Biomolecular condensates can both accelerate and suppress aggregation of α-synuclein

Biomolecular condensates present in cells can fundamentally affect the aggregation of amyloidogenic proteins and play a role in the regulation of this process. While liquid-liquid phase separation of amyloidogenic proteins by themselves can act as an alternative nucleation pathway, interaction of partly disordered aggregation-prone proteins with preexisting condensates that act as localization centers could be a far more general mechanism of altering their aggregation behavior. Here, we show that so-called host biomolecular condensates can both accelerate and slow down amyloid formation. We study the amyloidogenic protein α-synuclein and two truncated α-synuclein variants in the presence of three types of condensates composed of nonaggregating peptides, RNA, or ATP. Our results demonstrate that condensates can markedly speed up amyloid formation when proteins localize to their interface. However, condensates can also significantly suppress aggregation by sequestering and stabilizing amyloidogenic proteins, thereby providing living cells with a possible protection mechanism against amyloid formation.


Labelling of insulin
Insulin was labelled with FAM-NHS using the following method.Insulin was dissolved at 5 mg/ml concentration in sodium bicarbonate solution (0.1 M).FAM-NHS was dissolved in DMF at 10 mg/ml.Solution of FAM-NHS (54 μl) was added to solution of insulin (1.32 ml) and the mixture was stirred gently at 4 °C overnight.Subsequently, insulin was separated from unbound dye using Amicon Ultra-15 centrifugal filters with 3 kDa MWCO, by washing with 0.1 M carbonate buffer (4 times) 0.005 M carbonate buffer (5 times).

Partitioning of FAM-labelled insulin
Partitioning of FAM-labelled insulin was studied the same way as described for labelled protein in the main text.

ThT aggregation kinetics assays (insulin)
Aggregation assays were performed analogously to assays described in the main text.The same buffer composition was used (50 mM HEPES, 100 mM NaCl, 100 µM EDTA, 20 µM ThT).
Insulin was first dissolved in 10 mM hydrochloric acid and this stock was further diluted to obtain 50 μM insulin concentration in the aggregation assays.Other conditions remained as described in the main text.Kinetic parameters were extracted as described in the main text.

Preparation of samples and transmission electron microscopy (insulin)
Samples of insulin aggregates were prepared using samples from 384-well plate after the aggregation assay.Content of selected wells that shown aggregation in the ThT assay were mixed with a pipette and transferred onto a TEM grid (EM-Tec formvar carbon support film on copper, 300 square mesh, Micro to Nano, the Netherlands).Samples were blotted with filter paper, stained with 1.5 μl of 2% (w/w) sodium phosphotungstate solution (adjusted to pH 7.4), washed with 2 μl of water left to dry overnight.Imagining was performed using JEOL JEM-1400 FLASH.

Statistical analysis (supplementary)
Microscopy images were analysed using FIJI distribution of ImageJ.Violin plots were prepared according to the description under fig.S4 and S9.-c) and for the reference sample.Violin plots were prepared using Gaussian kernels with bandwidth determined automatically using Scott's method; density plots were cut at two bandwidth units past the extreme data points; violins are scaled to have the same area in supernatant-coacervate pairs.

Basic aggregation model
Typically for many amyloidogenic proteins, α-synuclein aggregation process may be considered an autocatalytic process.Our simple yet accurate model of α-synuclein aggregation is based on the secondary nucleation model proposed by Ferrone et al. (66) and involves 3 basic reactions: (i) primary nucleation of fibres from α-synuclein monomers, (ii) elongation of fibres by attaching monomers to one of the fibre ends, (iii) secondary nucleation catalysed by fibres: From this a set of differential equations describing concentration changes in the system can be derived: Solving this set of equations provide a kinetic trace of the aggregation process.Fitting the solution to the experimentally measured concentration of one of the species provides information about the protein aggregation rates.

Aggregation in droplets model
In case of partitioning into the coacervate droplets, the concentrations of monomer in the diluted and in the condensed phase is determined by the partition coefficient: where  P is the partition coefficient and [] cond and [] dil are the concentrations of the monomer in the condensed and the diluted phase respectively.Taking into account the equation describing the mass balance of monomers in the system: where  is the ratio of diluted phase volume to the condensed phase volume, we can write equations describing the concentrations of the monomers in the diluted and in the condensed phase: where  = 1+ + P . We assume the transport/partitioning process to be much faster than aggregation and to simplify the kinetic equations we assume further that the partitioning remains at equilibrium at every timepoint of the aggregation reaction.This leads to a set of differential equations describing aggregation process in the coacervate system with monomer partitioning: Again, similarly to the more simple case of aggregation in homogenous solution, solving the equations yields aggregation kinetic trace for both the diluted and the condensed phase.The proposed model is similar to the model previously described by Weber at al. (31), with the following main differences: we allow for different rate constants in the condensed and dilute phase, and we assume that the exchange of material between the droplet and the solution is infinitely fast.

