C-terminal Fragment of Presenilin Is the Molecular Target of a Dipeptidic γ-Secretase-specific Inhibitor DAPT (N-[N-(3,5-Difluorophenacetyl)-L-alanyl]-S-phenylglycine t-Butyl Ester)*

γ-Secretase is a multimeric membrane protein complex composed of presenilin (PS), nicastrin, Aph-1 and, Pen-2 that is responsible for the intramembrane proteolysis of various type I transmembrane proteins, including amyloid β-precursor protein and Notch. The direct labeling of PS polypeptides by transition-state analogue γ-secretase inhibitors suggested that PS represents the catalytic center of γ-secretase. Here we show that one of the major γ-secretase inhibitors of dipeptidic type, N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT), targets the C-terminal fragment of PS, especially the transmembrane domain 7 or more C-terminal region, by designing and synthesizing DAP-BpB (N-[N-(3,5-difluorophenacetyl)-l-alanyl]-(S)-phenylglycine-4-(4-(8-biotinamido)octylamino)benzoyl)benzyl)methylamide), a photoactivable DAPT derivative. We also found that DAP-BpB selectively binds to the high molecular weight γ-secretase complex in an activity-dependent manner. Photolabeling of PS by DAP-BpB is completely blocked by DAPT or its structural relatives (e.g. Compound E) as well as by arylsulfonamides. In contrast, transition-state analogue inhibitor L-685,458 or α-helical peptidic inhibitor attenuated the photolabeling of PS1 only at higher concentrations. These data illustrate the DAPT binding site as a novel functional domain within the PS C-terminal fragment that is distinct from the catalytic site or the substrate binding site.

A body of evidence suggests that ␥-secretase is a high molecular weight (HMW) 7 membrane protein complex, including presenilin (PS), nicastrin, Aph-1, and Pen-2 (1)(2)(3)(4)(5). PS is endoproteolyzed into two subunits termed N-terminal fragment (NTF) and C-terminal fragment (CTF) during the maturation process, and these fragments are believed to represent the active form of PS; each fragment has a conserved aspartate residue within TMDs that are critical for ␥-secretase activity. Transition-state analogue-type aspartyl protease inhibitors bind directly to these PS fragments (6,7), suggesting that PS is the catalytic component of ␥-secretase. Intriguingly, immobilized transition-state analogue inhibitors, which are predicted to occupy the catalytic site, can copurify ␥-secretase with its ␤-amyloid precursor protein-derived natural substrate, C83 (8,9). Furthermore, these inhibitors unexpectedly displayed a noncompetitive pharmacological inhibition profile (10). Finally, an ␣-helical peptide-based inhibitor binds to PS fragments in a distinct fashion to that of transition-state analogue inhibitors (11,12). These findings suggest that ␥-secretase has a "substrate binding site" that is topographically distinct from the catalytic site to which the substrate of this protease initially binds before the proteolysis.
A number of structurally diverse and potent ␥-secretase inhibitors have been so far reported in addition to the classical transition-state analogues (13,14). However, their molecular target(s) and mode(s) of action are poorly understood. Here, we have chosen one of these nontransition-state analogue ␥-secretase inhibitors, N-[N- (3,5-difluorophenacetyl)-L-alanyl]-(S)-phenylglycine t-butyl ester (DAPT) (15). Notably, DAPT does not inhibit proteolytic activity of signal peptide peptidase, which belongs to the same aspartic protease family, A22, in the MEROPS data base (16) and shares the consensus sequence around the active site with presenilin/␥-secretase (17). Considering the fact that several strong ␥-secretase inhibitors affect signal peptide peptidase activity (18,19), the characterization of DAPT would lead to the development of better ␥-secretase-targeted therapeutics against Alzheimer disease without side effects. We identified the molecular target of DAPT as PS CTF by rational derivatization and photocross-linking approach and found that the DAPT binding site is pharmacologically distinct from either the catalytic site or the substrate binding site. These results imply the existence of a novel functional domain within the ␥-secretase complex that is important for the catalytic activity and to which small molecule ␥-secretase inhibitors bind.
