Cytokine-Regulated Phosphorylation and Activation of TET2 by JAK2 in Hematopoiesis

Epigenetic changes including DNA cytosine methylation and hydroxymethylation play important roles in regulating differentiation of hematopoietic cells into committed progeny (1–3). The 5-hydroxymethylcytosine residue (5-hmC) is generated from 5-mC through dioxidation by the Ten-eleven translocation (TET) family of proteins (2, 4, 5). TET enzymes, especially TET2, are critical for myeloid and erythroid differentiation and inactivating mutations in this enzyme are associated with development of myeloid malignancies and altered hematopoietic differentiation (2, 4). TET enzymes require cofactors/substrates such as iron, oxygen, and α-ketoglutarate for their catalytic activity, and also ascorbate, which indirectly increase TET activity by acting on iron oxidation state leading to modulation of hydroxymethylcytosine levels in various models (1, 6, 7). Even though limited data exist on post-translational modification of TET enzymes, such as glycosylation and phosphorylation (8), these changes have not yet been linked to its activity or been shown to be a result of extracellular stimuli. More importantly, even though TET enzyme–catalyzed DNA hydroxymethylation is a critically important pathway for sustaining normal hematopoiesis, no mechanistic link between hematopoietic cytokine stimuli, TET activity, and DNA hydroxymethylation has been established.

Janus-activated kinases (JAK) are receptor-associated tyrosine kinases that are activated by cytokine receptors. Erythropoietin (EPO), stem cell factor (SCF), and FLT3 ligand (FLT3L) are important regulators of hematopoietic stem and progenitor cells whose cognate receptors utilize JAK2 to transduce signals intracellularly. Thus JAK2 activation is critical for cytokine-induced signaling and the activating V617F mutation in the JAK2 pseudokinase domain can lead to overactivation of signaling, leading to the development of myeloproliferative neoplasms (9–11). Accumulated data have established that the vast majority of activation signals downstream of JAK2 are due to STAT activation by JAK2. There have not been any reports linking JAK2 activation directly to modulators of epigenetic regulation.

In this report, we demonstrate that JAK2 is a kinase responsible for phosphorylating TET2 and that the sites of phosphorylation are at tyrosines 1939 and 1964 located within the catalytic core of TET2. We also show that TET2 phosphorylation is associated with increased TET activity, resulting in increased DNA hydroxymethylation. The two identified sites of tyrosine phosphorylation within TET2 are important in promoting erythroid differentiation and are absent in TET1 and TET3 family members, suggesting nonoverlapping functions for TET2. We also demonstrate that the JAK2^V617F-activating mutation is associated with increased hydroxymethylation and decreased genome cytosine methylation in primary myeloproliferative neoplasm samples and in transgenic mouse models with the Jak2^V617F mutation. These data provide a novel mechanistic link between cytokine-induced signaling and epigenetic regulation of hematopoiesis. These data also show the importance of novel post-translational modifications of TET2 in hematopoiesis.

Cytokines are important regulators of hematopoiesis and act via their cognate receptors to elicit downstream changes. Because TET enzymes and DNA hydroxymethylation have been shown to be important regulators of hematopoiesis, we wanted to determine whether TET enzymes are capable of responding to extracellular cytokine stimuli. We exposed human primary hematopoietic stem/progenitor cells and erythroid progenitors to the pleotropic cytokines, FLT3L and SCF, and to the lineage-specific cytokine EPO in a time-dependent fashion and examined TET enzyme activity and 5-hmC levels. We observed that TET2 activity gradually increased over a 60-minute time period when exposed to FLT3L, whereas TET2 activity peaked at 30 minutes and then declined by 60 minutes when erythroid progenitors were exposed to SCF or EPO (Fig. 1A–C). To ascertain whether cytokine-mediated increased TET activity also resulted in increased 5-hmC content, we exposed both early human stem/progenitor cells and erythroid progenitors to each of the three cytokines for 30 minutes prior to DNA isolation and determined 5-hmC levels by ELISA. We observed that 5-hmC levels were consistently increased following cytokine exposure in three independent biological replicates (Fig. 1D–F). These studies were extended by investigating whether thrombopoietin (TPO) and granulocyte colony-stimulating factor (G-CSF) are also capable of activating TET2 leading to increased 5-hmC levels; this analysis showed increased TET2 activity and 5-hmC levels upon cytokine stimulation (Supplementary Fig. S1A–S1D).

