Polycomb (PcG) regulation has been thought to produce stable long-term gene silencing. Genomic analyses in Drosophila and mammals, however, have shown that it targets many genes, which can switch state during development. Genetic evidence indicates that critical for the active state of PcG target genes are the histone methyltransferases Trithorax (Trx) and Ash1. This study has analyzed the repertoire of alternative states in which PcG target genes are found in different Drosophila cell lines and the role of PcG proteins Trx and Ash1 in controlling these states. Using extensive genome-wide chromatin immunoprecipitation analysis, RNAi knockdowns, and quantitative RT-PCR, it was shown that, in addition to the known repressed state, PcG targets can reside in a transcriptionally active state characterized by formation of an extended domain enriched in Ash1, the N-terminal, but not C-terminal moiety of Trx and H3K27ac. Ash1/Trx N-ter domains and transcription are not incompatible with repressive marks, sometimes resulting in a 'balanced' state modulated by both repressors and activators. Often however, loss of PcG repression results instead in a 'void' state, lacking transcription, H3K27ac, or binding of Trx or Ash1. It is concluded that PcG repression is dynamic, not static, and that the propensity of a target gene to switch states depends on relative levels of PcG, Trx, and activators. N-ter Trx plays a remarkable role that antagonizes PcG repression and preempts H3K27 methylation by acetylation. This role is distinct from that usually attributed to Trx/MLL proteins at the promoter. These results have important implications for Polycomb gene regulation, the 'bivalent' chromatin state of embryonic stem cells, and gene expression in development (Schwartz, 2010).

Key to the current understanding of PcG mechanisms is the fact that, while PcG proteins are present in most kinds of cells, the decision whether or not to repress a target gene depends crucially on whether that gene had been repressed in the previous cell cycle. This effect is responsible for the epigenetic maintenance of the repressed state and associated chromatin modifications. Similarly, through the action of Trx and Ash1, a PcG target gene that had not been repressed tends not to become repressed in the subsequent cell cycle and remains susceptible to transcriptional activators. By comparing PcG/TrxG and transcriptional landscapes in three lines of Drosophila cultured cells it was found that the full repertoire of chromatin states that PcG target genes can assume is not limited to the repressed state dominated by PcG mechanisms and the transcriptionally active state governed by TrxG proteins but in addition includes transcriptionally active 'balanced' states subjected to simultaneous or at least rapidly alternating control by both PcG and TrxG proteins, and a transcriptionally inactive 'void' state lacking both PcG and TrxG control. Thus, although PcG mechanisms first achieved fame for producing stable long-term silenced states in Drosophila homeotic genes, it is clear that, in the general case, PcG states are not necessarily stable nor long-term (Schwartz, 2010).

The results establish clearly that in robust PcG target regions (i.e. Class I PcG target regions) PcG and TrxG regulation are tightly coupled. Considering the role of MLL1 in the regulation of HOX genes and the similarity between PcG complexes in flies and mammals, it is expected that the same holds true for mammalian cells. It is possible that PcG and Trx recruitment to PRE/TREs share some DNA-binding proteins or DNA motifs. It will be important to determine whether PcG and Trx bind simultaneously or alternate over time. The nature of the Trx complex that binds to PRE/TREs remains enigmatic. To date the only Trx complex characterized biochemically is TAC1, purified from Drosophila embryos (Petruk, 2001). It is said to contain uncleaved full length Trx, anti-phosphatase Sbf1 and histone acetyltransferase dCBP. This study could detect no uncleaved Trx in the nuclei of cultured cells indicating that the Trx bound at PRE/TREs of repressed genes does not represent TAC1. Proteolytically cleaved human orthologs of Trx, MLL1 and MLL2 have been purified as part of complexes similar in composition to the yeast COMPASS4. Although the PRE/TRE binds both parts of the cleaved Trx, it lacks some COMPASS components and lacks H3K4 trimethylation, suggesting that it involves a different complex whose composition is yet to be characterized (Schwartz, 2010).