Antibodies and Immunological Methods-The rabbit polyclonal antibodies anti-PS1 CTF (G1L3) and anti-PS2 CTF (G2L) were raised as described (26,27). Anti-PS1 NT (28) against the N terminus of human PS1 and PS-C3 (29) against the C terminus of PS1 were kindly provided by Drs. G. Thinakaran (University of Chicago) and A. Takashima (RIKEN), respectively. Anti-biotin rabbit polyclonal antibody and anti-Xpress mouse monoclonal antibody were purchased from Bethyl and Invitrogen, respectively. The samples were analyzed by immunoblotting and two-site enzyme-linked immunosorbent assays as described (26, 27, 30 -32).
Evaluation of Inhibitory Potency-In vitro ␥-secretase assay and measurement of A␤ by sandwich enzyme-linked immunosorbent assay were performed as described previously (32). IC 50 values were calculated by plotting the enzyme-linked immunosorbent assay data on KyPlot software (KyensLab, Inc.) and fitted it to a sigmoidal function.
Preparation of Membrane Vesicle Lysates and Photoaffinity Labeling-HeLa S3 and other cell lines were cultured, and their membrane fractions were collected as described (31). The resulting membrane fractions were lysed in buffer A containing 20 mM Tris-HCl, pH 8.0, 1% CHAPSO with Complete protease inhibitor mixture (Roche Applied Science) and centrifuged at 100,000 ϫ g for 1 h. The supernatant was diluted with 20 mM Tris-HCl, pH 8.0, to a final 0.25% CHAPSO solution and used as solubilized ␥-secretase fraction. For photoaffinity labeling experiments, the solubilized fraction was incubated with 100 nM DAP-BpB at 4°C. In competition experiments, samples were prepared in the presence of 10 M DAPT (or other compounds with the indicated concentrations based on each IC 50 value determined by in vitro ␥-secretase assay) 40 min before the addition of DAP-BpB. The samples were irradiated on ice for 40 min using a B100A Lamp (UVP) at a distance of 7 cm. The labeled samples were incubated with streptavidin-Sepharose HP (Amersham Biosciences) overnight at 4°C and eluted from the resin by boiling for 3 min in the sample buffer. The labeled proteins were detected by SDS-PAGE/immunoblotting using antibodies against PS1 and biotin, respectively.
Protease Digestion Assay-Photocross-linked samples were prepared as detailed above except for using MBS buffer (10 mM MES, pH 6.0, 150 mM NaCl) for Asn-C digestion. These samples (ϳ1 ml) were then incubated with either trypsin (100 g, Invitrogen), Asp-N (1 g, Roche Applied Science), or asparaginyl endopeptidase (Asn-C; 50 microunits, Takara) at 37°C for 2 h (trypsin) or overnight (Asp-N and Asn-C). Digestion was stopped by the addition of soybean trypsin inhibitor (for trypsin) or 0.5% SDS (for Asp-N and Asn-C). Digested samples were captured with streptavidin-Sepharose and analyzed by SDS-PAGE/immunoblotting as described above.

Development of Photoactivable Derivative of DAPT-Previous
reports showed the direct binding of transition-state analogue ␥-secretase inhibitors to PS fragments, regarded as the catalytic components of ␥-secretase (6, 7). However, little is known about the mode of binding of other classes of ␥-secretase inhibitors, especially the dipeptidic compounds (i.e. DAPT, compound E, LY411575), that have been shown to potently inhibit ␥-secretase activity in vitro and in vivo (Ref. 13; Table 1). To address the molecular mechanism whereby the dipeptidics inhibit ␥-secretase activity, we developed a derivative DAP-BpB in which DAPT was coupled with a photoreactive benzophenone and a biotin moiety (Ref. 20; Fig. 1). The IC 50 value of DAP-BpB was similar to that of DAPT (ϳ100 nM) in an in vitro ␥-secretase assay using recombinant substrate C100FmH (5,32), suggesting that the affinity of this photoactivable derivative to ␥-secretase was at a comparable level with that of the parental compound (Fig. 1C). Notably, during the rational derivatization of DAPT, we found the incorporation of the benzophenonemethyl (p-benzoylbenzyl) amide in the C terminus of DAPT resulted in the significant increase of its inhibitory potency (up to ϳ30-fold) (21).
