Title: Polycombing the Genome: PcG, trxG, and Chromatin Silencing
Abstract: "Whereof one cannot speak, thereof one must be silent": thus spake 18Wittgenstein L "Wovon man nicht sprechen kann, darübermuss man schweigen," Annalen der Naturphilosophie. 1921; 14: 185-262Google Scholar. Tautological, perhaps, and certainly inapplicable to human affairs, yet a fair description of Polycomb group–dependent chromatin silencing. Polycomb group (PcG) proteins (reviewed by13Pirrotta V Curr. Opin. Genet. Dev. 1997; 7: 249-258Crossref PubMed Scopus (205) Google Scholar, 14Schumacher A Magnuson T Trends Genet. 1997; 13: 167-170Abstract Full Text PDF PubMed Scopus (140) Google Scholar) interact with many genes in Drosophila as well as in vertebrates, but their best-understood role is in the regulation of Drosophila homeotic genes. The pattern of expression of these genes is set in the early embryo by transient activators and repressors that define the segmental domains of expression. Shortly after gastrulation, when the early repressors disappear, the PcG protein complexes take over, establishing a silenced state at those genes that were initially repressed but not at those initially active. More remarkably, the PcG complexes preserve a memory of the early state of activity through many rounds of cell division so that once a gene has been silenced in a cell, it remains silent in its progeny. This implies that early events alter the chromatin in a heritable way, determining either the open or the closed state in the cellular descendants. PcG proteins are necessary to establish this epigenetic state, and their continued expression (with the exception of Esc) is required to maintain the state in later development. Note that to be epigenetically maintained, the repressed state must involve a durable and self-reproducing modification of the chromatin. The PcG complex is not simply reconstituted de novo every cell cycle since the reassembly process must discriminate between target sites that were repressed at earlier times and target sites that were not repressed. Furthermore, this mechanism affects all the known enhancers of a gene such as Ubx, though they are active at different times and in different tissues and are scattered over a distance of some 100 kb. Another set of proteins, the trithorax group (trxG) contributes to strong expression of the homeotic genes and counteracts, in some measure, the effect of PcG genes. Many of the proteins classified in this group are likely to act in diverse and independent ways not necessarily related to PcG silencing. The analysis of homeotic gene regulation shows that trxG mechanisms do not maintain continuous activity of homeotic genes, whose expression occurs only in specific tissues and specific stages in response to corresponding enhancers: more likely, their role is to stimulate this enhancer-dependent expression, possibly by keeping the genes "open" and accessible. Silencing by the PcG mechanism is mediated by Polycomb response elements (PREs), regulatory regions of several hundred nucleotides that are in vivo binding sites for PcG proteins and often for trxG proteins. In a reporter construct, a PRE can establish a silenced state affecting the activity of multiple genes contained in the construct or flanking its insertion site. The multitude of problems surrounding our current understanding of PcG silencing can be summarized by three questions: (1) how does the PcG complex form; (2) how does it silence gene expression; and (3) how does it maintain its cellular memory? Recent work has made important progress in answering these questions and the results have implications for the dynamics of nuclear organization and long-term stability of chromatin states. None of the known PcG proteins binds to DNA in vitro, but they can interact with one another forming multiprotein complexes. Perhaps the formation of a complex generates a DNA binding activity. PREs are compound structures that can be subdivided into multiple shorter regions that retain some independent PRE activity, but although some PREs contain targets for known DNA-binding proteins such as GAGA factor, it has not been possible to identify distinct consensus sequences common to all PREs. On the contrary, GAGA factor is associated with the promoters of many genes such as the heat shock genes or Ubx itself, where it does not induce silencing but is important for normal transcriptional activity. Many PRE properties suggest that the formation of the complex is a highly cooperative process and that at different genomic sites different subsets of PcG proteins are involved. Most likely, PcG complex formation is dependent on a mosaic of interactions either of the different PcG proteins with DNA or of multiple DNA-binding proteins that act as recruiters, not unlike the assembly of factors at enhancer modules. We know that targeting a single PcG protein by fusing it with a DNA-binding domain suffices to recruit a functional PcG complex and silence a reporter gene (11Müller J EMBO J. 1995; 14: 1209-1220Crossref PubMed Scopus (120) Google Scholar). Recruitment mechanisms are well known in yeast where DNA-binding proteins such as RAP1 