Aβ42 fibril formation from predominantly oligomeric samples suggests a link between oligomer heterogeneity and fibril polymorphism

Amyloid-β (Aβ) oligomers play a central role in the pathogenesis of Alzheimer's disease. Oligomers of different sizes, morphology and structures have been reported in both in vivo and in vitro studies, but there is a general lack of understanding about where to place these oligomers in the overall process of Aβ aggregation and fibrillization. Here, we show that Aβ42 spontaneously forms oligomers with a wide range of sizes in the same sample. These Aβ42 samples contain predominantly oligomers, and they quickly form fibrils upon incubation at 37°C. When fractionated using ultrafiltration filters, the samples enriched with smaller oligomers form fibrils at a faster rate than the samples enriched with larger oligomers, with both a shorter lag time and faster fibril growth rate. This observation is independent of Aβ42 batches and hexafluoroisopropanol treatment. Furthermore, the fibrils formed by the samples enriched with larger oligomers are more readily solubilized by epigallocatechin gallate, a main catechin component of green tea. These results suggest that the fibrils formed by larger oligomers may adopt a different structure from fibrils formed by smaller oligomers, pointing to a link between oligomer heterogeneity and fibril polymorphism.


Introduction
Protein aggregation is involved in a wide range of human diseases [1]. hallmark in these disorders. For example, amyloid-b (Ab) aggregation leads to the formation of senile plaques in Alzheimer's disease [2]; a-synuclein forms Lewy bodies in Parkinson's disease [3]; and the tangles in various tauopathies are the result of tau aggregation [4]. The end aggregation product of these proteins is called amyloid [5], which has well-defined histochemical, biophysical and biochemical characteristics [6,7]. To understand the pathology in these amyloid-related disorders, the process of protein aggregation that leads to amyloid formation has attracted extensive research efforts. At the same time, it has been well established that protein aggregation also leads to the formation of soluble intermediates, which are often called oligomers. These oligomers are shown to be more toxic than the amyloid fibrils in a variety of activity assays using cultured cells and transgenic animals (reviewed in [8]). Some studies suggest that oligomers eventually are converted to amyloid fibrils [9,10] or form the building blocks of amyloid fibrils [11 -15], and other studies show that the amyloid fibrils can also promote the formation of oligomers [16]. The relationship between oligomers and amyloid fibrils in the aggregation process is critical for a complete understanding of the underlying aggregation process and for a rational approach to develop therapeutic interventions.
In Alzheimer's disease, the senile plaques are mainly composed of the amyloid fibrils of Ab protein [2]. Ab is not a homogeneous species, consisting of two main isoforms: Ab40 and Ab42, and other minor truncated and modified Ab variants. Ab40 is the most abundant Ab isoform in the brain [17 -19], but Ab42 is the major Ab isoform in the amyloid plaques [20][21][22][23]. Experimental evidence suggests that Ab42 and Ab40 form interlaced amyloid fibrils [24] and they can seed each other's aggregation [25][26][27]. Ab fibrillization starts with a nucleation process in which small fibril nuclei are formed [28]. These fibril nuclei adopt the structure of the fibrils and are capable of growth by monomer addition. Fibrils are also capable of promoting nucleation through a secondary nucleation process [16]. Ab oligomers have been shown to adopt different structures from fibrils [29 -31]. The relationship between oligomerization and fibrillization is not fully understood. Some studies suggest a fibrillization model of nucleated conformational conversion [9,10,30]. This model starts with the formation of Ab oligomers, which then convert to fibril nuclei. There are also studies suggesting that Ab forms ringlike oligomers, which form the building blocks of amyloid fibrils [15,32,33]. It is unclear how Ab aggregation proceeds in the brain.
To this end, here we report that Ab42 aggregation starts with oligomerization. In phosphate-buffered saline (PBS), Ab42 forms oligomers almost immediately without fibril formation. Monomers account for only a small percentage of the Ab42 population. Upon incubation at 378C, the Ab42 sample with predominantly oligomers proceeds to form fibrils and display sigmoidal aggregation kinetics. We further show that the Ab42 sample enriched with smaller oligomers form fibrils at a faster rate than Ab42 sample enriched with larger oligomers. Ab42 fibrils formed by small and large oligomers appear to have different structural characteristics, suggesting a link between oligomer heterogeneity and fibril polymorphism.

