QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Bernstein, E. Articles by Allis, C. D. Search for Related Content PubMed PubMed Citation Articles by Bernstein, E. Articles by Allis, C. D. Related Content Chromatin and Gene Expression Post-transcriptional Control Social Bookmarking                   What's this?

REVIEW

RNA meets chromatin

Laboratory of Chromatin Biology, The Rockefeller University, New York, New York 10021, USA

	Abstract

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
In the universe of science, two worlds have recently collided—those of RNA and chromatin. The intersection of these two fields has been impending, but evidence for such a meaningful collision has only recently become apparent. In this review, we discuss the implications for noncoding RNAs and the formation of specialized chromatin domains in various epigenetic processes as diverse as dosage compensation, RNA interference-mediated heterochromatin assembly and gene silencing, and programmed DNA elimination. While mechanistic details as to how the RNA and chromatin worlds connect remain unclear, intriguing parallels exist in the overall design and machinery used in model organisms from all eukaryotic kingdoms. The role of potential RNA-binding chromatin-associated proteins will be discussed as one possible link between RNA and chromatin.

[Keywords: RNAi; dosage compensation; heterochromatin; transcriptional gene silencing]

We favor the general view that a complete appreciation of epigenetic regulation is likely to require a careful examination of both RNA and chromatin fields. One goal of this review is to expose potential links between these two research areas, with a focus on transcriptional gene silencing in a wide range of experimental models. We conclude with a speculative model for how a group of heterochromatin-associated proteins may participate in linking RNA and chromatin.

	Chromatin dynamics

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
The nucleosome is the fundamental repeating subunit of chromatin, consisting of an octamer of histone proteins (two copies of each of the four core histones H2A, H2B, H3, and H4) around which a defined segment (147 base pairs) of DNA is wound (Luger et al. 1997). Relatively unstructured histone "tail" domains are exposed on the nucleosomal surface and are rich in post-translational modifications such as serine and threonine phosphorylation, lysine acetylation, lysine and arginine methylation, lysine ubiquitination and sumoylation, and ADP-ribosylation (Strahl and Allis 2000; Zhang and Reinberg 2001). Emerging evidence suggests that covalent marks also exist in the histone-fold domains that correlate with the lateral surface of nucleosomes, near where critical histone:DNA interactions occur. Such modifications may mediate other aspects of chromatin structure (Cosgrove et al. 2004). Regardless of position or function, these covalent marks can be placed and removed by a variety of site-specific enzymes such as histone acetyltransferases (HATs) and histone deacetylases (HDACs), histone methyltransferases (HMTs), and the newly discovered "demethylating" activities (Cuthbert et al. 2004; Shi et al. 2004; Wang et al. 2004), ubiquitin ligases and deubiquitinases, and so on. Additional complexity and biological readout options are provided by the fact that certain modifications, such as methylation, can exist in mono-, di- and trimethylated forms, each of which is placed by a specific HMT. Finally, the notion that patterns of histone modifications can either coexist, on the same tail (cis) or on distinct tails (trans) that mediate potential "cross-talk" or "switching" between distinct marks and their binding partners (see below), has been formally proposed (Jenuwein and Allis 2001; Fischle et al. 2003a) and is currently under active investigation.

The sheer complexity of covalent histone modifications is multiplied by the existence of histone variants in many organisms, that give the cell added opportunities to change the overall composition of the nucleosome and its covalent modification potential (for review, see Kamakaka and Biggins 2005). In either the case of conventional or variant histones, the fundamental question of what covalent modifications actually do, remains. It has long been recognized that two general types of mechanisms exist that are certainly not mutually exclusive. These include trans mechanisms, which involve the binding of what we will refer to in this review as "effectors" that engage specific covalent marks in a context-dependent fashion, and cis mechanisms, which involve structural alterations in the chromatin fiber by changing internucleosomal contacts. Together, this variation may contribute to what has been referred to as a "histone code" (Turner 1993, 2000; Strahl and Allis 2000), which remains poorly understood.

View larger version (28K): [in this window] [in a new window]   Figure 1. Chromatin and RNA regulation across eukaryotic kingdoms. Noncoding RNAs and histone modifications are involved in the formation of "active" chromatin (euchromatin is represented in green shading) or "inactive" chromatin states (heterochromatin is shown in red shading). Representing the kingdom of Fungi is the fission yeast, S. pombe, which mediates silencing of repetitive regions such as centromeres, mat loci, and telomeres through an RNAi-like mechanism. The diagram represents a S. pombe centromere. The innermost repeats (imr) of the centromere proper are shown in relation to the dh-dg repeats of the pericentromeric region, in which transcripts were detected in mutants defective for the RNAi machinery. The heterochromatin of the centromere is marked primarily by H3K9me (red), and possibly by other unidentified histone modifications. Representing the Protist kingdom is Tetrahymena thermophila, a binucleated ciliate that dramatically rearranges its genome during its sexual life cycle. (Top) In the vegetative state, the micronucleus (MIC) is the transcriptionally silent, germline nucleus (red), containing 100% of the sequence complexity of the organism. The macronucleus (MAC) is the transcriptionally active, somatic nucleus that governs the phenotype of the cell (green; histone modifications include H3K4me). (Bottom) At the time of DNA elimination, which occurs during conjugation, the MIC (red) has divided to form the new MAC and MIC, and subsequently the old MAC will be destroyed. The new MAC is active (green) with discrete ring-like structures (marked primarily by H3K9 and K27 methylation, red) in which IES DNA will be eliminated. siRNAs play a role in DNA elimination and heterochromatin formation in this protist, as represented by the "scan RNA" model (see Fig. 2). In the Plant kingdom, various mechanisms are at play to silence transgenes and repetitive elements. Heterochromatic "knobs" (red), first described in maize, are rich in transposable elements and satellite repeats, and are common in plant genomes. These knobs are regulated by siRNAs, H3K9me (red), and the chromatin-remodeling ATPase, DDM1 in Arabidopsis. Whether other histone modifications, such as H3K27me, are enriched in these regions, is currently unknown. Representing the Animal kingdom, flies and mammals both undergo a process of dosage compensation; twofold up-regulation of transcription of the male X in flies and silencing of the female X chromosome in mammals. The roX RNAs are integral to the dosage compensation complex in flies, and the active X chromosome is marked by histone modifications, such as H3K4me and H4K16ac. The polytene chromosomes are represented, with the male X chromosome shown in green (active) and puffed due to transcription. The chromocenter, a transcriptionally inactive, tightly knit structure that contains the centromeres, is shown in red (silent) and is marked by H3K9me. In mammals, the female X chromosome is coated by the Xist RNA, which is critical for silencing, and is marked mainly by H3K9me, H3K27me, and H4K20me. siRNA-mediated TGS also exists in flies and mammals (not illustrated here), but does not seem to play a role in dosage compensation in either animal. Note: Methylation states (mono, di, tri) have been excluded for simplicity (see text for details and references).