are required to initiate the assembly of the telomeric and mating-type silencing complexes. They also operate in more localized types of gene silencing. The yeast α2 DNA-binding protein recruits a silencing complex including Tup-1, a WD repeat protein that acts as a repressor, possibly by interacting with adjacent nucleosomes (5Edmondson D.G Smith M.M Roth S.Y Genes Dev. 1996; 10: 1247-1259Crossref PubMed Scopus (402) Google Scholar). The Drosophila Dorsal protein, a transcriptional activator homologous to mammalian NF-κB, in concert with other DNA-binding factors, silences by recruiting other proteins, including Groucho, a WD repeat protein (4Dubnicoff T Valentine S.A Chen G Shi T Lengyel J.A Paroush Z Courey A.J Genes Dev. 1997; 11: 2952-2957Crossref PubMed Scopus (128) Google Scholar). Interestingly, the establishment of PcG silencing at some but not all PREs requires Esc, a WD repeat protein present in the early embryo but not needed to maintain silencing in later development. PcG silencing may then be a more elaborate version of a widely used silencing strategy, specialized perhaps to act at greater distances or to persist through cell division. Does the complex nucleated at the PRE spread to flanking sequences to silence a large chromatin domain? (Figure 1). This view was derived initially by analogy with position–effect variegation where the heterochromatic cytological appearance spreads from centric heterochromatin to invade euchromatic regions. It was supported by the strong dependence of silencing on the dosage of PcG genes and more recently by the demonstration that in yeast the silencing complex recruited at telomeres can spread up to 15–20 kb when the silencing proteins are overexpressed (6Hecht A Strahl-Bolsinger S Grunstein M Nature. 1996; 383: 92-96Crossref PubMed Scopus (440) Google Scholar). Chromatin cross-linking experiments initially indicated that Polycomb protein was associated with large tracts of the bithorax complex. However, more recent experiments using a refined technique show that PcG proteins are linked primarily to the vicinity of known PREs and decrease nearly to background levels within one or two kilobases (17Strutt H Cavalli G Paro R EMBO J. 1997; 16: 3621-3632Crossref PubMed Scopus (203) Google Scholar). If spreading occurs, it involves interactions with DNA that are more subtle or indirect than those involved at the PRE itself and are not detected by cross-linking experiments. The interactive properties of the PcG proteins are probably responsible for a number of features of PREs. One of these is the homing phenomenon: transposons containing a PRE often integrate near endogenous PRE sites, suggesting an interaction between endogenous and transgenic PcG complexes. Such interactions are also implied by the pairing effect, whereby the silencing of a PRE-containing transposon construct is often dramatically enhanced in flies homozygous for the transposon insertion. The pairing of the homologous chromosomes and consequent pairing of the two copies of the construct clearly increases the stability or silencing power of the PcG complex. More remarkably, PRE-containing transposons inserted at different sites or even on different chromosomes can, in some cases, interact with one another with similar synergistic effects on silencing (16Sigrist C.J.A Pirrotta V Genetics. 1997; 147: 209-221Crossref PubMed Google Scholar). In the nucleus, PcG complexes formed at one PRE can apparently "comb" the nuclear environment in search of related complexes with which to associate (Figure 2). The Drosophila genome contains at least 100 PcG-binding loci, but the activity of any one PRE depends on the interactions accessible from its genomic location. This is very similar to heterochromatic position effects where the silencing ability of a block of heterochromatin is strongly dependent on its proximity to centric heterochromatin (3Csink A.K Henikoff S Nature. 1996; 381: 529-531Crossref PubMed Scopus (276) Google Scholar). In yeast also, the vicinity of a silencer to telomeres enhances its silencing ability (7Maillet L Boscheron C Gotta M Marcand S Gilson E Gasser S.M Genes Dev. 1996; 10: 1796-1811Crossref PubMed Scopus (238) Google Scholar). The explanation is likely to be the same: interactions between complexes or local concentrations of the components of the silencing complex. The ability to search the nuclear environment implied by these phenomena can have surprising consequences. 