To determine the mechanism of TET activation and increased 5-hmC levels in response to hematopoietic cytokines, we first interrogated the TET2 protein sequence for potential sites of post-translational modifications. Using multiple algorithms (12), we performed in silico analysis and identified multiple phosphorylation consensus sequence motifs for several well-characterized tyrosine and serine/threonine kinases. Many of these potential phosphorylation sites were clustered within the catalytic core of the TET2 protein, including many of the sites predicted to be phosphorylated by JAK2 (Fig. 2A). Therefore, we focused on examining tyrosine phosphorylation at these sites, especially because JAK2 is a critical mediator of cytokine signaling cascades in hematopoietic stem and progenitors. First, we exposed human CD34⁺ early stem/progenitor cells to FLT3L and erythroid progenitors to either EPO or SCF followed by immunoprecipitation of total protein with an anti-TET2 antibody, followed by immunoblot analysis using an anti-phosphotyrosine antibody (Fig. 2B). We observed that exposure to each of the cytokines resulted in tyrosine phosphorylation of TET2 (Fig. 2B). To further interrogate this finding, we focused on erythroid progenitors and exposed primary human erythroid progenitors to EPO in a time-dependent fashion. This experiment showed that TET2 is phosphorylated by EPO and maintains such phosphorylation for up to 60 minutes (Fig. 2C). A similar pattern was also observed in the UT-7 erythroid cell line, further confirming this finding (Supplementary Fig. S2A). To establish whether blocking the actions of FLT3L and EPO lead to abolishing TET2 activity, we exposed primary stem/progenitor cells or erythroid progenitors to FLT3L or EPO in the presence or absence of specific inhibitors of FLT3 and JAK2 (activator of EPOR) and found that inhibiting these two kinases resulted in abrogation of TET activity to its basal levels, suggesting that TET activation is dependent on these two pathways (Fig. 2D and E).

We then focused on ascertaining whether TET2 is a direct substrate for JAK2, as suggested by the presence of multiple JAK2 consensus motifs and abolishment of TET2 activity in the presence of the JAK2 inhibitor. To determine this, we devised a cell-free in vitro kinase assay in which reconstituted recombinant GFP/FLAG-tagged TET2 protein (catalytic core) was used as a substrate for the recombinant active form of JAK2, as described in our Methods section. In parallel sets of in vitro assays, we also used recombinant SYK and p38 MAP kinases as controls for establishing specificity. The reaction products of these in vitro kinase assays were analyzed by immunoprecipitation with an anti-FLAG antibody (GFP/FLAG-tagged TET2 as the target) followed by immunoblot analysis using an anti-phosphotyrosine antibody. Our data showed that only JAK2 kinase was able to phosphorylate the TET catalytic core, and not SYK and p38 MAP kinases (Fig. 2F). Furthermore, the presence of the JAK2 inhibitor in the kinase assay abrogated TET2 phosphorylation, reinforcing the specificity of the in vitro kinase assays. Because p38 MAP kinase (used as a control) is a serine/threonine kinase, we also determined whether the catalytic domain of TET2 is a target for p38 MAP kinase by performing immunoblot analysis with an anti-phosphoserine–specific antibody, which showed no such phosphorylation (Fig. 2F). We then determined whether phosphorylation of the TET2 catalytic domain leads to activation of TET2 in a cell-free system. These experiments revealed that the phosphorylated TET2 catalytic domain is active, and its activity can be abolished by the presence of the JAK2 inhibitor during the in vitro kinase assay reaction (Fig. 2G). To further interrogate these findings, we performed a series of in vitro kinase assays, in which we used increasing concentrations of JAK2 to demonstrate that phosphorylation of TET2 by JAK2 is dose-dependent, reinforcing that TET2 is a direct substrate for JAK2 (Fig. 2H). Finally, we investigated whether JAK2 physically interacts with TET2 catalytic core by carrying out reciprocal immunoprecipitation-coupled immunoblot analysis. These experiments showed that JAK2 physically interacts with TET2 in multiple cell lines with erythroid characteristics (Fig. 2I and J; Supplementary Fig. S2B and S2C). We then investigated the localization of and interaction between these two proteins within the cytoplasmic and nuclear compartments using primary erythroid progenitors. These experiments revealed that TET2 was primarily localized within the nucleus, whereas JAK2 was localized both in the cytoplasm and in the nucleus, although the amount of JAK2 protein in the nucleus was less than in the cytoplasm (Supplementary Fig. S2D). Examination of the nuclear fraction showed that interaction between TET2 and JAK2 occurs within the nucleus and this interaction is reduced in the presence of the JAK2 inhibitor (Supplementary Fig. S2E).

