Dimerization of the transmembrane domain of amyloid precursor proteins and familial Alzheimer's disease mutants
© Gorman et al. 2008
Received: 31 July 2007
Accepted: 30 January 2008
Published: 30 January 2008
Skip to main content
© Gorman et al. 2008
Received: 31 July 2007
Accepted: 30 January 2008
Published: 30 January 2008
Amyloid precursor protein (APP) is enzymatically cleaved by γ-secretase to form two peptide products, either Aβ40 or the more neurotoxic Aβ42. The Aβ42/40 ratio is increased in many cases of familial Alzheimer's disease (FAD). The transmembrane domain (TM) of APP contains the known dimerization motif GXXXA. We have investigated the dimerization of both wild type and FAD mutant APP transmembrane domains.
Using synthetic peptides derived from the APP-TM domain, we show that this segment is capable of forming stable transmembrane dimers. A model of a dimeric APP-TM domain reveals a putative dimerization interface, and interestingly, majority of FAD mutations in APP are localized to this interface region. We find that FAD-APP mutations destabilize the APP-TM dimer and increase the population of APP peptide monomers.
The dissociation constants are correlated to both the Aβ42/Aβ40 ratio and the mean age of disease onset in AD patients. We also show that these TM-peptides reduce Aβ production and Aβ42/Aβ40 ratios when added to HEK293 cells overexpressing the Swedish FAD mutation and γ-secretase components, potentially revealing a new class of γ-secretase inhibitors.
Currently, almost 4.5 million individuals in the United States have Alzheimer's disease (AD) and this number is expected to increase to approximately 16 million by 2050 . Among these millions, 5–10% are a heritable form of the disease called familial AD (FAD). A subset of FAD is caused by mutations in the gene encoding for amyloid precursor protein (APP). APP degradation results in the formation of senile plaques, which are dense extracellular deposits localized to the limbic and association cortices and composed mainly of two amyloid peptides (Aβ40 and Aβ42) that are produced through proteolytic processing by β and γ secretases . First β-secretase cleaves the APP extracellular juxtamembrane domain, and then γ-secretase cleaves the APP transmembrane domain (TM) either after Val711 or after Ala713 producing Aβ40 and Aβ42, respectively. In normal individuals APP processing leads to low Aβ42/40 ratios, while individuals with FAD mutations in the APP-TM domain have an increased Aβ42/40 ratio and therefore an increased proportion of the more neurotoxic Aβ42 .
Several lines of evidence indicate that APP dimerizes in its native membrane environment. Chemical crosslinking has indicated that APP can dimerize and use of an obligate APP dimer (with disulfide-linkage at the TM domains) shows that dimeric APP is efficiently cleaved by γ-secretase . The APP-TM domain contains the motif GXXXG/A that is known to mediate dimerization of transmembrane helices in oligomeric membrane proteins . Given that more than half of the mutations in APP that cause FAD are localized to the TM domain , we hypothesized that these mutations are capable of perturbing the dimerization and this leads to the increased Aβ42/40 ratio seen in FAD.
To test this hypothesis, we have examined synthetic peptides corresponding to TM segments of APP (APP-TM peptides) in detergent micelles and phospholipid bilayers. We have found that this domain is capable of dimerization, and FAD mutations within the APP-TM domain affect the dimerization propensity. We find that the dimerization equilibrium constants of the APP-TM variants measured in vitro are correlated to two well known clinical pathological features of AD patients. The constants are correlated to the average age of onset of AD symptoms and to the Aβ42/Aβ40 ratios observed in FAD. We demonstrate that the addition of our synthetic APP-TM peptides to cell culture models reduces the production of both Aβ40 and Aβ42. We propose that APP-TM mutations increase monomerization, which leads to the increased Aβ42/40 ratios seen in FAD cell culture models.
Location of FAD mutations within the APP-TM sequence
FAD mutations in the APP-TM domain
FRET analysis of wild-type APP-TM peptide and FAD mutant derivatives also produces the signature hallmarks of FRET (i.e. donor quenching and acceptor sensitization). This indicates that wild-type and mutant peptides oligomerize in both SDS detergent micelles (data not shown) and phosphatidylglycerol bilayers (Fig. 2). SDS-PAGE showed that V717G populated both monomer and dimer states under conditions where wild-type formed stable dimers (Fig. 1B), suggesting that the FAD mutations reduce the stability of APP-TM dimers.
