Liquid droplet formation by HP1α suggests a role for phase separation in heterochromatin

Abstract

Gene silencing by heterochromatin is proposed to occur in part as a result of the ability of heterochromatin protein 1 (HP1) proteins to spread across large regions of the genome, compact the underlying chromatin and recruit diverse ligands. Here we identify a new property of the human HP1α protein: the ability to form phase-separated droplets. While unmodified HP1α is soluble, either phosphorylation of its N-terminal extension or DNA binding promotes the formation of phase-separated droplets. Phosphorylation-driven phase separation can be promoted or reversed by specific HP1α ligands. Known components of heterochromatin such as nucleosomes and DNA preferentially partition into the HP1α droplets, but molecules such as the transcription factor TFIIB show no preference. Using a single-molecule DNA curtain assay, we find that both unmodified and phosphorylated HP1α induce rapid compaction of DNA strands into puncta, although with different characteristics. We show by direct protein delivery into mammalian cells that an HP1α mutant incapable of phase separation in vitro forms smaller and fewer nuclear puncta than phosphorylated HP1α. These findings suggest that heterochromatin-mediated gene silencing may occur in part through sequestration of compacted chromatin in phase-separated HP1 droplets, which are dissolved or formed by specific ligands on the basis of nuclear context.

Extended Data Figure 1. Mass-spectrometric analysis of HP1α proteins. Cross-linking Mass Spectrometry of HP1α identifies extensive interactions between the Chromoshadow domain(CSD) and the hinge region

a, Phosphorylation of HP1α occurs almost exclusively at the N-terminus. Left panel: Annotated HCD product ion spectra of a quadruply phosphorylated, doubly charged HP1α peptide at Ser11, Ser12, Ser13, Ser14. Neutral loss of phosphoric acid from b-ions is indicated by b*. Right panel: Relative occupancy of observed HP1α phosphorylation sites as estimated by spectral counting. 41.7% of product ion spectra from peptides containing Serines 11–14 were observed quadruply phosphorylated (393 of 943 spectra). An additional 32.9% (310 spectra), 12.8% (121 spectra), and 8.5% (80 spectra) were identified triply, doubly, and singly phosphorylated, respectively, while only 4.1% (39 spectra) were observed with no phosphorylation. In contrast, phosphorylation was observed at other positions (Ser45, Thr132 and/or Ser135, Thr145, and Thr 188) with 1–2.5% occupancy (1059, 2243, 1586, 1042 total spectra observed for peptides containing these residues). b, Native MS charge state envelopes for WT, Phos- and nPhos- HP1α. c, Table with predicted and observed masses is also shown. The deconvoluted masses fit best to dimeric HP1α modified by 8-phosphates in Phos-HP1α and nPhos-HP1α samples. d, Cross-links were identified by separating cross-linked HP1α by SDS-PAGE and excising bands corresponding to monomeric and dimeric HP1α. Putative inter protein cross-links, diagrammed here, were identified by taking the set of cross-links that are unique to the dimer band (from three replicates). Only cross-links identified by 4 or more product ion spectra are shown for clarity.

Extended Data Figure 2. Phase separation is an isoform specific capability of phosphorylated HP1αthat is perturbed by GFP fusions

a, 1 μL of a solution of 400 μM of each protein was spotted on a plastic coverslip and imaged at 10×, scale bars are 50 μm. Buffer was 75 mM KCl, 20 mM HEPES pH 7.2, 1 mM DTT. Phos-HP1α is phosphorylated in the N-terminus and hinge, nE-HP1α has the N-terminal serines replaced with glutamates, Phos GFP-HP1α is a N-terminal GFP fusion phosphorylated in the N-terminus and hinge, Phos-HP1α-GFP is a C-terminal GFP fusion phosphorylated in the N-terminus and hinge, Phosbpm-HP1α has the ‘KRK’ hinge sequence mutated to alanines and phosphorylated in the N-terminus and hinge, Phos-HP1α-KCK has a C-terminal ‘GSKCK’ tag added and phosphorylated in the N-terminus and hinge. b, Complete comparison of saturation concentration measurements between spin down assay (left) and 340nm turbidity based measurement (right), some data is repeated from figure 1.

