1-(3′,4′-Dichloro-2-fluoro[1,1′-biphenyl]-4-yl)-cyclopropanecarboxylic Acid (CHF5074), a Novel γ-Secretase Modulator, Reduces Brain β-Amyloid Pathology in a Transgenic Mouse Model of Alzheimer's Disease without Causing Peripheral Toxicity

Alzheimer's disease (AD) is the most common form of dementia. The basic pathological abnormalities in AD brains are amyloid plaques, neurofibrillary tangles, and neuronal loss. Amyloid plaques consist of a proteinaceous core composed of 5- to 10-nm fibrils of β-amyloid peptide (Aβ) surrounded by dystrophic neurites, astrocytic processes, and microglial cells. Aβ consists of 40 to 42 amino acids generated by the cleavage of amyloid precursor protein (APP). Although the main form of Aβ is made of 40 amino acids, the 42-residue species (Aβ42) constitutes the major component of amyloid plaques, because it is more prone to aggregate into fibrils. Fibrillar Aβ contained in plaques is in dynamic equilibrium with soluble monomers and oligomers, which are neurotoxic (Gong et al., 2003).

Many epidemiological studies suggest that long-term use of nonsteroidal anti-inflammatory drugs (NSAIDs) may delay or prevent the onset of AD (Szekely et al., 2004). Despite this encouraging evidence, all randomized, controlled studies aiming to reduce inflammation in the brain of AD patients produced negative results. The failures include both steroidal anti-inflammatory agents such as prednisone (Aisen et al., 2000) and nonsteroidal anti-inflammatory drugs, e.g., hydroxychloroquine (Van Gool et al., 2001); and nonselective cyclooxygenase (COX) inhibitors, e.g., naproxen (Aisen et al., 2003) and COX-2-selective inhibitors, e.g., rofecoxib (Aisen et al., 2003; Reines et al., 2004) and celecoxib (Sainati et al., 2000). Negative results were also obtained in patients with mild cognitive impairment treated with rofecoxib (Thal et al., 2005) and in normal elderly subjects at risk of developing AD treated with naproxen or celecoxib (ADAPT Research Group et al., 2007). These multiple failures have questioned the role of inflammation in AD.

New insight into understanding the apparent discrepancy between epidemiological studies and clinical trials was the discovery that some NSAIDs decrease the production of Aβ42 (Weggen et al., 2001) and counteract the progression of Aβ pathology in transgenic mouse models of AD (Lim et al., 2000, 2001; Jantzen et al., 2002; Yan et al., 2003; van Groen et al., 2005; Wilcock et al., 2007). The proposed mechanism for this effect is an allosteric modulation of presenilin-1, the main component of the γ-secretase complex, that is responsible for the formation of Aβ (Eriksen et al., 2003; Takahashi et al., 2003; Weggen et al., 2003b; Beher et al., 2004; Lleó et al., 2004). The inhibition of Aβ42 production is independent from the anti-COX activity, and it depends on the chemical structure of the NSAIDs, with some compounds displaying activity (ibuprofen, sulindac, indomethacin, and flurbiprofen) and others not (naproxen, aspirin, meloxicam, and celecoxib) (Weggen et al., 2001; Morihara et al., 2002; Eriksen et al., 2003). This could explain the negative results of the large AD trials pursued to date, because they were conducted with compounds (naproxen, prednisone, hydroxychloroquine, rofecoxib, and celecoxib) not able to inhibit Aβ secretion in vitro. Interestingly, the only positive study in AD regards indomethacin (Rogers et al., 1993), shown to decrease Aβ42 production in vitro (Weggen et al., 2001) and plaque deposition in vivo (Sung et al., 2004).

The Aβ42-lowering NSAIDs differ from traditional γ-secretase inhibitors because they do not inhibit the γ-secretase-mediated cleavages of either APP at the alternative ϵ site, or Notch-1 at the S3 site (Weggen et al., 2001, 2003a). Unfortunately, the use of NSAIDs in AD is hampered by gastrointestinal toxicity, which is associated with COX inhibition.

