Amyloid-β oligomers have a profound detergent-like effect on lipid membrane bilayers, imaged by atomic force and electron microscopy

The ability of amyloid-β peptide (Aβ) to disrupt membrane integrity and cellular homeostasis is believed to be central to Alzheimer's disease pathology. Aβ is reported to have various impacts on the lipid bilayer, but a clearer picture of Aβ influence on membranes is required. Here, we use atomic force and transmission electron microscopies to image the impact of different isolated Aβ assembly types on lipid bilayers. We show that only oligomeric Aβ can profoundly disrupt the bilayer, visualized as widespread lipid extraction and subsequent deposition, which can be likened to an effect expected from the action of a detergent. We further show that Aβ oligomers cause widespread curvature and discontinuities within lipid vesicle membranes. In contrast, this detergent-like effect was not observed for Aβ monomers and fibers, although Aβ fibers did laterally associate and embed into the upper leaflet of the bilayer. The marked impact of Aβ oligomers on membrane integrity identified here reveals a mechanism by which these oligomers may be cytotoxic.

The amyloid cascade is a widely accepted description for the molecular basis of Alzheimer's disease (AD). 2 This hypothesis is centered on a small hydrophobic peptide, ␤-amyloid, which is typically 40 or 42 amino acids in length (A␤40/42). The selfassociation of A␤ and subsequent interaction with the lipid bilayer are thought to cause disrupted synaptic function; the resulting loss of cellular homeostasis ultimately leads to hyperphosphorylation of tau, cell death, and dementia (1). A␤ fibrous plaques are abundant in the AD brain and have high lipid content (2,3).
The interaction of A␤ with lipid membranes has been expertly reviewed (4 -6) and draws parallels with the mechanism of toxicity established for antimicrobial peptides (6). Importantly, A␤42 oligomers are reported to perforate the cellular membrane by forming single, discrete ion channel pores (7). Alternatively, a carpeting effect of A␤ has been proposed, which is thought to result in a general increase in membrane conductance either by membrane thinning or a lateral spreading of lipid headgroups (8 -10). Studies have also revealed that A␤42 can extract and incorporate lipid into growing A␤42 fibers by an effect likened to the action of a detergent, which has the potential to rupture the lipid bilayer (11). Some of these effects have been observed for other amyloid-forming proteins and include amylin (12)(13)(14), ␣-synuclein (15), and a prion protein fragment (16), which suggests a shared mechanism of membrane disruption. Nevertheless, a clear picture of the impact of A␤ on membranes is lacking, perhaps in part due to insufficient assembly state characterization and separation of A␤ oligomers from monomers and fibers.
Atomic force microscopy (AFM) is particularly well-suited to probe the topography of supported lipid bilayers (17). However, many AFM studies have focused on how the lipid bilayer can influence A␤ fiber assembly (5, 18 -21). AFM studies that investigate the impact of A␤ on the bilayer structure are only starting to be explored (22)(23)(24)(25).
Here, we report a systematic investigation of the effects of both A␤40 and A␤42 isoforms on lipid bilayers. The A␤ isoforms have been studied in three broad assembly states: monomers, prefibrillar heterogeneous oligomeric assemblies, and fibers. We have employed AFM and transmission electron microscopy (TEM) to show that these different A␤ structures have profoundly different impacts on membrane integrity, which can explain the diverse effects reported.

Results and discussion
Lipid bilayers were deposited onto muscovite mica surfaces from an aqueous mixture of phosphatidylcholine (PC), cholesterol, and GM1 ganglioside (68:30:2 by weight, a ratio chosen to mimic the typical composition of the extracellular face of a cellular membrane). A liposomal suspension in aqueous buffer was incubated with the negatively charged mica, allowing the lipid headgroups to adhere and fuse with the mica substrate to form a double-leaflet lipid bilayer. The remaining suspension was then blotted off the mica and air-dried. AFM images indicate this to be an effective method of generating flat, continuous patches of lipid bilayer with a sharp "cliff-edge" boundary between the membrane patch and lipid-free mica, as shown in Fig. S1. A mean bilayer height has been measured at 4.6 Ϯ 0.1 nm, as would be expected for this lipid composition.
To understand the nature of the interactions between monomeric, oligomeric, and fibrillar A␤40/42 with lipid bilayers, it is This work was supported by BBSRC Grant BB/M023877/1. The authors declare that they have no conflicts of interest with the contents of this article. This article contains supporting Experimental procedures and Figs. S1-S13. 1

