Direct Observation of Single Amyloid-β(1-40) Oligomers on Live Cells: Binding and Growth at Physiological Concentrations

Understanding how amyloid-β peptide interacts with living cells on a molecular level is critical to development of targeted treatments for Alzheimer's disease. Evidence that oligomeric Aβ interacts with neuronal cell membranes has been provided, but the mechanism by which membrane binding occurs and the exact stoichiometry of the neurotoxic aggregates remain elusive. Physiologically relevant experimentation is hindered by the high Aβ concentrations required for most biochemical analyses, the metastable nature of Aβ aggregates, and the complex variety of Aβ species present under physiological conditions. Here we use single molecule microscopy to overcome these challenges, presenting direct optical evidence that small Aβ(1-40) oligomers bind to living neuroblastoma cells at physiological Aβ concentrations. Single particle fluorescence intensity measurements indicate that cell-bound Aβ species range in size from monomers to hexamers and greater, with the majority of bound oligomers falling in the dimer-to-tetramer range. Furthermore, while low-molecular weight oligomeric species do form in solution, the membrane-bound oligomer size distribution is shifted towards larger aggregates, indicating either that bound Aβ oligomers can rapidly increase in size or that these oligomers cluster at specific sites on the membrane. Calcium indicator studies demonstrate that small oligomer binding at physiological concentrations induces only mild, sporadic calcium leakage. These findings support the hypothesis that small oligomers are the primary Aβ species that interact with neurons at physiological concentrations.


Introduction
Soluble oligomeric forms of the 39-to-42 residue amyloid-b peptide, the primary component of Alzheimer's disease plaques, may be a key player in synaptic dysfunction and cell death [1][2][3]. However, studying this system at physiological peptide concentrations presents challenges. Soluble Ab is only detected at nanomolar to picomolar levels in the human brain [4,5] which renders study at physiological concentrations by traditional biochemical methods problematic. Efforts to pinpoint the neurotoxic aggregates have also been complicated by the finding that at physiological concentrations, Ab exists as a mixture of metastable species [6][7][8]. Additionally, Ab-membrane interactions are complex and variable; externalized Ab may bind to specific cellular receptors or protein complexes (e.g. NMDA receptors [9,10] and a7 nicotinic acetylcholine receptors [11]), associate with phosphatidylserine in the membrane [12], or bind to and insert directly into the lipid bilayer [13][14][15].
Due to the complexities of this system, the primary mechanism by which soluble Ab induces toxicity remains unclear. Increased oxidative stress, interference with synaptic transmission, and impaired axonal transport may contribute to the death of affected neurons [1,16]. According to one strongly supported hypothesis, Ab initiates toxic effects by disrupting membrane integrity, either by direct formation of pores in neuronal cell membranes [14,17,18] or by carpeting the membrane, resulting in thinning of the lipid bilayer and loss of ion homeostasis [19,20]. The involvement of oligomeric Ab in toxicity is more strongly established. Levels of various soluble, oligomeric forms of Ab have repeatedly been shown to be a stronger indicator of disease state in humans than plaque load [4,5,21,22]. Stabilized versions of dimers reduce long-term potentiation in cultured neurons [23]. Experiments with cross-linked oligomers have shown that neurotoxicity increases nonlinearly with oligomer size, comparing dimers, trimers, and tetramers [24]. Finally, in one widely cited study, chemically purified oligomers were shown to induce instantaneous calcium leakage in cultured neuroblastoma cells [19].
If toxicity is indeed mediated primarily by extracellular Ab, as many of these studies suggest, binding to some component of the cell membrane is a necessary step in the pathway to functional abnormalities. Understanding the mechanism of binding for living cells exposed to physiological Ab concentrations requires novel approaches. Chemically stabilizing various Ab aggregates in solution may drastically alter normal aggregation rates [25] and membrane-binding behavior, obscuring mechanistic details. Furthermore, the harsh solvents used in certain of these procedures (e.g. hexafluoroisopropanol) may in themselves damage cell membranes, further complicating the problem [26]. A number of groups have begun to study Ab-membrane binding on live cells using fluorescence-based techniques. Using confocal microscopy, Bateman, et al. monitored the formation of fluorescently-labeled amyloid-b  aggregates on living PC12 cells [27]. At low micromolar concentrations, they observed formation of two distinct types of large membrane-bound aggregates. More recently, Nag et al. have studied FITC-labeled Ab40 bound to cell membranes after several minutes' exposure to near-physiological Ab concentrations [28]. Fluorescence correlation spectroscopy and fluorescence lifetime measurements were used to characterize the mobility and membrane insertion of the peptide, but no specific cell-bound oligomers were identified.
Previously, our group has utilized single molecule TIRF (Total Internal Reflection Fluorescence) microscopy to study the size distribution of amyloid-b(1-40) (hereafter referred to as Ab40) oligomers formed in solution at physiological concentrations [8]. We also recently used single molecule intensity measurements to examine which Ab40 oligomers correlate with conductance changes on a model membrane [29]. Here we present single molecule data showing that small Ab40 oligomers formed in solution bind directly to living SH-SY5Y neuroblastoma cells at physiological concentrations. These oligomers, ranging in size from dimers to hexamers and larger, are relatively immobile and widely distributed on the cell body. Using slide-localized oligomers as a cross section of the species present in solution, we compare the sizes of cell-bound aggregates to the distribution observed for Ab40 in solution. We find that cell-bound oligomers include a small proportion of distinctly larger (hexameric and greater) oligomers, demonstrating that oligomers may grow or colocalize following binding to the cell membrane.