Identification of the Target Protein of DAPT by Photocross-linking-To identify the target protein(s) of DAPT within the ␥-secretase complex, we mixed CHAPSO-solubilized lysate of HeLa S3 cells with

TABLE 1 The inhibitory potencies of the ␥-secretase inhibitors used in this study
DAP-BpB in the absence or presence of excess amount of DAPT and subjected the lysate to irradiation with long-wave near-UV light. Biotinylated products were then captured on streptavidin-Sepharose beads, separated by SDS-PAGE, and detected by an anti-biotin antibody ( Fig.  2A). Two biotinylated bands corresponding to relative molecular masses of 23 kDa (major) and 26 kDa (minor) were detected after DAP-BpB treatment and UV irradiation. The intensity of these bands was markedly reduced by pretreatment of lysates with excess DAPT, confirming that the photolabeling of these molecules was specific. The labeling of other bands (e.g. 16 and 14 kDa) was not diminished by DAPT, suggesting that the labeling of these bands was nonspecific. Because the sizes of these bands were similar to those of PS1 CTF and its phosphorylated species, we probed the DAP-BpB/UV-treated fraction with an anti-PS1 CTF antibody and found that these PS1 CTFs were recovered in this fraction and were abolished in the presence of DAPT (Fig. 2B). The apparent mobility shift of DAP-biotinylated PS1 CTF on gels was similar to that observed by labeling with other photoprobe (Ref. 8; see also Fig. 3A). The possibility that other proteins larger than 36 kDa were also labeled by DAP-BpB is not completely excluded. However, other ␥-secretase components (i.e. Nicastrin, Aph-1, Pen-2) were not labeled by DAP-BpB, 8 and these two biotinylated bands were separable by immunoaffinity chromatography using an anti-PS1 CTF antibody (data not shown). These data altogether indicated that DAP-BpB was specifically cross-linked chiefly with PS1 CTFs in the ␥-secretase complex. We also tested whether DAP-BpB binds to PS2, a paralogue of PS1 in mammalians that makes a minor contribution to ␥-secretase activity (36). Unexpectedly, DAP-BpB did not label endogenous PS2 CTF in HeLa S3 cells (Fig, 2C). These data suggest that PS1 CTFs are the major and direct target proteins of the dipeptidic ␥-secretase inhibitor DAPT.

DAPT Binding Site Resides in the C-terminal Region of PS1 CTF-PS1
CTF is generated by endoproteolysis of PS holoprotein after assembly of ␥-secretase complex (2)(3)(4). This polypeptide is comprised of two portions, a large hydrophilic loop (amino acids 299 -380) and a hydrophobic C terminus (amino acids 381-467), the latter harboring the catalytic aspartate residue in TMD7. To determine the relationship between endoproteolysis and DAPT binding and to locate the DAPT binding site within the PS1 CTF, we analyzed the truncated forms of PS1, PS1⌬exon9 (lacking amino acids 290 -319) and PS1⌬Loop (lacking amino acids 304 -371) (28,29). PS1⌬exon9 forms an active ␥-secretase complex as a holoprotein because it lacks the proximal cytosolic loop, 8 Y. Takahashi, T. Tomita, and T. Iwatsubo, unpublished data.   which contains the endoproteolytic cleavage site in PS. In contrast, PS1⌬Loop lacks most of the large loop domain but retains the endoproteolytic cleavage site, thus resulting in the formation of active stable fragments. Photocross-linking experiments revealed that both PS1⌬exon9 holoprotein and the CTF of PS1⌬Loop were labeled (Fig.  3A). These results indicate that the endoproteolysis of PS1 is dispensable for DAPT binding and that the binding site of DAPT resides within the TMD7 or more C-terminal portion of PS1 polypeptide (amino acids 372-467), which is highly homologous to PS2. We further tried to narrow down the DAPT binding site by protease digestion of the DAP-BpB-labeled PS1 CTF with various proteases (i.e. trypsin, Asp-N, Asn-C) (Fig. 3, B and C). All of these proteases yielded biotinylated ϳ10 -12-kDa bands (asterisk in Fig. 3C) despite that several different cleavage sites were predicted within the TMD7 and C terminus (Fig.  3B). Corresponding bands were detected by an antibody to the C terminus of PS1 (PS-C3) but not by an antibody against the PS1 loop region (G1L3), suggesting that the ϳ10 -12-kDa bands lacked the hydrophilic loop, whereas most of the C-terminal region of PS1 was preserved, exhibiting a high protease resistance (37). These results suggest that DAPT binds to the C-terminal region of PS1 CTF that contains TMD7 and -8 (residues 372-467) independently of its endoproteolysis.
