Attenuated Aβ₄₂ Responses to Low Potency γ-Secretase Modulators Can Be Overcome for Many Pathogenic Presenilin Mutants by Second-generation Compounds

Abstract

Sequential processing of the β-amyloid precursor protein by β- and γ-secretase generates the amyloid β-peptide (Aβ), which is widely believed to play a causative role in Alzheimer disease. Selective lowering of the pathogenic 42-amino acid variant of Aβ by γ-secretase modulators (GSMs) is a promising therapeutic strategy. Here we report that mutations in presenilin (PS), the catalytic subunit of γ-secretase, display differential responses to non-steroidal anti-inflammatory drug (NSAID)-type GSMs and more potent second-generation compounds. Although many pathogenic PS mutations resisted lowering of Aβ₄₂ generation by the NSAID sulindac sulfide, the potent NSAID-like second-generation compound GSM-1 was capable of lowering Aβ₄₂ for many but not all mutants. We further found that mutations at homologous positions in PS1 and PS2 can elicit differential Aβ₄₂ responses to GSM-1, suggesting that a positive GSM-1 response depends on the spatial environment in γ-secretase. The aggressive pathogenic PS1 L166P mutation was one of the few pathogenic mutations that resisted GSM-1, and Leu-166 was identified as a critical residue with respect to the Aβ₄₂-lowering response of GSM-1. Finally, we found that GSM-1-responsive and -resistant PS mutants behave very similarly toward other potent second-generation compounds of different structural classes than GSM-1. Taken together, our data show that a positive Aβ₄₂ response for PS mutants depends both on the particular mutation and the GSM used and that attenuated Aβ₄₂ responses to low potency GSMs can be overcome for many PS mutants by second generation GSMs.

Introduction

The amyloid β-peptide (Aβ)⁴ is a 37–43-amino acid secreted peptide and an invariant pathological hallmark of Alzheimer disease (AD). The 42-amino acid variant Aβ₄₂ has been suggested to be causative for the disease by triggering the amyloid cascade, a sequence of pathogenic events that ultimately leads to neurodegeneration and dementia in affected patients (1). The pathogenic peptide is generated by a sequential cleavage of the β-amyloid precursor protein (APP) by β- and γ-secretase (2). After β-secretase cleavage, γ-secretase cleaves the C-terminal fragment of APP that is left in the membrane by an intramembrane cleavage to release the various Aβ species (3–5). Although Aβ₄₂ is normally a minor species produced by this cleavage besides the major Aβ₄₀ species, its production is enhanced by familial AD (FAD) mutations in presenilin (PS) 1 and PS2, the catalytic component of γ-secretase (6), as well as by a subset of FAD mutations in APP. Targeting β- and γ-secretase by specific inhibitors is one of the current approaches toward an effective AD treatment (7). With respect to γ-secretase, however, γ-secretase inhibitors also block the cleavage of Notch1, a major physiological γ-secretase substrate and, thus, the generation of the Notch1 intracellular domain (NICD), which is a crucial signaling molecule controlling cell differentiation (7). Interfering with the cleavage of this substrate accounts for adverse side effects in animal studies (7) and probably in humans as well. γ-Secretase modulators (GSMs) that selectively lower Aβ₄₂, such as a subset of non-steroidal anti-inflammatory drugs (NSAIDs), are considered to be a safer approach than using inhibitors of γ-secretase that target the active site of the protease. These compounds lower Aβ₄₂ generation without affecting Notch1 intracellular domain generation, thus precluding potential Notch-related side effects (8). More potent drugs than NSAIDs, such as the NSAID-like GSM-1 and others effective in the nanomolar range, have been recently identified (9, 10).

The precise mechanistic mode(s) of action as well as the binding site(s) of GSMs, being on the substrate, the enzyme, and/or both, has not yet been fully resolved and may also differ for various types of GSMs (10, 11). Many GSMs typically shift the cleavage specificity of γ-secretase such that the inhibition of Aβ₄₂ generation is accompanied by an increased generation of the shorter Aβ₃₈ (8). For some GSMs, such as the NSAID flurbiprofen, an increase of Aβ₃₇ and even the shorter Αβ₃₃ and Aβ₃₄ species is observed (12). In addition to Aβ₄₂, the longer Aβ₄₃ (e.g. by flurbiprofen) and also the shorter Aβ₃₉ and Aβ₄₀ as well as the rare Aβ₄₁ species can be decreased by GSMs, suggesting a complex mechanism of GSM action (13). Of note, inverse modulators have also been identified that increase the longer Aβ₄₂ species while lowering the shorter Aβ₃₈ species (14). GSMs were reported to target the substrate by binding to a short region of the Aβ domain located in the N-terminal half of the APP transmembrane domain (TMD) (11). More recently, the binding of GSMs to Aβ has been reported to cause reduced dimerization of the APP TMD thereby lowering Aβ₄₂ generation (15, 16). However, controversial data challenging the view of substrate-targeting GSMs were also reported (10, 17) and supported a number of studies that had initially suggested that GSMs target the enzyme (12, 18–21). Mutational analyses of APP do not seem to support substrate binding of GSMs either (13). Point mutations disrupting key residues within the proposed GSM binding site in APP did not abolish the efficacy of GSMs (13). Likewise, both familial and synthetic mutations at the γ-secretase cleavage site domain showed the typical GSM responses (13). In contrast, although FAD mutations in PS still increase Aβ₃₈ in response to NSAIDs, lowering of Aβ₄₂ is not effective for the majority of these mutants (9, 22, 23). This was seen mostly for aggressive mutations with early disease onset that cause a strong increase of Aβ₄₂, such as the PS1 L166P mutant, which was also resistant to the potent compound GSM-1 (9). These data suggested that in particular, aggressive mutations lock γ-secretase in a conformation that renders the enzyme resistant to the Aβ₄₂-lowering activity of GSMs. Here, we report the unexpected finding that Aβ₄₂ can be lowered for many sulindac sulfide-resistant pathogenic and synthetic PS mutations by GSM-1 and other second-generation GSMs. Thus, a positive Aβ₄₂-lowering response not only depends on the respective mutation but also on the particular GSM.

