Equimolar Production of Amyloid β-Protein and Amyloid Precursor Protein Intracellular Domain from β-Carboxyl-terminal Fragment by γ-Secretase*

We showed previously that cells expressing wild-type (WT) β-amyloid precursor protein (APP) or coexpressing WTAPP and WT presenilin (PS) 1/2 produced APP intracellular domains (AICD) 49-99 and 50-99, with the latter predominating. On the other hand, the cells expressing mutant (MT) APP or coexpressing WTAPP and MTPS1/2 produced a greater proportion of AICD-(49-99) than AICD-(50-99). In addition, the expression of amyloid β-protein (Aβ) 49 in cells resulted in predominant production of Aβ40 and that of Aβ48 leads to preferential production of Aβ42. These observations suggest that ϵ-cleavage and γ-cleavage are interrelated. To determine the stoichiometry between Aβ and AICD, we have established a 3-[(3-cholamidopropyl)dimethylammonio]-2-hydroxy-1-propanesulfonic acid-solubilized γ-secretase assay system that exhibits high specific activity. By using this assay system, we have shown that equal amounts of Aβ and AICD are produced from β-carboxyl-terminal fragment (C99) by γ-secretase, irrespective of WT or MTAPP and PS1/2. Although various Aβ species, including Aβ40, Aβ42, Aβ43, Aβ45, Aβ48, and Aβ49, are generated, only two species of AICD, AICD-(49-99) and AICD-(50-99), are detected. We also have found that M233T MTPS1 produced only one species of AICD, AICD-(49-99), and only one for its counterpart, Aβ48, in contrast to WT and other MTPS1s. These strongly suggest thatϵ-cleavage is the primary event, and the produced Aβ48 and Aβ49 rapidly undergo γ-cleavage, resulting in generation of various Aβ species.

Preparation of C99-FLAG Substrate-A carboxyl-terminal fragment of APP (C99) was carboxyl-terminally fused with FLAG tag (C99-FLAG) and amino-terminally with signal peptide (MQLRNPELHLG-CALALRFLALVSWDIPGARA) of human ␣-galactosidase A. Resultant fragment was inserted into pFASTBAC TM 1 (Invitrogen). Sf9 cells were infected with recombinant baculovirus according to the manufacturer's instructions. Infected cells (60-ml culture) were harvested after 36 h and resuspended in 0.3-0.5 ml of Tris-buffered saline (50 mM Tris-HCl, pH 7.6, 150 mM NaCl). The suspension was mixed with equal amount of lysis buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 2% Nonidet P-40 and 2ϫ protease inhibitor mixture (Roche Diagnostics)) and incubated on ice for 1 h. After ultracentrifugation at 245,000 ϫ g for 20 min, the supernatant was agitated with 0.2 ml of ANTI-FLAG M2-agarose beads (Sigma) overnight. C99-FLAG was eluted from the beads by incubation with 0.2 ml of 100 mM glycine HCl, pH 2.7, for 10 min at room temperature, and the eluate was immediately neutralized by addition of 1/25 volume of 1 M Tris-HCl, pH 8.0. Concentrations of residual Nonidet P-40 in purified C99-FLAG were estimated to be 0.3-0.4% from elution volume. The eluted C99-FLAG was confirmed for its purity and quantified by Coomassie Brilliant Blue staining after gel electrophoresis.
