A Sensitive and Quantitative Assay for Measuring Cleavage of Presenilin Substrates*

The presenilin (PS) proteins are components of the (cid:1) -secretase activity, which is central in the pathogenesis of Alzheimer’s disease. Here we present a novel cell-based reporter gene assay for the quantification of PS-controlled (cid:1) -secretase cleavage of the Alzheimer amyloid precursor protein (APP). We show that this assay offers several advantages, including increased sensitivity and specificity, improved quantification of cleavage, and simultaneous detection of all (cid:1) -secretase cleavages in APP. Furthermore, the APP assay can be used in parallel with a similar assay that records (cid:1) -secretase cleavage of a Notch receptor. The use of these assays to analyze the effects of two known (cid:1) -secretase inhibitors and postulated PS active site mutants on APP and Notch processing demonstrated that inhibitors and mutants that differently affect Notch and APP cleavage can be identified rapidly. The possibility in using these assays for high throughput screening of candidate (cid:1) -secretase inhibitors for APP and Notch in parallel opens up new vistas to systematically search for novel inhibitors that selectively block APP cleavage while not affecting Notch signaling. plasmids by a subcloning approach. All con- structs were verified by sequencing using the ABI377 automated se-quencer (PerkinElmer Life Sciences). The construct encoding Notch (cid:2) E-GVP and the UAS-responsive reporter gene construct MH100 and the CMV- (cid:2) gal plasmid have been described previously. 2 Cell Transfection Experiments and Reporter Gene Assay— HEK293 cells and blastocyst-derived embryonic stem cells lacking PS1 and PS2 expression, BD8 cells, were cultured as described previously. All trans-fections with BD8 cells were carried out in 24-well tissue culture plates (Costar), and transfection of 293 cells was carried out in 10-cm cell culture dishes (Costar). For each well of the 24-well tissue culture dish, 200 ng of MH100, 50 ng of CMV- (cid:2) -gal plasmid, 100 ng of Notch (cid:2) E-GVP or C99-GVP plasmid, and either 100 ng of PS1, PS2, or empty pcDNA3 plasmids were mixed with PLUS reagent and LipofectAMINE according to the recommendations of the manufacturer (Invitrogen). The cells were har- vested and analyzed for reporter gene activity 48 h post-transfections. For the 10-cm dishes, 2 (cid:3) g of MH100, 400 ng of CMV- (cid:2) gal, and either 2 (cid:3) g of C99-GVPorNotch (cid:2) E-GVPweremixedandusedtotransfectthe293cells according to the LipofectAMINE PLUS protocol. The transfected 293 cells were split and seeded into 24-well plates, and (cid:1) -secretase inhibitors, were added. Twelve hours postdrug addition, the cells were harvested and analyzed as described previously.

PS 1 proteins and APP are frequently mutated in familial Alzheimer's disease (AD), and PS controls the ␥-secretase-mediated processing of APP. The level of ␥-secretase activity controls the amount of amyloid ␤-peptide (A␤) formation from the APP intermediate fragment C99 (see Fig. 1A). Familial ADassociated mutations in PS lead to a specific increase in production of the more fibril-prone A␤42 variant (1), a peptide species linked to the early pathological changes seen in AD (2,3). It is therefore of interest to develop pharmaceutical approaches that reduce the level of ␥-secretase activity acting on APP. However, a complicating factor for this approach is that PS also controls cleavage of Notch receptors (4), which are critical for many functions during development and in the adult organism. ␥-Secretase inhibitors aimed at reducing A␤ formation might thus also affect Notch cleavage and have unwanted side effects. Therefore, high throughput screening assays that record ␥-secretase cleavage of both APP and Notch would be useful for the identification of novel APP-specific ␥-secretase inhibitors. In this study we present a novel cell-based reporter gene assay for monitoring ␥-secretase cleavage on APP.

