Substrate determinants in the C99 juxtamembrane domains differentially affect γ–secretase cleavage specificity and modulator pharmacology


Address correspondence and reprint requests to Dirk Beher, Asceneuron SA, PSE-B EPFL, CH-1015 Lausanne, Switzerland. E-mails:;


The molecular mechanisms governing γ-secretase cleavage specificity are not fully understood. Herein, we demonstrate that extending the transmembrane domain of the amyloid precursor protein-derived C99 substrate in proximity to the cytosolic face strongly influences γ–secretase cleavage specificity. Sequential insertion of leucines or replacement of membrane-anchoring lysines by leucines elevated the production of Aβ42, whilst lowering production of Aβ40. A single insertion or replacement was sufficient to produce this phenotype, suggesting that the helical length distal to the ε–site is a critical determinant of γ-secretase cleavage specificity. Replacing the lysine at the luminal membrane border (K28) with glutamic acid (K28E) increased Aβ37 and reduced Aβ42 production. Maintaining a positive charge with an arginine replacement, however, did not alter cleavage specificity. Using two potent and structurally distinct γ–secretase modulators (GSMs), we elucidated the contribution of K28 to the modulatory mechanism. Surprisingly, whilst lowering the potency of the non-steroidal anti-inflammatory drug-type GSM, the K28E mutation converted a heteroaryl-type GSM to an inverse GSM. This result implies the proximal lysine is critical for the GSM mechanism and pharmacology. This region is likely a major determinant for substrate binding and we speculate that modulation of substrate binding is the fundamental mechanism by which GSMs exert their action.

Abbreviations used

Alzheimer's disease


anterior pharynx defective 1


amyloid precursor protein

amyloid-β peptide


C-terminal fragment




γ-secretase inhibitor


γ-secretase modulator




human embryonic kidney


immunoprecipitation mass spectrometry




non-steroidal anti-inflammatory drug


presenilin enhancer 2


presenilin 1/2


trifluoroacetic acid


transmembrane domain

γ-Secretase is an intramembrane-cleaving aspartyl protease ultimately linked to Alzheimer's disease (AD). This unusual protease performs the final cleavage step in the processing of the amyloid precursor protein (APP) thereby releasing amyloid-β (Aβ) peptides. Aβ peptides are generally thought to be causative of AD and represent the main constituent of senile plaques (Masters et al. 1985) and vascular amyloid (Glenner and Wong 1984), which are characteristic hallmarks of this neurodegenerative disease. Moreover, the vast majority of human mutations causing early-onset AD (EOAD) are found in the presenilin 1 (PS1) and presenilin 2 (PS2) genes and appear to raise the Aβ42/40 ratio [for review see (Weggen and Beher 2012)]. The identification of two aspartyl residues in the transmembrane domains of PS1 and PS2 as being critical for γ–secretase enzyme activity (Wolfe et al. 1999) and the subsequent demonstration that PS1/2 are the target for specific inhibitors (Esler et al. 2000; Li et al. 2000; Seiffert et al. 2000) of γ-secretase has further implicated this enzyme in the pathogenesis of AD. Genetic studies (Francis et al. 2002; Goutte et al. 2002) capitalizing on the knowledge that γ–secretase activity is required for Notch signaling (De et al. 1999) and the purification of PS1/2 complexes (Yu et al. 2000) provided a profound understanding of the composition of this enzyme complex. Subsequent co-expression studies have suggested that the minimally reconstituted activity requires endoproteolytically cleaved PS1 or PS2 subunits, nicastrin (NCT), presenilin enhancer-2 (PEN-2), and anterior pharynx defective 1 (APH-1) (Edbauer et al. 2003; Kimberly et al. 2003; Takasugi et al. 2003). An interesting twist to these reports was recently provided as it appears that PS1 alone can function as γ–secretase, albeit its proteolytic activity is roughly 10-fold reduced compared with the fully assembled complex (Ahn et al. 2010).

