Liquid-like condensates mediate competition between actin branching and bundling.

Cellular remodeling of actin networks underlies cell motility during key morphological events, from embryogenesis to metastasis. In these transformations, there is an inherent competition between actin branching and bundling, because steric clashes among branches create a mechanical barrier to bundling. Recently, liquid-like condensates consisting purely of proteins involved in either branching or bundling of the cytoskeleton have been found to catalyze their respective functions. Yet in the cell, proteins that drive branching and bundling are present simultaneously. In this complex environment, which factors determine whether a condensate drives filaments to branch or become bundled? To answer this question, we added the branched actin nucleator, Arp2/3, to condensates composed of VASP, an actin bundling protein. At low actin to VASP ratios, branching activity, mediated by Arp2/3, robustly inhibited VASP-mediated bundling of filaments, in agreement with agent-based simulations. In contrast, as the actin to VASP ratio increased, addition of Arp2/3 led to formation of aster-shaped structures, in which bundled filaments emerged from a branched actin core, analogous to filopodia emerging from a branched lamellipodial network. These results demonstrate that multi-component, liquid-like condensates can modulate the inherent competition between bundled and branched actin morphologies, leading to organized, higher-order structures, similar to those found in motile cells.

Steric pressure between glycosylated transmembrane proteins inhibits internalization by endocytosis.

Clathrin-mediated endocytosis is essential for the removal of transmembrane proteins from the plasma membrane in all eukaryotic cells. Many transmembrane proteins are glycosylated. These proteins collectively comprise the glycocalyx, a sugar-rich layer at the cell surface, which is responsible for intercellular adhesion and recognition. Previous work has suggested that glycosylation of transmembrane proteins reduces their removal from the plasma membrane by endocytosis. However, the mechanism responsible for this effect remains unknown. To study the impact of glycosylation on endocytosis, we replaced the ectodomain of the transferrin receptor, a well-studied transmembrane protein that undergoes clathrin-mediated endocytosis, with the ectodomain of MUC1, which is highly glycosylated. When we expressed this transmembrane fusion protein in mammalian epithelial cells, we found that its recruitment to endocytic structures was substantially reduced in comparison to a version of the protein that lacked the MUC1 ectodomain. This reduction could not be explained by a loss of mobility on the cell surface or changes in endocytic dynamics. Instead, we found that the bulky MUC1 ectodomain presented a steric barrier to endocytosis. Specifically, the peptide backbone of the ectodomain and its glycosylation each made steric contributions, which drove comparable reductions in endocytosis. These results suggest that glycosylation constitutes a biophysical signal for retention of transmembrane proteins at the plasma membrane. This mechanism could be modulated in multiple disease states that exploit the glycocalyx, from cancer to atherosclerosis.

Lipid droplets as substrates for protein phase separation.

Membrane-associated protein phase separation plays critical roles in cell biology, driving essential cellular phenomena from immune signaling to membrane traffic. Importantly, by reducing dimensionality from three to two dimensions, lipid bilayers can nucleate phase separation at far lower concentrations compared with those required for phase separation in solution. How might other intracellular lipid substrates, such as lipid droplets, contribute to nucleation of phase separation? Distinct from bilayer membranes, lipid droplets consist of a phospholipid monolayer surrounding a core of neutral lipids, and they are energy storage organelles that protect cells from lipotoxicity and oxidative stress. Here, we show that intrinsically disordered proteins can undergo phase separation on the surface of synthetic and cell-derived lipid droplets. Specifically, we find that the model disordered domains FUS LC and LAF-1 RGG separate into protein-rich and protein-depleted phases on the surfaces of lipid droplets. Owing to the hydrophobic nature of interactions between FUS LC proteins, increasing ionic strength drives an increase in its phase separation on droplet surfaces. The opposite is true for LAF-1 RGG, owing to the electrostatic nature of its interprotein interactions. In both cases, protein-rich phases on the surfaces of synthetic and cell-derived lipid droplets demonstrate molecular mobility indicative of a liquid-like state. Our results show that lipid droplets can nucleate protein condensates, suggesting that protein phase separation could be key in organizing biological processes involving lipid droplets.

Molecular mechanisms of steric pressure generation and membrane remodeling by disordered proteins.