EXPERIMENTAL PROCEDURES

Antibodies

Polyclonal and monoclonal anti-Aβ antibodies (3552) to Aβ1–40 and 2D8 to Aβ1–16 were described previously (24, 25) or obtained from Covance (4G8 to Aβ17–24). The C-terminal-specific anti-Aβ₃₈ antibody was obtained from Meso Scale Discovery, and C-terminal-specific anti-Aβ₄₀ (BAP24) and anti-Aβ₄₂ (BAP15) antibodies were kind gifts of Dr. Manfred Brockhaus (Roche Applied Science).

cDNA Constructs

cDNA constructs encoding synthetic and FAD-associated PS1 and PS2 mutants were generated by PCR-mediated mutagenesis of pcDNA4/HisC::PS1 (26) or pcDNA3.1/zeo(+)::PS2 (9), respectively, using oligonucleotide primers encoding the respective mutations.

Cell Lines, cDNA Transfections, and Cell Culture

HEK293 cells stably expressing Swedish mutant APP (HEK293/sw) were stably transfected with the indicated PS cDNAs using LipofectAmine2000 (Invitrogen) according to the manufacturer's instructions and cultured as described before (27). Pools of stably transfected cells with robust PS expression were investigated to avoid clonal variations (27).

Drug Treatment and Analysis of Aβ

Cultured cells were treated in six-well dishes with GSMs or vehicle (DMSO) at the indicated concentrations as described (9). Secreted Aβ species were analyzed by sandwich immunoassay specific for Aβ₃₈, Aβ₄₀, and Aβ₄₂ species (Meso Scale Discovery), Tris-Bicine urea SDS-PAGE, and MALDI-TOF mass spectrometry as described previously (13). IC₅₀ values of GSMs for inhibition of Aβ₄₂ were determined from dose-response curves by nonlinear regression analysis (sigmoidal dose-response with variable slope) using GraphPad Prism software.

Compound Synthesis

The TorreyPines compound N¹,N¹-diethyl-4-methyl-N³-{4-[4-(4-methyl-imidazol-1-yl)-phenyl]-thiazol-2-yl}-benzene-1,3-diamine was prepared by coupling of (5-diethylamino-2-methyl-phenyl)-thiourea with 2-bromo-1-[4-(4-methyl-imidazol-1-yl)-phenyl]-ethanone acetate in N,N-dimethylformamide according to patent WO2004110350, general procedure Example 3. Synthesis of the Eisai compound (4R,9aS,E)-7-(3-methoxy-4-(4-methyl-1H-imidazol-1-yl)benzylidene)-4-(3,4,5-trifluorophenyl)hexahydropyrido[2,1-c][1,4]oxazin-6(1H)-one was accomplished as previously outlined in the patent applications WO2007060821 and WO2009081959.

RESULTS

Effective Lowering of Aβ₄₂ Generated by the Aggressive PS1 G384A FAD Mutant with GSM-1

The NSAID-like compound GSM-1 is a second-generation GSM capable of reducing Aβ₄₂ generation with considerably higher potency (IC₅₀ = 180 nm in HEK293/sw cells) than typical GSMs of the NSAID class such as sulindac sulfide (IC₅₀ = 70 μm in HEK293/sw cells). Similar to NSAIDs but more potently than these, GSM-1 can also lower Aβ₄₂ generated by pathogenic mutant APP substrates (13). Unlike the APP mutants, pathogenic mutants in PS appear to be rather refractory to sulindac sulfide or other classical Aβ₄₂-lowering NSAIDs. Two exceptionally strong PS mutants, PS1 L166P and PS2 N141I, have been tested so far with GSM-1 and also do not respond to this highly potent compound (9). This has led to the concept that many pathogenic PS mutations, in particular those that generate high amounts of Aβ₄₂, lock γ-secretase in a GSM-resistant conformation such that the aberrant generation of Aβ₄₂ is maintained (9, 22, 23). Surprisingly, when we now analyzed the PS1 G384A mutant, which causes a similarly strong increase of the Aβ₄₂ to total Aβ ratio (total Aβ = Aβ₃₈ + Aβ₄₀ + Aβ₄₂) as the PS1 L166P mutant as assessed by a highly sensitive and specific Aβ immunoassay (Fig. 1A), we found that GSM-1 strongly reduced the Aβ₄₂ generation of this mutant in cultured cells (Fig. 1B). Consistent with previous results (9, 22, 23), the NSAID sulindac sulfide was only effective in lowering Aβ₄₂ for WT PS1 but not for the two mutants (Fig. 1B). Despite their differential Aβ₄₂ response, Aβ₃₈ was increased for both mutants by both drugs, demonstrating the modulatory activity of these compounds (Fig. 1C), which was substantially stronger for GSM-1 than for sulindac sulfide. Interestingly, for the PS1 G384A mutant, Aβ₄₀ was also robustly lowered by GSM-1 and somewhat by sulindac sulfide, whereas this Aβ species was much less affected by the drugs for WT PS1 or the PS1 L166P mutant (Fig. 1D). The unexpected behavior of the G384A mutant in response to GSM-1 was fully confirmed by a mass spectrometry analysis of the individual Aβ species (Fig. 1E), which further showed that Aβ₃₈ was the major Aβ species generated by this mutant in response to GSM-1.

