ABSTRACT
Biomolecules exist and function in cellular micro-environments that control their spatial organization, local concentration and biochemical reactivity. Due to the complexity of native cytoplasm, the development of artificial bioreactors and cellular mimics to compartmentalize, concentrate and control the local physicochemical properties is of great interest. Here, we employ self-assembling polypeptide coacervates to explore the partitioning of the ubiquitous cytoskeletal protein actin into liquid polymer-rich droplets. We find that actin spontaneously partitions into coacervate droplets and is enriched by up to ≈30-fold. Actin polymerizes into micrometer-long filaments and, in contrast to the globular protein BSA, these filaments localize predominately to the droplet periphery. We observe up to a 50-fold enhancement in the actin filament assembly rate inside coacervate droplets, consistent with the enrichment of actin within the coacervate phase. Together these results suggest that coacervates can serve as a versatile platform in which to localize and enrich biomolecules to study their reactivity in physiological environments.
SIGNIFICANCE STATEMENT Living cells harbor many protein-rich membrane-less organelles, the biological functions of which are defined by compartment composition and properties. Significant differences between the physico-chemical properties of these crowded compartments and the dilute solutions in which biochemical reactions are traditionally studied pose a major challenge for understanding regulation of organelle composition and component activity. Here, we report the spontaneous partitioning and accelerated polymerization of the cytoskeletal protein actin inside model polypeptide coacervates as a proof-of-concept demonstration of coacervates as bioreactors for studying biomolecular reactions in cell-like environments. Our work introduces exciting avenues for the use of synthetic polymers to control the physical and biological properties of bioreactors in vitro, enabling studies of biochemical reactions in cell-like micro-environments.
INTRODUCTION
The biological functions of intracellular organelles are defined by the composition and properties of the compartments, which often differ significantly from that of bulk cytoplasm. Well known examples include the acidic pH of lysosomes and the mitochondrial redox potential (1). While the compartmentalization of these organelles require a lipid bilayer as a physical barrier, recent work has shown that organelles can also form as phase-separated droplets that do not require such a membrane (2, 3). The physicochemical properties of membrane-less organelles likely regulate partitioning and reactivity of biomolecules, thereby serving an important role in their physiological function. The compositional complexity of individual cellular bodies, granules, and organelles pose a major challenge in discerning general mechanisms for partitioning and reaction regulation. One useful strategy has been to reduce compositional complexity by in vitro reconstitution of cellular bodies (4, 5). However, the sequence and structural complexity of natural biopolymers make systematic variation of micro-environment properties difficult.
A complementary approach is to selectively tune the physical and chemical properties of phase-separated micro-environments through the rational design of synthetic polymers that spontaneously phase separate via known mechanisms, and then use these materials as a platform to study biomolecule partitioning and reactivity. For instance, charged homopolymers (polyelectrolytes) form polymer-dense liquid phases via complex coacervation (6, 7) and localize charged proteins (8–10). Precise chemical control of polypeptide-based polyelectrolytes allows for fine-tuning of several physio-chemical properties of the coacervate phase (7, 11), including functional groups, water content, viscosity, and surface tension, thereby enabling systematic investigations of protein interactions and activities in controlled micro-environments (12, 13). Knowledge of the general mechanisms by which micro-environment properties tune protein partitioning and activity could provide needed insight into the function of membrane-less organelles as well as design principles for synthetic biology and engineering applications.
Here, we report the spontaneous partitioning and polymerization of the cytoskeletal protein actin inside model polypeptide coacervates (14, 15) as a proof-of-concept demonstration of coacervates as bioreactors for studying biomolecular reactions in cell-like physical environments. Our results establish polyelectrolyte complex coacervates as a viable platform to study mechanisms of partitioning and biochemical regulation by controlled perturbation of condensed-phase micro-environment.
RESULTS
We use a model coacervate system (15) composed of the polycation poly-L-lysine (pLK) and the polyanion poly-(L,D)-glutamic acid (pRE), typically with ~100 amino acids per polypeptide (see Supplementary Materials and Methods, Table S1). Phase separation at room temperature is rapid; initially clear aqueous solutions become visibly turbid in seconds upon mixing of pLK- and pRE-containing solutions at total polypeptide concentrations of 10 μM or more (Movie S1), and is driven primarily by the release of condensed counterions (16). The presence of a polydisperse size distribution of polypeptide-rich coacervate droplets in solution, ranging in size from ~0.4 < R < 4 μm, is confirmed directly by differential interference contrast (DIC) microscopy (Fig. 1A-C, Fig. S1). The round, droplet-like appearance of the condensed pLK/pRE coacervate phase is suggestive of a fluid phase (15). Under similar conditions, the surface tension has been measured to be γ ~ 1 mN/m (17). Consistent with liquid-like properties on the timescale of seconds and longer, merging pLK/pRE droplets rapidly coalescence into a single, larger droplet (Fig. S1). From coalescence observations, we estimate the inverse capillary velocity v−1 = η/γ = 1.6 ms/μm (Fig. S1, (5, 18)). This yields a viscosity of η = 1.6 Pa·s, ~1000-fold higher than water. Thus, this simple model system is sufficient to create viscous phase-separated droplets with picoliter volumes.
