Functional gamma-secretase inhibitors reduce beta-amyloid peptide levels in brain


and reprint requests should be addressed to Dale B. Schenk, Elan Pharmaceuticals, Inc., 800 Gateway Boulevard, South San Francisco, CA 94080, USA, e-mail:, or Patrick C. May, Lilly Neuroscience Division, Eli Lilly and Co., Lilly Corporate Center, Indianapolis, IN 46285, USA, e-mail:


Converging lines of evidence implicate the beta-amyloid peptide (Aβ) as causative in Alzheimer's disease. We describe a novel class of compounds that reduce Aβ production by functionally inhibiting γ-secretase, the activity responsible for the carboxy-terminal cleavage required for Aβ production. These molecules are active in both 293 HEK cells and neuronal cultures, and exert their effect upon Aβ production without affecting protein secretion, most notably in the secreted forms of the amyloid precursor protein (APP). Oral administration of one of these compounds, N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester, to mice transgenic for human APPV717F reduces brain levels of Aβ in a dose-dependent manner within 3 h. These studies represent the first demonstration of a reduction of brain Aβin vivo. Development of such novel functional γ-secretase inhibitors will enable a clinical examination of the Aβ hypothesis that Aβ peptide drives the neuropathology observed in Alzheimer's disease.

Abbreviations used

beta amyloid peptide


Alzheimer's disease


amyloid precursor protein


N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester


Dulbecco's modified Eagle medium


dimethyl sulfoxide


inhibitory concentration yielding 50% reduction


liquid chromatography mass spectrometry mass spectrometry


multireaction monitoring


presenilin 1


presenilin 2


structure activity relationship


sodium dodecyl sulfate

The central tenet of the amyloid hypothesis of Akheimer's disease (AD) is that the amyloid β-peptide (Aβ) is the prime causative agent of the disease process (Selkoe 1994). At autopsy, brain tissue of AD patients is defined by the large number of amyloid plaques that are composed primarily of Aβ (Wisniewski and Silverman 1997). The peptide is derived through proteolytic cleavage events from the amyloid precursor protein (APP) (Selkoe et al. 1996). A significant number of missense mutations that co-segregate with familial forms of AD reside either in APP itself or in presenilin 1 (PS-1) and presenilin 2 (PS-2) (Hardy 1996). These mutations share the common feature that they result in overproduction of the 42-amino-acid form of the Aβ peptide (Aβ42) (Citron et al. 1992; Suzuki et al. 1994; Borchelt et al. 1996), and have implicated Aβ generation and clearance as important potential therapeutic targets for AD.

Aβ peptide is formed from the APP molecule as a consequence of the action of two proteolytic activities, termed β- and γ-secretase, cleaving at the amino-terminus and carboxy-terminus of the peptide, respectively (Schenk et al. 1995; Selkoe et al. 1996). Although the protease responsible for γ-secretase activity has not yet been unambiguously identified, several groups have recently published the sequence of an enzyme with β-secretase activity (Sinha et al. 1999; Vassar et al. 1999; Yan et al. 1999). An additional proteolytic pathway exists wherein APP is cleaved in the middle of the Aβ peptide region by an α-secretase, thus abrogating the production of Aβ and producing secreted APPα (Esch et al. 1990). Inhibitors of either β- or γ-secretase activities have been suggested to theoretically reduce Aβ production.

Although most sporadic forms of AD are probably not a result of overproduction of Aβ (Mayeux et al. 1999), inhibition of peptide production might likely result in a reduction of the plaque burden. This pharmacological approach to AD is reminiscent of that taken with atherosclerosis wherein rare mutations, leading to the overproduction of circulating levels of cholesterol, result in disease; yet therapeutic treatments to reduce cholesterol levels, even in patients lacking such mutations, lead to clinical improvement (Watts and Burke 1996; Ballantyne et al. 1997).

