A vast number of diseases are characterized by the misfolding of specific proteins, causing these proteins to assemble into a wide array of extracellular and/or intracellular aggregates. Disease-associated protein aggregation often results in extended, β-sheet-rich, proteinaceous fibrillar structures, termed amyloids. These fibrils often have globular, soluble protein aggregate precursors, termed oligomers. These diseases include Alzheimer's disease (AD), Huntington's disease (HD), type 2 diabetes, and many others. 1 For example, one of the hallmarks of AD is the aggregation of the β-amyloid peptide (Aβ) into extracellular neuritic plaques, 2 and HD is caused by an expanded polyglutamine tract in the huntingtin (htt) protein that leads directly to its aggregation into inclusion bodies in affected neurons. 3 Amyloid deposits of amylin are a pathological characteristic of the pancreas in type 2 diabetes. 4 The two-dimensional liquid environment of lipid bilayers can profoundly influence protein structure and dynamic properties, thereby influencing the aggregation state of amyloid-forming proteins. 5 A common factor modulating amyloid formation appears to be the presence of lipid membranes, as fibril formation has been shown to be accelerated in a membrane environment for several amyloid-forming proteins, including amylin 6 and Aβ. [7][8][9][10][11] While lipid membranes may play a role in facilitating aggregation, 12,13 such membranes may be direct targets of toxic mechanisms associated with amyloid formation, such as membrane permeabilization associated with altered bilayer structure due to sequestration of membrane components into aggregates 12,[14][15][16][17][18][19][20][21] or the formation of unregulated pores. 22,23 Aβ is a cleavage product of the amyloid precursor protein (APP) and, as such, contains a portion of the APP transmembrane domain, which can facilitate Aβ's interaction with lipid membranes. The subcellular localization and interaction of htt-containing expanded polyQ domains with membranous surfaces have been well documented and suggest that these interactions play a role in HD pathogenesis. [24][25][26][27][28] Amylin has been shown to disrupt lipid bilayers by forming either defects 19 or pores. 29,30 The association of amyloidforming proteins with cellular membranes may represent a common step associated with a variety of mechanisms of toxicity. The exact chemical composition and lipid constituents contained within a lipid membrane can heavily influence the aggregation process. 31 For example, cholesterol 11,20,32,33 and oxysterols 34 have been shown to modulate the interaction of Aβ with lipid membranes. The interaction of amyloid-forming proteins with lipid membranes may also play an important role in modulating the effectiveness of small molecules designed to inhibit aggregation, as was demonstrated with (-)-epigallocatechin gallate (EGCG), which inhibits amyloid formation in the absence of lipid membranes 35,36 but is less effective at a phospholipid interface. 37 Here, we investigate the interaction of four amyloid-forming proteins, Aβ, htt, a synthetic polyQ peptide, and amylin (Figure S1 of the Supporting Information), with total brain lipid extract (TBLE) bilayers. These peptides represent two potentially different mechanisms of interaction with lipid membranes. The interaction of Aβ with lipid bilayers is driven by an amphipathic character associated with being a fragment of a transmembrane protein. 38 htt, synthetic polyQ peptide, and amylin all contain α-helical lipid binding domains. 39,40 In particular, we used a variation of in situ tapping mode atomic force microscopy (AFM), and scanning probe acceleration microscopy (SPAM), which allowed us to simultaneously determine the morphological and mechanical impact of exposure to amyloid-forming proteins on the bilayers. Tapping mode AFM is a wellestablished imaging technique capable of nanoscale resolution; however, the time-resolved tip/sample forces associated with tapping mode AFM contain information concerning mechanical surface properties, i.e., adhesion, rigidity, energy dissipation, and compressibility. The SPAM technique is able to reconstruct the tapping force for every cantilever oscillation cycle during imaging by taking the second derivative (or acceleration) of the cantilever deflection signal after Fourier transform-based filtering. 41 Furthermore, SPAM can operate in aqueous solution, making it highly applicable to biologically relevant samples. Numerical simulations of the entire tapping mode AFM scanning process are also provided to assist in the interpretation of the acquired tip/sample tapping forces.
