Acceleration of Amyloid Fibril Formation by Specific Binding of Aβ-(1–40) Peptide to Ganglioside-containing Membrane Vesicles*

The interaction of Alzheimer’s Aβ peptide and its fluorescent analogue with membrane vesicles was studied by spectrofluorometry, Congo Red binding, and electron microscopy. The peptide binds selectively to the membranes containing gangliosides with a binding affinity ranging from 10−6 to 10−7 m depending on the type of ganglioside sugar moiety. This interaction appears to be ganglioside-specific as under our experimental conditions (neutral pH, physiologically relevant ionic strength), no Aβ binding was observed to ganglioside-free membranes containing zwitterionic or acidic phospholipids. Importantly, the addition of ganglioside-containing vesicles to the peptide solution dramatically accelerates the rate of fibril formation as compared with that of the peptide alone. The present results strongly suggest that the membrane-bound form of the peptide may act as a specific “template” (seed) that catalyzes the fibrillogenesis processin vivo.

One of the histopathological hallmarks of Alzheimer's disease is the presence of insoluble amyloid deposits within the gray matter regions of the brain and the vascular walls of cerebral blood vessels (1). The principal component of these deposits is the ϳ4-kDa amyloid ␤ peptide (A␤), a product of proteolytic processing of a much larger amyloid precursor protein (2). While biological functions of A␤ are still poorly understood, rapidly accumulating evidence points to a causative (rather than merely consequential) role of the peptide in the pathogenesis of Alzheimer's disease. Such a causative link between A␤ and Alzheimer's disease is indicated by genetic studies which identified specific mutations in amyloid precursor protein (in close proximity to the amino or carboxyl terminus of A␤ or within the A␤ region) that are tightly linked to heritable forms of Alzheimer's disease (3)(4)(5). Further support is derived from in vitro studies which show that synthetic A␤ peptide is toxic to neuronal cells in culture (6 -9). However, despite recent important advances, the molecular mechanisms of A␤-induced neuronal cell death remain largely unknown.
To understand the neurotoxic action of A␤, it is essential to identify specific cellular components that interact with the peptide and mediate a biological response of the affected cells. A likely primary target of A␤ is the neuronal plasma membrane. Indeed, a rapidly growing number of observations indicate that the peptide may alter important physical and biological properties of the membrane (10 -17). The mechanisms of A␤-membrane interactions remain, however, elusive. Whereas some investigators have proposed the involvement of specific proteinaceous receptors (18,19), other studies postulate models based on the interaction of A␤ with the lipid bilayer matrix of the plasma membrane (14,15). Our present data show that A␤ binds with high affinity and selectivity to gangliosides. Furthermore, in the presence of ganglioside-containing membrane vesicles, there is a dramatic increase in the rate of fibril formation by the peptide. We postulate that the membrane-bound A␤ may act as a template that catalyzes the fibrillogenesis reaction in vivo.
Preparation of Membrane Vesicles-Small unilamellar phospholipid vesicles were prepared as described previously (22). Vesicles were kept at room temperature and used within 12 h after preparation. Ganglioside-containing vesicles were obtained by adding to sonicated POPC vesicles an appropriate amount of micellar ganglioside in buffer and incubating the mixture for several hours (23).
Peptide Binding Experiments-Peptide-membrane binding experiments were performed with [Trp 10 ]A␤-(1-40) by following changes in the fluorescence spectra of the sole tryptophan residue of the peptide upon its incubation with lipid vesicles. For this purpose, small aliquots of concentrated vesicle suspension were successively added to peptide solution in buffer (1.3 M peptide in PBS if not stated otherwise). After each addition of lipid the solution was thoroughly mixed and left to equilibrate for 10 min at room temperature (such an incubation period was found to be sufficient to establish equilibrium). Fluorescence spectra were measured on an SLM 8100 spectrofluorometer using a 3-mm quartz cell and an excitation wavelength of 280 nm. Each spectrum was corrected for light scattering effects (by subtracting lipid blanks in the same buffer) and for wavelength-dependent efficiency of the detection system. Fluorescence titration curves were analyzed in terms of the peptide-ganglioside dissociation constant, K d , defined as: . This equation was transformed into the following form containing directly measurable quantities, (adapted from Ref. 24) and fitted to the experimental data with a nonlinear regression analysis. Parameter p in the above equation denotes total peptide concentration, y is the ganglioside concentration, and n is the stoichiometry of binding. The quantity x represents the change (either the wavelength shift in the fluorescence emission maximum or the change in fluorescence intensity) of the fluorescence spec-* This work was supported in part by the Charles S. Britton Fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  tra at a given ganglioside concentration, and X max is the respective spectral change at saturation.
