BEI was defined by Equation 1, and the percentage of substrate remaining in the cerebrum (100 Ϫ BEI) was determined using Equation 2, as previously described (35). The apparent elimination rate constant, k app,el , was estimated by linear regression analysis of the semilogarithmic plot of (100 Ϫ BEI)% versus time. Kinetic parameters of A␤ clearance were calculated by using Equation 3.
k el,max , K m , and k el,NS represent the maximum elimination rate, half-saturation concentration, and nonsaturable elimination rate constant, respectively. C is the A␤ concentration in the injectate. Regression analysis was performed to estimate these parameters using JMP software (version 7; SAS Institute Inc., Cary, NC).
In Vitro Uptake Assay of A␤-The uptake of 125 I-A␤ into HMBEC was examined as reported previously (37). Briefly, HBMEC cultured onto collagen I-coated 24-well dishes were grown to 90 -100% confluence. Cells were washed three times with warmed ECF buffer and incubated at 37°C with 125 I-A␤ (0.1 nM) in ECF buffer for 20 min. Then, incubation medium was collected, and the cells were washed three times with icecold ECF buffer and an additional three times with acid wash buffer (28 mM CH 3 COONa, 120 mM NaCl, 20 mM sodium barbital, pH 3.0). Cells were solubilized with radioimmune precipitation buffer, and the protein amount in the cells was measured. The cell/medium ratio (l/mg of protein) is equal to 125 I counts in the cells (cpm/mg of protein)/ 125 I counts in the incubation medium (cpm/l).
Isolation of Brain Microvessels-Brain microvessels were isolated by previously described methods (38) with some modification. In brief, mouse brain devoid of leptomeninges was homogenized on ice in Dulbecco's modified Eagle's medium supplemented with 20 mM HEPES with a Teflon-glass homogenizer (0.25-mm clearance; Sansyo, Osaka, Japan) and resuspended with ice-cold 17% dextran (Sigma-Aldrich). Following centrifugation at 10,000 ϫ g for 15 min at 4°C, the precipitate was suspended with ice-cold culture medium (10% fetal bovine serum in Dulbecco's modified Eagle's medium) and filtered through sterilized glass beads (425-600 m, acid-washed; Sigma-Aldrich) on a 70-m nylon cell strainer (BD Biosciences, Tokyo, Japan). These glass beads were washed with the culture medium and transferred onto a plastic dish. After gentle shaking in culture medium, dissociated vessels from glass beads were collected, washed with phosphate-buffered saline, and solubilized in radioimmune precipitation buffer containing PIM. Purification of the microvessels was assessed by microscopic observation and immunoblotting using anti-CD31 (an endothelial cell marker), anti-␣-SMA (a smooth muscle cell marker), and anti-NeuN (a neuronal marker). In this evaluation, the rate of microvessels to brain parenchyma in this microvessels fraction was at least 20-fold higher when compared with that in the starting sample (data not shown).
Measurement of Fluvastatin Concentration-Fluvastatin level in each tissue was quantified using liquid chromatography-tandem mass spectrometry. In brief, animal serum or brain homogenate in saline was mixed with saturated sodium chloride and acetonitrile and centrifuged. Then, 0.1% acetic acid was added to the supernatant to prepare samples. Liquid chromatography was carried out on an HPLC apparatus (Agilent, Tokyo, Japan) equipped with an XTerra reverse phase column (RP18, 50 mm ϫ 3.0 mm, inner diameter 3.5 m; Waters, Tokyo, Japan) at a flow rate of 0.5 ml/min. The mobile phase consisted of 35% solvent A (0.1% acetic acid) and 65% solvent B (methanol). Electrospray data were acquired using a 4000 QTRAP triple-quadruple mass spectrometer (Applied Biosystems, Foster City, CA). The peak signals of transition from the parent ion to its major fragments (m/z 4103348) were measured. Quantification of fluvastatin was based on internal standardization using the peak area ratios of the analyte and the internal standard.