Numerous studies suggest that euchromatin is often positively correlated with hyperacetylation of lysine residues on each of the core histones, as well as other active marks (e.g., methylation of Lys 4, Lys 36, and Lys 79 in H3). Conversely, common features of heterochromatin include deacetylated core histones and the addition of repressive marks (e.g., various degrees of methylation of Lys 9 and Lys 27 in H3 and Lys 20 in H4), DNA methylation, and the recruitment of repressive binding effectors such as heterochromatin protein 1 (HP1) and other methyl-binding proteins. Interestingly, this more compacted, higher-order chromatin state can be spread along silent domains and stably inherited (for review, see Elgin and Grewal 2003), although mechanistic details as to how this occurs are not yet clear. Moreover, heterochromatin plays a critical role in establishing chromosomal domains, such as centromeric, pericentromeric, and telomeric regions, that are vital for proper chromosome segregation (Allshire et al. 1995; Kellum and Alberts 1995; Grunstein 1997a; Karpen and Allshire 1997; Ekwall et al. 1999) and in less well-appreciated events such as programmed DNA elimination that distinguishes the germline from soma in some organisms (Fig. 1; Mochizuki and Gorovsky 2004). Thus, the relationship between euchromatin and heterochromatin, in part dictated by covalent modifications of histone proteins, provides an elegant balance for the regulation of epigenetic states, and may have much more significance than simply governing gene expression.

	RNA intricacies

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
New evidence from a variety of model systems (Fig. 1) demonstrates that noncoding RNAs play a significant role in the control of epigenetic regulation, chromosomal dynamics, and long-range interactions, as well as phenomena that have an impact on normal cellular differentiation and organismal development. This includes mechanisms such as dosage compensation in flies and mammals, imprinting (for review, see Morey and Avner 2004; O'Neill 2005), and gene silencing by RNA interference (RNAi) phenomena in a variety of organisms.

Nowhere is the use of noncoding RNA in forming euchromatin versus heterochromatin better illustrated than with dosage compensation, a process that keeps the balance between the sex chromosomes (typically there exists two X chromosomes in females and one X chromosome and one Y chromosome in males). In both flies and mammals, noncoding RNAs are intimately involved in dosage compensation, as are changes in chromatin structure through histone modifications (Fig. 1; see below).

Considerable attention is currently being focused on RNAi-related silencing phenomena, processes that drastically alter gene expression among diverse species. Remarkable progress has been made in elucidating the mechanisms of RNAi that lead not only to post-transcriptional gene silencing (PTGS), but also to transcriptional gene silencing (TGS) (for review, see Hannon 2002; Grewal and Moazed 2003; Meister and Tuschl 2004; Sontheimer 2005; Tomari and Zamore 2005). This review will concentrate on emerging evidence that links RNAi-like mechanisms to the regulation of TGS through changes in chromatin.

Early evidence for an RNAi-like process directing epigenetic changes includes RNA-directed DNA methylation (RdDM) in recombinant viroid-infected plants where double-stranded RNA (dsRNA) directed at the promoter region of a gene causes de novo cytosine methylation at homologous DNA sequences, resulting in stable silencing (Wassenegger et al. 1994; Pelissier and Wassenegger 2000; Jones et al. 2001). Underscoring another fundamental mechanism, a role for RNA in the establishment of heterochromatin formation has been uncovered from a series of elegant studies in fission yeast (Hall et al. 2002; Volpe et al. 2002) and plants (for review, see Gendrel and Colot 2005). Here, small RNA molecules that are products of the RNAi machinery can achieve TGS through chromatin changes, particularly characterized by histone methylation, and/or, in particular systems, DNA methylation. RNAi-mediated transcriptional gene silencing studies have also been extended to other model systems, such as Drosophila and mammals, leading to the emerging view that this is an ancient and universal mechanism, although the intricacies and details vary from one organism to the next.

Thus, while traditionally separate fields, new evidence provides hints and clues to the ways in which chromatin and RNA may be intertwined. Throughout this review, we focus on RNA-directed chromatin-based alterations, particularly those that are brought about by noncoding RNAs.

	Dosage compensation: a strong link between RNA and chromatin

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
A growing number of studies suggest that noncoding RNAs are involved in a myriad of regulatory processes in the cell. One such process with a long history in the chromatin field is that of dosage compensation, a process that equalizes the output of transcription from the uneven number of X chromosomes found in male and females. Dosage compensation in both Drosophila and mammals is intimately linked to noncoding RNAs, which ultimately leads to changes in chromatin structure to either repress (mammals) or activate (Drosophila) transcription of X-linked genes (see Fig. 1). For clarity, each of these models is discussed separately.

Dosage compensation in flies: linking RNA with active histone acetylation marks

Transcription from the single male X chromosome in Drosophila is doubled in order to ensure equal levels of X-linked genes in males and females. If this twofold transcriptional activation fails to occur properly, a male-specific lethal phenotype ensues. A ribonucleoprotein complex (RNP) is involved in this process, namely, the dosage compensation complex (DCC, also referred to as the MSL complex). This complex is comprised of five core proteins, generally known as the MSL (male-specific lethal) proteins: MSL1, MSL2, MSL3, MLE (maleless, an RNA helicase), MOF (males absent on the first, a chromodomain-containing histone acetyltransferase or HAT), and a H3 Ser 10 kinase known as JIL, which interacts with the MSL proteins (for review, see Gilfillan et al. 2004). A key regulatory component of the DCC is MOF, which hyperacetylates H4 at Lys 16 and is, at least in part, responsible for the up-regulation of transcription from the X chromosome (Gilfillan et al. 2004). However, it remains unclear how this particular acetylation mark in H4, long known to be a residue critical for silencing in yeast in its unacetylated form (Kayne et al. 1988; Grunstein 1997b; Carmen et al. 2002), acts to bring about transcriptional up-regulation. It is tempting to speculate that Lys 16 acetylation either serves to recruit a positively acting effector to the male X chromosome or that this acetylation marks serves to displace a silencing factor. Relevant to the theme of this review, it remains of interest to know what role noncoding RNAs serve in this unique regulation of transcription (see below).

