Generation of C-terminally truncated amyloid-β peptides is dependent on γ-secretase activity

Authors


Address correspondence and reprint requests to Dirk Beher, Department of Biochemistry & Molecular Biology, Merck Sharp & Dohme Research Laboratories, The Neuroscience Research Centre, Terlings Park, Eastwick Road, Harlow, Essex CM20 2QR, UK. E-mail: dirk_beher@merck.com

Abstract

Aberrant production of amyloid-β peptides by processing of the β-amyloid precursor protein leads to the formation of characteristic extracellular protein deposits which are thought to be the cause of Alzheimer's disease. Therefore, inhibiting the key enzymes responsible for amyloid-β peptide generation, β- and γ-secretase may offer an opportunity to intervene with the progression of the disease. In human brain and cell culture systems a heterogeneous population of amyloid-β peptides with various truncations is detected and at present, it is unclear how they are produced. We have used a combination of surface enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) and a specific inhibitor of γ-secretase to investigate whether the production of all amyloid-β peptide species requires the action of γ-secretase. Using this approach, we demonstrate that the production of all truncated amyloid-β peptides except those released by the action of the nonamyloidogenic α-secretase enzyme or potentially beta-site βAPP cleaving enzyme 2 depends on γ-secretase activity. This indicates that none of these peptides are generated by a separate enzyme entity and a specific inhibitor of the γ-secretase enzyme should havethe potential to block the generation of all amyloidogenicpeptides. Furthermore in the presence of γ-secretase inhibitors, the observation of increased cleavage of the membrane-bound βAPP C-terminal fragment C99 by α-secretase suggests that during its trafficking C99 encounters compartments in which α-secretase activity resides.

Abbreviations used
AD

Alzheimer's disease

amyloid-β peptide

βAPP

β-amyloid precursor protein

BACE

β-site βAPP cleaving enzyme

CTF

C-terminal fragment

DMEM

Dulbecco's modified Eagle's medium

ECL

electrochemiluminescence

HEK293

human embryonic kidney 293

NSAID

non-steroidal anti-inflammatory drug

PBS

phosphate-buffered saline

SELDI-TOF MS

surface enhanced laser desorption/ionization time-of-flight mass spectrometry.

Amyloid-β (Aβ) peptide is the major proteinaceous constituent of extracellular protein deposits which are characteristic of Alzheimer's disease (AD). These deposits occur either as senile plaques in the brain parenchyma or as vascular amyloid around the brain blood vessels (Glenner and Wong 1984; Masters et al. 1985). Aβ peptides themselves are generated by processing of the β-amyloid precursor protein (βAPP) (Kang et al. 1987) which requires consecutive cleavages involving β- and γ-secretase enzymes (Haass et al. 1992). An alternative processing pathway leads to the release of the βAPP ectodomain as secretory βAPP by cleavage within the Aβ domain thereby excluding Aβ peptide formation (Weidemann et al. 1989). Members of the disintegrin and metalloprotease family (ADAM) such as ADAM10 and TACE appear to mediate this alternative cleavage termed α-secretase cleavage (Buxbaum et al. 1998; Lammich et al. 1999). The recently cloned β-secretase enzyme BACE1 (Asp-2) (Hussain et al. 1999; Sinha et al. 1999; Vassar et al. 1999; Yan et al. 1999; Lin et al. 2000) has been shown to generate the membrane bound βAPP C-terminal fragment (C99) intermediate, which is a prerequisite for the release of Aβ peptide by γ-secretase as the final processing step. BACE1 is the major β-secretase responsible for the generation of Aβ peptides by neurones (Cai et al. 2001) cleaving N-terminally at amino acids 1 and 11 of the Aβ peptide sequence. The homologue BACE2 (Asp-1) (Saunders et al. 1998; Yan et al. 1999; Hussain et al. 2000) cleaves preferentially near the α-secretase site after residues 19 and 20 of Aβ (Farzan et al. 2000). γ-Secretase activity seems to be mediated by a protein complex consisting of presenilins (Li et al. 2000a,b) and at least one other transmembrane protein, nicastrin (Yu et al. 2000), which has been identified recently as an essential component of these complexes. Recent biochemical studies have demonstrated that aspartyl protease transition state analogues inhibit γ-secretase enzyme activity in cell-based (Wolfe et al. 1998; Shearman et al. 2000) and in vitro models (Li et al. 2000a), bind selectively and directly to PS1 fragments (Esleret al. 2000; Li et al. 2000b; Seiffert et al. 2000), and block PS1 endoproteolytic processing (Beher et al. 2001). Remarkable evidence for a causative role of Aβ peptide deposition in the development of AD has been generated by the identification of autosomal dominant mutations in either the βAPP gene or the presenilin 1 and 2 genes leading to familial AD which is characterized by an early onset of disease (Goate et al. 1991; Levy-Lahad et al. 1995a,b; Rogaev et al. 1995). A common mechanism of all these mutations seems to be an elevation of the levels of a 42 amino acid Aβ species [Aβ(1–42)] in comparison with the shorter species Aβ(1–40) (Suzuki et al. 1994; Scheuner et al. 1996). Aβ(1–42) has a much higher tendency to form stable aggregates and according to the amyloid cascade hypothesis primary deposition of Aβ(1–42) is the causative event for the development of the disease (Hardy 1997). Several studies using mass spectrometry techniques have shown that in cell culture models and human brain a variety of different N- or C-terminally truncated peptides can be detected (Naslund et al. 1994, 1996; Gouras et al. 1998). To investigate the mechanisms of Aβ peptide generation and to determine the origin of truncated Aβ peptide species detected in cell culture models we have used surface enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) (Merchant and Weinberger 2000). In this technique, Aβ peptides are immunocaptured using monoclonal antibodies covalently immobilized onto the SELDI™ protein chip (Davies et al. 1999). Subsequent matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) of captured peptides allows the analysis of the pattern of distinct Aβ species present in the samples. Moreover, this method permits the detection of βAPP processing products not observed by conventional metabolic labelling techniques due to their lack of methionine residues. One example of this is Aβ(1–16), which is produced by sequential cleavage of βAPP by β- and α-secretase. We have applied this technique in combination with the use of highly specific inhibitors of βAPP processing to study the mechanism of α- and γ-secretase enzymes in cell culture models. Using these compounds we demonstrate that all truncated Aβ peptides species produced by various cell lines required γ-secretase activity with the exception of Aβ(1–15) and Aβ(1–16) peptides, which result from a combination of β- and α-secretase activities and Aβ(1–19), possibly a BACE2 cleavage product. Furthermore, we show that inhibition of γ-secretase results in an increase of α-secretase cleavage of the β-secretase cleaved membrane-bound βAPP C-terminal fragment C99. This suggests that when C99 is stabilized in the presence of γ-secretase inhibitors it is further trafficked to subcellular compartments in which α-secretase resides.