Antibodies and Other Reagents-The following antibodies were used: anti-APP N-terminal, anti-ADAM10/kuzbanian, anti-CD31 and anti-NeuN Anti-cathepsin-D antibody was developed as described previously (39). Chloroquine and mevalonate were purchased from Sigma-Aldrich. A human receptor-associated protein was purchased from Oxford Biomedical Research (Oxford, MI).
Statistics-All data were expressed as mean Ϯ S.E. Comparison of two groups was performed by two-tailed t test. Comparison among three or more groups was performed by analysis of variance (ANOVA) or repeated measures ANOVA followed by two-tailed t test. p values of less than 0.05 were considered significant. All statistical analyses were performed using JMP software. In Western blot experiments, the density of each band was measured by ImageJ (version 1.37). The density of protein bands was standardized to the density of loading controls and compared with those for control mice, to which 100% was assigned.

Reduction of A␤ Production by Fluvastatin in Brain through the Lysosomal Degradation of Endogenous APP-CTFs via
Enhanced Trafficking of APP-CTFs to Lysosomes-Initially, we investigated the in vivo effects of fluvastatin on A␤ production FIGURE 1. Reduction of A␤ and APP-CTF levels among APP metabolites in brain of C57BL/6 male mice by fluvastatin. A, A␤40 and A␤42 levels in the brain were determined by human/mouse A␤ ELISA after 4 weeks of treatment with or without 5 mg/kg/day fluvastatin (n ϭ 11/group). B, Western blot analysis of APP-FL and APP-CTFs (APP-CTF␣ and APP-CTF␤) in brain microsomal fraction and sAPP␣ and sAPP␤ in brain soluble fraction. Calnexin and ␤-tubulin were used as loading controls for the brain microsomal fraction and brain soluble fraction, respectively. C ϭ control; F ϭ fluvastatin. C, quantification of APP-FL, APP-CTF, sAPP␣, and sAPP␤ levels (n ϭ 6/group). *, p Ͻ 0.05, **, p Ͻ 0.01, N.S. ϭ not significant; two-tailed Student's t test.
in the brain of C57BL/6 mice. Fluvastatin treatment (5 mg/kg/ day) significantly reduced both A␤40 and A␤42 levels in the brain (p Ͻ 0.01; Fig. 1A). Higher dose fluvastatin treatment (10 and 20 mg/kg/day) also reduced A␤ level in the brain to a similar degree (supplemental Fig. 2A), although neither 5 mg/kg/ day nor 20 mg/kg/day fluvastatin treatment affect brain cholesterol level (supplemental Fig. 2B). To assess whether fluvastatin would affect ␣-, ␤-, or ␥-secretase cleavage of APP, we measured the levels of other APP metabolites (APP-FL, APP-CTF␣, APP-CTF␤, sAPP␣, and sAPP␤) in the brain. As shown in Fig. 1, B and C, fluvastatin treatment (5 mg/kg/day) significantly reduced both APP-CTF␣ and APP-CTF␤ (p Ͻ 0.01). Because these protein bands of APP-CTFs (APP-CTF␣ and APP-CTF␤) included some phosphorylated products (33), we quantified APP-CTF levels after dephosphorylation and found a reduction in the levels of C99, C89, and C83 (data not shown). In contrast, fluvastatin did not affect the levels of sAPP␣ and sAPP␤ (Fig. 1, B and C) and also failed to change the levels and activity of ␣-secretase (ADAM10 and ADAM17) and ␤-secretase (BACE1) in the brain (supplemental Fig. 3, A and B). These results suggest that the reduction of A␤ production by fluvastatin is independent of regulation of APP processing by ␣and ␤-secretase.