Ab42 quickly forms oligomers with a wide range of sizes
For all the studies in this work, we used recombinant Ab42 proteins with a protocol that we optimized over the years. The recombinant Ab42 protein in this work does not contain any extra residues. All the purification steps were performed in high-pH buffers in the presence of chemical denaturants to limit the formation of Ab aggregates. Following lyophilization, the Ab42 powder was treated with hexafluoroisopropanol (HFIP). HFIP treatment has become a common practice in Ab aggregation studies. To study aggregation, Ab needs to be dissolved in an aqueous buffer, which is often the PBS buffer. For this purpose, we first dissolved Ab42 protein in a high-pH denaturing buffer containing 7 M guanidine hydrochloride at pH 11. Both high pH and high concentrations of denaturant help inhibit Ab aggregation. Then we performed a buffer exchange step with a desalting column to transfer Ab protein to PBS buffer. We checked the morphology of the Ab sample after buffer exchange using transmission electron microscopy (TEM), which shows that the Ab42 sample contains abundant oligomers (figure 1a). These oligomers are largely globular, with diameters ranging from 4 to 15 nm (figure 1b). Most of the oligomers have diameters of 6-11 nm. Oligomers of similar sizes have also been previously reported for both Ab40 and Ab42 [13,34,35].
It was a surprise for us to see abundant oligomers in the Ab42 sample. The time it took from solubilizing Ab42 in the denaturing buffer to the preparation of TEM grids was approximately 1 h.
royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 190179 The Ab42 sample was kept on ice with the exception of the buffer exchange step, which took approximately 10 min. And the Ab42 concentration was approximately 50 mM in the high-pH denaturing buffer before the buffer exchange step and was approximately 20 mM after buffer exchange to PBS. As a reference to the larger context of Ab42 oligomer preparation, preparation of Ab-derived diffusible ligands (ADDLs) requires incubation of 100 mM Ab42 at 48C for 24 h [36]. Ab42 globulomers requires incubation of 100 mM Ab42 at 378C for 24 h [37]. Therefore, we expected that the Ab42 sample consists of predominantly monomers. However, this unexpected outcome provides an opportunity to study the oligomer distribution and fibril formation starting from these oligomers. Previously, fast oligomerization has also been observed for both Ab40 and Ab42 [9,13,34,35,38,39], suggesting that oligomerization may be the first step of Ab aggregation.
Next we studied the size distribution of Ab42 oligomers. Our approach was to take advantage of the ultrafiltration filters that are available in a wide range of pore sizes. We took aliquots of the Ab42 sample immediately after the buffer exchange step and filtered through ultrafiltration filters with seven molecular weight cutoffs: 10, 30, 50, 100, 300, 1000 kDa and 0.2 mm. Then we measured the concentration of the different filtrate samples using the fluorescamine method [40]. If all Ab42 proteins are monomers, which are 4.5 kDa in size, then the filtrate concentration from various filters would be very similar. Differences in filtrate concentration, on the other hand, would reflect the distribution of Ab42 oligomer size. Figure 1c shows the Ab42 concentrations after filtering through various ultrafiltration filters. The filtrate concentrations decrease with decreasing pore sizes of the filters, consistent with the existence of large oligomers of various sizes. If we assume that most of the oligomers larger than the nominal size of the filter will stay in the retentate, and most of the oligomers smaller than the nominal size of the filter will stay in the filtrate, then the concentration difference between two filtrates would reflect the amount of oligomers whose sizes are in between the two filter sizes. For example, the concentration of the unfiltered sample in figure 1c is 17.0 mM, and the concentration of the 0.2 mm filtrate is 11.5 mM, therefore, the concentration of the existing Ab42 aggregates that are larger than 0.2 mm is the concentration difference, 5.5 mM, or 32% of the total Ab42 sample before filtration. With the same calculation, we obtained the oligomer distribution of various size groups, and this is shown as a pie chart in figure 1c. Although this pie chart represents a rough estimate of oligomer size distributions, it still gives us an idea of the relative population of different oligomers. We were surprised to see that the concentration of the 10 kDa filtrate is only 1% of the unfiltered sample. Even the concentration of the small oligomers in 100 kDa filtrate is only approximately 15% of the total Ab. As a control, we filtered hen egg white lysozyme, a monomeric protein of 14 kDa, through 30, 50, 100, 1000 kDa, and 0.2 mm filters. The results for lysozyme show that the concentrations of the filtrate are greater than 90% of the unfiltered sample for all the filters (figure 1d). The 50 kDa, 1000 kDa and 0.2 mm filtrates are 94-96% of the unfiltered sample, and the 30 kDa and 100 kDa filtrates are 91 -92% of the unfiltered sample. The filtration studies of lysozyme provide support for the validity of the filtration method to study oligomer size distribution.