ACCELERATED COMMUNICATION
important to isolate and characterize the different A␤ assembly types, see "Experimental procedures"; Figs. S2 and S3 for details. We first deposited size-exclusion chromatographicallypurified monomeric A␤40 and A␤42 preparations (10 M) onto the lipid bilayer. After 60 min on the bilayer surface, the A␤ solution was blotted off, and the mica was allowed to air-dry for imaging. It is clear that neither monomeric A␤42 nor A␤40 has a detectable impact on the integrity of the bilayer, as seen in Fig. 1a and Fig. S4a. There is an even distribution of lipid with minimal disruption both on the lipid surface and at the lipid edge, where a very sharp cliffedge transition between the lipid-bilayer and the mica is observed. Moreover, the bilayers seen in these images are indistinguishable from the bilayer treated with aqueous buffer.
In contrast, prefibrillar heterogeneous oligomeric preparations of A␤42 and A␤40 (10 M monomer equivalent) have a very disruptive impact on the integrity of the bilayer (see Fig. 1b and Fig. S4b, respectively). In particular, there is widespread extraction of lipid from bilayer islands, suggesting a very strong detergent effect by A␤ oligomers. Furthermore, small clumps of lipid appear deposited on the surrounding mica in close proximity to the more substantial lipid islands. This is evident in membrane height cross-sections (see Fig. 1d and Fig. S4d). The impact of A␤42 and A␤40 mature amyloid fibers was much less pronounced than observed for the oligomeric preparations, with almost no A␤ fiber-induced lipid extraction and deposition (no detergent effect). The lipid bilayers remained predominantly intact in the presence of A␤ fibers, as shown in Fig. 1c and Fig. S4c.
We aimed to quantify the extent of the detergent effect for the three A␤ assembly types. This was achieved by measuring lipid coverage at the lipid-edge interface between regions of bilayer and lipid-free mica for multiple (n ϭ 20) preparations ( Fig. 1e and Fig. S4e). These data are presented as percentage of area lipid coverage versus distance from an island of lipid (for details, see Fig. S5). The mean lipid coverage data confirm that neither monomer nor fiber had an impact on the sharp cliffedge transition between the lipid bilayer and mica substrate. Conversely, a very poorly defined transition between the lipid bilayer and the mica was consistently observed upon the addition of A␤ oligomers. The highly disrupted islands of lipid bilayer were surrounded by small deposits of lipid that become less abundant and dispersed at further distances from the main island of lipid, suggesting an extraction of lipid away from the bilayer origin. We observed this consistently in numerous bilayers and A␤ preparations; mean values from typically 20 measurements are shown for each A␤ type. Further examples are shown for A␤42 and A␤40 in Figs. S6 and S7, respectively. Previous studies have utilized A␤-reconstituted liposomes to form an A␤-infused mica-supported bilayer, leading to membrane-spanning pore-like annular-oligomer structures, but with this method of bilayer preparation, the extraction and resulting deposition of lipid was not observed (26 -28).
In addition to the widespread disruption of lipid at the edges of the bilayer, extraction of lipid by A␤ oligomers also resulted in large holes in the membrane that span the bilayer, as reported previously (22). Quantification of these holes indicates a typical diameter of 50 nm (see Fig. 2). A similar range of hole sizes in the bilayer is observed for A␤40 (see Fig. S8). These holes, caused by lipid extraction, should not be conflated with the formation of ion channel pores. The diameters of A␤ ion channels have been shown to be between 1.7 and 2.4 nm based on measurements of their ionic conductance (7), imaging of annular A␤ oligomers that resemble channel structures (26 -28), and also modeling of A␤ barrel-like structures in the bilayer (29,30). The holes observed in Fig. 2 are much larger and would be  ACCELERATED COMMUNICATION: Amyloid-␤ and lipid membranes indistinguishable from membrane patch breakdown in ionic conductance measurements (7). These large holes are also very occasionally observed upon the addition of A␤ fibers; however, their occurrence is relatively rare and may be explained by the presence of a small number of oligomers in the A␤ fiber preparation.
