Reversible Disulfide Bond Cross-Links as Tunable Levers of Phase Separation in Designer Biomolecular Condensates

Biomolecular condensates (BCs) are membraneless hubs enriched with proteins and nucleic acids that have emerged as important players in many cellular functions. Uncovering the sequence determinants of proteins for phase separation is essential in understanding the biophysical and biochemical properties of BCs. Despite significant discoveries in the past decade, the role of cysteine residues in BC formation and dissolution has remained unknown. Here, to uncover the involvement of disulfide cross-links and their redox sensitivity in BCs, we designed a “stickers and spacers” model of phase-separating peptides interspersed with cysteines. Through biophysical investigations, we learned that cysteines promote liquid–liquid phase separation in oxidizing conditions and perpetuate liquid condensates through disulfide cross-links, which can be reversibly tuned with redox chemistry. By varying the composition of cysteines, subtle but distinct changes in the viscoelastic behavior of the condensates were observed. Empirically, we conclude that cysteines function neither as stickers nor spacers but as covalent nodes to lower the effective concentrations for sticker interactions and inhibit system-spanning percolation networks. Together, we unmask the possible role of cysteines in the formation of biomolecular condensates and their potential use as tunable covalent cross-linkers in developing redox-sensitive viscoelastic materials.


■ INTRODUCTION
−6 BCs reversibly achieve need-based spatiotemporal organization and control of cellular matter in an energy-independent manner. 7−10 BC formation is a density transition wherein a denser phase (or phases) enriched in biomolecules coexists with a biomolecule-deplete dilute phase above a threshold saturation concentration (C sat ). 9 The coacervation of biomolecules toward such a density transition is better captured by a phenomenon called liquid−liquid phase separation (LLPS), although many phase separation mechanisms are known. 11wo types of coacervations predominate BCs.Self-coacervation is unimolecular, involving polypeptides undergoing LLPS by themselves in specific ionic strengths, while complexcoacervation involves biomolecular scaffolds that partition other interacting partners called clients within the condensates. 12Irrespective of the coacervation type, the BCs' interactions involve weak, nonstoichiometric multivalent interactions, including cation−π, π−π, van der Waals, and hydrogen bonding. 12−14 LLPS, among associate polymers such as proteins, is best defined by a "stickers and spacers" model wherein stickers are amino acid residues that are involved in multivalent, weak, noncovalent interactions with one another, and spacers are disordered sequences containing noninteracting residues that spatially separate the stickers. 12,13The balance of the interaction strength between the stickers and the spacers' effective solvation volume determines the viscoelastic properties of the BCs formed. 15−22 Associative biopolymers such as proteins, either by self-coacervation or by complex-coacervation with clients such as RNA, form BCs best explained by the process of LLPS. 9,23,24−27 However, understanding the design and properties of redox-sensitive condensates remains limited. 18,28,29Only a handful of studies have investigated redox sensitivity in peptides and proteins based on the oxidation of methionine to sulfoxides and sulphones. 28,30The dearth of information is especially apparent in the use of cysteines as redox-modulating residues within the peptide sequence.−34 Therefore, it may seem counterintuitive to see the presence of cysteines among disorder-promoting amino acids such as arginine, glycine, serine, etc.However, nature seems to have accommodated disorder-promoting sequences interspersed with cysteine, especially in complex higher-order organisms such as eukaryotes, which are often involved in reactions with redox flux. 34Yet, the role of cysteines in the formation and dissolution of BCs has remained unknown.Previously, we demonstrated that cysteine-rich protein modules called granulins (GRNs) modulate LLPS of TAR-DNA binding protein (TDP-43) by tuning the redox state of cysteine, 35,36 suggesting that cysteines could play a role in coacervation and LLPS.Recently, cystamine-linked peptide synthons were used to demonstrate the effectiveness of redoxsensitive phase-separating peptides, 27 which further elucidates the potential of cysteines to be used in designer redox-sensitive biomolecular condensates.
Here, we set out to understand the following important questions: (i) Do covalent cross-links formed by disulfide bonds facilitate LLPS? (ii) Do disulfide bonds function as covalent stickers, spacers, or to generate extended networks?(iii) Do BCs formed show sensitivity to redox flux?By designing short peptides containing classical stickers and spacers interspersed with cysteines varying in positions and compositions, we learned that cysteine disulfide bonds promote LLPS reversibly, providing intriguing new clues in understanding the dynamics of BC formation and dissolution.

Design of Model Phase-Separating Peptides (PSPs).
We set out to answer these questions by designing simple peptide models that recapitulate phase-separating characteristics in proteins.39 Using PSP-1 as the basis, cysteine (C) residues were introduced, systematically varying their position and composition (Figure 1).Two cysteines were substituted for serines at the N-and C-terminal ends in PSP-2, while two cysteines replaced two glycine residues in the middle in PSP-3.PSP-4 and PSP-5 were single cysteine variants of PSP-2 and PSP-3 in which the N-terminal cysteine in PSP-2 was retained in PSP-4, and the cysteine in the middle of the spacer from PSP-3 was retained in PSP-5 (Figure 1).Together, the five peptides provided a minimal set of variations to investigate the contributions of cysteines in LLPS.All peptides were synthesized as C-terminal amides (see Methods for details), and target peptides were confirmed by electrospray ionization (ESI) mass spectrometry (Figure S1).