Extended Data Figure 3. Estimation of oligomeric potential by sedimentation velocity analytical ultracentrifugation

a, Representative sedimentation velocity runs from high concentration HP1 samples. Percentage of the loaded sample higher than 6S was quantified to estimate oligomeric species higher that a dimer. b, Table showing the comparison of high concentration AUC runs. Average sedimentation coefficient was quantified by integrating from 1–20S and higher order oligomers were estimated by integrating signal from 6–20S. c, Analytical ultracentrifugation c(S) analysis of fully phosphorylated HP1α and the fully phosphorylated basic patch mutant. d, Analytical centrifugation c(S) analysis of fully phosphorylated HP1α and the fully phosphorylated HP1α/β chimera (PhosNH-α/βchimera). Representative traces from three independent experiments are shown in a,b,c and d (n=3).

Extended Data Figure 4. Estimation of HP1α dimerization affinity by isothermal calorimetry and analytical ultracentrifugation

a, Isothermal calorimetry data showing the measured dimerization K_d for the HP1α CSD domain. The calculated K_d is 1.1 μM. b, An analytical ultracentrifugation isotherm used to estimate the dimerization K_d for WT HP1α. Estimated K_d for dimerization using an isodesmic association model is 1.12 μM.

Extended Data Figure 5. Scattering and Guinier fits of SAXS on WT and nPhos HP1α show homogeneous populations

a, Raw X-ray scattering intensity of WT (blue points) and nPhos-HP1 (green points) at 3.5 mg/mL (150μM) concentration. Black lines are Fourier transforms of the fitted interatomic distance distribution, P(r), with chi-square values of 1.186 and 1.199 for WT and nPhos, respectively. b, Guinier plots of WT (blue points) and nPhos-HP1 (green points) at 150μM. Black lines are linear fits to the data plotted as log-intensity versus q². The range of data used in the linear fits extend up to q*Rg of 1.3. The corresponding residuals for each fit are shown below as vertically-shifted horizontal lines for clarity.

Extended Data Figure 6. Phosphorylated HP1α elutes as an extended dimer when examined by SEC-MALS

a, Elution profiles of WT HP1 and nPhos-HP1 examined by SEC-MALS. The horizontal green, and blue lines correspond to the calculated masses for nPhos HP1 and WT HP1 respectively. b, MALS trace of fully phosphorylated HP1α run under identical conditions to those in a.

Extended Data Figure 7. Measuring Shogushin 1, Lamin B Receptor, H3K9me3 peptide affinity, and the effect of Shogushin peptide binding on oligomerization

a, b, Fluorescence anisotropy plots showing the K_d measurements (values in μM next to symbols for WT vs. CSDm) for LBR and Sgo1 peptide binding to WT HP1α and the I163A CSD mutant (CSDm) which can no longer form dimers. c, Comparative AUC runs of ~50 μM nPhos HP1α with and without 100 μM Shogushin peptide or LBR peptide. d, Fluorescence anisotropy plots with a 15mer trimethylated H3K9 peptide showing the relevant HP1 isoforms can bind the nucleosome tail.

Extended Data Figure 8. Effects of additional ligands on saturation concentrations

a, Bar graphs displaying the effects of 100 μM of the polyamine spermine along with the H3K9 and H3K9me3 peptides on phase separation behavior. b, Schematic of the assay used to quantify the partitioning of Cy3 labeled substrates into the two phases. The blue bars represent the total concentration of the labeled species before spin down; the orange bars represent the concentration of Cy3-labeled species remaining in the upper phase after spin down. The lower phase contains HP1α at a higher concentration than in the upper phase. Error bars represent standard error of the mean from three independent measurements. c, Model for phosphorylation or DNA driven HP1α phase separation. Phosphorylation or DNA binding relieves intra HP1 contacts and opens up the dimer. The location(s) of the intra- and inter-dimer contacts that change during this transition are not fully understood, but are predicted to involve interactions between the CTE, hinge, and NTE.

Extended Data Figure 9. Consequences of the interaction between HP1 and DNA

Wide-field TIRF microscopy images of DNA compaction by a, HP1β and b, HP1α bpm at different time points, scale bars are 5μm. Average kymograms for c, HP1β (N = 368) and d, HP1α bpm (N = 318) overlayed with fits for average compaction speed (dashed line) and standard deviation (solid lines). Individual kymograms showing compaction by e, WT HP1α and f, nPhos-HP1α at different protein concentrations.

