Conformational freedom and topological confinement of proteins in biomolecular condensates

Abstract

The emergence of biomolecular condensation and liquid-liquid phase separation (LLPS) introduces a new layer of complexity into our understanding of cell and molecular biology. Evidence steadily grows indicating that condensates are not only implicated in physiology but also human disease. Macro- and mesoscale characterization of condensates as a whole have been instrumental in understanding their biological functions and dysfunctions [1–3]. By contrast, the molecular level characterization of condensates and how condensates modify the properties of the molecules that constitute them thus far remain comparably scarce. In this minireview we summarize and discuss the findings of several recent studies that have focused on structure, dynamics, and interactions of proteins undergoing condensation. The mechanistic insights they provide help us identify the relevant properties nature and scientists can leverage to modulate the behavior of condensate systems. We also discuss the unique environment of the droplet surface and speculate on effects of topological constraints and physical exclusion on condensate properties.

Introduction

It is becoming increasingly clear that the condensed state of proteins that results from biomolecular condensation or liquid-liquid phase separation is not an exception but rather a fairly common state accessible to a wide range of proteins [4,5]. As we gain better understanding of the physical and biological properties of condensates, the question arises in what way the condensates modify the properties of the molecules that constitute them. Do proteins in the condensed state go through conformational changes? Are their dynamics restricted? Do they engage in persistent interactions? To undergo condensation, intermolecular protein:protein interaction must become energetically more favorable for the system overall than intramolecular interactions and interactions between the protein and the solvent. This general idea including a balance of energetic and entropic contributions for components and interactions was formalized in detail by Flory and Huggins through the Flory interaction parameter in polymer chemistry [6–8] and has been revisited more recently to describe biomolecular condensates [9]. Because intermolecular interactions become more favorable as protein concentration increases while intramolecular interactions don’t, concentration is a major driver of condensation. For well-folded globular proteins, however, intra-molecular interactions are highly specific and persistent, defining the molecule’s tertiary structure. By comparison, intrinsically disordered proteins (IDPs) or regions (IDRs), which do not adopt consistent tertiary or secondary structures, are characterized by a lower propensity for intramolecular interactions. They often display a heterogeneous conformational landscape which is sensitive to their environment. Proteins or conformations that can contribute multiple intermolecular interaction motifs exhibit multivalency, a hallmark driver of biomolecular condensation [2,10]. While structured domains capable of multivalent interactions have been observed to drive condensation [10,11], disordered proteins or proteins containing a combination of disordered and folded regions make up the majority of condensate systems that have been studied so far. Ultimately, the degree to which a protein is poised to undergo condensation depends on the relative energies of available intramolecular and intermolecular interactions and how they can be modulated by the environment, typically through changes in concentration, ionic strength, temperature, crowding and external interaction partners.

Protein interactions and conformational dynamics in condensates

One of the most studied proteins in the LLPS field is the RNA-binding protein FUS (Fused In Sarcoma), which plays a role in transcriptional activation, splicing, and RNA transport. Its N-terminal domain is a so-called low-complexity (LC) domain and is composed primarily of serine, tyrosine, glycine, and glutamine residues, containing only two charged amino acids. FUS LC mediates protein interactions, can undergo liquid-liquid phase separation, and has been implicated in neurodegenerative diseases. In two studies, Fawzi and colleagues focused on uncovering the molecular mechanism of FUS LC condensation at the residue level using an arsenal of NMR techniques, Raman spectroscopy, fluorescence microscopy and simulations [12,13]. Using pulsed field gradient NMR [14,15] the authors characterized diffusivity of dispersed and condensed FUS LC. Condensed FUS LC diffused about 500 times slower than its dispersed counterpart. But basic and advanced NMR techniques spanning multiple timescales as well as Raman spectroscopy could detect no evidence for structured states in this condensed phase, indicating that condensed FUS LC is structurally similar to its dispersed counterpart. To study the interactions that stabilize the condensed state, the authors used nuclear Overhauser effect-based methods that report on inter- but not intramolecular interactions [15]. The authors observed that all major residue types participate in the interactions and are distributed over the entire domain. Simulations further confirmed that the obtained results are compatible with disordered FUS LC engaging in promiscuous weak interactions but incompatible with structured conformations. The interactions include nonspecific hydrogen bonding, hydrophobic interactions, and π/sp² interactions. The heterogeneity and redundancy of the interaction modes likely allow for multivalency without reducing FUS LC’s conformational entropy too drastically. Indeed, no significant change in structure or dynamics occurred in FUS LC when the protein condensed. These results suggest that, under the given experimental conditions, dispersed FUS LC already adopts a conformational ensemble that is compatible with condensation and that its conformational freedom is not substantially reduced by the interactions driving condensation (Fig. 1a). While the critical concentration of FUS LC needed for condensation can be reduced by increasing the NaCl concentration, this effect seems to be linked to salting out of hydrophobic regions rather than conformational changes as the NMR spectra of dispersed monomers remained unchanged between 0 and 150 mM NaCl.

