Intrinsically disordered regions (IDRs) are key building blocks of biomolecular condensates, yet disorder per se is insufficient to drive phase separation. IDRs function conforms to the sticker–spacer theory: modular sticker motifs mediate multivalent, weak interactions, while flexible spacers supply conformational entropy to sustain liquid-like dynamics (Martin et al., 2020). This behavior is encoded by low-complexity sequences (LCRs) enriched for charged (R/K; D/E), polar (G/Q/N/S), and aromatic (Y/F) residues that frequently appear as short repeats (e.g., YG/S-, RG-, SY), furnishing targets for multivalent contacts (Nott et al., 2015; Ravindran et al., 2023).

IDRs do not rely on a single force; instead, they deploy a portfolio of weak, short-range interactions—electrostatic, hydrophobic, cation–π, π–π/sp²–π, and hydrogen bonding—that act in parallel to nucleate and stabilize condensates (Das et al., 2020; Murthy et al., 2021; Nott et al., 2015; Qamar et al., 2018). FUS exemplifies IDR-driven LLPS: its N-terminal SYGQ low-complexity segment, central RRM/zinc-finger domains, and C-terminal RGG repeats cooperate to promote condensation. [G/S]Y [G/S] motifs drive liquid-to-gel/solid transitions, while cation–π and hydrophobic interactions lower Csat. The RGG–SYGQ network further stabilizes droplets via hydrogen bonds, sp²–π stacking, and enhanced cation–π contacts (Kato et al., 2012; Murray et al., 2017; Murthy et al., 2021; Qamar et al., 2018; Wang and Li, 2024). These multimodal interactions endow FUS condensates with robustness and plasticity, yet their imbalance can drive pathological solidification. The functional spectrum of IDRs is highly diverse. For instance, in G3BP1, IDR1 acts as an autoinhibitory module that suppresses phase separation through intramolecular interactions, whereas IDR2 counteracts this inhibition to promote phase separation, collectively revealing a bidirectional control mechanism (Yang et al., 2020). Thus, the intricate modularity and regulatability of IDRs offer a promising foundation for developing targeted therapeutic interventions against condensation-associated pathologies.

Coexisting condensates in cells always show spatial exclusivity (De La Cruz et al., 2024; King et al., 2024),reflecting sequence-encoded selectivity. In the nucleolus, IDR features (e.g., D/E tracts vs. K-blocks with E-rich segments) and associated RNA/DNA-binding domains bias partitioning among FC/DFC (King et al., 2024). Moreover, experimental swap assays further demonstrate this principle: exchanging MED1’s charged IDR features with glutamine/tyrosine motifs redirected proteins toward FUS condensates, whereas the reciprocal replacement recruited proteins to MED1 condensates (De La Cruz et al., 2024). Another example comes from the distribution of charge and aromatic residues, where dense RGG clusters promote homotypic separation and exclusivity, while dispersed charges favor heterotypic mixing (Jo et al., 2022; Welles et al., 2024). These findings provide a theoretical foundation for modulating and targeting phase-separated condensates. However, the predictive capacity of sequence features for partitioning behavior is constrained by incomplete understanding of multiscale interactions; and the dynamic interplay between IDRs and other cellular components remains underexplored.

