Pen2 and Presenilin-1 Modulate the Dynamic Equilibrium of Presenilin-1 and Presenilin-2 γ-Secretase Complexes*

γ-Secretase is known to play a pivotal role in the pathogenesis of Alzheimer disease through production of amyloidogenic Aβ42 peptides. Early onset familial Alzheimer disease mutations in presenilin (PS), the catalytic core of γ-secretase, invariably increase the Aβ42:Aβ40 ratio. However, the mechanism by which these mutations affect γ-secretase complex formation and cleavage specificity is poorly understood. We show that our in vitro assay system recapitulates the effect of PS1 mutations on the Aβ42:Aβ40 ratio observed in cell and animal models. We have developed a series of small molecule affinity probes that allow us to characterize active γ-secretase complexes. Furthermore we reveal that the equilibrium of PS1- and PS2-containing active complexes is dynamic and altered by overexpression of Pen2 or PS1 mutants and that formation of PS2 complexes is positively correlated with increased Aβ42:Aβ40 ratios. These data suggest that perturbations to γ-secretase complex equilibrium can have a profound effect on enzyme activity and that increased PS2 complexes along with mutated PS1 complexes contribute to an increased Aβ42:Aβ40 ratio.

trolling the specificity of ␥-secretase activity for A␤40 and A␤42 production has not been resolved. It has been found that PS1 mutations affect the formation of ␥-secretase complexes (28). However, the precise mechanism by which individual subunits alter the dynamics of ␥-secretase complex formation and activity is largely unresolved. A better mechanistic understanding of ␥-secretase activity associated with FAD mutations has been hindered by the lack of suitable assays and probes that are necessary to recapitulate the effect of these mutations seen in cell models and to characterize the active ␥-secretase complex.
In our present studies, we have determined the overall effect of Pen2 and PS1 expression on the dynamics of PS1-and PS2containing complexes and their association with ␥-secretase activity. Using newly developed biotinylated small molecular probes and activity assays, we revealed that expression of Pen2 or PS1 FAD mutants markedly shifts the equilibrium of PS1containing active complexes to that of PS2-containing complexes and results in an overall increase in the A␤42:A␤40 ratio in both stable cell lines and animal models. Our studies indicate that perturbations to the equilibrium of active ␥-secretase complexes by an individual subunit can greatly affect the activity of the enzyme. Moreover they serve as further evidence that there are multiple and distinct ␥-secretase complexes that can exist within the same cells and that their equilibrium is dynamic. Additionally the affinity probes developed here will facilitate further study of the expression and composition of endogenous active ␥-secretase from a variety of model systems.
Cell Lines-N2A cell lines that stably express human wild type (WT) PS1 and FAD mutants were cultured as described previously (31). The HEK293-APP stable cell line (a gift from Dr. Elizabeth Chen, Merck Research Laboratories) was transfected with Myc-tagged Pen2 expressed from the pcDNA 4.1 vector and selected with Zeocin (Invitrogen) at 0.4 -0.8 mg/ml. Cultures were maintained in Dulbecco's modified Eagle's medium high glucose containing 10% fetal calf serum and antibiotics.
Membrane Preparation and in Vitro ␥-Secretase Assay-Membrane fractions were isolated from HeLa-S3 cells (National Cell Culture Center). Briefly 1-liter equivalents were resuspended in Buffer A (50 mM MES, pH 6.0, 150 mM KCl, 5 mM CaCl 2 , 5 mM MgCl 2 , and protease inhibitors) and lysed by passage through a French press. Nuclear debris were pelleted, and the resulting supernatant was ultracentrifuged 100,000 ϫ g for 1 h. The resulting pellet represented the total membrane fraction. Total membrane fractions were solubilized with 1% CHAPSO in Buffer A for 1 h 4°C. Non-solubilized material was removed by ultracentrifugation at 100,000 ϫ g for 1 h, and the resulting supernatant represented the CHAPSO-solubilized fraction. Protein concentration was determined with the DC Protein Assay kit (Bio-Rad) according to the manufacturer's instructions. ␥-Secretase activity was measured by electrochemiluminescence as described previously (18,33). CHAPSOsolubilized membrane was incubated in Buffer B (50 mM PIPES, pH 7.0, 150 mM KCl, 5 mM CaCl 2 , 5 mM MgCl 2 , and protease inhibitors) with 0.25% (v/v) CHAPSO, 1 M substrate, and 0.1% bovine serum albumin (v/v) in the presence or absence of ␥-secretase inhibitors for 2.5 h 37°C. The reaction mixture was incubated with biotinylated 4G8 and ruthenylated G2-10 or G2-11 in Buffer C (1ϫ phosphate-buffered saline, 0.5% (w/v) bovine serum albumin, and 0.5% (v/v) Tween 20) for 2 h at 25°C, and immunocomplexes were captured with magnetic streptavidin beads (Dynal). A␤40 and A␤42 production was measured by electrochemiluminescence on an Analyzer (Bio-Veris) and expressed as relative light units.