Materials and methods

Compounds

The hydroxamic acid-based zinc metalloprotease inhibitor batimastat, N4-hydroxy-2(R)-isobutyl-N1-[1(S)-methylcarbamoyl-2-phenylethyl)-3(S)-(thiophen-2-ylsulphanylmethyl)succinamide (Davies et al. 1993; Gearing et al. 1994) (Fig. 1a) was synthesized at Merck Research Laboratories, Rahway, NJ, USA, and the inhibitors of Aβ peptideproduction compound 1, 2(S)-{2(S)-[2-(3,5-difluorophenyl)acetylamino]propionylamino}-4-methyl pentanoic acid methyl ester, compound 2, 2(R)-{2(S)-[2-(3,5-difluorophenyl)acetylamino]propionylamino}-3-phenyl propionic acid methyl ester (Fig. 1a), exemplified in PCT WO98/22494, and compound 3 (1(S)-benzyl-4(R)-{1-[1(S)-carbamoyl-2-phenylethylcarbamoyl]-1(S)-3-methylbutylcarbamoyl}-2(R)-hydroxy-5-phenylpentyl)carbamic acid tert-butyl ester (Fig. 1a) (Shearman et al. 2000) were synthesized at Merck Sharp & Dohme, Terlings Park, UK. The metalloendopeptidase inhibitor phosphoramidon (Umezawa 1976) and the peptidominetic difluoroketone γ-secretase inhibitor MW167 (Wolfe et al. 1998; Moore et al. 2000) were obtained from commercial sources (Sigma and Enzyme Systems Products, Livermore, CA, USA, respectively).

Figure 1.

Compounds synthesized for this study and monoclonal antibodies utilized for SELDI-TOF MS immunocapture. Structures of the hydroxamic acid-based zinc metalloproteinase inhibitor batimastat, γ-secretase inhibitor compound 1, its less potent derivative compound 2 (a) (exemplified in PCT application WO 98/22494), compound 3 and the difluoroketone MW167. Monoclonal antibodies against the Aβ peptide sequence (b) and the amino acid sequences against which the antibodies were raised, or their recognition epitopes if known. The amino acid sequence of human Aβ(1–40) is shown and it is noteworthy that the monoclonal antibody G2-10 is specific for the free C-terminus of Aβ(1–40) ( Ida et al. 1996 ).

