Europium as an inhibitor of Amyloid‐β(1‐42) induced membrane permeation

Soluble Amyloid‐beta (Aβ) oligomers are a source of cytotoxicity in Alzheimer's disease (AD). The toxicity of Aβ oligomers may arise from their ability to interact with and disrupt cellular membranes mediated by GM1 ganglioside receptors within these membranes. Therefore, inhibition of Aβ–membrane interactions could provide a means of preventing the toxicity associated with Aβ. Here, using Surface Plasmon field‐enhanced Fluorescence Spectroscopy, we determine that the lanthanide, Europium III chloride (Eu3+), strongly binds to GM1 ganglioside‐containing membranes and prevents the interaction with Aβ42 leading to a loss of the peptides ability to cause membrane permeation. Here we discuss the molecular mechanism by which Eu3+ inhibits Aβ42‐membrane interactions and this may lead to protection of membrane integrity against Aβ42 induced toxicity.


Introduction
Alzheimer's disease (AD) is the most common form of dementia worldwide, with a global cost of over $600 billion in 2012 [1]. The pathological hallmarks of AD include the loss of neurons, the accumulation of neuritic plaques composed of extracellular, fibrillar Amyloid-beta peptide (Ab), and the deposition of intraneuronal neurofibrillary tangles composed of tau [2]. Ab  (Ab42) is the predominant variant deposited in AD brains.
The proteolytic cleavage of the integral membrane protein, Amyloid Precursor Protein (APP), at the N-and C-termini by band c-secretase respectively, results in the release of Ab [3]. This release of Ab from its transmembrane location means that the peptide retains an ability to interact with and/or insert into biological membranes. The propensity for peptides to form amyloid fibers is promoted by the b-strand conformation forming stable networks of hydrogen bonds that result in characteristic cross-b structures [4]. The hydropathic properties and charges on specific amino acids in the polypeptide chain also influence the peptide's ability to interact with surfaces such as membranes [5][6][7]. Therefore, subtle alterations in surface and membrane properties could dramatically affect both Ab-Ab and Ab-surface interactions.
Biological membranes are composed of complex mixtures of phospholipids, proteins, and membrane receptors. Gangliosides are a type of glycosphingolipid found in the outer leaflet of eukaryotic membranes that are composed of a hydrophobic, membraneembedded ceramide and a hydrophilic sialic acid moiety [8]. Gangliosides serve a variety of functions such as cell type-specific markers, differentiation and developmental markers, receptors, and as mediators of cell adhesion [9]. Expression of the GM1 ganglioside in the eukaryotic membrane is typically around 2% (w/w) [10], but can vary between 0.5% and 13% (w/w) depending on cell type and cell cycle stage [11][12][13][14]. The GM1 within lipid membranes has previously been shown to significantly affect the affinity of Ab42 for the membrane. We have previously shown that Ab oligomers possess high affinity and high avidity to GM1-containing membranes [15]. In our previous study, we showed that Ab  oligomers cause increased permeation of the membranes, whereby the peptide causes the formation of well-defined holes and defects [15]. GM1 has been shown to significantly affect the assembly state and oligomerization of Ab, and it was recently reported that increasing concentrations of GM1 induces a greater proportion of b-sheet structure within Ab [16,17]. GM1-bound Ab was proposed to act as a seed in the fibrillization of the peptide, and fibril extension occurred via consecutive binding of soluble Ab at the ends of the growing fibrils [18]. Therefore, GM1 may provide an ideal surface for Ab adsorption via effective hydrophobic and electrostatic interactions, and blocking of this receptor could significantly affect both Ab-membrane interactions and Ab-induced permeation.
Europium and other lanthanide complexes have been used to detect neutral sugars and glycolipids in ovarian cancer diagnostics [19] because the metal center of the europium cooperatively coordinates to the oligosaccharide and sialic acid moiety of gangliosides [19]. Europium complexes have received much attention as both therapeutic and diagnostic agents, including as angiogenic inhibitors [20], in vivo neuroimaging agents [21,22], as therapeutic agents in the treatment of ischemic heart disease [23], in myocardial infractions [24], and are reviewed in [25]. Williams et al. previously reported the potential use of europium (Eu 3+ ) as a means of inhibiting the binding of cholera toxin to tethered large unilamellar vesicles (LUV) [26]. Here, we report the effect of Eu 3+ complexed to GM1-containing membranes on the avidity and affinity of Ab42 binding to membranes in vitro, and the ability of Eu 3+ -coordinated membranes to resist Ab-induced membrane permeation. We also determine the sequence of events that occur during Ab-membrane binding, permeation and GM1 mediated seeding during Ab fibril formation.

