Background
γ-Secretase is responsible for cleavage of a number of type I membrane proteins, including amyloid precursor protein (APP) and Notch, and is comprised of presenilin 1 or 2, Aph1, Pen2 and Nicastrin [
1‐
5]. Proteolytic processing of APP by β- and γ-secretases results in production of amyloid β (Aβ) peptides. The major Aβ species are 40 and 42 amino acid long peptides, the latter of which is recognized as the more toxic species involved in Alzheimer's disease (AD) pathogenesis [
2,
6,
7]. Factors that modulate the Aβ
42/40 ratio can be classified into at least two categories; 1) substrate-targeting manipulations, such as FAD-linked mutations within the intramembranous region of the APP substrate [
8,
9], and 2) γ-secretase-targeting modifications, such as FAD-linked PS1 mutations [
10‐
12], Pen-2 N-terminus modification [
13] or expression of different Aph1 isoform [
14]. In addition, treatment with pharmacological agents, γ-secretase modulators (GSMs), could alter the Aβ
42/40 ratio [
15‐
17]. However, there is a controversy whether the primary target of these compounds is APP substrate [
18‐
20], or PS1/γ-secretase [
21‐
25].
Using Förster resonance energy transfer (FRET)/fluorescent lifetime imaging microscopy (FLIM) technique, we have previously demonstrated that PS1, a catalytic site of γ-secretase, could exist in a "closed" (close proximity between the PS1 N-terminus, C-terminus, and a large cytoplasmic loop domain) and "open" (longer distance between them) conformations [
26‐
28]. Although the detailed molecular mechanism responsible for different PS1 conformational states, and underlying the precision of APP cleavage by PS1/γ-secretase is currently unknown, we found that the "closed" conformation of PS1 is consistently linked to a higher Aβ
42/40 ratio, whereas the "open" conformation is associated with a lower Aβ
42/40 ratio [
26‐
28].
Notably, in addition to manipulations directly targeting components of the γ-secretase complex, mutations within the transmembrane region of the APP substrate have been shown to induce changes in the PS1 conformation. For example, APP with FAD-linked V717I or I716F mutations that increase the Aβ
42/40 ratio seem to associate with the PS1/γ-secretase earlier in the secretory pathway, alter the alignment of APP with PS1, and shift PS1 into a "close" NT and CT proximity conformation [
29]. Conversely, the APP V715F substitution, which dramatically decreases the Aβ
40 and Aβ
42 while increasing Aβ
38 levels, induced a structural rearrangements in PS1 reminiscent of that observed after the treatment with Aβ
42 -lowering non-steroidal anti-inflammatory drugs ("open" conformation) [
30].
Based on these findings, we hypothesized that APP-targeting manipulations may alter conformation of the APP molecule or it's positioning within the plane of the membrane. This alteration may change APP substrate presentation to the PS1/γ-secretase, and consequently induce a shift in PS1 conformation. Since previous structural analysis predicts that the APP cytoplasmic domain can associate with the membrane and alter its positioning in response to various stimuli [
31], in the current study we analyzed proximity between the membrane and APP-CT as readout of APP transmembrane positioning. Thus, to better understand the relationship between APP CT transmembrane positioning, PS1/γ-secretase conformation, and the Aβ
42/40 ratio, we asked whether APP-CT proximity to the membrane correlates with the Aβ
42/40 ratio and can be affected by the PS1 conformational change. The FLIM assay was utilized to monitor relative distance between the two fluorophores labelling membrane and APP-CT in intact cells. We found that FAD mutations within the APP transmembrane domain, that raise the Aβ
42/40 ratio, increase proximity of the APP-CT to the membrane. Interestingly, treatment of cells with GSMs, which are known to modify the Aβ
42/40 ratio and induce PS1 conformational change [
24,
26,
28], led to altered APP-CT and membrane proximity only in the presence of PS1/γ-secretase. Surprisingly, we found that Aβ
42/40 ratio-raising FAD-linked mutations in PS1 also affect the positioning of APP relative to the membrane in a manner similar to that of the FAD-linked APP mutations. These results suggest a reciprocal relationship between conformation of the APP-CT and/or its orientation relative to the membrane and PS1 conformation. Thus, factors that modulate either APP positioning or PS1 conformation could be exploited therapeutically to correct pathogenic Aβ
42/40 ratio, and thus prevent or slow down progression of AD.
