In vivo inhibition of Aβ production by memapsin 2 (β-secretase) inhibitors

Authors


Address correspondence and reprint requests to Jordan Tang, Protein Studies Program, Oklahoma Medical Research Foundation, 825 NE. 13th Street, Oklahoma City, OK 73104, USA.
E-mail: Jordan-Tang@omrf.ouhsc.edu

Abstract

We have previously reported structure-based design of memapsin 2 (β-secretase) inhibitors with high potency. Here we show that two such inhibitors covalently linked to a ‘carrier peptide’ penetrated the plasma membrane in cultured cells and inhibited the production of β-amyloid (Aβ). Intraperitoneal injection of the conjugated inhibitors in transgenic Alzheimer's mice (Tg2576) resulted in a significant decrease of Aβ level in the plasma and brain. These observations verified that memapsin 2 is a therapeutic target for Aβ reduction and also establish that transgenic mice are suitable in vivo models for the study of memapsin 2 inhibition.

Abbreviations used

amyloid-β

AD

Alzheimer's disease

APP

β-amyloid precursor protein

BBB

blood–brain barrier

CP

carrier peptide

DMSO

dimethyl sulfoxide

PBS

phosphate-buffered saline

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

The pathogenesis of Alzheimer's disease (AD) is intimately related to the presence of the neurotoxic amyloid-β peptide (Aβ) in the brain (Selkoe 1999; Walsh et al. 2002). Reduction of brain Aβ, which improved cognitive functions in transgenic AD mice (Morgan et al. 2000; Dodart et al. 2002), is therefore an important strategy for AD therapy (Dominguez and De Strooper 2002). Aβ, a peptide of 40/42 amino acid residues, is generated by the proteolysis of β-amyloid precursor protein (APP), first by the membrane-associated aspartic protease memapsin 2 (β-secretase or BACE; Lin et al. 2000; Sinha et al. 1999; Vassar et al. 1999; Yan et al. 1999), then by the intramembrane proteolysis of γ-secretase (Selkoe 1999). Both proteases are therefore therapeutic targets for the treatment of AD. The development of inhibitor drugs for memapsin 2 is encouraged by the lack of a deleterious response after gene deletion in mice (Cai et al. 2001; Luo et al. 2001; Roberds et al. 2001). However, to attain clinically useful memapsin 2 inhibitors is generally thought to be very challenging because they must be small enough (generally < 500 Da) to penetrate the blood–brain barrier (BBB) yet manifest high potency and other desirable pharmaceutical properties. Toward the development of such inhibitor drugs, it is important to demonstrate in cellular and animal models that memapsin 2 inhibitors can in fact achieve Aβ reduction. Not only would such results validate memapsin 2 as a therapeutic target, the cellular and animal models may also serve as future tools for drug development. Although several potent memapsin 2 inhibitors with Ki values in the nanomolar range have been reported (Ghosh et al. 2000; Ghosh et al. 2001; Turner et al. 2001; Ghosh et al. 2002), they are too large to traverse the BBB. Therefore, we sought a method by which to deliver a potent memapsin 2 inhibitor across the BBB for target validation studies. Here we describe the reduction of Aβ level in cells and in transgenic AD mice by the administration of relatively large memapsin 2 inhibitors linked to a ‘carrier peptide’ for cellular and BBB penetration.

Materials and methods

Synthesis of peptides and inhibitors

Conjugated inhibitor Fs[OM99-2]tat was synthesized at Research Genetics (Huntsville, AL, USA). Fs[OM00-3]DR9 and its peptide analogue lacking the transition state isostere (represented by Ψ; see Ghosh et al. 2000) were synthesized by SynPep (Dublin, CA, USA). Both OM00-3 and [OM00-3]DR9 were synthesized at the Molecular Biology Resource Facility, OUHSC. Fluorescein was conjugated to the amino terminus of OM00-3 by incubation with NHS-fluorescein (Pierce, Rockford, IL, USA). OM00-3, Fs[OM99-2]tat, Fs[OM00-3]DR9, and [OM00-3]DR9 incorporated the dipeptide transition-state isostere FMOC-LeuΨAla (Ghosh et al. 2000). All peptides were purified to > 90% by reversed-phase HPLC and dissolved in either dimethyl sulfoxide (DMSO) or phosphate-buffered saline (PBS) to concentrations of 50–100 mg/mL. Inhibition potency was determined according to established methods (Ermolieff et al. 2000).