Europium III chloride and other lanthanide metal ion preparation
Europium III chloride (Eu 3+ ), erbium III chloride, gadolinium III chloride, lanthanum III chloride, terbium III chloride and ytterbium III chloride (99.99% anhydrous, Sigma) were diluted to a stock concentration of 1 mM in HEPES pH 7.4, and then diluted to the working concentrations (10 and 100 lM).

Calcein-encapsulated large unilamellar lipid vesicles (LUVs)
Calcein encapsulated at a self-quenching concentration within the aqueous space of lipid vesicle lumen undergoes fluorescence dequenching as it diffuses into the external surrounding solution, resulting in an increase in calcein fluorescence intensity at 520 nm. Calcein-encapsulated LUVs were prepared as described previously [27]. Briefly, 40 mg mL À1 68:30:2 (w/w) DMPC/cholesterol/GM1 lipid films were rehydrated to 10 mg mL À1 with 200 mM calcein in HEPES pH 7.4 (Sigma-Aldrich Company Ltd., Dorset, UK), and vortexed vigorously for 30 min. The resulting suspension was passed 19 times through an Avestin extruder fitted with two stacked 100 nm polycarbonate membranes (GC Technology Ltd., Bedford, UK). Non-encapsulated calcein was removed from the LUVs using a 1 mL sephadex G-50 minicolumn preparation and spun in a 4°C controlled Mikro 22R centrifuge at 2000 RPM for 3 min to dispel the LUVs into the centrifuge tube. The eluted LUVs were washed a further two times. The LUVs were diluted to 1 mg mL À1 in HEPES pH 7.4, and stored at 4°C until use. 100 lM Eu 3+ was added to calcein loaded LUVs and mixed gently at 300 rpm for 30 min to allow complexation.
The encapsulation of calcein was confirmed in the fluorimeter by determining the release of the self-quenched calcein upon the addition of 0.5% (v/v) Triton X-100. If the calcein intensity did not exceed 100% of the starting fluorescence intensity prior to triton addition, then the LUVs were discarded. Percentage calcein release was calculated according to: where F is the measured fluorescence intensity of calcein at the specific time, F 0 is the initial measured fluorescence intensity at time = 0 min, F max is the maximum fluorescence intensity measured by the complete lysis of LUVs by the addition of the triton surfactant.

Fluorescence Spectroscopy
Fluorescence measurements were carried out on a Cary Eclipse fluorimeter (Varian Ltd., Oxford, UK). Samples were placed in a 1 cm path-length submicro quartz cuvette (Starna, Essex, UK) and the calcein fluorescence was monitored at various time points using an excitation wavelength of 490 nm. Calcein emission was monitored between 500 and 600 nm, with maximum fluorescence intensity at around 520 nm at a controlled temperature of 21°C. Excitation and emission slits were both set to 10 nm, and a scan rate set to 100 nm/min with 0.833 nm data intervals and an averaging time of 0.55 s. The photomultiplier tube was set to 400 V. Experiments were carried out in duplicate to confirm trends. Fluorescence intensities at the peak of 520 nm were plotted against time.

