Presenilin Is the Molecular Target of Acidic γ-Secretase Modulators in Living Cells

The intramembrane-cleaving protease γ-secretase catalyzes the last step in the generation of toxic amyloid-β (Aβ) peptides and is a principal therapeutic target in Alzheimer's disease. Both preclinical and clinical studies have demonstrated that inhibition of γ-secretase is associated with prohibitive side effects due to suppression of Notch processing and signaling. Potentially safer are γ-secretase modulators (GSMs), which are small molecules that selectively lower generation of the highly amyloidogenic Aβ42 peptides but spare Notch processing. GSMs with nanomolar potency and favorable pharmacological properties have been described, but the molecular mechanism of GSMs remains uncertain and both the substrate amyloid precursor protein (APP) and subunits of the γ-secretase complex have been proposed as the molecular target of GSMs. We have generated a potent photo-probe based on an acidic GSM that lowers Aβ42 generation with an IC50 of 290 nM in cellular assays. By combining in vivo photo-crosslinking with affinity purification, we demonstrated that this probe binds the N-terminal fragment of presenilin (PSEN), the catalytic subunit of the γ-secretase complex, in living cells. Labeling was not observed for APP or any of the other γ-secretase subunits. Binding was readily competed by structurally divergent acidic and non-acidic GSMs suggesting a shared mode of action. These findings indicate that potent acidic GSMs target presenilin to modulate the enzymatic activity of the γ-secretase complex.


Introduction
Alzheimer's disease (AD) is the most common age-related neurodegenerative disease with an estimated 5.4 million patients in the USA [1]. It is believed that progressive neurodegeneration and cognitive decline in AD are triggered by oligomerization and accumulation of toxic amyloid-b (Ab) peptides in the brain. The amyloid hypothesis is strongly supported by the analysis of earlyonset familial forms of AD (FAD), which has demonstrated that modest overproduction of the oligomerization-prone Ab42 peptides in the brain is sufficient to cause AD with complete penetrance [2].
The intramembrane-cleaving protease c-secretase is responsible for the last step in the proteolytic release of Ab42 peptides from the amyloid precursor protein (APP), and is a principal therapeutic target in AD [3]. c-Secretase is a multi-subunit aspartyl protease with the presenilin (PSEN) proteins, either PSEN1 or PSEN2, as its catalytic core. PSEN proteins encompass nine transmembrane domains (TMDs) and are endoproteolytically cleaved during assembly of the c-secretase complex into N-and C-terminal fragments that remain non-covalently associated. The PSEN fragments are incorporated together with three accessory proteins, nicastrin, anterior pharynx defective-1 (APH-1) and presenilin enhancer-2 (PEN-2), into high molecular weight complexes that display proteolytic activity [3]. Two critical aspartate residues in TMD6 and TMD7 of PSEN form the active center of c-secretase [4]. How c-secretase accomplishes the hydrolysis of peptide bonds in the hydrophobic environment of the membrane is only partially understood. TMDs 6 and 7 of PSEN around the catalytic aspartate residues form a hydrophilic cavity within the membrane that may allow access for water molecules required to hydrolyze peptide bonds [5,6]. In addition, it appears that the substrate initially binds to a distinct substrate-binding (docking) site on the outer surface of the c-secretase complex and is subsequently transported into the active site and cleaved [7].
Despite the limited knowledge of the catalytic mechanism of csecretase and the lack of high-resolution structural data, a large number of small molecule c-secretase inhibitors (GSIs) with excellent potency and in vivo properties have been developed [8]. Unfortunately, both preclinical and clinical studies have demonstrated that inhibition of c-secretase is associated with prohibitive side effects due to suppression of Notch processing and signaling [8,9]. c-Secretase modulators (GSMs) are small molecules that selectively lower generation of the highly amyloidogenic Ab42 peptides but spare Notch processing, and might be a safer alternative to GSIs [10]. The first GSMs were described in the class of non-steroidal anti-inflammatory drugs (NSAIDs) [11]. However, these NSAID-type GSMs suffered from low potency and brain penetration. Recently, GSMs with nanomolar potency and favorable pharmacological properties have been reported in two major structural classes: carboxylic acids with structural similarities to NSAIDs (acidic GSMs) and compounds based on bridged aromatics that do not resemble NSAIDs and lack a carboxylic acid group (non-acidic GSMs) [10].