We realize that studying oligomer size distribution with ultrafiltration filters seems to be a crude method at first glance. But we argue that ultrafiltration is actually an excellent approach for the following reasons. First, although each individual filter is not capable of resolving different sizes of oligomers, combination of filters with different molecular weight cutoffs can provide a picture that is not available through other means. Second, the time required for ultrafiltration is only 10-20 min and the filtration step can be performed at low temperatures using refrigerated centrifuge. This will help limit further oligomerization and fibrillization. One could argue that size exclusion chromatography may give better separation of oligomers of different sizes. In practice, we have not seen any reports that show discreet size exclusion peaks corresponding to Ab oligomers of different sizes. There are reports in which researchers have used fraction numbers as an indication of oligomer size. When used this way, size exclusion chromatography lacks the resolution required to distinguish Ab oligomers of different sizes. Ultrafiltration filters have been used by other groups as a way to separate Ab oligomers into different size groups, and have successfully shown that oligomers separated using ultrafiltration filters have different bioactivities [41,42].
We next investigated the sample-to-sample and batch-to-batch variations in terms of the size distribution of Ab42 oligomers. We studied three purification batches of Ab42. Each batch of Ab42 was prepared from independently cultured E. coli cells, and multiple tubes of Ab42 were obtained from each batch of preparation. Within each batch, we investigated three tubes of Ab42 sample. In total, nine tubes of HFIP-treated Ab42 powder were separately dissolved in the high-pH denaturing buffer called CG. The concentration of Ab42 in CG buffer was 50 mM for all nine tubes. After buffer exchange to PBS buffer, two aliquots of 400 ml from each sample were filtered through 0.2 mm or 100 kDa filters, and the concentration of the filtrate was measured using the fluorescamine method [40]. The results of measured concentrations are shown in figure 2. The concentrations of both the 0.2 mm and 100 kDa filtrates show a wide range even though the starting concentrations in CG buffer are all the same at 50 mM. These results suggest that the initial oligomerization process is controlled by a stochastic factor. As a result, the oligomer size distribution is not reproducible from one tube of sample to another. This stochastic control of oligomerization may be similar in principle to the stochastic nucleation for Ab aggregation proposed by Fändrich and co-workers [43].