Affinity Capture of Endogenous ␥-Secretase and Western Blotting-CHAPSO-solubilized membrane was incubated in the presence or absence of 2 M L458 in Buffer B for 0.5 h at 37°C. A biotinylated affinity probe at 20 nM was added for an additional 1.5 h at 37°C. Streptavidin-agarose (Pierce) was added to the reaction and incubated overnight 4°C. Captured complexes were washed in Tris-buffered saline with 0.1% (v/v) Tween 20 and eluted with 2ϫ Laemmli sample buffer. Samples were resolved by SDS-PAGE and transferred to a polyvinylidene difluoride membrane. The following antibodies were used for Western blotting: PS1-NTF (1:1000), PS1-CTF (1:1000), Aph1aL (1:250), Nct (1:1000), Pen2 (1:500), and PS2-CTF (1:1000). Anti-mouse or anti-rabbit horseradish peroxidaseconjugated (Amersham Biosciences) secondary antibodies were used in conjunction with standard electrochemiluminescence detection methods. In all cases, blots shown are representative of three or more experiments.
Photolabeling of ␥-Secretase-Solubilized membrane was incubated in the presence or absence of 2 M L458 in Buffer B for 0.5 h at 37°C. 20 nM biotinylated affinity probe was added for an additional 1 h at 37°C. Samples were then irradiated at 350 nm on ice for 0.75 h. The reaction was either denatured by addition of RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 0.1% (w/v) SDS, 1% (v/v) Nonidet P-40, and 0.5% (w/v) deoxycholic acid) or left in the native condition for 1 h at 25°C. Streptavidinagarose was added to the reaction mixture and incubated overnight at 4°C. Captured complexes were washed with RIPA buffer and eluted with 2ϫ Laemmli sample buffer. Samples were Western blotted as described above. In all cases, blots shown are representative of three or more experiments.

Pen2
Overexpression Causes an Increase in the A␤42:A␤40 Ratio in Cells and in Vitro-Pen2 is essential for ␥-secretase activity and is thought to stabilize the complex as well as promote endoproteolysis of PS (9,10). To further investigate the function of Pen2 on ␥-secretase activity and complex forma-tion, we generated two stable cell lines: HEK293 that overexpress APP (referred to as APP) and HEK293 that co-express APP and C-terminal Myc-tagged Pen2 (referred to as Pen2). Expression of Pen2 was confirmed by Western blotting with both anti-Myc and anti-Pen2 antibodies and showed considerable overexpression of Pen2 (Fig. 1a). Endogenous and tagged Pen2 was unresolved under our current SDS-PAGE condition. Furthermore tagged Pen2 co-immunoprecipitated with Nct ( Fig. 1b), suggesting that Pen2 is incorporated into the ␥-secretase complex as Pen2 incorporation is proposed as a final step in the formation of active ␥-secretase complexes (9). However, overexpression of Pen2 caused a reduction in the expression of the known components of ␥-secretase (Fig. 1c). This is consistent with earlier reports that showed that the C terminus of Pen2 is critical for complex formation (34) and that expression of a C-terminally tagged Pen2 (Pen2-Myc-His 6 ) resulted in destabilization of the PS1 heterodimer and mature Nct (8). There was no effect on the expression level of APP in the Pen2 cell line compared with the parental cell line (Fig. 1c).
We next determined the effect of Pen2-Myc expression on ␥-secretase activity for the 40 and 42 site cleavage of APP. Total membrane fractions isolated from the Pen2 and APP cell lines were incubated with the APP transmembrane substrate, and A␤40 and A␤42 cleavage products were measured (33). Production of A␤40 was greatly reduced in the Pen2 cell line, but there was no effect on the production of A␤42 (Fig. 2, left  panel). This resulted in an overall increase in the A␤42:A␤40 ratio in the Pen2 line (0.25) compared with the APP line (0.15) (Fig. 2, right panel). To verify the in vitro results, we examined the secretion of A␤40 and A␤42 peptides from both cell lines (Fig. 2, middle panel). Consistent with the in vitro data, we found that there was a marked reduction in A␤40 secretion in the Pen2 cell line but little effect on the secretion of A␤42 compared with the APP cell line. The A␤42:A␤40 ratio for the Pen2 cell line was 0.29 and was 0.09 for the APP cell line (Fig. 2, right panel). The very similar results in both the cellular and in vitro assays strongly suggest that overproduction of Pen2 alters ␥-secretase activity for A␤40 and A␤42 production, which is reminiscent of the effect of PS1 and PS2 FAD mutations.
