Simultaneous Single Molecule Fluorescence and Conductivity Studies Reveal Distinct Classes of Aβ Species on Lipid Bilayers

Abstract

The extracellular senile plaques prevalent in brain tissue in Alzheimer's disease (AD) are composed of amyloid fibrils formed by the Aβ peptide. These fibrils have been traditionally believed to feature in neurotoxicity; however, numerous recent studies provide evidence that cytotoxicity in AD may be associated with low molecular weight oligomers of Aβ that associate with neuronal membranes and may lead to membrane permeabilization and disruption of the ion balance in the cell. The underlying mechanism leading to disruption of the membrane is the subject of many recent studies. Here we report the application of single molecule optical detection, using fluorescently labeled human Aβ40, combined with membrane conductivity measurements, to monitor the interaction of single oligomeric peptide structures with model planar black lipid membranes (BLM). In a qualitative study, we show that the binding of Aβ to the membrane can be described by three distinctly different behaviors, depending on the Aβ monomer concentration. For concentrations much below 10 nM, there is uniform binding of monomers over the surface of the membrane with no evidence of oligomer formation or membrane permeabilization. Between 10 nM and a few 100 nM, the uniform monomer binding is accompanied by the presence of peptide species ranging from dimers to small oligomers. The dimers are not found to permeabilize the membrane but the larger oligomers lead to permeabilization with individual oligomers producing ion conductances of less than 10 pS/pore. At higher concentration, perhaps beyond physiologically relevant concentrations, larger extended and dynamic structures are found with large conductance (100's of pS) suggesting major disruption of the membrane.

One of the hallmarks of Alzheimer's disease (AD) is the formation of insoluble neurofibrillary tangles and senile plaques in brain tissue. The major component of these proteinaceous plaques is the 39-43 amino acid Aβ peptide, derived by proteolytic cleavage of the membrane spanning amyloid precursor protein (APP). The mechanism that underlies the pathogenicity of Aβ towards neuronal tissue is the subject of intense research.

In vitro studies of Aβ reveal that in aqueous solution the peptide progressively associates from its monomeric form via low molecular weight species to extended β-sheet fibrils, a process that is highly dependent upon experimental conditions (1, 2). It was originally assumed that Aβ becomes cytotoxic when it forms these large insoluble fibrillar aggregates (3, 4), historically referred to as the amyloid hypothesis. For the purposes of this paper we use the term cytotoxic, to mean that Aβ causes deviations from the cell's normal healthy homeostasis that relate to the disease (i.e., Alzheimer's) and/or possibly cell death.

In contrast with the amyloid hypothesis, recent studies have demonstrated that cytotoxicity may be associated with small oligomers of Aβ that bind to the neuronal membrane, leading to cell death possibly by membrane permeabilization (5-9). Generalizing this hypothesis, it has been suggested that other amyloid-associated diseases share this mechanism of cell toxicity, where peptide-generated disruption of the membrane (possibly producing well defined pores) allow an unrestricted influx of ions such as Ca²⁺ into cells, thereby stressing them either directly (10-18), or indirectly by triggering Ca²⁺ sensitive apoptosis signaling pathways (15, 19-21). A number of studies have tested this hypothesis through cell viability assays (20, 22-24), permeability studies using liposomes (25-27), and conductivity studies using lipid bilayers (8, 28-31).

Efforts to identify membrane-permeabilizing structures formed by amyloidogenic peptides have utilized a variety of techniques including atomic force microscopy (17, 19, 32-34), electron microscopy (9, 35, 36) and conductivity measurements (31, 37-40). A causal relationship between small oligomers of Aβ and toxicity has been hypothesized from these studies, but efforts to develop a more detailed mechanistic understanding of how Aβ assembles into toxic species and of the basis of the toxicity have been limited by significant experimental challenges since under physiologically relevant conditions the Aβ aggregates are metastable, present at extremely low concentrations, and highly heterogeneous. These features make it very difficult to detect and characterize the different structures as they form and to monitor their integration into the membrane thereby further hampering detailed mechanistic studies. Thus, in spite of many efforts, the identification of the toxic species and the specific mechanism by which they inflict toxicity has proven to be difficult (22, 28, 34, 41-44).