Penetration of inhibitors in cell lines and animal tissues

Fluorescently labeled inhibitors or fluorescein (control) were incubated with suspended cells for time intervals ranging from 10 to 30 min. Cells were fixed with paraformaldehyde and permeabilized in 0.2% Tween-20 in PBS for 6 min and incubated with antifluorescein-Alexa 488 antibody (Molecular Probes, Eugene, OR, USA) in order to enhance detection of intracellular inhibitor present from penetration. Flow cytometry (FACSCalibur) and confocal fluorescent microscopy (Leica TCS NT) were performed at the Flow and Image Cytometry Laboratory, OUHSC.

For in vivo cell and BBB penetration studies, Cd72c mice (background strain) were injected intraperitoneally with 0.3–10 nmol of fluorescein-labeled conjugated inhibitor or fluorescein control, anesthetized and perfused with neutral-buffered formalin (10%) before harvesting tissues. For flow cytometry, cells were released by abrasion over a 30 µm nylon screen, permeabilized in 0.2% Tween 20 in PBS, blocked with 1% normal rabbit serum, and incubated with antifluorescein-Alexa 488 antibody for assessment of inhibitor penetration. For confocal microscopy (Imaging Core Facility, OMRF), brains were harvested and incubated in OCT/PBS for at least 16 h followed by embedding in TFM (Triangle Biomedical Sciences, Durham, NC, USA). Brain hemispheres were sectioned (10 µm) using a Leica CM3050 research cryostat, fixed in 0.25% paraformaldehyde for 15 min, and incubated with antifluorescein-Alexa 488 antibody.

Inhibition of Aβ production in cell lines

Cultured cells, including human embryonic kidney (HEK293) cells, HeLa, and the neuroblastoma line M17, obtained from American Type Culture Collection (ATCC), were stably transfected with both human APP Swedish mutant (APPsw) and human memapsin 2 genes. All cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum and 1% penicillin/streptomycin. Two antibiotics, Zeocin (1 µg/mL) and G418 (250 µg/mL) were included in the media for maintenance of the stably transfected lines.

Either the parental lines (293, HeLa, or M17) or the stably transfected lines (293-D, HeLa-D, or M17-D) were plated on six-well plates and grown in a 37°C, 5% CO2 incubator until 90% confluent. Cells were then treated with or without 10 pmol of Fs[OM00-3]DR9 overnight then labeled by using [35S]TransLabel Protein Labeling Mix (100 µCi/mL) (ICN) in methionine- and cysteine-free Dulbecco's modified Eagle's medium for an additional 18 h. For treatment of cells, 10 pmole of the conjugated inhibitor was dispensed to cells 20 min prior to labeling, and likewise into labeling media.

Cells were lysed in 1 mL of RIPA buffer (10 mm Tris, pH 7.6, 50 mm NaCl, 30 mm sodium pyrophosphate, 50 mm NaF, and 1% NP-40) supplemented with 1 mm phenylmethylsulfonyl fluoride, 10 µg/mL leupeptin, 2.5 mm EDTA, 1 µm pepstatin, and 0.23 U/mL aprotonin. The total cell lysates were subjected to immunoprecipitation with 1 µL of 1 mg/mL of monoclonal antibody raised specifically against human Aβ17−24 (MAB 1561, Chemicon) with 20 µL of protein G-sepharose beads. Immunoprecipitated proteins were analyzed by using 10–20% gradient sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) (Invitrogen) and radiolabeled proteins were visualized by autoradiography. Quantitative results were obtained using the STORM phosphorimaging system (Amersham).

