From the classical colloidal theory to biomolecular condensation: implications in health and disease

1 Introduction

In 1912, the biologist William Hardy stated: “A large number of observations upon the filterability and the optical properties of colloidal solutions prove that they are coarse grained and consist of particles dispersed throughout a continuous fluid phase. For this reason these solutions are commonly described as multiphase systems in which the dispersed particles form one of the phases” (Hardy, 1912). From the perspective of molecular biophysics, this quotation is akin to the modern definition of biomolecular condensates (herein biocondensates) formed via liquid-liquid phase separation (LLPS). Classical colloids and biocondensates share common structural features and properties. Prominent physicochemical features of colloids are the size scale, shared with biocondensates, ranging from nano-to hundreds of nanometers; the biphasic behavior; the capability to disperse light (Tyndall effect); the impossibility of separating the particles from the solvent by conventional filtration; the surface energy characterized by a high surface-to-volume ratio; and the high stability, explained by the Derjaguin, Landau, Verwey and Overbeek (DLVO) theory (Agmo Hernández, 2023), originated primarily by van der Waals attraction and electrostatic repulsion forces (Brangwynne et al., 2015). In fact, biocondensates are biological colloids that can phase-separate in the micron-scale (Shin and Brangwynne, 2017; Vazquez et al., 2022).

In the late 19th century, pioneering work by Selmi, Graham and Van Bemmelen, among others, paved the way for colloidal theory and shaped our understanding of dispersed systems. However, it had limitations when applied to the study of biomolecules (Lyklema, 2015). Its approach highlighted factors like Brownian motion, multivalency scaffolding, percolation, and viscoelastic properties, as well as the effects of particle size on diffusion (Brangwynne et al., 2015). As the study of biomolecules progressed, it became clear that colloidal foundational principles needed a revision to encompass the behaviour of complex biological systems: biomolecules exhibit distinctive properties that the original colloidal theory does not fully capture (Sarangapani et al., 2015). Its major limitation lies in the scale and complexity of biomolecular interactions; biomolecules are highly dynamic and heterogeneous macromolecules. In addition, biomolecules possess intricate three-dimensional domains, intrinsically disordered regions (IDRs), sequence complexity, and variable shapes that affect their interactions and functionalities. Most of the current knowledge of phase condensation has arisen from the analysis of large liquid-like condensates, like membraneless organelles (MLOs). These microscopically observable structures arise from smaller and less understood condensates measuring tens to hundreds of nanometres, known as nanocondensates (Toledo et al., 2023a). Nanocondensates are considered important precursors in several nucleation processes, such as fibrillation, irreversible aggregation, amorphous aggregation, and ordered crystallization. Nanocondensation could be widespread in the cell, as recent evidence indicates that nearly 20% of the proteome could potentially be organized in nanocondensates (Keber et al., 2024). New techniques are being developed to focus specifically on nanocondensates (Houx et al., 2026).

A generic phase diagram for proteins in some temperature-dependent parameter and concentration plane shows a binodal behavior (Figure 1) (Alberti et al., 2019; Soranno, 2020; Villois et al., 2022). In the single-phase regime, two regions are delimited by the saturation curve: the undersaturated and the supersaturated regions. The nanocondensate-nucleation processes emerge in the supersaturated (metastable) region below the solubility curve. Supersaturation leads to phase separation, and it is certainly the only stably way to increase protein concentration above solubility limits (Toledo et al., 2023a).

2 The many roles of biocondensates in cellular processes

Classic organelles are compartments physically defined by a lipid bilayer membrane that separates their interior from the cytoplasm. The chemical composition of the organelle membrane regulates the entry and exit of most biological molecules through a specialised membrane-attached transport protein machinery. In contrast, biocondensates are MLOs that form through liquid-liquid phase separation and play a fundamental role in the organisation and compartmentalisation of cellular processes without a physical barrier (Figure 2). The liquid-like properties of biocondensates enable a dynamic exchange with the surrounding cytoplasm via a rapid assembling-disassembling mechanism. Starting with a uniform solution, the phenomenon of liquid-liquid phase separation depends on a demixing transition process that results in two coexisting phases. During this physicochemical process, macromolecules are drawn out from the initial solution, forming a liquid-like dense phase in swift exchange with a diluted phase (Vazquez et al., 2022).

MLOs are involved in almost all biological processes, from signalling to protein storage and degradation (Hirose et al., 2023; Li Y. et al., 2024). Abnormal condensate properties can produce harmful effects leading to pathological conditions (Figure 2) (Zhang S. et al., 2023). Aberrant biocondensate-related diseases involve several mechanisms including loss of protein’s function or gain in protein’s toxicity, and RNA foci dysregulation. This makes them promising targets for exploring novel drug-based research and therapies.