In contrast to the remarkable progress in the identity of γ-secretase, the actual molecular mechanisms underlying the intramembrane cleavage are still poorly understood. γ–Secretase performs at least two major cleavages: γ–cleavage approximately in the middle of the membrane bilayer and ε-cleavage close to the cytosolic membrane face (Fig. 1a). The existence of C–terminally elongated Aβ peptide species such as Aβ48/49 (Qi-Takahara et al. 2005) combined with the appearance of tri- and tetra-peptides covering the sequence starting from the ε- to the γ-site (Takami et al. 2009) favors a sequential cleavage model. Accordingly, the initial cleavage appears to occur at the ε-cleavage site and then progresses on the same face of the helical transmembrane domain (TMD) toward the γ–cleavage site. Early studies investigating cleavage specificity had identified the proximal half of the TMD as a critical determinant as insertions shifted the cleavage accordingly (Murphy et al. 1999; Lichtenthaler et al. 2002). Larger insertions at the distal half either had moderate effects or in the case of complete replacement of the triple lysine cytosolic membrane anchor with aspartic or glutamic acids, completely eliminated γ-cleavage (Murphy et al. 1999). Domain swapping studies suggested a lack of specific sequence requirements within the TMD although the luminal juxtamembrane domain influences binding or recognition by the enzyme (Zhang et al. 2002). Particularly, the juxtamembrane serine 26 and lysine 28 residue (S26 and K28 of the Aβ domain) have been identified as critical residues for γ–cleavage (Ren et al. 2007). In this study, we aimed to extend these findings and account for the recent evidence regarding the relationship between γ– and ε-cleavages. If the sequential trimming model was correct it would be reasonable to assume that more subtle changes around the initiation site such as insertion of helix elongating leucine residues would have an impact on cleavage specificity. Because of its previously identified role in governing enzyme binding (Ren et al. 2007) and processivity (Page et al. 2010; Kukar et al. 2011) we also investigated the luminal juxtamembrane K28 and its impact on the pharmacology of γ–secretase modulators.

Figure 1.

Schematic of the amyloid precursor protein (APP)/C99 amino acid sequence around the transmembrane domain and γ-secretase modulators (GSMs) utilized in this study (a) Schematic of the wild-type (WT) and mutant C99 polypeptides analyzed in this study. The residues encompassing the transmembrane domain are boxed in gray and the luminal and cytosolic membrane interfaces are indicated by dotted lines. Mutated amino acids are underlined and the positions of the lysine residues that have been mutated are indicated by the vertical bars. Arrows indicate the positions of the initiating ε- and corresponding γ–cleavages that generate Aβ40 and Aβ42 peptides. According to the sequential cleavage model, the Aβ40 and Aβ42 product lines are initiated by ε–cleavage at positions 49 and 48, respectively (Takami et al. 2009). (b) Chemical structures of the NSAID-type (GSM-1) and HA-type (E2012) GSMs used in our study.

Materials and methods


The heteroaryl-type GSM E2012 (US2006/0004013) was obtained from Haoyuan Chemexpress Co. (Shanghai, China) and the NSAID-type GSM-1 was synthesized at Syngene (Bangalore, India) according to methods described in WO2006/043064; G2-10 (EMD Millipore, Billerica, MA, USA), 6E10 (Covance, Princeton, NJ, USA), 4G8 (Covance) monoclonal antibodies were obtained from commercial sources and A387 was licensed from Torrey Pines Therapeutics (San Diego, CA, USA).

cDNA cloning and site-directed mutagenesis

For all the studies, a C99 (β-CTF) construct was utilized as described by Lichtenthaler et al. (1999a) that contains a modified signal peptide (SPA4CT DA). Removal of the signal peptide generates C99 identical to the β-CTF generated from APP by BACE1 cleavage. The C99 wild-type [SPA4CT DA (Lichtenthaler et al. 1999a)] construct was generated using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, Agilent Technologies, Inc., Santa Clara, CA, USA) (Primer sequences in Figure S1) according to the manufacturer's instructions. The template was a pcDNA3.1/Hygro(+) (Invitrogen, Carlsbad, CA, USA) vector with the SPA4CT insert originally described by Dyrks et al. (1992) and a Kozak sequence at the start codon. The C99 wild-type construct was used as a template to generate the various C99 mutants K28E, K28R, K/L, KK/LL, KKK/LLL, L-ins, LL-ins, and LLL-ins using either the Gene Tailor Site-Directed Mutagenesis system (Invitrogen) or QuikChange Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer's protocols (Primer sequences in Figure S1). The identity of the coding sequences of all constructs was confirmed by DNA sequencing.

Aβ analysis

Aβ40 and Aβ42 peptides secreted into conditioned media were quantified using the Mesoscale Discovery™ (MSD) electrochemical luminescence (ECL) technology as previously described (Wrigley et al. 2005). Aβ peptides were captured with biotinylated 4G8 monoclonal antibody on avidin-coated MSD plates followed by detection with ruthenlyated G2-10 or A387 monoclonal antibodies specific for the C-termini of Aβ40 and Aβ42, respectively. In addition to detecting Aβ peptides, this assay also detects the minor p3 species generated via the non-amyloidogenic/α-secretase pathway.