Figure. 1.

Condensate systems with varying degrees of conformational freedom in the condensed phase. (a) FUS LC retains its conformational freedom when transitioning from the dispersed to the condensed phase. (b) In the dispersed phase, Tau K18 is conformationally heterogeneous and has a propensity to adopt compact conformations. In the condensed phase, it preferentially adopts an elongated state that features increased torsional motion. (c) NPM1 can switch between a disordered monomeric state, and a compact as well as an extended pentameric state depending on ionic strength, phosphorylation state and the presence of binding partners (not shown). The propensity to undergo phase separation is influenced by the same factors and likely coupled to the conformational and oligomeric equilibria. The red and blue spheres represent one of its acidic and basic tracts, respectively. (d) As temperature and salt concentration increase, the entropic cost of solvating ELP’s hydrophobic domains (HD, yellow spheres) triggers phase separation. If the protein concentration is too low for phase separation, dispersed monomers undergo an intramolecular collapse. The figure was inspired by work from references [12,13,16,17,19,24] and created with BioRender.com.

In contrast to the example of FUS LC stand mechanistic studies of proteins that undergo clear changes in the conformational landscape in the condensed state. Majumdar et al. studied intramolecular proximity and conformational fluctuations of a fragment of the protein tau, termed K18 [16]. K18 adopts an ensemble of interconverting conformations with some degree of order-promoting structures and a propensity to adopt collapsed conformations. Condensation of K18 displays a lower critical solution temperature, i.e., condensation does not occur below a temperature of about 15 °C. The authors used pyrene as an intramolecular proximity probe; this fluorophore can form an excited-state dimer when two monomers are less than 10 Å apart. Upon heating the sample above the critical temperature, as condensation of K18 progressed, fluorescence of the excited-state dimer decreased, indicating a transition from a compact conformation in the dispersed state to more extended structures in the condensed state. Using time-resolved fluorescence anisotropy, the authors showed that the protein exhibits increased local backbone dynamics in the condensed state and speculated that both the unwinding of monomers and the increased fluctuations could contribute to the entropic driving force for condensation (Fig. 1b). It seems that the lower critical solution temperature represents the tipping point beyond which it is energetically more favorable for K18 to adopt the extended, high-entropy condensate state rather than its collapsed, dispersed conformation stabilized by intramolecular interactions.

An example of a more layered coupling between conformation and phase separation involves nucleophosmin (NPM1), an oligomeric nucleolar protein involved in ribosome biogenesis. Mitrea et al. showed that NPM1 adopts different conformations and oligomerization states depending on the ionic strength of the solvent, its phosphorylation state, and the availability of binding partners [17–19]. NPM1 has several acidic and basic tracts which can interact with each other within NPM1 molecules as well as with other proteins and ribosomal RNA. At intermediate salt concentrations and in the absence of binding partners, the protein can undergo homotypic LLPS. The resulting condensate composition depends on the availability of binding partners that can outcompete NPM1 self-association. Interestingly, the response of the protein to environmental triggers, here, ionic strength, could be gleaned from its dispersed state. Through the FRET efficiency as a proxy for compactness, the authors show that NPM1 adopts an increasingly extended conformation as salt concentration increases, indicating that electrostatic intramolecular interactions between acidic and basic tracts drive self-association and compaction (Fig. 1c). Deleting key tracts in the protein lifted this dependence between salt concentration and compaction. The study reveals an intricate control of NPM1’s fate depending on its environment. The ionic strength of the medium dictates the compaction of NPM1 whereas the available binding partners and other NPM1 molecules compete for interactions with the available acidic and basic tracts of extended NPM1 molecules.

Building on previous polymer physics work [20] and more recent work in the biological condensation field [21,22], Mittag, Pappu and colleagues adapted a predictive “stickers and spacers” model to describe the condensation behavior of prion-like domains by analyzing the valence and patterning of their aromatic residues [23]. In this work, they initially focused on the low complexity domain of heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1 LC). Similar to FUS LC, no persistent secondary structure could be observed for hnRNPA1 LC by solution NMR. According to the model, stickers (aromatic residues in this case) cluster together in weak interactions and their number corresponds to the valence of the proteins or peptides. Charged and polar residues do not interact but rather act as spacers between the aromatic stickers. Below the critical solution temperature, the cohesive interactions provided by the stickers drive intramolecular compaction of the peptide chains and above it they drive condensation. Both valence and patterning of spacers and stickers can be adapted to increase or decrease intermolecular cohesion. We note that, whether—and which—motifs can be considered stickers and spacers could be strongly context-dependent [5,21]. Furthermore, while this polymer physics model can provide a useful framework for understanding the thermodynamics of condensation behavior, testing its broad applicability is in early stages.

