Introduction
Strong evidence supports the notion that the aberrant formation and accumulation of β-amyloid peptides (Aβ), generated from the β-amyloid-precursor protein (APP) by the action of β- and γ-secretase, is a critical event in the pathology of familial and sporadic forms of Alzheimer's Disease (AD) [
1‐
3]. APP is initially cleaved by β-secretase to generate membrane-bound C-terminal fragments (CTFs), and a soluble N-terminal ectodomain sAPP
β . Subsequently, those CTFs are cleaved by the γ-secretase within the putative transmembrane domain to release heterogeneous β-amyloid peptides (Aβ) composed of 37 to 43 amino acids; the 40 and 42 amino acids long peptides referred to as Aβ
40 and Aβ
42, respectively, are more predominant and the latter one is the most toxic to neurons [
4]. Moreover, γ-secretase also cleaves APP at the ε-site (between L49 and V50), which resembles the γ-secretase S3 cleavage of Notch [
5‐
8], and is located further downstream the γ-site with only 2–5 residues inside the cytoplasmic membrane [
9]. Since the processing of Notch by the γ-secretase has important physiological implications, inhibition of its processing needs to be minimized in the development of γ-secretase inhibitors for AD therapy. Moreover, a variety of studies have already compiled a long list of potential γ-secretase cellular substrates that might further complicate selective inhibition of this enzyme. These substrates include CD44 [
10], LRP [
11], Erb4 [
12], Nectin [
13], E-cadherin [
14]), and two membrane-bound ligands of Notch (Delta and Jagged) [
15,
16].
Both biochemical and genetic approaches have led to the discovery that at least four membrane-bound proteins, presenilin (PS1 or PS2), nicastrin, aph1 and pen2, are needed to form an active γ-secretase complex [
17,
18]. Interestingly, only mutations in PS1 and PS2 but not in the other γ-secretase components are a common cause of the early onset familial Alzheimer's disease (FAD) [
3]. PS1 and PS2 FAD mutants have been shown to increase the ratio of Aβ
42 over total Aβ species [
19‐
22] by affecting γ-secretase activity through still poorly understood mechanisms. Notwithstanding, recent studies have demonstrated that a significant increase in absolute levels of Aβ
42 was only observed for half of the mutations [
23,
24], further strengthening the notion that the mechanistic effects of PS1 mutations on the onset of FAD are complex.
Ample evidences suggest that PS1 (or PS2) is not only an indispensable component of a functional γ-secretase, but also is its catalytic core. In fact, complete deficiency of PS1 in mice abolishes the processing of both APP to release Aβ and Notch at the S3 site to release NICD [
7,
25]. Strikingly, mutations of the aspartic acid residues to Ala or Glu in transmembrane segment 6 (TM6) (D257) and TM7 (D385) in PS1 or PS2 (numbering according to PS1) result in a dramatic reduction of γ-secretase activity [
26], implying the possibility of these two aspartic residues constitute the active site of the γ-secretase. Aspartyl protease inhibitors, such as L-685,458 [
27,
28] and difluoro ketones [
29], were found not only to inhibit γ-secretase activity but also to locate or extract presenilins from cell homogenates by approaches including photoaffinity crosslink and affinity chromatography.
PS1 has a putative topology of nine-transmembrane domain (TM1 to TM9), and is matured and activated by cleavage between TM6 and TM7 by a highly regulated endoproteolytic cellular process that leaves behind tightly associated N- (NTF) and C-terminal fragments (CTF) [
30]. Two conserved aspartic acids, D257 in TM6 and D385 in TM7, may potentially be positioned to face each other in an aqueous cavity within the lipid bilayer, and form the catalytic core [
31]. Although mutation of either of these two conserved aspartic residues PS1 and PS2 impairs γ-secretase activity [
32], they lack the two DTG (or DSG) triplets that form the catalytic pocket found in a typical aspartyl protease such as β-secretase [
33] or HIV protease [
34]. Similar to Presenilins, human signal peptide peptidase (SPP) possesses a pair of aspartic acids [
35]. Moreover, SPP, PS1 and PS2 share highly conserved YD257 and GXGD385 (PS1 numbering) motifs and constitute a unique family of aspartyl protease [
36].