DAPT Selectively Binds to the Active Form of ␥-Secretase-The low content of full-length endogenous PS1 in our solubilized ␥-secretase preparation precluded the assessment as to whether DAP-BpB binds to PS1 holoprotein. However, overexpression of PS1 in mammalian cells leads to accumulation of unprocessed PS1 polypeptides, which are less stable and form an inactive low molecular weight complex (4,30). We prepared solubilized ␥-secretase from HEK293 cells stably overexpressing human PS1; the level of PS1 holoprotein was similar to that of its fragments (Fig. 4A). In these cells PS1 CTF was specifically biotinylated, whereas PS1 holoprotein was not. This result indicated that the DAPT binding site is formed only within the CTF region of PS1 incorporated in a HMW complex but not in the holoprotein, the latter regarded as an inactive zymogen and incorporated in a low molecular weight complex (7). Next, we prepared solubilized ␥-secretase from membranes of HeLa S3 cells using a RIPA buffer containing 0.1% SDS and 1% Nonidet P-40 that disrupts the HMW complex (see Fig. 4E) and leads to PS NTF/CTF heterodimer disassembly (data not shown). PS1 CTF was never labeled by DAP-BpB in the RIPA-solubilized membrane fraction, further suggesting that preservation of HMW complex is required for the DAPT binding (Fig. 4B). In CHAPSO-solubilized membranes, ␥-secretase activity was enriched at a relative molecular mass of ϳ2 MDa (25). Consistent with these results, in 1% CHAPSO-solubilized HeLa S3 membrane fractions separated by size exclusion chromatography the whole ␥-secretase activity as well as PS1 CTF were enriched in fractions corresponding to more than ϳ1-2 MDa (Fig. 4, D and E, top). In contrast, no ␥-secretase activity was detected in RIPA-solubilized fractions (data not shown), and PS1 CTF was detected in under-220-kDa fractions (Fig. 4E, bottom). Thus, the labeling efficiency of PS1 CTF by DAP-BpB is highly correlated with the formation of HMW complex as well as the proteolytic activity, similar to that of the transition-state analogue-based photolabile inhibitor (25,36).
Then, to analyze the dependence of DAP-BpB labeling on ␥-secretase activity, we examined the labeling of dominant negative mutants of PS. Two intramembrane catalytic aspartate residues as well as a conserved proline residue in the C-terminal PALP motif are required for the endoproteolysis and the ␥-secretase activity (38 -40) but not for the formation of HMW complex. Intriguingly, no DAPT binding with PS1 polypeptides was observed in membranes from DKO cells expressing PS1/D257A, PS1/D385A, or PS1/P433L mutants (Fig. 5A).
To ascertain whether the endoproteolysis of PS is required for the labeling, PS1 carrying these loss-of-function mutations together with M292D mutation were examined (Fig. 5B). PS1/M292D restored the ␥-secretase activity in DKO cells and was labeled by DAP-BpB, whereas the fragment formation was completely abolished (Fig. 5C, data not shown; Ref. 41). Again, the loss-of-function mutations (i.e. D257A, D385A, and P433L) diminished the labeling by DAP-BpB on an uncleavable PS1 basis, strongly indicating that the preservation of these conserved residues is indispensable for the DAPT binding. Finally, we examined the labeling of an active form of ␥-secretase complex harboring PS2 as a catalytic subunit because exogenously overexpressed PS2 is able to fully restore the ␥-secretase activity in DKO cells (data not shown; Ref. 36). In contrast to endogenous PS2 in HeLa cells (see Fig.  2C), the PS2 CTF derived from overexpressed PS2 was biotinylated by DAP-BpB in lysates of DKO cells transfected with PS2 (Fig. 5C). Consistent with the result in PS1, labeling of PS2 holoprotein was never detected. To estimate the efficiency of DAP-BpB-labeling in a semiquantitative manner, we expressed N-terminally His/Xpress (HX)tagged PS1 or PS2 in DKO cells. These tagged proteins produced similar amounts of fragments and fully rescued ␥-secretase activity (data not shown), suggesting comparable levels of specific activity between PS1 and PS2. We conducted the DAP-BpB labeling experiment using solubilized fractions containing equal amounts of HX-PS1 NTF and HX-PS2 NTF, as quantitated by immunoblotting (Fig. 5D). The amounts of biotinylated PS1 CTF (i.e. the sum of non-phosphorylated and phosphorylated CTF) and PS2 CTF were nearly equivalent (Fig. 5E), suggesting that the labeling efficiency was similar between PS1 and PS2 under a condition in which comparable levels of ␥-secretase activities are observed. Thus, the lack of labeling of endogenous PS2 in HeLa cells might be due to the relatively limited level of endogenous PS2 participating in the formation of enzymatically active ␥-secretase complex (36). Taken together, these results suggest that DAPT selectively binds to the CTF of "catalytically active" forms of PS that are present within the fully assembled ␥-secretase complex.
