The control centers of biomolecular phase separation: How membrane surfaces, post-translational modifications, and active processes regulate condensation

Introduction

Compartmentalization of cells into distinct, functional volumes is essential for life. All cells require the partitioning of molecules into separate compartments to carry out diverse processes. Biomolecular condensates, composed of phase-separated protein and nucleic acid, are emerging as critical centers of cellular compartmentalization (Banani et al., 2017). These structures often behave as liquid-like droplets, in which component molecules are mobile and exchange with the surrounding medium (Brangwynne et al., 2015). In this review, we will refer to these entities as biomolecular condensates, phase-separated structures or networks, droplets, or clusters. Phase-separated structures can be found in physiological contexts throughout the cell, including nucleoli (Brangwynne et al., 2011) and P granules (Brangwynne et al., 2009), and display unique functional and biophysical identities (Langdon et al., 2018; Zhang et al., 2015). In addition to their essential roles in cell physiology, biomolecular condensates are involved in pathological aggregates that cause neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) (Patel et al., 2015). Phase separation is an intrinsic feature of many proteins – droplets will form spontaneously at sufficient protein concentration in vitro or when overexpressed in cells (Shin and Brangwynne, 2017). However, cells must control the natural tendency of proteins and nucleic acids to phase separate in order to build functional condensates at the correct time and location. The cellular mechanisms that regulate phase separation thus are a critical part of understanding how cells utilize biomolecular condensation to control diverse processes (Alberti et al., 2019).

The fundamental principles of regulating phase separation

Biomolecular phase separation is driven by multivalent interactions between proteins and nucleic acids, which collectively facilitate the formation of a condensed network that can display liquid-like or other related material properties (Pak et al., 2016; Shin and Brangwynne, 2017). A variety of different types of multivalent interactions can facilitate phase separation, including electrostatic, cation-π, π-π, and dipole-dipole interactions (Banjade et al., 2015; Brangwynne et al., 2015; Das and Pappu, 2013; Nott et al., 2015;

Vernon et al., 2018; Wang et al., 2018b). How do cells control these interactions in order to regulate phase separation? More broadly, what are the governing parameters that cells can manipulate in order to build and maintain biomolecular condensates with defined properties and identities? Some key parameters include the concentrations of the molecular components, the valencies and strengths of interaction between molecules, the initial nucleation or seeding event, environmental parameters such as ionic strength and pH, and thermodynamic parameters such as temperature. By toggling the key parameters of phase separation, cells can control when and where condensates assemble, the rates of assembly and disassembly, and the material and biophysical properties of condensates such as viscosity and surface tension.

While cells may have limited control over environmental and thermodynamic parameters, several key regulatory mechanisms enable control of the other governing parameters of phase separation. For example, membrane surfaces enable regulation over the concentrations required for phase separation, as well as the nucleation event (Fig. 1). Specifically, membrane surfaces restrict molecular diffusion to a two-dimensional plane, lowering the concentration threshold to phase separation and providing control over the timing and location of nucleation. In another important example, protein post-translational modifications (PTMs) facilitate control over molecular valency and strength of protein-protein interactions (Fig. 1). Finally, molecular chaperones and other active processes modify and restructure the architecture of condensates to control assembly and emergent properties (Fig. 1). These active processes transform condensates into “active matter” that can readily adapt to different environments and potentially transition between different functional states.

In this review, we discuss recent advances in our understanding of membrane surfaces, post-translational modifications, and active processes in regulating biomolecular phase transitions. In the process, we highlight areas in which our understanding of these regulatory mechanisms is lacking, and we suggest future research directions that will elucidate how cells control the assembly and function of biomolecular condensates in time and space.

Regulation of phase separation by membrane surfaces

Membrane surfaces are emerging as essential regulatory platforms for controlled assembly of biomolecular condensates. Specifically, membrane surfaces restrict the diffusion of proteins to two dimensions, reducing the threshold concentrations required for phase separation (Fig. 2A). The ability of membrane surfaces to concentrate proteins has long been recognized in the context of signaling cascades, in which membrane surfaces enhance signaling by increasing the local concentration of signaling proteins compared to the cytosol (Kholodenko et al., 2000). In fact, many signaling proteins assemble into macroscopic clusters at membranes upon activation by an extracellular cue, facilitating amplified signaling compared to individual proteins. Although these clusters have been studied extensively in diverse signaling pathways, it is only recently becoming appreciated that these clusters display the hallmarks of phase-separated systems (Banjade and Rosen, 2014; Case et al., 2019a; Huang et al., 2016; Su et al., 2016).