Given that TET2 is phosphorylated by JAK2, we sought to identify specific tyrosine residues within the TET2 catalytic domain that are phosphorylated by JAK2. Using the in vitro kinase platform, we used the recombinant TET2 catalytic domain under our assay conditions in the presence or absence of enzymatically active JAK2 prior to subjecting the TET2 catalytic core to mass spectrometry analysis. The results of repeated experiments revealed that tyrosine-1939 (Tyr-1939) and tyrosine-1964 (Tyr-1964) are phosphorylated, as seen by the 80-Da shift corresponding to gain of a phosphate group in each of the peptides representing this region of the TET2 protein, reflected by the peaks within the chromatograms (Fig. 3A–D). The control samples that were not exposed to JAK2 kinase showed no such shift (Fig. 3A–D). Interestingly, in silico analysis of amino acid sequences of TET1, TET2, and TET3 showed that the two identified tyrosine residues were present only in TET2, but not in TET1 or TET3 (Supplementary Fig. S3A). Furthermore, sequence alignment of the TET2 catalytic cores of multiple other species showed that the Tyr-1964 site was conserved among all the species examined, whereas Tyr-1939 was slightly less conserved (Supplementary Fig. S3B).

To further interrogate whether phosphorylation of Tyr-1939 and Tyr-1964 by JAK2 are interdependent, we performed in vitro mutagenesis, where the two sites were mutated individually or together by substituting tyrosine resides with alanine within the TET2 catalytic core. Using these constructs, we produced recombinant TET2 proteins and used them as substrates in the presence or absence of active JAK2 in our in vitro kinase assay. These experiments revealed that even when Tyr-1939 was substituted with alanine, JAK2 was able to phosphorylate the Tyr-1964 site. However, when the Tyr-1964 site was replaced with alanine, it abrogated phosphorylation of the Tyr-1939 site (Fig. 3E). As expected in the wild-type TET2 catalytic core, where both these tyrosine residues were intact, we observed a high level of tyrosine phosphorylation in the presence of JAK2 compared with the sample where we omitted the addition of active recombinant JAK2 (Fig. 3E). These results suggest that the presence of Tyr-1964 is important for robust phosphorylation of Tyr-1939 of TET2 by the JAK2 enzyme.

We then determined whether phosphorylation of Tyr-1939 and Tyr-1964 is important for the association of TET2 with JAK2. We transfected each of the two single mutants, double-mutant or wild-type TET2 constructs into the UT-7 erythroid cell line and performed immunoprecipitation-coupled immunoblot analysis. Each of the UT-7 cell lines carrying GFP/FLAG-tagged TET2 constructs were immunoprecipitated with an anti-FLAG antibody followed by immunoblot analysis with an anti-JAK2 antibody to detect JAK2. These experiments showed a marked disruption of interaction between JAK2 and TET2 when Tyr-1939 or Tyr-1964 was mutated, whereas the wild-type TET2 sustained the JAK2–TET2 interaction, suggesting that both Tyr-1939 and Tyr-1964 are critical for maintaining the interaction between JAK2 and TET2 (Fig. 3F).

