A hydrophobic low-complexity region regulates aggregation of the yeast pyruvate kinase Cdc19 into amyloid-like aggregates in vitro

Cells form stress granules (SGs) upon stress stimuli to protect sensitive proteins and RNA from degradation. In the yeast Saccharomyces cerevisiae, specific stresses such as nutrient starvation and heat-shock trigger recruitment of the yeast pyruvate kinase Cdc19 into SGs. This RNA-binding protein was shown to form amyloid-like aggregates that are physiologically reversible and essential for cell cycle restart after stress. Cellular Cdc19 exists in an equilibrium between a homotetramer and monomer state. Here, we show that Cdc19 aggregation in vitro is governed by protein quaternary structure, and we investigate the physical–chemical basis of Cdc19's assembly properties. Equilibrium shift toward the monomer state exposes a hydrophobic low-complexity region (LCR), which is prone to induce intermolecular interactions with surrounding proteins. We further demonstrate that hydrophobic/hydrophilic interfaces can trigger Cdc19 aggregation in vitro. Moreover, we performed in vitro biophysical analyses to compare Cdc19 aggregates with fibrils produced by two known dysfunctional amyloidogenic peptides. We show that the Cdc19 aggregates share several structural features with pathological amyloids formed by human insulin and the Alzheimer's disease–associated Aβ42 peptide, particularly secondary β-sheet structure, thermodynamic stability, and staining by the thioflavin T dye. However, Cdc19 aggregates could not seed aggregation. These results indicate that Cdc19 adopts an amyloid-like structure in vitro that is regulated by the exposure of a hydrophobic LCR in its monomeric form. Together, our results highlight striking structural similarities between functional and dysfunctional amyloids and reveal the crucial role of hydrophobic/hydrophilic interfaces in regulating Cdc19 aggregation.

Subcellular compartmentalization can be achieved via the formation of membrane-less organelles (1,2). The assembly and disassembly of many of these organelles is associated with the phase separation of proteins and RNA molecules (3)(4)(5)(6)(7). For instance, stress granules (SGs) 3 form in vivo upon stress stimuli, such as heat-shock and nutrient starvation, to protect sensitive proteins and RNA from aberrant degradation, thereby playing a crucial role in restoring viability when the stress is released (8). Many proteins associated with membrane-less compartments contain low complexity regions (LCR), which are repetitive and often disordered regions that are rich in specific amino acids, including G or Q/N motifs (6,9). Such sequences resemble features of the prion-like domains that induce aggregation of soluble peptides and proteins into amyloids. Indeed, the formation of intracellular foci and liquid phase transitions at high protein concentration could be critical factors for the nucleation and growth of certain pathological amyloid aggregates. (10). For instance, mutants of superoxide dismutase (SOD-1), an enzyme associated with amyotrophic lateral sclerosis, accumulate in SGs, triggering a transition to solid-like granules that are readily targeted by chaperones (11). Moreover, SGs have been suggested to be precursors of aberrant and irreversible aggregation (12). Phase separation similarly regulates human HnRNP2 in the nucleus, and its LCRs assemble in cross-␤ structures similar to amyloid fibrils (13). In contrast, phase separation of the RNA-binding protein FUS displays transient LCR-LCR interactions, without any conformational changes of the LCRs (14).
Although amyloid structures were originally discovered in association with a variety of pathological conditions (15,16), exciting recent findings now link amyloids to physiological functions (17)(18)(19)(20)(21), including the natural storage of peptide hormones (22) and the formation of adhesive microbe biofilms (23). Interestingly, the major isoform of yeast pyruvate kinase, Cdc19, was recently shown to assemble into functional amyloid-like structures (8) that are physiologically regulated, fully reversible, and localized to stress granules. Cdc19 binds RNA and under physiological conditions exists in an equilibrium between a homotetramer and a monomer state. Cdc19 contains a short LCR region that is hidden in the tetrameric structure but becomes solvent-exposed in monomers. The LCR region in monomers is phosphorylated when cells grow in the presence of glucose, which is thought to prevent its aggregation by intermolecular hydrophobic LCR-LCR interactions. However, upon dephosphorylation, Cdc19 monomers assemble into reversible cytoplasmic foci that contain solid aggregates with amyloid-like features (8). This system represents an interesting case in which cells are capable of generating reversible amyloid-like structures, in contrast to the essentially irreversible amyloids associated with the onset and development of several neurodegenerative disorders. The molecular mechanisms underlying the formation of protein-rich liquid compartments, functional amyloid-like structures, and pathological amyloids remain largely elusive.
In this work, we perform an in vitro biophysical analysis to compare the aggregates formed by Cdc19 with amyloid fibrils obtained with model amyloidogenic proteins such as human insulin and the peptide A␤42 strongly associated with Alzheimer's disease (24). We show that the nucleation of the Cdc19 aggregates in vitro is regulated by the equilibrium between the tetrameric and the monomeric forms of the protein and mediated by the hydrophobic LCR in the C terminus of the protein, which is prone to induce hydrophobic stacking interactions. Interestingly, the resulting aggregates share structural features with bona fide amyloids, but do not exhibit prion-like behavior, and cannot seed aggregation in bulk. Finally, we demonstrate that the exposed LCR of one monomer can interact not only with other LCRs of surrounding monomers but also with hydrophobic/hydrophilic interfaces, which can thereby trigger heterogeneous nucleation events. Taken together, these findings indicate that Cdc19 can assemble into solid aggregates with amyloid-like features and implicate regulation of LCR exposure as a governing feature of reversible aggregate assembly.