We were also interested in quantifying the extent by which the amyloid fibers embed into bilayers. AFM is particularly suited to measuring lipid bilayer heights, measured to be 4.6 Ϯ 0.1 nm, whereas single A␤ fibers are typically 5.3 Ϯ 0.1 nm in height (see Fig. 3 and Figs. S9 and S10). Measurements of the height of fibers laterally associated to the surface of the membrane indicate the extent to which the fibers embed and displace lipids within the bilayer. Fibers that laterally associate onto the surface of the membrane consistently protrude to a height of 2.5 Ϯ 0.05 nm above the lipid bilayer surface. This indicates that the fibers do indeed embed into the membrane to a depth where the upper leaflet of the bilayer is displaced, as shown in Fig. 3c. It seems that fibers do not displace lipid from the lower leaflet of the supported lipid bilayer; nor do fibers rest on the surface of the membrane to leave the bilayer unperturbed. Lateral association of amyloid fibers on the surface of a lipid bilayer has been reported using a range of techniques, including EM and fluorescence microscopy (11,13,(31)(32)(33). However, this is the first clear evidence that laterally associated A␤ fibers are capable of embedding and displacing the upper leaflet of the bilayer rather than simply residing on the charged surface of the bilayer.
To investigate the extent to which oligomers and curvy linear protofibrils embed into the lipid bilayer, we used AFM to record both topographical and phase-contrast information. Oligomers and curvy linear protofibrils generated here are typically at least 4 nm in diameter. Phase imaging can be used to distinguish differences in viscoelastic properties that in turn can be explained by different surface properties even when there is little change in height topology. In this way, A␤ oligomers can be observed even when they are fully embedded into the membrane (see Fig. S11). Phase imaging of bilayers exposed to A␤ oligomer revealed dark regions within the bilayer that are indicative of increased probe adhesion to membrane-inserted A␤; these oligomers were otherwise poorly visualized by height topography. The oligomer and curvy linear protofibrils have a widespread presence in the membrane and are sufficiently buried to be a similar height as the surrounding bilayer, displacing both leaflets of the membrane, which might then facilitate the observed lipid extraction.
We further investigated the impact of A␤ assemblies on the structure of lipid membranes within large unilamellar vesicles (LUVs) by negative-stain TEM imaging. In agreement with the AFM data, monomers of A␤ have little impact on the membrane, as the vesicles show little difference in appearance from vesicles imaged in the absence of A␤ ( Fig. 4a and Fig. S12). A␤42 oligomers clearly disrupt the shape of the vesicles, causing numerous discontinuities and curvatures within the bilayer ( Fig. 4b and Fig. S12). Peptide-induced membrane curvature has also been reported for anti-microbial peptides whose toxic action is linked to a detergent effect dependent on membrane insertion (34,35). In contrast, A␤ fibers cause less distortion in the vesicle appearance, with only the ends of the fibers causing membrane curvature (Fig. 4c and Fig. S13). This behavior is also apparent for ␤ 2 -microglobulin and amylin amyloid fibers imaged by TEM (13,32).
In summary, nanoscale imaging of supported lipid bilayers has revealed very different impacts on the membrane when challenged with A␤ monomers, heterogeneous oligomers, or fibers (Fig. 5). In the presence of isolated monomers of A␤40 or A␤42, the lipid membrane remains unperturbed, whereas fibers laterally embed into the bilayer to displace the upper leaflet. Notably, only A␤ oligomers cause extraction of lipid from the membrane, explaining the high levels of lipid deposited within extracellular A␤ plaques in the AD brain. In vivo, this detergent effect is likely to impact cellular homeostasis and integrity and is in line with the relative cytotoxicity of A␤ oligomers compared with fibrillar assembly states (36 -38). Cell viability assays report cytotoxicity for both A␤40 and A␤42, although A␤40 is less potent than A␤42 (38). Here we show there is not a significant difference in membrane disruption caused by A␤40 oligomers compared with A␤42, whereas only A␤42 oligomers form ion channels (7). The differences in ion channel formation and lipid bilayer disruption (Fig. 5) suggest there may be two or more mechanisms by which A␤ can disrupt neuronal homeostasis and trigger the cascade of events that culminate in dementia. Multiple modes of toxic action present numerous therapeutic targets; our study indicates that therapeutic molecules that prevent insertion of the A␤ oligomer into membranes may circumvent both mechanisms of toxicity.