Cysteines Lower the Saturation Concentrations for LLPS.The phase transition of a polypeptide from a homogeneous to a demixed solution containing two or more coexisting liquid-like phases occurs above a concentration threshold called the saturation concentration (C sat ).We first established phase boundaries and determined apparent C sat values for the self-coacervates by investigating the concentration, pH, and ionic strength landscape through turbidimetry analysis and confocal microscopy (Figure 2).PSP-1, the control phase-separating peptide without cysteine functionality, was assessed up to 100 mM between pH 7.0 and 12.0 and sodium chloride (NaCl) concentrations of 0−3.5 M (Figure 2a,b).The peptide showed that LLPS was above 80 mM in 3.5 M salt concentrations at pH 8.0 upon incubating at room temperature for at least 3 hours and showed a C sat value of 80 mM in 2.5 M NaCl (Figure 2b).PSP-2, within 30 min of incubation, showed a dramatic decrease in the phase boundary with LLPS occurring above 2.0 mM at pH 8.0, with an apparent C sat value of ∼2.0 mM in 2.5 mM NaCl (Figure 2c,d).PSP-3 also showed a similar phase boundary in pH scans (Figure 2e) and a slightly altered boundary in ionic strength scans with a C sat value of ∼1.2 mM in 2.5 M NaCl (Figure 2f).PSP-4 showed a phase boundary above 3.2 mM at pH8 (Figure 2g), with a C sat value of ∼3.2 mM in 2.5 M NaCl (Figure 2h).PSP-5 demonstrated no evidence of phase separation in the pH scans (Figure 2i) or ionic strength scans at or below 3.5 mM peptide concentration (Figure 2j).The peptide showed phase separation only above a C sat value of ∼16.0 mM in 2.5 M NaCl with 1% H 2 O 2 (Figure 2j).The respective differential interference contrast (DIC) and confocal fluorescence microscope images in specific buffer conditions (dotted circles in phase diagrams) for peptides PSP-1 (Figure 2k,l), PSP-2 (Figure 2m,n), PSP-3 (Figure 2o,p), PSP-4 (Figure 2q,r), and PSP-5 (Figure 2s,t) show the formed droplets along with the turbidity of the phase-separated solution (insets).Fluorescence recovery after photobleaching (FRAP) analysis on PSP-1 droplets showed a somewhat diminished 60% recovery, reflecting less fluidity and high viscoelasticity of the droplets likely indicated by prolonged time (3 h) taken for the droplets to form (Figure 2u).Droplet fluidity of other peptide coacervates analyzed by FRAP showed ∼80% recovery for PSP-2−5, more fluidity than PSP-1 (Figure 2u−y).Together, these results indicate that cysteines could play an important role in promoting condensate formation, as evidenced by the decreased C sat values of the peptides to phase separate by selfcoacervation.

Capping of Thiols Prevents Condensate Formation.
To uncover the precise role of cysteine thiols in condensate formation, we investigated the ability of peptides to phase separate when thiol groups are rendered incapable of forming disulfide bonds.To do so, iodoacetamide (IA) was used to cap thiols covalently in phase-separating conditions with nonphase-separating conditions as controls (Figure 3a). 36The amount of IA was also varied (0.5 molar equivalence or excess) such that the cysteines were capped partially (one of the two cysteines) or completely (both cysteines).In all cases, the peptides were first capped with IA in non-phase-separating buffer conditions (high concentrations without salt) and then were either diluted to non-LLPS (low concentrations, no salt) or LLPS conditions (>C sat ; 2.5 mM NaCl).The samples were then investigated by matrix-assisted laser desorption ionization-time-of-flight (MALDI-ToF) mass spectrometry and confocal fluorescence microscopy using 1% of FITC-tagged peptides in the samples.PSP-2 and PSP-3, which contain two cysteine residues, under non-phase-separating conditions, showed heterogeneous mixtures when capped partially with IA, as confirmed by MALDI-ToF spectra (first panel; Figure 3b,c).In excess conditions, PSP-2 displayed both singly-and doubly-capped peptides, while PSP-3 showed only a doublycapped peptide (second panel; Figure 3b,c).Images of the samples in a confocal microscope showed no droplets under these conditions as expected (insets in the top two panels; Figure 3b,c).In LLPS conditions, when partially capped, mass spectra of PSP-2 showed the presence of uncapped and singly capped peptides (third panel; Figure 3b).Under these conditions, the peptides showed condensate formation (inset in the third panel; Figure 3b).However, under fully-capped and phase-separating conditions containing singly-and doubly- capped populations, the peptide failed to form condensates (fourth panel; Figure 3b).Similarly, under partial capping conditions, PSP-3 showed the presence of a heterogeneous mixture of uncapped, singly-capped, and doubly-capped peptides, which formed condensates (third panel; Figure 3c).However, when fully-capped with excess IA, PSP-3 failed to phase separate (fourth panel; Figure 3c), which confirms that some populations of disulfide-bonded peptides are required for LLPS.PSP-4 and PSP-5, which contain single cysteines, showed no phase separation under non-LLPS conditions regardless of partial or full-capping (first and second panels; Figure 3d,e).When partially capped under LLPS conditions, both peptides showed condensate formation (third panels; Figure 3d,e), but when fully-capped, they failed to form droplets (fourth panels; Figure 3d,e).From these data, one can infer that thiol oxidation to disulfide cross-links is crucial in driving the observed LLPS and forming the condensates.
