|Home | About | Journals | Submit | Contact Us | Français|
Activation of Janus Kinase 2 (JAK2) by chromosomal translocations or point mutations is a frequent event in haematological malignancies1-5. JAK2 is a non-receptor tyrosine kinase that regulates a number of cellular processes by inducing cytoplasmic signalling cascades. Here we show that JAK2 is present in the nucleus of haematopoietic cells and directly phosphorylates Y41 on histone H3. Heterochromatin protein 1 alpha (HP1α), but not HP1β, specifically binds to this region of H3 and phosphorylation of H3Y41 by JAK2 prevents this binding. Inhibition of JAK2 activity in leukaemic cells reduces both the expression of the haematopoietic oncogene lmo2, and the phosphorylation of H3Y41 at its promoter, whilst simultaneously increasing the binding of HP1α at the same site. Together, these results identify a previously unrecognised nuclear role for JAK2 in the phosphorylation of H3Y41 and reveal a direct mechanistic link between two genes, jak2 and lmo2, involved in normal haematopoiesis and leukaemia1-8
The DNA of most nucleated eukaryotic cells is packaged within chromatin9. The core histones within nucleosomes are subject to numerous post-translational modifications including phosphorylation9. Only kinases responsible for serine and threonine phosphorylation of non-variant histones have been reported9, and thus far tyrosine phosphorylation of non-variant histones remains uncharacterised.
Janus kinase 2 (JAK2) is a non-receptor tyrosine kinase that is critical for haematopoiesis, adipogenesis, immune and mammary development10. Recently, constitutive activation of JAK2 has been noted as a sentinel event in several different haematological malignancies3-5,11-14. The most prevalent gain of function mutation in JAK2 results from a missense mutation in its JH2 autoregulation domain (JAK2 V617F)11-14. JAK2 signalling is implicated in diverse biological processes such as cell cycle progression, apoptosis, mitotic recombination, genetic instability and alteration of heterochromatin15-19. Mechanistic insights into these potential oncogenic events are elusive but the presiding opinion is that the biological/oncogenic effects of JAK2 are mediated by cytoplasmic signalling pathways 1,2.
We considered the possibility that the V617F mutation may perturb the subcellular localisation of JAK2. Figure 1A and supplementary figure 1 show that JAK2 is present within the nucleus of three cell lines (HEL, UKE1 and SET2) carrying the JAK2 V617F mutation20. However, K562 cells, which express wild-type JAK2, also contain the enzyme in the nucleus. Nuclear JAK2 levels are higher in HEL and SET2 as these lines contain multiple copies of JAK2 (figure 1A and supplementary figure 1)20. Nuclear JAK2 was also observed in primary cells positive for the CD34 stem cell antigen obtained from a patient with JAK2 V617F-positive post-polycythaemic myelofibrosis. Transfection of JAK2 into a JAK2 null background, γ-2A cells,21 independently confirms the nuclear localisation of JAK2 and serves to validate the specificity of the antibodies used in IF (figure 1B and supplementary figure 2). Finally, subcellular fractionation experiments on HEL cells also demonstrate that JAK2 is indeed in the nucleus (figure 1C). Taken together these results demonstrate that a significant fraction of JAK2 is present in the nucleus of haematopoietic cells irrespective of JAK2 mutation status.
To explore the role of JAK2 within the nucleus we investigated the possibility that histones could be a substrate. Figure 2A and supplementary figure 3 show that recombinant JAK2 (rJAK2) can specifically phosphorylate histone H3 and that this activity is inhibited by the JAK2 inhibitor TG10120922.