FIGURE 1.

Pathogenic Aβ₄₂ generation of PS1 G384A can be lowered by GSM-1. A, conditioned media of HEK293/sw cells stably expressing WT PS1, PS1 L166P, or PS1 G384A were analyzed for levels of secreted Aβ₃₈, Aβ₄₀, and Aβ₄₂ species by sandwich immunoassay. Each species is plotted as a percentage of the total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) measured for each cell line. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the Aβ₄₂/Aβ_total ratio changes relative to WT PS1 (***, p < 0.001). B–D, conditioned media of HEK293/sw cells stably expressing WT PS1, PS1 L166P, or PS1 G384A treated with sulindac sulfide (50 μm), GSM-1 (1 μm), or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₄₂, Aβ₃₈, and Aβ₄₀. Changes in the ratios of Aβ₄₂ (B), Aβ₃₈ (C), and Aβ₄₀ (D) species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) are shown as a percentage of those obtained for vehicle-treated cells, which were set to 100%. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). E, 4G8-immunoprecipitated total Aβ species from conditioned media of HEK293/sw cells stably expressing WT PS1, PS1 L166P, or PS1 G384A treated with GSM-1 (1 μm) or vehicle control (DMSO) were subjected to MALDI-TOF mass spectrometry analysis. The intensities of the highest peaks were set to 100% in the spectra.

Taken together, the behavior of the PS1 G384A mutant demonstrates that a pathogenic PS mutant can be unresponsive to a classical NSAID-type GSM while still being responsive to the Aβ₄₂-lowering activity of another GSM such as GSM-1. We further conclude that the capability of GSM-1 to lower Aβ₄₂ can be differentially affected by PS FAD mutations of similar pathogenic strength and that the lack of an Aβ₄₂-lowering response to GSMs may not be generally observed for very strong PS FAD mutants.

Mutants of the Active Site Domain of γ-Secretase Respond Differentially to GSMs

Because the PS1 G384A mutant is distinguished from other FAD mutants by its immediate location directly to the active site aspartate 385 and may thus behave differently from other FAD mutants, we next investigated whether mutants of the γ-secretase active site domain might be generally permissive to this particular GSM. The active site domain of γ-secretase is characterized by conserved YD and GLGD motifs in TMD6 and TMD7 of PS, respectively, that include the catalytic aspartate residues and a PAL motif at the N-terminal end of TMD9. Besides the N-terminal aspartate in TMD6, the latter two sequence motifs are highly conserved in PS and related aspartyl proteases as GXGD and PXL consensus sequences (28).

To investigate whether and how active site domain mutants in PS1 respond to GSM-1, known familial and synthetic mutants of the above sequence motifs were selected. For the YD motif, the Y256S FAD mutant was chosen for analysis. This mutant is located directly adjacent to the TMD6 active site aspartate, similar to the G384A mutant in the immediate vicinity to the TMD7 aspartate. In addition, this mutation also generates very high, comparable amounts of Aβ₄₂ (9). For the GLGD motif, besides the G384A mutant described above, the G382A mutant was chosen. Both mutants are the only known substitutions of the glycine residue, which are functionally tolerated for APP processing (29, 30). Considerably more amino acid substitutions are tolerated for APP processing for the more flexible X position of the motif (Leu-383), but only a few mutants at this position cause an increased production of Aβ₄₂.⁵ From these, the L383W mutant displayed the strongest increase of Aβ₄₂ generation, so this mutant was chosen for analysis. Again, only very few substitutions are functionally tolerated at the PAL motif (31, 32) for which the P433A, A434C, and L435V mutants were selected. As expected, the selected mutants increased the Aβ₄₂/Aβ_total ratio to very different extents (Fig. 2A). As shown in Fig. 2B, all mutants responded substantially less well to the low potency GSM sulindac sulfide compared with WT PS1. Strikingly, although the Y256S mutant caused, like the G384A mutant, one of the strongest increases of Aβ₄₂ among the PS mutations (9) (Fig. 2A), the levels of Aβ₄₂ could be robustly reduced by GSM-1 (Fig. 2B). Both Aβ₄₂ and (as assessed by mass spectrometry and Tris-Bicine urea SDS-PAGE) also Aβ₄₃, which is the preferentially generated long Aβ species of the G382A mutant (30), could be lowered by GSM-1 (Fig. 2, B, D, and E). In contrast, GSM-1 was unable to effectively lower Aβ₄₂ generated by the L383W mutant, whereas all PAL motif mutants responded robustly to GSM-1 (Fig. 2B). The lack of a positive Aβ₄₂ response of the L383W mutant was not due to a general failure of GSM action as sulindac sulfide and GSM-1 were capable of raising Aβ₃₈ levels for this and all other mutants, which for certain mutants (see e.g. PS L435V) was very strong (Fig. 2C). Thus, several but not all active site domain mutants other than the G384A mutant, including another aggressive FAD mutant (Y256S), are also responsive to the Aβ₄₂-lowering capacity of GSM-1. Therefore, we confirm that the responses to GSM-1 are apparently independent of the levels of Aβ42 generated by the individual mutants.