Condensates with Disulfide Cross-Links Are Reversible under Redox Gradients.If disulfide bonds are crucial for condensate formation, we questioned whether the condensates can reversibly form and dissolve under a redox flux.To investigate this, peptide condensates were generated in respective phase separation conditions (3.5 mM peptides, 2.5 M NaCl, and pH 8.0 for PSP-2 and PSP-3; 3.5 mM peptide, 2.5 M NaCl, and pH 8.0 for PSP-4, and 18 mM peptide, 3.0 M NaCl, 1% H 2 O 2 , and pH 8.0 for PSP-5).All peptides showed turbidity due to droplet formation observed by confocal microscopy (Figure 4, oxidized panels).The condensates were then titrated with increasing concentrations of dithiothreitol (DTT) as a reducing agent.Even low concentrations of DTT significantly reduced the number of droplets of PSP-2 condensates formed under oxidizing conditions (Figure 4a).Increasing the concentration of DTT to 3 and 5 mM completely dissolved the condensates almost instantaneously (Figure 4a), further affirming the significance of disulfide bonds in condensate formation.To see whether dissolved condensates can be assembled by changing the reducing environment to an oxidized one, 2% hydrogen peroxide was added to the same reaction mixture, and imaged under a confocal microscope.PSP-2 droplets reappeared in the solution (2% H 2 O 2 ; Figure 4a), with the solution turning turbid within 10 min (inset within Figure 4a; 2% H 2 O 2 ).FRAP analysis of the droplets before adding DTT and after reoxidation showed nearly identical recovery kinetics, suggesting that the reformed droplets had similar viscosity and liquidlike character (Figure 4b).Nearly identical behavior was also observed with the peptides PSP-3, PSP-4, and PSP-5 (Figure 4c−h).As expected, PSP-1, the peptide lacking cysteines, did not show a change in the droplet morphology upon the addition of DTT (Figure S2).It has to be borne in mind that the presence of cysteines without stickers and spacers is not enough to promote phase separation, as we have previously shown in other cysteine-rich peptides. 35,36The results here unequivocally establish that disulfide bonds facilitate the formation of peptide self-coacervates that can be controlled by redox gradients.
Condensate Fluidity and Viscosity Subtly Vary with the Positional Variance of Disulfide Cross-Links.To fully understand whether the peptide condensates form systemspanning networks, interactions that are key for viscoelasticity, 24,40 we monitored and measured the fluidity, morphology, and viscosity changes of the peptide condensates for 10 days at room temperature.Immediately after formation, the droplets of PSP-1 showed somewhat mitigated FRAP recovery, suggesting that they are highly percolation-prone, as mentioned above (Figure 2u).After 1 day, PSP-1 droplets showed significant coalescence, which further increased in 3 days (Figure 5a).Percolation was confirmed by the gel-like characteristic of the droplets formed (Figure 5b).PSP-1 also showed percolation above 250 mM without salt immediately after incubation (Figure 5c).This behavior is called percolation without phase separation. 11Nevertheless, after 3 days of incubation, the samples showed negligible FRAP recovery (Figure 5d), suggesting that the droplets had crossed the percolation threshold to form system-spanning networks as expected for a model condensate.PSP-2 formed droplets with a surface area of ∼1−5 μm 2 on the first day of incubation and, over 10 days, showed an average increase in the surface area to 30−40 μm 2 (Figure 5e,m).This increase can be attributed mainly to the coalescence of the droplets, but importantly, the Confocal microscopy images of FITC-labeled peptide self-coacervates in their respective LLPS conditions were titrated with DTT to reduce peptides and then reoxidized with hydrogen peroxide, followed by subsequent FRAP analysis.Confocal imaging with turbidity insets and FRAP analysis of the peptides before reduction (indicated as "before" in FRAP analysis), after reduction, and after reoxidation (indicated as "after" in FRAP analysis) for PSP-2 (a and b), PSP-3 (c and d), PSP-4 (e and f), and PSP-5 (g and h).Insets show the images of Eppendorf tubes with samples in respective buffer conditions (number of independent repeats = 3, scale bar = 20 μm).
droplet morphology was maintained during the incubation period.The condensate viscosity was measured by FRAP analysis, which showed a rather consistent fluorescence recovery for 10 days, suggesting a liquid-like behavior (Figure 5i).During this time, the parameters, such as the rate constant for first-order exponential recovery (k), remained at ∼0.06 s −1 , while the time taken to reach half recovery (t 0.5 ) showed an insignificant increase from 18 to 21 s (Figure 5q,r).These data were consistent with the percentage recovery that decreased from 87 to 65% in the 10-day period (Figure 5s).PSP-3 and PSP-4 showed a droplet surface area of ∼10 μm 2 that remained unchanged for 10 days (Figure 5f, g, n, and o).The viscosity of the condensates showed a minimal change by FRAP (Figure 5j,k).While the rate constant, k, for PSP-3 decreased from 0.07 to 0.04 s −1 , a decrease from 0.09 to 0.06 s −1 was observed for PSP-4 (Figure 5q).Similarly, the t 0.5 values averaged between 21 and 18 s for PSP-3 and PSP-4, respectively (Figure 5r).The percentage of FRAP recovery decreased from 80 to 70% for PSP-3 and decreased from 90 to 75% for PSP-4 within the 10day incubation period (Figure 5s).PSP-5 condensates, on the other hand, showed distinct differences from the others.It showed the largest increase in the droplet surface area from 5 to 100 μm 2 in 10 days (Figure 5h,p), indicating that these condensates have greater fluidity based on coalescence alone.However, while k and t 0.5 did not show significant changes that averaged at 0.06 s −1 and 24 s, respectively, the percentage recovery decreased from 84 to 59% (Figure 5q−s).Interestingly, the largest decrease in FRAP recovery was observed between 8 and 10 days, suggesting that the PSP-5 condensates may have started to form percolation networks or, in other words, become more viscoelastic (Figure 5s).It is important to note that since all the peptides were air oxidized before the start of these reactions, which occurs within 30 min of resuspension in buffer, the changes in FRAP recovery solely reflect the viscoelasticity and percolation changes of the droplets.These data reveal that the positional and compositional variance of cysteines in the peptides subtly alter the fluidity and gelation of the condensates but largely remain unchanged, except for PSP-5.In other words, the disulfide bond cross-links inhibit "gelation" and help maintain the fluidity of the condensates.This point is further elaborated on in the Conclusions section.
