The spatial segregation of heterochromatin into distinct, membrane-less nuclear compartments involves the binding of the heterochromatin protein 1 (HP1) to H3K9me2/3-rich genomic regions. While HP1 exhibits liquid-liquid phase separation properties in vitro, its mechanistic role in vivo on the structure and dynamics of heterochromatin remains largely unresolved. Here, using biophysical modeling, we systematically investigate the mutual coupling between self-interacting HP1-like molecules and the chromatin polymer. We reveal that the specific affinity of HP1 for H3K9me2/3 loci facilitates coacervation in nucleo, and promotes the formation of stable heterochromatin condensates at HP1 levels far below the concentration required in vitro to observe phase separation in purified protein assays. These heterotypic HP1-chromatin interactions give rise to a strong dependence of the nucleoplasmic HP1 density on the HP1-H3K9me2/3 stoichiometry, consistent with the thermodynamics of multicomponent phase separation. The dynamical crosstalk between HP1 and the visco-elastic chromatin scaffold also leads to anomalously-slow equilibration kinetics, which strongly depend on the genomic distribution of H3K9me2/3 domains and result in the coexistence of multiple long-lived, microphase-separated heterochromatin compartments. The morphology of these complex coacervates is further found to be governed by the dynamic establishment of the underlying H3K9me2/3 landscape, which may drive their increasingly abnormal, aspherical shapes during cell development. These findings compare favorably to 4D microscopy measurements of HP1 condensates that we perform in live Drosophila embryos, and suggest a general quantitative model of heterochromatin formation based on the interplay between HP1-based phase separation and chromatin mechanics.

Flies homozygous for the expression of ectopic GFP-HP1a on the second chromosome (+, GFP-HP1a/CyO; +) were allowed to lay embryos on apple juice plates supplemented with yeast paste at 25°C. Embryos were collected by hand and dechorionated in 50% bleach before being mounted for live imaging. Embryos were imaged using a Zeiss 880 Airyscan microscope with a 63x/1.4 oil immersion objective. Time-lapse image stacks were collected every 30 seconds with a Z-spacing of 0.36μm. Imaging at least part of nuclear cycle 13 (NC 13) allowed the start of NC14 (time=0 sec) to be clearly defined as the first time point where circular nuclei appear following mitosis. All movies were Airyscan-processed using Zen 2.3 software and the first 72 time points of NC 14 (max36 minutes) were analyzed using Arivis software. Briefly, HP1a foci were segmented from the overall nuclear signal using an intensity threshold > 75% of the fluorescence intensity and a lower size threshold of 0.03 μm³ corresponding to approximately 10 voxels (0.085 μm x 0.085 μm x 0.36 μm). All HP1a segments found to be percolating in 3-dimensional space were merged into a single segment, and population statistics were calculated for each time point.

The imaging data made accessible here are the 4-D zeiss confocal modies converted to .tif stacks in FIJI but unanalyzed and unaltered in any other way.  

These files can be opened in ImageJ/ FIJI and Arivis. Please check the file names for each file to make sure that the time and z axis of the data have been properly read by whichever software you would like to use. The time and z index is appended to each file name, so a cursory read of these will orient the user.