Cellular membranes, which are densely crowded by proteins, take on an elaborate array of highly curved shapes. Steric pressure generated by protein crowding plays a significant role in shaping membrane surfaces. It is increasingly clear that many proteins involved in membrane remodeling contain substantial regions of intrinsic disorder. These domains have large hydrodynamic radii, suggesting that they may contribute significantly to steric congestion on membrane surfaces. However, it has been unclear to what extent they are capable of generating steric pressure, owing to their conformational flexibility. To address this gap, we use a recently developed sensor based on Förster resonance energy transfer to measure steric pressure generated at membrane surfaces by the intrinsically disordered domain of the endocytic protein, AP180. We find that disordered domains generate substantial steric pressure that arises from both entropic and electrostatic components. Interestingly, this steric pressure is largely invariant with the molecular weight of the disordered domain, provided that coverage of the membrane surface is held constant. Moreover, equivalent levels of steric pressure result in equivalent degrees of membrane remodeling, regardless of protein molecular weight. This result, which is consistent with classical polymer scaling relationships for semi-dilute solutions, helps to explain the molecular and physical origins of steric pressure generation by intrinsically disordered domains. From a physiological perspective, these findings suggest that a broad range of membrane-associated disordered domains are likely to play a significant and previously unknown role in controlling membrane shape.

Clathrin senses membrane curvature.

The ability of proteins to assemble at sites of high membrane curvature is essential to diverse membrane remodeling processes, including clathrin-mediated endocytosis. Multiple adaptor proteins within the clathrin pathway have been shown to sense regions of high membrane curvature, leading to local recruitment of the clathrin coat. Because clathrin triskelia do not bind to the membrane directly, it has remained unclear whether the clathrin coat plays an active role in sensing membrane curvature or is passively recruited by adaptor proteins. Using a synthetic tag to assemble clathrin directly on membrane surfaces, here we show that clathrin is a strong sensor of membrane curvature, comparable with previously studied adaptor proteins. Interestingly, this sensitivity arises from clathrin assembly rather than from the properties of unassembled triskelia, suggesting that triskelia have preferred angles of interaction, as predicted by earlier structural data. Furthermore, when clathrin is recruited by adaptors, its curvature sensitivity is amplified by 2- to 10-fold, such that the resulting protein complex is up to 100 times more likely to assemble on a highly curved surface compared with a flatter one. This exquisite sensitivity points to a synergistic relationship between the coat and its adaptor proteins, which enables clathrin to pinpoint sites of high membrane curvature, an essential step in ensuring robust membrane traffic. More broadly, these findings suggest that protein networks, rather than individual protein domains, are likely the most potent drivers of membrane curvature sensing.

Membrane fission by protein crowding.

Membrane fission, which facilitates compartmentalization of biological processes into discrete, membrane-bound volumes, is essential for cellular life. Proteins with specific structural features including constricting rings, helical scaffolds, and hydrophobic membrane insertions are thought to be the primary drivers of fission. In contrast, here we report a mechanism of fission that is independent of protein structure-steric pressure among membrane-bound proteins. In particular, random collisions among crowded proteins generate substantial pressure, which if unbalanced on the opposite membrane surface can dramatically increase membrane curvature, leading to fission. Using the endocytic protein epsin1 N-terminal homology domain (ENTH), previously thought to drive fission by hydrophobic insertion, our results show that membrane coverage correlates equally with fission regardless of the hydrophobicity of insertions. Specifically, combining FRET-based measurements of membrane coverage with multiple, independent measurements of membrane vesiculation revealed that fission became spontaneous as steric pressure increased. Further, fission efficiency remained equally potent when helices were replaced by synthetic membrane-binding motifs. These data challenge the view that hydrophobic insertions drive membrane fission, suggesting instead that the role of insertions is to anchor proteins strongly to membrane surfaces, amplifying steric pressure. In line with these conclusions, even green fluorescent protein (GFP) was able to drive fission efficiently when bound to the membrane at high coverage. Our conclusions are further strengthened by the finding that intrinsically disordered proteins, which have large hydrodynamic radii yet lack a defined structure, drove fission with substantially greater potency than smaller, structured proteins.

Molecular Mechanisms of Membrane Curvature Sensing by a Disordered Protein.

