C-terminal PAL motif of presenilin and presenilin homologues required for normal active site conformation


Address correspondence and reprint requests to Alison Goate, Department of Psychiatry, Campus Box 8134, Washington University School of Medicine, St. Louis, MO 63110, USA
E-mail: goate@icarus.wustl.edu


The Alzheimer's disease-associated β-amyloid peptide is produced through cleavage of amyloid precursor protein by β-secretase and γ-secretase. γ-Secretase is a complex containing presenilin (PS) as the catalytic component and three essential cofactors: Nicastrin, anterior pharynx defective (APH-1) and presenilin enhancer-2 (PEN-2). PS and signal peptide peptidase (SPP) define a novel family of aspartyl proteases that cleave substrates within the transmembrane domain presumptively using two membrane-embedded aspartic acid residues for catalysis. Apart from the two aspartate-containing active site motifs, the only other region that is conserved between PS and SPP is a PAL sequence at the C-terminus. Although it has been well documented that this motif is essential for γ-secretase activity, the mechanism underlying such a critical role is not understood. Here we show that mutations in this motif affect the conformation of the active site of γ-secretase resulting in a complete loss of PS binding to a γ-secretase transition state analog inhibitor, Merck C. Analogous mutations in SPP significantly inhibit its enzymatic activity. Furthermore, these mutations also abolish SPP binding to Merck C, indicating that SPP and γ-secretase share a similar active site conformation, which is dependent on the PAL motif. Exploring the amino acid requirements within this motif reveals a very small side chain requirement, which is conserved during evolution. Together, these observations strongly support the hypothesis that the PAL motif contributes to the active site conformation of γ-secretase and of SPP.

Abbreviations used

β-amyloid peptide


Alzheimer's disease


anterior pharynx defective


amyloid precursor protein


C-terminal fragment


Dulbecco's modified Eagle's medium


PS1 Δexon 9


endoplasmic reticulum


familial Alzheimer's disease


fetal bovine serum




human embryonic kidney


high molecular weight


mouse embryonic fibroblast




Notch intracellular domain


N-terminal fragment


phosphate-buffered saline


presenilin enhancer-2



PS1/2 K/O

presenilin 1/presenilin 2 knockout


sodium dodecyl sulfate-polyacrylamide gel electrophoresis


signal peptide peptidase




3,3′5,5′-Tetramethyl benzidine



Cerebral deposition of β-amyloid (Aβ) peptides is believed to be the central event in the pathogenesis of Alzheimer's disease (AD). Aβ is generated through sequential cleavage of the β-amyloid precursor protein (APP) by β-secretase and γ-secretase. The latter is a prototypic member of a family of aspartyl proteases that are responsible for regulated intramembranous proteolysis of type I transmembrane (TM) proteins (Martoglio and Golde 2003). γ-Secretase functions as a high molecular weight (HMW) complex that consists of presenilin (PS), which is presumed to harbor the active site and three essential cofactors: nicastrin (NCT) (Yu et al. 2000), anterior pharynx defective (APH-1) (Goutte et al. 2002) and presenilin enhancer-2 (PEN-2) (Francis et al. 2002). Mutations in PS are the most common cause of autosomal dominant familial AD (FAD). PS is a polytopic membrane protein that requires interaction with three cofactors and endoproteolysis to generate heterodimeric complexes consisting of PS N-terminal fragment (NTF) and PS C-terminal fragment (CTF), which are the active form of PS. Two aspartic acids in PS (D257 and D385 per PS1 sequence) are believed to constitute the active site of γ-secretase (Wolfe et al. 1999). Five PS homologues have been identified in the human genome based upon sequence conservation (Grigorenko et al. 2002; Ponting et al. 2002; Weihofen et al. 2002). One of these homologues has since been identified as signal peptide peptidase (SPP), an aspartyl protease that mediates intramembranous cleavage of certain signal peptides (Weihofen et al. 2002). PS and SPP share similar overall topology (although reversed membrane orientation) (Martoglio and Golde 2003; Friedmann et al. 2004; Nyborg et al. 2004a), both appear to be aspartyl proteases that have partially overlapping inhibitor profiles (Weihofen et al. 2003; Nyborg et al. 2004b) and both require ectodomain shedding of their substrates before they will cleave substrates within the TM domain (Lemberg and Martoglio 2002). PS, SPP and the other homologues share little primary sequence homology except around the active site aspartyl residues and around a P433A434L435 motif (per PS1 sequence) at the C-terminus. We and others have previously shown that single point mutations to this motif (P433L, A434D and L435R) completely abolish γ-secretase activity (Tomita et al. 2001; Wang et al. 2004). When the P433L mutation was introduced into the PS1 Δexon 9 (ΔE9) variant (Perez-Tur et al. 1995), a naturally occurring mutant that does not undergo endoproteolysis, but is active as a full-length (FL) protein, it also abolished γ-secretase activity of ΔE9 (Wang et al. 2004). The mechanisms that dictate such an essential role for the PAL sequence with respect to activity are not clear. This motif was originally proposed to be important for stabilization and formation of the γ-secretase complex with cofactors (Tomita et al. 2001). This hypothesis, however, does not explain why this motif is also completely conserved in SPP, which, unlike PS, does not appear to require cofactors for activity. Furthermore, we have reported experimental evidence that does not support such a role for this motif (Wang et al. 2004). In our studies, the P433L PS1 mutation did not affect protein stability, trafficking through the Golgi apparatus, binding to substrates or incorporation into the HMW γ-secretase complex as assessed by glycerol gradient centrifugation (Wang et al. 2004). However, because APH-1 and PEN-2 are both small molecules, a small change in the molecular weight of the complex may not be detected with glycerol gradient experiments. We went on to examine the effects of the P433L mutation on presenilin association with each cofactor by performing co-immunoprecipitation experiments. Our data showed that the P433L mutation has no effect on wild-type (wt)PS1 or ΔE9 association with NCT, APH-1 or PEN-2, nor does it affect NCT maturation (Wang J., unpublished data). These results were recently confirmed by Kaether et al. with a GFP-tagged PS1 (Kaether et al. 2004). In this study Kaether et al. suggested that the PAL motif was part of an endoplasmic reticulum (ER) retention signal. However, this mechanism does not explain the complete loss-of-function phenotypes of the PAL motif mutants.