SPFS tethered large unilamellar vesicles
40 mg mL À1 stock solutions of the bilayer constituents were solubilized in 2:1 (v/v) chloroform/methanol and stored at À20°C until use. A mixed lipid solution was prepared in a glass vial from the stock solutions containing (50:13:2:2:30:2:1 w/w) DMPC/DM PE/DMPG/DMPS/Cholesterol/GM1/Biotin-PE and the solvent was removed using a stream of dry nitrogen and vacuum desiccated overnight. The lipid films were rehydrated with 5 lM BODIPY dye (6-(((4,4-difluoro-5-(2-pyrrolyl)-4-bora-3a,4a-diaza-sinda cene-3-yl)styryloxy)acetyl) aminohexanoic acid, sulfotetrafluorophenyl ester, sodium salt) (excitation wavelength = 648 nm, emission wavelength = 660 nm) (Molecular Probes, Invitrogen Ltd., UK) hydrolyzed to form the carboxylic acid dissolved in HEPES pH 7.4, and vortexed vigorously for 30 min. BODIPY is not a selfquenching dye at the concentrations used in these experiments, therefore is applicable to use with the SPFS system. The resulting suspension was passed 19 times through an Avestin extruder fitted with two stacked 100 nm polycarbonate membranes (GC Technology Ltd., Bedford, UK). Changes in mass (ng mm À2 ) at the sensor surface are calculated using: where (A) is the absorbance in ng mm À2 , (Dh/r) is the surface concentration correlation of 0.1868°and Dh is the shift in resonance minimum. The fluorescence flux through the permeated membranes (D * ) were calculated as previously described [15], and the K D (Eq. (4)) was calculated using the k off (Eq. 3a) and k on (Eq. 3b) dissociation and association equations, respectively using Origin 7 data analysis software (OriginLab): The equilibrium dissociation constant (K D ) was determined, and all changes in reflectivity were normalized. The apparent dye diffusion coefficient (D * ), that is a product of the fluorophore flux through permeated membranes, is calculated from the linearized change in fluorescence over time (extrapolated linearized slope of the dye diffusion rate), and takes into account vesicle membrane internal volume and membrane area and thickness.

SPFS bilayer preparation
For SPFS, the procedure for tethering of the LUVs can be found in detail in Ref. [15]. Briefly, gold-coated SFL6 glass were immersed in an ethanolic solution of 0.05 mM 11-mercaptoundecanoic-(8-b iotinoylamido-3,6-dioxaoctyl) amide and 0.95 mM 11-mercapto-1-undecanol (99%, Sigma) and allowed to self-assemble onto the gold substrates for 16 h. The biotin surface was coupled with streptavidin (500 nM in HEPES pH 7.4). The biotin tagged vesicles were capture to the streptavidin surface and binding followed by SPR. Vesicles were tethered for 30 min before rinsing in buffer to remove no encapsulated fluorophore and non-bound vesicles from the system. A detailed schematic of the SPFS setup, flow-cell setup and extensive background to SPFS principles can be found in [34]. A 1 mM stock of europium III chloride in HEPES pH 7.4 was prepared, and then diluted to the working concentration (10-100 lM) prior to injection into the flow cell and allowed to adsorb to the GM1-containing LUVs for 30 min prior to rinsing of the flowcell with fresh HEPES pH 7.4. Solutions were circulated around the flow-cell at a rate of 1.8 mL min À1 using a peristaltic pump for optimized analyte delivery, while minimizing the mass transport effects and the shear force on the bilayer vesicles.

Transmission electron microscopy
A 4 ll droplet of the peptide (80 lM) was adsorbed onto formvar/carbon coated 400 mesh copper grids (Agar Scientific, Essex, UK) for 60 s, and blotted dry. 4 ll of 0.22 lm filtered water was added to the grid and immediately blotted, then negatively stained with 4 ll of 2% uranyl acetate adsorbed for 60 s and blotted dry.
The grid was allowed to air dry before examination on a Hitachi 7100 microscope (Hitachi, Germany) fitted with a Gatan Ultrascan 1000 CCD camera (Gatan, Abingdon, UK). Aliquots of samples at the stock concentration were taken at time points for each experiment to monitor fibrillization. Measurements were made using ImageJ [35].

Solubilized Ab42 forms small, soluble oligomers that assemble to form fibers
Transmission electron microscopy (TEM) confirmed the oligomeric and fibrillar morphology of the Ab42 peptide following dissolution and then after 24 h incubation, respectively. A substantial in depth characterization of both peptide preparation protocol and Ab-membrane permeation has previously been carried out using calcein release, TEM, circular dichroism, SPFS, and atomic force microscopy [15,27]. We showed that Ab42 used immediately after preparation formed small, circular oligomeric assemblies ranging in size between 1 and 5 nm, [15,27]. Electron micrographs of 24 h incubated peptide showed larger assemblies composed of curvilinear oligomers, and small protofibrils ranging between 15 and 30 nm, and small fibers ranging between 140 and 240 nm in length and 5 nm in width [27]. Moreover, Ab oligomers in the presence of Eu 3+ were observed to form fibrils indistinguishable from fibrils formed in the absence of Eu 3+ , therefore, Eu 3+ does not adversely affect the morphology of the assembling peptides or alter the aggregation pathway ( Supplementary Fig. 1).