In this work we present results of a single molecule study of Aβ binding to a model planar black lipid membrane (BLM) system. Using fluorescently labeled Aβ combined with the ability to measure membrane electrical conductance, we show that Aβ binding and membrane permeabilization in this system falls into three distinct morphological classes: 1) At relatively low peptide concentrations, (<10 nM) we find uniform binding of Aβ to the membrane surface that involves monomers only and with no detectable membrane conductance (<1pS). 2) At intermediate peptide concentrations, between 10 nM and several hundred nM, we find small oligomers (from dimers to ~14mers), where the dimers produce no detectable conductance but the larger oligomers result in conductance values of <10 pS per oligomer. 3) At the highest peptide concentrations (>>100 nM), we find large structures of Aβ that slowly move across the surface and are associated with large electrical conductivity (~100's of pS). From the data we conclude that the interactions of Aβ with the membrane surface are highly nonlinear with respect to peptide concentration, and that if this behavior occurs on the surface of neurons the toxicity, if due to permeabilization, would depend profoundly on the intracellular Aβ concentration.

Material and Methods

SMS-Conductivity System

Figure 1 shows the experimental setup used to measure membrane conductivity and single molecule fluorescence simultaneously. This system is comprised of an in-house built inverted confocal microscope utilizing an Olympus 60x 1.45 NA objective. The output of an 80MHz Coherent Ti:Saphire laser is frequency doubled to 460nm (Inrad, Northvale, NJ) and used for single photon excitation of fluorescently labeled Aβ bound to the surface of the Black Lipid Membrane (BLM). A Chroma (Rockingham, Vermont) Q500XRU dichroic filter is used to separate excitation and HL488 emission. Images are recorded by raster scanning the surface of the BLM with a piezo-electric stage (Physik Instrumente, Auburn, MA) with emission detected by an avalanche photodiode (APD; EG&G Ortec SPCM-AQR-16). A CCD camera was used to determine that all oligomeric structures were within the area to be raster scanned. Maximum resolution is set by the diffraction limit, approximately 0.5 μm. Ag/AgCl electrodes measure current through the BLM bilayer formed on the surface of agarose layered on the coverslip.

Figure 1.

The single molecule BLM system is comprised of an in-house built inverted confocal microscope utilizing an Olympus 60x 1.45 NA objective. Frequency doubled Ti:sapphire output was used for single photon excitation of fluorescently labeled Hilyte-488 Aβ bound to the surface of the Black Lipid Membrane. Images are recorded by raster scanning the surface of the BLM with a piezo-electric stage with emission detected by an avalanche photodiode. A CCD camera was used to determine that oligomeric structures were within the area to be raster scanned. Maximum resolution is set by the diffraction limit, approximately 0.5 μm. Ag/AgCl electrodes measure current through the BLM bilayer formed on the surface of agarose layered on the coverslip.

Black lipid membranes

BLM were generated using a solution of diphytanoyl phosphatidylcholine and diphytanoyl phosphatidylserine (DiphPC/DiphPS; Avanti, AL), 70/30, in methylcyclohexane. The membrane was formed by the procedure outline by Ide et al (45), in which an aqueous solution of Aβ is covered by cyclohexane containing 10mg/ml of the lipids. The interface between the phases has a multilamellar membrane that is forced by the probe onto an agarose surface coating the microscope coverslip to form a unilamellar membrane under very low hydrostatic pressure within the probe channel. Formation of a single bilayer was monitored by membrane capacitance (~1μF/cm²) utilizing the Warner 535 Patch Clamp Amplifier.

Bilayer Conductivity Measurements

BLM conductance was monitored by a Warner DG535 bilayer voltage amplifier (Hamden, CT), with digitization and data acquisition using a home written software and a National Instrument 6070 AD card using Labview. Analysis was performed by a procedure utilizing Igor Pro analysis routines. A 4 pole Bessel filter was applied to the signal to improve signal/noise. The Ag/AgCl electrodes were chlorinated by a 30-minute treatment with bleach.

Aβ Sample Preparation

Salts and buffers were obtained from Sigma-Aldrich (St. Louis, MO). Fluorescently labeled and unlabeled Aβ40 were obtained from Anaspec (San Jose, Ca). Fluorescently labeled Aβ40 was N-terminally labeled with Hilyte 488 (Aβ40-HL488) by Anaspec. Aβ40 was distributed into smaller aliquots by dissolving the lyophilized powder in 1% ammonium hydroxide, sonicated for 1 minute then re-lyophilized into 20μg samples. Aβ was then dissolved at 50μM in distilled water, vortexed for 1 min, then brought to a concentration of 25μM with 2X buffer resulting in a final buffer composition of 5 mM sodium phosphate, 100 mM sodium chloride pH 7.4. Final peptide concentrations of between 1 and several 100 nM were produced by further dilution with buffer.

