Our results demonstrate that efficient H3K27me3 production by PRC2 requires its association with PHF1. The interaction between PHF1 and the PRC2 complex we observed in vitro also occurs in vivo at the promoters of the Ezh2 target genes tested. In similarity to the results seen with Ezh2 and its associated proteins, PHF1 is absent from the coding region of these genes. It is interesting that the regions occupied by PHF1 and Ezh2, and therefore by H3K27me3, across the MYT1 locus correspond to highly conserved regions in the genome (as shown earlier; see reference 44
). PHF1 functions as a transcriptional repressor, and its targeting to an artificial promoter results in elevated levels of the Ezh2 complex components and of H3K27me3 at the promoter. Yet the presence of Ezh2 and the presence of PHF1 at their common target gene promoters occur independently of each other. The loss of PHF1 at the HOXA6, HOXA9, and HOXA11 promoters results in a decrease in the level of H3K27me3, an increase in the level of H3K27me2, and a deregulation of these genes, even though the levels of Ezh2 are unchanged. Further analysis also showed that this decrease in the level of H3K27me3 is a global phenomenon.
To our knowledge, this is the first time that PHF1 has been defined as a factor affecting Ezh2 activity. On a gel filtration column, PHF1 and PRC2 overlapped but did not perfectly coelute as would be expected of components of a stable complex. Therefore, unlike the core components Eed and SirT1 that affect PRC2/PRC3/PRC4 activity through changes in Ezh2 substrate specificity, PHF1 is not an integral component of these complexes but is rather an associated protein that does not alter Ezh2 substrate specificity but instead stimulates its trimethyltransferase activity.
Upon knockdown of PHF1, there is a loss of H3K27me3 and a concomitant increase in H3K27me2 levels at some of the Hox loci. These same genes were deregulated despite the increased presence of H3K27me2. Ezh2-KD also deregulated these genes to the same extent as PHF1-KD, and there was no additional derepression observed in a double knockdown. Thus, both PHF1 and Ezh2 are required to maintain the repressed state of these Hox genes, as loss of either protein causes complete derepression. Our studies are in agreement with those of Nekrasov et al., who reported that the Drosophila
Pcl protein is required for H3K27me3 at polycomb target genes (28a
). Their studies examining the Ubx and Abd-B loci showed that loss of Pcl results in a decrease in H3K27me3 levels and a concomitant increase in H3K27me1 and H3K27me2 levels. Another recent study by Cao et al. suggests that mPHF1 is required for global H3K27me2 and H3K27me3 formation (6a
). While we clearly detected a decrease in global H3K27me3, we did not observe any changes in global H3K27me2 levels but instead observed an increase in H3K27me2 levels at polycomb target genes that were depleted of PHF1.
Our results underscore the role of H3K27me3 in repression while raising questions regarding the role of H3K27me2. The results with respect to the abundance of H3K27me2 (from 40 to 60% of total histone H3, depending on the studies and on the organism) do not support the idea of its being a functional repressive mark. In the case of Drosophila, a large fraction of the genome bears the H3K27me2 mark without E(z) association. While the genomic distribution of H3K27me2 is not yet clear in the mammalian case, we suggest that H3K27me2 is an intermediate H3K27 methylation state that marks genes as being potentially repressible and that the concerted action of Ezh2 (in the context of its associated proteins) and PHF1 is required to achieve complete repression (Fig. ). While the association of PRC2 with its target genes may be required for H3K27me3 catalysis, the presence of PHF1 at these loci allows PRC2 to efficiently catalyze H3K27me3.
More insights into the roles of H3K27me2 and H3K27me3 will likely come about upon the identification of the downstream effectors and their specificities. Of particular interest is the chromodomain protein, Polycomb (Pc), which has been shown to bind both H3K27me2 and H3K27me3 marks without apparent discrimination in vitro, although our studies suggest that Pc binds preferentially to H3K27me3 in vivo, as measured indirectly by the presence of Bmi1. Another interesting possibility is that in the absence of PHF1, Ezh2 is able to catalyze H3K27me3 only inefficiently; perhaps the establishment of these low levels of H3K27me3 stabilizes PHF1 at the site (through the TUDOR or PHD domains as discussed below), with resultant stimulation of Ezh2 activity (Fig. ).
In vitro reconstitution experiments have indicated that the addition of PHF1 stimulates the activity of PRC2 to increase twofold. While the major product formed in an in vitro reaction with PRC2 is H3K27me1, minor amounts of H3K27me2 and H3K27me3 were also formed. We observed that addition of PHF1 resulted in a specific increase in H3K27me3 levels without changes in H3K27me1 or H3K27me2 levels. Interestingly, PHF1 contains a TUDOR domain as well as two PHD fingers. The TUDOR domain of 53BP1 was shown to bind histone H3K79me2 as well as H4K20me2 (2
). Similarly, the PHD finger of ING2 was shown to bind H3K4me3 (41
). The presence of both the TUDOR and PHD fingers raises the possibility that PHF1 could bind to methylated H3K27 and promote Ezh2-dependent trimethylation at this residue. Whether PHF1 exhibits preferential binding to methylated histone residues and whether PHF1 can discriminate between the levels of methylation remains to be seen. The PHD fingers of some proteins were also shown to contain ubiquitin E3 ligase activity (9
). Our biochemical studies failed to demonstrate PHF1-mediated ubiquitination of histone proteins, but whether the PHD fingers of PHF1 are able to function in this manner in the presence of additional factors remains an open question.
Of note, overexpression of Ezh2 as well as other PRC components has been correlated with cellular transformation. In fact, one of the Eed isoforms (Eed2) is present specifically in cells that are either undifferentiated or tumorigenic and Eed1 regulates a specific set of genes compared to other Eed isoforms (18
). The expression level of hPCL3, the human homologue of Polycomblike, is also increased in several types of cancers (47
). More recently, the PHF1 gene was reported to be rearranged in three cases of endometrial stromal sarcoma, resulting in the production of two types of chimeric proteins (JAZF1/PHF1 and EPC1/PHF1) that retained the complete open reading frame of PHF1 but under the control of the JAZF1 (Juxtaposed with another zinc finger 1) and the EPC1 (Enhancer of polycomb 1) promoters, respectively (26
). This suggests that in addition to the production of an abnormal fusion protein, another cause of pathogenicity might be the altered regulation of the PHF1 gene. Given that PHF1 interacts with the PRC2 complex in vitro and modulates the ability of Ezh2 to catalyze the repressive H3K27me3 mark in vitro and in vivo, PHF1 becomes another likely candidate for investigation with regard to cancer progression.