Two noncoding RNAs exist in the DCC, known as roX1 and roX2, which are essential components of the DCC complex and whose expression is male-specific. They are functionally redundant; only loss of both roX1 and roX2 dramatically affects male viability (Gilfillan et al. 2004). The MSL components are thought to assemble at 35 chromatin-entry sites along the male X chromosome, two of which encode the roX genes themselves. These sites serve as nucleation sites for spreading of the MSL complex into the flanking chromatin (Kelley and Kuroda 2000). The roX RNAs play an essential role in targeting the DCC to the X chromosome by mechanisms that remain unclear. Deletion of roX1 and roX2 results in both mislocalization of the DCC complex and the acetylation of histone H4 at Lys 16 by the MOF histone acetyltransferase (Gu et al. 1998; Meller and Rattner 2002).

Interestingly, three of the DCC components are reputed RNA-interacting proteins, although the exact significance of these findings is not yet clear. The MLE helicase has an RNase-sensitive association with the X chromosome (Richter et al. 1996), and the chromodomain of MOF has been shown to bind RNA in vivo and in vitro and may represent a targeting mechanism for the DCC to the male X chromosome (Akhtar et al. 2000). Another component of the DCC complex, MSL3, contains two CDs, and this protein was also found to bind RNA in an in vitro setting (Akhtar et al. 2000). These findings may underscore a more general property of chromodomain-containing proteins: Do these proteins contain motifs permitting a novel mechanism for the targeting of proteins by RNA (see Fig. 3A,B)? In particular, is there a mechanism wherein RNA-directed recruitment of histone-modifying activities to a target locus (or in this case, entry sites along a chromosome) is likely to be involved in the critical regulation of gene dosage using either RNA:RNA or RNA:DNA base-pairing interactions (Fig. 3C)? Extending these ideas to mammals, we imagine that HMTs that catalyze more repressive, silencing marks on the inactive X chromosome, act in a similar fashion through a critical RNA association (see below). Support for this model comes from studies in flies, in which mutants lacking subunits of the DCC, particularly the RNA-binding components, show reduced binding to the male X chromosome (Richter et al. 1996; Gu et al. 1998). Understanding the precise mechanisms, or motif(s), that link the suspected RNA to chromatin remains an important challenge for the future.

Dosage compensation in mammals: linking RNA with repressive histone methylation marks

Remarkably, dosage compensation is handled quite differently in XX female mammals, where one of the two X chromosomes is inactivated to provide gene dosage between the sexes. In general, two types of X-chromosome inactivation have been well characterized in mice: (1) an imprinted inactivation mechanism that occurs in extra-embryonic tissues that selects the paternal X chromosome to be silenced, and (2) a random mechanism in the embryo proper that occurs later in development and allows the female cell to "choose" which chromosome to inactivate (Heard 2004). We focus on the latter process. Random X inactivation is a multistep process involving choice of which chromosome to inactive, establishment of the inactive X state (Xi), and maintenance of its silent state in subsequent cell divisions (Avner and Heard 2001).

Unlike in flies, where multiple chromatin-entry sites serve as docking sites for the DCC and from where the complex spreads in cis, inactivation in mammals is initiated at a single site known as the X inactivation center (Xic). The Xic produces multiple noncoding RNAs including Xist and its antisense RNA counterpart Tsix. Xist RNA is expressed exclusively from the X chromosome to be inactivated and is retained in the nucleus, where it coats the X chromosome. Xist RNA is 17 kb (depending on species) and is capped, spliced, and poly-adenylated. The primary Tsix transcript covers at least 40 kb, overlaps the full length of Xist, and regulates its expression; it too undergoes complex processing. Transcription of Tsix is thought to exert a repressive effect on the accumulation of Xist RNA at the onset of X inactivation, by mechanisms that are currently under investigation (Heard 2004).

Interestingly, while the initiation of X inactivation is absolutely dependent on the Xist RNA, this RNA is not required for maintenance of the heterochromatic state of the Xi (Penny et al. 1996; Heard 2004). Deletions of Xist cause embryonic lethality in mice and in embryonic stem (ES) cells, and mutations in conserved regions of Xist, such as the 5' stem-loop A-repeats, prevent the X chromosome from being inactivated (Penny et al. 1996; Marahrens et al. 1997; Wutz et al. 2002). Once Xist initiates X inactivation, a chain of chromatin-remodeling events occurs in order to ensure silencing. This includes the generation of hypoacetylated isoforms of histones H3 and H4, and methylation to varying degrees of residues Lys 9 and Lys 27 in H3 and Lys 20 in H4 (Fig. 1). Moreover, the inactive X incorporates a specialized H2A histone variant known as macroH2A (see below), although the significance of this variant residing in facultative heterochromatin is not known. In keeping with other forms of specialized heterochromatin, the inactive X is also marked by DNA methylation and replicates late in S phase (Heard 2004). Thus, this chromosome, but not its genetically identical sister, has become transformed into an epigenetically silent chromosome during development. This occurs, in part, by a redundancy of silencing marks that ensure its propagation into a stable heterochromatic state.

Enzymes that are responsible for "writing" repressive histone marks on the inactive X chromosome (e.g., dimethyl Lys 9 and trimethyl Lys 27 in H3 and monomethyl Lys 20 in H4) are slowly being uncovered. One such enzyme, E(z), a HMT that is a member of the Polycomb group (PcG) gene family, has been implicated in regulating Lys 27 methylation on the Xi (Plath et al. 2003; Silva et al. 2003). The PcG genes are required for maintaining stable repression of various genes during Drosophila development, including the homeobox (Hox) genes (for review, see Ringrose and Paro 2004). Collectively, there are 30–40 members of this family, and mutations in these genes lead to homeotic transformations. The mouse counterparts of the Drosophila PcG genes ESC (Extra sex combs) and E(z) (Enhancer of zeste) are Eed (Embryonic ectoderm expression) and Enx1, respectively. These polypeptides form HMT complexes that methylate Lys 27 predominantly, and to a lesser extent, Lys 9 of histone H3 (Kuzmichev et al. 2002). Eed is a WD40-repeat protein required for the methyltransferase activity of (E)z. In mice deficient for Eed, E(z) is not recruited to the Xi, nor is there any detectable Lys 27 methylation (Silva et al. 2003). Interestingly, in these mice, even though the X chromosome is still coated by Xist, sporadic X-linked gene reactivation and changes in chromatin structure have been observed, including H3 lysine acetylation and Lys 4 methylation, two "active" marks (Silva et al. 2003). Together, these data suggest that, although there is likely some redundancy between "repressive" marks, Lys 27 methylation plays a critical role in this epigenetic silencing. The HMT responsible for H3 Lys 9 methylation on the Xi is still somewhat of a mystery, although it has recently been shown that G9a dimethylates the Lys 9 hotspot in the Xic that lies 5' to Xist (Rougeulle et al. 2004). However, mice deficient for G9a have proper regulation of Xist expression and stable maintenance of the Xi (Ohhata et al. 2004). The HMT that monomethylates H4 Lys 20 on the Xi is currently unknown, although PR-SET7 is a likely candidate.