␥-Secretase Assay and Detection of A␤ and AICD-Microsomal fractions of CHO cells were obtained as described previously (15). Briefly, the harvested cells were homogenized in buffer A (20 mM PIPES, pH 7.0, 140 mM KCl, 0.25 M sucrose, 5 mM EGTA) using a glass/Teflon homogenizer. The homogenates were centrifuged at 800 ϫ g for 10 min to remove nuclei and cell debris. The postnuclear supernatants were recentrifuged at 100,000 ϫ g for 1 h. The resulting pellets representing the microsomal fractions were suspended in a buffer (50 mM PIPES, pH 7.0, 0.25 M sucrose, 1 mM EGTA). Their protein concentrations were adjusted at 10 mg/ml. The membranes were solubilized by the addition of equal volume of 2ϫ NK buffer (50 mM PIPES, pH 7.0, 0.25 M sucrose, 1 mM EGTA, 2% CHAPSO (Sigma; catalogue number C3649; lot numbers 013K5314 and 015K5313), 2 mM diisopropyl fluorophosphate, 20 g/ml antipain, 20 g/ml leupeptin, 20 g/ml TLCK, 10 mM phenanthroline, and 2 mM thiorphan) and incubated on ice for 1 h (20,21). After centrifugation at 100,000 ϫ g for 1 h, the supernatants were saved (1% CHAPSO lysate). 1% CHAPSO lysate was diluted with 3 volumes of CHAPSO-free buffer (50 mM PIPES, pH 7.0, 0.25 M sucrose, 1 mM EGTA, 1 mM diisopropyl fluorophosphate, 10 g/ml antipain, 10 g/ml leupeptin, 10 g/ml TLCK, 5 mM phenanthroline, and 1 mM thiorphan) containing defined amounts of C99-FLAG. Furthermore, 0.1% phosphatidylcholine (catalogue number P3556, lot number 034K5218; Sigma) was added into the diluted lysate, which significantly enhanced the activity of ␥-secretase but did not alter the proportion of A␤40/42 produced. Residual Nonidet P-40 in the substrate should be estimated carefully before starting the reaction. The Nonidet P-40 concentrations at 0.2% and above in the reaction mixture abolished ␥-secretase activity. Its final concentrations in all reactions in this study were kept less than 0.06%, which exhibited no noticeable suppression in A␤ and AICD pro-ductions. After incubation at 37°C for 3 h, lipids were extracted with chloroform/methanol (2:1), and protein residues were subjected to quantitative Western blotting with defined amounts of synthetic A␤ as a control. For AICD standard, we used CTF50 synthetic peptide (AICD-(50 -99); Calbiochem) or AICD-(50 -99)-FLAG expressed in Escherichia coli as described below. For detection of A␤, 82E1, a monoclonal antibody end-specific for the amino terminus of human A␤ (IBL, Gunma, Japan) was used (23). To visualize AICD, UT-421 (gift of Dr. T. Suzuki, Hokkaido University) raised against carboxyl-terminal sequence of APP was used (24). All blots in this study were immersed in boiled phosphatebuffered saline for 5 min before blocking in 5% skim milk, which significantly enhanced the detectability of the antigens.
Quantification of Authentic AICD-(50 -99)-FLAG-AICD-(50 -99)-FLAG was expressed in E. coli and affinity-purified with ANTI-FLAG M2 beads. The protein concentrations of authentic AICD-(50 -99)-FLAG were determined using both amino acid compositional analysis that determines the total protein amount, and sequence analysis that assesses the specific amount of AICD-(50 -99)-FLAG. The appropriate amounts of AICD-(50 -99)-FLAG were purified by reverse-phase liquid chromatography on super-Phenyl column (Tosoh, Tokyo, Japan) with 0 -48% gradient of acetonitrile in 0.1% trifluoroacetic acid. Peak fractions were collected and rechromatographed under the same condition. Compared between the peak areas of the first and second chromatogram, chromatographic recovery was estimated (about 58%). Pooled fractions from the second chromatography were dried and hydrolyzed in 6 N HCl vapor at 110°C for 20 h. The acid hydrolysate was derivatized with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate and quantified as described by Shindo et al. (25). For amino acid sequence analysis, an appropriate amount of AICD-FLAG was subjected to Edman degradation using a Procise cLC protein sequencing system (Applied Biosystems, Foster City, CA) to obtain the amino-terminal 10-residue sequence. Repetitive yield and initial yield were calculated from the sequence.
Detection of Longer A␤s Produced by ␥-Secretase-Various longer A␤ species were separated on a Tris, Tricine, 8 M urea gel with minor modifications (22,27). A 10% T/3% C separation gel, pH 8.45, containing 8 M urea (gel system I) was used to separate A␤37 through A␤45. A 12% T/3% C separation gel, pH 8.90, containing 8 M urea (gel system II) was used to separate A␤46 through A␤49. The spacer and stacking gels did not contain urea. Following transfer, the blots were probed with 82E1 to detect only A␤s that begin at Asp-1 and developed using an ECL system. Intensities of the bands were quantified using a LAS-1000 plus luminescent image analyzer (Fuji Film, Tokyo, Japan).