Comparison of freshly dissolved HL647cAb40 and unlabeled Ab40 by gel filtration HPLC
In order to examine the cell membrane binding behavior of Ab40, nanomolar concentrations of Ab40, fluorescently labeled at the N-terminus, were applied to SH-SY5Y neuroblastoma cells. The N-terminus of Ab is solvent-exposed and therefore unlikely to be involved in the b-sheet formation which drives fibrillization [30,31]. Also, work done in our laboratory and others indicates that various forms of N-terminally labeled Ab40 behave similarly to unlabeled Ab40 in terms of fibrillization [32], ability to permeabilize synthetic membranes [8,29] as well as rat basophilic leukemia cell-derived membrane blebs ( Figure S1), and toxicity to cultured cells [33]. The current studies were performed using Ab40 labeled with HiLyte Fluor 647 by a C2-maleimide linkage at the N-terminus (termed HL647cAb40). To confirm that this peptide behaves similarly to the unlabeled peptide on solubilization, we ran freshly prepared HL647cAb40 and unlabeled Ab40 on HPLC. Both versions of the peptide elute primarily as a single peak with dimer-to-trimer molecular weight ( Figure S2).

HL647cAb40 at physiological concentrations rapidly binds to SH-SY5Y neuroblastoma cells
To examine the questions of how Ab40 binds to cell membranes and what species are present following binding, we applied HL647cAb40 to live neuroblastoma cells at near physiological peptide concentrations and measured the integrated intensity of cell-bound particles. Cells were imaged following a 10 minute exposure to 50 nM HL647cAb40 in media; in order to obtain cross-sections through the cell membranes, the objective was focused at a z-plane that fell halfway between the apical and basal portions of the cell membrane.
Representative images of cells exposed to HL647cAb40 and unlabeled Ab40 (Figure 1) reveal both internal and surfacelocalized fluorescent particles in the HL647cAb40 samples. Some form of membrane binding is believed to be a key step in toxicity directly mediated by Ab40, and while several groups have observed internalization of fluorescently labeled Ab40 and Ab42 by neuroblastoma cell lines [33,34], only one has reported binding to live cells at physiological concentrations [28]. During analysis, we focused on measuring the size of those particles that were edgelocalized (as this is where membrane-bound oligomers could be most clearly identified).