Cell lines and transient over-expression studies

Human embryonic kidney 293T/17 (HEK293T/17) cells (American Type Culture Collection) are a subclone derived from the original HEK293T cell line and were cultured in Dulbecco's modified Eagle medium with l-glutamine supplemented with 10% heat inactivated fetal bovine serum, 100 units/mL penicillin, and streptomycin, respectively. HEK293T/17 cells were transiently transfected with wild-type or various mutant C99 constructs using Fugene HD transfection reagent (Roche Applied Science, Penzberg, Germany) according to the manufacturer's instructions. Twenty-four hours post-transfection, media were removed and fresh media added. Forty-eight hours post-transfection, media were collected for Aβ analysis and the corresponding cells lysed in 50 mM Tris HCl pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, 1% Triton X-100, 0.2% sodium dodecyl sulfate, 1x EDTA free protease inhibitor cocktail (Roche Applied Science).

For dose-response studies, cells were plated at 100 000 cells/well/100 uL in 96-well microtiter plates 48 h post-transfection. After two hours of plating, compounds were added with a final dimethyl sulfoxide (DMSO) concentration of 0.5% and Aβ secretion quantified after 16 h incubation. For data analyses, all values were expressed as % vehicle control with the background defined by the signal in the presence of 1 μM of the potent γ-secretase inhibitor XIX (Calbiochem, Darmstadt, Germany). For all compounds shown, cell viability was not affected up to 10 μM as revealed by the Alamar blue cell viability assay.

SDS-PAGE and Western blot analysis

For Western blot analysis, cell lysates were resolved on a 10-20% tricine gel (Invitrogen) under reducing conditions as described by the manufacturer and transferred onto nitrocellulose membranes (Invitrogen). Membranes were probed for C99 with the 6E10 (Covance) monoclonal antibody followed by detection with a goat anti-mouse IRDye 800CW antibody (Li-COR Biosciences Inc., Lincoln, NE, USA) and quantified on a LI-COR Odyssey infrared imager (LI-COR Biosciences Inc.).

Immunoprecipitation mass spectrometry (IP-MS)

HEK293T/17 were transiently transfected with wild-type C99 or various mutant constructs. Twenty-four hours post-transfection, the media were removed and fresh media added the following day and collected for IP-MS after 16 h incubation with the transfected cells. Three microgram 4G8 and 6E10 monoclonal antibody, respectively, and 60 ng isotopically labeled 13C, 15N Aβ40 (Anaspec; mass = 4355.3 Da) or recombinant 15N-labeled (rPeptide; mass = 4382.9 Da) internal standard were added to 1.5 mL media and incubated for 3 h at 22°C followed by capture with anti-IgG coated magnetic beads (Invitrogen) overnight at 4°C. After washing the beads according to Beher et al. (2002), bound peptides were eluted in 4 μL of α-cyano-4-hydroxycinnamic acid matrix solution (50% MeCN, 0.1% TFA) and 2 μL dried onto a target plate. The corresponding spectra were collected on a Voyager-DE PRO Workstation (Applied Biosystems, Carlsbad, CA, USA) matrix-assisted laser desorption/ionization mass spectrometer in linear positive ion mode averaging five times 100 shots for each sample. For semiquantitative data analysis, the individual peak intensities were normalized to the internal calibrant by calculating the ratio of peptide versus calibrant peak intensity. Changes after compound treatment and upon expression of mutant γ-secretase substrates were expressed relative to the DMSO controls and wild-type substrates, respectively. For the spectra shown in Fig. 5b, bound peptides were first eluted into 2 μL of 50% MeCN, 0.1% TFA solution. One microliter eluate was then mixed with 1 μL of α-cyano-4-hydroycinnamic acid matrix solution (50% MeCN, 0.1% TFA) on the target plate and dried.

Statistical methods

The standard deviation (SD) of the data shown in Table 1 was calculated using the build-in function of Microsoft Excel. For the remainder of the data comparing either individual mutants to wild-type controls (Fig. 3) or individual drug treated samples to vehicle controls (Fig. 5) the Student's t-test (two-tailed distribution) function of GraphPad Prism (La Jolla, CA, USA) was used for statistical analysis. Details on number of replicates and independent experiments are given in the respective Figure Legends.

