To investigate the role of histone methylation in APL, we analyzed whether trimethylation of histone H3 on lysine 9 was present at PML-RAR target promoters. We analyzed the RARβ2 promoter, which is considered a “canonical” PML-RAR target gene (5
). This promoter contains a retinoic acid-responsive element in proximity of the transcription start site (5
). For a model system, we used U937-PML-RAR cells (hematopoietic precursors carrying the PML-RAR cDNA under a zinc-inducible promoter); we have previously shown recruitment of other chromatin-modifying enzymes (HDACs and DNMTs) to the RARβ2 promoter by PML-RAR in these cells (5
). Chromatin immunoprecipitation analysis of U937-PML-RAR cells revealed a robust, PML-RAR-dependent trimethylation of lysine 9 of histone H3 of the promoter (Fig. ). This modification was accompanied by a strong transcriptional repression of the gene by the fusion protein (Fig. ). Histone methylation was observed at another PML-RAR target gene directly repressed by the fusion protein (NFE2 [Fig. ]). In mammals, H3-K9 trimethylation is mainly-if not exclusively-due to the action of SUV39H1 (22
). Therefore, we investigated whether PML-RAR associated with SUV39H1. Coimmunoprecipitation experiments showed that anti-PML-RAR antibodies are able to precipitate endogenous SUV39H1 only upon expression of the fusion protein, demonstrating the existence of a PML-RAR/SUV39H1 complex in vivo (Fig. ).
FIG. 1. PML-RAR recruits the histone methyltransferase SUV39H1 and induces histone methylation and transcriptional silencing of its target genes. (A) PML-RAR induces H3-K9 trimethylation of the RARβ2 and NFE2 promoters. U937-PML-RAR cells (P/R) and control (more ...)
We next checked by ChIP whether SUV39H1 was recruited to PML-RAR target genes. Since the available antibodies against SUV39H1 are unable to work efficiently in ChIP studies, we adopted a transient-transfection approach. Figure shows that transiently transfected, epitope-tagged SUV39H1 was recruited to the endogenous RARβ2 promoter only when PML-RAR was coexpressed.
Histone methylation assays confirm that, in the context of the PML-RAR/SUV39H1 complex, SUV39H1 is still able to methylate histone H3 lysine 9. Immunoprecipitation experiments were carried out in 293T cells transfected with epitope-tagged SUV39H1 and/or PML-RAR. Anti-PML-RAR antibodies specifically immunoprecipitated an endogenous histone H3 methyltransferase activity, which was more evident when SUV39H1 was overexpressed (Fig. ). Identical results were obtained in U937-PML-RAR cells (data not shown). Taken together, these results indicate that PML-RAR can recruit an active SUV39H1 on the RARβ2 promoter and induce trimethylation of H3-K9 that correlates with transcriptional silencing.
To determine whether SUV39H1 is actively involved in PML-RAR-mediated transcriptional repression of the RARβ2 promoter, we transiently transfected 293T cells with a RARβ2 promoter-based reporter construct. Cotransfection of PML-RAR resulted in repression of reporter activity (Fig. ). SUV39H1 slightly, but reproducibly, increased the transcriptional repression mediated by the fusion protein and did not affect reporter activity in the absence of PML-RAR. In contrast, increasing amounts of H324K, a construct carrying an inactivating point mutation within the HMT domain of SUV39H1 or of SUV39H1ΔSET, a deletion construct that lacks the catalytic domain, were able to completely abrogate the transcriptional repression mediated by PML-RAR (Fig. and data not shown). H324K and the mutant lacking the HMT domain were still able to associate with PML-RAR in immunoprecipitation studies, suggesting that they might act as dominant-negative mutants, competing with endogenous HMTs (data not shown).
FIG. 2. Differential requirement for SUV39H1 in the transcriptional repression of PML-RAR target genes. (A) Transient-transfection analysis. 293T cells were transfected (+) with a luciferase-based reporter containing the RARβ2 promoter and expression (more ...)