Preparation of Defect-Free Bilayer Patches. Lyophilized total brain lipid extract (porcine) was purchased from Avanti Polar Lipids (Alabaster, AL) and resuspended in phosphate-buffered saline (PBS) (pH 7.3) at a concentration of 1 mg/mL. By using an acetone/dry ice bath, bilayers and multilayer lipid sheets were formed by five cycles of a freezethaw treatment. 8 Vesicle formation within the lipid suspensions was promoted by bath sonication for 30 min. All experiments were performed with the same lot of lipids. Twenty-five microliters of the suspended vesicle solution, diluted with 25 μL of PBS (pH 7.3), was added directly to the AFM fluid cell and the cell placed onto freshly cleaved mica. Vesicles were allowed to absorb to the mica and fuse together to form a supported bilayer (40 μm × 40 μm) that was relatively defectfree as assessed by an AFM image (Figure S2 of the Supporting Information). Once the bilayer formed, two or three washes with PBS were performed to remove excess lipid vesicles. For these experiments, the lipid:peptide ratio was approximately between 5 and 15. The uncertainty is due to several factors. First, TBLE is a complex mixture of lipids, so there is not a straightforward way to calculate the moles of lipid components. Second, despite efforts to be as consistent as possible, the efficiency of the washing step after formation of the bilayer may vary, resulting in slightly different amounts of total lipid in the fluid cell.
Preparation of Protein Samples. Wild-type Aβ 1-40 (AnaSpec, San Jose, CA) and human amylin(1-37) (AnaSpec) were prepared using published protocols that result in monomeric solutions. 19,[42][43][44] Aβ 1-40 and amylin were treated separately with hexafluoro-2-propanol (HFIP) to dissolve preexisting aggregates and seeds. The resulting solutions of Aβ 1-40 or amylin were placed under vacuum to remove the HFIP, resulting in peptide films. Aβ 1-40 films were resuspended in 10 μL of dimethyl sulfoxide (DMSO) and vortexed to make 2000 μM stock solutions, which were dissolved directly into PBS buffer (pH 7.3) to a final concentration of 20 μM. The amylin films were resuspended in PBS and vortexed, resulting in a final concentration of 50 μM. htt exon1-51Q was purified from Escherichia coli as a glutathione S-transferase (GST)-htt fusion protein by protocols that result predominately in monomeric protein. 45,46 The fusion proteins were purified from lysates using a BioLogic LP system (Bio-Rad) equipped with a GST affinity column. Gel electrophoresis was used to determine purity and to verify the relevant fractions. The GST moiety was cleaved with Factor Xa (Promega, Madison, WI), which initiates aggregation. Prior to the addition of the cleaving agent, solutions of GST-htt fusion proteins were centrifuged at 20000g for 30 min at 4 °C to remove any preexisting aggregates. The concentration of htt exon1-51Q was determined using a Bradford assay, and solutions of htt exon1-51Q were prepared, with a final concentration of 20 μM. Synthetic polyQ peptides (Keck Biotechnology Resource Laboratory, New Haven, CT) were dissolved for 3 h in a 1:1 mixture of trifluoroacetic acid (TFA) and HFIP to a concentration of 0.5 mg/mL, which results in a monomeric stock based on published protocols. 47 After the mixture had been vortexed, the solvent was evaporated off, producing a thin peptide film. The polyQ peptide film was resuspended in ultrapure water adjusted to pH 3 with TFA to a concentration of 2.0 mg/mL. This stock solution was diluted in PBS to a final concentration of 20 μM and a pH of 7.3. For Aβ, amylin, and polyQ peptide preparations, aliquots of each solution were spotted on mica as soon as they were prepared to verify the absence of aggregates via AFM imaging in air. Bovine serum albumen (BSA) (Thermo Scientific) was diluted in PBS to a final concentration of 20 μM and used directly.