The aggregation of Cdc19 is regulated by the quaternary state
We previously showed (8) that yeast cells expressing Cdc19-GFP form small fluorescent foci upon glucose depletion that co-localize with SGs. Glucose readdition dissolved foci, and growth resumed when Cdc19 was resolubilized. Aggregate formation in vivo was likely driven by monomers, as a Cdc19-R369E variant containing a mutation at the subunit interface that destabilizes the tetramers formed reversible aggregates with increased kinetics. Conversely, a Cdc19-R49A variant (⌬PEP) with a modification in a critical residue required for binding the metabolite phosphoenolpyruvate (PEP) stabilized the homotetramer state and failed to form reversible aggregates upon glucose starvation. To confirm the predicted quaternary states of WT and the Cdc19 variants in vitro, we analyzed the recombinantly expressed and purified Cdc19 proteins by sizeexclusion chromatography coupled with multiangle light scattering (SEC-MALS) (Fig. 1A). In addition to the R369E mutant, we also examined a Cdc19 variant in which the residues Thr 372 , Thr 376 , Ser 377 , and Ser 385 have been replaced by alanine (4A) to mimic the unphosphorylated LCR. As expected, the SEC-MALS experiments confirm that Cdc19-⌬PEP is present as

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
tetramers, whereas the 4A and R369E mutants are monomeric, and WT Cdc19 is a mixture of tetramers and monomers in equilibrium.
Cdc19 contains a LCR composed of 23 amino acids. Only three LCR residues are charged, and 10 amino acids have a high hydrophobicity score (Fig. 1B), suggesting that the sequence has poor solubility in aqueous solutions. LCR regions in the homotetramer assume an ␣-helical fold within the core of the oligomer, and these domains are therefore protected from the environment. However, tetramer destabilization and shift toward the monomeric species lead to exposure of the LCR, as demonstrated by a binding assay based on the dye 8-anilinonaphthalene-1-sulfonic acid (ANS) (Fig. 1C). The fluorescence intensity of ANS increases upon binding to hydrophobic patches (25). We measured the fluorescence intensity of ANS in the same solutions analyzed by SEC-MALS to correlate the quaternary structure of the protein with the total amount of exposed hydrophobic surface. The fluorescence intensity measured with Cdc19-⌬PEP was comparable with the control buffer, whereas the monomeric Cdc19 -4A and R369E mutants exhibited a significant fluorescence increase. As expected, WT Cdc19 showed intermediate values between Cdc19-⌬PEP and the monomeric variants. These results confirm that the equilibrium shift toward monomeric species leads to solvent exposure of the LCR, which in turn increases the total hydrophobic surface of the protein. Exposure of the LCR in the monomeric species can therefore potentially trigger aggregation by inducing intermolecular hydrophobic LCR-LCR interactions.
We observed formation of WT Cdc19 aggregates in vitro within a few hours at room temperature in aqueous buffer. These clumps were visible by optical microscopy and were stained by the dye thioflavin T (ThT), a common reporter of the intermolecular ␤-sheet structures of amyloid fibrils (Fig. 1D). These results suggest that Cdc19 aggregates exhibit structural similarities with amyloids. Kinetic analysis of the aggregation process by the ThT fluorescence assay revealed that Cdc19-⌬PEP remains soluble as a tetramer for several hours, whereas WT Cdc19 and the monomeric 4A and R369E mutants formed ThT-positive aggregates within 2 h (Fig. 1D). The aggregation profiles of WT Cdc19 and the 4A and R369E monomers were essentially identical, indicating that the WT Cdc19 rapidly shifted toward monomer under the in vitro conditions. More-over, the fluorescence values at complete conversion were proportional to the initial protein concentrations (Fig. S1), indicating that the fluorescence intensities were essentially proportional to the amount of aggregates present in the system. We further validated these results by measuring the soluble fraction at the end of the aggregation process. To this purpose, aggregates were removed by centrifugation, and the concentration of the soluble species was analyzed by UV absorbance at 280 nm. As shown in Fig. 1E, most of the tetrameric Cdc19-⌬PEP mutant remained in solution, whereas the conversion of the soluble WT Cdc19 and the 4A and R369E mutants into insoluble aggregates was essentially complete.