Capture of the ␥-Secretase Complex with Biotinylated Active Site-directed Inhibitors Depends on the Length of the Biotin Linker-To better understand the role of tagged Pen2 in the formation of ␥-secretase complexes, we set out to examine the active ␥-secretase complexes present in the Pen2 and APP cell lines. It is known that only a small percentage of PS is engaged in catalytically active complexes (35,36). Therefore, traditional Western blotting and immunoprecipitation methods would be inadequate for our study because they fail to distinguish between catalytically active and inactive complexes. To overcome this, we attempted to develop small molecular affinity probes that would allow us to capture and characterize the active ␥-secretase complex under native conditions. L458 is a potent transition state analog that selectively binds to catalytically active ␥-secretase, and its analogs have been used to characterize ␥-secretase (17,36) (Fig. 3a). We first attempted to use L646, a biotinylated ␥-secretase inhibitor, to isolate the ␥-secretase complex (1). CHAPSO-solubilized  . For the cellular assay (middle panel), APP and Pen2 cells were grown in 96-well plates, and the amount of A␤40 and A␤42 secreted into the medium was measured in the electrochemiluminescence (ECL) assay. Background was defined as the amount of A␤40 and A␤42 secreted when cells were grown in the presence of 1 M L458. The concentration of A␤40 and A␤42 was determined with peptide standard curves (average Ϯ S.E., n ϭ 3). The A␤42:A␤40 ratio for the APP (black bar) and Pen2 (gray bar) cell lines for both the in vitro and cellular (secreted A␤) assays are shown in the right panel.
membranes were incubated with L646 in the presence or absence of excess L458, and the bound complexes were captured with streptavidin-agarose, resolved by SDS-PAGE, and Western blotted for the core components of ␥-secretase. Surprisingly L646 was incapable of capturing PS1 under non-denaturing condition even though it has been shown previously to directly label PS1-NTF upon photoactivation (1) (Fig. 3b, lanes  3 and 4). We reasoned that steric hindrance precluded L646 from interacting with both streptavidin and the ␥-secretase complex under native conditions. Therefore, we designed and synthesized a series of L458 analogs with varying linker lengths (Fig. 3a). In addition, L646, compound 1, and compound 3 also contain a benzophenone moiety that allows covalent photoincorporation of these probes into the ␥-secretase complex. Initially we determined their inhibitory potency against ␥-secretase and found that these compounds effectively block ␥-secretase activity with IC 50 values in the low nanomolar range, indicating that biotinylation did not adversely affect the activity of the compounds (supplemental Table S1). We next examined the ability of these compounds to capture PS1. We found that only compound 3 and compound 4 were able to capture both PS1-NTF and PS1-CTF (Fig. 3b). Because com-pound 4 captured PS1 with greater efficiency than compound 3, we determined whether compound 4 was able to capture the remaining components of ␥-secretase. Indeed compound 4 was also able capture the remaining core components of active ␥-secretase: Pen2, Nct, and APH1a (Fig. 3c). Inclusion of excess non-biotinylated L458 was able to completely block capture of ␥-secretase by both compounds 3 and 4, thereby indicating that the capture was specific in nature ( Fig. 3, b and c, ϩ L458). These studies suggest that probes that are able to interact with both streptavidin-agarose and ␥-secretase require linkers of at least 34 Å between biotin and the L458 backbone. To verify this, we used two of the photoreactive probes, compounds 1 and 3, to determine whether they were capable of capturing ␥-secretase under denatured and native conditions following cross-linking. CHAPSO-solubilized HeLa membranes were incubated with either compound 1 or 3 in the presence or absence of L458 and irradiated. Following photolysis, the labeled complexes were captured with streptavidin-agarose under either native (ϪRIPA) or denatured (ϩRIPA) conditions. First, we demonstrated that both compounds 1 and 3 were able to photoinsert into PS1-NTF and that the labeled protein could be captured under denaturing conditions (ϩRIPA). Second, as predicted, only compound 3 was able to capture PS1-NTF under native (ϪRIPA) conditions (Fig. 4a). In all cases, capture was completely blocked by inclusion of excess L458, indicating specific capture.