We sought to further investigate the mechanism by which this motif contributes to γ-secretase activity. We show that mutations in the PAL motif abolish wtPS1 or ΔE9 binding to a γ-secretase transition state analog inhibitor. These mutations also greatly inhibit SPP activity as well as association with the same inhibitor. We further analyzed the structural requirement at this motif and found that the size of the side chain was critical for these residues. Collectively, our findings suggest an involvement of this motif in the normal conformation of the active site of γ-secretase and SPP.

Materials and methods

Plasmids, cells and antibodies

cDNA constructs for wtPS1, PS1/ΔE9, PS1/P433L, PS1/A434D, ΔE9/P433L, C99-6myc, APPΔNL, NotchΔE, SPPCTV5His (wtSPP) and pGL3 5x ATF6 luciferase reporter have been described previously (Cai et al. 1993; Wang et al. 2000, 2004; Nyborg et al. 2004b); The GFP construct was obtained from Gibco (Rockville, MD, USA). Expression construct (pAG3 SPPsub) encoding an SPP substrate gpUL40 (Lemberg and Martoglio 2002) fused at the N-terminus with the cytoplasmic fragment of the basic-leucine zipper DNA-binding protein ATF6 was described elsewhere (Nyborg et al. 2004a). The QuickChange™ site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) was used to introduce point mutations into wtPS1 (P433G, P433A, P433V, P433F, P433D, P433N, A434L, A434G, A434V, A434I, A434C, L435A, L435G, L435V, L435I and L435F), ΔE9 (P433A, P433V, A434G, A434V, A434D, L435G, L435F, L435R, P436L and D257A) and wtSPP (P317L, A318D and L319R) constructs. All constructs were confirmed by sequence analysis.

Human embryonic kidney 293 cells (HEK) (ATCC, Rockville, MD, USA) and presenilin 1/presenilin 2 knockout (PS1/2 K/O) mouse embryonic fibroblasts (MEFs) (provided by B. De Strooper) were grown at 37°C under 5% CO2 in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) and 100 µg/mL penicillin/streptomycin (Invitrogen Corporation, Carlsbad, CA, USA). Pooled stable cell lines expressing wtPS1, P433L, ΔE9, ΔE9/P433L, ΔE9/A434D, ΔE9/L435R, ΔE9/P436L and ΔE9/D257A were generated by transfecting HEK parental cells with the respective cDNA constructs under G418 selection and maintained in DMEM medium with 1000 µg/mL G418 (Research Products International, Mt Prospect, IL, USA). Stable cell lines expressing wtSPP, SPP/P317L, SPP/A318D and SPP/L319R were generated in HEK cells under Blasticidin (Invitrogen) selection and maintained in DMEM with 7.5 µg/mL Blasticidin.