Consistent with genetic evidence, the presence of Ash1 and Trx at PcG target regions is linked to their transcriptional activity. However the two proteins show important differences in their behavior: binding of Ash1 is limited to transcriptionally active (fully derepressed or balanced) PcG targets and is not detected at completely repressed target loci. Trx is more complex. Both N-ter and C-ter parts of the protein associate with PREs regardless of the transcriptional status of their target genes and bind in the vicinity of TSS specifically when a target gene is transcriptionally active. In addition, the N-ter moiety of Trx together with Ash1 forms broad domains that encompass transcriptionally active PcG target genes. The different behavior of N-ter and C-ter parts of Trx may account for the discrepancy between reports of the co-localization of Trx and PcG proteins at many chromosomal sites and reports claiming that Trx binds exclusively to transcriptionally active target genes. The different accounts are due to the use of anti-Trx antibodies specific to different parts of the protein (Schwartz, 2010).

Whether C-ter or N-ter specific antibodies were used, the number of Trx bound regions detected in these experiments is small compared to the number of active genes. This argues against a general role for Trx in transcription and is consistent with the limited number of regions detected on polytene chromosomes by various antibodies directed against C-ter or N-ter Trx. In marked contrast to these observations, Schuettengruber (2009) has recently reported exclusive association of N-ter but not C-ter moiety of Trx with TSS of most active transcription units in the chromatin of embryonic cells. The same report also asserted that in embryonic cells Trx C-ter but not Trx N-ter is bound at PREs. Remarkably the Trx N-ter specific antibody used by Schuettengruber is reportedly the same as one of the two used in experiments from this lab. While the possibility cannot be excluded that the behavior of the N-ter moiety of Trx in embryos is totally different from that in cultured or salivary gland cells, it is suspected that more likely the preparation of the antibody used by Schuettengruber cross-reacted with some general transcription factor particularly abundant in embryonic cells. This emphasizes the importance, even the necessity, of using two or more independent antibodies to verify genome-wide ChIP results (Schwartz, 2010).

The RNAi knockdown experiments show that the broad binding of Ash1 and Trx N-ter within transcriptionally active PcG target regions is interdependent. This is consistent with the reported dissociation of Ash1 from polytene chromosomes of the salivary gland cells subjected to Trx RNAi and the severe reduction of Trx N-ter binding to polytene chromosomes of ash1 mutant larvae. Despite interdependence in binding there is no compelling evidence that Ash1 and Trx N-ter are in the same protein complex. Although an interaction between Trx and Ash1 has been reported, it was said to require the intact SET domain of Trx, which is absent from its N-ter moiety. This study has not found that Trx N-ter and Ash1 co-precipitate from nuclear extracts, strengthening the impression that the two peptides do not interact directly (Schwartz, 2010).

A histone mark associated with Ash1/Trx N-ter domains is H3K27ac. Acetylation of H3K27 can antagonize PcG activity by competing with the placement of the H3K27me3 mark, which in our model is needed for effective contact of the PRE complex with the promoter, as well as for stable PcG binding. In fact, targeting a histone H3 acetylase to a PRE is sufficient to prevent the epigenetic maintenance of repression indicates that in Drosophila the HAT responsible for bulk acetylation of H3K27 is CREB-binding protein (dCBP), encoded by the nejire gene. Direct association of both Trx and Ash1 with dCBP has been previously reported. Consistent with this, the Trx knock-down experiment, which also impairs Ash1 binding, shows that either or both proteins promote acetylation of K27 in the chromatin of active PcG targets. It also shows, however, that the level of H3K27ac is not directly related to the amount of Ash1 or Trx N-ter bound. It is possible that a small amount of Trx and/or Ash1 remaining on the chromosomes after RNAi depletion is sufficient to target enough HAT activity to maintain nearly normal levels of H3K27ac. Alternatively H3K27 acetylation may be produced by a process that is not mechanistically linked to Ash1 or Trx but is promoted by the two. A global reduction of immunostaining of polytene chromosomes with anti-H3K27ac antibodies and a global elevation of immunostaining with anti-H3K27me3 antibodies in trx mutant larvae was recently reported. The current study did not detect any global changes in either H3K27ac or H3K27me3 levels. It is possible that the effects are generally weak and could only be detected on polytene chromosomes that consist of thousands of chromatin fibers bundled together. It is noted, however, that the reported changes of H3K27 acetylation and trimethylation levels involved numerous chromosomal sites, most of which, according to current data, do not stably bind PcG or TrxG, suggesting that in these cases the effect of trx mutation may have been indirect (Schwartz, 2010).