These studies suggest that the biotin moiety of the ligand becomes partly occluded by the protein interface of the active site when compounds with short linkers, such as compound 1, bind to ␥-secretase under native non-denaturing conditions (Fig. 4a, ϪRIPA). This partial occlusion prevents streptavidinagarose from binding to the biotin moiety of the ligand, and therefore labeled ␥-secretase is not captured. However, denaturing conditions destroy the ␥-secretase complex and thereby eliminate steric hindrance between the protein interface of the active site and streptavidin-agarose allowing isolation of the labeled components (Fig. 4a, ϩRIPA). To support this conclusion, we determined the inhibitory potency of the biotinylated analogs in the presence or absence of soluble streptavidin ( Fig.  4b and supplemental Table S1). Inclusion of soluble streptavidin had no effect on the potency of L458 (Fig. 4b, left panel) but dramatically increased the IC 50 values of L646, compound 1, and compound 2 (Ͼ220-fold) (Fig. 4c, middle panel, and supplemental Table  S1). However, inclusion of soluble streptavidin had much less of an impact on the potency of compound 3 and compound 4 (Ͻ20-fold) (Fig.  4c, right panel, and supplemental Table S1). It appears that the effect of streptavidin on the interaction of these inhibitors with ␥-secretase is inversely correlated with the length of biotin linker. These data suggest that the active site of ␥-secretase is a fairly deep binding pocket, which is consistent with studies that have indicated that the active site of ␥-secretase is a hydrophilic pore 20 -40 Å in length within the lipid bilayer (37,38). Furthermore these data also indicate that affinity probes with linkers of at least 34 Å are required to capture ␥-secretase under native non-denaturing conditions (Fig. 4c).
Overexpression of Pen2 Increases the Formation of PS2-containing Active ␥-Secretase Complexes-After establishing that compound 4 was able to fully capture active ␥-secretase complexes under native conditions, we characterized the active ␥-secretase complexes in the Pen2 and APP cell lines using this probe. First we assessed the affinity of ␥-secretase for L458 from both cell lines. We determined that the IC 50 values for inhibition of solubilized ␥-secretase from the Pen2 and APP cells were 1.5 Ϯ 0.1 and 1.3 Ϯ 0.3 nM, respectively, indicating that overexpression of Pen2 did not affect the affinity of ␥-secretase complexes for L458 (supplemental Fig. 1). We then used compound 4, based on the L458 backbone, to analyze the ␥-secretase complexes in the APP and Pen2 cell lines. Solubilized membrane fractions from both cell lines were incubated with compound 4 in the presence or absence of excess L458, and streptavidin-agarose was then used to capture the compound 4⅐␥-secretase complexes. Bound complexes were washed, resolved by SDS-PAGE, and immunoblotted for ␥-secretase components: PS1-NTF, Nct, Pen2, and PS2-CTF (Fig. 5a). When we compared the active ␥-secretase complexes from the two cell lines, we immediately noticed two remarkable features. First, Pen2 cells possessed significantly more PS2-containing complexes but considerably fewer PS1 complexes than the APP cells even in light of the fact that Pen2 cells express less total PS2 than APP cells (Figs. 5a and 1c). This finding further supports the assertion that the total amount of ␥-secretase subunits in cell membrane is not necessarily correlated with the level of protein engaged in the active complexes. Therefore separation of the catalytically active complexes from the non-enzymatic complexes helps to accurately characterize the relationship between ␥-secretase complex composition and activity. These results suggest that overexpression of Pen2 alters the dynamics of PS1-and PS2containing ␥-secretase complexes and favors the formation of PS2-containing complexes. Second, Pen2 overexpression seems to alter the relative stoichiometry of subunits within the active ␥-secretase complex. Pen2 overexpression caused a decrease in the incorporation of Nct into the active complexes but an increase in the incorporation of Pen2 (Fig. 5a). Although the amount of each component that is associated with the PS1 and PS2 complexes is unknown, it is clear that the relative stoichiometry of the ␥-secretase complexes is strikingly different between the Pen2 cells and the APP cells. Quantification of the relative amounts of captured proteins revealed that the Pen2 ␥-secretase complexes contain only ϳ25-30% of PS1-NTF and Nct but 250 and 400%, respectively, of Pen2 and PS2-CTF as compared with the APP ␥-secretase complexes (Fig. 5b). Furthermore equivalent results were seen even when compound 4 was used at 100 nM, indicating that there was a sufficient excess of compound 4 available for capture (data not shown). Together these data indicate that the APP and Pen2 cell lines display distinct ␥-secretase activity and possess different ␥-secretase complexes and/or different ratios of the components.

Overexpression of PS1 FAD Mutants Causes a Shift in Equilibrium from PS1-containing Complexes to PS2-containing
Complexes-Mutations in PS1 linked to early onset AD are associated with an increase in the A␤42:A␤40 ratio (31). This prompted us to consider the possibility that overexpression of PS1 FAD mutants would alter the dynamics of PS1 and PS2 complex formation and mimic the effect seen with Pen2 overexpression. We performed analogous experiments as with the Pen2 cells using established cell lines that expressed either WT or mutated PS1. We characterized N2a cell lines stably expressing either PS1 WT or PS1 harboring the FAD mutations M146L or C410Y. Cell membranes isolated from each cell line were assayed for A␤40 and A␤42 production in the in vitro ␥-secretase assay (Fig. 6a, left panel) (31,39). Both PS1 M146L and C410Y mutations caused an increase in the A␤42: A␤40 ratio (0.20 and 0.36, respectively) compared with PS1 WT (0.08) (Fig. 6a, right panel) in our in vitro assay; this is consistent with previous cellular studies (31,39). M146L slightly augmented A␤40 production (138% of WT levels) and to a much larger magnitude increased A␤42 production (346% of WT levels), whereas C410Y greatly decreased A␤40 production (29% of WT levels) without significantly affecting A␤42 levels (128% of WT levels). Irrespective of the divergent effect on A␤40 and A␤42 production, the end result of overexpression of either PS1 FAD mutant was an increase in the A␤42: A␤40 ratio.
