γ-Secretase Substrate Concentration Modulates the Aβ42/Aβ40 Ratio

Mutation of the amyloid precursor protein (APP), presenilin-1, or presenilin-2 results in the development of early onset autosomal dominant forms of Alzheimer disease (AD). These mutations lead to an increased Aβ42/Aβ40 ratio that correlates with the onset of disease. However, it remains unknown how these mutations affect γ-secretase, a protease that generates the termini of Aβ40 and Aβ42. Here we have determined the reaction mechanism of γ-secretase with wild type and three mutated APP substrates. Our findings indicate that despite the overall outcome of an increased Aβ42/Aβ40 ratio, these mutations each display rather distinct reactivity to γ-secretase. Intriguingly, we found that the ratio of Aβ42/Aβ40 is variable with substrate concentration; increased substrate concentrations result in higher ratios of Aβ42/Aβ40. Moreover, we demonstrated that reduction of γ-secretase substrate concentration by BACE1 inhibition in cells decreased the Aβ42/Aβ40 ratio. This study indicates that biological factors affecting targets such as BACE1 and APP, which ultimately cause an increased concentration of γ-secretase substrate, can augment the Aβ42/Aβ40 ratio and may play a causative role in sporadic AD. Therefore, strategies lowering the Aβ42/Aβ40 ratio through partial reduction of γ-secretase substrate production may introduce a practical therapeutic modality for treatment of AD.

Mutation of the amyloid precursor protein (APP), presenilin-1, or presenilin-2 results in the development of early onset autosomal dominant forms of Alzheimer disease (AD). These mutations lead to an increased A␤42/A␤40 ratio that correlates with the onset of disease. However, it remains unknown how these mutations affect ␥-secretase, a protease that generates the termini of A␤40 and A␤42. Here we have determined the reaction mechanism of ␥-secretase with wild type and three mutated APP substrates. Our findings indicate that despite the overall outcome of an increased A␤42/A␤40 ratio, these mutations each display rather distinct reactivity to ␥-secretase. Intriguingly, we found that the ratio of A␤42/A␤40 is variable with substrate concentration; increased substrate concentrations result in higher ratios of A␤42/A␤40. Moreover, we demonstrated that reduction of ␥-secretase substrate concentration by BACE1 inhibition in cells decreased the A␤42/A␤40 ratio. This study indicates that biological factors affecting targets such as BACE1 and APP, which ultimately cause an increased concentration of ␥-secretase substrate, can augment the A␤42/A␤40 ratio and may play a causative role in sporadic AD. Therefore, strategies lowering the A␤42/A␤40 ratio through partial reduction of ␥-secretase substrate production may introduce a practical therapeutic modality for treatment of AD.
␥-Secretase cleaves the amyloid precursor protein (APP) to generate the C termini of ␤-amyloid (A␤) 2 peptides, generally 40 or 42 amino acids in length (A␤40 and A␤42, respectively). A␤ peptides are believed to be a major causative factor in the pathogenesis of Alzheimer disease (AD) (1). A␤42 is more prone to aggregation than A␤40 (2), and therefore biological or environmental factors that promote increased A␤42 production accelerate the pathological cascade leading to AD. Expression of A␤42, rather than A␤40, in Drosophila and mice leads to the formation of A␤ plaques (3,4). Furthermore, mouse model studies suggest that the ratio of A␤42/A␤40, rather than total amount of A␤, correlates with the load of characteristic AD plaques in the brain (5,6). Moreover, evidence suggests A␤40 may play a beneficial role in that it antagonizes A␤42 aggregation (5,6). Therefore, inhibition of ␥-secretase activity that specifically generates A␤42 or reduction of the Ab42/A␤40 ratio would be an appealing strategy for treatment of AD. However, despite intensive studies on ␥-secretase, the mechanism of cleavage specificity for ␥-secretase is still unknown.
APP was the first gene found to be linked with inherited AD (7). Each mutation surrounding the ␥-secretase cleavage site appears to alter the production of A␤40 and A␤42. Suzuki et al. (8) demonstrated that mutating APP at Val-46 to Phe or Ile increased the ratio of secreted A␤42 to A␤40 in transfected cells. An increased ratio of A␤42/A␤40 was also observed with other mutations (9 -11). De Jonghe et al. (9) found certain mutations enhanced the stability of the ␥-secretase substrates known as C-terminal fragments (␤CTF and ␣CTF), and Ancolio et al. (12) reported that the V44M mutation influences ␣-secretase cleavage. Therefore, secretion of A␤ from APP mutation-transfected cells is affected by a multitude of factors. How these mutations affect APP binding and reaction with ␥-secretase remains unknown.
