The quest for mammalian Polycomb response elements: are we there yet?

Abstract

A long-standing mystery in the field of Polycomb and Trithorax regulation is how these proteins, which are highly conserved between flies and mammals, can regulate several hundred equally highly conserved target genes, but recognise these targets via cis-regulatory elements that appear to show no conservation in their DNA sequence. These elements, termed Polycomb/Trithorax response elements (PRE/TREs or PREs), are relatively well characterised in flies, but their mammalian counterparts have proved to be extremely difficult to identify. Recent progress in this endeavour has generated a wealth of data and raised several intriguing questions. Here, we ask why and to what extent mammalian PREs are so different to those of the fly. We review recent advances, evaluate current models and identify open questions in the quest for mammalian PREs.

Keywords: Drosophila; Epigenetics; Mammal; Polycomb; Polycomb response element; Trithorax.

Fig. 1

Composition of PcG Complexes in flies and vertebrates. The Polycomb repressive complex 2 (PRC2) and Polycomb repressive complex 1 (PRC1) family of complexes are shown. Core subunits are shown in green for PRC2 and blue for PRC1. Alternate subunits, derived from multiple genes and if more than two, are shown in grey. Accessory proteins are shown in orange. Top: mouse complexes; bottom: Drosophila complexes. Selected histone modifications are shown: Red hexagons: histone H3 tail trimethylated at lysine 27 (H3K27me3); yellow ovals: histone H2A monoubiquitinated at lysine 119 (vertebrates) or 118 (fly). (H2AK119/118 Ub). a PRC2 consists of four core subunits, EZH1/2 (fly E(Z)), EED (fly ESC), SUZ12 (fly SU(Z)12), and RbAp46/48 (or RBBP7/4; fly NURF55) (Cao et al. ; Czermin et al. ; Kuzmichev et al. ; Muller et al. 2002), and three accessory proteins, PCL (Walker et al. 2010), JARID2 (Herz et al. ; Kalb et al. ; Landeira et al. ; Li et al. ; Pasini et al. ; Peng et al. ; Shen et al. 2009), and AEBP2 (Cao and Zhang ; Kalb et al. 2014). Alternate translation start site usage results in four different EED isoforms (not shown in the figure), which have different preferred histone substrates (Kuzmichev et al. 2004). PRC2 dimethylates and trimethylates histone H3 at Lys27 (H3K27me3) through the SET domain of EZH1/2 (fly E(Z)) (Cao et al. ; Czermin et al. ; Kuzmichev et al. ; Muller et al. 2002). In addition, PRC2 can bind H3K27me3 via EED (Hansen et al. ; Margueron et al. 2009). b Canonical PRC1 consists of four core subunits, RING1A/B (fly dRING), CBX (fly PC), PCGF (fly PSC or SU(Z)2), and PHC (fly PH) (Gil and O’Loghlen ; Simon and Kingston 2009). PRC1 catalyses H2AK119Ub1 (in flies H2AK118Ub1) through its RING1A/B (fly dRING) subunit (Cao et al. ; de Napoles et al. ; Scheuermann et al. ; Wang et al. 2004a). Canonical PRC1 can bind H3K27me3 via the chromodomain of CBX2 or 7 (fly PC) (Bernstein et al. ; Fischle et al. ; Min et al. 2003); however, different CBX proteins have different preferences for modified histone tails (Bernstein et al. 2006b), see main text and Fig. 2 for details. c Top: one class of vertebrate non-canonical PRC1s consists of three core subunits, RING1A/B, PCGF, and RYBP or YAF2 and various accessory proteins. The complexes are distinguished by different PCGF subunits. The complex containing PCGF1 (PRC1.1) also contains the histone H3K36 demethylase KDM2B. Other PCGF subunits copurify with other accessory proteins (orange) (Gao et al. 2012). Bottom: Drosophila dRAF is the most similar to vertebrate PRC1.1 and consists of dRING, PSC, and the histone H3K36 demethylase dKDM2 (Lagarou et al. 2008). Further non-canonical PRC1s exist and are reviewed in Gil and O’Loghlen (2014) and Simon and Kingston (2013). See main text and Table 1 for detail on molecular properties

Fig. 2

Evidence for different molecular mechanisms mediated by Drosophila and vertebrate PcG proteins. a Different proteins of the PRC1 complex mediate chromatin compaction in Drosophila and mouse (Grau et al. 2011). Purified PRC1 (see Fig. 1) from both fly and mouse can compact nucleosomal arrays in vitro; however, a different protein mediates this activity in the two species. Fly and mouse homologs of the proteins involved are shown. Red regions show domains required for compaction in each case, which are overrepresented in basic amino acids. Other domains and degree of conservation between mouse and Drosophila are indicated. b Alignment of the chromodomains of Drosophila Polycomb (PC, amino acids 15–77) and five mouse homologs (CBX, amino acids 1–62) redrawn from Bernstein et al. (2006b) and coloured according to the ClustalX colour scheme http://www.jalview.org/help/html/colourSchemes/clustal.html. On the right of the alignment, in vitro binding preferences of the different chromodomains from Bernstein et al. (2006b) are shown. Histone binding was addressed using modified peptides, Kds ranged between 12 and 49 μM. RNA binding was non-sequence specific. RNA-binding activity of the Drosophila PC chromodomain has not been reported to our knowledge