Ultimately, testing the amyloid hypothesis in the clinic requires molecules capable of entering the CNS and producing a sustained reduction in brain Aβ levels. Although a number of pharmacological agents affect Aβ production in tissue culture (Higaki et al. 1995; Wolfe et al. 1998), the reduction of brain Aβ levels has not yet been reported. While several compounds inhibit the γ-secretase cleavage of APP in cultured cells (Higaki et al. 1995; Klafki et al. 1995; Citron et al. 1996; Yamazaki and Ihara 1998; Figueiredo-Pereira et al. 1999), these molecules are either low-potency peptide aldehydes or relatively non-specific protease inhibitors with no utility for testing in either animal or clinical studies. In this report, we describe a novel class of small molecule inhibitors of γ-secretase activity that potently block Aβ production in vitro. When a prototypical compound from this drug-like series is administered to PDAPP mice (Games et al. 1995), a significant reduction of brain Aβ is achieved within 3 h. This is the first report demonstrating an in vivo reduction of brain Aβ. Compounds derived from this class of novel functional γ-secretase inhibitors should ultimately yield drug candidates that will permit a direct examination of the amyloid hypothesis in AD patients.

Materials and methods

Compound synthesis

Compounds were synthesized as described in patent applications WO9822441-A2 and WO9822494-A2 filed by Athena Neurosciences, Inc. (Elan Pharmaceuticals Inc.) and Eli Lilly and Co.

In vitro Aβ reduction assays

Human embryonic kidney cells (American Type Culture Collection CRL-1573), transfected with the gene for APP751 (HEK 293) were used for routine Aβ reduction assays. The Aβ peptides secreted from these cells have been characterized previously (Dovey et al. 1993). Cells were plated in 96-well plates (Costar 3595, Corning, NY, USA) and allowed to adhere overnight in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum. For compound screening and dose–response testing, compounds were diluted from stock solutions in dimethylsulfoxide (DMSO) to yield a final concentration equal to 0.1% DMSO in media. Cells were pre-treated for 2 h at 37°C with compounds, media were aspirated off and fresh compound solutions applied. After an additional 2-h treatment period, conditioned media were drawn off and analyzed by a sandwich ELISA (266–3D6) specific for total Aβ (Seubert et al. 1992). Reduction of Aβ production was measured relative to control cells treated with 0.1% DMSO and expressed as a percentage inhibition. Data from at least six doses in duplicate were fitted to a four-parameter logistical model using XLfit software (IDBS, Emeryville, CA, USA) in order to determine potency.

Human and PDAPP mouse neuronal cultures were prepared as described previously (Seubert et al. 1992; Seubert et al. 1993). Cultures were grown in serum-free media to enhance their neuronal characteristics, and appeared to be greater than 90% neurons after maturation prior to use. Conditioned media to establish baseline Aβ values were collected by adding fresh media to each well and incubating for 24 h at 37°C in the absence of compound. Cultures were then treated with fresh media containing compound at the desired range of concentrations for an additional 24 h at 37°C, and conditioned media collected. For the measurement of total Aβ, samples were analyzed with the same ELISA (266–3D6) as used for the HEK 293 cell assays described above. Analyses of samples for Aβ42 production were performed by a separate ELISA (21F12–3D6) that utilizes a capture antibody specific for the Aβ42 C-terminus. Inhibition of production for both total Aβ and Aβ42 were determined by the difference between the values for the compound treatment and baseline periods. After plotting percentage inhibition versus compound concentration, data were analyzed with XLfit software, as above, to determine potency.