Charged proteins spontaneously partition into coacervate droplets
Using a previously published protocol, proteins are mixed with the cationic pLK prior to initiation of phase separation by the addition of anionic pRE (8). It was previously found that the negatively-charged protein BSA localizes preferentially to pLK/pRE coacervates, and is uniformly distributed within them (Fig. 1A-D, Fig. S2). This preference for the coacervate phase is described quantitatively by a partition coefficient, defined as the ratio of fluorescence intensity inside to outside the coacervates (4). We find an average partition coefficient of PCavg ≅ 8, whether BSA is added to solution prior to or following phase separation (Fig. 1E, Fig. S3), indicative of spontaneous partitioning.
Here, we study the partitioning of actin, a cytoskeletal protein that self-assembles to form linear filaments (F-actin). Actin monomers and the chemically inert BSA are globular proteins of similar size (42 and 66 kDa, respectively) and carry comparable negative charge (isoelectric points of 5.23 and 5.60) (19). We find that actin partitions to pLK/pRE coacervates and immediately observe linear structures localized preferentially to the coacervate periphery (Fig. 1A-D). Integrating the total actin intensity within the droplet, we find an average partition coefficient that is 4-fold higher than that for BSA (Fig. 1E). Interestingly, the partition coefficients for BSA and actin are the same whether one or both proteins are present in solution (Fig. 1E, Fig. S2). This suggests that, under the conditions explored here, BSA and actin do not compete directly for space in the coacervate. Both partitioning and peripheral localization of actin are robust to the order of addition (Fig. S4).
Self-assembled F-actin of canonical structure localizes to the coacervate periphery
To test whether the linear actin structures are bona fide F-actin, we stained with fluorescently-labeled phalloidin (647-phalloidin). Phalloidin is a small, uncharged toxin recognized for its ability to specifically bind to F-actin (20). 647-phalloidin was introduced into the solution after the coacervate formation and actin assembly, and found to localize along the linear actin structures. Confocal fluorescence micrographs at both the coverslip surface (Fig. 2A,C) and droplet midplane (Fig. 2B,D) reveal strong co-localization of phalloidin fluorescence to the linear actin structures with a Pearson’s correlation coefficient of 0.86 (Fig 2E, Supplementary Materials and Methods). This provides strong evidence that these linear actin-rich structures are composed of F-actin of canonical structure. Given the brightness of the F-actin structures, and previous work demonstrating that concentrations of polycations (and pLK in particular) lower than those in the coacervate phase are sufficient to bundle F-actin (21), we presume that the structures visible in Figs. 1 and 2 are F-actin bundles, rather than individual filaments.
Actin assembly is enhanced in coacervates
Having demonstrated that the actin polymerization proceeds in pLK/pRE coacervates, we next ask to what extent the coacervate micro-environment impacts the reaction rate. Actin is a convenient model protein for this purpose owing to the existence of established spectroscopic tools for quantitatively monitoring assembly kinetics (22). In particular, the fluorescence intensity of the pyrene fluorophore increases ~20-fold when pyrene-labeled monomers are incorporated into filaments and is a well-established method to track actin assembly (22), as depicted in the schematic in Fig. 3A. In solution, the polymerization time course of 1.5 μM actin shows a characteristic lag phase, indicative of the kinetically slow filament nucleation step (23), followed by a phase of rapid growth and then saturation once a steady-state is reached (Fig. 3A) (22). At this actin concentration, the initial lag phase is typically ~10 min and steady-state is reached in ~120 min (Fig. 3B, black). The presence of pLK/pRE coacervates eliminates the lag phase and steady-state is achieved within 10 minutes (Fig. 3B, red). Thus, actin filament assembly is stimulated significantly by pLK/pRE coacervates.