From the view of a simplistic unified model, two main components are involved in the biocondensate: the scaffold and the client (Rana et al., 2024). Cellular biocondensates can contain hundreds to thousands of different components and, in most cases, only a few of these, the scaffold components, suffice to form a biocondensate. The scaffold components form a network of multivalent interactions constituting the core of the condensate. The client components are homo or heterotopically interacting biomolecules recruited to the preformed condensate, establishing multiple interactions with the multivalent scaffold. The client components typically undergo dynamic exchange with the surrounding environment, providing a regulatory mechanism.

In most cases, the scaffold components are proteins with low complexity sequences. Due to their inherent low complexity sequence, intrinsically disordered proteins (IDPs) have totally or in part, undefined tridimensional structures (Gonçalves-Kulik et al., 2022). Intrinsically disordered regions (IDRs) represent ∼14% of the proteome of Bacteria and Archaea and increase up to 44–54% in Eukaryotes (Perdigão et al., 2015). The lack of a well-defined structure gives IDPs the plasticity necessary to establish multiple and dynamic long-range weak interactions, providing a driving force for LLPS (Borcherds et al., 2021). Besides, several factors such as concentration, ionic strength, pH, osmolarity, crowding, temperature, hydrostatic pressure, and low molecular-weight molecules influence assembly-disassembly dynamics of biocondensates (Figure 2) (Alberti and Hyman, 2016). The following sections highlight the role of inorganic compounds and covalent modifications in biocondensate dynamics.

3 The importance of being inorganic: the role of protonation state

Water is the most abundant inorganic molecule, making up 70% or more of the total cell volume. Hydronium ion concentration, represented as the pH, exerts fundamental roles in biomolecular conformation, oligomeric state, catalytic activity, and protein-protein interactions via changes in the protonation state of titratable residues. In the same way, pH alters LLPS by influencing the electrostatic interacting network (Jin et al., 2022; Notarmuzi and Bianchi, 2024; Zhang et al., 2024).

Aging alters intracellular pH homeostasis (Kurz et al., 2004; Kenyon, 2010; Gems and Partridge, 2013). This phenomenon is well exemplified in yeast, where replicative aging is asymmetric: mother cells age, while daughter cells are rejuvenated (Henderson et al., 2014). This is due to the asymmetric distribution of the plasma membrane H⁺-ATPase (Pma1), the major regulator of cytosolic pH. Pma1 accumulates in aging mother cells, but it is largely absent in daughter cells. Consequently, vacuoles become less acidic in aging mother cells, which limits lifespan. Similarly, by analyzing a large sample cohort, Decker et al. found that the pH of the brain and cerebrospinal fluid decreases with aging (Decker et al., 2021). Following the same pattern, extracellular pH is further decreased in patients with Alzheimer’s disease (AD).

Several studies found amyloid deposits of the pathogenic low-complexity domain (LCD) of the transactive response DNA-binding protein (TDP-43) in patients with aging-related diseases like amyotrophic lateral sclerosis or AD (Kao et al., 2015; Josephs et al., 2017; Nemani et al., 2018). Under physiological conditions of pH, ionic strength and crowding, TDP-43 LCD can undergo LLPS at low concentration and form larger structures as protein concentration increases (Babinchak et al., 2019). However, a low decrease in pH can trigger the formation of amyloid fibrils even in the absence of liquid-like droplets (Babinchak et al., 2019), also showing hydrogel behavior in non-LLPS conditions (Lin et al., 2024). Moreover, the analysis of droplet formation in vitro through pH-jump from pH 11.0 to 7.5 shows an increase in the radius of the droplets over time, denoting its strong pH dependence (Van Lindt et al., 2021). This behavior is consistent with the formation of TDP-43 puncta observed in HEK239T cells at neutral pH (Al Ojaimi et al., 2024), the value where the maximum solubility was predicted (Pintado-Grima et al., 2022).

4 Enzymatic activity inside biocondensates

Catalytic biochemical reactions are known to be highly sensitive to temperature and pH, and most of them are performed by well-behaved folded proteins. Nevertheless, some IDPs exhibit enzymatic activity, often through a substrate-induced folding-upon-binding mechanism (Chakrabortee et al., 2010; Kim et al., 2018; Zhang Y. et al., 2023). While most enzymes have an optimal activity near neutral pH (Talley and Alexov, 2010), some evolved to function at extreme pH, like pepsin which has a narrow range of functionality at very low pH (pH 1.5–2.0).

