HP1a: A Structural Chromosomal Protein Regulating Transcription

Histone modification

Assembly of heterochromatin in most eukaryotes appears to depend on core histone deacetylation and histone H3K9 di- and tri-methylation. HP1a facilitates this process by reading the K9 methyl mark and recruiting relevant enzymes to the site. This can include 5mC DNA methylation, common in mammals. The process of heterochromatin formation can be considered a cascade of chromatin modifications; for example, H3K9 must be deacetylated before it can be methylated (the mark read by HP1a) (see [3] for an illustration). In Drosophila, the histone deacetylase and histone methyltransferase can be recovered as a complex [25]. However, there is growing evidence that these steps in histone modification can occur as distinct events. For example, the two S. pombe HP1a orthologues, Swi6 and Chp2, exist in distinct complexes, both of which promote silencing. Chp2 associates with SHREC histone deacetylase, required for H3K14 deacetylation, a shift that limits Pol II access to the template. Swi6 is required for the RNAi-dependent silencing mechanism, which utilizes local transcripts for targeting and recruits the H3K9 HMT Clr4 ([26]; see [27] for review). These dual pathways also occur in Neurospora crassa. Here, H3K9me3 binds HP1 in a complex that directs DNA methylation, resulting in silencing. A distinct HP1 complex works in parallel to cause silencing by maintaining histone hypoacetylation [28].

Why should these changes promote silencing? The chromatin fiber is continually in flux, with relatively unstable nucleosomes (generally marked by the core histone variants H2A.X and H3.3) around the transcription start sites (TSS), where accessibility is detectable as a nuclease hypersensitive site (HS site). Core histone acetylation (particularly in the N-terminal tails of histones H3 and H4) is prominently associated with the active state, presumably because this modification neutralizes electrostatic interaction between the acidic DNA and basic histones, facilitating unfolding of the chromatin fiber. Indeed, comparing the same transgene in heterochromatin to the gene in its normal home in euchromatin in Drosophila has shown a loss of HS site accessibility and a more regularly spaced nucleosome array at the heterochromatic copy. Accessibility is partially restored by HP1a depletion, confirming that the effect is due to heterochromatin packaging (see [3] for original citations). A model based on histone turnover makes sense here; the histone deacetylase Clr3 (required for silencing in S. pombe) can inhibit histone turnover and promote nucleosome occupancy across heterochromatin domains, while Epe1, which promotes the active state, promotes histone exchange [29]. Thus, histone hypoacetylation, by itself, can promote silencing. What about the role of H3K9me2/3? The presence of any methyl groups precludes acetylation at this residue, and binding of HP1a could interfere with demethylation and subsequent acetylation, thus contributing to a stable heterochromatin state. In addition, HP1 can provide a binding platform for proteins that further drive assembly and maintenance of the heterochromatin state, such as H3K9 methyltransferase. However, in some mammalian systems, the HP1 proteins are not required to maintain silencing of endogenous retroviruses (the test applied for heterochromatin formation in this study); the HMT is sufficient, suggesting that H3K9me3 might block transcription by other mechanisms [30].

Chromatin fiber condensation

HP1a binding also suggests mechanisms for chromatin fiber condensation to promote silencing, presumably also by blocking pol II access. Because HP1a forms homodimers through the CSD, leaving the two CDs available to bind H3K9me3, HP1a dimers could crosslink the nucleosome array (Fig. 2A). Crosslinking adjacent nucleosomes could impose uniform nucleosome spacing, which is observed in heterochromatin; crosslinking two turns of a 30nm fiber could preclude the unwinding necessary for transcription. Indeed, sedimentation analysis comparing human HP1α binding to unmethylated or H3K9 trimethylated nucleosome arrays shows significantly higher array self-association with the methylated arrays, an effect requiring the CSD [31]. Recent work in S. pombe has suggested that Swi6 binds more effectively to H3K9me3 by dimerizing through CD-CD contacts [32]. This leaves the two CSDs available to dimerize with two additional Swi6 molecules, creating a tetramer that could bridge to nearby H3K9-methylated nucleosomes (see Fig. 2B). The importance of the CD-CD contacts in tetramer formation on the nucleosome was not anticipated based on earlier studies of metazoan HP1 CDs in solution; CD-CD dimerization was not observed in earlier solution studies or crystal structures [15]. In particular, a recent, detailed biophysical characterization of recombinant mammalian HP1β binding to recombinant mononucleosomes disclosed no other constraints on interaction, or multimerization interfaces, besides the H3K9me3 mark and the CD in creating flexible and dynamic nucleosome binding [33]. These results suggest that CD-CD interaction may be limited to Swi6. The CD-CD association postulated for Swi6 is supported by genetic analysis; missense mutations with increased CD interaction in vitro result in increased tetramerization of Swi6 (which is thought to occur on the nucleosome), and are correlated with increased silencing and heterochromatin spreading [32]. A conformational shift in Swi6 has been proposed to promote heterochromatin spreading [34].