Table 1. Potency of γ-secretase modulators in cells expressing a subset of mutated C99 substrates
C99 constructIC50 Aβ42 (nM) GSM-1IC50 Aβ42 (nM) E2012
  1. HEK293T/17 cells transiently expressing either wild-type (WT) or various mutant C99 polypeptides were plated into 96-well plates (105 cells/well) and treated for 16 h with increasing concentrations of either GSM-1 or E2012. Aβ42 secretion was quantified by the MSD ECL immunoassay. The signal obtained in wells treated with 1 μM of the potent γ–secretase inhibitor XIX (Calbiochem) was subtracted from the signal in the presence of 0.5% DMSO vehicle to define baseline Aβ42 secretion. Data shown are the means from two independent experiments with at least two replicates each ± SD. Note that although the specific ECL signal for Aβ42 detection in the conditioned media of K28E mutant expressing cells was low (average 50–60 counts) the intra- and inter-experimental reproducibility was acceptable (IC50 = 2380 ± 908 nM). Individual dose–response curves for this particular set of experiments are shown in Figure S3.

WT202 ± 4243.6 ± 7.7
K28E2380 ± 908Aβ42 elevation
K28R168 ± 4374.1 ± 23
K/L 235 ± 37 57.7 ± 14
KK/LL174 ± 7154.4 ± 28
L-ins218 ± 2345.5 ± 25
LL-ins231 ± 5872.8 ± 25


Extension of the distal helix portion by sequential lysine replacement or leucine insertion favors Aβ42 generation without affecting the sensitivity for γ-secretase modulation

To address the contribution of the most distal part of the APP TMD to γ-secretase cleavage specificity, we extended the TMD at the cytosolic face of the membrane by sequential replacement of the cytoplasmic triple lysine anchor with leucines (K/L replacement mutants). In addition, to be able to relate our findings to an earlier report (Sato et al. 2009), sequential insertion of leucines N–terminal of the cytoplasmic triple lysine sequence was conducted (L-ins mutants) (Fig. 1a). For these mutagenesis studies, we chose the direct γ-secretase substrate APP C99 which has routinely been used in many studies (Lichtenthaler et al. 1997, 1999b, 2002) as it avoids secondary effects of the mutations on either BACE1 cleavage or APP biogenesis. To analyze the effects of the various mutations in detail, the corresponding expression constructs were transiently transfected into HEK293T/17 cells. Western blot analysis confirmed appreciable expression of all mutant polypeptides (Fig. 2). Quantification of Aβ40 and Aβ42 secretion into the media revealed that overall for the K/L replacement and L-ins mutants a decrease in Aβ40 (≥ 50%) was accompanied by a concomitant increase in Aβ42 (≥ 200%) (Fig. 3). The notable exception was the triple lysine replacement mutant KKK/LLL where a decrease in secreted Aβ40 was not associated with an increase in secreted Aβ42.

Figure 2.

Western blot analysis of C99 expression. 6E10 monoclonal antibody Western blot immunodetection of wild-type (WT) and mutant C99 polypeptides transiently expressed in human embryonic kidney (HEK) 293T/17 cells. Duplicate transfections were individually analyzed to confirm the reproducibility of the transfection method. Note that all polypeptides are expressed at detectable levels. The numbers on the left side of each blot indicate molecular mass standards in kilodalton.

Figure 3.

Proximal and distal juxtamembrane mutations influence the relative production of Aβ40 and Aβ42. Aβ secretion from HEK293T/17 cells transiently expressing various C99 polypeptides was quantified by the MSD electrochemiluminescence (ECL) immunoassay. Differences in substrate expression were corrected by normalization to the wild-type (WT) C99 levels after quantification of the corresponding Western blots of cell lysates by infrared densitometry. Data shown are means ± SEM from two independent experiments (each consisting of two transfections and technical replicates). Statistical analysis was conducted using Student's t-test (two-tailed distribution); **p < 0.005, ***p < 0.0001, n.s. = not significant.

Subsequent immunoprecipitation and mass spectrometric (IP-MS) analyses demonstrated that the reduction in Aβ40 generation was coupled with a concomitant increase in both Aβ42 and especially the shorter Aβ peptides, Aβ38 and Aβ39 (Fig. 4a). This was a common pattern for almost all of the leucine replacements and insertions. No correlation between the Aβ profile and the stepwise increased length of the distal helix was seen beyond the introduction of a single leucine at position 53. This implied that the main structural determinant leading to the altered Aβ phenotype was the introduction of a single leucine. The only exception was the KKK/LLL mutant which preserved a cleavage pattern comparable to the wild-type C99 polypeptide.

Figure 4.

