γ-Secretase Modulator (GSM) Photoaffinity Probes Reveal Distinct Allosteric Binding Sites on Presenilin

Background: Potent GSMs have been identified that lower Aβ42; however, the mechanism of modulation is not well understood.

Results: The photoaffinity probe E2012-BPyne specifically labels PS1-NTF at a unique site.

Conclusion: Acid and imidazole GSMs bind to distinct sites on PS1-NTF and are differentially affected by L458.

Significance: Our results provide evidence for multiple binding sites within γ-secretase that confer specific modulatory effects.

Abstract

γ-Secretase is an intramembrane aspartyl protease that cleaves the amyloid precursor protein to produce neurotoxic β-amyloid peptides (i.e. Aβ42) that have been implicated in the pathogenesis of Alzheimer disease. Small molecule γ-secretase modulators (GSMs) have emerged as potential disease-modifying treatments for Alzheimer disease because they reduce the formation of Aβ42 while not blocking the processing of γ-secretase substrates. We developed clickable GSM photoaffinity probes with the goal of identifying the target of various classes of GSMs and to better understand their mechanism of action. Here, we demonstrate that the photoaffinity probe E2012-BPyne specifically labels the N-terminal fragment of presenilin-1 (PS1-NTF) in cell membranes as well as in live cells and primary neuronal cultures. The labeling is competed in the presence of the parent imidazole GSM E2012, but not with acid GSM-1, allosteric GSI BMS-708163, or substrate docking site peptide inhibitor pep11, providing evidence that these compounds have distinct binding sites. Surprisingly, we found that the cross-linking of E2012-BPyne to PS1-NTF is significantly enhanced in the presence of the active site-directed GSI L-685,458 (L458). In contrast, L458 does not affect the labeling of the acid GSM photoprobe GSM-5. We also observed that E2012-BPyne specifically labels PS1-NTF (active γ-secretase) but not full-length PS1 (inactive γ-secretase) in ANP.24 cells. Taken together, our results support the hypothesis that multiple binding sites within the γ-secretase complex exist, each of which may contribute to different modes of modulatory action. Furthermore, the enhancement of PS1-NTF labeling by E2012-BPyne in the presence of L458 suggests a degree of cooperativity between the active site of γ-secretase and the modulatory binding site of certain GSMs.

Introduction

Abnormal accumulation of β-amyloid (Aβ)⁴ peptides in brain regions critical for learning and memory has been implicated as causative for Alzheimer disease (AD) progression (1, 2). The genesis of a heterogeneous pool of Aβ peptides occurs via sequential proteolytic processing of the amyloid precursor protein (APP) by two aspartyl proteases: β-secretase followed by C-terminal cleavage within the membrane by γ-secretase (3, 4). γ-Secretase activity relies on the assembly of an active enzyme complex that is composed of a quartet of proteins: nicastrin, presenilin 1 or 2, Pen-2, and Aph-1a or -1b (5). As a result of this biochemical and molecular complexity, γ-secretase represents a challenging target for drug discovery.

Selectively modulating γ-secretase activity has emerged as a potential disease-modifying treatment strategy for AD because γ-secretase modulators (GSMs) can reduce the formation of pathogenic Aβ42 species (6–8). Unlike γ-secretase inhibitors (GSIs) (9), which inhibit the production of all Aβ peptides to the same degree, GSMs do not affect the total amount of Aβ produced. Additionally, GSMs do not result in an accumulation of APP C-terminal fragments (10), but instead shift the cleavage site specificity, leading to a reduction in Aβ42/40 and an increase in the shorter nontoxic forms of Aβ peptides such as Aβ38 and/or Aβ37. Importantly, and in contrast to GSIs, GSMs do not broadly inhibit the cleavage of other γ-secretase substrates that are critical for normal cellular signaling such as Notch (11, 12).

GSMs can be generally divided into two categories: NSAID-derived carboxylic acid-containing GSMs and non-NSAID imidazole-containing GSMs (13–15). The first generation GSMs include the NSAIDs: ibuprofen, indomethacin, sulindac sulfide, and flurbiprofen (6–8, 16). The NSAID-derived GSMs selectively lower the formation of Aβ42 with a concomitant increase in the generation of Aβ38, with little to no effect on the proteolysis of other γ-secretase substrates such as Notch. These compounds were important chemical tools that provided evidence that γ-secretase could be modulated to reduce the more pathogenic Aβ42 species, but that the in vitro potencies are weak (Aβ42 IC₅₀ > 50 μm) and brain penetration is limited. Several next generation NSAID-like GSMs with improved in vitro potency and brain penetration have recently been reported, including GSM-1, GSM-10h, and JNJ-40418677 (17–19). The non-NSAID-derived imidazole GSMs have a slightly different profile in that they lower the production of Aβ42 and Aβ40 while increasing the levels of Aβ38 and Aβ37 to differing degrees. Several members from this class have been disclosed, including E2012, NGP 555, RO-57, and AZ4800 (20–23).