FIGURE 2.

Differential response of active site mutants to GSM-1. A, conditioned media of HEK293/sw cells stably expressing WT PS1 or mutants of the YD, GLGD, and PAL active site motifs of PS1 were analyzed for levels of secreted Aβ₃₈, Aβ₄₀, and Aβ₄₂ species by sandwich immunoassay. Data were plotted as described in Fig. 1A. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the Aβ₄₂/Aβ_total ratio changes relative to WT PS1 (*, p < 0.05; **, p < 0.01; ***, p < 0.001). B and C, conditioned media of HEK293/sw cells stably expressing WT or mutant PS1 described in A that had been treated with sulindac sulfide (50 μm), GSM-1 (1 μm) or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ (B) and Aβ₃₈ species (C) to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1, B and C, bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses for WT PS1 to sulindac sulfide or GSM-1, respectively. D, 4G8-immunoprecipitated total Aβ species from conditioned media of HEK293/sw cells stably expressing PS1 G382A that had been treated with GSM-1 (1 μm) or vehicle control (DMSO) were subjected to qualitative MALDI-TOF mass spectrometry analysis. Note that GSM-1 also causes a reduction of Aβ₄₃ generated by the PS1 G382A mutant. The intensities of the highest peaks were set to 100% in the spectra. In B and C, solid and dashed horizontal lines denote the levels of the responses for WT PS1 to sulindac sulfide or GSM-1, respectively. E, total Aβ species from conditioned media of HEK293/sw cells stably expressing WT PS1 or PS1 G382A that had been treated with GSM-1 (1 μm) or vehicle control (DMSO) were immunoprecipitated using antibody 3552, separated by Tris-Bicine urea SDS-PAGE, and analyzed by immunoblotting using antibody 2D8.

The Attenuated Response to Low Potency Aβ₄₂-lowering GSMs Observed for Many PS Mutants Can Be Overcome by GSM-1

The above results show that all but one (PS1 L383W) of the analyzed PS mutations of the active site domain are permissive to GSM-1. To investigate whether the response to this potent compound is a more common behavior of PS mutants, we analyzed additional FAD mutants of different pathogenic strength within each individual TMD of PS1. For this study PS1 ΔIM83/84 (TMD1), N135D, N135S, and M146L (all TMD2), S170F and L173W (both TMD3), G206V (TMD4), M233V (TMD5), A246E (TMD6), G378E, L381V, and F386S (all TMD7), L424R (TMD8), and P436Q (TMD9) were analyzed. In addition, we also analyzed additional mutations located in hydrophilic loop (HL) regions of PS1 such as T116N, P117L (HL1), A285V, L286V, and Δexon9 (HL6). Fig. 3A shows the relative increases of the Aβ₄₂/Aβ_total ratios measured for the mutants selected. The strongest increase was observed for the M233V mutant, whereas the A246E mutant caused the weakest increase of the Aβ₄₂/Aβ_total ratio. With respect to their response to Aβ₄₂ reduction by GSM-1, all mutants responded to the compound (Fig. 3B), although to different extents. Strikingly, aggressive mutants with extreme increases of Aβ₄₂ generation such as P117L or M233V (Fig. 3A) were again clearly responsive to GSM-1. Five mutants (T116N, M146L, M233V, A246E, A285V, L286V) showed a similar response to WT PS1, whereas several mutants showed a ∼3–4-fold weaker response than WT PS1. The N135S mutant showed the weakest response to GSM-1 (∼5-fold less than WT PS1). Its response was similar to that which was typically reached for WT PS1 with sulindac sulfide. In agreement with our previous results (9), many mutants were clearly attenuated or even resistant to the Aβ₄₂-lowering activity of sulindac sulfide. With respect to modulation of Aβ₃₈ levels, all mutants responded to GSM-1, increasing the levels of this short Aβ species at least ∼3–4-fold as for WT PS1. For some mutants, exceptionally high (e.g. ∼27-fold for the L424R mutant) Aβ₃₈ increases were observed (Fig. 3C). Sulindac sulfide showed a very similar but less pronounced behavior to GSM-1 on the generation of the short Aβ species, consistent with being a less potent GSM (Fig. 3C).