Immunoprecipitation mass spectrometry characterization of peptides produced from the different C99 γ-secretase substrates. (a) Amyloid-β (Aβ) peptides secreted from HEK293T/17 cells transiently expressing various C99 polypeptides were immunoprecipitated and analyzed on a MALDI-TOF mass spectrometer. The pattern of secreted peptides from wild-type (WT) C99 was compared with K28E, L insertion, and K/L exchange mutants. For semiquantitive analysis (Figure S2), 60 ng isotopically labeled 13C, 15N Aβ40 (mass = 4355.3) was included in all media. The assignment of the peptides was based on their masses taking the changes introduced by the point mutations into account. Note that no appreciable levels of shorter peptides were observed except for the K28E mutant (Aβ32, Aβ33, and Aβ34; see Fig. 5b). Note the consistent increase in Aβ42 and Aβ38 for the L-ins, K/L, and KK/LL mutants, whereas the KKK/LLL exchange mutant exhibits an Aβ peptide profile similar to WT C99. (b) Aβ peptides secreted from HEK293T/17 cells transiently expressing WT C99 or K28R polypeptides were immunoprecipitated and analyzed on a MALDI-TOF mass spectrometer. For semiquantitive analysis (Figure S2), 60 ng isotopically labeled recombinant 15N-labeled Aβ40 (mass = 4382.9 Da) was included in all media since the mass of the standard used in Fig. 4a (13C, 15N-labeled Aβ40; mass = 4355.3) was close to the peptide mass of K28R Aβ40 (mass = 4357.9 Da). STD, isotopically labeled Aβ40 peptide standard; *Na-adducts with a mass + 22; **not identified.

Since the mutations at the distal helix portion revealed an increase in Aβ38/39 and Aβ42, we were interested to know whether this part of the substrate affects the sensitivity to γ–secretase modulators (GSMs). For this analysis, we determined the Aβ42 lowering potencies for two potent GSMs in cells expressing wild-type or mutant APP C99 substrates. GSM-1 and E2012 were chosen as representatives of the NSAID-type class and HA-type class of GSM (Fig. 1b), respectively. As evident in Table 1, none of the representative insertion and replacement mutants analyzed showed any change in their sensitivity to either of the two GSMs.

Lysine 28 in the C99 substrate controls γ–secretase cleavage specificity and the pharmacology of GSMs in cells

To address the contribution of lysine 28 (K28) at the luminal juxtamembrane region of the Aβ domain to the γ-secretase cleavage mechanism, we performed site-directed mutagenesis at this position. This residue had previously been identified as being critical for γ-secretase activity (Ren et al. 2007) and cleavage specificity/processivity (Page et al. 2010) (Kukar et al. 2011). However, neither of the previous studies had investigated whether native Aβ production requires a specific side chain or solely a positive charge at position 28. To address this question, K28 was exchanged to an arginine (K28R) and compared with the glutamic acid (K28E) mutant (Fig. 1a). The change from the primary amine to the guanidine functionality in the K28R mutant provides a structurally distinct side chain whilst preserving the positive charge. Both K28E and K28R mutants were transiently expressed at levels comparable to wild-type (WT) C99 in HEK293T/17 cells (Fig. 2). However, compared with the wild-type C99 substrate, the K28E mutant showed a decrease in both Aβ40 and Aβ42 production (Fig. 3). IP-MS profiling revealed a strong increase in Aβ37 (Fig. 4a) which is consistent with a previously published study (Page et al. 2010). On the other hand, maintaining the positive charge at this position with the K28R mutant restored the processing pattern of the wild-type construct as revealed by both the immunoassay (Fig. 3) and IP-MS data (Fig. 4b).

One discussed potential mode of action for GSMs besides targeting the enzyme complex (Beher et al. 2004; Lleo et al. 2004; Uemura et al. 2010; Ebke et al. 2011; Ohki et al. 2011) was substrate binding (Kukar et al. 2008). Since the requirement for a carboxylic acid in NSAID-type GSMs led to the speculation that K28 could be a potential interaction site (Kukar et al. 2008), we analyzed responsiveness of the K28E and K28R mutants to both the NSAID- and HA-type GSMs (Table 1). Potencies for GSM-1 and E2012 remained unchanged with respect to the K28R mutant. In contrast, with the K28E mutant the potency of GSM-1 was reduced approximately 10-fold and for E2012 an unexpected increase in Aβ42 production was observed. To investigate the underlying mechanism, we repeated this experiment using single concentrations of both GSM-1 and E2012 and quantified Aβ40 and Aβ42 production in wild-type and K28E C99 expressing cells (Fig. 5a). Remarkably, both Aβ40 (~twofold) and Aβ42 (~sixfold) production were elevated in K28E C99 expressing cells upon exposure to E2012.