Although significant progress has occurred with regard to the identification and characterization of GSMs with drug-like properties, the location of the GSM binding site(s) and the mechanism that confers the selective modulation is not well understood. NSAID-derived GSMs have been proposed to selectively bind to the C-terminal fragment of APP within the Aβ region (24, 25). However, this proposal has been disputed because these NSAID-based GSMs are prone to form aggregates that can bind nonspecifically to Aβ (26). In contrast, we and others have reported that photoaffinity probes from the piperidine acetic acid class of GSMs specifically label the N-terminal fragment of presenilin 1 (PS1-NTF) (27–29). The exact target of the imidazole GSM series is also controversial. Recently, an immobilized imidazole GSM derivative based on NGP 555 was reported to bind to Pen-2 (21). In contrast, a biotinylated photoaffinity probe based on the imidazole GSM RO-57 was found to specifically label PS1-NTF and PS2-NTF (22). Clearly, more research is needed to further clarify the mechanism and binding sites of these GSMs.

Here, we demonstrate that the photoaffinity probe E2012-BPyne derived from the imidazole GSM E2012 specifically labels PS1-NTF in cell membranes. We observed similar results when live HeLa cells or primary neuronal cultures were photolabeled with E2012-BPyne. Moreover, the labeling of PS1-NTF is not competed by the acid GSM-1. We also observed that the labeling of PS1-NTF by E2012-BPyne is enhanced in the presence of the active site-directed GSI L458. Taken together, our results provide evidence for multiple binding sites within the active γ-secretase complex that confer specific and selective modulatory effects on lowering pathogenic species of Aβ.

EXPERIMENTAL PROCEDURES

Compounds

The preparation of E2012-BPyne, [³H]E2012-BP, and RO-57-BPyne is described in the supplemental material. E2012 (US2006/0004013), GSM-1 (WO2006/043064), GSM-5 (27), BMS-708,163 (30), L-685,458 (L458), L-682,679 (L679) (31), and L458-BPyne (32) were prepared according to published methods. The helical peptide inhibitor that targets the substrate docking site (pep11, Boc-^dVG(Aib)^dV^dV^dI(Aib)^d(Thr(Bzl))^dV(Aib)-OMe) and the corresponding biotinylated peptide photoprobe (pep11-Bpa-Bt, Biotin-Acp-^dVG(Aib)^dV^dV^d(Bpa)(Aib)^d(Thr(Bzl))^dV(Aib)-OMe) were purchased from BEX (Tokyo, Japan) (33).

Cell-free γ-Secretase Activity Assay Using C100-FLAG as Substrate

The in vitro γ-secretase assay was modified from Li et al. (34) and Oborski et al. (35). In brief, HeLa P2 cell membranes (300 μg) prepared as described by Oborski et al. (35) were incubated with 0.25% CHAPSO and vehicle/drugs followed by the addition of 0.3 μm C100-FLAG substrate and incubation at 37 °C for 75 min. The reaction was stopped with radioimmunoprecipitation assay (RIPA) buffer. Samples were then diluted 10-fold in 5 m guanidine HCl and 50 mm Tris-HCl (pH 8) and incubated for an additional 30 min at 50 °C. Each reaction was subjected to solid phase extraction using an Oasis 60-μm HLB plate (60 mg; Waters Corp., Milford, MA), eluted in 2% NH₄OH and 90% MeOH (v/v), and evaporated to dryness. Aβ peptides were measured by ELISA time-resolved fluorescence. De novo Aβx-40 and Aβx-42 were captured using the cleavage-specific antibodies RN1219 and 10G3 (Rinat, Pfizer Inc., San Francisco, CA), respectively, with 4G8-biotin (Covance, Princeton, NJ) and streptavidin-europium (PerkinElmer Life Sciences) to report. Aβ38 was captured with anti-amyloid 6E10 mouse monoclonal antibody (Covance) and reported with anti-Aβ38 cleavage site rabbit antibody R341-biotin (36) and streptavidin-europium. Time-resolved fluorescence of europium (excitation 340 nm, emission 615 nm) was measured using the EnVision multilabel plate reader (PerkinElmer Life Sciences).

Photoaffinity Labeling of HeLa, H4-APP, and ANP.24 Membranes with Clickable GSMs Followed by Western Blot Analysis