A key example is the transmembrane protein linker for the activation of T cells (LAT). LAT becomes phosphorylated upon activation of the T cell receptor, facilitating the assembly of a membrane-bound network with the proteins Grb2 and Son of Sevenless (SOS) (Houtman et al., 2006). Specifically, the SH2 domain in Grb2 binds phosphotyrosines in LAT, while the SH3 domains in Grb2 bind the proline-rich domain in SOS. Once bound to the membrane, SOS undergoes a conformational change that facilitates activation of Ras and propagation of downstream signals (Iversen et al., 2014). Recent work found that this protein network phase separates in vitro on supported lipid bilayers to form macroscopic, membrane-bound clusters (Huang et al., 2016; Su et al., 2016; Su et al., 2017), consistent with predictions from earlier theoretical work (Nag et al., 2009). These clusters were found to increase the membrane dwell time of SOS, increasing the probability of SOS undergoing the conformational changes necessary for Ras activation (Huang et al., 2016). Membrane-bound clusters thereby enable cells to distinguish between transient SOS interactions with the membrane and true signals. A recent study provided a more in-depth examination of this “kinetic proofreading” mechanism for SOS activation in membrane-bound clusters (Huang et al., 2019). Moreover, these membrane-bound structures were found to recruit kinases and exclude phosphatases, promoting downstream signaling toward actin polymerization (Su et al., 2016). In particular, clusters promoted the ability of the protein Nck to recruit the actin regulatory proteins N-WASP and Arp2/3 to promote actin polymerization at the membrane. Thus, these phase-separated clusters facilitate localized actin assembly at membrane surfaces in response to specific extracellular stimuli.

Importantly, phase-separated clusters have been demonstrated to promote actin assembly at membrane surfaces in other contexts. Specifically, a study found that the nephrin-Nck-N-WASP system, which regulates actin polymerization in kidney podocyte cells (Jones et al., 2006), also undergoes phase separation at membrane surfaces (Banjade and Rosen, 2014). Nephrin is a transmembrane protein with a disordered cytoplasmic tail containing three phosphotyrosines. The SH2 domains in Nck interact with the phosphotyrosines in nephrin, and the SH3 domains in Nck interact with the proline-rich domain of the actin regulatory protein N-WASP. Collectively, these proteins form a multivalent, phase-separated network (Li et al., 2012) capable of driving local actin assembly at membrane surfaces (Banjade and Rosen, 2014). Importantly, lower concentrations of the proteins were required to observe phase-separated clusters at membrane surfaces in comparison to condensates formed in solution (Banjade and Rosen, 2014; Li et al., 2012). This finding supports a role for membrane surfaces in reducing the threshold to phase separation. More recent work also found that phase separation of this network increases the dwell time of N-WASP and the actin filament nucleator Arp2/3 at membrane surfaces, leading to increased actin assembly at the site of phase separation (Case et al., 2019b). Moreover, the stoichiometry of the components within membrane-bound clusters was found to influence the dwell times of N-WASP and Arp2/3, thereby facilitating fine tuning of actin assembly (Case et al., 2019b). Thus, another regulatory function of membrane surfaces may be to locally control the stoichiometry of the components of a phase-separated network in order to control the emergent functions of different condensates in a context-dependent manner. Collectively, these studies demonstrate that membrane surfaces are capable of assembling and regulating functional condensates that promote robust signaling and actin polymerization.