The ability of proteins to sense membrane curvature is essential for the initiation and assembly of curved membrane structures. Established mechanisms of curvature sensing rely on proteins with specific structural features. In contrast, it has recently been discovered that intrinsically disordered proteins, which lack a defined three-dimensional fold, can also be potent sensors of membrane curvature. How can an unstructured protein sense the structure of the membrane surface? Many disordered proteins that associate with membranes have two key physical features: a high degree of conformational entropy and a high net negative charge. Binding of such proteins to membrane surfaces results simultaneously in a decrease in conformational entropy and an increase in electrostatic repulsion by anionic lipids. Here we show that each of these effects gives rise to a distinct mechanism of curvature sensing. Specifically, as the curvature of the membrane increases, the steric hindrance between the disordered protein and membrane is reduced, leading to an increase in chain entropy. At the same time, increasing membrane curvature increases the average separation between anionic amino acids and lipids, creating an electrostatic preference for curved membranes. Using quantitative imaging of membrane vesicles, our results demonstrate that long disordered amino acid chains with low net charge sense curvature predominately through the entropic mechanism. In contrast, shorter, more highly charged amino acid chains rely largely on the electrostatic mechanism. These findings provide a roadmap for predicting and testing the curvature sensitivity of the large and diverse set of disordered proteins that function at cellular membranes.

The Physics of Entropic Pulling: A Novel Model for the Hsp70 Motor Mechanism.

Hsp70s use ATP to generate forces that disassemble protein complexes and aggregates, and that translocate proteins into organelles. Entropic pulling has been proposed as a novel mechanism, distinct from the more familiar power-stroke and Brownian ratchet models, for how Hsp70s generate these forces. Experimental evidence supports entropic pulling, but this model may not be well understood among scientists studying these systems. In this review we address persistent misconceptions regarding the dynamics of proteins in solution that contribute to this lack of understanding, and we clarify the basic physics of entropic pulling with some simple analogies. We hope that increased understanding of the entropic pulling mechanism will inform future efforts to characterize how Hsp70s function as motors, and how they coordinate with their regulatory cochaperones in mechanochemical cycles that transduce the energy of ATP hydrolysis into physical changes in their protein substrates.

Instability of the Human Cytochrome P450 Reductase A287P Variant Is the Major Contributor to Its Antley-Bixler Syndrome-like Phenotype.

Human NADPH-cytochrome P450 oxidoreductase (POR) gene mutations are associated with severe skeletal deformities and disordered steroidogenesis. The human POR mutation A287P presents with disordered sexual development and skeletal malformations. Difficult recombinant expression and purification of this POR mutant suggested that the protein was less stable than WT. The activities of cytochrome P450 17A1, 19A1, and 21A2, critical in steroidogenesis, were similar using our purified, full-length, unmodified A287P or WT POR, as were those of several xenobiotic-metabolizing cytochromes P450, indicating that the A287P protein is functionally competent in vitro, despite its functionally deficient phenotypic behavior in vivo Differential scanning calorimetry and limited trypsinolysis studies revealed a relatively unstable A287P compared with WT protein, leading to the hypothesis that the syndrome observed in vivo results from altered POR protein stability. The crystal structures of the soluble domains of WT and A287P reveal only subtle differences between them, but these differences are consistent with the differential scanning calorimetry results as well as the differential susceptibility of A287P and WT observed with trypsinolysis. The relative in vivo stabilities of WT and A287P proteins were also examined in an osteoblast cell line by treatment with cycloheximide, a protein synthesis inhibitor, showing that the level of A287P protein post-inhibition is lower than WT and suggesting that A287P may be degraded at a higher rate. Current studies demonstrate that, unlike previously described mutations, A287P causes POR deficiency disorder due to conformational instability leading to proteolytic susceptibility in vivo, rather than through an inherent flavin-binding defect.

Binding Properties of the Transforming Growth Factor-ß Coreceptor Betaglycan: Proposed Mechanism for Potentiation of Receptor Complex Assembly and Signaling.

Transforming growth factor (TGF) ß1, ß2, and ß3 (TGF-ß1-TGF-ß3, respectively) are small secreted signaling proteins that each signal through the TGF-ß type I and type II receptors (TßRI and TßRII, respectively). However, TGF-ß2, which is well-known to bind TßRII several hundred-fold more weakly than TGF-ß1 and TGF-ß3, has an additional requirement for betaglycan, a membrane-anchored nonsignaling receptor. Betaglycan has two domains that bind TGF-ß2 at independent sites, but how it binds TGF-ß2 to potentiate TßRII binding and how the complex with TGF-ß, TßRII, and betaglycan undergoes the transition to the signaling complex with TGF-ß, TßRII, and TßRI are not understood. To investigate the mechanism, the binding of the TGF-ßs to the betaglycan extracellular domain, as well as its two independent binding domains, either directly or in combination with the TßRI and TßRII ectodomains, was studied using surface plasmon resonance, isothermal titration calorimetry, and size-exclusion chromatography. These studies show that betaglycan binds TGF-ß homodimers with a 1:1 stoichiometry in a manner that allows one molecule of TßRII to bind. These studies further show that betaglycan modestly potentiates the binding of TßRII and must be displaced to allow TßRI to bind. These findings suggest that betaglycan functions to bind and concentrate TGF-ß2 on the cell surface and thus promote the binding of TßRII by both membrane-localization effects and allostery. These studies further suggest that the transition to the signaling complex is mediated by the recruitment of TßRI, which simultaneously displaces betaglycan and stabilizes the bound TßRII by direct receptor-receptor contact.

