Considerations and challenges in studying liquid-liquid phase separation and biomolecular condensates (original) (raw)

. Author manuscript; available in PMC: 2020 Jan 24.

Abstract

Evidence is now mounting that liquid-liquid phase separation (LLPS) underlies the formation of membraneless compartments in cells. This realization has motivated major efforts to delineate the function of such biomolecular condensates in normal cells and their roles in contexts ranging from development to age-related disease. There is great interest in understanding the underlying biophysical principles and the specific properties of biological condensates with the goal of bringing insights into a wide range of biological processes and systems. The explosion of physiological and pathological contexts involving LLPS requires clear standards for their study. Here, we propose guidelines for rigorous experimental characterization of LLPS processes in vitro and in cells, discuss the caveats of common experimental approaches, and point out experimental and theoretical gaps in the field.

Introduction

Membraneless compartments such as the nucleolus were described as early as the 1830s (Valentin, 1836; Wagner, 1835). Evidence is now mounting that liquid-liquid phase separation (LLPS) or condensation underlies the formation of membraneless bodies such as nucleoli in cells. The list of cell compartments thought to be formed via the process of LLPS is growing rapidly and touches a myriad of cell functions. In addition to punctate membraneless bodies, other subcellular structures are also formed via LLPS and share similar underlying interactions and physical properties. These structures include heterochromatin (Larson et al., 2017; Strom et al., 2017), the transport channel in the nuclear pore complex (Schmidt and Görlich, 2016), and membrane receptor clusters at the cell membrane (Su et al., 2016). Compartments that form through condensation have recently been termed “biomolecular condensates” (Banani et al., 2017; Shin and Brangwynne, 2017). The realization that LLPS can drive compartmentalization has expanded our understanding of cell biology and has motivated major efforts to delineate the function of membraneless compartments.

Through study of these condensates, we have also gained insights into the molecular basis of disease. Early examples of work in this area focused on the influence of stress granules on cell survival and their link to the development of amyotrophic lateral sclerosis (ALS) (Bentmann et al., 2012; Kim et al., 2013; Li et al., 2013; Ramaswami et al., 2013). Stress granule proteins form liquid condensates that can mature and solidify, i.e. they can undergo liquid-to-solid transitions, or can nucleate aggregation (Lin et al., 2015; Molliex et al., 2015; Murakami et al., 2015; Patel et al., 2015) and disease-associated mutations accelerate this transition (Kato et al., 2012; Kwon et al., 2013; Molliex et al., 2015; Patel et al., 2015). A connection between cataract development and LLPS was made over three decades ago (Tanaka et al., 1977) and, with the growing general realization that LLPS may favor aggregation, we can now gain more detailed insight into the molecular mechanisms of age-related disease initiation and progression (Alberti and Carra, 2018). Thus, pathological processes related to neurodegenerative diseases and aging are also being viewed anew as manifestations of LLPS-driven processes. It is increasingly clear that understanding the biophysical principles underlying the formation of biomolecular condensates is vital for investigation of the physiology and pathophysiology of a wide range of biological processes and systems.

With the rapid identification of different membraneless compartments and proteins linked to LLPS comes the responsibility for the rigorous characterization of biological phase separation. Properties exhibited by a protein or set of proteins in vitro cannot always be extrapolated to functionally-relevant LLPS in cells. Care should be taken to ask the question whether an observed phase separation process observed in vitro is relevant to the biological process under study. Herein, we propose guidelines for rigorous experimental characterization of LLPS processes in vitro and in cells, discuss the caveats of common experimental approaches, and point out experimental and theoretical gaps in the field. We consider in vitro assays as those in which minimal components are combined in buffers to recapitulate LLPS and in vivo assays as studies performed in cells or animals. Importantly, many of the suggestions stem from our own experiences working with phase-separating systems in our laboratories. The goal, therefore, is to provide a practical primer for studying LLPS and not a full description of the concepts of phase separation. For a more detailed description of the biophysics and an exhaustive coverage of the available experimental and theoretical approaches we will refer the interested reader to published literature.

What is liquid-liquid phase separation?

When solutions of macromolecules such as proteins or nucleic acids undergo LLPS, they condense into a dense phase that often resembles liquid droplets, and this dense phase coexists with a dilute phase (Figure 1A, bottom). The driving force underlying LLPS is the exchange of macromolecule/water interactions for macromolecule/macromolecule and water/water interactions under conditions for which this process is energetically favorable. Discussions of the thermodynamic driving forces of LLPS and a framework for how the multivalent interactions of polymers influence the process have been presented elsewhere (Banani et al., 2017; Harmon et al., 2017; Martin and Mittag, 2018; Ruff et al., 2018). Importantly, whether a solution undergoes phase separation depends strongly on the concentration and identities of the macromolecules and the solution, and the environmental conditions including temperature, salt type and concentration, co-solutes, pH and the volume excluded by other macromolecules. Macromolecules thus undergo stimulus-responsive phase separation (Franzmann and Alberti, 2018; Nott et al., 2015; Ruff et al., 2018).

Figure 1: Schematic phase diagram.

The coexistence line (black) separates the one-phase and two-phase regime and is a function of environmental conditions such as temperature, pH etc. The system does not undergo phase separation beyond the critical point. (A) At concentrations below csat, the system is in the one-phase regime. At any condition within the two-phase regime, the system demixes into a light phase (with c=cL) and a dense phase (with c=cD). All conditions on a single tie line (the orange line is an example) result in two phase systems with fixed light phase and dense phase concentrations, cL and cD, respectively; only the volume fractions of the two phases, fL and fD, change relatively to each other (examples 2 – 4). The volume fractions resulting from demixing of condition 3 can be calculated by the lever rule and are fL = D/T (i.e. the ratio of the lengths of D and T) for the light phase and fD = L/T for the dense phase. In equilibrium, csat and cL are equivalent, but when phase separation is nucleated, csat and cL can differ during the ripening dynamics of the system. (B) The spinodal (grey line) indicates the region of instability in which the system must undergo demixing via spinodal decomposition. In the area between the coexistence line, or binodal, and the spinodal, the system demixes when nucleated.

Phase diagrams are generated by experiments that define the set of conditions that result in a single, well-mixed phase and the conditions that promote phase separation (Figure 1). The generation of a phase diagram involves systematically changing two conditions, for example concentration and salt, and assessing in which conditions a dense phase is detectable. If a system is in the two-phase regime, i.e. if light and dense phases are coexisting, so-called “tie lines” result from connecting the light phase and dense phase concentrations under the given conditions. All conditions on a single tie line result in two-phase systems with fixed dilute and dense phase concentrations; only the volume fractions of the two phases change relatively to each other (Figure 1A, bottom). Within the coexistence line, or binodal, a second line, the spinodal exists (Figure 1B). The spinodal indicates the region of instability in which the system must undergo demixing via spinodal decomposition. In the area between the binodal and the spinodal, the system demixes when nucleated (Figure 1B, bottom). Generating phase diagrams can provide powerful insights such as how valency and chemical properties of molecules can modulate phase separation and whether phase separation can occur in physiologically relevant contexts. It is important to note that the simplified density transitions captured in phase diagrams may not represent what happens in the complex environment of cells. The density of molecules may not change because of the high concentration of macromolecules in cells, but instead only their spatial distribution may change.

Liquids, solids and gels can all emerge from LLPS.

It is increasingly clear that LLPS can produce assemblies of varied material states in cells. If the assemblies resulting from LLPS are liquid, this implies long-range molecular disorder and only short-range order, i.e. within the first few layers of molecules, their arrangement is regular but further out, the variation in molecular orientation and distances increases (Shin and Brangwynne, 2017). Liquid assemblies can fuse, coalesce, and drip, which are typical emergent properties of liquids and determined by their surface tension (Widom, 1988). If the dense phases have liquid-like properties, polymer molecules are often mobile within the dense phase and between the dense and light phases (for a more detailed description of the biophysics of liquids see (Hyman et al., 2014)). The formation of the condensed phases through LLPS is often reversible. However, they can also undergo further transitions, e.g. gel-transitions, in particular if they exist at conditions deep within the two-phase regime, far away from the binodal coexistence line (i.e. at high quench depth) (Shin and Brangwynne, 2017). They can further transition to hydrogels formed by amyloid-like fibers, which often form irreversibly or need a change in condition such as high salt or denaturant to dissolve (Kato et al., 2012; Murray et al., 2017). The driving force for these transitions is likely sequence-encoded and there is emerging evidence that individual protein sequences have evolved to use liquid-to-solid transitions for function. One example is the yeast protein Pab1, which undergoes LLPS in heat-shocked cells and forms non-dynamic, glassy assemblies that may be adaptive (Riback et al., 2017). Given these possible transitions, LLPS does not have to result in liquid, freely-fusing assemblies; instead, biological molecules can adopt a continuum of material properties (Alberti and Hyman, 2016; Weber and Brangwynne, 2012).

Predicting and analyzing the features of molecules that undergo LLPS

The ability to undergo LLPS may be a universal property of proteins and nucleic acids under specific conditions, many of which may never be encountered in a normal cell. LLPS, in this way, resembles the formation of amyloid, which is a generic state of proteins (Knowles et al., 2014). Importantly, only a small subset of proteins is able to form amyloids under physiological conditions, and these particular amyloid-forming proteins are highly relevant in physiological as well as pathological contexts (Fowler et al., 2007). In the same manner, LLPS may not be accessible to many proteins under physiological conditions and only particular protein sequences appear to have the ability to phase separate under the conditions that exist in living cells. Currently, our ability to identify genuine and biologically relevant LLPS is still limited and this should make us careful in interpreting the results from phase separation studies performed in vitro.

Despite these limitations, tremendous progress has been made in recent years in understanding the molecular signatures common in molecules that can phase separate in physiological conditions. One concept that has been very useful is the concept of scaffolds and clients (Banani et al., 2016). Scaffold molecules are considered the drivers of phase separation, whereas molecules that partition into condensates formed by scaffolds are called clients. Phase separation of scaffold proteins and the partitioning of clients is now appreciated to require the formation of a network of interactions, often between proteins and frequently between proteins and RNA. Two archetypes of protein architectures promote the formation of such networks. One type is characterized by multiple folded domains (e.g., SH3 domains in Nck) which interact with short linear motifs (SLiMs) in other proteins (e.g., proline-rich motifs in N-WASP) (Li et al., 2012). A second type of weak multivalent interaction that can mediate LLPS is characterized by the presence of intrinsically disordered regions (IDR) with multiple interacting motifs, or “stickers” (Elbaum-Garfinkle et al., 2015; Nott et al., 2015; Smith et al., 2016; Wang et al., 2018).

Both archetypes have multivalency in common, i.e., the proteins interact through multiple interacting domains or motifs. Experiments to genetically manipulate the valency of a protein proved to be very insightful in defining the domains and motifs that drive protein phase separation and showed that the saturation concentration csat, i.e. the concentration above which the system starts to phase separate, decreases dramatically with an increase in valency (Banani et al., 2016; Li et al., 2012; Nott et al., 2015; Wang et al., 2018). Where specific RNAs have been studied in driving LLPS, an additional source of multivalency in the protein-RNA interactions has been found (Langdon and Gladfelter, 2018; Langdon et al., 2018; Lee et al., 2015; Zhang et al., 2015). Many IDR-containing proteins harbor multiple domains for interacting with RNAs and target RNAs contain multiple possible binding sites for the protein. Thus, there are many routes to forming multivalent interactions and these interactions underlie the ability of particular proteins and nucleic acids to undergo LLPS in the constraints of the chemistry of live cells.