Although the nephrin- and LAT-based networks are the best-characterized examples of phase separation at membrane surfaces, recent studies suggest that membranes may act as a general platform to regulate phase separation in other contexts. In a notable example, phase separation was recently shown to regulate the organization of synaptic vesicles (Milovanovic et al., 2018). Specifically, the protein synapsin was found to form a condensate capable of capturing and accumulating synaptic vesicles at nerve terminals. In another recent report, phase separation was shown to regulate protein organization at presynaptic membranes prior to synaptic transmission (Wu et al., 2019). In this work, the scaffold proteins RIM and RIM-BP, involved in clustering voltage gated calcium channels in the active zone of presynaptic neurons (Sudhof, 2012), were found to undergo phase separation in vitro on supported lipid bilayers, forming two-dimensional, membrane-bound clusters (Wu et al., 2019). These clusters also enriched voltage-gated calcium channels, supporting a role in helping to organize presynaptic membranes. Collectively, this work suggests that phase separation serves a general function of organizing presynaptic active zones.

Endomembranes and biomolecular phase separation

Studies of phase separation at membrane surfaces have primarily focused on condensates that assemble at the plasma membrane. However, endomembrane surfaces are also emerging as platforms for regulating and organizing condensates throughout the cell. For example, a recent study showed that the RNA-binding protein TIS11B phase separates to form a meshwork that is intertwined with the endoplasmic reticulum (ER) (Ma and Mayr, 2018). This complex forms a distinct subcellular compartment that facilitates protein-protein interactions essential for trafficking of proteins to the cell surface. In another example, influenza viral ribonucleoprotein particles, which display qualities of a liquid-liquid phase separation, were found to assemble at ER exit sites (Alenquer et al., 2019). Collectively, these reports suggest that the ER membrane surface may serve an essential role in regulating when and where condensates assemble in diverse processes. Importantly, the ER occupies a substantial fraction of the cell volume and rapidly explores the cytoplasm (Nixon-Abell et al., 2016; Valm et al., 2017), suggesting that the ER membrane surface may act as a key regulator of condensate assembly and function throughout the cell.

One mechanism by which the ER could regulate condensate assembly in space and time is by contacting and communicating with other organelles and the plasma membrane (Fig. 2B). These membrane contacts serve important functions in signaling, exchange of materials, and organelle biogenesis (Fernández-Busnadiego et al., 2015; Phillips and Voeltz, 2016; Valm et al., 2017). While the apposed membrane surfaces do not directly touch or fuse, protein assemblies at these interfaces facilitate tethering and signaling (Phillips and Voeltz, 2016). An interesting hypothesis is that the ER may regulate the assembly of different condensates via specific interactions with other organelles. Future work will reveal how condensates may be spatially patterned and temporally controlled by specific organelle interfaces.

It is also likely that biomolecular condensates may regulate and maintain organelle contact sites to facilitate signaling and transfer of materials (Fig. 2B). For example, recent work suggests that a protein which organizes the interface between the ER and ER-Golgi intermediate compartments may undergo a phase separation that regulates the early secretory pathway between these organelles (Hanna et al., 2017; Johnson et al., 2015). Specifically, the protein TFG binds to coated vesicles trafficking from the ER to the ER-Golgi intermediate compartment. The authors hypothesize that TFG may form a phase separated network that efficiently concentrates coated vesicles, thereby ensuring efficient secretion by preventing the dispersal of vesicles away from the space between the organelles (Hanna et al., 2017). These findings suggest that an important function of biomolecular condensates may be to stabilize and regulate organelle contact sites, though studies are needed to examine this hypothesis in other contexts.

The membrane surfaces of other organelles may also play essential roles in regulating the positioning of biomolecular condensates within cells. Previous work found that a ribonucleoprotein complex is co-transported with endosomes along microtubules, suggesting that some RNA-based condensates rely on docking to endosomal membranes in order to be shuttled over long distances (Baumann et al., 2012). A more recent study showed that another RNA condensate is transported long distances by docking to lysosomes (Liao et al., 2019). Specifically, the protein annexin A11 acts as a tether that attaches RNA granules to the lysosomal membrane, thereby facilitating granule hitchhiking as lysosomes are transported on microtubules. Collectively, these studies suggest an essential role of various endomembrane surfaces in regulating the positioning of biomolecular condensates in cells. Importantly, subcellular positioning was found to dramatically influence the material states of ribonucleoprotein condensates in neurons, and a disease-linked mutant with altered material properties also displayed disrupted transport (Gopal et al., 2017). Together, these findings suggest that the biophysical properties and functions of condensates are intimately linked with their transport and subcellular positioning. How subcellular location regulates condensate material properties, and how endomembrane surfaces help to regulate condensate assembly and properties during co-transport, are questions for further study.