Because the PS1 FAD mutants affected the A␤42:A␤40 ratio in a manner similar to that of Pen2 overexpression, we characterized the active ␥-secretase complexes in the PS1 FAD mutants compared with PS1 WT. We first looked at expression levels of the known components of ␥-secretase by Western blotting. Compared with the human WT PS1-expressing cells, the M146L-expressing cells exhibited a similar level of PS1, Pen2, and Nct but had significantly higher amounts of PS2 (Fig.  6b). Although overexpression of C410Y also resulted in higher levels of PS2, there was a decrease in expression of PS1, Nct, and Pen2 (Fig. 6b). After we determined that ␥-secretase in these cells exhibited the same affinity for L458 (data not shown), we assessed the relative amount of active PS1versus PS2-containing ␥-secretase complexes between the cell lines. We again used compound 4 to capture the active ␥-secretase complexes from CHAPSO-solubilized membrane fractions. Clearly both FAD PS1 and PS2 are capable of incorporating into the active ␥-secretase complex. Furthermore M146L and C410Y had increased amounts of PS2-containing ␥-secretase complexes compared with WT (Fig. 6c). M146L had slightly more PS1, but roughly equal amounts of Nct and Pen2 engaged in active complexes compared with WT, whereas C410Y had much less PS1, Nct, and Pen2. These data suggest that the amount of PS2containing complexes is positively correlated with the A␤42: A␤40 ratio generated by these cell membranes. Additionally these data, like the Pen2 overexpression data, suggest the possibility of differential stoichiometries of subunits of the various active ␥-secretase complexes. For example, M146L overexpression resulted in a gross increase in PS2-containing active complexes but did not affect the levels of active PS1-containing complexes nor did it alter the incorporation of the critical subunits Nct and Pen2 into active complexes. Clearly the stoichiometry of subunits in the active ␥-secretase complexes is dynamic.
To verify the assertion that PS1 WT and PS1 FAD mutants differentially affect the dynamics of the PS1 and PS2 ␥-secretase complexes, we conducted similar studies using three additional PS1 FAD stable expressing N2a cell lines. The L286V, H163R, and E280A mutations reduced ␥-secretase-mediated A␤40 production to 15, 23, and 11%, respectively, of WT levels (supplemental Fig. 2a, upper panel). These FAD mutations also caused a reduction in the production of A␤42 (58, 54, and 84% of WT production, respectively) (supplemental Fig. 2a, middle  panel). The net effect of these mutations resulted in an increase in the A␤42:A␤40 ratio (0.32, 0.19, and 0.64 for L286V, H163R and E280A, respectively, compared with 0.08 for WT) (supplemental Fig. 2a, lower panel). We again used compound 4 to characterize the active PS1 and PS2 ␥-secretase complexes in FIGURE 6. Overexpression of PS1 FAD mutants causes an increase in the A␤42:A␤40 ratio and shifts the equilibrium from PS1-containing active ␥-secretase complexes to PS2-containing complexes. a, PS1 FAD mutations increase the A␤42:A␤40 ratio. CHAPSO-solubilized membrane from PS1 FAD mutant-overexpressing N2a cells was assayed for in vitro production of A␤40 and A␤42 (left panel) using an APP transmembrane substrate. The A␤42:A␤40 ratios are shown in the right panel. Background was defined as activity remaining in the presence of 1 M L458. Activity is graphed as background-subtracted units/min/g of membrane assayed. b, expression of ␥-secretase components is altered by expression of PS1 FAD mutants. CHAPSO-solubilized membrane from PS1 WT and two PS1 FAD mutants (M146L and C410Y) were Western blotted for the indicated components of ␥-secretase. c, formation of PS2-containing ␥-secretase complexes is increased by expression of PS1 FAD mutants. Active ␥-secretase was captured from CHAPSO-solubilized membrane of PS1 WT, PS1 M146L, and PS1 C410Y cells with 20 nM compound 4 in the presence (ϩ) or absence (Ϫ) of 2 M L458. Bound complexes were captured with streptavidinagarose, resolved by SDS-PAGE, and blotted for the indicated components of ␥-secretase. ECL, electrochemiluminescence. these cell lines. Although similar amounts of PS1 were detected in all four cell lines (supplemental Fig. 1b, upper panel), the amount of PS1 engaged in the active ␥-secretase complex is considerably different. WT-and L286V-expressing cells had similar amounts of PS1-containing active ␥-secretase, whereas H163R and E280A had much fewer PS1-containing active complexes (supplemental Fig. 2b, lower panel). H163R and E280A contained roughly the same amount of PS2 compared with WT, whereas L286V showed increased PS2 levels compared with WT (supplemental Fig. 2b, upper panel). However, all three FAD mutants contained much more active PS2-containing ␥-secretase complexes compared with WT (supplemental Fig.  2b, lower panel). Furthermore using these PS1 FAD mutants and PS1 WT, we determined the relationship among 1) the ratios of A␤42:A␤40 and the compound 4-captured PS2:PS1 (referred to (PS2:PS1) captured ), 2) the ratios of A␤42:A␤40 and the total PS2:PS1 (referred to (PS2:PS1) total ), and 3) the ratios of (PS2:PS1) captured and (PS2:PS1) total. We found that 1) the ratio of A␤42:A␤40 strongly correlates with the ratio of (PS2: PS1) captured (r 2 ϭ 0.906) but not (PS2:PS1) total (r 2 ϭ 0.018) and 2) there is little correlation between (PS2:PS1) total and(PS2: PS1) captured (r 2 ϭ 0.041) (supplemental Fig. 3). These data demonstrate three critical points. First, the total amount of PS1 and PS2 protein does not always correlate with their incorporation into active ␥-secretase complexes. Therefore, isolation of the active ␥-secretase complex is necessary for understanding the formation and dynamics of the complexes. Second, PS1 WT and FAD mutants have distinct effects on the equilibrium of PS1 and PS2 ␥-secretase complexes, supporting our previous conclusion that ␥-secretase complexes are dynamic and that perturbations to complex equilibrium can affect enzyme activity. Third, despite the different amounts of active PS1 and PS2 present in different cell lines, it appears that an increased ratio of captured PS2:PS1 leads to an elevated ratio of A␤42:A␤40, which correlates with the amount of characteristic AD plaques in mouse models (40,41) and with the age of onset of familial Alzheimer disease (42). Therefore, these studies suggest that it is the relative amount of active PS1 and PS2 ␥-secretase complexes that plays a critical role in the determination of the ratio of A␤42:A␤40.
FAD Knock-in Mouse Models of AD Show an Increase in PS2containing ␥-Secretase Complexes-To further validate our conclusion that perturbations to the PS1 and PS2 complex equilibrium results in increased A␤42:A␤40 ratios, we determined whether FAD knock-in mouse models also exhibited a biochemical phenotype similar to that seen in the cell models. Knock-in mouse models of AD have been shown to have increased A␤42:A␤40 ratios and accelerated AD plaque deposition compared with wild type controls (31,43,44). We chose two knock-in models that harbor PS1 mutations, M146V and ⌬E10, to validate our findings that PS1 mutations can cause a shift in the equilibrium of PS1-and PS2-containing ␥-secretase complexes. In previous studies, we have shown that ⌬E10 mice have altered ␥-secretase specificity for A␤40 and A␤42 production that leads to a 1.7-fold increase in the A␤42:A␤40 ratio (40). We determined that M146V mice also show a similar increase in the A␤42:A␤40 ratio. Solubilized membrane from WT or PS1 M146V brain was assayed for in vitro A␤40 and A␤42 production using the APP transmembrane substrate as done in the above cellular assays. M146V mice showed decreased A␤40 but increased A␤42 production compared with WT (Fig. 7a) that led to an increase in the A␤42:A␤40 ratio (1.5-fold).
We then examined the total amount of PS1 and PS2 in the membrane fractions from total brain. We found that there was very little difference in the expression levels of PS1 or PS2 between either of the knock-in mice and their respective wild type controls (Fig. 7b). We next determined the relative ratio of PS1versus PS2-containing active ␥-secretase complexes in brain from the knock-in mice. ␥-Secretase was captured from CHAPSO-solubilized membrane fractions of brain with compound 4, and the amount of bound PS1 and PS2 was determined by Western blotting. M146V mice showed an increase in both PS1-and PS2-containing active ␥-secretase complexes compared with WT control (Fig. 7c). ⌬E10 mice showed no change in PS1-containing complexes but a large increase in PS2-containing complexes (Fig. 7c). In both cases it was evident that the equilibrium between PS1-and PS2-containing ␥-secretase complexes was altered in the knock-in mice, and both exhibited an increase in the A␤42:A␤40 ratio. These data further support a positive correlation between shifting the equilibrium toward PS2-containing complexes at the expense of PS1containing complexes and alteration of the A␤42:A␤40 ratio.