We next checked whether SUV39H1 was involved in transcriptional repression of endogenous PML-RAR target genes. U937-PML-RAR cells were transduced with retroviral vectors coding for SUV39H1 or H324K, and PML-RAR expression was induced by zinc treatment. We analyzed by quantitative PCR the expression levels of the endogenous RARβ gene and of several additional PML-RAR target genes. These genes were previously identified in a systematic screening for PML-RAR targets and represent direct transcriptional targets of the fusion protein (1
; also data not shown). For the RARβ gene, we confirmed the results obtained in transfection studies: repression of the endogenous gene was further increased by SUV39H1 overexpression, and the H324K mutant relieved transcriptional repression by the fusion protein (Fig. ). Interestingly, the other genes tested showed distinct patterns of transcriptional modulation (Fig. ). PSCD4 and NFE2 behaved similarly to RARβ, since SUV39H1 overexpression increased transcriptional repression by the fusion protein and H324K relieved repression (Fig. ). The fusion protein repressed MYO1F, but SUV39H1 overexpression (both of the wild type and of the mutant) did not show significant effects (Fig. ). Taken together, these results suggest differential requirements for SUV39H1 in transcriptional repression of PML-RAR target genes.
We investigated the structural determinants of the association of PML-RAR with SUV39H1 in transient-transfection assays, followed by coimmunoprecipitation studies. Mammalian 293T cells were cotransfected with expression vectors for epitope-tagged SUV39H1 and for PML-RAR (or variants). PML-RAR specifically interacted with SUV39H1, as observed by coimmunoprecipitation studies using either anti-SUV39H1 (tag), or anti-PML-RAR antibodies (Fig. ). Wild-type RAR was not able to associate stably with SUV39H1 (Fig. ). In contrast, wild-type PML associated with SUV39H1 (Fig. ). We mapped this association to the carboxy-terminal region of PML; in fact, using amino- and carboxy-terminal deletions of PML, we showed that only the carboxy-terminal region is required for the association with SUV39H1 (Fig. ). Interestingly, the carboxy-terminal region of PML is absent in PML-RAR. The fusion protein retains the N-terminal region (which contains the tripartite motif, including the coiled-coil oligomerization domain [24
]) and maintains the ability to associate with wild-type PML (which in leukemic cells is produced from the remaining wild-type PML allele not involved in the chromosomal translocation) (11
). To verify whether the association of PML-RAR with SUV39H1 was indirect and mediated by wild-type PML bridging the two proteins, we performed in vitro binding studies and immunoprecipitation experiments in pml−/−
cells. Bacterially expressed PML-RAR failed to interact with in vitro-translated SUV39H1, suggesting an indirect association of the two proteins; wild-type PML, however, was not able to allow the formation of a ternary complex (data not shown). In support of PML-independent mechanisms, coimmunoprecipitation studies performed in pml−/−
mouse embryonic fibroblasts confirmed the association of PML-RAR with SUV39H1 in the absence of PML (Fig. ).
FIG. 3. Studies of association of PML-RAR and RAR with SUV39H1. 293T cells were cotransfected (+) with expression vectors for PML-RAR, wild-type RAR, and Myc-SUV39H1 as indicated above the blots. Whole-cell extracts were precipitated with the indicated (more ...)
FIG. 4. The association of PML-RAR with SUV39H1 requires oligomerization of RAR. 293T cells were cotransfected (+) with the expression vectors indicated above the blots. Whole-cell extracts were immunoprecipitated (IP) with the appropriate antibodies (more ...)
