Introduction

Intracellular droplets formed by liquid-liquid phase separation are driven by proteins with low-complexity domains (LCDs) that take the role of multivalent and transient interaction with other molecules1,2. LCDs are regions composed of a limited number of amino acids and are known to be involved in both liquid-liquid phase separation (LLPS) and liquid-to-solid phase transition3,4. Indeed, in vitro experiments have shown that by maintaining the droplets for an extended period, amyloid-like aggregates are formed within the droplets for several proteins with LCDs5,6,7,8,9. This phenomenon may underlie diseases associated with protein aggregation. Therefore, understanding the mechanism of amyloid-like aggregate formation from droplets and how it is controlled could lead to significant advances in the treatment and drug discovery for diseases related to protein aggregation. Until now, in vitro studies of droplet solutions separated into condensed and dilute phases have primarily relied on microscopy-based techniques. This preference arises because solutions containing droplets scatter light, which limits the spectroscopic analysis that can be conducted. As a result, the mechanism of amyloid nucleation inside the droplets remains largely unknown. In this context, observing the influence of other molecules on amyloid formation within droplets offers indirect but valuable insights into the underlying mechanism.

Intracellular droplets are assemblies of a wide variety of biomolecules10,11. These droplets contain many proteins with low-complexity domains12,13, and the heterotypic interactions among these proteins determine the physical properties of the droplet14. Many studies have reported the effect of heterotypic interactions on droplet properties from mixing two or more proteins with low-complexity domains capable of forming droplets. These interactions have been shown to alter droplet properties, such as droplet formation ability, droplet dynamics, stability, and maturation15,16,17,18. Several studies have also examined the effects of heterotypic interactions on amyloid formation within droplets. For instance, two amyloidogenic proteins interact inside or on the surface of droplets and accelerate progress to nucleation and co-aggregation16,19. Additionally, small molecules interact with amyloidogenic proteins and have been reported to suppress amyloid formation by promoting droplet maturation (gelation)20,21. Increasing droplet stability and preventing structural transitions to amyloid formation is one approach to suppressing amyloidogenesis, with amino acid substitutions shown to inhibit amyloid formation by stabilizing droplets22. Although few reports describe coexisting proteins that suppress amyloid fibril formation within droplets, peptide analogs interacting with amyloidogenic proteins have been found to inhibit amyloid formation under non-droplet conditions23,24. This suggests that transient and weak interactions within droplets, which help to stabilize them, may also play a role in suppressing amyloid nucleation.

The yeast prion protein Sup35 from the budding yeast Saccharomyces cerevisiae (SC) is a translation termination factor consisting of three domains (N, M, C). The C domain is necessary and sufficient for translation termination. The N and M domains (Sup35NM) are intrinsically disordered regions with unknown functions. The N domain is a prion domain known to form the core of intermolecular interactions during amyloid formation25,26,27, and LCD in which Gln, Gly, Asn, and Tyr residues make up approximately 80% of the total domain (Fig. 1). In addition, The N domain contains two amyloid initiation sites at the N- and C-terminal regions, which are regions with concentrated Asn residues26. The Gln and Asn residues in the N domain contribute to the stabilization of the amyloid core structure by forming hydrogen bonds between their side chains28. The M domain is a highly charged LC domain, where residues Lys, Glu, and Asp account for about 40% of the domain (Fig. 1). The arrangement of positive and negative charges in the M domain is biased, with more positive charges on the N-terminus and more negative charges on the C-terminus, which is favorable for intermolecular electrostatic interactions.

Fig. 1: Amino acid sequence and composition of budding yeast SC, KL, and CA-Sup35NM proteins.
figure 1

a Amino acid sequence alignment of SC, KL, and CA-Sup35 N-domain (left) and M-domain (right). b Pie charts showing the abundance ratios of major amino acids in the N-domain (left) and M-domain (right).

