A Noncompetitive BACE1 Inhibitor TAK-070 Ameliorates Aβ Pathology and Behavioral Deficits in a Mouse Model of Alzheimer's Disease

We discovered a nonpeptidic compound, TAK-070, that inhibited BACE1, a rate-limiting protease for the generation of Aβ peptides that are considered causative for Alzheimer's disease (AD), in a noncompetitive manner. TAK-070 bound to full-length BACE1, but not to truncated BACE1 lacking the transmembrane domain. Short-term oral administration of TAK-070 decreased the brain levels of soluble Aβ, increased that of neurotrophic sAPPα by ∼20%, and normalized the behavioral impairments in cognitive tests in Tg2576 mice, an APP transgenic mouse model of AD. Six-month chronic treatment decreased cerebral Aβ deposition by ∼60%, preserving the pharmacological efficacy on soluble Aβ and sAPPα levels. These results support the feasibility of BACE1 inhibition with a noncompetitive inhibitor as disease-modifying as well as symptomatic therapy for AD.


Introduction
The accumulation of amyloid-␤ peptides (A␤) in the brain is strongly implicated in the pathogenesis of Alzheimer's disease (AD), and considered as a prime target for the disease-modifying therapy of AD (Selkoe and Schenk, 2003). A␤ is proteolytically produced through sequential cleavages by ␤and ␥-secretases from amyloid precursor protein (APP). The ␤-secretase cleavage of APP is executed by a membrane-bound aspartic protease, ␤-site APP-cleaving enzyme 1 (BACE1), which is considered to be the rate-limiting step in the production of A␤ (Cole and Vassar, 2008), whereas a majority of APP is cleaved by ␣-secretase at the midportion of A␤ sequence in a way to preclude A␤ production, by competing with BACE1.
␥-Secretase generates the C termini of A␤ with different length, e.g., A␤ 40 or A␤ 42 , the latter being considered as the pathogenic species (Iwatsubo et al., 1994). Inhibition of ␥-secretase may potentially cause side effects, because genetic knock-out (KO) of presenilin 1 and 2, the catalytic subunits of ␥-secretase, leads to embryonic lethality due to failure in activation of Notch, which is essential for development and differentiation (Shen et al., 1997;Wong et al., 1997;Donoviel et al., 1999). Furthermore, cognitive deficits associated with synaptic degeneration have been documented in PS1/PS2 conditional KO mice with or without APP transgenic background (Saura et al., 2004(Saura et al., , 2005Chen et al., 2008). In contrast, BACE1 KO mice do not show such fatal phenotypes despite its complete ablation, except for partial hypomyelination at the developmental stage (Hu et al., 2006;Sankaranarayanan et al., 2008) or schizophrenialike behavior in homozygous BACE1 KO mice (Savonenko et al., 2008), whereas cognitive deficits are ameliorated on APP transgenic background (Ohno et al., 2004(Ohno et al., , 2006(Ohno et al., , 2007. Furthermore, it has been well documented that the protein levels or activities of BACE1 are upregulated in the brains of patients with sporadic AD (Stockley and O'Neill, 2007). Therefore, BACE1 is considered as a promising target for the mechanism-based therapy for AD. So far, several BACE1 inhibitors have been reported (Hussain et al., 2007;Sankaranarayanan et al., 2009;Silvestri, 2009), although no compound that is orally active and highly penetrable to brain tissues with functional ameliorations has been documented.
We conducted a cell-based assay in the IMR32 human neuroblastoma cell line for small chemical compounds that reduce the secretion of A␤ and increase that of sAPP␣, the latter being recognized as neurotrophic with ameliorative effects on cognitive behaviors (Isacson et al., 2002;Postina, 2008). Finally we discovered a nonpeptidic compound, (R)-6-[(1,1Ј-biphenyl)-4-ylmethoxy]-1,2,3,4-tetrahydro-N,N-dimethyl-2-naphthalene-ethan-amine hydrochloride monohydrate (TAK-070) (Fig. 1), as a novel noncompetitive BACE1 inhibitor. TAK-070 ameliorated A␤ pathology and behavioral deficits in Tg2576, an APP transgenic model mice of AD, although the reduction in A␤ levels was modest, unlike those observed by complete ablation of BACE1. We propose that the partial reduction in A␤ as well as increase in sAPP␣ by a noncompetitive BACE1 inhibition may be sufficient to modify amyloid pathology and ameliorate cognitive deficits, without causing potential adverse events by complete BACE1 ablation.

Compound
The chemical TAK-070 was made by Takeda Pharmaceutical Company Limited (Takeda), and the chemical structure is shown in Figure 1. The chemical synthesis and related information are described in the patent of JP-A 11-80098 (WO98/38156 ). ␥-Secretase inhibitor IX (DAPT) was purchased from Calbiochem.

Cell cultures and sample preparation
IMR32 human neuroblastoma cell line was obtained from American Type Culture Collection (ATCC), and mouse Neuro-2a neuroblastoma cells stably expressing human Swedish mutant APP (N2aAPPsw cells) were generated as described previously (Tomita et al., 2002). For ELISA analysis, cells were cultured on 48-well multi-plates at 5 ϫ 10 4 cells/cm 2 to reach near total confluence in DMEM (Nikken Biomedical Laboratory) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) (Wako) in a humid atmosphere containing 10% CO 2 . The culture medium was replaced with DMEM/0.2% bovine serum albumin (BSA) (Wako) containing various concentrations of TAK-070, and the cells were cultured for 24 h. The conditioned media were subjected to ELISA quantitation.