Cell culture and cell lines

HEK293 cells stably expressing full-length human βΑPP695 (Shearman et al. 2000) or SH-SY5Y neuroblastoma cells stably expressing the truncated human SPA4CT construct (Dyrks et al. 1993) were propagated under standard conditions using appropriate antibiotic selection as described. Aβ peptides secreted into the media were quantified by an electrochemiluminescence (ECL) assay in a 96-well plate format [(Yang et al. 1994; Khorkova et al. 1998), Origen M-SeriesTM analyser, Igen] as described (Beher et al. 2001). Cytotoxicity was measured by a colourimetric cell proliferation assay (CellTiter 96™ AQ assay; Promega, Madison, WI, USA) utilizing the bioreduction of (3-carboxymethoxyphenyl)-2-(4-sulphophenyl)-2H-tetrazolium (Owen's reagent) to formazan according to the manufacturer's instructions.

Antibodies

Hybridoma cell clones secreting the monoclonal antibodies G2-10 (IgG2bκ), and W0-2 (IgG2aκ), were licensed from the University of Heidelberg (Heidelberg, Germany). Antibodies were isolated from the respective tissue culture media by Cymbus Biotechnology Ltd. (Hampshire, UK). Affinity-purified monoclonal antibodies 4G8 and 6E10 were purchased from Signet Pathology Systems Inc. (Dedham, MA, USA). The antigens against which themonoclonal antibodies were raised or their epitopes described in original publications (Kimet al. 1988; Kim et al. 1990; Ida et al. 1996) are shown schematically in Fig. 1(b). Full-length βAPP andβAPP C-terminal fragments were detected by polyclonal antiserum R7334 (raised against residues 659–694 of βAPP695, provided by M. Kounnas, MRL San Diego, San Diego, CA, USA).

Immunoprecipitation western blot analysis

HEK293 cells stably expressing full-length human βΑPP695 were incubated overnight with 10 µm of either compound 1 or 2. Cell lysates were prepared by suspending the cells in 1 mL lysis buffer [20 mm HEPES, pH 7.3 containing 1% Nonidet P-40, 0.5% deoxycholate, 1 mm EDTA, EDTA-free protease inhibitor cocktail (Roche Molecular Biochemicals, Indianapolis, IN, USA)] and incubation for 30 min on ice with repeated mixing. Insoluble material was removed by centrifugation for 10 min at 20 000 g and 10 µg 4G8 monoclonal antibody added to each sample. After incubation for 2 h at 4°C, 30 µL of a 1 : 1 protein G agarose slurry (Roche Molecular Biochemicals) was added followed by capture of immunocomplexes overnight at 4°C. Non-specifically bound proteins were removed by washing twice with 10 mm Tris-HCl, pH 7.5, 0.15 m NaCl, 2 mm EDTA, 0.2% Nonidet P-40 and once with 10 mm Tris-HCl, pH 7.4. Immunoprecipitated proteins were separated on 10–20% Tris-Tricine precast gradient gels (Invitrogen, Groningen, the Netherlands) and detected using the polyclonal antiserum R7334 as described (Beher et al. 2001).

Synthetic amyloid-β peptides

Synthetic peptides corresponding to the various human Aβ sequences were purchased from California Peptide Research Inc. (Napa Valley, CA, USA). All peptides were of  > 95% purity. Stock solutions were prepared at 100 µm in 100% dimethyl sulphoxide (Me2SO), and aliquots thereof stored at − 80°C.

Surface enhanced laser desorption/ionization time-of-flight (SELDI-TOF) mass spectrometry

Amino-coupling of monoclonal antibodies to preactivated SELDI chips

Monoclonal antibodies (0.5 µg) diluted in phosphate-buffered saline (PBS) were applied to an individual spot of the SELDI protein chipcoated with a preactivated surface (PS1 ProteinChip array, Ciphergen Biosystems Inc., Fremont, CA, USA). After amino-coupling for 3 h at room temperature, free reactive sites were blocked by incubation for 30 min at room temperature with 1 m ethanolamine–HCl, pH 7.5. The chip was washed with PBS/0.05% Triton X-100 for 15 min at room temperature and rinsed once with PBS.