Selection of Eu 3+ as inhibitory agent as opposed to other lanthanide ions
Surface Plasmon field enhanced Fluorescence Spectroscopy (SPFS) was used to screen potential candidate lanthanide ions for complexation to GM1-containing membranes. Initially, increasing concentrations (0-10 lM) of the positive control cholera toxin (Ctx) were added to the tethered LUVs that had no lanthanide metal ions coordinated, and the mass adsorption of the control protein measured (Supplementary Fig. 2). This measurement was to determine the appropriate concentration of Ctx to add to the lanthanide coordinated LUVs for comparison. A concentration dependent increase in Ctx adsorption was observed and when equilibrium was reached at 10 lM Ctx equilibrium. Therefore, 10 lM Ctx was used to ensure a like-for-like comparison between the lanthanide species. Erbium III chloride, europium III chloride, gadolinium III chloride, lanthanum III chloride, terbium III chloride or ytterbium III chloride were added to the tethered membrane vesicles and allowed to complex with the membrane, and the amount of metal ion adsorption determined (Fig. 1). To test the abilities of the lanthanide ions to resist peptide-induced permeation, Ctx was injected into the system and the permeation determined (Fig. 1). The Ctx specifically associates with gangliosides, therefore this positive control was used to ensure that maximal permeation of the membranes was observed in order select the most appropriate lanthanide ion to use in this study.
The addition of the six different lanthanide metal ions resulted in varying levels of adsorption to the GM1 containing membranes (Fig. 1). Gadolinium showed the lowest affinity binding to the tethered membranes, followed by ytterbium and erbium. Terbium showed the third highest amount of metal ion binding to the membranes, followed by lanthanum. The highest amount of metal ion binding was observed with europium ions (Fig. 1). ANOVA statistical analysis shows a significant difference between the groups of the metal ions in the amount of metal ion binding to the GM1containing membranes (P < 0.05).
The addition of the 10 lM Ctx to lanthanide-complexed membranes resulted in a significant reduction in permeation of the membranes compared to Ctx alone (Fig. 1). All lanthanides tested using the positive control Ctx toxin resulted in similar abilities to reduce membrane permeation. ANOVA statistical shows no significant difference between the lanthanides ability to cause reduced permeation (P > 0.05). However, there is a statistical significance between the lanthanides ability to reduce permeation compared to when no lanthanides are present (P < 0.05), whereby on average the lanthanide complexed LUVs showed a 40% reduction in permeation compared to when no lanthanides were present. We selected europium ions in this study because it showed the highest mass adsorption to the membranes, while similarly causing a reduction in the ability for the positive control to cause membrane permeation. Selection of europium trivalent ions appeared as the most suitable candidate to ensure the greatest binding to the biomimetic membranes for the subsequent studies using the Ab peptide to permeate membranes. As each of the lanthanide ions had the ability to resist peptide-induced permeation, it would be intuitive to use the lanthanide with the greatest mass adsorption. It should be noted that Ab42 may bind to varying degrees to the other lanthanide ions but as a proof of concept that lanthanide ions resist Ab association to the membranes we dedicated this study to a single lanthanide ion to fully study the lanthanide-Ab molecular mechanism.
The addition of 10 lM Ctx to 0-500 lM Eu 3+ resulted in concentration dependent saturation of the membrane surface (Supplementary Fig. 3). Increasing the coordination of Eu 3+ above 10 lM resulted in a very slight increase in Ctx adsorption to the membrane surface compared to membranes coordinated with 10 lM Eu 3+ . This could be an effect of oversaturation of the membrane surface that results from a kind of steric hindrance between the metal ions overcrowding the GM1 receptors.