We have synthesized new NSAID derivatives endowed with selective Aβ42-lowering activity but devoid of COX inhibitory activity, thus suitable for chronic use in AD patients (Peretto et al., 2005). Within this new chemical series, CHF5074 (Fig. 1) proved to be of major interest. In human neuroglioma cells (H4swe), CHF5074 preferentially lowers Aβ42 secretion, with an IC₅₀ of 40 μM (Imbimbo et al., 2007). CHF5074 does not display inhibitory activity on COX-1 and COX-2 enzymes when used at concentrations up to 100 and 300 μM, respectively (Imbimbo et al., 2007). At 100 μM, CHF5074 does not alter the expression profile of several Notch intracellular domain-responsive genes (Imbimbo et al., 2007). In rats, CHF5074 is orally well absorbed (50%), and it is slowly eliminated from plasma (t_1/2 = 20.7 h) (Peretto et al., 2005). In young Tg2576 transgenic mice, short-term oral dosing with CHF5074 causes dose-dependent reduction of plasma Aβ42.

We performed the present study to determine the effect of chronic treatment with CHF5074 on brain Aβ levels and amyloid plaque deposition in the Tg2576 mouse model of AD. In addition, we histologically examined peripheral tissues for signs of toxicity, in particular those that may be related to COX inhibition or altered Notch signaling.

Materials and Methods

Animals and Treatment. All procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the National Institutes of Health. Tg2576 transgenic mice (male and female) expressing Swedish mutated form of human APP (APPswe) (Hsiao et al., 1996) were bred in-house. All animals were maintained on a 12:12-h light/dark cycle with unrestricted access to food and water until use. Twenty-eight mice of 9.5 to 10.5 months of age were treated with CHF5074 (375 ppm in the diet) or standard diet (controls) for 17 weeks. At week 17, animals were sacrificed by decapitation, and brains were rapidly removed and divided in the two hemispheres. One hemisphere was used for measurements of Aβ plaques, and the other hemisphere was used for extractable Aβ. Blood samples were collected in EDTA-coated tubes and centrifuged at 800g for 20 min to separate plasma. Plasma samples were divided into two aliquots of approximately 100 μl each, and they were stored at –80°C until Aβ or CHF5074 assay. Peripheral tissues, i.e., liver, kidneys, spleen, esophagus, stomach, duodenum, jejunum, ileum, cecum, colon, and rectum, were collected for histological evaluations.

Brain Aβ Extraction. Aβ40 and Aβ42 were extracted from brain by using both the SDS and formic acid (FA) methods. Because γ-secretase modulators are thought to affect mainly SDS-extractable Aβ pools, i.e., the newly formed Aβ (Yan et al., 2003; Dvir et al., 2006), a double SDS extraction was performed in an attempt to maximize SDS-extractable Aβ. One hemisphere was first homogenated in 2% SDS in water [100 mg/ml (wet w/v)] containing protease inhibitors (Complete Protease Inhibitor Cocktail; Roche Diagnostics, Basel, Switzerland) and an aliquot of 500 μl was centrifuged at 16,000g for 10 min. Supernatants were collected (first SDS aliquot) and stored at –80°C. The pellet was resuspended in 2% SDS containing protease inhibitors and centrifuged at 100,000g for 1 h at 4°C. The supernatant was collected (second SDS aliquot) and stored at –80°C. Pellets obtained after the second SDS extraction were mixed with 500 μl of 70% FA and centrifuged at 100,000g at 4°C for 1 h. Supernatants were collected and stored at –80°C for assessment of Aβ levels.