AFM and SPAM Imaging Conditions. In situ AFM experiments were performed with a Nanoscope V MultiMode scanning probe microscope (Veeco, Santa Barbara, CA) equipped with a closed-loop "vertical engage" J-scanner and a sealable tapping fluid cell. Images were acquired with V-shaped oxide-sharpened silicon nitride cantilevers (Veeco) with nominal spring constants of ∼0.5 N/m and a nominal tip radius of <10 nm. Scan rates were set at 1-2 Hz with cantilever drive frequencies ranging from ∼8 to 10 kHz. The free amplitude of the cantilever was ∼25 nm, and the tapping amplitude was set at 70% of this free amplitude. For SPAM experiments, 5 μm × 1.25 μm images were captured with 256 × 64 pixel resolution. Using a signal access module (Veeco) and CompuScope 14100 data acquisition card (Gage, Lachine, QC), cantilever deflection trajectories were captured at 2.5 MS/s and 14 bit resolution with a range of 2 V. The captured cantilever trajectory was filtered using a Fourier transformbased harmonic comb filter. 41 The second derivative of the resulting filtered cantilever deflection trajectory was taken and multiplied by the effective mass, m eff , of the cantilever to obtain the time-resolved-based tapping force between the tip and sample. m eff was determined using a thermal tuning method, 48 performed in solution, yielding the spring constant and resonance frequency of the cantilever, which can be used to determine m eff based on the following equation:
Numerical Simulations of the Complete AFM Experiment. The cantilever motion was described as a single degree of freedom damped driven harmonic oscillator. 49,50 ω
where m eff is the effective mass of a cantilever, b is the damping coefficient, k is the cantilever spring constant, a 0 is the drive amplitude, ω is the drive frequency, D 0 is the resting position of the cantilever base, F ext is the tip/sample force, and z is the position of the cantilever with respect to the surface. In tapping mode, the tip/surface separation distance changes as the cantilever oscillates. While the tip and surface are not in contact for the majority of the oscillation cycle (noncontact regime), the tip intermittently contacts (or taps) the surface during each cycle. In solution, the AFM tip in the proximity of a surface would experience van der Waals forces and electric double-layer forces, which can be described by the Derjaguin-Landau-Verway-Overbeek (DLVO) theory. 51 Our experimental system, however, has a high salt concentration in the buffer, resulting in a short Debye length and negligible forces arising from the electric double layer. 52 As a result, the external force associated with the noncontact regime was approximated using the van der Waals interaction between a sphere and flat surface: 53
where A is the Hamaker constant, R tip is the tip radius, and a DMT is the interatomic distance parameter of a Derjaguin-Muller-Toporov (DMT) potential. 54 When the separation distance between the tip and surface is less than or equal to a DMT , the tip is in contact with the surface, and the external force was modeled with a DMT potential where E tip , E sample , ν tip , and ν sample are the Young's moduli and Poisson coefficients of the tip and sample, respectively.
To simulate the entire scanning process, we implemented a feedback loop containing an integral gain in the model. In practice, the cantilever deflection, rather than the position with respect to the surface, is monitored experimentally. As a result, the feedback loop in the model monitored the oscillation amplitude of the deflection (y) of the cantilever, which is related to the position by ω =-+ y zD a t sin( ) 00 (6) The measured amplitude of the deflection signal was compared to a specified set point amplitude, and the cantilever position relative to the surface was adjusted to maintain the set point. The feedback loop feature can be used to simulate the process of acquiring an AFM scan line in which the surface topography is variable. The resulting model is capable of capturing several features of the cantilever deflection signal and tip/sample force (Figure S3 of the Supporting Information). 55,56 ■ RESULTS AND DISCUSSION Amyloidogenic Proteins Alter Membrane Morphology. To determine the impact of exposing lipid membranes to amyloid-forming proteins, supported TBLE bilayers were used as a model surface. TBLE is comprised of a physiologically relevant ratio of membrane components (cholesterol, gangliosides, sphingolipids, isoprenoids, and both acidic and neutral phospholipids), providing a relevant model surface. The supported bilayers were formed by vesicle fusion on freshly cleaved mica 57 and systematically exposed to a variety of amyloid-forming proteins (Aβ 1-40 , htt exon1-51Q, synthetic polyQ peptide, or amylin). Observations were limited to 40 μm × 40 μm patches of bilayer that were defect-free, at a scale determined by AFM analysis, prior to exposure to amyloidforming proteins. A freshly prepared bilayer was used in each experiment, and the addition of protein indicated our zero time point for exposure. No additional protein was added after the initial injection. Each protein was freshly prepared and was in a predominately monomeric state upon initial injection, although some small aggregates may have been present. Control experiments were performed by exposing TBLE bilayers to protein-free solutions that were identical to vehicle solvents used for each protein. None of these vehicle solutions altered the morphology or mechanics of the TBLE bilayer as assessed by AFM.