Quantitation of sAPP␣ and A␤ by ELISA
To quantitate human sAPP␣, we used LN27 that recognizes the N-terminal portion of APP (Zymed) as a capture antibody. ELISA plates (high binding, clear plate, Greiner) were filled with LN27 (0.5 g/ml, 75 l/well) in carbonated buffer (100 mmol/L, pH 9.6) and incubated at 4°C overnight. After washing the plates with PBS (Invitrogen) three times, each well was blocked with 100 l of BlockAce solution (Dai-Nippon) diluted fourfold (v/v) for Ͼ2 h. After washing the plates with PBS twice, 50 l samples or standards prepared from conditioned media containing sAPP␣ were mixed with 50 l of buffer A (20 mmol/L phosphate buffer, pH 7.2, 10% BlockAce, 0.2% protease-free BSA, 0.05% thimerosal, 0.4 mol/L NaCl, 0.076% CHAPS, 2 mmol/L EDTA-2Na, 0.2% SDS, and 4 mmol/L DTT) in each well. Buffer A contains DTT to break the S-S bond of sAPP␣ to enhance the recognition by the LN27 antibody. The mixture was incubated in the plate overnight at 4°C. After washing the plates with PBS four times, BAN50-HRP (75 l/well) [which recognizes the C-terminal portion of human sAPP␣ (Asami-Odaka et al., 1995)] diluted in the detection buffer (20 mmol/L phosphate buffer, pH 7.2, 1% protease-free BSA, 2 mmol/L EDTA-2Na, 0.05% thimerosal, and 0.4 mol/L NaCl) was added to each well. The plates were incubated at room temperature for 3-4 h. After washing the plates with PBS six times, substrates were added and the reaction mixtures were developed. To measure sAPP␣ in brain lysates, the homogenate buffer free of detergents was used to preclude contamination of membrane-associated APP.
A␤ 40 or A␤ 42 was quantitated by two-site sandwich ELISA using a capture antibody BNT77, which recognizes the midportion of A␤ without detecting A␤ 17-40/42 (i.e., the cleaved products by ␣and ␥-secretases) (Fukumoto et al., 1999), and the detector antibodies of BA27-HRP or BC05-HRP that specifically detect the C termini of A␤ 40 or A␤ 42 , respectively, as described previously (Asami-Odaka et al., 1995). TMB substrate (Pierce) was used as a chromogenic substrate. After stopping the reaction with phosphoric acid solution (1 mol/L, 75 l/well), the enzymatic products were measured using a multi-label counter at OD450 (WALLAC Arvo Sx; PerkinElmer Life Sciences).

Cell-based assay for ␣-secretase activity
The assay (Doedens et al., 2003) was performed with a slight modification. N2aAPPsw cells were cultured in DMEM supplemented with 10% FCS until grown to confluence. The cells were collected by PBS (Ϫ) (Ca 2ϩ , Mg 2ϩ free) buffer and centrifuged for 5 min at 300 ϫ g. After washing with PBS (Ϫ), the cells were suspended in PBS (Ϫ) at a final concentration of 4 ϫ 10 7 cells/ml. The enzymatic reaction was initiated by combining an equal volume (100 l) of cell suspension and reaction mixture at a final cell concentration of 2 ϫ 10 7 cells/ml, 10 mol/L each of leupeptin (Peptide Institute), aprotinin (Roche Diagnostics), and ␣-secretase fluorogenic substrate [MCA-HQKLVFFA (K-DNP), Bio-Source], with vehicle of DMSO, TAK-070 (final concentration: 3 mol/L), or (Ϫ)-epigallocatechin-3-gallate (catechin, Wako) (final concentration: 20 mol/L). After each incubation time point, the cells were centrifuged, the cell-free supernatants of each 100 l were added to a 96-well black plate (Greiner), and fluorescence intensity after cleavage by ␣-secretase was measured (excitation 320 nm, emission 400 nm) (WALLAC Arvo Sx; PerkinElmer Life Sciences).

Cell-free assay for BACE1 activity
A statine substrate analog inhibitor PI (TEEISEVNXVAEF; X ϭ statine) (Sinha et al., 1999) and the fluorogenic substrate for BACE1 [Nma-SEVKMDAEK(Dnp)RR-NH 2 ] were purchased from the Peptide Institute. The substrate was dissolved in 125 mmol/L acetic acid. TAK-070 and PI were dissolved in dimethylformamide (DMF). Assays were performed in black 96-well microplates (Greiner) in a final volume of 50 l. Each well contained 25 l of acetate buffer (pH 5.5, 50 mmol/L), 10 l of recombinant BACE1, 10 l of substrate (250 mol/L), and 5 l of various concentrations of compounds at a final DMF concentration of 0.5%. The assay mixtures were incubated at 37°C for 20 h. After incubation, the fluorescence of the enzymatic product was measured at 460 nm (excitation at 325 nm) using Fluoroskan Ascent (Labsystems). The percentage of inhibition was calculated by an equation of 100 ϫ [1 Ϫ (test Ϫ blank)/ (control Ϫ blank)], where test, control, and blank are fluorescence intensities in the presence of a compound, absence of a compound, and absence of both the BACE1 enzyme and a compound, respectively. IC 35 values were calculated by linear regression analysis using a BSAS program. To clarify the inhibition profile, double-reciprocal (Lineweaver-Burk) plot analysis was performed using 10 l substrate of 100, 150, 250, 500, or 1000 mol/L (a final concentration of 20, 30, 50, 100, or 200 mol/L, respectively) and 5 l of TAK-070 of 100 or 300 mol/L (a final concentration of 10 or 30 mol/L) in total assay solution of 50 l. The reciprocal of change in the fluorescence value in the presence of TAK-070 at each concentration was plotted on the vertical axis, and the reciprocal of the substrate concentration was plotted on the longitudinal axis.