The following antibodies were used: polyclonal antibody 7N14 (against PS1 residues 1–14); monoclonal antibody NT1 (against an epitope in PS1 residues 41–49, a gift from Dr P. Mathews); 00/2 PS1 loop antibody was a gift from Janetta Culvenor (Evin et al. 2001). Anti-myc monoclonal antibody 9E10 was purchased from Sigma (St. Louis, MO, USA); anti-V5 monoclonal antibody was purchased from Invitrogen. Monoclonal antibodies 2G3 (Aβ 1–40), 21F12 (Aβ 1–42) and 3D6B (Aβ 1–5) were from Eli Lilly (Indianapolis, IN, USA) and have been previously described (Johnson-Wood et al. 1997).

Inhibitor affinity precipitations

Membranes from HEK cells stably expressing different PS1 and ΔE9 variants were prepared and the inhibitor affinity precipitation assay was performed as previously described (Beher et al. 2003). Briefly, membranes were solubilized in 1% (w/v) CHAPSO MES-buffer [50 mm MES pH 6.0, 0.15 m NaCl, 5 mm MgCl2, 1 × EDTA-free protease inhibitor cocktail (Roche Molecular Biochemicals, Lewes, East Sussex, UK)], and adjusted to a final detergent concentration of 0.5%. Endogenous biotinylated proteins were precleared by incubation with streptavidin-coupled magnetic beads (Dynal, Bromborough, Wirral, UK). Protein concentrations of the samples were determined using the bicinochonic assay kit (Perbio Science, Tattenhall, Cheshire, UK) according to the manufacturer's instructions in a 96-well format. All samples were adjusted to 0.2 mg/mL protein with 0.5% CHAPSO MES-buffer prior to incubation of 0.45 mL of solubilized enzyme with 0.1 µm of the biotinylated aspartyl protease transition-state analogue inhibitor Merck C as described previously (Beher et al. 2003). Non-specific binding was analyzed by adding a 100-fold excess of the non-biotinylated inhibitor Merck A (Shearman et al. 2000) in addition to the affinity ligand. Enzyme–inhibitor complexes were captured with streptavidin-coupled magnetic beads, and following washing, the specifically bound proteins were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophores (SDS-PAGE) prior to immunoblotting. Generally the precipitate was compared to 4% of the input. L = load/input (4% of total), S = specific capture, N = non-specific capture.

For SPP inhibitor binding experiments, HEK cells stably expressing wtSPP, SPP/P317L, SPP/A318D and SPP/L319R were collected and solubilized in lysis buffer (1% CHAPSO, 50 mm MES pH 6.0, 0.15 m NaCl, 5 mm EDTA, 1 × protease inhibitor cocktail). After solubilization, samples were diluted with an equal volume of buffer without CHAPSO to yield 0.5% CHAPSO before subsequent affinity precipitation assays as described above. SPP and PS1-CTF were detected from the same precipitation and blot. SPP was detected by western blotting with anti-V5 antibody and PS1-CTF was detected by blotting with the 00/2 PS1 loop antibody (Evin et al. 2001). Individual polypeptides were detected by infrared imaging using Alexa Fluor 680-conjugated goat anti-rabbit or anti-mouse F(ab′)2 fragments (Molecular Probes, Paisley, UK) and a LI-COR®. Odyssey infrared imager (LI-COR Biosciences Inc., Lincoln, NE, USA).

Signal peptide peptidase activity assay

The luciferase reporter assays were performed exactly as described (Nyborg et al. 2004a). Briefly, HEK cells were transiently cotransfected with 2 µg of DNA mixture containing 0.02 µg of pRL-SV40 Renilla expression construct (Promega, Madison, WI, USA), 0.25 µg of pGL3 5x ATF6 reporter construct, 0.25 µg of pAG3 SPPsub, various SPP variants using FuGENE™ 6 transfection reagent (Roche, Indianapolis, IN, USA). Cells were lysed with passive lysis buffer (Promega, Madison, WI, USA). Firefly luciferase and Renilla activities were measured using the Dual-Luciferase® kit (Promega) and a Veritas microplate Luminometer (Turner Biosystems, Sunnyvale, CA, USA) with Veritas 2.0.40 software package. Transfections were performed in triplicate. Results were normalized to the Renilla activity control.