Eu 3+ complexed to LUVs in solution inhibits oligomeric Ab42 induced permeation
We have previously shown that lipid bilayers containing GM1 are highly susceptible to permeation by oligomeric Ab42 [27]. A calcein release assay was employed to observe Ab-induced permeation of LUVs in solution of a simple membrane composition. The addition of oligomeric Ab42 alone resulted in an immediate release of 37% of the total encapsulated calcein from the LUVs ( Table 1), indicative of Ab-induced membrane permeation (Fig. 2) and in agreement with previous results [27]. Next, LUVs were preincubated with Eu 3+ and this was followed by the addition of oligomeric Ab42. Eu 3+ -complexed membranes resulted in the release of only 0.7% of the total encapsulated calcein ( Fig. 2 and Table 1). Therefore, incubation of the Ab42 with the Eu 3+ -coordinated LUVs showed a 56-fold decrease in membrane permeation relative to LUVs incubated in the absence of Eu 3+ (Fig. 2). Fibrillar Ab42 has previously been shown to cause less permeation to membranes than oligomeric Ab42 [27]. The addition of fibrillar Ab led to 3% of the total encapsulated calcein release from LUVs in the absence of Eu 3+ (Table 1), significantly less than permeation induced by oligomeric Ab42. The addition of fibrillar Ab42 to Eu 3+ -complexed LUVs resulted in the release of 0.7% of the total encapsulated calcein within the membranes, which corresponded to a 4-fold decrease in permeation. These results demonstrate that Eu 3+ protects the membranes from permeation by both oligomeric and fibrillar Ab42.

High affinity complexation between Eu 3+ and GM1-containing membranes
In order to investigate the membrane protective mechanism of Eu 3+ against Ab, Surface Plasmon resonance was used to follow binding of Ab42 to tethered LUVs in the presence and absence of Eu 3+ . Initially, the affinity of the Eu 3+ with GM1 containing membranes was considered. Two concentrations of Eu 3+ (10 or to the tethered LUVs resulted in an equilibrium dissociation constant of K D = 0.48 lM (±0.04) and the adsorption of 100 lM Eu 3+ resulted in a K D = 0.63 lM (±0.18) (Fig. 3), which represents strong binding of a similar strength to the specific binding between cholera toxin and GM1 [26].

