Session S23: Macromolecular Phase Separation in Biology I

Protein condensates as aging Maxwell fluids

Liquid-like protein condensates (LLPCs) are intracellular compartments that segregate material without the use of a membrane. The liquid-like behavior of the condensates is a defining characteristic and the viscosity, surface tension and other material properties determine how segregated species diffuse into and within condensates; they, thus, critically impact the biological function of the condensates. It has become increasingly clear that some LLPCs do not have time-independent material properties, but can, instead, transition to more solid, gel-like materials. Here, we present our efforts to quantify these new materials as they age in vitro. We measure the visco-elastic material properties of several proteins, (PGL-3, FUS and DDX4), by means of a combination of active and passive microrheology. At early times, we find that the droplets behave much like simple liquids but gradually become more elastic. Surprisingly, the changing mechanical properties can all be scaled onto a single master curve using one characteristic time scale which grows as the sample ages. This and other features we observe bear a striking resemblance to the behaviors observed in materials with glass-like aging. We consider protein condensates as soft glassy materials with age dependent material properties that we call Maxwell glasses. To gain insight into the molecular origins of this behavior, we present electron microscopy images of the condensates at different ages. Furthermore, we demonstrate how salt concentration tunes the characteristics of the aging process. Lastly, we speculate on possible molecular origins that might lead to the glass-like arrest we observe and how such arrest could be used for modulation of cellular biochemistry.   

Surface Fluctuations and Coalescence of Nucleolar Droplets in the Human Cell Nucleus

The nucleolus is a membraneless organelle embedded in chromatin solution inside the cell nucleus. By analyzing the surface dynamics and fusion kinetics of nucleoli in live human cells, we find that the nucleolar surface exhibits subtle, but measurable, shape fluctuations and the radius of the neck connecting two fusing nucleoli grows as r(t)~t^1/2 [1]. This is consistent with liquid droplets with low surface tension ~10^-6 Nm^-1 coalescing in a fluid of higher viscosity ~10³ Pa s, i.e. chromatin solution. We find the neck velocity, dr/dt, is comparable to the velocity of chromatin solution [2]. Surprisingly, nucleolar coalescence occurs in an active fluid, yet can be described by coalescence theory for passive liquid droplets, suggesting the measured quantities might be effective quantities. Our study presents a noninvasive approach, using natural probes to investigate material properties of the cell as well as to understand the physical interactions between nucleoli and chromatin solution [1,3].
1. Caragine CM et al, Phys. Rev. Lett. (2018)
2. Zidovska A et al, Proc. Natl. Acad. Sci. (2013)
3. Caragine CM et al, eLife (2019)   

Phase separation in the nucleus is limited by chromatin mechanics

Liquid-liquid phase separation is a fundamental mechanism underlying biological organization. While conventional theory predicts that a single phase-separated condensate would be energetically favored, both natural and synthetic condensates in cells typically appear as multiple dispersed droplets with suppressed growth dynamics. Here, we combine coarse-grained molecular dynamics simulations and theory of liquid-liquid phase separation to show that mechanical interactions with chromatin can constrain the size of droplets in the nucleus. The "gel-like" chromatin suppresses both droplet coalescence and ripening dynamics, resulting in a reduced scaling exponent for mean droplet radius versus time. Our work highlights the impact of the local mechanical environment on biomolecular condensate formation and growth, and further elucidates the role of mechanics in fundamental biological processes taking place in the cell nucleus.   

Subdiffusive Dynamics of Optogenetic Droplets Report on Local Chromatin Mechanics

DNA is organized into chromatin, a complex material which stores information and controls gene expression. Liquid-liquid phase separation is believed to be a principal mechanism governing its organization. Previous work with optogenetically activatable droplets demonstrated that liquid condensates displace chromatin. Here, we show that droplet growth dynamics are directly inhibited by chromatin. Generally, droplet radii follow a power law scaling with time, such that R~t^β. We observe an anomalously slow coarsening exponent during interphase but recover dynamics more consistent with classical theory when chromatin is condensed during mitosis. We show that this slowed growth is due to subdiffusion of individual condensates, a clear signature of elastic behavior in the nucleus. We further apply this framework to elucidate local mesoscopic mechanics of the nucleus in various biologically relevant contexts. Thus, our work demonstrates the use of engineered intracellular condensates as “probes” of local mesoscale nuclear organization.   