Transfection and Aβ enzyme-linked immunosorbent assay

PS1/2 K/O MEFs in p60 culture plates were transfected with 7.5 µg of DNA mixture (3.75 µg APPΔNL and 3.75 µg PS1/ΔE9 variants) using FuGENE™ 6 transfection reagent. After 16 h, transfection medium (Optimem reduced serum medium + 10% FBS + 100 µg/mL penicillin/streptomycin) was replaced with recovery medium (DMEM + 10% FBS + 100 µg/mL penicillin/streptomycin + 2% Glutamine + 1% Na-pyruvate + 1% NEAA + 2% MEM essential amino acids + 10 mm HEPES). After recovery for 6 h, recovery medium was replaced with 1.3 mL transfection medium supplemented with 10 µm phosphoramidon (Sigma). The medium was harvested after 24 h and store at −80°C for Aβ measurement.

Aβ is measured with sandwich ELISA assay, using 2G3 (1–40) and 21F12 (1–42) as the capture antibodies and 3D6B as the reporter antibody (Johnson-Wood et al. 1997; DeMattos et al. 2001). Briefly, the 2G3 and 21F12 monoclonal antibodies were coated at 20 µg/mL into 96-well immunoassay plates overnight at 4°C. The plates were then aspirated and blocked with 1% dry milk in phosphate-buffered saline for 1 h at 37°C. After washing five times in phosphate-buffered saline (PBS), the samples and standards were added to the plates and incubated at 4°C overnight with constant shaking. The plates were washed five times with PBS/0.05% Tween 20 and incubated with reporter antibody 3D6B (1:2000 dilution in 0.5% bovine serum albumin/PBS/Tween 20) for 2 h at room temperature. After washing five times in PBS/Tween 20, the plates were incubated with avidin–horseradish peroxidase for 1.5 h at room temperature. Signals were developed with 100 µL/well TMB (3,3′,5,5′-Tetramethyl benzidine) (Sigma) and read at 650 nm with a SpectraMAX 190 microplate reader (Molecular Devices, Sunnyvale, CA, USA).


The PAL motif mutations affect γ-secretase active site conformation

The observations that mutations in the PAL motif exhibit exactly the same phenotypes as the active site aspartate mutants, and that PAL and the two active site aspartates are the most highly conserved regions in both PS and SPP suggest that the PAL sequence may affect the active sites of both enzymes. To test this hypothesis, we determined whether mutations within the PAL motif affected PS1 and ΔE9 binding to γ-secretase transition-state analogue inhibitors, which are directed to the γ-secretase active site. These inhibitors have previously been shown to covalently label PS1 NTF/CTF fragments, but not the holoprotein, providing strong evidence that PS1 NTF/CTF form the active site of γ-secretase (Li et al. 2000). These transition state analogue inhibitors also label the full-length PS1ΔE9 variant, which is active as a holoprotein (Li et al. 2000). Since the wtPS1 holoprotein is thought to be inactive and predicted not to bind to the inhibitor, we also examined the effects of the PAL mutations in the ΔE9 background. All three mutations within the PAL motif (P433L, A434D, L435R) and D257A were inactive in wtPS1 (Wang et al. 2004) and the ΔE9 background in a rescue assay in PS1/2 K/O cells (Fig. 1a). CHAPSO-solubilized membranes from HEK cells stably expressing different PS1 and ΔE9 variants were incubated with the biotinylated γ-secretase transition state analogue inhibitor Merck C (Beher et al. 2003), a derivative of the inhibitors used by Li et al. (2000). The inhibitor complexes were captured by addition of streptavidin-coupled magnetic beads and subjected to SDS-PAGE and immunoblotting with anti-PS1-NTF antibody. We observed that PS1-NTF (overexpressed or endogenous), but not PS1-FL or P433L-FL, was isolated with the inhibitor (Fig. 1b). As previous published (Li et al. 2000), ΔE9 holoprotein bound prominently to the inhibitor. Mutations at the active site aspartates abolished PS1 (Wrigley et al. 2004) and ΔE9 binding to the inhibitor as expected (Fig. 1b). Strikingly, ΔE9 PAL motif mutants (ΔE9/P433L, ΔE9/A434D and ΔE9/L435R) all completely abolished ΔE9 association with the inhibitor (Fig. 1b). In contrast, an identical substitution at the adjacent proline (P436L) had no effect on ΔE9 binding to the inhibitor. Consistent with this observation, the P436L mutation did not affect the γ-secretase activity of wtPS1 (Wang et al. 2004) or ΔE9 (Fig. 1a). In addition, although this P436 is conserved among PS homologues across species, it is not conserved in SPP and SPP homologues. These observations support the hypothesis that the PAL mutants affect the conformation of the active site of γ-secretase.