The Eed/E(z) complex is important for early maintenance of the inactive X; however, there are other Xist-independent mechanisms for sustaining this inactive state. Due to their role in developmentally regulated repression in Drosophila, it had been proposed that the PcG genes might also be involved in the maintenance of silencing (Heard 2004). Recently, it has been shown that members of the PRC1 (Polycomb repressive complex 1) including Cbx2 (a mammalian Polycomb homolog containing a chromodomain), Polyhomeotic 1 and 2 (Phc), and Bmi1, are localized to the inactive X chromosome; some of these are recruited in a cell cycle-dependent manner (Plath et al. 2004; Hernandez-Munoz et al. 2005). Interestingly, accumulation of these proteins is not solely regulated by H3 Lys 27 methylation, but is dependent on the presence of the Xist RNA. The mechanistic link(s) between Xist RNA and the chromatin alterations described above are not known, and remain an important area for future research.

Overall, the inactive X is characterized by a series of covalent histone modifications that easily fit into the category of repressive marks (Fig. 1). It becomes of interest to compare these modifications to those on the active X chromosome or genes on the inactive X that "escape" inactivation. By chromatin immunoprecipitation (ChIP) analyses, promoter regions have been systematically compared on the inactive X with the homologous regions on the active X. While H3 Lys 9 and Lys 27 methylation is generally linked to gene silencing, genes on the active X display H3 Lys 4 methylation and hyperacetylation of histones H3 and H4 (Goto et al. 2002). Moreover, the chromatin signature of genes escaping X inactivation includes H3 Lys 4 methylation, hyperacetylation of H3, and hypoacetylation of CpG islands surrounding these genes (Goodfellow et al. 1988; Boggs et al. 2002; Filippova et al. 2005). Exactly how this separation of inactive and active chromatin along the X chromosome is achieved is not known, but it may be regulated by insulator and/or boundary elements (Filippova et al. 2005). In some ways, these findings are reminiscent of the large-scale genomic indexing relationships between Lys 4 and Lys 9 methylation that occur along domains of the chicken -globin locus (Litt et al. 2001) and the mating-type locus in fission yeast (Noma et al. 2001).

	Heterochromatin meets RNAi

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
RNA interference is a relatively novel silencing process that was discovered serendipitously in Caenorhabditis elegans when researchers were attempting to use antisense RNA approaches to knock down gene function (Guo and Kemphues 1995). Unexpectedly, sense RNA gave essentially the same knock-down effects as antisense RNA, an inexplicable observation at the time. Mello and Fire then demonstrated that double-stranded RNA was the trigger for the observed potent gene silencing (Fire et al. 1998). This seminal discovery led to the identification of RNAi as a widespread gene-silencing mechanism used by many organisms. Besides being a valuable research tool, the mechanism of RNAi is a fascinating one. The effects of RNAi can be threefold; this process can result in messenger RNA degradation, translational inhibition, chromatin changes, or possibly even a combination of these.

A brief discussion of the principal components of the RNAi machinery is necessary in order to appreciate RNAi-mediated TGS mechanisms. It has been recognized for some time that the "canonical" RNAi pathway uses small interfering RNAs (siRNAs) to induce degradation of cognate messenger RNAs in a sequence-specific manner (for review, see Hannon 2002; Meister and Tuschl 2004). This involves the processing of dsRNA by Dicer, the RNA substrate of which may be exogenously introduced by experimental manipulation, viruses, endogenous parasitic elements, or repetitive sequences. The enzyme Dicer belongs to the RNase III family of ribonucleases, which creates, through its action, small interfering RNAs (siRNAs) that are central to RNAi and related phenomena. These small RNAs are then loaded into dedicated machinery, known as the RISC complex (RNA-induced silencing complex), whose nuclease activity is encoded by the Argonaute-2 protein (Liu et al. 2004; Meister et al. 2004; Song et al. 2004). RNAi is also involved in the regulation of translation, in which endogenous microRNA precursors (pre-miRNAs) are sequentially processed by the Drosha and Dicer RNase III enzymes, yielding microRNAs (miRNAs). miRNAs bind the 3'-UTR of their target genes and inhibit translation by a currently unknown mechanism (for review, see Carmell and Hannon 2004; Cullen 2004). Through various cloning studies, it is becoming clear that scores of miRNAs are encoded in the genome in a variety of organisms, ranging from viruses to plants to mammals (for review, see He and Hannon 2004; Pfeffer et al. 2004). miRNAs are responsible for key processes of development such as patterning, cell symmetry, and proliferation (Brennecke et al. 2003; Johnston and Hobert 2003; Palatnik et al. 2003; Juarez et al. 2004; Kidner and Martienssen 2004), and are implicated in disease (Poy et al. 2004). However, RNA degradation and translational inhibition, both PTGS mechanisms, are not the focus of this review. Instead, we focus on RNAi-mediated TGS effects such as chromatin modifications, particularly histone methylation. Like PTGS, this silencing is instigated or "guided" by a subset of small RNAs generated by the RNAi machinery.

The idea that an RNAi-like mechanism could target locus-specific domains for chromatin alterations was not anticipated. RNAi has now been demonstrated to have transcriptional gene-silencing effects, most notably through the establishment of heterochromatin in fission yeast. RNAi-mediated chromatin effects have also been uncovered in organisms as diverse as Tetrahymena, Drosophila, and mammals, but the detailed mechanisms have yet to be revealed. The discovery that RNAi plays a role in heterochromatin formation came from landmark genetic studies that investigated the effects of defective RNAi machinery on the silencing that occurs at centromeres and mating-type loci in Schizosaccharomyces pombe (Hall et al. 2002; Volpe et al. 2002; see below).

RNAi and repetitive elements

In all complex genomes, endogenous repetitive elements, as well as multiple copies of defective and intact transposable elements, form a significant fraction of the genomic complexity. These elements, which have been described as "junk" DNA, are often assembled into condensed, transcriptionally silent chromatin states (heterochromatin), in part due to the combined action of histone and/or DNA covalent modifications, notably methylation (see below for discussion of centromeres). Suppression of these elements is essential for genetic stability in two ways: Intact transposons are potential mutagens if integrated into coding genes, and repetitive sequences are potential sites for nonhomologous crossovers.