Peptides Show Conformational Differences in Phase-Separating and Homogeneous Buffer Conditions.To see whether the peptides adopt different conformations within condensates and in bulk solution, we investigated these by 1 H nuclear magnetic resonance (NMR) spectroscopy and far-UV circular dichroism (CD).In all our experiments, the NMR spectra were collected from peptide samples in LLPS and non-LLPS conditions without separating the dense and dilute phases.Therefore, the spectra will contain signals from both phases depending on their partition function.PSP-1, the control peptide in non-LLPS conditions, showed the expected aromatic chemical shifts at 7.0 and 7.3 ppm from the tyrosine residues and backbone amide protons between 7.6 and 8.7 ppm (black; Figure 6a).Chemical shifts of other protons were evident between 3.0 and 4.2 ppm (Figure 6b) and between 1.6 and 2.1 ppm (Figure 6c).PSP-2 showed significant changes in the chemical shifts in all protons under LLPS and non-LLPS conditions (red; Figure 6 a−c).Similarly, the PSP-3 (orange) and PSP-4 (green) peptides also showed changes in 1 H chemical shifts between LLPS and non-LLPS conditions (Figure 6a−c).PSP-5, under non-LLPS conditions, showed large differences in chemical shifts in the aliphatic region (1.5− 2.2 ppm) compared with other peptides in similar conditions (Figure 6c), possibly suggesting a different conformation.More interestingly, PSP-5 showed greater differences in the amide region (7.8−8.6 ppm) between LLPS and non-LLPS conditions than in other regions (blue; Figure 6a−c), indicating that conformation of the peptide within the condensates is significantly different from that in the dilute phase.Together, the data indicate that all the peptides adopted different conformations within the condensate and in dilute phases.The NMR results were also supported by far-UV CD, which provided time-averaged conformation of the peptide.All peptides showed a positive maximum at 232 nm, indicating a turn conformation (Figure 6d−g).In addition, PSP-2 showed a negative minimum at 218 nm, indicating a β-sheet conformation in LLPS conditions that was absent in non-LLPS conditions (pink and blue; Figure 6d).Reduction of the LLPS sample with DTT led to the loss of β-sheet, but the turn conformation remained (black; Figure 6d).PSP-3 did not show an appreciable deviation from the turn conformation in LLPS, non-LLPS, and reducing conditions (Figure 6e).PSP-4 demonstrated a change from turn to β-sheet only under reducing conditions (Figure 6f).PSP-5 showed only a turn conformation in non-LLPS conditions, but a β-sheet conformation alongside turns in LLPS conditions (Figure 6g).This conformation change was reflected in the backbone amide shifts in NMR.Under reducing conditions, the peptide showed a partially disordered conformation (black; Figure 6g).Together, the data indicate that peptides adopt distinct conformations within and outside the condensates.

Complex-Coacervation with RNA Markedly Decreases the C sat values and Renders the Condensates
Insensitive to Redox Flux.In cells, BCs, such as stress granules, p-bodies, and others, are formed primarily via complex-coacervation involving several proteins and RNA molecules. 41,42To assess the propensity of the designed peptides to undergo complex-coacervation with RNA, we sought to establish a phase diagram by scanning LLPS conditions across a concentration landscape of peptides and poly-A RNA under both an oxidizing and reducing environment (Figure 7).Surprisingly, PSP-1, the control peptide lacking cysteines, showed condensates in the presence of RNA with a low C sat of 75 μM in 250 μg/mL of RNA under oxidizing conditions (−DTT; Figure 7a), three orders of magnitude reduction from the observed self-coacervation C sat of 80 mM observed in the absence of RNA.As expected, under reducing conditions, the peptide did not show any change in the phase boundary since PSP-1 is devoid of cysteine residues (+DTT; Figure 7a).PSP-2-RNA condensates also showed a C sat value between 50 and 75 μM in 250 μg/mL of RNA in oxidizing conditions, which is almost three orders of magnitude lower than that for self-coacervation (2.0 mM) (−DTT; Figures 7b and S3a,b).More interestingly, the phase boundary did not change in fully reducing conditions, suggesting that the condensates are insensitive to disulfide bond cross-links (+DTT; Figure 7b).An identical phase boundary and redox insensitivity were also observed for PSP-3, PSP-4, and PSP-5 with respective C sat values between 50 and 75 μM in 250 μg/mL of RNA in oxidizing conditions (− and + DTT; Figure 7c−e).In oxidizing conditions, PSP-1 samples, in the condition indicated in a dotted circle in Figure 7a (250 μM with 200 μg/mL RNA), showed turbidity and numerous small droplets in the range of 0.5−3 μm diameter (Figure 7f).Nearly identical behavior was observed with peptides PSP-2, PSP-3, PSP-4, and PSP-5 under similar conditions (Figure 7g−j).PSP-1 showed somewhat attenuated FRAP recovery (∼60%), potentially revealing increased viscosity within the condensates (Figure 7k).Droplets of RNA coacervates with other peptides showed FRAP recoveries between 60 and 80% (Figure 7l−o).The data revealed that the peptides efficiently undergo complex-coacervation with RNA.Furthermore, significantly smaller droplets were observed with complex-coacervates than self-coacervates of the peptides (Figure S3c).Among the peptides, PSP-4 and PSP-5 showed the largest, and PSP-2 and PSP-3 showed more modest droplet size differences (Figure S3c).Surprisingly, complex-coacervation with RNA renders the condensates insensitive to redox flux, likely due to multivalent electrostatic interactions overcompensating the effects of disulfide cross-links.This point is discussed in further detail below.