One possible mechanism of the reduction of A␤ production is through the enhancement of lysosomal degradation of APP-CTFs in the brain. To test this hypothesis, we performed intracerebroventricular infusion of lysosomal inhibitors ( Fig. 2A). Leupeptin administration did not affect sAPP␣ and sAPP␤ levels, indicating that ␣and ␤-secretase cleavages of APP were unaffected (data not shown). Leupeptin increased APP-CTF and A␤ levels in the brain, suggesting that the lysosomal degradation of APP-CTFs is successfully suppressed (Fig. 2, B and D). Moreover, leupeptin blocked fluvastatin-mediated reductions of APP-CTF and A␤ levels to the levels of leupeptin onlytreated mice (Fig. 2, B and D). Similarly, treatment with another lysosomal inhibitor, E64, also blocked fluvastatin-mediated reductions of APP-CTF and A␤ levels to the levels of E64 only-treated mice (Fig. 2, C and E) without affecting sAPP␣ and sAPP␤ levels (data not shown). These results indicate that fluvastatin reduces A␤ production in the brain through up-regulation of the lysosomal degradation of APP-CTFs. Meanwhile, fluvastatin did not affect the levels as well as the activity of lysosomal proteinases (cathepsin-B, -D, and -L; supplemental Fig. 4) and the level of Lamp1, a lysosomal receptor protein (supplemental Fig. 4A). This suggests that an increase in the lysosomal degradation of APP-CTFs is not due to up-regulation of lysosomal activity.
Next, we examined the hypothesis that fluvastatin enhances the trafficking of APP-CTFs to lysosomes. It is noteworthy that fluvastatin treatment shifted the subcellular distribution of APP-CTFs to heavier fractions (Fig. 3B), which showed considerable overlap with cathepsin-D, a lysosome marker (Fig. 3C), and partial overlap with Rab7, a late endosome marker (Fig. 3E), as assessed by iodixanol gra-

FIGURE 2. Effects of lysosomal inhibitors on reduction of APP-CTF and A␤ levels by fluvastatin in brain.
A, after mice were treated with or without 5 mg/kg/day fluvastatin for 4 weeks, leupeptin or E64 was administered for 7 days into the cerebral ventricle via an osmotic minipump. Fluvastatin treatment was continued until the end of experiment. B, Western blot analysis of APP-FL and APP-CTFs in brain of each mouse group (control ϩ ACSF, fluvastatin ϩ ACSF, control ϩ leupeptin, and fluvastatin ϩ leupeptin). Graph represents quantification of APP-CTF␤ level. Calnexin was used as a loading control. C, Western blot analysis of APP-FL and APP-CTFs in each group (control ϩ ACSF, fluvastatin ϩ ACSF, control ϩ E64, and fluvastatin ϩ E64). Graph represents quantification of APP-CTF␤ level. Calnexin was used as a loading control. expo., exposure. D, A␤42 level in each group (control ϩ ACSF, fluvastatin ϩ ACSF, control ϩ leupeptin, and fluvastatin ϩ leupeptin). E, A␤42 level in each group (control ϩ ACSF, fluvastatin ϩ ACSF, control ϩ E64, and fluvastatin ϩ E64). ACSF was used as vehicle for the osmotic minipump. A␤42 level was determined by human/mouse A␤42 ELISA. *, p Ͻ 0.05, **, p Ͻ 0.01, N.S. ϭ not significant; ANOVA followed by two-tailed Student's t test. n ϭ 4/group. JULY 16, 2010 • VOLUME 285 • NUMBER 29 JOURNAL OF BIOLOGICAL CHEMISTRY 22095 dient centrifugation. Meanwhile, fluvastatin did not affect the subcellular distribution of APP-FL (Fig. 3A). Rab7 localizes to late endosomes and controls the late step of endosomal traffic to lysosomes. It acts downstream of Rab5, a regulator of membrane traffic in the early endocytic pathway (40). Then, we examined the subcellular distribution of several organelle makers. Fluvastatin shifted the subcellular distribution of Rab7 to heavier fractions (Fig. 3E), whereas that of Rab5a, an early endosome marker, remained unaffected (Fig. 3D). Fluvastatin did not change the subcellular distribution of cathepsin-D (Fig. 3C) and calnexin, an endoplasmic reticulum marker (Fig. 3F). To evaluate the effect on other proteins, we measured the distribution of amyloid precursor-like protein 1 (APLP1). Indeed, fluvastatin changed the level and subcellular distribution of APLP1-CTFs ( Fig.  3H and supplemental Fig. 5) without affecting those of APLP1-FL ( Fig. 3G and supplemental Fig. 5). Statins inhibited the modification of isoprenylation of small GTPases, such as Ras, Rho, and Rab, leading to reduced levels of those proteins in the microsomal fraction (41). We examined levels of these protein family members. Among several Rab and Rho proteins, fluvastatin treatment particularly reduced the levels of Rab5a and Rab5b in the brain microsomal fraction ( Fig. 4 and supplemental Fig. 6). These data suggest that fluvastatin might increase the trafficking of APP-CTFs from early endosomes to late endosomes and lysosomes, associated with the change in Rab7 subcellular distribution and the reduction of endosomal Rab5 levels.