RNAi-like mechanisms are now known to play a critical role in mediating heterochromatic gene silencing and can prevent the mobilization of transposable elements (Matzke et al. 2000; Sijen and Plasterk 2003). Clear evidence has come from studies in C. elegans in which RNAi-deficient worms show high rates of transposition (Ketting et al. 1999; Tabara et al. 1999). In Drosophila, I elements (similar to mammalian LINE elements) can be silenced by previous introduction of transgenes expressing a small region of the transposon (Jensen et al. 1999). Defects of RNAi were also found to relieve silencing of tandem transgene arrays in Neurospora crassa and Arabidopsis (Cogoni and Macino 1999; Mourrain et al. 2000). In the mouse embryo, knock-down of Dicer results in an increase in retrotransposon (IAP and MuERV-L) transcript abundance, suggesting that RNAi constrains expression of repetitive parasitic sequences in the preimplantation embryo (Svoboda et al. 2004a). Furthermore, in an effort to identify endogenous targets of RNAi, the sequencing of small RNAs has revealed sequences corresponding to endogenous transposons and other repetitive sequences in Drosophila and plants (Hamilton et al. 2002; Llave et al. 2002; Aravin et al. 2003). Although the underlying mechanism of these processes is still under investigation, together these results indicate that RNAi has evolved, in part, to maintain genomic stability and may be a conserved mechanism across species.

Because RNAi is central to heterochromatin formation, the intuitive belief that silent chromatin is not transcribed (and therefore, that RNA is not available or required to initiate silencing) needs to be challenged. "Readthrough" transcription of the aforementioned repetitive elements, for example, can produce dsRNAs that initiate silencing. Bidirectional transcription has been detected in mutant plants that are defective for silencing, including met1, a DNA methyltransferase, and ddm1 (decrease in DNA methylation), a SWI2/SNF2-related chromatin-remodeling protein (Lippman et al. 2004). Once dsRNA is generated, it can then be processed into small RNAs through the Dicer family of RNase III enzymes and fed into the appropriate silencing machinery. Underscoring the generality of these mechanisms, centromeric repeats are transcribed in S. pombe (Volpe et al. 2002) as are micronuclear germline sequences in Tetrahymena (Chalker and Yao 2001), a nucleus often described as being "silent" (see below).

An intriguing hypothesis has been put forward to explain how tandem arrays are important for the maintenance of silencing (Martienssen 2003). In some organisms, including C. elegans, plants, and fission yeast, a mechanism is in place to amplify the RNA signal, namely, an RNA-dependent RNA polymerase (RdRP). Given that mutations of the RdRP in various systems abolish silencing, tandem repeats may be distinguished from dispersed repeats by their inherent ability to sustain RdRP activity over multiple rounds of RNAi (due to the fact that the primer for the RdRP, the small RNAs, would be unlimited in tandem arrays). However, dispersed or single-copy elements would eventually lose RNAi capability since their supply of primers would eventually be depleted. This model is tantalizing; however, it does not account for the fact that RdRPs do not necessarily require a primer and cannot be applied to Drosophila and mammals that appear to lack RdRP enzymes (Makeyev and Bamford 2002).

Recent mapping studies in plants and mammals have suggested that repetitive elements are strongly correlated with histone methylation and the production of RNA. In Arabidopsis, mapping across large chromosomal regions containing heterochromatic knobs (Fig. 1) has revealed a robust correlation between DNA repeats, DNA methylation, noncoding RNA, and histone H3 Lys 9 methylation (Lippman et al. 2004). In addition, a profile analysis of histone methylation states (H3 Lys 9, Lys 27, and H4 Lys 20 in the mono-, di-, and trimethylated forms) has been performed on repeat-associated sequences in the mouse genome (Martens et al. 2005). Tandem repeats such as the major and minor satellite sequences were examined, as well as other repetitive elements, including various transposons. In accord with the model described above, tandem repeats, but not interspersed elements, were found to give rise to dsRNAs. Interestingly, the level of these RNAs is elevated in cells deficient for the SUV39H histone methyltransferases (Martens et al. 2005). This finding leads to the intriguing possibility that recruitment of RNAi processing complexes is impaired in the absence of H3 Lys 9 methylation at these repetitive regions, an analogous situation to the silencing of centromeres in fission yeast (see below).

RNAi and centromeres

Heterochromatin stabilizes repetitive DNA sequences or multiple copies of transposable elements at centromeres, telomeres, and other regions of the genome by preventing recombination between homologous sequences (Karpen and Allshire 1997; Csink and Henikoff 1998). S. pombe has provided an excellent model system in which to investigate the role of RNAi-induced TGS at these regions. Using this organism, several groups have demonstrated that small RNAs play a critical role in regulating heterochromatin formation. The first of these studies examined the loss-of-function effects of various RNAi genes, including dicer (dcr1), Argonaute 1 (ago1), and the RNA-dependent RNA polymerase (rdp1), each the sole genes of their kind in this yeast (Hall et al. 2002; Volpe et al. 2002). Disruption of these genes resulted in the elimination of silencing of a ura4 marker gene inserted into the inner and outer repeats of the centromere, similar to the effects of disrupting the swi6 (HP1 homolog) and clr4 (H3 Lys 9 HMT) genes (Ekwall et al. 1996; Volpe et al. 2002). In addition, a corresponding loss of H3 Lys 9 methylation was observed. The loss of silencing at the centromeres results in catastrophe for the cell, including abnormal chromosome segregation (Ekwall et al. 1995; Partridge et al. 2002; Provost et al. 2002). Similar studies, performed at the mating-type locus (mat), which contains sequences homologous to the centromeric repeats, illustrated that the RNAi machinery was also required for silencing at this region. However, while RNAi is involved in the establishment and maintenance of silencing at the centromeres, it is involved exclusively in the establishment of the mat locus (Hall et al. 2002).

The RNAs involved in establishing this transcriptional silencing have been highly sought after. Long, noncoding RNAs homologous to the centromeric repeats were found to accumulate in the dcr1, ago1, and rdp1 mutant cells, but not in wild-type cells. Nuclear run-on experiments suggested that these RNAs are continuously being made in wild-type cells, but are actively diced into small RNAs and thus are not detected (Volpe et al. 2002). The small RNAs homologous to the centromeric repeats have been detected by various means in fission yeast, including small RNA cloning studies that revealed abundant species homologous to the centromeric sequences (Reinhart and Bartel 2002).