Figure 1.

Mutations in the PAL motif abolish PS1 and ΔE9 binding to a γ-secretase transition state analogue inhibitor. (a) PS1/2 K/O cells were transfected with C99-6myc and different ΔE9 variants as indicated. Cell lysates were western blotted with anti-myc antibody 9E10 to assess myc-tagged substrates (APP-CTFα/β) and cleavage product APP-CTFγ (upper panel). Cell lysates were also immunoprecipitated with 7N14 and blotted with NT1 to assess the expression of ΔE9 variants (lower panel). (b) CHAPSO-solubilized membranes from HEK cells stably expressing wtPS1, P433L, ΔE9, ΔE9/P433L, ΔE9/A434D, ΔE9/L435R, ΔE9/P436L and ΔE9/D257A were subject to inhibitor affinity precipitation assays exactly as described in Materials and methods. L, load/input (4% of total); S, specific capture; N, non-specific capture.

The PAL motif is essential for signal peptide peptidase activity

To further test the hypothesis that the PAL motif is involved in the active site of both γ-secretase and SPP, we tested whether this motif is also important for SPP activity. It is striking that PS and SPP are very dissimilar polypeptides apart from sharing the common feature of being intramembrane-cleaving aspartyl proteases. They and their substrates have exactly opposite membrane topology (Martoglio and Golde 2003; Friedmann et al. 2004; Nyborg et al. 2004a). Most importantly, they are regulated quite differently. PS apparently requires three additional proteins as cofactors and activation through endoproteolysis. SPP, on the other hand, does not appear to require any other proteins, or endoproteolysis for activity. If the PAL motif is important for regulatory mechanisms that are not shared by PS and SPP, we would expect that this motif would be dispensable for SPP activity. We made analogous mutations in the SPP PAL motif (P317L, A318D and L319R) and examined the activities of these mutants using our recently developed cell based reporter assay for SPP activity (Nyborg et al. 2004a). In this assay, cleavage of the substrate by SPP releases the N-terminal ATF6 cytoplasmic domain, which translocates to the nucleus and activates an ATF6 luciferase reporter gene. Luciferase activity was normalized to Renilla activity to control for transfection efficiency and protein expression. Transient transfection of HEK cells with SPPCTV5His (full-length SPP containing V5 and His6 epitope tags at the C-terminus, designated wtSPP hereafter) dramatically increased SPP activity, whereas similar levels of expression of SPP proteins containing the PAL mutations only slightly increased SPP activity over the endogenous level (Fig. 2a). This slight increase in SPP activity is also observed with the aspartate mutants (Nyborg and Golde, manuscript in preparation). We have previously shown that SPP forms a homodimer which appears to be the active form of SPP (Nyborg et al. 2004b). Similar to the wtSPP, SPP PAL mutants also exist predominantly as 95 kDa homodimers (Fig. 2b), indicating that the loss of function of the PAL mutants is not due to a defect in dimer formation. These results demonstrate that the PAL motif is required for SPP activity.

Figure 2.

Mutations in the PAL motif in SPP dramatically decrease SPP activity and abolish SPP binding to the γ-secretase transition state analogue inhibitor. (a) Experiments were performed as described in Material and methods. HEK cells were transiently transfected with 2 µg of total DNA mixture containing 0.02 µg of pRL-SV40 Renilla expression construct, 0.25 µg of pGL3 5x ATF6 reporter construct, 0.25 µg of pAG3 SPPsub and control empty vector (Endogenous) or different SPP variants as indicated. Cells were lysed and firefly luciferase and Renilla activities were measured. Results were expressed as the ratio of firefly luciferase/Renilla activity. Transfections were performed in triplicate. ‘No sub’ indicates a background activity when no substrate was transfected. (b) Cell lysates were run on a 10–20% SDS-PAGE and subjected to western blot with anti-V5 antibody to evaluate the expression of exogenous SPP proteins. (c) HEK cells expressing wtSPP, SPP/P317L, SPP/A318D and SPP/L319R were collected and subject to inhibitor affinity precipitation assays exactly as described in Materials and methods. L, load/input (1.4% of total); S, specific capture; N, non-specific capture.