Eu 3+ -complexed to LUVs inhibits the interaction of oligomeric Ab42 to the membrane surface
Surface Plasmon field-enhanced Fluorescence Spectroscopy (SPFS) was used to concurrently measure Ab adsorption to the membrane surface, the avidity of these interactions and the fluorescence flux through the permeated membranes [15,26]. 10 lM oligomeric Ab42 was injected into the flow-cell and the change in mass monitored. 10 lM Eu 3+ significantly reduced Ab42 adsorption to the tethered LUVs, while 100 lM Eu 3+ almost entirely prevented the adsorption of Ab42 (Fig. 4). The Eu 3+ and Ab42 mass binding and equilibrium dissociation constants are shown in Table 2. The adsorption of 10 lM Eu 3+ resulted in a 6% decrease in the binding between Ab42 and the LUVs, whereas the addition of 100 lM Eu 3+ resulted in a 94% decrease in binding between Ab42 and the LUVs. The peptide was non-specifically adsorbed to the membrane surface and was easily washed off into solution. Thus, the addition of 100 lM Eu 3+ to LUVs inhibited the ability of Ab42 to bind to membranes in a concentration dependent manner.
The association binding data (k on ) were further analyzed by fitting to mathematical exponential curve model to determine the number of analyte-ligand binding events that occur during the binding of Ab to the LUV surface, and these values are summarized in Table 3. The exponential fitting function provides a means to statistically and mathematically determine the best fit for the binding between Ab and the membranes, and provides a more sophisticated fitting procedure that can distinguish curves that do not follow the simple bimolecular binding. A monophasic response (single exponential curve function) indicates that a single    Table 3.
1:1 binding event between the analyte (Ab) and ligand (LUV) is occurring, such as the binding of single Ab molecule to one GM1 receptor. A biphasic response (2 exponential curves) indicates two distinct and different binding events occurring during the association of Ab to the membrane surface, for example, the specific association of the Ab peptide to the GM1 receptor followed by GM1 induced seeding fibrillization of Ab on the membrane surface, as has previously been reported [18]. From the model fitting and the analysis of the Ab42 oligomermembrane binding in the absence of europium (0 lM Eu 3+ ), it was revealed that there were two distinct binding events between Ab42 and the membranes (show as t1 and t2 in Table 3), because the best curve fit was found to be to two consecutive exponential curves. Phase (I) the initial 1:1 specific binding between Ab42 to the GM1-containing membranes (t1). This is then followed by Phase (II), non-specific binding of Ab42 to the membranes and/or Ab elongation as a result of GM1-induced seeded fibrillization (Fig. 4). Because the binding is relatively strong, a non-specific adsorption of Ab to the LUVs is unlikely. Self-assembly and fibrillization of Ab42 on the membrane surface would be more in line with the observed binding, and we hypothesize that the specific Ab-bound GM1 is acting as a seed for further Ab molecules to aggregate. The first event occurs between 11 and 430 min (Phase I, t1), and corresponds to the immediate binding of the Ab to the GM1-containing membrane surface ( Fig. 4 and Table 3). The second exponential curve fits from the beginning of 430 min (Phase II, t2), which corresponds to the GM1-Ab complex acting as a seed for further Ab fibrillization ( Fig. 4 and Table 3).
The binding of oligomeric Ab42 to 10 lM Eu 3+ complexed-LUVs revealed that the best fit was also obtained using the biphasic exponential function, therefore, two binding events are occurring similar to when no europium is present. The first binding phase occurs between 60 and 505 min (t1 Table 3), corresponding to the initial binding of Ab to the membrane surface (Fig. 4), but took significantly longer to occur compared to when europium was absent (81% increase in the initial t1 time phase) ( Table 3). This is attributed to the low concentration of europium blocking a significant proportion of the GM1 binding sites and also altering the net charge of the membrane surface, therefore, the Ab took longer to bind to the membranes because of the alteration in membrane surface elicited by the low concentration of europium. The second phase occurs at 505 min corresponded to the beginning of the Ab fibrillization resulting from the GM1-seeding effect ( Fig. 4 and Table 3).
Analysis of Ab42 association to 100 lM europium coordinated membranes showed a poor fit to a mono-, bi-and tri-phasic exponential fitting and were therefore all excluded. This showed that Ab42 did not bind significantly to 100 lM Eu 3+ coordinated LUVs, suggesting that the Ab binds non-specifically to membranes coordinated with high concentrations of Eu 3+ , as all the GM1 binding sites are occupied by the metal ions and there is a significant alternation in membrane charge caused by the positive metal ions that results in electrostatic repulsion between the Ab and the membrane surface. The exponential model fitting calculations shows that increasing Eu 3+ concentrations significantly affected the binding of Ab42 to membrane surfaces, and that Eu 3+ caused an increase in the time taken for Ab42 to initially associate with LUVs.
10 lM Eu 3+ caused a 5.4 fold delay in the initial association and formation of defects in the membrane compared to when Eu 3+ was absent (Table 3, t1 and t2).