Mass Spectrometric Analysis of AICDs-After incubating 0.25% CHAPSO lysate of CHO cells with C99-FLAG in the presence of 0.1 mM bestatin, 10 M amastatin, 0.1 M arphamenine A, the uncleaved excess C99-FLAG was largely removed by immunoprecipitation with 4G8, a monoclonal antibody raised against 17-28 residues of A␤ (epitope, 17-24 residues) (Signet Laboratories, Dedham, MA). The produced AICD was immunoprecipitated with Anti-FLAG M2-agarose beads and extracted with 30% acetonitrile in 1% trifluoroacetate. Masses of the peptides were determined with a matrix-assisted laser desorption ionization-TOF-mass spectrometer, Autoflex (Bruker Daltonics, Ibaraki, Japan). Samples were prepared by the thin layer method using sinapinic acid as a matrix (14).

RESULTS
CHAPSO-solubilized ␥-Secretase Assay System-The relationship between ␥-cleavage and AICD production has been a matter of debate.
Studies by multiple research groups implied that ␥-cleavage was largely independent of -cleavage (28 -31). On the other hand, Pinnix et al. (32) noted a potential link between A␤ and AICD productions. In their paper, they identified and considered AICD as a counterpart of A␤. Recently, we and others (14,15,33) identified novel -cleavage that liberates AICD and assumed that -cleavage was the primary event. We also assumed that equal amounts of total A␤ and AICD are produced by ␥-secretase (14,15,33). Most of quantitative studies on A␤ and AICD productions were performed in the presence of large amounts of ␣CTF, which is processed to p3 and AICD but not to A␤. In addition, levels of A␤ and AICD were assessed by methods based on different principles, FIGURE 1. A␤ species generated by the solubilized ␥-secretase assay system. A, the microsomal fraction of CHO cells was solubilized with CHAPSO and incubated with 500 nM C99-FLAG; the reaction was terminated at the time indicated by placing a reaction tube on ice. The A␤s produced, together with synthetic authentic A␤s, were separated on gel I (top panel) and gel II (bottom panel), followed by Western blotting with 82E1, a monoclonal antibody specific for the amino terminus of A␤. A␤40, A␤42, A␤43, and A␤45 were produced in a time-dependent manner. A␤46 and longer A␤s were stuck at the bottom of gel I. Gel II showed that the longer A␤s were A␤48 and A␤49 and that they were also produced in a time-dependent manner. B, defined amounts of C99-FLAG were incubated with CHAPSO-solubilized microsomal fraction of CHO cells. Protein samples were separated on gel I and II. Against increasing concentrations of C99-FLAG, intensities of each A␤ species were plotted as percentage of 50 pg of synthetic A␤45. Retaining efficiencies of various A␤s were postulated to be the same. C, data represent means Ϯ S.D. of three independent experiments. Apparent K m and V max values are summarized in Table 1. D and E, the CHAPSOsolubilized microsomal fraction was incubated with 500 nM C99-FLAG in the presence of L-685,458 (D) and DAPT (E) at the indicated concentrations. Production of all A␤ species was uniformly suppressed by both ␥-secretase inhibitors in a dose-dependent manner. This contrasts with findings in cell-based or cell-free assay systems, in which DAPT induces differential accumulation of A␤43 and A␤46. Arrowheads indicate carboxyl-terminally truncated, C99-FLAG fragments (27).
such as enzyme-linked immunosorbent assay for A␤ released into media and indirect luciferase transactivation assay for AICD. To our knowledge, rigorous quantitative determination by conventional Western blotting of the produced A␤ and AICD has never been performed.
To examine the stoichiometry of A␤ relative to AICD, we took advantage of the CHAPSO-solubilized ␥-secretase assay system (20,21). C99 was fused at the carboxyl terminus with FLAG tag (C99-FLAG) and expressed in Sf9 cells. After C99-FLAG was purified by Anti-FLAG M2 (see "Experimental Procedures"), and its purity was confirmed, and its amounts were quantified by Coomassie Brilliant Blue staining after gel electrophoresis (see Fig. 2B). Purified 0.5 M C99-FLAG was incubated in 0.25% CHAPSO-lysate of CHO cells. A␤ species produced by our assay system were separated on gel system I (22,27) (Fig. 1A). The amount of each A␤ species was estimated from densitometric scanning of the Western blot and increased in a time-dependent linear fashion (data not shown). Each A␤ species appeared to be generated by random cleavage depending on the interaction between susceptible sites and catalytic sites. Robust signals for A␤42 and A␤43 were detected as well as for A␤40, which contrasts with the major A␤ species in cell-based and cell-free assays (1,14,34,35) (Fig. 1A). In addition to A␤45, A␤ species longer than A␤45 were also detected and stuck at the bottom of gel (Fig. 1A, top panel). Gel system II (22,27) revealed that these longer species were A␤48 and A␤49 (Fig. 1A, bottom panel, and see also Fig.  4B). Against increasing concentrations of C99-FLAG, band intensities of all A␤ species were plotted as the percentage of intensity of synthetic A␤45 (50 pg) on each blot ( Fig. 1, B and C). The production rates of all A␤ species appeared to follow the Michaelis-Menten type curve (Fig.  1C). The apparent K m (K app ) and V max values for various A␤ species were summarized in Table 1. As can be seen in Fig. 1, A and B, this solubilized ␥-secretase assay system was characterized by abundant A␤40, A␤42, and A␤43 and small or trace amounts of A␤45, A␤48, and A␤49. The majority of A␤ species produced by the assay system consisted of A␤40, A␤42, and A␤43 (see below).