As aforementioned, biocondensates usually present a dense phase characterized by a crowded environment, where the intrinsic dynamics of both globular proteins and IDPs can be altered (Candotti and Orozco, 2016; Bonucci et al., 2021; Das et al., 2024). To test the effect of biocondensation on enzyme and substrate dynamics, Saini and Mukherjee (Saini and Mukherjee, 2023) designed a highly efficient heterotypical biocondensate involving folded horseradish peroxidase and glucose oxidase that exhibited increased catalytic activity up to 205-fold. Surprisingly, this occurred at a more acidic pH than in a non-crowded environment (pH 6.0–6.5 versus pH 4.0) (Schomburg et al., 1994). Hence, biocondensates can be rationally designed as factories to enhance specific enzymatic activities.

The confinement within biocondensates can also protect enzymes from degradation or inhibition (Fiedler et al., 2010; Singha et al., 2019). This further optimises the catalytic efficiency, even for enzymes that depend on conformational rearrangements.

Protein intrinsic dynamics can also be modulated by changing the polarity or solvent-composition within the condensate displacing the equilibrium to the open or close conformation (Zhou and Pang, 2018; Vazquez et al., 2022). For instance, Stoffel and collaborators (Stoffel et al., 2025) conjugated lipase 2 from Bacillus thermocatenulatus with the LCD regions of DEAD-box protein Laf1, and the resulting low micrometer-size condensate recruited the chimera at millimolar concentrations in the dense phase. The chimera protein in the dense phase exhibited altered protonation state, expanding the pH range at which the protein shows high activity (Testa et al., 2021; Choi et al., 2023; Dindo et al., 2024). Notably, nanocondensates may exhibit a significant increase in enzymatic activity compared to larger condensates, as was recently shown by the same group (Gil-Garcia et al., 2024).

5 Catalytically active amyloids

Biomolecular condensates and amyloid formation are connected via a phase transition pathway from dynamic liquid states to solid structures (Figure 2). For instance, the amyloid-β peptide Aβ40 can undergo liquid-liquid phase separation to form an intermediate condensed state, in which high peptide concentrations accelerate nucleation and the rapid assembly of amyloid fibers prior to transitioning to a solid state (Morris et al., 2024). According to Vendruscolo and Fuxreiter (2022), there are two aggregation pathways: a deposition route involving increasingly organized disordered oligomers and a condensation route involving a liquid droplet state prior to forming solid amyloids. Sequence determinants for aggregation in condensates include droplet-forming propensity, intrinsic aggregation tendency, and binding entropy.

Amyloid protofibrils are soluble, high-molecular-weight intermediates in the aggregation process of proteins such as the amyloid-β. These represent a transitional state in the liquid-solid phase separation process between smaller oligomers and larger, insoluble mature fibrils. They can grow through a hierarchical colloidal assembly process in which pre-formed spherical oligomers act as “pearls”, aligning and fusing with each other (Xu et al., 2001; Modler et al., 2003). According to this model, amyloid formation is a colloidal phenomenon governed by physical forces, such as dipole moments and DLVO forces. These forces favor linear growth over amorphous precipitation.

Amyloids can be either functional, beneficial aggregates, like storage bodies, or unchain the formation of disease-linked amyloid fibers. Structural frustration—which relies on the degree of energetically unfavorable interactions in a protein conformation (Gianni et al., 2021) — emerged as a key concept and a hotspot to define the fate of these aggregates. Amyloid fibers are highly frustrated ensembles that can be regarded as the Last Universal Common Ancestor for protein folding, due to its repetitiveness, self-replication and information-transfer capability (Maury, 2018). It is generally accepted that amyloid fibrils are thermodynamically more stable and rigid structures than the classical native conformations, due to their tightly packed hydrophobic core (Buell, 2022). Nevertheless, they exhibit some degree of flexibility (Okumura and Itoh, 2016; Vugmeyster et al., 2016) — especially in the flanking regions which leads to conformational polymorphism—allowing the formation of well-defined tertiary functional binding sites lacking in monomeric counterparts (Bhopatkar and Kayed, 2023).

These supramolecular assemblies have garnered significant attention due to their catalytic activity (Marshall and Korendovych, 2021). In a proof-of-concept work, Rufo et al., designed a battery of simple minimalist heptapeptides that can coordinate with Zn²⁺ intermolecularly and form efficient catalysts for ester hydrolysis (Rufo et al., 2014). Thenceforth, an increasing number of physiological-relevant amyloids like Aβ42 (Arad et al., 2021); α-synuclein (Horvath and Wittung-Stafshede, 2023); the islet amyloid polypeptide (Sawazaki et al., 2024); and the glucose blood-level regulator glucagon (Arad et al., 2022; Nandi et al., 2024), tested positive for several catalytic activities, including hydrolysis, esterolysis, and dephosphorylation.