Eu 3+ -complexed to LUVs inhibits oligomeric Ab42-induced permeation of LUVs
The addition of oligomeric Ab42 to Eu 3+ complexed LUVs resulted in decreased rates of BODIPY fluorophore diffusion (Fig. 5). A simple theoretical model was previously developed to fit the decrease in fluorescenceÀtime curves (apparent dye diffusion coefficient (D ⁄ )), and detailed in Ref. [26]. The addition of Ab42 oligomers to LUVs resulted in D ⁄ = 1.72 Â 10 À15 m 2 s À1 (±1.5 Â 10 À15 ). In contrast, when Ab42 oligomers were added to 10 lM Eu 3+ -complexed LUVs, the D ⁄ fell to D ⁄ = 2.18 Â 10 À16 m 2s À1 (±1.2 Â 10 À16 ). Therefore, 10 lM Eu 3+ coordinated LUVs resulted in an 87% decrease in membrane permeation rate. Table 2 Comparison of europium and Ab42 binding and permeation constants. Comparison of the mass of Eu 3+ adsorption to LUVs and subsequent equilibrium dissociation constants (K D ), mass adsorption of Ab42 and apparent dye diffusion coefficient (D ⁄ ) of oligomeric Ab42 with 0, 10 and 100 lM Eu 3+ coordinated GM1-containing LUVs (± are the standard error of the mean).  Table 3 Exponential function fitting for Ab42 oligomer to the Eu 3+ -coordinated LUVs. Exponential functional expression and statistical analysis to determine mono-and bi-phasic association of the oligomeric Ab with Eu 3+ -coordinated membranes. a0 and a1 = mono-and bi-phasic exponent per-factors respectively, the larger the per-factor signifies a larger contribution to that particular phase in the binding process. t1 and t2 = characteristic time constants for mono-and bi-phasic exponential fit ( Table 2. 100 lM Eu 3+ complexed LUVs resulted in D ⁄ = 3.92 Â 10 À17 m 2 s À1 (±2.69 Â 10 À17 ), showing an 82% decrease in permeation compared to in the absence of Eu 3+ and a 98% decrease compared to 10 lM Eu 3+ coordinated LUVs (Table 2).
Interestingly, although the 10 lM Eu 3+ did not appear to have a strong effect on binding of Ab42 to tethered LUVs, this lower Eu 3+ concentration still had a strong effect on the peptide binding lagphase and peptide assembly on the membrane (Table 3). There was a significant effect of Eu 3+ on resulting BODIPY diffusion rates, which may indicate a link between self-assembly of Ab42 on the membrane and the mechanism of Ab42-induced membrane permeation. We posit that as Ab42 begins to self-assemble into higher order oligomers (phase II Figs. 4 and 5), diffusion of the fluorophore is reaching equilibrium and permeation is completed.