L-685,485 and DAPT suppressed A␤ production in a dose-dependent manner (Fig. 1, D and E). As expected, L-685,485, a transition state analogue inhibitor, uniformly suppressed production of all A␤ species (Fig. 1D). Most unexpectedly, DAPT, a nontransition state analogue inhibitor, did not result in a build up of A␤43 and A␤46 (22,27,33,36) but uniformly suppressed the production of all A␤ species (Fig. 1E). This may indicate that membrane integrity is a prerequisite for DAPT-induced accumulation of A␤43 and A␤46, as this phenomenon was observed in cell-based assay (27) and cell-free assay using a microsomal fraction. 5 Stoichiometric Relationship between A␤ and AICD-Purified C99-FLAG was incubated with CHAPSO-solubilized microsomal fraction from CHO cells. The A␤ at about 4 kDa on the SDS gel was quantified by Western blotting as reported previously (15). Total AICD was quantified using synthetic CTF50 (AICD-(50 -99), eight residues smaller than the substrate AICD-FLAG) as a control. The highest molecular mass band at about 15 kDa and the lowest molecular mass band at about 9 kDa before incubation represent presumably amino-terminally extended C99-FLAG with residual signal peptide and truncated C99-FLAG, respectively ( Fig. 2A, bottom panel). Coomassie Brilliant Blue staining cannot detect those fragments, and thus only trace amounts of the fragments contaminate the reaction mixture (Fig. 2B). The A␤ band at about 4 kDa on the SDS gel most likely represents A␤40, A␤42, and A␤43. A␤s longer than A␤43 have slower mobilities (data not shown and see Ref. 22) and can be separated from the major A␤ band at about 4 kDa. The longer A␤s appear to be present in very small amounts, as they constituted around 10% of the total amount estimated from densitometric scanning of the Western blot (gel I; Fig. 1A and Table 1), assuming that the retention efficiency for each A␤ species is equal. Thus, A␤ here indicates virtually the sum of A␤40, A␤42, and A␤43. Time course analysis showed that equal amounts of A␤ and AICD increased similarly in a time-dependent manner (Fig. 2, A and C); the kinetics of A␤ and AICD production was statistically indistinguishable (Fig. 2D). These observations indicate that cleavage of C99 provides equal amounts of A␤ (mostly A␤40, -42, and -43) and AICD. The rates of A␤ and AICD production apparently followed Michaelis-Menten type curve (Fig. 2D). The apparent K m (K app ) values of C99 for A␤ and AICD production were 507.93 Ϯ 26.45 and 468.72 Ϯ 129.26 nM, respectively (Table 2). Those values are roughly consistent with K m values reported previously for crude and purified ␥-secretase (20,21,37). The apparent V max values for A␤ and AICD were 433.53 Ϯ 51.94 and 419.62 Ϯ 19.39 pM/min, respectively ( Fig. 2D and Table 2). Those values were more than 30-fold higher than reported previously (21,35). Such high activity of ␥-secretase in our system made it possible to use conventional Western blotting to accurately quantify the amounts of A␤ and AICD produced.