DISCUSSION
Three mutated genes (APP, PS1, and PS2) have been linked to early onset AD. Elucidating the function of these proteins and the effect of their mutations on ␥-secretase activity offers a unique opportunity to investigate the reaction mechanism of ␥-secretase and the molecular pathogenesis of AD. The small molecule approaches used in our present studies have provided critical insights into the effect of PS1 FAD mutations on ␥-secretase activity and the dynamics of ␥-secretase complex formation.
PS1 mutations that increase the ratio of A␤42:A␤40 in cell and animal models suggest that they affect ␥-secretase activity. However, whether these cellular effects are attributed to ␥-secretase itself and/or other cellular factors remains to be investigated. In other words, can these cellular observations be biochemically recapitulated in an in vitro system? This study provides biochemical evidence that PS1 mutations directly alter the rate of ␥-secretase production of A␤40 and A␤42. Furthermore although there was a similar increase in the A␤42: A␤40 ratio, the PS1 mutations differentially affected ␥-secretase activity for A␤40 and A␤42 production. We observed three distinct effects on A␤40 and A␤42 production by PS1 FAD mutants: 1) M146L increased both A␤40 and A␤42 production, 2) C410Y reduced A␤40 and increased A␤42 production, and 3) L286V, H163R, and E280A reduced both A␤40 and A␤42 production. Our studies suggest that these mutations mediate an increase in the A␤42:A␤40 ratio through distinctive mechanisms. Our data also indicate that expression of C-terminally Myc-tagged Pen2 increases the A␤42:A␤40 ratio in a manner similar to that of PS1 FAD mutants and therefore represents yet another distinctive mechanism by which ␥-secretase activity can be modulated. Understanding how PS1 mutations and Pen2 overexpression alter ␥-secretase complex formation may reveal a unifying factor for understanding their diverse effects on ␥-secretase activity.
Our current findings reveal that expression of PS1 FAD mutants or C-terminal Myc-tagged Pen2 results in increased formation of active PS2 ␥-secretase complexes. These findings clearly suggest that PS1 FAD mutants have a reduced ability to compete for the pool of shared obligatory subunits (Nct, Pen2, and Aph1) with PS2 and therefore lead to increased PS2 ␥-secretase complex formation (Fig. 8). The precise mechanism of this reduced ability requires further investigation but potentially results from a reduced affinity for subunit interactions because of conformational changes in PS1. Because of the complexity of ␥-secretase as a multisubunit membrane-bound enzyme complex, there has been a lack of both an in vitro reconstitution system and high resolution structural information of the complex. This has made it extremely challenging to obtain kinetic data that would indicate exactly how PS1 FAD mutations affect subunit interactions and complex assembly. Furthermore the expression of C-terminal Myc-tagged Pen2 also led to an increase in PS2-containing complexes suggesting that PS2 may normally have an increased ability to compete with PS1 for the shared cofactors when the formation and natural dynamic of complex equilibrium are perturbed.
PS1 and PS2 are highly homologous (67% identical) but are engaged in mutually exclusive complexes (13,35). Both PS1-and PS2-containing complexes are catalytically active; however, PS1-containing ␥-secretase complexes display considerably higher specific activity than PS2 complexes (28,35). Therefore, shifting the equilibrium toward an increase in PS2-containing ␥-secretase complex at the expense of PS1containing complexes should result in a reduction in overall ␥-secretase activity; this hypothesis is supported by our current observations. Another critical question is how changing the complex equilibrium alters the ratio of A␤42:A␤40. Our preliminary studies indicating that the A␤42:A␤40 ratio displayed a strong positive correlation with the relative amount of active PS2:PS1 complexes suggest that PS2 ␥-secretase complexes might be capable of processing the APP substrate with a higher A␤42:A␤40 ratio (Fig. 8, pathway 1). Concomitantly PS1-mutated ␥-secretase also has altered specificity for A␤42 and A␤40 production (Fig. 8, pathway 2). Therefore, the alteration of ␥-secretase specificity for A␤40 and A␤42 production could result from a combination of the increase in PS2 complex formation and altered activity by the mutated PS1 complex. Indeed our previous study showed that the A␤42:A␤40 ratio in PS2-expressing cells was 1.4-fold higher than that in PS1-expressing cells, but this was not seen in biochemical studies (35). Therefore, it is noteworthy to point out that investigation of the relationship between PS1-and PS2-mediated ␥-secretase activity for A␤40 and A␤42 production is technically challenging. Utilizing a single form of PS (PS1 or PS2) in the absence of the FIGURE 7. Two independent knock-in mouse models of FAD have increased A␤42:A␤40 ratios and show a shift toward PS2 containing active ␥-secretase complexes. a, M146V knock-in mice exhibit altered A␤42:A␤40 ratios compared with WT mice. CHAPSO-solubilized membrane of total brain from PS1 M146V mice and its respective WT control were assayed for in vitro production of A␤40 (black bar) or A␤42 (gray bar) using an APP transmembrane substrate. Background was defined as activity remaining in the presence of 1 M L458. Activity is graphed as background-subtracted units/min/g of membrane assayed (average Ϯ S.E., n ϭ 3). The A␤42:A␤40 ratios are shown below (average Ϯ S.E., n ϭ 3). b, FAD knock-in mice have comparable basal expression of ␥-secretase subunits as compared with WT. CHAPSO-solubilized membrane from total brain of PS1 M146V mice (left panel) and PS1 ⌬E10 (right panel) and their respective WT controls was Western blotted for expression of PS1-NTF and PS2-CTF. c, FAD knock-in mice favor an increase in PS2-containing ␥-secretase complex formation compared with WT. Active ␥-secretase was captured from CHAPSO-solubilized membrane of total brain from PS1 M146V mice (left panel) or PS1 ⌬E10 mice (right panel) and their respective WT controls with 20 nM compound 4 in the presence (ϩ) or absence (Ϫ) of 2 M L458. Bound complexes were captured with streptavidin-agarose, resolved by SDS-PAGE, and Western blotted for PS1-NTF and PS2-CTF. ECL, electrochemiluminescence.