Cells (HeLa and H4-APP overexpressing human wtAPP) were lysed by passing through M110L Microfluidizer (Microfluidics) and spun at 800 × g for 10 min. The resulting supernatant was centrifuged (100,000 × g, 1 h). The pellets were washed once, resuspended in 50 mm MES, 150 mm KCl, 5 mm CaCl₂, 5 mm MgCl₂, Complete protease inhibitors (Roche Applied Science), pH = 6, and stored at −80 °C. Membranes from ANP.24 cells overexpressing Aph-1a, nicastrin, and PS1 were prepared as described previously (37). Membranes (800 μg) diluted to a volume of 1.2 ml with PBS in 6- or 12-well plates were preincubated with competitor compounds or DMSO control for 30 min at 37 °C. The photoprobe was added at the designated concentrations for 1 h at 37 °C followed by UV irradiation (365 nm) for 30 min at 4 °C. After photocross-linking, membranes were precipitated by centrifugation at 100,000 × g for 30 min and resuspended in PBS with the aid of a Qiagen TissueLyser II (3 min, 25 shakes/min). CHAPSO (0.25%) was added followed by the click chemistry reagents (1 mm tris(2-carboxyethyl)phosphine, 1 mm CuSO₄, 0.1 mm tris-(benzyltriazolylmethyl)amine, and 0.1 mm biotin azide in 5% t-butyl alcohol with 1% DMSO), and the mixture was shaken for 1.5 h at room temperature. The membranes were pelleted by centrifugation at 150,000 × g for 30 min, washed once with PBS, and solubilized overnight in radioimmunoprecipitation assay (RIPA) buffer (1 ml) followed by affinity enrichment with streptavidin magnetic beads (Pierce). Proteins were eluted with 2 mm biotin in 2× lithium dodecyl sulfate (LDS) sample buffer (Life Technologies) and separated on a 4–12% NuPAGE Bis-Tris gel in MES running buffer (Life Technologies), transferred to nitrocellulose (Life Technologies), probed with anti-PS1-NTF antibody (Covance PRB-354P), and visualized with an Odyssey Infrared Imager (Li-COR Biosciences). Experiments were repeated at least three times, and a representative image is shown in Figs. 3–7.

FIGURE 3.

E2012-BPyne specifically labels PS1-NTF in HeLa membranes and PS1ΔE9 in proteoliposomes. A, photoaffinity labeling of E2012-BPyne (2 μm, lanes 1–3 and 1 μm, lanes 4–6) in HeLa membranes in the presence of absence of 20 μm (lanes 2 and 5) and 50 μm E2012 (lanes 3 and 6) followed by click chemistry with biotin-azide, streptavidin pulldown, and Western blot analysis with PS1-NTF antibody. B, E2012-BPyne (2 μm) does not label nicastrin, PS1-CTF, Aph-1a, or Pen-2, but it does specifically label PS1-NTF and the signal peptide peptidase dimer (SPP dimer) (50 μm E2012 was used for the competition experiment in lane 3). Input in lane 1 was 5 μg of HeLa membrane (0.8%). SPP mono, signal peptide peptidase monomer. C, photoaffinity labeling of E2012-BPyne (200 nm) in PS1ΔE9 proteoliposomes in the presence or absence of E2012 (5 μm) followed by click chemistry with TAMRA-azide, in-gel fluorescence, and Coomassie Blue staining.

FIGURE 4.

Competition studies reveal distinct binding sites on PS1-NTF for GSMs and GSIs. A and B, photolabeling of PS1-NTF with 1 μm E2012-BPyne in the presence or absence of 50 μm E2012, GSM-1, BMS-708,163, or L458 in membranes from HeLa cells (A) or H4-APP cells (B). C, photoaffinity labeling of GSM-5 (2 μm) in HeLa membranes in the presence or absence of GSM-1 (50 μm), E2012 (50 μm), BMS-708,163 (1 and 50 μm), and L458 (1 and 50 μm). D, photoaffinity labeling of E2012-BPyne (2 μm) in HeLa membranes in the presence or absence of E2012 (50 μm), L458 (1 μm), and the docking site inhibitor pep11 (50 μm). E, photoaffinity labeling of a biotinylated pep11 photoprobe pep11-Bpa-Bt (0.5 μm) in HeLa membranes in the presence or absence of pep11 (25 μm), L458 (1 μm) and E2012 (1–30 μm).

FIGURE 5.

The active site-directed GSI L458 potentiates the photoaffinity labeling of E2012-BPyne. A, photolabeling of PS1-NTF with 2 μm E2012-BPyne in the presence or absence of 50 μm E2012, 0.1 nm to 10 μm L458, or 10 μm L679 in membranes from HeLa cells. B and C, dose-responsive photolabeling of E2012-BPyne (30, 100, 300, 600 nm) in HeLa membranes in the presence of 100 nm L458 (B) and with competition by E2012 (50 μm) (C). D, photolabeling of PS1-NTF with 1 μm RO-57-BPyne in the presence or absence of E2012 (50 μm) and L458 (1 μm) in HeLa cell membranes. E, effect of L458 on the binding of [³H]E2012-BP in HeLa cell membranes. K_d and B_max were calculated (n = 3) as described under “Experimental Procedures” using SigmaPlot 8.0 software.

FIGURE 6.

Photoaffinity labeling of E2012-BPyne and L458-BPyne in ANP.24 cell line that accumulates full-length PS1 (FL PS1). E2012-BPyne (1 μm) and L458-BPyne (20 nm) preferentially label active PS1-NTF over inactive FL PS1, which is blocked in the presence of E2012 (50 μm) and L458 (1 μm), respectively.

FIGURE 7.

E2012-BPyne labels PS1-NTF in live cells. A and B, photolabeling of PS1-NTF with E2012-BPyne (0.1, 0.5, 2 μm) in the presence or absence of 1 μm L458 in live HeLa cells (A) and primary neuronal cultures (B). 14 DIV, 14 days in vitro.