To further test the hypothesis that PS and SPP share a similar active site conformation, and that the PAL motif is essential for this conformation, we examined the effect of the PAL motif mutations on SPP binding to the same inhibitor Merck C. As shown in Fig. 2(c), wtSPP binds normally to the inhibitor. It is noticeable that both the dimer and monomer forms of SPP are in the inhibitor pulled-down precipitates. It appears that part of the bound dimer was converted to monomer during handling process. In contrast, all three mutants exhibited binding to the inhibitor at levels similar to non-specific background levels. As a control, endogenous PS1-CTF was pulled downed by the inhibitor in all cell lines (Fig. 2c). Previously, we showed that a γ-secretase transition photo affinity probe that is capable of labeling the active γ-secretase complex (Esler et al. 2002) is capable of labeling the SPP dimer as well (Nyborg et al. 2004b). Here we show that another active site directed inhibitor of γ-secretase not only labels SPP but mutations in the PAL motif of SPP abolish labeling. This experiment clearly demonstrated that the PAL motif is also essential for SPP active site conformation, providing support for the hypothesis that this motif is involved in the catalytic activity shared by both enzymes, but not their differing regulatory mechanisms.

Structural requirements at the PAL motif for presenilin endoproteolysis and γ-secretase activity

Since the PAL motif appears to be directly involved in the catalytic activity of γ-secretase and SPP, we determined the amino acid side-chain requirement at these residues to support enzyme activity. We made a series of conservative substitutions at these three residues and analyzed their effects on γ-secretase activity. Most amino acid substitutions at residue P433 abolished PS1 endoproteolysis as well as γ-secretase cleavage of both APP and Notch (Figs 3a, d–f). The P433G mutant showed trace levels of endoproteolysis and cleavage of C99 on long exposures of the blots (Fig. 3a, long exposure & Exp#2). The only amino acid tested that can replace a proline at this position, and retain activity, is an alanine. While P433A is active, P433V is completely inactive. These data suggest that a very small side chain is required at position P433. Requirements at A434 position are also very stringent (Figs 3b, d–f). Among the many conservative mutations tested, only A434G and A434C are still active, also suggesting that a small side chain is required at this position as well. In contrast, the side chain requirement at the L435 residue is less stringent. Conservative substitutions at this position (L435A, L435V and L435I) all undergo endoproteolysis and cleave APP (Figs 3b, d, f), but the activities of the L435A and L435V substitutions were impaired for Notch cleavage (Figs 3b and e). Similar discrepancies between the effect of specific mutations on APP and Notch cleavage have recently been reported for several PS1 and PS2 mutants (Brunkan et al. 2005a; Walker et al. 2005). The activity of the L435G substitution was significantly impaired and the L435F mutant, a naturally occurring mutant, identified in a patient with early onset AD (Rogaeva et al. 2001), was completely inactive. The requirement at position L435 appears to be a nonaromatic hydrophobic residue. These results were confirmed when similar amino acid substitutions were examined in the ΔE9 background (Figs 3c and g).

Figure 3.