Ab42 samples enriched with smaller oligomers form fibrils faster than Ab42 samples enriched with larger oligomers
We then investigated how these Ab42 samples form fibrils after ultrafiltration. We studied three filtrate samples using the 100 kDa, 1000 kDa and 0.2 mm filters. These Ab42 filtrates all started from a single tube of HFIP-treated Ab42, which was dissolved in the high-pH denaturing buffer, buffer exchanged to PBS, and then split into three aliquots for ultrafiltration with 100 kDa, 1000 kDa and 0.2 mm filters. The 100 kDa filtrate represents Ab42 monomers and small oligomers; 0.2 mm filtrate has a majority of larger oligomers; and the 1000 kDa filtrate represents a sample that is devoid of largest oligomers. After concentration measurements using fluorescamine, we set up fibrillization experiments at 378C without agitation. Two Ab42 concentrations at 1.7 and 3.4 mM, and three repeats for each concentration, were incubated in a plate reader to allow continuous monitoring of aggregation. The aggregation curves are shown in figure 3a. The aggregation data were fitted to a sigmoidal equation to obtain the lag time, half time and apparent fibril growth rate (see Material and methods and   figure S1). The 100 kDa filtrate samples form fibrils faster than both the 1000 kDa and 0.2 mm filtrates, at both 1.7 mM and 3.4 mM concentrations. At lower concentration (1.7 mM), the 1000 kDa filtrate forms fibrils faster than the 0.2 mm filtrate, but the difference between the 1000 kDa and 0.2 mm filtrates disappeared at higher concentration (3.4 mM), suggesting that faster aggregation may obscure differences in aggregation kinetics between some samples. We also noticed that the 100 kDa filtrate has faster apparent fibril growth rate than both the 1000 kDa and 0.2 mm filtrates (figure 3b). Furthermore, the fibril growth rate for the 100 kDa filtrate more than doubles from 1.7 to 3.4 mM Ab42. By contrast, the 1000 kDa and 0.2 mm filtrates show marginal increase when Ab42 concentration is increased from 1.7 to 3.4 mM. Two main factors contribute to the apparent fibril growth rate: the rate of fibril elongation and the concentration of fibril nuclei. Because the same Ab42 protein was used, the rate of fibril elongation is likely similar for different filtrates. Therefore, we speculate that the high fibril growth rate for the 100 kDa sample is mainly due to higher concentrations of fibril nuclei than the 1000 kDa and 0.2 mm filtrates.
We examined the aggregation products after 24 h of incubation using TEM. The electron micrographs show fibrillar morphology for all three filtrate samples (figure 3c), confirming that the aggregation reactions we are studying here are fibrillization.
We considered the possibility that the differences in aggregation kinetics resulted from concentration differences in the aggregation reaction. In other words, we may underestimate the concentration of the 100 kDa filtrate, or overestimate the concentration of the 1000 kDa and 0.2 mm filtrates. We do not believe that this is likely for the following reasons. First, we used fluorescamine to measure the concentration of the Ab filtrate [40]. Fluorescamine is a small molecule of 276 Da. It reacts with the primary amine groups of N-terminus or lysine. In oligomers, it is possible that the primary amines are buried inside the protein core and are thus not accessible to fluorescamine. As a result, this would lead to an underestimate of Ab concentration. If this were the case, the concentration of larger oligomers would be underestimated compared to smaller oligomers. The aggregation kinetics show that samples with larger oligomers form fibrils slower, in contrast to this possibility. Second, the fluorescence intensity at aggregation plateau suggests that our concentration measurement is accurate. Xue et al. [44] showed that thioflavin T fluorescence is a quantitative measure of fibril quantity. For Ab42 aggregation at 1.7 mM, all three filtrates reach similar fluorescence intensity at aggregation plateau. For Ab42 aggregation at 3.4 mM, the 100 kDa filtrate reached a higher fluorescence intensity than the 1000 kDa and 0.2 mm filtrates. We speculate that the 100 kDa filtrate forms fibrils of a different structure from the 1000 kDa and 0.2 mm filtrates, and this point is further discussed below. Third, even if we compare the 100 kDa filtrate at 1.7 mM with the 1000 kDa and 0.2 mm filtrates at 3.4 mM, the 100 kDa filtrate still forms fibrils faster, and has a faster apparent fibril growth rate, suggesting that the difference in fibrillization kinetics between smaller and larger oligomers is not due to inaccurate concentration measurements.