Cdc19 aggregates exhibit amyloid-like features
We next characterized the structures of the Cdc19 aggregates by comparing their biophysical properties with amyloid fibrils obtained from two model amyloidogenic proteins, human insulin and the peptide A␤42. Insulin amyloids were formed under conventional acidic conditions (26,27), whereas A␤42 fibrils were obtained as described previously (28,29) to guarantee reproducibly with this unstable peptide. We measured the secondary structure content of dried samples of Cdc19 aggregates and insulin amyloids by FTIR spectrometry ( Fig. 2A). A␤42 fibrils could not be analyzed with this method because of the low concentrations available. The spectra of Cdc19 aggregates and insulin amyloids exhibited very similar shapes, and in particular they both displayed a peak at ϳ1630 cm Ϫ1 , which is typical of intermolecular antiparallel ␤-sheet strands. Circular dichroism spectra confirmed that the Cdc19 aggregates contained a large amount of ␤-sheet structures, which was comparable with the secondary structure of amyloids from human insulin and A␤42 (Fig. 2B) and differed from the ␣-helical structures of monomeric Cdc19 (8). We therefore conclude that Cdc19 aggregates share not only tinctorial (ThT staining) but also structural properties (FTIR and CD spectra) with amyloids. However, analysis of the same samples by transmission EM imaging revealed different morphologies of Cdc19 aggregates and insulin fibrils at the mesoscale (Fig. S2), although fibrillar structures can also be observed within the Cdc19 aggregates.
One of the most important features of amyloid fibrils is their high thermodynamic stability, which results largely from the

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
high number of hydrogen bonds that monomeric proteins can form along the amyloid structure (16). Amyloids are therefore irreversible even upon moderate denaturing conditions. However, Cdc19 aggregates in cells are reversible, indicating unresolved questions of thermodynamic stability and energy required by cells to dissolve these aggregates. We compared the stability of Cdc19 aggregates and A␤42 fibrils by incubating the solutions at different concentrations of SDS at 40°C for 2 days and monitored the structural changes of the aggregates by dynamic light scattering (Fig. 2C). Both types of aggregates showed negligible change in the light scattering correlation function, indicating that both types of structures are resistant to the applied denaturation treatments. This experiment was repeated by varying the incubation times and type of denaturant (e.g. guanidinium hydrochloride), and in all cases no significant difference was observed between Cdc19 aggregates and A␤42 fibrils. Although these experiments cannot provide quantitative parameters of the thermodynamic stability, we conclude that Cdc19 aggregates obtained in vitro are not significantly less stable than amyloid fibrils.