Discussion
The cytoplasmic domain of APP is believed to function in multiple signalling pathways ranging from apoptosis to gene transcription regulation [
35]. APP intracellular domain has been shown to interact with various molecules and contribute to axonal transport [
36], neurite outgrowth and arborization [
37], and signalling events in the cell [
38,
39]. Thus, alterations in APP-CT structure in pathological settings could ultimately interfere with these events and accelerate neuropathological changes. A potential relationship between the APP-CT conformation and Aβ production has been suggested. For example, structural studies of the APP-CT have demonstrated that although APP-CT does not adopt a stable folded conformation, it has a transient preordered structure, whose conformation can be altered by phosphorylation [
40,
41]. In addition, a recent structural model predicts that APP-CT might be associated with the membrane, and suggests that APP-CT association with and dissociation from the membrane might regulate interactions of APP with various proteins [
31], and could contribute to an altered Aβ
42/40 ratio. A new γ-secretase activating protein has been recently described that interacts with both PS1 and APP C-terminal fragment (but not with the Notch substrate), affects Aβ production, and may alter the structural relationship between γ-secretase and APP CT [
42].
A model in which there is successive release of tri-peptides has been proposed for differential production of the Aβ
40 and Aβ
42 species [
43,
44]. According to this model, Aβ
49 produced by PS1/γ-secretase dependent ε-cleavage of APP at the membrane-cytosol interface is converted to Aβ
40 after successive release of tri-peptides, whereas initial ε-cleavage of APP at the Aβ
48 site is converted to the Aβ
42. In the present study, we employed the FLIM assay in intact cells to demonstrate that APP-CT positioning relative to the membrane, or a conformational change of the APP/C99 cytoplasmic domain, correlates with changes in the Aβ
42/40 ratio. We found that Aβ
42/40 ratio-raising mutations in PS1 or in the APP transmembrane region altered APP positioning within the membrane by bringing the APP C-terminus closer to the membrane. Thus, a conformational change of the APP cytoplasmic domain, which we observed in the current study, may affect the initial APP cleavage at the ε-site by altering APP substrate presentation to PS1/γ-secretase at the membrane-cytosol interface.
Interestingly, we found that PS1/γ-secretase itself has a profound effect on APP-CT positioning relative to the membrane. First, our data indicate that in the absence of PS1/γ-secretase or when APP-PS1/γ-secretase interaction is inhibited by HP treatment, FRET between APP-CT-RFP and myrGFP-membrane is absent, suggesting that the distal part of the APP-CT is located relatively far away from the membrane. Surprisingly, it appears as if interaction of APP with the PS1/γ-secretase affects the orientation of APP-CT by bringing it into close proximity to the membrane (FRET present). Moreover, interactions with FAD mutant PS1/γ-secretase further change the positioning of the APP-CT to obtain an even closer proximity relative to the membrane. Although the precise mechanism of how mutations in PS1 affect APP positioning relative to the membrane is unknown, it is possible that in the process of APP substrate alignment with the topographically altered mutant γ-secretase active site, changes in the APP-CT membrane proximity occur. This is in agreement with a cross-linking experiment demonstrating that aggressive FAD-linked PS1 mutations cause alterations in topography of the γ-secretase active site [
45].
GSMs have been shown to affect both PS1 conformation [
13,
24,
26‐
28], as well as APP positioning in the membrane (current study). There remains an uncertainty over the primary target of GSMs, with some studies showing that GSMs target γ-secretase, either PS1 itself or other components, such as Pen2, [
21,
22,
42], whereas others propose that GSMs directly bind to the APP substrate [
18‐
20]. In our current study, we did not observe any effect of GSMs on APP-CT positioning in the absence of PS1 and PS2. However, we could not exclude the possibility that GSMs binding may have a subtle effect within the range of non-FRETing distance (>10 nm) from the membrane in PS1/2 dKO cells, thus rendering the APP-CT positioning change undetectable by our current method. It is also possible that some GSMs could still bind to APP CTF in the absence of presenilins but it requires the complex formation for the conformational shift to occur. We have recently reported that the modulatory effect of GSMs is implemented through the "allosteric site" located within the γ-secretase complex itself, although substrate docking to γ-secretase is needed to allow GSM access to this site [
24]. Thus, the most likely scenario is that these GSMs primarily target PS1/γ-secretase or the PS1/APP interface, and the change in APP positioning within the membrane is a secondary response to the change of PS1 conformation.