Animals
All animals were housed in rooms maintained at 24°C with a 12 h light/ dark cycle. Food (chow containing TAK-070; Oriental Yeast) and tap water were provided ad libitum. In each experiment, mice were randomly grouped, avoiding differences in body weight among groups. All experiments using animals were reviewed and approved by the Internal Animal Care and Use Committee of Takeda Pharmaceutical Research Laboratories.

Long-term treatment of Tg2576 by TAK-070
Male and female Tg2576 mice at 7 months of age (n ϭ 16 -17 for each group, n ϭ 8 -9, male; n ϭ 8, female) were used for long-term treatment with TAK-070. Tg2576 mice were fed chow containing TAK-070 (56 ppm, corresponding to ϳ7 mg/kg/d, p.o., when evaluated at 6 months of treatment) for 6 months and a week from 7 months of age, or chow without TAK-070 (vehicle control). Male Tg2576 mice at 8 months of age (n ϭ 9) were used as a young control. After decapitation, the brains were removed and the left cerebral hemisphere was immediately frozen on dry ice and stored at Ϫ80°C until biochemical assays; the right hemisphere was fixed in 4% paraformaldehyde for 24 h, embedded in paraffin, and subjected to immunohistochemical analysis. Biochemical quantitation of A␤ and sAPP␣ was performed as follows: the cerebral cortex was initially homogenized with ice-cold Tris-extraction buffer and centrifuged as in the short-term treatment study to obtain the supernatants for quantitation of soluble A␤ and sAPP␣. The pellet was then homogenized in a 19-fold volume of ice-cold 70% formic acid, and centrifuged at 44,000 ϫ g for 5 min. The supernatant was further diluted, neutralized with 1 mol/L Tris-based solution, and the levels of insoluble A␤ 40 and A␤ 42 were quantitated by ELISA.

Immunohistochemistry
Immunohistopathological analysis was performed on two distinct coronal sections from the right hemisphere at the level of the hippocampus and thalamus of Tg2576 mice. Sample preparation and quantitation of A␤ plaques were conducted under blinded conditions for the examiner. Four-micrometer-thick sections were deparaffinized and pretreated with 99% formic acid for 5 min. The section was blocked with 10% fetal calf serum for 30 min and then reacted with BAN50 (0.5 g/ml) at 4°C overnight. BAN50-positive plaques were visualized with Dako REAL En-Vision Detection Kit (Dako) using diaminobenzidine as a chromogen. The amyloid burden with a diameter more than ϳ30 m (percentage of immunopositive areas that comprised the total area) and the number of plaques throughout the right cerebral neocortices were quantitated using Vanox (AH-2, Olympus) connected to a digital video camera (Prog Res 3012, Carl Zeiss) and image analysis software (Win ROOF, Mitani).

Y-maze and Morris water maze tests
Male Tg2576 mice of 18 weeks of age were divided into three groups, i.e., vehicle-treated (n ϭ 14), TAK-070 1 mg/kg treated (n ϭ 14), and TAK-070 3 mg/kg treated (n ϭ 14). Wild-type littermates (n ϭ 15) were used as a nontransgenic control group. Tg2576 mice were treated with TAK-070 (1 or 3 mg/kg, p.o.) or vehicle (0.5% methylcellulose; MC) once a day for 9 d before the behavioral test. Each mouse was treated with drugs after all trials were completed every day during the test period. Each mouse was sequentially subjected to Y-maze test on day 10, and then in Morris water maze test from day 11 to day 13. On day 14, the mice were decapitated. The brains were dissected out on ice immediately and stored at Ϫ80°C.
Y-maze test. To measure spontaneous alternation behavior and exploratory activity, a black Y-maze with arms of 40 cm length, 3 cm width, with 12.5 cm walls was used. Each animal underwent one trial, during which the animal was placed into one of the three alleys and allowed free exploration of the maze for 5 min, and alternations and total numbers of arm choices were recorded. Spontaneous alternation, expressed as a percentage, refers to ratio of arm choices differing from the previous two choices to the total number of arm entries.
Morris water maze test. The water maze pool comprised a circular plastic water tank, 120 cm in diameter and 20 cm in depth. The pool was filled with water at room temperature to a height of 15 cm. A transparent acrylic platform (10 ϫ 10 cm), its top surface being 0.5 cm below the surface of water, was located in a constant position in the middle of one quadrant from the center and edge of the pool, and was invisible for mice inside the pool. Each mouse was given four trials daily for 3 consecutive days with an interval of ϳ20 min. The sequence of the starting points was randomly selected. The escape latency and the swimming distance for mice to find the hidden platform were automatically recorded by the computer analyzing system (Target/2, Neuroscience). The value for each session was defined as the mean of four trials. The probe test was not conducted because the deficits were too modest to evaluate the effects of compounds.