Cdc19 aggregates do not have prion-like behavior
Amyloid conversion into insoluble aggregates is typically the consequence of a cascade of several reactions of nucleation and growth (15,30). Chemical kinetic analysis emerged as a powerful tool to identify the underlying mechanisms from the measurement of macroscopic aggregation profiles (31,32). The approach builds on the analysis of different aggregation reactions performed at different initial concentrations of proteins (33). This analysis provides important information about the reaction orders and the relative contribution of the individual microscopic reactions of nucleation and growth on the overall aggregation process. We monitored the aggregation profiles of the Cdc19 -4A and R369E mutants at different initial concentrations by recording the intensity of ThT fluorescence over time (Fig. 3, A and B). The individual profiles exhibited a sig-moidal shape that resembled the reaction profile that is typically observed during the formation of amyloid fibrils from solutions of soluble peptides and proteins. An initial lag-phase of ϳ20 min was followed by rapid growth until a plateau, corresponding to full monomer conversion. Repetition with independent samples showed reproducibility (Fig. S3). The halftimes of the aggregation profiles scaled as a function of the initial protein concentration according to a power law with exponent equal to Ϫ0.11 and Ϫ0.21 for the Cdc19 -4A and the R369E mutants, respectively (Fig. 3, C and D). These very low values indicated that the aggregation process is essentially independent of the initial monomer concentration. Aggregation profiles of WT Cdc19 showed similar behavior, with a scaling exponent equal to Ϫ0.31 (Fig. S4).
We next investigated whether Cdc19 aggregates have seeding properties and could abolish the lag time observed in the unseeded aggregation profiles. We performed aggregation assays with freshly purified soluble Cdc19 -4A and R369E proteins in the absence and presence of 10% monomer equivalent Cdc19 aggregates at two different soluble protein concentrations (7.2 and 1.8 M). We did not observe any difference in the aggregation profiles with either sonicated or nonsonicated seeds (Fig. 3, E and F, and Figs. S5 and S6), indicating that the Cdc19 aggregates cannot accelerate the aggregation kinetics. This behavior is different from the situation typically observed with amyloids, in which preformed aggregates exhibit prionlike behavior and accelerate the conversion of soluble monomers into insoluble fibrils by promoting microscopic reactions of elongation and secondary nucleation events.

Hydrophobic/hydrophilic interfaces regulate Cdc19 aggregation
The essentially zero order reaction suggests the presence of a saturated process within the network of microscopic reactions that lead to the formation of Cdc19 aggregates. Moreover, the absence of seeding capability indicates that no reactions of

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
elongation or fibril-dependent secondary nucleation occur in bulk. These findings can be explained by the possible role of surfaces in triggering the aggregation process, which has been observed with some amyloid systems (34 -36). To test this hypothesis, we performed aggregation assays with the Cdc19 -4A and R369E mutants in sealed glass capillaries, in which the air/water interface was removed and the protein interactions with the hydrophilic material of the walls was limited (Fig. 4A). Aggregation was monitored by recording ThT fluorescence using an epifluorescence microscope. Under these conditions we did not observe aggregation after 3 days of incubation at room temperature (Fig. 4, B and C). By contrast, when we incubated the same protein solution at room temperature in a multiwell plate, the formation of aggregates was observed within a few hours, as shown by images acquired with ThT staining under an epifluorescence microscope (Fig. 4, D  and E). A control solution of A␤42 at 20 M in sealed capillaries was observed to form amyloids within 1 day (Fig. S7). These data support the important role of hydrophobic/hydrophilic interfaces in triggering the aggregation of Cdc19.
To further confirm this result, we performed kinetic assays in the presence of hydrophobic and hydrophilic nanoparticles (NPs) (36) (Fig. 5). In the presence of hydrophobic NPs, both the WT Cdc19 and the R369E mutant formed more particles of smaller sizes compared with the aggregates generated in the absence of NPs, indicating an increase in the number of nucleation events. By contrast, the aggregation of Cdc19-⌬PEP was unaffected by the presence of hydrophobic NPs. The addition of hydrophilic PEGylated NPs did not alter the aggregation behavior of either monomeric or tetrameric Cdc19 protein, thereby indicating the absence of interactions between Cdc19 species and hydrophilic interfaces. These results, together with the analysis shown in Fig. 1, demonstrate that the exposure of the hydrophobic LCR upon destabilization of the tetramer drives Cdc19 aggregation, because hydrophobic LCRs are prone to interact not only with other monomeric proteins but also with other hydrophobic interfaces present in the system, thereby promoting heterogeneous primary nucleation events.