Structure of the Hsp110:Hsc70 nucleotide exchange machine.

Hsp70s mediate protein folding, translocation, and macromolecular complex remodeling reactions. Their activities are regulated by proteins that exchange ADP for ATP from the nucleotide-binding domain (NBD) of the Hsp70. These nucleotide exchange factors (NEFs) include the Hsp110s, which are themselves members of the Hsp70 family. We report the structure of an Hsp110:Hsc70 nucleotide exchange complex. The complex is characterized by extensive protein:protein interactions and symmetric bridging interactions between the nucleotides bound in each partner protein's NBD. An electropositive pore allows nucleotides to enter and exit the complex. The role of nucleotides in complex formation and dissociation, and the effects of the protein:protein interactions on nucleotide exchange, can be understood in terms of the coupled effects of the nucleotides and protein:protein interactions on the open-closed isomerization of the NBDs. The symmetrical interactions in the complex may model other Hsp70 family heterodimers in which two Hsp70s reciprocally act as NEFs.

Nuclear Magnetic Resonance Structural Mapping Reveals Promiscuous Interactions between Clathrin-Box Motif Sequences and the N-Terminal Domain of the Clathrin Heavy Chain.

The recruitment and organization of clathrin at endocytic sites first to form coated pits and then clathrin-coated vesicles depend on interactions between the clathrin N-terminal domain (TD) and multiple clathrin binding sequences on the cargo adaptor and accessory proteins that are concentrated at such sites. Up to four distinct protein binding sites have been proposed to be present on the clathrin TD, with each site proposed to interact with a distinct clathrin binding motif. However, an understanding of how such interactions contribute to clathrin coat assembly must take into account observations that any three of these four sites on clathrin TD can be mutationally ablated without causing loss of clathrin-mediated endocytosis. To take an unbiased approach to mapping binding sites for clathrin-box motifs on clathrin TD, we used isothermal titration calorimetry (ITC) and nuclear magnetic resonance spectroscopy. Our ITC experiments revealed that a canonical clathrin-box motif peptide from the AP-2 adaptor binds to clathrin TD with a stoichiometry of 3:1. Assignment of 90% of the total visible amide resonances in the TROSY-HSQC spectrum of (13)C-, (2)H-, and (15)N-labeled TD40 allowed us to map these three binding sites by analyzing the chemical shift changes as clathrin-box motif peptides were titrated into clathrin TD. We found that three different clathrin-box motif peptides can each simultaneously bind not only to the previously characterized clathrin-box site but also to the W-box site and the ß-arrestin splice loop site on a single TD. The promiscuity of these binding sites can help explain why their mutation does not lead to larger effects on clathrin function and suggests a mechanism by which clathrin may be transferred between different proteins during the course of an endocytic event.

Mechanical stress-activated integrin α5ß1 induces opening of connexin 43 hemichannels.

The connexin 43 (Cx43) hemichannel (HC) in the mechanosensory osteocytes is a major portal for the release of factors responsible for the anabolic effects of mechanical loading on bone formation and remodeling. However, little is known about how the Cx43 molecule responds to mechanical stimulation leading to the opening of the HC. Here, we demonstrate that integrin α5ß1 interacts directly with Cx43 and that this interaction is required for mechanical stimulation-induced opening of the Cx43 HC. Direct mechanical perturbation via magnetic beads or conformational activation of integrin α5ß1 leads to the opening of the Cx43 HC, and this role of the integrin is independent of its association with an extracellular fibronectin substrate. PI3K signaling is responsible for the shear stress-induced conformational activation of integrin α5ß1 leading to the opening of the HC. These results identify an unconventional function of integrin that acts as a mechanical tether to induce opening of the HC and provide a mechanism connecting the effect of mechanical forces directly to anabolic function of the bone.

A parametrically constrained optimization method for fitting sedimentation velocity experiments.