Several studies have confirmed that GSMs modulate enzyme activity in cell-free c-secretase assays [12,13,14]. However, aside from this compelling evidence that GSMs directly interact with the c-secretase complex, their molecular mechanism remains largely undefined. Photo-affinity labeling studies have identified PSEN as the molecular target of different classes of GSIs [7,15,16,17]. In contrast, previous attempts to elucidate the molecular target of GSMs have produced conflicting results and both the substrate amyloid precursor protein (APP) and subunits of the c-secretase complex have been proposed [18,19,20,21]. Using a novel photoprobe based on a potent piperidine GSM, we have now demonstrated that PSEN is the molecular target of acidic GSMs in living cells.

Results
We recently presented the GSM BB25, derived from Merck patent WO20006043064, which lowered Ab42 levels in cellular assays with an IC 50 = 87 nM and served as a starting point for the synthesis of our GSM photo-probe AR243 [22] (Figure 1). In this compound class, potent acidic GSMs such as GSM-1 with good bioavailability, brain penetration and Ab42-lowering efficacy in AD mouse models have been described [23,24,25]. Because of inconsistent results in previous attempts to establish the GSM target [18,19], we applied strict criteria to the design of the photoprobe. These included potency in the nanomolar range, small size to avoid steric hindrance, and incorporation of a diazirine photoreactive group to allow cross-linking in living cells. Structure activity relationships that we obtained during the synthesis of BB25 enabled us to synthesize the photo-probe AR80 matching these criteria (Figure 1). The preserved GSM activity of AR80 (IC 50 = 190 nM) confirmed the tolerance of the scaffold for the introduction of the diazirine group on the para position of the lower aromatic ring, providing a non-invasive probe due to near complete similarity with the parent compound BB25. Subsequently, the biotin moiety for affinity purification was placed on the alkynyl side chain, yielding the photo-probe AR243. Briefly, compounds were synthesized via a copper-catalyzed three component coupling reaction between a racemic piperidine 1, a diazirinyl-benzaldehyde 2 and a terminal alkyne 3 ( Figure 2A).
To establish the bioactivity of the photo-probe, CHO cells with stable overexpression of wild type APP and PSEN1 were treated with AR243 and Ab levels in culture supernatants were measured by sandwich ELISA. This demonstrated that AR243 behaved like a typical GSM and lowered Ab42 levels with an IC 50 = 290 nM, Figure 1. Chemical structures and Ab42-lowering activities of c-secretase modulators used in this study. IC 50 values were determined after treatment of CHO cells with stable co-expression of wild type human APP and PSEN1 as described in the Experimental Procedures. Note that AR366 is an inverse c-secretase modulator with Ab42-raising activity. doi:10.1371/journal.pone.0030484.g001 comparable to the potency of the parent compound BB25 ( Figure 2B).