How condensates can emerge from a network of multivalent domain/motif interactions is easy to understand because the molecular basis of these multi-point interactions is well understood from high-resolution structures. However, how IDRs mediate LLPS is less well understood and far less intuitive, and thus warrants a brief description of the current state of the art. IDRs are a type of protein domain that is frequently found in phase-separating proteins (Elbaum-Garfinkle et al., 2015; Li et al., 2012; Lin et al., 2015; Nott et al., 2015; Patel et al., 2015). IDRs often do not have many aromatic and aliphatic amino acids, which typically build the core of folded domains, and do not adopt a single folded structure that corresponds to a single low energy state. Instead, these proteins sample a range of conformations with similar energies that are determined by the specific primary sequence of the IDR (Das et al., 2015; Jensen et al., 2013; Mittag and Forman-Kay, 2007). The primary sequence also determines the phase behavior of these IDRs. Our understanding of the sequence determinants of phase separation in IDRs is still rudimentary, but it is clear that different flavors of IDRs exist that determine the type of stimulus the IDR responds to (Franzmann and Alberti, 2018; Martin and Mittag, 2018; Ruff et al., 2018). The sequence also likely determines the emergent properties of its dense phase, i.e., dense phase concentration (Wei et al., 2017) and material properties such as viscoelasticity (Weber, 2017).

Sequence variations that fine-tune the phase behavior are thought to include the IDR length, the number, patterning and type of stickers, and the identity of sequences connecting stickers, i.e., so-called “linkers” or spacers (Harmon et al., 2017; Wang et al., 2018). Typical determinants of physico-chemical and conformational properties include the fraction, patterning and identity of hydrophobic residues. While hydrophobic residues are rare in average IDRs, they represent adhesive elements in phase-separating IDRs and mediate condensation upon changes in temperature (Lin et al., 2017; Pak et al., 2016; Urry et al., 1992). Additional determinants of physico-chemical and conformational properties are the fraction and patterning of charged residues (Das and Pappu, 2013; Mao et al., 2010; Marsh and Forman-Kay, 2010; Müller-Späth et al., 2010; Wei et al., 2017). Pairs of highly, but oppositely charged proteins (or proteins and nucleic acids) can condense together in a process called complex coacervation (Aumiller and Keating, 2016; Overbeek and Voorn, 1957; Pak et al., 2016).

One feature shared by some IDRs is that they are composed of low-complexity sequence regions (LCRs), i.e. regions in which specific amino acids are overrepresented compared to the amino acid proportions found in the proteome. One of the most common types of LCRs are prion-like LCRs. Prion-like LCRs are largely devoid of charged residues and enriched for polar amino acid residues, such as serine, tyrosine, glutamine or asparagine allowing their identification by composition-based algorithms (Alberti et al., 2009). Prion-like proteins were first described in connection with infectious aggregation-prone proteins called prions. However, recent findings suggest that prion-like LCRs are also frequently involved in protein phase separation (Franzmann and Alberti, 2018; Molliex et al., 2015; Patel et al., 2015). This suggests that algorithms to detect prion proteins not only identify aggregation-prone proteins, but primarily proteins that form condensates via phase separation. A slightly different type of LCR are the RGG domains that frequently occur in RNA-binding proteins (Chong et al., 2018). In addition to the small polar residues of prion-like LCRs, they contain a significant proportion of arginines, and the RGG boxes after which they are named mediate LCR/RNA interactions. The sequence features of prion-like domains and other LCRs are thought to promote weak interactions through charge-charge, pi-pi and cation-pi interactions. The structural basis of these interactions is currently debated.

Generally, it is useful to identify disordered regions in proteins with predictive algorithms (see Box 1 and Figure 2), to analyze their physico-chemical properties and to use the results of these analyses to generate hypotheses as to the physical origin of the phase behavior. Taking FUS as an example, we illustrate the use of these predictive algorithms in Figure 2. The domain architecture of FUS is known and shown for reference (Figure 2, top). Using a combination of IUPred and PLAAC (or a different set of disorder predictors from Box 1), we identify the first ~250 residues as disordered. This prediction is in agreement with the first folded domain, the RRM, starting at residue 287. Additional disordered regions are identified on the left and right side of the folded zinc finger motif between residues ~365 and 420 and from residue 450 to the C-terminus. PLAAC identifies the QGSY- and G-rich regions in the N-terminus as well as the first RGG domains as prion-like LCRs. CIDER (Box 1) reveals a lack of charged residues in the N-terminus (as low fraction of charged residues (FCR) and net charge per residue (NCPR)) and a high fraction of positively charged arginines in the RGG domains (as a slightly positive NCPR and non-zero FCR). As CIDER and PLAAC cannot differentiate between folded and disordered domains, the proper demarcation of folded and disordered domains is thus critical, and additional information about folded domains can be found in various databases, e.g. in D2P2 (Box 1). From this sequence analysis, we would predict that the N-terminal QGSY- and G-rich domains as well as the RGG domains may mediate LLPS, but that FUS condensation may involve different types of amino acid motifs.

Box 1: Sequence analysis and prediction of phase separation properties.

General sequence analysis tools

UNIPROT (https://www.uniprot.org/): Information on sequence, functional annotation, tissue/ subcellular localization and disease association.

BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi): Identify homologs and orthologs.

ProtParam (https://web.expasy.org/protparam): Calculate basic protein properties including molecular weight, pI, and amino acid composition; can be applied to domains.

CIDER (http://pappulab.wustl.edu/CIDER/): Calculation of distribution and patterning of charged and hydrophobic residues to classify disordered sequences.

Disorder predictors

We recommend using Meta-predictors, which integrate results from several individual predictors, and screen for known folded domains to reduce the possibility of false positives or negatives.

MobiDB (http://mobidb.bio.unipd.it/about): Combines annotations from external databases, indirect evidence from experimental data and predictions to provide an overall picture of folded domains, disordered regions, secondary structure, and regions with low sequence complexity.

D2P2 (http://d2p2.pro/): Database that aggregates predictions of disorder and binding sites, and information on Pfam domains and PTMs.

DisMeta (http://www-nmr.cabm.rutgers.edu/bioinformatics/disorder/): Meta-predictor of protein disorder.

Phase separation predictors

Pi-Pi predictor (https://cdn.elifesciences.org/articles/31486/elife-31486-supp-v1.zip): Predicts the driving force of pi-pi contacts for phase separation in a given sequence.

PLAAC (http://plaac.wi.mit.edu/): Predictor of prion-like domains.

ZipperDB (https://services.mbi.ucla.edu/zipperdb/): Predicts fibril-forming segments in proteins, which have been argued to contribute to phase separation or can mediate maturation from liquid to solid states.

Figure 2: Bioinformatic analysis of the amino acid sequence of FUS to identify protein regions that are involved in phase separation.

Schematic representation of the human FUS/TLS domain structure is shown on top. QGSY: region enriched for the residues glutamine (Q), glycine (G), serine (S) and tyrosine (Y) depicted in green, G-rich: region enriched for glycine residues depicted in blue, NES: nuclear export sequence depicted in magenta, RRM: RNA recognition motif depicted in yellow, RGG: region enriched for residues arginine and glycine depicted in orange, ZN: Zinc finger domain depicted in purple, NLS: nuclear localization sequence depicted light blue. IUPred: Prediction of intrinsic disorder, PLD: Prediction of prion-like region (PLAAC), FOLD: Intrinsic disorder prediction by PLAAC (black) and PAPA (purple). Fold index is shown in gray. Pi-Pi: Pi interaction prediction NCPR: Net charge per residue (sliding window size of 10), FCR: Fraction of charged residues, HYDRO: hydrophobicity (Kyte & Doolittle, sliding Windows size of 9).

Hypotheses based on sequence analyses then have to be tested experimentally. Indeed, experimental analysis of FUS phase behavior demonstrated the importance of the disordered regions and aromatic and arginine-rich motifs within them in driving phase separation of FUS (Wang et al., 2018). Not only does this strategy of sequence-directed experimental analysis provide insight into the phase behavior of the protein in question, but it also serves to build our knowledge base of sequence/phase behavior relationships and enables further development of predictors. A key unsolved issue in biological phase separation is the degree to which particular material properties have been selected for in evolution; therefore, intensive work is needed to link primary sequence features, emergent physical properties of condensates and their functions.

Reconstitution of LLPS with minimal components

The step between a careful analysis of protein sequence and the unequivocal demonstration that a cellular structure forms via LLPS is challenging. One step in the process of investigating biological condensation is to reconstitute a simplified assembly of the critical components in vitro and to test whether the assembly forms via LLPS. If the driving force for assembly can be identified (e.g. a protein interaction via a specific interface), then removal of this driving force (e.g. by mutating the interface) will suppress LLPS in vitro. Use of these same mutants then allows testing in cells of whether the assemblies are formed via the same interactions and hence LLPS. This in vitro strategy has the inherent advantage that it is carried out with purified components, and one can test explicitly which components are critical for the process. A detailed collection of methods to characterize condensates forming via LLPS in vitro and in cells can be found in Mitrea et al (Mitrea et al., 2018).

How can one demonstrate that an assembly is formed via LLPS in vitro? A simple clue is the presence of spherical droplets that contain concentrated components (see Table 1 for a minimal set of criteria). This behavior indicates that the component is able to condense into droplets and that the droplets can fuse and relax into a spherical shape. A more definitive approach is to define a saturation concentration csat; where when c < csat, the protein is diffuse in solution; and for c > csat, dense droplets appear whose volume fraction fD increase when the concentration is increased further, while the concentration in the light phase stays constant (Figure 1A). This behavior is not expected for other types of self-association processes, e.g. dimerization, oligomerization into defined species or higher-order self-association without phase separation. A constant csat and a growing dense phase volume fraction fD with increasing total concentration are strong support for LLPS and can be tested microscopically via scattering or centrifugation. Addition of RNA to proteins that are thought to bind RNAs either specifically or non-specifically can substantially alter and often lower the csat for a protein (Elbaum-Garfinkle et al., 2015; Lin et al., 2015; Maharana et al., 2018; Molliex et al., 2015; Zhang et al., 2015). The direct observation of fusion processes is helpful for demonstrating liquid behavior of the dense phase, but is not a strict criterion because some condensed droplets undergo liquid-to-solid transitions, and therefore do not fuse. A constant csat and cD (concentration of the dense phase, see Figure 1A) can help identify LLPS as a transient state in the formation of rapidly solidifying condensates.

Table 1.

A minimum set of experiments to support identification and characterization of liquid-liquid phase separation in vitro and in cells. Not all systems will be amenable to all experiments.