Aβ Measurements in Brain Extracts and Plasma. Levels of Aβ40 and Aβ42 in plasma and in brain extracts were measured with commercial ELISA kits (The Genetics Company, Zurich, Switzerland). The microtiter plates were coated with capturing purified monoclonal antibodies specifically recognizing the C terminus of human Aβ40 (clone G2-10, reactive to amino acid residues 31–40, isotype IgG2b, κ) or Aβ42 (clone G2-13, reactive to amino acid residues 33–42, isotype IgG1, κ). As detection antibody, a monoclonal biotin-conjugated antibody recognizing the N terminus of human Aβ (clone W0-2, reactive to amino acid residues 4–10, isotype IgG2a, κ) was used. The assay was linear in the range 25 to 500 pg/ml, and the detection limit was 25 pg/ml. Before Aβ40 and Aβ42 ELISA determinations, the two brain SDS aliquots (50 μl each) were pooled and diluted 1:800 and 1:200 in standard diluent for Aβ40 and Aβ42 assay, respectively (SDS-extracted Aβ). Supernatants obtained following extraction with FA were neutralized with 20 vol of 1 M Tris, pH 11.0 (FA-extracted Aβ). Total Aβ40 and Aβ42 levels were the sum of the double SDS- and FA-extracted fractions. Plasma was diluted 1:4 for Aβ42 and 1:20 for Aβ40 determinations.

Immunohistochemistry. For Aβ plaque measurements, brain hemispheres were fixed by immersion in Alcolin (DiaPath, Martinengo, Italy) for 2 weeks. The hemibrains were then cut coronally into two blocks, and then tissue was dehydrated and embedded in paraffin. A series of coronal sections at the levels of anterior striatum, anterior hippocampus, and middle hippocampus were cut at 10 μm in thickness. Before immunohistochemistry for Aβ, the slides were immersed in 10% buffered formalin overnight. Sections were subsequently incubated in 80% FA for 15 min to expose the epitope. Staining of Aβ plaques was performed by immunohistochemistry using 1:500-diluted 6E10 antibodies (Signet Laboratories, Dedham, MA), as primary antibody, and biotin-conjugated mouse-on-mouse kit Peroxidase (Dako Denmark A/S, Glostrup, Denmark), as revealing system. Peroxidase activity was detected by treatment with 3,3′-diaminobenzidine (Hoffman-La Roche, Nutley, NJ). Slides were photographed using a digital Nikon DS microscope color camera. Digital images were analyzed using NIS-Elements software (Nikon, Tokyo, Japan). Each image was analyzed using the automated target detection mode. Images size was 1280 × 960 pixels and target area had a size of 68,000 μm². The software determined numbers of plaques, mean areas of plaques, and plaque area fraction (immunopositive area/total area used as scan object). We performed nine counts for each of the three above-mentioned levels. Analyses were performed only in the cortex and hippocampus using 10× objective. Glial filaments in CA1 region of hippocampus were stained with specific glial fibrillar acidic protein (GFAP) antibodies (Z0334; Dako Denmark A/S), using 1:800 dilution. For the counts in this region, we used a 20× objective and a target area of 17,300 μm².

Histological Examinations. Tissue samples of liver, kidneys, spleen, esophagus, stomach, duodenum, jejunum, ileum, cecum, colon, and rectum, fixed in 10% buffered formalin, were trimmed, dehydrated, embedded in paraffin wax, and sectioned at 5 μmin thickness. Slides were stained with hematoxylin and eosin, and they were examined by a blinded skilled pathologist for the qualitative evaluation of any treatment-related changes. Samples of ileum from all animals were sectioned at 5 μm in thickness and stained with Alcian Blue to reveal mucopolysaccharides and glycosaminoglycans in the goblet cells. Staining was rated according the following ordinal scale: 0, no specific positive reaction detected; 1, few scattered cells reacting weakly positive; 2, many weakly positive cells; 3, many strongly positive cells; and 4, extremely strong positive cells.

Pharmacokinetics. CHF5074 levels in plasma and in brain samples were measured by liquid chromatography as described previously (Imbimbo et al., 2007). In brief, samples were prepared by adding 300 μl of acetonitrile and 40 μl of 40% phosphoric acid to 100 μl of plasma or brain homogenate and by placing the mixture in a vortex for 5 s. Plasma and brain samples were then centrifuged at 14,000 rpm for 5 min, and the supernatants (15 and 50 μl, respectively) were injected into the high-performance liquid chromatography system. Equipment systems with fluorescence (Waters 474; Waters, Guyancourt, France) or mass spectrometry (API 2000; Applied Biosystems, Foster City, CA) detectors were used. The chromatographic conditions were adapted to each compound to obtain good peak separation and detection sensitivity. A mixture of 20 mM ammonium formate buffer/acetonitrile/methanol was used as mobile phase for the fluorescence detector. The assay was linear between 400 and 20,000 ng/g in brain and between 100 and 8500 ng/ml in plasma, with limits of quantitation of 400 ng/g in brain and 100 ng/ml in plasma.