A freshly formed bilayer had a smooth appearance with a root-mean-square (rms) roughness of 0.06 ± 0.03 nm. Initially, when the lipid bilayer was exposed to freshly prepared solutions of Aβ 1-40 (20 μM), htt exon1-51Q (20 μM), synthetic polyQ peptide (20 μM), or amylin (50 μM), patches of increased surface roughness developed, indicating disruption of the bilayer structure (Figures 1 and2). When the rms roughness of these regions was measured (excluding the unperturbed regions of the image to normalize for the surface area actually occupied by roughened morphology), the increased roughness induced in the bilayer by all four amyloid-forming proteins was statistically significant (p < 0.05) (Figure 2a). Histograms of the height values associated with every pixel within the region also demonstrate that the bilayer morphology was significantly altered upon exposure to the amyloid-forming proteins (Figure 2b). TBLE not exposed to any protein or TBLE exposed to BSA had tight distributions of height values centered around 0 nm, as zero height corresponds to the top of the unperturbed bilayer. With exposure to amyloid-forming proteins, the histogram deviates from zero to larger values of height, indicating the extent of the bilayer roughness. Small aggregate complexes were often observed within these roughened domains for each protein, but fibrils were not observed (Figure 1, blue arrows). These aggregate complexes were globular in morphology. While these complexes are most likely comprised of both peptide and lipid components, they may represent discrete oligomers. The concentrations used in this study are much higher than those typically observed in vivo for these proteins. This was done to allow for observations over an experimentally feasible length of time. As aggregation is highly dependent on concentration, this must be taken into consideration when interpreting these results.
The disrupted regions of bilayers exposed to Aβ 1-40 took ∼12-18 h to develop and exhibited the largest surface roughness (rms = 1.95 ± 0.38 nm) (Figure 1b). Aβ 1-40 formed aggregate complexes (4-16 nm in height) larger than those formed by the other amyloid-forming proteins in the presence of the bilayer. Other reported globular aggregates of Aβ tend to be smaller than the observed aggregate complexes. Amyloidderived diffusible ligands (ADDLs) have been observed to be 3-5 nm in height, 58 which is on the small end of the range of aggregates observed in this study. Much smaller Aβ oligomers are commonly observed, such as Aβ*56, which has been reported to be ∼1 nm in height. 59 The larger size of the observed complexes may indicate that either lipid components indeed accumulate within the complex or the TBLE lipid environment is promoting the formation of a larger globular aggregate.