Immunocapture of Aβ peptides from conditioned cell culture media and SELDI-TOF mass spectrometry

SH-SY5Y cells stably expressing SPA4CT or HEK293 cells stably expressing βAPP695 were cultured overnight in phenol red-free Dulbecco's modified Eagle's medium (DMEM) medium, 50 mm HEPES, pH 7.3 containing 2.5% and 5% fetal bovine serum, respectively. Conditioned media were collected, insoluble particles removed by centrifugation for 10 min at 3000 g and after dilution with an equal volume of 25 mm HEPES, pH 7.3, and 1 mm EDTA, media were incubated overnight at 4°C with an antibody-coupled SELDI protein chip. Non-specifically bound peptides were removed by extensive washing: twice with wash A (10 mm Tris-HCl, pH 7.5, 0.15 m NaCl, 2 mm EDTA, 0.2% Nonidet P-40); once with wash B (10 mm Tris-HCl, pH 7.5, 0.5 m NaCl, 2 mm EDTA, 0.2% Nonidet P-40); once with wash C (10 mm Tris-HCl, pH 7.5, 0.15 m NaCl, 2 mm EDTA, 0.1%n-octyl-β-d-glucopyranoside), and once with 5 mm HEPES, pH 7.3 followed by a brief rinse in H2O. A bovine insulin standard for internal calibration was dissolved in a 1 : 5 diluted saturated α-cyano-4-hydroxycinnamic acid solution in 40% acetonitrile, 0.25% trifluoroacetic acid. The standard (0.5 µL, 25fmol bovine insulin) was applied to each spot of the SELDI chip, and after air drying, samples were analysed on a SELDI mass analyser PBS II with a linear time-of-flight mass spectrometer (Ciphergen Biosystems Inc) using time-lag focusing. All spectra were calibrated internally using the single and double positively charged species of the bovine insulin peptide and normalized to the average intensity height of peak for the single charged bovine insulin peptide.

Results

Compound 1 displays the expected pharmacological profile of a γ-secretase inhibitor

Compound 1 was characterized in cell-based models for its potency to inhibit Aβ(40) peptide production (Fig. 2). To facilitate the detection of Aβ peptides, we have used human SH-SY5Y neuroblastoma cells stably transfected with SPA4CT that produce significant amounts of Aβ. The SPA4CT construct encodes a version of the β-secretase cleaved fragment of βAPP (C99) and needs only the action of γ-secretase for Aβ release after removal of the signal peptide (Dyrks et al. 1993). Therefore, by using a cell line transfected with this construct, γ-secretase mechanisms can be studied independently from β-secretase cleavage. In a 96-well plate format assay Aβ production was inhibited with a potency in the low nanomolar range (IC50 7.2 nm) (Fig. 2a). As a control, compound 2 was synthesized, which shows a strongly reduced activity for inhibition of γ-secretase (IC50 13.5 µm) (Fig. 2a). A change of the stereochemistry combined with an altered amino acid at the C-terminal position of the dipeptide has led to this significant decrease in potency.

Figure 2.

Specific inhibition of Aβ production by the prototypical γ-secretase inhibitor compound 1. The γ-secretase inhibitor compound 1 and its less potent derivative compound 2 were analysed for their ability to inhibit Aβ production in human SH-SY5Y neuroblastoma cells stably expressing SPA4CT. Aβ(40) peptide release into the media during the incubation period was quantified using ECL in a 96-well plate format (a). Reduction of Aβ production was measured relative to Me 2 SO-treated controls (a) and the data represents the average of two independent experiments (duplicate measurements in each; if error bars are not visible they are smaller than the actual symbol). IC 50 values were calculated by non-linear regression fit analysis using GraphPad P rism ™ software (GraphPad Software Inc., San Diego, CA, USA). Cytotoxicity was measured by a colourimetric cell proliferation assay (CellTiter 96™ AQ assay, Promega) according to the manufacturer's instructions (b). Note that compound 2 with an opposite stereochemistry at the C-terminal amino acid of the dipeptide and a phenylalanine residue instead of leucine is a ∼ 1900-fold less potent inhibitor. To analyse the stabilization of βAPP CTFs, after treatment of HEK293 βAPP 695 cells with 10 µ m of either compound 1 or 2, cell lysates were immunoprecipitated with monoclonal antibody 4G8 and both full-length βAPP and βAPP-CTFs stained with the antiserum R7334 (c). The open arrowhead indicates β-CTF (C99) and the filled arrowhead α-CTF (C83) polypeptides. As expected, α-CTF was not immunoprecipitated by monoclonal antibody W0-2 (data not shown). Aβ(40) peptide production in the corresponding media was quantified by ECL (d) and cytotoxicity measured by a colourimetric cell proliferation assay (e).