Studies in H4swe and HEK293swe Cells. Aβ42 and Aβ40 were measured in the culture medium of H4swe human neuroglioma cells and of HEK293swe human embryonic kidney 293 cells, both expressing APPswe, as reported previously (Camacho et al., 2004; Facchinetti et al., 2006). In brief, cells were seeded onto 24-well plates (2 × 10⁵ cell/well) in culture medium (Opti-MEM; Invitrogen, Carlsbad, CA) containing 10% fetal bovine serum, and the cells were allowed to grow to confluence for 24 h, in 5% CO₂, 95% air in a humidified atmosphere. Cells were exposed overnight to increasing concentrations of CHF5074 (0.03–100 μM) and/or (R)-flurbiprofen (10–300 μM) in culture medium (Opti-MEM) without serum. Both compounds were dissolved in dimethyl sulfoxide (0.1% final concentration). At the end of the treatment, Aβ40 and Aβ42 in culture media were measured by ELISA. Cell viability was assessed with the 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide assay. For Notch assays, HEK cells were seeded in six-well plates and cultured at 37°C in the presence of 5% CO₂. After 24 h of incubation, cultures (60% confluence) were transfected using FuGENE-6 (Roche Diagnostics) with 1 μg of pCS2NotchΔE plasmid. After 48 h, medium was discarded, and a minimum volume of Dulbecco's modified Eagle's medium (Invitrogen) supplemented with increasing concentrations of CHF5074 (5–200 μM) was added to the cells for 16 h. After this period, 10 μM lactacystin was added to the cells to block degradation of Notch intracellular domain. Cells were lysed in 1% Triton X-100, and postnuclear fractions were isolated by centrifugation at 10,000g at 4°C for 15 min. Protein concentrations were determined by the Bradford assay (Pierce Chemical, Rockford, IL). Proteins were resolved in 10% Bis-Tris SDS-polyacrylamide gel electrophoresis gels (Invitrogen), and then they were transferred to nitrocellulose membranes for Western blot detection with anti-myc monoclonal antibody 9E10 (Sanver Tech, Heerhugowaard, The Netherlands), Anti-cleaved Notch (val 1744; Westburg, Leusden, The Netherlands), and anti-β-actin (Sigma-Aldrich, St. Louis, MO). Signals were detected using Odyssey IR detection (LI-COR Biosciences, Bad Homburg, Germany) and quantified.

Statistics. Student's t test was used to evaluate treatment differences in plasma or brain Aβ40 and Aβ42 levels and in Aβ plaque load. When appropriate, data were transformed (log or square root) to normalize the results and to obtain homoscedasticity. Treatment differences in rating of goblet cell staining were tested with the Mann-Whitney U test. Calculations were done with the statistical software SigmaStat 2.0 for Windows (SPSS Inc., Chicago, IL). Results were presented as percentage of vehicle and as mean ± S.E.M.

Results

In Vitro Effects of CHF5074 on Aβ Secretion. The effects of CHF5074 on Aβ secretion were evaluated in H4swe, a human neuroglioma cell line expressing APPswe, mimicking the Tg2576 mouse model. CHF5074 concentrations ranging from 0.03 to 100 μM were evaluated. No fetal bovine serum was added to the medium to avoid the avid protein binding of CHF5074 (98.4 ± 0.4% with 10% serum; P. Petrillo, personal communication). At concentrations up to 30 μM, CHF5074 showed no or modest cytotoxicity as assessed by the 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide assay (18% at 30 μM). However, at 100 μM a marked cytotoxicity was noted (49%), so that Aβ levels were not considered at this drug concentration. At noncytotoxic concentrations, both Aβ species were reduced by CHF5074, but with different sensitivities (Fig. 2). The estimated IC₅₀ values were 3.6 μM for Aβ42 and 18.4 μM for Aβ40. (R)-Flurbiprofen decreased Aβ42 and Aβ40 only at high concentrations, and formal IC₅₀ values were not calculated. At the maximal concentration of 300 μM, (R)-flurbiprofen reduced Aβ42 levels to 24.2 ± 3.0% of vehicle and Aβ40 levels to 75.6 ± 4.3% of vehicle. No significant cytotoxicity was detected with (R)-flurbiprofen at the concentrations tested (16% at 300 μM).

