Phase transition of a disordered nuage protein generates environmentally responsive membraneless organelles

Abstract

Cells chemically isolate molecules in compartments to both facilitate and regulate their interactions. In addition to membrane-encapsulated compartments, cells can form proteinaceous and membraneless organelles, including nucleoli, Cajal and PML bodies, and stress granules. The principles that determine when and why these structures form have remained elusive. Here, we demonstrate that the disordered tails of Ddx4, a primary constituent of nuage or germ granules, form phase-separated organelles both in live cells and in vitro. These bodies are stabilized by patterned electrostatic interactions that are highly sensitive to temperature, ionic strength, arginine methylation, and splicing. Sequence determinants are used to identify proteins found in both membraneless organelles and cell adhesion. Moreover, the bodies provide an alternative solvent environment that can concentrate single-stranded DNA but largely exclude double-stranded DNA. We propose that phase separation of disordered proteins containing weakly interacting blocks is a general mechanism for forming regulated, membraneless organelles.

Graphical abstract

Figure 1

Ddx4 Spontaneously Self-Assembles to Form Organelles in Live Cells (A) Evolutionary relationships between the disordered regions of Ddx4 homologs and their domain architectures. Disordered regions (green) and locations of DEAD-box helicase domains (brown) are indicated. (B) Schematic showing the DEAD-box helicase domain of Ddx4 replaced with YFP before being transfected into HeLa cells. Ddx4^YFP organelles appear over time. (C) Differential interference contrast (DIC) and corresponding extended focus fluorescence intensity images of a HeLa cell expressing Ddx4^YFP. Ddx4^YFP forms dense, spherical organelles in the nucleus. Cells were stained with antibodies to visualize nucleoli, PML bodies, nuclear speckles, and Cajal bodies as indicated, revealing that Ddx4 organelles are entirely distinct from these other bodies. (D) The variation in total droplet volume with time is explained by the Avrami equation for nucleated growth (Supplemental Experimental Procedures Section 5). The time is measured from the appearance of the first droplet.

Figure 2

Ddx4^YFP Organelles Are Internally Mobile and Respond Rapidly to Changes in Environmental Temperature and Tonicity (A) Fluorescence recovery after photobleaching (FRAP) of a Ddx4^YFP organelle in a live HeLa cell at 37°C. Sample bleaching is indicated with a gray bar. 50% of the fluorescence signal is recovered within approximately 2.5 s post-bleach, corresponding to a diffusion coefficient of 3 ± 1 × 10⁻¹³ m² s⁻¹. (B) Cold shock induces condensation of sub-nuclear Ddx4^YFP droplets at low expression levels. Extended focus fluorescence intensity images showing the nucleus from a time series analysis of a HeLa cell expressing Ddx4^YFP undergoing cold shock. Images are shown at 2-min intervals. Prior to cold shock treatment, Ddx4^YFP had not reached the critical concentration for phase separation at 37°C and was diffuse in the nucleoplasm (first two frames). Rapid exchange of growth media at 37°C for media cooled on ice (time = 0) induced small Ddx4^YFP droplets to condense rapidly within the nucleus (purple line, number of droplets; blue line, total volume of droplets). Following cold shock, the number of Ddx4^YFP droplets decreased through a combination of coalescence and dissolution as the temperature rose. Scale bar, 5 μm (see Movie S2). (C) Extended focus fluorescence intensity image slices showing a section of the nucleus from a time series analysis of a HeLa cell containing Ddx4^YFP droplets undergoing osmotic shock. Images are shown at 2-min intervals. Axis labels, data colors, and scale as in (B). See Movies S3 and S4.

Figure 3

The N Terminus of Ddx4 Reversibly Forms Organelles In Vitro (A) Schematic showing the relationship between constructs of Ddx4 and the wild-type protein. Ddx4^N1 (residues 1–236) and Ddx4^N2 contain only the disordered N terminus. (B) DIC (left) and YFP fluorescence (right) images of (i) Ddx4^YFP organelles inside HeLa cells (scale bar, 2 μm) and (ii) 60:1 Ddx4^N1:Ddx4^YFP organelles formed in vitro at 150 mM NaCl (scale bar, 10 μm). (C) FRAP curve of a 10 μm diameter droplet containing Ddx4^N1 and recombinant, purified Ddx4^YFP at a molar ratio of 60:1 in 150 mM NaCl buffer at 20°C. The bleach period is indicated with the gray bar. 50% of the fluorescence signal is recovered after approximately 1 min, corresponding to a diffusion coefficient of 4 ± 1 × 10⁻¹³ m² s⁻¹. (D) Time series analysis of bright-field microscopy images of Ddx4^N1 (202 μM protein, 200 mM NaCl) with varying temperature, shown at 50 s intervals (scale bar, 50 μm). At 50°C, the sample was monophasic with low turbidity. Temperature was linearly decreased (4°C min⁻¹) from 50°C to 22°C. At 36°C, the turbidity of the sample rapidly increased concomitant with the emergence of an incipient dense phase containing concentrated Ddx4^N1. After holding at 22°C for 1 min, the sample was reheated to 50°C. At approximately 45°C during reheating, the condensed phase was completely dissolved and the turbidity of the solution returned to its initial turbidity. The thermal cycle was repeated with the same sample in situ (light green line), revealing that the changes in the droplet are fully reversible.