Disulfide Cross-Linked Condensates Can Be Compartments for Molecular Cargo in Peptide Self-Coacervates.Given the tunability of the condensate properties based on disulfide cross-links, we questioned whether the condensates could host molecular cargo limited by their pore size and electrostatic charges on the cargo.To do so, we first evaluated the pore size of the condensates formed by peptides containing different disulfide bond cross-links.Using fluorescently labeled dextran with varying sizes, the pore sizes were measured by their ability to partition 40 (Figure S4).Based on this analysis, we determined that all peptide condensates are porous with pore sizes larger than 100 nm (Figure S4).To test whether partitioning is limited by molecular charge, fluorescein and tetramethylrhodamine methyl ester (TMR-OMe) dyes, which have similar sizes but with negative and positive charges at pH 8.0, respectively, were used.We measured the "encapsulation efficiency (EE)" by a colorimetric assay schematically shown in Figure 8a (see Methods).Fluorescein and TMR-OMe were partitioned within the droplets of PSP-2, PSP-3, PSP-4, and PSP-5 and visualized on a confocal microscope (Figure 8b).UV−visible spectrometry was utilized to quantify the amount of dye in the supernatant (and therefore by deduction of the amount within the droplets) (Figure 8c,d).The amount of dye in the dense phase was then calculated indirectly from the total amount.The dilute phase in all peptide condensates showed the significant presence of fluorescein (Figure 8c).In contrast, TMR-OMe showed low absorbance (Figure 8d), indicating that substantially more TMR-OMe was accommodated within the droplets than fluorescein.The calculated EE suggested that all peptides accommodated TMR-OMe at least 3-fold better than fluorescein (Figure 8e).While PSP-2, PSP-3, and PSP-5 condensates accommodated ∼90−95% of TMR-OMe, PSP-4 accommodated only 80% (Figure 8e).In contrast, all peptides accommodated only 18−35% fluorescein (Figure 8e).Since all peptides are highly porous (>100 nm; Figure S4), partitioning is not limited by the size of the two dyes but by the molecular charge on the cargo.The peptide prefers positively charged TMR-OMe over negatively charged fluorescein.Fluorescein seems to be less preferred within the droplets, yet the peptides complex-coacervate with negatively charged RNA.This is likely due to the compatibility of positively charged TMR-OMe for cation−π interactions, especially at substoichiometric quantities.

■ CONCLUSIONS
The results presented here provide several novel conclusions on the role of cysteine disulfide bonds in the formation and viscoelasticity of biomolecular condensates, which has remained unknown thus far.First, disulfide bonds formed by the cysteines interspersed within a canonical sticker spacer framework of peptide promote condensate formation by decreasing the peptides' saturation concentrations (C sat ).Second, the redox chemistry of thiols undergoing disulfide bonds controls the reversibility of the condensates formed.The empirical observations presented here raise an important question: are cysteines covalent stickers or auxiliary spacers in the stickers and spacers model of peptide condensates?The answer to this question is far from trivial.If disulfide bonds are stickers, then the strong covalent bonds violate the prerequisite of multivalent weak interactions for condensate formation.On the other hand, if the cysteines are present within the spacers, they are likely to affect the effective solvation volume, which is also key for LLPS. 15However, we believe that cysteines function neither as stickers nor spacers but as nodes for crosslinks that enhance intermolecular sticker−sticker interactions by decreasing the effective concentrations.They also help maintain liquid-like characteristics for a prolonged period of time.This can be reconciled from the four categories of peptides that can be classified I−IV based on their cysteine compositions (Figure 9).Classes I and II can form an extended network of covalently bonded disulfide cross-links at high peptide concentrations in addition to some dimers in an oxidized state.Classes III and IV can exclusively form dimers. Therefore, class I and II peptides (PSP-2 and PSP-3), with their two cysteines on the termini and interior forming disulfide networks (Figure 9), render the stickers interactions to transition to predominantly intramolecular from intermolecular in the reduced state.The formation of the disulfide network significantly decreases the effective concentrations of the sticker interactions, which is manifested in the reduction of C sat , as seen in Figure 2. As expected, the disulfide networkforming PSP-2 and PSP-3 peptides have C sat values lower than those of the dimer-forming PSP-4 and PSP-5 peptides.
Although not substantial, we observe specific changes in varying the position or composition of cysteines within the peptides.PSP-5 (class IV) showed noticeable deviations from other peptides in greater viscoelasticity and higher C sat .By comparing class I (PSP-2) and class III (PSP-4), it is evident that even a single terminal cysteine near the stickers is adequate to replicate the effects induced by two terminal ones (Figure 9).However, a comparison of class II (PSP-3) and class IV (PSP-5) suggests that the elimination of even one cysteine from the spacer renders the peptide incapable of forming sufficient disulfide cross-links to lower C sat and efficiently inhibit the percolation network over time (Figures 5 and 9).In other words, when present within the spacer regions, more than one cysteine is required to bring about the effect a single cysteine near the stickers.This, in turn, suggests that disulfide cross-links near the stickers could be more important in facilitating the sticker interactions within the disulfide bond networks.This effect is also evident when class III and class IV peptides are compared, which contain single cysteines that cannot form intermolecular disulfide bond networks.Yet, the class III peptide's (PSP-4) ability to undergo phase separation at a lower C sat value than class IV's (PSP-5) further confirms the significance of the disulfide bonds near the stickers.It is also likely that the terminal cysteine provides a more extended peptide conformation to facilitate better sticker interactions, a possibility supported by the 1 H NMR chemical shifts and CD (Figure 6).