Phase separation and endocytosis

Recent work suggests that biomolecular phase separation may serve an important role in endocytic trafficking. Specifically, actin assembly is essential for shaping the membrane and driving vesicle scission during endocytosis in yeast (Galletta et al., 2010). Recent work suggests that a potentially phase-separating network composed of SH3-proline rich domain interactions concentrates actin nucleating factors to promote actin polymerization at sites of endocytosis (Sun et al., 2017). This result further supports a central function of membrane-bound condensates in enhancing actin assembly at membrane surfaces (Banjade and Rosen, 2014; Huang et al., 2016; Su et al., 2016). Additionally, intrinsically disordered protein domains are prevalent in membrane trafficking (Pietrosemoli et al., 2013), and many of these disordered domains are regulated by phosphorylation, a key post-translational modification that controls phase separation (Hofweber and Dormann, 2019; Miao et al., 2018). Thus, cells may regulate the phosphorylation state of endocytic proteins in order to tune condensate assembly at the membrane surface at various times during the formation of trafficking vesicles. These condensates may enhance the recruitment of other components of the endocytic protein network, or even directly remodel the underlying membrane. Indeed, previous work has shown that phase-separated mixtures of synthetic polymers are capable of driving membrane remodeling (Li et al., 2011), suggesting that biomolecular condensates may also directly shape membranes by unknown mechanisms.

Biomolecular condensates and lipid phase separation

In addition to being influenced by and remodeling membrane surfaces, membrane-bound condensates may also control the organization of the underlying lipids (Fig. 2C). Specifically, cellular membranes are organized into micro- or nano-scale domains of distinct lipid and protein composition (Eggeling et al., 2009; Owen et al., 2012; Rayermann et al., 2017; Toulmay and Prinz, 2013). These domains serve important functions in cell physiology and pathology (Lingwood and Simons, 2010; Sezgin et al., 2017). Interestingly, protein networks such as actin have been shown to organize underlying lipid domains (Honigmann et al., 2014; Koster et al., 2016). Therefore, an intriguing function of biomolecular condensates composed of protein and nucleic acid may be to regulate the organization of underlying, phase-separated lipid domains. Inversely, phase-separated lipid domains may provide specialized environments at the membrane that regulate assembly of protein and nucleic acid-based condensates. It will likely be worthwhile to examine the possible inter-dependence of lipid organization and protein and nucleic acid-based condensation.

Outlook on biomolecular condensates and membrane surfaces

While it is clear that membrane surfaces play an essential role in regulating phase separation, substantial work remains to understand how membranes control condensate assembly throughout the cell. Specifically, it is unclear if membranes simply reduce the concentration threshold to phase separation, or if specific factors that nucleate condensation are also present on membrane surfaces. Cells may control the organization of these factors on endomembrane networks in time and space in order to regulate when and where different condensates are assembled.

Moreover, it will be important to examine the specific mechanisms by which proteins and nucleic acids are recruited to endomembrane surfaces prior to phase separation. For example, some membrane-associated condensates are attached to membranes via resident, transmembrane proteins, as is the case for the LAT and nephrin-based networks. However, other condensates may assemble at membrane surfaces by binding to the membrane peripherally, rather than via a transmembrane protein. This peripheral interaction with the membrane could be facilitated by specific, lipid-interacting domains (Lemmon, 2008). Other proteins may require binding to a separate, membrane-associated molecule. For example, some proteins may be recruited by binding to membrane-associated RNAs, while others may associate with resident membrane proteins.

Once recruited to membrane surfaces, how are biomolecular condensates organized and maintained by membrane-associated proteins? Intriguingly, a component of the protein quality control machinery that is anchored in endomembranes (Caplan et al., 1992) has been shown to influence the dynamics and functions of different condensates (Lee et al., 2015; Walters et al., 2015). These findings suggest that cells may utilize active, energy-consuming protein systems to regulate condensate properties and identities. In the next section, we discuss active cellular mechanisms that regulate biomolecular phase separation.