The catalytic efficiency and selectivity of the aggregates are not as high as those of optimized enzymes. However, their structural versatility and tunability make them exciting models for studying the evolution of catalysis within highly stable folds and adaptation to ancient physicochemical environments. Through this lens, frustration appears to be a feasible parameter for identifying potential fibrillation proneness via high-throughput proteomic and bioinformatic analysis and sparks the interest in exploring ways to design drugs to inhibit or enhance amyloid fibril activity.

6 Metal ions as low molecular-weight regulators of biocondensates

The metallome plays fundamental roles in cell metabolism (Aulakh et al., 2025) and also in biocondensate homeostasis (Figure 2). Two of the most important cations are magnesium and zinc. With a millimolar concentration in eukaryotic cells, magnesium ion plays a significant role in RNA-based biocondensates stabilizing RNA structure, folding, and promoting intermolecular interactions with other RNA molecules and proteins (Roden and Gladfelter, 2021). Phase separation of defined RNA structures, like riboswitches, can be promoted by Mg²⁺ ions using crowding agents such as polyethyleneglycol. Nevertheless, the presence of Mg²⁺ ions alone is enough to promote LLPS in the absence of a crowder, evidencing the importance of this cation in stabilizing RNA-based biocondensates (Poudyal et al., 2021). Also, Mg²⁺ ion diminishes the internal dynamic of heterotypical RNA-protein biocondensates, showing a liquid-gel transition that can be reversed in the presence of ATP (Yewdall et al., 2022), which acts as a chelator.

On the other hand, zinc has a major impact in liquid-to-gel transitions mostly in protein-based biocondensates (Shih et al., 2022). Particularly, Zn²⁺ ions are crucial for the correct biosynthesis, storage and release of insulin in pancreatic β-cells (Fukunaka and Fujitani, 2025). In connection with this, insulin biogenesis starts with the oxidative folding of its single-chain precursor proinsulin, and proinsulin can oligomerize into a R₆ hexamer (Emdin et al., 1980; Kiselar et al., 2011). Further, our group recently demonstrated that proinsulin can self-populate nanocondensates in vitro, under conditions that mimic the environment of the Golgi apparatus and the immature secretory granules (Toledo et al., 2023b). Importantly, these nanocondensates can form microcondensates in presence of the secretory granule transmembrane protein ICA512. In this regard, Zn²⁺ ions may be important for the subsequent liquid-solid phase separation of insulin-Zn²⁺ crystals. Zinc ions can also bind to preformed insulin fibrils, altering its kinetics and morphology (Frankær et al., 2017). In this context, ion-induced aggregation of relevant therapeutic hormones for drug delivery and biomanufacturing is a promising strategy.

The accumulation of Fe³⁺, Mn²⁺, and Al³⁺ ions plays a significant role in the neuronal environment, leading to the formation of toxic amyloid species. In particular, iron promotes protein aggregation and the production of reactive oxygen species through Fenton-type chemistry, thereby exacerbating lipid peroxidation (Joppe et al., 2019; Xu et al., 2024). Exposure to manganese has been linked to the accelerated maturation of α-synuclein condensates (Xu et al., 2022). Trivalent cations, such as Al³⁺, exert a potent influence on the structural polymorphism of amyloid assemblies, often promoting the formation of insoluble fibrils over amorphous aggregates (Ricchelli et al., 2005).

The interaction between Cu²⁺ ions and amyloidogenic proteins is governed by a bimodal regulatory mechanism that is strictly dependent on stoichiometric ratios and acts as either a kinetic catalyst or a structural inhibitor. At sub-stoichiometric concentrations, copper acts as a scaffold that neutralizes electrostatic repulsions primarily via histidine residues and stabilizes high-affinity binding sites, thereby accelerating the nucleation kinetics of the amyloid-β peptide (Sarell et al., 2010; Abelein et al., 2022). Conversely, supra-stoichiometric levels induce a rapid, disordered collapse of the protein chain. This favors the formation of amorphous aggregates or traps the system in non-fibrillar, curvilinear protofibrillar states (Tian et al., 2024). Recent literature highlighting the role of several other important metal ions of biological relevance can be found in references (Hong et al., 2020; Vazquez et al., 2022; Sołtys et al., 2023).

