A biophysical basis for the spreading behavior and limited diffusion of Xist

Abstract

Xist RNA initiates X inactivation as it spreads in cis across the chromosome. Here, we reveal a biophysical basis for its cis-limited diffusion. Xist RNA and HNRNPK together drive a liquid-liquid phase separation (LLPS) that encapsulates the chromosome. HNRNPK droplets pull on Xist and internalize the RNA. Once internalized, Xist induces a further phase transition and "softens" the HNRNPK droplet. Xist alters the condensate's deformability, adhesiveness, and wetting properties in vitro. Other Xist-interacting proteins are internalized and entrapped within the droplet, resulting in a concentration of Xist and protein partners within the condensate. We attribute LLPS to HNRNPK's RGG and Xist's repeat B (RepB) motifs. Mutating these motifs causes Xist diffusion, disrupts polycomb recruitment, and precludes the required mixing of chromosomal compartments for Xist's migration. Thus, we hypothesize that phase transitions in HNRNPK condensates allow Xist to locally concentrate silencing factors and to spread through internal channels of the HNRNPK-encapsulated chromosome.

Keywords: HNRNPK; RGG domain; RepB; Xist RNA; cis-limited spreading; condensates; deformability; droplets; liquid-liquid phase separation; repeat B.

Figure 1.. HNRNPK undergoes LLPS.

(A) Schematic representation of HNRPNK variants. (B) Droplets formed by HNRNPK-GFP at 1.25–10 μM. (C) Inability to form droplets by 5 μM HNRNPK-GFP mutants. (D) Photobleaching (arrow) and recovery of HNRNPK-GFP droplets (10 μM) over time in vitro. Half-life, t^1/2. N= 23 droplets. Three independent experiments. (E) Photobleaching (arrow) and recovery of HNRNPK-GFP droplets over time in U2OS cells. (F) Quantification of panel E. Data collected from 33 droplets in three independent experiments. (G) HNRNPK-GFP (10 μM) droplets fusing over time. Zoom-ins show fusing droplets and circularity. (H) HNRNPK-GFP droplet formation at indicated KCl concentrations and pHs. (I) Quantification of droplet sizes shown in H. **, P<0.01; ****, P<0.0001; ns, not significant (unpaired Student t-test). Error bars, ±SEM. (J) Effect of 5% 1,6-HD and 2,5-HD treatment on HNRNPK-GFP droplets in vitro (upper, 5 μM) and in U2OS cells (lower) for 5 mins. (K) Quantification of droplet sizes in (J). ***, P<0.001; ns, not significant (unpaired Student t-test). Error bars, ±SEM. (L) RNA immunoFISH of Xist RNA (red) and native HNRNPK (green), +/−10-minute detergent treatment in MEF. N>100 cells in three independent experiments (M) Average Xist and HNRNPK intensities from 110 overlapping Xist-HNRNPK domains from three different experiments.

Figure 2.. The 6A mutation in HNRNPK’s RGG domain abolishes LLPS.

(A) Intrinsically disordered region prediction of HNRNPK through IUPred3. (B) Residue substitutions in 6A mutant HNRNPK-GFP cause LLPS loss at 10 μM. (C) Absence of droplet formation in U2OS cells with transient expression of 6A HNRNPK-GFP. (D) 2D surface plots of boxed areas in (C). (E) Kurtosis analysis of (D). ****, P<0.0001 (unpaired Student t-test). Error bars, ±SEM. N=128 WT; N=127 6A in three replicates. (F) Schematics of live imaging of droplets at various times (T). (G) Real time images of 10 μM WT and 6A HNRNPK-GFP droplets at indicated time points. (H) Droplets area measured at indicated time points. Error bars, ±SEM. (I) RNA EMSA: 2-fold serial dilutions (0–1.2 μM) of untagged WT HNRNPK shifted with 20 fmol RepB-WT or RepB-MU. (J) RNA EMSA: 2-fold serial dilutions (0–1.2 μM) of WT or Mutant HNRNPK (ΔHK3-RGG or 6A) shifted with 20 fmol RepB-WT. Left, ΔKH3-RGG mutant HNRNPK. Right, 6A mutant HNRNPK. (K) Binding curves for data in (I) and (J), with indicated $K_d$.

Figure 3.. Repeat B is internalized by HNRNPK droplets and enhances LLPS.

(A, B, C) Co-phase separation of RepB-WT or RepB-MU (0–100 nM; Cy5-labelled) with HNRNPK-GFP (0.25–2 μM). Top panels: GFP+Cy5 merged. Bottom panels: Cy5 channel only. Quantitation of HNRNPK with RepB-WT (B) or RepB-MU (C) signals in (A). Data collected from 35 acquisition areas (300 × 300 pixels). Error bars, ±SEM. ****, P<0.0001; ***, P<0.001; **, P<0.01; ns, no significant difference (unpaired Student t-test) from three replicates. (D) RepB-WT vs -MU (20 nM) phase separation with HNRNPK-GFP (5 μM). (E) Line plot quantification of fluorescence intensity for (D). (F) The photobleaching and recovery kinetics of HNRNPK-GFP (5 μM) with RepB-WT vs -MU (20 nM). (G) Quantification of (F). N=20 (HNRNPK + Rep-MU), 18 (HNRNPK + Rep-WT), and 19 (HNRNPK) droplets in three replicates. Error bars, ±SEM.