Discussion
The interactions between Ab42 and membranes have been hypothesized to play an important role in the mechanism of cyto-toxicity in AD. Our previous work showed that soluble oligomeric Ab42 binds to GM1-containing membranes and causes the permeation of phospholipid membranes [15,27]. Here, we demonstrate that (i) Eu 3+ shows the greatest mass adsorption to GM1 containing membranes compared to the other lanthanide ions tested, and all the ions tested share similar abilities to resist Ctx-induced permeation in control studies; (ii) the interaction between Ab42 and membranes is inhibited by the formation of complexes formed between Eu 3+ and GM1 gangliosides; (iii) Calcein-encapsulated LUVs incubated with Eu 3+ prevented the permeation of the membranes caused by Ab42; (iv) Ab42 showed decreased binding to Eu 3+ preincubated membranes, consequently inhibiting Ab42 membrane-induced permeation; We are also able to correlate the events that coincide between Ab binding, permeation of the membranes and disturbance of biomimetic membrane homeostasis. This is highly significant as these events have not been previously measured simultaneously and directly linked, giving us a clearer picture of the molecular events associated with the binding and permeation of membranes by Ab. Fig. 6. Schematic diagram of the proposed mechanism of europium inhibition of Ab binding to membranes. The administration of europium blocks Ab42 from interacting with the GM1-containing membranes, and inhibition of subsequent membrane permeation. Europium is proposed to selectively form cooperative complexes with the oligosaccharide and sialic acid residues via the europium metal center via electrostatic and hydrogen bonds that results in a coordination shell. The schematic is adapted from [19] and is not to scale.
The calcein release assay and BODIPY diffusion from tethered LUVs reveals that Eu 3+ -complexed LUVs resist Ab42-induced permeation by reducing the binding of Ab42. Our results highlight the significant role for GM1 and membrane properties in the mechanisms of Ab toxicity. Arising from the findings reported here, a schematic of a proposed mechanism of Eu 3+ inhibition of Ab42membrane interactions is shown in Fig. 6. GM1 has been shown to play a pivotal role in Ab42-membrane interactions and Ab42induced permeation. The polar moiety of the GM1 provides a complementary surface for the polar amino acids to allow the formation of hydrogen bonds. GM1 gangliosides also provide a site of Ab attachment to membranes [36][37][38][39], and Ab has been shown to colocalize in GM1-rich lipid raft microdomains [39]. Eu 3+ was previously shown to covalently interact with the oligosaccharide and sialic acid moiety of the GM1 [19]. Therefore, if Eu 3+ is blocking the GM1 site, it is no longer available to participate in Ab interactions, aid Ab polymerization that may lead to permeation of membranes. These results underline the importance of the Ab toxic mechanism being mediated by GM1.
The decrease in permeation upon the addition of Ab42 to Eu 3+ -coordinated membranes may be attributed to the change in the membrane charge resulting from the coordination of the positive trivalent Eu 3+ ions to the membrane surface. Increasing the attractive electrostatic interactions between Ab and GM1 results in increased insertion pressure due to the Ab penetration into the membrane surface [40]. The solvent exposed aromatic residues of Ab are able to stack onto the sugar rings of the GM1 driven by the net positive charge of the GM1 sugar rings that are in close proximity to the p-electron cloud of the amino acid aromatic ring [41]. Here, the positively charged Eu 3+ ions would cause electrostatic repulsion toward the highly solvent exposed arginine residue 5 of Ab42, which lies next to the aromatic phenylalanine at position 4 that can form stacking interactions with the GM1 sugar rings. By changing the net charge of the membrane with the coordination of positively charged Eu 3+ , the N-terminus of Ab42 is repelled from the membrane surface. It has been demonstrated that dipolar compounds that shield the membranes negative charge can prevent the association of Ab peptides with membranes and decrease Ab-induced cell toxicity [42], similar to the mechanism we propose here to occur with Eu 3+ .
The mechanism by which Ab causes membrane damage has been proposed to occur via three potential mechanisms; poreformation, detergent-like and carpeting [43]. SPFS measures changes in mass in real-time, therefore, any subtle changes are monitored and visualized simultaneously. If the addition of oligomer Ab42 resulted in total or partial lysis or removal of the lipid bilayer by a detergent-like mechanism, this would be observed as a decrease in the SPR signal. However, this is not observed (Fig. 4), therefore, a detergent-like mechanism of membrane permeation is not supported by our data. We have previously shown that Ab42 rapidly forms defects and holes in planar phospholipid bilayers [15]. In support of this, here we demonstrate that the Phase I of Ab42 binding to the membranes continues for 400 min. This establishes that, even though the initial formation of defects occurs very quickly, Ab continues to bind to the membranes for a much longer time causing the formation of stable defects within the membrane. Phase II of binding and fluorophore diffusion that we attribute to Ab selfassembly and fibrillization, would therefore not heavily influence the formation of membrane defects and permeation processes because they have already occurred much earlier during Phase I. This may suggest that mechanistically, the initial binding of Ab and the primary formation of membrane defects are the critical determinant in Ab-induced permeation of the membranes, and that Ab self-assembly and fibrillization that occur during phase II are of less importance. We hypothesize that the initial binding of Ab and formation of defects that correspond to Phase I (Fig. 4) is the time period where the most significant damaging events occur. These initial defects during Phase I are significant enough to cause dramatic membrane damage. We attribute Phase II (Fig. 4) to Ab self-assembly and fibrillization, and during this time period, membrane homeostasis has already been significantly disrupted. A schematic of these events is shown in Fig. 7. Therefore, prevention of the interaction of Ab with membranes could be an important factor in preventing and reducing Ab membrane damage.
Here we show that lanthanide ions, and in particular Eu 3+ represents a potential Ab-membrane interaction inhibitor that could provide a novel strategy of altering the course of Ab42 assembly and aggregation on the membranes and may effectively reduce the cytotoxicity associated with Ab42. Fig. 7. Schematic of the phases that occur during Ab42-induced membrane permeation. (a and b) Ab42 in close proximity to the GM1-containing membrane surface becomes associated with the GM1 gangliosides with high affinity (as determined by SPR Fig. 4). (c) Ab42 begins to permeate the first lipid bilayer within 30 s of contact with the membrane (as determined by AFM in Ref. [15]). (d) Ab42 permeated through both leaflets of the membrane bilayer and allows non-specific diffusion between the internal space into the extracellular environment (as demonstrated by SPFS Fig. 5). (e) Ab assembly and fibrillization occurs, whereby Ab42-bound GM1 acts as a seed for further Ab42 assembly and the formation of higher order oligomers and fibers.