We show here that, in addition to the contact of the CD with methylated H3K9, HP1 proteins use the CSD to associate with histone H3 at the level of the entry/exit point of the nucleosome barrel. This region on histone H3 that we termed Shadock is also contacted by Brg1 and Brm, the catalytic subunits of the hSWI/SNF complex, and chromatin remodeling can favor loading of HP1 proteins onto chromatin. We provide also evidence suggesting that this loading ultimately decreases the efficiency of hSWI/SNF remodeling.
FRAP experiments have shown that the CSD domain is required for the binding of HP1 proteins to native chromatin in vivo
, with actually a stronger contribution than the CD when observation are made in euchromatic regions 
. In addition, deletion of the CSD abolishes HP1-mediated transcriptional repression in transfection assays 
. In the same type of experiment, we show that a point mutation in the CSD disrupting the interaction with the H3 Shadock region (I126F) is sufficient to interfere with the repressing activity. Furthermore, our REA assays show that the CSD is required for inhibition of chromatin remodeling by hSWI/SNF and that this remodeling complex contacts the same region on H3 as the HP1 proteins. Taken together, these observations strongly suggest that HP1 chromatin binding and repression activities are largely mediated by the histone binding activity of the CSD.
ChIP and ChIP reChIP assays show that Brm is required for the loading of HP1γ to the IFIT3 promoter during transcriptional activation. We note that on this promoter, we have another example of the switch from HP1α/HP1ß to HP1γ during transcriptional activation, also observed on the HIV1 LTR and on the Survivin promoter 
. This switch could suggest that, in vivo
, histone H3K9 methylation is the determining factor for the recruitment of HP1α and HP1ß, while recruitment of HP1γ would be more dependent on chromatin opening by hSWI/SNF or the RNAPII. The role of HP1 proteins on active promoters is still enigmatic. The transient recruitment of Brg1 to the HIV1 promoter in sync with HP1γ recruitment shown in suggests that HP1 proteins could be involved in limiting remodeling on activated promoters. However, HP1 proteins may also have a role further downstream in connection with mRNA maturation 
The presence of HP1α on the IFIT1 and IFIT3 promoters and thereby the direct regulation of these genes by the HP1 protein was established by ChIP experiments (data not shown). However, the ChIP approach did not allow us to estimate the impact of hSWI/SNF remodeling on the recruitment of HP1α, as depletion of Brm leads to repression of the IFIT1 and IFIT3 promoters. Under these conditions, it was not possible to part between enhanced recruitment due to increased histone H3K9 methylation and decreased recruitment associated with reduced hSWI/SNF activity. However, our in vitro nucleosome-array binding assay suggests that HP1 proteins other than HP1γ can benefit from chromatin remodeling to bind nucleosomes. It must here be noted that this assay does not allow us to discriminate between histone and DNA binding. Nevertheless, repression of hSWI/SNF remodeling does not appear to rely on DNA binding as an HP1α mutant no longer binding DNA still represses. In addition, HP1γ does not bind to DNA in our hands and prefers nucleosomes (data not shown). Yet, it is efficient in repressing hSWI/SNF remodeling. Finally, we note that methylation of histone H3 on K9 is not required for repression of hSWI/SNF remodeling in vitro, further suggesting that the contact of the CSD with the Shadock, and not other contacts, is important for the repressing activity of HP1 proteins.
The interaction of HP1α CSD with histone H3 required HP1 dimerization and was disrupted by the mutation of V46 in the PXVXL-like sequence in the Shadock. It is therefore possible that this interaction can be structurally compared to that of other molecular partners of CSDs such as CAF1 or TIF1 proteins 
. Interestingly, the rH3(35–66)V46A and rH3(44–66) mutants bind HP1γ but not HP1α, while H3 constructs with a full Shadock region bind both HP1 proteins. These observations show that HP1α and HP1γ have neighboring but distinct sites of interaction on the histone. While this manuscript was in revision, it was shown that phosphorylation of histone H3 on tyrosine 41 by JAK2 compromises the binding of the chromoshadow-domain of HP1α to an H3 peptide spanning from aa 31 to 56 
. This modification is outside the HP1γ binding site and it therefore seems possible that HP1α and HP1γ are differentially regulated by post-translational modifications in the H3 Shadock region.
The region of histone H3 contacted by the HP1 proteins also associates with the hSWI/SNF subunits Brg1 and Brm. This is compatible with earlier studies in yeast showing that residues K56 and L61 are involved in SWI/SNF recruitment 
. In yeast, it was suggested that the H3 αN helix is targeted and remodeled by the SWI/SNF complex 
. More recently, mutagenesis within this region was shown to affect the efficiency of yeast SWI/SNF remodeling in vitro 
. Our competition experiments with polypeptides mimicking the H3 αN helix further suggest that interaction of Brg1 with this region is essential for the chromatin remodeling activity of the hSWI/SNF complex. Consistent with this, the sequences downstream of the helicase domain of Brg1/Brm that mediate the interaction with the H3 Shadock domain have earlier been reported as essential for efficient remodeling in vitro 
. We therefore speculate that the competing binding of Brg1/Brm and HP1 proteins to a same region of histone H3 during unwanted remodeling can at least in part explain the inhibiting effect of HP1 on chromatin opening by hSWI/SNF.
binding assays suggest that Brg1 has an affinity for the globular domain of H3 that is higher than that of HP1α (Figure S4
). In a mechanism based on competition of Brg1 and HP1 proteins for binding to overlapping sites on histone H3, this difference in affinity may explain that a 7-fold excess of HP1α over Brg1 is required to obtain 50% inhibition of hSWI/SNF remodeling in the REA assays. It must finally be noted that the activity of the hSWI/SNF complex is also inhibited by the Polycomb Group (PcG) class II complex in vitro 
. This complex, involved in gene silencing, includes a CD protein that, like HP1 proteins, binds methylated histone H3 tails with a preference for methylation on K27 rather than K9 
. However, Polycomb proteins contain no CSDs and it seems that PRC1 relies essentially on a structuring effect on the nucleosomal template, repressing remodeling by creating more condensed chromatin 
hSWI/SNF and HP1 proteins have many common target promoters including E2F1, human thymidine kinase, c-Myc, Sox2, Cyclin E, and the MMTV and HIV1 LTRs 
. Likewise, we show here that several interferon-inducible genes that require hSWI/SNF for their activation are under the negative control of HP1α and HP1β. We note however that not all hSWI/SNF target genes we tested were affected by knock-down of HP1 proteins, including for example DraL and SPARC. These genes were both expressed at relatively high levels in the cells we used and may therefore not have any repressive structure on their promoter (data not shown). All considered, we suggest that on promoters where SWI/SNF functions as a repressor, the opening of the chromatin may, as previously suggested, be the event initiating HP1 stable recruitment 
. Where SWI/SNF functions as an activator, HP1 proteins recruited by methylated histone H3 tails may instead function as sensors of unwanted SWI/SNF activity, binding nucleosome domains uncovered by the remodeling, and thereby block the reaction (see proposed model in ). Finally, after the activation, additional HP1 proteins may be involved in controlling excessive remodeling activity. In that sense, HP1 proteins could very generally function as gatekeepers using the exposure of domains internal to the nucleosome to detect and restrict chromatin opening.
Model suggesting how HP1 proteins may gain access to internal nucleosomal regions and control remodeling by SWI/SNF.