We examined whether the addition of APP-TM peptides to cells expressing APP interferes with the production of Aβ40 and Aβ42. As it has been shown that cationic peptides, such as those derived from the HIV-TAT protein, can be taken up spontaneously by cells [13, 14], we reasoned that the cationic APP-TM peptides may also be taken up by similar mechanisms. Once inside the cell, these peptides, which have identical TM sequences to endogenous APP, could potentially interfere with the γ-secretase activity via a competitive inhibitor mechanism and alter production of Aβ42 and Aβ40. HEK293 cells overexpressing APP and γ-secretase components were incubated with APP-TM peptides and the conditioned media was assayed for Aβ42 and Aβ40 levels using a commercial assay (BioSource). The APP-TM peptides, which share sequence identity with Aβ42 and Aβ40, do not interfere in this assay (see Materials and Methods). Only wild-type and V717G peptides could be examined in this experiment because the other mutant peptides, V717F and V714I, were only soluble in the presence of SDS, which is toxic to cells.
Our studies indicate that APP dimerization propensity, measured in vitro, is correlated with two previously identified in vivo pathological features of FAD: the Aβ42/40 ratio and the age of onset of AD symptoms. While our biophysical studies indicate that APP-TM peptides form dimers that are sensitive to APP-FAD mutations, validation of the predicted structure will require high-resolution techniques, such as NMR spectroscopy. However, when the locations of FAD mutations in the TM domain of APP are mapped onto our predicted structure of the APP-TM an intriguing feature arises. Out of the 11 known FAD mutations in the APP-TM domain, 7 occur within the predicted dimer interface and the remaining 4 are adjacent to the interface (Table 1) [8, 16–22]. These mutations are ideally located to disrupt APP dimerization. Our experiments on FAD mutant derivatives indicate that such mutations can in fact result in loss of APP-TM dimerization (Fig. 4). The dissociation constants of our APP-TM peptides may differ from full-length APP; however, what is important is the relative difference between the wild type dissociation constant and that of the FAD APP-TM variant. Previoulsy, Aβ42/Aβ40 ratios determined using primary culture neurons were compared to the mean age of onset and found to exhibit a strong inverse correlation . Here we compared our dissociation constants to these two important clinical features of AD. Interestingly, our in vitro Kd measurements are directly correlated to the Aβ42/Aβ40 ratios and the in vitro Ka measurements are equally correlated to the mean age of disease onset. These correlations suggest that the relative production of Aβ42 increases as the equilibrium between dimeric and monomeric APP shifts toward monomer. The exact mechanism responsible for causing the shift in Aβ42/Aβ40 ratios between wild-type and FAD mutants is currently unclear. It is likely, however, that the preferential cleavage at either the Aβ40 or Aβ42 sites is related to the process by which the γ-secretase complex recognizes monomeric versus dimeric APP.
There is evidence that APP dimers are physiologically relevant as it has previously been shown that APP dimers exist in vivo and that these dimers are substrates for β- and γ-secretase. Scheuermann et al  have provided direct biochemical evidence for the presence of APP dimers in cellular environments using chemical cross-linking experiments on full-length and deletion mutants of APP. They identified two extracellular domains in APP that mediate dimerization. A third dimerization domain was detected and its location was determined to be between the start of the TM domain and the end of APP at the C-terminus. It is likely that the APP-TM domain studied here represents the precise location of this third dimerization domain. The X-ray crystal structure of the extracellular E2 domain of APP  demonstrates that it is dimeric and likely corresponds to the first extracellular dimerization domain detected by Scheuermann et al . Scheuermann et al. also developed a site-directed obligate dimer of APP where engineered Cys residues placed at the N-terminal of the TM domain tether together two monomers with a disulfide bond. When expressed in cells, this obligate APP dimer is efficiently cleaved by β- and γ-secretases to form Aβ. There is, therefore, significant biochemical data that indicates that APP dimers exist in vivo and are substrates for β- and γ-secretases.
Dimerization appears to be a common theme among several membrane proteins involved in Aβ metabolism and AD. Both BACE , the membrane-bound protease responsible for β-site cleavage of APP, and presenilin 1 , the catalytic component of γ-secretase, are dimeric in their native membrane environments. While the dimeric nature of these proteases does not provide evidence for APP dimerization, this common theme is nevertheless noteworthy.