A method for fitting sedimentation velocity experiments using whole boundary Lamm equation solutions is presented. The method, termed parametrically constrained spectrum analysis (PCSA), provides an optimized approach for simultaneously modeling heterogeneity in size and anisotropy of macromolecular mixtures. The solutions produced by PCSA are particularly useful for modeling polymerizing systems, where a single-valued relationship exists between the molar mass of the growing polymer chain and its corresponding anisotropy. The PCSA uses functional constraints to identify this relationship, and unlike other multidimensional grid methods, assures that only a single molar mass can be associated with a given anisotropy measurement. A description of the PCSA algorithm is presented, as well as several experimental and simulated examples that illustrate its utility and capabilities. The performance advantages of the PCSA method in comparison to other methods are documented. The method has been added to the UltraScan-III software suite, which is available for free download from http://www.ultrascan.uthscsa.edu.

A role for an Hsp70 nucleotide exchange factor in the regulation of synaptic vesicle endocytosis.

Neurotransmission requires a continuously available pool of synaptic vesicles (SVs) that can fuse with the plasma membrane and release their neurotransmitter contents upon stimulation. After fusion, SV membranes and membrane proteins are retrieved from the presynaptic plasma membrane by clathrin-mediated endocytosis. After the internalization of a clathrin-coated vesicle, the vesicle must uncoat to replenish the pool of SVs. Clathrin-coated vesicle uncoating requires ATP and is mediated by the ubiquitous molecular chaperone Hsc70. In vitro, depolymerized clathrin forms a stable complex with Hsc70*ADP. This complex can be dissociated by nucleotide exchange factors (NEFs) that release ADP from Hsc70, allowing ATP to bind and induce disruption of the clathrin:Hsc70 association. Whether NEFs generally play similar roles in vesicle trafficking in vivo and whether they play such roles in SV endocytosis in particular is unknown. To address this question, we used information from recent structural and mechanistic studies of Hsp70:NEF and Hsp70:co-chaperone interactions to design a NEF inhibitor. Using acute perturbations at giant reticulospinal synapses of the sea lamprey (Petromyzon marinus), we found that this NEF inhibitor inhibited SV endocytosis. When this inhibitor was mutated so that it could no longer bind and inhibit Hsp110 (a NEF that we find to be highly abundant in brain cytosol), its ability to inhibit SV endocytosis was eliminated. These observations indicate that the action of a NEF, most likely Hsp110, is normally required during SV trafficking to release clathrin from Hsc70 and make it available for additional rounds of endocytosis.

Unraveling protein-protein interactions in clathrin assemblies via atomic force spectroscopy.

Atomic force microscopy (AFM), single molecule force spectroscopy (SMFS), and single particle force spectroscopy (SPFS) are used to characterize intermolecular interactions and domain structures of clathrin triskelia and clathrin-coated vesicles (CCVs). The latter are involved in receptor-mediated endocytosis (RME) and other trafficking pathways. Here, we subject individual triskelia, bovine-brain CCVs, and reconstituted clathrin-AP180 coats to AFM-SMFS and AFM-SPFS pulling experiments and apply novel analytics to extract force-extension relations from very large data sets. The spectroscopic fingerprints of these samples differ markedly, providing important new information about the mechanism of CCV uncoating. For individual triskelia, SMFS reveals a series of events associated with heavy chain alpha-helix hairpin unfolding, as well as cooperative unraveling of several hairpin domains. SPFS of clathrin assemblies exposes weaker clathrin-clathrin interactions that are indicative of inter-leg association essential for RME and intracellular trafficking. Clathrin-AP180 coats are energetically easier to unravel than the coats of CCVs, with a non-trivial dependence on force-loading rate.

Liquid-like condensates that bind actin drive filament polymerization and bundling.

Liquid-like protein condensates perform diverse physiological functions. Previous work showed that VASP, a processive actin polymerase, forms condensates that polymerize and bundle actin. To minimize their curvature, filaments accumulated at the inner condensate surface, ultimately deforming the condensate into a rod-like shape, filled with a bundle of parallel filaments. Here we show that this behavior does not require proteins with specific polymerase activity. Specifically, we found that condensates composed of Lamellipodin, a protein that binds actin but is not an actin polymerase, were also capable of polymerizing and bundling actin filaments. To probe the minimum requirements for condensate-mediated actin bundling, we developed an agent-based computational model. Guided by its predictions, we hypothesized that any condensate-forming protein that binds actin could bundle filaments through multivalent crosslinking. To test this idea, we added an actin-binding motif to Eps15, a condensate-forming protein that does not normally bind actin. The resulting chimera formed condensates that drove efficient actin polymerization and bundling. Collectively, these findings broaden the family of proteins that could organize cytoskeletal filaments to include any actin-binding protein that participates in protein condensation.