EXPERIMENTAL PROCEDURES
DNA Constructs-Silent mutations were first incorporated in a pcDNA3-C99 construct to create an AscI site immediately 3Ј of the nucleotides encoding the transmembrane-spanning region of C99. A cDNA encoding the DNA binding/transactivation domains Gal4/VP16 (GVP) was subsequently amplified by PCR and cloned into the AscI site of C99 generating the C99-GVP construct. Both PS1 and PS2 cDNAs were obtained in the pcDNA3 backbone. All PS1 mutants were generated by PCR using Pfu Turbo polymerase according to the QuikChange mutagenesis protocol (Stratagene, La Jolla, CA). The PS1 D257A/ D385A double mutant construct was generated out of the PS1 D257Aand PS1 D385A-encoding plasmids by a subcloning approach. All constructs were verified by sequencing using the ABI377 automated sequencer (PerkinElmer Life Sciences). The construct encoding Notch ⌬E-GVP and the UAS-responsive reporter gene construct MH100 and the CMV-␤gal plasmid have been described previously. 2 Cell Transfection Experiments and Reporter Gene Assay-HEK293 cells and blastocyst-derived embryonic stem cells lacking PS1 and PS2 expression, BD8 cells, were cultured as described previously. 2 All transfections with BD8 cells were carried out in 24-well tissue culture plates (Costar), and transfection of 293 cells was carried out in 10-cm cell culture dishes (Costar). For each well of the 24-well tissue culture dish, 200 ng of MH100, 50 ng of CMV-␤-gal plasmid, 100 ng of Notch ⌬E-GVP or C99-GVP plasmid, and either 100 ng of PS1, PS2, or empty pcDNA3 plasmids were mixed with PLUS reagent and LipofectAMINE according to the recommendations of the manufacturer (Invitrogen). The cells were harvested and analyzed for reporter gene activity 48 h post-transfections. For the 10-cm dishes, 2 g of MH100, 400 ng of CMV-␤gal, and either 2 g of C99-GVP or Notch ⌬E-GVP were mixed and used to transfect the 293 cells according to the LipofectAMINE PLUS protocol. The transfected 293 cells were split and seeded into 24-well plates, and ␥-secretase inhibitors, were added. Twelve hours postdrug addition, the cells were harvested and analyzed as described previously. 2

RESULTS AND DISCUSSION
In this report we describe a new sensitive and quantitative assay for ␥-secretase-mediated cleavage of APP. To generate a reporter APP protein, from which ␥-secretase cleavage can be recorded, we incorporated a Gal4 DNA-binding/VP16 transactivation (GVP) domain into the C99 form of APP (Fig. 1A), which is an immediate substrate for ␥-secretase in vivo (5). Cleavage of the resulting hybrid protein, hereafter referred to as C99-GVP, liberates the C-terminal region, including the GVP moiety, which is translocated to the nucleus due to the nuclear localization signals in GVP. In the nucleus, GVP specifically signals through a UAS-luciferase reporter gene by virtue of the strong transactivation domain and specific binding to a UAS promoter (Fig. 1A). The assay has several attractive features. First, it should be specific for ␥-secretase cleavage * This work was supported by Stiftelsen för ålderssjukdomar vid Karolinska Institutet, Loo och Hans Ostermans stiftelse, AMFs sjukförsä kringsfond (to J. L.), and Erik Rönnbergs Stiftelse (to J. N.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
since the GVP moiety is inserted only 13-15 amino acid residues C-terminal of the ␥-secretase cleavage sites, thus minimizing unspecific cleavage events. Second, it should be very sensitive since VP16 is a strong transactivation domain on the UAS promoter. Third, it will record the total ␥-secretase cleavage activity, i.e. ␥-cleavage generating both A␤40, the more abundant A␤ variant (6), and the pathogenic A␤42 species as well as minor cleavage products generated by processing at FIG. 1. Proteolytic processing of C99-GVP is dependent on PS and sensitive to ␥-secretase inhibitors. A, schematic view of the processing events of APP and the ␥-secretase assay. The three cleavage sites in APP (␣, ␤, and ␥) are depicted. The A␤ domain is shown in red. C99 is a direct substrate for ␥-secretase cleavage. Following cleavage at the ␥-secretase site, the intracellular domain containing the GVP domain translocates to the nucleus and initiates luciferase expression from the UAS