We next investigated the reproducibility of our observation with a different Ab preparation batch ( figure 4a,b), and an Ab sample without HFIP treatment (figure 4c,d). In both cases, we observed that the 100 kDa filtrate forms fibrils faster than the 0.2 mm filtrate, and the 1000 kDa filtrate forms fibrils at a rate that is in between the 100 kDa and 0.2 mm filtrates. Similarly, we also observed that the fibril growth rate of the 100 kDa filtrate increases dramatically from 1.7 to 3.4 mM Ab42, while the 1000 kDa and 0.2 mm filtrates show marginally higher fibril growth rate at higher Ab42 concentrations.
In addition to the faster fibrillization rate, smaller Ab42 oligomers often show higher thioflavin T fluorescence at aggregation plateau, especially at the higher concentration that we studied (figures 3 and 4). Xue et al. [44] previously showed that thioflavin T fluorescence is a quantitative measure of fibril amount. Since we are comparing fibril formation at the same Ab42 concentration, the difference in thioflavin T fluorescence suggests a possibility that the underlying fibril structure, not just the amount of fibrils, is different for smaller and larger oligomers. Although the best evidence may come from the structures of fibrils formed by the smaller and larger oligomers, structural determination of these fibrils is a challenging task in itself and is beyond the scope of this work. However, there are other ways to study the structure based on the relationship between protein structure and its thermodynamic stability. In a separate study of the Guo laboratory (unpublished data, GP & ZG, 2019), we have found that epigallocatechin gallate (EGCG), a main component of green tea catechins, is capable of solubilizing Ab fibrils, and this effect depends on how the fibrils were prepared. Therefore, we investigated how EGCG affects the fibrillization of the 100 kDa and 0.2 mm filtrates. Figure 5 shows the aggregation of 5 mM Ab42, filtered through 100 kDa or 0.2 mm filters, in the presence of 25 mM EGCG. For the 0.2 mm filtrate, EGCG leads to a steady decrease in thioflavin T fluorescence after the fibril growth phase (figure 5a). By contrast, the thioflavin T fluorescence for the 100 kDa filtrate remained stable after the aggregation has reached plateau (figure 5b). These results suggest a difference in stability for the fibrils formed by the royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 190179 100 kDa and 0.2 mm filtrates. While the 0.2 mm fibrils can be solubilized, although not completely, by EGCG, the 100 kDa fibrils are resistant to EGCG solubilization at this combination of Ab and EGCG concentrations. We attribute this difference to the underlying structure of the 100 kDa and 0.2 mm fibrils.
The fibrillization kinetics of different Ab42 oligomers have implications regarding the mechanism of Ab aggregation. Here we consider three potential pathways of Ab aggregation that involves Ab oligomers. In the first pathway (figure 6a), Ab aggregates to form oligomers, and then oligomers grow in size and eventually form fibrils. Our results are not in favour of this pathway because larger oligomers form fibrils at a slower rate (figures 3 and 4). In the second pathway (figure 6b), Ab    In the third pathway (figure 6c), Ab aggregates to form oligomers of different sizes, and different oligomers proceed to form fibrils without interconversion between different oligomers. This would explain the different thioflavin T fluorescence amplitude from different oligomers upon fibrillization, and the potential structural difference between these fibrils ( figure 5). The step of oligomer to fibril formation is via nucleated conformation conversion, which has been proposed previously for Ab [9,10]. Similar mechanisms have also been proposed for other amyloid systems [45,46]. The contribution of our model here is to include different types of oligomers and to suggest that different oligomers form fibrils in parallel fibrillization pathways. Therefore, our model suggests that Ab42 samples of smaller oligomers would form a higher concentration of fibril nuclei, consistent with the higher apparent fibril growth rate (figures 3 and 4). The shorter lag time for the fibrillization of smaller oligomers is likely due to a faster conversion rate from oligomers to fibril nuclei. Previously, we have also suggested that the mechanism of this nucleated conformational conversion is a b-strand rotation, which converts antiparallel b-structure in oligomers to parallel b-sheet structure in fibril nuclei [30]. A similar structural conversion was also proposed by Fu et al. [10] based on NMR studies. Our model of Ab aggregation suggests a link between oligomer heterogeneity and fibril polymorphism. The issue of fibril polymorphism has been long recognized in the amyloid research community [47]. Different fibril polymorphs have been observed even in the same sample prepared either in vitro [48] or in vivo [49], and may lead to different pathology in Alzheimer's disease. Different underlying structures of Ab fibrils have been found in different clinical subtypes of Alzheimer's disease patients [50 -52]. Our results suggest that Ab42 oligomers of different sizes may form fibrils of different structures. Therefore, oligomer heterogeneity leads to fibril polymorphism. The heterogeneity of oligomers is likely a result of the stochastic nature of the oligomerization process. Due to the presence of two hydrophobic regions (the central hydrophobic cluster and the C-terminal region) in Ab sequence and their importance in Ab aggregation, the initial oligomerization process is likely driven by hydrophobic collapse. Similar to protein folding where the transition state is similar to the native state of a folded protein [53], the structures of the oligomers also dictate the final structures of fibrils that they eventually form. Given the diversity of Ab42 oligomers that have been identified in vitro and in vivo, and the highly polymorphic nature of Ab42 fibrils, our results here point to the potential convergence of these two lines of research inquires.