Methods
Cell Lines and Pharmacological treatments
PS1/PS2 double knockout (PS1/2 dKO) mouse embryonic fibroblasts (MEFs), and PS1/2 dKO cells reconstituted by stable expression of wild-type or FAD-linked PS1 (L166P, Delta9 and A246E) were a generous gift from Dr. Bart Destrooper [
46]. APP/APLP2 dko MEFs were generous a gift from Dr. Koo. Chinese Hamster Ovary (CHO) cells were obtained from ATCC. Cells were cultured with Opti-MEM (Invitrogen) supplemented with 5% fetal bovine serum. The cells plated into four-chamber slides were transfected with various constructs, and were subjected to microscopy (FRET, FLIM) analyses. To evaluate the effect of GSMs on APP, the cells were treated for 24 hours with either 100 μM fenofibrate or 400 μM ibuprofen. To inhibit the interaction between APP and PS1/γ-secretase, cells were treated for 24 hours with 100 nM helical peptide (a gift from Dr. M. Wolfe, BWH, Boston, MA), which was designed to mimic a portion of the APP transmembrane domain and competes with APP for binding to PS1/γ-secretase [
34]. Control cells were treated with a vehicle (either DMSO or ethanol).
Constructs
Human APP 695 isoform was tagged with either green or red fluorescent protein at its C-terminus to generate APP-GFP and APP-mRFP constructs, respectively. APP mutations (V717I, V717K, I716F) were inserted using Quick Change Site-Directed mutagenesis kit (Stratagene), according to the manufacturer's instructions. GFP with myristoylation signal at the N-terminus was described previously [
33]. EGFP-C3 empty vector (Clontech, Madison, WI) was used as a control for expression of EGFP without a membrane targeting signal. Wild-type PS1, as well as E318G and G384A mutant PS1 constructs were described previously [
27].
Fluorescence Lifetime Imaging Microscopy (FLIM)
FLIM was used as an approach to monitor proximity between the myrGFP labeled membrane and APP-CT-RFP. Briefly, cells expressing myrGFP only were used as a negative control to determine the baseline myrGFP lifetime. The degree of GFP donor lifetime shortening due to presence of FRET was used as an indicator of the proximity between the GFP donor and RFP acceptor fluorophores in myrGFP and APP-RFP cotransfected cells. The FLIM software uses the Levenberg-Marquardt algorithm to fit the raw data from each pixel to two-exponential fluorescence decay curves to record presence of a shorter than baseline GFP lifetime (for details please see [
24,
27].
Alternatively, the plasma membrane was labelled with CM-DiI (Molecular Probes) as an acceptor fluorophore in the FLIM assay, and GFP-donor was fused to the APP CT. In this case, cells expressing APP-CT-GFP but not labelled with the CM-DiI served as a negative control. To label the membrane, cells were incubated with 1 μg/ml of CM-DiI dissolved in PBS for 15 minutes at 4°C and fixed with 4% paraformaldehyde, prior to the FLIM analysis. Data analysis was performed using SPC Image (Becker&Hickl, Berlin, Germany), in which donor fluorophore lifetimes are determined by fitting the data to one (negative control) or two (experimental conditions) exponential decay curves. In two component analysis, GFP lifetime (negative control with no acceptor fluorophore) is monitored first, and its value is "fixed" as t1 lifetime. The second, shorter lifetime representing FRET is calculated by the system as t2 value. This t2 value was used for comparisons between different experimental conditions. Thus, "non-FRETing" component (t1 lifetime representing APP molecules that do not interact with PS1, and thus position in the membrane in such a way that does not support FRET) is excluded from the lifetime comparisons.
ELISA
To measure the effect of APP mutations on Aβ production, wild-type APP-RFP, or APP-RFP, with V717I, V717K or I716F mutations were transfected into CHO cells cultured in 35 mm dishes. 6 hours after the transfection, culture medium was exchanged with 1 ml of fresh OPTI-MEM with 1% FBS, and cells were grown for an additional 24 hours. The conditioned medium was subjected to ELISA analysis using human β-amyloid (1-40 and 1-42) ELISA kit (WAKO, JAPAN), according to the manufacturer's instruction.
Statistical analysis
StatView for Windows, Version 5.0.1 (SAS Institute, Inc) was employed to perform statistical analysis using Fisher's PSLD analysis of variance (ANOVA). Samples were considered significantly different at p < 0.05.
Acknowledgements
We are grateful to Dr. Bradley T. Hyman for helpful discussions, to Ms. Mary Banks for technical assistance, to Drs. Koo (UCSD) and DeStrooper (VIB, Belgium) for PS1/2 dko and APP/APLP2 dko MEF cells, respectively; and to Dr. Wolfe (BWH, Boston) for HP. This work was supported by AG026593 and AG15379 (OB).
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KU and PJ carried out the FLIM study. KU analyzed the data and drafted the manuscript. KCF carried out the ELISA, NNG performed western blotting. OB conceived the study, and participated in its design, discussion of the data, and helped to write the manuscript. All authors read and approved the final manuscript.