Fluorescence Lifetime Measurements

Time correlated single photon counting measurements (TCSPC) were done using standard procedures (46) in reverse counting mode. A Coherent Verdi V-10 diode laser pumped a Coherent model Mira 900 Ti:Sapphire laser (Santa Clara, CA) operating with 150 fs pulsewidth. The output was frequency doubled using an Inrad model 5-050 Ultrafast Harmonic Generating system (Northvale, NJ). The excitation rate was 4 MHz, controlled by a Coherent 9200 Pulse Picker. Fluorescence emission was detected with either a Hamamatsu R3809U-50 microchannel plate photomultiplier or a Hamamatsu H7421 photomultiplier (Bridgewater, NJ) input into a Tennelec quad Constant Fraction Discriminator (model TC455Oak Ridge, TN). Analog output from an Ortec Time-amplitude converter (model 457, Oak Ridge, TN) was digitized by a system developed in our laboratory utilizing National Instruments hardware (NI6070 12bit A/D converter, NI6602 counter) and controlled by Labview Software (Steve Parus, Chemistry Department) that simultaneously records total emission and fluorescence anisotropy. Anisotropy measurements utilized a Hinds PEM-90 photoacoustic modulator (Hillsboro, OR) to monitor horizontal and vertical polarizations at 42 kHz. Time resolved studies with BLM membranes also utilized a Picoquant Timeharp 200 (Berlin, Germany) card running at 80 MHz excitation rate.

Results

Identification of Membrane-Permeabilizing Aβ structures

To identify membrane-bound structures produced by Aβ and quantify their associated membrane permeabilization efficacy, we utilized Aβ40 that had been fluorescently labeled with Hilyte 488 (Anaspec) and applied single molecule optical detection to monitor individual peptide oligomers bound to the BLM while simultaneously determining the electrical conductivity associated with the same sample.

Our experimental setup, shown in Figure 1, provides the capability to individually image labeled Aβ oligomers on the surface of a 50 μm radius BLM and correlate the optical image with the integrated conductivity of the membrane. To confirm that peptide labeling with the fluorescent dye has not altered its membrane permeabilizing behavior, both labeled and unlabeled peptides (100 nM) were incubated for 4 hours in aqueous solution and each was then added (separately) to the membrane preparations. Conductivity was monitored (Figure 2) and the results confirm that the labeled Aβ-HL488 preserves the membrane permeabilizing capability of the unlabeled peptide. There is substantial variability in BLM conductivity with both unlabeled and labeled Aβ between different sample preparations with no apparent differences in the protocol. This effect, coupled with the limited stability of BLMs (tens of minutes in the absence of Aβ peptide) makes quantitative studies challenging at this point. Thus, while we observe that both labeled and unlabeled Aβ peptide affect membrane permeabilization with comparable conductivity and time scale, a more detailed comparison of the relative permeabilizing ability of labeled and unlabeled peptides has not yet been accomplished.

Figure 2.

Conductivity plots for 100 nM unlabeled Aβ40 and Hilyte 488 labeled Aβ40 added to a BLM after a 4 hour incubation in buffer at 20°C. The BLM was maintained at 50mV potential with transient changes to 100mV potential (arrows).

A direct way to evaluate the number of labeled Aβ molecules in a given membrane-bound oligomer is by counting the number of steps observed when the oligomer is photobleached by the laser light. While we have used this approach in the past (47), we found it to be inadequate for the current studies since some of the peptide under study is rapidly diffusing and these species are present in all the BLM raster scanned surfaces where they provide a fluorescence background. This makes the determination of oligomer size by the bleaching technique impractical since during a bleaching experiment the constant infusion of new background fluorescence creates an uninterpretable bleaching profile. In addition, as the oligomers become larger, they also become progressively more difficult to analyze for size by photobleaching, since the relative size of individual bleaching steps diminishes.