Dual Effects of Statin on A␤ Metabolism
To investigate the mechanisms of the reduction in APP-CTFs by fluvastatin, we employed primary neurons from C57BL/6 mouse embryos. As shown in Fig. 5, A and  B, fluvastatin significantly reduced APP-CTF levels in primary neurons, without affecting APP-FL levels. In contrast, in vitro experiments demonstrated that treatment with chloroquine, another lysosomal inhibitor, significantly inhibited the reduction in APP-CTFs, consistent with in vivo experiments (Fig. 5, C and D). The addition of mevalonate inhibited the reduction in APP-CTF levels by fluvastatin (Fig. 5, E and F), indicating that up-regulation of APP-CTF degradation by fluvastatin is mediated by the isoprenoid pathway. The doses of fluvastatin (1 and 3 nM) used in this study were equivalent to the level of fluvastatin in the brain of fluvastatin-treated mice ( Table 1).
Enhancement of Brain A␤ Clearance by Fluvastatin through Increase in LRP1 at BBB-As our previous study showed that fluvastatin reduced A␤ accumulation in the brain in an A␤ intracerebroventricular injection model (25), we next focused on A␤ clearance from the brain in C57BL/6 mice using the BEI method. The estimated half-life of intact 125 I-A␤ and [ 14 C]inulin in the brain was 43.6 and 187.4 min, respectively (Fig. 6A). A␤ clearance was composed of saturable and nonsaturable components (Fig. 6B), indicating an active process involving A␤ clearance. Indeed, fluvastatin significantly increased A␤ clearance rate at an injected concentration of 10 M (Fig. 6C), whereas it failed to affect A␤ clearance rate at 20 nM and 1 M. As the injected drug was diluted at least 30-fold by diffusion immediately after administration and spread farther (42) and the soluble A␤ level in the brain of AD patients is near or over 100 nM (43)(44)(45), our present study might be relevant in the pathological state of AD rather than in the normal physiological state (physiolog-  . Levels of small GTPases in microsomal and soluble fractions in brain with or without fluvastatin. Western blot shows levels of Rab small GTPases (Rab4, Rab5a, Rab5b, Rab7) in brain microsomal fraction (A) and soluble fraction (C). C ϭ control; F ϭ fluvastatin. B, quantification of Rab protein family levels in brain microsomal fraction. Calnexin was a loading control for the microsomal fraction. D, quantification of Rab protein family levels in brain soluble fraction. ␤-tubulin was a loading control for the soluble fraction. *, p Ͻ 0.05, N.S. ϭ not significant; two-tailed Student's t test. n ϭ 6/group. JULY 16, 2010 • VOLUME 285 • NUMBER 29

JOURNAL OF BIOLOGICAL CHEMISTRY 22097
ical A␤ level is about 1 nM). Because we previously showed that fluvastatin did not affect the activity and level of major A␤-degrading enzymes such as neprilysin and insulin-degrading enzyme in the brain (25), we focused on the levels of apolipoprotein E, P-glycoprotein and the low density lipoproteinrelated protein 1 (LRP1), which are reported to be involved in A␤ clearance from the brain (46). Levels of ApoE as well as P-glycoprotein in the soluble and membrane fractions of the brain and brain microvessels were not affected by fluvastatin (Fig. 6, D and E). However, LRP1 was significantly increased by fluvastatin in brain microvessels (Fig. 6, D and E), which might contribute to the increase in A␤ clearance.