Microscopic detection of LLPS:

The most common way to initially detect LLPS is light microscopy. To start, macromolecules are observed under conditions where no assemblies are observed. Then, with a change in conditions to favor assembly (e.g., by adding or diluting salt, changing the pH or temperature, adding RNA) followed by incubation of the sample for a fixed amount of time (minutes up to hours) the solution can be imaged microscopically for the presence of liquid droplets. When comparing different samples, the incubation time and imaging parameters (e.g., detection on the cover glass or immediately above the glass) should be held constant. The reason for this is that droplets settle on the surface of a microscope slide. This can lead to over- or underestimation of the volume fraction of the dense phase. Interaction with the glass surface can also lead to wetting or de-wetting of the surface and to changes in droplet material properties. For this reason, we advise comparing the behavior on glass slides with a variety of coatings, e.g., polyethylene glycol (PEG) or lipids (Alberti et al., 2018). Different microscopic modalities are suitable for imaging including differential interference contrast (DIC) or fluorescence microscopy. Use of fluorescently labeled components allows estimation of the concentration of the dense phase if standard curves for the fluorophore at different concentrations are established under identical imaging conditions. However, certain RNAs and RNA-binding proteins can be crosslinked by light so care should be taken in terms of the acquisition conditions in samples with these components to ensure there are not experimentally-induced gelation effects.

Detection of LLPS via turbidity measurements:

Mesoscale assemblies on the order of 10s to 100s of nanometers in diameter in solution scatter visible light and can be detected by optical density measurements (typically at wavelengths of 340 nm or 400 nm) or direct static light scattering. The absolute values of scattered light report on a convolution of number and size of scattering particles and thus do not lend themselves to a direct comparison of the driving force for phase separation of different protein constructs or mutants. Instead, this approach allows determination of the onset of scattering of a dilution series, and thus identification of the concentration above which droplets form, csat. Alternatively, the onset of scattering can be monitored as a function of temperature for a given concentration (Amiram et al., 2011; Lyons et al., 2013). One caveat here is that simple turbidity measurements detect a variety of assemblies and do not differentiate their shape, size or the mechanism underlying their formation. Turbidity measurements should thus be used in conjunction with microscopy.

Determine cL by centrifugation:

Centrifugation of systems in the two-phase regime and the associated separation of light and dense phases, has been shown to allow precise measurements of the concentrations of two coexisting phases, i.e. the light or dense phase concentrations cL and cD, as a function of temperature. For this approach, LLPS of the stock solution is induced by changing the solution conditions, the phase-separated sample is incubated for a defined time and the dense phase sedimented by centrifugation (Taratuta et al., 1990). An aliquot of the light phase is removed and its concentration determined spectroscopically (e.g., by measuring absorbance at 280 nm, in which case the solution should be diluted to prevent reformation of droplets, or with a colorimetric assay or fluorescence intensity when using a fluorophore-labeled protein). Notably, this assay is well suited to test for a constant cL if different protein stock concentrations are used (Elbaum-Garfinkle et al., 2015; Mackenzie et al., 2017; Wang et al., 2018), one of the defining properties of LLPS. It is noteworthy that processes such as aggregation or fibrillization that may compete with LLPS, may make it challenging to determine cL accurately. Thus, microscopic analysis that can distinguish liquid droplets from aggregates or fibers should always accompany other methods.

Practical notes, controls and limitations of in vitro reconstitution assays

The inherent advantage of in vitro phase separation assays is that LLPS can be monitored with a fully defined set of components. To exclude the possibility that contaminants are driving phase separation, the proteins or RNAs should be pure. Additionally, they should be homogeneous in size and not aggregated. We describe criteria for generating, storing and handling samples for in vitro phase separation assays.

Protein production considerations

Proteins should be heterologously overexpressed in E. coli, yeast or insect cells, or generated via in vitro transcription/translation assays. A typical BL21 strain is often sufficient for efficient expression, but many other strains and expression systems with different properties are available for optimization (Alberti et al., 2018; Sørensen and Mortensen, 2005).

Disordered protein regions, which are prevalent in phase-separating proteins, are sensitive to degradation. The IDR-containing proteins have to be purified as quickly as possible to limit the time available for degradation and aggregation, and initial purification steps should be carried out in the presence of protease inhibitors. Alternatively, protein expression can be optimized to maximize partitioning into inclusion bodies, which can be resolubilized and yield pure protein without exposing the protein to proteases of the expression host. Proteins purified under denaturing conditions have to be exchanged into a physiological buffer before carrying out phase separation assays, and care needs to be taken to avoid aggregation, which may mask or even potentiate phase separation. In some cases, a cleavable solubility tag can be helpful such as maltose-binding protein (MBP) that is cleaved prior to analysis of the protein (Burke et al., 2015).

Some phase-separating proteins such as FUS or TDP-43 are highly aggregation-prone when made from bacteria. One reason for this may be the absence of crucial post-translational modifications (PTMs) such as phosphorylation and methylation. Purification from insect cells is thus an attractive possibility because the samples incorporate the required PTMs and thus do not undergo aberrant protein behavior (Alberti et al., 2018). Use of bacterial expression systems bypasses most uncertainty around PTMs but when proteins are expressed from eukaryotic systems, it is strongly advised to determine the PTMs to better understand what the underlying driving forces are for LLPS. This can also improve our understanding of how cells may use PTMs to regulate LLPS. For example, phosphorylation and methylation of FUS and acetylation of DDX3X reduce the driving force for LLPS, respectively (Hofweber et al., 2018; Monahan et al., 2017; Qamar et al., 2018; Saito et al., 2019).

Phase separation should be avoided during the purification process if the protein tends to undergo liquid-to-solid transitions. However, if dense phases can be disassembled without loss of material, several cycles of phase separation and dilution can be used as a purification method (Quiroz and Chilkoti, 2015). The purity and integrity of protein samples should be assessed by SDS-PAGE and the identity by immunoblotting and/or mass spectrometry. Pure protein samples should be kept in a storage buffer that prevents phase separation and then should be flash-frozen. Whenever possible we recommend thawing a protein only once because we have observed that multiple freeze-thaw cycles can increase the aggregation propensity of a protein. The required protein stock buffer conditions depend on the protein sequence and phase behavior but often are a buffer with pH in the physiological range and either high or very low salt to prevent phase separation. A typical buffer will contain a buffering component (e.g., 50 mM HEPES pH 7.5), a salt component (e.g., 300 mM NaCl, 500 mM KCl, or even no salt) and a reducing agent (e.g., 1 mM TCEP, or 5 mM DTT). To determine whether phase separation occurs under physiological conditions, the salt concentration should be adjusted to 150 mM NaCl or KCl. We do not recommend the use of buffers with multivalent ions such as phosphate as these buffers can strongly influence the phase behavior of proteins. Optimization of buffer conditions is an important step to achieving stable protein behavior and reducing aggregation propensity. It is important to realize that as salt, pH or protein concentrations deviate from physiological conditions, caution should be used in interpretation of in vitro phase separation experiments. However, a dependence of the phase behavior on solution conditions provides information on the interactions mediating phase separation and can thus be useful. Dependence on solution conditions may also suggest stimulus responsiveness to changes in intracellular pH and PTMs and can thus be an important clue to the potential of a protein to mediate adaptive responses (Franzmann et al., 2018; Kroschwald et al., 2018; Riback et al., 2017). Further in-depth tips on how to handle proteins for phase separation assays can be found elsewhere (Alberti et al., 2018).

RNA considerations

Many phase separation assays also require the presence of RNA. In vitro transcription assays are a good source for long RNAs, whereas short RNAs can be purchased commercially. To visualize RNA recruitment to droplets, fluorescently labeled nucleotides can be added to in vitro transcription reactions at low concentrations to favor sparse or single molecule labeling of RNAs (Langdon and Gladfelter, 2018). Good RNA handling practices such as working to avoid RNase contamination and avoiding cycles of freeze-thawing RNAs are important along with checking for stability of the synthesized RNA after storage in the freezer. Some assays make use of total polyA-mRNA purified from a given cell and this bulk RNA is commercially available.

Buffer, additives, crowders and experimental conditions considerations

Phase separation is exquisitely sensitive to changes in physical-chemical conditions. Even small differences in temperature, protein, nucleic acid or salt concentration can lead to different outcomes. Therefore, buffers and protein concentrations should be carefully considered and controlled, and physiologically relevant salt concentrations and compositions used. Every effort should be made to measure the physiological concentrations of components in the respective subcellular compartment (nucleus or cytoplasm) and consider these concentrations when interpreting in vitro data.

A related issue is the use of synthetic macromolecular crowders such as polyethylene glycol, dextran or ficoll. In many cases, the amount of crowding agent added to an experiment may exceed the crowding predicted to exist in intracellular contexts (Luby-Phelps, 2013; Mitchison, 2019). It is important to realize that the underlying mechanisms of how crowders enhance phase separation are still unclear. One prominent idea is that crowders increase the effective protein concentration and thus mimic the conditions in the crowded cytoplasm (Minton, 2001; Zimmerman and Minton, 1993). However, there are other potential explanations. It could also be that crowders change the water activity or affect the effective valency of interacting proteins by promoting oligomerization (Senske et al., 2014; Wirth and Gruebele, 2013). Thus, crowders should be used with caution. If crowders are used, our recommendation is to test a range of different synthetic and biological crowders to exclude artifactual chemical effects.

Small molecule effects on phase separation

As efforts focus on understanding the function of compartments formed via LLPS, it is increasingly common to include enzymes into in vitro reconstituted, phase-separating systems. This may be done with the goal to enzymatically modify the phase separating biomolecules or to test whether LLPS alters the activity of the enzyme. In turn, as controls for enzyme experiments, high concentrations of small molecule chemicals (such as kinase inhibitors, methyltransferase inhibitors etc.) are commonly added to phase separation assays. However, in addition to their biological effects on the enzymes, these chemicals can also have direct chemical effects on phase separation. One specific example is ATP (Patel et al., 2017). ATP not only affects protein phase separation as an energy source, but it can also act as a hydrotrope that directly alters protein solubility. Such chemicals should be tested at various different concentrations and they should ideally be compared to chemically similar derivatives that lack the specific biological activity (nonhydrolyzable analogs in the case of ATP).

General considerations regarding the complexity of in vitro phase separation experiments

In vitro, IDRs are frequently sufficient to mediate phase separation. Examples of IDRs that phase separate at high concentrations are the IDRs of Ddx4, LAF-1, FUS, hnRNPA1 and Whi3 (Elbaum-Garfinkle et al., 2015; Lin et al., 2015; Molliex et al., 2015; Nott et al., 2015; Patel et al., 2015; Zhang et al., 2015). These observations have led to the widespread view that these IDRs function as protein-autonomous units that drive phase-separation through homotypic interactions. However, it is becoming increasingly clear that heterotypic interactions with other regions of the same polypeptide or other proteins can also drive phase separation (Wang et al., 2018). It will be critical to test how the rules of assembly that have been established for simple one- or two-component systems stand up to more complex mixtures of heterotypically interacting proteins that are thought to more commonly exist in cells. Complex mixtures of IDR interactions will likely be critical for establishing both the specific material properties of phase-separated assemblies and the specific molecular composition of an assembly.

Analysis of the physical properties of condensates

Material states of biological condensates vary widely.