Recently, Marchesi proposed a novel mechanism of Aβ toxicity that involves homo- and hetero-dimerization of α-helical membrane segments . Marchesi proposes that a certain fraction of Aβ molecules generated by β- and γ-secretases remain inserted in the membrane bilayer in a helical conformation and can utilize its GXXXG/A motif to form homodimers or heterodimers with other GXXXG/A membrane proteins, such as APP or Aph1. It is speculated that accumulation of homo- and hetero-dimers of Aβ interferes with biochemical processes that occur in the membrane including receptor, channel, and secretase activities. This interference may lead to downstream events that culminate in eventual neuronal dysfunction. While Marchesi's proposal and the hypothesis reported here are fundamentally different, they both invoke the contribution of homo- and hetero-typic interactions of membrane proteins in AD pathology. The heterotypic association of wild-type and mutant APP-TM peptides that we have observed (Fig. 5) reinforces the idea that such interactions could occur with other molecules, such as Aβ and Aph1.
The addition of APP-TM peptides to Aβ-producing HEK293 cells causes dramatic reductions in Aβ production, and intriguingly, the V717G mutant caused significantly greater inhibition of Aβ production than the wild-type peptide (Fig. 7). While elucidation of the mechanism of inhibition requires further study, several feasible possibilities arise involving the interaction of the peptides with the γ-secretase complex. The γ-secretase complex can potentially interact with monomeric or dimeric forms of the APP-TM peptides or form heteromeric complexes of peptide with endogenous APP. Both monomeric and homodimeric forms of the APP-TM peptides could act as competitive inhibitors of γ-secretase leading to the observed accumulation of γ-secretase substrates (i.e. α- and β-stubs) and decrease in γ-secretase products (i.e. Aβ). Mutational and chemical crosslinking studies are underway to test these hypotheses. It should be noted that peptides analogous of the APP-TM domain that are too short to span the plasma membrane have also been shown to inhibit γ-secretase activity .
We have provided in evidence using a cell culture model that the extent of APP dimerization can determine the Aβ42/40 ratio and disruptions of dimerization induced by at least one FAD mutation in the APP-TM domain increases the Aβ42/40 ratio. These studies have to be replicated in vivo to substantiate the hypothesis that dimeric APP causes preferential production of the less neurotoxic Aβ40. A detailed understanding of the pathological and physiological processing of APP would be a valuable asset in the development of novel therapeutic strategies, such as small molecule stabilizers of the dimeric form of APP, aimed at controlling Aβ production.
Wild-type: Lys-tag(W/EEDANS)GAIIGLMVGGVVIATVIVITLVML Lys-tag
V717G: Lys-tag(W/EEDANS)GAIIGLMVGGVVIATVIGITLVML Lys-tag
V717F: Lys-tag(W/EEDANS)GAIIGLMVGGVVIATVIFITLVML Lys-tag
T714I: Lys-tag(W/EEDANS)GAIIGLMVGGVVIAIVIGITLVML Lys-tag
The sequences of the APP-TM peptides are shown above, and the central block of residues given in the one-letter code correspond to residues 700 to 723 of human APP and contain the TM segment. All peptides contain N- and C-terminal oligo-lysine tags, which have been shown previously to increase aqueous solubility of TM segments . Two versions of each peptide were synthesized. The first version contained a Trp residue that followed the N-terminal Lys-tag; the second version differed from the first in that a Glu residue with the fluorescent group EDANS covalently linked to its sidechain (Glu-EDANS; Molecular Probes) was used instead of the Trp residue. The fluorescence of the Trp residue and the EDANS group were used to follow dimerization of the peptides through fluorescence experiments discussed below.
All peptides were prepared by solid-phase synthesis on a PerSeptive Biosystems 9050 Plus peptide synthesizer as previously described . The peptides were cleaved with 88% trifluoroacetic acid/5% phenol/5% distilled H2O/2% tri-isopropylsilane (v/v). Cleaved peptides were precipitated with ice-cold diethyl ether, dissolved in distilled H2O and lyophilized. Crude peptide powder was dissolved in water and purified by reverse phase C4-HPLC. Mass spectroscopy was used to confirm the molecular weight of the purified peptide. The stock peptides solutions were stored in distilled H2O at 4°C.
Detailed modeling procedures for TM helix oligomerization are described elsewhere [6, 30]. The APP-TM sequence was built into uniform α-helices with backbone torsion angles of Φ = -65° and ψ = -40°. Sidechain rotamers were chosen using the backbone-dependent rotamer library program SCWRL . The Monte Carlo (MC) search procedures and parameters for potential helix packing were described previously . Homo-dimeric APP-TM models were selected from four hundred independent MC simulations. The selected model structures were clustered by Cα RMSD using the NMRCLUSTER program . The median model from the largest cluster was selected as our final predicted structure.