Swi6/HP1 also contributes to higher order chromatin condensation by recruiting other factors, notably cohesin, which is required for mitotic sister chromatid cohesion [35]. While such a direct interaction has not been demonstrated in higher eukaryotes, recent studies in mammalian cells have shown that HP1 interaction with the HMT Suv4-20h2 leads to cohesin recruitment at pericentric heterochromatin, with associated compaction [36].

Transcript degradation

Heterochromatin domains generally show low levels of associated RNA Pol II, but many recent studies show that transcription can occur, and can be essential to targeting heterochromatin formation through the RNAi system (see below). Several reports have documented nucleic acid binding by HP1, most often involving the hinge region, but sometimes either the CD or CSD. A recent study in S. pombe, looking primarily at the MAT loci and telomeres, argues that Swi6 can capture transcripts from heterochromatic reporter genes and direct them to the nuclear exosome for degradation. Extensive genetic alteration of the hinge region that abolished RNA binding allowed persistence of the heterochromatic transcripts [37], but may have altered other Swi6 properties as well. An ability of Swi6 to facilitate silencing at several levels could potentially create a “fail safe” system where silencing is critical. It will be important to see if similar HP1 family protein-dependent transcript degradation is detected in metazoa as well. The HP1-family protein Rhino/HP1d is required for targeted post-transcriptional degradation of certain transposon RNAs in the Drosophila female germline, and is required for nuage organization, suggesting a role in RNA processing, albeit for piRNA production [38].

Targeting of heterochromatin formation

Clearly, heterochromatin formation is a potent silencing mechanism; thus the cell must target domains that should be silenced, while maintaining appropriate access for transcription. Hypotheses as to targeting mechanisms generally turn on either protein recognition of specific DNA sequences or RNA-based recognition (often invoking RNAi-based recognition of transient transcripts) to localize either HP1 or appropriate histone modification enzymes to initiate the heterochromatin formation cascade. An RNAi-based mechanism for heterochromatin targeting is well-established in some fungi (e.g., S. pombe) and plants based on the recognition of transcribed repetitious sequences. In general, the data indicate that Argonaute family proteins carrying small RNAs can target heterochromatin proteins, including H3K9 HMTs and HP1, by sequence recognition of transient transcripts (reviewed in [39]). Similarly, growing evidence implicates a piRNA system in silencing of transposable elements (TEs) and their remnants in most metazoa, particularly in the female germline [40]. For example, depletion of either HP1a or PIWI (an Argonaute protein that can bind piRNAs and enter the nucleus) in the Drosophila female germline leads to increased expression of the same set of TEs [41]. De-repression of TEs upon PIWI depletion correlates with increased Pol II occupancy, whereas expression of piRNAs that target a reporter construct results in increased association of H3K9me3 and HP1a [42]. Analysis of Drosophila ovarian somatic cells has also demonstrated Piwi-dependent transcriptional silencing of transposons, through a mechanism dependent on Maelstrom (an HMG protein) to block RNA Pol II recruitment [43]. In vitro, PIWI binds directly to HP1a through a PxVxL domain [44], but whether or not this occurs in vivo remains an open question, as a PIWI missense mutation that disrupts the in vitro interaction fails to compromise function in vivo [41].