To identify the molecular target of AR243, cellular membranes were prepared from neuronal N2a-ANPP cells, which stably overexpress APP and all four subunits of the c-secretase complex [26], and solubilized in 0.25% CHAPSO. These conditions have previously been shown to preserve the integrity and bioactivity of the c-secretase complex [7,16]. Subsequently, membrane preparations were incubated with 0.5 mM of the photo-probe or DMSO vehicle, UV irradiated, and putative target proteins were affinity purified using streptavidin beads. The purified material was resolved on SDS gels and probed by Western blotting with antibodies against the four subunits of the c-secretase complex or the substrate APP. Western blotting with the monoclonal antibody PSN2 demonstrated that the GSM photo-probe AR243 bound the N-terminal fragment of PSEN1 (PSEN1-NTF) ( Figure 3). Omitting the UV irradiation or incubation of membranes with a 20-fold higher concentration of compound AR80 (IC 50 = 190 nM), which was structurally identical to the photo-probe but lacked the biotin moiety, did not result in a Western blotting signal. As expected, adding an excess of the parent compound BB25 (100 mM) during incubation of membrane preparations with the photo-probe efficiently reduced the labeling signal. Western blotting for the C-terminal fragment of PSEN1 (PSEN1-CTF) and the remaining c-secretase subunits APH-1, PEN-2 and Nicastrin did not demonstrate binding by the photo-probe AR243. Likewise, neither APP nor its membrane-bound C-terminal fragments (APP-CTFs), which are generated after shedding of the APP ectodomain and are the proximate c-secretase substrates, were labeled by the photo-probe ( Figure 3). Next, we sought to confirm binding of the GSM photo-probe to endogenously expressed PSEN1-NTF. Since the PSN2 antibody is specific for human PSEN1, we used membrane preparations from human HEK293T cells and performed the photo-affinity labeling as described above. Western blotting demonstrated labeling of endogenous PSEN1-NFT but not PSEN1-CTF by AR243 ( Figure 4). Control experiments including displacement of the photo-probe by an excess of parent compound BB25 demonstrated specificity of the binding to PSEN1-NTF.
Diazirine photo-reactive groups are activated by long-wave UV light $300 nm, which permits the use of diazirine-based photoprobes in living cells without the induction of acute cellular toxicity. Hence, to provide critical confirmation for our in vitro photo-labeling results, we treated N2a-ANPP cells with 0.2 mM of the photo-probe AR243 or DMSO vehicle and irradiated the live cells with UV light ( Figure 5A). Subsequently, cellular membranes were prepared and target proteins of the photo-probe were captured and analyzed as described for the cell-free experiments above. Western blotting with PSN2 antibody demonstrated efficient labeling of PSEN1-NTF, which was abolished by omitting UV irradiation ( Figure 5B). As expected, incubation of cells with DMSO or a 100-fold higher concentration of compound AR80 lacking the biotin moiety did not result in a positive labeling signal. In accordance with the labeling results in vitro, Western blotting did not demonstrate binding of the photo-probe AR243 to PSEN1-CTF in vivo ( Figure 5B). None of the compound treatments nor the UV irradiation caused cellular toxicity as monitored by alamar-BlueH reagent (data not shown). However, because we observed dose-dependent cytotoxicity with BB25 concentrations .10 mM, we did not pursue competition experiments in living cells.