Liquid-liquid phase separation (LLPS) stabilizes the genome by modulating key processes such as DNA damage signaling, heterochromatin organization, telomere maintenance and assembly of critical cellular structures. While studies show that condensates like TopBP1 amplify ATR–Chk1 signaling and stabilize replication forks (Frattini et al., 2021); and that 53BP1 compartments modulate repair pathway choice (Zhang et al., 2022),a crucial question remains: how do these hubs prevent spurious activation? The enormous capacity for amplification could potentially lead to excessive signaling from minor lesions. The finding that SLX4-SUMO condensates regulate the timely resolution of repair intermediates (Alghoul et al., 2023) hints at built-in termination mechanisms. At telomeres, shelterin components such as TRF1 and TRF2 facilitate the formation of telomric condensates that promote higher-order compaction and T-loop formation. Furthermore, the role of LLPS extends to the assembly of critical cellular structures. A prime example occurs during mitosis, where the phase separation of NuMA, regulated by Aurora A kinase, drives the precise assembly of the mitotic spindle (Mengjie et al., 2021). In heterochromatin maintenance, LLPS driven by HP1 proteins, reinforced by H3K9me3 deposition and ATRX–DAXX–mediated H3.3 incorporation, facilitates the compaction and functional partitioning of heterochromatic regions. Notably, 53BP1–HP1α condensates contribute to heterochromatin stability largely independent of canonical double-strand break repair, suggesting a structural role in genome organization beyond acute damage handling (Larson et al., 2017; Zhang et al., 2022). These compartments function not only structurally but also protectively by limiting spurious engagement of DNA damage response factors, thereby preventing illegitimate repair and preserving chromosome ends (Jack et al., 2022). This critical perspective suggests that LLPS functions as a double-edged sword in heterochromatin, balancing compaction with accessibility, a tension that is not yet fully explored. Notably, aberrant transcriptional condensates can drive tumorigenesis by inducing chromosomal structural abnormalities. For instance, in leukemia, the aberrant transcriptional condensates formed by the NUP98-HOXA9 fusion protein orchestrate tumorigenesis through a dual mechanism: mimicking super-enhancer functionality to hyperactivate proto-oncogenes while concurrently remodeling the 3D chromatin architecture through anomalous looping, ultimately leading to uncontrolled cellular proliferation (Jeong Hyun et al., 2021).

Liquid-liquid phase separation (LLPS) has emerged as a pivotal mechanism in gene transcription regulation, functioning as a bidirectional switch that dynamically assembles membrane-less compartments to control genomic output. On the activating front, coactivators like BRD4 and Mediator form condensates at super-enhancers that enrich the transcriptional machinery, facilitating Pol II initiation and productive elongation, with Pol II-CTD modifications further tuning engagement within these hubs to promote promoter escape (Boehning et al., 2018; Boija et al., 2018; Sabari et al., 2018). Conversely, repressive condensates regulate gene silencing by compacting chromatin and reducing its accessibility. Polycomb assemblies (e.g., PRC1/CBX2) undergo phase separation to achieve acute compaction, which is then stabilized by sustained H3K27me3 deposition to maintain long-term repression. Additionally HP1α droplets nucleate at H3K9me3 sites to reinforce heterochromatin organization (Larson et al., 2017; Plys et al., 2019). Beyond the behavior of individual protein complexes, the material state of chromatin itself is increasingly recognized as an active participant in this process; the condensation of nucleosome arrays, driven by histone tails or modulated by small molecules like Ru1, alters viscoelasticity, reconfigures enhancer-promoter communication, and ultimately shifts transcriptional output (Gibson et al., 2019; W. J. Wang et al., 2021). These findings offer more dynamic perspective than traditional static models for understanding chromatin reorganization. Future efforts should focus on the integration of LLPS biophysics with mechanistic studies of transcriptional control.

Despite distinct compositions, major RNA–protein condensates share a common assembly principle: multivalent interactions mediated by intrinsically disordered regions (IDRs) and RNA initiate core formation, while interfacial energy differences drive multiphase organization. For instance, the nucleolus nucleates via high rRNA transcription, organizing into a tripartite condensate (FC, DFC, GC) scaffolded by electrostatic/π-interaction networks involving NPM1, rRNA, and arginine-rich proteins like SURF6; compartmentalization is refined by interfacial tension, facilitating ordered rRNA processing (Feric et al., 2016; Ferrolino et al., 2018; Lafontaine et al., 2021; Mitrea et al., 2018). Similarly, nuclear speckles form through LLPS driven by IDR-rich scaffolds like SON and SRRM2, with phosphorylation dynamically regulating architecture (Ilik et al., 2020; Prashant et al., 2024; Xu et al., 2022). In contrast, stress granules (SGs) assemble via multi-step condensation of untranslated mRNPs, stabilized by G3BP1, CAPRIN1, and TIA1 networks (Jain et al., 2016; Wheeler et al., 2016; Yang et al., 2020).