Bilayer disruption occurred within 10-25 min of exposure to htt exon1-51Q (Figure 1c). Despite the apparent more aggressive interaction of htt exon1-51Q with the bilayer, the associated rms roughness with disruption was smaller (0.66 ± 0.15 nm) and the size of observed aggregate complexes was typically smaller (2-10 nm in height) compared to those of Aβ 1-40 . While these aggregate complexes associated with htt exon1-51Q may contain some lipid components, they were similar in size to oligomers observed for a htt exon1 construct with 53Q repeats 60 and to htt exon1 purified from transgenic mice. 61 While the rate of interaction with the bilayer appeared to be slower (1-3 h), the synthetic polyQ peptide interacted with the membrane in a fashion similar to that of htt exon1-51Q, with an rms roughness of 0.54 ± 0.13 nm and aggregate complex heights ranging between 2 and 5 nm (Figure 1d). The slower apparent rate of aggregation on the bilayer of the synthetic polyQ peptide compared to the htt exon1-51Q protein was potentially caused by the length of the polyQ domain, which has been shown to alter the rate of aggregation. 60 As amylin interacted with the bilayer with the slowest apparent rate, the concentration was increased to 50 μM, which resulted in membrane disruption within 3 h (Figure 1e). The rms roughness of the amylin-induced disrupted bilayer regions was 0.94 ± 0.36 nm, and the observed aggregate complexes were ∼2-7 nm in height. Similarly sized oligomers of amylin (3.6 nm) were reported from aggregation studies performed in the presence of a DOPC/DPPC/Chol lipid bilayer system, but this was observed at a much lower concentration (1 μM) and after a longer incubation time (30 h). 62 These aggregate complexes that formed on the TBLE bilayer were also much larger than oligomers of amylin that are commonly observed during the lag phase of aggregation in the absence of a lipid membrane. 62,63 Complicating the comparison of these observed aggregates with those reported in the literature is the fact that the heights of the aggregate complexes associated with each protein were measured with respect to the bilayer surface. As a result, the actual heights of the aggregate complexes associated with the exposure to all four peptides may be larger than measured here if the complex is fully or partially inserted into the bilayer. As a control, bilayers were exposed to 50 μM bovine serum albumin (BSA), and no bilayer disruption (rms roughness of 0.07 ± 0.01 nm) or protein aggregation was observed over 24 h (Figure 1f).
The differences in the magnitude of roughening caused by each protein may be attributable to the mechanism of lipid binding associated with each protein. The random coil in the transmembrane and extracellular domains of Aβ can result in deeper insertion into the bilayer, resulting in a larger roughness measurement. The amphiphilic α-helix in htt exon1-51Q, synthetic polyQ peptide, and amylin allow for lipid binding with less insertion into lipid membranes when compared to that with Aβ 1-40 . 21 Partial insertion of amylin into lipid membranes, stabilizing its α-helical character, has been suggested by molecular dynamics simulations. 64 The lateral morphology of the disrupted regions of the bilayer associated with the different amyloid-forming proteins varied. The disrupted regions of the bilayer associated with Aβ 1-40 were often interconnected by meandering, elongated regions. htt exon1-51Q and amylin were associated with round regions of increased surface roughness that were typically between 300 and 500 nm in diameter. The synthetic polyQ peptide also formed these round regions of disruption, but often these regions were as large as 1-2 μm in diameter and generally were fewer in number.
Collectively, these observations suggest that the disruption of membrane structure is a common feature of the interaction between amyloid-forming proteins and lipid bilayers. This disruption should alter the local mechanical properties of the bilayer, compromising its ability to function properly as a membrane. Mechanical properties of lipids have long been thought to correlate with dysfunction and disease [65][66][67] because it is hypothesized that elasticity and adhesion of membranes directly influence cell stability. Cell and/or lipid permeability correlating with dysfunction has previously been suggested with Aβ, 68,69 mutant htt, 28,70 and amylin. 71,72 As a result, analysis of the tip/sample tapping forces was conducted to determine the impact of these amyloid-forming proteins on the mechanical properties of the bilayer.