Freshly prepared HL647cAb40 solutions contain primarily monomers and dimers
To study how Ab40 oligomerization and binding occur at physiological concentrations, we wished to begin with monomeric Ab40. We therefore examined the size distribution of oligomers present in freshly dissolved HL647cAb40 samples by recording photobleach trajectories on dry, spin-coated peptide ( Figure S3; Figure S4). Of spots with clean photobleach traces, 83% 63% bleach as monomers and 15% 63% bleach as dimers (Figure 2A). These results led us to conclude that the majority of freshly dissolved HL647cAb40 is monomeric or dimeric.
Fluorescence intensity calibration enables determination of oligomer stoichiometry in the confocal mode To measure oligomer size on living cells, we developed a protocol to correlate particles' fluorescence intensity values with the number of Ab40 monomers they contain. When laser power is below saturation, the fluorophore emission varies linearly with excitation power. Under these conditions, the slope of total intensity from a given volume versus the number of molecules present in the volume yields intensity-per-molecule. The fluorescence intensity of an oligomer can be divided by this value to yield the number of Ab40 monomers present in the oligomer (assuming the quantum yields of the free dye and peptide-bound dye are the same, as discussed below).
To determine the intensity of a single HL647 dye molecule in solution, a series of confocal scans were performed on various dilutions of HL647 hydrazide. The average intensity of a 1.6 mm by 1.6 mm by 1 mm volume element was measured at 10 nM HL647 hydrazide concentration increments from 0 nM to 60 nM (0 to 102 dye molecules per volume element). In this regime, intensity per number of molecules in a volume element is linear, with a best-fit correlation coefficient of 0.996558. The slope of this line represents an intensity-per-molecule value that can be used to extract HL647cAb40 oligomer size ( Figure S5).
The validity of this calibration method requires the assumption that the fluorescence yields of HL647cAb40 and free HL647 hydrazide in solution are identical. Collisional quenching from the attached peptide is one mechanism that might reduce the efficiency of photon emission of the fluorophore in HL647cAb40. Fluorescence lifetime measurement was used to eliminate this possibility. For dynamic quenching, the ratio of intensity of the quenched fluorophore (I q ) to that of the unquenched fluorophore (I 0 ) is equal to the ratio of the fluorophore lifetimes (t q and t 0 , respectively) under each condition [35]: Accordingly, fluorescence lifetimes of HL647 hydrazide and HL647cAb40 were measured in imaging buffer. Fitting the fluorescence decay curves using the Exponential Series Method yielded lifetimes of 1.5560.24 ns for the free dye and 1.6460.09 ns for HL647cAb40. Hence, the two lifetimes are equivalent within experimental error. Collisional quenching from Ab40 does not reduce the quantum yield of the dye in HL647cAb40.
Other factors may also affect the relative fluorescence efficiencies of the free dye and the peptide-bound dye, when HL647cAb40 aggregates or associates with surfaces. One possibility is that oligomerization of the HL647cAb40 peptide, its insertion into a biological membrane, or its interaction with a surface results in significant quenching. However, we have previously addressed these questions using Ab40 N-terminally labeled with HiLyte Fluor 488 and established that oligomers up to 20 monomers in size have fluorescence decays that fit well to a single exponential, and fluorescence lifetime is not altered for aggregates of this size even after binding to (and insertion in) lipid bilayers [29]. As HL647cAb40, like HL488Ab40, has an Nterminally located fluorophore, it is reasonable that low-order HL647cAb40 oligomers also retain the lifetime of the dye under these conditions.
To further verify the accuracy of this method, a distribution generated with our calibration was compared to one compiled from photobleach trajectories taken on a total internal reflection fluorescence (TIRF) microscope. Samples for both measurements were prepared by dilution of HL647cAb40 to 50 nM in media over a kilned, autoclaved glass slide, which results in oligomers binding to the slide ( Figure S3; Figure S6). The oligomer size distributions obtained using the two methods were highly similar ( Figure 2B). In both samples, approximately 70% of oligomers fall in the monomer-to-dimer range, with roughly 30% trimers to hexamers. These distributions differ substantially from that described above for freshly prepared peptide (98% monomer/ dimer); however, they do resemble those presented in the literature upon for Ab40 in media or physiological buffers at nanomolar concentrations [6][7][8]36]. We note that unlike our earlier studies on dried samples [8], where we demonstrated significant variations in dipole orientation between monomers in a given oligomer, these measurements were made in solution. The data shows that this change allows orientational motion of the fluorophore emission dipoles, leading to the same time-averaged intensity for each monomer in an oligomer. The equal magnitudes of the photobleach steps for oligomers in solution observed here ( Figure  S6) support this assumption.