Immunoprecipitation of APP fragments

Metabolic labeling and immunoprecipitation were carried out as described previously (Oltersdorf et al. 1990). Cells (HEK 293) were plated in 6-well plates (Coming 25810, Coming, NY, USA) and grown until nearly confluent. Cells were pre-treated with compound in media with 0.1% DMSO for 2 h before changing to a fresh compound solution in methionine-free media containing 10% dialyzed serum and 100 µCi/mL L-[35S]methionine for an additional 2 h. After radiolabeling, conditioned media were removed and immunoprecipitated with R1282 antiserum (Haass et al. 1992) to analyze secreted APPα and Aβ. Cells were lyzed in 50 mm Tris: pH 8.0, 0.15 m NaCI, 20 mm EDTA, 1% sodium deoxycholate, 1% Triton X-100 and 0.1% sodium dodecyl sulfate (SDS). Crude lysates were centrifuged at 10 000 g at 4°C for 5 min to remove nuclei and debris. Cleared cell lysates were normalized for protein content and immunoprecipitated with αBX6 (Oltersdorf et al. 1990) to assess the degree of cell-associated 10- and 12-kDa carboxy-terminal fragments of APP and the maturation of full-length APP.

To assess these fragments in cortices from PDAPP mice, frozen brain samples were solubilized in Laemmli sample buffer without mercaptoethanol, and sheared through a 27-gauge needle. Protein concentrations were normalized to a constant concentration of 0.8 mg/mL and mercaptoethanol was added to 5%. Samples were electrophoresed on 10–20% tricine SDS gels and analyzed by immunoblotting with αBX6 (Oltersdorf et al. 1990). The identity of the 12-kDa carboxy-terminal β-secretase-cleaved product was confirmed by reactivity to antibody 3D6, which requires the first five amino acids of Aβ to be exposed (data not shown).

In vivo Aβ reduction studies

All studies were conducted with three- to four-month-old heterozygous PDAPP transgenic mice overexpressing the APPV717F mutant form of the amyloid precursor protein (Games et al. 1995). These animals have been previously shown to exhibit many of the neuropathological features of AD, and to produce high levels of Aβ in a regionally specific manner (Johnson-Wood et al. 1997).

Each treatment group (n = 10) consisted of equal numbers of age-matched male and female animals that were fasted overnight prior to treatment. Both treatment and control groups were dosed at a volume of 10 mL/kg with compound formulated in corn oil, 5% (v/v) ethanol or vehicle alone. Tissues were processed and all Aβ and APP measurements were made as described previously (Johnson-Wood et al. 1997). After removal of the brain, the cortex from one hemisphere was homogenized, extracted with 5 m guanidine, 50 mm Tris – pH 8.0, centrifuged, and the supernatant was used for Aβ measurements. Cortex from the other hemisphere was snap frozen for analysis of compound levels. Aβ levels were expressed as ng/g of wet tissue weight, and percentage reductions were calculated relative to the mean Aβ level of tissue from vehicle-treated control animals. Data were analyzed with Mann–Whitney non-parametric statistics to assess significance.

Measurement of compound levels in tissue

Analysis of compound concentrations in brain tissue was performed by liquid chromatography mass spectrometry mass spectrometry (LC/MS/MS) analysis following organic solvent extraction of brain homogenates. Cortex samples were homogenized in guanidine as described for the Aβ measurements above. Compound was extracted from the guanidine homogenates with hexane 2% (v/v) butanol containing an internal standard. For quantitation, a standard curve was constructed with samples containing compound added to brain homogenates of control animals that were extracted in parallel. The organic layer was evaporated to dryness and reconstituted in mobile phase (35% methanol, 0.1% formic acid). Samples were injected onto the LC/MS/MS and quantitated with a multireaction monitoring (MRM) experiment. The analysis was performed with a PE-Sciex, API 300 triple quadruple mass spectrometer (Applied Biosystems, Foster City, CA, USA) with a Turbolon Spray interface, equipped with dual pumps (Shimadzu 10 A, Columbia, MD, USA) and a Perkin-Elmer series 200 autosampler (Perkin-Elmer Analytical Instruments, Norwalk, CT, USA). Separation was performed with a Hypersil BDS-C18 (2 × 50 mm, 3 µm) column (Keystone, Bellefonte, PA, USA). Data acquisition (sample control versus 1.2) and analysis (multiview versus 1.2) was carried out with PE-Sciex software on a Macintosh 8500.


