Low-resolution structure of a vesicle disrupting α-synuclein oligomer that accumulates during fibrillation

Abstract

One of the major hallmarks of Parkinson disease is aggregation of the protein α-synuclein (αSN). Aggregate cytotoxicity has been linked to an oligomeric species formed at early stages in the aggregation process. Here we follow the fibrillation process of αSN in solution over time using small angle X-ray scattering and resolve four major coexisting species in the fibrillation process, namely monomer, dimer, fibril and an oligomer. By ab initio modeling to fit the data, we obtain a low-resolution structure of a symmetrical and slender αSN fibril in solution, consisting of a repeating unit with a maximal distance of 900 Å and a diameter of ∼180 Å. The same approach shows the oligomer to be shaped like a wreath, with a central channel and with dimensions corresponding to the width of the fibril. The structure, accumulation and decay of this oligomer is consistent with an on-pathway role for the oligomer in the fibrillation process. We propose an oligomer-driven αSN fibril formation mechanism, where the fibril is built from the oligomers. The wreath-shaped structure of the oligomer highlights its potential cytotoxicity by simple membrane permeabilization. This is confirmed by the ability of the purified oligomer to disrupt liposomes. Our results provide the first structural description in solution of a potentially cytotoxic oligomer, which accumulates during the fibrillation of αSN.

Parkinson Disease (PD) is a common neurodegenerative disorder of the brain. Hallmarks of PD are massive death of dopaminergic neurons and formation of intracellular Lewy bodies (LBs). LBs mainly consist of large fibrillar inclusions of α-synuclein (αSN), a 140-residue natively unfolded protein (1). Numerous in vitro studies of αSN fibrillation have shown that fibril formation is a nucleated polymerization and that oligomers form transiently in the lag phase (1–5). Such oligomers, rather than fibrils or monomers, have been suggested to be the neurotoxic species (4, 6), however whether the oligomers are on- or off-pathway in fibril formation and whether the cytotoxic species corresponds to the nucleus of fibrillation remain unclear. Neurotoxicity is proposed to arise from a pore-like membrane permeabilization (6) or by destabilization of the membrane allowing nonspecific ion-transport (7). Unambiguous structural information of the cytotoxic species is difficult to obtain due to sample coexistence of different species, sensitivity to sample handling and potential surface-binding artifacts. The morphology of purified oligomers from wildtype and mutant αSN have been studied by atomic force microscopy (AFM) and electron microscopy (EM) (4, 5), but purified oligomers may be structurally and functionally distinct from those in equilibrium with monomers and fibrils. Different conditions can lead to different oligomers (4, 5, 8) with varying ability to disrupt artificial cell membranes (9–12). Ideally, structural studies should be conducted on an unperturbed ensemble of all potential on- and off-pathway species during the fibril formation process. This is possible with small angle X-ray scattering (SAXS), a noninvasive technique that allows the study of fibrillation in solution under physiologically relevant conditions. Due to signal additivity SAXS can provide low-to-medium-resolution structural information from relatively simple polydisperse solutions as recently illustrated for fibrillation (13–15). Here we use SAXS to obtain structural information about the species present during the fibrillation of αSN at neutral pH and 37 °C. We fibrillate αSN above the supercritical concentration (SCC), which, according to theory, causes accumulation of the fibrillation nucleus (16) making structural analysis of this species feasible. From our data we can reconstruct both the low-resolution structure of such an accumulating oligomeric species, as well as the mature fibrils. We find that an oligomer purified by gel filtration under the same conditions can permeabilize calcein-filled DOPG ((1,2-di-oleoyl-sn-glycero-3-[phospho-rac-(1-glycerol)]) vesicles, highlighting its potential cytotoxicity as previously suggested (2, 3, 17). The oligomeric species has a wreath-shaped structure suggesting that a simple permeabilization mechanism could explain a cytotoxic effect. The time profile for the accumulation and decay of this oligomer is consistent with an on-pathway role for the oligomer and we propose an oligomer-driven αSN fibrillation mechanism.