Full size image

Furthermore, Sup35NM is a region of rapid molecular evolution29,30, and there is a significant sequence difference between evolutionarily distant budding yeast species. This sequence variation is the molecular basis of the species barrier, which prevents cross-species infection between heterologous yeast strains. However, changes in amino acid composition are small, and the Gln, Gly, Asn, and Tyr-biased composition of the N domain, as well as the charge arrangement bias in the M domain, are conserved across species (Fig. 1). Therefore, many Sup35 proteins derived from budding yeasts retain their ability to form amyloid31. Sup35 has been shown to form intracellular droplets in yeast under conditions of carbon depletion and low pH32, and more recently, it has also been observed to form droplets under osmotic stress33. The conservation in the amino acid composition might contribute to the preservation of droplet-forming ability. In contrast, the N-domain composition of Sup35NM from fission yeast Schizosaccharomyces pombe (SP) differs significantly. The Gln, Asn, Gly, and Tyr content decreases to about 36%, though the Gln, Asn, and Tyr residues are present at approximately twice the average of typical proteins34. The ability of SP-Sup35 to form droplets under carbon depletion and low pH conditions has been observed32 and is considered within a range that retains phase separation ability despite sequence changes. In vitro studies on Sup35 droplet formation have shown that low salt concentration and weak acidic pH are optimal for SC-Sup35 and SP-Sup3532,35, and amyloid formation of SC-Sup35NM from droplets has been observed a few hours after droplet formation9. In yeast cells, Sup35NM from the budding yeast Ogataea methanolic formed amyloid-like fibrillar aggregates from droplets induced by osmotic stress, some of which have been observed to be heritable amyloids across generations33.

In this study, to investigate the effect of proteins with LCDs colocalized in droplets on amyloid formation, we used three types of Sup35 N-terminal intrinsically disordered regions (Sup35NM) derived from the budding yeasts S.cerevisiae, Kluyveromyces lactis (KL), and Candida albicans (CA) as a model. We prepared an in vitro environment in which the two Sup35NM coexisted and examined the resulting effect. Although these different yeast-derived Sup35NM proteins do not naturally coexist in cells, they possess distinct amino acid sequences that act as a species barrier, preventing interspecies amyloid infection (Fig. 1). Each of these proteins is capable of forming amyloid independently, yet they do not interfere with each other and do not cause co-aggregation, making them suitable for this study. In addition, since these Sup35NMs have similar amino acid compositions (Fig. 1), we anticipated that the intermolecular interactions during droplet formation would be similar, which led us to use this set of Sup35NM.

Our results show that when two types of Sup35NM proteins were mixed, droplets formed in which both types of Sup35NM coexisted. Amyloid formation from these mixed Sup35NM droplets was delayed compared to amyloid formation from single-Sup35NM droplets. This delay is believed to result from repeated formation of transient weak heterotypic interactions between molecules in the amyloid initiation site, suppressing the structural changes necessary for amyloid nucleation. Additionally, a similar effect was observed with SP-Sup35NM from the fission yeast, which has a significantly different amino acid composition, suggesting that various proteins utilizing similar interactions for droplet formation may also delay amyloid formation.

Results

Droplet formation of Sup35NM proteins from different yeast species

First, we measured the temperature dependence of droplet formation of the SC-Sup35NM proteins using turbidity measurements and optical microscopy in the presence of polyethylene glycol (PEG, MW: 20,000) as a crowder (Fig. 2). SC-Sup35NM exhibited upper critical solution temperature (UCST) behavior36 conducive to droplet formation at low temperatures. A sharp transition in turbidity was observed over a narrow temperature range (e.g., 30–40 °C under 20% PEG conditions) (Fig. 2a left), indicating high temperature sensitivity. To assess whether these characteristics are conserved in evolutionarily distant yeast species, we analyzed the KL-Sup35NM and CA-Sup35NM. While slight differences were observed in the onset temperatures of droplet formation, both CA-Sup35NM and KL-Sup35NM also increased turbidity at lower temperatures, similar to SC-Sup35NM, and exhibited comparable high temperature sensitivity (Fig. 2a). Optical microscopy confirmed droplet formation at 20 °C for all three proteins, whereas no droplets were observed at 60 °C, where turbidity did not increase (Fig. 2b). The droplet formation of all Sup35NM was dependent on PEG concentration, with higher PEG concentrations promoting droplet formation.