Figure 5.

The K28E point mutation changes the pharmacology of a HA-type γ-secretase modulator (GSM) to an inverse GSM. (a) HEK293T/17 cells transiently expressing either wild-type (WT) or K28E C99 were treated overnight with dimethyl sulfoxide (DMSO) vehicle (0.5%) or 10 μM of either GSM-1 or E2012. The following day Aβ40 and Aβ42 peptides secreted into the media were quantified using the MSD electrochemiluminescence (ECL) immunoassay. Data shown are means ± SEM from two transfections with two technical replicates each. Statistical analysis was conducted using Student's t-test (two-tailed distribution); **p < 0.005, ***p < 0.0001, n.s. = not significant. Note that with the K28E C99 substrate E2012 raises both Aβ40 and Aβ42. (b) HEK293T/17 cells transiently expressing either WT or K28E C99 were treated overnight with DMSO vehicle (0.5%) or 10 μM of either GSM-1 or E2012. The following day Aβ42 peptides secreted into the media were immunoprecipitated and analyzed on a MALDI-TOF mass spectrometer. For reference, 60 ng isotopically labeled 13C, 15N Aβ40 was included in all media as internal standard (STD). The data confirm that GSM-1 still acts as an Aβ38 raising GSM with the K28E substrate, whereas E2012 shows an inverse GSM pharmacology raising Aβ40 and Aβ42.

Subsequent IP-MS analyses revealed differences in the pharmacology of GSM-1 and E2012 (Fig. 5a). In a wild-type C99 background GSM-1 reduced Aβ42 production, whilst increasing Aβ38. On the other hand, E2012 reduced both Aβ40 and Aβ42 while predominantly increasing Aβ37 and to lower extent Aβ38. When K28E C99 expressing cells were exposed to GSM-1, despite the reduced potency observed before (Table 1) this compound still induced the shift of γ-cleavage to Aβ38 (Fig. 5b). In contrast, E2012 acted as a so called ‘inverse’ GSM (Narlawar et al. 2007) on the K28E substrate elevating levels of both Aβ42 and Aβ40, fully consistent with the previous immunoassay data (Fig. 5a). It is noteworthy that the K28E mutant also produces to some extent more of Aβ32, Aβ33, and Aβ34 compared with the wild-type substrate (Fig. 5b). The production of these short Aβ peptide species appears to be responsive to E2012 treatment, whereas the dominant Aβ37 peak remains unchanged.


Despite its pivotal role in the production of Aβ peptides and therefore ultimately in the pathogenesis of AD, the underlying enzymatic mechanism controlling the intramembrane γ–secretase cleavage of the APP-derived C99 substrate is not well understood. It has been proposed that initial γ-secretase cleavage(s) at the ε–site(s) gives rise to two distinct Aβ product lines, wherein subsequent stepwise processing every three to four residues leads to the generation of either Aβ40 / Aβ37 or Aβ42 / Aβ38 peptides. Terms such as ‘processivity’ have subsequently arisen (Quintero-Monzon et al. 2011) to describe this successive trimming model (Takami et al. 2009). As a result of the complexity of γ-secretase processing, the identification of critical regions in the C99 substrate which control the production of the various Aβ peptides is of great importance. Herein, we made the observation that the far distal juxtamembrane region interfacing the membrane-anchoring triple lysine (KKK) sequence contains important elements determining the overall cleavage specificity. The insertion of a single leucine upstream of the triple lysine sequence is sufficient to shift cleavage specificity from the Aβ40 product line to the Aβ42/Aβ38 product line. Since we did not analyze the production of every Aβ peptide in our studies we use the term cleavage specificity to encompass the final Aβ peptide products bearing in mind that any change here could also reflect a change in processivity. Previous studies were mainly focused on the proximal domain N-terminal of the γ-site and it is well documented that its length relative to the γ–cleavage site plays an important role in the cleavage specificity of γ-secretase (Lichtenthaler et al. 1999b; Murphy et al. 1999). In contrast, apart from the analysis of various APP FAD mutations located close to the γ-site, the contribution of the distal part of the APP TMD has not been well studied. It is noteworthy, that either moderate effects or a complete block of Aβ production by large insertions or replacement of the triple lysine membrane anchor with charged sequences such as DDD and EEE have previously been observed (Murphy et al. 1999). Our systematic study revealed a major shift from Aβ40 generation to Aβ42 and Aβ38 production which appeared to be independent of the α-helical periodicity. There was no correlation of the length of Aβ peptides generated with the increased number of insertions and replacements. This is in contrast to the reported effects of proximal insertions (Lichtenthaler et al. 2002). It turns out that a single exchange at the transition of the TMD to the membrane anchor sequence is sufficient to change the cleavage from Aβ40 toward Aβ42 and shorter peptides such as Aβ38 and Aβ39. The coherent phenotype obtained with both the K/L exchange and L-ins mutations rules out that these changes are caused by a distortion of the membrane anchor itself. Our main conclusion is that the γ-secretase enzyme is highly sensitive to changes in substrate topology introduced at this position. In the context of a membrane-embedded helix a single insertion can lead to far reaching structural changes by modifying the tilt or pitch of the TMD. This assumption is further supported by the phenotype of the KKK/LLL triple mutant which essentially harbors a deletion of the cytosolic membrane anchor. Removal of the conformational restraint by the membrane anchor leads to an Aβ peptide pattern in IP-MS that is comparable to the wild-type substrate.