In Vitro Effects of CHF5074 on Notch Processing. The effects of CHF5074 on Notch processing was evaluated in HEK293swe cells. CHF5074 concentrations ranging from 5 to 200 μM were evaluated in absence of serum. At 5 μM, no inhibition of Notch processing was observed (Fig. 3). At concentrations of 15 μM and higher, a marked cytotoxicity was noted, making it impossible to estimate an IC₅₀ value for Notch endoproteolysis inhibition.

Effect of Chronic Treatment with CHF5074 on Brain and Plasma Aβ Levels in Aged Tg2576 Mice. Animals of 9.5 to 10.5 months of age were treated for with CHF5074 for 17 weeks. At the end of the treatment, total brain Aβ40 levels in the group of animals receiving vehicle were 960.9 ± 190.3 pmol/g (Fig. 4, top) with SDS and FA fractions being 442.4 ± 58.0 and 518.5 ± 150.0 pmol/g, respectively. The animals receiving CHF5074 had total brain Aβ40 levels of 488.2 ± 88.8 pmol/g (SDS and FA fractions, 252.8 ± 37.7 and 223.2 ± 54.6 pmol/g, respectively), herein displaying a significant reduction of 49.2 ± 9.2% compared with the vehicle group (Student's t test, p = 0.017). Total levels of Aβ42 in the vehicle group were 285.2 ± 48.5 pmol/g (Fig. 4, top), with SDS and FA fractions being 133.5 ± 23.0 and 151.7 ± 39.1 pmol/g, respectively. In the group receiving CHF5074, total brain levels of Aβ42 were 161.2 ± 27.6 pmol/g (SDS and FA fractions, 83.4 ± 15.7 and 77.8 ± 20.1 pmol/g, respectively). This was equivalent to a 43.5 ± 9.7% reduction compared with the vehicle group (Student's t test, p = 0.027).

To evaluate the effects of CHF5074 on the peripheral Aβ levels, we also measured plasma concentrations of Aβ40 and Aβ42 at the end of the 17-week treatment. No significant differences in the levels of both peptides were observed between CHF5074 treated and control mice (Fig. 4, bottom).

Effect of Chronic Treatment with CHF5074 on Brain Plaque Burden in Aged Tg2576 Mice. Paraffin-embedded sections were immunolabeled for Aβ using the 6E10 antibody. Aβ immunoreactivity in representative coronal sections of cortex and hippocampus of vehicle- and CHF5074-treated mice is shown in Fig. 5, A and B, respectively. Quantification of the immunoreactivity demonstrated significantly lower plaque area fraction in CHF5074-treated animals compared with controls in both cortex (Fig. 5C, –52.2 ± 5.6%; Student's t test, p = 0.0003) and hippocampus (Fig. 5D, –76.7 ± 6.4%; Student's t test, p = 0.004). Number of plaques was also significantly reduced by the CHF5074 treatment compared with controls in cortex (Fig. 5C, –48.9 ± 6.6%, Student's t test, p = 0.0004) and hippocampus (Fig. 5D, –66.2 ± 10.3%, Student's t test, p = 0.037). The mean size of the remaining plaques was not significantly different between the two treatment groups in either cortex (Fig. 5C, –15.4 ± 4.9%; Student's t test, p = 0.151) or hippocampus (Fig. 5D, –43.2 ± 14.1%; Student's t test, p = 0.100).