Fibril Formation Assays-The progress of amyloid A␤-(1-40) fibril formation was followed by a Congo Red binding assay (21,25). HFIPdisaggregated peptide was incubated at 37°C in PBS (250 g/ml) in the presence and absence of membrane vesicles. Ten-microliter aliquots of each sample were withdrawn at 24-h intervals and transferred into the wells of microtiter plates containing 240 l of 7 M Congo Red in PBS. After a 30-min incubation at room temperature, the absorbance was measured at 540 and 480 nm using a THERMOmax microplate reader. For electron microscopy studies, a drop of each 24-h incubation sample was placed on a carbon-coated copper grid and negatively stained with 2% aqueous uranyl acetate. Grid preparations were visualized using a Jeol 100CX transmission electron microscope operating at 80 keV.

RESULTS
Membrane Binding-To study the interaction of A␤-(1-40) with membrane vesicles of different lipid composition, we prepared a peptide analogue in which Tyr at position 10 was replaced by Trp. Tryptophan residue provides a convenient spectroscopic probe which allows the measurement of peptidemembrane binding by fluorescence spectroscopy. The properties of the fluorescent peptide were found to be essentially identical to those of the parent molecule (20).
The fluorescence emission spectrum of HFIP-disaggregated peptide in PBS has a maximum at 347 nm and is indicative of a polar environment of the Trp residue. Incubation of the peptide with membrane vesicles consisting solely of phosphatidylcholine did not result in any measurable spectral change, suggesting the lack of peptide interaction with these vesicles. Similarly, no alterations in peptide fluorescence (in PBS) were observed upon addition of vesicles prepared from phosphatidylserine or phosphatidylglycerol (Fig. 1). 2 However, the fluorescence spectra changed drastically when the vesicles were doped with gangliosides. As shown in the inset within Fig. 1, upon addition of POPC vesicles containing 3 mol % ganglioside G M1 to peptide solution in PBS, the emission maximum is shifted to a shorter wavelength, and there is an enhancement of the fluorescence intensity. The observed blue shift reflects the increase in hydrophobicity of the tryptophan microenvironment and is indicative of peptide binding to the membrane. The titration curve obtained by measuring changes in the wavelength of the fluorescence emission maximum at increasing concentrations of vesicles shows that peptide binding to ganglioside G M1 is saturable (Fig. 1) and can be characterized by 1:1 stoichiometry with a dissociation constant of 1.4 ϫ 10 Ϫ6 M. The binding affinity did not change appreciably when PBS was replaced with low ionic strength buffer (Table I). Effects qualitatively similar to those illustrated in Fig. 1 were also observed for other members of the ganglioside family, including gangliosides G D1a , G T1b , and G M2 . However, the dissociation constants were found to differ significantly (in the range between 2 ϫ 10 Ϫ7 M to 5 ϫ 10 Ϫ6 M), indicating the following order of peptide affinity for different gangliosides: G D1a ϭ G T1b Ͼ G M1 Ͼ G M2 (Table I).
The results described above indicate that peptide binding to membrane vesicles is mediated through specific recognition by the gangliosides. Consistent with this, no peptide association with the membranes was detected when ganglioside G M1 was replaced by asialoganglioside G M1 (glycolipid lacking sialic acid portion of the head group) (Fig. 1). To further test the role of the sugar moiety in peptide binding to ganglioside-containing membranes, we titrated [Trp 10 ]A␤-(1-40) with free G M1 -pentasaccharide, as well as with sialic acid. While in these cases the position of the emission maximum remained unchanged, a concentration-dependent, saturable quenching of tryptophan fluorescence was observed clearly indicating A␤ peptide interaction with the free sugars (Fig. 2). (The different response of peptide fluorescence upon binding to free sugars and gangliosides is understandable since only the latter interaction leads to membrane-dependent increases in the hydrophobicity of the Trp microenvironment.) Analysis of the titration curve revealed that the affinity of the peptide for free G M1 -oligosaccharide is only modestly (5 times) lower than that for membraneassociated G M1 -ganglioside. Much weaker, although measurable, peptide binding was observed when sialic acid was titrated into the peptide solution ( Fig. 2 and Table I).