Figure 4.. HNRNPK-RepB condensates internalize, entrap, and concentrate XCI factors.

(A) Full-length or truncated RFP-tagged SMCHD1, SPEN, YY1, and Alexa-561-labelled RING1 B. SM-CT, SMCHD1 C-terminal domain; SP-NT, SPEN N-terminal domain. (B) At 10 μM, the four proteins could not form droplets by themselves in three independent experiments. (C) RFP-tag alone is not internalized by HNRNPK-RepB droplets. HNRNPK, 10 μM. RFP tag, 10 μM. RepB-WT, 20 nM. Left, images of representative droplets. Right, quantification of fluorescence (across line). Three independent experiments. (D) Internalization of XCI factors (5 μM) by droplets formed by HNRNPK alone (5 μM) or by HNRNPK-RepB-WT (HNRNPK, 5 μM; RepB-WT, 20 nM). Three independent experiments. (E) Quantification of fluorescence across lines shown in (D). (F) FRAP: HNRNPK-GFP recovery was tracked following photobleaching +/− indicated XCI factors and +/− RepB. Data from 12 × 10 mins bleach-track objects for groups 1–11 in three independent experiments. **, P<0.01; ****, P<0.0001; ns, not significant (unpaired Student t-test). Error bars, ±SEM. (G) FRAP recovery kinetics of (F). Data from 12 droplets in three independent experiments. Error bars, ±SEM. ****, P<0.0001 (unpaired Student t-test). (H) FRAP: RFP-tagged XCI factors were tracked following photobleaching in HNRNPK droplets +/−RepB. Data from 12 × 10 mins bleach-track objects for groups 1–8 in three independent experiments. ****, P<0.0001(unpaired Student t-test). Error bars, ±SEM. (I) FRAP recovery kinetics of (H). Data from 12 droplets in three independent experiments. ****, P<0.0001 (unpaired Student t-test). Error bars, ±SEM. (J) FRAP recovery kinetics for Cy5-labelled RepB-WT following photobleaching in HNRNPK droplets +/− indicated XCI factors. Data from 12 droplets in three independent experiments. Error bars, ±SEM. (K) Model: RepB-HNRNPK condensates internalize, entrap, and concentrate XCI factors.

Figure 5.. RepB increases the deformability, adhesiveness, and wetting of HNRNPK condensates.

(A) BFP assay measures deformability of HNRNPK-GFP droplets +/−RepB. Four steps of BFP assay shown (see Results for description). (B) Young’s modulus of indicated HNRNPK-GFP condensates. ****, P<0.0001 (unpaired Student t-test). N=120 (HNRNPK-GFP), 158 (HNRNPK-GFP+RepB-MU), 194 (HNRNPK-GFP+RepB-WT) droplets in three independent experiments. Error bars, ±SEM. (C) Measurement of rupture force and adhesiveness. Top panels: Idealized BFP dynamics +/− RNA internalization. Steps correspond to (A). Bottom panels: Measurement of rupture force (right) and adhesion frequency (left) of HNRNPK-GFP droplets +/− RepB-WT or RepB-MU. ****, P<0.0001 (unpaired Student t-test). Adhesion frequency: N=156 (group 1), 141 (group 2), 180 (group 3) events in three independent repeats. Rupture force measurement: N=291 (HNRNPK-GFP+RepB-WT), 205 (HNRNPK-GFP+RepB-MU) events in three independent repeats. Error bars, ±SEM. (D) Flow chamber assay using microfluidics. HNRNPK-GFP droplets are introduced into chamber under flow (at 744 dyn/cm²) onto a glass surface +/− RepB coating. (E) Representative images of HNRNPK-GFP condensates before and after 5 min at 744 dyn/cm² flow over glass without RNA (1), with RepB-MU (2), or RepB-WT (3). Three replicates. (F) Droplet-surface contact ratio after versus before flow for groups 1, 2, and 3 above. ****, P<0.0001 (unpaired Student t-test). Error bars, ±SEM. N=124 (group 1), 134 (group 2), 130 (group 3) droplets in three independent repeats. (G) Cartoons for HNRNPK-GFP droplets under 6 conditions. (H) Contact surface area of condensates for conditions 1–6 in (G). ****, P<0.0001; ns, not significant (unpaired Student t-test). Error bars, ±SEM. N=137 (group 1), 119 (group 2), 115 (group 3) 285 (group 4), 131 (group 5), 99 (group 6) droplets in three independent repeats. (I) Feret’s diameter of captured droplets under 6 conditions shown in (G). ****, P<0.0001; ns, not significant (unpaired Student t-test). Error bars, ±SEM. N=102 (group 1), 109 (group 2), 108 (group 3) 255 (group 4), 110 (group 5), 95 (group 6) droplets in three independent repeats. (J) Summary: Enhanced wetting behavior of HNRNPK-GFP condensates in the presence of RepB.