To investigate whether a binding site within the PSEN1-NTF is unique to our photo-probe or likely common to potent secondgeneration GSMs, we performed competition experiments with other acidic and non-acidic GSMs in our cell-free system using membranes from N2a-ANPP cells. JNJ-40418677, an acidic GSM with considerable structural diversity to the photo-probe and an IC 50 = 290 nM, abolished binding of the photo-probe to PSEN1-NTF ( Figure 6A). In addition, three non-acidic GSMs with IC 50 values below 100 nM competed binding of the photo-probe to variable degrees. Whereas the non-acidic compound 1 reduced binding of the photo-probe as effectively as the acidic compound JNJ-40418677, compound 2 and E-2012 caused more modest reductions ( Figure 6B). Hence, in aggregate, the results of our competition studies indicate that acidic and non-acidic GSMs might share an overlapping binding site within the PSEN1-NTF. Finally, earlier studies had demonstrated that derivatization of acidic GSMs at the carboxylic acid group resulted in inverse GSMs that increase Ab42 levels and concomitantly decrease the production of shorter peptides such as Ab38 [27]. Accordingly, through esterification we obtained the inverse GSM photo-probe AR366, which increased Ab42 levels in our cellular assay with an IC 50 = 1.1 mM while dose-dependently lowering Ab38 levels. Incubation of membranes from N2a-ANPP cells with 500 nM of AR366 also resulted in photo-affinity labeling of PSEN1-NTF suggesting that acidic GSMs and inverse GSMs either target the same or a closely related binding site ( Figure S1). In conclusion,  . The acidic c-secretase modulator photo-probe AR243 targets the N-terminal fragment of PSEN1. Cellular membranes were prepared from neuronal N2a-ANPP cells, solubilized in 0.25% CHAPSO, and incubated with 0.5 mM of the photo-probe or DMSO vehicle. After UV irradiation, putative target proteins were affinity purified using streptavidin beads as described in the Experimental Procedures. Western blotting of the purified material demonstrated that the GSM photo-probe AR243 bound the N-terminal fragment of PSEN1 (PSEN1-NTF). Omitting the UV irradiation or incubation of membranes with a 20-fold higher concentration of compound AR80, which lacked the biotin moiety, did not result in a Western blotting signal. Adding an excess of the parent compound BB25 (100 mM) during incubation with the photo-probe efficiently reduced the labeling signal. Western blotting did not demonstrate binding of the photo-probe AR243 to the C-terminal fragment of PSEN1 (PSEN1-CTF) or any of the remaining c-secretase subunits APH-1, PEN-2 and Nicastrin. Likewise, neither full-length APP nor its membrane-bound C-terminal fragments (APP-CTFs) were labeled by the photo-probe. The asterisk depicts an unspecific protein band of slower mobility that is present under all conditions. Material that was bound non-specifically to the streptavidin beads is visible in the supernatant of the first washing steps but is completely removed after the third wash. Input represents 0.02% (PSEN1-NTF) or 0.2% (all other proteins) of the total membrane material. doi:10.1371/journal.pone.0030484.g003 Figure 4. The c-secretase modulator photo-probe AR243 targets endogenous PSEN1 in human HEK293T cells. Photo-affinity labeling studies were performed with membrane preparations from human HEK293T cells as described in Fig. 1. Western blotting of purified target proteins demonstrated labeling of endogenous PSEN1-NFT but not PSEN1-CTF by AR243. Co-incubation with an excess of parent compound BB25 (100 mM) caused displacement of the photo-probe, demonstrating specificity of the binding to PSEN1-NTF. Input represents 0.02% of the total membrane material. doi:10.1371/journal.pone.0030484.g004 our findings identify PSEN as the molecular target of potent acidic GSMs using established cell-free conditions and living cells.

Discussion
Limiting Ab production and accumulation in the brain through inhibition of c-secretase remains a rational approach for treatment or prevention of AD. However, GSIs that function as classical protease inhibitors and reduce enzymatic activity have displayed mechanism-based toxicity, which appears to be largely attributable to suppression of Notch processing [8]. In contrast, GSMs selectively lower production of the amyloidogenic Ab42 peptides and have been shown to spare processing of Notch [11,13].