Effect of Chronic Treatment with CHF5074 on Brain Gliosis in Aged Tg2576 Mice. To determine whether CHF5074 treatment affected reactive astrocytosis associated to Aβ deposition, immunostaining of GFAP with Z0334 antibody was performed in the CA1 region of hippocampus of the mice. GFAP immunoreactivity in representative sections of vehicle- and CHF5074-treated mice are shown in Fig. 6, top. Quantitation of immunostaining did not show significant differences between treatment groups in either the number of immunoreactive glial filaments or in total area occupied by the immunoreactive filaments (Fig. 6, bottom), although reduction in mean dimension of glial filaments in CHF5074-treated animals compared with controls approached statistical significance (–42.8 ± 6.3%; Student's t test, p = 0.057).

Histological Evaluation after Chronic Treatment of CHF5074 in Peripheral Tissues of Aged Tg2576 Mice. Long-term treatment with CHF5074 was well tolerated by aged Tg2576 mice, because no deaths were observed in the treated group, and body weight gain of the two groups was similar (data not shown). To determine whether there were any toxic effects caused by CHF5074 exposure, histological sections of liver, kidneys, spleen, esophagus, stomach, duodenum, jejunum, ileum, cecum, colon, and rectum were examined for any microscopic abnormalities. No significant differences were found upon histopathological examination of tissues obtained from vehicle- or CHF5074-fed animals. Figure 7 shows representative images of liver, kidneys, and spleen of vehicle- and CHF5074-treated animals. In the ileum, rating of goblet cells staining did not reveal differences between the two groups (median score of 3 in both control and treated group). Representative microphotographs of goblet cells for CHF5074-treated and control groups are reported in Fig. 8. Overall, these findings indicate that chronic treatment with CHF5074 in the diet did not cause COX-mediated toxic effects in the gastrointestinal tract and Notch-mediated cell differentiation abnormalities in the ileum.

CHF5074 Plasma and Brain Levels after Chronic Treatment in Aged Tg2575 Mice. Based on diet consumption, the estimated ingested dose of CHF5074 was 60.8 ± 1.7 mg/kg/day. At the end of 17-week treatment, mean drug levels were 228.1 ± 10.1 μM in plasma and 6.4 ± 0.4 μMin the brain, with a mean brain-to-plasma ratio of 3.0 ± 0.3%.

Discussion

Different studies have shown that a subset of NSAIDs acts in vitro as γ-secretase modulators, i.e., they differentially inhibit the production of Aβ42 and Aβ40. The ability of these NSAIDs to lower brain Aβ levels and plaque deposition in vivo has been reported for ibuprofen (Lim et al., 2000, 2001; Yan et al., 2003), indomethacin (Sung et al., 2004) and flurbiprofen derivatives (Jantzen et al., 2002; van Groen et al., 2005; Wilcock et al., 2007). Our data represent the first demonstration that chronic administration of a γ-secretase modulator devoid of anti-COX activity can significantly reduce the deposition of Aβ in the brain. We have demonstrated a marked decrease in both the number and total area occupied by amyloid deposits in the cortex and hippocampus of CHF5074-treated mice compared with their vehicle-treated controls. The average size of the remaining plaques was not significantly affected by the CHF5074 treatment. This suggests that CHF5074 is effective in inhibiting the initiation of the formation of plaques but less effective in inhibiting the growth of the plaques. Similar results were reported after chronic treatment with ibuprofen in the diet (375 ppm for 4 months; Yan et al., 2003) and indomethacin in the drinking water (10 mg/l for 8 months; Sung et al., 2004) in Tg2576 mice of comparable age (15 months) at sacrifice. Another study in 10-month-old Tg2576 mice receiving ibuprofen in the diet for 6 months (375 ppm) produced also similar results (Lim et al., 2000). Comparable effects were also recently described in 15-month-old Tg2576 mice after chronic treatment with the γ-secretase inhibitor MRK-560 (3 mg/kg/day for 3 months; Best et al., 2007).