The microarray data show that narrower peaks of H3K27ac are also found near numerous active promoters, at genes not known to be PcG targets. A role of H3K27ac at these promoters may be to antagonize dimethylation of H3K27, which is abundantly distributed throughout the genome and may have a general negative effect on transcription (Schwartz, 2010).

In Drosophila cultured cells most PcG target genes are either completely repressed or fully derepressed or entirely devoid of both PcG and TrxG regulation. However in about 5% of cases, exemplified by the Psc-Su(z)2 or inv-en loci, binding of PcG complexes does not result in complete transcriptional silencing and can coexist with binding of Ash1/Trx N-ter. Whether a PcG target gene is capable of and will assume a 'balanced' state may depend on the nature of the PRE, its binding complexes, and the promoter of the target gene. More likely, however, the major determinants are the nature and concentration of activators and repressors that act in concert with PcG/TrxG. Consistent with this idea, in imaginal disc cells, which are controlled by much more complex regulatory networks than cultured cells, simultaneous presence of both PcG and TrxG proteins at the transcriptionally active PcG target genes appears to be more common. Interestingly a common feature of the 'balanced' chromatin state in both cultured and imaginal disc cells is the confinement of Ash1/Trx N-ter binding to the regions immediately around the promoters. This is taken as a hint that the formation of a broad Ash1/Trx N-ter domain starts in the vicinity of the TSS (Schwartz, 2010).

Much has been said about the 'bivalent' state, i.e. containing both PcG repressive marks and transcriptional activity marks, as characteristic of genes in mammalian embryonic stem cells. In Drosophila, in cases such as those of the Psc-Su(z)2 or inv-en loci, the balanced action of PcG and TrxG results in chromatin states similar in appearance to the 'bivalent' state. It is supposed that, like the 'balanced' chromatin state of Drosophila PcG targets, the 'bivalent' domains of embryonic stem cells would be associated with both PcG proteins and mammalian orthologs of Trx and Ash1 (Schwartz, 2010).

PcG target genes may also assume a state lacking both PcG and TrxG proteins. The fact that in several instances the same locus resides in this 'void' chromatin state in cultured cells of completely different origin argues against it being a product of genomic aberrations. Several experimental observations also indicate that the 'void' state is not a peculiarity of cultured cells. Thus, in salivary glands the hh gene lacks PcG binding and H3K27me3 but remains transcriptionally inactive, as in D23 cells. More recently a comparison of embryonic and imaginal disc cells showed that in many cases lack of PC binding was not accompanied by transcriptional activity. The void chromatin state might be simply interpreted as a derepressed region that is transcriptionally inactive because the needed activator is absent. However, this does not explain why in this state Trx is also absent from the PRE, implying that neither PcG nor Trx binding is the default state of the PRE or that some specific condition prevents the recruitment of both. No other known repressive marks such as H3K9 methylation has been detected at these sites (Schwartz, 2010).

In line with the finding that the lack of PcG repression in the 'void' chromatin state does not automatically lead to the activation of a target gene is the observation that PC knock-down elicits a very specific genomic response. Remarkably Sex combs reduced (Scr), Antennapedia (Antp) and Abd-B, the HOX genes whose derepression in heterozygous +/Pc^- flies gives the famous Polycomb phenotype, are also among the genes most sensitive to PC knockdown in BG3 cells. The sensitivity of these genes cannot be explained by intrinsically poor recruitment of PcG proteins as Abd-B and Scr are controlled by multiple strong PREs capable of robust repression when placed next to a reporter gene. It is supposed that the reason for the differential sensitivity to PC levels lies in the availability of the corresponding transcriptional activators. It is propose that in the BG3 cells the transcriptional activators of sensitive PcG target genes are present but at levels insufficient to override repression under normal conditions. The knockdown of PC lowers the threshold required for derepression. It is suggested that a general role of PcG and TrxG mechanisms is to modulate the constraints on the levels of transcriptional activators required to switch the expression of PcG target genes. This concept helps to explain why, despite the implication of the PcG system in the control of all morphogenetic pathways, the reduction of PcG levels during differentiation of mammalian cell lineages or tissue regeneration in flies results in the execution of very specific genomic programs (Schwartz, 2010).