7 Chemical modification inside biocondensates: isopeptide bond formation

Post-translational modifications (PTMs) refers to the rearrangement or addition of covalent bonds in proteins, expanding their conformational landscape, localization, functionality, thermodynamic stability, net charge, and interactions with other biomolecules. The homeostasis of protein modifications is important for protein function and hence for human health (Figure 2). Abnormal regulation of PTMs leads to changes in protein properties and loss of functions, which is closely related to the development of several diseases (Zhong et al., 2023; Li W. et al., 2024; Ramazi et al., 2024).

Reversible protein phosphorylation, one of the best-known PRMs, can promote or repress condensate formation by changing net charge, protein conformation and/or modulating protein-protein or protein-RNA interactions (Snead and Gladfelter, 2019). Phosphorylation targets the hydroxyl groups of Ser, Thr and Tyr residues and the imidazole ring of histidine. Sridharan and collaborators (Sridharan et al., 2022) mapped numerous phosphorylated sites enriched in both soluble and condensate proteins in HeLa cell lysates, including a subset of phosphosites that modulate protein-RNA interactions. Notably, multi-phosphorylation of the key RNA-splicing factor heteronuclear ribonucleoprotein A1 reduces its ability to localize to nuclear clusters, suggesting a phosphorylation-induced alteration in RNA-protein condensates interactions. Conversely, multi-phosphorylation of nucleophosmin 1, an essential nucleolar protein, lowers its partitioning to the nucleolus, thereby decreasing protein-RNA interaction and increasing its retention in the nucleoplasm.

Phosphorylation of biocondensates can also regulate virus infection. Nuclear bodies (NBs) formed during the last stages of human adenovirus infection require the L1-52/55 (52k) protein for capsid assembly (Charman et al., 2023). Phosphorylation of the IDR region of 52k in Ser residues positively modulates liquid-like biocondensation and promotes viral particle assembly (Grams et al., 2024). In the same capacity, phosphorylation of a Ser/Arg-rich region of the nucleocapsid N protein of SARS-CoV-2 destabilises a distal Leu-rich helix (Favetta et al., 2024), promoting its partial unfolding, and reduces protein-RNA condensation, enhancing viral infection and evasion of the host immune system.

Although protein ubiquitination and related PTMs such as SUMOylation and neddylation have been known for more than 50 years, the understanding of how these PTMs influence phase separation is a recent development. An intricate crosstalk between protein degradation and function regulation takes place (Jin, 2019), upon which LLPS acts as a link. This is exemplified by the formation of nuclear bodies of the promyelocytic leukemia protein, where SUMOylated proteins are regulated by ubiquitination (Sha et al., 2019). Further research should focus on the fine-tuning properties of PTMs in biocondensate as cellular rheostats. Outstanding articles on the role of ADP-ribose addition to proteins (PARylation) in biocondensate organization can be found elsewhere (Alemasova and Lavrik, 2022; Rhine et al., 2023; Liu et al., 2025).

8 Conclusion and perspectives

Spontaneous biomolecules segregation into concentrated liquid-like droplets constitutes a new paradigm in protein biophysics, with deep implications for our understanding of normal and abnormal cellular biochemistry. Condensation implies a higher level of organization in the protein-energy landscape. The folding and function of single protein molecules, as described by the still-valid Anfinsen’s dogma, are now seen as part of a collective of molecules orchestrated to optimize physical and biological properties, allowing proteins to reach concentrations beyond their solubility limits (see Figure 1). The assembly-disassembly dynamics of biocondensates are tightly modulated by environmental factors like pH, temperature, cofactors (metal ions), ionic strength, PTMs, crowding and others. This provides a complex mechanism for regulating cellular processes adding a new dimension to the relationship between protein sequence, structure and cellular function.

Phase separation frequently represents the initial step in the protein aggregation pathway within the folding energy landscape and commonly results in the formation of amyloid fibrils. Importantly, the emergence of catalytic activities within these supramolecular scaffolds reframe amyloidogenic transitions as shifts toward non-canonical enzymatic states rather than only pathological misfolding. Consequently, the therapeutic frontier should evolve from inhibiting aggregation to embracing the concept of interfacial pharmacology (Kochman et al., 2025; Fleming and Schuijers, 2026). By targeting the rheological properties and thermodynamic stability of biocondensates, it becomes possible to move beyond aggregate dissolution and toward the deliberate reprogramming of the protein energy landscape.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Diana Mitrea, DeMix Solutions, LLC, United States

Reviewed by: Eva Zerovnik, Institut Jožef Stefan (IJS), Slovenia

Hector Garcia Seisdedos, Spanish National Research Council (CSIC), Spain