While chemical development has yielded a second generation of GSMs with nanomolar potencies and improved brain permeability, the understanding of the molecular mechanism of GSMs is still rudimentary [10]. Photo-affinity labeling and cross-competition studies have demonstrated that, depending on their chemical structure and mechanism of action, GSIs target either the active site of c-secretase, the docking site or an allosteric site, which are all located within the PSEN proteins [7,15,16,17]. Surprisingly, previous photo-affinity labeling studies have suggested that the binding site of NSAID-type GSMs resides within the substrate APP and not within PSEN or one of the three other c-secretase subunits [19]. In these studies, photo-probes based on the NSAIDtype GSM flurbiprofen and the inverse GSM fenofibrate with benzophenone as a photo-reactive moiety were shown to bind APP, either in the presence of purified recombinant APP-CTFs or in crude membrane preparations of APP overexpressing cells. However, the concept of GSMs targeting the substrate APP contradicted important earlier observations, which suggested interaction of GSMs with the enzyme complex and specifically with PSEN. (i) Kinetic studies have demonstrated non-competitive inhibition for NSAID-type GSMs in c-secretase in vitro assays [12,13]. This indicated that the inhibitory effect of GSMs on Ab42 production could not be overcome by increasing substrate concentrations, which would be expected for compounds with a primary binding site within the substrate. (ii) While GSMs do not impair the release of the intracellular signaling domain from Notch-1, which is critical for Notch signaling, it has been convincingly demonstrated that NSAID-type GSMs modulate csecretase cleavage events in the middle of the Notch-1 TMD [28]. These are analogous to the cleavage events in the APP TMD that generate the Abpeptides suggesting that proteolytic processing of at least APP and Notch is affected in a similar manner, thereby questioning the specificity of GSMs for APP. (iii) The aspartyl intramembrane-cleaving protease signal peptide peptidase (SPP) has been demonstrated to harbor a binding site for NSAID-type The active GSM photo-probe contains a diazirine photo-reactive group and a biotin moiety for affinity purification. Live cells were treated with the photo-probe and subsequently irradiated in two steps at 365 and 302 nm. Note that the use of longwave UV light $300 nm did not induce acute cellular toxicity. Upon UV irradiation, the photo-probe forms a stable adduct with its specific target. Cellular membranes were prepared and solubilized, and target proteins of the photo-probe were captured and affinity purified using streptavidinfunctionalized beads. After extensive washing, the purified material was eluted and resolved on SDS gels and probed by Western blotting. (B) N2a-ANPP cells were incubated with 200 nM of AR243 or DMSO vehicle, and target proteins of the GSM photo-probe were purified as described above. Western blotting with the antibody PSN2 demonstrated binding of the photo-probe to PSEN1-NTF. Incubation of cells with 20 mM of compound AR80 that lacked the biotin moiety, or omission of UV irradiation did not result in a labeling signal. Consistent with the in vitro labeling results, Western blotting with an antibody against the PSEN1-CTF did not demonstrate binding of the photo-probe confirming that the binding site is located within the PSEN1-NTF. doi:10.1371/journal.pone.0030484.g005 GSMs [29,30]. SPP is homologous to PSEN but functions as a homodimer without accessory proteins, and c-secretase and SPP substrates do not show any overlap. This suggested that c-secretase contains a binding site for GSMs that is conserved in SPP and, accordingly, should be present in PSEN [29,30]. (iv) Circumstantial evidence for the interaction of GSMs with the enzyme has been provided by the observation that some FAD-associated PSEN but not APP mutations render cells resistant to GSMs [22,25,31,32].
Using a novel GSM photo-probe with nanomolar potency, we have now identified the PSEN1-NTF as the molecular target of acidic GSMs. In contrast, we did not observe binding of the photoprobe to the substrate APP or any of the other c-secretase subunits. Importantly, our GSM photo-probe with a diazirine photo-reactive moiety allowed us to confirm targeting of the PSEN1-NTF in living cells. A photo-probe with inverse Ab42raising activity similarly labeled PSEN1-NTF indicating that acidic GSMs and inverse GSMs likely target similar or identical binding sites. Under established cell-free conditions that preserve the integrity of the c-secretase complex, binding of the photoprobe to PSEN1-NTF was readily competed by unrelated acidic and non-acidic GSMs indicating that our findings are not limited to the specific class of piperidine GSMs. These findings are to some extent consistent with a recent study by Kounnas et al. [18]. In this study, a potent non-acidic GSM was coupled to a solid support and used for affinity chromatography. The affinity ligand retained predominantly the c-secretase subunit PEN-2 and smaller amounts of PSEN1-NTF and PSEN1-CTF from cellular lysates [18]. However, this was only observed when cells were solubilized in 1% Triton X-100 or b-D-maltoside, detergent conditions that do not support c-secretase activity, whereas solubilization in CHAPSO did not result in PEN-2 retention. Furthermore, it was not investigated whether the affinity ligand preserved its GSM activity after immobilization [18]. Notably, our results are entirely consistent and complementary to a recent study by Ebke and colleagues, who identified the PSEN1-NTF as the molecular target of a distinct class of GSMs, non-acidic aminopyrimidines [21]. During preparation of this manuscript, we further learned that Ohki and colleagues had identified the PSEN1-NTF as the target of acidic, piperidine GSMs using a photo-probe that is similar but structurally clearly distinct from our probe [33].