Analysis of the Mechanism of ␥-Secretase Inhibition by DAPT Using a Competition for DAP-BpB Labeling-Based on enzymatic and biochemical analyses, it has been suggested that ␥-secretase harbors an "initial substrate binding site" which are distinct from the catalytic site (Refs. 8 -11; see Fig. 7). Thus, the molecular mechanism by which ␥-secretase is inhibited by small molecule inhibitors might be related to one or the other of the following; (i) occupation of the catalytic site to block the catalysis, (ii) occupation of the initial binding site to block the access of substrates, or (iii) different or indirect mechanisms (e.g. allosteric effect). To gain insights into the molecular mechanism of inhibition by DAPT as well as its binding region within PS CTF, we performed a competition assay of labeling by DAP-BpB. In this paradigm, co-existing inhibitors that occupy similar position to DAP-BpB within the ␥-secretase complex displace the photoprobe from its target and diminish labeling (Fig. 6). As labeling competitors, we chose different classes of compounds; dipeptidic (compound E, YO01027) (22,24), arylsulfonamide derivative (HF14057) (23), transition state analogue (L-685,458) (25), and Aib-containing ␣-helical peptide (peptide 15) (11). Pharmacological and biochemical analysis suggest that L-685,458 and peptide 15 directly target the catalytic and substrate binding sites, respectively (8 -12). In contrast, the molecular target(s) of arylsulfonamide inhibitors still remains unknown (10,42). We used these competitors at concentrations equivalent to 10 or 100 (or higher, if needed) times their IC 50 values in vitro. First we tested the dipeptidic compounds and found that they inhibit the DAP-BpB labeling in a similar manner to DAPT as expected, suggesting that these compounds bind to the same site to which DAPT binds. Interestingly, despite its distinct structural feature, arylsulfonamide HF14057 exhibited a similar labeling competition profile as DAPT, suggesting that arylsulfonamide might also compete for the DAPT binding site. In contrast, neither L-685,458 nor peptide 15 displaced the photoprobe at similar relative concentrations (10 and 100 times IC 50 ) as DAPT. However, they attained an almost complete inhibition at ϳ1,000 -10,000 times the IC 50 concentration. These results are consistent with the finding that the labeling of PS1 polypeptides by a transition-state analogueand a helical peptide-based photoligand was reduced by competition with DAPT (12,43). Thus, these results suggest that the DAPT binding site is distinct from either the "catalytic site" or the substrate binding site, although these sites may partially overlap or are located very close to each other.

DISCUSSION
In this study we found a stringent structural requirement for DAPT binding to PS CTF that should not only form a HMW complex with other co-factor proteins but also stay in an "active" state. Beher and co-workers (44) reported that the catalytic aspartates are necessary for binding of the transition state analogue inhibitor, L-685,458, to PS1. It is possible that these catalytic aspartates also contribute to the direct interaction of PS with DAPT. However, our observation that a high concentration of L-685,458 was required for the displacement of DAP-BpB suggests that DAPT targets a distinct site from the catalytic site. Thus, although the aspartate mutant PS is capable of forming stable HMW complexes together with other components, the integrity of the catalytic machinery (i.e. the catalytic site and the DAPT binding site) might be somewhat compromised in these mutants (45). We also think that the PALP motif is unlikely to form the DAPT binding site despite the fact that P433L mutant was not labeled by DAP-BpB, because this motif is also conserved among the signal peptide peptidase family and required its activity, on which DAPT has no effect (18,19,46).