To assess reaction kinetics quantitatively, we estimate the assembly rate 1/t1/2, defined as the inverse of the time at which the pyrene fluorescence intensity reaches half of its relative change during the course of actin polymerization (Fig. 3A, Supplementary Materials and Methods). The actin assembly rate 1/t1/2 increases from 0.03 to >1 min-1 as the total pLK concentration increases from 0.3 to 30 μM, while maintaining a pLK:pRE ratio of 1 (Fig. 3C). Above 30 μM, the assembly rate saturates. Thus, the actin assembly rate is enhanced by nearly two orders of magnitude in the presence of coacervates (Fig. 1, 2, 4).
Polylysine and coacervates stimulate actin assembly via distinct mechanisms
One possible explanation for the enhanced assembly rate is polycation-mediated F-actin nucleation. Polylysine has been shown to promote formation of antiparallel actin dimers (24) that nucleate F-actin (25, 26). Spontaneous assembly of pyrene-labeled actin in the presence of pLK shows a concentration-dependent increase in the rate of actin assembly (Fig. 3B,C, blue data). It is tempting to compare the filament formation rate in solution directly to the rates observed within coacervates. However, the local pLK concentration within the coacervate phase is actually much higher, on the order of 1-3 M, such that the pLK concentrations reported in Fig. 3C should not be directly compared. Furthermore, pLK/pRE interactions within the coacervate could limit pLK-mediated antiparallel dimer formation.
To test whether pLK-stabilized antiparallel actin dimers contribute to the assembly of F-actin in pLK/pRE coacervates, we monitored pyrene excimer fluorescence (24). In the absence of pLK, 1.5 μM actin displays no change in pyrene excimer fluorescence during the nucleation-dominated early phase of assembly (Fig. 3D, black). In the presence of pLK, excimer fluorescence is highest during the initial nucleation phase, and decays rapidly as assembly proceeds (Fig. 3D, blue). This excimer fluorescence time course is the hallmark of actin assembly mediated by pLK-stabilized antiparallel actin dimers (24). Importantly, in the presence of pLK/pRE coacervates, excimer fluorescence does not have these features characteristic of anti-parallel dimer-mediated nucleation events (Fig. 3D, red). These data strongly suggest that pLK-mediated nucleation is not the dominant mechanism by which actin assembly is enhanced in pLK/pRE coacervates.
Partitioning increases the local protein concentration in coacervates
A direct consequence of partitioning is that the local actin concentration in the coacervate phase is higher than that in the polymer-dilute phase. Thus, an alternate mechanism underlying enhanced assembly rates is an increased local actin concentration, clocal, within coacervates.
We tested this possibility by varying the global actin concentration, cglobal, from 0.01 μM to 1.5 μM. The threshold monomer concentration, or critical concentration c*, required for polymerization of Mg-ATP-actin is ≈0.1 μM (27, 28). If actin is concentrated in coacervate droplets ≈30-fold via partitioning, we would expect actin assembly within coacervates at global actin concentrations of ≈0.003 μM. Importantly, we observe coacervate-associated F-actin at global actin concentrations as low as 0.05 μM (Fig. 4A-C). Interestingly, we observe peripherally-biased partitioning of actin to pLK/pRE coacervates at all actin concentrations examined, even at the lowest concentration (0.01 μM) for which no filaments are clearly discernible. We note that the density of peripherally localized F-actin changes as a function of the global actin concentration. Whereas isolated filaments or bundles are visible at 0.05 μM, an F-actin shell too dense to resolve individual structures forms at 1.5 μM (Fig. 4A-C).
To systematically characterize the localization of actin fluorescence, we examined fluorescence intensity line scans through the midplane, depicted in Fig. 4B and shown in Fig. 4D. These curves may be divided into three regions: the droplet exterior (red), periphery (blue) and interior (black). Under each experimental condition, fluorescence is highest at the droplet periphery, followed by the droplet interior. The lowest fluorescence intensities are consistently observed exterior to droplets (Fig. 4E). The fluorescence increases nearly linearly with actin concentration both inside and outside the coacervate droplets (Fig. 4E).
In addition to the average partition coefficient (Fig. 1E), we report two additional partition coefficients derived from these intensity profiles; one for the ratio of droplet periphery (maximum observed intensity) to average intensity in the exterior of the droplet (PCPeriph) (Fig. 4F, blue/red), and a second for the ratio of droplet interior intensity (near the center of the droplet) to the average intensity in the exterior (PCInt) (Fig. 4F, black/red). Both partition coefficients are greater than unity, indicative of partitioning of actin to the polymer-dense coacervate phase from the polymer-dilute phase. PCPeriph and PCInt both tend to reach plateaus for actin concentrations above 0.1 μM; PCPeriph values grow almost 10-fold before stabilizing at ≈45 once the global actin concentration reached 0.1 μM, while PCInt values increase over the range of actin concentrations investigated in the current study, and appear to approach a plateau value of ≈10. The saturation of the PCPeriph with global actin concentration suggests that exchange of protein between the polymer-dense and dilute phases occurs readily, as has been reported in other liquid phase-separated systems (4).