A recent publication by Sato and colleagues (Sato et al. 2009) provides important structural data on the LLL-ins mutation arguing that a helix-to-coil transition close to the ε-site is important for proteolysis. Their overall model, however, needs to be modified taking our data into account. In the absence of an analysis of alternative Aβ species, a decrease in Aβ40 production with the LLL-ins mutation was interpreted as a loss of γ–cleavage (Sato et al. 2009). Our approach using a combination of specific Aβ42 and Aβ40 immunoassays and IP-MS revealed interesting mechanistic details. Accordingly, the helix-to-coil transition is certainly important for Aβ40 generation but not γ-secretase activity in general. Increased helical length by a single leucine is sufficient to promote the generation of the more pathogenic Aβ42 and simultaneously shorter products such as Aβ38. Considering the available structural data, we argue that the helix-to-coil transition close to the ε-site is of great importance for the cleavage specificity of γ–secretase. These findings are relevant for AD since the most distal mutation to cause familial AD reported to date is a single point mutation at the first membrane-anchoring lysine 53 (K/N) (Theuns et al. 2006). Similar to our K/L mutant an increase in Aβ42 was observed although it can be presumed that the introduction of a polar amino acid introduces different structural changes compared with a hydrophobic leucine.

Considering the proximal juxtamembrane region, we focused our investigation on lysine 28 (K28) of the C99 substrate which has been of interest to various groups (Page et al. 2010; Kukar et al. 2011). We addressed the unanswered question of whether preserving a positive charge at position 28 is sufficient to conserve the normal cleavage pattern. Our data reveal that exchanging the side chain from a primary amine to a guanidine did not lead to any qualitative changes in the pattern of Aβ peptides being secreted. In contrast, the K28E exchange favored production of Aβ37 at the expense of Aβ40 and Aβ42 as previously reported (Page et al. 2010). Quite unexpectedly and in contrast to K28R, the K28E mutation markedly changed the pharmacological response to GSM-1 by lowering its potency for Aβ42 by approximately tenfold. In addition, E2012 treatment caused a surprising increase in Aβ40 and Aβ42 production in K28E expressing cells. A more detailed analysis by IP-MS confirmed that in principle, GSM-1 still exerts its modulatory mechanism with the K28E substrate as evident by the elevation of Aβ38. It is fair to conclude that although lowering the affinity of GSM-1, this mutation did not prevent the shift of cleavage to a position further N–terminal in the C99 TMD. In sharp contrast, the mechanism of action of the HA-type GSM E2012 was drastically changed to an inverse GSM. This was apparent by the elevation of Aβ40 and Aβ42 at the expense of several shorter species such as Aβ32, Aβ33, and Aβ34. A previous study has provided evidence that the increase in Aβ37 observed with the K28E mutant is only observed in cells, whereas in a cell-free system this substrate behaves like wild-type C99 (Page et al. 2010). This could imply that additional proteases lead to the generation of the very abundant Aβ37 peak as a result of the change in the Aβ peptide sequence itself. Our data suggest a similar interpretation since upon treatment with E2012 a strong inverse modulatory phenotype is observed which does not appear to lead to any changes in the Aβ37 peak. Nevertheless, the production of the Aβ37 peptide still depends on an initial γ-secretase processing step since its production can be blocked with a potent γ–secretase inhibitor (data not shown).