Inhibition of Aβ production in AD mice

Transgenic Tg2576 mice (6–15 months of age) obtained from Taconic (Germantown, NY, USA) were injected intraperitoneally with 200 µg of [OM00-3]DR9 per 25 g body weight or equimolar amount of its peptide analogue as a control, diluted in 200 µL PBS. OM00-3 (Turner et al. 2001) was dissolved in DMSO to 50 mg/mL and likewise diluted into H2O. Blood was sampled from animals prior to injection and at time intervals following injection either from the orbital sinus of anesthetized animals or from the saphenous vein with collection into heparinized capillary tubes. Plasma was separated from cells and stored at − 70°C until analysis for Aβ40 and Aβ42 by sandwich ELISA (BioSource International). The brain was removed and one hemisphere was homogenized in RIPA buffer (Phiel et al. 2003; buffer prepared as described in the protocol of Santa Cruz Biotechnology http://www.scbt.com/support/protocol/01; 150 mg/mL wet weight) followed by sonication. After centrifugation at 100 000 × g, the pellet was further extracted with four volumes of a cold guanidine-HCl buffer (Johnson-Wood et al. 1997; 5 m guanidine-HCl in 50 mm Tris-HCl, using a protocol from Biosource International, Camarillo, CA, USA) and centrifuged. Supernatants were separately immunoprecipitated with monoclonal antibody MAB 1560 recognizing Aβ1−17 (Chemicon), subjected to SDS–PAGE on 10–20% gradient Tricine gels (Invitrogen) and transferred to polyvinylidene difluoride membranes for immunoblotting with MAB 1561 (Chemicon). For detection of C-terminal fragments of APPsw (C83 and C99), brain homogenates were immunoprecipitated with monoclonal antibody MAB 1561 recognizing Aβ17−24 (Chemicon) and immunoblotted with MAB 1561 or AB 5352 directed towards the cytosolic domain of APP (Chemicon). Additionally, RIPA brain homogenates were electrophoresed similarly and immunoblotted with antigen affinity-purified polyclonal antibody to memapsin 2 (Lin et al. 2000) for visualization of endogenous murine memapsin 2. This polyclonal antibody directed towards human memapsin 2 cross-reacted with murine endogenous memapsin 2 (98% amino acid identity in catalytic domain) as found in characterization of human memapsin 2 transgenic animals (data not shown). Bands from immunoblots were quantified by optical scanning using the PC software Image Gauge, version 3.46. All animal procedures followed guidelines for humane care and use of animals provided by the IACUC.

Results

Covalently linked ‘carrier peptides’ (CP) are known to facilitate the transport of proteins and DNA across the cell membrane (Fawell et al. 1994; Schwarze et al. 1999; Wender et al. 2000). We tested if CPs can assist the transport of memapsin 2 inhibitors into the cells and across the BBB. Two inhibitors (Fig. 1a), Fs[OM99-2]tat (Ki = 39 ± 4 nm) and Fs[OM00-3]DR9 (Ki = 1.7 ± 0.5 nm) were developed based on the previously reported transition-state inhibitors of memapsin 2, OM99-2 (Ki = 1.6 nm, Ghosh et al. 2000) and OM00-3 (Ki = 0.3 nm, Turner et al. 2001). Each contained a covalently linked FITC (Fs) group and a CP (Fig. 1a) of either a 12-residue tat fragment (Schwarze et al. 1999; Wender et al. 2000) or a nine-residue poly d-arginine (Wender et al. 2000), DR9. Incubation of HEK293 cells with Fs[OM99-2]tat resulted in an increase of fluorescence relative to cells incubated with fluorescein alone, as demonstrated by flow cytometry (Fig. 1b). Furthermore, the fluorescence intensity of the incubated cells correlated with the inhibitor concentration in the range of 4–400 nm. Conjugated inhibitor Fs[OM00-3]DR9 likewise penetrated cells, whereas Fs[OM00-3], without the CP moiety (DR9), did not (Fig. 1b), demonstrating that the CP was necessary for transporting the inhibitor across the plasma membrane. Confocal microscopy showed that cells incubated with Fs[OM99-2]tat were intensely fluorescent as compared to cells incubated with fluorescein alone (Fig. 1c), confirming that the conjugated inhibitor had entered the cells. Interestingly, the intracellular fluorescence intensity was not uniformly distributed (Fig. 1c) which suggests sequestration of the inhibitor within discrete subcellular compartments. The transport of conjugated inhibitors was observed in several cell lines including M17, a neuronal cell line (data not shown).

Figure 1.

Cell penetration by carrier peptide-conjugated memapsin 2 inhibitors. (a) Structures of conjugated inhibitors (Ψ, transition-state isostere; Ghosh et al. 2000). (b) Top row, flow cytometry analysis of HEK293 cells after incubation with Fs[OM99-2]tat or controls; bottom row, flow cytometry analysis of whole blood cells, splenocytes, and brain cells isolated from mice at indicated times after injection of Fs[OM99-2]tat or control Fs. (c) Fluorescence confocal microscopy images of HeLa cells incubated with Fs[OM99-2]tat or control Fs, and (d) of hemispheric sagittal brain sections isolated from mice at indicated times after injection of inhibitors or control Fs. Cells were stained with anti-Fs antibody for detection (scale bars: c, 10 µm; d, 20 µm).