While mutagenesis of residues surrounding the GXGD motif is reported [
31,
37,
38], the sequence and structural contexts following these two essential aspartyl proteases are not understood. In this study, we artificially introduced either one DTG triplet or both DTG triplets in PS1 and asked whether the DTG triplet could alter γ-secretase catalytic property due to a structural resemblance to a classical aspartyl protease. We found that neither the single D385TG nor the double triplet D257TG/D385TG were able to convert PS1 into a classic aspartyl protease. However, D385TG in particular, dramatically elevated Aβ
42 production, reinforcing the importance of residues surrounding D385 in the γ-secretase catalytic activity. More interestingly, PS1-D385TG significantly depressed the Notch S3-cleavage in releasing NICD, suggesting that the cleavage of APP at the γ-site and S3-site in Notch are differentially regulated. We also found that, unlike wt PS1 and other PS1 familial mutants, PS1-D385TG did not exhibit a biphasic profile on processing of APP in response to the inhibition by the γ-secretase inhibitors, L-685,458 or fenchylamine. We, therefore, propose a possible model in which PS1-D385TG adopts a conformation, within the γ-secretase complex, that differs from that of wt-PS1 and favors the production of Aβ
42 but not NICD. This is the first demonstration that a mutation at F386I387 to TG causes such a dramatic effect in γ-secretase activity and inhibition. Because of its unique enzymatic profiles, PS1-D385TG will be valuable in a comparative structural exploration by approaches including electron microscope [
39] as well as in rational design of drugs that specifically inhibit Aβ production.
Discussion
The triplets D257LV and D385FI, found in human PS1 TM 6 and 7, respectively, are conserved across species, and mutations of D257 and/or D385 to Ala result in essentially a complete loss of activity [
26], suggesting that PS1 may function like an aspartyl protease. In addition, PS1 shares the YD257 and GxGD385 motifs with a unique family of aspartyl proteases referred to as signal peptidases (SPP) [
35]. While the role of the amino acids comprising the YD257 and GxGD385 motifs have been studied, the couple of amino acids following D385, in particular F386, have not been tested. Since the two conserved catalytic aspartic acids D257 and D385 are not in the context of the two DTG (or DSG) catalytic triplets of a classical aspartyl protease [
51], we artificially replaced L258V259 and/or F386I387 with TG in PS1 to generate two PS1 mutants D257TG and D385TG, and the double-mutant D257TG/D385TG; we then investigated the effect of these mutations on γ-secretase catalytic activity in the presence or absence of γ-secretase inhibitors. Our data shows that expression of the PS1-D257TG in a stable cell line Swedish APP did not significantly alter the production of Aβ
40 or Aβ
42 (data not shown). However, when the same experiment was carried out with PS1-D385TG or PS1-D257TG/D385TG, levels of Aβ
42 were increased at least 2.5 fold compared to that with wt PS1 whereas Aβ
40 remained largely unchanged. It appears that the mutation of F386I387 to T386G387 greatly favors the formation of Aβ
42.
Equally interestingly, the F386I387 to T386G387 mutation results in a significant reduction in processing of Notch at the S3 site APP ε-site (ITL49↓V50ML) (Chen, 2001; Weidemann, 2002). Indeed, our preliminary experiments based on a reporter-depending assay suggested that both PS1-D385TG and PS1-D257TG/D385T also reduced processing at the APP ε-site (data not shown); yet, this latter observation needs to be fully validated. It appears that the mutation of F386I387 to T386G387 greatly favors the formation of Aβ42 and decreases the cleavage at both APP ε- and Notch S3-site.
We have shown that this enhanced production of Aβ
42 in PS1-D385TG is not due to the conversion of PS1 into a classic aspartyl protease because the mutated triplet D385TA still increases production of Aβ
42 to a certain extent (Table
3). In light of these and other findings (Tolia et al, 2006;Yamasaki et al., 2006), it appears that mutating I387 may have a minor effect on the γ-secretase activity. Instead, the effect is driven by the F386 to T386 mutation. The presenilin and signal peptidase consensus motif GxGD has been studied with regard to its role in the context of the γ-secretase activity. Interestingly, the clinical PS1 G384A mutant increased Aβ
42 production by a factor of 6, but had no significant effect on Notch processing [
52]. On the contrary, phenylalanine at position × of the GxGD385 motif in PS1 is suggested to be responsible for altering Notch processing [
38]. Together with our findings that the mutation of F386 to T carries most of the responsibility for the observed change in substrate preference, we support the notion that the sequences surrounding the D385 may be critical in determining γ-secretase substrate recognition and docking, enzyme specificity and substrate cleavage rate. Although D257 is also implicated as a critical residue in cleaving APP at the γ-secretase site, PS1-D257TG has less effect on the elevation of Aβ 42 and reducing Notch cleavage in our experiment. This weak effect is likely due to the presence of wild type PS1 in our cell lines as suggested by others [
53].