In this study, we have developed a simplified system using small peptide substrates that allow for the mechanistic characterization of APP FAD mutations. Kinetic analyses (k cat /K m ) suggest that all of these APP mutations enhance the preference of ␥-secretase for the 42-site over the 40-site cleavage and these changes are generally caused by k cat , rather than K m . Importantly, we have revealed that the amount of ␥-secretase substrate governs the ratio of A␤42/A␤40 in vitro and in cells. Increased substrate concentrations result in higher ratios of A␤42/A␤40.

EXPERIMENTAL PROCEDURES
Peptide and Compound Synthesis-Wild type (WT) and mutant APP-TM peptides were synthesized and purified by Midwest Biotech (Fig. 1). Biotinylated standard peptides (P40 and P42) were synthesized in our laboratory using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase chemistry on a peptide synthesizer (Protein Technologies, Inc.). Peptides were purified by reversed-phase high performance liquid chro-matography using a semi-preparative C18 column. The identity of these peptides was confirmed by liquid chromatography tandem mass spectrometry analysis (Agilent Technologies). Compound 3 was synthesized as described (13).
In Vitro ␥-Secretase Assay-␥-Secretase assay was the same as described previously (14) except for substrate identity. Briefly, the wild type or mutant APP-TM peptides were incubated with HeLa cell membrane in the presence of CHAPSO (0.25%) in buffer A (50 mM PIPES, pH 7.0, 5 mM MgCl 2 , 5 mM CaCl 2 , 150 mM KCl) at 37°C for 2.5 h. The reactions were stopped by adding radioimmune precipitation buffer (150 mM NaCl, 1.0% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0). Then the reaction mixtures were incubated with ruthenylated G2-10 or G2-11 antibody, which specifically recognizes products P40 and P42, respectively. The antibody reactions were detected by electrochemiluminescence (ECL). The G2-10 and G2-11 antibodies were ruthenylated with ruthenium (II) tris-bipyridine N-hydroxysuccinimide ester (Biovarice Inc.) according to the manufacturer's instructions. The concentrations of P40 and P42 were determined using synthetic peptide standards. The K m and V max were determined from the Michaelis-Menten equation Cell-based Assay for A␤ Production-N2A cells that stably express APP Swedish mutation were incubated with ␥-secretase or BACE1 inhibitors. Twenty-four hours later, the conditioned medium was removed and assayed for A␤ production. A␤40 and A␤42 were detected with biotinylated 6E10 paired with ruthenylated G2-10 or G2-11 antibodies, respectively. A␤x40 and A␤x42 were detected with biotinylated 4G8 paired with ruthenylated G2-10 or G2-11 antibodies, respectively (15). Concentration of A␤ peptides is calculated from standard curves that are generated using synthetic A␤40 and A␤42 peptides using the ECL assay.

RESULTS AND DISCUSSION
␤CTF, a ␤-secretase-generated C-terminal fragment of APP, has been utilized as a substrate of ␥-secretase to biochemically characterize this protease (14,16). Prior studies suggest that this substrate binds to the active site and docking site of ␥-secretase (17). To elucidate the reaction mechanism, we developed a sensitive in vitro assay in which the substrate solely interacts with the active site of ␥-secretase. We chose a biotinylated peptide substrate that contains a transmembrane domain (APP-TM) plus three lysine residues of APP (Fig. 1A). Similar experimental conditions as used for the C100 FLAG assay were used in this APP-TM substrate assay (14). Our peptide substrate was incubated with HeLa cell membrane in the presence of 0.25% CHAPSO, and the P40 and P42 products (reference to products resulting from the cleavage at the 40 and 42 sites) were detected with G2-10 and G2-11 antibodies, respectively. The ␥-secretase activity against this substrate was characterized to determine optimal conditions (supplemental Fig. S1). In our assay both the substrate and products are biotinylated. To eliminate the possibility of signal saturation under high concentrations of substrate due to the lack of streptavidin magnetic beads, we assayed 5, 10, and 25 l of total ␥-secretase reaction mixture, each containing increasing amounts of biotinylated peptide. All three assay conditions resulted in the same substrate dependence (supplemental Fig. 1C), demonstrating that streptavidin beads are not limited under our assay conditions. Therefore, the ␥-secretase cleavage reaction with APP-TM as a substrate displayed a steady state kinetic mechanism. P42 production was 11.5% of the total sum of P40 and P42. ␥-Secretase activity for production of both P40 and P42 was inhibited by L-685,458, with an IC 50 of 1.5 and 1.3 nM, respectively. The three products, P38, P40, and P42, were confirmed by liquid chromatography tandem mass spectrometry analysis (supplemental Fig. 2). Under current assay conditions, our mass spectrometry analysis was not able to detectand ⑀-cleavages (18,19). Nevertheless, this study focuses on ␥-cleavage sites (P40 and P42) that have been well characterized in association with AD. Hence, this novel assay using the APP-TM substrate recapitulates native characteristics and is a suitable assay to characterize ␥-secretase activity for A␤40 and A␤42 production.