We then examined the stoichiometry between A␤ and AICD in reaction mixtures containing MTAPP and MTPS1/2. The CHAPSO lysates were reacted with C99-FLAG containing FAD mutations T714I, V717F, and L723P, and an artificial mutation I716F (I45F, according to A␤ numbering) (38). As shown in Fig. 3A, the MTAPPs tested in this study caused substantial to profound reductions in A␤ and AICD productions (Fig. 3A). Nevertheless, these MTAPPs did not alter the stoichiometry between A␤ and AICD (Fig. 3B). Similarly, MTPSs led to significantly reduced production of A␤ and AICD but maintained the one-to-one stoichiometry indicated for WTPS1/2 (Fig. 3, C and D). In G384A MTPS1, an additional signal was detected above the A␤ band (Fig. 3C) and was identified as a mixture of A␤48 and A␤49 on gel II (data not shown). Densitometric analysis showed that the longer A␤s produced by G384A MTPS1 amount to more than 40% of total A␤s, as compared 5 M. Takami, S. Funamoto, Y. Ihara, unpublished data.

TABLE 1 Apparent K m (K app ) and V max values for each A␤ species
Varying amounts of C99-FLAG were incubated in 0.25% CHAPSO lysate from the microsomal fraction of CHO cells for 3 h. Protein samples were subjected to Western blotting using 82E-1. The intensity of each A␤ species was plotted as % of intensity of 50 pg of synthetic A␤ 45 (see Fig. 1B). Percentages of total A␤ for each A␤ species were calculated as % of the sum of A␤ species generated at 0.5 and 2.5 M C99-FLAG. Data represent the mean Ϯ S.D. (n ϭ 3). AU indicates arbitrary units.

K m (K app )
V max % of total A␤ with less than 10% in WTPS1. Signals from longer A␤s were included in the quantification of A␤ for G384A MTPS1. Weak additional signals for longer A␤s were also observed in WTPS2 and N141I MTPS2 and similarly included in quantification of A␤ (Fig. 3, C and D). Identification of Longer A␤s, A␤48 and A␤49, Produced by ␥-Secretase-Our previous observations on -cleavage itself led to predict the two counterparts of AICDs, namely A␤48 and A␤49 (14). Consistent with this assumption, the cells expressing A␤48 or A␤49 secreted A␤40 and A␤42 into media in a PS-dependent manner (15).
By using this solubilized system, we sought to confirm ␥-secretasedependent production of A␤48 and A␤49. ␥-Secretase complex that has been known to contain mature (highly glycosylated) nicastrin (39) was immunoprecipitated with anti-nicastrin antibody from 0.25% CHAPSOsolubilized MEF lysate. This partially purified ␥-secretase complex was incubated with C99-FLAG, and the reaction mixtures were subjected to gel system II to detect A␤48 and A␤49. The immunoprecipitate from WTMEF by anti-nicastrin antibody produced A␤ and AICD from C99-FLAG, whereas that by preimmune serum generated negligible levels of A␤ and AICD (Fig. 4A). Precipitate from PS1/2-deficient MEF contained only an immature form of nicastrin and generated no products (Fig. 4A). Gel II clearly shows the two distinct bands for A␤48 and A␤49 in the immunoprecipitate from WTMEF but none in that from PS1/2deficient MEF (Fig. 4B). Furthermore, A␤48 and A␤49 were profoundly reduced by the addition of L-685,458, in a dose-dependent manner (data

TABLE 2 Effect of FAD mutations of APP on apparent K m (K app ) and V max values for A␤ and AICD
Defined amounts of mtC99-FLAG substrate were incubated with 0.25% CHAPSO lysate from the microsomal fraction of CHO cells for 3 h. Protein samples were subjected to Western blotting with synthetic A␤ and E. coli-expressed AICD (see text) being the authentic standards. A␤ and AICD produced from wtC99-FLAG or mtC99-FLAG were visualized with 82E-1 and UT-421, respectively, and their intensities were quantified (see Fig. 3A). Data represent the mean Ϯ S.D. (n ϭ 3). not shown). These data indicate that ␥-secretase does produce A␤48 and A␤49 as well as other A␤s. A␤46 was undetectable in this CHAPSO-solubilized system, for which currently we cannot offer an appropriate explanation (22,27,33,36).
The CHAPSO lysate of CHO cells was incubated with C99-FLAG at 37°C. Uncleaved C99-FLAG was removed by the pretreatment with 4G8, and AICD-FLAG was subsequently precipitated with Anti-FLAG M2-agarose. The immunoprecipitate was subjected TOF-mass spectrometry. Most surprisingly, even under this solubilized condition, only two species of AICD, AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG, were detected (6, 14) (Fig. 4C), which contrasts with the presence of various A␤ species generated, as shown by the gel system I (Fig. 1A).
