ATP-induced cross-linking of a biomolecular condensate

Abstract

DEAD-box helicases are important regulators of biomolecular condensates. However, the mechanisms through which these enzymes affect the dynamics of biomolecular condensates have not been systematically explored. Here, we demonstrate the mechanism by which the mutation of a DEAD-box helicase's catalytic core alters ribonucleoprotein condensate dynamics in the presence of ATP. Through altering RNA length within the system, we are able to attribute the altered biomolecular dynamics and material properties to physical cross-linking of RNA facilitated by the mutant helicase. These results suggest that mutant condensates approach a gel transition when RNA length is increased to lengths comparable to eukaryotic mRNA. Lastly, we show that this cross-linking effect is tunable with ATP concentration, uncovering a system whose RNA mobility and material properties vary with enzyme activity. More generally, these findings point to a fundamental mechanism for modulating condensate dynamics and emergent material properties through nonequilibrium, molecular-scale interactions.

Figure 1

Trapping an intermediate along the ATP-hydrolysis pathway induces slowed dynamics and altered material properties of a DEAD-box helicase-RNA condensate. (a) Mutation scheme for the LAF-1 DEAD-box helicase using well-characterized DEAD-box mutations. (b) Transient LAF-1-RNA association does not enable strong RNA-RNA interactions, but long-lived protein-RNA clamping results in RNA-RNA cross-linking. Inter-RNA linkages can potentially be formed through simultaneous RNA binding of the LAF-1 helicase domain and the RGG sequence to separate RNA molecules or through RGG-RGG interactions of LAF-1 on separate RNA molecules. (c) LAF-1 condensates cocondense with PolyU RNA in the presence of ATP. Scale bar = 5 μm. (d) Representative fluorescence recovery after photobleaching (FRAP) assay to study macromolecular diffusion in biomolecular condensates. Scale bar = 2 μm (e) FRAP curves and (f) extracted fit parameters for LAF-1 protein in LAF-1-PolyU condensates indicate that clamp-like mutant ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$, but not slowed ATPase mutant ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DAAD}$, experiences altered dynamics in the presence of ATP. Two timescales are apparent in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ recovery, with the faster timescale matching that of ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$. (g) FRAP curves and (h) extracted fit parameters for PolyU RNA in LAF-1-PolyU condensates indicate over an order of magnitude decrease in RNA diffusion in clamp-like ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates in the presence of ATP. FRAP data shown in (e)–(h) are from at least 12 droplets per condition and error bars are the standard deviation over all droplets measured (Table S2). (i) Tracking of fluorescent tracer particles in biomolecular condensates to measure rheological properties. Scale bar = 2 μm. (j and k) MSD plots for beads embedded in LAF-1-PolyU condensates indicate altered material properties for clamp-like mutant ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ in the presence of ATP, with diffusive, fluid-like dynamics for all other conditions. MSD data are from bead trajectories compiled across at least three droplets.

Figure 2

RNA length-viscosity dependence indicates RNA cross-linking within ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates. (a) Capillary electrophoresis results of heat- and magnesium-fragmented PolyU RNA for various times. Error bars within the inset represent the standard deviation of the mean RNA length over 3 fragmentation replicates. (b and c) MSDs of fluorescent tracer particles in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ and ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates have different responses to RNA length in the presence of ATP. (d) Effective viscosity for condensates containing ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ or ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ (blue and red, respectively), PolyU RNA of the indicated length, and 5 mM ATP, as extracted using the Stokes-Einstein relation. A sharp increase in effective viscosity for ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates is observed to begin at an RNA length of roughly 250 nucleotides (nt), consistent with near-critical gel-like behavior (32,33). Effective viscosities for individual droplets are shown as open circles, while average viscosities for a given condition are shown as closed circles. Measurements were taken from at least three condensates per condition and error bars are the standard deviation of individual droplet measurements (Table S3). (e) RNA-length-dependent cross-linking, wherein longer RNAs result in exponentially larger complexes. The system should eventually tend toward a percolated gel-like RNA network as a critical RNA chain length is reached (46,47).