It is clear that the material properties of intracellular condensates can vary substantially. These structures can be highly fluid and liquid-like on a continuum to more viscous, viscoelastic or porous solids or gels (Weber, 2017). These variable material states can arise due to the specific molecular components involved in the condensation and as a consequence of time and maturation of the droplets and quench depth, i.e., how deep in the two-phase regime the system is. The presence of RNA—either specific or non-specific sequences—can influence the material properties of droplets; however, whether RNA fluidizes or solidifies droplets is context dependent (Elbaum-Garfinkle et al., 2015; Zhang et al., 2015), presumably due to contributions of both valency and electrostatics. In several contexts, droplets have been shown to become more solid-like with time or with mutations that promote stable protein interactions or abrogate the ability of proteins to bind RNAs (Maharana et al., 2018; Molliex et al., 2015; Patel et al., 2015). Additionally, within the more gel-like states, the degree to which the solid state is reversible is an important feature to consider, as the implications of irreversibility for physiology and pathology are potentially important. Despite the growing descriptions of variable physical states that can be detected in reconstituted systems, the actual function of a specific material state in cells remains unclear. The degree to which a particular viscosity or viscoelasticity has been selected for during evolution or is an emergent property of the condensing components and not necessarily tuned for the function of the structure, is not yet known. Therefore, it remains critical to characterize and manipulate the material states of liquid- or gel-like compartments with the goal of ultimately understanding if and how material states relate to function.

Monitoring the physical properties of condensates in vitro.

There are multiple options for measuring the material state of droplets in the experimental setups used for in vitro reconstitution of LLPS (Mitrea et al., 2018). The most straightforward assay to determine the material state in vitro is to measure the inverse capillary velocity, which is a ratio of the viscosity to surface tension. This ratio can be estimated by filming droplets fusing using fluorescence or transmitted light microscopy and measuring the time it takes for two droplets to completely fuse into one droplet for a number of differently sized droplets. Combining measurements of inverse capillary velocity with passive microrheology, in which viscosity can be directly calculated, allows for the inference of the surface tension of droplets (Elbaum-Garfinkle et al., 2015; Zhang et al., 2015). Measurement of contact angles (the angle between the coverglass and the droplet surface) can also provide important information on surface tension and the chemical nature of the surface of droplets (Feric et al., 2016).

Passive microrheology involves embedding beads inside droplets and monitoring their diffusivity by determining the mean squared displacement (MSD) of the beads. There are many considerations for microrheology including the material of and size of the bead, the passivation of the bead surface, the need for a stable microscope set-up to minimize drift, and a means of adding beads to droplets and ensuring that analyzed beads are in the central regions of droplets so as to avoid boundary-effect behavior. A microfluidics-based approach has been implemented that aids in incorporation of beads and avoiding boundary effects by forming only two distinct liquid phases under flow (by fusing all droplets) and this helps avoid some artifacts commonly encountered and promotes incorporation of tracer beads into the dense phase (Taylor et al., 2016). Surveying the behavior of a variety of sizes of beads can help determine if, when and where the material may show elastic properties. Ideally these measurements are applied to simple liquids, but it is not always known if condensates are pure simple liquids or if they have structural heterogeneity. If structures appear to behave as highly viscous liquids or gels, AFM or optical tweezers may be a relevant approach to further measure the stiffness of the materials (Patel et al., 2015; Wang et al., 2018). The relationship of assembly morphology and secondary and tertiary structure of the component proteins is accessible by a combination of atomic force microscopy (AFM) with infrared nanospectroscopy (IR) (Dazzi et al., 2012; Qamar et al., 2018). Relevant to interpretations of microrheology is estimating the pore size of the polymer mesh in droplets. This can be done with fluorescently labeled dextrans of varying sizes and determination of the threshold for incorporation (Wei et al., 2017) with care given to artifacts that may arise from interactions between the probe and droplet components.

Fluorescence recovery after photobleaching (FRAP) is often also performed on droplets as an assessment of their liquidity, and different components can vary substantially between one another in the rate of recovery due to different mobilities. This difference is especially pronounced when comparing proteins and RNAs, in which the latter appear relatively less mobile. While a highly accessible technique, there are various limitations and challenges in analysis of FRAP data. One commonly made mistake is to assume that the recovery rate of a full-FRAP reports only on the exchange rate between the dilute and the dense phase. However, the FRAP recovery rate also depends on other parameters such as the size of the photobleached droplet, mobility within a droplet and size of bleach areas and these are often not taken into account. However, it can still be useful for assessing extremes in material state or changes in the material state through time. If the structure is large enough, a half-FRAP can also be performed in addition to a full-FRAP (Brangwynne et al., 2009). In a half-FRAP, the re-arrangement of fluorescence from the bleached to the unbleached area gives more direct information on the internal mobility of molecules within a given structure. Importantly, FRAP can also be useful in assessing if the droplets are spatially homogeneous based on the pattern of recovery. As noted below, FRAP should not be used as a definitive diagnostic for determining if LLPS is the mechanism of assembly of a structure.

For more accurate estimates of diffusivity at the scale of individual molecules within liquid states, fluorescence correlation spectroscopy (FCS) can be used in cases where components can be sparsely labeled and are sufficiently mobile to be detected. Variations on FCS have also been used to assess how dilute droplets are (Wei et al., 2017), which can be a useful parameter if theoretical models are being developed and to accurately assess concentrations. Additionally, application of polarized fluorescence microscopy may prove useful in detecting anisotropic subcompartments of fibrous or solid-like structures that may co-exist with liquid-like states, but may not be detectable from bulk fluorescence (McQuilken et al., 2015; Yoshizawa et al., 2018).

What if assemblies are not liquid-like?

Given the potentially metastable nature of liquid assemblies, they can transition into glassy or gel-like states that do not have classical liquid properties (Hughes et al., 2018; Murray et al., 2017). Importantly, some sequences likely have evolved to favor these transitions to encode mechanically resistant assemblies (Franzmann et al., 2018; Kroschwald et al., 2018; Riback et al., 2017). The diffusion of components is arrested in gel-like states, and they are typically unable to fuse and coalesce. However, if the assemblies are originally formed via LLPS, they typically have spherical shapes, often resulting in many deformed spherical structures that stick together in large clusters (Molliex et al., 2015; Roberts et al., 2018). If dynamic arrest happens shortly after phase transition, the assemblies may be too small to accurately determine their shape by light microscopy. However, at early time points, the protein concentration in the supernatant of the arrested assemblies still reflects csat of the LLPS process. Determining the supernatant concentration for a variety of total protein input concentrations in the assembly assay can thus be used to assess whether the arrested assemblies are formed via LLPS.

Designation and analysis of condensates in live cells

Demonstrating LLPS in vivo.

A major challenge in the field is having accurate metrics for demonstrating definitively that a specific protein or structure is indeed a phase-separated body in the context of the cell. Under certain conditions, many proteins and RNAs are capable of LLPS in vitro when at sufficient concentrations and/or in artificial buffer conditions. Furthermore, it is common to overexpress a protein and see a large, spherical droplet and surmise that the endogenously expressed protein must also be forming liquid-like droplets at lower concentrations that are simply below the detection limit of the light microscope. However, as phase separation requires crossing a saturation concentration, caution should be exercised when interpreting over-expression data. Efforts should be made to find additional metrics other than overexpression to support a claim that a compartment is indeed phase separated as opposed to simply a macroscopic punctum.

Currently, the commonly accepted criteria for defining a phase-separated structure are that it is spherical, fuses and recovers from photobleaching. While the geometry and capacity to fuse support a liquid-like state, it is critical to keep in mind that recovery rates from FRAP have many different interpretations, especially when performed within the constraints of live cell imaging of endogenous proteins where structures may be small and moving in space. Fast FRAP can result for a variety of reasons, including reversible binding of proteins to porous, solid structures. Similarly, treatment with 1,6-hexanediol, which has emerged as an indicator of LLPS, has caveats. The rationale for using hexanediol to determine the material properties of an assembly came from a study where the selectivity barrier formed by FG-nucleoporin gels was disrupted (Ribbeck and Görlich, 2002). While hexanediol can interfere with LLPS of many IDRs, this is neither necessary nor sufficient since phase separation involves various types of interactions including pi/pi, cation/pi and electrostatic interactions not all of which are sensitive to hexanediol. Hexanediol should also be used with caution when used on live cells because it changes the permeability of membranes and thus can lead to additional artifacts (Kroschwald et al., 2017). Fast FRAP recovery and sensitivity to hexanediol are insufficient to unequivocally demonstrate that a structure is formed via LLPS.

What are reliable metrics to further characterize structures as products of phase separation? A good approach would be to try to recapitulate phase diagrams in living cells that show a concentration dependent threshold for assembly, i.e., csat (Patel et al., 2015; Rai et al., 2018). Mapping phase diagrams in cells requires the variation of the expressed protein levels often beyond endogenous levels. While challenging, the measured csat can then be directly compared to the cellular concentration of a protein in a given condition, revealing whether phase separation can take place. One caveat is that it may be impossible to detect diffraction-limited assemblies with simple methods and super-resolution microscopy may have to be used. However, it is currently unclear what size structures need to reach to even be considered as phase separated.

In many contexts, LLPS is driven by multivalent interactions in protein-protein or protein-RNA complexes and it has been established that changes in valency can shift the phase landscape and material properties of condensates. Thus, genetic manipulations that increase and decrease valency should impact quantifiable aspects of phase-separation such as concentration thresholds, but also location, size, or material state of the droplets. It can also be useful to generate chimeric proteins in which the sequences thought to drive phase separation such as IDR are exchanged for other sequences that are assembly-competent to assess if function can be rescued. Such sequences could be protein domains that are known to oligomerize such as the self-assembling matrix protein of the muNS virus (Kroschwald et al., 2015). Recently, optogenetic tools have been developed to locally induce condensation through controlled changes in multivalency and, in some situations, this can be a useful tool. Several optogenetic systems are available (Bracha et al., 2018; Dine et al., 2018; Shin et al., 2017, 2018), which use engineered multivalency to drive phase separation in specific regions of the cell. Such systems will be very useful to study the nucleation of membraneless compartments in living cells, in particular if the induced phase separation is linked to a functional readout and the tools can be tuned to recapitulate physiological protein concentrations (Shin et al., 2017). However, because current optogenetic tools impose an artificial multivalency on top of a naturally-evolved multivalency, the results should be interpreted with caution and ideally combined with additional experiments.

Probing the physical properties of in vivo condensates.

The ability to assess material properties is far more limited in cells. However, when droplets in vivo are sufficiently large, inverse capillary velocities can be determined by time-lapse microscopy that captures the relaxation behavior of fusing droplets. This approach can be useful to compare the properties of different droplets even if absolute viscosities are not determined. FRAP may also be used, but as cautioned above, there are many interpretations of recovery and this measurement does not definitively prove a liquid-like state that has arisen through an LLPS process.

Future work is critically needed in the form of probes that can be used in vivo to assess the viscosity and porosity of condensates in the context of cells. Genetically encoded nanoparticles (GEMS) have recently been shown to be effective probes for microrheology of the cytosol and future work to target these particles to condensates may enable their use as passive microrheology probes. The adaptation of GEMS for active microrheology using thermally induced intracellular flows also appears within reach (Delarue et al., 2018; Mittasch et al., 2018). Additionally, FRET probes sensitive to crowding targeted to specific condensates may be a useful tool as well to monitor density of droplets (Boersma et al., 2015; Liu et al., 2017).