Figure 6.. Loss of LLPS of HNRNPK disturbs Xist spreading and Polycomb recruitment on Xi.

(A, B) Confocal analysis with image deconvolution and 2D projections of 3D stacks of indicated cell lines. Staining for Xist RNA, HNRNPK or GFP, and H2AK119ub1 (A) or H3K27me3 (B) are colorized as indicated. Right panels, magnifications of a single Xi. (C) Allelic ChIP-seq for H2AK119ub1 in MEF samples indicated. Composite (comp), Xi and Xa tracks are shown. (D) Boxplots displaying H2AK119ub1 coverage genes subject to XCI in MEFs. **** P<0.0001, ***, P<0.001; **, P<0.01; *, P<0.05; ns, no significant difference (Wilcoxon rank-sum test). (E) Left panels: Confocal images of RNA immunoFISH of Xist and HNRNPK in WT and 6A mutant. Representative images exhibiting overlap between immunofluorescence (IF) of HNRNPK-WT or 6A and Xist with RNA FISH in fixed mouse ES cells. Right panels: Average RNA FISH signal and average HNRNPK immunofluorescence signal centered on RNA FISH foci. N= 97 (HNRNPK-WT),105 (HNRNPK-6A). At least 3 biological replicates were performed. (F) Quantification of Xist dispersal in WT and 6A mutant. ****, P<0.0001 (Mann-Whitney test). N>100 cells per condition. (G) Confocal images of RNA immunoFISH of Xist, H2AK119ub1 and H3K27me3 in WT and 6A mutant. (H) Quantification of cells with normal, weak, or no signals for panel (G). n>100 cells per condition. (I) Xist CHART-seq in WT and 6A mutant ESCs. Gray track, difference (Δ) in Xist coverage (6A minus WT). (J) Boxplots displaying Xist coverage in genes subject to XCI, non-expressed genes, and escapees in WT and 6A mutant ESCs. ****, P<0.0001; ns, not significant (Wilcoxon rank-sum test).

Figure 7.. LLPS drives gene silencing and long-range “mixing” of Xi chromatin.

(A) Scatterplots of allelic gene expression for mus (Xi) and cas (Xa) in 6A vs WT ESCs at differentiation day 14. P values, Wilcoxon rank-sum test. (B) CDF plots of allelic expression of mus (Xi) and cas (Xa) in 6A vs WT ESCs at differentiation days 0 and 14. P values, Wilcoxon rank-sum test. (C) Hi-C contact heatmaps of Xi at 250-kb resolution for indicated MEF cell lines. Two biological replicates showed similar results. Combined results are shown. Scale shown under each. Corresponding Pearson correlation heatmaps shown in Fig. S7D. (D) Log2-fold differences in contact frequencies for the indicated pairwise comparisons. (E) Inter-megadomain crossover frequencies. Bar plots display fraction of long-range interactions (>10 Mb) across megadomain boundary in each cell line. Two replicates were analyzed. Error bars, one standard deviation. *, P<0.05; **, P<0.01; ***, P<0.001 (one-tailed Student t-test). (F) Domain strength scores for each cell line. *, P<0.05; **, P<0.01; ***, P<0.001 (Wilcoxon rank-sum test). (G) 4DHiC modeling of chromosomes in 3D. Red, A-compartment; Blue, B-compartment (defined by PC1 or PCA). Note partitioning of A/B chromatin on opposite Xi hemispheres for Chr13 (chromosome 13) and Xa. When 3D Xi structure is mapped onto the PC1 of the Xa, Xi chromatin appears “mixed” due to reorganization into megadomains. (H) 3D Megadomains structure of WT and ΔRepB MEFs. Megadomains in the WT Xi become visible when the Xi structure is mapped onto its own PC1. Mixing is reduced in ΔRepB MEFs. Right graph, coefficient of megadomain mixing for each case. ****, P<0.0001 (two sample t-test). (I) 3D Xi structures mapped onto their own PC1 of K mutant MEFs. Note loss or reduced mixing in the blue megadomain in shK and shK+6A cells. Right graph, degree of mixing for each case. ***, P<0.001; **, P<0.01; *, P<0.05; ns, no significant difference (two sample t-test). (J) Model: HNRNPK-RepB LLPS regulates cis-limited spreading of Xist and protein partners. ChrX, X-chromosome.