Fig. 3

Similarities and differences in Drosophila and vertebrate Hox gene regulation. a The Drosophila Antennapedia (ANT-C) and Bithorax (BX-C) complexes and the mouse HoxD complex are drawn approximately to scale, based on Duboule (2007) and Maeda and Karch (2009). Dark bars indicate exons (introns not shown for HoxD due to scaling); light bars and vertical arrowheads in ANT-C and BX-C indicate experimentally verified PREs (Ringrose and Paro and references therein). Genes and regulatory regions with a common colour are most closely related in sequence between fly and mouse, and thus belong to the same paralogy group (Duboule 2007). Note that the colour coding is not intended to reflect the different regulatory regions of ANT-C and BX-C as in Maeda and Karch (2009). b Pattern of histone H3 lysine 27 methylation at mouse HoxD (left) and fly BX-C (right) in specific tissues over developmental time. Left: summary of data from Soshnikova and Duboule (2009b). In embryonic stem cells, H3K27me3 covers the entire HoxD locus (top). In tail buds of E8.5 embryos (middle) and E9.5 embryos (bottom), Hox genes are sequentially activated leading to clearing of H3K27me3 from the locus. Right: summary of data from Bowman et al. (2014) and Maeda and Karch (2009). In early (0–2 h) embryos (top), the BX-C very probably lacks H3K27me3 and PcG proteins, based on indirect evidence ((Orlando et al. ; Petruk et al. 2012); see main text for details). In parasegment 7 of stage 5 (2–3 h) embryos (middle), appropriate Hox genes are activated and repressed by the gap and pair rule gene products (Maeda and Karch 2009). In the same parasegment of later (post 5 h) embryos, repressed domains gain H3K27me3 (Bowman et al. 2014)

Fig. 4

Recruitment of mammalian PcG complexes. a Relationship between occurrence of gene promoters, CpG islands, KDM2B and RING1B, according to Deaton and Bird (2011), Farcas et al. (2012), He et al. (2013), and Wu et al. (2013). b The RING1B subunit is a component of multiple different complexes, including both canonical and non-canonical PRC1 (Gao et al. 2012) see main text for details. c Factors influencing PcG recruitment. A stretch of GC- and CpG-rich DNA is shown (yellow). Various motifs for sequence-specific DNA-binding proteins can exist within this DNA (dark yellow), and several of these are themselves GC-rich (see Table 2). All of these motifs may also exist in otherwise GC-poor DNA. Proteins that can bind directly to DNA and have been shown or suggested to have role in PcG recruitment are shown in orange. PRC1: indicates all versions of PRC1 except the special case of PRC1.1 which is recruited by KDM2B. Arrows indicate that the DNA-binding protein in question does not copurify with the complexes but has been shown to interact by Co-IP. TA: activating transcription factor. See main text for details

Fig. 5

Motif occurrences in mammalian PREs. a A selection of mammalian PREs that have been verified to recruit PcG proteins in transgenic assays are shown (see Table 3 for details): HoxC11-12, HoxB4-5 (Woo et al. 2013), HoxD11-12 (Woo et al. 2010), PREkr (Sing et al. 2009), HoxD10 (Schorderet et al. 2013), DBE (Cabianca et al. 2012). Above each element, the % GC is shown, with CpG islands marked in dark grey, according to the following criteria: window size 100; minimum length of an island 200; minimum observed/expected CpG 0.6; minimum % GC 50.0. NB with these settings the HoxD10 PRE scores a short 200 bp GpG island; however, this was not detected by the more stringent settings used by Schorderet et al. (2013) and is designated as having no CpG island in Table 3 according to the authors of that study. Motifs for the DNA-binding proteins shown were scored as regular expressions with no mismatch allowance, as follows: REST: NTCAGCACCNNGGACAGCNCC; CP2: GCNCNANCCAG; RUNX:TGYGGT; YY1: GCCAT; GAF: GAGAGA, using the IUPAC code for non-conserved nucleotides as described in the legend to Table 2. b Occurrence per kb of motifs in the PREs shown and in random sequence (black). To generate random sequence, the total sequence of all elements shown (10.67 kb) was shuffled and searched for motifs. The mean of four iterations is shown

Fig. 6

Different modes of PcG binding and their resulting ChIP-binding profiles. On the left are shown different modes of dynamic binding of PcG proteins to PREs. On the right are shown the ChIP profiles that would result from each mode of binding. a Spreading. PcG proteins are recruited by a PRE and subsequently spread up and downstream (left), resulting in a broad ChIP peak (right) from which the PRE is not identifiable. b Looping. PcG proteins are recruited by a PRE and subsequently loop to the promoter via higher order interactions (left), resulting in two ChIP peaks (right) only one of which is a bona fide PRE. c Dynamic changes. In the example shown, PcG proteins are recruited by a PRE and are subsequently delivered to a different location (left), resulting in a ChIP peak at the site of delivery (in this example, the gene) but not at the site of entry (in this example, the PRE) (right). Variations on this theme include different profiles in different cell types, in which only a subset of multiple PREs may be occupied in different tissues or at different times

Fig. 7

Summary of similarities and differences relevant for fly and vertebrate PREs. The figure summarises the main points of this review. For PRC1, PRC2, TRXG, and target genes, key similarities and differences are listed, discussed in detail in the main text. For DNA-binding factors and PREs, open questions are identified, discussed in the conclusion section of the review