The most beneficial site to inhibit the production of Aβ is clearly in the intracellular compartments of neurons in the central nervous system. Thus we tested the ability of conjugated memapsin 2 inhibitors to penetrate cells and BBB in vivo. Inhibitor Fs[OM99-2]tat was injected intraperitoneally in mice and fluorescence was clearly detected by flow cytometry or fluorescence microscopy (Fig. 1b,d) in brain cells and cells of other tissues, relative to those isolated from control animals injected with fluorescein alone. These results indicated that the conjugated inhibitor Fs[OM99-2]tat had also penetrated cells in mice. The presence of fluorescence in brain cells from in vivo experiments implies also the penetration of the BBB by the conjugated inhibitors. Inhibitor Fs[OM00-3]DR9 was more efficient in cell penetration both in vitro (Fig. 1b) and in vivo where the fluorescence of the inhibitor was observed in brain cells 3 h after injection (Fig. 1d), compared to 8 h required for observation of Fs[OM99-2]tat. These results confirm the previous report that DR9 is a better carrier than tat (Wender et al. 2000).

We investigated the inhibition of secreted Aβ by conjugated inhibitors in cultured HEK293 cells stably transfected with both memapsin 2 and Swedish mutant APP genes. In the presence of 10 pmol of Fs[OM00-3]DR9, both C99 (C-terminal 99 residues of APP resulting from cleavage by memapsin 2) and Aβ in the autoradiogram of SDS–PAGE were completely abolished. OM00-3 without the CP failed to inhibit either C99 or Aβ production in the same conditions (Fig. 2). These findings suggest that Fs[OM00-3]DR9 inhibited intracellular memapsin 2 activity.

Figure 2.

Memapsin 2 inhibitor [OM00-3]DR9 reduced Aβ and C99 (the C-terminal 99 residues of APP) produced in HEK293 cells. HEK293 cells, stably transfected with both Swedish mutant APP (APPsw) and memapsin 2 (Mep2), were metabolically labeled with [35S]Met in the presence or absence of [OM00-3]DR9. After 18 h, C99 and Aβ were immunoprecipitated, subjected to SDS–PAGE and visualized by autoradiogram (see arrows). Treatment with inhibitor OM00-3 that lacks the CP (lane 5) did not inhibit the production of C99 or Aβ.

We further studied the influence of the conjugated inhibitor on plasma Aβ levels in the Tg2576 transgenic AD mouse (Hsiao et al. 1996; Kawarabayashi et al. 2001). Nine mice in each group were injected intraperitoneally with various doses of inhibitor [OM00-3]DR9 (Fig. 1a, Ki = 1.7 ± 0.5 nm). The Fs group was omitted in this inhibitor since it was employed as a reporter group to track inhibitor penetration in the initial experiments. At 2 h following the injection, plasma Aβ40 showed a significant dose-dependent reduction relative to Aβ40 from control mice injected with PBS (Fig. 3a). To study the duration of inhibition, eight Tg2576 mice were injected intraperitoneally with inhibitor [OM00-3]DR9. The plasma Aβ40 level dropped to about one-third of the initial value at 2 h following injection, consistent with presence of Fs[OM00-3]DR9 in the brain in the same range of time (Fig. 1d). The inhibition had a relatively short half-life near 3 h, with the plasma Aβ40 level then gradually rising to the initial value by 8 h (Fig. 3b), consistent with the observed disappearance of fluorescent inhibitor Fs[OM00-3]DR9 from brains of mice at 8 h postinjection (data not shown). Collectively, these data established a dosage and a course of reduction and recovery of plasma Aβ40 by memapsin 2 inhibition in vivo. Injecting either the unconjugated inhibitor OM00-3 (Fig. 3b) or the peptide analogue of [OM00-3]DR9 without the transition-state isostere (Fig. 3c) did not reduce plasma Aβ40 levels. The latter established that the CP was not responsible for the observed inhibition. Injection of a mixture of the peptide analogue of [OM00-3]DR9 and the inhibitor OM00-3 did not decrease plasma Aβ40 (Fig. 3c), indicating that CP must be linked to the inhibitor to achieve an in vivo inhibition. We also demonstrated that the percentage of Aβ40 relative to total Aβ (sum of Aβ40 and Aβ42) was constant at 73 ± 8% and 75 ± 5% for Aβ levels ranging from nearly 1000 to over 5000 pg/mL in treated and untreated animals, respectively. These observations established that the measured Aβ40 changes may be taken as the change of total Aβ in the observed range of inhibition. As the observed duration of inhibition had been relatively short, we sought to establish a maximal inhibition level of this inhibitor by multiple injections. Experiments with four injections at 2-h intervals in mice of 6–9 months of age significantly inhibited the Aβ40 production to an average high of 71% (ranging 67% to 78%) of the initial value (Fig. 3d). The difference in Aβ40 values of the experimental group and the control group receiving PBS or the peptide analogue of [OM00-3]DR9 were statistically significant at time points 2 h following a given injection. We also tested the inhibition of Aβ40 in older Tg2576 animals (age 12–15 months) treated with [OM00-3]DR9 over a 6-h period. Significant inhibition (reduction to 19% and 6%, respectively) of Aβ40 was seen in plasma sampled at both 2.5 h and 6.5 h following the first administration (Fig. 3d).