Small oligomers form in solution and bind to the membranes of living SH-SY5Y cells at physiological concentrations
Even a short incubation in media at 37uC results in some formation of oligomers in the trimer-to-hexamer size range ( Figure 2B). Comparing the size distribution obtained for on-cell oligomers with the on-slide distribution ( Figure 2C) reveals a small decrease in the numbers of dimers and trimers observed on cells. Conversely, particles with intensities corresponding to oligomer sizes from heptamers to 14-mers comprise a significant component of the on-cell distribution (10.6%), while the largest measurable oligomers observed on slides are heptamers (1.5% of the total onslide distribution). The oligomer size distribution on the cell surface is thus shifted towards larger aggregates. (Note: These histograms exclude rarely detected spots with full width at halfmaximum exceeding 250 nm and particles saturating the detector at more than three pixels-20-mers and larger by integrated intensity).
A prominent concern in live-cell single molecule microscopy involves signal contamination by autofluorescence. As an autofluorescence control, we performed our analysis on cells treated with unlabeled Ab40. Figure 2C (inset) shows that the vast majority of detected spots in the HL647cAb40 sample indeed represent fluorescently labeled peptide. Additionally, all of the autofluorescent spots detected by our analysis in the unlabeled Ab40 sample have integrated fluorescence intensities that place them in the monomer-to-tetramer range.
Cell perimeter-localized HL647cAb40 oligomers are membrane-bound and susceptible to potassium iodide quenching Potassium iodide quenching studies were performed to determine whether edge-localized HL647cAb40 oligomers were solvent-accessible. Iodide, a well-known collisional quencher, has formerly been used to characterize live cell-associated Ab40 aggregates [27]. We imaged cells treated for 10 minutes with 50 nM HL647cAb40 before and after the addition of 300 mM potassium iodide ( Figure 3A). Under these conditions, cellperimeter localized peptide was highly susceptible to quenching. Interestingly, the majority of large internalized aggregates in cells exposed to 50 nM peptide overnight did not quench. Subsequent experiments verified that potassium iodide is able to permeate cells and quench fluorophores localized to the cytoplasm (see Figure  S7). Late stage aggregate quenching resistance may represent a fundamental change in aggregate structure [27] or an inability of iodide ions to permeate the endosome-lysosome system.
We also stained cell membranes with a lipophilic membrane dye, DiO, and performed colocalization studies. The vast majority of the HL647cAb40 signal is colocalized with DiO ( Figure 3B), suggesting that HL647cAb40 in and on the cell is nearly all membrane-associated, bound to membrane-associated proteins, or enclosed in membrane-bound vesicles.

HL647cAb40 and unlabeled Ab40 induce minimal calcium leakage
Application of either extracellular Ab at supraphysiological concentrations or of chemically prepared oligomers [17,19] has previously been shown to result in dramatic calcium influx (3-fold or greater increases in calcium indicator fluorescence), and this effect was observed within seconds in a majority of cells. To test whether freshly prepared Ab permeabilized cell membranes to calcium at physiological Ab levels, SH-SY5Y cells were loaded with the fluorescent calcium indicator Fluo4-AM and exposed to 50 nM Ab. The calcium ionophore ionomycin [19] was used as a permeabilization control. Only a small proportion (roughly 10%) of cells treated with unlabeled Ab40 ( Figure 4A) or HL647cAb40 ( Figure 4B) exhibited any increase in fluorescence as compared to controls ( Figure 4C). Observed fluorescence increases were less than 2-fold and occurred gradually over the course of 3 to 5 minutes.