Results

The Supercritical Concentration of αSN.

SCC was determined by fibrillating αSN at different concentrations. The time required for the fibril formation reaction to reach 50% completion (T_50%) was plotted against total αSN concentration in a double logarithmic plot (Fig. 1A). The intercept between the two linear regions defines SCC to be ∼8 mg/mL. We used a concentration well above SCC (12 mg/mL) for the SAXS experiments.

Fig. 1.

Fibrillation of αSN above SCC reveals a sigmoidal curve and β-sheet formation. (A) Double logarithmic plot of the time required to reach 50% completion of the fibrillation versus the protein concentration. Linear regression yields two linear segments and the intercept between the lines defines SCC. Error bars represent the standard deviation with N = 3. (B) ThT-emission as a function of time, where each data point represents ThT-emission of a given sample that was subsequently used for SAXS solution measurements. The fibrillation experiments run for a total of 22.5 hr. (C) Far-UV CD spectrum of freshly prepared αSN (gray spheres) and mature fibrils after 22 hr (black spheres). (D) ATR-FTIR spectrum of mature fibrils after 22 hr. Both techniques confirm the appearance of extensive β-sheets.

SAXS Data Collection.

We recorded solution SAXS spectra in parallel with the emitted fluorescence of the general fibril specific dye Thioflavin T (ThT) to correlate structural changes (SAXS) with the appearance of fibrils (ThT). To approximate physiological conditions experiments were carried out in PBS buffer (pH 7.4) at 37 °C. The ThT fluorescence described a standard sigmoidal curve over time (Fig. 1B), consistent with a nucleation-dependent polymerization process. The far-UV circular dichroism (CD) of a freshly prepared αSN revealed a minimum of ∼198 nm, whereas the minimum had shifted to ∼218 nm of the mature fibril. This is consistent with the conversion of random coil αSN into β-sheet rich structures (Fig. 1C). Attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR) (Fig. 1D) showed a distinct amide 1 band at 1624 cm^-1 (∼21% of the total spectral area) that is characteristic of amyloid structure (18). We recorded SAXS spectra at ∼30 min intervals over 22.5 hr (Fig. S1 A and B). Representative spectra with an interval of ∼1.5 hr are shown in Fig. 2A. A qualitative visualization of the process is obtained from the average radius of gyration (R_g) of the scattering species and the maximal dimension (D_max) of any scattering species present at a given timepoint (Fig. 2B). At intermediate time points (7–10 hr) intermediate sized species are present, whereas very large species are formed at later stages. This general evolution is confirmed by dynamic light scattering (DLS) (Fig. S2). By visual inspection of the data curves, it is evident that the structures evolve, also after no further development is evident in ThT measurements (Fig. S1 A and B).

Fig. 2.

Analysis of SAXS data of the fibrillation process shows a time dependent increase in R_g and D_max and SVD reveals four components. (A) Selected SAXS spectra showing the evolution of fibrillation from 0–22.5 hr in intervals of ∼1.5 hr. Curves are translated arbitrarily for viewing purposes. (B) R_g (black spheres) and D_max (white spheres) plotted versus time. (C) Plot of the logarithm to the eigenvalues resulting from the singular value decomposition. The plot suggests 4 principal components.

Decomposition of SAXS Data and Identification of an On-Pathway Oligomeric Species.