promoter. B, C99-GVP is only a substrate for ␥-secretase in the presence of PS. ␥-Secretase-mediated cleavage of C99-GVP was analyzed in transiently transfected cells deficient for PS1 and PS2 (BD8 cells, bars [1][2][3][4] or in cells containing both PS1 and PS2 genes (HEK293 cells, bar 5). No cleavage is observed in BD8 cells transfected with C99-GVP in the absence of PS (bar 2), while substantial processing is recorded from the UAS-luc reporter gene after co-transfection of PS1 cDNA (bar 3) and PS2 cDNA (bar 4). Similar levels of cleavage were observed in the HEK293 cells transfected with C99-GVP-encoding cDNA (bar 5). The experiments were performed in triplicates and repeated at least three times. Data are presented as -fold activation of mock-transfected cells and show mean Ϯ S.E. **, p Ͻ 0.01 and *, p Ͻ 0.05 versus mock. C, the processing of C99-GVP is dependent on but saturable to the amount of PS expressed. Processing of C99-GVP in BD8 cells was analyzed in the presence of different amounts of transfected PS1-encoding cDNA. Note that above 10 ng of transfected PS1-encoding cDNA only a very minor increase in processing activity on the C99-GVP substrate was observed. Data are presented as -fold activation of mock-transfected cells and show mean Ϯ S.E. of two independent experiments with triplicates of each treatment. ***, p Ͻ 0.001 and *, p Ͻ 0.05 versus 0 ng of PS1 cDNA. Below is shown a Western blot examining the PS1 expression levels in the transfected BD8 cells using the antibody Ab14 (1:1000). D, ␥-secretase inhibitors block the processing of C99-GVP in a dose-dependent manner. The effect of two ␥-secretase inhibitors, MW167 and L-685,458, on the processing of C99-GVP was analyzed in transiently transfected HEK293 cells. Both inhibitors caused a dose-dependent decrease in ␥-secretase activity as measured by luciferase activity. The experiment was repeated at least three times with duplicates of each treatment. Data are presented as percent activity where 100% activity corresponds to the reporter gene activity obtained in the absence of ␥-secretase inhibitors. The graph shows mean Ϯ S.E. Parts of the extracts were analyzed by Western blot for C99-GVP expression using the 369 antibody (1:1000). Note that the slightly higher expression of C99-GVP in the presence of 50 -75 M MW167 does not result in an elevated reporter gene activity. This indicates the specificity of the assay for recording ␥-secretase activity. ***, p Ͻ 0.001; **, p Ͻ 0.01; and *, p Ͻ 0.05 versus C99-GVP with Me 2 SO. n, number of experiments; FL, full length; NTF, N-terminal fragment.
Novel Assay for ␥-Secretase Activity 6764 adjacent residues will be measured simultaneously. This is a marked improvement over current ␥-secretase assays that rely primarily on analyzing secreted A␤ species with specific antibodies while excluding the large, and structurally distinct (7), intracellular pool of A␤. Moreover, the C99-GVP construct harbors the native APP cytoplasmic tail, ensuring correct intracellular trafficking by the sorting signals contained within the tail (8,9). Finally, the assay records cleavage only from the C99-GVP protein and is thus insensitive to the levels of endogenous ␥-secretase substrates in the cell.
We first examined the specificity of the C99-GVP assay by recording ␥-cleavage in cells expressing different combinations of PS genes. After transfection of C99-GVP into HEK293 cells, which express endogenous levels of wild-type PS1 and PS2, a strong activation of the UAS-luciferase gene was observed (Fig.  1B). In contrast, transfection of C99-GVP into BD8 cells, which lack PS1 and PS2 expression, did not activate the reporter gene (Fig. 1B), indicating that cleavage is completely dependent on the presence of PS proteins. To test the assay for PS1 and PS2 individually, we transfected expression constructs of PS1 and PS2 into the BD8 cells and measured the cleavage of C99-GVP. This resulted in strong reporter gene activation in both cases, demonstrating that both PS1 and PS2 are capable of mediating ␥-cleavage of C99-GVP (Fig. 1B). These results are in agreement with recent reports from Drosophila in which a similar construct was successfully used for genetic analysis of PScontrolled ␥-cleavage (10).