12Pal-Bhadra M Bhadra U Birchler J.A Cell. 1997; 90: 479-490Abstract Full Text Full Text PDF PubMed Scopus (184) Google Scholar found that increasing the number of transgene copies, inserted at different genomic sites, causes a progressive decrease in their collective level of expression. Such effects were observed using a chimeric transgene containing the white promoter and the Adh transcriptional unit and affected in parallel the expression of the endogenous Adh gene but not of the endogenous white gene. Most surprisingly, this cosuppression is dependent on PcG genes and is associated with the emergence of new PcG protein–binding sites at the target genes as their number of copies in the genome increases. The simplest interpretation is that at least some genes contain sequences that act as weak PREs. A single copy of such genes is not significantly repressed, but when the number of copies in the nucleus increases, they begin to find one another, driven by homology as well as by interactions between transiently bound PcG proteins. Such trans-pairing effects could eventually stabilize the formation of PcG complexes and result in a degree of silencing. As expected, paired copies of the transgene are more effective in silencing one another than dispersed copies since pairing of the homologous chromosomes facilitates the search for interacting sequences. In either case, the search takes time, as chromosomes emerge from mitosis, and is limited by the length of interphase and by the position of individual sequences in the nucleus (8Marshall W.F Fung J.C Sedat J.W Curr. Opin. Genet. Dev. 1997; 7: 259-263Crossref PubMed Scopus (84) Google Scholar). In Drosophila, even the pairing of homologs occurs gradually in the course of development, presumably because it is prevented by the short cell cycles during embryonic development. In larvae, the lengthening interphases would allow both pairing and trans-interactions, resulting in more stable PcG complexes and more efficient silencing. PcG complexes formed at a PRE affect enhancers or promoters over distances of 20–30 kb. It is often assumed that they package chromatin into a more compact form, rendering the DNA inaccessible to transcriptional activators. Though silenced loci appear somewhat more condensed in polytene chromosomes, the idea that silencing is caused by packaging up the chromatin is borrowed principally from heterochromatic silencing since heterochromatin is both more condensed and underreplicated. In fact, for both PcG targets and heterochromatin, there is no compelling evidence that condensation is the cause of silencing rather than a consequence of transcriptional silence. Could PcG complexes coat the chromatin or otherwise block the access to DNA? 9McCall K Bender W EMBO J. 1996; 15: 569-580Crossref PubMed Scopus (82) Google Scholar found that a reporter gene inserted within the Ubx transcription unit was efficiently silenced in parallel with Ubx itself while phage T7 RNA polymerase could still recognize its promoter inserted in the same place. Since T7 polymerase is a much smaller protein than the large apparatus required to initiate transcription from a Pol II promoter, perhaps the difference lies in the ability of the smaller T7 protein to slip through PcG complexes. However, the cross-linking experiments suggest that PcG proteins do not spread to coat the gene but are principally associated with the immediate surroundings of the known PREs, while enhancers and promoters lie tens of kilobases away. Enhancers act at a distance by a mechanism most commonly envisioned as a looping of the enhancer-activator complex to contact the promoter complex. A similar looping model could be applied to silencing by the PRE. An intermediate "hop and skip" model, combining the features of looping and spreading, envisions the formation of a core complex at the PRE which can then interact and stabilize weaker complexes formed at frequently occurring but weak proto-PRE elements lying along the path from PRE to enhancers and promoters (16Sigrist C.J.A Pirrotta V Genetics. 1997; 147: 209-221Crossref PubMed Google Scholar). In this model the PRE would proceed by a series of short loops from one such way station to another until it is within striking distance of the promoter complex (Figure 1B). If the PcG complex does not package chromatin, how could it effect silencing? Here our ignorance becomes virtually complete. PcG proteins might interact directly with promoter or enhancer proteins or, in keeping with current trends, we might imagine that they recruit enzymes that alter the state of acetylation of histones and remodel the chromatin, rendering the DNA less accessible. Whatever effects the silencing, a massive pulse of activator can even displace a preexisting PcG complex (19Zink D Paro R EMBO J. 1995; 14: 5660-5671Crossref PubMed Scopus (173) Google Scholar). Most likely, then, the formation of a silencing complex and the binding of an activator involve mutually incompatible chromatin states. A characteristic feature of PcG complexes is their self-maintaining property or cellular memory. If the PcG complex is an extended structure, we might suppose that parts of the silenced region could undergo DNA replication while other parts with their complexed proteins constitute a sufficient nucleus to reassemble the silenced state as the replication wave passes through. However, if the PRE is excised from a reporter construct during development, using the FLP recombinase, silencing cannot be maintained (1Busturia A Wightman C.D Sakonju S Development. 