Discussion
Our biochemical studies have revealed features of the conversion of soluble Cdc19 proteins into amyloid-like aggregates. The shift of the equilibrium from the homotetramer toward the monomeric species plays a key role in inducing aggregation (Fig. 6). The C terminus LCR of Cdc19 in the homotetramer state is buried within the oligomeric structure, and monomerization leads to exposure of the hydrophobic LCR (Fig. 1). This biochemical feature strongly resembles the behavior of transthyretin, a transport protein whose aberrant aggregation is associated with several forms of systemic amyloidosis (37). Similar to Cdc19, transthyretin is present mainly as a homotetramer. Tetramer destabilization into monomers induces the formation of an aggregation-prone species, which aggregates into amyloid fibrils (38). Indeed, stabilization of the tetrameric structure by small molecules has revealed a successful strategy to inhibit the formation of amyloids (39 -41). In contrast with the pathological insoluble structures of transthyretin, Cdc19 forms functional aggregates that are readily resolubilized to allow efficient reinitiation of cell growth following stress relief (8). Thus, in contrast to the aberrant formation of amyloids, cells have developed mechanisms to control Cdc19 aggregation. One example is represented by phosphorylation. The LCR region of Cdc19 can be phosphorylated (8), which increases solubility of the domain and may thereby prevent intermolecular hydrophobic LCR-LCR interactions under normal conditions. Changes in phosphorylation upon starvation or heat shock can trigger aggregation (8), raising the question of whether phosphorylation or other post-translational modifications represent a generic mechanism shared by other aggregation-prone proteins containing LCRs.
From a structural point of view, Cdc19 investigated in this study in vitro forms aggregates that share features of amyloids,

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
in particular secondary ␤-sheet structure, thermodynamic stability, and ability to be stained by the thioflavin T dye (Figs. 2  and 3). In contrast with amyloids, however, they do not exhibit prion-like behavior in that the Cdc19 aggregates are not able to seed aggregation (Fig. 3). This behavior is similar to the "downhill" polymerization mechanism observed with transthyretin (38). The analysis of the microscopic mechanism of Cdc19 aggregation provides an overall reaction order close to zero, indicating the presence of a saturated process (Fig. 3). Moreover, analysis of aggregation reactions in containers of different materials and in the presence of hydrophobic nanoparticles reveals that this saturated process is represented by nucleation and/or growth events occurring at the hydrophobic/hydrophilic interfaces (Figs. 4 and 5). This significant observation is consistent with the role of the hydrophobic LCR exposure in inducing aggregation. Exposed LCRs can promote intermolecular interactions not only between different monomeric proteins but also between monomeric and hydrophobic/ hydrophilic interfaces. This result is not surprising because hydrophobic/hydrophilic interfaces are well known to modu-

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
late the formation of several functional and dysfunctional amyloids. Examples include human insulin (42), ␣-synuclein, which is the protein associated with Parkinson's disease (34,35), and class I hydrophobin proteins, which are extremely surface active and form fibrillar rodlets in filamentous fungi (43,44). In particular, the amyloidogenic LCR of the hydrophobin EAS from Neurospora crassa (43) exhibits a similar composition of the LCR of Cdc19. In the case of Cdc19, the absence of seeded aggregation in bulk suggests that the growth events may occur at the interface, similar to what has been proposed for ␣-synuclein (34). Because protein-surface interfaces may potentially be relevant in the Cdc19 cellular aggregation process, it will be interesting to examine which intracellular surfaces are able to initiate Cdc19 aggregation in vivo.
Finally, although our in vitro analysis demonstrates that Cdc19 can assemble aggregates that biochemically and biophysically resemble amyloids, it remains to be demonstrated whether a similar structural organization and LCR stacking also occurs in vivo. Because the accumulation of insoluble amyloids is at the origin of several neurodegenerative disorders, it will be essential to further investigate the molecular mechanisms underlying the ability of cells to revert the formation of such stable aggregates into soluble monomers.
In summary, we have compared functional amyloid-like aggregates of the yeast pyruvate kinase Cdc19 with pathological amyloids. The molecular "switch" for Cdc19 aggregation is a change of protein quaternary structure, shifting equilibrium between homotetramer and monomer formation. The monomer species exposes a hydrophobic LCR that is prone to induce intermolecular interactions with surrounding proteins and with hydrophobic/hydrophilic interfaces, thereby triggering aggregation. We have shown that Cdc19 aggregates share amyloid features with fibrils from human insulin and the Alzheimer's disease-associated A␤42 peptide, including secondary ␤-sheet structure, thermodynamic stability, and ability to be stained by the thioflavin T dye. However, Cdc19 amyloid-like aggregates do not have prion-like behavior and are not able to recruit monomers in bulk. These results highlight striking structural resemblances between functional and dysfunctional amyloids and suggest that modulating access to the hydrophobic LCR of Cdc19 is a potential mechanism to control aggregation. In particular, based on our findings we expect that hydrophobic-hydrophilic interfaces in cells may be important for aggregation control, in addition to post-translational modifications such as phosphorylation.
The recombinant peptide A␤42 was expressed and purified following the procedure described in Ref. 29. Aggregates were obtained by incubating the peptide solution at 37°C in a 96-well half-area nonbinding polystyrene plate (Corning). Human insulin was kindly donated by NovoNordisk (Bagsvaerd, Denmark). Insulin amyloids were generated by dissolving the lyophilized peptide in 20 mM HCl solutions at pH 1.6 and incubating the samples in multiwell plates at 60°C for 12 h with shaking at 400 rpm in a plate reader (Clariostar, BMG Labtech).

