Generation of the Amyloid-β Peptide N Terminus inSaccharomyces cerevisiae Expressing Human Alzheimer's Amyloid-β Precursor Protein*

The Alzheimer's amyloid-β precursor protein (βAPP) is a type 1 membrane-spanning protein from which the Alzheimer's disease amyloid-β peptide (Aβ) is proteolytically derived. To date, attempts to identify the enzymes responsible for Aβ generation have failed. Here we report the accumulation of Aβ-immunoreactive peptides in yeast expressing human βAPP. Characterization of these peptides by metabolic labeling, immunoprecipitation with Aβ-specific antibodies, and N-terminal radiosequencing indicates that these peptides include the Aβ peptide at their N termini. The Aβ-like peptides generated in yeast were recovered predominantly as 8- and 12–14-kDa species. A 4-kDa species was recovered either when a protease-deficient strain was used to prevent breakdown or when the 8- and 12–14-kDa species were treated with disaggregating agents. The likely existence in yeast of enzymes generating the Aβ N terminus indicates that the molecular identification of yeast β-secretase-like enzymes may be accomplished using genetic screens or empirical approaches based upon the sequenced genome of Saccharomyces cerevisiae.

The Alzheimer's amyloid-␤ precursor protein (␤APP) is a type 1 membrane-spanning protein from which the Alzheimer's disease amyloid-␤ peptide (A␤) is proteolytically derived. To date, attempts to identify the enzymes responsible for A␤ generation have failed. Here we report the accumulation of A␤-immunoreactive peptides in yeast expressing human ␤APP. Characterization of these peptides by metabolic labeling, immunoprecipitation with A␤-specific antibodies, and N-terminal radiosequencing indicates that these peptides include the A␤ peptide at their N termini. The A␤-like peptides generated in yeast were recovered predominantly as 8-and 12-14-kDa species. A 4-kDa species was recovered either when a protease-deficient strain was used to prevent breakdown or when the 8-and 12-14-kDa species were treated with disaggregating agents. The likely existence in yeast of enzymes generating the A␤ N terminus indicates that the molecular identification of yeast ␤-secretase-like enzymes may be accomplished using genetic screens or empirical approaches based upon the sequenced genome of Saccharomyces cerevisiae.
Alzheimer's disease (AD) 1 is characterized by an intracranial amyloidosis that develops in an aging-dependent manner. This amyloidosis appears to be dependent on the production of the amyloid-␤ peptide (A␤) from the amyloid-␤ precursor protein (␤APP) (1,2). ␤APP is a type 1 transmembrane protein, which can be proteolytically processed through two mutually exclusive pathways. Cleavage of a lysine-leucine bond within the lumenal portion of the A␤ region of ␤APP, a reaction catalyzed by a membrane-associated proteolytic activity termed ␣-secretase, releases the lumenal portion of ␤APP as a soluble protein (␤APP s␣ ) and precludes the formation of A␤ (3). Alternatively, A␤ is produced by the successive proteolytic processing of ␤APP by at least two activities termed ␤-secretase and ␥-secretase. A␤ includes the final 28 residues of the N-terminal lumenal domain and the first 12-14 residues of the transmembrane domain of ␤APP. All known genetic alterations underlying familial AD increase the accumulation of A␤ in the brain, suggesting that an important early step in the pathogenesis of AD involves production or deposition of A␤. This invariant pathological phenotype has made A␤ metabolism a potential target for therapeutic intervention.
Isolation of the enzymes responsible for A␤ generation has proven extremely challenging. We have sought to overcome many of the barriers inherent in mammalian enzyme discovery by searching in a simpler organism for a gene homologous to one of the mammalian secretases. For these studies, we chose yeast because of the ease of genetic manipulation. More specifically, we selected the budding yeast Saccharomyces cerevisiae because it is the only yeast for which the genome has been fully sequenced. Combined genetic and biochemical approaches have identified several important proteolytic processing enzymes in yeast, which have proven homologous to enzymes that catalyze the same reaction in higher eukaryotes (4,5).
It was previously demonstrated that yeast could execute an ␣-secretase-like cleavage of human ␤APP (6, 7), a reaction recently discovered to be catalyzed by yapsins (yeast glycosylphosphatidylinositol-linked aspartyl proteases, Mkc7 and Yap3) (8,9). Here we report that yeast also possess one or more ␤-secretase-like activities, which metabolize human ␤APP 751 into peptides bearing the A␤ sequence at their N termini.