Figure 3.

Memapsin 2 inhibitor [OM00-3]DR9 reduced plasma Aβ40in vivo. (a) Reduction is dosage-dependent (asterisks, p < 0.01) and optimal at 2 h (b) Time course of plasma Aβ40 in mice injected separately with [OM00-3]DR9 (•), DMSO (◆) and unconjugated OM00-3 (▴). While [OM00-3]DR9 produced a maximal inhibition at 2 h, the controls did not produce a significant effect (asterisks, p < 0.05, double asterisks, p < 0.01). (c) Plasma Aβ40 was reduced only by conjugated inhibitor [OM00-3]DR9 (200 µg per 25 g body weight; asterisk, p < 0.01), but not by controls that included unconjugated OM00-3, the peptide analogue of [OM00-3]DR9 (labeled peptide), or both. (d) Plasma Aβ40 from mice receiving repeated injection (arrows) of [OM00-3]DR9 was decreased relative to controls (asterisks, p < 0.01). Solid data points are from mice in the age range of 6–9 months injected with PBS (▴), peptide analogue of ELDLAVEF (◆) and [OM00-3]DR9 (•), respectively. Open data points are from mice in the age range of 12–15 months injected with PBS (▵) and [OM00-3]DR9 (○), respectively.

We further studied the influence of the conjugated inhibitor on Aβ levels in the AD Tg2576 mouse brain. After four administrations of [OM00-3]DR9 in 2-h intervals, the brain tissues of Tg2576 mice (12–15 months of age) that had demonstrated significant inhibition of plasma Aβ (Fig. 3d) were recovered and extracted sequentially by RIPA buffer and a guanidine buffer. These extractions were designed to recover soluble Aβ and more solidly deposited and less soluble Aβ, respectively (Phiel et al. 2003). Western blot analysis shows that the brain Aβ extracted by RIPA buffer from the inhibitor treated group was significantly reduced (examples from a typical pair is shown in Fig. 4a). Quantification by scanning of the bands revealed that the Aβ was inhibited by about 70% of the untreated control (n = 6, p = 0.02) (Fig. 4b). ELISA analysis of Aβ40 in RIPA extracts produced similar results (n = 6, p = 0.05) Fig. 4(b). The amount of Aβ40 in guanidine extracts did not show a significant difference as a result of inhibitor treatment, although a slight reduction was seen in the western blot analysis (Fig. 4b). These observations suggest that the conjugated inhibitor had significantly suppressed the Aβ level in RIPA buffer extracted tissue, which represents the more recently produced Aβ in the brain of Tg2576 transgenic mice.

Figure 4.

Effect of conjugated inhibitors on Aβ in Tg2576 mouse brain extracts. Immunoblotting of Aβ of from control mice (receiving injection of PBS, n = 6), and mice receiving four injections of [OM00-3]DR9 (n = 6). Tissue was successively extracted with detergent-buffer (RIPA) and 5 m guanidine-HCl buffer (guanidine) (see Methods) then subjected to SDS–PAGE and immunoblotting. (a) Electrophoretic bands of brain Aβ from representative mice. (b) Upper panel: Bar graphs represent the quantitation from optical scanning of Aβ bands from immunoblotting (see a). Left side, Aβ in RIPA-buffer extracts shows significant difference of Aβ levels (asterisk, p = 0.02). Right side, Aβ in subsequent guanidine-buffer extracts (p > 0.16). Lower panel: Bar graphs represent the same as above except that the Aβ determinations were done with ELISA analysis. The difference of Aβ in RIPA extracts is statistically significant (asterisk, p < 0.05).