Regulation of biomolecular condensates by active processes

Once phase separation occurs, cells must maintain and remodel biomolecular condensates to meet specific cellular requirements. How do cells actively modify condensates to satisfy different needs? Moreover, how do cells ensure that different, functionally-distinct condensates remain discrete and do not merge? Active, energy-consuming mechanisms are likely required to maintain the function and identities of condensates.

Previous work has revealed that cells utilize active processes to regulate the material properties of the cytoplasm (Parry et al., 2014). Specifically, ATP and GTP-consuming processes dramatically influence the diffusive motion of organelles and other particles by maintaining the cytoplasm in a fluid-like state. Similarly, the properties of biomolecular condensates are also tuned and modified by active cellular processes. In particular, work has shown that depletion of ATP can lead to dramatic changes in the physical and functional properties of phase-separated structures (Brangwynne et al., 2011; Feric et al., 2016; Jain et al., 2016). Biomolecular condensates may therefore be considered a type of active matter, in which energy is consumed to physically rearrange and restructure the molecular components to maintain a functional state. In the next sections, we discuss active regulation of biomolecular condensates by molecular chaperones and RNA helicases. Importantly, we distinguish these active molecules from others that chemically modify proteins and nucleic acids, such as enzymes that post-translationally modify proteins. Though post-translational modifications represent an important category of active regulation, we provide a separate discussion of such processes later in this review.

Regulation of condensates by molecular chaperones

Molecular chaperones are a key class of proteins that actively regulate cellular biochemistry. Chaperones bind diverse proteins and consume energy to remodel protein-protein interactions and assist in protein folding (Akerfelt et al., 2010; Tyedmers et al., 2010). While chaperones are well-known regulators of protein quality control and aggregation prevention, the role of chaperones in regulating the material properties and functions of biomolecular condensates is only beginning to be understood.

Stress granules are the best-studied biomolecular condensates regulated by molecular chaperones. Stress granules sequester protein and RNA following cellular stress, and display the hallmarks of liquid-like condensates (Kedersha et al., 2013). Several studies have shown that molecular chaperone components deposit with stress granules and regulate granule dynamics, material properties, and disassembly during stress recovery (Cherkasov et al., 2013; Jain et al., 2016; Kroschwald et al., 2015; Mateju et al., 2017; Wallace et al., 2015; Walters et al., 2015). Importantly, stress granule components are not targeted for degradation and remain functional following chaperone-mediated dissolution (Cherkasov et al., 2013; Wallace et al., 2015). Molecular chaperones thereby help to modify and restructure stress granules such that the components remain active and functional after stress recovery. Moreover, a study found that the formation of an aberrant, solid-like stress granule promoted by a disease-linked protein is inhibited by molecular chaperones (Mateju et al., 2017). This finding indicates that chaperones serve a protective role by preventing the formation of pathological aggregates that form from misregulated phase transitions.

While studies on the role of chaperones in regulating phase transitions have primarily focused on stress granules, chaperones likely regulate phase separation in other contexts. For example, chaperone components have been implicated in modifying ribonucleoprotein granules that regulate early development (Hubstenberger et al., 2015). Another recent report found that transportin acts as a chaperone that actively regulates FUS phase separation in neuron terminals (Qamar et al., 2018). Although transportin is unique from the traditional chaperones involved in protein quality control, this finding highlights that cells may utilize diverse, active mechanisms to regulate the material properties and functions of condensates throughout the cell.

The studies cited above primarily indicate that chaperones promote condensate dissolution. Specifically, chaperones appear to act as binary switches that turn phase separation off. However, chaperone systems may function more subtly in cells to tune and modify the properties and functions of condensates to meet changing requirements. For example, chaperones may adjust the porosity of condensates to promote more rapid exchange of material with the surrounding medium. Chaperones may thereby act as “rheostats” that actively tune and adjust biomolecular condensates. The biophysical influence of chaperones on condensate material properties, and the functional consequences of chaperone-mediated remodeling in cells, are topics for further investigation.