Specific inhibition of βAPP γ-secretase cleavage is expected to result in an accumulation of metabolic intermediates that have been produced by the action of both α- and β-secretase. To study the stabilization of all βAPP derived γ-secretase substrates by functional γ-secretase inhibitors, we have used HEK293 cells stably transfected with βAPP695. Upon treatment of these cells with compound 1, C83 (α-CTF) and C99 (β-CTF) accumulated in cell lysates (Fig. 2c). Consistent with this, when conditioned media from the same cells were analysed for Aβ peptide production, an equivalent inhibition of its production was observed (Fig. 2d). Compound 2 was inactive for both βAPP-CTF accumulation and inhibition of Aβ peptide production. The actual potencies of both compounds are considerably lower in the HEK293-βAPP695 cell line (compound 1: IC50 = 580 nm; compound 2: IC50 >> 10 µm) which is a general observation for many inhibitors (Shearman et al. 2000). To confirm that neither of the observed inhibitory effects are due to cytotoxicity, the cell viability was measured by a colourimetric cell proliferation assay. In both cell lines, SH-SY5Y neuroblastoma (Fig. 2b) and the HEK293 cells (Fig. 2e) no overt cytotoxicity was observed atany compound concentration used for the individual experiments.

Aβ peptide profile from conditioned media of SH-SY5Y cells stably transfected with SPA4CT

To study Aβ peptide generation almost independently of β-secretase we determined the Aβ peptide profiles in the conditioned media of SH-SY5Y cells stably transfected with SPA4CT. A large variety of different peptides was captured by the monoclonal antibodies 6E10 and W0-2, which recognize epitopes at the N-terminus of the Aβ peptide (Figs 1b and 3). The spectrum of Aβ peptides (Fig. 3 and Table 1) includes several truncated versions around the γ-secretase cleavage site [e.g. Aβ(1–37), (1–38), (1–39)] and peptide species clustering around α-secretase or potential BACE2 cleavage sites [e.g. Aβ(1–15), (1–16), (1–19), (1–20)]. The latter ones were not captured by 4G8 antibody, as expected, as the epitope for this monoclonal antibody has been described originally to reside between residues 17–24 of the Aβ peptide (Kim et al. 1988). Based on our data, peptides containing residues up to amino acid 20 of Aβ do not contain the complete epitope and therefore are not recognized by 4G8. G2-10 captured exclusively Aβ(1–40) as predicted from the known epitope of this antibody (Figs 1b and 3). Additional peaks of very low intensity are caused by non-specific binding, as these were also obtained when mouse IgG was used for capture.

Figure 3.

SELDI-TOF MS spectra of Aβ peptides immunocaptured from conditioned media of SH-SY5Y cells stably transfected with SPA4CT. Aβ peptides were immunocaptured from conditioned media using monoclonal antibodies 6E10, W0-2, 4G8, G2-10 or purified non-immune mouse IgG. Captured peptides were directly analysed by SELDI-TOF MS. All spectra were normalized to the average intensity height of the peak for the single charged bovine insulin peptide species and calibrated internally using the single and double positively charged species of bovine insulin. The error of masses of all identified peptides was ≤ 400 ppm compared with the calculated average isotopic masses. *Indicate oxidized peptides. Masses of Aβ peptides are increased by 16 mass units as expected for methionine 35 sulphoxide derivatives. As mentioned in the footnote to Table 1 , all peptides contain the additional sequence Leu-Glu at their N-termini and the last C-terminal amino acid of the peptides indicates the last amino acid of the native Aβ sequence. The data are representative of at least two independent experiments. Additional peaks of very low intensity are either caused by non-specific binding (n.s.), as these were also obtained when mouse IgG was used for capture or their identity is unknown (n.i.).

Table 1.  Soluble Aβ peptides captured by monoclonal 6E10 antibody from the media of SH-SY5Y cells stably transfected with SPA4CT
PeptideMass observedMass calculatedAmino acid sequence
  1. Aβ peptides were identified using the masses obtained from the 6E10 spectrum (Fig. 3). Their masses are in accordance with the other spectra shown (Fig. 3) and the epitopes of the antibodies used for immunocapture (Fig. 1b). Note that due to the cloning of the original SPA4CT construct, Aβ peptides derived from the AβCT protein contain the two additional amino acids leucine and glutamic acid at the N-terminus (Dyrks et al. 1993). Therefore the peptides are labelled as LE-Aβ and the last C-terminal amino acid of the identified peptides indicates the number of the last amino acid in the native Aβ sequence [e.g. LE-Aβ(1–40) ends at position 40 of native Aβ but contains 42 amino acid residues, due to the additional LE sequence included].