An important breakthrough in understanding the mechanism of TGS came with the biochemical purification of complexes that contain the small RNAs crucial to this process. A RNP complex known as RITS (RNA-induced transcriptional silencing)—which contains Ago1, Chp1 (a chromodomain-containing protein), and a novel protein of unknown function referred to as Tas3—was purified from S. pombe (Verdel et al. 2004). The RNAs in this complex are thought to direct sequence-specific targeting of the RNAi machinery to homologous sequences (see Fig. 3C). In addition, ChIP experiments suggest that the RITS complex is bound to all known heterochromatic loci, including centromeres, telomeres, and the mat locus (Volpe et al. 2002; Noma et al. 2004). RITS association with these loci is dependent on H3 Lys 9 methylation, and the chromodomain of Chp1 is thought to recognize and bind to this histone modification as a "docking" effector. Importantly, loading of this complex with small RNAs and tethering it to chromatin is essential for silencing and for the generation of additional siRNAs required to bring about heterochromatin formation. The RITS complex has been found to interact with another complex through Ago1 in an siRNA-dependent manner. This complex, RDRC (RNA-dependent RNA polymerase complex), contains Rdp1, a novel RNA helicase and an intriguing member of the poly A polymerase family (Motamedi et al. 2004). Cells defective for these proteins fail to localize RITS to centromeric regions and initiate heterochromatin assembly. Moreover, both RITS and RDRC localize to noncoding centromeric transcripts (Motamedi et al. 2004).

The collective model arising from these exciting studies suggests a critical role for RNAi in the establishment of heterochromatin at repetitive loci. In such a model, these repeats generate RNAs that can be double-stranded in nature or form dsRNA intermediates, which are then processed by the RNAi machinery in the nucleus to produce small RNAs. These RNAs may serve as "guides" for targeting histone methyltransferase complexes to the chromatin, such as the H3 Lys 9 HMT, Clr4, although this mechanism is not yet understood. Once such a guiding RNA mark is positioned, through base-pairing with its complementary DNA sequence or nascent transcript (as seems the case for centromeric regulation), complexes analogous to RITS can dock onto modified sites through methyl-histone recognition by chromodomain-containing proteins (see Figs. 2, 3). This mechanism would facilitate silencing and subsequent spreading by HP1 and its homologs, and may represent a self-enforcing loop to maintain silencing at specific loci, such as centromeres and other highly repetitive regions of the genome. Although H3 Lys 27 methylation has not yet been reported in fission yeast, it is appealing to investigate whether RNAi can direct this inactive chromatin mark in other systems as well (Fig. 1).

The mechanism of centromeric silencing described above may be conserved among species. The link between RNAi and heterochromatin formation has certainly been established in plants. For example, an Argonaute family member in Arabidopsis, Ago4, is required for locus-specific siRNA accumulation as well as DNA and histone methylation (Zilberman et al. 2003). In addition, it has been described in mammalian cells that an RNA component is required for the integrity of pericentric heterochromatin (Maison et al. 2002), and embryonic stem cells deficient for the RNAi machinery, particularly Dicer, have centromeric silencing defects (Kanellopoulou et al. 2005; see below). Interestingly, the only known PcG-binding sites in mammals consist of pericentromeric heterochromatin repeats (Saurin et al. 1998). Determining the sequential order of these proposed steps and learning the identity of all binding adaptors or effectors remain important areas for future investigation.

RNAi and DNA elimination

DNA elimination is considered by some to be the "ultimate form of gene silencing" (Mochizuki and Gorovsky 2004). Ciliated protozoans such as Tetrahymena and Paramecium are binucleated unicellular organisms that rearrange their genomes significantly during the developmental process of conjugation (Figs. 1, 2; for review, see Prescott 1994). During vegetative growth, the nonsexual phase of the life cycle, these ciliates contain two functionally distinct nuclei within the same cytoplasm, namely, a micronucleus (MIC) and a macronucleus (MAC). The macronucleus is responsible for transcriptional activity and is thus analogous to the somatic nucleus of a metazoan; it is responsible for determining the phenotype of the cell and is polyploid. The micronucleus, in contrast, is transcriptionally silent and diploid; it is considered the germline nucleus containing 100% of the sequence complexity of the organism. During the sexual process of conjugation, the micronucleus divides to form the next generation of macro- and micronuclei, and concomitant with formation of new macronuclei, the old (parental) macronucleus is destroyed (Fig. 1). During the differentiation of new macronuclei (also known as anlagen), 10%–15% of the germline genome is eliminated by deletion of several thousand discrete internally eliminated sequences (IESs). Thus these two nuclei, both derived from the same zygotic nucleus, differ markedly in their genomic organization and sequence content.

Although several cis-acting DNA elements involved in such DNA rearrangements have been identified (Coyne et al. 1996; Wuitschick and Karrer 2003), these DNA sequences fail to explain how a heterogeneous collection of IESs are recognized and eliminated in a stage-specific fashion. Thus, it has long been suspected that epigenetic mechanisms may play a role in the general process of programmed DNA elimination. Supporting evidence was obtained when primary DNA sequences from the parental MAC were shown to dictate which sequences are eliminated in the newly forming MAC, a form of epigenetic communication that lacked a molecular explanation. This hypothesis was tested experimentally by introducing IES DNA sequences into the parental MAC (in which these sequences have already been eliminated and thus no longer exist), which resulted in inhibition of DNA elimination of the IES in the newly forming MAC (Chalker and Yao 1996).

Recently, important molecular insights have been gained into the process of DNA elimination in ciliates by demonstrating that this remarkable genomic-downsizing process is also mediated by an RNAi-like mechanism (Mochizuki et al. 2002; Yao et al. 2003; Garnier et al. 2004). Pivotal to these findings was the discovery of a remarkably abundant population of small (28-nt) RNAs specifically expressed during the conjugation pathway prior to the timing of large-scale DNA elimination (Mochizuki et al. 2002). These small RNAs were found to be enriched in micronuclear-limited sequences, and may be required for recognizing IESs in the developing macronucleus. Furthermore, an Argonaute family member, Twi1, is required for DNA elimination. In Twi1-deficient cells, small RNAs are detected at very low levels, suggesting a role for Twi1 in the generation and/or stability of these RNAs (Mochizuki et al. 2002). An RNAi-based mechanism for the process of genome-wide rearrangements was further demonstrated by the injection of dsRNA corresponding to sequences that are not normally eliminated in the MAC (Yao et al. 2003). The introduction of this RNA early in conjugation triggered efficient deletion of corresponding sequences in the newly forming MAC, suggesting that RNA may mediate the long-acknowledged communication between the old MAC and the developing new MAC (see internuclear "communication" illustrated in Fig. 2). In some ways, these studies are reminiscent of work in fission yeast where addition of a dsRNA hairpin can trigger the silencing of ectopic sequences by inciting heterochromatin assembly (Schramke and Allshire 2003).