Estimating an oligomer size from its fluorescence intensity relative to that of a monomeric species was also not adequate since we found the fluorescence to be progressively quenched as oligomers grew. To evaluate an oligomer's size, therefore, we used the lifetime-normalized fluorescence intensity and compared the oligomer's fluorescence intensity (F) and its associated fluorescence lifetime (τ) to those of individual HL488 dye molecules adsorbed on glass. The number of peptides in the oligomer (N) was calculated as:

$$\mathrm{N}=[{\mathrm{F}}_{\text{oligomer}}∕{\mathrm{F}}_{\text{dye}}]\cdot [{\tau }_{\text{dye}}∕{\tau }_{\text{oligomer}}].$$

We observed that discrete species up to 18-20 do not require lifetime normalization (i.e. have intrinsic lifetime of dye) while the extended structures can have highly attenuated lifetimes.

A wide-field CCD camera was used to verify that all the fluorescent species detected were contained within the raster scanned area, and an example of oligomer size determination using the above equation is presented in Fig. 3A. At an excitation power of 5 μW and a diffraction limited beam profile, a single Hilyte 488 dye molecule yields a mean value of 9,000 (±2000) counts per second above background with a 4.1 ns fluorescence lifetime. The intensity of the prominent fluorescence peak observed in Fig 3A (16,000 cps; its value adjusted by subtracting the fluorescence background due to rapidly diffusing species) was divided by intensity observed for the single dye molecule (9,000 cps) resulting in species that most closely represents a dimer of Aβ. For this experiment, the dimer had the same lifetime as the free dye in ensemble measurements or when isolated on a glass surface (or as a monomer on the membrane); τ_dye = 4.1 ± 0.15 ns.

Figure 3. Examples of membrane-bound Aβ species.

The panels show the fluorescence images (linear scale of counts/second) obtained by scanning the BLM surface with 460 nm excitation and monitoring through a 500 nm dichroic filter. The resolution in these 40×40um² scanned areas is near the diffraction limit (~0.5 μm). The emission intensity and fluorescence lifetime of an individual HL488 dye molecule is used to assess the number of fluorescently labeled Aβ molecules associated with a discrete pores on the BLM surface. In the case of pores observed in this figure, the fluorescence lifetime has been within 10% of the lifetime of the free chromophore. Based on these numbers we estimate that the structure on the BLM in panel (A) contains a single peak with 2 monomers and has conductivity less than 0.3 pS (ie., below our detection limit). Rescanning of this surface indicated this peak exhibited a diffusion rate below the current measurement limit of 5×10^-4 μm²/sec. (B) Non-conducting BLM surface with membrane bound Aβ40-HL488. Average fluorescence intensity corresponds to approximately 4-5 dye molecules per μm² surface area (upper red surface). Lower surface (blue) shows the background from the BLM in absence of labeled Aβ40-HL488 (C) Two distinct, small, oligomers. Based on the lifetime-normalized fluorescence intensity, both most closely matched Aβ dimers. No membrane conductivity was detected. (D) Example of a single membrane permeabilizing oligomer with a conductivity of 2 pS. The lifetime-normalized fluorescence intensity corresponds to an Aβ hexamer. (E) Several discrete structures are observed with a total conductivity of 80 pS, the most intense fluorescence peaks correspond to structures of approximately 12-14 dye molecules. (F) At 250 nM Labeled Aβ40-HL488, extended and a discrete structure coexist on the membrane. The total conductivity is 150 pS. (G) Histogram of 257 discrete species including 103 species observed without corresponding conductivity data. The data at 20 oligomers represents species larger than 20.

Experiments to characterize Aβ–HL488 interactions with the phospholipid membrane revealed three distinct classes of bound peptide species, as depicted in Figure 3B-3F; 1. Small, monomeric, species that are homogeneously distributed throughout the membrane surface and shown in Figure 3B; 2. Small, but discrete, oligomers seen as fluorescent spikes in Figure 3C-3E; and 3. Spatially extended structures whose size is above the diffraction limit, as seen in Figure 3F.

The class-1 Aβ species is present in all the samples shown in Figure 3B, and produces an elevated fluorescence background (red surface) over that of peptide-free membrane samples (blue surface), that arises from monomeric peptide binding and freely diffusing across the membrane surface. This elevated fluorescence is evenly distributed across the membrane surface, is stable for a period of several hours and does not change membrane conductivity, which remains at basal level (below 0.3 pS). This result suggests that Aβ binding promotes membrane stabilization.