LE-Aβ(1–15)2068.52069.1LEDAEFRHDSGYEVHHQ
LE-Aβ(1–16)2196.62197.3LEDAEFRHDSGYEVHHQK
LE-Aβ(1–17)2309.92310.5LEDAEFRHDSGYEVHHQKL
LE-Aβ(1–19)2556.22556.8LEDAEFRHDSGYEVHHQKLVF
LE-Aβ(1–20)2703.52703.9LEDAEFRHDSGYEVHHQKLVFF
LE-Aβ(1–33)3916.13916.3LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIG
LE-Aβ(1–34)4029.34029.4LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGL
LE-Aβ(1–37)4316.14316.8LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVG
LE-Aβ(1–38)4373.44373.9LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGG
LE-Aβ(1–39)4472.44473.0LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGV
LE-Aβ(1–40)4571.34572.1LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVV
LE-Aβ(1–42)4755.84756.4LEDAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIA

Inhibition of γ-secretase blocks the generation of all truncated peptide species except α-secretase processing products

Treatment of SPA4CT transfected SH-SY5Y neuroblastoma cells with the γ-secretase inhibitor compound 1 (10 µm) resulted in a complete inhibition of the production of Aβ (1–40) and Aβ(1–42), and all truncated species with the exception of Aβ(1–15) and Aβ(1–16) (Fig. 4a). The latter species are derived by α-secretase cleavage of SPA4CT and their relative amounts were increased in the media. This is documented by a much stronger signal intensity for these peaks relative to the bovine insulin standard. It is noteworthy, that it is inappropriate to compare concentrations of different peptide species from their peak heights, since they are ionized with different efficiencies (Wang et al. 1996). With the system used for this study, synthetic Aβ(1–42) peptide spotted directly onto a chip gave a signal intensity approximately seven times less when compared with identical amounts of synthetic Aβ(1–40) (data not shown). The data can be analysed semiquantitatively, however, by comparing the peak height of the same peptide to an internal standard since the relative peak height of a peptide depends on the amount of peptide on the spot (Wang et al. 1996). If normalized to the single charged species of the bovine insulin standard a 2.1-fold increase of Aβ(1–15) and a 3.9-fold increase of Aβ(1–16) was observed. This finding indicates that upon inhibition of the γ-secretase pathway the alternative α-secretase cleavage of C99 is increased, most likely because more substrate molecules are available. To verify these findings, cells were treated with batimastat, a zinc metalloproteinase inhibitor (Davies et al. 1993) that is known to inhibit the α-secretase cleavage of βAPP (Parvathy et al. 1998). Treatment of cells with 25 µm batimastat decreased the production of α-secretase derived peptides Aβ(1–15)(1.9-fold) and Aβ(1–16) (3.9-fold). Further, this observation confirmed the identity of these peaks as α-secretase processing products in addition to the assignment based on the detected masses. The combination of inhibitors of both α- and γ-secretase enzymes resulted in the disappearance or reduction of all peptides which are produced by processing of the SPA4CT protein.

Figure 4.

Inhibition of γ-secretase in SH-SY5Y cells stably transfected with SPA4CT results in an increase of α-secretase cleavage products. SPA4CT transfected SH-SY5Y neuroblastoma cells were incubated with either vehicle Me 2 SO (control) or compounds overnight. Aβ peptides were immunocaptured from conditioned media of these cells using monoclonal antibody 6E10 and directly analysed by SELDI-TOF MS. The spectra were normalized to the peak height of the bovine insulin standard prior to data analysis. The data are representative of at least two independent experiments. The grey box highlights the mass range of Aβ(1–15) and Aβ(1–16) peptides. Additional peaks of very low intensity are either caused by non-specific binding (n.s.), as these were also obtained when mouse IgG was used for capture (data not shown) or their identity is unknown (n.i.). *Indicate oxidized peptides. Mass spectra obtained after treatment with 10 µ m γ-secretase inhibitor compound 1, 25 µ m zinc metalloproteinase inhibitor batimastat, or both 10 µ m compound 1 and 25 µ m batimastat are shown in (a). SPA4CT cells were treated with 10 µ m of compound 1, 10 µ m of compound 3 and MW167 at 25 µ m or 50 µ m concentration to verify that the observed changes are specific for γ-secretase inhibition (b). A higher magnification of the mass/charge range from 4300 to 4900 mass units is shown for media of control and MW167-treated (25 µ m ) cells (c). Note the presence of methionine sulphoxide derivatives with an increased mass by ∼16 mass units (ox.).