Figure 4

Quantitative Analysis and Interpretation of the Ddx4^N1 Phase Transition (A) The temperature at which the phase transition is observed, T_P, was determined as a function of protein concentration and ionic strength at pH 8. At a given ionic strength, the Flory-Huggins model of polymer phase separation quantitatively describes each curve. This yields two fitting parameters, the enthalpy and entropy changes of the transition, which report on the microscopic interactions between molecules. (B) The interaction parameters varied in a predictable way with increasing salt. The enthalpic contribution to the interaction parameter (i) was found to decrease as a function of increasing NaCl. This is quantitatively explained by fitting the curve to a screened coulomb potential (light blue, Equation S19). The non-ionic component of the enthalpy is close to zero, −0.058 ± 0.137 kJ mol⁻¹, the relative permittivity within the condense phase was 45 ± 13, and the average spacing between opposite charges is 13 ± 2 Å. The entropic contribution to the interaction parameter (ii) decreases slightly with increasing salt, fitted to Equation S20. The error bars represent the SE in the fitted parameters (Figure 4A). (C) The entropy and enthalpy values are correlated, suggesting that when the interactions are destabilized at higher salt, the chains in the interior of the droplet become more mobile. The error bars represent the SE in the fitted parameters (Figure 4A). (D) Schematic representation of dissolution of the Ddx4 condensed phase and expansion of the monomer in the disperse phase through increasing ionic strength or temperature. Ddx4^N1 protein chains depicted as green lines. Transition point (T_p) is indicated with a dashed gray line. The ionic interactions within the droplets are attenuated with increasing salt, as is the residual structure within the protein in the dispersed phase. Corresponding bright-field images are shown on the right. Scale bar, 10 μm.

Figure 5

Post-Translational Modification by Arginine Methylation Alters the Phase Transition of Ddx4^N1 (A) Sequence logo (weblogo.berkeley.edu) depicting the amino acid motifs surrounding arginine residues of Ddx4^N1 predominantly targeted by PRMT1. Arginine residues to be converted to aDMA are highlighted in dark red and with two small ellipses. The amino acid numbers of the modified arginine residues are shown within their respective sequence contexts. Asterisks highlight aDMA sites identified in Ddx4^N1Me with 95% probability (Scaffold score) from a combination of trypsin and GluC digestion of recombinant, purified Ddx4^N1Me. aDMA at sites 146 and 147 was identified at ∼65% probability (Scaffold score). (B) Schematic and mass reconstruction of +TOF MS spectra of Ddx4^N1 (green; 25.833 kDa) and Ddx4^N1Me (dark red). In the latter, a series of peaks was observed between 1 and 20 methyl additions. The major peaks indicate complete aDMA modification at 5 and 6 sites, respectively. (C) A schematic of aDMA together with an insert showing the ¹H-¹³C HSQC NMR spectrum of the ^θCH₃ of Ddx4^N1Me. The chemical shifts of the methyl groups verify that the modification is aDMA (see Figure S5). (D) The phase-transition temperatures of Ddx4^N1Me (dark red) are shifted compared to the unmodified form under the same conditions (light green). Modification with aDMA at a mixture of 5–6 aDMA sites reduces the transition temperature by 25°C, an effect on the phase transition comparable to increasing the ionic strength by 100 mM.

Figure 6

The Sequence Features that Enable Droplet Formation by Ddx4 and Their Distribution within the Human Genome (A) Sliding net charge (10 amino acid window, black) is shown for (i) Ddx4^N1 and (ii) a charge-scrambled mutant, Ddx4^N1CS, obtained by swapping the positions of positive residues (blue bars) and negative residues (red bars) to minimize any persistence of blocks of charge. (iii) A mutant where nine phenylalanine residues, whose placement was highly conserved, were mutated to alanine (Ddx4^N1FtoA, see Figure S6). The positions of the nine phenylalanine residues (yellow circles) mutated to alanine are indicated. (B) Representative fluorescence images from cell imaging experiments reveal that Ddx4^N1CS and Ddx4^N1FtoA do not form organelles in cells under physiological conditions. Residual HeLa nucleoli are still observed as fluorescence-depleted regions within the cell nucleus. (C) The human genome was surveyed for sequences with similar physical properties to the Ddx4 disordered termini. 1,556 sequences out of 14,198 were identified to have [F/R]G spacings in their sequence that are similar to the Ddx4 ortholog family. The top 10% of these are indicated (dotted line). A significant number of proteins associated with forming non-membrane organelles were present in this group. (D) Similar plots from the yeast (i) and E. coli (ii) genomes revealing a number of proteins closely associated with nucleic acid biochemistry.

Figure 7

Proteinaceous Organelles Differentially Solubilize Nucleic Acids (A) Ddx4^N1 organelles were allowed to form under near-physiological conditions at a total concentration of 162.5 μM. (i) Double- and (ii) single-stranded 32-nt DNAs (dsDNA and ssDNA, respectively) tagged with atto647N were added at a concentration of 1 μM. In the case of dsDNA, the majority of the material was excluded from the droplets. The reverse effect was observed for ssDNA. (B) The average and SD (error bar) confocal fluorescence emission intensities from both inside and outside the organelles were used to quantify the partition equilibrium coefficient and its corresponding free energy (Equation 1, Figure S7).