Effect of Membrane Binding on A␤ Fibrillization-The Congo Red assay is based on the observation that the dye binds to amyloid fibrils, shifting toward higher wavelength the maximum of its absorption spectrum (21,25). In this study, we have used the ratio of the absorbance at 540 and 480 nm as a measure of A␤ fibril formation. The ratio parameter increases linearly with the amount of fibrillar peptide and, in our experience, is more reproducible than the absorbance differencebased parameters used in other studies (21,25).
In agreement with previous reports (21,25), the kinetics of fibril formation by HFIP-disaggregated A␤-(1-40) is very slow. 2 Consistent with previous data (22,31), binding of A␤-(1-40) to acidic phospholipids could be detected by fluorescence spectroscopy only under the conditions of very low ionic strength (10 mM phosphate buffer, no NaCl) or at acidic pH. However, this nonspecific, purely electrostatic interaction is beyond the scope of the present study.  No fibrils were formed up to 5-6 days of peptide incubation in PBS (Fig. 3). However, the rate of fibrillogenesis was greatly increased in the presence of ganglioside G M1 -containing membrane vesicles (Fig. 3). In the latter case, massive Congo Red binding (corresponding to approximately 56% of the maximum binding) was detected already after 1 day of incubation. Simultaneous experiments performed in the presence of gangliosidefree POPC vesicles did not indicate any increase in fibril formation (data not shown for brevity). In preliminary studies, we noted that ganglioside-containing POPC vesicles alone can bind a limited amount of the Congo Red dye. However, this binding is negligible at the ganglioside concentrations used in the studies presented herein.
A␤-(1-40) fibril formation was also studied by transmission electron microscopy. Consistent with the Congo Red binding data, no fibrillar structures were detected following 1-day peptide incubation in PBS alone (Fig. 4A) or in the presence of ganglioside-free POPC vesicles (data not shown). However, following 1 day of incubation, mixtures in the presence of ganglioside G M1 -containing membranes exhibited, in addition to the vesicles, numerous fibrillar structures. The fibrils varied in length and had an average diameter of approximately 9 nm. Notably, the fibrils were for the most part associated with the membrane vesicles, and many of them appeared to originate directly from the vesicular surface (Fig. 4B). DISCUSSION A growing number of observations indicates that the neurotoxic action of A␤ is mediated by peptide-induced perturbation of the functional and structural properties of neuronal plasma membranes. Some of the reported membrane effects of the peptide include changes in bulk membrane fluidity, perturbation of the interface between lipids and proteins, inactivation of membrane-bound enzymes, formation of new or modulation of pre-existing membrane channels, and activation of free radicalgenerating pathways (10 -17, 26 -30). However, the molecular mechanisms of A␤-membrane interactions as well as the nature of acceptor molecules responsible for A␤ binding to the membrane surface remain largely unknown. The goal of this study was to characterize the interaction of A␤ with the lipid components of neuronal plasma membrane. To this end, we have used a fluorescent analogue of A␤-(1-40) in which the sole Tyr residue was substituted with Trp. Given the similarity of the aromatic residues, such a substitution is usually considered to have minimum effect on the properties of proteins and peptides. Indeed, no differences were found in the biophysical properties of A␤-(1-40) and [Trp 10 ]A␤-(1-40) (20). The advantage of using a Trp-labeled peptide is that its membrane binding can be assessed directly from changes in fluorescence spectra upon addition of membrane vesicles, with no need for physical separation of the free and bound species. Furthermore, from a structural point of view, Tyr 3 Trp substitution is less perturbing compared with other chemical modifications commonly used for A␤ labeling, including radioiodination and attachment of extrinsic fluorescent probes.