To confirm further the stoichiometry between A␤ and AICD produced by ␥-secretase, immunoprecipitated ␥-secretase was used to avoid possible contamination with proteases. In addition, instead of  D, left panel). In G384A MTPS1, an additional band, which was included for the quantification of A␤, was detected above the A␤ band and identified as a mixture of A␤48 and A␤49 (data not shown). As found, MTPS1/2 produced reduced amounts of A␤ and AICD, but both were equivalently generated (right panel). The stoichiometry between A␤ and AICD was not altered across various C99s and PS1/2s (Kruskal-Wallis test). Asterisks in C indicate the samples for which 4-fold more than other samples was loaded onto the gel.  bilized membrane (Fig. 4D, left panel). More importantly, purified ␥-secretases from three cell lines similarly exhibited one-to-one stoichiometry between A␤ and AICD (Fig. 4D, right panel). Again, only two AICDs were detectable (6, 14) (Fig. 4C).
Similarly, the effects of MTPS1/2 on the A␤ and AICD species generated were investigated. As was the case with WTC99-FLAG, WTPS1/2 showed robust production of A␤42 and A␤43 (Fig. 6A) and generated AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG, with the latter giving a stronger signal (Fig. 6B). Although M146L MTPS1 produced A␤40 as did WTPS1, M233T and G384A MTPS1 and N141I MTPS2 similarly exhibited an increase in the A␤42/A␤40 ratio on gel I (Fig. 6A). In addition, it is of note that each MTPS1/2 invariably provides higher signal intensity for AICD-(49 -99)-FLAG relative to that for AICD-(50 -99)-FLAG (Fig. 6B). Most interestingly, M233T MTPS1 produced only AICD-(49 -99)-FLAG, as did T714I and V717F MTC99-FLAG (see Fig. 5B). After incubation, the reaction mixture was subjected to the gel system II. Although both A␤48 and A␤49 were detectable in the WTPS1-containing reaction mixture, only A␤48 was identified in the M233T MTPS1-containing reaction mixture (Fig. 6C). . A␤ (A) and AICD (B) species generated from WT or MTC99-FLAG. A, each WT or MTC99-FLAG at 500 nM was incubated with a CHAPSOsolubilized lysate of CHO cells, and the reaction mixture was subjected to gel system I. In WTC99-FLAG, robust signals for A␤42 and A␤43 as well as A␤40 were detected on the blot. V717F mutation resulted in similar proportions of A␤ species. In contrast, T714I and I716F mutations led to predominant production of A␤42. L723P mutation generated neither A␤ nor AICD in our hands. Arrowheads indicate carboxyl-terminally truncated C99-FLAG fragments. B, the generated AICD species were immunoprecipitated and subjected to TOF-mass spectrometry. AICD-(50 -99)-FLAG and AICD-(49 -99)-FLAG were identified from a WTC99-FLAG-containing reaction mixture with the former signal being stronger. Most interestingly, only AICD-(49 -99)-FLAG was detected from the reaction mixtures containing T714I and V717F MTC99-FLAG. In artificial mutant I716F, both AICD-(50 -99)-FLAG and AICD-(49 -99)-FLAG were identified despite predominant production of A␤42. These indicate that FAD mutations downstream of the ␥-cleavage sites cause significant alterations in the -cleavage sites.  A␤ (A and C) and AICD (B) species generated by MTPS1/2. The CHAPSO lysates of CHO cells expressing MTPS1/2 were incubated with C99-FLAG. WTPS1 showed robust production of A␤42 and A␤43 in addition to production of A␤40 (A). Asterisks indicate the samples for which the volume that was loaded onto the gel was 4-fold more than that of the other samples. Although in M146L MTPS1, the proportions of A␤ species produced were similar to those in WTPS1/2, G384A MTPS1 and N141I MTPS2 similarly exhibited an increase of A␤42/A␤40 ratio (A). TOF-mass spectrometry showed higher signal intensity for AICD-(49 -99)-FLAG relative to that for AICD-(50 -99)-FLAG in MTPS1/2, which is contrast with WTPS1/2 (B). M233T MTPS1 produced A␤42 predominantly (A) and only AICD-(49 -99)FLAG (B). Gel system II shows that both A␤48 and A␤49 were detected in case of WTPS1, but only A␤48 was detected in M233T MTPS1, probably as a counterpart of AICD-(49 -99)-FLAG (C). Arrowheads indicate carboxyl-terminally truncated C99-FLAG.