For complex-coacervates of the peptides with RNA, the disulfide cross-links significantly decrease C sat compared to peptide self-coacervates.Both under our oxidized and reduced experimental conditions, the phase boundary remains innocuous to redox flux (Figure 7).The possible underlying reason for the redox insensitivity of complex-coacervates is the following: Since multivalent electrostatic interactions dominate RNA-peptide coacervation, we conjecture that the sum of energies contributed by the electrostatic interactions between RNA backbone and peptides exceeds the net decrease in effective concentrations induced by disulfide cross-links.In addition, these interactions may also decrease the cation−π interactions between lysines and tyrosines.Furthermore, the overall negative charge on the peptide-RNA condensates may also prevent DTT from interacting with the condensates (and S−S bonding) due to negative charge repulsion between the condensates and the DTT.However, several parameters, such as RNA concentrations, ionic strength, and pH, are likely to influence these dynamics, and our ongoing experimental analysis will decipher these aspects and will be reported later.Nevertheless, from the results presented here, one can unambiguously conclude that the peptide condensates are redox-sensitive to varying degrees, especially the selfcoacervates.
The data presented here provide insights into a fundamental understanding of cysteine's role in the LLPS phenomenon, which may underlie mechanisms of several proteins in biology, especially those involved in pathologies involving LLPS.The results also showcase how cysteines can be incorporated into phase-separating model peptides and the design of customized, specific redox-tunable soft materials.
Peptide Synthesis.Peptides were synthesized on a Liberty Blue 2.0 automated peptide synthesizer (CEM) through standard 9fluorenyl methoxycarbonyl (Fmoc)-based solid phase peptide synthesis.Peptide synthesis was performed at a 0.25 mmol scale using Rink Amide ProTide Resin (0.65 mmol/g loading, 100−200 mesh).Deprotection of Fmoc protecting groups was carried out using 20% v/ v piperidine in DMF.Each amino acid addition was carried out using Fmoc-protected amino acids (0.2 M), DIC (1M), and Oxyma (1 M) in DMF.After the final Fmoc deprotection, the resin beads were washed 3x using DCM.The peptide underwent global deprotection and cleavage from the resin beads through gentle shaking in TFA/ TIS/H 2 O/EDT (92.5:2.5:2.5:2.5)cleavage cocktail for 4 h at room temperature.Peptides were then precipitated in cold diethyl ether and chilled for 4 h at −20 °C.Following this, samples were centrifuged, and the diethyl ether supernatant was decanted from the resulting peptide pellet.The peptide pellet was then resuspended in diethyl ether and chilled overnight at −20 °C.Centrifugation and decanting of the diethyl ether supernatant were performed again before allowing the peptide pellet to air-dry.Crude peptides were purified by reversephase high-performance liquid chromatography (HPLC) on a Prodigy HPLC system (CEM) with a water/acetonitrile gradient (containing 0.1% TFA).The mass and identity of the eluting fractions containing the desired PSP peptides were confirmed using electrospray ionization (ESI)-mass spectrometry (MS) on a Thermo Scientific Orbitrap Exploris 240.
Turbidity Assay.Turbidity measurements were performed on a BioTek Synergy H1 microplate reader.Samples were allowed to equilibrate at room temperature for approximately 10 min before each measurement.Phase diagrams were generated using a boundary value of 1.40 A.U. at 600 nm.Data processing was done by using Origin 8.5.Three independent data sets were collected and averaged for each measurement.
Coating Glass Slides and Coverslips.Microscopic glass slides and coverslips were cleaned with 70% ethanol by sonicating for 15 min.Coverslips and glass slides were then allowed to air-dry.Glass slides and coverslips were submerged in a coating solution (20% 613 Tween20) for 30 min.To remove the extra coating solution, the glass slides and coverslips were rinsed eight-ten times with Milli-Q water.Glass slides and coverslips were then dried at 37 °C overnight.Lens paper was used to wrap dry-coated glass slides and coverslips and was stored at room temperature until further use.
Preparation of RNA.Lyophilized powdered Poly-A RNA was acquired from Sigma and dissolved in RNase-free, sterile water.The prepared stock was stored at −80 °C until use.
Confocal Microscopy and FRAP Analysis.A Leica STELLA-RIS-DMI8 microscope at 63× magnification was used to capture confocal microscopy images of the peptide droplets on coated glass slides.In all reactions, droplets were allowed to settle for a few minutes before imaging.0.5% FITC was added to the reaction (after forming droplets) for imaging.The internal dynamics of the self-and complex condensates were investigated using fluorescence recovery after photobleaching (FRAP).For 5 s, the liquid droplets were exposed to a laser intensity of approximately 90% to achieve photobleaching.The recovery of fluorescence was then observed for 60 s.The kinetics of fluorescence recovery were normalized and plotted against time using Origin 8.5.