Histone H3 contains three tyrosine residues that are highly conserved. One of these, H3Y41, is positioned at the N-terminus of the first helix of H3 (the αN1-helix) where the DNA enters the nucleosome, and is juxtaposed to the major groove of the DNA double helix (figure 2B). Given its position within the nucleosome, we considered that this residue might be the target of JAK2 kinase activity. We therefore raised an antibody to phosphorylated H3Y41 (H3Y41ph) and verified its specificity (supplementary figures 4-6). Using this antibody we show that rJAK2 phosphorylates H3Y41 when core histones, purified histone H3 or recombinant H3 are used as substrates (figure 2C, lanes 1-6). This phosphorylation was markedly reduced by the JAK2 inhibitor TG101209 (figure 2C, lanes 9&10). Mutation of H3Y41 to phenylalanine demonstrates that this tyrosine is a target of JAK2 in vitro (figure 2C, lanes 7&8), and it confirms the specificity of the antibody. To ensure that cellular JAK2 can also phosphorylate H3Y41 we immunoprecipitated JAK2 from HEL cells and used it in phosphorylation experiments. Supplementary figure 7 shows that endogenous JAK2 phosphorylates H3Y41 and that the TG101209 inhibitor blocked this activity.
To assess if tyrosine phosphorylation of H3Y41 is present in vivo, chromatin preparations from six cell lines were probed with the H3Y41ph antibody. Notably, H3Y41 phosphorylation was more abundant in the cell lines that contain active JAK2 signalling (SET2, HEL, UKE1 and K562)20,23. In contrast, H3Y41ph was significantly reduced in HL60 cells and γ2A cells, which both lack detectable JAK223,21 (figure 2D). Cytokine stimulation of K562 cells with leukaemia inhibitory factor (LIF) activates JAK2, as evidenced by JAK2 phosphorylation, and leads to a concomitant increase in H3Y41ph suggesting a role for JAK2 in this pathway (Figure 2E). Similar results were noted with PDGF-BB in K562 cells. A direct role for JAK2 in this pathway is demonstrated by the ability of the JAK2 inhibitor TG101209 to block the PDGF-BB mediated increase in H3Y41ph (supplementary figure 8). Finally, we demonstrate that stimulation of murine BaF3 cells with IL3 (a cytokine that exclusively signals via JAK2 in these cells) also leads to an increase in H3Y41ph (supplementary figure 8). Taken together these data suggest that activation of JAK2 is upstream of H3Y41ph.
Whilst the presence of H3Y41ph in both HL60 and γ2A cells indicates that JAK2 is not the only tyrosine kinase responsible for this post-translational modification, transfection of JAK2 into JAK2-null (γ2A) cells demonstrates that this enzyme is one of the cellular kinases responsible for this post-translational modification on histone H3 in vivo (figure 2F). Finally, to provide further evidence that JAK2 phosphorylates H3Y41 in vivo, we used two specific, chemically distinct JAK2 inhibitors, TG101209 and AT9283, that have been extensively characterised22,24. Analysis of chromatin preparations from HEL cells grown in the presence or absence of the JAK2 inhibitors demonstrated that H3Y41ph was markedly reduced following four hours of exposure to either of the specific JAK2 inhibitors and that these changes were not a consequence of broad effects on cell cycle or apoptosis (supplementary figure 9A&B). Moreover, JAK2 inhibition leads to a rapid and sustained loss of H3Y41ph. The reduction in H3Y41ph is observed within 15 minutes and by one-hour an 80% decrease is observed (figure 2G-H). The rapidity of this in vivo response when coupled to the in vitro data strongly suggests that JAK2 directly phosphorylates H3Y41 in vivo.
JAK2 has recently been implicated in the DNA damage response15,19 and the alteration of heterochromatin16. Analysis of histones from cells that were subject to ionising radiation indicates that H3Y41ph is not responsive to DNA double strand breaks (supplementary figure 6). Given the connection between HP1 and the JAK pathway in Drosophila,16 we next investigated whether increased JAK2 activity in haematopoietic cells affects heterochromatin by regulating the binding of heterochromatin protein 1 (HP1). When we compared the binding of HP1α and β to chromatin in permeabilised nuclei from haematopoietic cells containing active JAK2 (HEL cells) we find that there is significant amount of soluble, non chromatin bound HP1α present in HEL cells whereas HP1β is essentially all chromatin bound (Figure 3A). Since the binding of HP1α and HP1β were analysed within the same population of cells, the difference in their apparent chromatin binding affinities cannot be attributed to H3K9me. Our results therefore raised the possibility that in HEL cells JAK2 signalling may weaken HP1α binding and/or stabilise the binding of HP1β.