Consistently, PS1-D257TG and PS1-D385TG mutants were also affected differently by γ-secretase inhibitors in our studies when compared to the wild type or clinical mutants. Both wild type and most PS1 variants showed biphasic curves, stimulation at low inhibitor concentration followed by inhibition at higher concentration, to the γ-secretase inhibitor L-685,458 while the PS1-D385TG mutants showed normal dose-dependent inhibitory curves within the test range. This stimulation/inhibition behavior has been reported for a variety of apparently unrelated inhibitors, including aspartyl protease transition state inhibitors, both in cells and in cell-free assay of γ-secretase activity [
54‐
57]. Indeed, increases of Aβ
42 up to 8-fold were observed at the peak elevation, and Aβ 42 constituted as much as 50% of the total secreted Aβ [
54]. Among the various mechanisms suggested to justify this phenomenon there is one involving an allosteric modulation of the γ-secretase complex by a process similar to that responsible for the increase of Aβ 42 production by FAD mutations in presenilins, possibly related to structural changes; and another one related to inhibitor binding to either PS allosteric sites or the cell membrane which would cause structural changes in the preseniln structure to favor Aβ 42 production prior to reaching inhibitory concentrations. No matter which stimulation/inhibition mechanism is operative, our PS1-D385TG mutant distinguishes itself from wild type and other mutants in that this mutation has changed the enzyme structure sufficiently to have a significant effect in its ability to recognize and cleave substrates and be inhibited. On the other hand, the mutation leading to PS1-D257TG has much less influence on inhibition than PS1-D385TG; yet it may be able to induce some minor structural changes since it is inhibited by lower concentrations of inhibitor than the wild type enzyme (data not shown).
It has been reported that the γ-secretase is composed of one of each of the following proteins: presenilin, nicastrin, aph1 and pen2 [
58]. If this stoichiometry is correct, our results would favor a model which comprises a γ-secretase catalytic site with D257 and D385 as catalytic residues and F386 involved in substrate docking with other residues possibly from the GLGD385 motif. However, a different stoichiometry involving two molecules of presenilin per γ-secretase complex has been reported (Clarke et al. 2006; Schroeter et al, 2003). If the latter stoichiometry is true, an additional model could also be possible where two C-terminal fragments or two N-terminal fragments form a homodimer with two D385s or two D257s, respectively, as catalytic residues. Based on the larger effect exhibited by mutations around D385 than those produced by mutations around D257, we suggest that the mutations F386I387 to T386G387, particularly F386 to T, following the catalytic residue D385 in TMD7 of PS, are involved in APP/Notch substrate selection and in substrate docking at the active site of γ-secretase.
Earlier work [
48] showed that, in cultured cells, certain NSAIDs were able to decrease Aβ
42 secretion with concomitant increase in the Aβ
38, suggesting a correlation between Aβ
42 decrease/increase and Aβ
38 increase/decrease. Recent work [
59] has examined 10 clinical mutants of PS1 and shown that Sulindac Sulfide could increase the formation of Aβ
38 without affecting Aβ
42. Yet, these investigators observed that wt PS1 and even more PS1 M146L had a significant decrease of Aβ
42 and increase of Aβ
38 upon treatment with 50 μM sulindac sulfide. These latter observations are consistent, at least qualitatively, with those of the present study. Specifically, we show (Figure
4) that Sulindac Sulfide treated wt PS1, PS1 M146V, and PS1-D385TG, result in less Aβ
42 production with a significant Aβ
38 increase. Yet, while the two former proteins have a complete competence to produce Aβ
42, the latter one has totally lost its ability to form Aβ
39, and only partially that of making Aβ
42 and Aβ
40. The nature of these differences might be related to induction of conformational changes on these PS1 molecules by Sulindac Sulfide treatment [
60] and are apparently different than those induced on most of the clinical mutants explored in the previous studies as reflected by the different effects on Aβ production [
59].
In summary, we have, for the first time, demonstrated that a mutation at F386I387 to TG in PS1 causes a dramatic effect on γ-secretase activity with respect to higher productions of Aβ42 compared to wt PS1, and differential proteolytic cleavages of its two studied substrates by increasing cleavage of APP after residue 42 (yielding Aβ42) while decreasing the cleavage at the Notch S3-site. The knowledge gained from this study provides useful information in rational design and in the development of a protocol to screen compounds that only block the γ-secretase activity toward APP to release Aβ but not the other substrates. PS1-D385TG can be used as a tool for comparative studies of structure and conformation in the γ-secretase complex.
Competing interests
The authors declare that they have no competing interests.