The reasons for the divergent results of photo-affinity labeling studies with NSAID-type GSMs and the second-generation piperidine GSMs are not known. However, given the low potency of the parent compounds, the photo-probes based on NSAID-type GSMs were applied at high concentrations (10-100 mM) in the photo-affinity labeling studies, raising concerns for non-specific interactions. Moreover, many GSMs share common structural features with known aggregation inhibitors such as extended conjugated aromatics and anionic moieties that are characteristic of binders to hydrophobic patches. Indeed, some GSMs have been reported as weak Ab aggregation inhibitors and, vice-versa, potent aggregate binders such as X-34 and chrysamine G as moderate GSMs [19]. Noteworthy, the benzophenone cross-linking moiety used in the photo-affinity labeling studies with NSAID-type GSMs also shows strong structural similarity with fenofibrate and, in addition, with potent amyloid binders in the anthraquinone family [34,35]. Hence, the cross-activities in c-secretase modulation and/ or APP-binding might reflect the high hydrophobicity of the putative targets and are not mutually exclusive in view of the available data [36,37]. Clearly, we cannot exclude the possibility that NSAID-type GSMs target a different binding site than piperidine-based GSMs. Ebke et al have performed competition studies with the NSAID GSM sulindac sulfide and observed strongly reduced binding of their non-acidic GSM photo-probe indicative of overlapping binding sites [21]. Competition experiments are generally not able to distinguish between direct and allosteric competition for a specific binding site. Therefore, it remains possible that sulindac sulfide has a specific binding site within the substrate APP and reduces binding of the photo-probe to PSEN through allosteric interaction. However, because of the superior properties of the photo-probes based on secondgeneration GSMs, the overall structural similarities between NSAID GSMs and second generation acidic GSMs, the prior evidence from cell biological studies indicating a binding site of NSAID GSMs within PSEN, and the new evidence from competition experiments, the modulatory activity of GSMs appears to be strongly linked to interaction with PSEN. Future studies aided by tandem mass spectrometry might allow to fully define the binding sequence of GSM photo-probes [38]. In addition, fluorophore-coupled photo-probes that can be crosslinked in live cells might be useful to investigate whether GSMs target specific subcellular pools of c-secretase. Such studies could achieve additional insights into the molecular mechanism of GSMs and Ab42 generation, and might provide a basis for rational chemical design of GSMs.

Dose-response experiments, IC 50 determinations, and cytotoxicity assays
To determine IC 50 values for the Ab42-lowering activity of GSMs used in this study, PS70 cells were cultured in serumcontaining medium and treated for 24 h with 8 increasing concentrations of each GSM or DMSO vehicle in triplicates. Ab40, Ab42 and Ab38 peptide levels in conditioned cell culture supernatants were measured in a previously described sandwich ELISA assay [40]. For calculation of IC 50 values, a non-linear curve-fit with variable slope model was applied to the results.
Cell viability was assessed using the alamarBlue H reagent (Invitrogen). N2a-ANPP cells were seeded at low density in 96-well plates (4000 cells/well) and cultured for 24 h. The cells were then treated in duplicates with increasing concentrations of the respective compounds or DMSO as vehicle for an additional 24 h. 20 ml alamarBlue H was added to cells cultured in 200 ml medium and incubated overnight. Absorbance was measured with a Paradigm TM microplate reader at 570 nm, using 600 nm as the reference wavelength, and percent viability of vehicle control was calculated.