Functionally, RNA condensates modulate cellular metabolism and stress responses through the selective concentration and release of RNAs and ribonucleoproteins (RNPs). In the nucleus, the nucleolus exemplifies this principle by forming a biomolecular condensate that provides a viscoelastic environment enhancing pre-rRNA processing efficiency (Feric et al., 2016; Lafontaine et al., 2021). Under stress conditions, cells employ phase separation to reconfigure the splicing machinery, precisely tuning mRNA output through widespread intron retention. A key illustration of this is the CLK1/2 kinase–driven remodeling of nuclear speckles, which promotes intron retention to fine-tune mRNA output (Prashant et al., 2024; Wang B. et al., 2024; Xu et al., 2022), highlighting how phase separation integrates transcriptional and post-transcriptional control. In the cytoplasm, stress granules (SGs) transiently sequester mRNAs to arrest translation. DYRK3-mediated dissolution of SGs enables translational recovery upon stress alleviation (Jeon et al., 2024). However, the functional implications of SGs remain context-dependent, as persistent SGs do not always correlate with cellular rescue, suggesting complex regulation beyond mere condensation. At a systems level, RNA concentration bidirectionally regulates condensates: low levels stabilize via electrostatic complementarity to enhance transcription, while high levels dissolve them (Henninger et al., 2021). Beyond translational control, emerging evidence indicates that SGs act as immunomodulatory hubs by linking translational arrest to innate immune signaling. For instance, TRIM25–G3BP1 co-condensation facilitates the tuning of the RIG-I/MDA5 pathway (Onomoto et al., 2014; Shang et al., 2024). Nevertheless, the specificity of immune modulation through phase separation is debated, particularly in viral infections, where pathogens may exploit LLPS to evade immune detection.

Membrane-less condensates establish high-activity reaction hubs that amplify signaling outputs and sharpen response thresholds. At the plasma membrane, multivalent receptor–adaptor assemblies (e.g., EGFR–Grb2) exhibit condensate-like properties that potentiate downstream activity, suggesting a general mechanism for signal amplification (Lin et al., 2022). This principle extends to the nucleus, where YAP-derived condensates enrich transcription factors like TEAD1 to drive proliferative programs (Cai et al., 2019). Innate immune responses are similarly regulated through opposing LLPS modalities: while polyubiquitin-driven condensation of NEMO amplifies NF-κB signaling, STING forms inhibitory condensates that prevent hyperactivation (Du et al., 2022; Yu et al., 2021). These findings reveal that LLPS enables bidirectional control of signaling pathways. However, a key limitation across these studies is the heavy reliance on overexpression systems and engineered constructs, leaving open the question of whether such condensates form at endogenous expression levels and under physiological conditions. This gap is particularly relevant in disease contexts: although LLPS-enabled super-enhancers (e.g., containing LSD1, BRD4, and FOXA1) sustain oncogenic signaling in castration-resistant prostate cancer (Li et al., 2023), and aberrant glycogen phase separation promotes hepatocarcinogenesis via YAP activation (Liu et al., 2021), the direct causal role of phase separation in tumorigenesis, versus its role as a correlative epiphenomenon, remains inadequately tested. Thus, while IDR-driven phase separation is proposed to constitute a central layer of signaling control, the field must critically evaluate its mechanistic necessity in vivo.