Size-exclusion chromatography coupled with multiangle light scattering
The quaternary state of the WT Cdc19 and of the mutants was analyzed by injecting 50 l of samples on a Superdex 200 10/300 GL (GE Healthcare) size-exclusion column assembled on an Agilent 1200 Series at a flow rate of 0.5 ml/min. The molecular weight of the fractionated proteins was measured in-line with an 18 angle light scattering detector (Dawn Heleos II, Wyatt).

ANS assay
The relative amount of hydrophobic surface exposed by the different Cdc19 proteins was measured by an ANS-binding assay. ANS fluorescence intensity was recorded in a plate reader (Clariostar, BMG Labtech) by monitoring the emission signal at 505 nm after excitation at 365 nm. 30 M ANS was added to 10.8 M solutions of WT, Cdc19 4A, R369E, and ⌬PEP in 100 mM Tris buffer at pH 7.4 with 200 mM NaCl, 1 mM MgCl 2 , 10% (v/v) glycerol, and 1 g/liter NaN 3 . A control experiment was performed in the absence of protein in the same buffer and the same ANS concentration. Protein solutions were analyzed immediately after thawing to avoid the presence of aggregation and correlate the results with the analysis performed by SEC-MALS.

Secondary structure, stability, and size of the aggregates
The secondary structure of Cdc19 aggregates was analyzed by FTIR spectrometry on a Cary 630 instrument (Agilent Technologies) after drying the samples in situ by evaporation. The structure of the aggregates in solution was characterized by CD spectroscopy using a Jasco-815 spectrophotometer (Jasco). Far UV spectra were recorded from 250 to 205 nm in a quartz cuvette with 0.1-cm path length. The stability of aggregates obtained from Cdc19 (5 M) and A␤42 (30 M) was measured by incubating the samples in the presence of SDS at 1, 2, and 5% Under physiological conditions Cdc19 is present in an equilibrium between homotetramers and monomers. Monomeric Cdc19 exposes a hydrophobic LCR (in red) that in the presence of hydrophobic/hydrophilic interfaces triggers its aggregation into amyloid-like structures. These amyloid-like aggregates share several features with pathological amyloids (secondary ␤-sheet structure, thermodynamic stability, and staining by the thioflavin T dye) but do not exhibit prion-like behavior.

Pyruvate kinase Cdc19 forms amyloid-like aggregates in vitro
at 40°C. The stability of the aggregates was evaluated over time by measuring the size by dynamic light scattering with a Zetasizer instrument (Malvern).