We have previously demonstrated that the oncogenic potential of PML-RAR depends on its oligomerization ability, which is mediated through the coiled-coil domain of the PML moiety and which enhances the ability of monomeric, wild-type RAR to interact with transcriptional coregulators, such as nuclear receptor corepressor (NCoR)-silencing mediator of retinoic and thyroid hormone (SMRT) or DNMTs (5
). We hypothesized that (as for HDACs and DNMTs) the association of PML-RAR with SUV39H1 might depend on the stabilization of a weak RAR/SUV39H1 association, through an oligomerization-dependent mechanism. We tested two PML-RAR deletion constructs: ΔCC-PML-RAR lacks the PML coiled-coil domain (and it is therefore unable to oligomerize), while CC-RAR retains only the coiled-coil domain of PML (and it is able to form oligomers, to associate with NCoR-SMRT and DNMTs, and to block differentiation) (18
). As shown in Fig. , SUV39H1 is able to associate with CC-RAR, but not with ΔCC-PML-RAR, indicating that oligomerization is required for SUV39H1 recruitment. To confirm the relevance of oligomerization (and to exclude the possibility that the coiled-coil region of PML was acting as an interacting surface for SUV39H1), we analyzed another construct, which contains the p53 tetramerization domain fused to RAR (p53-RAR). This chimeric protein has been previously shown to behave as PML-RAR in its capacity to recruit NCoR-DNMTs and to block hematopoietic differentiation (18
). SUV39H1 is indeed able to associate with p53-RAR (Fig. ). Taken together, these results show that PML-RAR depends on its ability to form oligomeric complexes to associate with SUV39H1 and that this interaction occurs indirectly. The significance of the association of wild-type PML with SUV39H1 remains to be investigated in other studies.
To analyze the biological relevance of the association of SUV39H1 with PML-RAR, U937-PML-RAR cells were infected with an empty retroviral vector (as a control) or retroviral vectors expressing wild-type SUV39H1 or H324K. Transduced cells were induced to express PML-RAR and then stimulated to differentiate with vitamin D and TGF-β (9
). We chose to use different concentrations of zinc sulfate to achieve variable expression levels of PML-RAR (Fig. ). With lower zinc concentrations, we obtained PML-RAR expression levels comparable to those observed in the promyelocytic cell line NB4, a cell line derived from an APL patient, and considered a “marker” for PML-RAR levels observed in the disease state; at higher zinc concentrations, the levels are in excess over those observed in NB4 cells (Fig. ). After 2 days of vitamin D and TGF-β treatment, myeloid differentiation was scored by fluorescence-activated cell sorting (FACS) analysis of surface differentiation markers (CD14 and CD11b). At lower concentrations of PML-RAR, the fusion protein inhibited differentiation weakly, and SUV39H1 cooperated dramatically in blocking differentiation (Fig. , left panel). SUV39H1 alone did not show any effects on differentiation (Fig. , left and right panels). The enzymatic activity of SUV39H1 was required for the enhancement of the differentiation block by PML-RAR, as the catalytically inactive SUV39H1 was almost completely ineffective (Fig. , left panel). As shown in Fig. , high doses of PML-RAR overcame the requirement for SUV39H1 overexpression and caused a strong decrease in the expression of the differentiation marker CD14, which was slightly increased by the expression of SUV39H1 (Fig. , left panel). In contrast, CD11b expression was not affected by PML-RAR expression, but coexpression with SUV39H1 led to a strong reduction in CD11b mRNA levels (Fig. , right panel). Taken together, our data suggest that SUV39H1 cooperates with PML-RAR to mediate an efficient block of myeloid differentiation and that this cooperation requires the enzymatic activity of SUV39H1. Surprisingly, the catalytically inactive H324K mutant did not counteract the biological effects observed at high levels of expression of PML-RAR (Fig. , left panel). We looked therefore at the expression of PML-RAR targets upon treatment with vitamin D and TGF-β. All of the analyzed genes (NFE2, PSCD4, and MYO1F) were induced by vitamin D/TGF-β treatment at a comparable degree (three- to fourfold induction [data not shown]). Interestingly, high levels of PML-RAR expression led to repression of these genes, and SUV39H1 overexpression led to a variable, further increase in transcriptional repression (Fig. ). Overexpression of H324K had no effect on PML-RAR-mediated repression of NFE2 and PSCD4 and caused a slight enhancement of repression in the case of MYO1F (Fig. ). These results are in contrast to those obtained analyzing the levels of expression of the same genes in the absence of vitamin D/TGF-β treatment, where H324K was very effective in relieving transcriptional repression by PML-RAR (Fig. ). Intriguingly, they might be ascribed to distinct chromatin modifications triggered by PML-RAR under conditions where transcription of its target genes would be induced by differentiating stimuli (such as vitamin D and TGF-β) as opposed to basal levels of expression. This hypothesis is currently being tested in our laboratory. We therefore conclude that the lack of a convincing rescue of the differentiation block by the dominant-negative SUV39H1 mutant is due to the lack of a dominant-negative effect on transcriptional repression of different PML-RAR targets whose transcription has been induced by differentiating stimuli.