Novel object recognition test
Male Tg2576 mice of 5 months of age were divided into two groups, vehicle treated (n ϭ 14) and TAK-070 3 mg/kg treated (n ϭ 15). As a nontransgenic control group, wild-type littermates (n ϭ 15) were used. Tg2576 mice were treated with TAK-070 (3 mg/kg, p.o.) or vehicle (0.5% MC) once a day for 15 d before the test. During the test, each mouse was treated with TAK-070 or vehicle after all trials were completed.
Each mouse was subjected to the novel object recognition test from day 16 to day 17. In the acquisition session on day 16, the same two objects were placed in the back corner of the test box (30 ϫ 30 ϫ 30 cm). The mouse was then placed in another corner of the box and the time exploring each object was recorded for 5 min. After 24 h later on day 17, animals were placed back into the same box, except that one of the familiar objects used during the acquisition was replaced with a novel object. The animals were then allowed to explore freely for 5 min. A preference ratio of the time exploring the novel object to the time exploring both objects was calculated as an index of cognitive function.

Statistical analysis
Statistical analysis was performed by the one-tailed Williams' test for analysis of multiple groups in dose-response study, by Tukey's test for analysis of multiple groups in no dose-response study or Student's t test for analysis of two groups under the BSAS program.

TAK-070 reduced A␤ secretion and increased that of sAPP␣ in cell cultures
We treated human IMR-32 neuroblastoma cells with TAK-070 for 24 h, and measured the levels of A␤ and sAPP␣ in the conditioned media by ELISA. We observed a concentration-dependent suppression of the secretion of A␤, with minimum effective concentrations (MECs) for A␤ 40 and A␤ 42 of ϳ100 and ϳ1000 nmol/L, respectively (Fig. 2 A). TAK-070 also stimulated sAPP␣ production in a concentration-dependent manner with MEC of ϳ100 nmol/L. The percentage reduction in the levels of A␤ 40 and A␤ 42 , and percentage increase in that of sAPP␣ by treatment with 3 mol/L TAK-070 were ϳ50, ϳ70, and ϳ30%, respectively. Similarly significant effects at submicromolar to micromolar ranges of TAK-070 on APP processing (ϳ25% reduction in A␤ secretion and ϳ90% increase in sAPP␣ at 3 mol/L TAK-070) were observed in mouse Neuro-2a neuroblastoma cells stably overexpressing human APP carrying Swedish-type familial Alzheimer mutation (APPsw; N2aAPPsw cells) (Fig. 2 B).

TAK-070 inhibited BACE1 activity in cultured cells
We next examined the effects of TAK-070 in N2aAPPsw cells by immunoblot analysis. Treatment with TAK-070 (3 mol/L) significantly decreased the secreted level of both human Swedish sAPP␤ and mouse endogenous sAPP␤, N-terminal counterparts of APP generated by BACE1 cleavage, by ϳ16 and ϳ19%, respectively. Simultaneously, the levels of human and mouse endogenous sAPP␣ were increased by ϳ70% and ϳ30%, respectively (Fig. 3A). We then examined the effects of TAK-070 on the levels of membrane-bound APP and its C-terminal stubs (e.g., C83 and C99), BACE1, and ADAM10 [a neuronal ␣-secretase candidate (Jorissen et al., 2010)] in lysates of N2aAPPsw cells. TAK-070 decreased the level of C99 by ϳ15%, in contrast to the prominent increase in the levels of C83 and C99 (by ϳ2.1and ϳ7.1-fold, respectively) by inhibition of ␥-secretase by DAPT (Fig. 3B). TAK-070 treatment did not significantly affect the protein levels of APP, C83, BACE1, or ADAM10 (Fig. 3B). The levels of mouse sAPP␤ in the conditioned media of TAK-070-treated naive N2a cells also was decreased (supplemental Fig. S1, available at www.jneurosci.org as supplemental material). We further examined the effects of TAK-070 on ␣-secretase activity using a cell-based, peptide cleavage assay (Doedens et al., 2003). Although (Ϫ)-epigallocatechin-3-gallate induced the enzymatic activity, in line with the reported increase in the active form of ADAM10 (Obregon et al., 2006), TAK-070 did not show any incremental effects on the ␣-secretase-cleaved product (supplemental Fig. S2, available at www.jneurosci.org as supplemental material), suggesting that TAK-070 is not an ␣-secretase activator.