Regulation of condensates by RNA helicases

Finally, helicases are another, essential class of energy-consuming proteins that actively regulate biomolecular condensates. Specifically, the DEAD-box RNA helicases, which are key regulators of nearly all processes involving RNA (Cordin et al., 2006), are the best-studied helicases involved in regulating phase separation. Similar to molecular chaperone components, DEAD-box RNA helicases have been shown to localize to stress granules (Jain et al., 2016), and are suggested to regulate the dynamics and disassembly of stress granules and other ribonucleoprotein condensates (Hilliker et al., 2011; Hubstenberger et al., 2013; Kim and Myong, 2016; Mugler et al., 2016). However, the mechanisms by which helicases control the material states and emergent functions of biomolecular condensates are only beginning to be understood.

Molecular chaperones and helicases facilitate powerful control of phase separation by physically altering molecular structures and intermolecular interactions. However, other active enzymes can also introduce chemical modifications that alter the dynamics and architecture of biomolecular condensates. Specifically, protein post-translational modifications (PTMs) facilitate a wide variety of alterations to condensate properties and functions, many of which are rapid and reversible. We next discuss our growing understanding of the role of PTMs in regulating biomolecular phase separation.

Regulation of phase separation by post-translational modifications

PTMs are well-recognized as key regulators of the assembly and properties of biomolecular condensates (Bah and Forman-Kay, 2016; Hofweber and Dormann, 2019). Specifically, PTMs can modulate protein valency and interaction strength to either promote or oppose phase separation in different contexts, thereby tuning the emergent properties and overall functions of condensates (Brangwynne et al., 2015). Here we discuss key examples of PTMs that regulate phase separation, and suggest future research avenues that will clarify how PTMs control the assembly and properties of condensates.

Phosphorylation is one of the most well-characterized PTMs that influence biomolecular phase transitions (Aumiller and Keating, 2016). Because this PTM is rapid and reversible, phosphorylation can serve as a mechanism for quickly regulating and tuning phase separation in response to different cues. Phosphorylation can either promote or suppress phase separation in different contexts (Fig. 3) (Hofweber and Dormann, 2019). For example, the kinase DYRK3 has been shown to promote the dissolution of stress granules, revealing the importance of phosphorylation in regulating the cellular stress response (Wippich et al., 2013). More recent work found that DYRK3 acts as a “dissolvase” that prevents the formation of several different condensates during mitosis, suggesting that phosphorylation serves a general function of inhibiting phase separation at key stages of growth and division (Rai et al., 2018). More specifically, phosphorylation is a key suppressor of phase separation in neuronal proteins such as FUS (Monahan et al., 2017), CPEB4 (Guillen-Boixet et al., 2016), and TDP-43 (Wang et al., 2018a) (Fig. 3A). In particular, phosphorylation disrupts electrostatic interactions within low-complexity, unstructured domains in FUS and CPEB4 (Guillen-Boixet et al., 2016; Monahan et al., 2017). In the case of TDP-43, a single phosphomimetic mutation in the well-folded N-terminal domain showed reduced phase separation, indicating that phosphorylation can also disrupt interactions among globular domains. However, the unstructured, C-terminal domain of TDP-43 is also multiphosphorylated in disease states, and it remains unclear whether phosphorylation of this domain enhances or suppresses phase separation of TDP-43 (Hofweber and Dormann, 2019).

Phosphorylation can also promote phase separation in other cases. An important example is the microtubule-binding protein Tau, which is hyperphosphorylated in disease states (Kopke et al., 1993). Recent studies found that Tau phosphorylation alters the net charge and the charge distribution to promote electrostatic interactions that favor phase separation (Fig. 3B) (Ambadipudi et al., 2017; Wegmann et al., 2018), thereby helping to explain the formation of disease-linked Tau aggregates. Phosphorylation can also promote phase separation by increasing the valency of binding with other proteins (Fig. 3C). Notable examples include the transmembrane proteins nephrin and LAT, which become phosphorylated on multiple tyrosine residues in response to extracellular stimuli (Banjade and Rosen, 2014; Huang et al., 2016; Li et al., 2012; Su et al., 2016). The nephrin and LAT binding partners Nck and Grb2, respectively, can then bind to these phosphotyrosines via SH2 domains, facilitating the formation of a multivalent, phase-separated network. Decreasing the number of phosphotyrosines in nephrin and LAT inhibits phase separation of these proteins, revealing the importance of phosphorylation in modulating multivalent interactions (Banjade and Rosen, 2014; Li et al., 2012; Su et al., 2016). Collectively, these studies reveal that phosphorylation can promote or inhibit phase separation via diverse mechanisms, highlighting the importance of examining the specific regulatory effects of PTMs on different condensates.