FIG. 5. SUV39H1 cooperates with PML-RAR in the differentiation block of hematopoietic precursor cell lines. (A) (Left) U937-PML-RAR cells were transduced with retroviral vectors for SUV39H1, H324K, or the empty vector as a control. Cells were induced to produce (more ...)
Chromatin modifications may occur in a hierarchical fashion, and specific posttranslational changes may depend on previously occurring modifications at the same or other sites. Deacetylation of K9 of histone H3 is required for its subsequent methylation in the same lysine residue (23
). In the absence of the fusion protein, H3-K9 histone acetylation levels at the RARβ2 promoter and at other tested PML-RAR targets were already low in U937-PML-RAR cells, making it impossible to verify whether PML-RAR-mediated recruitment of HDACs played a direct role in subsequent events (data not shown). Kinetics of DNA and histone methylation (which occur at relatively later time points [>8 to 12 h]) by ChIP did not convincingly demonstrate a hierarchical relationship between these two events (data not shown). It was previously shown that demethylating agents (such as decitabine) did not lead to a complete rescue of the block in vitamin D/TGF-β-mediated differentiation caused by PML-RAR (5
). We performed similar experiments in U937-PML-RAR cells overexpressing SUV39H1 (Fig. ). As previously observed, decitabine treatment significantly, but not completely, rescued differentiation in the presence of PML-RAR (Fig. ). Interestingly, SUV39H1 cooperated with PML-RAR in blocking differentiation also in decitabine-treated cells (Fig. ). Conversely, decitabine continued to partially release the differentiation block in SUV39H1-overexpressing cells (Fig. ). H324K was completely inactive in these assays, showing that the effect of SUV39H1 requires its catalytic activity (data not shown). Taken together, these results suggest that histone and DNA methylation can occur independently in PML-RAR-expressing cells and that they both contribute to transcriptional differentiation and differentiation block.
FIG. 6. PML-RAR requires histone and DNA methylation for efficient block of differentiation of hematopoietic precursors. (A) U937-PML-RAR cells were transduced with the specified vectors. Transduced cells were treated with decitabine (+) for 48 h, and (more ...)
Considering the known biochemical and biological properties of PML-RAR and our present findings, we propose that the main action of the fusion protein is to impose an heterochromatin-like structure on its target genes, making these genes (normally involved in hematopoietic differentiation) unable to respond to differentiating stimuli (Fig. ). The strategy exploited by PML-RAR to impair differentiation of its target cell is therefore to “convert” genomic loci which are physiologically set to respond to differentiating stimuli into nonresponsive, heterochromatin-like areas which block the transcriptional response to these stimuli. Potentially, the same strategy might function in other transformed cells, due to the action of other leukemia-associated fusion proteins acting through similar mechanisms, or to the effect of other transforming oncogenes in solid tumors. Recently, histone methylation by the H3K79 hDOT1L histone methyltransferase has been involved in leukemogenesis (20
). The pharmacological modulation of the changes in chromatin structure of cancer cells may therefore represent an extremely powerful strategy for reversion of the tumor phenotype.