other does not reflect the endogenous dynamics of PS1-and PS2-containing ␥-secretase complexes that compete for the known shared cofactors (Nct, Aph1, and Pen2) when present in the same cells (32). Therefore, the contribution of PS1 mutations to the production of A␤42 and A␤40 should be carefully interpreted with consideration to PS2 involvement in the formation of A␤ peptides. Alternatively PS1 ␥-secretase complexes may preferentially process A␤40 peptides, whereas PS2 complexes show no preference for A␤40 and A␤42 production (Fig. 8, pathway 3). A loss of PS1 ␥-secretase complexes because of an increase in PS2 ␥-secretase complexes would selectively reduce A␤40 peptide production and thereby account for the observed A␤42:A␤40 ratio increases we observed. Similarly a loss of A␤40-generating PS1 complexes because of PS1 FAD mutations would also increase the A␤42:A␤40 ratio in the absence of PS2 expression. This is in fact supported by previous studies that showed that expression of PS1 FAD mutations in PS1-and PS2-null cells was capable of elevating the ratio of A␤42 to A␤40 (28).
Taken together, these findings reveal a novel mechanism of ␥-secretase regulation by dynamic alteration of the equilibrium of PS1-and PS2-containing ␥-secretase complexes (Fig. 8).
These data provide a possible mechanism for multiple "loss of function" PS mutations distributed within different regions of the protein (25)(26)(27); these distinct mutations could exert their effect on A␤40 production by perturbation of the complex equilibrium. In other words, shifting the equilibrium from PS1 ␥-secretase to PS2 ␥-secretase alters overall ␥-secretase activity and leads to an elevated ratio of A␤42:A␤40 (Fig. 8). In addition to favoring an increase in PS2-containing complexes, PS1 FAD mutations also result in formation of a mutated PS1 ␥-secretase complex with altered specificity for A␤40 and A␤42 production as well as leading to increased A␤42:A␤40 ratios (Fig. 8).
Additionally the requirement of affinity probes with longer linkers for capture of native ␥-secretase has indicated that that the active site is buried deeply within the hydrophobic membrane. This conclusion is in agreement with other biochemical and structural studies (36 -38) that have shown the active site of ␥-secretase resides in a hydrophilic pore 20 -40 Å in length within the lipid bilayer. Moreover conjugation of L685,458 to biotin by either hydrophobic or hydrophilic linkers had little effect on the inhibitory potency of L-685,458, suggesting that this inhibitor can easily access the active site of ␥-secretase regardless of the biophysical properties of linkers. It is likely that L-685,458 accesses the ␥-secretase active site through a channel-like mechanism. It is of note to point out that conformation of the ␥-secretase complex is dependent on detergent concentration because ␥-secretase in CHAPSO at concentrations above the critical micelle concentration, in the presence or absence of digitonin, can be captured by a transition state inhibitor immobilized to solid support through a six-atom hydrophilic linker (15,17). ␥-Secretase captured under high detergent conditions such as these may represent a different form of the complex. Therefore we believe our panel of affinity probes with varying linker lengths is a valuable tool for the functional characterization of active and endogenous ␥-secretase and allows us to further examine the relationship between ␥-secretase specificity and endogenous PS1-and PS2-containing ␥-secretase complexes in normal individuals and AD patients. This ability will facilitate elucidation of the pathogenesis of AD and aid in the development of therapeutics.