To verify that the observed changes are specific for γ-secretase inhibition and are not due to non-specific off-target activities of compound 1 we compared a selection of well described and characterized γ-secretase inhibitors in our assay system. Compound 3 (Shearman et al. 2000) and MW167 (Wolfe et al. 1998) are known as nanomolar and micromolar γ-secretase inhibitors, respectively. Essentially treatment of SH-SY5Y cells with either compound 1, compound 3 or increasing concentrations of MW167 leads to an increase in the production of Aβ(1–15) and Aβ(1–16) peptides (Fig. 4b; peaks highlighted). This was accompanied by a complete inhibition of the production of all other Aβ peptide species after treatment of cells with either compound 1 or 3. In accordance with previous reports (Wolfe et al. 1999; Moore et al. 2000) if MW167 was used at 25 µm concentration a selective inhibition of the production of Aβ(1–40) was observed whereas the production of Aβ(1–42) remained unchanged (Figs 4b and c). Furthermore, the production of all the C-terminally truncated species ending at residues > 16 was inhibited. Interestingly, this treatment appeared to enhance the production of a longer Aβ peptide variant, Aβ(1–43) which is not observed in untreated cells (Figs 4b and c). Only if MW167 was used at 50 µm concentration was the production of all Aβ peptides except Aβ(1–15) and Aβ(1–16) inhibited. Note that this passage of cells secretes considerably more Aβ when compared with the cells used for the experiments shown in Fig. 4(a) (higher peak intensities of Aβ peptides compared with the standard) which may explain the higher signal intensity obtained for the Aβ(1–20) peptide in the spectra.

To exclude any artificial metabolism associated with the overexpression of a truncated βAPP derivative, a similar study was performed using HEK293 cells stably transfected with βAPP695(Figs 5a and b). Again, under control conditions a variety of truncated Aβ peptide species were captured using monoclonal 6E10 antibody. In addition to Aβ(1–40) and Aβ(1–42), the C-terminally truncated peptides [e.g. Aβ(1–37), (1–38), (1–39)], and furthermore the N-terminally truncated peptides Aβ(5–38) and Aβ(5–40) were detected. The products of subsequent β- and α-cleavage of βAPP, Aβ(1–15) and Aβ(1–16) were just above the detection limit under these conditions (Fig. 5b). Treatment with 10 µm compound 1 mirrors the situation observed with the SPA4CT cells, by completely inhibiting the production of almost all Aβ peptide species. The production of the β-/α-cleavage product Aβ(1–16) was strongly increased whereas a further remaining peptide, Aβ(1–19), appeared to be unchanged. Position 19 has been identified as one of the major cleavage sites for BACE2 (Asp-1), a β-secretase homologue (Farzan et al. 2000) which suggests that this peptide is derived by a subsequent cleavage of βAPP by BACE1 and BACE2.

Figure 5.

Inhibition of γ-secretase in HEK293 cells stably transfected with βAPP 695 results in an increase of α-secretase cleavage products. HEK293 cells stably overexpressing βAPP 695 were treated with vehicle Me 2 SO (control) 10 µ m compound 1 or 10 µ m phosphoramidon as indicated. Aβ peptides were immunocaptured from conditioned media using monoclonal antibody 6E10 and directly analysed by SELDI-TOF MS (a). The spectra were normalized to the peak height of the bovine insulin standard prior to data analysis. Note that the cells used for this experiment are transfected with wild-type βAPP 695 leading to the production of genuine Aβ peptides. A higher magnification of the mass/charge range from 1600 to 2400 mass units is shown for media of control and compound-treated cells (b). The data are representative of at least two independent experiments. The grey box highlights the mass range of Aβ(1–15) and Aβ(1–16) peptides.

To examine whether the truncated Aβ peptides are generated by subsequent metalloprotease cleavage of Aβ(1–40) and Aβ(1–42), HEK293 βAPP695 cells were treated with 10 µm phosphoramidon. This compound is known to inhibit several metalloproteases such as neprilysin (Emoto and Yanagisawa 1995), angiotensin-converting enzyme and endothelin converting enzyme (Kukkola et al. 1995) and increases Aβ(1–40) and Aβ(1–42) in conditioned media of various cell lines (Fuller et al. 1995; Eckman et al. 2001). This treatment, however, did not change the qualitative pattern of Aβ peptides detected in the conditioned media but as expected increased the overall amounts of secreted Aβ peptides.