Discussion
Physiologically meaningful experiments on Ab binding to live cells and the resulting toxicity have historically been difficult for reasons previously discussed. Emerging single molecule techniques represent one avenue for overcoming these barriers. Here, we adapt a conventional confocal laser scanning microscope to perform single molecule measurements on slide-localized and cellbound single Ab40 oligomers. By optimizing our imaging parameters, we identified a linear regime in which integrated intensity level for a given volume element was directly proportional to number of molecules present. We confirmed the accuracy of this method by another technique commonly used for measuring oligomer size, total internal reflection fluorescence (TIRF) single molecule photobleaching [37,38].
Using this calibration, we find that HL647cAb40 rapidly forms oligomers in the trimer to hexamer range upon dilution to 50 nM in physiological buffers and exposure to surfaces. Importantly, our treatment did not utilize harsh solvents, unnaturally high peptide concentrations, or chemical modification of oligomer structure, and our initial freshly prepared Ab40 samples contain over 80% monomers (Figure 2A). SH-SY5Y neuroblastoma cells exposed to 50 nM HL647cAb40 both bind the peptide and internalize it, as previously reported [33,34]. Binding to the membrane or membrane-associated structures is a prerequisite for internalization, and if membrane-binding of a specific oligomer population were found to correlate with toxicity, such a species (or its binding site on the membrane) could be specifically targeted. Accordingly, we limited our quantitative study of cell-associated HL647cAb40 to cell perimeter-localized particles.
Significantly, cell-bound HL647cAb40 oligomers include a small number of aggregates (roughly 10%) that are greater in size than the largest particles present on slides ( Figure 2C). Decreases of approximately 20% in both monomers/dimers and trimers/ tetramers are observed on cells as compared to the on-slide data. These decreases reflect a 140% increase in pentamers/hexamers and an 8-fold increase in heptamers to 14-mers. The largest ''oligomers'' may simply be smaller aggregates clustered in specific, high-density peptide binding domains. Alternatively, these structures may represent large oligomers formed from the onmembrane association of several slow-moving, smaller oligomers or from addition of single monomers to bound oligomers. Our recent data indicates that at low (2 nM) Ab40 concentrations, oligomers form slowly on supported anionic lipid bilayers from self-association of diffusing monomers. However, at 100 nM, large oligomers appear much more quickly (within 2 hours), which may indicate that small oligomers binding directly from solution can recruit other species both from solution and from on the membrane (Hao Ding, unpublished observations).
We found no evidence for a diffusing, uniformly distributed population of monomers and dimers. However, if the binding of mobile labeled monomers to the membrane surface is nonuniform, or densities are lower than approximately 1 monomer per 3 mm 2 , our analysis will not detect such monomers above cellular autofluorescence. Nag et al. recently reported the presence of diffusible, membrane-bound fluorescein-labeled Ab40 on PC12 cells exposed to near-physiological peptide concentrations [28]. Our laboratory has also observed uniformly bound pools of diffusible Ab40 at densities up to 5 to 8 monomers per mm 2 on synthetic lipid bilayers [29]. This disparity may result from our use of a different cell line and a lower concentration of Ab40 than Nag et al. (50 nM vs. 150 to 350 nM) and from differences in membrane composition between living cells and model membranes.
Occasional very high-intensity fluorescent structures are visible in our images (see Figure 1). These probably represent peptide aggregates containing many more than 20 monomers. Such large species have not been rejected as possible mediators of Ab toxicity. However, given the relative rarity of these species and the lack of obvious morphologic abnormalities in the cells to which they are bound, we have focused our analysis on oligomers of measurable size.
The majority of cell-bound oligomers are immobile on a time scale of several seconds. This immobility is likely related to Ab40 binding or insertion sites on the membrane, a number of which have been suggested in the literature [9,11,12,39,40]. Ab oligomers may bind to specific lipid microdomains, such as regions enriched in externalized phosphatidylserines [12], cholesterol-rich regions [39], or lipid rafts [40]. Whether such mechanisms would preclude detectable diffusion of cell-bound Ab is unclear. Lipid microdomains are dynamic nanoscale structures [41], with lipid confinement times in the tens to hundreds of milliseconds. High-affinity Ab40 binding to such sites may stabilize microdomains, possibly by interacting with intracellular anchor proteins. Another likely explanation is that oligomers bind to or associate with specific membrane-integral proteins or receptors [9,11,23,42] that are temporarily restricted to specific locations within the cell membrane [43,44], immobilizing the Ab within an area small enough that positional fluctuations are undetectable. Several of these receptors are expressed in SH-SY5Y cells and could be the binding sites of the small oligomers we observe on cells (e.g. NMDA NR1 [45], a7nAChR [46], and EphB2 [47]).
Substantial evidence supports the hypothesis that Ab-induced toxicity stems from the formation of calcium-permeable Ab pores in cell membranes. Interestingly, treatment of SH-SY5Y cells with 50 nM fresh unlabeled Ab40 or HL647Ab40 results in minimal, sporadic calcium leakage (Figure 4). Superficially, these findings may seem contrary to the massive, immediate calcium influxes observed by other groups [17,19,48]. However, such studies have in general been performed at micromolar Ab42 concentrations [12,17] or using Ab42 prepared with solvents known to destabilize membranes [19,26]. Our own studies of Ab40-liposome interactions indicate that membrane binding and permeabilization occur in separate stages and depend on distinct membrane characteristics [15]. The low-level calcium leakage seen here may simply reflect how extracellular Ab40 interacts with healthy cell membranes at physiological levels. In an effort to address these complexities, this work represents a detailed analysis of the oligomeric size of Ab species capable of binding to living cells at physiological concentrations.