To address whether the assay was quantitative, we transfected different amounts of PS1 into BD8 cells. The level of reporter gene activation increased with the amount of PS expressed, primarily in the range from 10 pg to 10 ng of added PS1 plasmid DNA (Fig. 1C). Interestingly, above 10 ng, only a minor increase in cleavage activity was observed. This may suggest that the production of the activated form of PS required for cleavage is limiting (i.e. by adding too much PS, the proteolytic machinery generating the N-and C-terminal PS fragments is saturated) or, alternatively, that a component other than PS in the active ␥-secretase complex is the rate-limiting determinant of cleavage (11). We also tested whether the assay responded quantitatively to the addition of two known ␥-secretase inhibitors, MW167 (12) and L-685,458 (13). Addition of both MW167 and L-685,458 resulted in a dose-dependent decrease in ␥-secretase activity (Fig. 1D).
As discussed above, ␥-secretase inhibitors that do not discriminate between APP and Notch cleavage may have dangerous side effects if used in a therapeutic setting. Both APP and Notch are cleaved at defined sites in the intramembraneous region, and thus it should be possible to use a very similar GVP-based assay to monitor ␥-secretase-mediated cleavage (site 3 cleavage) of Notch ( Fig. 2A). We have recently shown that a truncated Notch receptor that acts in a ligand-independent constitutive manner (14) and harbors the GVP moiety inserted immediately C-terminal of the site 3 cleavage site (Notch ⌬E-GVP) is cleaved specifically in the presence of PS in FIG. 2. Processing of Notch ⌬E-GVP is PS-dependent and susceptible to ␥-secretase inhibitors. A, overview of the ligand-induced processing events of the Notch receptor and the ␥-secretase assay. Notch ⌬E is a truncated form of a Notch receptor, which is cleaved at site 2 and site 3 in a ligand-independent constitutive manner. Following cleavage at the ␥-secretase site (site 3) the intracellular domain containing the GVP domain translocates to the nucleus and initiates luciferase expression from the UAS promoter. B, PS proteins are required for Notch ⌬E-GVP processing. Cleavage of Notch ⌬E-GVP in BD8 cells was analyzed in the presence of different amounts of transfected PS1encoding cDNA. As for C99-GVP (Fig. 1C), only a very minor increase in ␥-secretase-mediated cleavage of the Notch ⌬E-GVP substrate was observed above 10 ng of transfected PS1-encoding cDNA. Data are presented as -fold activation of mock-transfected cells and show mean Ϯ S.E. of two independent experiments with triplicates of each treatment. ***, p Ͻ 0.001 and **, p Ͻ 0.01 versus 0 ng PS1 cDNA. C, ␥-secretase inhibitors inhibit processing of Notch ⌬E-GVP. As for C99-GVP, both MW167 and L-685,458 caused a dose-dependent decrease in ␥-secretase activity. The experiment was repeated at least three times with duplicates of each treatment. Data are presented as percent activity where 100% activity corresponds to the reporter gene activity obtained in the absence of ␥-secretase inhibitors. The graph shows mean Ϯ S.E. of three independent experiments with duplicates of each treatment. ***, p Ͻ 0.001; **, p Ͻ 0.01; and *, p Ͻ 0.05 versus Notch ⌬E-GVP with Me 2 SO. Protein levels of Notch ⌬E-GVP in the presence of different concentrations of MW167 were assessed by Western blotting using the monoclonal c-Myc antibody 9E10 (1:200). luc, luciferase.