The key finding of the present study is that A␤ peptide interacts selectively with membrane gangliosides. This interaction is characterized by a relatively high affinity and a considerable degree of specificity with respect to the structure of the glycolipid oligosaccharide moiety. In addition to the sialic acid group, which is a prerequisite for the effective recognition of A␤, other structural elements of the glycolipid appear to play a role in the peptide-ganglioside interaction. Thus, the observed 3-fold tighter binding of A␤ to G M1 as compared with G M2 points to a stabilizing role of the terminal galactose residue (which is absent in G M2 ). The interaction is further strengthened (by a factor of approximately 6) in the presence of a second sialic acid residue, as in G D1b . While further studies are needed to fully elucidate structural and mechanistic aspects of A␤-ganglioside binding, it is notable that this binding shows very little sensitivity to ionic strength. This characteristic clearly differentiates A␤ interaction with gangliosides from that observed between the peptide and acidic phospholipids such as phosphatidylserine or phosphatidylglycerol. The latter interaction appears to be driven by nonspecific electrostatic effects; it is completely abolished in the presence of higher (150 mM) salt concentration (22,31). The apparent lack of A␤-(1-40) binding to phospholipids under physiologically relevant conditions is at odds with the recent hypothesis that A␤ peptide exerts its neurotoxic effect by a relatively nonspecific mechanism which involves direct interaction with the phospholipid bilayer to form Ca 2ϩ channels (14,15). However, the general "channel hypothesis" is not necessarily without merit. Experiments are currently under way to explore whether peptide incorporation into the membrane could be mediated by specific binding to gangliosides or other surface receptors.
While present only in relatively small quantities in most tissues, gangliosides are abundant components of neurons. They constitute about one-tenth of total neuronal membrane lipids (32,33) and appear to be especially highly concentrated in pre-and postsynaptic membranes (34). Functionally, gangliosides have been implicated in a number of important neurobiological events such as neurodifferentiation, neuritogenesis, synaptogenesis, synaptic transmission, and neuronal survival after injury. We postulate that oligosaccharide-specific interaction of A␤ with gangliosides may play a role in A␤induced neuronal degeneration. In particular, gangliosides are likely to function as high avidity "receptors" that capture the peptide and tether it to the cell surface. Once bound to the membrane surface, the peptide may engage in relatively nonspecific interactions with other membrane components, initiating the cascade of events that lead to membrane pathology (35) and eventually, neuronal cell death. It should be noted that A␤-ganglioside binding affinity is somewhat (4 -5 times) lower than that reported for peptide binding to putative proteinaceous receptors such as the receptor for advanced glycation end products or serpin-enzyme complex receptor (18,19). However, the modestly lower affinity could be easily compensated by a very high surface density of gangliosides. The proposed role of gangliosides as A␤ receptors is consistent with the finding that treatment with neuroaminidase greatly decreases binding of A␤ peptides to PC12 cells (36).
A striking consequence of ganglioside-mediated binding of A␤ to the membrane is the rapid acceleration of ␤-amyloid fibril formation. We suggest an important significance of this finding because a correlation appears to exist between biological effects of A␤ and its aggregation state (7)(8)(9)(10). Furthermore, it is believed that the fibrillar peptide itself represents the neurotoxic species. The mechanism of ganglioside-mediated A␤ fibrillization likely involves an initial step in which the glycolipid-bound peptide self-associates on the membrane surface, undergoing a conformational transition to a ␤-sheet structure. Such a conformational transition has indeed been demonstrated in our recent circular dichroism study (23). Surface-associated (␤sheet-rich) peptide microaggregates could then act as specific template ("seeds" (37)) which recruit peptide molecules from solution and promote fibril formation by the ␤-sheet augmentation mechanism. The role of ganglioside-bound A␤ as a physiological "seeding agent" is strongly supported by the recent observation that ganglioside G M1 -bound peptide constitutes an integral component of diffuse plaques associated with early stages of Alzheimer's disease (38). Furthermore, the proposed involvement of the membrane surface in A␤ fibrillogenesis is consistent with the in situ observation that A␤ is localized along neuronal plasma membranes (especially pre-synaptic regions) in early diffuse plaques (39).