Amyloid-b(1-40) Preparation
Synthetic unlabeled amyloid-b(1-40), HiLyte Fluor 647-labeled amyloid-b(1-40) (HL647cAb40) and HiLyte Fluor 647 hydrazide were from Anaspec. HL647cAb40 was synthesized with an additional N-terminal cysteine residue, to which HiLyte Fluor 647 was attached by a C2 maleimide linkage. Amyloid-b (Ab) peptides were dissolved in 1% NH 4 OH at 0.1 mg/mL and vortexed for 30 seconds. Peptides were then lyophilized and stored at 220uC. To prepare fresh Ab, single aliquots were dissolved in 10 mM sodium phosphate buffer, pH 7.4, at a concentration of 1 to 2 mM and pipetted 5 to 8 times to mix. Freshly prepared Ab was used within 30 minutes to 1 hour.

Cell Culture
SH-SY5Y neuroblastoma cells (ATCC) were maintained in phenol-red free 1:1 DMEM/Ham's F12 (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen) and 200 units/mL penicillin/streptomycin (Invitrogen). Prior to plating, 25 mm circular No. 1 cover glasses (Fisher Scientific) were kiln-baked at 500uC for 2 hours and then autoclaved. For live-cell imaging, cells were plated onto coverslips at a density of 28,000 cells per cm 2 and imaged 2 to 4 days following plating.

Overcoming Autofluorescence
Initial experiments using amyloid-b(1-40) N-terminally labeled with HiLyte Fluor 488 revealed that autofluorescence of this cell line in the FITC channel was high enough to prevent definitive detection of single molecules on the cell membrane. Comparison of cellular autofluorescence levels at multiple wavelengths and of the fluorescence characteristics of available probes led us to identify the near infrared region of the spectrum as optimum for single molecule experiments. Kiln-baking slides prior to plating, imaging on days 2 to 4 following plating, and use of phenol-red free media were also key to reducing autofluorescent background.

Measurement of Fluorescence Lifetimes
Fluorescence lifetime measurements were performed on HiLyte Fluor C2 maleimide-labeled Ab40 (HL647cAb40) and HiLyte Fluor 647 hydrazide, as described [29]. Briefly, peptide and dye were dissolved to approximately 800 nM (as measured by absorption) in 10 mM sodium phosphate buffer, pH 7.4, and diluted to 200 nM in Hanks' Balanced Salt Solution (HBSS) (Invitrogen). Reverse-mode time-correlated single photon counting measurements were then performed. Data were analyzed by the Exponential Series Method in TimeMaster software from Photon Technology International.

Spin-Coated Sample Preparation
Freshly solubilized HL647cAb40 was diluted to 0.5 nM in 10 mM sodium phosphate, pH 7.4, and a droplet of this was spincoated onto a kilned glass slide. A series of 120 confocal microscopy scans were then taken with the objective lens focused at the surface of the glass slide, and the intensity of each surfacebound fluorescent particle was plotted versus scan number.

On-slide, in Solution, and On-cell Sample Preparation
Prior to imaging, cells on coverslips were exposed to 50 nM HL647cAb40 or unlabeled Ab40 for 10 minutes at 37uC. Freshly dissolved peptide was added directly into media and pipetted over coverslips twice to mix. Following peptide exposure, coverslips were washed three times in Hanks' Balanced Salt Solution (HBSS), placed into a sampleholder and covered in 1 mL of HBSS for imaging.