A singular value decomposition (SVD) analysis in the program SVDPLOT (19) estimates the minimum number of independent scattering components significantly contributing to the scattering (Fig. 2C). We identified four species present in detectable amounts in equilibrium throughout the fibrillation process. Monomers and dimers of αSN exist in equilibrium upon dissolution (5, 20, 21), thus we assign three of the species to monomer, dimer and fibrils and conclude that a fourth species coexists during the fibrillation process. We then used the program OLIGOMER (19). This program fits any experimental scattering curve by a linear combination of a number of input-scattering curves (I_exp = A ∗ I_a + B ∗ I_b + …+N ∗ I_n; where I_exp is the observed scattering from the mixture of species; A,B…N are volume fractions of individual species a,b,…n; I_n is the scattering of the nth species extrapolated to zero concentration and n is the total number of species) with an appropriate weight (volume fractions of the individual components, determined by a least squares fit). It is thus possible to follow the evolution of the volume fractions of individual species present during fibrillation. For each experimental data curve at any time point, the program OLIGOMER produces the corresponding best fit and isolates residual scattering, which should be indistinguishable from the noise level of the data when an adequate description of the experimental data is obtained from the linear combination. Hence, to isolate the scattering signal from the structural nucleus, experimental scattering of the other components were obtained separately: (I) Scattering signals from pure monomers were obtained at pH 10.5 where the protein is monomeric (Fig. 3A); (ii) the scattering signal from a mixture of monomers and dimers was obtained from the sample in PBS measured at time t = 0 hrs, whereas (iii) a scattering curve from the repeating unit of mature fibrils is obtained at the end of the fibrillation process (21 hr and 45 min). Because the SVD (Fig. 2C) had pointed to the presence of four significantly scattering species in solution, an attempt to fit the data at all time points with only three input curves was not expected to yield adequate fits to the experimental data points, but was used to obtain a first approximation of the scattering from the unknown fourth species, represented by the residual scattering curves. Visual inspection revealed that these residual curves had a similar shape. This suggests that a fourth distinct component is present in solution. An average of all residual curves significantly larger than the noise level was calculated, isolating the scattering that was present in the data and which was not described by monomers, dimers and fibrils, in a completely model-free manner. No assumptions regarding the overall shape of the fourth species in solution were made at any point. This average curve was then included as a fourth input curve for the program OLIGOMER. The procedure was repeated twice refining the estimate of the fourth component representing the structural nucleus.

Fig. 3.

(A) Gel filtration chromatrogram of 12 mg/mL αSN at pH 7.4 (black spheres) and pH 10.5 (gray spheres). Insert shows zoom of oligomer peak at pH 7.4, which disappears at pH 10.5. (B) Development of the fibrillation process monitored by SAXS and ThT-emission. SAXS-derived volume fractions are plotted as a function of incubation time, illustrated by blue spheres (sum of monomer and dimer fractions), pink spheres (volume fraction of the fourth component), and green spheres (the volume fraction of fibril). The measured ThT-emission as a function of time is illustrated by black spheres. (C) The ability of purified αSN oligomers to permeabilise DOPG vesicles containing calcein. Black spheres illustrate the time profile of calcein release after addition of oligomers isolated from gel filtration experiments. The signal is normalized relative to signal from Triton-X induced complete rupture of the vesicles. As negative control we use monomer obtained by gel filtration (white spheres).

The obtained time course of the volume fractions of monomers and dimers, the fourth intermediate species and mature fibrils confirmed the expected accumulation of an intermediate species during fibril elongation prior to the formation of mature fibrils (Fig. 3B). We assign this species to the oligomeric structural nucleus, which we shall refer to as the oligomer. Given that the experiment is preformed at SCC indeed, the fourth species is present from the beginning of the experiment and increases in volume fraction before the rise of the ThT signal (Fig. 3B). The onset of the increase in volume fractions of mature fibrils coincides with the increase in ThT-fluorescence. This indicates that ThT does not bind to the oligomer, but only to mature fibrils. The volume fraction of the oligomer reaches a maximum of about 30–40% after 6–10 hr. We also observe the strongest fibril accumulation in this time period.

A Membrane-Permeabilizing Oligomer.