Photolabeling of PS1ΔE9 Proteoliposomes with Clickable GSMs Followed by In-gel Fluorescence

The photoreactive probe E2012-BPyne (200 nm) was incubated with 25 μg of PS1ΔE9 proteoliposomes, which were prepared as described previously (38), and 0.25% CHAPSO for 1 h at 37 °C in the presence or absence of 5 μm E2012 in 250 μl of PBS followed by UV irradiation at 350 nm for 30 min. Cross-linked proteins were labeled with tetramethyl rhodamine (TAMRA) using copper-catalyzed azide alkyne cycloaddition click chemistry with 1 mm CuSO₄, 1 mm tris(2-carboxyethyl)phosphine, 0.1 mm tris-(benzyltriazolylmethyl)amine, 50 μm TAMRA-azide, in PBS with 5% t-butyl alcohol, 2% DMSO, and shaking for 1 h at 25 °C. Proteins were then precipitated with 1 ml of cold acetone at −20 °C for 30 min and washed once with 500 μl of cold acetone. Precipitated proteins were centrifuged at 15,000 × g for 10 min, the acetone was removed, and the protein pellet was air-dried for 10 min. The protein pellets were resolubilized in 50 μl of PBS buffer with 1% SDS, and 5 μl of sample was loaded on to an SDS-PAGE gel for protein band separation and then scanned for fluorescent bands. The same gel was then stained with Coomassie Blue to compare the total amount of protein loaded in each sample.

Photoaffinity Labeling in Live HeLa Cells

HeLa cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 20 mm l-glutamine, 20 units/ml penicillin, and 20 μg/ml streptomycin (all from Gibco) at 37 °C, 5% CO₂. HeLa cells were plated at 4 × 10⁵ cells/well in 6-well plates and allowed to reach 90% confluence. The cells were treated with L458 or DMSO control for 30 min followed by the photoprobe at the designated concentrations for 1 h at 37 °C, 5% CO₂ and then UV-irradiated (365 nm) for 15 min at 4 °C. Following photocross-linking, cells were washed three times with cold PBS, and lysed by sonication in 1 ml of PBS + Halt protease inhibitor mixture (Thermo Scientific). Membranes were precipitated by centrifugation at 100,000 × g for 30 min and processed as described above.

Photoaffinity Labeling in Primary Neuronal Cultures

Primary neuronal cultures were prepared from Sprague-Dawley rat pups by dissecting the cortex in Hanks' balanced salt solution supplemented with 2 mm GlutaMAX, 7 mm HEPES, 100 units/ml penicillin, and 100 μg/ml streptomycin (all from Gibco). Cortices were mechanically dissociated with a Pasteur pipette followed by enzymatic dissociation with 20 units/ml papain, 100 units/ml DNase I, and 0.5 mm EDTA in Earle's balanced salt solution (all from Worthington Biochemical) for 15–20 min at 37 °C. The reaction was stopped by the addition of minimum essential medium Eagle with Earle's balanced salts, 2 mm GlutaMAX, 100 units/ml penicillin, 100 μg/ml streptomycin (all from Gibco), 0.278% additional glucose (Sigma), and 10% heat-inactivated FBS (HyClone). Neurons were plated at 8–10 × 10⁷ cells per 100-mm poly-d-lysine-coated dishes and maintained in Neurobasal medium with B27, 2 mm GlutaMAX, 100 units/ml penicillin, and 100 μg/ml streptomycin. Neurons were treated with E2012-BPyne and photoaffinity-labeled as described previously for live HeLa cells.

Determination of Relative Equilibrium Binding Constants with [³H]E2012-BP in the Presence and Absence of L458

[³H]E2012-BP, where the alkyne reporter group of E2012-BPyne has been replaced with a methyl group containing three ³H, was synthesized and used as a radioligand (supplemental material). Binding to γ-secretase was assayed by measuring the counts associated with the cell membranes following photocross-linking. HeLa membranes (140 μg/200-μl assay volume) were incubated with radioligand at varying concentrations in the absence or presence of L458 (1 μm) in a 24-well microplate for 1.5 h at 37 °C. Nonspecific binding was defined by preincubating the cell membranes with an excess of the cold parent compound E-2012 (25 μm). Samples were irradiated at 365 nm for 35 min to photocross-link bound [³H]E2012-BP to the target site and then washed four times with 3 ml of PBS, 0.25% CHAPSO followed by centrifugation at 150,000 × g for 1 h. The final pellets were suspended in 10 ml of EcoLume liquid scintillation mixture and counted in a Wallac liquid scintillation counter. Apparent K_D and B_max values were calculated using the equation, y = B_max × x/(K_d + x) (where y and x represent the specific binding and ligand concentration, respectively), by SigmaPlot 8.0 software.