Fig. 2: Droplet formation of budding yeast SC, KL, and CA-Sup35NM proteins.
figure 2

a Temperature and PEG concentration dependence of turbidity (wavelength 500 nm) of SC, KL, and CA-Sup35NM. b Optical microscope images of droplets of SC, KL, and CA-Sup35NM under 15% PEG conditions at 20 °C (top) and 60 °C (bottom). Scale bar = 10 mm.

Full size image

For comparison, we also examined droplet formation in SP-Sup35NM. Although the N domain of SP-Sup35NM differs significantly from that of budding yeast in terms of amino acid sequence and composition(S_Fig.1), it contains twice the amount of Glu, Asn, and Tyr compared to the average protein composition34. Unlike budding yeast Sup35NM, SP-Sup35NM did not exhibit high temperature sensitivity, instead showing a monotonic change in turbidity with temperature (S_Fig. 2a). Optical microscopy confirmed the formation of spherical droplets derived from SP-Sup35NM, similar to those formed by budding yeast Sup35NMs (S_Fig. 2b). These results indicate that the ability to form droplets is even conserved in SP-Sup35NM, while the specific amino acid composition of budding yeast Sup35NM appears to be critical for conferring temperature sensitivity.

Amyloid formation from droplets of Sup35NM proteins from different yeast species

Amyloid-like aggregates have been reported to form from droplets of several proteins with LCDs, and SC-Sup35NM is one of them9. To investigate whether amyloid formation occurs in all Sup35NM droplets, cysteine residues added to the C-termini were labeled with Alexa Fluor 488, and the behavior after droplet formation was observed using fluorescence microscopy in the presence of 15% PEG (Fig. 3a top). Immediately after droplet formation, three Sup35NM variants were concentrated in spherical droplets. By day three, the droplets had completely disappeared, revealing fibrillar aggregates for SC-Sup35NM and CA-Sup35NM, and spherical aggregates with short protruding fibers for KL-Sup35NM (Fig. 3a bottom). To monitor the kinetics of droplet-to-amyloid transition at 30 °C, Thioflavin T (ThT), a dye that specifically binds to amyloid fibrils and emits fluorescence, was used. An increase in ThT fluorescence with a lag phase was observed for all three Sup35NMs (Fig. 3b). Under non-droplet-forming conditions (5% PEG), spontaneous amyloid formation exhibited a slow onset. However, it was significantly accelerated under droplet-forming conditions (15% and 20% PEG), suggesting that it promotes amyloid formation (Fig. 3b). The promotion of droplet formation even under the 10% PEG condition, where no increase in turbidity was observed at 30 °C, is likely attributable to the formation of fine droplets that are undetectable by turbidity measurements. In contrast, SP-Sup35NM did not form amyloid fibrils under either droplet-forming or non-droplet-forming conditions (S-Fig. 2c). This result supports the previous study showing that SP-Sup35NM cannot induce amyloid when overexpressed in yeast37.

Fig. 3: Liquid to solid transition of SC, KL, and CA-Sup35NM proteins.
figure 3

a Fluorescence microscopy images of Alexa Fluor-488-labeled SC, KL, and CA-Sup35NM under 15% PEG conditions, captured at 30 minutes (top) and 3 days (bottom) after droplet formation. Scale bar = 10 mm. b Amyloid formation kinetics of SC, KL, and CA-Sup35NM at 30 °C monitored by ThT fluorescence under 5, 10, 15, and 20% PEG conditions.

Full size image

To confirm the amyloid nature of these aggregates, seeding experiments were conducted (S_Fig. 3). By adding preformed amyloid fibrils as seeds to a monomer solution, the structure of the seed serves as a template and undergoes an autocatalytic rapid structural change from the monomer to amyloid. All Sup35NM variants showed a seeding effect on their respective monomer, confirming that the aggregates formed from droplets are indeed amyloid fibrils. Previous reports have not observed cross-seeding between SC-Sup35 and CA-Sup3530, and while cross-seeding between SC-Sup35 and KL-Sup35 has been observed in yeast cells31, it has not been observed in vitro38. Our experiments also showed no seeding effect for heterologous monomers, consistent with previous studies showing no cross-seeding in vitro. This lack of cross-seeding is likely due to the low sequence homology between these heterologous Sup35NM variants, which prevents their coexisting within amyloid fibrils.