The tendency of proteins to adopt compacted conformations under environmental conditions that are detrimental to condensation, effectively shielding their interaction motifs from potential interaction partners, is common but not universal. Reichheld et al. studied structure and dynamics of an elastin-like polypeptide (ELP) that undergoes phase separation with a lower critical solution temperature [24]. ELP features hydrophobic domains that are highly disordered but transiently populate β-turns. In these domains, heat and high concentrations of salt promote a hydrophobic collapse which triggers condensation. The authors used solution and solid-state NMR to study changes in structure and interactions upon condensation. They found that, in the condensed state, nonspecific intermolecular contacts between ELP monomers became more frequent and diffusion slowed down, as expected. But local structural elements remained unchanged and disorder was retained, indicating that protein chains in the condensate experience an environment that is qualitatively similar to a concentrated solution of monomers. When the authors studied ELP at condensate-promoting conditions (high temperature and high salt concentrations) but keeping the concentration below its critical value, they found that the monomers exhibited faster diffusion and reduced relaxation rates indicating increased compaction of monomers. In this case, the hydrophobic effect drives condensation by rendering protein:solvent interactions energetically unfavorable, destabilizing conformations of ELP that expose hydrophobic patches. In the absence of interactions partners this is apparent in the ELP monomers’ collapse, which shields their hydrophobic domains from the solvent, whereas they condense into droplets if interaction partners are present (Fig. 1d).

We’d like to mention that the exquisite sensitivity of IDPs, LLPS, and condensation behavior to the environment implies that, given the appropriate conditions, one could observe many changes in behavior that are not necessarily directly physiologically relevant, although they might still provide useful information about the underlying physical and chemical principles of the system being studied. This also means that representations of transitions such as those we draw in Figure 1 always depend on the perspective (i.e., the initial experimental conditions) from which observations were made. In the examples above, proteins can switch between condensate-compatible and incompatible states depending on environmental conditions like temperature or ionic strength. The existence of such an equilibrium could be an opportunity to design new drugs that aim at promoting or preventing formation of certain condensates. For example, a drug could aim at stabilizing condensate-incompatible states characterized by strong intramolecular interactions, which incidentally are structurally more homogeneous and might be more druggable than disordered states. This idea is somewhat similar to the concept of polyphasic linkage, where ligands preferentially interact with proteins in the dispersed phase compared to the condensed phase [25]. Recently, polyphasic linkage was successfully applied using the protein Profilin. Profilin reduced phase separation as well as aggregation of N-terminal fragments of huntingtin through preferential binding of dispersed monomers and oligomers [26]. As polyphasic linkage modulates the relative populations of the dispersed and condensed phases, we wonder if it is possible to modulate the relative populations of condensate-compatible and incompatible conformational states using ligands.

The environment of the droplet surface

So far, we’ve talked about proteins being either in the dispersed phase, i.e., in contact with the solvent, or the condensed phase. The boundary between both phases, the droplet surface, constitutes an environment where molecules are exposed to both the bulk of the dense phase and the solvent of the dilute phase. This makes the surface a unique environment where both protein:solvent and protein:protein interactions occur. Thus, a protein’s physical characteristics (including conformation and dynamics) and function could be quite different at these interfaces. Also, from a biological viewpoint, the surface of a droplet plays a functional role because it can promote or discourage droplet fusion, growth and content exchange with the dilute phase, regulating droplet size and composition. Being able to fuse and grow in size is expected to be important for the functioning of some condensates [27] and must be avoided for others [28,29]. Therefore, both the environment and the functional demands at the surface of a droplet can be different compared to the dense phase. A recent study attests to the importance of the interfacial properties for the regulation of condensate size. Folkmann et al. investigated the mechanisms that control the size of P granules, condensates in the Caenorhabditis elegans germline [30]. They studied the effect of the MEG-3 protein on condensates of PGL-3, one of the proteins constituting the liquid like phase of P granules. MEG-3 had been reported previously to adsorb onto the surface of PGL-3 droplets without entering into the bulk of the dense phase in newly fertilized zygotes [31,32]. In absence of MEG-3, PGL-3 condensates fused readily over a timescale of 180 minutes, increasing the average droplet size. When MEG-3 was added, however, the increase in size of the droplets was reduced in a dose-dependent manner without affecting the internal dynamics of the PGL-3 condensates. Another example of formation of dynamic surfaces was provided during condensate dissolution in a model peptide-RNA system showing reentrant phase behavior. Here, Banerjee, Milin et al. observed formation and loss of droplet substructures and corresponding interfaces in a complex and dynamic process [33]. Given the unique environment of the phase boundary and the importance of its properties for condensate size, we expect more examples to emerge of condensates featuring a layer of molecules whose nature or behavior differs markedly from those of the dense phase. These considerations of droplet interfaces are expected to be widely relevant in cells, as a number of cellular membraneless organelles have complicated substructures with numerous mesoscale and nanoscale interfaces [34].