Image Processing and Analysis.Confocal microscopy images were processed and analyzed by using a custom pipeline implemented in Fiji (version 1.54f) and R (version 4.1.2) to quantify the size distribution of the phase-separated droplets over time.In Fiji, the images were preprocessed by applying Huang's autothresholding method 43 for binarization and by removing noise using a despeckle filter.Morphological operations (erosion and dilation) and the watershed algorithm 44 were used for droplet segmentation.The segmented droplets were analyzed, excluding those touching the edge of the frame, and their properties (area, mean intensity, perimeter, and shape descriptors) were measured and exported as CSV files.In R, the CSV files from Fiji were processed by using custom scripts.The tidy verse, 45 ggsci, and scales packages were utilized for data organization, visualization, and statistical analysis.The droplet counts and size distributions were summarized for each combination of peptide, time point, and replicates.Boxplots were generated to visualize the droplet size distributions across time points for each peptide.The image analysis pipeline assumed that the phase-separated droplets were spherical and did not account for the potential deviations from this shape.Additionally, the segmentation process may have introduced errors for closely spaced or overlapping droplets, leading to potential undercounting or inaccurate size measurements.
Dye Partitioning and Encapsulation.Peptide solutions were oxidized and diluted as described above.To each solution, 5 μM of dye was added and incubated for 10 min.Standard solutions for each equivalent dye and buffer concentration were prepared, and their UV−vis absorbances were recorded.The peptides were centrifuged at 15,000×g for 10 min.Of the 100 μL volume, 40 μL was drawn as the supernatant, and the UV−vis absorbance was measured.The encapsulation efficiency of the dye was calculated using the following equation, where A sup is the absorbance of the supernatant, and A tot is the total absorbance at 490 and 550 nm for fluorescein and TMR-OMe, respectively.Droplet Pore Size Determination.The porosities of the peptide self-coacervates were determined based on the established method from the Banerjee lab. 40Peptide condensates were formed in respective phase separation conditions with 3.5 mM peptide (except PSP-5, 18 mM) at pH 8.0 with 50 mm Tris, 2.5 M NaCl (except PSP-5, 3 M) containing 1% FITC-labeled peptide.The droplets were then incubated with tetramethyl rhodamine (TMR)-labeled dextran beads of molecular weights of 40 and 100 kDa with approximate diameters of 4.5 and 9 nm at a final concentration of 500 nM and imaged on a confocal microscope.The partitioning of dextran beads within the droplets was observed to deduce the pore size information.
Circular Dichroism (CD).CD measurements were performed on a JASCO 810 spectrophotometer.The peptide samples were added to a 0.1 mm path length cuvette and were scanned from 200 to 260 nm.A 0.1 mm cuvette was used to minimize the light scattering signals at the high concentration of peptides used to avoid linear dichroism obscuring the observed CD signals.This mixture was allowed to incubate for 5 min before each measurement.Three scans were averaged with a resolution of 1 nm.The raw data were background subtracted, smoothed with the Savitzky−Golay algorithm, normalized, and plotted by using Origin 8.0.
Nuclear Magnetic Resonance (NMR) Spectroscopy.PSP-1, PSP-2, PSP-3, and PSP-4 peptides were dissolved in 50 mM sodium phosphate (NaPi) buffer and 2.5 mM sodium chloride (NaCl), along with 10% deuterated water (D2O), and pipetted into 5 mm Wilmad NMR tubes.LLPS samples of all peptides except PSP-5 were first prepared at 3.5 mM peptide concentrations and then diluted to 1.1 mM for non-LLPS samples.The LLPS sample of PSP-5 was prepared in the same buffer at 18 mM concentration, and the non-LLPS sample was obtained by diluting it to 6 mM.The samples also included 10 mM sodium trimethylsilylpropanesulfonate (DSS) as an internal standard for chemical shift calibration.NMR measurements were performed using an 11.75 T magnet (500 MHz, 1H NMR frequency) on a Bruker spectrometer.Using Bruker default "zgpr" presaturation pulse sequence, the radio frequency carrier frequency (O1) was optimized to minimize the solvent signal.Then, we used the O1 value on the subsequent excitation sculpting pulse sequence "zgesgp" for water suppression and collected the 1 H spectra.A recycle delay time (D1) of 1 s was employed, and signals were averaged over 128 scans.All experiments were conducted at a constant temperature of 298 K. Spectra were processed by Fourier transform, and 1 H signal intensities were normalized.Chemical shifts were calibrated for the water peak at 4.696 ppm using TopSpin 3.7 (Bruker) and analyzed using customized Mathematica scripts.

Figure 1 .
Figure 1.Phase-separating peptides used in this study.All synthetic peptides contain a free amine at the N-terminus and an amide at the C-terminus.A schematic representation indicating the positions of the cysteines in the peptide is shown in parentheses (right).

Figure 2 .
Figure 2. Phase boundaries for peptide self-coacervates.Respective phase diagrams for concentration vs pH (in 2.5 M NaCl) and concentration vs salt NaCl (at pH 8.0) measured based on turbidity measured as absorbance at 600 nm (contour plot on the left) for PSP-1 (a and b), PSP-2 (c and d), PSP-3 (e and f), PSP-4 (g and h), and PSP-5 (i and j).Approximate C sat values corresponding to 2.0 M NaCl concentration are indicated in green lines.Respective bright-field and confocal images of peptides in 50 mM tris buffer and 2.5 mM NaCl, pH 8.0, in phase-separating concentrations of the peptide (indicated as dotted circles in the phase diagram) for PSP-1 (k and l), PSP-2 (m and n), PSP-3 (o and p), PSP-4 (q and r), and PSP-5 (s and t).(n = 3, scale bar = 20 μm).Images of Eppendorf tubes under phase-separating conditions are shown as insets.Corresponding FRAP recovery analyses for PSP-1 (u), PSP-2 (v), PSP-3 (w), PSP-4 (x), and PSP-5 (y).The insets show a representative region of interest used in FRAP analysis.(n = 3, scale bar in FRAP insets = 2 μm).