It has been shown that sequential extraction of brain homogenates using SDS and FA solubilizes most of the total amyloid pool from brains of Tg2576 mice (Kawarabayashi et al., 2001), except for some plaque core and vascular amyloid. At 15 months of age, the SDS-extractable Aβ represents approximately 10% of the Aβ pool extracted by FA (Kawarabayashi et al., 2001). Because γ-secretase modulators are thought to mainly affect SDS-extractable Aβ pools, i.e., the newly formed Aβ (Yan et al., 2003; Dvir et al., 2006), a double SDS extraction was performed in an attempt to maximize this Aβ fraction. Indeed, the double SDS-extracted Aβ in our study represented 46 to 47% of the total extracted Aβ. The lowering effects of the CHF5074 treatment on the double-SDS fraction (–43% for Aβ40 and –38% for Aβ42) was similar to that found for the FA fraction (–57% for Aβ40, –49% for Aβ42). Overall, the CHF5074-induced decrease of total extracted Aβ40 (–49%) and Aβ42 (–44%) in the brain hemispheres agreed well with the decrease in plaque number (44–66%) and total plaque area (52–77%) in cortex and hippocampus assessed by immunostaining, suggesting that these biochemical measures are an accurate reflection of overall amyloid load. On the contrary, we did not observe a significant reduction in plasma Aβ42 or Aβ40 levels in CHF5074-treated animals. This is in contrast with what was found in a previous study in young plaque-free Tg2576 mice in which a significant reduction in Aβ levels was observed in plasma but not in brain after a 4-day treatment (Imbimbo et al., 2007). In that study, the doses of CHF5074 were higher (100 and 300 mg/kg/day) than in the present study (61 mg/kg/day), and they were given by oral gavage. In addition, brain Aβ was extracted with only formic acid. It could be that the effects of CHF5074 on central and peripheral Aβ vary depending from the age of the animal, the dose, and the duration of the treatment.

The mechanism(s) through which chronic treatment with CHF5074 ameliorates amyloid pathology remains to be elucidated. In the present study, CHF5074 concentrations reached in the brain (6.4 ± 0.4 μM) were enough to efficiently inhibit Aβ42 secretion (cell-based IC₅₀ = 3.6 μM). These brain concentrations are not higher than those attained in young transgenic Tg2576 mice after short-term treatment (10.2 ± 3.7 μM, after 100 mg/kg/day for 4 days) where we failed to demonstrate a decrease in brain Aβ levels (Imbimbo et al., 2007). This is in line with previous reports indicating that short-term administration (3–8 days) of Aβ-lowering NSAIDs in young Tg2576 transgenic mice does not attain significant effects on brain Aβ levels (Lanz et al., 2005; Peretto et al., 2005; Stock et al., 2006). Alternatively, long-term treatments (3–8 months) with ibuprofen (Lim et al., 2000, 2001; Yan et al., 2003) indomethacin (Sung et al., 2004), or flurbiprofen derivatives (Jantzen et al., 2002; van Groen et al., 2005; Wilcock et al., 2007) have shown a significant reduction in brain Aβ pathology in transgenic mice. It is possible that γ-secretase modulators do not act centrally but rather peripherally by increasing Aβ42 clearance, and, in turn, by enhancing efflux of Aβ42 from the brain. However, the lack of the significant decreasing effect of CHF5074 on plasma Aβ42 does not support this hypothesis. It cannot be excluded that CHF5074 may decrease Aβ deposition acting through other mechanisms such as activation of the peroxisome proliferator-activated receptor-γ (Camacho et al., 2004; Sastre et al., 2006) for which the compound displays a K_i value of approximately 10 μM (G. Villetti, personal communication).

Plaque formation and maturation are normally accompanied by inflammatory processes. GFAP immunostaining suggested a reduction of the extension of glial filaments of Tg2576 mice chronically treated with CHF5074. This preliminary finding may indicate a decrease in the activation of glia. More detailed immunohistochemistry work with morphological evaluation and quantification would be required to confirm this hypothesis, because there is no evidence in a reduction in number of glial filaments and their total area.