LLPS directly orchestrates the spatial organization of the cytoskeleton by unifying force-producing components. For instance, reconstituted PSD condensates promote F-actin bundling and spine growth through charged surface interactions on Homer EVH1 (Chen et al., 2023). This finding is complemented by studies on abLIM1, whose condensates integrate actin nucleation and crosslinking to assemble tension-bearing cortical networks under low-salt conditions (Sen et al., 2022). A parallel logic applies to microtubule regulation: TPX2–tubulin co-condensation enhances branching nucleation on pre-existing microtubules, while importin-α/β suppresses condensation to impose spatiotemporal precision (King and Petry, 2020). Collectively, these examples illustrate how phase separation links upstream signals to the architectural programming of the cytoskeleton. This mechanism paves the way for innovative drug development targeting major diseases such as cancer metastasis and neurodegenerative disorders and facilitates the creation of novel smart biomaterials.

LLPS plays a dual role in proteostasis, facilitating both the targeted degradation of ubiquitinated substrates and the formation of adaptive stress-responsive compartments. p62/SQSTM1 undergoes K63-ubiquitin–driven phase separation to initiate selective autophagy (Sun et al., 2018), while in the nucleus, p62 condensates function as proteasomal “reactors” that enrich K48-ubiquitylated substrates to accelerate degradation (Yasuda et al., 2020). Under stress, the proteasome itself forms reversible nuclear condensates (SIPANs) that co-assemble substrates, receptors, and machinery into a single functional unit (Uriarte et al., 2021). Similarly, RNP granules such as P-bodies and stress granules triage misfolded clients and silenced mRNPs, with their dysfunction leading to proteotoxic stress (Nostramo et al., 2019). The material properties of these granules are actively maintained by chaperone systems like HSPB8–BAG3–HSP70, which prevent pathological solidification (Ganassi et al., 2016); and regulated by factors such as NS1-BP, which stabilizes stress granules to enable their autophagic clearance (Jeon et al., 2024). These studies collectively position LLPS as a central organizer of protein quality control. However, they also reveal a critical, unresolved tension: the same physicochemical forces that promote adaptive compartmentalization can, under dysregulated conditions, drive a liquid-to-solid transition that underlies neurodegenerative pathology. This underscores the need to better understand the physiological checks and balances that maintain functional phase separation and prevent its collapse into disease-associated aggregates.

LLPS serves as a universal organizing principle across the central dogma, coordinating genome stability, transcriptional regulation, dynamic RNP organelle assembly, and protein homeostasis, as integrated in Figure 3.

On short timescales (t < τ_relax), biomolecular condensates predominantly display elastic responses; over longer periods (t > τ_relax), viscous flow becomes dominant. This macroscopic behavior can be provisionally captured by classical Maxwell-type viscoelastic models. Importantly, condensate “aging” involves more than a simple rise in viscosity. Research indicates that aging is principally driven by time-dependent internal network remodeling, governed by sequence-specific intermolecular interactions—such as hydrogen bonds and π–π stacking—that function as physical crosslinks (Alshareedah et al., 2024). These interactions directly strengthen network connectivity and elevate the elastic modulus (G′), rather than merely increasing macroscopic viscosity. Consequently, aging manifests as a marked extension of the relaxation time (τ_relax), slowed fusion and rounding dynamics, and systematic shifts in rheological moduli as well as fluorescence recovery after photobleaching (FRAP) kinetics (Jawerth et al., 2020; Mukherjee et al., 2024); Such network remodeling has direct functional repercussions. Aging can reduce molecular exchange rates and substrate accessibility within condensates, leading to time-dependent declines or alterations in biochemical activities, including catalytic function (Kang et al., 2025). This indicates that the balance between stability and dynamics, maintained through metastable free-energy landscapes and multivalent interactions, is inherently dynamic and precarious. Notably, aging may progress toward irreversible transitions. Under pathological conditions, liquid-liquid phase separation (LLPS) itself can serve as a platform for aberrant aggregation. For example, α-synuclein undergoes LLPS to form condensates that dramatically concentrate the protein, creating a local environment conducive to irreversible conformational changes and drives the irreversible solidification of liquid condensates into amyloid fibrillar aggregates (Ray et al., 2020). Beyond internal interactions and aging, condensate phase behavior is modulated by a spectrum of factors including interfacial effects, viscoelastic relaxation, the competition between nucleation and spinodal decomposition pathways, local crowding, and chaperone activity (Visser et al., 2024).