Novel Assay for ␥-Secretase Activity 6765 cultured cells. 2 We show here that Notch ⌬E-GVP, like C99-GVP, can respond in a graded fashion to varying amounts of PS (Fig. 2B) and ␥-secretase inhibitors (Fig. 2C). As in the case for C99-GVP, addition of PS1 above 10 ng does not significantly increase Notch cleavage efficiency. In addition, preliminary experiments suggest that expression of increasing amounts of Notch ⌬E results in a dose-dependent attenuation of the processing of C99-GVP (data not shown). These data further demonstrate the similarities in processing between APP and Notch. Both ␥-secretase inhibitors displayed similar potency in inhibiting Notch ⌬E-GVP and C99-GVP processing. The IC 50 value for MW167 was calculated to be ϳ50 M, which is slightly higher than the previously reported value of 20 M (20). This discrepancy in IC 50 values probably reflects the different methodologies used in the two investigations. L-685,458 inhibited C99-GVP and Notch ⌬E-GVP processing with an IC 50 value of 200 nM. The IC 50 for C99-GVP was comparable to the IC 50 reported for L-685,458 in intact cells (48 -775 nM depending on the cell line used and whether secreted A␤40 or A␤42 was measured) (13). Thus, the similarity of the data obtained in this study and other reports confirm the validity of the C99-GVP reporter gene assay to address ␥-secretase activity on APP.
To learn whether the assay was capable of recording differences between Notch and APP processing, we used artificial missense mutations in PS that affect two aspartate residues at positions 257 and 385 thought to be involved in the PS active site that have been shown to differentially alter APP and Notch processing (17). We generated single or double aspartate to alanine mutants at positions 257 and 385 in PS1 and tested the effects in the C99 and Notch GVP assays in the PS-deficient BD8 cells (Fig. 3). Transient expression of the D257A, D385A, and D257A/D385A mutants into BD8 cells resulted in no detectable signaling from the C99-GVP reporter construct. In contrast, the D385A mutant rescued 15% of the signaling for Notch ⌬E-GVP, while the other two mutants did not activate the reporter gene. This finding is interesting in light of previous data. Our data on APP processing support the original finding by Wolfe and co-workers (18) that the aspartate residues are critical in PS-mediated cleavage of APP but disagree with data from Haass et al. (17). The D385A mutant has in all previous reports been shown to inhibit Notch processing in a potent way (17,18). The residual cleavage activity of the D385A PS1 mutant in our assay was therefore unexpected but may reflect the sensitivity of the GVP-based assay and that it was conducted in a PS-null background, avoiding possible interference of wild-type, endogenously expressed PS. Moreover, recent data have emerged that challenge the view that PS proteins are direct executors of the ␥-secretase cleavage (19). 2 Irrespective of the finer details, our data show that the GVPbased assays for APP and Notch are able to identify situations in which APP and Notch processing differ.
In conclusion, we provide a novel assay to monitor ␥-secretase processing of APP and show that it can be used together with a similar assay for Notch processing. The APP assay is highly sensitive and quantitative and can simultaneously record all ␥-secretase-mediated cleavages. The fact that the GVPbased assays for APP and Notch are based on transfected cells also render these assays suitable for large scale testing of multiple compounds for ␥-secretase inhibition in an unbiased high throughput fashion. In principle, it should be possible to screen several thousands of compounds for their effect on ␥-secretase activity in a few days. The observation that these assays can score differences in processing between APP and Notch makes the assays very attractive as tools to identify novel ␥-secretase modulators that differentially affect APP and Notch processing.
FIG. 3. PS1 active site mutants differentially affect ␥-secretase-mediated cleavage of C99-GVP and Notch ⌬E-GVP. cDNAs encoding PS1 D257A, D385A, or D257A/D385A were transfected into BD8 cells together with either C99-GVP or Notch ⌬E GVP. ␥-Secretasemediated processing was recorded as luciferase activity where the activity obtained in the presence of wild-type PS1 was set to 100%. Note that the expression of none of the mutants resulted in any detectable processing from either the C99-GVP or the Notch ⌬E-GVP construct, except for the D385A mutant, which rescued cleavage from the Notch ⌬E GVP construct by 15%. The graph shows mean Ϯ S.E. of three independent experiments with triplicates of each transfectant. ###, p Ͻ 0.001 versus C99-GVP without PS1 cDNA. ϩϩϩ, p Ͻ 0.001 versus Notch ⌬E-GVP without PS1 cDNA. ***, p Ͻ 0.001 versus C99-GVP ϩ PS1 wild-type cDNA. § § §, p Ͻ 0.001 versus Notch ⌬E-GVP ϩ PS1 wild-type cDNA. WT, wild type.