Yeast Media, Growth, and Radiolabeling-Yeast strains were grown in synthetic complete medium (13) lacking uracil and containing either 2% (w/v) glucose (SGluC-ura) or galactose (SGalC-ura). Cells were grown at 30°C in SGluC-ura to a density of ϳ1 ϫ 10 7 cells/ml. Cells were washed twice with SGalC-ura and equal amounts of cells were resuspended in 5 ml of SGalC-ura and incubated at 30°C for 2-4 h in the presence of EXPRESS-[ 35 S]methionine/cysteine labeling mix (250 * This work was supported by the National Institutes of Health Grants AG10491 (to S. G.) and AG09464 (to P. G.), the American Federation for Aging Research (to M. S. and H. X.), the New York State Office of Mental Health, and the Research Foundation for Mental Hygiene (to S. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Preparation of Yeast Cell Lysates-Labeled cells were harvested by centrifugation and washed twice in sterile H 2 O. Acid washed glass beads (Sigma) (ϳ250 l) were added to the cells with 100 l of extraction buffer (0.5% deoxycholate, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 50 g/ml aprotinin, 20 g/ml leupeptin, 2 g/ml pepstatin, 50 mM Tris, pH 7.4). The mixture was vortexed at full speed for 2 min and then centrifuged at 14,000 rpm in a microcentrifuge for 5 min at 4°C. An equal volume of a solution (Buffer A) containing 20 mM sodium phosphate, pH 7.4, 200 mM sodium chloride, 2% Triton X-100, 2 mg/ml sodium azide, 1 mM phenylmethylsulfonyl fluoride, and 50 g/ml aprotinin was added, and the supernatant fractions were subjected to immunoprecipitation.
Immunoprecipitation-A 400-l aliquot of each supernatant was incubated with 5 l of 369 and immunoprecipitated with protein A-Sepharose (Amersham Pharmacia Biotech) to preclear full-length ␤APP and C-terminal fragments of ␤APP. The resulting supernatants were then incubated with either 1 l of 4G8, 1 l of 26D6, or 3 l of 6E10 and proteins immunoprecipitated with 3 l of rabbit affinity-purified antibody to mouse IgG (Cappel) followed by protein A-Sepharose. Sepharose pellets were washed twice in a solution containing 150 mM sodium chloride, 10 mM Tris-HCl, 5 mM EDTA, 0.1% Triton X-100, and 100 mg/ml bovine serum albumin, once in phosphate-buffered saline, and resuspended and heated for 3 min at 100°C in 25 l of tricine loading buffer (Novex, San Diego, CA). Samples were subjected to SDS-polyacrylamide gel electrophoresis on 10 -20% tricine gels (Novex) and autoradiography (Kodak). For quantitative analyses, autoradiographic densities were quantified using a Bio-Rad PhosphorImager and software version 2.0.
Sequence Analysis-Cells (ϳ5 ϫ 10 7 ) were doubly labeled with 0.5 mCi/ml L-[2,3,4,5,6-3 H]phenylalanine (NEN Life Science Products) and 250 Ci/ml [ 35 S]methionine (NEN Life Science Products) for 4 h at 30°C in SGalC-ura. Lysates were prepared, and proteins were immunoprecipitated with 4G8 as described above. Samples were subjected to SDS-polyacrylamide gel electrophoresis on 10 -20% tricine gels. Labeled peptides were transferred to polyvinylidene difluoride sequencing membranes (Millipore) at 35 V for 90 min. After autoradiography, bands were excised from the membrane and eluted with acetonitrile, and the peptides subjected to 12-22 cycles of Edman degradation in an automated sequencer (Rockefeller University Protein Facility). Fractions were collected and radioactivity was measured by liquid scintillation spectrometry (Beckman LS 5801).
Disaggregation Assays-Cells were harvested and lysed as described above except that the extraction buffer was replaced by 100 l of either 70% formic acid or 100% hexafluoroisopropanol (HFIP). Lysates were sonicated for 4 ϫ 30 s followed by centrifugation at 14,000 rpm at 4°C for 1 min to remove unbroken cells. Formic acid-treated lysates were re-equilibrated to pH 7 with 2 M Tris-HCl, pH 8.8. HFIP-treated lysates were lyophilized and redissolved in 6 M urea. Samples were adjusted to 0.5% SDS and heated at 100°C for 3 min. An equal volume of Buffer A was added to samples, and proteins were immunoprecipitated overnight using antibody 4G8. Lysates were sequentially immunoprecipitated with 369 (preclearing) followed by 4G8. The proteins recovered by 4G8 were subjected to SDS-PAGE and transferred to nitrocellulose. The 8-and 12-14-kDa bands were excised and eluted from the nitrocellulose after which Edman degradation was performed. Fractions were diluted in liquid scintillation fluid, and the tritium incorporation was measured by liquid scintillation spectrometry. Radioactivity was present in fractions 4, 19, and 20 of both species, as would be expected for a peptide bearing A␤ at its N terminus. CPM, counts/minute.