1997; 124: 4343-4350PubMed Google Scholar). The PcG chromatin complex does not organize flanking chromatin in a self-renewing structure. It is probably dissociated at each mitotic cycle, requiring the PRE not only for initiating but also for maintaining the silenced state. To account for the cellular memory, that is, the reconstitution of the complex only at PREs that were previously silenced but not those that had no previous complexes, we might suppose that some residual proteins remain associated with the PRE to "mark" it for rapid reassembly, or that the PRE chromatin has been modified by the silencing, for example by deacetylating the nucleosomes. Similarly, the PREs of "open" genes might be marked by some proteins that prevent the de novo assembly of PcG complexes (10Michelotti E.F Sanford S Levens D Nature. 1997; 388: 895-899Crossref PubMed Scopus (132) Google Scholar) or by a modification of the chromatin such as acetylation. The state of acetylation would constitute a marker with the required properties. During DNA replication, the semiconservative partitioning of the old nucleosomes on the daughter DNA molecules could provide the link with the previous chromatin state, provided that the presence of acetylated nucleosomes activates a function to acetylate the newly deposed nucleosomes and maintain the fully acetylated state. This might be a role for Trx and related proteins (Figure 3). 2Cavalli G Paro R Cell, in press. 1998; Google Scholar now add a surprising new dimension to the question of cellular memory. They used a lacZ reporter gene construct activated by a GAL4 UAS and containing the Fab-7 PRE from the bithorax complex. The transposon construct also contains the white gene as a marker to identify the transgenic flies. In these flies, the PRE represses the basal expression of lacZ and strongly silences the white gene, resulting in weak and variegated eye color. Massive production of GAL4 from another construct driven by a heat shock promoter activates the lacZ transgene and displaces the PcG complex from the PRE. Remarkably, when the activation is induced during embryonic development, the derepressed state of both lacZ and white persists through larval and pupal development. The PRE is evidently so completely stripped of PcG proteins that it cannot reestablish silencing. A simple interpretation of these results might be that silencing complexes can only be established de novo in the early embryo. However, if GAL4 activation takes place during larval development, the derepression is only transient and silencing returns. This could be explained by increased stability of silencing complexes as trans-interactions become possible with longer interphases. In larval cells, PcG proteins might not be completely displaced by GAL4 or might reassemble more easily at the PRE. Hypersilenced states, as well as derepressed states, can be inherited by cell progeny. For PREs such hypersilenced states are induced by growing the embryos at higher temperature (in contrast, for unknown reasons, heterochromatic silencing is alleviated by higher temperatures). When they are returned to lower temperature for the remainder of development, the resulting adults still display higher levels of silencing. These results suggest that some components of the PcG complex remain associated with the chromatin or modify it in a way that persists through DNA replication and mitosis. Yet, none of these explanations can easily account for the further finding of 2Cavalli G Paro R Cell, in press. 1998; Google Scholar that the unsilenced configuration is not only maintained through many cell divisions but to some extent also through meiosis, fertilization, and development of the next generation. The derepressed state resulting from GAL4 induction in germ line cells apparently survives the extensive chromatin reconfiguration that takes place in gametogenesis, and in some way repression fails to be reestablished in up to one-fourth of the G1 embryos. One explanation for this failure would require a maintenance mechanism for the active state. That is, activation would result in a chromatin modification by a mechanism that is self-renewing every cell cycle and that persists through meiosis. Cavalli and Paro propose that trxG proteins might maintain the "open" state, preventing the reconstitution of the silencing complex. In fact, GAGA factor and Trx bind to Fab-7 and other PREs both in the silent and in the active state. Perhaps they lie in wait for the opportunity to institute the open state when the silencing complex is displaced. However, the fact that derepression in larvae is only transient argues against a simple version of this scenario. A more adventurous speculation is that the silent state is the normal state of chromatin in the germ line, where most somatic genes, including homeotic genes, would be inactive. The massive expression of the GAL4 activator in this experiment is obtained by induction of a heat shock promoter, hence GAL4-dependent activation occurs also in the germ line. As a consequence, the zygote would begin life with the reporter construct in a derepressed state. In fact, a global silencing system exists in the germ line cells of C. elegans (15Seydoux G Mello C.C Pettit J Wood W.B Priess J.R Fire A Nature. 1996; 382: 713-716Crossref PubMed Scopus (245) Google Scholar). Clearly more surprises are in store. What is needed now is a better understanding of the structural and molecular changes associated with silenced chromatin.