Taken together, our data are open to multiple interpretations given the fact that the exact mode of action of GSMs remains unknown. Despite a claim for direct substrate binding (Kukar et al. 2008) compelling photoaffinity labeling evidence now favors a direct interaction of both NSAID- and HA-type of GSMs with the catalytic subunit of γ-secretase enzyme (PS1/2) (Uemura et al. 2010; Ebke et al. 2011; Ohki et al. 2011). These data confirm earlier observations pointing toward an allosteric mechanism (Beher et al. 2004; Lleo et al. 2004). One potential underlying mechanism could be a change of the energetic barriers for the different sequential lineages or product lines, meaning a GSM elevating Aβ38 promotes increased conversion or processivity from Aβ42, whilst a GSM elevating Aβ37 increases conversion or processivity from Aβ40. This is in part supported by the pharmacology of GSM-1 (elevates Aβ38 at the expense of Aβ42) and E2012 (reduces both Aβ40 and Aβ42 and elevates Aβ37 and Aβ38). Structural changes introduced by mutagenesis of the distal TMD, however, did not influence the overall response to either GSM. This observation may argue against the increased processivity scenario as the GSMs could be expected to be less efficient in a situation where the Aβ38/42 lineage is highly favored, which was not the case. How mechanistically the pharmacology of E2012 is dramatically changed with the K28E mutant is challenging to envisage but not unprecedented since fenofibrate also changes its pharmacology from an Aβ40-lowering inverse GSM to an Aβ40-raising compound with the K28E mutant (Page et al. 2010). We speculate that the K28E mutant could change the way the substrate is presented to the native enzyme (Fig. 6c). This would be similar to the allosteric change introduced by the binding of the HA-type GSM (Fig. 6b) to γ-secretase and explains the enhanced production of shorter Aβ peptides. The allosteric change introduced into γ-secretase by the HA-type GSM potentially counteracts the effects of the conformational change in the K28E substrate and thus restores the wild-type cleavage pattern.

Figure 6.

Potential mechanism underlying the change in pharmacology of the HA-type γ-secretase modulator (GSM) with the K28E C99 substrate. (a) γ-Secretase cleavage of wild-type C99 occurs approximately in the middle of the membrane releasing Aβ40 and Aβ42 peptides. (b) Binding of the HA-type GSM to γ–secretase changes the conformation of the enzyme affecting the way the substrate is presented to the active site. One potential mechanism could be a tilt of the substrate relative to the enzyme which would move the γ-secretase cleavage further N-terminal of C99. This might explain the enhanced generation of Aβ37 and Aβ38 at the expense of Aβ40 and Aβ42. (c) The K28E mutation changes the way the substrate interacts with γ-secretase thereby shifting the cleavage to shorter peptides such as Aβ33. (d) The allosteric change in γ-secretase upon binding of the HA–type GSM counteracts the effects of conformational change in the substrate and restores the wild-type cleavage pattern in the K28E substrate. GS, γ-secretase; GSM, HA-type GSM E2012.

Taken together with the observation that the potency of GSM-1 is reduced, we view our data in support of a model where the different classes of GSM (HA- vs. NSAID-type) bind to different sites or via different modes to the γ-secretase enzyme. This view is supported by recently reported binding and compound cross-competition studies (Borgegard et al. 2012) indicating different sites of interaction for distinct GSMs. Although the reduction of GSM-1 potency could be interpreted as evidence for substrate binding per se, we rather view this as evidence for modulation of substrate binding as the underlying mechanism of action of GSMs. With this respect, the region around K28 had been identified as a critical regulator of γ-secretase cleavage in earlier studies (Ren et al. 2007), which could be viewed as evidence that this region directly interacts with the enzyme. Therefore, it is perfectly conceivable that modulation of this interaction is a key mechanism underlying GSM activity and that any structural change in this domain has a profound impact on the pharmacology of these compounds. It is noteworthy that structural elements along the GXXXG dimerization motif (Munter et al. 2007) further distal to K28 also contribute to the pharmacology of GSMs which were identified using substrate chimeras (Sagi et al. 2011).

In summary, we have identified critical determinants in the proximal and distal juxtamembrane domains of the APP C99 substrate which differentially affect γ-secretase cleavage specificity and modulator pharmacology. The observation that replacement of lysine 28 with a glutamic acid has a profound impact on the pharmacology of GSMs could be used to further explore the mechanism of action of different types of GSM in development.


The authors declare a conflict of interest since all authors were full-time employees of Merck Serono SA at the time this study was performed. DB is a founder of Asceneuron SA and owns a share in the company. The authors thank Julien Fabrègue for preparing ruthenium-labeled antibodies, Damien Begue for the collection and analysis of the initial mass spectra.