Glycosylation is another essential PTM that plays a role in regulating phase separation. Studies have revealed that O-linked N-acetylglucosaminylation (O-Glc-NAc) glycosylation inhibits the formation of aggregates of the proteins Tau, hnRNPA1, and α-synuclein (Marotta et al., 2015; Roth and Khalaila, 2017; Wang et al., 2016; Yuzwa et al., 2012). Importantly, O-Glc-NAc modifies serine and threonine residues, where phosphorylation also occurs. Thus, there is likely an interplay between these two PTMs in regulating phase separation. Future studies will help to characterize the biophysical influence of the O-Glc-NAc modification on the properties and functions of condensates, and to examine how phosphorylation and O-Glc-NAc modifications within the same protein differentially regulate phase separation. More broadly, other types of glycosylation modifications may regulate phase separation in unknown ways. For example, large sugar chain modifications may regulate phase separation by sterically hindering interactions within a phase-separated network. The influence of bulky sugar chains on biomolecular condensates remains to be explored, but may be a versatile source of specific, tunable biophysical properties.

A variety of other PTMs regulate phase separation by disrupting cation-π interactions that are thought to be relevant for formation of many biomolecular condensates (Nott et al., 2015; Wang et al., 2018b). For example, arginine methylation (Lorton and Shechter, 2019) is a key PTM that inhibits cation-π interactions in several phase-separating RNA binding proteins (Fig. 3A) (Hofweber and Dormann, 2019; Wang et al., 2018b). Examples include FUS (Hofweber et al., 2018; Qamar et al., 2018), the RNA helicase Ddx4 (Nott et al., 2015), and hnRNPA2 (Ryan et al., 2018). While arginine methylation has been demonstrated to suppress phase separation, it remains unclear whether methylation may promote phase separation via other mechanisms (Hofweber and Dormann, 2019).

Similar to methylation, arginine citrullination inhibits cation-π interactions to oppose phase separation of RNA-binding proteins like FUS (Fig. 3A) (Qamar et al., 2018). Specifically, conversion of arginine to citrulline by the protein arginine deiminase enzyme family neutralizes a positive charge. A recent report found that a protein arginine deiminase family member inhibits the phase separation of several ALS-related proteins, including FUS, supporting a role for citrullination in regulating phase separation and protecting against the formation of pathological aggregates (Tanikawa et al., 2018). While studies on the role of citrullination in controlling phase separation have primarily focused on ALS-related proteins, the role of this modification in regulating other proteins remains to be investigated.

Lysine acetylation can also oppose phase separation by neutralizing cationic amino acids and disrupting cation-π interactions (Fig. 3A) (Saito et al., 2019). For example, recent work found that acetylation inhibits phase separation of Tau (Carlomagno et al., 2017; Ferreon et al., 2018), suggesting that cells may utilize this modification to protect against the formation of pathological Tau aggregates. Another recent study found that acetylation disrupts phase separation of a stress granule protein (Saito et al., 2019). Importantly, this study found that the degree of acetylation strongly affects the material properties of condensates, revealing that altering the number of acetyl modifications can progressively modify droplet properties (Saito et al., 2019). This finding suggests that any modification which disrupts cation-π interactions may also tune droplet properties in a manner dependent on the number of modifications. Precise biophysical studies will help to characterize how the number and type of modifications on a protein regulates the properties of the resulting condensate.

While lysine acetylation inhibits phase separation of some proteins, this modification can also promote phase separation in other cases. For example, acetylation of the protein TDP-43 in the RNA-binding domain disrupts protein-RNA interactions, promoting the formation of TDP-43 aggregates (Cohen et al., 2015). Thus, acetylation can have differing effects on phase separation in a context dependent manner. It is therefore essential to examine the location on a protein where the modification occurs in order to understand the type of interaction that is affected.

