Regulation of biomolecular condensate dynamics by signaling

Abstract

Biomolecular condensates are mesoscopic biomolecular assemblies devoid of long range order that contribute to important cellular functions. They form reversibly, are stabilized by numerous but relatively weak intermolecular interactions, and their formation can be regulated by various cellular signals including changes in local concentration, post translational modifications, energy consuming processes and biomolecular interactions. Condensates formed by liquid-liquid phase separation are initially liquid but are metastable relative into hydrogels or irreversible solids stabilized by stronger, more permanent interactions that have been associated with protein aggregation diseases. As a consequence of this a series of cellular mechanisms are available to not only regulate biomolecular condensation but also the physical properties of the condensates.

Graphical abstract.

Biomolecular condensates can have important cellular functions

Biomolecular condensates are membrane-less structures that form through liquid-liquid phase separation (LLPS) of biomolecules, typically proteins. Biomolecular condensation (BC) relies on multivalent interactions, which are provided either by the presence of multiple interacting globular domains separated by flexible linkers or by multiple interacting residues or motifs in intrinsically disordered regions [1,2].

The dynamicity of the resulting condensates is provided by the weak and therefore transient nature of the non-covalent interactions that stabilize them, which vary depending on the system but can include electrostatic (between anionic and cationic residues), hydrophobic, π-π (aromatic-aromatic) and cation-π interactions [3]. Not only the nature and number of interactions is relevant but also their distribution in the sequence, as has been shown to be the case for charged [4] or aromatic residues [5].

Since their discovery, new functions [6,7] for biomolecular condensates are continuously being proposed. A generic important role of BC is the control of cellular processes in space and time [1]: selectively concentrating specific molecules in condensates can indeed facilitate biomolecular interactions and accelerate biochemical reactions [8], thus promoting or activating signalling processes [9,10]. Of course, the phenomenon can also have the opposite effect as condensation can deplete the cytoplasm/nucleoplasm of specific components, thus inhibiting their functions [11,12].

In biphasic systems, provided that solution conditions remain unaltered, the concentrations of phase-separating biomolecules inside and outside the condensates are constant and only the relative volumes of the two phases vary when changes in protein concentration occur. As a consequence BC provides cells with a mechanism to buffer protein concentrations against changes in protein levels due to, for example, alterations in protein synthesis or degradation [7,13]: biomolecular condensates can thus be seen as reservoirs of specific biomolecules that can spontaneously i.e. without work replenish the cytoplasm or nucleoplasm upon consumption.

The strength of the weak intermolecular interactions that drive BC can be exquisitely sensitive to solution conditions such as temperature, pH, ionic strength or the concentration of specific solutes such as ATP [14]. Hydrophobic (entropic) as well as electrostatic (enthalpic) interactions, for example, are strongly affected by temperature, although in opposite ways [15]. BC thus provides cells with a mechanism to respond to environmental changes such as to heat or pH stress [16–18] in ways that allow them to adapt to such external factors.

In summary BC offers a variety of mechanisms to regulate cellular functions and indeed has already been shown to contribute to the regulation of transcription [19–21], translation [22,23] and protein degradation [24]. There is little doubt that, with time, more mechanisms and functional roles for BC will be revealed, highlighting even further the need for highly efficient means of modulating the cellular thermodynamic and kinetic stability of biomolecular condensates by various types of signals, that we will here review.

Regulation of biomolecular condensation

A key feature of condensation processes, in contrast to other phase transitions of biological relevance that can irreversibly change cell fate such as crystallization or protein aggregation into amyloid fibrils, is their reversibility. This property, that is key for efficient signaling, allows cells, at least in principle, to swiftly regulate the position of the phase equilibrium by modifying either of the parameters that control it.

As mentioned above, LLPS is governed by multivalent interactions between specific residues, motifs or domains of biomolecules [1]. Multivalency is indeed one of the key parameters that controls LLPS propensity: increasing the number of valencies promotes condensation while decreasing it hampers LLPS [9,10,25,26]. Proteins can be modified in this way, in different timescales and by various means, to modulate the propensity of biomolecules to phase separate as well as the functions associated with this process.

In relatively slow timescales processes such as alternative splicing can modify the number of valencies of phase separating biomolecules and, thus, their propensity to condensate. For example alternative splicing events modulate this propensity, as well as the properties of the associated condensates, by either introducing or removing a specific region of sequence, as in Ddx4 [25], altering the number of repeats in a sequence, as for Tau [27], or generating multiple isoforms of the protein, as in hnRNPs [28,29] or PML [30,31].