Topological and entanglement aspects in biological condensation

Much attention has been paid to the particular types of interaction motifs that drive condensation. We discuss here whether condensation might also occur or be modulated independently of such “sticky” interactions, merely through topological constraints and entanglement. The general idea is inspired from the polymer physics literature dating several decades back. For example, in the 1970s, de Gennes mentioned the concept of an “Olympic gel”, where the gel network is held together by the topological constraints of interlocking polymer rings (hence “Olympic”) rather than by cohesive interactions [35]. A biological example of this type of topologically constrained gel is the kinetoplast DNA of Leishmania tarentolae, consisting of thousands of interlocked DNA circles [36–38]. Numerous other motifs could function as “mechanical bonds” in condensates [39] such as the wide variety of transiently knotted macromolecular structures found in biology [40,41], possibly including proteins like huntingtin and α-synuclein [42,43].

The idea of topological constraints can be viewed more generally from the perspective of physical exclusion. In the Olympic gel, the exclusion concept relates to non-crossing of chains leading to the linked circles or loops. A related and prior concept from the polymer physics developments several decades back, is the tube model of Edwards and Doi describing the physics of polymer melts and concentrated solutions [44,45]. Here, a given polymer chain can be considered to exist in a “tube” that is formed by the entangled network of other surrounding polymer chains. An interesting prediction of this model was provided by de Gennes: the motion of individual polymer chains would follow reptation behavior through the above-mentioned (flexible) tube. Perkins et al. provided direct evidence for this behavior using single-molecule optical tweezers experiments [46]. Such reptation should be an important aspect of the motion and related function for long polymeric IDPs and RNA species within condensates [47].

Chain non-crossing results in chain entanglement, allowing long polymer chains to form more robust networks and condensates. This idea was invoked in the work of Guillén-Boixet et al. [48] which indicated that condensates formed by the stress granule protein G3BP can reduce entanglement and formation of stable clusters of long RNAs. Another example that may reflect a contribution of RNA entanglement is the work by Ma et al. [49]. They showed that long unstructured mRNA regions in TIS granules form mesh like condensates that can inhibit fusion of smaller condensates, while still containing mobile protein components.

Concluding remarks

In this perspective, we have briefly highlighted a few key concepts illustrated by recent studies in the area of protein condensation. While important biological themes such as the ones we noted are developing in this research area, this still-emerging field awaits broader theoretical, experimental and in silico studies to provide critical support for the generality of these ideas as well as testing contrasting hypotheses and their functional outcomes. For example, improved and wider studies of protein conformational distributions, dynamics and interactions in condensed states are anticipated. Of note is that condensed states of these systems can be heterogeneous and dynamic, and methods to probe these properties as a function of these different states would be critical to develop and use. A variety of methods will be developed and applied in this regard [50–54]. We note the particular strengths of fluorescence and other spectroscopic methods for this purpose, due to their abilities to probe such dynamic species and processes with high spatiotemporal resolution in native environments in cells and tissues and with single-molecule sensitivity.

We nonetheless note that higher resolution studies may need modifications and refinements in existing methods as compared to their more commonly used modalities for studies of homogeneous solutions. This is relevant not only to the dynamically changing material state distributions of dense phase soft matter, but also to molecular layers at or close to interfaces. Indeed, a large body of physical chemistry research demonstrates the potential of such surfaces and interfacial layers to dramatically influence molecular properties, chemistry and catalysis. Along these lines, we believe that conceptual understanding and experimental progress will continue to draw heavily upon polymer and colloid physics and physical chemistry and will develop in new directions that are unique to biological condensates. Finally, in addition to providing new insights into biological function, concepts and tools for this research will likely continue to be leveraged in the development of therapeutic strategies against multiple diseases [3] and tunable and environment-responsive soft materials for medical and industrial applications [55]. Examples along these lines include tunable and tissue integrating networked peptide materials [56], engineered cellular organelles allowing spatially controlled production of proteins with expanded genetic codes [57], potential for networked soft matter with long-range (“allosteric”) effects [58], feedback-integrating [34,59–62] or spatiotemporally controlled nucleation-based [63] responses to biological stimuli, control of localization and activity of small-molecule therapeutics [64], and potential for modulation of condensation related larger-scale changes in tissue [65], to note just a few from a rapidly expanding list.

Highlights.

• We discuss concepts and recent studies focused on structure, dynamics and interactions of proteins undergoing condensation.

• We discuss the unique environment of the droplet surfaces and their relevance.

• We also speculate on the effects of topological constraints and entanglement on condensate properties.