Proteins and RNA can drive phase transitions through multivalent interaction modules. These typically consist of sticker motifs, such as aromatic or arginine residues, and flexible or disordered spacer linkers. The sticker–spacer framework proposes that reversible associations between stickers can drive two distinct yet often coupled transitions: liquid–liquid phase separation, governed by net attraction as quantified by a renormalized chi parameter and occurring above a saturation concentration (Csat); and a percolation transition, driven by network connectivity and defined by a percolation threshold (c_perc), which leads to the formation of a system spanning physical gel. When Csat is lower than c_perc, and c_perc in turn lies below the concentration in the dense phase, phase separation and percolation become tightly coupled. The resulting condensates thus form reversible physical microgels that exhibit time dependent viscoelastic properties rather than irreversible solidification (as seen in Figure 4) (Choi et al., 2020; Harmon et al., 2017; Mittag and Pappu, 2022). Since 2015, studies have revealed that the pathological aggregation of many neurodegenerative disease related proteins, such as tau, α synuclein, and TDP 43, is closely associated with phase transitions (Cao et al., 2019; Fitzpatrick et al., 2017; Li et al., 2018). These proteins can form fibrils through two primary nucleation pathways: classical nucleation in homogeneous solution (percolation transition), and condensate mediated nucleation. The latter is governed by a two threshold model in which liquid–liquid phase separation (LLPS) is initiated at a lower concentration, generating protein rich condensates. A subsequent, higher concentration threshold within these condensates then triggers the liquid to solid transition, dramatically accelerating fibril formation (Mathieu et al., 2020; Molliex et al., 2015; Patel et al., 2015; Ray et al., 2020; Wegmann et al., 2018). Notably, biomolecular condensates can exert dual regulatory effects on nucleation. For example, full length α synuclein, due to its amphipathic nature, accumulates at condensate interfaces where elevated local concentrations reduce the nucleation energy barrier and promote fibrillation. In contrast, truncated variants of α synuclein can experience nucleation suppression within certain condensates, likely due to stabilizing molecular interactions (Lipiński et al., 2022). In hematology, pathological liquid phase separation acts as a critical nucleation precursor that accelerates the aberrant polymerization and fibrillation of sickle cell hemoglobin (HbS), thereby driving the core pathology of erythrocyte rigidification and vaso occlusion (Galkin et al., 2002). In oncology, mutant p53 undergoes nuclear LLPS, forming prion like condensates that solidify into amyloid aggregates and facilitate gain of function tumorigenesis (Petronilho et al., 2021). Moreover, oligomerization may lower nucleation barriers and reduce saturation thresholds (Dao et al., 2018). Although the resulting inclusions can transiently buffer toxic oligomers and exert a short lived neuroprotective effect, sustained proteostasis stress or failure of quality control pathways ultimately drives their maturation into fibrillar deposits that perpetuate aggregation and toxicity (Alberti and Dormann, 2019; Tompkins and Hill, 1997). This dynamic highlights a central conceptual tension in the field: the dual role of phase transitions as both adaptive mechanisms and drivers of pathology. Critically, it raises the fundamental question of when and how a physiological response shifts toward becoming pathological.