Features of the Time-Resolved Tip/Sample Tapping Forces Can Be Used To Measure Mechanical Properties. It has been established that features of the time-resolved tip/ sample force associated with tapping mode AFM are dependent on the mechanical properties of the sample. 41,52,[73][74][75][76][77][78][79][80] While AFM topography images of protein/lipid bilayer systems were being obtained, the cantilever deflection signal was captured and used to recover the time-resolved tip/sample tapping force. Features of the force associated with tapping events changed during the imaging processes, suggesting differences in mechanical properties along the surface. In particular, the maximal tapping force (F max ) and minimal tapping force (F min ) per oscillation cycle varied. F max is defined as the peak or largest positive force experienced between the tip and surface during one tapping event, and F min is defined as the largest negative force experienced between the tip and surface (Figure S3a of the Supporting Information). To demonstrate the impact of surface properties on the tip/sample force interaction associated with tapping mode AFM in fluids and to aid in the interpretation of these force interactions, numerical simulations of the entire imaging process were performed as described in Materials and Methods. 41,55,56 For changes in the tip/sample force to be useful in determining the mechanical properties of lipid bilayers disturbed by amyloid-forming proteins, the tip/sample tapping force must be independent of the surface roughness. To determine the role of surface roughness on tip/sample tapping forces, simulations were performed using 500 nm long model surfaces that consisted of two regions of rms roughness of 0.1 nm (similar to the undisturbed bilayer) flanking a region in which the rms roughness was varied (0.5 to 2.0 nm) in successive simulations (Figure S4 of the Supporting Information). The range of rms roughness used in these simulations was comparable to that observed for bilayers disrupted by amyloid-forming proteins. As there is an inherent response time associated with the ability of the feedback loop to restore the tapping amplitude when the tip encounters an abrupt change in surface morphology, the tapping force trajectory appears choppier when the rms roughness increases. However, the average F max and F min do not change as a function of rms roughness (Figure 3a,b). This indicates that as long as the feedback loop is optimized, changes in tapping forces are not dependent on sample topography.
After establishing that changes in F max and F min are independent of surface topography, we performed simulations to determine how the relative rigidity of the surface and its adhesion to the AFM probe altered features of the tip/sample force interaction. Changes in the rigidity of a surface can be modeled by changing the Young's modulus of the sample (E sample in eq 5). As the surface free energy (and adhesive force) between the tip and surface is related to the Hamaker constant by
2.1 10 DMT 21 (7) the adhesive interaction between the tip and surface can be modeled by changing the value of the Hamaker constant. For these simulations, 500 nm long model surfaces that contained a region in which the mechanical properties of the surface could be changed in successive simulations were used. This variable region was flanked by regions that had a Young's modulus of 5.5 GPa and a Hamaker constant of 1.0 aJ. First, a series of simulations were run with the Young's modulus systematically increasing from 0.5 to 5 GPa in the variable region (the Hamaker constant was not changed) (Figure S5 of the Supporting Information). Increasing the Young's modulus results in a less compliant surface and a reduction in the constant, κ eff , in eq 5. As a result, F max increased in magnitude with larger values of the sample Young's modulus with a power law dependence (Figure 3c). F min was independent of changes in the surface Young's modulus (Figure 3d). Next, a series of simulations were run with various values of the Hamaker constant (0.1-1.75 aJ) in the variable region of the model surface, and changes in F max and F min were observed (Figure S6 of the Supporting Information). As the Hamaker constant was increased, the attractive component of the tip/sample force became larger, and the magnitude of F min increased (became more negative) in a linear fashion (Figure 3e). The larger attractive force also caused the tip to push deeper into the surface. As a result, the Hertzian portion of the tip/sample force during contact increased, causing F max to increase in magnitude linearly with larger Hamaker constants (Figure 3f). Collectively, the simulations with varying values of the Hamaker constant or Young's modulus demonstrated that unique features of the tip/ sample tapping force can be used to determine relative changes in the mechanical properties of the bilayer surface. Amyloidogenic Proteins Alter the Mechanical Properties of Lipid Bilayers. To elucidate the impact of amyloidforming proteins on lipid bilayers, we spatially analyzed the mechanical properties of the membranes on the basis of the time-resolved tip/sample imaging forces associated with tapping mode AFM. To accomplish this, we captured the entire tip deflection trajectory while obtaining an entire AFM image and converted the trajectory into the force interaction by SPAM. This allowed for specific tapping events to be correlated with morphological features of the surface. Spatially resolved maps of F min (Figure 4) and F max (Figure 5) that correspond to each topography image in Figure 1 of lipid bilayers exposed to amyloid-forming proteins were