RESULTS

Design of Potent Clickable Photoaffinity Probe Based on E2012 Imidazole GSM

To investigate how GSMs may mechanistically modulate the genesis of Aβ peptides, we generated novel GSM-based photoaffinity probes (27). Here, we describe the design and synthesis of a clickable photoprobe based on the imidazole GSM E2012 (Fig. 1). We substituted the phenyl group on the right-hand side of E2012 with a benzophenone photoreactive group to affect covalent capture of the binding partner(s) upon UV irradiation. In addition, we replaced the methyl group of the methoxyphenyl moiety with a propargyl group to enable click chemistry-mediated conjugation with reporter groups (39, 40). The resulting clickable photoprobe E2012-BPyne lowered Aβ42 and Aβ40 and increased Aβ38 with efficacy similar to the parent GSM E2012 in a cell-free HeLa membrane assay using C100-FLAG as a recombinant APP substrate (Fig. 2A). The Aβ42 IC₅₀ values of E2012 and E2012-BPyne were 146 and 296 nm, respectively (Fig. 2B). In addition, the in vitro potencies of E2012 and E2012-BPyne in a whole-cell assay using Chinese hamster ovary cells overexpressing wtAPP (CHO-APP) were 144 and 255 nm, respectively (data not shown). Therefore, the clickable probe design resulted in a photoaffinity probe that had similar in vitro potency as the parent compound. In contrast to most biotinylated compounds that may have limited cell permeability, E2012-BPyne was cell-penetrant, enabling direct labeling in live cells and in primary neuronal cultures. The clickable strategy that we employed also expands the utility and versatility of photoaffinity probes in that labeled proteins can be tagged with any number of azide-linked reporter groups using click chemistry for target pulldown and fluorescent imaging applications (27, 32, 40–42).

FIGURE 1.

Structures of GSMs and GSIs used in this study.

FIGURE 2.

E2012-BPyne remains a potent γ-secretase modulator with an Aβ profile similar to E2012. A, representative Aβ profile of E2012-BPyne in cell-free γ-secretase activity assay using membranes from HeLa cells and recombinant C100-FLAG as the APP substrate. B, corresponding Aβ42 and Aβ40 IC₅₀ and Aβ38 EC₅₀ values for E2012-BPyne and E2012. IC₅₀ values were generated by nonlinear regression curve fit (mean ± S.E., n = 3).

Imidazole Photoprobe E2012-BPyne Labels PS1-NTF Specifically

To investigate the target of E2012-BPyne within the γ-secretase complex, we performed photoaffinity labeling studies in HeLa cell membranes, which are known to have high γ-secretase activity (43). E2012-BPyne was incubated with HeLa cell membranes in the presence or absence of E2012 and UV-irradiated to initiate photocross-linking to nearby proteins (Fig. 3A). The labeled proteins were tagged with biotin via click chemistry with biotin-azide and isolated by affinity chromatography with streptavidin. The eluted proteins were separated by SDS-PAGE, and Western blot analysis was utilized to identify various known components of γ-secretase. We identified PS1-NTF as a target protein for E2012-BPyne within the γ-secretase complex. The labeling of PS1-NTF by E2012-BPyne was concentration-dependent with more prominent labeling occurring with increasing concentrations of E2012-BPyne. Additionally, the labeling of E2012-BPyne to PS1-NTF was specific because little to no signal was observed when increasing concentrations (20–50 μm) of the parent compound E2012 were included in the reaction. We were not able to detect specific labeling for any other γ-secretase components such as PS1-CTF, Aph-1, Pen-2, or nicastrin (Fig. 3B).

To determine whether E2012-BPyne retained activity in the absence of any other components or substrates of γ-secretase, we conducted photoaffinity labeling studies in a reconstitution system consisting of recombinant PS1ΔE9 in proteoliposomes (27, 38). E2012-BPyne was incubated with PS1ΔE9 proteoliposomes in the presence or absence of E2012 followed by UV irradiation. The labeled proteins were fluorescently tagged via click chemistry with TAMRA-azide and detected by in-gel fluorescence (Fig. 3C). We found that E2012-BPyne (200 nm) robustly labeled PS1ΔE9 proteoliposomes. Importantly, the labeling by E2012-BPyne was completely blocked in the presence of E2012 (5 μm), confirming the direct and specific interaction of E2012-BPyne with PS1ΔE9 in this reconstituted proteoliposome system.

In addition, we wanted to determine whether E2012-BPyne labeled signal peptide peptidase, an aspartyl intramembrane protease with active site homology to PS1. We found that E2012-BPyne specifically labeled signal peptide peptidase, albeit faintly relative to PS1-NTF (Fig. 3B).

Competition Studies Reveal Distinct Binding Sites on PS1-NTF for Acid and Imidazole GSMs

We next investigated whether or not the acid and imidazole classes of GSMs would bind to similar or distinct sites within the γ-secretase complex. We performed photoaffinity labeling studies with E2012-BPyne in HeLa membranes in the presence or absence of either E2012 or GSM-1. As previously observed, the labeling of PS1-NTF by E2012-BPyne was significantly attenuated in the presence of E2012, whereas GSM-1 had no effect on PS1-NTF labeling (Fig. 4A). We also investigated whether overexpression of APP and hence β-secretase cleaved CTF (β-CTF) would influence the selective binding of E2012-BPyne to PS1-NTF by performing photoaffinity labeling experiments with membrane fractions prepared from H4 cells overexpressing wild type APP. Similar to the results we observed with HeLa membranes, E2012 was able to attenuate the binding of E2012-BPyne to PS1-NTF, whereas GSM-1 had no effect (Fig. 4B). We also performed photoaffinity labeling studies with GSM-5, our previously reported photoaffinity probe from the acid GSM series (27), in the presence or absence of either GSM-1 or E2012 (Fig. 4C). GSM-1 specifically blocked the labeling of PS1-NTF by GSM-5; however, E2012 had no effect on the labeling of PS1-NTF by GSM-5. Taken together, these results suggest that the different classes of GSMs bind to distinct sites on PS1-NTF within the γ-secretase complex.