Amino acid requirement at the PAL motif for PS1 endoproteolysis and γ-secretase activity. PS1/2 K/O cells were cotransfected with C99-6myc (upper panels) or NotchΔE (middle panels) and different PS1 or ΔE9 variants as indicated. Cell lysates were blotted with anti-myc antibody 9E10 to detect APP-CTFα/β and cleavage product APP-CTFγ (upper panels), NotchΔE and cleavage product NICD (middle panels). The cell lysates were also immunoprecipitated with 7N14 and blotted with NT1 to assess the expression and endoproteolysis of PS1 variants (lower panels). (a) Amino acid substitutions at P433; Long Exposure: a longer exposure of the shown blot; Exp#2: a result from another independent experiment, showing wtPS1 and P433G. (b) Amino acid substitutions at A434 and L435; •, non-specific band. (c) Selected amino acid substitutions in ΔE9. (d) Semi-quantitative densitometric analysis of CTFγ production. The densities of both CTFγ and CTFα/β were quantified using Quantity one software and ratio of CTFγ/CTFαβγ was plotted. (e) Semi-quantitative densitometric analysis of NICD production. Ratio of NICD/(NotchΔE + NICD) was plotted. (f) Semi-quantitative densitometric analysis of PS1 endoproteolysis. Ratio of PS1-NTF/(PS1-FL + PS1-NTF) was plotted. (g) Semi-quantitative densitometric analysis of CTFγ production in ΔE9 mutants. Error bars represent SEM. Asterisks indicate a significant difference from the wtPS1 (d, e, f) or ΔE9 (g), as determined by one-way anova with Bonferroni correction (*p < 0.05; **p < 0.01; ***p < 0.001). Each value was calculated from measurements of three or more independent experiments.

We also measured the activities of the mutants towards Aβ production. To get a definitive answer without the interference of endogenous PS, we performed the experiments in PS1/2 K/O cells. The profile of total Aβ (Aβ40 + Aβ42) production was similar to APP-CTFγ(Fig. 4a). All the inactive mutants did not show any detectable Aβ40 or Aβ42, suggesting that these mutants exhibit a complete loss of function. P433G and L435G mutants produced very low amounts of Aβ40 and no detectable Aβ42, consistent with the notion that these two mutants are severely impaired. The P433G mutant was shown to produce high amounts of Aβ42 in a recently published study (Nakaya et al. 2005). The reason for the discrepancy between the published data and our own results is not known, but the authors utilized a retrovirus-mediated gene expression system which could produce an artifactual result. Most mutants which have activity showed a decrease in total Aβ, but did not affect the Aβ42/Aβ40 ratio (Fig. 4b). The L435A mutant showed a slightly decreased Aβ42/Aβ40 ratio. The A434C mutant produced six-fold more Aβ42 than wtPS1 (Aβ42 represented 35% of total Aβ with A434C compared with 6% of total Aβ with wtPS1). This result confirmed published genetic studies, which suggested that A434C was an FAD mutation (Devi et al. 2000; Rogaeva et al. 2001). Although the L435F mutation was also indicated as an FAD mutation (Rogaeva et al. 2001), our Aβ analysis did not detect any Aβ produced from this mutant. The Aβ result was consistent with APP-CTFγ, Notch intracellular domain (NICD) and endocleavage data (Fig. 3), which all indicated that this is a complete loss-offunction mutation. It is, however, possible that L435F could produce trace amounts of Aβ42, which are beyond the detection limits of our assay, yet sufficient to cause an FAD phenotype in patients. Alternatively, since no segregation data was available for the L435F mutation it may be a chance occurrence that it was associated with an individual with FAD. We also tested representative mutations in the ΔE9 background. Similarly, active mutants (ΔE9/P433A, ΔE9/A434G) produced Aβ, although at a lower level. These mutants did not modify the FAD phenotype of ΔE9. Inactive mutants, including ΔE9/L435F, did not produce any Aβ42 (Fig. 4b).

Figure 4.

Aβ analysis of PAL mutants. PS1/2 K/O cells were transfected with APPΔNL and different PS1 or ΔE9 variants as indicated. Secreted Aβ40 and Aβ42 were measured exactly as described in Materials and methods. Total Aβ (Aβ40 + Aβ42) (a) as well as Aβ42/Aβtotal ratio (b) are shown. The results of GFP or inactive mutants (P433V, P433L, P433F, P433D, P433N, A434D, A434L, A434V, A434I, L435F, ΔE9/P433V, ΔE9/A434V, ΔE9/L435F) were not plotted because they did not produce any detectable Aβ. Error bars represent SEM. Asterisks indicate a significant difference from the wtPS1 or ΔE9, as determined by one-way anova with Bonferroni correction (*p < 0.05; **p < 0.01; ***p < 0.001). N.D., not detectable. The data are representative of three independent experiments.