Looking to the future, quantitative phase microscopy can assess changes in the refractive index at the surfaces of droplets which should vary depending upon the material state of structures (Steelman et al., 2017). If sufficient differences at interfaces exist, it may even be possible to use optical tweezers to manipulate and measure the material properties of structures in the context of living cells. Finally, new methods to determine the density of condensates such as refractive index tomography (Schürmann et al., 2018) and more advanced imaging techniques such as Brillouin microscopy (Scarcelli et al., 2015) hold great promise to allow the mapping of the material properties of condensates non-invasively inside living cells.

In addition to probing the physical properties of in vivo condensates, it is important to address whether those specific properties matter for function (see below). Therefore, manipulations that can alter the putative condensate in some capacity in terms of its physical character and also impact its function are ideal. This is easier said than done because genetic perturbations that impact protein architecture may also impact function directly, not just due to altered condensation.

The utility and limits of polymer chemistry as a model for guiding theory

With the insight that LLPS has pervasive roles in biology, one goal for the field is to progress from the characterization of the phase behavior of individual proteins to a general framework that explains and predicts phase behavior of macromolecules. In an ideal scenario, the saturation concentration, stimulus responsiveness and material properties of the condensed phase could be predicted from the sequence. While we are still far removed from this goal, we can make use of theories on phase separation developed by polymer physicists and crystallographers and expand these theories for application to complex mixtures of biopolymers. A major long-term goal for integrating theory is to enable dissection of mechanisms controlling specific features and driving forces of phase separation. Simulations provide insights into details of molecular interactions that are not easily accessible experimentally. The marriage of experiments with theory and simulations will become all the more critical as more complex mixtures of biological molecules are reconstituted that better represent the actual molecular heterogeneity of condensates in cells.

Several theoretical frameworks are useful for describing phase separation and extrapolating from experimental data. Flory-Huggins theory describes enthalpically-driven demixing of homopolymers from solution under conditions of poor solvent quality which will favor polymer-polymer interactions (Flory, 1942; Huggins, 1942). The theory has been applied to predict the critical point of phase separation from experimental data and has yielded insights into the influence of solvent conditions on LLPS (Nott et al., 2015). The Overbeek and Voorn extension of Flory-Huggins theory explicitly considers electrostatic effects that are required for the description of complex coacervation between oppositely charged polymers (Overbeek and Voorn, 1957). Both are mean-field theories that best describe homopolymers and do not consider explicit sequence-dependent effects or polymer fluctuations. The recently developed random phase approximation explicitly considers sequence patterning of charged residue and has been successfully used to model experimental data of DDX4 (Lin et al., 2016). The development of theories with explicit sequence consideration shifts away from a homopolymer perspective and allows the evaluation of questions about the extent to which heteropolymeric proteins conform to the behavior of homopolymers. While the phase behavior of some proteins seems to be well described with mean-field homopolymer theories, emerging data suggest that other protein sequences encode emergent dense phase properties inaccessible to homopolymers. For example, the phase behavior of the LAF-1 RGG domain, which forms a semidilute dense phase as detected using ultra-fast FCS, requires a theoretical description accounting for three-body interactions and large-scale conformational fluctuations of the protein polymers (Wei et al., 2017). Stepping away from mean-field theory and considering the role of fluctuations may also be required to capture phase behavior near the critical point, where fluctuations are strongest.

Another avenue to deeper insight into phase behavior is the use of simulation approaches. The simulation of individual proteins has become possible even for disordered proteins and has been successfully used to derive insights into their sequence/function relationships (Das et al., 2015). Systems that are large enough to phase separate typically contain hundreds to thousands of molecules and require coarse-graining where the system is represented in a simplified manner (e.g. one bead per amino acid residue or series of amino acids) and only a subset of interactions are modeled. Model parameterization that results in the faithful recapitulation of accessible experimental behavior is the main challenge for coarse-graining (Ruff et al., 2015). The recently described slab sampling method can be used for the direct extraction of thermodynamic phase diagrams of intrinsically disordered proteins (Dignon et al., 2018) and has great promise for simulating polymers with detailed sequence information.

Coarse-grained simulations of multi-component systems have recently provided insight into the physical principles that may drive the formation of multilayered membraneless organelles such as nucleoli and nuclear speckles (Fei et al., 2017; Feric et al., 2016). The simulations provided support for the interplay of sequence-encoded interactions between different proteins or between protein and RNA; this was possible for up to 5 components, which shows the power of such simulations to systematically probe the interactions in multicomponent systems. The combination of favorable and unfavorable interactions between different components can result in the non-random organization that was observed experimentally (Fei et al., 2017; Feric et al., 2016).

Simulation and theory are important complementary approaches to the experimental characterization of phase separation. In turn, the characterization of phase behavior in cells will be a rich source of data for the parameterization of faithful simulation approaches and for emerging new theories that describe the phase behavior of complex biomolecular mixtures with properties that have so far not been observed in simpler synthetic polymer systems.

What does it all mean? Identification of function

To date, many proteins have been shown to phase separate in vitro under idealized conditions. Frequently, the very same proteins form assemblies in living cells, especially when those proteins are overexpressed. However, the fact that a given protein forms assemblies at high concentrations is not necessarily evidence that the phase separation ability of this protein is functionally relevant. To demonstrate this requires carefully designed experiments to manipulate the phase separation of a protein without altering its other functions or properties. The basis of such an experiment could be an in vitro phase separation assay. Guided by sequence analysis, mutations could be introduced into the protein to alter its phase separation properties. However, mutagenesis of phase-separating proteins may not be as straightforward as that of structured proteins. To change the phase behavior of a low complexity protein, for example, it may be necessary to introduce multiple mutations to significantly alter protein multivalency (Wang et al., 2018; Nott et al., 2015). Once variants with specific phase separation defects have been identified, they could be introduced into cells to replace the wild-type protein. The resulting cell lines could then be tested for the ability to promote the formation of membrane-less compartments under physiological conditions or in response to a perturbation. Ideally, these experiments should be combined with a functional assay to determine whether defects in LLPS go along with defects in protein functionality.

One problem with such genetic experiments is the time scales on which effects are observed. The phenotypes observed upon expression of a variant protein might not necessarily be the result of altered phase separation behavior, but could also be due to indirect effects. For example, one can imagine that expression of the variant could lead to a stress response and this could indirectly affect the formation of a condensate in living cells. One way around this problem could be to inject fluorescently labeled proteins directly into living cells. The use of fluorescence time-lapse microscopy would then allow observation of protein phase separation in real-time. Such experiments have successfully been used to investigate the role of RNA in regulating the phase behavior of prion-like proteins in the nucleus of living cells (Maharana et al., 2018). As discussed above, optogenetics offers another approach to manipulate phase separation in living cells.

Another possibility to investigate the functional effects of phase separation is the use of cellular extracts. Experiments with extracts have been very important in helping us understand fundamental cellular processes such as translation and transcription. Extract experiments have the advantage that they allow the use of in vitro reconstitution biochemistry. For example, one could assemble an RNA-binding protein into condensates in vitro and then test the functional effect of condensate formation on protein activity in an in vitro translation assay or similarly combine transcriptional machinery for in vitro transcription assays. Such experiments could build an important bridge between experiments in vitro and in vivo.

The physicochemical properties of LLPS suggest a variety of possible functions of phase separation processes in cells (Alberti, 2017; Holehouse and Pappu, 2018). In the following, we provide diverse examples for the possible functional repertoire of condensates, however, evidence for all of these functions is still being acquired and there is much need to identify the full functional consequences of condensates (Figure 4).

Figure 4:

An initial functional repertoire of biomolecular condensates.

1. LLPS can be used for sensing and fast, adaptive, and reversible responses. Small changes in solution conditions result in fast and decisive (i.e. infinitely cooperative) condensation, and this biophysical response is faster than the initiation of transcriptional or translational programs. LLPS has been shown to be used for adaptive responses to heat or pH stress (Franzmann et al., 2018; Kroschwald et al., 2018; Riback et al., 2017).
2. LLPS can be used to buffer concentrations of proteins. Increasing the concentration of a macromolecule results in its condensation above the saturation concentration. A further increase results in a larger volume fraction of the dense phase (with constant dense phase concentration) and a smaller volume fraction of the dilute phase, i.e. the concentration in the dilute phase remains constant. Other self-assembly processes such as protein oligomerization do not generally produce this buffering behavior. In cells, excess protein may be stored in membraneless organelles and will enter the dilute phase as needed when protein levels drop.
3. LLPS can locally concentrate molecules in condensates to activate reactions, signaling processes and nucleation of cytoskeletal structures. Increasing the local concentration of a key enzyme or protein complex in a condensate could accelerate biochemical reactions. For example, the miRISC complex has enhanced decapping activity when it is phase separated (Sheu-Gruttadauria and MacRae, 2018), and T-cell receptor clusters formed by 2D phase separation with signaling adaptors at the membrane nucleate actin polymerization more effectively (Su et al., 2016). Similarly, centrosomal components concentrate tubulin to promote microtubule nucleation and growth (Woodruff et al., 2017). Lastly, returning to one of the earliest identified membraneless organelles, there is growing evidence that the liquid-like state of the nucleolus may be important for the assembly of ribosomes (Feric et al., 2016; Lee et al., 2016; Mitrea et al., 2016).
4. LLPS may sequester molecules to prevent reactions or for inactivation. If one critical component is recruited into a dense phase, but all other components for an enzymatic reaction or signaling event remain in the dilute phase, the reaction or signaling event will be inhibited or slowed down.
5. LLPS can mediate the localization of proteins to preexisting phase separated, non-membrane-bound organelles. Biomolecular condensates show selective properties, admitting some proteins and RNAs and excluding others. The rules by which molecules are sorted into different coexisting condensates are largely unclear, although a role for sequence features in RNAs and proteins and relative surface tension has been revealed (Feric et al., 2016; Langdon et al., 2018). It is also becoming clear that many proteins found in membraneless organelles are able to undergo LLPS at close to physiological conditions, even if they are not strictly required for the organelles to form in cells. Phase separation may thus mediate targeting of molecules to preexisting organelles, as has recently been proposed for UBQLN2 (Dao et al., 2018) and SPOP (Bouchard et al., 2018).
6. LLPS can result in the formation of materials with specific viscoelastic properties and the condensation process could be used to do work. For example, condensation could generate mechanical forces that shape cellular structures such as membranes (Bergeron-Sandoval et al., 2017). Such mechanisms may turn out to be important for morphogenesis.
7. LLPS can result in the formation of physico-chemical and mechanical filters where pore sizes are determined by the number and dynamics of the cross-links between the macromolecules that make up the condensate. This has been demonstrated for the nuclear pore (Schmidt and Görlich, 2016).

While the possible roles for condensates in biology are numerous and exciting to contemplate, making a definitive determination of a functional role is not trivial. Each functional category described above (any others yet to be uncovered or imagined) would require the design of tailored experiments to determine the contribution of the condensate. Some will require the genetic manipulation of valency, others the meticulous characterization of the material properties of a condensate in vitro and in vivo. We now have an increasing arsenal of tools and methods at hand to delineate the functions of condensates in a given cellular context.