A␤ peptides and solubilization
Lyophilized A␤40 and A␤42 peptides were purchased from Cambridge Research Biochemicals and EZBiolab Inc. A␤ pep-   . The mechanisms of ␤-amyloid mediated membrane disruption depends on A␤ structure. A␤ monomers, oligomers, and fibers have distinct impacts on membrane integrity. a, A␤ monomers have little impact on membrane structure. b, only oligomeric A␤42 has the ability to form ion channel pores in cellular membranes (7). c, oligomers of both A␤42 and A␤40 have a widespread detergent effect causing lipid extraction, curvature, and rupture of the membrane bilayer. d, the ability of A␤ to form ion channels or have a detergent effect is lost on assembly into amyloid fibers; instead, fibers can laterally embed into the membrane and displace the upper leaflet of the bilayer. The fiber structure is taken from Protein Data Bank entry 2MXU (44).
ACCELERATED COMMUNICATION: Amyloid-␤ and lipid membranes tides were synthesized using solid-phase Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry, producing a single elution band in HPLC with correct mass verified by MS. Unless otherwise stated, all other chemicals were purchased from Sigma-Aldrich.
Peptides were solubilized in ultra-high quality (UHQ) water (0.7 mg/ml at pH 10) and left at 4°C for 12 h. Samples were then centrifuged for 15 min at 16,000 ϫ g at 4°C to remove any high-molecular weight aggregates. A␤ concentration was determined by absorbance of stock solutions at 280 nm (⑀ ϭ 1280 M Ϫ1 cm Ϫ1 ).