Figure 3

Increased effective cluster size for RNA in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates. (a and b) LAF-1 protein FRAP curves for LAF-1 and ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates, respectively, with varying PolyU lengths and 5 mM ATP. (c) FRAP recovery timescales for LAF-1 protein FRAP from (a) and (b). Good agreement is seen between the ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ recovery timescale and the first timescale of the ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ fit, with no significant dependence on RNA length. ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$’s second recovery timescale scales with RNA length. FRAP measurements were performed on at least 12 droplets per condition and error bars are the standard deviation over all droplets measured (Table S2). (d) ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ FRAP curves are best fit by a double exponential. The free, diffusing protein species has a fast recovery timescale, while the protein bound to the supramolecular protein-RNA complex will have a slower timescale that is a combination of the complex recovery timescale and the protein release timescale. This second timescale should then increase with RNA length. (e and f) FRAP curves for labeled PolyU RNA in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ and ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates, respectively, with varying PolyU lengths and 5 mM ATP. (g) RNA FRAP recovery timescales fit to single exponentials. A stronger RNA length dependence on RNA recovery timescale is seen for the ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ mutant, consistent with the formation of supramolecular ribonucleoprotein complexes. The dotted line represents the expected scaling of recovery timescale with length based on predicted polymer size (35). FRAP measurements were performed on at least 12 droplets per condition and error bars are the standard deviation over all droplets measured (Table S2). (h) The RNA will have a single timescale that is either the average recovery of the supramolecular protein-RNA structures or the network’s recovery timescale. This second timescale will increase with RNA length faster than would be expected for single RNA monomers.

Figure 4

Titratable cross-link strength through ATP concentration. (a) FRAP experiments for labeled ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ in condensates containing PolyU RNA and varying concentrations of MgATP show a sharp transition to slowed dynamics above 0.1 mM ATP. (b and c) Recovery timescale (open square) and mobile fraction (closed diamond) fit parameters for the first (b) and second (c) timescales of ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ protein FRAP as a function of MgATP concentration. Consistent recovery timescales are observed for the first timescale, while the second, longer timescale increases with ATP concentration. (d) Titrating ATP results in tunable FRAP recovery dynamics for PolyU RNA in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates. (e) Recovery timescale (open square) and mobile fraction (closed diamond) fit parameters of FRAP curves from (d) as a function of MgATP concentration. A monotonic increase in PolyU recovery timescale (τ) is observed with increasing ATP concentration. FRAP measurements in (a)–(e) were performed on at least 12 droplets per condition and error bars are the standard deviation over all droplets measured (Table S2). (f) Titrating ATP concentration with the ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ variant results in tunable network properties via tunable average cross-link strengths. If a single protein can form an RNA cross-link, then titrating ATP will increase the overall number of long-lived cross-links (top row). If the cross-links are formed by multiple ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ proteins, then titrating ATP concentration will result in a distribution of cross-link lifetimes with an increasing average lifetime (bottom row). (g) MSD as a function of time lag for 100 nm fluorescent tracer particles in ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ and ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ condensates containing PolyU RNA of average length 310 nt and varying concentrations of MgATP. ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ and ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$ experience opposite responses to increased ATP concentration. (h) Effective viscosity for condensates containing ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$ or ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{DQAD}$, PolyU RNA of an average length of 310 nt, and the indicated concentration of MgATP. Error bars are the standard deviation over individual droplets, whose effective viscosities are shown in open circles, with measurements from at least three separate droplets per condition (Table S3). The dotted line indicates a hyperbolic fit to the data, reminiscent of ${\mathrm{L}\mathrm{A}\mathrm{F}\text{-}1}^{WT}$’s ATPase Michaelis-Menten fit.