How strictly, then, is the DAPT binding site defined? Is it located solely within the PS CTF with the help of other ␥-secretase components, and does it maintain the proper conformation of the former? Or alternatively, does it correspond to the interface between PS CTF and other components? In our present study specifically labeled bands corresponding to PS NTF, nicastrin, Aph-1, and Pen-2 were not detected by the immunoblot analysis using an anti-biotin antibody. To further address this issue, it would be helpful to use alternative photoprobes in which the benzophenone moiety is introduced at a different position within the structure of DAPT, although the derivatives that we designed in this way lost their inhibitory potencies and, consequently, the affinities to ␥-secretase (21) 9 . Taken together, it is most likely that the binding site of dipeptidictype ␥-secretase inhibitors is located in PS CTF, especially within the TMD7 or more C-terminal portion, and that its formation requires a fine-tuning of the structures around the catalytic machinery, whose integrity is maintained by the conserved aspartates and the PALP motif.
We also tried to dissect the inhibitory mechanism of DAPT by using various classes of inhibitors as competitors against photolabeling. We found that one of the arylsulfonamides competes with the binding of DAP-BpB with a similar relative concentration as that of DAPT as well as to its dipeptidic relatives, which was in good agreement with the results of enzymatic analysis (10,42). Thus, dipeptides and arylsulfonamides might share an identical binding site, which may be formed around the PS CTF. However, it is also possible that arylsulfonamides bind to a different site but allosterically affect the binding of DAPT to its cognate binding site. A systematic crosslinking study using photoactivable arylsulfonamide derivatives would be necessary to address this point. Intriguingly, we observed a weak but significant inhibition of DAP-BpB labeling by high concentrations of L-685,458 and peptide 15 in our labeling competition assays. This indicates that the DAPT binding site is distinct from, but overlaps with both the catalytic and substrate binding sites, or alternatively, these compounds may slightly alter the conformation of the DAPT binding site. Another possibility would be that both the transition state analogues and helical peptides are capable of binding to the DAPT binding site with a relatively low affinity in addition to binding to their cognate sites. However, in good agreement with our results, previous reports showed that DAPT reduced the labeling efficiency of other photoaffinity probes based on a transition-state analogue or a helical peptide inhibitor (12,43). Collectively, these results may support the "overlapping" hypothesis. According to a model proposed by Tian et al. (10), substrates are cleaved by ␥-secretase in the following manner (Fig. 7); (i) the substrate initially enters the substrate binding site and (ii) subsequently moves toward the catalytic site through a route connecting the two sites. During this process the ␣-helical structure of substrate might be unwound by an as yet unknown mechanism in a way to facilitate cleavage by ␥-secretase (47,48). (iii) Partially unwound substrate reaches the catalytic site and undergoes cleavage at multiple sites. Based on this model, it is most plausible to speculate that the inhibitory effect of DAPT may be executed by blocking the substrate movement or the unwinding process through either a direct or an indirect mechanism by altering the conformation of ␥-secretase (Fig. 7C). This model may account for the recent finding that cultured cells treated by DAPT, but not L-685,458, intracellularly accumulate very low amounts of the longer A␤ species, A␤1-46, whereas most of the A␤ secretion is inhibited (49,50). These results also support our hypothesis that the inhibition mechanism by DAPT is distinct from that by transition-state analogues. To obtain further support for this hypothesis, an exact determination of the DAPT binding site by mass spectrometry at the single amino acid level would be needed.
In summary, we have defined a novel functional domain in the C-terminal region of PS CTF with which small molecule ␥-secretase inhibi-  Substrates carrying ␣-helical TMD structure entered the "initial binding site" (1) then go on to the catalytic site undergoing an unwinding process (2,3). During this process the helical conformation near the scissile bond might be unwound. Then the substrates go on to the proteolysis, and the resulting fragments are liberated (4). C, inhibition mode of DAPT on ␥-secretase complex. DAPT might inhibit substrates to go through the "unwinding machinery" contributed by the PS1 CTF by occupying this site or alter its conformation indirectly by binding to the "allosteric site." tors interact that is distinct from either the catalytic site or the substrate binding site. Recently, Wolfe and co-workers (51) reported that selectivity of ␥-secretase substrates is modulated by binding of ATP to PS CTF, suggesting that PS CTF plays an important role not only as an active site domain but also as a contributor to the substrate docking site in intramembrane proteolysis. Nevertheless, whatever the precise molecular mechanism of PS CTF is, our findings are pharmacologically relevant and could have major therapeutic implications, as DAPT is a ␥-secretase-specific inhibitor. Further investigations using such chemical biological approach would provide several clues to the development of ␥-secretase-targeted therapeutics for Alzheimer disease.