In this work, we investigated the mechanism by which the essential PAL motif in PS contributes to γ-secretase activity. It is now clear that this motif is not involved in γ-secretase high molecular weight complex formation (Kaether et al. 2004; Wang et al. 2004). Our data support the hypothesis that this motif is important for the proteolytic mechanism shared by SPP and PS. First, mutations at this motif completely abolish PS and ΔE9 binding to a transition state analog inhibitor directed to the active site (Fig. 1). Second, like the two aspartates that form the active site, this motif is also essential for SPP activity, which is regulated very differently from PS (Fig. 2a). Third, the same transition state analog inhibitor binds to SPP, which is also dependent on the PAL motif (Fig. 2c). This motif could constitute part of the γ-secretase active site, or alternatively, it may be distant from the active site but necessary to ensure the correct conformation of the active site. A naturally occurring PS1 FAD mutant within the TM1 domain (V96F) appears to also change γ-secretase active site conformation (Brunkan et al. 2005b). The TM1 region, however, is highly divergent in SPP and SPP homologues. Therefore, the latter mechanism does not explain the high conservation between PS and SPP. While proof of the hypothesis will require solution of the presenilin structure, current data are more consistent with the hypothesis that the PAL motif is directly involved in the active site of both PS and SPP.

We also determined the amino acid requirements at these positions for γ-secretase activity. Surprisingly P433 can be replaced by an alanine. It suggests that a possible structural change that could be introduced by a proline is not required for PS1 endoproteolysis and γ-secretase activity. This proline is conserved among all known homologues including distant bacterial species such as Thermoplasm acidophilum and Thermoplasma volcanium. The only exception is Halobacterium sp. Interestingly, it contains an alanine at this position (Ponting et al. 2002). This is consistent with our mutagenesis data. At position A434, it appears that either a very small side chain or no side chain is required. Also consistent with our mutagenesis data, which show that glycine can replace alanine, bacteria contain a glycine at this position (Ponting et al. 2002). The requirement at L435 appears to be a hydrophobic hydrocarbon side chain, introducing a bulky aromatic ring (L435F) or charged residue (L435R) abolished the activity. Taken together, the consistency of our functional data and evolutionary data suggests that this motif is essential for the formation of a functional active site.

It was recently proposed that this motif may be part of an ER retention signal (Kaether et al. 2004). Although mutation of this motif results in release of PS1 from the ER, our cell surface biotinylation experiments (data not shown) with the PAL motif mutants did not detect any differences in the steady state levels of cell surface PS1. These results suggest that either a smaller proportion of the mutant complexes actually make it to the cell surface or that the turnover of the complexes at the cell surface may be different for mutant vs. wild type PS complexes.

For this motif to be directly involved in an intramembrane-cleaving activity, it would be expected to be within a TM domain or to be membrane associated. Topological models of SPP place the PAL motif in the last transmembrane domain (Martoglio and Golde 2003; Friedmann et al. 2004; Nyborg et al. 2004a). Although early studies placed the PAL sequence of PS within a cytosolic domain, hydropathy plots suggest that the sequence is within a hydrophobic region (Li and Greenwald 1996; Lehmann et al. 1997; Nakai et al. 1999; Dewji et al. 2004). Several recent experimental and computational studies have demonstrated that the C-terminal domain of PS is extracytoplasmic and that PS, like SPP, has nine TM domains (Henricson et al. 2005; Laudon et al. 2005; Oh and Turner 2005). These observations suggest that PS and SPP have identical topology, although with an opposite membrane orientation. In both molecules these new topological models place the PAL sequence within the final TM domain and close to the cytosolic edge of the membrane, in the case of PS. This putative location would appear to place the PAL sequence, in both PS and SPP, in a similar location within the membrane to the active site aspartyl residues and close to the putative cleavage sites of their substrates. Solving the PS and SPP structures will provide insight into the molecular mechanism of this unusual class of aspartyl proteases and facilitate the design of specific γ-secretase inhibitors for the treatment of AD.


We thank Dr B. De Strooper for PS1/2 K/O cells, Dr P. Mathews for the NT1 antibody, J. Culvenor for 00/2 PS1 loop antibody, Dr R. Kopan for the NotchΔE construct and Dr R. De Mattos for antibodies and protocols for the Aβ ELISA used in this study. We thank Dr A. Brunkan for the ΔE9/D257A construct, M. Martinez for technical assistance and J.S.K. Kauwe for assistance in Aβ analysis. This work was supported by a Missouri ADRD research grant and John Douglas French Alzheimer's foundation fellowship to JW, an NIH grant AG17050 to AG, the Mayo Foundation, an NIH/NINDS grant NS39072 to TEG and NS44734 to ACN.