Total Internal Reflection Fluorescence (TIRF) Microscopy
Single molecule TIRF microscopy was performed on an Olympus IX-71 inverted microscope. A HeNe red laser (Uniphase) was focused onto the back focal plane of a 60x, 1.45 NA Olympus PlanAPO TIRFM objective; laser power at the plane of the coverslip was measured at 300 mW. Through-theobjective TIRF was performed by translation of a mirror just upstream of the objective lens. A multi-band pass SEMRock dichroic mirror was used to separate excitation from emission signal; a 620/60 excitation band pass filter (Chroma Technology Corporation) and 700/75 emission filter (Chroma Technology Corporation) were included in the setup. Images were acquired on a back-illuminated Ixon EMCCD camera, model DV887ACS-BV (Andor) and analyzed in ImageJ.

Epifluorescence Microscopy
Epifluorescence images for selected calcium indicator studies were acquired on the same microscope as was utilized for TIRF studies, using the 488 nm line of an Argon laser (Coherent) and a FITC filter cube.

Fluorescence Intensity Calibration
Fluorescence intensity per dye molecule was obtained by imaging free HL647 hydrazide in HBSS at concentrations ranging from 0 nM to 60 nM. These scans were acquired with the same settings as were utilized for live-cell imaging. Specifically, the calibration was performed as follows: 1. Single slice confocal scans were performed on free HL647 hydrazide in solution at 10 nM concentration increments, ranging from 0 nM to 60 nM. 2. A ''volume element'' was defined as a box of length 40 pixels, width 40 pixels, and height 1 pixel (or 1.6 mm by 1.6 mm by 1 mm). 3. The total fluorescence intensity of a volume element at each concentration was measured. 4. The average number of dye molecules present in a volume element at each concentration was calculated. 5. Total fluorescence intensity of a volume element versus the number of molecules present in a volume element was plotted for each concentration. 6. A line was fit to these data, and the slope represents the measured fluorescence intensity per molecule in the system (see Figure S5).
Calibration accuracy in determining oligomer size was confirmed by photobleach step counting, as discussed in Results.

Single Molecule Confocal Laser Scanning Microscopy
Confocal laser scanning microscopy was performed on an Olympus FluoView 500 microscope mounted on an Olympus IX-71 frame. Images were acquired in line scan mode using an Olympus PlanAPO 60x, 1.42 NA oil immersion objective. The 488 nm line of an Argon laser was used to obtain FITC channel and DIC (Differential Interference Contrast) images, with illumination power of 140 mW, and a 633 nm HeNe red laser was used for HL647 excitation at a power of 28 mW. A BA660IF emission filter and Cy5 channel PMT settings of 850 volts, 6.0 gain, and 6% offset were used for all single molecule experiments. Scans were taken at medium speed (10.67 seconds per scan). Scans of a 43 mm by 43 mm field of view were taken at 1024 by 1024 pixel resolution, yielding an image plane pixel size of 42 nm by 42 nm. Axial resolution (full width at half-maximum in the zdirection) was set to 1 mm by adjusting the confocal aperture diameter to 590 mm. Imaging was performed at 37uC under 5% CO 2 within 1 hour of exposure to Ab40.

Data Analysis in ImageJ
For live cell data, an outline representing the cell membrane was drawn around the inner edge of the cell's image in the DIC scan (see Figure S8). This outline was then overlaid on the fluorescence scan. All diffraction-limited fluorescence spots whose maxima fell on or within 1 mm outside of this line were designated as ''cellbound oligomers''. Irregularly shaped fluorescence spots, spots saturating the detector at four or more pixels, and spots associated with a clear non-membrane artifact (i.e., bigger than a resolution element) on the DIC image were rejected from the analysis. Remaining ''cell-bound oligomers'' were boxed with a 40 pixel by 40 pixel (approximately 1.6 mm by 1.6 mm) region of interest. The total photodetector output for such a square represents the intensity of a ''volume element'' containing the oligomer-a box 1.6 mm in length, 1.6 mm in width, and 1 mm in height with the oligomer at its center. The total fluorescence counts for this region of interest were recorded. An adjacent off-cell square of the same dimensions was designated as ''local background''. The total counts for this box were then subtracted from the total counts present in the ''oligomer'' box to yield an integrated fluorescence intensity value for the oligomer (see Figure S5). Dividing this value by the intensity per dye molecule yields an oligomer size, in monomers, for each individual particle included in the analysis.