The lifetime-normalized fluorescence intensity of the class-1 membrane bound Aβ-HL488 shown in Figure 3B corresponds to 5 monomers of the labeled peptide per μm². Assuming a surface area of ca 70A² per lipid molecule we calculate a lipid/peptide ratio of ca 2.8×10⁵. Hence, only a small fraction of the lipid appears to be actively binding the Aβ. Analysis of the fluorescence intensity from unit areas of 0.5 μm², done across the 40×40 μm² surface shown in Figure 3B, reveals intensity variations that are only marginally above those expected for Poisson statistics, which is consistent with a uniform distribution of monomeric fluorophores. The diffusion constant of Aβ 40-HL488 on a POPC/PG model membrane is 2.3 μm²/sec (Ding et al, manuscript in preparation) as determined by incubating 50 nM Aβ with the membrane, washing solution peptide off and then performing a FRAP measurement. This is consistent with a rapid movement of labeled Aβ leading to homogeneous fluorescence intensity in our 50 ms raster scanned data. Our observations support a model where monomeric Aβ binds to the membrane with high affinity and is delocalized to form a rapidly diffusing uniform layer (at low peptide/lipid ratio), but does not induce measurable conductivity (above background of 0.3 pS). It should be noted that in the work of Ding et. al. (manuscript in preparation) incubation of POPC/PG membranes at higher Ab40-HL488 concentrations also resulted in discrete and immobile fluorescence species consistent with these observations on agarose supported BLM surfaces. High affinity binding of Aβ1-40 to model membranes has also been reported by Kremer & Murphy (48).

Examples of class-2 Aβ oligomers are shown in Figure 3C-3E revealing fluorescent species that most closely match two dimers (3C), a hexamer (3D), and many oligomers (3E), respectively. These discrete membrane-bound species are all below the optical diffraction limit, however using their lifetime-normalized fluorescence intensities we conclude that dimeric Aβ shows no conductivity. The smallest structure that we observed to possess a detectable conductivity ( ~ 2 pS; Figure 3D) is about a hexamer. While systematic errors (e.g., steric hindrance of complete rotational averaging, partial static quenching, etc) could modify this number somewhat, we note that structures of comparable size resembling pores, composed of Aβ peptides, have been previously detected by AFM (17) though those structures could not be tested for membrane interaction or permeabilization, We further note that we only observe conductivity when we simultaneously observe immobilized oligomers > 5-8 monomers. We note that even the non-conducting dimer shown in Figure 3A is effectively immobilized (with a diffusion constant below the current measurement limit of 5×10^-4 μm²/sec). The dramatically reduced mobility of these species may be associated with interactions of the oligomers with the underlying agarose surface. Finally, we note that this data is in reasonable agreement with the Aβ40 oligomer size of 6-8 monomers in the model proposed by Arispe et. al. (8).

Interestingly, recent work by Selkoe's group (49) shows that the dimeric and possibly trimeric (50) forms of Aβ leads to modified plasticity of the synaptic junction in neuronal cells. While we can only quantitate membrane permeabilization by relatively larger oligomers, it is possible that the dimeric species aggregates on the neuronal membrane to form larger, and toxic, oligomers. While neuronal toxicity may arise from mechanisms other than membrane permeabilization (the property determined in our experiments), this hypothesis is supported by the observation that Aβ both permeabilizes the cell allowing the influx of calcium and, by some mechanism, causes cell death as discussed above. We also note that our membrane system is a simplified model for the neuronal membrane and may not fully represent the latter's behavior.

It is also interesting to note that we observe significant heterogeneity in the size and conductance of these small membrane-bound oligomers. For example, some of the structures shown in Figure 3E correspond to 12-14 dye molecules and, from the total conductivity of 80 pS, we estimate that each of these possesses higher conductivity than the structures in Figure 3D. The data clearly shows the existence of multiple types of conducting Aβ oligomeric structures.

When higher Aβ concentrations (>>100nM) were incubated with the BLM we observed a new, much larger, lipid-bound peptide species, denoted above as class 3. Figure 3F shows a membrane surface after incubation for a few minutes with 250 nM Aβ40-HL488 revealing large areas of disrupted membrane (well beyond the diffraction limit) that evolve in time and that exist simultaneously with discrete (class 2) structures of Aβ on the membrane. The combined membrane conductivity of this sample was 150 pS and that is clearly mostly due to the extended structure. The observation that the two Aβ oligomer types can coexist on the BLM indicates that conditions for generating these disparate species are not mutually exclusive. A histogram of the size of the discrete structures is illustrated in Figure 3G (including 103 species without corresponding conductivity data). It can be seen that the larger oligomer species are increasing less common.