Preparation of Ab42 proteins
Detailed protocols for Ab42 preparation have been described previously [25,44]. Briefly, Ab protein was expressed in Escherichia coli as a fusion protein, GroES-ubiquitin-Ab, and was purified using a nickel column in a high-pH denaturing buffer. After purification, the fusion protein was cleaved to obtain full-length Ab. Full-length Ab was then buffer exchanged to 30 mM ammonium acetate ( pH 10). At this stage, Ab42 proteins in ammonium acetate from several purifications were pooled together, and then were aliquoted to small volumes in 1.5-ml microcentrifuge tubes for lyophilization. This pooled Ab42 sample is called a purification batch. Each batch of Ab consists of multiple tubes of Ab powder, and each tube of Ab is identical to other tubes within the same batch. The lyophilized Ab powder was then treated with HFIP by incubating at approximately 100 mM Ab concentration at room temperature with shaking (1000 r.p.m.) for 24 h. The HFIP was allowed to dry overnight in a chemical hood. The resulting Ab film is stored at 2808C.  royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 190179 a 5-ml HiTrap desalting column (GE Healthcare). Depending on the requirement of Ab volume, multiple tubes of Ab may be prepared and pooled together after buffer exchange step. Following buffer exchange, 400 ml of Ab was pipetted to ultrafiltration filters of different molecular weight cutoff sizes ranging from 10 kDa to 0.2 mm and centrifuged at 48C at 14 000g. For all the experiments described here, the ultrafiltration filters of 10, 30, 50, 100, 300 and 1000 kDa were obtained from Sartorious (Vivaspin 500 product line), and the 0.2 mm filters were obtained from PALL (Nanosep product line). All filters were pre-washed with PBS buffer. The centrifugation was monitored so that most of the sample is in the filtrate and the retentate volume is minimal.

Measurement of Ab concentration using the fluorescamine method
Ab concentration was measured using the fluorescamine method as previously described [40]. A standard curve using hen egg white lysozyme (Fisher Bioreagents, crystalline powder) was prepared for each experiment of concentration measurement. The range of amine concentrations for the standard curve is 0.3-7 mM. The volume of each assay was 50 ml, containing 500 mM fluorescamine. The assay for Ab was performed the same way as the lysozyme standard. The Ab sample was also diluted when needed so that the fluorescence reading is in the middle of the concentration range of the standard curve. The concentration of each Ab sample was measured in triplicates.

Ultrafiltration of hen egg white lysozyme
As a control experiment, we prepared a lysozyme solution at 10.17 mM. Then 400 ml of aliquots were filtered through filters of five molecular weight cutoff sizes: 30 kDa, 50 kDa, 100 kDa, 1000 kDa and 0.2 mm. These filters were obtained from the same manufacturers as described above for Ab ultrafiltration experiments. Following centrifugation, the lysozyme concentration of each filtrate was measured using absorbance at 280 nm and an extinction coefficient of 38 mM 21 cm 21 [54]. This ultrafiltration experiment was repeated for two additional times to obtain final concentration data in triplicates.