We examined the inhibition patterns of L-685,458 and Compound E, which represent two classes of structurally distinct ␥-secretase inhibitors, against this peptide substrate. Double reciprocal (Lineweaver-Burke) analyses (Fig. 1, B and C) showed that the plots of L-685,458 intersect on the 1/v axis, whereas the plots of Compound E interface on the left of the 1/[s], which are indicative of competitive and non-competitive inhibition, respectively. In other words, L-685,458, an expected transition state analog, is a competitive inhibitor against substrate APP-TM (Fig. 1B), whereas Compound E, which contains a benzodiazepine moiety, displays non-competitive behavior (Fig. 1C). These findings strongly suggest that this small peptide substrate (APP-TM) solely binds to the active site of ␥-secretase. This differs from the C100 substrate (17), which interacts with both the active site and the docking site of ␥-secretase (Fig. 1D). Furthermore, it suggests that L-685,458 and Compound E bind to different sites on ␥-secretase, as previously reported (20,21). This study provides experimental evidence that L-685,458 is a transition state inhibitor, further validating prior conclusions that the presenilins contain the active site of ␥-secretase (22).
We next evaluated the effect of APP mutations on P40 and P42 production. Three APP mutant substrates were selected (Fig. 1). Among them, two were FAD mutants that reside at positions 44 (V44M, named French mutation V715M) and 45 (I45V, named Florida mutation I716V) (Fig. 1A). The V45F mutation is a non-natural mutant discovered by phenylalanine mutagenesis of the transmembrane domain of APP (23). At a concentration of 0.5 M, each substrate was incubated with HeLa cell membrane and the initial rate of P40 and P42 production was determined (Table 1). Our findings indicate that two mutations (I45F and I45V) reduced P40 production and increased P42 production, whereas the V44M mutation augmented the production of both P40 and P42. Interestingly, the I45F mutation led to production of higher quantities of P42 than P40, which is in contrast to all other mutations. Nevertheless, the overall outcome of each of these mutations is an increased ratio of P42/P40.
To elucidate the mechanism of ␥-secretase catalysis, the kinetic parameters of ␥-secretase for these substrates were determined (Table 2). Surprisingly, there were different K m values for P40 and P42 production and all these substrates exhibited a higher apparent K m for P42 than that for P40. If these differences truly reflect the affinity between substrate and ␥-secretase, we predict that the ratio of P42 to P40 will be altered as substrate concentration is varied. In other words, a substrate concentration that saturates for P40 production, but not for P42, will lead to a higher ratio of P42 to P40. Thus, we determined the ratio of P42 to P40 at different substrate concentrations of WT and I45F-mutated APP-TM. Ratios of P42 to total product for WT substrate were 13 and 22% at concentra-  tions of 0.5 and 2 M, respectively, and for the I45F substrate were 57 and 75%, respectively ( Fig. 2A). This observation strongly suggests that ␥-secretase has distinct affinities to the same substrate for P40 and P42 production. There was no significant difference between K m values of WT and the mutated substrates for P40 and P42 activities. These mutations significantly influence V max (maximum velocity) for both P40 and P42 production ( Table 2). V max values for I45F, V44M, and I45V substrates for production of P40 are 0.25-, 1.15-, and 0.70-fold of WT, respectively. For P42, they are 1.96-, 4.56-, and 2.73-fold of WT, respectively. These point mutations conclusively affect catalysis, rather than affinity for substrate. Importantly, we next examined the catalytic efficiency (k cat /K m or V max /K m ) of ␥-secretase against each substrate. k cat /K m is a secondorder rate constant that informs how the enzyme performs when the substrate concentration is low and indicates an enzyme's preference for different substrates (also called specificity constant). Considering that V max ϭ k cat * E] total (total concentration of ␥-secretase) and the same concentrations of ␥-secretase were used for each substrate, the ratio of V max /K m (mutation) to V max /K m (WT) are equal to the ratio of their k cat /K m . We analyzed the V max /K m of these substrates from two perspectives. First, we compared mutated substrate versus WT. For P40 production V44M was a better substrate (2.12-fold) than WT, while both I45F (0.24-fold) and I45V (0.78-fold) were poorer substrates. For P42, all three mutations were better substrates, ranging from 2.12-to 3.60-fold compared with WT. Although each mutation displays a unique effect on P40 and P42 production, the overall tendency for all three mutations was to increase the relative specificity for P42 production. Secondly, we determined the relative k cat /K m for each substrate for P40 and P42 production. The relative specificity for P40 over P42 is 9.21, 1.01, 5.42, and 2.53 for  Table 2. Inset, substrate concentration from 0.01 to 2 M. FIGURE 3. The cellular relationship between the P42/P40 ratio and substrate concentration. A, structure of Compound 3. B, effect of Compound 3 on A␤40 and A␤42 production. The N2A cells that stably express APP Swedish mutation were treated with Compound 3, and secreted A␤40 and A␤42 were determined with biotinylated 6E10/ruthyluated G2-10 or biotinylated 6E10/ruthyluated G2-11, respectively (n ϭ 6). C, effect of Compound 3 on A␤x40 and A␤x42 production. Secreted A␤40 and A␤42 were determined with biotinylated 4G8/ruthyluated G2-10 or biotinylated 4G8/ruthyluated G2-11, respectively (n ϭ 6). D, relative ratio of A␤42/ A␤40 or A␤x42/A␤x42. *, p Ͻ 0.05; **, p Ͻ 0.01.