FIGURE 3.

Many sulindac sulfide-resistant PS1 FAD mutants respond to GSM-1. A, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated PS1 FAD mutants (ΔIM = ΔI83/M84, ΔE9 = ΔExon9) were analyzed for levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂ species by sandwich immunoassay. Data were plotted as described in Fig. 1A. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance of the Aβ₄₂/Aβ_total ratio change (two-tailed unpaired Student's t test) relative to WT PS1 (*, p < 0.05; **, p < 0.01; ***, p < 0.001). B and C, conditioned media of HEK293/sw cells stably expressing WT PS1 or mutant PS1 described in A that had been treated with sulindac sulfide (50 μm), GSM-1 (1 μm), or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ (B) and Aβ₃₈ (C) species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1, B and C. Bars represent the mean of three to six experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses for WT PS1 to sulindac sulfide or GSM-1, respectively. In A–C, numbers of the TMDs and hydrophilic loops (HLs) are denoted in Roman and Arabic letters, respectively.

Inspection of the GSM response profiles of the various mutants compared with WT showed that these were very similar for a number of mutants. This suggests that the attenuated responses of these mutants can be overcome with more potent drugs such as GSM-1. Additional dose-response experiments showed that the Aβ₄₂ generation of several aggressive mutants that responded well to GSM-1 (PS1 P117L, M233V, Y256S, and G384A) could also be further lowered by doubling the dose of sulindac sulfide (100 μm; i.e. ∼1.5 times the IC₅₀) (Fig. 4A). This increase of drug dose was, however, still not effective for mutants with attenuated Aβ₄₂ response to GSM-1 (PS1 ΔIM and ΔE9)- or GSM-1-resistant mutants (PS1 N135I, a synthetic PS1 homolog of the GSM-1-resistant PS2 N141I mutant (see below), L166P, and L383W). As expected, the PS1 M146L mutant showed a dose-dependent and enhanced Aβ₄₂ response compared with WT PS1 in agreement with previous results (18). Dose-response experiments with GSM-1 using a selection of the above mutants tested showed that Aβ₄₂ could be further lowered in a dose-dependent manner for PS1 G384A or PS1 ΔE9, whereas the GSM-1-resistant mutants remained rather unresponsive even at a 10-fold higher concentration of GSM-1 (10 μm; i.e. ∼55 times the IC₅₀) (Fig. 4B). Taken together, these data suggest a similar mode of action of these compounds and indicate that the differences in their potency may relate to differences in the affinity to their target binding site, which is individually affected by the mutations.

FIGURE 4.

Differential dose responses to GSMs of PS mutants for Aβ₄₂ generation. A, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated mutant PS1 variants that had been treated with either 50 or 100 μm sulindac sulfide or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1B. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses of WT PS1 for 50 and 100 μm sulindac sulfide, respectively. B, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated mutant PS1 variants that had been treated with either 0.1, 1, or 10 μm GSM-1 or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1B. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses of WT PS1 for the respective GSM-1 concentrations.

Homologous PS Mutations Can Respond Differently to GSMs

We next asked whether and how PS2 mutations would respond to GSM-1 and how their response would compare with corresponding mutants in PS1. From the very few FAD mutations that have been identified in PS2, we chose the “Italian” M239V (TMD5) mutant for analysis besides the GSM-1-resistant “Volga-German” N141I (TMD3) (9) and its synthetic GSM-1-resistant PS1 N135I homolog (see Fig. 4B). The Italian mutant was chosen as it represents the homolog of the GSM-1-responsive PS1 M233V mutant analyzed above (see Fig. 3), thus, allowing a direct comparison of the GSM response of these two PS variants. To further compare PS1 and PS2 mutants with each other, we additionally generated a direct homolog of the aggressive and GSM-1 resistant PS1 L166P mutant (PS2 L172P) and a direct PS2 homolog of the similarly aggressive but GSM-1-responsive PS1 G384A mutant (PS2 G365A). As shown in Fig. 5A, all mutants increased the Aβ₄₂/Aβ_total ratio, which in several cases were similar to those of the aggressive PS1 L166P mutant. Except for the PS1 N135I, PS1 L166P, and M233V mutants, which displayed stronger increases of Aβ₄₂ generation than their PS2 homologs, the other mutant pair (PS1 G384A/PS2 G365A) was comparable in its respective pathogenic activities.

FIGURE 5.

Homologous PS mutants can elicit distinct Aβ₄₂ response to GSM-1. A, conditioned media of HEK293/sw cells stably expressing WT PS1, WT PS2, or the indicated mutants were analyzed for levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂ species by sandwich immunoassay. Data were plotted as described in Fig. 1A. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the Aβ₄₂/Aβ_total ratio changes relative to WT PS1 (***, p < 0.001). B and C, conditioned media of HEK293/sw cells stably expressing WT PS1, WT PS2. or the indicated PS variants that had been treated with sulindac sulfide (50 μm), GSM-1 (1 μm), or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ (B) and Aβ₃₈ (C) species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1, B and C. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses for WT PS1 to sulindac sulfide or GSM-1, respectively.