Given our findings we searched whether an additional binding site for HP1α or HP1β lies around H3Y41, the site phosphorylated by JAK2. Figure 3B and supplementary figure 10 A & B show that HP1α binds specifically to an unmodified H3 peptide encompassing amino acids 31-56, and this binding is markedly reduced when the peptide is phosphorylated at H3Y41. In contrast, HP1β binds neither the unmodified nor the modified peptide. The integrity of the H3Y41ph peptide is demonstrated by the fact that the H3Y41ph antibody binds only the modified peptide. Thus, these in vitro data demonstrate that phosphorylation of H3Y41 selectively destabilises the binding of HP1α from this region of H3. Furthermore, the binding of HP1α to the Y41 region of H3 is specifically mediated by its chromoshadow domain (CSD), therefore utilising an alternative binding domain to its interaction with H3K9me (figure 3C)25. Indeed, H3K9me peptides in trans neither cooperate with nor compete for binding of HP1α to the Y41 region of H3 (supplementary figure 10C). Importantly, the interaction between the CSD of HP1α and H3 is inhibited by the presence of H3Y41ph (figure 3D).
To provide further evidence that HP1α recognises chromatin via a second H3 binding site (other than H3K9me), we used a peptide competition assay. This assay was previously used to demonstrate binding of HP1α to H3K9me25. Figure 3E and F show that an H3K9me3 peptide displaces HP1α from nuclear heterochromatic speckles and a peptide spanning H3 residues 31-56 displaces HP1α to a similar extent. In contrast, the H3(31-56) peptide phosphorylated at Y41 (H3Y41ph) is unable to displace HP1α from heterochromatin, consistent with our in vitro experiments showing that HP1α is unable to bind H3 phosphorylated at Y41.
These results indicate that HP1α has a second binding site on H3 that tethers its CSD to chromatin and that this contact is disrupted by phosphorylation of Y41. We next asked whether binding of HP1α is modulated by JAK2 signalling in vivo. Permeabilised nuclei were prepared from HEL cells cultured with or without JAK2 inhibitors. Figure 3G demonstrates that inhibition of JAK2 results in retention of chromatin-bound HP1α and prevents its displacement into the soluble fraction (compare lanes 2 and 3 to 1). Notably, in contrast to H3Y41ph (figure 2G), the level of H3K9me3 is unaffected by JAK2 inhibition consistent with a role for JAK2 kinase in the release of HP1α from chromatin as a consequence of H3Y41 phosphorylation.
To investigate the biological function of the interplay between H3Y41ph and HP1α we performed genome wide expression profiling of HEL cells grown with or without the JAK2 inhibitor TG101209 for 4 hours to identify genes regulated by JAK2. Figure 4 and supplementary figure 11 A&B show the genes whose expression is most reduced by inhibition of JAK2. A number of these genes (e.g Pim1, Bcl-xL, CyclinD) have previously been identified as transcriptional targets of JAK2, as part of the canonical JAK2/STAT5 pathway, which is a major pathway operational in normal and dysregulated erythroid cells 17,26. These genes and many others in our profiling contain canonical STAT5 binding sites. In addition, our profiling approach also identified several JAK2 regulated genes which do not contain predicted STAT5 binding sites27.
One of the top 0.5% of genes down-regulated by JAK2 inhibition is the lmo2 oncogene, which has been linked to JAK2 signalling.28 The connect between lmo2 expression and JAK2 inhibition was independently confirmed with a second JAK2 inhibitor, AT9283 (supplementary figure 11C). Lmo2 is essential for normal haematopoietic development and has been implicated in leukemogenesis6,8. We therefore employed chromatin immunoprecipitation to assess changes to the chromatin structure of the lmo2 gene following JAK2 inhibition. Down-regulation of lmo2 expression (corroborated by reduced levels of H3K4me3) was accompanied by reduced levels of H3Y41ph together with a reciprocal increase in the binding of HP1α (figure 4B and supplementary figure 12) at sites surrounding the lmo2 transcriptional start site. In contrast, HP1β did not increase on the lmo2 promoter following JAK2-inhibition (data not shown) and the promoter of β2microglobulin, a housekeeping gene, and two sites upstream of the lmo2 promoter showed no change in H3K4me3, H3Y41ph or in HP1α binding (figure 4B and supplementary figure 13). Collectively, these results demonstrate that JAK2 signalling results in H3Y41 phosphorylation and the exclusion of HP1α at the lmo2 promoter.