The small, class 2, Aβ structures are static and do not change over several hours. We have on occasion observed the discrete species to remain intact after 16 hours. The BLM stability is dependent upon the characteristics of the Teflon probes used to form the membranes. However we have consistently observed that the extended class 3 structures are dynamic and can evolve with time. This is demonstrated by the progression seen in the consecutive scans in Fig. 4A and in the corresponding evolution of conductivity shown in Fig. 4B. Here, 1 μM Aβ40-HL488 was incubated for several minutes with the BLM. The initial scan in Fig. 4A is associated with a average membrane conductance of 230 pS, while in the subsequent scans, taken over the following 30 minutes, the fluorescence profile evolves and changes in several locations on the membrane. The overall conductance significantly increases until the membrane is entirely disrupted and the conductivity reading is off scale, which occurs at the very termination of Figure 4B (after the final scan in Figure 4A was completed). The heightened fluorescence intensity of the extended structures is generally associated with shortened fluorescence lifetimes (below 1 nanosecond compared to 4.1 nanosecond for small HL488 Aβ oligomers), which may reflect a different environment for the HL488 or be the result of efficient intra-aggregate energy transfer due to closer positioning of the fluorophores.

Figure 4. Time progression of extended pore structures.

1 μM Aβ40-HL488 is incubated with the BLM and forms extended pore structures that are dynamic in time. Raster scans of the 40 × 40 μm² BLM surface were initiated at approximately 1, 22 and 42 minutes (corresponding to solid lines labeled 1,2,3 in 4B) with a holding potential of 50mV, and show BLM surfaces with conductivities averaging (1) 230 pS, (2) 280 pS and (3) 600 pS respectively. The dashed lines in (A) indicate regions that underwent significant evolution in surface morphology. These membranes often disintegrate when the conductivity exceeds several nS for a 50 μm radius BLM. Since our resolution of conductivity is nominally 0.3 pS, the variation seen in (B) represents real changes in BLM conductivity. While some changes in fluorescence lifetime have been noted in the extended structures, it is not presently clear how much the intensity increase is associated with changes in the fluorescence quantum yield or increased concentration of labeled Aβ40–HL488.

To complement the measurements with fluorescence labeled Aβ, we performed experiments using ThT that fluoresces upon binding to the beta structure of fibrils. To study the evolution of these structures in ensemble measurements, Figures 5A&B show ensemble steady state and time-resolved fluorescence and fluorescence anisotropy measurements of ThT bound to Aβ40 that has been incubated for different lengths of time in solution. ThT was added to freshly solubilized Aβ40 at 1 μM and the sample was then incubated for 4, 22 and 46 hours at 37°C in 10 mM sodium phosphate, 100mM sodium chloride pH 7.4. Initially, for ThT added to freshly solubilized 1 μM, the Aβ40 has a fluorescence lifetime of 30 psec that is indistinguishable from that of free ThT (data not shown). During incubation, the steady state ThT fluorescence is progressively enhanced and the fluorescence lifetime becomes broadly distributed indicative of a heterogeneous mixture. The fluorescence polarization anisotropy indicates binding to a species with a rotational correlation time of about 15 nsec, quite distinct from that of monomeric Aβ40 in solution (ca 5 ns) and corresponding to oligomers with molecular weights in the range of 25-40 kD. Further incubation of the peptide in solution (for 46 hr) shows ThT binding to a species with a very long rotational correlation time (> 50 ns, the upper limit of the operating conditions of the instrument) consistent with a large (fibrillar) structure(s). Aβ40 incubated for long periods has an attenuated ability to form membrane permeabilizing pore structures, as most of the peptide is incorporated into fibrils.

Figure 5. Interaction of Aβ with BLM followed by ThT binding.

(A) steady state fluorescence of 2 μM ThT (excitation 440 nm) freshly added with Aβ40 incubated 4, 22 and 46 hours at 37°C in 10 mM sodium phosphate, 100mM sodium chloride pH 7.4. (B) Time resolved fluorescence decay and time resolved anisotropy of ThT added to Aβ after 4 hours (blue) and after 46 hours (red) in solution. Excitation of ThT at 425 nm, emission monitored at 480 nm. The formation of Aβ fibrils is indicated by the large increase in the ThT fluorescence lifetime (and quantum yield) and the time independent anisotropy decay reflecting a rotational correlation time in excess of the fluorescence decay window (>50 ns rotational correlation time). The ThT binding species at 4 hours is best fit with a 15nsec rotational correlation time. (C) ThT binding to unlabeled Aβ40 on 40 × 40 μm² BLM membrane. Fresh Aβ40 was incubated with membrane for 10 min, and ThT (5 μM) was added. Conductivity of the BLM was 20 pS.

