Genome-wide activities of Polycomb complexes control pervasive transcription

Abstract

Polycomb group (PcG) complexes PRC1 and PRC2 are well known for silencing specific developmental genes. PRC2 is a methyltransferase targeting histone H3K27 and producing H3K27me3, essential for stable silencing. Less well known but quantitatively much more important is the genome-wide role of PRC2 that dimethylates ∼70% of total H3K27. We show that H3K27me2 occurs in inverse proportion to transcriptional activity in most non-PcG target genes and intergenic regions and is governed by opposing roaming activities of PRC2 and complexes containing the H3K27 demethylase UTX. Surprisingly, loss of H3K27me2 results in global transcriptional derepression proportionally greatest in silent or weakly transcribed intergenic and genic regions and accompanied by an increase of H3K27ac and H3K4me1. H3K27me2 therefore sets a threshold that prevents random, unscheduled transcription all over the genome and even limits the activity of highly transcribed genes. PRC1-type complexes also have global roles. Unexpectedly, we find a pervasive distribution of histone H2A ubiquitylated at lysine 118 (H2AK118ub) outside of canonical PcG target regions, dependent on the RING/Sce subunit of PRC1-type complexes. We show, however, that H2AK118ub does not mediate the global PRC2 activity or the global repression and is predominantly produced by a new complex involving L(3)73Ah, a homolog of mammalian PCGF3.

Figure 1.

Widespread distribution of H3K27 methylation. (A) Representative genome region showing the distribution of mono- (blue), di- (gray), and trimethylated H3K27 (green). Pc (red), and Pol II (purple) mapped by ChIP-chip in Bg3 cells (the modENCODE Project) or Sg4 cells (this study; Supplemental Fig. 1A). The H3K27me2 track shows the enrichment threshold based on the signal distribution (see Supplemental Fig. 1B–C). (B,C) Meta-gene profiles of H3K27me2 and H3K27me1 for PcG target genes and five subgroups of non-PcG target genes with different expression levels (e.g., NP-Q1 indicates the top 20 percentile active non-PcG target genes) in Bg3 cells (RNA tiling array data from the modENCODE Project). The x-axis represents the positions relative to the transcription start site (TSS; 0.0) and transcription end site (TES; 1.0). ChIP-chip signal values are scaled to the distribution with a mean of zero and a standard deviation of one. The same results normalized to the H3 distribution are shown in Supplemental Figure 1. (D–F) Meta-gene profiles of UTX, Pol II, and H3K27me2 for three subgroups of genes with different UTX binding levels in Sg4 cells.

Figure 2.

Global changes of H3K27me2/3 levels after E(z) inactivation. (A) Growth curves of EZ2-2 and Ras3 (w.t.) cells after shift to 31°C. (B) Global H3K27me2/3 levels in EZ2-2 cells at 25°C and 31°C assayed by Western blot. Threefold serial dilutions of nuclear extract were loaded. Total H3 is the loading control. (C,D) H3K27me3 and H3K27me2 levels at PcG target and non-PcG target regions from EZ2-2 cells at 25°C and 31°C. H3K27me2/3 levels were measured by ChIP-qPCR from at least two independent experiments. The relative enrichment values were calculated based on the qPCR values normalized to input DNA. Error bars, SE. The four intergenic regions are identified by the distance in kb from the TSS (e.g., u12k). (E,F) The levels of Pc and E(z) at PcG target regions and two non-PcG target regions assayed by ChIP-qPCR from EZ2-2 cells grown at 25°C and 31°C. (G) H3K27me2 levels in highly expressed (HighExp 10%), intermediate (MidExp 10%), weakly expressed (LowExp 10%), and silent (no RNA-seq reads) non-PcG target genes and PcG target genes were analyzed from EZ2-2 H3K27me2 ChIP-seq data at 25°C and 31°C. (H) H3K27me2 levels in transcriptionally active intergenic non-PcG target (Top 10%) and silent regions (no RNA-seq reads) were analyzed from EZ2-2 H3K27me2 ChIP-seq data at 25°C and 31°C.

Figure 3.