Droplet formation in heterogonous mixtures of Sup35NM proteins

Cellular droplets contain a number of biomolecules10,11, and it is believed that these droplets often include many proteins with LCDs that share similar amino acid compositions12,13. The heterotypic interactions among different molecules within the droplets are thought to influence properties. To investigate the effects of other proteins with LCDs coexisting in the droplets, we mixed different Sup35NM variants and examined the effects on droplet and amyloid formation. Two Sup35NM variants were mixed in a 1:1 ratio (KL/SC and CA/SC), and droplets were induced by adding a PEG. Turbidity intensity increased due to the doubled protein concentration (Fig. 4a). The temperature and PEG concentration dependence profiles remained consistent with those observed when the variants were used individually, indicating no significant changes in droplet properties (Fig. 4a). To confirm the co-existence of the two Sup35NM variants within droplets, SC-Sup35 was labeled with Alexa Fluor 594, while KL- and CA-Sup35NM were labeled with Alexa Fluor 488 at the C-terminal cysteines. Fluorescence microscopy revealed the coexistence of SC-/KL-Sup35NM and SC-/CA-Sup35NM within the same droplets (Fig. 4b). Furthermore, the coexistence of SP-/KL-Sup35NM and SP-/CA-Sup35NM in droplets was also confirmed (S-Fig. 5a). These results suggest that, unlike in amyloid fibrils, high sequence homology is not required for coexistence in droplets.

Fig. 4: The mixture of SC/KL-Sup35NM and SC/CA-Sup35NM forms droplets where the two Sup35NM variants coexist.
figure 4

a Temperature and PEG concentration dependence of turbidity (wavelength 500 nm) of SC/KL-Sup35NM mixture (top) and SC/CA-Sup35 mixture (bottom). b Fluorescence microscope images of heterotypic Sup35NM mixture droplets under 15% PEG conditions. SC-Sup35NM was labeled with Alexa fluor-594 (SC-Sup35NM-AF488), while KL and CA-Sup35NM were labeled with Alexa fluor-488 (KL-Sup35NM-AF488, CA-Sup35NM-AF488). Scale bar = 10 mm.

Full size image
Fig. 5: Amyloid formation from mixed droplets of two Sup35NM variants is delayed.
figure 5

a Amyloid formation kinetics of SC/KL-Sup35NM (top) and SC/CA-Sup35 (bottom) mixed droplets under 20% and 15% PEG conditions. Black lines indicate the amyloid formation of single Sup35NM (5 μM), while red lines represent mixed Sup35NM (total 10 μM). b Fluorescence microscope images of amyloid fibrils formed from mixed Sup35NM droplets under 15% PEG conditions. SC-Sup35NM was labeled with Alexa fluor-594 (SC-Sup35NM-AF594), KL and CA-Sup35NM were labeled with Alexa fluor-488 (KL-Sup35NM-AF488, CA-Sup35NM-AF488).