WT, I45F, V44M, and I45V, respectively. These studies indicate that ␥-secretase generally prefers the 40-site cleavage when the WT substrate concentration is low and suggest that ␥-secretase is responsible for the 40-site processing under physiological conditions. Mutations alter the preference of ␥-secretase for 40versus 42-site cleavage, and these changes are generally attributable to k cat rather than K m . Moreover, we calculated the overall effect on the rate of ␥-secretase for P40 and P42 production based on the V max and K m values of WT (Fig. 2B). This analysis revealed that an increase in substrate concentration not only results in generation of more total amount of A␤ but also leads to an increased ratio of A␤42/A␤40, which may be directly related to the pathogenesis of AD.
The next critical question is whether substrate concentration correlates with the ratio of A␤42/A␤40 at the cellular level. BACE1 is responsible for generation of ␤CTF, a ␥-secretase substrate; therefore, we chose to control the ␤CTF level through inhibition of BACE1 activity. N2A cells that stably express APP Swedish mutation were treated with a BACE1 inhibitor, Compound 3 (13) (Fig. 3A) at 1 and 3 M. The amounts of secreted A␤40, A␤x40 (that includes A␤1-40 and A␤17-40), A␤42, and A␤x42 were estimated using an ECL assay with synthetic peptides as standards. Compound 3 at 1 and 3 M inhibits 33 and 39% of A␤40 and 47 and 55% of A␤42, respectively (Fig. 3B). Similarly, this inhibitor at 1 and 3 M suppresses 44 and 52% of A␤x40 and 61 and 72% of A␤x42, respectively. Treatment of APP Swedish mutation cells with BACE1 inhibitor at 1 and 3 M reduced the relative ratio (compared with "no treatment" assigned a value of 1) of A␤42/A␤40 to 0.81 and 0.73 and A␤x42/A␤x40 to 0.70 and 0.58, respectively (Fig. 3C). Western analysis confirmed that this compound has no effect on the level of APP (supplemental Fig. S3). These cellular studies therefore reinforce our assertion that the A␤42/A␤40 ratio is dependent on the substrate concentration of ␥-secretase. This finding corroborates other studies showing that up-regulation of BACE1 activity by Par4 elevates the ratio of A␤42/A␤40 (24). Our discovery illustrates that the A␤42/ A␤40 ratio is variable with ␥-secretase substrate concentration and provides critical insight into the pathogenesis of sporadic AD. Any biological or environmental factors that promote increased levels of ␥-secretase substrate likely will result in an increase in the A␤42/A␤40 ratio, which is similar to the effect of presenilin-1, presenilin-2, and APP mutations (25)(26)(27). Recent studies have shown that BACE1 activity is increased in sporadic AD brains and is correlated with A␤ load (28,29). Our studies suggest that higher BACE1 activity in AD patients increases the production of ␤CTF and ultimately leads to a higher ratio of A␤42/A␤40, which is associated with the pathological state of the disease. A slight increase in ␤CTF production resulting from BACE1 cleavage leads to a mild elevation of the A␤42/ A␤40 ratio (see Fig. 2B) and could chronically be detrimental to neuronal cells. Multiple factors have been found to regulate BACE1 expression and activity, such as Par4 (24) and HIF-1␣ (30). The role of these proteins in regulation of BACE activity in AD patients needs to be investigated.
In summary, this study demonstrates that in addition to the clinical mutations of presenilin-1, presenilin-2, and APP, ␥-secretase substrate concentration can increase the ratio of A␤42/A␤40, which may play a critical role in the pathogenesis of sporadic AD (Fig. 4). Moreover, BACE inhibitors that reduce ␤CTF production are capable of lowering the ratio of A␤42/ A␤40 in addition to reducing the total amount of A␤. Partial inhibition of BACE1 activity could reduce the A␤42/A␤40 ratio and represent a practical strategy for AD therapies.