Photo-crosslinking and competition experiments
For photo-affinity labeling in living cells, N2a-ANPP were treated overnight with 0.2 mM of the GSM photo-probe AR243. Subsequently, cells were UV-irradiated on ice for 30 min at 365 nm and for 30 min at 302 nm. As negative controls, UV irradiation was omitted or cells were incubated with DMSO vehicle or compound AR80, which was structurally identical to AR243 but lacked the biotin moiety. Cells were washed and harvested in PBS and incubated with hypotonic buffer (10 mM Tris pH 7.4, 1x protease inhibitor mix) for 10 min on ice. To prepare cellular membranes, cells were passed 10 times through a 30G needle and centrifuged at 8006g, 10 min, 4uC. The postnuclear supernatant was centrifuged at 18.0006g, 45 min, 4uC. The resulting membrane pellet was washed once with MES buffer (50 mM MES, 150 mM NaCl, 5 mM MgCl 2 , 5 mM CaCl 2 , pH 6.0), centrifuged at 18.0006g, 45 min, 4uC, and again dissolved in MES buffer. The total protein concentration of the membrane preparations was determined by BCA assay, concentrations and volumes of each sample were adjusted, and 5x RIPA buffer was added to a final concentration of 1x RIPA buffer (150 mM NaCl, 1.0% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 250 mM Tris HCl, pH 8.0). Solubilization was allowed to continue for 45 min at 4uC with gentle shaking. Insoluble debris was pelleted by centrifugation at 18.0006g, 45 min, 4uC. Protein concentrations were again determined and adjusted to equal levels prior to the addition of streptavidin magnetic beads (Invitrogen). Samples were incubated overnight at 4uC with gentle shaking, and the streptavidin beads were washed 3 times for 30 min with 1x RIPA buffer. Bound material was eluted with Laemmli buffer and incubation at 65uC for 10 min with intermittent vortexing. The eluted material was separated on 12% Bis-Tris gels and analyzed by Western blotting. Chemiluminescence was recorded with a LAS 3000 ECL camera system (Fuji Photo Film GmbH, Duesseldorf, Germany) or on CL-XPosure film (Fisher Scientific GmbH, Schwerte, Germany).
For photo-affinity labeling in vitro, membranes were prepared from N2a-ANPP or HEK293T cells as described above. Prior to the addition of compounds, membrane preparations were incubated with streptavidin magnetic beads to deplete endogenously biotinylated proteins. Membranes were then incubated with 0.5 mM of the GSM photo-probe AR243 or DMSO vehicle for 1 h at 4uC with gentle shaking. In competition experiments, the photo-probe was co-incubated with 100 mM of acidic or nonacidic GSM competitor compounds. 1% CHAPSO was used for solubilization. Samples were diluted to contain 0.25% CHAPSO and UV-irradiated. After the photo-crosslinking, to ensure exposure of the bound biotin-tagged GSM, 1x RIPA buffer was added to unfold/denature proteins. The pull-down and analysis of labeled target proteins using streptavidin magnetic beads was performed as described above. Figure S1 The photo-probe AR366 inversely modulates c-secretase and targets the N-terminal fragment of PSEN1. (A) Chemical structure of AR366, an inverse GSM photo-probe. AR366 was obtained through esterification of the carboxylic acid group in AR243. Note that, in contrast to AR243, the biotin moiety for affinity purification was placed on the alkynyl side chain via an amide bond. AR366 increased Ab42 levels with an IC 50 = 1.1 mM. (B) Incubation of solubilized membranes from N2a-ANPP cells with 500 nM of AR366 resulted in photo-affinity labeling of PSEN1-NTF suggesting that acidic GSMs and inverse GSMs either target the same or a closely related binding site. Input represents 0.02% of the total membrane material. (TIF)