Oligomeric and monomeric αSN were isolated via gel filtration (Fig. 3A) of a fibrillating sample after 6 hr incubation, where significant amounts of the oligomer are present (Fig. 3B). To examine whether the oligomer could permeabilise DOPG vesicles, we monitored the release of the fluorescence probe calcein entrapped in vesicles at self-quenching concentrations. This assay does not distinguish between different permeabilization mechanisms (e.g., pore-like mechanism, membrane destabilization, or detergent-like mechanism), but measures whether the vesicles release calcein. It is known that the αSN monomer disrupts vesicles in a detergent-like mechanism at high αSN∶lipid ratios when incubated with giant unilamellar vesicles formed by DOPG (2). We confirm this (Fig. S3A), but see no rupture in neutral 1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC) vesicles (Fig. S3B). Therefore, samples containing only monomer were diluted to concentrations comparable to those of the oligomer samples and used as a negative control against DOPG vesicles. We did not see any significant permeabilization resulting from addition of the diluted monomer or buffer alone. In contrast, addition of the oligomer fraction at similar w/w concentrations resulted in an approximately 40% release of calcein (Fig. 3C). Hence, the species detected in the SAXS experiment shows membrane-permeabilizing properties and could be cytotoxic. There is a distinct discrepancy between the estimated volume fraction of the oligomeric species seen by SAXS and that of the gel-filtration experiments (30–40% versus 1–2%) (Fig. 3 A and B). Contact to the matrix of the gel-filtration column and the actual separation of individual species can cause a significant shift in equilibrium between the individual species. It has been proposed that the oligomeric species is in a pseudostable state that converts to monomers when using gel filtration (22). DLS analysis of fibrillating αSN (22) (Fig. S2) confirms the presence of large amounts of an intermediate species. This further emphasizes the importance of using a noninvasive method like solution SAXS, which allows investigation without isolating individual species.

The Wreath-Shaped Oligomer.

The scattering from the oligomeric species was employed to reconstruct its low-resolution structure. The pair distance distribution function P(r) (Fig. 4A) yields the maximum distance within the molecule of D_max = 180 ± 30 Å. The low-resolution shape generated ab initio by DAMMIF (Fig. 4B) without applying symmetry restraints shows that the oligomer is a slightly elongated annular species with a central cavity resembling a closed wreath. This general appearance agrees well with previous electron microscopy observations of isolated oligomers (5). The present SAXS-derived solution model is more detailed and reveals two arms forming a wreath 180 Å long and 90 Å wide, each arm with an approximate diameter of 45 Å (Fig. 4B). A comparison of the fit of our presented model, a spherical shape and the previously published insulin nucleus is shown in Fig. S4. The average excluded volume of multiple models obtained by independent DAMMIF runs (460 ± 30 nm³) corresponds to ∼16 molecules of αSN per oligomer. Previous estimates of the number of protomers in oligomers isolated via gel filtration suggested 21–38 molecules (23).

Fig. 4.

Low-resolution SAXS model of the oligomer, fibril, and an oligomer-driven fibril formation. (A) The P(r) of the early and late fibrils (large spheres, light and dark colors respectively), and the cross section of the early fibril (Insert, black spheres) compared with the P(r) of the oligomer [Insert, (white spheres)]. (B) SAXS-derived structure of the αSN oligomer. The average structure (mesh representation) and the filtered averaged structure (surface representation) are displayed and superimposed. The filtered structure has a volume corresponding to the average volume of individual models, and the difference between the average and filtered structure indicates the general level of differences between individual models. The model is showed in two orientations, rotated 90° around the longest axes. (C) The SAXS-derived model of the symmetric, early fibril that exists in solution in equilibrium with native species and a fourth component. A single repeating unit is shown in cyan, with the averaged model and the filtered averaged models superimposed in mesh-representation. The principle of the repeats building the mature fibrils is shown to the right, where three repeats of the filtrated model are shown. Repeats two and three have been translated 880 Å vertically with respect to the first repeat, and the model to the right is rotated 90° around a vertical axis with respect to the left model. (D) Model for the elongation of fibrils. In rose/purple colors, 26 oligomers constituting one repeating unit of the mature fibril (averaged and filtered model shown in cyan mesh) are displayed in surface representation. Below, the 26 oligomers are superimposed with the fibril repeating unit, whereas the two models are separated above. The lowest representation is rotated 90° around a horizontal axis with resepect to the top two models. Length scales are as in Fig. 4C.