Cells can also spatio-temporally regulate BC, as well as the material properties of the associated condensates [32], in a faster timescale. This can generally be achieved either by changing protein concentration, by modifying the strength of the interactions that stabilize the condensates, by post-translational modifications, or by active energy consuming processes that dissolve the condensates and/or control their material properties:

Regulation by concentration changes

It is clear from first principles that BC strongly depends on the concentration of the relevant biomolecule. A straightforward way to modulate whether condensates form is therefore to alter the overall protein concentration - by altering transcription, translation and protein degradation - or locally by confining the relevant protein at specific cellular foci.

The activation domains of transcription factors, for example, can form transcriptional condensates that recruit members of the transcription machinery and activate transcription when their local concentrations increase upon enhancer binding [33]. Increases in local concentration can also be caused by protein binding to membranes: actin polymerization, for example, starts from phase separated clusters formed on membrane surfaces upon activation [9,10,34,35], TIS granule formation by TIS11B on the endoplasmic reticulum enables translation of certain mRNAs [36], and condensation of SynGAP/PSD-95, Synapsin1 and RIM/RIM-BP on membranes is involved in the control of synaptic transmission [37–39].

Changes in cellular localization that concentrate specific proteins in specific cellular compartments can also trigger condensation: blocking nuclear import of hnRNPA1, for example, results in its accumulation in the cytoplasm and in the formation of stress granules [40]. Similarly, mislocalization of FUS in the cytoplasm also facilitates the formation of stress granules [41]. Finally changes in global protein concentration can of course be due to changes in cell volume rather than to changes in protein levels. This intriguing phenomenon has been described as a factor controlling the assembly of nucleoli by FIB-1, where fluctuations in cell volume control the formation of these nuclear membraneless organelles [42].

Regulation by biomolecular interactions

It is useful, in discussing the mechanisms by which BC can be regulated by biomolecular interactions, to differentiate between scaffold and client proteins. The former are proteins that can condensate by themselves and provide the essential multivalent interactions that drive the process. The latter are proteins that do not condensate, or would do so only under extreme conditions, but that can be found in condensates and influence the condensation process of scaffolding proteins [26].

Indeed client molecules can both enhance phase separation or repress it. Perhaps the most widely studied class of client molecules are RNAs, that alter the phase separation properties of multidomain proteins harboring RNA binding domains (RRMs) such as hnRNPA1 [40,43] and Whi3 [44,45]. If the RNAs contain more than one binding site for RRMs i.e. if they are multivalent they can enhance the condensation properties of the scaffolding protein by transiently linking two protein molecules and thus be found in condensates. Importantly, the magnitude of the enhancement depends on the number of binding sites/valencies, the distance between them in the RNA, and stoichiometry i.e. on the relative concentration of protein and RNA [46].

Biomolecules other than RNAs can play similar roles by also interacting directly with scaffolding proteins. Like RNA they will enhance condensation and be found in the condensates if they are multivalent and interact with domains, residues or motifs that do not drive the condensation of the scaffolding protein. They will instead repress condensation if they are univalent and interact with valencies that are key for condensation by competing against the interactions that drive this process. Importantly, biomolecules can also be found in condensates not because they directly interact with scaffolding or client proteins but due to the favourable physicochemical properties of the condensates, although the mechanistic basis for this phenomenon is not yet well-understood [47]. Finally, the presence of certain client molecules in condensates can exclude others, implying that BC can regulate the functions of proteins that do not condensate themselves [48,49].

Biomolecular interactions may also alter the phase separation properties of specific proteins by inducing conformational changes in them. This is for example the case for nuclear hormone receptors, a class of transcription factors that only become transcriptionally active after activation by hormones that includes the androgen receptor, the estrogen receptor [50,51] and the glucocorticoid receptor [52]. In these cases the ligand does not enhance condensation due to its multivalency, instead it does so by changing the conformation of a domain, likely generating a new valency. Since these proteins are therapeutic targets for important indications such as prostate and breast cancer, respectively, and inflammation there is substantial interest in understanding the roles that BC plays in their function.

Regulation by post-translational modifications

Cellular signalling processes can result in protein post-translational modifications (PTMs). These fast and reversible means of modifying the chemical properties of protein side chains can in principle modify BC processes by altering the strength of the intermolecular interactions in which they are involved or by altering global properties of the protein such as charge patterning or conformation. PTMs known to modify BC include Ser/Thr/Tyr phosphorylation, Arg methylation, Lys acetylation, glycosylation, citrullination, ubiquitination, parylation or SUMOylation and in some cases the mechanistic basis of their regulatory power is established.