Compound screening and cellular Aβ reduction assay

Human kidney 293 cells overexpressing human APP751 (HEK 293) were previously characterized for their ability to secrete Aβ (Dovey et al. 1993). These cells secrete both the 40 and 42 residue forms of the peptide, as well as other minor truncated peptides. Approximately 25 000 compounds were screened for their inhibition of cellular Aβ production using a sandwich ELISA, which detects the predominant forms of Aβ (Seubert et al. 1992). Active compounds were secondarily tested for selectivity to ensure a lack of effect on inhibition of [35S]-methionine incorporation into total protein. A chemical optimization program was initiated around one active compound, the N-arylalanine ester (1), which emerged from this screen (Fig. 1). Numerous simple ester and amide analogs were evaluated, leading to identification of the isobutyl ester (2). Formation of the amide homologous to the dipeptide (5) led to nanomolar active compounds in the cellular assay for the N-(3,4-dichlorophenyl)alanine series.

Figure 1.

Potency determinations for the ability of compounds to reduce the total Aβ production from HEK 293 cells. Results show the progression of the SAR studies that led from the initial screening lead, compound 1, to the most active, compound 7 (DAPT), which was administered in the subsequent in vivo efficacy studies. Note that the (S)-enantiomer, DAPT, is greater than 1000-fold more potent than its corresponding (R)-enantiomer, compound 8.

The 3,4-dichlorophenyl group was also the focus of considerable chemical modification. The activity of the naphthalene analog (3) led to the synthesis of compound 4, wherein the phenylacetyl group was applied as a novel isostere for the naphthalene group. The resulting compound 4 maintained low µm cellular potency. Further optimization led to compound 7 {N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT)}, a compound which was evaluated in the PDAPP mouse.

Concentration-response analyzes for selected compound effects upon Aβ production in HEK 293 cells are shown in Fig. 2, and demonstrate that two orders of magnitude were gained in cellular potency. These data show that complete inhibition of Aβ production could be achieved with these compounds during the 4-h course of treatment. Figure 3 presents the concentration response for DAPT upon both total Aβ and Aβ42 production in human primary neuronal cultures. Both measures of Aβ production are similarly inhibited with potencies (Aβ total IC50 = 115 nm, Aβ42 IC50 = 200 nm) of 5–10-fold lower than is observed in HEK 293 cells. Similar dose responses were seen with PDAPP neuronal cultures suggesting that somewhat higher compound levels may be required to affect Aβ production from neuronal sources.

Figure 2.

HEK 293 cell dose–response curves for compound 1 (●), compound 2 (▪), compound 6 (▴) and DAPT (◆) show the composite data from which the IC50s were determined. Data points are the mean of multiple determinations (from n = 2 to n > 100). Compounds were diluted from DMSO stock solutions to yield final concentrations in media with 0.1% DMSO. After pre-treatment for 2 h at 37°C, media were aspirated off and fresh compound solutions applied. Following a 2-h treatment period, conditioned media were analyzed with a sandwich ELISA specific for total Aβ. Reduction of Aβ production was measured relative to control cells treated with 0.1% DMSO and expressed as a percentage inhibition. Data from at least six doses in duplicate were fitted with a four-parameter logistical model to measure IC50s using XLfit software (IDBS, Emeryville, CA, USA).

Figure 3.

Dose–response curves for DAPT on human neuronal cultures. Following 24-h treatment with compound, total Aβ (●) and Aβ42 (▪) measurements made with specific ELISA assays were compared with baseline data from conditioned media collected for 24 h prior to treatment in order to generate percentage inhibition data. Data were fitted with a four-parameter logistical model to measure IC50s using XLfit software (IDBS, Emeryville, CA, USA).