DOPG vesicles with incorporated peptides were prepared via sonication of dried lipid films in the presence of peptide solution at a molar ratio of 1 mole of peptide: 20 moles of DOPG as previously described . The samples contained 50 mM Tris, 10 mM NaCl (pH 7.0) as the buffer. Gel filtration experiments on these samples indicated that the majority of the peptides were incorporated into the vesicles and absorbance measurements indicated that similar amounts of donor- and acceptor-labeled peptides were incorporated. Purified peptide-incorporated vesicle fractions were used for fluorescence measurements.
Mixed peptide-SDS micelles were prepared by the addition of concentrated SDS to peptide solutions, such that the final concentrations were 20 mM SDS, 40 mM Tris, 8 mM Acetate (pH 7.0) and 0 – 90 μM peptide. The samples were sonicated and incubated overnight in the dark at room temperature before making fluorescence measurements.
Fluorescence measurements were made at room temperature using a PTI QM-1 fluorescence spectrophotometer. A quartz cuvette with a 2 mm excitation pathlength and a 1 cm emission pathlength was used. The excitation wavelength was 285 nm and the bandpass was 4 nm. We obtained a quantitative measure of the degree of peptide oligomerization by integrating the EDANS emission signal from 472–496 nm and dividing this value with the integrated Trp emission signal from 350–375 nm. This ratio of acceptor fluorescence to donor fluorescence is denoted either as Acceptor/Donor ratio or A/D ratio and can be used to compare peptide oligomerization at varying concentrations.
Samples contained 20 mM SDS, 40 mM Tris, 8 mM Acetate (pH 7.0), 5% glycerol, and 0 – 25 μM peptide. They were boiled for 5 minutes and subjected to SDS-PAGE using precast 4–12% Nupage gels (Invitrogen) and a standard running buffer. SeeBlue®Plus2 molecular weight markers from Invitrogen were used as standards.
Circular Dichroism (CD) spectra were recorded on an Aviv-62DS CD-spectrometer at 25°C. Spectra were obtained from 260 nm to 198 nm with a 1.0 nm step size, a 1 nm bandwidth, a 1 mm pathlength and an averaging time of 8 seconds. The peptide concentration was 80 μM.
Where AT = total APP-TM peptide concentration, F = A/D ratio, fraction dimer = (Fobs-Fmin)/Fmax-Fmin), and % dimer = 100% (fraction dimer). The goodness-of-fit to the model were assessed using the chi-squared test, nonlinear least squares correlation, and examination of residuals.
Stably transfected HEK293 cells overexpressing the Swedish double mutant (K670N/M671L) of APP751, presenilin 1, nicastrin, PEN-2 and APH-1 were grown to confluency in DMEM media containing 10% fetal bovine serum . The cells were grown for another 48 hours with the media changed twice at equivalent time intervals. Cells were then incubated overnight in fresh media containing 0, 20 or 100 μM APP-TM peptides. The media was collected and subjected to an immunoassay to determine the Aβ42 and Aβ40 concentrations. The ELISA BioSource immunoassay human β-amyloid 1–40/1–42 colorimetric kit (catalogue number KHB3482/KHB3481) was used and the standard protocol was followed. This sandwich ELISA technique makes use of an immobilized Aβ antibody that binds both Aβ40 and Aβ42 at an N-terminal epitope. The bound Aβ40 and Aβ42 are then detected using a C-terminal antibody specific for either Aβ40 or Aβ42. The APP-TM peptides, which share sequence identity with both Aβ40 and Aβ42, do not interfere in this analysis because they lack the N-terminal epitope responsible for capturing Aβ40 and Aβ42 from the media. Therefore, this assay is ideally suited to detect any disturbances in the normal processing of endogeneuous full length APP in the presence of APP-TM peptides.
Sulforhodamine B (SRB) cytotoxicity assay was used to assess whether the APP-TM peptides possessed any intrinsic toxicity. Stably transfected HEK293 cells overexpressing Swedish-mutant APP751 and γ-secretase components were plated at 15,000 cells/well in DMEM media containing 10% fetal bovine serum. The cells were grown for another 48 hours with the media changed twice at equivalent time intervals. Cells were then incubated overnight in fresh media containing 0, 10, 20, and 100 μM APP-TM peptides, and the SRB cytotoxicity assay was performed as previously described .
We thank Nancy F.L. Ng, Elyssa Elton and Rishi Rakhit for helpful comments. Funding was received from Canadian Institutes of Health Research, ALS Society (Canada) and Muscular Dystrophy Association (Canada) (A.C.); NIH Grant GM063919 (J.U.B.); Alzheimer Society (Canada) (P.M.G.)
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.