Aβ levels were significantly reduced in both brain and blood of Tg2576 animals upon treatment with [OM00-3]DR9 (Figs 3 and 4), likely attributed to blocking of β-secretase activity in vivo. To exclude other factors that could potentially affect Aβ level, we examined the expression levels of both endogenous memapsin 2 and full-length human APPsw and observed no significant influence of inhibitor on the expression of either proteins (Fig. 5a,b). To assess the influence of inhibitor on the activity of α-, β- and γ-secretases, we also examined APP fragment C99 (product of β-secretase) and C83 (product of α-secretase) in the brain extracts of mice treated with the inhibitor and the controls. A substantial decrease in the level of C99 was observed in treated animals, evident in western blots (Fig. 5c). The quantification of bands showed that the reduction of C99 by the inhibitor averaged 55% (Fig. 5d; p-value = 0.035), which agreed well with the significant Aβ reduction observed in the brain homogentates of the inhibitor treated mice (Fig. 4b). The observation of C99 reduction indicated that the inhibition of Aβ production was indeed in β-secretase but not in γ-secretase activity that would have resulted in the accumulation of C99. C83 bands in the Western blot of C99 using antibody MAB 1561 (see Materials and methods) were very faint (Fig. 5c). We found that they could be enhanced using another antibody (see Materials and methods). The resulting western band of C83 did not show significant difference between the brain extracts of inhibitor treated mice and the controls (Fig. 5d; p-value = 0.36). These observations indicated that the brain α-secretase activity was not significantly affected by the administration of [OM00-3]DR9 in mice.

Figure 5.

Memapsin 2 inhibitor [OM00-3]DR9 specifically inhibits β-secretase activity in vivo. Tg2576 animals (3 months old) were treated with either [OM00-3]DR9 or PBS as a control (see Materials and methods) with seven injections every hour. Brain hemispheres were harvested and the RIPA buffer homogenates were electrophoresed and immunostained for (a) memapsin 2 or (b) APPsw protein. (c) Homogenates were immunoprecipitated and immunostained with MAB 1561 (see Materials and methods) to reveal C99 and C83. (d) The quantification of C99 (solid bars) and C83 (open bars) represents immunoblots probed with MAB 1561 or with AB 5352, respectively (data not shown) and indicated that administration of inhibitor [OM00-3]DR9 significantly reduced C99 (p < 0.05, n = 8) but not C83 (p > 0.05, n = 5).

Discussion

Memapsin 2 inhibitors are particularly attractive candidates for Aβ reduction therapy since its cleavage of APP represents the initial step in the biogenesis of Aβ. The inhibition of this step would lead to the elimination of all steps in the pathogenesis of AD. In vivo inhibition of Aβ production in the brain of AD mice by a memapsin 2 inhibitor has not been previously achieved. Our current results are consistent with the important role of this enzyme in amyloidogenesis and confirm it as a therapeutic target. The inhibition was also quite efficient as about 64% of Aβ reduction was observed with a single intraperitoneal injection of an inhibitor of moderate potency. The kinetics of plasma Aβ reduction is in general consistent with the presence of fluorescent inhibitor Fs[OM00-3]DR9 in the brain 3 h after injection (Fig. 1d). Multiple injections to simulate a longer half-life inhibitor produced maximal inhibition of about 90% in the plasma and about 70% in the brain.

In measuring Aβ in the brain, we sought to determine the effect of the conjugated inhibitor on brain Aβ levels over the 8-h period when four injections of inhibitor were made. Obviously, Aβ from deposition prior to the administration of inhibitor may be present in brain samples and would skew the data. To this end, we used RIPA buffer extraction of brain tissue to measure relatively soluble, thus, recently synthesized Aβ (Phiel et al. 2003). The statistically significant difference of Aβ in RIPA extracted brains of these two groups (Fig. 4b) therefore can be attributed to the inhibition of Aβ production in the brain. The subsequent extraction with a guanidine-based buffer was designed to further solubilize Aβ that had been deposited in a less soluble form over a longer time period (Phiel et al. 2003). The fact that the statistically significant difference from inhibitor administration was only seen for the RIPA extracts and the fact that guanidine extracts contained at least eightfold more Aβ than that in RIPA extracts (Fig. 4b) suggests that the guanidine extracts contained Aβ deposited during the period before the start of inhibitor administration.