The studies discussed so far have focused on PTMs that modify “scaffold” proteins, which are the proteins that drive phase separation (Banani et al., 2017; Banani et al., 2016). However, “client” proteins, which partition into condensates but are not the primary drivers of phase separation, can also strongly influence the properties and functions of condensates (Banani et al., 2017; Banani et al., 2016). Recent reports found that PTMs on certain client proteins can strongly influence phase separation. In one example, a ubiquilin protein involved in targeting ubiquitinated proteins for degradation was found to undergo phase separation under cellular stress (Dao et al., 2018). However, binding to ubiquitinated client substrates opposed phase separation. This result thereby reveals how ubiquilin becomes solubilized in order to transport ubiquitinated substrates from condensates into the proteasome for degradation. In contrast, another study found that the protein TDP-43 binds client substrates modified with poly(ADP-ribose) (PAR), and that binding to PAR promotes phase separation of TDP-43 (McGurk et al., 2018). Moreover, PAR binding also drives TDP-43 accumulation in stress granules, which protects TDP-43 from undergoing disease-associated hyperphosphorylation. Collectively, these findings suggest that client PTMs can inhibit or promote phase separation in different contexts, thereby providing another level of additional, powerful control over biomolecular phase separation.

Collectively, the studies discussed here reveal that PTMs act as a complex regulatory network over biomolecular phase separation. While it is clear that disrupting certain interactions with PTMs can strongly influence whether phase separation will occur or not, substantial work remains to understand how these modifications influence the properties and identities of condensates. Specifically, PTMs do not necessarily act as binary switches that turn phase separation on or off. Rather, these modifications can serve as rheostats that fine-tune the material state. In particular, cells may control the number or position of different protein modifications in order to gradually modify a phase-separating network, thereby progressively altering condensate material properties. By finely regulating these properties, cells may thereby control the emergent functions of condensates in a context-dependent manner. Future work is needed to better understand how protein modifications precisely regulate biomolecular condensates throughout the cell.

Conclusions

Here we have reviewed three essential mechanisms by which cells regulate biomolecular phase separation in time and space: membrane surfaces, PTMs, and active molecules such as chaperones and helicases. All of the mechanisms discussed here likely work in concert to control phase separation. For example, the phosphorylation PTM controls the valency of the nephrin and LAT protein networks that phase separate at membrane surfaces. Therefore, cells must utilize PTMs to regulate condensate assembly at membranes. In line with this thinking, lipid PTMs on proteins, such as farnesyl or myristoyl groups, may regulate phase separation by providing a tether between condensates and cellular membranes. The interplay between PTMs and membrane surfaces in regulating phase separation is an exciting topic for further investigation. PTMs may also coordinate with chaperone complexes and helicases to regulate condensates. Specifically, PTMs may influence whether active molecules are able to access and remodel interactions within a phase-separated system. Importantly, this potential function of PTMs could be useful in cases of cellular stress, in which condensates may become more quiescent and energy is conserved. Finally, some molecular chaperone components are anchored in endomembranes (Caplan et al., 1992), suggesting that phase separation at membrane surfaces may also be coupled to active remodeling by resident chaperone complexes. How membrane surfaces and active mechanisms cooperate to drive phase separation is unknown. Collectively, these examples suggest how several regulatory platforms may function in concert to control condensate assembly and properties. Regulatory input at multiple, independent levels may help provide robust control over the material states and identities of biomolecular condensates.

Finally, a variety of mechanisms beyond those discussed here help to regulate biomolecular phase transitions. For example, RNA itself is a key driver of phase separation, and can act as a scaffold for physiological condensate assembly (Shevtsov and Dundr, 2011) and drive the formation of pathological aggregates (Jain and Vale, 2017). Moreover, RNA modifications are critical regulators of phase separation (Drino and Schaefer, 2018; Ries et al., 2019). The interplay between the regulatory mechanisms discussed here and others is an important topic for future study.

There is a growing understanding that the ability to undergo phase separation is an inherent feature of many proteins and nucleic acids throughout the cell. However, the mechanisms by which cells control this process to produce functional, phase-separated structures are only beginning to be uncovered. Fortunately, a substantial body of work has revealed some of the essential cellular platforms that regulate biomolecular phase transitions. An important task for future research will be to elucidate how these diverse and complex regulatory systems coordinate to ensure robust phase separation, and to prevent the formation of pathological aggregates that cause disease.