DISCUSSION
We present proof-of-concept experiments demonstrating that a polypeptide-based complex coacervate can be used as a model bioreactor to control the localization and activity of the self-assembling cytoskeletal protein actin. We find that actin partitions spontaneously to the coacervate phase, and that its partitioning is not influenced by BSA. Strong partitioning of actin to pLK/pRE coacervates increases the local actin concentration, contributing substantially to a >50-fold increase in the actin assembly rate at the highest concentrations of actin and coacervate. Actin filaments of canonical structure localize to the coacervate periphery, effectively forming core-shell particles, with the actin shell density controlled by the actin concentration.
Partitioning vs. encapsulation of client proteins
Previous work interpreted the preferential localization of the client protein BSA to the pLK/pRE coacervate phase as “encapsulation” (8, 29, 30). The implication of this language is that exchange of client molecules between the coacervate and dilute phases is either non-existent or so small as to be negligible, as with encapsulation within lipid vesicles or emulsion droplets (31, 32). Indeed, Black et al. argued that entry of the large (66 kDa) client into the coacervate phase requires the formation of an intermediate electrostatic complex between the client and an oppositely-charged polyelectrolyte in solution prior to phase separation, and that client release is triggered by pH-induced dissolution of the coacervate phase (8).
Our present results are more consistent with a molecular view termed partitioning (4), where the partition coefficient reflects the equilibration of steady fluxes of client molecules into and out of the coacervate phase. For instance, we observe partitioning of BSA within ≈30 s upon addition to pre-formed pLK/pRE coacervates (Fig. S3). Given the very low polypeptide concentration in the dilute phase (< 30 nM pLK), this suggests that recruitment of the client to the coacervate does not require the formation of an intermediate complex with a polyelectrolyte. Additionally, the saturation of the partition coefficient for actin concentrations above 0.1 μM (Fig. 4F) is indicative of an equilibration between the client concentrations in the dilute and coacervate phases, which necessarily requires exchange.
Equilibrium partitioning in synthetic polypeptide coacervates is reminiscent of other recent in vitro work wherein client proteins of low-valency spontaneously partition into liquid phase-separated structures composed of high-valency scaffold proteins (4), as well as in coacervates formed from natural biopolymers (33). This is particularly interesting in that binding is mediated by specific low-affinity protein-protein interactions in the former case, in contrast to the non-specific electrostatic interactions presumed in the case of coacervates. This suggests that the capacity to selectively partition client molecules may be a general property of condensed liquid-like phases, independent of the interactions driving partitioning.
Origin of peripheral F-actin localization
Below, we examine three non-mutually exclusive physical mechanisms for the peripheral localization of F-actin in coacervates droplets: filament buckling, macromolecular depletion, and interfacial adsorption.
F-actin does not appear to protrude from micron-sized coacervate droplets, suggesting that coacervate surface tension may play a role in confining F-actin. One mechanism for peripheral filament localization is that surface tension causes filaments to buckle once the contour length exceeds the droplet diameter. A comparison of the energy required to increase the coacervate surface area to accommodate a protruding filament of length L and diameter d with a cylindrical cap, EArea = πd (L − 2R)γ, with the energy required to bend the filament into a circular arc with radius R equal to that of the droplet,
, yields the shortest length greater than 2R for which bending is energetically favorable: where kB is the Boltzmann constant, T is temperature, and lp = 10 μm is the persistence length of F-actin (34). In a 1-μm diameter coacervate droplet with the surface tension γ = 1 mN/m (17) at room temperature, bending is preferable for filaments longer than ≈1 μm. Coacervate surface tension is thus sufficient to bend F-actin with contour lengths larger than the droplet diameter. However, the observation that filaments and bundles even shorter than the droplet diameter are peripherally localized (Fig 4 A-C, 0.05 μM panels) indicates that surface tension-induced buckling cannot be the sole cause.
The peripheral localization of F-actin is reminiscent of the well-known crowding of F-actin to interfaces observed in the presence of macromolecular crowding agents (35, 36), which arises from depletion interactions (37). To assess this, we estimate the osmotic pressure needed to crowd F-actin to an interface to be Π* ≅ 450 Pa (Supplementary Information). We estimate the osmotic pressure of the coacervate interior as that arising from a solution of flexible polymers characterized by a mesh size (38), ξ, as Π = kBT/ξ3. This suggests that a coacervate with mesh size ξ ≤ 20 nm would generate sufficient osmotic pressure to drive peripheral localization of F-actin. We estimate the mesh size of our pLK/pRE coacervates to be 2-3 nm (Supplementary Information), which supports the plausibility of a depletion-based mechanism. Noting the empirical observation that long filaments crowd more readily than short ones, macromolecular depletion could preferentially crowd long, high aspect ratio filaments and bundles, leaving short filaments, actin monomer, and BSA uniformly distributed.