Noncompetitive BACE1 inhibition by TAK-070 in a cell-free assay
To confirm that TAK-070 has a direct inhibitory effect on BACE1, we developed a cell-free assay, using recombinant full-length human BACE1 and a quenching type fluorogenic BACE1 substrate based on ϳ10 aa residues flanking the ␤-cleavage site of wild-type human APP. TAK-070 inhibited the BACE1 activity in a concentration-dependent manner, with IC 35 of ϳ3.15 mol/L and MEC of ϳ100 nmol/L (Fig.  4A), the latter being a similar effective concentration to that in cell culture studies ( Fig.  2A,B). Under the same experimental conditions, a peptidic BACE1 inhibitor (TEE-ISEVNXVAEF; X ϭ statine) inhibited BACE1 activity with IC 35 value of 38.8 nmol/L, which was consistent with the previously published data (Sinha et al., 1999). To further examine the inhibitory profile of TAK-070, we conducted a Lineweaver-Burk plot analysis by incubating the fluorogenic BACE1 substrate with recombinant full-length human BACE1 in the presence of 10 or 30 mol/L TAK-070. All fitted lines converged at an identical point on the x-axis with an estimated K m value of 156 mol/L (Fig. 4B), indicating that TAK-070 inhibits BACE1 in a noncompetitive manner. The K i value estimated from the y-axis values with an intercept of (1 ϩ [I]/K i )/V max was 19 mol/L. TAK-070 did not inhibit other aspartic proteases (e.g., cathepsin D and E, renin, and ␥-secretase (Takahashi et al., 2003)), nor activated enzymatic activity of human TACE in cell-free assays even at the concentration of 100 mol/L (data not shown), in agreement with the cell culture data described above.

Binding of TAK-070 to full-length BACE1, but not to its extracellular domain
To gain further insight into the mechanism of the noncompetitive BACE1 inhibition by TAK-070, we examined the binding of TAK-070 to BACE1 using a surface plasmon resonance assay. Since TAK-070 inhibited the proteolytic activity of full-length BACE1 [BACE1 (1-501)] in a noncompetitive manner, but not that of the truncated BACE1 (1-454), lacking the transmembrane domain (data not shown), we first compared the binding of TAK-070 to BACE1 (1-501) or truncated BACE1 (1-454). Surface Plasmon resonance assay clearly showed that TAK-070 was specifically bound to BACE1 (1-501) in a concentration-dependent manner (0.5-8 mol/L), but not to BACE1 (1-454) within the same concentration range (Fig. 5A). To further narrow down the binding site of TAK-070 within the C-terminal region of BACE1, we examined the binding of TAK-070 to a series of C-terminally truncated BACE1, i.e., BACE1 (1-460), (1-465), (1-471), and (1-474). The binding of TAK-070 to BACE1 (1-460) and (1-465) was completely lost, whereas BACE1 (1-474) retained a comparable affinity to TAK-070 as BACE1 (1-501), and the binding of BACE1 (1-471) was partially impaired (Fig. 5B). These data suggest that the critical region within the C terminus of BACE1 for binding to TAK-070 resides around residues 465-474, a subdomain of the membrane spanning region. We also examined the binding of TAK-070 to recombinant proteins of APP (18-688) containing Kunitz-type protease inhibitor domain and the BACE1cleavage site or sAPP␤, and found that neither APP (18-688) nor sAPP␤ showed significant binding to TAK-070 (5  's t test). B, Immunoblots of APP, C-terminal fragments of APP (C99 and C83), BACE1 and high-and low-molecular-weight forms of ADAM10 (pro-and matured forms, respectively) from lysates of N2aAPPsw cells treated with vehicle, DAPT (3 mol/L), or TAK-070 (3 mol/L) are shown. Levels of ␣-tubulin are shown as an internal control. Note that all the immunoblot data are obtained from a single membrane replica with identical exposure. Values for C99 and C83 (right) are mean percentages of band intensities analyzed by densitometry relative to those in vehicle control (ϮSEM) in the three independent experiments. *p Ͻ 0.05, **p Ͻ 0.01, ***p Ͻ 0.001, versus vehicle control (Student's t test). mol/L) (supplemental Fig. S3, available at www.jneurosci. org as supplemental material).
These data from cell-based and cell-free studies collectively indicate that TAK-070 is a direct, noncompetitive inhibitor for BACE1 that acts by binding to the noncatalytic site of BACE1, presumably to the transmembrane domain.