DiO Labeling and Colocalization
For membrane colocalization studies on HL647cAb40, cells were first incubated with or without 50 nM HL647cAb40 in 1 mL media for 10 minutes. An aliquot of 6 mL of 1 mM Vybrant DiO cell-labeling solution (Invitrogen) was then added and cells were swirled to mix. Cells were incubated at 37uC prior to washing and imaging as described above. Colocalization analysis was performed in ImageJ using an open-source macro entitled Colocalization (written by Pierre Bourdoncle, available on the web at http://rsbweb.nih.gov/ij/plugins/). Pixels where the ratio of HL647 to DiO fluorescence intensity was greater than 40% were designated ''colocalized'' regions.

Cell-bound Oligomer Quenching
For potassium iodide oligomer quenching, a 3 M solution of potassium iodide was prepared in HBSS and added onto cells to a final concentration of 300 mM. Cells were imaged before and after 2 minutes exposure to potassium iodide.

Calcium Leakage Experiments
SH-SY5Y cells were loaded with fluorescent calcium indicator Fluo4-AM (Invitrogen) at room temperature for 15 minutes at a concentration of 1.7 mM. Cells were then incubated in HBSS for an additional 15 minutes prior to washing and loading into sampleholders. 50 mL of 1 mM unlabeled Ab40 or HL647cAb40 were added by pipette to 950 mL HBSS in the sampleholder, for a final concentration of 50 nM. To induce maximal calcium leakage, ionomycin in HBSS was added by pipette to a final concentration of 6 mM. Fluo4-AM DF for each frame n was calculated as: where F c,n represents cell body fluorescence in frame n, F bg,n represents fluorescence of an off-cell background region in frame n, F c,i represents initial cell body fluorescence, and F bg,i represents initial background fluorescence.

Supporting Information
Figure S1 Unlabeled Ab40 and HL647Ab40 cause dye to leak out of cell-derived blebs. Blebs are prepared from rat basophilic leukemia (RBL) cells, as described in [49].  Figure S2 HPLC spectra for HL647cAb40 and Ab40. The HPLC scan 215 nm absorbance spectrum is shown for 8 mM HL647cAb40 (dark solid line, HL647AB40) and unlabeled Ab40 (gray dashed line, UnlabAB40). Both peptides elute as a single peak with dimer-to-trimer molecular weight. An additional minor peak, visible in the void volume for the HL647cAb0, may consist of peptide aggregates greater than 200 kDa; however, no fluorescent spots of the size and intensity expected for such large particles were observed in spin-coated samples examined by single molecule microscopy (see Figure S3). A limited number of large aggregates were detected in on-slide, in-solution samples, but these were a relatively small component (less than 10%) of the species observed.
(TIF) Figure S3 Confocal images of single HL647cAb40 molecules. The confocal image on the left depicts single molecules of dry HL647cAb40. A droplet of 0.5 nM HL647Ab40 in 10 mM sodium phosphate buffer, pH 7.4, was spin-coated onto a kilned glass slide. The confocal image on the right shows single molecules of HL647cAb40 adhered to a glass slide after ten minutes' incubation of the slide with 50 nM peptide in media. Frames are 43 mm by 43 mm.
(TIF) Figure S4 Typical confocal mode photobleach traces for HL647cAb40 diluted to 0.5 nM and spin-coated onto a kilned glass coverslip. An example monomer trace is shown at top left; an example dimer is shown at top right. Example trajectories for two spots that did not photobleach in digital steps are shown at bottom left and right. Approximately 65% to 70% of particles bleached in clean, digital steps. Of these, 83% 6 3% bleached as monomers, and 15% 6 3% bleached as dimers. Of the particles that did not bleach in single steps, 5% 6 1% had intensities greater than was typical for observed monomers and dimers.
(TIF)  Figure S6 Example confocal image and TIRF photobleach trace for HL647cAb40 trimers. Example of a 40 by 40 pixel (1.6 mm by 1.6 mm) region of interest (ROI) in the confocal mode; this ROI contains a slide-bound HL647cAb40 particle with the integrated intensity of a trimer (left). To the right is an example photobleach trace taken from a Total Internal Reflection Fluorescence (TIRF) film. Three discrete photobleach steps can easily be identified, marking the particle as a trimer.