Compared with WT PS2, the PS2 mutants were clearly attenuated (PS2 L172P, PS2 M239V, PS2 G365A) or, in agreement with our previous results (9), resistant (PS2 N141I) with respect to the inhibition of Aβ₄₂ generation by GSM-1 (Fig. 5B). The substantially attenuated response of the PS2 M239V mutant to the Aβ₄₂-lowering capacity of GSM-1 was surprising as Aβ₄₂ could be very effectively lowered for its PS1 homolog (PS1 M233V), consistent with our results above (Fig. 3B). Likewise, with its substantially attenuated Aβ₄₂ response to GSM-1, the PS2 G365A mutant behaved strikingly differently than its PS1 G384A homolog. Thus, the PS1 M233V/PS2 M239V and PS1 G384A/PS2 G365A mutant pairs behaved differently from the GSM-1-resistant PS1 N135I/PS2 N141I and PS1 L166P/PS2 L172P mutant pairs (Fig. 5B).

As expected, several PS1 and PS2 mutants showed attenuated Aβ₄₂ responses to the less potent drug sulindac sulfide. Interestingly, although both WT PSs were similarly responsive to sulindac sulfide, WT PS2 was considerably less responsive to GSM-1 than WT PS1. Moreover, the analyzed PS2 mutants showed attenuated responses to sulindac sulfide compared with WT PS2. Similar sulindac sulfide responses were observed for the PS1 G384A mutant and its PS2 homolog, whereas their GSM-1 responses were strikingly different (Fig. 5B). Finally, as expected, all mutants elicited robust increases of Aβ₃₈ generation in response to the used GSMs (Fig. 5C).

Taken together, some but not all homologous mutations in PS1 and PS2 show comparable Aβ₄₂ responses to GSM-1. This suggests that the respective spatial environment, which might be subtly different between homologous PS mutations, is an important determinant as to whether a given mutant PS will show an effective Aβ₄₂ response to GSM-1.

The Side Chain of a Particular Residue Individually and Differentially Affects Aβ₄₂ Changes in Response to GSM-1

The above data suggest that the PS1 L166P mutant belongs to a group of few and rather exceptional mutants with respect to their resistance to the highly potent GSM-1. They further suggest, considering the differential responses of the PS1 N135D/S/I mutants (Figs. 3 and 4) as well as the differential behavior of a subset of homologous PS1 and PS2 mutations (Fig. 4), that the nature of the amino acid exchange and/or its structural context within PS can have a substantial influence as to whether GSM-1 will be effective as an Aβ₄₂-lowering drug or not. However, drastic amino acid exchanges per se such as leucine to proline (P117L, L166P) or others, do not necessarily block the GSM-1 response (Figs. 1 and 3).

To further explore the influence of the amino acid side chain on the efficacy of GSM-1, we compared the L166P mutation with a number of additional well characterized amino acid substitutions of this residue (27) including a conservative mutant (L166V). Consistent with our previous results (27), the individual mutations increased the Aβ₄₂/Aβ_total ratios to various extents (Fig. 6A). Strikingly, with respect to their Aβ₄₂ response, many of the PS1 Leu-166 mutants were resistant to GSM-1 (Fig. 6B). The L166C mutant showed some response to GSM-1, which, however, was still ∼4-fold weaker than for WT PS1. Only the conservative L166V mutant responded effectively to GSM-1 and showed an Aβ₄₂ inhibition comparable with WT PS1. As expected, the mutants with poor GSM-1 response also showed a strongly attenuated Aβ₄₂ response to sulindac sulfide (Fig. 6B). The conservative L166V mutant displayed an Aβ₄₂ response to sulindac sulfide comparable with WT PS1. Despite their poor Aβ₄₂ response, all mutants increased Aβ₃₈ in response to GSM-1 and sulindac sulfide (Fig. 6C). The poor response of the L166W mutant that showed only a mild increase of Aβ₄₂ generation (Fig. 6A) further confirmed that the response to GSM-1 is independent of the Aβ₄₂ levels generated by the particular mutant. These data indicate that Leu-166 of PS1 is a critical residue for maintaining an enzyme and/or substrate conformation that supports an Aβ₄₂-lowering shift in γ-secretase cleavage specificity in response to GSM-1.

FIGURE 6.

The Leu-166 residue of PS1 is critical for a positive Aβ₄₂ response to GSMs. A, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated PS1 Leu-166 variants were analyzed for levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂ species by sandwich immunoassay. Data were plotted as described in Fig. 1A. Bars represent the mean of three experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the Aβ₄₂/Aβ_total ratio changes relative to WT PS1 (**, p < 0.01; ***, p < 0.001). B and C, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated PS1 Leu-166 variants and treated with sulindac sulfide (50 μm), GSM-1 (1 μm), or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ (B) and Aβ₃₈ (C) species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1, B and C. Bars represent the mean of three to six experiments with error bars indicating the S.E. Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses for WT PS1 to sulindac sulfide or GSM-1, respectively.