The Active Site-directed GSI L458 Potentiates the Photoaffinity Labeling of E2012-BPyne

Because we observed differential binding of representative GSM classes to PS1-NTF, we next investigated whether or not different classes of structurally diverse GSIs would affect the labeling of E2012-BPyne to PS1-NTF. BMS-708,163, a representative sulfonamide GSI (30), had little to no effect on PS1-NTF labeling by E2012-BPyne or GSM-5 (Fig. 4, A–C). Consistent with this result, we previously reported that E2012 and GSM-1 do not block the photolabeling of PS1-NTF by 163-BPyne, a photoprobe derived from BMS-708,163 (44). Surprisingly, we observed a robust enhancement of PS1-NTF labeling by E2012-BPyne in the presence of the active site-directed inhibitor L458 (Fig. 4, A and B), which was not observed with the acid photoprobe GSM-5 (Fig. 4C). Moreover, the ability of L458 to enhance the binding of E2012-BPyne and labeling of PS1-NTF was not affected by the presence of excess APP holoprotein because a similar degree of signal was observed when the photoaffinity labeling reaction was conducted in either HeLa or H4-APP membranes (Fig. 4B). We also investigated whether the labeling of E2012-BPyne would be influenced by pep11, a helical peptide inhibitor that has been reported to target the substrate docking site on γ-secretase (33). In contrast to L458, pep11 did not affect the labeling of PS1-NTF by E2012-BPyne (Fig. 4D). In addition, we analyzed the ability of L458 and E2012 to block the labeling of PS1-NTF by the pep11-Bpa-Bt photoprobe (33). Unlike the parent pep11, L458 and E2012 did not prevent the photolabeling by pep11-Bpa-Bt (Fig. 4E). These results indicate that E2012 does not bind the same site as BMS-708,163, L458, or pep11. However, binding of L458 in the catalytic pocket of the enzyme seems to allosterically influence the binding site of E2012 in a manner that enhances the labeling of E2012-BPyne.

To further investigate the enhancement of E2012-BPyne binding to PS1-NTF in the presence of L458, we performed photoaffinity labeling studies in the presence of various concentrations of L458. We observed a concentration-dependent enhancement of E2012-BPyne binding to PS1-NTF in the presence of L458 (Fig. 5A). The enhanced binding of E2012-BPyne to PS1-NTF was apparent at low concentrations of L458 (1–10 nm), which is consistent with the reported potency of L458 in HeLa membranes (Aβ42 IC₅₀ = 0.3 nm). The extent of E2012-BPyne labeling increased with increasing concentrations of L458 up to 100 nm, which is a concentration where L458 binding should be saturated. To confirm that the enhancement of E2012-BPyne labeling to PS1-NTF by L458 was the result of specific labeling to an active γ-secretase complex, we performed photoaffinity labeling studies in the presence of L679, a considerably less active epimer of L458 (Aβ IC₅₀ > 1 μm) (31). When L679 (10 μm) was included in the photoaffinity labeling reaction, only a slight enhancement of labeling to PS1-NTF by E2012-BPyne was observed (Fig. 5A). Next, we performed the photoaffinity labeling using a dose response of E2012-BPyne in the presence of 100 nm L458 (Fig. 5B). We observed significant photolabeling of PS1-NTF with 100 nm E2012-BPyne, and the extent of labeling increased in a concentration-dependent manner. The majority of the labeling was blocked upon pretreatment with E2012 (Fig. 5C).

In contrast to the results with E2012-BPyne, previous studies have observed that L458 had little effect on, or partially competed with, the imidazole photoaffinity probe RO-57-BpB (22). It may be that these two series of imidazole-containing GSMs have different or only partially overlapping binding sites. To explore this further, we prepared a clickable version of the RO-57-BpB photoprobe, referred to as RO-57-BPyne. Consistent with previously reported results (22), we observed that the labeling of PS1-NTF by RO-57-BPyne was competed by E2012 and that the presence of L458 had little to no effect (Fig. 5D).

We next determined how the binding of L458 affects the E2012 interaction site within γ-secretase using a variation of a radioligand binding assay employing the radiolabeled photoprobe [³H]E2012-BP. We determined the apparent K_d and B_max in the presence and absence of L458 (Fig. 5E). The co-incubation of [³H]E2012-BP with L458 led to a 10-fold reduction in the K_d (K_d = 3.9 nm with L458 versus 39.7 nm without L458) and a 1.5-fold increase in the B_max (B_max = 69510 with L458 versus 47540 without L458) when compared with the radioligand alone. These results suggest that the enhancement in photoaffinity labeling of E2012-BPyne is primarily driven by an increase in the affinity or photolabeling efficiency of [³H]E2012-BP that occurs in the presence L458.