Converting Stochastic Assembly into an Assembly Line: Non-Equilibrium Droplet Dynamics Assists Ribosome Formation

The nucleolus is a large liquid-like membraneless organelle responsible for the majority of the processing of ribosomal components and the assembly of ribosomes, which involves hundreds of proteins. It has been suggested that one of the primary functions of the nucleolus is to concentrate these proteins with the ribosomal RNA (rRNA), thereby significantly enhancing the binding rates and enzyme reaction speed for ribosome assembly and post-translational modifications.

Here we expand on this idea by considering the non-equilibrium effects that arise from having a constant rRNA flow outward from the center of a nucleolus and an inward flow of ribosomal protein (rProtein) from the nucleoplasm. We show numerically and analytically that the binding of specific rProteins to rRNA can be localized within a well-defined radial shell inside the nucleolus instead of being homogeneously distributed. By giving the different rProteins different physical properties, the different rProteins can be confined to bind to the rRNA at different radial distances from the transcription centers within the nucleolus. Thus, as rRNA molecules diffuse outward through the nucleolus, the rProteins can be added in sequential order like an assembly line.   

Reentrant Liquid Condensation of Ribonucleoprotein–RNA Complexes

Intracellular Ribonucleoprotein (RNP) granules are membrane-less liquid condensates that dynamically form, dissolve, and mature into a gel-like state in response to a changing cellular environment. RNP condensation is largely governed by attractive inter-chain interactions mediated by low-complexity domains. Using an archetypal disordered RNP, fused in sarcoma (FUS), here we employ atomistic simulations to study how RNA, a primary component of RNP granules, can modulate the phase behavior of RNPs by controlling both droplet assembly and dissolution. Electrostatic interactions are found to be the primary driving force behind condensate formation. Monotonically increasing RNA concentration initially leads to droplet assembly via complex coacervation and subsequently triggers an RNP charge inversion, which promotes disassembly. We construct phase diagrams based on Droplet density and Shannon entropy calculations, wherein three distinct regimes can be identified based on RNA and peptide concentrations. Increasing salt concentration is found to inhibit the formation of liquid condensates and narrow the coexistence region. The internal organization and dynamics of the condensates are investigated as a function of RNA/peptide concentrations, RNA chain length and salt concentration.
  

Size selection of phase-separated liquid droplets in strain-stiffening elastic networks

Membraneless organelles are formed via phase separation inside cells, but it remains unclear how cytoskeleton affects this process. Recent experiments showed the size of separated droplets is controlled by the stiffness of the surrounding elastic network. This is in contrast to phase separation of liquid mixtures, which continue to coarsen indefinitely (Ostwald ripening). Motivated by these observations, we developed a model to investigate the coupling between the separating liquid mixture and the elastic network. We find that the elastic energy of distorted network effectively modifies the nucleation barrier and the critical nucleus size for the separating mixture. For a neo-Hookean elastic network we find that the coarsening proceeds indefinitely. On the other hand, for networks with sufficiently strong strain-stiffening considered in experiments we find that the elastic energy arrests the coarsening once droplets reach certain size. Furthermore, finite element simulations indicate that the interactions between droplets via the distorted elastic network are rather short-ranged, which explains why observed droplets in experiments achieve a uniform size.   

A hydrodynamic instability drives TPX2 protein droplet formation on microtubules and leads to branching microtubule nucleation

Microtubules are protein polymers with a variety of roles in biological cells. During cell division, many microtubules are generated by branching from the surface of preexisting microtubules. Recent work in vitro and ex vivo shows that the protein TPX2 drives the nucleation of branched microtubules by bringing augmin and the gamma-tubulin ring complex (gTuRC) to microtubules. It has also been shown that TPX2’s ability to condense into a liquid phase is important for branching. Using atomic force microscopy, fluorescence imaging, electron microscopy, and hydrodynamic theory, we show that the dynamics of liquid TPX2 are crucial for nucleating branched microtubules. Initially, TPX2 coats the surface of microtubules in seconds, producing a cylindrical, liquid tube around the microtubule with an even thickness of 13-17 nm. Then, this layer loses stability due to surface tension, producing discrete droplets regularly spaced by 140-250 nm via a Rayleigh-Plateau mechanism. These droplets bind augmin and gTuRC to the microtubule surface, thus localizing where branches can form. Together, our work explains how the liquid phase of TPX2 leads to branching microtubule nucleation.   

Motif Sequences and Intracellular Phase Separation

Intrinsically Disordered Proteins (IDPs) lack a unique folded structure, and yet perform diverse and important functions inside cells. Recent work suggests that some IDPs promote the formation of membrane-less organelles via phase separation, helping cells spatially organize their biomolecules. Classical theories of phase separation focus on homopolymers, but IDPs have evolved particular sequences of interacting motifs. How does an IDP’s motif sequence determine its physical properties? We propose a statistical physics model of IDPs to elucidate the relationship between motif sequence, the phase boundary, and the partitioning of proteins between phases.