The effects of phosphorylations on BC, for example, have been studied for different proteins and appear to be system-specific. In Tau [27,53], FMRP [23], LAT [10], or nephrin/N-WASP/NCK [9,34] phosphorylations enhance condensation whereas in CPEB4 [22], FUS [54–56], and TDP-43 [57] the effect is the opposite. This can be rationalized by considering the nature of the intermolecular interactions stabilizing each condensate. For example Tau condensation, driven by electrostatic interactions, is favoured by phosphorylations because they decrease the charge of an excessively positively charged region of sequence. By contrast in CPEB4 condensation, that does not appear to be driven by electrostatic interactions, phosphorylations greatly increase the global charge of the domain that drives it, likely increasing the repulsion between monomers and hence stabilizing the monomer state.

Methylation of arginine residues has also been studied in the context of BC, its effect being a decrease of the LLPS propensity of the studied systems, such as FMRP [23], hnRNPA2 [58], FUS [59,60] or DDX4 [25]. In the case of FUS, for example, arginine methylation decreases the strength of the cation-π interactions which are responsible for the assembly of the protein into biomolecular condensates. Lysine acetylation has been shown to impair LLPS of DDX3X [61] or Tau [62]. A plausible explanation of this phenomenon is the fact that acetylation of lysine residues causes the loss of the positive charge of this amino acid, therefore impairing both electrostatic and cation-π interactions, which have been shown to be two of the main interactions drivers of LLPS.

Regulation of condensate properties by energy consumption

Energy consuming processes can play a role in the regulation of BC and of the material properties of condensates. ATP-driven molecular chaperones, for example, can promote the disassembly of stress granules at the end of the cellular stress response [63–66]. In addition lowering cellular ATP levels decreases the dynamicity of the stress granules formed by G3BP, preventing them from fusing [67].

RNA helicases such as the DDXs family and the RVB complex, that carry out functions similar to those of molecular chaperones, can also be found in stress granules. The Ded1 RNA helicase, for example, is important for the release of mRNAs from stress granules [67–70].

Finally ATPases such as Cdc48 can be found as components of biomolecular condensates and their ATPase activity can contribute to regulating the formation and dissolution of these assemblies. At the end of the cellular stress response such domains can enable the dissolution of the stress granules and dissipation of the components into the cytoplasm [67,71]. Similarly, processing bodies [72] or stress granules [70] formation has been shown to be regulated by the ATPase activity of some of their components.

Misregulation of biomolecular condensation

The physical properties of membraneless organelles formed by BC can be very different [1]. Condensates formed by LLPS are liquid i.e. behave as fluids whereas others, such as Balbiani bodies in oocytes [73] and some stress granules [67,74], are better described as solid assemblies. The prevalent view is that the liquid character of the former can be key for their functions but that these fluid assemblies are metastable relative to solid ones, characterized by stronger, more permanent intermolecular interactions perhaps equivalent to those driving protein folding.

The process by which biomolecular condensates progressively lose their liquid character in a liquid to solid transition is usually called maturation. In cases where the condensates do not change morphology upon analysis by techniques such as fluorescence microscopy it has been termed gelation, where it is thought that the components of the condensate do not change molecular structure, or do so only locally, but lose their ability to diffuse within the condensate and exchange with protein molecules in their exterior.

In other cases it has become possible to observe that the approximately spherical, liquid condensates formed by LLPS abruptly change morphology to become more dense and fibrillar, even star-shaped [75]. In this scenario it is thought that the components of the condensate undergo a substantial change in molecular structure leading to the adoption of a conformation akin to that stabilizing amyloid fibrils [13].

The maturation process has been proposed as a disease mechanism, mainly in neurodegeneration. This loss of liquid character can have negative consequences because it is necessary for the actual function of the condensate, because it does not allow for the phase equilibrium to be regulated by the various mechanisms listed above and, finally, when the solid assemblies that it produces are inherently toxic.

A number of examples of liquid to solid transitions associated with protein malfunctions or disease mechanisms have been put forward in the literature. Prolonged stress, for example, causes nucleoli to transition from liquid to solid and therefore irreversible glassy solids [65] and long persistence of stress granules composed by G3BP together with TDP-43 and/or FUS can promote the formation of inclusions that behave as hydrogels and are equivalent to those thought to cause amyotrophic lateral sclerosis (ALS) [63,76]. Similarly, chronic activation by 17β-estradiol of the estrogen receptor α leads to the gelation of phase separated MegaTrans enhancers, which causes a decrease in transcriptional activity [51].

