Condensates with different properties have significantly different growth patterns and coarsening kinetics.

(A) A schematic for the use of a chemical dimerizer to induce condensate formation. (B) The ratio of the disordered protein condensate partition coefficient, viscosity, and surface tension to the coiled-coil condensate. (C,D) FRAP images and curves for a coiled-coil protein condensate (in C) and a disordered protein condensate (in D). Black lines are the exponential fits. Inset schematics are the predicted structure of the coiled-coil Mad1 protein (in C) and the disordered RGG domain of LAF-1 protein (in D) from AlphaFold2. (E,F) Fusion images and plots of aspect ratio over time for coiled-coil condensates (in E) and discorded protein condensates (in F). Black lines are the exponential fits. (G) Representative U2OS cell nucleus (magenta with DNA staining) containing condensates (green) formed by the coiled-coil protein imaged over time. Box indicates condensate that shrinks. (H) Condensate radius vs time for the six coiled-coil protein condensates shown in G. (I) Representative U2OS cell nucleus (magenta with DNA staining) containing disordered protein condensates (green) imaged over time. Boxes indicate condensates that shrink over time. (J) Condensate radius vs time for the six disordered condensates shown in I. (K) Fraction of growth types of the coiled-coil and the disordered protein condensates. Fusion events were scored as the coalescence of condensates, ripening events are characterized as the number of condensates shrinking while the remaining condensates grow, and diffusion-based growth is scored as continuous growth in the absence of ripening and can occur alongside fusion events (n.s., no significance; ***, p<0.001). (L) Change of average condensate radii over time. Condensate radius was normalized to the average condensate size at nucleation and time is normalized to the time nucleation occurs in the cell. Dashed lines are linear fits yielding indicated slopes. Scale bar, 5 μm.

Distinct growth patterns emerge from the interplay between condensate surface tension and stiffness of the chromatin network.

(A) Radii vs time for two condensates (labelled in yellow and blue) undergoing diffusive growth (suppressed ripening). The solid lines indicate model fits (see Methods for details) to the experimental data (solid circles). (B) Radii vs time for two condensates (labelled in yellow and blue) undergoing Ostwald ripening. Here the condensates have a larger surface tension compared to (A), with all other parameters fixed. (C) Phase diagram for condensate growth behavior as a function of normalized mean stiffness and surface tension , showing regimes of diffusive growth (stable) and Ostwald ripening. The parameter values used for panel A and B are , 4πD/3V = 10−5 μm−1 s−1, and the renormalized surface tension values are for panel A and for panel B. For panel C the values are and and other relevant parameters are same as in panel A & B.

Mechanical heterogeneity of the surrounding elastic network induces elastic ripening and slow growth of condensates.

(A) Phase portrait showing the dependence of time-derivative of the condensate size difference, , as a function of ∆R. Three distinct growth patterns emerge depending on the slope and the intercept of vs ∆R. (B) Phase diagram in the plane of surface tension and mean stiffness showing the parameter regimes for suppressed ripening, Ostwald ripening and elastic ripening. (C) Time evolution of the sizes of three growing coiled-coil condensates, exhibiting suppressed ripening. Solid lines are model fits to the experimental data. (D) Time evolution of the sizes of five disordered condensates, showing instances of ripening. Solid lines are model fits to the experimental data. (E) ∆R vs time for ripening droplets in disordered and coiled-coil condensates show qualitatively distinct trends. (Inset) Characterization of ripening events by obtaining the rate δ (in units of hour−1) by fitting linear theory (Eq. 5) to experimental data. (F) The numerical solution of the model (Eq. 4) for multiple condensates growing in a heterogeneous stiffness landscape predicts power-law scaling of mean condensate sizes during growth. We use high and low values of mean stiffness (⟨E⟩) and coefficient of variation (CVE) for proteins with low (coiled-coil) and high (disordered) surface tension to predict how different stiffness distributions affect condensate growth. Here ⟨R0⟩ is the mean initial radius of the condensates and the characteristic timescale t0 = 600 seconds. The parameters used are provided in Table. 1.

Chromatin heterogeneity affects the growth of coiled-coil condensates more than disordered condensates.

(A) Images of representative nuclei (magenta with SPY650DNA staining) with coiled-coil condensates (green) nucleated with LacI in a U2OS cell, a HeLa cell, and a TSA-treated U2OS cell, and that nucleated with Hotag3 in a U2OS cell (scale bar, 5 μm). (B) Distributions of the chromatin intensity for HeLa, U2OS, and TSA treated U2OS cells. (C) The variance of chromatin intensity for HeLa, U2OS, and, TSA treated U2OS cells. (D) Quantification of the growth types of coiled-coil condensates nucleated with LacI in HeLa, U2OS, and TSA treated U2OS cells, and that nucleated with Hotag3 in U2OS cells. (E) Quantification of the growth types of the disordered protein condensates in HeLa, U2OS, and TSA treated U2OS cells. (F) The average radii over time for coiled-coil condensates nucleated with LacI in HeLa, U2OS, TSA treated U2OS cells, and that nucleated with Hotag3 in U2OS cells. (G) The average radii over time for disordered condensates in HeLa, U2OS, TSA treated U2OS cells. In (F) and (G), the radii were normalized to the initial average droplet size by the cell and the time was normalized to the time condensates were nucleated.

Parameter values used in numerically solving the model.