GSMs of Different Structural Classes Act Similarly as GSM-1

Finally, we asked whether GSMs lacking any structural relationship to NSAID-type GSMs would be effective on aggressive PS mutations. To address this question, we analyzed the Aβ₄₂ response of several mutants, which showed a good (P117L, M146L, M233V, Y256S, G384A) or moderate response (ΔE9) to GSM-1 or which were resistant to this compound (N135I, L166P, and L383W) (see Fig. 4) to other potent second generation GSMs developed by Eisai or TorreyPines Therapeutics (10). These compounds are structurally distinct from sulindac sulfide and GSM-1 and also lack the carboxylic acid that is common for these NSAID and NSAID-like GSMs (Fig. 7A). Determination of their IC₅₀ values in dose-response experiments revealed that they are ∼4-fold more potent than GSM-1 with respect to inhibition of Aβ₄₂ generation in HEK293/sw cells (Fig. 7A). To evaluate their activity on the PS mutants selected, the compounds were used at 1 μm, like GSM-1 in the above experiments. As shown in Fig. 7B, the Eisai compound was also not able to lower Aβ₄₂ for the GSM-1-resistant PS1 N135I, L166P, and L383W mutants, whereas the other mutants responded effectively to this compound. Very similar results were obtained for the TorreyPines GSM (Fig. 7B). As expected for their modulatory activity, the compounds increased the generation of Aβ₃₈ (Fig. 7C). Thus, the analyzed mutants showed very similar Aβ₄₂ responses to second generation GSMs structurally different from GSM-1. We conclude that potent second generation GSMs of distinct structural classes are suitable to lower Aβ₄₂ for a number of PS mutants.

FIGURE 7.

Potent second-generation GSMs of different structural class act similar to GSM-1. A, structures and dose-response curves of the acidic GSMs sulindac sulfide and GSM-1 and the non-acidic GSMs from Eisai and TorreyPines Therapeutics are shown. To determine GSM potencies, conditioned media of HEK293/sw cells treated with vehicle (DMSO) or increasing concentrations of GSM were analyzed for Aβ₃₈, Aβ₄₀, and Aβ₄₂ levels by sandwich immunoassay. Data are shown as percentage of the mean value of the vehicle control, which was set 100%. Error bars indicate the S.E. of the mean of three independent experiments. IC₅₀ values for Aβ₄₂ inhibition were determined by non-linear regression analysis (sigmoidal dose-response with variable slope). B and C, conditioned media of HEK293/sw cells stably expressing WT PS1 or the indicated mutant PS1 variants that had been treated with 1 μm of either Eisai or TorreyPines compound or vehicle control (DMSO) were analyzed by sandwich immunoassay for the levels of Aβ₃₈, Aβ₄₀, and Aβ₄₂. Changes in the ratios of Aβ₄₂ (B) and Aβ₃₈ (C) species to total Aβ (i.e. the sum of Aβ₃₈, Aβ₄₀, and Aβ₄₂) were plotted as described in Fig. 1, B and C. Bars represent the mean of three experiments with error bars indicating the S.E. (#, below detection level). Asterisks indicate the significance (two-tailed unpaired Student's t test) of the GSM responses relative to vehicle control (**, p < 0.01; ***, p < 0.001). Solid and dashed horizontal lines denote the levels of the responses for WT PS1 to the Eisai or the TorreyPines compound, respectively.

DISCUSSION

Many FAD mutations in PS, in particular aggressive ones, are known to resist low potency Aβ₄₂-lowering drugs such as sulindac sulfide and others (9, 22, 23). Two previously analyzed mutants, PS1 L166P and PS2 N141I, were also shown to be resistant to the more potent second-generation compound GSM-1 (9). Together, these data have led to the concept that Aβ₄₂ generation may not be inhibitable by GSMs for strong FAD mutations. Unexpectedly, we now found that several but not all aggressive mutations in PS can be effectively modulated by GSM-1. The Aβ₄₂-responsive mutants identified were widespread over the entire PS molecule. Although their Aβ₄₂ responses to GSM-1 varied in their strength and rather depended on the particular mutation, they were apparently independent of the levels of Aβ₄₂ they generated. Besides the known GSM-1-resistant PS1 L166P and PS2 N141I mutations, only a few other PS mutations were identified to confer resistance to GSM-1. Interestingly, homologous PS1 and PS2 mutations could display differential responses, suggesting that a positive Aβ₄₂ response depends on the particular mutation and its spatial environment in PS.

For many mutants, the Aβ₄₂ response profiles for sulindac sulfide and GSM-1 were similar, suggesting that these modulators act also mechanistically very similar. Accordingly, with respect to drug action, GSM-1 may have a higher affinity to the GSM binding site than for example the low potency drug sulindac sulfide, thus being more potent than this compound. PS mutations may weaken the binding affinity of a low potency GSM but not of a high potency GSM, thus differentially affecting the modulatory capacity of certain GSMs on Aβ₄₂ generation. In that regard, the PS1 L166P mutation is apparently rather an exception as it neither responded to GSM-1 nor to other potent GSMs of different structural class (see below). Moreover, the PS1 Leu-166 residue appears to be critical for GSM-1 action, as a number of the analyzed additional mutations were either not, or at best only moderately, responsive to GSM-1. Only the conservative L166V mutant showed a comparable Aβ₄₂ response to WT PS1.

