Phase separation is increasingly recognized in facultative heterochromatinization of Polycomb target genes; however, the mechanisms underlying this process remain obscure. Using single-molecule imaging and tracking, we show that individual condensates in mouse embryonic stem cells (mESCs) contain approximately 3 CBX2 molecules and numerous Polycomb repressive complex (PRC)1 and PRC2 subunits and indicate that the composition and dynamics of condensates are developmentally regulated. We reveal that CBX2 clusters PRC2 and controls the spatial distribution of both PRC2 and H3K27me3. Using genomic approaches, we demonstrate that CBX2 binds to condensate initiation sites, which are enriched for PRC2 nucleation sites. CBX2 deletion causes PRC2 and H3K27me3 to redistribute from their regular targets. By developing a separation-of-function variant, we determine that CBX2 relies on its self-clustering ability to function. These findings collectively support a phase-separation model driven by nucleation and bridging, in which low-abundance proteins self-cluster to initiate condensate assembly, a process tightly coupled to function. • 3 CBX2 molecules with many Polycomb proteins per condensate in mESCs • Polycomb condensate assembly and CBX2 dynamics are cell-type specific • CBX2 self-clustering orchestrates PRC2 clustering and H3K27me3 deposition • CBX2 self-clustering is required for cellular differentiation Ingersoll et al. uncover that approximately three CBX2 molecules drive PRC2 clustering and H3K27me3 deposition in mESCs. This process supports a model in which few CBX2 molecules self-cluster on chromatin, initiating condensate assembly while concurrently recruiting and stabilizing PRC2 at nucleation sites for H3K27me3 deposition and cellular differentiation.
Ingersoll et al. (Sun,) studied this question.