Interface-aggregation model
Another model was developed for a case where aggregation-prone protein accumulates in the coacervate-diluted phase interface.Binding of the monomers to the coacervate interface can be described by equation: where  B is the binding constant, [] int is the concentration of interface-bound monomers and [] is the concentration of available binding sites ([] = [] tot − [] int ).Again, taking into account the mass balance equation for monomers, we can write equations describing the concentration of free and surface-bound monomers.Since the aggregation reaction occurs now only in the diluted phase (or in the interface, which is treated as a part of the diluted phase), we can omit the change of volume: We assume that the surface can act as a nucleation site, requiring one monomer from the surface and one monomer from the solution to react.If we further assume that the fibres formed at the interface can grow by attaching monomers from the solution, they can participate in secondary nucleation and that they remain attached to the interface, we can write a set of differential equations for this system: where  h is the reaction rate constant of the interface-catalysed nucleation and, for clarity, [] dil and [] int symbols were used instead of full equations dependent on [] tot .
The (local) concentration of monomers at the interface, [] int , can be estimated from partitioning experiments (fig.2) to be roughly 200 and 300 µM for the pLys/pGlu and pLys/ATP systems, respectively, which is low compared to the local concentration of pLys/pGlu or pLys/ATP inside the coacervates.Therefore, the use of a binding model that assumes single-layer adsorption seems justified.

Fig. S2 .
Fig. S2.Critical salt concentration of coacervate systems without and with αSyn variants.All coacervate systems were tested in with FL-αSyn, αSyn-108 and NACore.Differences between selected samples were tested for statistical significance (student's t-test) in coacervate droplets-supernatant control pairs."ns" indicates values above 0.05, single asterisk indicates α<0.05.

Fig. S3 .
Fig. S3.The kinetics of aggregation of different αSyn variants is altered by coacervates.Lines correspond to single aggregation experiments (ThT fluorescence intensity) of different αSyn variants in buffer (reference, grey traces), in the presence of coacervates (coloured traces), or in the presence of coacervate supernatants (dark traces).

Fig. S4 .
Fig. S4.Characteristics of αSyn fibrils aggregated in absence and presence of coacervates.(A) Distribution of fibril thickness formed by different αSyn variants in the absence (blank) or presence of coacervate systems (n=50).Violin plots were prepared using Gaussian kernels with bandwidth determined automatically using Scott's method.(B) TEM images of the fibrils formed by different αSyn variants in the absence (blank) or presence of coacervate systems.Blue marks indicate places where the diameter was measured.

Fig. S5 .
Fig. S5.Variability in lag time and maximum aggregation rate is altered by coacervates.Standard deviation of aggregation parameters for all protein variants and all coacervate systems (supernatant -s, coacervate -c) and for the reference sample.

Fig. S6 .
Fig. S6.Fitting of aggregation models to αSyn-108 and NACore aggregation.(A) Aggregation of αSyn-108 in the presence of different systems; supernatant traces with fitted curves are shown in grey (and average in red) and coacervate traces are shown in colour.(B) Aggregation of NACore in the presence of different systems; supernatant traces with fitted curves are shown in grey and coacervate traces are shown in colour.Proposed models for aggregation in the presence of coacervate systems can explain similar aggregation kinetics in the presence of droplets without partitioning, but fails to explain slower aggregation in the presence of droplets with low to moderate partitioning.

Fig. S7 .
Fig. S7.Coacervates also interact differentially with insulin.(A) Confocal microscope images of coacervate systems with FAM-labelled insulin, colourised artificially.Ratio of positive to negative charge of the coacervate components is indicated in the brackets.(B) Partition coefficient of FAM-labelled insulin determined from microscopy experiments for different coacervate systems and different charge ratios of coacervate components.

Fig. S9 .
Fig. S9.Analysis of insulin aggregation kinetics.Distribution of the lag times (tlag) and of the maximum aggregation rates (vmax) for insulin and all coacervate systems (supernatant -s, coacervate -c) and for the reference sample.Violin plots were prepared using Gaussian kernels with bandwidth determined automatically using Scott's method; density plots were cut at two bandwidth units past the extreme data points; violins are scaled to have the same area in supernatant-coacervate pairs.

Fig. S10 .
Fig. S10.Characteristics of insulin aggregates in the absence and presence of coacervates.(A) Confocal microscope fluorescence images and transmission images collected at the end of ThT aggregation assay (fig.S12).Apart from image for pLys/pGlu at 1:1 charge ratio, which was still in the growth phase, images show samples that reached aggregation plateau or were in the final stage of the growth phase.(B) TEM images of insulin aggregates formed in the presence of different coacervate systems.Insulin aggregates appear as fine fibrils.

Fig. S11 .
Fig. S11.An intramolecular FL-αSyn FRET probe reports on fibril formation.Fluorescence spectra of the FL-αSyn-based FRET probe in solution (in bulk), shortly after preparing the solution (t=0) and after 48 hours of incubation at 37 °C (t=48 h).

Fig. S13 .
Fig. S13.Coacervate surface area affects FL-αSyn aggregation kinetics in the presence pLys/pGlu coacervates.(A) Aggregation traces of FL-αSyn in the presence of different amount of pLys/pGlu coacervates.(B) Aggregation traces of FL-αSyn in the presence of coacervates dispersed in solution and fused at the bottom of the plate after centrifugation.
S3)    where:  n ,  + ,  2 are the reaction rates of the corresponding reactions,  and  2 are the nucleation numbers of primary and secondary nucleation (the lowest number of oligomers required to form a fibre nucleus), and [], [] and [] are the concentration of monomers, concentration of fibres (so 2 • [] reflects the number concentration of fibril ends) and concentration of monomeric units incorporated in fibres (proportional to fibre mass concentration and the surface available for secondary nucleation catalysis).