TAK-070 reduced A␤ and increased sAPP␣ in the brains of Tg2576 mice
We then examined whether TAK-070 is effective on A␤ and sAPP␣ in the brains of Tg2576 mice, a transgenic mouse model of AD that overexpresses APPsw. We first performed a short-term treatment, feeding young female Tg2576 mice with chow containing TAK-070 (5.6 and 56 ppm, corresponding to 0.87 and 8.2 mg/kg/d, p.o., respectively) starting at 2 months of age for 7 weeks. All mice survived without any differences in body weight and food consumption among cohorts. Oral administration of TAK-070 significantly reduced the levels of soluble A␤ 40 and A␤ 42 in Tris buffer-soluble fractions of the cerebral cortex (average Ϯ SEM: 7707 Ϯ 334 and 1825 Ϯ 100 fmol/g wet weight, respectively, in vehicle group) by ϳ16 -23%, and increased that of sAPP␣ by ϳ15-21% at both doses (Fig. 6 A).
We next conducted a long-term treatment of Tg2576 mice with TAK-070. We started treatment at the age of ϳ7 months, just before Tg2576 mice develop the A␤ deposition as amyloid plaques (at ϳ8 months). Tg2576 mice were fed with chow containing 56 ppm TAK-070 until 13 months of age for ϳ6 months. Tg2576 mice tolerated chronic treatment with TAK-070, and the mean survival rates were at similar levels after ϳ6 months treatment by vehicle or TAK-070 (81% or 94%, respectively), without any differences in body weights and food consumption between cohorts.
We first quantitated the levels of Tris-soluble A␤ in the brains of untreated 13-month-old Tg2576 mice, which were dramatically increased by 68% and 129%, respectively for A␤ 40 and A␤ 42 , compared with those at 8 months (Fig. 6 B). Notably, the level of sAPP␣ was decreased by 32% at 13 months. Consistent with the results in young Tg2576 mice (Fig. 6 A), TAK-070 reduced the levels of Tris-soluble A␤ 40 and A␤ 42 by ϳ15 and ϳ25%, respectively, and increased that of sAPP␣ by ϳ22% even after the 6 months of treatment (Fig. 6 B).
We next quantitated the levels of insoluble A␤ that was extracted from the Tris-insoluble pellets by formic acid denaturation. The levels of insoluble A␤ 40 and A␤ 42 in untreated Tg2576 mice were markedly increased at 13 months by ϳ35-fold and ϳ23-fold, respectively, compared with those of young control mice (6367 Ϯ 720 and 3513 Ϯ 317 pmol/g wet weight, in 8 months of Tg2576). No gender differences were noted in the extent of age-related A␤ increase in our cohort (data not shown). Chronic TAK-070 treatment significantly reduced the levels of insoluble A␤ 40 and A␤ 42 by ϳ30% (Fig. 6C).
We then analyzed the effects of TAK-070 on the formation of A␤ plaques using immunohistochemistry and unbiased morphometric analysis. The numbers of A␤ plaques in the cerebral neocortex and hippocampus in TAK-070-treated cohort were markedly reduced  compared to those in the vehicle-treated mice (Fig. 6D). Quantitative analysis demonstrated that the A␤ burden (i.e., percentage area covered by A␤ immunoreactivity), as well as the number of plaques per area, were reduced by ϳ60% upon treatment with TAK-070 (Fig. 6E), in agreement with the biochemical data.

TAK-070 ameliorated behavioral deficits in Tg2576 mouse model of AD
We finally assessed the effects of TAK-070 on the behavioral deficits in Tg2576 mice. For this purpose, we conducted three different types of behavioral tests, i.e., Y-maze test, Morris water maze test and a novel object recognition test in relatively young (ϳ5 months old) Tg2576 mice, in which behavioral impairments, along with synaptic deficits, have been documented at this stage, preceding A␤ deposition (Westerman et al., 2002;Ohno et al., 2004;Jacobsen et al., 2006).
We initially conducted Y-maze test, which has been considered as a test for spatial memory. The total arm entries of vehicle-treated Tg2576 mice (n ϭ 14) were not significantly different from those of the wild-type control mice (n ϭ 15). Treatment with TAK-070 for 9 d did not affect the total arm entries in Tg2576 mice (data not shown), suggesting that repeated treatment with TAK-070 did not have any effects on the basal level of exploring activity. However, the spontaneous alternation in vehicle-treated Tg2576 was significantly reduced to ϳ50%. This reduction was recovered by treatment with TAK-070 in a dosedependent manner, and the ameliorating effect was significant at both dosages of 1 (n ϭ 14) or 3 mg/kg (n ϭ 14) (Fig. 7A).
We then assessed the effects of TAK-070 on impairments in spatial memory by sequentially subjecting the same cohorts to the Morris water maze test. The ability of Tg2576 mice to find an invisible platform was impaired compared to that in wild-type mice. On training day 2, significant differences in both escape latency and swimming distance remained between Tg2576 and wild-type mice, whereas they diminished on day 3. Treatment Figure 6. Effects of TAK-070 on A␤ and sAPP␣ levels in the brains of Tg2576 mice. A, Levels of Tris-soluble A␤ 40 , A␤ 42 , and sAPP␣ in the cerebral cortices of young female Tg2576 mice after short-term administration. Values are mean percentages (ϮSEM) relative to levels in vehicle control (n ϭ 15 for both cohorts). *p Ͻ 0.025, versus vehicle control (one-tailed Williams test). B, Levels of Tris-soluble A␤ 40 , A␤ 42 , and sAPP␣ in cerebral cortices of 13-month-old Tg2576 mice after long-term treatment. The number of 13-month-old Tg2576 mice with vehicle or TAK-070 (56 ppm, corresponding to ϳ7 mg/kg/d, p.o.) were 13 (male 6, female 7) and 16 (male 10, female 6), respectively after 6 months treatment. Values are mean percentages (ϮSEM) relative to levels in young controls (8-month-old nontreated Tg2576, n ϭ 9). C, Levels of Tris-insoluble, formic acid-extractable A␤ 40  with TAK-070 reduced the latency (Fig. 7B), as well as the distance (Fig. 7C), in a dosedependent manner. On training day 2, the reduction in the swimming distance in TAK-070-treated Tg2576 mice (3 mg/kg) was statistically significant ( p Ͻ 0.025, Williams' test). No significant effects were observed on the swimming speed between the vehicle-and TAK-070-treated mice (data not shown). On the next day of Morris water maze test, we obtained brains from all Tg2576 mice and measured the brain levels of Tris buffer-soluble A␤ peptides, which were decreased by ϳ9 -16% for A␤ 40 , and ϳ8 -12% for A␤ 42 , by administration of 1 and 3 mg/kg TAK-070, respectively, compared with those in vehicle-treated mice. These values were at similar levels to those observed in short-term treatment (see Fig. 6A).
We further assessed the effects of TAK-070 on recognition memory by a novel object recognition test using new cohorts. After a 15 d successive treatment with vehicle (n ϭ 15; wild type mice, n ϭ 14; Tg2576) or TAK-070 (3 mg/kg, p.o., n ϭ 15; Tg2576), all mice were subjected to an acquisition trial on day 1, in which mice were allowed to get access to the two identical objects in the test box. As expected, all mice equally interacted with both objects in the exploration (data not shown). On the following day, one of the two objects was replaced with a novel one and retention test was conducted. Whereas wild-type mice more frequently interacted with a novel object than a familiar object, with the novel object preference ratio of 78% (Fig. 8A,B), vehicle-treated Tg2576 mice showed a markedly decreased preference ratio of 44% (Fig. 8B), indicating an apparent impairment in recognition memory in Tg2576. By contrast, TAK-070 treatment significantly recovered the preference ratio to a normal range of 71% (Fig. 8B).

