Characterization of design grammar of peptides for regulating liquid droplets and aggregates of FUS

Leveraging a large language model to predict protein phase transition: A physical, multiscale, and interpretable approach

Protein phase transitions (PPTs) from the soluble state to a dense liquid phase (forming droplets via liquid-liquid phase separation) or to solid aggregates (such as amyloids) play key roles in pathological processes associated with age-related diseases such as Alzheimer's disease. Several computational frameworks are capable of separately predicting the formation of droplets or amyloid aggregates based on protein sequences, yet none have tackled the prediction of both within a unified framework. Recently, large language models (LLMs) have exhibited great success in protein structure prediction; however, they have not yet been used for PPTs. Here, we fine-tune a LLM for predicting PPTs and demonstrate its usage in evaluating how sequence variants affect PPTs, an operation useful for protein design. In addition, we show its superior performance compared to suitable classical benchmarks. Due to the "black-box" nature of the LLM, we also employ a classical random forest model along with biophysical features to facilitate interpretation. Finally, focusing on Alzheimer's disease-related proteins, we demonstrate that greater aggregation is associated with reduced gene expression in Alzheimer's disease, suggesting a natural defense mechanism.

An aberrant fused in sarcoma liquid droplet of amyotrophic lateral sclerosis pathological variant, R495X, accelerates liquid-solid phase transition

Intracellular aggregation of fused in sarcoma (FUS) is associated with the pathogenesis of familial amyotrophic lateral sclerosis (ALS). Under stress, FUS forms liquid droplets via liquid-liquid phase separation (LLPS). Two types of wild-type FUS LLPS exist in equilibrium: low-pressure LLPS (LP-LLPS) and high-pressure LLPS (HP-LLPS); the former dominates below 2 kbar and the latter over 2 kbar. Although several disease-type FUS variants have been identified, the molecular mechanism underlying accelerated cytoplasmic granule formation in ALS patients remains poorly understood. Herein, we report the reversible formation of the two LLPS states and the irreversible liquid-solid transition, namely droplet aging, of the ALS patient-type FUS variant R495X using fluorescence microscopy and ultraviolet-visible absorption spectroscopy combined with perturbations in pressure and temperature. Liquid-to-solid phase transition was accelerated in the HP-LLPS of R495X than in the wild-type variant; arginine slowed the aging of droplets at atmospheric conditions by inhibiting the formation of HP-LLPS more selectively compared to that of LP-LLPS. Our findings provide new insight into the mechanism by which R495X readily forms cytoplasmic aggregates. Targeting the aberrantly formed liquid droplets (the HP-LLPS state) of proteins with minimal impact on physiological functions could be a novel therapeutic strategy for LLPS-mediated protein diseases.

Interfacial and intrinsic molecular effects on the phase separation/transition of heteroprotein condensates

Liquid-liquid phase separation (LLPS) and droplet formation by LLPS are key concepts used to explain compartmentalization in living cells. Protein contact to a membrane surface is considered an important process for protein organization in a liquid phase or during transition to a solid or liquid dispersion state. The direct experimental comprehensive investigation is; however, not performed on the surface-droplet interaction and phase transition. In the present study, we constructed simple and reproducible experiments to analyze the structural transition of aggregates and droplets in an ovalbumin (OVA) and lysozyme (LYZ) complex on glass slides with various coatings. The difference in droplet-surface interaction may only be important in the boundary region between aggregates and droplets of a protein mixture, as shown in the phase diagram. Co-aggregates of OVA-LYZ changed to droplet-like circular forms during incubation. In contrast, free l-lysine resulted in the uniform droplet-to-solid phase separation at lower concentrations and dissolved any structures at higher concentrations. These results represent the first phase-diagram-based analysis of the phase transition of droplets in a protein mixture and a comparison of surface-surface and small molecular-droplet structure interactions.

Suppression of TDP-43 aggregation by artificial peptide binder targeting to its low complexity domain

TAR DNA-binding protein 43 (TDP-43), aggregation prone protein, is a potential target of drug discovery for amyotrophic lateral sclerosis. The molecular binders, targeting the disordered low complexity domain (LCD) relevant to the aggregation, may suppress the aggregation. Recently, Kamagata et al. developed a rational design of peptide binders targeting intrinsically disordered proteins based on contact energies between residue pairs. In this study, we designed 18 producible peptide binder candidates to TDP-43 LCD by using this method. Fluorescence anisotropy titration and surface plasmon resonance assays demonstrated that one of the designed peptides bound to TDP-43 LCD at 30 μM. Thioflavin-T fluorescence and sedimentation assays showed that the peptide binder suppressed the aggregation of TDP-43. In summary, this study highlights the potential applicability of peptide binder design for aggregation prone proteins.