Of course malfunctional BC can be caused by the dysregulation of the PTMs that in turn regulate it. This is for example the case for FUS hypomethylation, that promotes formation of hydrogels impairing proper RNP granule function and is associated with frontotemporal lobar degeneration (FTLD) [60], Tau hyperacetylation, that impairs Tau condensation and thus consequent microtubule assembly [62] and Tau hyperphosphorylation, that increases LLPS propensity and is related with Alzheimer's disease [27,53].

Strong evidence for a link between condensate maturation and disease onset comes from the observation that disease causing mutations accelerate condensate maturation. This has been shown for hnRNPA1[40,43], hnRNPA2[58] or FUS[75,77] in ALS and FTD. Also, familial ALS variants of TDP-43 promote formation of irreversible aggregates [78]. Additionally, disease related expansions of polyglutamine tracts have also been shown to correlate with the degree of maturation, as in the tract present in huntingtin, associated with Huntington's disease [79].

Finally mis-splicing can also alter biomolecular condensates properties in disease. Splicing of FXR1, for example, influences the condensates function for development, and mis-splicing in regions promoting condensation has been implicated in multi-minicore myopathy disease [80]. Also, alternative splicing of Tau generates isoforms with different Tau repeats, which influence its LLPS propensity and subsequent amyloid aggregation [27].

Concluding remarks

It is being increasingly recognized that biomolecular condensates play key roles in various important cellular functions. Their functionality relies on the fact that their formation is reversible and can therefore be regulated by cellular signals. This regulation can be achieved by different means, such as changes in the local or global concentration of phase separating biomolecules, changes in the strength of the interactions that stabilize the condensates, energy-consuming processes or biomolecular interactions. The physiological relevance of this regulation is clear since misregulation of BC in cells, as well as disease-associated mutations, can cause changes in the dynamic properties of condensates render them solid, irreversible, non-functional and, potentially, toxic.

Highlights.

• Biomolecular condensation is a reversible physical process that produces mesoscopic biomolecular assemblies defining unique cellular environments

• Cellular signals regulate biomolecular condensation mainly by modulating local protein concentration and the strength of the weak, transient intermolecular interactions that stabilize the condensates

• Failures in the timely regulation of biomolecular condensation and of the dynamic properties of the condensates underlie various pathologies associated with protein aggregation

Figure 1.

Schematic view of the liquid-liquid phase separation process, where some of the intramolecular interactions responsible for the stabilization of the monomeric state become intermolecular in the stabilization of the phase separated state. This is a fast and reversible process that takes places once the concentration of the biomolecule increases above the saturation concentration for phase separation, resulting in the formation of two different phases: a light phase, at the saturation concentration, and a protein rich phase or dense phase, where the protein concentration can reach up to 1000x the saturation concentration. The interactions that play an important role in the stabilization of the dense phase can be of different types, the most widely characterized being hydrophobic (with a positive entropy due to the release of water molecules), electrostatic (between a positive and negative charge), π-π (between aromatic systems) and cation-π (between a positive charge and an aromatic ring). Liquid-liquid phase separation in biological systems gives rise to the formation of biomolecular condensates, which have been described to have key cellular functions, such as the control of biochemical reactions in space and time, a buffering mechanism for concentration changes and to adapt to environmental factors such as heat or pH.

Figure 2.

Biomolecular condensation is a fast and reversible process characterized by the constant exchange of monomeric biomolecules between the interior (dense phase) and the exterior (light phase) of the condensates. However, these biomolecular condensates are metaestable and can therefore, in slower timescales, undergo liquid to solid transitions due to the loss of exchange of monomeric molecules with the solution. This is known as maturation of the condensates, an irreversible phenomenon that has been linked to disease in some cases. Cells regulate the formation and dissolution of the condensates, as well as their physical properties, by various means, including changes in concentration, in the strength of biomolecular interactions, post-translational modifications or energy consuming processes. Additionally, in longer timescales processes such as disease mutations, misregulation of PTMs (hyper- or hypo-), alternative splicing or polyCAG expansions have been shown to modify the propensity of the molecule to undergo LLPS as well as the maturation process, and in many cases they have also been related to diseases due to either a loss of function of the condensates or an acceleration of the maturation process.