Finally, beyond the experimental tools being developed, we need to get better at extracting information about functional phase separation from the vast amount of natural sequence variation available to us in the post-genomic era. Nature has already done numerous experiments to adapt the phase behavior of a protein to the functional constraints in a living cell. We need innovative tools to be able to read this information and identify the key features and selective pressures underlying phase separation. Currently, such analyses are difficult due to the fact that sequences of phase-separating proteins are very hard to align. New ways to align these sequences and innovative computational approaches such as evolutionary couplings based on sequence covariation using approaches suitable for disordered sequences (Toth-Petroczy et al., 2016) will bring us closer towards this goal.

Figure 3:

Overview of experimental approaches used to evaluate properties of assemblies formed by LLPS

Acknowledgements

We thank Rohit Pappu, Christoph Weber, and Richard Kriwacki for comments on the manuscript and insightful discussions. We are grateful to Titus M. Franzmann for preparing Figure 2. S.A. acknowledges funding by the Max-Planck Society, the ERC (No. 725836 and 643417), the BMBF (01ED1601A, 031A359A), the JPND (CureALS), the Human Frontier Science Program RGP0034/2017, and the Volkswagen Foundation. T.M. acknowledges funding by NIH grant R01GM112846, St. Jude Children’s Research Hospital and the American Lebanese Syrian Associated Charities. A. G. acknowledges funding from the National Institutes of Health GM R01-GM081506, and the HHMI Faculty Scholars program.