Discussion

We have applied the technique of SELDI-TOF MS in combination with specific inhibitors of α- and γ-secretase to analyse the biogenesis of Aβ peptides released into tissue culture media. A variety of different soluble Αβ peptide species were found in the media of human SH-SY5Y neuroblastoma cells stably transfected with SPA4CT. This observation is in good accordance with Wang et al. (1996), who obtained their data using either wild-type or βAPP overexpressing mouse Neuro2A cells. Besides Aβ(1–42) and Aβ(1–40), a wide range of C-terminally truncated species were observed, clustering around α-secretase (or potential BACE2) and γ-secretase sites. All truncated peptides except Aβ(1–15) and Aβ(1–16) in SPA4CT and additionally Aβ(1–19) in βAPP695 cells are derived directly from the γ-secretase pathway, as it has been shown that their production is inhibited upon treatment with the prototypical γ-secretase inhibitor compound 1 and other reference γ-secretase inhibitors. This rules out the possibility that any of these species are produced by direct cleavage of C99 at these alternative positions by an enzyme activity unrelated to γ-secretase. There is, however, a slight discrepancy between the cellular systems used, since in the SPA4CT overexpressing cells the production of putative BACE2-derived peptides Aβ(1–19) and Aβ(1–20) is inhibited by γ-secretase inhibitor treatment. The finding that their production is abolished upon treatment with three structurally diverse γ-secretase inhibitors rather rules out the possibility that they are BACE2 processing products, and it is more likely that they are produced by alternative pathways in this specific cell line.

There are several possibilities to explain how truncated Aβ derivatives could be generated, but as yet the exact mechanisms remain unclear. Spiking of conditioned media with peptides has shown that only a small proportion of exogenously added peptide was truncated (Wang et al. 1996), and thus it seems unlikely that either proteinases or exopeptidases in the media are responsible for the generation of these peptides. This is further supported by our finding that treatment with the metalloprotease inhibitor phosphoramidon did not change the pattern of Aβ peptides detectable in the conditioned media. This rules out that any of the enzymes inhibited by this compound, such as neprilysin, angiotensin-converting enzyme or endothelin converting enzyme, are involved to a marked extent in the generation of truncated Aβ peptides. Otherwise one would have expected to observe a selective inhibition of their formation. The possibility that these truncated Aβ peptides are mainly generated by γ-secretase during intracellular processing of βAPP may be explained by the fact that the γ-secretase enzyme contains unique features. Substrate cleavage seems not to depend on a specific sequence motif for recognition (Lichtenthaler et al. 1997; Lichtenthaler et al. 1999), rather cleaving at a specific position relative to the membrane bilayer (Lichtenthaler et al. 1999; Murphy et al. 1999; Lichtenthaler et al. 2002). The production of truncated peptides was observed upon mutagenesis of the βAPP transmembrane region (Murphy et al. 1999; Lichtenthaler et al. 2002) and the production of these species under native conditions could reflect an intrinsic property of an enzyme being able to cleave its substrates at multiple positions. This is further highlighted by our novel finding that the difluoroketone peptidomimetic inhibitor MW167 when used at 25 µm concentration can induce the production of the longer peptide Aβ(1–43). It is noteworthy, that recent evidence has been generated that non-steroidal anti-inflammatory drugs (NSAIDS) can selectively lower Aβ(1–42) and correspondingly increase the production of Aβ(1–38) (Weggen et al. 2001). Taken together these data suggest that a variety of compounds may have the potential to modulate γ-secretase cleavage specificity.

Interestingly, under our experimental conditions, inhibition of the γ-secretase pathway led to an increase of C99 cleavage by α-secretase as demonstrated by the strong increase of α-secretase cleavage fragments Aβ(1–15) and Aβ(1–16). The production of Aβ(1–15) and Aβ(1–16) was reduced by theα-secretase inhibitor batimastat, confirming that both peptides are derived from the alternative α-secretase pathway. The detection of two α-secretase cleavage products is in accordance to the original reports describing the identification of the α-secretase cleavage site on βAPP by amino acid sequencing of peptides derived from purified secretory βAPP (Esch et al. 1990; Wang et al. 1991). Using this approach, a heterogenous population of secretory βAPP molecules either ending at glutamine 15 or lysine 16 of Aβ was observed.

Taken together our data suggest that cleavage of βAPP by BACE1 generates C99 which is mainly converted to Aβ by subsequent γ-cleavage. As alternative pathways, even C99 can reach compartments where it can be processed by α-secretase leading to the release of Aβ(1–15/16) peptides. Most importantly, treatment of cells with a single γ-secretase inhibitor blocks the generation of all amyloidogenic βAPP processing products. This indicates that a specific inhibitor of γ-secretase has generally the potential to prevent the formation of amyloid deposits, which would be a reasonable preventive strategy to combat AD.

Acknowledgements

The authors wish to thank Drs Timothy Harrison and Luis Castro from the Department of Medicinal Chemistry for valuable discussions. Joseph F. Leone (Merck Research Laboratories, Rahway) and Dr Alan Nadin (Department of Medicinal Chemistry) are acknowledged for the synthesis of batimastat and compound 3, respectively.

Ancillary