Preferential Labeling of E2012-BPyne to PS1-NTF

To investigate whether E2012-BPyne preferentially binds to an active γ-secretase complex (i.e. PS1-NTF) or is capable of labeling inactive full-length PS1 (FL PS1), we performed photoaffinity labeling in ANP.24 cells where both PS1-NTF and FL PS1 are present. ANP.24 cells co-overexpress PS1, Aph-1, and nicastrin, but only have endogenous expression of Pen-2 (45). Because Pen-2 is required for the endoproteolysis of FL PS1 into the catalytically active components PS1-NTF and PS1-CTF, ANP.24 cells accumulate stable FL PS1, whereas only endogenous levels of PS1-NTF are present. First, we demonstrated that the active site-directed GSI photoaffinity probe L458-BPyne (32) only labeled PS1-NTF in ANP.24 cells as expected (Fig. 6). Next, we utilized E2012-BPyne and also observed preferential labeling to PS1-NTF. In contrast, previous studies showed that GSM-1-BpB labeled both PS1-NTF and FL PS1 (28). Therefore, E2012 preferentially interacts and labels active γ-secretase, whereas GSM-1 appears to interact with both active and inactive γ-secretase complexes.

E2012-BPyne Labels PS1-NTF in Live Cells

Encouraged by recent studies that have used clickable photoaffinity probes for photocross-linking studies in live cells (41, 46, 47), we performed photoaffinity labeling with E2012-BPyne in live HeLa cells and primary cortical neurons to interrogate binding to γ-secretase in a more native cellular environment (Fig. 7). E2012-BPyne was administered to live HeLa cells or neurons with or without L458 followed by UV irradiation. The cells were lysed and centrifuged to precipitate the membranes. Click chemistry with biotin-azide was performed to biotinylate the labeled proteins followed by enrichment using streptavidin and Western blot analysis. Similar to the results we obtained from HeLa membranes, E2012-BPyne (2 μm) labeled PS1-NTF and L458 dramatically enhanced the labeling of PS1-NTF such that prominent labeling was observed even with very low concentrations of E2012-BPyne (100 nm) (Fig. 7A). Importantly, we also observed a similar result when we applied our clickable photoaffinity protocol to primary neuronal cultures, where we observed a similar enhanced photolabeling of PS1-NTF in the presence of L458 (Fig. 7B).

DISCUSSION

GSMs have emerged as potential disease-modifying treatments for AD because they reduce the formation Aβ42, but leave the total output of Aβ and the APP intracellular domain (AICD) as well as intracellular domains from other γ-secretase substrates unchanged. Thus far our knowledge of the GSM binding site(s) and mechanism of modulation is limited. Further clarity with regard to potential similarities and/or differences possessed by various classes of GSMs would expedite the advancement of this class of compounds as potential AD therapeutics. Previously, we reported that clickable photoaffinity probes based on the piperidine acetic acid GSM-1 bind directly to PS1-NTF (27). In addition, we demonstrated that the allosteric interaction of piperidine acetic acid GSMs (i.e. GSM-1) with PS1-NTF resulted in a conformational change leading to an increase in labeling of the active site-directed GSI photoprobe GY4.

Here, we extend our initial work by generating and characterizing an imidazole GSM photoprobe based on the structure of E2012, which provides insight into the interaction of GSMs with γ-secretase. First, we demonstrate that E2012 binds directly to PS1-NTF, which is also the target of GSM-1 (27, 28) and RO-57 (22) (Fig. 3). Second, we provide evidence that E2012 occupies a binding site on PS1-NTF that is distinct from the acid class of GSMs (i.e. GSM-1), the sulfonamide class of GSIs (i.e. BMS-708,163), and the helical peptide inhibitor pep11 (Fig. 4). The inability of GSM-1, BMS-708,163, and pep11 to effectively compete the specific labeling of E2012-BPyne to PS1-NTF and the inability of E2012 and BMS-708,163 to compete the specific labeling of GSM-5 to PS1-NTF supports the notion of distinct binding sites within an active γ-secretase complex. Third, we demonstrate that co-incubation with the active site-directed GSI L458 resulted in a dramatic enhancement in the labeling of PS1-NTF by E2012-BPyne in membranes as well as primary neuronal cultures, whereas the labeling of PS1-NTF by the acid photoprobe GSM-5 was not affected (Figs. 4 and 5). The differential labeling profile of E2012-BPyne and GSM-5 in the presence of L458 further supports the hypothesis that E2012 and GSM-1 have distinct binding sites on PS1-NTF. Our data suggest that binding of L458 in the active site of γ-secretase induces a conformational change resulting in enhanced binding and/or photocross-linking of E2012-BPyne to PS1-NTF (Fig. 8). Further work is required to more precisely understand the mechanism of this potentiation.

FIGURE 8.

Model depicting conformational change induced by L458 binding that enables better labeling of PS1-NTF by E2012-BPyne. Binding of L458 to the active site of γ-secretase causes a conformational change that induces a better fit for E2012-BPyne, which leads to increased efficiency of photolabeling.