Figure 5C is intended to complement the higher concentration data in Fig. 4, showing data obtained with unlabeled Aβ40 incubated in the presence of ThT. The results were obtained with a freshly dissolved 100 nM Aβ40 sample interacting with BLM for 10 min in the presence of 5 μM ThT. Discrete ThT-stained species are detected by their fluorescence and the resultant permeabilization leads to a total BLM conductivity of 20 pS. The ThT fluorescence shows that the binding of Aβ to the model membrane leads to formation of beta structure. Furthermore, given the concentration and conductivity, we believe this may correspond to some of the small structures seen with fluorescently labeled Aβ in Fig. 3C-3E.

From the results presented above we conclude that interaction of Aβ with the phospholipid membrane greatly accelerates oligomer and β-structure formation (both appear within minutes compared with hours in solution). It is possible, then, that both the small, class 2, and extended, class 3, peptide aggregates bind ThT and therefore are likely to contain at least some β-structure.

Discussion

The association of Aβ with lipid membranes is driven by a combination of electrostatic and hydrophobic interactions (26, 51). Monomeric Aβ in solution is predominantly disordered (1, 2) but develops increasing amounts of secondary structure upon binding to negatively charged lipids (26, 51). Recent work in our laboratory using circular dichroism in ensemble studies (Wong et. al., (52)) indicates that at low protein/lipid ratios (<1/50) Aβ assumes a mostly α-helical conformation while higher protein/lipid ratios induce a transition to β–sheet conformation. At the low protein/lipid ratios the α–helical structure is supported by the interaction of positively charged residues of amphipathic α–helices with negatively charged lipids (53), while at higher Aβ/lipid the β sheet structure is stabilized by protein-protein interactions on the membrane surface. At moderate peptide concentrations, free diffusion on the membrane surface and association of α-helical monomers supports the formation of oligomers that can then adopt a new conformation capable of permeabilizing the lipid membrane (54, 55). This model suggests that membrane permeabilizing species can be generated by lipid induced conformational changes of Aβ followed by additional peptide-peptide interactions inducing further structural rearrangement of the Aβ.

Our observation that Aβ disrupts lipid membranes by two distinct mechanisms is consistent with reports of heterogeneous channel populations based on conductivity measurements (13, 24, 37). For example, Kourie et. al. used electrical conductivity measurements of Aβ bound to lipid membranes and reported dramatically heterogeneous conductance and gating characteristics. (37, 38). The ability of membrane active peptides to form oligomers with different permeabilization capabilities has previously been observed with the antimicrobial peptide cecropin which allows the conductance of ions at concentrations of peptide lower than those necessary for the passage of larger molecules such as β-galactosidase (39). High protein/lipid ratios have also been reported to facilitate Aβ incorporation into the hydrophobic region of the membrane resulting in membrane thinning, and the subsequent reduction of the dielectric barrier to ion translocation (29, 56).

Our measurements, however, show conductivity arising when distinct peptide aggregates are bound to the membrane as seen in the low concentration data of Fig. 4. It is possible, though, that the large class 3 structures produce large-scale membrane “thinning”. It is important to also note that at physiologically relevant concentrations of Aβ (this peptide has been shown to form toxic species at concentrations as low as 10 nM (34, 57)), we only observe class 2 aggregates on the membrane and these discrete species are thus most likely the relevant ones at physiological concentration of Aβ40.

A critical, and hitherto unresolved, question is whether membrane permeabilizing Aβ species evolve on the phospholipid membrane (16) or develop in solution and subsequently bind to the membrane. Numerous studies have shown that the formation of β–sheet rich amyloid fibril by Aβ40 is preceded by a complex series of peptide aggregation events that take place in solution and involve soluble Aβ40 oligomers (58-60). While in solution Aβ oligomers contain relatively little β structure, it has been suggested that the membrane-bound toxic forms of Aβ adopt a β–sheet structure (61-63). An increase in ThT fluorescence supports this hypothesis since increased fluorescence of this dye reflects binding to the β-sheet of amyloid fibrils (64) or protofibrils (6).