To evaluate the effect of fluvastatin, we employed HBMEC. As shown in Fig. 7A, fluvastatin significantly increased the expression of LRP1 in HBMEC in a dose-dependent manner (p Ͻ 0.01), whereas cotreatment with mevalonate attenuated it (Fig. 7B), indicating that up-regulation of LRP1 by fluvastatin is due to an isoprenoid-dependent mechanism. To analyze the functional properties of LRP1, the initial internalization step of A␤ clearance at the BBB was examined using these cells as described previously (37). Expectedly, fluvastatin upregulated A␤ uptake, whereas the addition of mevalonate attenuated it (Fig. 7C). Moreover, receptorassociated protein as well as anti-LRP1 antibody, which inhibit LRP1 function, blocked the effect of fluvastatin (Fig. 7, D and E), indicating that the increased uptake of A␤ by fluvastatin is mediated by LRP1. In this experiment, we selected the doses (10 -100 nM) to achieve a serum concentration of fluvastatin close to that in treated mice (Table 1) because LRP1 located at the BBB is influenced by the serum concentration.
Finally, we examined the effect of fluvastatin in APP transgenic mice (APP23). Fluvastatin treatment reduced brain A␤ levels (Fig. 8A). Moreover, we confirmed that fluvastatin also increased brain microvessel LRP1 level (p Ͻ 0.01; Fig. 8B). These results suggest that up-regulation of LRP1-mediated clearance at the BBB could be involved in the regulation of A␤ metabolism in this AD mouse model.

TABLE 1 Drug concentration in tissues following fluvastatin administration
Fluvastatin levels in serum and brain of C57BL/6 mice were determined by liquid chromatography-tandem mass spectrometry after 4 weeks of treatment with or without 5 mg/kg/day fluvastatin (mean Ϯ S.E.).

Treatment
Fluvastatin level in serum (n ‫؍‬ 5) Fluvastatin level in brain (n ‫؍‬ 4) statin and pravastatin increased the sAPP␣ level in the brain of young TgCRND8 mice. They proposed up-regulation of ␣-secretase processing, resulting in reduced A␤ production, whereas Burns et al. (20) reported that simvastatin, lovastatin, and atorvastatin reduced the level of APP-CTF␤ among A␤ precursors (including sAPP␣) in the brain of wild-type mice and proposed a reduction of ␤-secretase processing. Several reasons for this inconsistency can be speculated. Variations in APP (wild type, Swedish, Indiana) or its level might shift the locus of APP processing and affect the effects of statins on APP metabolism (50), producing different results among wild-type mice and APP23 and TgCRND8 mice. In previous studies, which focused on the effect of statins on cholesterol synthesis, high doses of statins were administered mostly to reduce the serum cholesterol level (20,47,48,51). In addition, as pleiotropic effects of statins are different, the results could be explained by the kind of statin used. Especially, fluvastatin has a potent antioxidant action and more potent pleiotropic effects on blood vessel walls and several tissues (52). Indeed, our previous study demonstrated a different preventive effect of fluvasta-tin from that of other statins in an A␤-induced cognitive impairment model (25). Our study demonstrated that fluvastatin up-regulated lysosomal degradation of endogenous APP-CTFs in brain. Previously, Ostrowski et al. (17) reported up-regulation of lysosomal degradation of APP-CTFs by statins via an isoprenoid-dependent mechanism in H4 neuroglioma cells expressing wild-type APP or APP-CTF␤. They proposed that Rho proteins might be involved in the degradation of APP-CTFs by statins. In this study, we failed to detect any change in the levels of Rho family proteins (RhoA, Rac1/2/3 and Cdc2) in the brain microsomal fraction and RhoA-GTP in the brain (supplemental Fig. 6). Further analyses are required to ascertain the involvement of other Rho members such as RhoB and RhoC.