Full size image

Amyloid formation is delayed under heterogonous mixtures of Sup35NM proteins

We next examined the effect of the heterologous Sup35NM protein mixtures on spontaneous amyloid formation within droplets. Using ThT fluorescence assays, we compared the amyloid formation rates in mixed protein droplets with corresponding single-protein droplets. Our results showed that amyloid formation by heterogonous mixtures was delayed compared to the single-Sup35NM under the droplet-forming conditions (15%-20% PEG), with the delay becoming core pronounces at higher PEG concentrations (Fig. 5a, S-Fig. 4). In many cases, amyloid formation occurred as a one-step reaction, wherein both Sup35NM molecules simultaneously transitioned into the amyloid state. Especially in SC/CA mixture, the slope of the fiber extension in mixed droplets was gentler than observed for single-protein droplets. This may be due to competitive inhibition at fiber ends caused by the binding of different Sup35NM variants. Conversely, under non-droplet-forming conditions (5% PEG), amyloid formation occurred more rapidly in the mixed condition than in single-protein (S-Fig. 4). It is likely that increased total protein concentration resulting from the mixing of proteins may have led to the formation of turbidity undetectable fine droplets in the solution, which could have promoted amyloid formation. These results suggest that the delay in amyloid formation of Sup35NM occurs specifically within the droplets. Furthermore, a similar delay was observed in the droplets that coexisted with the SP-Sup35NM (S_Fig. 5b). To further explore the specificity of this effect, we mixed KL- or CA-Sup35NM with an equal amount (mg/mL) of bovine serum albumin (BSA) as a negative control (S_fig. 6). BSA, while capable of forming droplets in vitro39, lacks LCDs. No delay in amyloid formation was observed under 15% PEG, indicating that the delay is specific to the mixture between Sup35NM variants. Fluorescence microscopy of amyloid fibrils formed in mixed droplets revealed distinct patterns (Fig. 5b). Under the SC/KL mixed condition, two-wavelength fluorescence signals overlapped only at the droplet remnants, and no overlap was observed in the fibrils themselves. This indicates that each amyloid fibril formed independently. In contrast, under the SC/CA mixed condition, two-wavelength fluorescence signals overlapped across all amyloid fibrils. However, as no cross-seeding was observed in the SC-/CA-Sup35NM seeding experiments (S_Fig. 3), this suggests that independently formed protofibrils became entangled. Given multiple possible amyloid structures26,27,40, we cannot rule out the possibility that a specific structure allowing co-existence of the two different Sup35NM has been acquired.

Discussion

In this study, we used three Sup35NM proteins with different amino acid sequences derived from evolutionarily distant budding yeast species and confirmed that their droplet-forming abilities are conserved. Furthermore, we observed that these Sup35NM proteins could coexist in crowder-induced droplets and their coexistence delays amyloid formation. These Sup35NM proteins have no significant changes in their amino acid composition and are thought to maintain the droplet state through shared molecular interactions, including π-π interactions involving Tyr residues in the N domain, electrostatic interactions mediated by Lys, Glu, and Asp residues in the M domain, and cation-π interactions between Tyr and Lys residues. These conserved interactions enable the three Sup35NM proteins to colocalize within the same droplet. However, since amyloid co-aggregation requires high sequence homology, the amyloid fibrils formed in these coexisting droplets are independent. Based on these findings, we propose that proteins with LCDs that cannot co-aggregate but share similar droplet-forming interactions can coexist within droplets, thereby delaying amyloid formation.

Additionally, SP-Sup35NM derived from fission yeast demonstrated similar effects, coexisting with droplets and delaying amyloid formation. Although the M domain of SP-Sup35 retains a highly charged, biased composition, similar to that of budding yeast M domains, the amino acid composition of the N domain differs significantly. However, the SP-Sup35 N domain still has the number of Gln, Asn, and Tyr residues, which is approximately twice the average found in general proteins, suggesting that it utilizes interactions analogous to those of budding yeast Sup35NM to form droplets. These results indicate that the amino acid composition necessary for the amyloid-suppression effect does not require high similarity. Instead, any protein with LCDs capable of forming droplets via similar interactions may exhibit this effect. Consequently, in-cell droplets containing multiple proteins with LCDs may harbor several proteins that contribute to amyloid suppression.

In our experiments, droplets were formed by adding a crowder, PEG (mw 20,000), and amyloid formation from these droplets was subsequently investigated. Molecular crowding is known to have a significant effect on both droplet formation and amyloid formation41. Under crowded conditions, the excluded volume effect promotes intermolecular interactions, facilitating both droplet and amyloid formation. First, the attractive forces arising from the excluded volume effect facilitate protein-protein interactions, forming droplets, thereby reducing the volume occupied by individual proteins. This effect promotes the formation of amyloids, which occupy a smaller volume per protein. However, amyloid nucleation becomes a rate-limiting step, leading to a time lag. Once nucleation occurs, the transition from droplets to amyloids proceeds rapidly.

In contrast, proteins concentrated within droplets due to the excluded volume effect exhibit increased viscosity, which restricts protein movement. Since amyloid nucleation is a process involving multiple molecules, a reduction in molecular collisions within droplets likely suppresses amyloid formation. In this study, the concentration of the Sup35NM mixture was twice that of the single-protein solutions. The increased concentration within droplets may have further elevated viscosity, potentially delaying amyloid formation. A control experiment using BSA, a protein lacking an LCD, showed no delay in amyloid formation. However, the mixture of two Sup35NM proteins, both intrinsically disordered, may have resulted in higher viscosity compared to the mixture of Sup35NM and BSA, a globular protein. Therefore, it cannot currently be excluded that part of the observed suppression of amyloid formation may be attributable to solvent conditions rather than solely to protein-protein interactions.

Methods

Expression and purification of Sup35NM protein

For bacterial expression of SC, KL, CA, and SP-Sup35NM, pET29b vectors, including a C-terminal 7x histidine-tag. were used. Sup35NM was overexpressed in the bacterial strain BL21(DE3) and purified by nickel-nitrilotriacetic acid histidine-tag affinity chromatography and cation-exchange chromatography under denaturing conditions (8 M Urea). For purification of cysteine mutants, 1 mM dithiothreitol was added to the buffer.

Turbidity measurement

Turbidity measurements were conducted using an EPOCK2 absorbance plate reader (BioTek) at a wavelength of 500 nm. This wavelength was selected to minimize deviations from the linear relationship between droplet volume and absorbance caused by protein absorption or excessively high absorbance values. Measurements were performed in a single direction, from high to low temperatures. An 18-minute equilibration period was set prior to each measurement to ensure temperature stabilization. The samples for measurement were prepared on a heat block set to 70 °C and subsequently transferred to a preheated plate (70 °C). Measurements started immediately, without allowing the sample to cool at any point. For single-protein measurements, 5 µM Sup35NM was prepared in a 10 mM sodium phosphate buffer (pH 7.0) containing 5–20% (w/v) polyethylene glycol (PEG). For mixed protein measurements, 5 µM Sup35NM was mixed with an additional 5 µM Sup35NM variants (final concentration: 10 µM) under the same buffer and PEG conditions.

Amyloid formation kinetics monitored by ThT fluorescence

The kinetics of Sup35NM amyloid formation were monitored using thioflavin T (ThT) fluorescence42. Spontaneous amyloid formation was initiated in 10 mM sodium phosphate buffer (pH 7.0) containing 5–20% (w/v) polyethylene glycol (PEG; average Mw: 20,000) and 2 µM ThT under quiescent conditions. For single-protein measurements, 5 µM Sup35NM was used. For mixed protein measurements, 5 µM Sup35NM was mixed with an additional 5 µM Sup35NM variant, resulting in a final concentration of 10 µM. The kinetics were recorded using fluorescence plate readers (SH-9000Lab, CORONA ELECTRIC, and Infinite 200, TECAN) with ThT as a fluorescent probe. ThT fluorescence was excited at 445 nm, and emission was detected at 485 nm. In the seeding experiment, amyloid fibrils spontaneously polymerized under 15% (w/v) PEG were used as seeds, which were pre-treated by sonication to fragmentation. The seed was added to the monomer solution at a concentration corresponding to 5% of the total protein amount. Subsequent measurements were conducted under the same conditions.

Microscopic measurements

Optical microscope observations were performed using an Eclipse (Nikon). Sample solutions were placed on concave glass slides and sandwiched between cover glasses. Observations were conducted using a 40–100× objective lens, and the temperature was controlled with a Peltier stage. Fluorescence microscopy observations were carried out using a BZ-X (KEYENCE) and an IX71 (OLYMPUS). Sample solutions were placed in 96-well glass plates for observation. Fluorescent labeling was achieved by covalently modifying proteins containing a cysteine residue introduced at the C-terminus with Alexa Fluor maleimide dyes (488 or 594). Labeled proteins were mixed at a 1:9 ratio with unlabeled proteins for fluorescence observations.