The above data suggest that sequence elimination in ciliates occurs via an RNAi-like mechanism. Similar to RNAi-mediated silencing in fission yeast, the chromatin in which DNA elimination takes place is heterochromatic in nature, and is marked by methylated histones and enrichment in chromodomain-containing effector proteins. Interestingly, a small collection of stage-specific and actively synthesized polypeptides was identified in Tetrahymena that are collectively referred to as Pdd (Programmed DNA degradation) proteins (Madireddi et al. 1994, 1996). Pdd1 contains multiple copies of a chromodomain motif, and the other three polypeptides include Pdd2, Pdd3 (another chromodomain family member), and Twi1 (formerly p90) (Madireddi et al. 1994; Smothers et al. 1997; Nikiforov et al. 1999, 2000; Y. Liu, E. Bernstein, and C.D. Allis, unpubl.). The connection between heterochromatin formation and programmed DNA elimination has been strengthened by the finding that Pdd1 and Pdd3 bind methylated Lys 9 in H3, a distinctive mark of IES elements that is observed prior to DNA elimination (Taverna et al. 2002). Interestingly, DNA to be eliminated in also marked by Lsy27 methylation (S.D. Taverna, Y. Liu, and C.D. Allis, unpubl.).

View larger version (60K): [in this window] [in a new window]   Figure 2. Tetrahymena "Scan RNA" model. The model depicted represents the "scan RNA" model put forth by Gorovsky and colleagues (Mochizuki and Gorovsky 2004). This scan model proposes that the entire micronuclear genome or regions immediately surrounding IESs are bidirectionally transcribed in the MIC. This transcription occurs prior to DNA elimination (depicted by asterisk); the MIC is active (green) for a short period when it changes morphology and elongates. During DNA elimination, the MIC resumes its silent state (shown in red in mating pair). The small (28-nt) RNAs are the products of dsRNA processing by Dicer-like ribonucleases, and are thought to be escorted by Twi1 from the site of dicing (presumably in the micronucleus) through the old MAC prior to its destruction (red), in which IESs have previously been eliminated. Those sequences having a homologous macronuclear DNA sequence are proposed to be degraded, while only those scnRNAs without homologous sequences in the old MAC (IESs) remain in association with Twi1. These scnRNAs are then transported again (presumably in association with Twi1) to the new MAC (green; active except for regions of DNA elimination, see red rings in mating pair), where they direct DNA elimination of IESs by unknown excision machinery. It is likely that Twi1 and associated scnRNAs interact with the Pdd proteins, two of which (Pdd1 and Pdd3) dock on H3K9me sites. Nucleosomes and histone tails are not drawn to scale.

Although a silencing complex akin to RITS has yet to be identified in Tetrahymena, an analogous complex likely exists that contains scnRNAs, Twi1, and various Pdd proteins (Fig. 2). We envision that the small RNAs, acting as the sequence-specificity determinants, would direct the relevant complexes to its IES target sites. The chromodomain-containing Pdds would recognize the heterochromatic marks, including H3 Lys 9 methylation (and perhaps other repressive histone modifications), at sequences to be eliminated (Fig. 2). It is currently unknown how chromatin-modifying enzymes are targeted to these sites, but a HMT could very well be a component of such a complex. Subsequently, unknown excision machinery (possibly similar to the RISC complex, containing a endonuclease for excision) would be recruited to eliminate these sequences from the genome. This drastic form of gene silencing in ciliates, as well as the heterochromatization of centromeres in fission yeast, demonstrates how striking the epigenetic consequences of RNAi-directed processes can be.

Transcriptional gene silencing in flies

RNA-directed silencing mechanisms have been investigated in Drosophila, including both PTGS and TGS mechanisms. For the scope of this review, we concentrate on the latter. It has been demonstrated that the introduction of repetitive transgenes in Drosophila can result in both repression of the transgene and the endogenous locus, a phenomenon known as cosuppression. Cosuppression, an RNAi-related process, occurs in many organisms including plants and fungi (for review, see Bernstein et al. 2001). Transgene cosuppression was one of the first lines of evidence to suggest the potential for RNA-directed chromatin changes in Drosophila (Pal-Bhadra et al. 2002). In the case of white-Adh repetitive fusion transgenes, silencing is mediated by the PcG proteins, including Pc and Pc-like, as mutation of these genes relieves silencing. Furthermore, some of the PcG gene products are strongly recruited to the transgene insertion sites (Pal-Bhadra et al. 1997).

Comparable to fission yeast, mutations in the RNAi machinery also relieve heterochromatin-mediated silencing in Drosophila. Using mini-white transgene arrays and white transgenes in heterochromatic regions, Pal-Bhadra et al. (1997) demonstrated that mutations in piwi, aubergine (both Argonaute family members), and homeless cause a loss of silencing, or in other words, suppress PEV (Pal-Bhadra et al. 2004). In addition, this loss of silencing is correlated with a reduction in H3 Lys 9 methylation and delocalization of HP1 and HP2, most dramatically observed in the hls mutants. Interestingly, mutations in hls increase the expression of some retrotransposons and genomic repeats, and mutant embryos are resistant to injection of dsRNA (Aravin et al. 2001; Kennerdell et al. 2002). Although the exact molecular function of hls is unknown, its domain structure may be revealing. It contains not only a DEAD-box RNA helicase motif, but also a tudor domain. Intriguingly, tudor domains have been suggested to be a potential chromatin-binding module, as well as a potential RNA-binding motif, somewhat like the well-characterized chromodomain (Fig. 3; see below).

Transcriptional gene silencing in mammals

Transcriptional gene silencing mechanisms in mammals, analogous to those in plants and S. pombe, are still somewhat of a mystery, unlike PTGS, which has been well established in mammals. The heritable and epigenetic side of TGS is just being exposed, and mechanistic details are not yet clear. Two recent reports have suggested that gene silencing, mediated by DNA methylation, can be induced by promoter-directed siRNAs in mammalian cells (Kawasaki and Taira 2004; Morris et al. 2004). Taira and colleagues also found that siRNAs targeted to a promoter could induce H3 Lys 9 methylation in various mammalian cell lines (Kawasaki and Taira 2004). However, other reports have illustrated the inability to direct siRNA-induced DNA methylation. For example, siRNAs targeted to the coding region of the Huntingtin gene are not able to direct DNA methylation of the corresponding genomic locus, but are capable of reducing mRNA levels (Park et al. 2004). Another report proved unable to detect DNA methylation in mouse oocytes targeted with a 500-bp hairpin RNA (Svoboda et al. 2004b). However, the differences in experimental manipulations (i.e., targeting a promoter vs. a coding region, nuclear import of siRNAs) may be a critical factor in the outcome and interpretation of transcriptional silencing in these systems.