On the other hand, we found a marked shift in the intracellular distribution of APP-CTFs to heavier fractions. This result suggests that the up-regulation of lysosomal degradation of APP-CTFs is mediated by enhanced trafficking of APP-CTFs from early endosomes to late endosomes and subsequently lysosomes (Fig. 9A). Furthermore, the present FIGURE 6. Effects of fluvastatin on brain A␤ clearance and A␤ clearance-related proteins. A, time course of 125 I-A␤ and [ 14 C]inulin elimination rate from mouse brain by BEI method (n ϭ 3/each point). B, concentration-dependent inhibition by unlabeled A␤-(1-40) (cold-A␤) of 125 I-A␤ elimination rate from mice brain (n ϭ 5/each point). Kinetic parameters of the concentration elimination rate constant curve were calculated as k el,max ϭ 12.9 ϫ 10 Ϫ3 (M/min), K m ϭ 0.97 (M), and k el,NS ϭ 6.7 ϫ 10 Ϫ3 (min Ϫ1 ). C, effects of fluvastatin treatment (5 mg/kg/day) on A␤ elimination rate with vehicle or with indicated concentrations of cold-A␤ (n ϭ 5/group). D, Western blot analysis of ApoE in brain microsomal and soluble fractions, P-glycoprotein, and LRP1 in brain microvessels of C57BL/6 mice after fluvastatin treatment. Calnexin was a loading control for the brain microsomal fraction. ␤-Actin was a loading control for the brain soluble fraction and brain microvessels. C ϭ control; F ϭ fluvastatin; LRP1-LC ϭ LRP1-light chain. E, quantification of level of apolipoprotein E in brain microsomal and soluble fractions and P-glycoprotein and LRP1 in brain microvessels of C57BL/6 mice (n ϭ 5/group). APOE, apolipoprotein E; P-gp, P-glycoprotein. *, p Ͻ 0.05, **, p Ͻ 0.01, N.S. ϭ not significant; two-tailed Student's t test.
study documented a shifted subcellular distribution of Rab7 and reduced levels in Rab5. These changes are noteworthy because several groups propose that abnormal endosomal function is involved in the pathogenesis of AD (53). In the diseased state, APP is thought to be stacked in endosomes, where it is cleaved sequentially by ␤and ␥-secretase, leading to A␤ generation and deposition (50). The increase of lysosomal degradation of APP-CTFs by fluvastatin might be mediated via modification of the endosomal-lysosomal pathway. Indeed, we found that other proteins (i.e. APLP1) were similarly affected by fluvastatin. Complete suppression or strong activation of Rab5 function through overexpression of several Rab5 constructs causes a marked change in early endosome dynamics and APP metabolism (54 -56). However, it remains to be determined how a partial reduction of Rab5 affects them. Recent studies showed that Rab5 and Rab7 act sequentially and cooperatively in the endosomal-lysosomal pathway (54,57). Therefore, moderate change in Rab5 function through the inhibition of isoprenoid pathways by low levels of statin might be involved in shifting the subcel- . Graph represents quantification of LRP1-LC level (n ϭ 8/group). ␤-Actin was used as a loading control. C, the uptake assay of 125 I-A␤ into HBMEC after treatment with or without fluvastatin (100 nM) in the presence or absence of mevalonate (200 M) for 20 h (n ϭ 6/group). D, the uptake assay of 125 I-A␤ into HBMEC in the absence or presence of 400 nM receptor-associated protein (RAP) after treatment with fluvastatin (100 nM) for 20 h (n ϭ 6/group). E, the uptake assay of 125 I-A␤ into HBMEC in the presence of anti-LRP1 N20 antibody or control IgG (each 160 g/ml) after treatment with fluvastatin (100 nM) for 20 h (n ϭ 9/group). *, p Ͻ 0.05, **, p Ͻ 0.01, N.S. ϭ not significant; ANOVA with two-tailed Student's t test. FIGURE 8. Effects of fluvastatin on A␤ levels and LRP1 levels in APP23 mice. A, A␤40 and A␤42 levels in the brain were determined by human/mouse A␤ ELISA after 4 weeks of treatment with or without 5 mg/kg/day fluvastatin (n ϭ 8/group). B, Western blot analysis of LRP1 in brain microvessels of APP23 mice with or without 5 mg/kg/day fluvastatin. ␤-Actin was used as a loading control. Graph represents quantification of level of LRP1 (n ϭ 5/group). C ϭ control; F ϭ fluvastatin; LRP1-LC ϭ LRP1-light chain. *, p Ͻ 0.05, **, p Ͻ 0.01, two-tailed Student's t test. lular distribution of Rab7, resulting in the degradation of APP-CTFs (and APLP1-CTFs). Further studies are needed to clarify this hypothesis.
Our study also showed that fluvastatin enhanced A␤ clearance from the brain at high A␤ levels. In addition, LRP1 level, which is involved in A␤ clearance, was increased at the BBB of wild-type mice and also APP23 mice. Moreover, the in vitro model of A␤ clearance also supports that fluvastatin increases LRP1-mediated A␤ clearance at BBB. LRP1, a member of the low-density lipoprotein receptor family, is thought to be involved in A␤ clearance through an efflux process across the BBB (46) and/or a degradation process (58). The involvement of LRP1 in A␤ clearance was reported by several groups (37, 58 -60). However, it remains to be resolved what proportions of LRP1-mediated clearance contribute to the total A␤ clearance in the normal brain and in the AD or the pre-AD brain. Up-regulation of LRP1 at the BBB was associated with an increase in A␤ clearance at high A␤ levels and also with the reduction of APP23 A␤ levels. These results imply the potential effect of fluvastatin in the pathological state of AD (Fig. 9B) rather than in the normal physiological state. Although the present study did not sufficiently address details of how fluvastatin increases LRP1mediated A␤ clearance, as it is known that statins promote ligand uptake by the low density lipoprotein receptor through an increase in the receptor level (61), similar mechanisms might exist. In the present study, we believe that the local concentration of statins might be important in understanding the therapeutic value of statins as the levels of flu-vastatin are quite different between brain and serum (1 nM in the brain versus 100 nM at the BBB; Table 1 and Fig. 9).
Overall, the present study demonstrated that fluvastatin reduced the A␤ level in vivo. This phenomenon might be due to up-regulation of the lysosomal degradation of endogenous wild-type APP-CTFs, a shift in the intracellular distribution of APP-CTFs to heavier fractions and changes of Rab proteins, leading to a reduction in A␤ production, and an increase in LRP1 in brain microvessels, leading to an increase in A␤ clearance from the brain at high A␤ levels. Our in vitro studies indicated that these effects are mediated by an isoprenoid-dependent mechanism. A recent study suggested that an increase in isoprenoid levels might be involved in the pathogenesis of AD (62). Additionally, other pleiotropic effects such as altered levels of phospho-Akt and endothelial nitric oxide synthase might be associated with improved learning and memory (49). The present results would have important implications in the development of disease-modifying therapy for AD as well as in basic understanding of A␤ metabolism, and further studies are needed to clarify the therapeutic mechanisms of statins in AD.