Rational design of phase separating peptides based on phase separating protein sequence of p53

Artificial phase-separating (PS) peptides can be used in various applications such as microreactors and drug delivery; however, the design of artificial PS peptides remains a challenge. This can be attributed to the limitation of PS-relevant residues that drive phase separation by interactions of their pairs in short peptides and the difficulty in the design involving interaction with target PS proteins. In this study, we propose a rational method to design artificial PS peptides that satisfy the requirements of liquid droplet formation and co-phase separation with target PS proteins based on the target PS protein sequence. As a proof of concept, we designed five artificial peptides from the model PS protein p53 using this method and confirmed their PS properties using differential interference contrast and fluorescence microscopy. Single-molecule fluorescent tracking demonstrated rapid diffusion of the designed peptides in their droplets compared to that of p53 in p53 droplets. In addition, size-dependent uptake of p53 oligomers was observed in the designed peptide droplets. Large oligomers were excluded from the droplet voids and localized on the droplet surface. The uptake of high-order p53 oligomers into the droplets was enhanced by the elongated linker of the designed peptides. Furthermore, we found that the designed peptide droplets recruited p53 to suppress gel-like aggregate formation. Finally, we discuss aspects that were crucial in the successful design of the artificial PS peptides.

Mechanism underlying liquid-to-solid phase transition in fused in sarcoma liquid droplets

The RNA-binding protein fused in sarcoma (FUS) forms ribonucleoprotein granules via liquid-liquid phase separation (LLPS) in the cytoplasm. The phase separation of FUS accelerates aberrant liquid-solid phase separation and leads to the onset of familial amyotrophic lateral sclerosis (ALS). We previously found that FUS forms two types of liquid condensates in equilibrium, specifically LP-LLPS (i.e., normal type) and HP-LLPS (i.e., aberrant type), each with different partial molar volumes. However, it is unclear how liquid condensates are converted to the pathogenic solid phase. Here, we report a mechanism underlying the aberrant liquid-to-solid phase transition of FUS liquid condensates and the inhibition of this transition with small molecules. We found that the liquid condensate formed via HP-LLPS had greatly reduced dynamics, which is a common feature of aged wild-type FUS droplets and the droplet-like assembly of the ALS patient-type FUS variant. The longer FUS remained on the HP-LLPS, the harder it was to transform it into a mixed state (i.e., one-phase). These results indicate that liquid-to-solid phase transition, namely the aging of droplets, is accelerated with HP-LLPS. Interestingly, arginine suppressed the aging of droplets and HP-LLPS formation more strongly than LP-LLPS formation. These data indicate that the formation of HP-LLPS via the one-phase state or LP-LLPS is a pathway leading to irreversible solid aggregates. Dopamine and pyrocatechol also suppressed HP-LLPS formation. Our data highlight the potential of HP-LLPS to be used as a therapeutic target and arginine as a plausible drug candidate for ALS-causing FUS.

Rational peptide design for regulating liquid-liquid phase separation on the basis of residue-residue contact energy

Since liquid-liquid phase separation (LLPS) of proteins is governed by their intrinsically disordered regions (IDRs), it can be controlled by LLPS-regulators that bind to the IDRs. The artificial design of LLPS-regulators based on this mechanism can be leveraged in biological and therapeutic applications. However, the fabrication of artificial LLPS-regulators remains challenging. Peptides are promising candidates for artificial LLPS-regulators because of their ability to potentially bind to IDRs complementarily. In this study, we provide a rational peptide design methodology for targeting IDRs based on residue-residue contact energy obtained using molecular dynamics (MD) simulations. This methodology provides rational peptide sequences that function as LLPS regulators. The peptides designed with the MD-based contact energy showed dissociation constants of 35-280 nM for the N-terminal IDR of the tumor suppressor p53, which are significantly lower than the dissociation constants of peptides designed with the conventional 3D structure-based energy, demonstrating the validity of the present peptide design methodology. Importantly, all of the designed peptides enhanced p53 droplet formation. The droplet-forming peptides were converted to droplet-deforming peptides by fusing maltose-binding protein (a soluble tag) to the designed peptides. Thus, the present peptide design methodology for targeting IDRs is useful for regulating droplet formation.

Structure-dependent recruitment and diffusion of guest proteins in liquid droplets of FUS

Liquid droplets of a host protein, formed by liquid–liquid phase separation, recruit guest proteins and provide functional fields. Recruitment into p53 droplets is similar between disordered and folded guest proteins, whereas the diffusion of guest proteins inside droplets depends on their structural types. In this study, to elucidate how the recruitment and diffusion properties of guest proteins are affected by a host protein, we characterized the properties of guest proteins in fused in sarcoma (FUS) droplets using single-molecule fluorescence microscopy in comparison with p53 droplets. Unlike p53 droplets, disordered guest proteins were recruited into FUS droplets more efficiently than folded guest proteins, suggesting physical exclusion of the folded proteins from the small voids of the droplet. The recruitment did not appear to depend on the physical parameters (electrostatic or cation–π) of guests, implying that molecular size exclusion limits intermolecular interaction-assisted uptake. The diffusion of disordered guest proteins was comparable to that of the host FUS, whereas that of folded proteins varied widely, similar to the results for host p53. The scaling exponent of diffusion highlights the molecular sieving of large folded proteins in droplets. Finally, we proposed a molecular recruitment and diffusion model for guest proteins in FUS droplets.