The requirement for RNAi in centromere silencing has been well established in fission yeast, and thus, it was hypothesized that RNAi would play a pivotal role in regulating mammalian centromeres. RNA has been implicated in the structure and maintenance of mouse pericentromeric heterochromatin (Maison et al. 2002). Evidence for RNAi-directed centromere regulation was demonstrated in chicken–human hybrid DT40 cells containing human chromosome 21, in which Dicer had been inactivated by homologous recombination (Fukagawa et al. 2004). These Dicer-deficient cells have defects in sister-chromatid cohesion, and de-repress centromeric heterochromatin, as witnessed by the accumulation of transcripts from the human satellite sequences (Fukagawa et al. 2004). Moreover, HP1 proteins were found to be partially delocalized in the nucleus, suggesting a disruption of heterochromatin by the RNAi machinery. Similar results were observed for Dicer-deficient embryonic stem cells (Kanellopoulou et al. 2005). Centromere-derived transcripts were up-regulated in Dicer null cells, and DNA methylation was partially abolished at the minor satellite repeats. These cells also had reduced levels of H3 Lys 9 methylation and diffuse staining of HP1 and HP1. However, another report examining the loss of Dicer in ES cells discovered that transcripts derived from the centromeric major satellite accumulate, but retain the presence of cy tosine methylation (G. Hannon, pers. comm.).

Taken together, these results suggest that DNA methylation may not be the main component compromised by the loss of Dicer, but rather the loss of histone modifications such as H3 Lys 9 methylation. In support, transcription across the satellite repeats seems to be controlled, at least in part, by histone modification status. Mouse ES cells deficient in the histone methyltransferase SU(VAR)3–9 have a slight increase in the accumulation of transcripts derived from the major satellites (Lehnertz et al. 2003). Noteworthy is the fact that, thus far, centromeric small RNAs have not been identified from the many cloning studies performed in mammalian cell lines and tissues. It may be that they are in low abundance in wild-type cells; however, proving their existence will strengthen the role of RNAi-mediated chromatin regulation in mammals.

	Missing links: heterochromatin-associated proteins that may interact with RNA

Top Abstract Chromatin dynamics RNA intricacies Dosage compensation: a strong... Heterochromatin meets RNAi Missing links: heterochromatin... Intriguing links between RNA... Unanswered questions and future... Summary Acknowledgments References

 
The RNAi machinery contains RNA-binding proteins that guide and transport RNA, such as the Argonaute family of proteins that bind small RNAs through their PAZ (piwi/argonaute/zwille) domain (Lingel et al. 2003; Song et al. 2003; Yan et al. 2003; Ma et al. 2004). However, it remains a formal possibility that chromatin-associated proteins bind RNA as well. Such interactions may help to guide and stabilize chromatin changes, by providing an additional means by which chromatin effectors can form a close association to chromatin, transcribing or otherwise. Various such chromatin-associated proteins have been implicated in RNA binding, including HP1, the chromodomains of MOF and MSL, PcG proteins, as well as others (see below). Thus, while coding RNA has had a longstanding link to actively transcribing chromatin, the use of noncoding RNA to guide chromatin regulators or to associate them more stably to chromatin, by mechanisms that remain unclear, are relatively new concepts. Here, we present a speculative model wherein these chromatin regulators may provide a direct link to RNA. In proposing this model, we acknowledge that there may be little experimental support in favor of these ideas and that other models are formally possible. In presenting these ideas, we attempt to review both supporting and opposing viewpoints. Importantly, we look forward to the next generation of experiments aimed at testing these ideas in both in vitro and in vivo settings.

Chromodomain-containing proteins

The chromodomain (CD) was originally identified in the chromatin-associated, architectural proteins Heterochromatin Protein 1 (HP1) and Polycomb (Pc) of Drosophila melanogaster, leading to the proposal that they represented a conserved motif that might be involved in mediating protein:protein interactions (Paro and Hogness 1991). The chromodomain is found in a wide range of chromatin-associated proteins, mostly those involved in heterochromatin and/or transcriptional repression. These include the histone H3 Lys 9 histone methyltransferase SU(VAR)3–9 (suppressor of variegation), chromomethyltransferases (CMTs) in plants that also encode DNA methyltransferases, as well as the aforementioned HP1 and Pc proteins. However, chromodomains are also found in proteins known to be involved in transcriptional activation. These include members of the MYST family of HATs: MOF, a histone H4 Lys 16 acetyltransferase involved in dosage compensation in flies (see above), human Tip60, a transcriptional coactivator involved in DNA repair and apoptosis (Ikura et al. 2000), and Esa1, a major essential histone H4 acetyltransferase in Saccharomyces cerevisiae with known functions in transcriptional control as well as DNA repair (Bird et al. 2002). This brief survey suggests a possible dichotomy of function for the chromodomain, or more likely, a conserved role in chromatin binding in which the outcome, either gene activation or silencing (or other), is dictated by the protein itself and those polypeptides and/or nucleic acids with which it associates.

Excellent progress has been made in demonstrating that the chromodomain is, indeed, a protein:protein interaction module, specifically by its ability to bind to methylated histone peptides. For example, the CD of HP1 has been shown to bind the histone H3 tail at di or trimethylated Lys 9, while interestingly, the CD of Pc has the ability to specifically bind trimethylated Lys 27 on the same histone tail despite an identical sequence immediately surrounding the target methylated lysine (Bannister et al. 2001; Jacobs et al. 2001; Lachner et al. 2001; Fischle et al. 2003b). Moreover, atomic structures of both protein–peptide complexes have been determined, which suggest a common binding fold using two features: (1) The H3 peptide itself binds by inserting itself as a missing "rung" in a -stranded core architecture; and (2) the positively charged methyl ammonium ion is often "caged" by three highly conserved aromatic residues of the CD (Jacobs and Khorasanizadeh 2002; Nielsen et al. 2002; Min et al. 2003). However, CDs may also be more permissive in their binding capabilities. Several reports have suggested that CDs also bind to nucleic acids, both RNA and DNA (Akhtar et al. 2000; Bouazoune et al. 2002). Here, we focus our attention on potential RNA-binding properties of the chromodomain, as well as that of other chromatin-associated proteins.

The CDs of the Drosophila DCC

As previously mentioned, the dosage compensation complex of Drosophila contains roX RNAs as well as two CD-containing proteins, MOF and MSL. These proteins were shown to interact with the X chromosome in an RNase-sensitive manner and to bind RNA in vitro (Akhtar et al. 2000). RNase treatment of Drosophila S2 cells resulted in loss of MOF staining on the X chromosome, and electromobility shift assays revealed an intriguing interaction between MOF and RNA. A mutation in one of the caging aromatic residues of the MOF chromodomain resulted in a loss of RNA binding, suggesting that these residues may also be important for nucleic acid binding (Akhtar et al. 2000). Although it has since been shown that MOF CD interaction with the roX RNAs plays only a minor role in targeting it to the X chromosome in vivo, the MOF CD:RNA interaction still potentially contributes to its localization (Morales et al. 2004