Global transcriptional increase after E(z) inactivation. (A,B) Expression levels of PcG target genes and silent non-PcG target genes in EZ2-2 cells at 25°C and 31°C were assayed by RT-qPCR and shown relative to RpL32. Since the expression of RpL32 increases approximately twofold at 31°C (Supplemental Fig. 5A), this normalization is conservative. (C) Expression levels of a panel of silent intergenic non-PcG target regions in EZ2-2 cells at 25°C and 31°C. The Ras3 panel shows expression levels in a control wild-type cell line (Ras3) at the same temperatures. (D) Global derepression in coding regions after E(z) inactivation. The log ratio of the expression levels at 31°C and 25°C was calculated for non-PcG targets with different expression levels and for PcG targets. For silent genes, one expression tag read was added to the number of RNA-seq tags to estimate the approximate fold-differences. The bottom and top of each box represent the first and third quartiles, and the mark inside the box is the median value. Whiskers are extended to 1.5× interquartile range. (E) Preferential transcriptional increase at H3K27me2-enriched genes. Expression fold-change between 31°C and 25°C was calculated for the non-PcG target genes with the lowest and the highest levels of H3K27me2.

Figure 4.

Attenuation of intergenic derepression by depletion of Trithorax group proteins. Expression levels in a panel of intergenic regions were measured by RT-qPCR analysis with total RNA from EZ2-2 cells at 25°C and 31°C incubated with dsRNA targeting LacZ (mock treatment), Utx (A), trr (C), or CBP (E) for 12 d. The expression levels are normalized by the DNA amount in input cells. (B,D,F) The efficiency of the knockdown experiments was monitored by the mRNA levels of the targets.

Figure 5.

H3K27ac and H3K4me1 levels increase upon E(z) inactivation. (A,B) H3K27ac and H3K4me1 in intergenic regions were assayed by ChIP-qPCR in EZ2-2 cells at 25°C and 31°C. (C) Box plots of H3K27ac levels in non-PcG targets with different expression levels and PcG targets at 25°C and 31°C. (D,E) The increase in H3K27ac and H3K4me1 levels after E(z) inactivation is negatively correlated with those levels at 25°C. The ratio of levels at 31°C and 25°C was calculated for non-PcG target genes with different levels at 25°C.

Figure 6.

PRC1 involvement in intergenic repression. (A) Western blot analysis shows the loss of total H2AK118ub after 12 d of RING/Sce RNAi on S2 cells. LacZ dsRNA was used as control; twofold serial dilutions of nuclear lysate were loaded; and total histones stained with Coomassie blue served as loading controls. (B) H2AK118ub levels in PcG and intergenic non-PcG target regions were assayed by ChIP-qPCR after treatment of S2 cells with dsRNA for LacZ (control) or RING/Sce. H2AK118ub levels in three PcG target regions, one active gene region, and five intergenic non-PcG target regions were measured by ChIP-qPCR. (C) The H2Aub1 levels of intergenic non-PcG (transcriptionally active or silent) and PcG target regions in EZ2-2 cells were determined from the 25°C and 31°C ChIP-seq data. (D) H3K27me2 levels after RING/Sce knockdown in S2 cells were assayed by ChIP-qPCR compared with control treatment with LacZ dsRNA. (E) Decrease of global H2AK118ub level after RNAi depletion of l(3)73Ah in S2 cells. (F) Expression levels in intergenic and genic non-PcG target regions were measured by RT-qPCR with total RNA from mock-treated and Pc, E(z), or L(3)73Ah-depleted S2 cells. Changes in mRNA levels of l(3)73Ah, Pc and E(z) after knockdown are shown in Supplemental Figures 7G, 7B, and 3B, respectively. (G) Distribution of H2AK118ub (blue) in the genomic region surrounding the bithorax complex as determined by ChIP-seq. For comparison, the distributions of Pc (red), H3K27me3 (green), and H3K27me2 (gray) are also shown, indicating the absence of H2AK118ub at the entire BX-C (from Ubx to Abd-B) but presence in other regions both with and without PC or H3K27me2/3.