[Development of peptide binder design method for disease-related phase separation proteins]

Neurodegenerative diseases such as dementia and Alzheimer's disease are caused by liquid-liquid phase separation (LLPS) proteins. LLPS is a phenomenon in which a dense liquid phase of proteins is formed in a liquid phase in which proteins are dispersed at a low concentration. The concentrated proteins enable highly efficient chemical reactions, but at the same time, there is a risk of forming insoluble aggregates that cause diseases. In fact, neurodegenerative disease-related proteins form insoluble aggregates, which cause great damage to nerves, resulting in memory and motor disorders. Drug discovery requires the design of drug candidates that can strongly bind to the intrinsically disordered region of a phase-separated protein and control the phase-separated state. This paper mainly introduces our research on peptide design that binds to phase-separated proteins. For peptide drug discovery, it is necessary to efficiently search for drug candidates among a huge number of peptides. As an efficient search method for peptides that control phase-separated proteins, we searched for amino acids that can control liquid-liquid phase separation, and devised a method for designing peptides containing effective amino acids. It was demonstrated that this method can be used to control the LLPS and solid aggregate formation of the neurodegenerative disease-related protein FUS. Furthermore, we devised a method for rationally designing a peptide that binds complementarily to the intrinsically disordered region of the target, and demonstrated the functional control of the cancer disease-related protein p53. Finally, we discuss the possibility of peptide drug discovery for disease-related LLPS proteins.

Heterotypic amyloid interactions: Clues to polymorphic bias and selective cellular vulnerability?

The number of atomic-resolution structures of disease-associated amyloids has greatly increased in recent years. These structures have confirmed not only the polymorphic nature of amyloids but also the association of specific polymorphs to particular proteinopathies. These observations are strengthening the view that amyloid polymorphism is a marker for specific pathological subtypes (e.g. in tauopathies or synucleinopathies). The nature of this association and how it relates to the selective cellular vulnerability of amyloid nucleation, propagation and toxicity are still unclear. Here, we provide an overview of the mechanistic insights provided by recent patient-derived amyloid structures. We discuss the framework organisation of amyloid polymorphism and how heterotypic amyloid interactions with the physiological environment could modify the solubility and assembly of amyloidogenic proteins. We conclude by hypothesising how such interactions could contribute to selective cellular vulnerability.

Single-Molecule Microscopy Meets Molecular Dynamics Simulations for Characterizing the Molecular Action of Proteins on DNA and in Liquid Condensates

DNA-binding proteins trigger various cellular functions and determine cellular fate. Before performing functions such as transcription, DNA repair, and DNA recombination, DNA-binding proteins need to search for and bind to their target sites in genomic DNA. Under evolutionary pressure, DNA-binding proteins have gained accurate and rapid target search and binding strategies that combine three-dimensional search in solution, one-dimensional sliding along DNA, hopping and jumping on DNA, and intersegmental transfer between two DNA molecules. These mechanisms can be achieved by the unique structural and dynamic properties of these proteins. Single-molecule fluorescence microscopy and molecular dynamics simulations have characterized the molecular actions of DNA-binding proteins in detail. Furthermore, these methodologies have begun to characterize liquid condensates induced by liquid-liquid phase separation, e.g., molecular principles of uptake and dynamics in droplets. This review discusses the molecular action of DNA-binding proteins on DNA and in liquid condensate based on the latest studies that mainly focused on the model protein p53.

Molecular principles of recruitment and dynamics of guest proteins in liquid droplets

Despite the continuous discovery of host and guest proteins in membraneless organelles, complex host–guest interactions hinder the understanding of the molecular grammar governing liquid–liquid phase separation. In this study, we characterized the localization and dynamic properties of guest proteins in liquid droplets using single-molecule fluorescence microscopy. Eighteen guest proteins of different sizes, structures, and oligomeric states were examined in host p53 liquid droplets. Recruitment did not significantly depend on the structural properties of the guest proteins, but was moderately correlated with their length, total charge, and number of R and Y residues. In contrast, the diffusion of disordered guest proteins was comparable to that of host p53, whereas that of folded proteins varied widely. Molecular dynamics simulations suggest that folded proteins diffuse within the voids of the liquid droplet while interacting weakly with neighboring host proteins, whereas disordered proteins adapt their structures to form tight interactions with the host proteins. Our study provides insights into the key molecular principles of the localization and dynamics of guest proteins in liquid droplets.