Structure of the Repeating Unit of αSN Fibrils.

The P(r) function of the fibril repeating unit reveals a maximal distance of 900 Å (Fig. 4A). The ab initio model of the fibril reveals a slender and symmetrical fibril. The models of the fibril repeat (Fig. 4C) have an average excluded volume of 11,900 nm³, which corresponds to 413 αSN molecules. The diameter of the fibril is ∼180 Å corresponding to the oligomer maximal distance. The symmetric fibril appearance allows us to test the assumption that the fibril repeats can be modeled individually, which is a general prerequisite for modeling the repeating unit at all. The fibrils have an apparent threefold repeat and a truncated model with 2/3 length was created by ab initio modeling, based on a P(r) function forced to zero at 2/3 of D_max. The resulting model is very similar to the central part of the full-length unit (Fig. S5).

Discussion

We have shown that a wreath-shaped oligomer accumulates during αSN-fibrillation at SCC and that the rate of fibril formation is highest when the oligomer reaches the maximal concentration. This is consistent with a scenario in which the oligomers are direct precursors to, and building blocks for, the fibril. Thus we consider Fig. 3B to be good—but not conclusive—evidence for an on-pathway oligomer mechanism. We note that an off-pathway oligomer (which dissociates to monomers to form fibrils) may also accumulate as observed; however, only within a certain range of rate constants governing its formation and decay. A third option is the accumulation of an on-pathway oligomer, which forms at higher rate constants than the dissociation into monomers that add to the growing fibrils after a nucleation. However, structural considerations support the theory that the oligomeric species is the building block for fibrils, though we cannot exclude that monomers assist in the stacking of the oligomer. We do not detect the presence of an additional structural species in solution, neither monomers in a nonnative state nor any additional on- or off-pathway oligomeric species. If such species are present during the process, they are only present in amounts that are below the limit of detection with the current method. Previous SAXS studies of αSN solutions have revealed that dopamine stabilizes a αSN trimer (24). Under the experimental conditions applied here, we do not see such trimers, they are not visible by gel filtration and are not necessary to fit the data at different timepoints.

A feasible packing mechanism of the oligomers into the fibrils would be a simple stacking of the wreaths with surface contacts formed between the relatively flat sides of each wreath. To test this hypothesis the P(r) function of the cross section of the fibril repeating unit was calculated and compared to P(r) function of the oligomer (Fig. 4A). The comparison reveals that the distributions of distances in the cross section and in the oligomer are similar, which supports the stacking model. The observed dimensions of the ab initio structure of the wreath and the diameter of the fibril are similar. The estimated volume of the fibril repeating unit corresponds to ∼26 times the volume of the oligomer. If stacking 26 wreaths (corresponding to 416 αSN molecules per fibril repeat, assuming 16 αSN molecules per wreath) the length of the stacked model is well in accordance with that of the ab initio model. A model has been generated (Fig. 4D) suggesting such a packing of the αSN fibrils. Twenty-six oligomers are stacked and each wreath rotated around the long axis of the fibril by 360°/26, and translated along the same axis by 45 Å (the estimated average thickness of one wreath). Finally individual wreaths have been translated along the short axes of the fibril to fit into the density. The result is well in agreement with the SAXS-derived ab initio model. If this model is correct, the wreaths stack arm by arm (see Fig. 4 B and D), and each form what on a macroscopic level would appear as one protofibril. In this model, what appears to be separate protofibrils are not formed independently followed by lateral association; rather the protofibrils are formed together by the two arms in the stacking wreaths.

Previous observations (5, 8) show the coexistence of several oligomeric species during αSN fibrillation and different morphologies of mature αSN fibrils have been observed using AFM (25, 26). The following differences may be important: firstly, this study applies SCC, far above the concentrations applied during, for example, AFM and tunneling electron microscopy. Secondly, surface contacts may influence both the equilibrium between species and the actual appearance of individual species. When applying SAXS to fibrillating solutions, we have the advantage of a noninvasive method, which works directly in solution, with no prior separation of individual species.