L458 only binds to active γ-secretase complexes, which require PS1 endoproteolysis to generate the active site catalytic aspartic acid residues (31, 38, 43). The fact that E2012-BPyne labeling is enhanced in the presence of L458 may suggest that E2012-BPyne preferentially binds to an active γ-secretase complex. To explore this further, we performed photolabeling studies in ANP.24 cells, which co-express PS1, Aph-1, and nicastrin, but not Pen-2 (45). These cells accumulate FL PS1, which represents inactive γ-secretase, and contain endogenous levels of PS1-NTF, which represents active γ-secretase. We found that both the active site-directed GSI photoprobe L458-BPyne and the imidazole GSM photoprobe E2012-BPyne preferentially label PS1-NTF (Fig. 6). In contrast, our results (Fig. 4C) and those of others (28) demonstrate that the binding of L458 has little to no effect on the labeling of PS1-NTF by the acid photoaffinity probes (GSM-5 and GSM-1-BpB). Additionally, Ohki et al. (28) have reported that GSM-1-BpB, a biotinylated photoaffinity probe based on GSM-1, labels both PS1-NTF and the inactive full-length PS1 D385A mutant. Collectively, these results suggest that E2012-BPyne preferentially labels the active γ-secretase complex, whereas GSM-1-BpB can bind to and label an inactive γ-secretase complex as well.

Our data also suggest that E2012-BPyne occupies a similar yet distinct site than that of RO-57-BpB. Both photoprobes are competed by E2012; however, unlike E2012-BPyne, the labeling of RO-57-BPyne is not affected by the presence of L458 (Fig. 5D). Given the structural similarities of these imidazole-based GSMs, it is surprising that such differential effects are observed in the presence of L458. Consistent with the subtle differences in binding profiles described above, E2012 and the RO imidazole GSMs also have slightly different Aβ profiles (22, 23). Although both compounds reduce Aβ42 and Aβ40, E2012 increases Aβ37 > Aβ38, and RO-02 increases Aβ38 > Aβ37. Subtle differences in Aβ profiles have also been reported among other imidazole GSMs (23). Further studies to understand the subtle differences in binding and Aβ profiles of these different imidazole GSMs are clearly warranted.

Several previous studies demonstrating specific labeling of PS1-NTF with different GSM photoprobes have used high concentrations of the GSM competitor compounds (100 μm) (22, 28, 29). These competition experiments should be interpreted with caution because the high concentrations utilized for competition may have confounded the results due to the limited solubility and/or aggregation of the parent compound. For example, several NSAID-based GSMs are known to aggregate at concentrations greater than 50 μm (26). These colloidal aggregates could lead to nonspecific competition because they have been shown to envelop proteins, leading to nonspecific inhibition (48, 49). In our studies, we used competitor compounds at concentrations less than or equal to 50 μm. By using a novel clickable photoprobe design, we were able to generate photoaffinity probes that had similar potencies in whole-cell and cell-free γ-secretase assays when compared with the competitor compounds. For example, using 2 μm of E2012-BPyne photoprobe, we were able to compete the photolabeling of PS1-NTF with a 25-fold excess of E2012 (50 μm), minimizing the potential for nonspecific artifacts that may be interpreted as direct competition.

Furthermore, the clickable photoprobe E2012-BPyne was cell-permeable, allowing photoaffinity labeling studies in living cells and primary cultures of cortical neurons. We observed that E2012-BPyne labeled PS1-NTF in live HeLa cells with similar efficiency when compared with HeLa membranes, especially in the presence of L458. We found similar results in primary cortical neuronal cultures, which gives us additional confidence that our studies are relevant to brain-derived γ-secretase. Thus E2012-BPyne is a suitable tool for additional studies examining the modulation of γ-secretase in living cells where endogenous factors that may influence enzyme activity are present.

Detailed in vitro studies have now documented qualitative differences in the de novo genesis of various Aβ peptides that occur as a result of mutations in PS or APP and that are associated with familial forms of AD. Indeed, the qualitative kinetic differences that occur in the profile of Aβ peptides as a result of these familial forms of AD-associated mutations further support the hypothesis that primary amino acid changes in PS, which forms the catalytic active site for γ-secretase, significantly affect the generation of pathological species of Aβ (37, 50). Here, we utilized novel photoaffinity probes and a source of native γ-secretase from two different cell lines and primary cultures of cortical neurons to document the ability of different GSMs to bind directly to PS1-NTF at different sites. Additionally, we made the surprising observation that the labeling of E2012-BPyne to PS1-NTF is enhanced in the presence of very low concentrations of the active site-directed GSI L458, suggesting a degree of cooperativity between the γ-secretase active site and the GSM binding site(s). Although we did not observe any significant difference in the ability of E2012-BPyne to bind and label PS1-NTF in cells overexpressing wild type APP, it is possible that various γ-secretase substrates could also influence specific GSM binding sites, and this is an important area for further investigation (51, 52). In summary, these results support the hypothesis that multiple binding sites exist within the γ-secretase complex, each of which contribute to different modes of modulatory or inhibitory action. Importantly, the chemical biology tools used herein strengthen our ability to characterize the various classes of GSMs and are helping to elucidate the mechanism of these enigmatic modulators. Additional studies to more precisely define their binding sites on PS1-NTF are underway in these laboratories.