References

1. Alberti S (2017). The wisdom of crowds: regulating cell function through condensed states of living matter. J. Cell Sci 130, jcs.200295. [DOI] [PubMed] [Google Scholar]
2. Alberti S, and Carra S (2018). Quality Control of Membraneless Organelles. J. Mol. Biol [DOI] [PubMed]
3. Alberti S, and Hyman AA (2016). Are aberrant phase transitions a driver of cellular aging? Bioessays 38, 959–968. [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Alberti S, Halfmann R, King O, Kapila A, and Lindquist S (2009). A systematic survey identifies prions and illuminates sequence features of prionogenic proteins. Cell 137, 146–158. [DOI] [PMC free article] [PubMed] [Google Scholar]
5. Alberti S, Saha S, Woodruff JB, Franzmann TM, Wang J, and Hyman AA (2018). A user’s guide for phase separation assays with purified proteins. J. Mol. Biol [DOI] [PMC free article] [PubMed]
6. Amiram M, Quiroz FG, Callahan DJ, and Chilkoti A (2011). A highly parallel method for synthesizing DNA repeats enables the discovery of “smart” protein polymers. Nat. Mater 10, 141–148. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Aumiller WM Jr, and Keating CD (2016). Phosphorylation-mediated RNA/peptide complex coacervation as a model for intracellular liquid organelles. Nat. Chem 8, 129–137. [DOI] [PubMed] [Google Scholar]
8. Banani SF, Rice AM, Peeples WB, Lin Y, Jain S, Parker R, and Rosen MK (2016). Compositional Control of Phase-Separated Cellular Bodies. Cell [DOI] [PMC free article] [PubMed]
9. Banani SF, Lee HO, Hyman AA, and Rosen MK (2017). Biomolecular condensates: organizers of cellular biochemistry. Nat. Rev. Mol. Cell Biol 18, 285–298. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Bentmann E, Neumann M, Tahirovic S, Rodde R, Dormann D, and Haass C (2012). Requirements for stress granule recruitment of fused in sarcoma (FUS) and TAR DNA-binding protein of 43 kDa (TDP-43). J. Biol. Chem 287, 23079–23094. [DOI] [PMC free article] [PubMed] [Google Scholar]
11. Bergeron-Sandoval LP, Heris HK, and Hendricks AG (2017). Endocytosis caused by liquid-liquid phase separation of proteins. bioRxiv
12. Boersma AJ, Zuhorn IS, and Poolman B (2015). A sensor for quantification of macromolecular crowding in living cells. Nat. Methods 12, 227–229, 1 p following 229. [DOI] [PubMed] [Google Scholar]
13. Bouchard JJ, Otero JH, Scott DC, Szulc E, Martin EW, Sabri N, Granata D, Marzahn MR, Lindorff-Larsen K, Salvatella X, et al. (2018). Cancer Mutations of the Tumor Suppressor SPOP Disrupt the Formation of Active, Phase-Separated Compartments. Mol. Cell [DOI] [PMC free article] [PubMed]
14. Bracha D, Walls MT, Wei M-T, Zhu L, Kurian M, Avalos JL, Toettcher JE, and Brangwynne CP (2018). Mapping Local and Global Liquid Phase Behavior in Living Cells Using Photo-Oligomerizable Seeds. Cell 175, 1467–1480.e13. [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Brangwynne CP, Eckmann CR, Courson DS, Rybarska A, Hoege C, Gharakhani J, Jülicher F, and Hyman AA (2009). Germline P granules are liquid droplets that localize by controlled dissolution/condensation. Science 324, 1729–1732. [DOI] [PubMed] [Google Scholar]
16. Burke KA, Janke AM, Rhine CL, and Fawzi NL (2015). Residue-by-Residue View of In Vitro FUS Granules that Bind the C-Terminal Domain of RNA Polymerase II. Mol. Cell 60, 231–241. [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Chong PA, Vernon RM, and Forman-Kay JD (2018). RGG/RG Motif Regions in RNA Binding and Phase Separation. J. Mol. Biol [DOI] [PubMed]
18. Dao TP, Kolaitis R-M, Kim HJ, O’Donovan K, Martyniak B, Colicino E, Hehnly H, Taylor JP, and Castañeda CA (2018). Ubiquitin Modulates Liquid-Liquid Phase Separation of UBQLN2 via Disruption of Multivalent Interactions. Mol. Cell [DOI] [PMC free article] [PubMed]
19. Das RK, and Pappu RV (2013). Conformations of intrinsically disordered proteins are influenced by linear sequence distributions of oppositely charged residues. Proc. Natl. Acad. Sci. U. S. A 110, 13392–13397. [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Das RK, Ruff KM, and Pappu RV (2015). Relating sequence encoded information to form and function of intrinsically disordered proteins. Curr. Opin. Struct. Biol 32, 102–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
21. Dazzi A, Prater CB, Hu Q, Chase DB, Rabolt JF, and Marcott C (2012). AFM-IR: combining atomic force microscopy and infrared spectroscopy for nanoscale chemical characterization. Appl. Spectrosc 66, 1365–1384. [DOI] [PubMed] [Google Scholar]
22. Delarue M, Brittingham GP, Pfeffer S, Surovtsev IV, Pinglay S, Kennedy KJ, Schaffer M, Gutierrez JI, Sang D, Poterewicz G, et al. (2018). mTORC1 Controls Phase Separation and the Biophysical Properties of the Cytoplasm by Tuning Crowding. Cell 174, 338–349.e20. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Dignon GL, Zheng W, Kim YC, Best RB, and Mittal J (2018). Sequence determinants of protein phase behavior from a coarse-grained model. PLoS Comput. Biol 14, e1005941. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Dine E, Gil AA, Uribe G, Brangwynne CP, and Toettcher JE (2018). Protein Phase Separation Provides Long-Term Memory of Transient Spatial Stimuli. Cell Syst 6, 655–663.e5. [DOI] [PMC free article] [PubMed] [Google Scholar]
25. Elbaum-Garfinkle S, Kim Y, Szczepaniak K, Chen CC-H, Eckmann CR, Myong S, and Brangwynne CP (2015). The disordered P granule protein LAF-1 drives phase separation into droplets with tunable viscosity and dynamics. Proc. Natl. Acad. Sci. U. S. A 112, 7189–7194. [DOI] [PMC free article] [PubMed] [Google Scholar]
26. Fei J, Jadaliha M, Harmon TS, Li ITS, Hua B, Hao Q, Holehouse AS, Reyer M, Sun Q, Freier SM, et al. (2017). Quantitative analysis of multilayer organization of proteins and RNA in nuclear speckles at super resolution. J. Cell Sci 130, 4180–4192. [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Feric M, Vaidya N, Harmon TS, Mitrea DM, Zhu L, Richardson TM, Kriwacki RW, Pappu RV, and Brangwynne CP (2016). Coexisting Liquid Phases Underlie Nucleolar Subcompartments. Cell [DOI] [PMC free article] [PubMed]
28. Flory PJ (1942). Thermodynamics of High Polymer Solutions. J. Chem. Phys 10, 51–61. [Google Scholar]
29. Fowler DM, Koulov AV, Balch WE, and Kelly JW (2007). Functional amyloid--from bacteria to humans. Trends Biochem. Sci 32, 217–224. [DOI] [PubMed] [Google Scholar]
30. Franzmann T, and Alberti S (2018). Prion-like low-complexity sequences: Key regulators of protein solubility and phase behavior. J. Biol. Chem [DOI] [PMC free article] [PubMed]
31. Franzmann TM, Jahnel M, Pozniakovsky A, Mahamid J, Holehouse AS, Nüske E, Richter D, Baumeister W, Grill SW, Pappu RV, et al. (2018). Phase separation of a yeast prion protein promotes cellular fitness. Science 359. [DOI] [PubMed] [Google Scholar]
32. Harmon TS, Holehouse AS, Rosen MK, and Pappu RV (2017). Intrinsically disordered linkers determine the interplay between phase separation and gelation in multivalent proteins. Elife 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Hofweber M, Hutten S, Bourgeois B, Spreitzer E, Niedner-Boblenz A, Schifferer M, Ruepp M-D, Simons M, Niessing D, Madl T, et al. (2018). Phase Separation of FUS Is Suppressed by Its Nuclear Import Receptor and Arginine Methylation. Cell 173, 706–719.e13. [DOI] [PubMed] [Google Scholar]
34. Holehouse AS, and Pappu RV (2018). Functional Implications of Intracellular Phase Transitions. Biochemistry [DOI] [PMC free article] [PubMed]
35. Huggins ML (1942). Some Properties of Solutions of Long-chain Compounds. J. Phys. Chem 46, 151–158. [Google Scholar]
36. Hughes MP, Sawaya MR, Boyer DR, Goldschmidt L, Rodriguez JA, Cascio D, Chong L, Gonen T, and Eisenberg DS (2018). Atomic structures of low-complexity protein segments reveal kinked β sheets that assemble networks. Science 359, 698–701. [DOI] [PMC free article] [PubMed] [Google Scholar]
37. Hyman AA, Weber CA, and Jülicher F (2014). Liquid-Liquid Phase Separation in Biology. Annu. Rev. Cell Dev. Biol 30, 39–58. [DOI] [PubMed] [Google Scholar]
38. Jensen MR, Ruigrok RWH, and Blackledge M (2013). Describing intrinsically disordered proteins at atomic resolution by NMR. Curr. Opin. Struct. Biol 23, 426–435. [DOI] [PubMed] [Google Scholar]
39. Kato M, Han TW, Xie S, Shi K, Du X, Wu LC, Mirzaei H, Goldsmith EJ, Longgood J, Pei J, et al. (2012). Cell-free formation of RNA granules: low complexity sequence domains form dynamic fibers within hydrogels. Cell 149, 753–767. [DOI] [PMC free article] [PubMed] [Google Scholar]
40. Kim HJ, Kim NC, Wang Y-D, Scarborough EA, Moore J, Diaz Z, MacLea KS, Freibaum B, Li S, Molliex A, et al. (2013). Mutations in prion-like domains in hnRNPA2B1 and hnRNPA1 cause multisystem proteinopathy and ALS. Nature 495, 467–473. [DOI] [PMC free article] [PubMed] [Google Scholar]
41. Knowles TPJ, Vendruscolo M, and Dobson CM (2014). The amyloid state and its association with protein misfolding diseases. Nat. Rev. Mol. Cell Biol 15, 384–396. [DOI] [PubMed] [Google Scholar]
42. Kroschwald S, Maharana S, Mateju D, Malinovska L, Nüske E, Poser I, Richter D, and Alberti S (2015). Promiscuous interactions and protein disaggregases determine the material state of stress-inducible RNP granules. Elife 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Kroschwald S, Maharana S, and Alberti S (2017). Hexanediol: a chemical probe to investigate the material properties of membrane-less compartments. Matters
44. Kroschwald S, Munder MC, Maharana S, Franzmann TM, Richter D, Ruer M, Hyman AA, and Alberti S (2018). Different Material States of Pub1 Condensates Define Distinct Modes of Stress Adaptation and Recovery. Cell Rep 23, 3327–3339. [DOI] [PubMed] [Google Scholar]
45. Kwon I, Kato M, Xiang S, Wu L, Theodoropoulos P, Mirzaei H, Han T, Xie S, Corden JL, and McKnight SL (2013). Phosphorylation-regulated binding of RNA polymerase II to fibrous polymers of low-complexity domains. Cell 155, 1049–1060. [DOI] [PMC free article] [PubMed] [Google Scholar]
46. Langdon EM, and Gladfelter AS (2018). A New Lens for RNA Localization: Liquid-Liquid Phase Separation. Annu. Rev. Microbiol 72, 255–271. [DOI] [PubMed] [Google Scholar]
47. Langdon EM, Qiu Y, Ghanbari Niaki A, McLaughlin GA, Weidmann C, Gerbich TM, Smith JA, Crutchley JM, Termini CM, Weeks KM, et al. (2018). mRNA structure determines specificity of a polyQ-driven phase separation. Science [DOI] [PMC free article] [PubMed]
48. Larson AG, Elnatan D, Keenen MM, Trnka MJ, Johnston JB, Burlingame AL, Agard DA, Redding S, and Narlikar GJ (2017). Liquid droplet formation by HP1α suggests a role for phase separation in heterochromatin. Nature 547, 236–240. [DOI] [PMC free article] [PubMed] [Google Scholar]
49. Lee C, Occhipinti P, and Gladfelter AS (2015). PolyQ-dependent RNA–protein assemblies control symmetry breaking. J. Cell Biol jcb.201407105. [DOI] [PMC free article] [PubMed]
50. Lee K-H, Zhang P, Kim HJ, Mitrea DM, Sarkar M, Freibaum BD, Cika J, Coughlin M, Messing J, Molliex A, et al. (2016). C9orf72 Dipeptide Repeats Impair the Assembly, Dynamics, and Function of Membrane-Less Organelles. Cell 167, 774–788.e17. [DOI] [PMC free article] [PubMed] [Google Scholar]
51. Li P, Banjade S, Cheng H-C, Kim S, Chen B, Guo L, Llaguno M, Hollingsworth JV, King DS, Banani SF, et al. (2012). Phase transitions in the assembly of multivalent signalling proteins. Nature 483, 336–340. [DOI] [PMC free article] [PubMed] [Google Scholar]
52. Li YR, King OD, Shorter J, and Gitler AD (2013). Stress granules as crucibles of ALS pathogenesis. J. Cell Biol 201, 361–372. [DOI] [PMC free article] [PubMed] [Google Scholar]
53. Lin Y, Protter DSW, Rosen MK, and Parker R (2015). Formation and Maturation of Phase-Separated Liquid Droplets by RNA-Binding Proteins. Mol. Cell 60, 208–219. [DOI] [PMC free article] [PubMed] [Google Scholar]
54. Lin Y, Currie SL, and Rosen MK (2017). Intrinsically disordered sequences enable modulation of protein phase separation through distributed tyrosine motifs. J. Biol. Chem [DOI] [PMC free article] [PubMed]
55. Lin Y-H, Forman-Kay JD, and Chan HS (2016). Sequence-Specific Polyampholyte Phase Separation in Membraneless Organelles. Phys. Rev. Lett 117, 178101. [DOI] [PubMed] [Google Scholar]
56. Liu B, Åberg C, van Eerden FJ, Marrink SJ, Poolman B, and Boersma AJ (2017). Design and Properties of Genetically Encoded Probes for Sensing Macromolecular Crowding. Biophys. J 112, 1929–1939. [DOI] [PMC free article] [PubMed] [Google Scholar]
57. Luby-Phelps K (2013). The physical chemistry of cytoplasm and its influence on cell function: an update. Mol. Biol. Cell 24, 2593–2596. [DOI] [PMC free article] [PubMed] [Google Scholar]
58. Lyons DF, Le V, Bidwell GL 3rd, Kramer WH, Lewis EA, Raucher D, and Correia JJ (2013). Structural and hydrodynamic analysis of a novel drug delivery vector: ELP[V5G3A2–150]. Biophys. J 104, 2009–2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
59. Mackenzie IR, Nicholson AM, Sarkar M, Messing J, Purice MD, Pottier C, Annu K, Baker M, Perkerson RB, Kurti A, et al. (2017). TIA1 Mutations in Amyotrophic Lateral Sclerosis and Frontotemporal Dementia Promote Phase Separation and Alter Stress Granule Dynamics. Neuron 95, 808–816.e9. [DOI] [PMC free article] [PubMed] [Google Scholar]
60. Maharana S, Wang J, Papadopoulos DK, Richter D, Pozniakovsky A, Poser I, Bickle M, Rizk S, Guillén-Boixet J, Franzmann T, et al. (2018). RNA buffers the phase separation behavior of prion-like RNA binding proteins. Science [DOI] [PMC free article] [PubMed]
61. Mao AH, Crick SL, Vitalis A, Chicoine CL, and Pappu RV (2010). Net charge per residue modulates conformational ensembles of intrinsically disordered proteins. Proc. Natl. Acad. Sci. U. S. A 107, 8183–8188. [DOI] [PMC free article] [PubMed] [Google Scholar]