The lowering of RIPA buffer-soluble Aβ in the brain by the conjugated inhibitor suggests that it is due to the inhibition of memapsin 2 activity in the brain for the following reasons. First, the inhibitors used were transition-state analogues (Ghosh et al. 2000; Turner et al. 2001) specific for aspartic proteases and had a high potency toward memapsin 2. The fact that both C99 and Aβ are inhibited by the conjugated inhibitor in HEK293 cells (Fig. 2) and in mice (Fig. 5) indicates that the inhibition was on the β-secretase step but not on γ-secretase. Second, memapsin 2 gene deletion in mice completely abolished β-secretase activity (Cai et al. 2001; Luo et al. 2001; Roberds et al. 2001), indicating that no other aspartic protease besides memapsin 2 significantly participates in the cleavage of β-secretase site of APP leading to the production of Aβin vivo. In view that inhibition was observed for both the brain Aβ and the plasma Aβ, it is of interest to consider if the Aβ levels in these two compartments are related. It has been established previously that Aβ in young Tg2576 AD mice is expressed almost entirely in the brain (Kawarabayashi et al. 2001). We have confirmed this by Northern blots for human APP mRNA expressed from this mouse strain (results not shown). In PDAPP mice, an intravenously injected monoclonal antibody versus Aβ that did not penetrate BBB resulted in an increased Aβ efflux from brain to plasma (DeMattos et al. 2001; DeMattos et al. 2002). These observations suggested that some of the observed plasma Aβ reduction may have originated from the inhibition of brain Aβ production. This possibility is indirectly supported by the results in Fig. 3d where two different age groups of mice expected to have different permanent Aβ deposition in the brain produced a very similar inhibitory response in plasma Aβ. In addition, we have also observed a moderate correlation of Aβ levels in the plasma and brain of untreated animals in the data presented in Figs 3 and 4 (r = 0.726, p < 0.025, results not shown) and the difference of Aβ in RIPA extracts of the two groups attained better statistical significance if the ELISA data in Fig. 4b were normalized by the plasma levels in preinjection plasma Aβ (data not shown). These observations are consistent with the notion that the plasma Aβ in our studies had largely originated in the brain. However, the current data do not directly establish the origin of the plasma Aβ or the efflux of brain Aβ into plasma. Whether the plasma Aβ can be taken as an indicator of brain Aβ production in Tg2576 mice would require more extensive investigation. It should be noted that the reduction of Aβ in both plasma and brain observed in this study is independent of the relationship between the Aβ pools in each compartment.

CPs had previously been shown to facilitate the transport of natural macromolecules such as protein and DNA across the cell membrane (Fawell et al. 1994; Schwarze et al. 1999; Wender et al. 2000). Current results demonstrate for the first time that inhibitors containing a non-peptidic transition-state isostere can be delivered across the cell membrane and the BBB with the assistance of a CP. The conjugated inhibitors described here may be employed to study memapsin 2 functional mechanisms in cells or animal models and the cell-penetration strategy can likely be extended to design inhibitors for the study of other cellular systems. Although the current studies were aimed at the target validation, our results nevertheless raise an interesting possibility that CPs may be employed for the delivery of AD therapeutics and others destined to the central nervous system. One may argue for the advantage of such an approach in that the parental inhibitors need not be small enough for BBB penetration, so the drug can be selected from a wider repertoire of candidate compounds based on potency, selectivity and other drug properties. On the other hand, the CP-conjugated inhibitors are not likely to be orally delivered and their pharmacokinetic potential, toxicity, and immunologic response have yet to be studied. The observation that non-peptidic bonds can be transported suggests that the design of new non-peptidic carriers may overcome some of these drawbacks.

Acknowledgements

This work was in part supported by grants from the NIH and the Pioneer Award from the Alzheimer's Association as well as the Institute for the Study of Aging. JT is holder of the JG Puterbaugh Chair in Biomedical Research at the Oklahoma Medical Research Foundation. Authors thank Drs JD Capra and JA Hartsuck for the constructive criticism of the manuscript and the following for their assistance: Angela Irwin, Chris Thompson, Dr Robert Turner, Dr Yukun Wang, and Johnson Yohannan as well as the Imaging Core Facility and the Live Animal Resources Center, Oklahoma Medical Research Foundation, and the Molecular Biology Resource Facility, and Flow and Image Cytometry, University of Oklahoma Health Sciences Center.

Ancillary