We find that motif sequences strongly influence the physical properties of model IDPs. Intuitively, each sequence has its own preference for inter- versus intra-protein bonds, which favor a condensed versus a dilute phase, respectively. As a result, the concentration of IDPs required for phase separation depends strongly on the motif sequence. Our work demonstrates the emergence of spatial order from conformational disorder, a process which may play a key role in intracellular organization.
  

Coupling signaling cascades to membrane criticality

Cellular biology has long understood spatial organization to be a crucial feature for determining function. Recent evidence suggests that thermodynamic phase separation may explain a range of structures in Eukaryotic cells. In particular, proximity to a liquid-liquid critical point may underlie membrane domains that are often termed lipid rafts. Such domains have been implicated in the functioning of many signaling cascades by localizing components to particular domains. Addition of ligand may lead to the formation of signaling platforms, and perturbations to membrane criticality often modify signaling outcomes. In order to create a theory that explains the interplay between signaling cascades and membrane criticality, here we present a model and Monte-Carlo simulation framework for proteins coupled to their surrounding lipid membrane using a 2D Ising model. We have additionally developed schematic diagrams that predict the effects of domain size on the outcome of various simple signaling cascade motifs. Our model naturally explains how changes in domain size arising from perturbations to membrane criticality can lead to changes in the rates of interaction amongst signaling components, eventually leading to altered signaling outcomes.   

Wetting of Critical Membranes by Protein Droplets

Phase-separated liquid-droplets of protein and RNA have recently been found as ubiquitous structures in cells – coordinating reactions, organizing cellular machinery, and protecting sensitive biomolecules. In two dimensions, the plasma membrane is organized by heterogeneities in lipid composition, with membrane domains posited to be a consequence of proximity to a miscibility critical point. These structures are linked by components that interact with both 2D and 3D phases such as lipidated proteins that partition into specific membrane phases. This leads protein droplets to localize to particular membrane domains, as seen in the post-synaptic structure, the immunologic synapse, and components of cell adhesion. Here, we consider the underlying thermodynamics of this interaction, constructing a minimal Landau theory describing the wetting of protein droplets to a near-critical membrane. The resulting phase diagram shows non-standard wetting behavior, where surface criticality greatly enhances a prewetting-like regime where bulk and surface phase separate together. We buttress these theoretical predictions with simulations of nearly critical Ising surfaces coupled to coacervating lattice polymers and link existing experimental observations to this phase diagram.   

Probing the Dynamics of Optically Induced Protein Droplets with Single-Walled Carbon Nanotubes

Eukaryotic cells exhibit high levels of internal spatial organization, including a wide variety of membrane-less compartments. Protein-protein and protein-RNA interactions have been implicated as the primary factors responsible for these membrane-less structures, but how they are maintained and utilized by the cell is still poorly understood. Many proteins essential to the formation of these structures feature long intrinsically disordered regions (IDRs) that interact with each other and can induce demixing, as seen in many recent in vitro studies. However, quantitative characterization of these structures in vivo remains challenging due to the complexity of intracellular environments. Here, we use single-walled carbon nanotubes (SWNTs), which fluoresce in the near infrared and are photostable, to probe the local environment of optically induced protein droplets in vivo. The dynamics of SWNTs can reveal internal organization within these droplets and the influence of the intracellular environment on their mechanical properties.   

Measuring protein concentrations in biomolecular condensates via quantitative phase microscopy

Many compartments in eukaryotic cells are protein-rich biomolecular condensates formed via phase separation from the cyto- or nucleoplasm. Although knowledge of condensate composition is essential for a full description of condensate properties and potential functions, measurements of composition pose a number of technical challenges. To address these, we use quantitative phase microscopy and optical diffraction tomography to measure the refractive index of model condensates, from which the protein concentration may be inferred. Here, model condensates are formed by phase separation of purified protein constructs derived from the primarily disordered RNA-binding domain (RBD) of TAF15. Surprisingly, we find that phase separation of TAF15(RBD) is attenuated only weakly by salt (0.05-3 M KCl) or temperature (10-50 °C), suggesting that Coulombic and entropic interactions, respectively, play only minor roles in controlling the phase equilibria. Interestingly, we also find that partition coefficients determined by fluorescence microscopy dramatically underestimate protein concentrations in condensates. A simple model including inner filter and excited-state saturation effects suggests that the discrepancy stems primarily from reduced fluorescence quantum yields in condensates.