A third, non-mutually exclusive mechanism for peripheral localization of F-actin is filament adsorption to the coacervate/bulk interface. For example, electrostatic interactions between F-actin and the coacervate could drive adsorption in a process akin to that of polyelectrolyte-mediated emulsion stabilization (39). Alternately, a difference in the interfacial tensions between F-actin and the solution and the coacervate, respectively, could drive localization of filaments to the coacervate/solution interface, such as seen in Pickering emulsions (40). In support of an adhesion-based mechanism, we note that filaments occasionally wrap around coacervate droplets when assembling in solution (Fig. S5, Movie S2), indicative of an attractive interaction between F-actin and the coacervate/solution interface.
Importantly, not all actin fluorescence is peripherally localized. Actin fluorescence intensity in the center of pLK/pRE coacervates is diffuse and enriched by as much as 10-fold compared to the surrounding solution (Fig. 4F). Interior fluorescence increases with global actin concentration, which is inconsistent with the peripheral localization of all filaments. In that case, the interior fluorescence would correspond to solely actin monomers, which we would predict to have a constant local concentration of clocal = c* = 0.1 μM at steady-state. This suggests that the coacervate interior contains a mixture of monomers and filaments. Given that a 250-nm filament (~90 actin subunits) cannot be resolved with conventional light microscopy, an interior including both monomers and short filaments is consistent with the diffuse fluorescence signal we observe. The aforementioned mechanisms of peripheral localization all depend on F-actin length. As such, the presence of short filaments in the interior does not qualitatively distinguish between them. Understanding how peripheral localization is regulated will be an exciting avenue for future studies.
Mechanism of actin assembly enhancement
The assembly of actin within coacervates for global actin concentrations below the critical concentration is largely predicted by the 30-fold increase in the local actin concentration via partitioning into the coacervate phase. Using measured partition coefficients, we calculate the concentrations at which filament assembly within coacervates is expected and find that partitioning is sufficient to explain filament assembly down to 0.05 μM (red dashed line, Fig. 4F). However, our measured partitioning is not quite sufficient to polymerize actin within coacervates at the lowest concentration (0.01 μM), yet we observe strong peripheral intensity (and interpret this to be polymerized actin). It is possible that the coacervate environment may alter the reaction rate kinetics of actin assembly (41, 42), as has been seen for transcription in cell-lysate coacervates (43). Indeed, since the coacervate-phase volume fraction (40%) and viscosity (2 Pa-s) are similar to those of the cytoplasm, this system may serve as a useful platform to study biochemical reactions in a more physiological environment.
Implications for biochemical reaction regulation
The high local concentrations generated by partitioning provide for an elegant means to both spatially localize and enhance the rates of biochemical reactions. Spontaneous partitioning to a condensed liquid-like phase substantially reduces the quantities of protein needed to study reactions under more physiological conditions. Of particular interest is the possibility of having direct control over partition coefficients and other local physicochemical properties of the coacervate phase as a means to control biochemical reactivity.
In summary, we have illustrated spontaneous partitioning of proteins inside coacervate droplets, leading to markedly increased actin assembly rates and spatial confinement of filaments. Assembly rate enhancements reported here are qualitatively consistent with a model in which these enhancements were contributed by an increase in the local effective concentration of actin monomers in the coacervate phase. Our work introduces exciting avenues for the use of synthetic polymers to control physical and biological properties of bioreactors, and address questions in biology about the biochemistry of molecules in cell-like micro-environments.
MATERIALS AND METHODS
Solutions containing pLK/pRE coacervates and fluorescently-labeled proteins were imaged on a spinning disk confocal microscope. Details of all experimental methods and analysis can be found in the SI Materials and Methods.
ACKNOWLEDGEMENTS
The authors thank C. Suarez, K. Weirich, T. Witten, and J. Vieregg, as well as the Gardel, Kovar, and Tirrell labs, for helpful discussion and suggestions. We thank L. Li for assistance with gel permeation chromatography measurements. DRK, MLG, and MVT acknowledge support from the University of Chicago MRSEC (NSF DMR-1420709). PMM thanks the University of Chicago MRSEC for a graduate fellowship.