The internal microenvironment of biomolecular condensates is formed through dynamic interactions among proteins, RNA, and solutes. These interactions govern key condensate properties—such as biochemical reactivity, molecular sorting, and material exchange—and enable the selective recruitment of specific small molecules, which in turn influence condensate stability and assembly. Mechanistically, small molecules can alter the physical state of condensates by modulating their internal environment. Hydrophilic compounds, for example, raise polarity and lower viscosity, while “proteophilic” molecules like ATP exclude water, thereby increasing hydrophobicity and viscosity. Such effects suggest that cells can fine-tune the material state of membraneless organelles through changes in the concentration of endogenous regulators (Pan et al., 2025). However, under pathological conditions, small molecules may act as drivers that promote the conversion of liquid-like condensates into gel-like or solid states (Pan et al., 2025). Studies using purified α-synuclein have further revealed how small molecules can selectively modulate phase separation. Polyphenols such as quercetin and EGCG promote the formation of liquid condensates while potently inhibiting their conversion into solid amyloid fibrils. These molecules bind α-synuclein via hydrophobic interactions, enhancing its effective multivalency to drive phase separation, while simultaneously stabilizing conformations that hinder β-sheet assembly required for fibrillation. In contrast, dyes like thioflavin T act as potent inducers that rapidly initiate phase separation and accelerate aggregation (Sarkar et al., 2025). These findings illustrate that small molecules can precisely control the dynamics and phase states of biomolecular condensates, providing a rationale for interfering with pathological protein aggregation. This insight underpins the emerging therapeutic strategy termed Condensate-Modifying Therapeutics (C-mods). Rather than targeting classic protein active sites, C-mods are rationally designed small molecules that directly regulate the assembly, material state, and function of biomolecular condensates—often by tuning properties such as hydrophilicity/hydrophobicity, charge distribution, and three-dimensional conformation (Usman et al., 2025; Wang X. et al., 2024). C-mod activity is highly context-dependent. For instance, polar aliphatic molecules like 1,6-hexanediol can dissolve condensates (e.g., Cajal bodies, stress granules) by disrupting weak hydrophobic interactions, yet under certain conditions they can also promote condensate formation or chromatin compaction through competitive interference with multivalent interactions (Hamad et al., 2021; Trnka et al., 2021). Polar aromatic C-mods frequently operate by weakening inter-chain hydrogen bonds within condensates or via competitive binding, thereby preventing protein oligomerization or abnormal conformational transitions—offering new intervention routes for neurodegenerative diseases, cancer, and viral infections (Usman et al., 2025). Beyond synthetic compounds, ions and endogenous metabolites also exhibit dual regulatory capacities. In viral infection, Zn²⁺ promotes phase separation of the viral nucleocapsid protein (N), supporting replication and RNA packaging. Host antiviral responses can counteract this by upregulating Cu²⁺, which competitively inhibits Zn²⁺, disrupting viral morphology and reducing pathogenicity (Monette and Mouland, 2020). Similarly, ATP acts as a key endogenous modulator: at low concentrations it generally enhances phase separation, whereas high concentrations inhibit it. The adenine ring engages in π-π stacking or hydrophobic interactions with protein regions, while the triphosphate chain influences solvent environment; age-related declines in ATP levels compromise the reversibility of phase separation (Meyer and Voigt, 2017; Rice and Rosen, 2017). In summary, C-mods regulate liquid–liquid phase separation through multiple, reversible, and context-sensitive mechanisms. This approach moves beyond the classical “lock-and-key” drug-design model, opening new avenues for treating condensate-related pathologies.

The study of biomolecular condensates is evolving from descriptive observation to quantitative prediction, yet significant challenges persist. Current techniques lack sufficient resolution to dynamically probe condensate organization in living systems, necessitating advanced methods that can precisely manipulate key physicochemical parameters. Fundamental questions remain about cellular control over phase transitions, requiring integrated theoretical frameworks combining Flory-Huggins and active matter principles. Therapeutically, while “molecular glues” and “wedges” offer promising approaches for modulating pathological condensates, achieving specificity presents substantial hurdles. Ultimately, tackling these challenges will demand intensified cross-disciplinary efforts to translate mechanistic insights into clinical applications.