The data presented here demonstrate a novel nuclear role for JAK2 outside its established involvement as an initiator of cytoplasmic signalling cascades. In the nucleus, JAK2 mediates the phosphorylation of H3 and displaces HP1α from a novel binding site surrounding H3Y41. Since HP1α can also associate with methylated H3K9, binding to the H3Y41 region via its CSD may allow HP1α to further stabilise its association with chromatin. Alternatively, HP1α may utilise this new binding site on H3 to localise at distinct loci, where H3K9 methylation is absent. In either scenario, phosphorylation of H3Y41 by JAK2 would destabilise the binding of HP1α, but not HP1β, to chromatin.
The displacement of HP1α by JAK2 is likely to be tightly regulated in normal cells. However, in malignancies driven by constitutive activation of JAK2, unregulated displacement of chromatin-bound HP1α may override potential tumour suppressive functions of HP1α29,30. This suggestion is supported by the fact that enforced over-expression of HP1 ameliorates the leukaemic phenotype of overactive JAK signalling in Drosophila melanogaster16. The data presented here provide a direct mechanistic explanation for the regulation of HP1α by the JAK2 pathway and identify the euchromatic oncogene lmo2 as a direct target for JAK2. Interestingly, the lmo2 locus lacks a predicted STAT5 binding site. Moreover, chromatin immunoprecipitation (ChIP) analyses indicates that STAT5 does not bind the Lmo2 locus (A. Wood and B. Göttgens, manuscript submitted) and manipulation of STAT5 in haematopoietic progenitors does not alter the expression of lmo231 (Jan Jacob Schuringa personal communication). Together these observations raise the possibility that regulation of lmo2 by JAK2 may not require STAT5.
In addition to transcriptional regulation of lmo2 it is possible that dysregulated displacement of HP1α by H3Y41 phosphorylation may have other oncogenic consequences (Figure 4C). HP1α is recognised to reduce mitotic recombination29, repress transcription of heterochromatic genes32 and preserve centromeric architecture leading to faithful sister chromatid segregation30. Indeed, the phenotypic consequences of constitutive JAK2 activation in haematopoietic malignancies (increased gene expression, mitotic recombination and genetic instability)1,2,15 are consistent with reversal of these functions.
Cell culture and isolation of peripheral blood stem cells were performed using standard methodology15. Immunofluorescence images were captured with an Olympus Fluoview FV1000 microscope and cells were prepared and stained as previously described15,25. Cell fractionation, immunoprecipitation, western blotting and kinase assays were performed using standard methodology25. Peptides (supplementary table 1) were synthesized by Almac Sciences and used for binding/competition assays as previously described25. Information about parameters for STAT5 binding site analysis and antibodies used is available in full methods summary (supplementary information).
We thank P. Flicek, S. Wilder, B. Huntly, SJ. Dawson and all the members of the A.R.G, B.G and T.K labs, in particular P. Hurd, B. Xhelmace, E.J. Baxter and P. Beer, for helpful discussions. We thank A. Wood for sharing unpublished data, J. LeQuesne for help with image analysis and J. Lyons, M. Squires and Astex Therapeutics, Cambridge, UK for kindly providing AT9283. This work was supported by PhD fellowship grants to M.A.D from The General Sir John Monash Foundation, Cambridge Commonwealth Trust and Raymond and Beverly Sackler. The Green (A.R.G) laboratory is funded by the UK Leukaemia Research Fund, the Wellcome Trust, Leukemia & Lymphoma Society of America and NIHR Cambridge Biomedical Research Centre. The Göttgens (B.G.) laboratory is funded by the Leukemia Research Fund, Cancer Research UK and an MRC studentship to SDF. The Kouzarides (T.K.) laboratory is funded by grants from Cancer Research UK (CRUK) and the 6th Research Framework Programme of the European Union (Epitron and SMARTER).
Author Information: We declare competing financial interests; TK is a director of Abcam Ltd and ARG is on the clinical advisory board for Astex Therapeutics, Cambridge, UK.