An RNAi-based system seems well-adapted to recognizing TEs, which can change rapidly, are quite variable, and must be silenced to maintain genome integrity. What about regions with satellite DNA? Many mammalian transcription factors (e.g. Pax3 and Pax9) have binding sites in pericentric heterochromatin; indeed, a high concordance has been found between Suv39h-dependent heterochromatic repeat regions and transcription factor binding sites in mouse. Pax3 and Pax9 depletion results in de-repression of major satellite transcripts, impairment of heterochromatic marks, and defects in chromosome segregation [45]. While in this case targeting may be determined by specific protein binding, loss of Pax3/Pax9 results in bidirectional transcription up-regulation, resulting in dsRNA [45], which could once again feed into the piRNA system.

In some cases, binding of specific transcription factors to euchromatic sites appears to trigger formation of small blocks of heterochromatin to achieve silencing. For example, phosphorylated HERS (histone gene-specific epigenetic repressor in late S phase) binds to the histone gene regulatory regions in Drosophila, and anchors HP1a and Su(var)3-9 to induce silencing of this repeated gene cluster [46]. Similarly, the TIF1β/KAP1 co-repressor targets HP1 to specific loci in the euchromatic arms to silence those genes [47,48]. In contrast to the PEV associated with pericentric heterochromatin, there is very little spreading of the silencing marks in these cases, suggesting a silencing state that differs in some key characteristics from heterochromatin formation. However, analysis of the modENCODE data makes clear that as Drosophila cells mature, large “facultative heterochromatin” domains enriched for H3K9me2/3 and HP1a accumulate in the euchromatic arms [49,50]; unfortunately, there has been little study of this process to date.

In summary, the evidence points to an HP1-based silencing mechanism in which HP1 binds histone tails in a way that constrains nucleosome arrays to occlude the chromatin template. The mechanism for HP1 targeting depends on the H3K9 methyl mark and, at least in some cases, on the specificity of the RNAi pathway. The mechanism by which these interactions contribute to the defining cytological appearance of heterochromatin is unknown.

Maintaining heterochromatin structure

As mentioned above, genes normally packaged in heterochromatin appear to require that HP1-rich environment for optimal expression. Drosophila genes normally found in pericentric heterochromatin can be silenced by juxtaposition (through rearrangement) with a euchromatic domain. Protein factors found in heterochromatin are required for optimal expression of these genes in their native state (reviewed in [51]). Similarly, most of the genes on the heterochromatic fourth chromosome lose expression when HP1a is depleted [54]. It has been argued [55] that heterochromatin formation is required for piRNA production, based on the loss of expression upon depletion of dSETDB1 from a piRNA cluster that maps to a heterochromatic domain. These results suggest that HP1a is required to maintain a necessary feature of the chromatin structure at and around active genes that are surrounded by heterochromatin. One clue comes from direct studies of two fourth chromosome genes, Dyrk3 and Caps, which show a loss of HS sites (indicating loss of promoter accessibility) on HP1a depletion [56]. But how this might be accomplished remains unknown.

Promoting transcript elongation and processing

In addition to its association with heterochromatin, Drosophila HP1a is also associated with a small set of sites in the euchromatic arms of the major autosomes, including conspicuous chromosome puffs, sites with high transcription levels [57]. Chromosome-wide binding studies using cultured cells find HP1a enriched across the entire transcription units of certain exon-rich active genes in euchromatin [58]. HP1a stimulates transcript elongation at a small set of euchromatic sites, and is associated with heterogeneous nuclear ribonuclear proteins (hnRNPs) [59], suggesting a role for HP1a in transcript processing in flies. The mammalian HP1-family protein CBX3/HP1γ is required for the full activity of some euchromatic genes, as evidenced by reduced expression upon CBX3/HP1γ knockdown. The knockdown leads to defective recruitment of splicing factors, with accumulation of unspliced transcripts for these genes [60]. These results argue for a positive role for HP1a in transcript elongation/processing for some euchromatic genes, the opposite of the result reported above from S. pombe looking at heterochromatic transcripts [61].

Taken together, the context-dependent effects of HP1 family proteins on gene expression may be understood mechanistically by thinking of these proteins and their associated complexes as chromatin organizers. In some cases, the resulting chromatin occludes promoters and/or regulatory DNA, while in other cases, the resulting chromatin facilitates access to promoters and/or regulatory DNA for activators. By extension, Pol II elongation could be either impeded or enhanced by HP1-dependent architecture, and RNA processing efficiency directly or indirectly affected [62].