Our in vitro studies in H4swe cells confirmed previous studies showing that CHF5074 preferentially inhibits Aβ42 versus Aβ40 secretion (Imbimbo et al., 2007). However, the estimated in vitro potency of CHF5074 on Aβ42 secretion was considerably higher than in previous studies (3.6 versus 40.0 μM). This 10-fold difference is probably due to the absence of fetal bovine serum in medium of the present experiments, which avoided avid binding of CHF5074 to proteins contained in fetal bovine serum. We do not know which of the two IC₅₀ values are the most relevant in vivo. The 5-fold preferential inhibitory activity of CHF5074 on Aβ42 observed in vitro was not observed in vivo. In Tg2576 mice of 12 to 15 months of age, diffuse plaques are mainly formed of Aβ42 (Kawarabayashi et al., 2001), meaning that more Aβ42 than Aβ40 is accumulated in the brains of these animals. This may counterbalance the higher potency of CHF5074 in preferentially inhibiting Aβ42 secretion. Alternatively, it could simply be that in vitro selectivity of CHF5074 is not high enough to translate in an in vivo measurable effect.

CHF5074 was very well tolerated in young Tg2576 mice when administered in the diet at 375 ppm for 17 weeks (estimated assumed dose of 61 mg/kg/day). No deaths occurred during the 17-week treatment period. Body weight and body weight gain over time were very similar to those of the vehicle-treated group. At week 17, macroscopic and histological examinations of potential peripheral target organs did not show significant abnormalities. In the ileum, there were no differences in goblet cell numbers and distribution between the vehicle-dosed animals and the animals receiving CHF5074. The spleen of the CHF5074-treated animals did not show histological abnormalities. Finally, histological evaluation of other portions of the gastrointestinal tract revealed no abnormalities. Overall, our data demonstrate that chronic treatment with CHF5074 can significantly reduce the induction of amyloid deposits in the brains of Tg2576 mice without causing histopathological changes in peripheral organs that would be associated with inhibition of processing of alternative substrates by this enzyme complex. Indeed, different reports have shown that the subset of NSAIDs acting as γ-secretase modulators can distinguish between the proteolytic activities on APP and Notch (Weggen et al., 2001, 2003b). This characteristic differentiates γ-secretase modulators from traditional γ-secretase inhibitors that in addition to APP may affect processing of alternative substrates for γ-secretase, particularly the Notch family of receptors. Genetic deletion of the Notch pathway in mice results in defects in the gastrointestinal tract with a phenotype similar to that observed using a γ-secretase inhibitor (van Es et al., 2005). Intestinal goblet cell metaplasia and cell population changes in the ileum have been observed with several γ-secretase inhibitors (Searfoss et al., 2003; Milano et al., 2004; Wong et al., 2004). In addition, γ-secretase inhibitors have been associated to alteration of lymphopoiesis (Wong et al., 2004) and atrophy of the thymus (Hadland et al., 2001). Toxicity of the γ-secretase inhibitors may be linked to the specific chemical structure, since it has been shown that MRK-560, a potent APP and Notch cleavage inhibitor (IC₅₀ values = 4.32 and 3.44 nM, respectively), administered for 3 months to aged Tg2576 mice significantly reduced brain Aβ pathology without causing histopathological changes in the ileum, spleen, or thymus (Best et al., 2007). Our present studies in HEK293swe cells indicated that CHF5074 concentrations inhibiting APP processing (5 μM) do not affect Notch processing. Previous studies in H4swe cells have shown that CHF5074, at the concentration of 100 μM, did not inhibit the expression of 20 Notch intracellular domain-responsive genes (Imbimbo et al., 2007). The lack of toxic effects on gastrointestinal tissues and on spleen observed in this study seems to confirm the good selectivity on CHF5074 for APP cleavage. Further formal chronic toxicological studies are needed to confirm these preliminary findings.

In conclusion, this study shows that 4-month treatment with CHF5074 in the diet (61 mg/kg/day) seems to be safe and well tolerated in aged Tg2576 transgenic mice. CHF5074 treatment markedly reduced total brain Aβ40 and Aβ42 levels and plaque burden in cortex and hippocampus, both in terms of plaque area fraction and number of plaques. CHF5074 is therefore a novel γ-secretase modulator that has the potential to be a safe and promising therapeutic agent for AD treatment.