The results of the current study demonstrate that the interaction of Aβ with phospholipid membranes involves distinct and potentially competing mechanisms for forming oligomers and membrane permeabilization. The monomeric species do not induce any measurable conductance. We also observed membrane bound Aβ dimers; however, under the experimental conditions used in our study, these did not induce measureable conductivity in the membrane. The neurotoxicity of the dimeric species reported by Shankar et al. (49) may, of course, be due to differences between the model membrane used here and the cellular membrane, or it is possible that in vivo dimers and trimers (50) may more readily assemble into larger-conducting species on the cell membrane. Alternatively, the neurotoxicity reported in (49) may be due to a mechanism other than membrane permeabilization.

The membrane disruption by the extended class 3 structure resembles the one observed for antimicrobial peptides referred to as the ‘Carpet Mechanism’ (Shai-Matsuzaki-Huang; SMH model, or the “self-promoted uptake” of Hancock (65)), where the antimicrobial peptide carpets the bilayer leading to sequestration of lipids in a detergent-like interaction (65, 66). Peptides that interact with membranes through the SMH mechanism generally do so at micromolar concentrations (66), well above physiological levels for Aβ. Hence, the discrete small membrane permeabilizing species and the more extended membrane disrupting structures seen by us may be initiated and facilitated under different conditions.

An extended model presented by Sokolov et al. (56) describes an Aβ-membrane interaction reminiscent of the SMH membrane disrupting mechanism. The SMH, or ‘self-promoted uptake’ model, describes a process whereby a protein homogeneously carpets a membrane surface disrupting membrane structure and increasing membrane conductance. Our observation of the time course for aggregate evolution on the BLM (Figure 4) revealed a dynamic rearrangement in the fluorescence intensity suggestive of a lateral mobility of Aβ-HL488 to be recruited into a structure associated with enhanced ionic conductivity. It remains unresolved whether the extended disrupted membrane structures are derived from the coalescence of discrete pores or recruitment of monomeric membrane bound Aβ-HL488.

Our observation of binding of Aβ to the membrane and the rapid appearance of β-structured ion conducting oligomers (relative to the oligomerization rate in solution) is in line with Huang's two-state model. This model describes a process whereby antimicrobial peptides develop pores by a process initiated by monomer binding to the membrane surface followed by surface diffusion and subsequent assembly into discrete pore structures (54). In the case of Aβ, β-sheet interactions in an oligomer may be required to stabilize the structure and to provide stability against dissociation back to monomers once it binds to the membrane.

Interestingly, there is a major difference in the background of the images seen in Figs. 3B and 3C in the transition from uniform and highly diffusive binding to the formation of small oligomers. Namely, the background level shows a significant increase in the fluorescence fluctuations. The intensity peaks are not large enough to correspond to the dimers seen in the large stationary peaks, but are larger than expected for monomers. Since the images in this work are obtained by raster scanning (50 msec/pixel) and are not wide field images, rapidly diffusing species may produce smaller intensity peaks due to movement of the chromophore through the observation point. Therefore our explanation is that these fluctuations correspond to small oligomers that have either associated briefly reducing their mobility and then dissociated or simply continued to diffuse on the surface with reduced mobility and have not yet inserted into the membrane and become fixed (as have the dimers and larger oligomers that are shown as strong peaks in the data). Additional work is necessary to correlate the degree of membrane permeabilization with oligomer size and mobility.

In summary, we observed Aβ interaction with BLM membranes ranging from a mechanically stabilizing and non-conductive interaction at low peptide concentrations to the formation of discrete low conductance structures at physiologically concentrations, and ultimately at higher concentrations, to more delocalized high conductance structures. Our results show that the variations seen in in vitro experiments may reflect a critical dependence on experimental conditions and parameters.

Abbreviations

Aβ40

Aβ 1-40

Aβ42

Aβ 1-42

APD

avalanche photodiode

BLM

black lipid membrane

DiphPC

diphytanoyl-phosphatidylcholine

DiphPS

diphytanoyl-phosphatidylserine

FRAP

fluorescence recovery after photobleaching

HL488

hilyte 488 fluorophore

ms

millisecond

ns

nanosecond

nm

nanometer

pS

picosiemen

ps

picosecond

SMH

Shai-Matsuzaki-Huang

ThT

thioflavin T