Discussion
We show that TAK-070 is an orally active BACE1 inhibitor that effectively lowers the levels of soluble A␤ and increases that of sAPP␣, inhibits cerebral deposition of insoluble A␤, and rescues behavioral deficits in vivo in a transgenic mouse model of AD. Notably, the partial inhibition in the levels of soluble A␤ eventually resulted in a significant reduction in A␤ deposition after a 6 month chronic treatment, preserving the pharmacological efficacy at a similar level to that in a short-term treatment. We also suggest that TAK-070 exerts a unique noncompetitive inhibitory activity by interacting presumably with the transmembrane region of BACE1 outside the catalytic domain.
Multiple lines of genetic, clinical, and cell biological evidence support the causative role of A␤ in the pathogenesis of AD (for review, see Selkoe and Schenk, 2003). In contrast, sAPP␣ has been reported to have neurotrophic effects, e.g., promotion of synapse formation or amelioration of cognitive deficits [for review, see Isacson et al. (2002) and Postina (2008)]. In our present study, untreated, aged Tg2576 mice had lower brain levels of sAPP␣ and higher soluble A␤ with aging, in agreement with previous observations that BACE1 activity is upregulated with aging in the brains of animals as well as humans (Fukumoto et al., 2004;Zohar et al., 2005). Hence, manipulation of APP processing by BACE1 inhibition in a way to reduce A␤ and increase sAPP␣ would be a rational strategy for the treatment and prevention of AD.
The chemical structure of TAK-070 differs markedly from that of peptide-based BACE1 inhibitors (for review, see Silvestri, 2009). However our cellular and cell-free assay data clearly indicated that TAK-070 is a bona fide BACE1 inhibitor. Cell-free study showed that TAK-070 directly and specifically inhibited full-length BACE1 without affecting other aspartic proteases. TAK-070 reduced levels of secreted A␤ and sAPP␤, together with an increase in sAPP␣ in cultured cells (Fig. 3), which are in agreement with the previous results of antisense oligonucleotide study for BACE1 (Vassar et al., 1999). The Lineweaver-Burk plot analysis revealed that TAK-070 is a noncompetitive inhibitor (Fig. 4), which was supported by the surface plasmon resonance assay. TAK-070 did bind to the full-length BACE1 (1-501) and truncated BACE1 (1-471 and 474), but not to the truncated BACE1 (1-454, 460, and 465) (Fig. 5). This suggests that TAK-070 inhibits BACE1 activity in a unique mode of interaction by binding to the ϳ10 aa residues in the C-terminal region (residues 465-474) within the transmembrane domain, but not to the catalytic center (located in residues 93-96 and 289 -293). Surface plasmon resonance assay also showed that TAK-070 does not interact with APP(18-688) or sAPP␤ (supplemental Fig. S3, available at www. jneurosci.org as supplemental material). This suggests that TAK-070 does not affect APP processing by binding to subdomain of APP containing the BACE1-cleavage sites. We were not able to completely rule out the possibility that TAK-070 interacts with the transmembrane domain of APP, like benzofurancontaining compounds that bind C99 (Espeseth et al., 2005). However, TAK-070 failed to inhibit A␤ secretion from HEK293 cells overexpressing C99 (data not shown), supporting the notion  that TAK-070 does not target C99 in APP. In addition, the possibility that TAK-070 is an ␣-secretase activator was excluded by (1) the lack of increase in the protein levels of ␣-secretase candidate, i.e., ADAM10 (Fig. 3), (2) lack of inhibition of TACE activity using a peptidic substrate in a cell-free assay (data not shown), and (3) the lack of increase in ␣-secretase activity in a cell-based assay (supplemental Fig. S2, available at www.jneurosci.org as supplemental material).
The potency of TAK-070 to reduce the A␤ secretion in cell cultures was modest (i.e., ϳ25% reduction was achieved at 3 mol/L with a MEC of ϳ0.1-0.3 mol/L in N2aAPPsw cells) (Fig. 2). These results were in agreement with the relatively modest BACE1-inhibitory effect in the cell-free assay with IC 35 of ϳ3.15 mol/L and MEC at ϳ0.1 mol/L (Fig. 4). Interestingly, however, we observed similar levels of reduction in soluble A␤ by ϳ20% in the brains of Tg2576 mice (Fig. 6 A, B). Although small chemicals generally have less potency in brains, hampered by the blood-brain-barrier and cell-penetration issues, this relatively high potency of TAK-070 is likely to be attributable to the highly lipophilic structure bearing N-alkyl-amine moiety. In fact, a single administration of TAK-070 in rat (3 mg/kg, p.o.) yielded effective concentration of ϳ2 mol/L in brain with the T max of ϳ24 h using 14 C-TAK-070, and the brain exposure levels in shortterm-and long-term-treated Tg2576 mice were ϳ8 mol/L and ϳ6 -11 mol/L, respectively (56 ppm of TAK-070, corresponding to ϳ7-8 mg/kg) (Fig. 6) (our unpublished observations). Furthermore, it has been reported that full-length BACE1, forming a high-molecular-weight complex associated with lipid, exhibits higher enzymatic activity than that of C-terminally truncated BACE1 (1-454) (Marlow et al., 2003;Westmeyer et al., 2004). This may support the view that lipophilic TAK-070 effectively reaches the membrane-associated BACE1 complex.