Mechanistic studies of APP fragments

DAPT was used to probe the mechanism of action involved in the inhibition of Aβ secretion. Metabolic labeling analyzes were performed on HEK 293 cells pre-labeled with [35S]-methionine and treated with various concentrations of DAPT, followed by immunoprecipitation with site-specific APP antibodies. Figure 4(a) demonstrates that DAPT inhibits the appearance of Aβ into conditioned media in a concentration-dependent manner as revealed by immunoprecipitation with a polyclonal anti-Aβ antibody (R1282) (Haass et al. 1992). This antibody also recognizes APPα, the secreted form of APP that contains 16 amino acids of Aβ at its C-terminus. Figure 4(a) indicates that secretion of APPα was largely unaffected by DAPT, although some modest increases seem to be elicited at the highest concentrations. Similar immunoprecipitation studies with APPβ-specific antibodies showed that concentrations of DAPT, which inhibited Aβ production over 90%, effected only a modest (20%) reduction in APPβ in the culture media (data not shown).

Figure 4.

(a) HEK 293 cells treated with increasing concentrations of DAPT secrete decreasing levels of Aβ into the conditioned media. Aβ was immunoprecipitated from the conditioned media from HEK 293 cells with R1282 antiserum after metabolic labeling with [35S]-methionine for 2 h in the presence of DAPT. Lanes 1–6 = 0, 0.4, 2, 10, 50 and 250 nm DAPT, respectively. While secreted Aβ was fully inhibited in a dose-dependent manner, secreted APPα remained largely unaffected; bands were assessed with a Personal Densitometer (Molecular Dynamics, Sunnyvale, CA). (b) HEK 293 cells treated with increasing concentrations of DAPT show an increased stabilization of 10- and 12 kDa carboxy-terminal fragments of APP in the cell lysates, indicative of interference with γ-secretase processing. Carboxy-terminal fragments were immunoprecipitated from the lysates of HEK 293 cells with αBX6 antiserum after metabolic labeling with [35S]-methionine for 2 h in the presence of DAPT. Lanes 1–6 = 0, 0.4, 2, 10, 50 and 250 nm DAPT, respectively. The dose-dependent stabilization of the carboxy-terminal fragments of APP, without effect on full-length APP, suggests a mechanism of action involving the inhibition of functional γ-secretase activity either directly or indirectly.

To further examine the effects of DAPT on APP processing, PDAPP neuronal cultures were treated for 24 h and the conditioned media were analyzed for total Aβ, Aβ42, APPα and APPβ with specific ELISA assays, which have been characterized previously(Seubert et al. 1993). In addition, to assess the reversibility of treatment, the cultures were washed free of external compound and conditioned media collected for a subsequent 24-h period. The results of these studies are summarized in Fig. 5. Whereas total Aβ and Aβ42 were inhibited in a concentration-dependent and reversible manner by DAPT (Figs 5a and b), APPα appears unaffected by DAPT treatment (Fig. 5c). Although APPβ was reduced by about 30% by DAPT treatment, this effect was not concentration-dependent and was reversed by the removal of the compound (Fig. 5d). Thus, although APPβ processing may be slightly diminished by DAPT, inhibition of either β-secretase or α-secretase activity does not appear to be the primary mechanism of action of this compound series on Aβ production.

Figure 5.

PDAPP mouse neuronal cultures treated with DAPT for 24 h (shaded gray) followed by the removal of compound for an additional 24 h (shaded black). Using specific ELISA assays, several products of APP processing were followed: (a) total Aβ, (b) Aβ42, (c) APPα and (d) APPβ. ELISA data following 24-h treatments were compared with baseline data from conditioned media collected for 24 h prior to treatment in order to generate percentage control data.

If the primary mechanism of action of these compounds is via functional inhibition of γ-secretase, then both the 10- and 12-kDa carboxy-terminal fragments of APP should accumulate in cells as result of a lack of γ-secretase cleavage of these fragments. Consistent with this proposed mechanism, treatment of HEK 293 cells with DAPT leads to a concentration-dependent increase in both the 10- and 12-kDa fragments subsequently immunoprecipitated from cell lysates (Fig. 4b). All compounds from this series had a similar profile when analyzed by this type of APP immunoprecipitation in HEK 293 cells, untransfected, primary cultured human neurons, as well as primary neurons from PDAPP mice (data not shown). Collectively these findings argue that this class of compounds reduce Aβ production by the inhibition of γ-secretase activity, but cannot distinguish whether such inhibition is direct or indirect.