These observations strongly suggest that A␤48 is a counterpart of AICD-(49 -99)-FLAG exclusively generated by M233T MTPS1. These also warrant the accuracy of identification of longer A␤ by the present gel systems.

DISCUSSION
␥-Secretase cleaves type I membrane proteins within their transmembrane domains (40). The type I membrane proteins as the substrates of ␥-secretase are characterized by ectodomain shedding, which is mediated by a wide variety of membrane-bound proteases (sheddases) (41). The relationship between ectodomain shedding and intramembrane proteolysis is well known for the processing of APP, in which ectodomain shedding occurs through ␣or ␤-cleavage (1). The cleavages at the ␣and ␤-sites are necessary and sufficient for the execution of ␥-cleavage, which results in A␤ production. However, the relationship between ␥and -cleavages has remained unclear (42). Although several reports implied that ␥and -cleavages are independent processes (28 -31), we found previously that there is a significant correlation between ␥and -cleavages (14,15). Here we have further examined the relationship by taking advantage of a solubilized ␥-secretase assay system with high specific activity, in which use of recombinant C99-FLAG substrate allowed us to exclude contamination by C83derived AICD. Our major finding using this CHAPSO-solubilized assay system is that equal amounts of A␤ and AICD are produced from C99 and that only two AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG are detectable. Even in MTAPP and MTPS1/2, the stoichiometry between A␤ and AICD is almost identical.
In the this study, we identified nothing but AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG, despite detection of various A␤ species, including A␤40, A␤42, A␤43, A␤45, A␤48, and A␤49. This is surprising because the present CHAPSO-solubilized system warrants free collision of enzyme and substrate. One may point out that very hydrophobic peptides often cannot be detected by TOF-mass spectrometry, and thus longer AICDs, even if they exist, may not be detected. However, we previously detected AICD-(46 -99), which was produced by the membrane prepared from cells transfected with artificial MTAPP (Ew; -cleavage sites are replaced by a stretch of tryptophan) (43). This strongly suggests that no significant amounts of longer AICDs are produced in the assay system.
In our previous report, we detected A␤46 and A␤48 but failed to identify A␤49 in the cell lysates (27). But in the solubilized assay system, small or trace amounts of A␤48 and A␤49 were invariably detected. At present, we cannot explain the discrepancy between the present data and previous report. Possibly, the absence or presence of the membrane may have an effect on the processing of C99 by ␥-secretase. However, this study has clearly shown that the immunoprecipitated ␥-secretase generates A␤48 and A␤49 in a presenilin/ nicastrin-dependent manner. In addition, production of A␤48 and A␤49 was prevented by L-685,458 in a dose-dependent manner (data not shown). It is most likely that A␤48 and A␤49 are counterparts of AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG, respectively, and potential intermediates for A␤40, A␤42, and A␤43. In this context, it is of note that, even if -cleaved AICD-(49 -99)-FLAG and AICD-(50 -99)-FLAG are predominant in the reaction mixture, only a trace or very small amounts of A␤48 and A␤49 can be detected. This indicates that, once generated, A␤48 and A␤49 promptly undergo ␥-cleavage, which occurs at the carboxyl termini of Ile-45, Thr-43, Ala-42, and Val-40, presumably depending upon the affinity between the cleavage sites of the substrate and catalytic sites of ␥-secretase (Fig. 1A). Thus, the abundance of various A␤ species in the reaction mixture may reflect varying affinities between susceptible sites and catalytic sites of ␥-secretase. The order of susceptibility to ␥-secretase (in decreasing order) on the carboxyl terminus is Val-40, Ala-42, Thr-43 Ͼ Ͼ Ile-45 Ͼ Leu-49 Ͼ Thr-48. Most interestingly, A␤46 is undetectable, and the carboxyl terminus of Val-46 is least susceptible in this assay system, which contrasts sharply with the ready detectability of A␤46 in the cell-free system (22,27). In addition, DAPT treatment of solubilized ␥-secretase uniformly suppressed the levels of A␤40, A␤42, and A␤43 but failed to accumulate A␤43 and A␤46 (22,27) (Fig. 1E). Thus, one should bear in mind that the present ␥-secretase assay system can simulate only a part of cell-based or cell-free system. Nevertheless, MTC99 and MTPS1/2 produced a higher proportion of A␤42 and a higher proportion of AICD-(49 -99), thus still maintaining the important characteristics of MTAPP and MTPS1/2.
