Figure 1.
Core components of the major PcG complexes in vertebrate cells overlap with CpG islands.
(A) PRC1 complexes catalyze monoubiquitylation of histone H2A lysine 119 (H2AK119ub1). The catalytic subunits are RING1A/B and one of six PCGF protein homologues, PCGF1–6. PRC1 exists in canonical and noncanonical forms. Canonical PRC1 complexes comprise RING1A/B, PCGF2/4, the CBX protein subunit that has a chromodomain (CD) that binds specifically to PRC2-mediated H3K27me3, and the PH subunit. In noncanonical PRC1, CBX proteins are substituted by the RYBP/YAF2 subunit and the PH subunit is absent. Association of additional subunits with noncanonical PRC1 occurs in a manner dependent on the associated PCGF protein. Thus, PCGF1 complexes, discussed extensively herein, include the additional subunits BCOR and KDM2B. KDM2B has a zinc finger CxxC domain (CXXC) that mediates binding to unmethylated CpG dinucleotides. PRC2 complexes methylate histone H3 lysine 27 (H3K27me1/2/3) and comprise the catalytic EZH2 protein and the core subunits EED, SUZ12, and RBAP48. JARID2 is a substoichiometric subunit that has been implicated in PRC2 targeting. (B) PcG complexes occupy CpG islands at target gene promoters: an example from genome-wide ChIP-seq analysis of mouse embryonic stem cells illustrating a PcG target gene, Lhx4. The Bio-CAP procedure provides a molecular readout of the density of unmethylated CpG dinucleotides with peaks corresponding to CpG islands. ChIP-seq for PRC1 (RING1B) and PRC2 (EZH2) subunits illustrates that PcG protein occupancy closely mirrors unmethylated CpG density.
Figure 2.
Comparison of instructive and responsive models for recruitment of PcG complexes to CpG islands at target gene promoters.
DNA methylation and hypomethylation are indicated with filled and open lollipops, respectively. Repressed PcG–associated CpG islands are colored red and active TrxG–associated CpG islands green. (A) Classical instructive models invoke that sequence-specific DNA binding transcription factors (TF) and/or long noncoding RNAs (lncRNA) interact biochemically with PcG complexes and thereby target these complexes to defined promoters at which PcG-mediated histone modifications inhibit RNA polymerase II (RNAPII) activity. At active gene promoters (large arrow), TFs directly recruit TrxG proteins that in turn deposit histone modifications linked to gene activation. (B) In the responsive model, it is proposed that both PcG (PRC1-KDM2B and PRC2) and TrxG (CFP-SET1 and MLL) complexes stochastically sample unmethylated CpG island chromatin irrespective of the transcriptional state of a given gene (i). This occurs either via CxxC zinc finger protein binding to unmethylated CpG or, in the case of PRC2, via sensing the absence of otherwise pervasive histone modifications, like H3K36me2. The outcome of PcG and TrxG sampling would then be responsive to the transcriptional state of the associated gene (ii). In the absence of transcription, the PcG protein–occupied chromatin state would accumulate by default (aided by positive feedback loops) and antagonize TrxG activity. Conversely, at transcribed genes, TFs, RNAPolII, and transcription would favor accumulation of TrxG factors with their associated activities, in turn antagonizing the function of PcG complexes. Importantly, in the responsive model, the capacity of PcG proteins to sample CpG island chromatin would permit them to respond to the transcriptional state of potential target genes without a requirement for direct interactions with TFs or ncRNAs.
Figure 3.
A simplified representation of the responsive model emphasizing that it has properties of a classical bistable switch.
A schematic of the proposed bistable switch (center panel). The ground state at CpG island chromatin, PcG occupancy (left panel), is the default that is overcome only when TF-mediated gene activation reaches a critical threshold. Positive feedback mechanisms between PRC1 and PRC2 reinforce PcG occupancy and antagonize the TrxG state. Beyond the activation threshold, the system switches to the TrxG state (right panel) that is in turn reinforced by positive feedback mechanisms between RNAPII and TrxG factors. This, together with TrxG-mediated antagonism of PcG activities, ensures stability of the TrxG state such that stochastic fluctuations in gene activation signal would not trigger a switch back to the PcG state. However, when gene activation signals drop below a critical threshold, the CpG island will switch back to the default PcG state.