Aβ reduction studies in transgenic mouse brain

To assess in vivo efficacy, DAPT was administered to PDAPP mice (100 mg/kg s.c.) and the levels of DAPT and Aβ were examined in the brain cortex. Peak DAPT levels of 490 ng/g were achieved in the brain 3 h after treatment (Fig. 6), and levels greater than 100 ng/g (∼200 nm) were sustained throughout the first 18 h. These brain concentrations of DAPT are in excess of its IC50 for lowering Aβ in neuronal cultures (115 nm),and resulted in a robust and sustained pharmacodynamic effect in vivo. Specifically, the reduction of Aβ in the cortex paralleled compound levels with a peak inhibition of 40% occuring 3 h after treatment. A reduction of greater than 20% was observed up to 18 h after treatment. To assess selectivity in vivo, mice were administered DAPT, or its less active enantiomer (compound 8), at 100 mg/kg s.c. and killed 3 h later. In contrast to the robust efficacy observed with DAPT (45% reduction of cortical Aβ), its enantiomer yielded only a marginal 2% reduction (Fig. 7).

Figure 6.

Following subcutaneous administration of 100 mg/kg DAPT to PDAPP mice, compound levels in the brain (▪) exceed 100 ng/g within 1 h and persist up to 18 h after administration, with peak levels of 490 ng/g observed after 3 h. Similarly, the cortical total Aβ levels (▴) are significantly reduced between 1 and 18 h following treatment (Mann–Whitney non-parametric analysis, p < 0.05, n = 10), with a peak reduction of 40% measured after 3 h. Aβ data are presented as percentage reduction relative to the mean of two vehicle-treated control groups killed 1 and 24 h after administration. All data are the mean of data from 10 animals per group. Reduction of brain Aβ levels closely reflected the amount of compound in the brain, suggesting that soluble Aβ is rapidly turned over in these animals.

Figure 7.

Enantiospecific effects upon cortical total Aβ levels in PDAPP mice. Young PDAPP mice were subcutaneously administered 100 mg/kg of either DAPT or its enantiomer, compound 8. Mice were killed 3 h later and cortical guanidine homogenates assessed for total Aβ. Results are expressed as percentage reduction of Aβ levels relative to a vehicle-treated control group (mean + SEM, n = 10). Significance was assessed with Mann–Whitney non-parametric analysis (*p < 0.05).

Based on pilot studies that confirmed the peak reduction of Aβ levels at 3 h after oral administration, efficacy of DAPT was evaluated in PDAPP mice killed 3 h after receiving 10, 30 and 100 mg/kg of compound by oral gavage. DAPT reduced the cortical total Aβ in a dose-dependent manner with a 50% reduction occuring at 100 mg/kg (Fig. 8). Similar reductions were also observed in cortical Aβ42 3 h after the oral administration of DAPT (Fig. 8).

Figure 8.

The effect of DAPT on total Aβ (shaded gray) and AB42 (shaded black) levels in PDAPP mouse cortex 3 h after the oral administration of 10, 30 or 100 mg/kg. Both the total Aβ and Aβ42 were reduced in a dose-dependent manner. Cortical samples were processed as in Fig. 5 and Aβ levels measured with specific ELISAs. Results are expressed as the percentage reduction of Aβ levels relative to a vehicle-treated control group (mean + SEM, n = 10). Significance was assessed with Mann–Whitney non-parametric analysis (*p < 0.05).

These findings indicate that a substantial proportion of the soluble forms of Aβ are rapidly turned over in vivo, as changes in Aβ levels were observed within a few hours of compound administration, in accord with a previous report of a short half-life for soluble Aβ in the rodent brain (Savage et al. 1998). As anticipated, levels of APPα and APPβ in the brains of mice treated with DAPT remained relatively unaffected (data not shown). In addition, western blot analyzes of treated brain tissue showed a stabilization of carboxy-terminal APP fragments similar to that seen in cultured cells (Fig. 9), confirming that the compounds are reducing Aβ levels through the inhibition of γ-secretase-like activity in vivo, as well as in vitro.