TAK-070 exhibits ceiling effects on reduction in A␤ and increase in sAPP␣ (Figs. 2, 6 A), which may partly be explained by the noncompetitive inhibitory profile for BACE1. We have also observed similar plateau effects in normal rats with a minimum effective dose of 0.1 mg/kg after 4 week administration (our unpublished observation). However, long-term treatment with TAK-070 led to more pronounced A␤-lowering effects on insoluble A␤ (Fig. 6C-E) than on soluble A␤ (Fig. 6 A, B). This finding dovetails with the observation in BACE1 heterozygous KO crossed with PDAPP transgenic mice, in which soluble A␤ levels were lowered only by 12% at a young age, whereas A␤-accumulation was eventually reduced by ϳ50 -90% with synaptic amelioration in elderly animals (McConlogue et al., 2007). Together, these results strongly suggest that partial inhibition of BACE1, causing partial reduction in A␤ and increase in sAPP␣, has sufficient pharmacological efficacy on normalization of APP processing and cognitive functions.
Behavioral deficits in Tg2576 mice have been reported to occur before the deposition of A␤ plaques (Westerman et al., 2002;Ohno et al., 2004;Jacobsen et al., 2006), which may be due to the accumulation of toxic forms of A␤, e.g., oligomers, that leads to the deterioration of synaptic functions and behaviors (Walsh et al., 2002;Cleary et al., 2005;Venkitaramani et al., 2007). In the present study, relatively young (ϳ5 months) Tg2576 mice showed impairment in behaviors both in Y-maze and novel object recognition tests, whereas the deficits in Morris water maze test were modest, with no differences in the acquisition trial on day 3 between Tg2576 and wild-type cohorts. TAK-070 ameliorated all these behavioral deficits by a short-term treatment at biochemically effective doses (1-3 mg/kg, p.o.) (Figs. 7, 8). TAK-070 had ameliorative effects in the Y-maze and Morris water maze tests that reflect the hippocampal-dependent learning, in line with observations in BACE1 homozygous KO/APP transgenic bigenic mice (Ohno et al., 2004(Ohno et al., , 2006(Ohno et al., , 2007. However, there were pivotal differences: TAK-070 treatment affected neither the total number of arm entry in Y-maze test (Ohno et al., 2004) nor the swimming speed in Morris water maze test (Ohno et al., 2006), which were documented to be abnormal in BACE1homozygous KO regardless of APP-transgenic background. Furthermore, BACE1-homozygous KO in nontransgenic background have been reported to show cognitively deteriorative (Ohno et al., 2004(Ohno et al., , 2006(Ohno et al., , 2007, schizophrenia-like (Savonenko et al., 2008), or hypomyelination (Hu et al., 2006;Sankaranarayanan et al., 2008) phenotypes, underscoring the necessity of BACE1 activity for physiological functions, probably due to multiplicity of substrates for BACE1 (for review, see Marks and Berg, 2008). Also in nontransgenic aged rats, TAK-070 ameliorated behavioral deficits in the water maze test (our unpublished observation). Hence, TAK-070 appears to be pharmacologically effective and safe by partial BACE1 inhibition, avoiding adverse events due to complete inhibition of BACE1.
It is noteworthy that the pharmacological effects of orally administered TAK-070 for ϳ6 months on the brain levels of soluble A␤ and sAPP␣ were similar to those in short-term treatment ( Fig. 6 A, B). Under the chronic treatment, mice were tolerable to TAK-070 and survived comparable to vehicle control after ϳ6 months. These profiles should be a merit of this compound, considering the long period of AD medication. The sustained efficacy of TAK-070 markedly differs from those documented in other BACE1 inhibitors (Sankaranarayanan et al., 2008) or on the higher efficacy of a compound in the presence of inhibitors of P-glycoprotein (Hussain et al., 2007), that determines exposure levels of compounds in brains.
In sum, the successful treatment by a noncompetitive BACE1 inhibitor, TAK-070, provides strong support for the validity of partial BACE1 inhibition as a disease-modifying as well as symptomatic therapy for AD. TAK-070 will also provide a clue for the elucidation of the mechanism of noncompetitive regulation of the activity of BACE1.