Transmission electron microscopy
Ab samples were applied on to glow-discharged copper grids (400 mesh formvar/carbon film, Ted Pella) and stained with 2% uranyl acetate. The EM grids were then examined using a FEI T12 electron microscope with an accelerating voltage of 120 kV.

Fibrillization of Ab42 samples following ultrafiltration
For each of the three aggregation experiments reported in figures 3 and 4a,b, one tube of Ab42 (HFIPtreated for figures 3 and 4a, non-HFIP-treated for figure 4b) was dissolved in 1 ml of CG buffer, and then buffer exchanged to PBS. Following buffer exchange, three aliquots of 400 ml Ab42 were filtered through the 0.2 mm, 1000 kDa and 100 kDa ultrafiltration filters at the same time using a refrigerated centrifuge. Centrifugation was monitored so that most of the volume is in the filtrate and the retentate volume is minimal. Then the concentration of the filtrates was determined using the fluorescamine method as described above. Ab42 fibrillization assays were set up at two Ab42 concentrations: 1.7 and 3.4 mM. A 50-ml assay was prepared by mixing 5 ml of thioflavin T stock (200 mM in PBS, freshly prepared), Ab42 filtrate samples and PBS buffer. The thioflavin T control was prepared using 5 ml of thioflavin T stock and 45 ml of PBS buffer. The volumes of the Ab42 filtrate and PBS were calculated based on the final Ab42 concentration. Each aggregation condition (different filters and different Ab42 concentrations) was performed either in triplicates or quadruplicates as indicated in the figure legends. All the mixing steps were performed on ice. Finally, the fibrillization mixtures (50 ml each) were transferred to a black 384-well Non-binding Surface microplate with clear bottom (Corning product 3655) and sealed with a polyester-based sealing film (Corning product PCR-SP). The fibrillization was started by incubation at 378C in a Victor 3 V plate reader (Perkin Elmer) without agitation. The thioflavin T fluorescence was measured from the bottom of the plate with an excitation filter of 450 nm and an emission filter of 490 nm. The aggregation data are reported as fold change in fluorescence by dividing the average of thioflavin T fluorescence at each time point of measurements.
royalsocietypublishing.org/journal/rsos R. Soc. open sci. 6: 190179 3.7. Fibrillization of Ab42 samples in the presence of EGCG One tube of HFIP-treated Ab42 was dissolved in CG buffer and buffer exchanged to PBS. Then aliquots of 400 ml samples were filtered through the 0.2 mm and 100 kDa ultrafiltration filters. The concentration of Ab filtrates was determined using the fluorescamine method described above. EGCG was obtained from Tocris (greater than 98% purity, catalogue number 4524) and was dissolved in deionized water. The concentration of EGCG was calculated based on the amount (weight) and water volume. The final aggregation was set up to have 5 mM of Ab42 (either 0.2 mm filtrate or 100 kDa filtrate), 25 mM EGCG and 20 mM thioflavin T. All the mixing steps were performed on ice, and the aggregation was monitored using a plate reader as described above.

Fitting of fibrillization data
To obtain the kinetic parameters, the fibrillization data were fitted to the equation below, which was based on the sigmoidal equation described by Nielsen et al. [55] with modifications.
where S is the fluorescence intensity, S I and k I are the fluorescence and slope in the lag phase, S F and k F are the fluorescence and slope in the plateau phase, t is the time of aggregation, t 50 is the time to reach 50% of maximal fluorescence, and t is the time it takes for the fluorescence signal to change from 26.9 to 50%, or from 50 to 73.1%, of the maximum fluorescence. The lag time of aggregation is given by t 50 22t, and the maximum fibril growth rate is given by 1/4t. The fits are shown in electronic supplementary material, figures S1 -S3.
H.W., G.P. and F.H. carried out experiments, analysed data and revised the manuscript. Z.G. conceived and supervised the study, designed the experiments and drafted the manuscript. All authors gave final approval for publication.