Extended Data Figure 10. Additional micrographs of NIH3T3 cells transduced with HP1

NIH3T3 cells transduced with .3 μg of HP1 proteins and imaged under identical conditions. a, nPhos, b, CSDm, c, WT HP1α

Figure 1

Phase separation by HP1α. a, Schematics of HP1α mutants. CTE-C-Terminal Extension, CSD-ChromoShadowDomain, H-Hinge, CD-ChromoDomain, NTE-N-Terminal Extension. b, Left panel: Phase separation of nPhos-HP1αat 4°C, 75 mM KCl, 20 mM HEPES pH 7.2. Right panel: Micrograph of phase separated nPhos-HP1α taken at 10×. Scale bar is 50 μm. c, Turbidity assay using a sigmoid function to measure saturation concentration. Dotted vertical lines indicate calculated saturation concentration. d, Saturation concentrations for nPhos-HP1α and nPhos-ΔCTE-HP1α with and without Sgo or LBR peptides. e, Saturation concentrations of different HP1α proteins using spin-down assay (inset). Measurements entailed three independent experiments (n=3), error bars reflect standard error of the mean (s.e.m).

Figure 2

NTE Phosphorylation promotes HP1α oligomerization and conformational change. a, Sedimentation Velocity AUC analysis of 300 μM WT HP1α and nPhos-HP1α. b, P(r) distributions of WT HP1αand nPhos-HP1α obtained SAXS. A 2-fold dilution (green solid vs. black dashed lines) does not significantly change the Dmax for nPhos-HP1α suggesting data reports predominantly on dimeric state (Extended Data Fig. 5). Inset: models describing two possible conformations for the HP1α dimer generated as described in Methods. c, Model for how HP1α switches between a compact and extended state: the N-terminal phosphates interact with basic hinge residues to stabilize inter-dimer contacts in the extended state and promote higher-order oligomerization. Traces from three independent experiments shown in a and b (n=3).

Figure 3

Consequences of interactions with DNA. a, DNA binding causes droplet formation with WT HP1α but not HP1β. Mutating HP1α hinge residues (bpm) or disrupting CSD dimer (CSDm) inhibits DNA-driven phase separation. b, Schematic of DNA curtains. A fluid lipid bilayer in the flow cell allows diffusion of tethered DNA strands to nanofabricated barriers with buffer flow (black arrow indicates direction) c, Cartoon of DNA compaction assay. DNA labeled with intercalating dye YOYO (yellow stars) is compacted (vertical arrow) over time (horizontal arrow) with the addition of HP1α. Wide-field TIRF microscopy images of DNA compaction by d, WT-HP1α and e nPhos-HP1α at different time points, scale bars are 5 μm. Average kymograms for f, WT-HP1α (N = 422) and g nPhos-HP1 α (N = 371) overlayed with fits for average compaction speed (dashed line) and standard deviation (solid lines). h, Individual kymograms showing compaction by WT HP1α (top) and nPhos-HP1α (bottom). i, four panels showing time-course of DNA compaction by WT HP1α in trans, time at bottom in seconds. Asterisks indicate formation of individual puncta. j, Overlayed trajectories of DNA compaction by HP1α (WT, nPhos, and BPM) and HP1β.

Figure 4

Partitioning of specific macromolecules into HP1α phase and behavior of HP1α molecules in cells. a, Micrographs of phase separated nPhos-HP1α droplets with either Cy3 labeled or YOYO labeled macro-molecules visualized using Cy3 fluorescence or YOYO fluorescence, respectively. For each panel a representative micrograph is shown from three independent experiments. Scale bar is 50 μm. b, NIH3T3 cells transduced with Cy3 labeled HP1 proteins and classification of puncta distribution. Right, top plot: average number of distinct puncta per cell. Right, bottom plot: percentage of cells that have at least one large puncta. A large puncta is defined as having a diameter > 5μm in any direction within XY dimension of a Z projection. Scale bar is 10 μm. Error bars represent standard error of the mean. nPhos-HP1α (N=36), WT-HP1 α(N=38), and CSDm-HP1α(N=26). c, Model for the role of regulated phase separation in chromatin organization.