Figure 3 .
Figure 3. Prevention of tdisulfide bond oxidation inhibits condensate formation.(a) Schematic reaction involving iodoacetamide (IA) capping of thiols.(b−e) The peptides PSP-2, PSP-3, and PSP-4 at 9 mM and PSP-5 at 30 mM concentrations were incubated for 2 h in water at room temperature (non-LLPS conditions) either with 0.5 molar excess of IA for partial capping or with a 2-fold molar excess of IA for complete capping.The peptide stocks were then diluted to 3.0−3.5 mM (except PSP-5, which was 18 mM) either in LLPS conditions (50 mM Tris, 2.5 M NaCl, pH 8.0) or non-LLPS conditions (50 mM Tris, pH 8.0) followed by the analysis by MALDI-ToF mass spectrometry (MS) and confocal microscopy.MS spectra with partial and full-capping in non-LLPS conditions (top and second panels, respectively) and LLPS conditions (third and bottom panels, respectively) for (b) PSP-2, (c) PSP-3, (d) PSP-4, and (e) PSP-5.Insets in the panels show corresponding confocal microscopy images (scale bar = 20 mm).

Figure 4 .
Figure 4. Peptide self-coacervates are redox-sensitive and reversible.Confocal microscopy images of FITC-labeled peptide self-coacervates in their respective LLPS conditions were titrated with DTT to reduce peptides and then reoxidized with hydrogen peroxide, followed by subsequent FRAP analysis.Confocal imaging with turbidity insets and FRAP analysis of the peptides before reduction (indicated as "before" in FRAP analysis), after reduction, and after reoxidation (indicated as "after" in FRAP analysis) for PSP-2 (a and b), PSP-3 (c and d), PSP-4 (e and f), and PSP-5 (g and h).Insets show the images of Eppendorf tubes with samples in respective buffer conditions (number of independent repeats = 3, scale bar = 20 μm).

Figure 6 .
Figure 6.Conformations of the peptide show differences in bulk and condensed phases.Structural analysis of the peptides.(a−c) 1 H NMR spectra of the peptides in respective LLPS and non-LLPS conditions in 50 mM phosphate, 2.5 M NaCl, pH 8.0 at room temperature as follows: PSP-1 non-LLPS = 1 mM; PSP-2 non-LLPS = 1 mM, PSP-2 LLPS = 3.5 mM; PSP-3 non-LLPS = 1 mM; PSP-3 LLPS = 3.5 mM; PSP-4 non-LLPS = 1 mM; PSP-4 LLPS = 3.5 mM; PSP-5 non-LLPS = 3.5 mM; PSP-5 LLPS = 18 mM.The spectra show the aromatic and backbone amide region (a) and other regions (b and c).Vertical lines indicate chemical shifts that vary significantly between LLPS and non-LLPS conditions.(d) CD spectra for PSP-2 under non-LLPS conditions, at LLPS conditions, and at reducing conditions (with 4 mM DTT).(e) CD spectra for PSP-3 at non-LLPS conditions, at LLPS conditions, and at reducing conditions (with 4 mM DTT).(f) CD spectra for PSP-4 at non-LLPS conditions, at LLPS conditions, and at reducing conditions (4 mM DTT).(g) CD spectra for PSP-5 at non-LLPS conditions, at LLPS conditions, and at reducing conditions (60 mM DTT).

Figure 7 .
Figure 7. Peptide-RNA complex-coacervates are redox-insensitive.(a−e) Phase boundaries of poly-A RNA and PSP-1−5 complex-coacervates, respectively, in oxidizing (−DTT) and reducing (+DTT) conditions in 50 mM Tris and 150 mM NaCl, pH 7.4.Approximate C sat values corresponding to 250 μg/mL RNA are indicated in green lines.(f−j) Confocal microscopy images of FITC-labeled peptides in concentration conditions indicated with dotted circles in panels (a−e; 250 μM peptide, 200 μg/mL poly-A RNA in 50 mM Tris, 150 mM NaCl, pH 7.4).Insets show the pictures of Eppendorf tubes with the corresponding samples.(k−o) Results from fluorescence recovery after photobleaching (FRAP) performed on the droplets shown in panels (f−j).The insets show a representative region of interest used in FRAP analysis (n = 3, scale bar of images = 10 μm, FRAP insets = 2 μm).

Figure 8 .
Figure 8. Peptide condensates show different encapsulation efficiencies for partitioning payloads.(a) Schematic of the assay performed to obtain encapsulation efficiency (EE) for fluorescein and tetramethyl rhodamine (TMR-OMe) payloads.(b) Bright-field images before adding dye (top) and confocal microscopy images of dye partitioning within PSP-2, PSP-3, PSP-4, and PSP-5 peptide condensates (bottom).The scale bar is 20 μm.(c,d) Average of at least three independent UV spectra of the supernatants was obtained after centrifugation (see (a)) after incubation with fluorescein and TMR-OMe.(e) Measured EE for fluorescein (blue) and TMR-OMe (red) dyes within the peptide droplets was calculated from the absorbance measurements in (c and d).

Figure 9 .
Figure 9. Overall conclusions derived from this work.The schematic shows four classes of cysteine-containing (yellow circles) peptides and a control lacking cysteine residues.The most prominent disulfide bond networks are shown in the schematic droplets below each class.The salient features of the condensates, as determined in this study, are listed below for comparison.