Two-Step Amyloid Aggregation: Sequential Lag Phase Intermediates

The self-assembly of proteins into fibrillar structures called amyloid fibrils underlies the onset and symptoms of neurodegenerative diseases, such as Alzheimer’s and Parkinson’s. However, the molecular basis and mechanism of amyloid aggregation are not completely understood. For many amyloidogenic proteins, certain oligomeric intermediates that form in the early aggregation phase appear to be the principal cause of cellular toxicity. Recent computational studies have suggested the importance of nonspecific interactions for the initiation of the oligomerization process prior to the structural conversion steps and template seeding, particularly at low protein concentrations. Here, using advanced single-molecule fluorescence spectroscopy and imaging of a model SH3 domain, we obtained direct evidence that nonspecific aggregates are required in a two-step nucleation mechanism of amyloid aggregation. We identified three different oligomeric types according to their sizes and compactness and performed a full mechanistic study that revealed a mandatory rate-limiting conformational conversion step. We also identified the most cytotoxic species, which may be possible targets for inhibiting and preventing amyloid aggregation.


S1 Transmission Electron Microscopy (TEM)
TEM measurements were performed using a Libra 120 Plus transmission electron microscope (Carl Zeiss SMT, Germany) operated at 120 kV and equipped with a LaB6 filament and an SSCCD 2 k × 2 k direct coupling camera.
The protein samples (N47A-SH3-DA) were diluted in a buffer that promoted aggregation (0.10 M NaCl and 0.10 M Gly, pH 3.2) and were incubated at 37 °C for 45 days. Ten microliter aliquots were removed at appropriate times during the incubation of the aggregation sample, and they were rapidly frozen in liquid nitrogen to stop the aggregation. Prior to the TEM measurements, these aliquots were deposited on Formvar 300-mesh copper grids (ANAME, Spain). After a 5 minute incubation, the copper grids were washed twice with deionized water and negatively stained with a 1% (w/v) uranyl acetate solution for 1 minute. The excess stain was removed with tissue paper. Finally, the samples were dried and used for the TEM analysis.  The burst-wise FRET efficiency, E, was obtained with F A488 and F FRET (emissions detected at the 470-nm excitation time window) using eq. S1. The FRET signal must be corrected for the spectral crosstalk of the donor dye in the acceptor detection channel (β = 0.0174), the fraction of direct excitation of the acceptor by the donor laser (α = 0.017), and detection correction factor (γ = 0.95) [4][5] .
For each aggregate burst, the fluorescence lifetime of the donor, τ A488 , was obtained by fitting a single exponential decay function to the photon decay trace using iterative deconvolution with a simulated instrument response function (IRF). The maximum likelihood estimator was employed as a criterion for optimizing the fits because this criterion works better than others in cases with low photon counts. The value of τ A488 represents an orthogonal estimation of the FRET efficiency through the quenching caused by the A488 fluorophore because E = 1 -τ A488 /τ 0 , where τ 0 is the fluorescence lifetime of the donor in the absence of the acceptor, which is 4.0 ns for A488 6 . Hence, τ A488 involves an estimation of the FRET efficiency independent from the burst intensities and is a much more reliable approach than the estimation of both E and oligomer size from the same burst-wise intensity, a method that was previously employed to study α-synuclein oligomers [7][8][9] .
Finally, we used the total intensity of the directly detected A488 and A647N (F A488 and F A647N ) divided by the average intensity of a monomer to estimate the apparent oligomer size of the aggregates detected. This simple method has been used previously S6/S19 [10][11] ; however, it presents an important problem when there is intra-oligomer FRET: the quenching caused by the donor's intensity because FRET can cause an underestimation of the size. Cremades and colleagues corrected this issue using the burst-wise FRET efficiency E as the correction factor based on the donor's emission intensity [8][9] . This method is not free of problems because the FRET efficiency and the oligomer size are For molecules exhibiting τ A488 ≥ τ 0 , 6 the quenching and FRET processes are negligible. For these molecules, the apparent oligomer size of each individual event was calculated in the conventional way as the total intensity of the burst by considering both channels and dividing by the average fluorescence intensity of a monomer (eq. S2) 10-11 : For molecules exhibiting τ A488 ≤ τ 0 , the overall intensity in the A488 channel was corrected by the amount of quenching using a factor that accounted for the lifetime ratio, τ 0 /τ A488 (eq. S3): In these equations, F mon is the average fluorescence emission intensity of a monomer. as well as the disappearance of the type 1 aggregates. By fitting the experimental data to single exponential decay functions (Fig. 3A, main text), we obtained the initial rate as S9/S19 the derivative of the relative population of oligomer with time, at time = 0 ( , where [C] is the concentration of each type of oligomer, 1, 2, or 3) 12 . The apparent order of the reaction was obtained from the slope of the logarithmic plot of the initial rates versus the total protein concentration (Fig. 3B in main text). The initial rates and the apparent values of the reaction order are compiled in Supplementary Table S1.
Supplementary Table S1. Initial rates (absolute values) of the disappearance (type 1) and formation (types 2 and 3) of the different types of oligomers along with their corresponding apparent reaction order. At least 10 images were collected from various areas for each sample.
The FLIM images were analyzed using the SymPhoTime 32 software package and a S10/S19 similar protocol to the protocol used in the SMF-PIE traces but in a pixel-by-pixel fashion. The PIE excitation scheme allows dual-color excitation but is separated into two different time windows; therefore, simultaneous images of two different fluorophores can be reconstructed. We performed spatial pixel binning from 3×3 to 5×5 on the raw image to increase the number of photons per pixel. By selecting the time windows and the detection channels for the donor images and the FRET and acceptor images, the PIE scheme allowed for the one-step reconstruction of the donor dye FLIM image, FRET fluorescence image, and directly excited acceptor FLIM image ( Supplementary Fig. S4). For the analysis of specific aggregates, e.g., the aggregates that showed similar τ A488 values, we selected those particles that were present in all three donors, FRET images, and acceptor images above a certain photon count threshold ( Supplementary Fig. S4). high concentration of sample (from 0.5 to 2 nM) that contained the same number of S11/S19 time channels as the dataset to be analyzed. Only the pixels that surpassed a threshold of 5 counts at the maximum channel were analyzed to discard the background fluorescence. After the FLIM analysis, the images were displayed as pseudo-color plots depending on the τ A488 value (Supplementary Fig. S5). The fluorescence lifetime distributions of τ A488 shown in Fig. S5 are constructed only from pixels containing aggregates (coincident pixels in the three images: F A488 , F FRET , and F A647N ).

Supplementary
Supplementary Figure S5 A

S3.2 Kinetic studies using FLIM. A home-coded
Fiji is just ImageJ 14 macro was used as an alternative analysis for the kinetic studies performed using FLIM data 14 to obtain the average τ A488 within the aggregates in the images. In detail, after exporting the output matrix data of the FLIM analysis, a Gaussian smoothing function was applied (s.d. = 1, in pixels). Then, the three channels (F A488 , F FRET , and F A647N ) were semiautomatically segmented based on the pixel intensity using an isodata algorithm ('getAutoThreshold' method or 'make binary' function in Fiji and binary masks created, that is, 0 for background and 1 for oligomeric species). By multiplying each intensitymasked channel, we obtained a final image, which only contains the values for the coincident events. Finally, we multiplied the final masked images by the τ A488 to obtain a final image where each intensity-pixel was associated with the corresponding τ A488 (Fig. 4, main text). For the kinetic study, the average value of τ A488 in all the pixels containing oligomers was assessed. S14/S19