Ubiquitin-driven protein condensation stabilizes clathrin-mediated endocytosis.

Clathrin-mediated endocytosis is an essential cellular pathway that enables signaling and recycling of transmembrane proteins and lipids. During endocytosis, dozens of cytosolic proteins come together at the plasma membrane, assembling into a highly interconnected network that drives endocytic vesicle biogenesis. Recently, multiple groups have reported that early endocytic proteins form flexible condensates, which provide a platform for efficient assembly of endocytic vesicles. Given the importance of this network in the dynamics of endocytosis, how might cells regulate its stability? Many receptors and endocytic proteins are ubiquitylated, while early endocytic proteins such as Eps15 contain ubiquitin-interacting motifs. Therefore, we examined the influence of ubiquitin on the stability of the early endocytic protein network. In vitro, we found that recruitment of small amounts of polyubiquitin dramatically increased the stability of Eps15 condensates, suggesting that ubiquitylation could nucleate endocytic assemblies. In live-cell imaging experiments, a version of Eps15 that lacked the ubiquitin-interacting motif failed to rescue defects in endocytic initiation created by Eps15 knockout. Furthermore, fusion of Eps15 to a deubiquitylase enzyme destabilized nascent endocytic sites within minutes. In both in vitro and live-cell settings, dynamic exchange of Eps15 proteins, a measure of protein network stability, was decreased by Eps15-ubiquitin interactions and increased by loss of ubiquitin. These results collectively suggest that ubiquitylation drives assembly of the flexible protein network responsible for catalyzing endocytic events. More broadly, this work illustrates a biophysical mechanism by which ubiquitylated transmembrane proteins at the plasma membrane could regulate the efficiency of endocytic internalization.

Liquid-like condensates mediate competition between actin branching and bundling.

Cellular remodeling of actin networks underlies cell motility during key morphological events, from embryogenesis to metastasis. In these transformations there is an inherent competition between actin branching and bundling, because steric clashes among branches create a mechanical barrier to bundling. Recently, liquid-like condensates consisting purely of proteins involved in either branching or bundling of the cytoskeleton have been found to catalyze their respective functions. Yet in the cell, proteins that drive branching and bundling are present simultaneously. In this complex environment, which factors determine whether a condensate drives filaments to branch versus becoming bundled? To answer this question, we added the branched actin nucleator, Arp2/3, to condensates composed of VASP, an actin bundling protein. At low actin to VASP ratios, branching activity, mediated by Arp2/3, robustly inhibited VASP-mediated bundling of filaments, in agreement with agent-based simulations. In contrast, as the actin to VASP ratio increased, addition of Arp2/3 led to formation of aster-shaped structures, in which bundled filaments emerged from a branched actin core, analogous to filopodia emerging from a branched lamellipodial network. These results demonstrate that multi-component, liquid-like condensates can modulate the inherent competition between bundled and branched actin morphologies, leading to organized, higher-order structures, similar to those found in motile cells.

Liquid-like VASP condensates drive actin polymerization and dynamic bundling.

The organization of actin filaments into bundles is required for cellular processes such as motility, morphogenesis, and cell division. Filament bundling is controlled by a network of actin-binding proteins. Recently, several proteins that comprise this network have been found to undergo liquid-liquid phase separation. How might liquid-like condensates contribute to filament bundling? Here, we show that the processive actin polymerase and bundling protein, VASP, forms liquid-like droplets under physiological conditions. As actin polymerizes within VASP droplets, elongating filaments partition to the edges of the droplet to minimize filament curvature, forming an actin-rich ring within the droplet. The rigidity of this ring is balanced by the droplet's surface tension, as predicted by a continuum-scale computational model. However, as actin polymerizes and the ring grows thicker, its rigidity increases and eventually overcomes the surface tension of the droplet, deforming into a linear bundle. The resulting bundles contain long, parallel actin filaments that grow from their tips. Significantly, the fluid nature of the droplets is critical for bundling, as more solid droplets resist deformation, preventing filaments from rearranging to form bundles. Once the parallel arrangement of filaments is created within a VASP droplet, it propagates through the addition of new actin monomers to achieve a length that is many times greater than the initial droplet. This droplet-based mechanism of bundling may be relevant to the assembly of cellular architectures rich in parallel actin filaments, such as filopodia, stress fibers, and focal adhesions.