We have isolated an oligomeric species, which clearly disrupts DOPG lipid vesicles, in contrast to native monomers at similar concentrations. For simplicity we have only tested the oligomer on 100% anionic lipids, however, it is known that also vesicles with lower levels of anionic lipids can also be disrupted by an isolated αSN oligomer (21, 27). Our data increase the likelihood that the oligomer could be cytotoxic in vivo and we have modeled the wreath-shaped oligomeric species that we suggest is responsible for this function. The shape and size of the wreath immediately suggests a simple perforation mechanism. The wreath is approximately 45 Å thick, which allows it to span the hydrophobic part of a phospholipid bilayer [∼30 Å thick (28)] and has a central hole, which if inserted into a membrane could cause uncontrolled flux of ions. Further characterization of, for example, the hydrophilicity of the surface of the oligomer is needed before speculating further about the mechanism of cytotoxicity. In conclusion, we have generated an ab initio model of an accumulating wreath-shaped αSN oligomer. The oligomer membrane permeabilizing properties and structural similarity to the mature fibrils, suggests that αSN fibrils, in themselves nontoxic, are formed by simple stacking of cytotoxic oligomeric species.

Experimental Procedures

Materials.

Chemicals were analytical grade (Merck), except Thioflavin T (ThT) (Sigma), phospholipids (Avanti Phospholipids) and calcein (Fluka). Unless otherwise stated experiments were conducted in PBS (20 mM phosphate, 150 mM NaCl, pH 7.4)

Conditions for αSN Fibrillation in Conjunction with SAXS.

Recombinant αSN was expressed and purified according to ref. 29 (see SI Text, SI Experimental Procedures). One hundred fifty μL of 12 mg/mL αSN, 40 μM ThT in PBS were incubated in a sealed clear bottom 96-well plate (Nunc, Thermo Fisher Scientific). The plate was incubated at 37 °C in an Infinite 200 fluorescence plate reader (Tecan) and shaken orbital for 20 min per hour at 99 rpm including a 3 mm diameter glass microsphere (Glaswarenfabrik Karl Hecht GmbH & Co KG) to increase reproducibility (30). Fluorescence was measured at 30 min intervals (excitation/emission 450/485 nm). Subsequently, 150 μL sample was withdrawn from one well for SAXS data recording. For each SAXS timepoint sample was taken from a new well.

Determination of Supercritical Concentration.

Triplicates of 12 samples in the concentration range 0.44–16 mg/mL αSN were fibrillated as described above. To provide T_50% values data was fitted and data from replicate wells were fitted to an empirical equation (31) with flat baselines.

Circular Dichroism of αSN.

Spectra were recorded as described in ref. 15 using a freshly prepared sample of 0.5 mg/mL αSN in PBS as well as 12 mg/mL αSN, fibrillated for 22 hr and then diluted to 0.5 mg/mL. Four spectra were recorded (scan speed 100 nm/ min, bandwidth 1 mm and 2 sec response time) using a 2 mm, Helma 106-QS cell in a Jasco J-810 CD-spectrometer at 37 °C. Spectra were averaged and buffer subtracted.

ATR-FTIR of Fibrillated αSN.

An FTIR spectrum was measured using 12 mg/mL αSN fibrillated for 22 hr using a Tensor 27 (Bruker) FTIR spectrophotometer with a Deuterated Triglycine Sulfate (DTGS) Midinfrared detector and a Golden Gate single reflection diamond attenuated total reflectance (ATR) cell (Specac). Two μL αSN fibrils were dried on the ATR crystal. Spectra were recorded from 4,000–1,000 cm^-1, nominal resolution of 2 cm^-1, 64 accumulations.

Calcein Release Assay.