Despite differences in the response of the PS mutants with respect to the capacity to lower Aβ₄₂, sulindac sulfide and GSM-1 elicited in all cases an increase of Aβ₃₈ generation. Consistent with being the more potent drug, the increases of Aβ₃₈ generation induced by GSM-1 were also much stronger than that of sulindac sulfide. The question of how it is possible that for a given mutation Aβ₄₂ cannot be lowered by a certain GSM, whereas Aβ₃₈ can still be increased (9, 23), is also not readily explainable by our present study. This question can probably only be resolved when suitable cross-linkable second-generation GSM reagents become available to identify their binding site(s), which once known may shed light on these mechanistic issues. In the absence of such reagents, mutational analyses are reasonable experimental tools to probe GSM binding sites. Unfortunately, no mutation has so far been identified that blocks the modulatory capacity of GSMs on all Aβ species, which would be indicative that a common GSM binding site was destroyed. Although our data can, thus, also not fully clarify whether GSMs target the substrate, the protease, and/or both, the differential responses of GSMs to PS mutations may be more compatible with a GSM binding site(s) in γ-secretase, probably within the catalytic subunit PS itself. With respect to this view, we found that, interestingly, WT PS2 displayed attenuated Aβ₄₂ responses to GSM-1 as compared with WT PS1. In addition, all analyzed PS2 mutants were resistant to GSM-1. These differences in GSM response between the two PSs may also be easier to reconcile with an enzyme-located GSM binding site.

The Aβ₄₂ responses of several PS mutants analyzed that were positive for GSM-1 were also positive for other potent second-generation compounds from Eisai and TorreyPines Therapeutics (10). Likewise the GSM-1-resistant mutants such as PS1 L166P were also resistant to the Eisai and the TorreyPines compounds. This may be surprising as these compounds are structurally different from GSM-1, indicating that the GSM binding site is rather flexible and can thus also accommodate, probably by an induced-fit mechanism, structurally distinct GSMs with high affinity. Another interesting feature of GSM-1-responsive mutants was that besides Aβ₄₂, Aβ₄₀ could also be inhibited in several cases. This was very prominent for the PS1 G384A mutant, for which Aβ₃₈ became the principal species in response to GSM-1, and also for the Eisai and TorreyPines compounds (Table 1). The reduction of Aβ₄₀ generation is typical for GSMs of these potent compound classes (10, 33).

TABLE 1.

Summary of PS mutants for which inhibition of Aβ₄₀ generation was observed and their Aβ₄₀ responses to GSMs

The relative levels of the Aβ₄₀/Aβ_total ratios as determined by Aβ immunoassay in response to GSM treatment compared to vehicle control (set to 100%) are indicated (+++, 0–25%; ++, >25–50%; +, >50–75%; −, >75%).

PS1 Mutant	Aβ40 response to GSMs
	50 μm Sulindac sulfide	1 μm GSM-1	1 μm Eisai	1 μm TorreyPines
Wild type	–	–	++	++
ΔIM83/84	–	++
P117L	–	–	–	+
N135I	–	–	+	++
M146L	–	–	++	++
L166P	–	–	+	++
L166V	–	+
L166W	–	+
M233V	–	+	+	+
Y256S	–	+++	+++	+++
ΔExon9	–	+	–	++
L381V	–	++
G382A	–	++
L383W	–	–	–	–
G384A	–	+++	+++	++
P433A	–	++
A434C	–	+
P436Q	–	+

In conclusion, our data show that the resistance of many PS mutations to low potency drugs can be overcome by GSMs with higher potencies such as GSM-1 (see Table 2 for a summary). Potent GSMs that may become available for AD therapy could therefore also be useful to treat patients suffering from FAD caused by PS mutations when a careful preclinical analysis of their suitability with respect to the particular mutation carried by the patient is undertaken. In that respect, our data reinforce the notion that not all transgenic AD mouse models are suitable for preclinical studies with GSMs and thus have to be carefully selected. Our study indicates that their suitability depends both on the particular GSM to be tested and the particular PS mutant expressed as transgene. Thus, although PS1 L166P and PS2 N141I mouse models (34, 35) may not be suitable for GSM testing in vivo (9), other AD mouse models such as the APP23/PS45 (swAPP/PS1 G384A) mouse with similarly fast Aβ deposition (36) are expected to be better suited for the in vivo analysis of GSMs in preclinical settings.

TABLE 2.

Summary of PS mutations and their Aβ₄₂ responses to GSMs

The relative change in the Aβ₄₂/Aβ_total ratio as determined by Aβ immunoassay for each mutant compared to wild type is indicated by arrows (↑ = 1.5-3-fold; ⇈, >3-4.5-fold; ⇈↑, >4.5–6-fold; ⇈⇈, >6-fold increase of the Aβ₄₂/Aβ_total ratio). Double horizontal arrow, no change. Additionally, the relative levels of the Aβ₄₂/Aβ_total ratios in response to GSM treatment compared to vehicle control are indicated (+++ = 0–25%; ++, >25–50%; +, 50–75%; −, >75%). For FAD mutations, the average age of onset is also indicated. N.D., not determined.

* Data are from the Alzheimer Disease and Frontotemporal Dementia Mutation Database.