The increased proportion of AICD-(49 -99) produced by MTC99 and MTPS1/2 in the CHAPSO-solubilized ␥-secretase assay system is consistent with our previous report on AICD production in the cell-free system (14). Interestingly, T714I and V717F MTC99 generated only AICD-(49 -99). It would be of interest to identify longer A␤s in these MTC99-containing reaction mixtures. Unfortunately, we were unable to identify corresponding longer A␤s species because they have altered mobilities on the gel, because of amino acid substitutions. In our experience it is not possible to identify longer A␤s by TOF-mass spectrometry, presumably because of their hydrophobicity.
Alternative proof that -cleavage is followed by ␥-cleavage could be provided by a reciprocal experiment. We attempted to generate substrates CTF-(41-99)-FLAG and CTF-(43-99)-FLAG, potential counterparts of A␤40 and A␤42, respectively, and we examined whether AICD was produced from those substrates. When those CTF-FLAG substrates, aminoterminally fused with signal peptide of ␣-galactosidase A, were expressed in Sf9, most of purified CTF-FLAG fragments still had uncleaved signal peptide at the amino terminus, which were not appropriate for the reciprocal experiment. Although we failed to do this experiment, Shah et al. (44) recently proposed a very interesting model for the mechanism of ␥-cleavage. They found that nicastrin acts as a gatekeeper in ␥-secretase complex by binding to an amino terminus of the substrate (44). According to their model, the amino terminus of the substrate should protrude from membrane to the extracellular/luminal side. In view of this, it is unlikely that CTF-(41-99) and CTF-(43-99) act as ␥-secretase substrates, because of absence of their ectodomains. Taken together, our data support the view that -cleavage precedes ␥-cleavage and triggers subsequent ␥-cleavage to produce various A␤ species.
The most striking feature of our assay system is high specific activity of ␥-secretase. Apparent V max values of A␤ and AICD were ϳ400 pM/min. These values are roughly more than 30-fold higher than those reported previously (21,35). This high activity of our assay system makes possible unambiguous detection of A␤ and AICD by conventional Western blotting. In fact, we successfully determined the stoichiometry between A␤ and AICD and detected minor A␤ species, such as A␤45, A␤48, and A␤49, which are probably unable to be detected by enzyme-linked immunosorbent assay, by combination of new gel systems I and II. The reasons for higher activity of ␥-secretase in our assay system would be as follows. First, 0.25 M sucrose was added to the reaction mixture. In the presence of such polyol, ␥-secretase might be stabilized and active in cleaving substrates for a longer time. We examined the effects of 0.25 M sucrose on ␥-secretase activity, and we observed 4 -5-fold A␤ production in the presence of sucrose, but such an increase is not enough to account for the present high activity. 6 Second, the substrate C99-FLAG was expressed in and purified from Sf9 cells. We thought that protection of the transmembrane domain of substrate with native lipids would be important to take a right conformation with which the substrate readily interacts with ␥-secretase. Even after solubilization with Nonidet P-40, C99-FLAG may still retain cellular lipids derived from Sf9 cell membranes. In this condition, the addition of phosphatidylcholine to the reaction mixture further enhanced ␥-secretase activity. This cannot be realized by the bacterial expression system in which the expressed substrate was recovered as insoluble inclusion bodies probably caused by interaction through its exposed transmembrane domain. Once inclusion bodies were formed, it would be difficult for the substrate to become soluble and to take a right conformation, even if solubilized, by subsequent addition of lipid or detergent. Thus, we believe that the substrate expressed in and purified from Sf9 cells greatly contributes to the high specific activity of our solubilized assay system. Accordingly, C99-FLAG substrate from Sf9 cells was compared with that from E. coli in terms of A␤ production. Roughly 7-fold more A␤ was produced with C99-FLAG derived from Sf9 cells, compared with that from E. coli. 6 Third, the absence of a formyl group at the amino terminus of the substrate may also contribute to high activity of the assay system. As the substrate fused with a signal peptide was expressed in Sf9 cells, the purified substrate should possess a free amino terminus after removal of signal peptide. When expressed in E. coli, a protein is usually N-formylated, and this modification leads to significantly reduced ␥-secretase activity (44). It would be possible that the free amino terminus of the substrate is required for its efficient interaction with ␥-secretase.