generated. The force images have distinct contrast between the undisturbed regions of the bilayer and the disrupted regions of the bilayer associated with all of the tested amyloid-forming proteins. Histograms of F min and F max for every tapping event can be sorted on the basis of topography images, so that every individual tapping event can be associated with specific regions in the surface. The domains in which proteins have associated with lipid, resulting in disruption, were associated with smaller magnitudes of both F min and F max compared to those of the undisturbed lipid domains. On the basis of simulations, there are several potential implications of these changes in the tip/sample force interaction associated with disrupted regions of the bilayer. The shifts in F min and F max were independent of the rougher morphology caused by the amyloid-forming proteins. As F min was not altered by changes in the Young's modulus of the surface in the simulation, the observed shifts in F min associated with exposure to amyloid-forming proteins were primarily due to a weakened adhesive interaction between the disrupted regions of the bilayer and the AFM tip. The shifts in F max associated with disrupted regions of the bilayer are potentially related to changes in the bilayer's Young's modulus and/or the Hamaker constant associated with the interaction between the tip and surface. Simulations suggested that shifts in F max due to changes in the surface free energy can be estimated from the observed change in F min . Upon comparison of the results associated with F min and F max , the observed shift in F max could not be completely accounted for by what would be expected on the basis of the observed shift in F min . As a result, a portion of the observed decrease in F max associated with disrupted regions of the bilayer can be associated with a decreased Young's modulus. Collectively, these results indicate that interaction between lipid bilayers and amyloid-forming proteins results in regions of altered membrane morphology (increased roughness) and mechanical properties (decreased Young's modulus of the bilayer and a weakened adhesive interaction between the tip and surface).
A potential explanation for these observed changes in bilayers exposed to amyloid-forming proteins is a decrease in the efficiency of the packing of the lipid components within the bilayer in response to protein insertion or binding and aggregation. This would lead to a rougher bilayer that is more easily compressed and a decrease in the number of potential hydrogen bonds formed between the tip and bilayer surface, leading to the weaker adhesive interaction. Previous studies highlight the effects lateral tension has on lipid structure and stability. 81,82 These studies suggest that because of the material properties of the lipid bilayer (thickness, intrinsic curvature, and Young's modulus) the membrane can serve as an allosteric regulator of membrane protein function. Furthermore, lipids naturally compress and/or stretch to minimize the interaction between hydrophobic domains of a transmembrane protein and the aqueous environment. Thus, once the amyloidforming proteins associate with the bilayer, the bilayer may stretch itself to minimize the hydrophobic effect, allowing more space between the two tail ends of the lipid leaflets. A larger space between the two leaflets would be consistent with our results because this would result in greater compressibility. The primary limitation associated with the use of supported TBLE bilayers is the potential that interactions with the underlying mica substrate would influence the physical properties of the bilayer. While there is generally a thin water gap (∼0.2-1 nm) between the substrate and the supported bilayer, [83][84][85] extensive hydrodynamic coupling between the bilayer and the substrate is still possible, reducing the diffusion coefficient of the supported bilayer. 86 As there is an apparent link between compressibility and lateral fluidity, 87 hydrodynamic coupling with mica may result in an altered compressibility of the bilayer; however, the relative changes, with regard to mechanical properties, between the undisturbed and disrupted regions of the bilayer would be similar.
Our results and simulations demonstrate that F min and F max of the tip/sample tapping force can be directly related to material properties of the surface. As a result, the mechanical properties of biological surfaces, such as lipid bilayers, can be mapped with nanoscale spatial resolution while simultaneously obtaining the surface topography in a tapping mode AFM experiment. Currently, this technique is limited to measuring relative shifts in properties within a single AFM image because of difficulties in controlling for variations in probe properties and imaging parameters. Despite these limitations, we have determined the morphological and mechanical changes associated with exposing lipid bilayers to amyloidogenic proteins. As membrane elasticity directly influences membrane stability, such mechanical changes can lead to membrane dysfunction and disease. 66 Aβ 1-40 , htt exon1-51Q, synthetic polyQ peptide, and amylin all caused regions of morphological disruption of TBLE bilayers. While the extent and size of these regions varied, disrupted regions of the bilayer had similar alterations in their mechanical properties. Specifically, these regions were more compressible and adhered more weakly to the AFM probe. The similar physical impact of these amyloid-forming proteins on bilayers suggests that this may represent an intrinsic amyloidlipid interaction with potential implications for a common toxic mechanism.