Calcein-containing vesicles were prepared according to ref. 32. αSN oligomers were isolated using gel filtration (SI Text, SI Experimental Procedures). Two fractions of the isolated oligomer peak (2 × 0.5 mL) were divided into 3 × 150 μL and loaded into a clear-bottomed 96-well plate. Two μL calcein-containing vesicles were added and the release of calcein was measured using a Tecan fluorescence plate reader at room temperature with excitation/emission 490/520 nm. Data were collected for 30 min preceded by 5 sec of auto-mixing. Monomeric αSN at equal w/w concentration was used as control. Full rupture of the calcein vesicles was measured by adding 0.1% (w/v) Triton-X.

Small Angle X-ray Scattering Data.

Data were collected at X33 at the European Molecular Biology Laboratory on DORIS III (DESY) (33) at a wavelength of 1.5 Å, using a MAR345 Image Plate detector, in the momentum transfer range 0.006 < s < 0.498 Å^-1 (s = 4π sin θ/λ, θ is half the scattering angle). Exposure times were 2 min. Repeated exposure showed no detectable radiation damage. Data analysis was performed in ATSAS 2.1 (34). The average molecular masses of solutes were estimated from the extrapolated relative forward scattering, I(0). P(r) functions were estimated using the indirect Fourier transformation program GNOM (35) yielding D_max and the average R_g within the particles.

Singular Value Decomposition (SVD) Analysis and Decomposition of Data.

All measured datasets were used for a SVD analysis in the program SVDPLOT (19) to estimate the minimum number of independent scattering components (Fig. 2C). The program OLIGOMER (19) was employed to isolate scattering curves from the intermediate species, and to calculate the evolution of volume fractions of species. Initially, three experimental input curves were employed: (i) monomers at pH 10.5; (ii) a mixture of monomers and dimers in PBS, and (iii) mature fibrils at 21.75 hr. Significant residual scattering isolated by the program was averaged at all timepoints and used as a first estimate representing the scattering from the intermediate species in a subsequent run of the program. Two additional refinements (the fourth species represented by the average input-scattering curve plus residual scattering) produced the final scattering curve. The isolation of scattering from the fourth species is thus model-free. The R_g and D_max of the distinct oligomeric species were evaluated as above, and the excluded volume was taken from the ab initio models (see below).

Ab Initio Modeling.

Ab initio structures of the fibril repeat and oligomer were obtained using the programs DAMMIN (36) and DAMMIF (37) respectively. A simulated annealing protocol is used to build a compact bead model of uniform scattering length density, which will minimize the discrepancy between the experimental and calculated curves at low resolution (up to s about 0.2–0.3 Å^-1). The initial search volume for the fibril repeating unit was estimated based on the shape of the P(r) function, and an ellipsoid with half-axes of 475, 150, and 150 Å was used using 35,000 spheres. Twenty individual jobs were computed, and averaged using the program DAMAVER (38). A final averaged and filtered model with an excluded volume in accordance with individual models was calculated. Initial models of the fibril repeating unit clearly indicated P2-symmetry and the models were recalculated applying symmetry, with equally good fits between calculated and experimental scattering curves. The resulting averaged model was the starting volume for twenty refined individual models, which were averaged. After two refinements, the represented model was obtained (Fig. 4B). The fibril repeat is symmetric with three turns per repeat. A model of the symmetric fibril in a search volume at app. 2/3 length (ellipsoid with half-axes 310, 150, 150 Å) was calculated from a P(r) function forced to 0 at 620 Å (Fig. S5). Modeling of fibrils at the latest timepoints resulted in less symmetric fibrils and both initial modeling and refinement was performed in P1 (Fig. S5). The starting volume for calculation of the oligomer shape was a sphere with radius of 95 Å (equal to D_max from the P(r) function) and refined over three runs using 20 individual models in each case. No symmetry restraints were applied. In all cases, the maximal s-value defining the resolution of the reconstruction was selected from a resulting normalized spatial discrepancy between individual models below 1.0 (38). In Fig. S6. a model of the least and most typical model is superimposed to visualize the differences and similarities between individual models.