Imaging
Transmission EM images were acquired on a FEI Morgagni 268 microscope. 5 l of samples were spotted on carbon support films: 400 mesh copper grids (Quantifoil, Jena) for 30 s, washed with distilled water, and negative-stained with a 2% uranyl acetate solution.
The formation of aggregates by Cdc19 was also monitored by adding ThT into the solution and recording the ThT fluorescence signal on an epifluorescence microscope (Nikon Eclipse Ti-E). The solutions were illuminated using a 455-nm High power LED light source (Omicron LedHUB light engine) equipped with 426 -446 nm excitation and 460 -500 nm emission filters (49001 ET-CFP; Chroma Technology Corporation) and a 10ϫ (NA 0.13) objective (Nikon). Images were collected using an Andor Zyla sCMOS camera (Andor). For kinetic experiments the solutions were loaded in squared glass capillaries with inner diameter of 0.2 mm (CM Scientific) and sealed with wax. The capillaries were incubated at room temperature over a glass slide and imaged for 3 days at intervals of 2 h.

Aggregation profiles
Aggregation profiles over time were recorded by monitoring the increase of the fluorescence signal of the dye thioflavin T upon binding to the Cdc19 aggregates. Before each assay, an aliquot was thawed and purified via gel filtration chromatography to remove the presence of potential seeds. The purification was performed on a Superdex 200 10/300 GL column (GE Healthcare) using a Åkta explorer system (GE Healthcare). The fractions corresponding to the soluble form of Cdc19 were collected and immediately stored on ice. Aggregation reactions were triggered by incubating the samples at 37°C in multiwell plates, using 96-well half-area nonbinding polystyrene plates (Corning). The ThT fluorescence signal was recorded in a plate reader (Clariostar, BMG Labtech) by monitoring the emission signal at 490 nm after excitation at 450 nm. Unless otherwise noted, kinetic assays were performed in 100 mM Tris buffer at pH 7.4 with 200 mM NaCl, 1 mM MgCl 2 , 10% glycerol, 1g/liter NaN 3 , and 20 M ThT. Within an individual experiment, the fluorescence measurements were performed in triplicate. Three independent repetitions were performed by using three different protein aliquots.
For seeded aggregation assays, the seeds were generated by incubating solutions of Cdc19 -4A and Cdc19-RE at 7.2 M at 37°C until complete conversion of soluble proteins into aggregates monitored by ThT fluorescence was reached. The presence of aggregates was further verified by optical microscopy. The supernatant containing an excess of ThT was removed from the seeds by centrifugation for 20 min at 15,000 ϫ g. The pellet was then resuspended in 200 l of the same buffer, and the proteins were homogeneously mixed by pipetting. The seeds were used both untreated and after applying sonication with a Bandelin Sonopuls HD 2070 homogenizer equipped with a MS 37 tip. The seeds were sonicated for 30 s at 50 and 80% of the instrument maximum power, alternating 10 s of pulsing with 10 s of pause. The seeds were added to 7.2 and 1.8 M Cdc19 -4A and RE solutions at 10% monomer equivalent. Control aggregation experiments for seeded assays were performed by adding the same buffer of the protein solution.

Polymeric nanoparticles
Hydrophobic and PEGylated NPs were produced and characterized as described (36), using batch and semibatch emulsion polymerization techniques. For the PEGylated NPs, 500 mg of PEG methyl ether methacrylate were dissolved in 45 ml of distilled water in a three-neck flask. The system was incubated in an oil bath equipped with a glass condenser and a thermocoupler and purged with a nitrogen flux for 20 min at room temperature under stirring. After the temperature was raised to 70°C and the radical initiator was introduced (potassium persulfate, 50 mg dissolved in 2.5 ml of distilled water), methyl methacrylate (2 g) was added with a syringe pump over 1 h, and the reaction was allowed to continue for 4 h. For hydrophobic NPs, a batch emulsion polymerization was employed mixing 75 mg of SDS and 2.5 g of hydrophobic monomer (butyl acrylate) in 45 ml of water. The solution was first flushed with nitrogen and then heated to 70°C. Potassium persulfate (30 mg dissolved in 2.5 ml of water) was used as the radical initiator. The reaction was stopped after 4 h, and the dispersions were cleaned with ion-exchange resins (Dowex Marathon M3; Sigma-Aldrich) to remove the surfactant, as well as all the possible electrolytes. Hydrophobic and hydrophilic NPs have diameters of 70 and 57 nm, respectively, as measured by dynamic light scattering (Zetasizer, Malvern, UK). The NPs were added to protein solutions at a final volume fraction of 0.8 and 0.5% for PEGylated and HB NPs, respectively.