Figure 9.

PDAPP mouse brain samples treated in vivo with DAPT show a stabilization of 10- and 12-kDa carboxy-terminal fragments of APP, indicative of interference with γ-secretase processing. Carboxy-terminal fragments were immunoprecipitated with αBX6 antiserum from the lysates of cortex after four doses of vehicle or DAPT subcutaneously over a 21-h period. Lanes 1 and 2 – vehicle treated, lanes 3 and 4 – 100 mg/kg DAPT, duplicate animals. The stabilization of the carboxy-terminal fragments of APP suggests a mechanism of action involving the inhibition of functional γ-secretase activity either directly or indirectly.


The role of Aβ in the pathogenesis of AD remains of central importance to emerging therapeutic strategies seeking to alter disease progression (Schenk et al. 1995). A complex and rich literature has emerged on the biological effects of the Aβ, particularly in tissue culture where the peptide is both neurotoxic and inflammatory (Mattson et al. 1992; Behl 1997; Mattson et al. 1998). Although rigorous proof that the peptide exerts these activities in the AD brain are difficult to obtain, increased inflammation and neuronal dysfunction are clearly apparent in various mouse models of the disease, driven by the overexpression of APP where increased Aβ is found (Frautschy et al. 1998; Hsiao et al. 1996; Masliah et al. 1996).

Regardless of the way in which Aβ exerts its deleterious effects on the brain, the amyloid hypothesis predicts that decreased production of the Aβ peptide in the brain will elicit a therapeutic benefit. The most direct way of reducing Aβ production is through the inhibition of the specific proteases driving peptide production. Although a large number of toxic compounds will reduce the secretion of many proteins and peptides, the class of compounds exemplified by DAPT effects only the secretion of the Aβ peptide without significantly affecting the secretion of either the α and β forms of APP. This finding, together with the observation that carboxy-terminal fragments of APP increase with treatment, suggests that the pharmacological action of these compounds is via the functional inhibition of γ-secretase.

Whether these compounds affect cleavage of other, putative γ-secretase targets, such as Notch (De Strooper et al. 1999; Steiner et al. 1999; Struhl and Greenwald 1999), is presently unclear. Nevertheless, efficacy was achieved in the absence of cellular toxicity, as determined by either total protein synthesis or general secretion, even when cells were treated continuously for 24 h and observed 24 h after the compound was removed. Thus, at least in vitro, functional blockade of γ-secretase leading to an accumulation of its substrates, e.g. carboxy-terminal fragments, did not result in cellular toxicity. In addition, there were no signs of in vivo toxicity in these studies.

The optimization of the initial lead compound 1 resulted in the identification of DAPT, a functional γ-secretase inhibitor. Upon parenteral administration of DAPT, reasonable blood and brain levels were attained and a significant, dose-dependent reduction of brain Aβ levels was achieved. Peak efficacy was observed at 3 h after treatment, but significant reductions in Aβ were maintained up to 18 h after treatment. The demonstration of a sustained pharmacodyamic effect on brain Aβ levels after a single acute administration of DAPT is an important first step in the in vivo testing of the amyloid hypothesis.

Future pre-clinical studies in PDAPP mice will be necessary to assess whether DAPT, or related functional γ-secretase inhibitors, are capable of altering β-amyloidosis and neuropathology upon prolonged treatment. In addition, these experiments will allow the assessment of any physiological or toxicological implications of chronic γ-secretase inhibition. Such chronic studies will be critical for validating the utility of functional γ-secretase inhibitors as therapeutic agents for the treatment of AD. Ultimately, compounds derived from this class of molecules should yield clinical candidates for directly testing whether the sustained reduction of Aβ levels in the brain will attenuate the cognitive decline and alter the disease progression in AD.