Purification of monomeric A␤ by size-exclusion chromatography
Size-exclusion chromatography (SEC) was utilized to separate A␤ monomer from amyloid assemblies of higher molecular weight. Superdex 75 10/300 GL column (GE Healthcare) was used to purify and elute a monomeric fraction of solubilized A␤ preparations using AKTA FPLC. Stock A␤ solution (90 M) in UHQ water was loaded onto the column, and the monomer peak was eluted with a solution of NaCl (160 mM), HEPES (30 mM) using a 0.5 ml/min flow rate at 4°C. Eluted monomer had a typical concentration of 30 M. SEC-purified A␤ was deemed to be seed-free, evidenced by a single elution band and an absence of observable assemblies by negative-stain TEM. Seedfree A␤, termed here as monomeric, also had no thioflavin T (ThT) fluorescence signal and exhibited a clear lag phase in the nucleation polymerization reaction (see Fig. S2). Monomeric samples were used directly after SEC elution or immediately stored at Ϫ80°C.

A␤ oligomer and fiber preparations
Fiber preparations-Amyloid fiber preparations of A␤40 and A␤42 (10 M monomer-equivalent) were generated in a 96-well plate in aqueous buffer containing NaCl (160 mM) and HEPES (30 mM) at pH 7.4, as described previously (39) (see supporting information for further details). Assembly kinetics were monitored with ThT fluorescence (10 M). Adjacent sample wells with no ThT added were used to generate assemblies for use in all experiments. To remove any low-molecular weight oligomers, fiber preparations were centrifuged at 16,100 ϫ g for 5 min. The supernatant, which contained small diffusible oligomers, was removed, and the fiber pellet was resuspended. At equilibrium, A␤ assemblies had the typical amyloid fibrous appearance according to AFM and TEM (see Fig. S3). Fiber preparations have a low level of A␤ monomer and oligomer content.
Oligomer preparations and characterization-Similarly, A␤40 and A␤42 assemblies with predominantly circular oligomeric structure were obtained from the well plate toward the end of the lag phase, as monitored by ThT fluorescent dye in separate wells. Oligomeric samples were used immediately or stored at Ϫ80°C to halt further assembly. The heterogeneous lag-phase mixture contains a range of oligomer sizes, whereas appreciable monomeric A␤ is still present, as indicated by SEC elusion profiles (Fig. S2d). According to TEM, these oligomeric preparations are heterogeneous in nature but contain a large number of circular oligomeric assemblies that range in size, with typical diameters between 10 and 12 nm, identified by TEM and AFM (see Fig. S3).
TEM micrographs of negatively stained prefibrillar oligomeric assemblies of A␤42 were further characterized using singleparticle analyses. A total data set of 3997 single particles were grouped into 25 class averages (see Fig. S3a). From this data set, 70% were broadly circular in appearance and ranged in diameter between 10 and 12 nm. Using the relationship, volume (nm 3 ) ϭ mass (kDa)/1.27 (nm 3 /kDa), and assuming spherical protein assemblies, approximate masses of between 400 and 700 kDa are indicated. Whereas SEC suggests smaller oligomers are also present (Fig. S2d), the remaining particles appeared elongated up to 16 nm, with some approaching curvy linear status. Some much longer curvy linear protofibrillar assemblies were indeed present, typically 100 nm long. In agreement with our SEC elution profiles (Fig. S2d), it has been shown much of A␤ remains monomeric in the prefibrillar lagphase preparations (40). The nature of the heterogeneous A␤ assemblies present in lag-phase preparations has been reviewed previously (40).

Lipid bilayer preparation
Lipid bilayers were prepared using methods similar to those described previously (31). A lipid mixture of egg PC (Avanti Polar Lipids Inc.), GM1 ganglioside (Avanti Polar Lipids Inc.), and cholesterol (Sigma-Aldrich) solubilized to 5 mg/ml at a ratio of 68:2:30 (w/v), in a 2:1 (v/v) mixture of chloroform and methanol. Organic solvent was then gently removed under a stream of nitrogen and left overnight in a fume hood to form a lipid film.
For AFM studies, the dried lipid film of PC/GM1/cholesterol was rehydrated in an aqueous buffer (160 mM NaCl, 30 mM HEPES, pH 7.4) to achieve a 0.6 mg/ml liposome solution. This vesicle suspension was then tip-sonicated at low power for three 30-s periods, until the cloudy lipid solution was clear, ready for deposition onto mica.
For TEM lipid vesicle studies, LUVs were generated by the extrusion method (31). A 1 mg/ml lipid solution was immediately passed through a lipid extruder (Avanti Lipids Ltd.), which contained a 100-nm polycarbonate filter (Avanti Polar Lipids Inc.), seven times to generate a uniform size. LUVs were typically 100 -200 nm in diameter.

Preparation of supported lipid bilayers on mica and application of A␤
Mica discs (Agar Scientific Ltd.) were cleaved five times with sticky tape, and both the mica and the lipid vesicle suspension were heated to 60°C in a benchtop oven. The heated suspension was then pipetted onto mica discs in a Corning sterile 6-well plate (Sigma-Aldrich). The plate was then sealed in a humid box and incubated at 60°C for 30 min before decreasing the temperature to 25°C at a rate of 5°C every 15 min until reaching room temperature. Slow cooling of the sample enables large micro-islands of bilayer to form on the mica (Fig. S1).
The mica was then washed three times with UHQ water, ready for exposure to A␤. Typically, 200 l of A␤ (10 M) with NaCl (160 mM) and HEPES (30 mM), pH 7.4, was applied by pipette to evenly coat the surface of the mica. The A␤ droplet ACCELERATED COMMUNICATION: Amyloid-␤ and lipid membranes was incubated in a sealed container for 1 h to minimize evaporation and maintain salt/A␤ concentration. The droplet was then removed using filter paper, and the mica surface was washed again three times with UHQ water. Mica samples were finally left to dry in ambient air conditions under the cover of pierced parafilm. The samples were then imaged by AFM. A␤ was also deposited directly onto the mica in the absence of lipid bilayer using a similar protocol.

AFM and TEM; quantification and characterization
Details of the measurements made using both AFM and TEM can be found in the supporting information. In particular, AFM data quantification of the detergent effect, bilayer hole diameters, (41), lateral fiber embedding, and single particle analysis (42, 43) of TEM micrographs are described.