62. Marsh JA, and Forman-Kay JD (2010). Sequence determinants of compaction in intrinsically disordered proteins. Biophys. J 98, 2383–2390. [DOI] [PMC free article] [PubMed] [Google Scholar]
63. Martin EW, and Mittag T (2018). The relationship of sequence and phase separation in protein low-complexity regions. Biochemistry [DOI] [PMC free article] [PubMed]
64. McQuilken M, Mehta SB, Verma A, Harris G, Oldenbourg R, and Gladfelter AS (2015). Polarized Fluorescence Microscopy to Study Cytoskeleton Assembly and Organization in Live Cells. Curr. Protoc. Cell Biol 67, 4291–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
65. Minton AP (2001). The influence of macromolecular crowding and macromolecular confinement on biochemical reactions in physiological media. J. Biol. Chem 276, 10577–10580. [DOI] [PubMed] [Google Scholar]
66. Mitchison TJ (2019). Colloid Osmotic Parameterization and Measurement of Sub-Cellular Crowding [DOI] [PMC free article] [PubMed]
67. Mitrea DM, Cika JA, Guy CS, Ban D, Banerjee PR, Stanley CB, Nourse A, Deniz AA, and Kriwacki RW (2016). Nucleophosmin integrates within the nucleolus via multi-modal interactions with proteins displaying R-rich linear motifs and rRNA. Elife 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
68. Mitrea DM, Chandra B, Ferrolino MC, Gibbs EB, Tolbert M, White MR, and Kriwacki RW (2018). Methods for Physical Characterization of Phase-Separated Bodies and Membrane-less Organelles. J. Mol. Biol 430, 4773–4805. [DOI] [PMC free article] [PubMed] [Google Scholar]
69. Mittag T, and Forman-Kay JD (2007). Atomic-level characterization of disordered protein ensembles. Curr. Opin. Struct. Biol 17, 3–14. [DOI] [PubMed] [Google Scholar]
70. Mittasch M, Gross P, Nestler M, Fritsch AW, Iserman C, Kar M, Munder M, Voigt A, Alberti S, Grill SW, et al. (2018). Non-invasive perturbations of intracellular flow reveal physical principles of cell organization. Nat. Cell Biol 20, 344–351. [DOI] [PubMed] [Google Scholar]
71. Molliex A, Temirov J, Lee J, Coughlin M, Kanagaraj AP, Kim HJ, Mittag T, and Taylor JP (2015). Phase separation by low complexity domains promotes stress granule assembly and drives pathological fibrillization. Cell 163, 123–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
72. Monahan Z, Ryan VH, Janke AM, Burke KA, Rhoads SN, Zerze GH, O’Meally R, Dignon GL, Conicella AE, Zheng W, et al. (2017). Phosphorylation of the FUS low-complexity domain disrupts phase separation, aggregation, and toxicity. EMBO J 36, 2951–2967. [DOI] [PMC free article] [PubMed] [Google Scholar]
73. Müller-Späth S, Soranno A, Hirschfeld V, Hofmann H, Rüegger S, Reymond L, Nettels D, and Schuler B (2010). From the Cover: Charge interactions can dominate the dimensions of intrinsically disordered proteins. Proc. Natl. Acad. Sci. U. S. A 107, 14609–14614. [DOI] [PMC free article] [PubMed] [Google Scholar]
74. Murakami T, Qamar S, Lin JQ, Schierle GSK, Rees E, Miyashita A, Costa AR, Dodd RB, Chan FTS, Michel CH, et al. (2015). ALS/FTD Mutation-Induced Phase Transition of FUS Liquid Droplets and Reversible Hydrogels into Irreversible Hydrogels Impairs RNP Granule Function. Neuron 88, 678–690. [DOI] [PMC free article] [PubMed] [Google Scholar]
75. Murray DT, Kato M, Lin Y, Thurber KR, Hung I, McKnight SL, and Tycko R (2017). Structure of FUS Protein Fibrils and Its Relevance to Self-Assembly and Phase Separation of Low-Complexity Domains. Cell 171, 615–627.e16. [DOI] [PMC free article] [PubMed] [Google Scholar]
76. Nott TJ, Petsalaki E, Farber P, Jervis D, Fussner E, Plochowietz A, Craggs TD, Bazett-Jones DP, Pawson T, Forman-Kay JD, et al. (2015). Phase transition of a disordered nuage protein generates environmentally responsive membraneless organelles. Mol. Cell 57, 936–947. [DOI] [PMC free article] [PubMed] [Google Scholar]
77. Overbeek JT, and Voorn MJ (1957). Phase separation in polyelectrolyte solutions; theory of complex coacervation. J. Cell. Physiol Suppl 49, 7–22; discussion, 22–26. [PubMed] [Google Scholar]
78. Pak CW, Kosno M, Holehouse AS, Padrick SB, Mittal A, Ali R, Yunus AA, Liu DR, Pappu RV, and Rosen MK (2016). Sequence Determinants of Intracellular Phase Separation by Complex Coacervation of a Disordered Protein. Mol. Cell 63, 72–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
79. Patel A, Lee HO, Jawerth L, Maharana S, Jahnel M, Hein MY, Stoynov S, Mahamid J, Saha S, Franzmann TM, et al. (2015). A Liquid-to-Solid Phase Transition of the ALS Protein FUS Accelerated by Disease Mutation. Cell 162, 1066–1077. [DOI] [PubMed] [Google Scholar]
80. Patel A, Malinovska L, Saha S, Wang J, Alberti S, Krishnan Y, and Hyman AA (2017). ATP as a biological hydrotrope. Science 356, 753–756. [DOI] [PubMed] [Google Scholar]
81. Qamar S, Wang G, Randle SJ, Ruggeri FS, Varela JA, Lin JQ, Phillips EC, Miyashita A, Williams D, Ströhl F, et al. (2018). FUS Phase Separation Is Modulated by a Molecular Chaperone and Methylation of Arginine Cation-π Interactions. Cell 173, 720–734.e15. [DOI] [PMC free article] [PubMed] [Google Scholar]
82. Quiroz FG, and Chilkoti A (2015). Sequence heuristics to encode phase behaviour in intrinsically disordered protein polymers. Nat. Mater 14, 1164–1171. [DOI] [PMC free article] [PubMed] [Google Scholar]
83. Rai AK, Chen J-X, Selbach M, and Pelkmans L (2018). Kinase-controlled phase transition of membraneless organelles in mitosis. Nature 559, 211–216. [DOI] [PubMed] [Google Scholar]
84. Ramaswami M, Taylor JP, and Parker R (2013). Altered ribostasis: RNA-protein granules in degenerative disorders. Cell 154, 727–736. [DOI] [PMC free article] [PubMed] [Google Scholar]
85. Riback JA, Katanski CD, Kear-Scott JL, Pilipenko EV, Rojek AE, Sosnick TR, and Drummond DA (2017). Stress-Triggered Phase Separation Is an Adaptive, Evolutionarily Tuned Response. Cell 168, 1028–1040.e19. [DOI] [PMC free article] [PubMed] [Google Scholar]
86. Ribbeck K, and Görlich D (2002). The permeability barrier of nuclear pore complexes appears to operate via hydrophobic exclusion. EMBO J 21, 2664–2671. [DOI] [PMC free article] [PubMed] [Google Scholar]
87. Roberts S, Harmon TS, Schaal J, Miao V, Li K. (jonathan), Hunt A, Wen Y, Oas TG, Collier JH, Pappu RV, et al. (2018). Injectable tissue integrating networks from recombinant polypeptides with tunable order. Nat. Mater [DOI] [PMC free article] [PubMed]
88. Ruff KM, Harmon TS, and Pappu RV (2015). CAMELOT: A machine learning approach for coarse-grained simulations of aggregation of block-copolymeric protein sequences. J. Chem. Phys 143, 243123. [DOI] [PMC free article] [PubMed] [Google Scholar]
89. Ruff KM, Roberts S, Chilkoti A, and Pappu RV (2018). Advances in Understanding Stimulus Responsive Phase Behavior of Intrinsically Disordered Protein Polymers. J. Mol. Biol [DOI] [PubMed]
90. Saito M, Hess D, Eglinger J, Fritsch AW, Kreysing M, Weinert BT, Choudhary C, and Matthias P (2019). Acetylation of intrinsically disordered regions regulates phase separation. Nat. Chem. Biol 15, 51–61. [DOI] [PubMed] [Google Scholar]
91. Scarcelli G, Polacheck WJ, Nia HT, Patel K, Grodzinsky AJ, Kamm RD, and Yun SH (2015). Noncontact three-dimensional mapping of intracellular hydromechanical properties by Brillouin microscopy. Nat. Methods 12, 1132–1134. [DOI] [PMC free article] [PubMed] [Google Scholar]
92. Schmidt HB, and Görlich D (2016). Transport Selectivity of Nuclear Pores, Phase Separation, and Membraneless Organelles. Trends Biochem. Sci 41, 46–61. [DOI] [PubMed] [Google Scholar]
93. Schürmann M, Cojoc G, Girardo S, Ulbricht E, Guck J, and Müller P (2018). Three-dimensional correlative single-cell imaging utilizing fluorescence and refractive index tomography. J. Biophotonics 11. [DOI] [PubMed] [Google Scholar]
94. Senske M, Törk L, Born B, Havenith M, Herrmann C, and Ebbinghaus S (2014). Protein stabilization by macromolecular crowding through enthalpy rather than entropy. J. Am. Chem. Soc 136, 9036–9041. [DOI] [PubMed] [Google Scholar]
95. Sheu-Gruttadauria J, and MacRae IJ (2018). Phase Transitions in the Assembly and Function of Human miRISC. Cell 173, 946–957.e16. [DOI] [PMC free article] [PubMed] [Google Scholar]
96. Shin Y, and Brangwynne CP (2017). Liquid phase condensation in cell physiology and disease. Science 357, eaaf4382. [DOI] [PubMed] [Google Scholar]
97. Shin Y, Berry J, Pannucci N, Haataja MP, Toettcher JE, and Brangwynne CP (2017). Spatiotemporal Control of Intracellular Phase Transitions Using Light-Activated optoDroplets. Cell 168, 159–171.e14. [DOI] [PMC free article] [PubMed] [Google Scholar]
98. Shin Y, Chang Y-C, Lee DSW, Berry J, Sanders DW, Ronceray P, Wingreen NS, Haataja M, and Brangwynne CP (2018). Liquid Nuclear Condensates Mechanically Sense and Restructure the Genome. Cell 175, 1481–1491.e13. [DOI] [PMC free article] [PubMed] [Google Scholar]
99. Smith J, Calidas D, Schmidt H, Lu T, Rasoloson D, and Seydoux G (2016). Spatial patterning of P granules by RNA-induced phase separation of the intrinsically-disordered protein MEG-3. Elife 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
100. Sørensen HP, and Mortensen KK (2005). Advanced genetic strategies for recombinant protein expression in Escherichia coli. J. Biotechnol 115, 113–128. [DOI] [PubMed] [Google Scholar]
101. Steelman ZA, Eldridge WJ, Weintraub JB, and Wax A (2017). Is the nuclear refractive index lower than cytoplasm? Validation of phase measurements and implications for light scattering technologies. J. Biophotonics 10, 1714–1722. [DOI] [PMC free article] [PubMed] [Google Scholar]
102. Strom AR, Emelyanov AV, Mir M, Fyodorov DV, Darzacq X, and Karpen GH (2017). Phase separation drives heterochromatin domain formation. Nature 547, 241–245. [DOI] [PMC free article] [PubMed] [Google Scholar]
103. Su X, Ditlev JA, Hui E, Xing W, Banjade S, Okrut J, King DS, Taunton J, Rosen MK, and Vale RD (2016). Phase separation of signaling molecules promotes T cell receptor signal transduction. Science 352, 595–599. [DOI] [PMC free article] [PubMed] [Google Scholar]
104. Tanaka T, Ishimoto C, and Chylack LT Jr,(1977). Phase separation of a protein-water mixture in cold cataract in the young rat lens. Science 197, 1010–1012. [DOI] [PubMed] [Google Scholar]
105. Taratuta VG, Holschbach A, Thurston GM, Blankschtein D, and Benedek GB (1990). Liquid-liquid phase separation of aqueous lysozyme solutions: effects of pH and salt identity. J. Phys. Chem 94, 2140–2144. [Google Scholar]
106. Taylor N, Elbaum-Garfinkle S, Vaidya N, Zhang H, Stone HA, and Brangwynne CP (2016). Biophysical characterization of organelle-based RNA/protein liquid phases using microfluidics. Soft Matter 12, 9142–9150. [DOI] [PMC free article] [PubMed] [Google Scholar]
107. Toth-Petroczy A, Palmedo P, Ingraham J, Hopf TA, Berger B, Sander C, and Marks DS (2016). Structured States of Disordered Proteins from Genomic Sequences. Cell 167, 158–170.e12. [DOI] [PMC free article] [PubMed] [Google Scholar]
108. Urry DW, Gowda DC, Parker TM, Luan CH, Reid MC, Harris CM, Pattanaik A, and Harris RD (1992). Hydrophobicity scale for proteins based on inverse temperature transitions. Biopolymers 32, 1243–1250. [DOI] [PubMed] [Google Scholar]
109. Valentin G (1836). Repertorium Für Anatomie Und Physiologie, Verlag von Veit Und Comp, Berlin 1, 1–293. [Google Scholar]
110. Wagner R (1835). Einige Bemerkungen und Fragen über das Keimbläschen (vesicular germinativa). Müller’s Archiv Anat Physiol Wissenschaft Med 373–377.
111. Wang J, Choi J-M, Holehouse AS, Lee HO, Zhang X, Jahnel M, Maharana S, Lemaitre R, Pozniakovsky A, Drechsel D, et al. (2018). A Molecular Grammar Governing the Driving Forces for Phase Separation of Prion-like RNA Binding Proteins. Cell [DOI] [PMC free article] [PubMed]
112. Weber SC (2017). Sequence-encoded material properties dictate the structure and function of nuclear bodies. Curr. Opin. Cell Biol 46, 62–71. [DOI] [PubMed] [Google Scholar]
113. Weber SC, and Brangwynne CP (2012). Getting RNA and protein in phase. Cell 149, 1188–1191. [DOI] [PubMed] [Google Scholar]
114. Wei M-T, Elbaum-Garfinkle S, Holehouse AS, Chen CC-H, Feric M, Arnold CB, Priestley RD, Pappu RV, and Brangwynne CP (2017). Phase behaviour of disordered proteins underlying low density and high permeability of liquid organelles. Nat. Chem 9, 1118–1125. [DOI] [PMC free article] [PubMed] [Google Scholar]
115. Widom B (1988). Note on the interfacial tension of phase-separated polymer solutions. J. Stat. Phys 52, 1343–1351. [Google Scholar]
116. Wirth AJ, and Gruebele M (2013). Quinary protein stzructure and the consequences of crowding in living cells: leaving the test-tube behind. Bioessays 35, 984–993. [DOI] [PubMed] [Google Scholar]
117. Woodruff JB, Ferreira Gomes B, Widlund PO, Mahamid J, Honigmann A, and Hyman AA (2017). The Centrosome Is a Selective Condensate that Nucleates Microtubules by Concentrating Tubulin. Cell 169, 1066–1077.e10. [DOI] [PubMed] [Google Scholar]
118. Yoshizawa T, Ali R, Jiou J, Fung HYJ, Burke KA, Kim SJ, Lin Y, Peeples WB, Saltzberg D, Soniat M, et al. (2018). Nuclear Import Receptor Inhibits Phase Separation of FUS through Binding to Multiple Sites. Cell 173, 693–705.e22. [DOI] [PMC free article] [PubMed] [Google Scholar]
119. Zhang H, Elbaum-Garfinkle S, Langdon EM, Taylor N, Occhipinti P, Bridges AA, Brangwynne CP, and Gladfelter AS (2015). RNA Controls PolyQ Protein Phase Transitions. Mol. Cell 60, 220–230. [DOI] [PMC free article] [PubMed] [Google Scholar]
120. Zimmerman SB, and Minton AP (1993). Macromolecular crowding: biochemical, biophysical, and physiological consequences. Annu. Rev. Biophys. Biomol. Struct 22, 27–65. [DOI] [PubMed] [Google Scholar]