RESULTS AND DISCUSSION
Previous work (6,7) demonstrated that human ␤APP can be expressed in S. cerevisiae and successfully trafficked through the secretory pathway. In the current study, S. cerevisiae expressing human ␤APP 19 -751 fused to a yeast ␣-mating pheromone precursor, prepro-␣-factor (Fig. 1), were metabolically labeled for 4 h with [ 35 S]methionine in galactose-containing growth medium. After this labeling period, the yeast were harvested, and the media were collected. Cell lysates were precleared with antiserum 369 to remove full-length and Cterminal fragments. Using antibody 4G8 for immunoprecipitation, a prominent 8-kDa species and a much weaker 12-14-kDa species were recovered from the lysates (Fig. 2), but not the media (data not shown), of cells (CRY1) expressing ␤APP. Parallel immunoprecipitations were performed on the same strain with antibody 6E10 or 26D6, each of which recognizes an epitope within A␤ N-terminal to the ␣-secretase cleavage site. The results obtained with 6E10 or 26D6 were similar to those obtained with 4G8 (Fig. 2). The "Swedish" mutation, which results in a KM to NL tandem missense substitution at the ␤-secretase cleavage site of ␤APP, was introduced into the ␤APP plasmid by site-directed mutagenesis. In mammalian cell lines expressing Swedish ␤APP (11), there is increased production of A␤. Similarly, introduction of the Swedish mutation into yeast increased the production of the 8-and 12-14-kDa bands (Fig. 2), suggesting that these species were generated by a yeast enzyme with an authentic ␤-secretase-like activity.
To further characterize the identities of the 8-kDa and 12-14-kDa bands, their N termini were radiosequenced by labeling ␤APP-expressing yeast with [ 35 S]methionine and [ 3 H]phenylalanine. Lysates from these cells were precleared with 369, immunoprecipitated with 4G8, separated on 10 -20% tricine gels, and transferred to nitrocellulose. The bands were eluted from nitrocellulose, subjected to Edman degradation, and ana-lyzed by liquid scintillation spectrometry. The radioactivity profiles of the first 22 residues of the 8-kDa species (Fig. 3, left) and of the 12-14-kDa species (Fig. 3, right) suggest N termini identical to that of A␤, because 3 H radioactivity was detected at residues 4, 19, and 20.
It seemed possible that the absence of a typical 4-kDa A␤ species might be attributable to protease activity, to aggregation, or to both. Evidence in support of the first possibility was obtained by screening a series of ␤APP-expressing yeast strains, deficient in one or another protease, for their ability to generate an A␤-immunoreactive species. In a strain deficient for Kex2 (BFY-106-4D), an A␤-immunoreactive species with apparent molecular mass of 4 kDa was detected (Fig. 4A). N-terminal radiosequencing of this 4-kDa species was consistent with the N terminus of A␤ (Fig. 4B) One explanation for the appearance of the 4-kDa band only in a Kex2-deficient strain could be that the presence of Kex2 results in degradation of A␤. Alternatively, the absence of Kex2 could affect the activity of an as yet unidentified enzyme, which promotes the formation of the 4-kDa species, or lead to a missorting of ␤APP to an inappropriate intracellular compartment, such as the vacuole, where proteases are highly active. Any of these possibilities could account for the generation of the 4-kDa A␤-like species in yeast strains deficient for Kex2. Evidence in support of the possibility of peptide aggregation was obtained by sonication of yeast lysates in the presence or absence of formic acid or HFIP. Following either treatment, the recovery of the 8-and 12-14-kDa species was reduced (but not abolished), and a 4-kDa band became apparent (Fig. 5), suggesting that the larger species represent, at least in part, dimers and trimers of A␤. The determination of the exact C terminus of this 4-kDa species will be required to determine whether this peptide corresponds to FIG. 4. Analysis of a 4-kDa A␤-like species from a Kex2-deficient yeast strain. A, wild type (Kex2 ϩ ) and Kex2-deficient (Kex2 Ϫ ) strains were transformed with wild type human ␤APP and analyzed as described under "Experimental Procedures." The 4-kDa A␤-like species is indicated by an arrowhead. B, Kex2 Ϫ cells were metabolically labeled, and lysates were prepared and analyzed as described in the legend to Fig. 3. Radioactivity was present in fractions 4, 19, and 20 of the 4-kDa species, as would be expected for a peptide bearing A␤ at its N terminus. CPM, counts/minute.

FIG. 5. Effect of disaggregating agents on electrophoretic mobilities of A␤-like species in yeast A.
Cells were harvested after 2 h of metabolic labeling in SGal-ura. The cells were lysed, and the proteins were extracted with 0.5% deoxycholate/0.5% Nonidet P-40 (lanes 1 and  3), 70% formic acid (lane 2), or 100% HFIP (lane 4) as described under "Experimental Procedures." Proteins were sequentially immunoprecipitated with antibody 369 (preclearing) and antibody 4G8. 4G8-immunoprecipitated proteins were subjected to SDS-PAGE and autoradiography. B, relative intensities of the three A␤-like bands were determined using NIH Imagequant software, version 2.0. ND refers to a signal below the limit of detection. A␤ 1-40 or 1-42, both of which occur in mammalian cells. Preliminary studies, using an high pressure liquid chromatography system, which is capable of resolving these two variants, suggest that the 4-kDa species is predominantly similar to A␤42. 2 The discovery that yeast strains can process full-length human ␤APP into an A␤-like species, together with the ease of genetic manipulation in yeast, should make it possible to conduct genetic screens designed to identify A␤ N terminus-generating activities and to study ␤APP trafficking.