To further interrogate the nature of the interaction between TET2 and JAK2, we performed surface plasmon resonance (SPR) binding analysis using the Biacore instrument. This instrument allows determination of the binding affinity of a given protein to specific peptide motifs under dynamic flow conditions. To determine the affinity of TET2-phosphorylated motifs to JAK2, we synthesized four different biotinylated peptides surrounding the two tyrosine residues (Y1939 and Y1964) with four different permutations, as described in our Methods section. The first peptide was synthesized without the phosphate moiety on either one of the two tyrosine residues. The second peptide contained the phosphate moiety on both tyrosines, the third peptide contained the phosphate moiety only on Tyr-1939, and the fourth peptide contained the phosphate moiety only on Tyr-1964. These peptides were immobilized on sensor chips via streptavidin, as described in our Methods section, in each of the channels available on the chip. Recombinant JAK2 (experimental) and p38 MAP kinase (control) were used as analytes to determine the binding affinity of each of these proteins to the four peptides on the chip. Various concentrations (0–500 nmol/L) of each analyte were applied to the channels in the chip at the flow rate and temperature as specified in our Methods section to generate sensorgrams, which were then used to plot the response to various concentrations of JAK2 for each of the four peptides being tested (Fig. 3G–J). We also included a series of samples that contained the JAK2 inhibitor (Fig. 3G–J). These plots were then used to determine the binding constants for JAK2 interaction with each of the four peptides. The results of these experiments showed that JAK2 binding to TET2 was most effective (highest binding affinity) when both Tyr-1939 and Tyr-1964 were phosphorylated (K_D = 127; Fig. 3H). All the other permutations of phosphorylation or nonphosphorylation exhibited very weak binding (K_D = 352,265, 345) compared with the dual phosphorylated peptide (Fig. 3H). The presence of JAK2 inhibitor abolished all binding (Fig. 3G–J), similar to recombinant p38 MAP kinase, which was used as a negative control (Fig. 3G–J). These data showed that phosphorylation of TET2 by JAK2 within the catalytic domain of TET2 provides docking sites for sustained JAK2 interaction with TET2.

Our previous work (5) had demonstrated that several transcription factor binding sites within the genome acquire 5-hmC during erythroid differentiation, implying TET2 is recruited to these sites by transcription factors that normally bind these sites. Our recent published work demonstrated that the binding site motif for KLF1 was among the top transcription factor binding sites that gained 5-hmC (4, 5). Therefore, wereasoned that KLF1 is a likely candidate for interactions with TET2 for potential recruitment to the specific cis regions within the genome. We used the erythroid cell line UT-7, which is known to abundantly express KLF1, to ascertain whether TET2 binds KLF1. First, we transfected GFP/FLAG-tagged TET2 catalytic domain into UT-7 cells and performed reciprocal immunoprecipitation-coupled immunoblot analysis using cells cultured under steady-state conditions (Fig. 4A and B). These experiments showed that indeed KLF1 interacts with TET2. To determine whether this interaction between TET2 and KLF1 is responsive to EPO, we deprived the UT-7 cells of EPO or exposed the cells to EPO after a period of EPO derivation and performed immunoprecipitation-coupled immunoblot analysis, which showed increased association between TET2 and KLF1 upon exposure to EPO (Fig. 4C). We then determined whether TET2 phosphorylation at the JAK2 sites was a prerequisite for KLF1 binding to TET2 by using the Biacore instrument and the methodology we had established as described previously. These experiments showed that double phosphorylated peptides bind with the highest affinity (K_D = 48.2) and nonphosphorylated peptides bind with less affinity (K_D = 74.7; Fig. 4D and E). Phosphorylation at either Tyr-1939 or Tyr-1964 site alone showed very low binding affinity (Fig. 4F and G). Once again, using p38 MAP kinase as a control showed no binding (Fig. 4D–G).

To ascertain the functional role played by the two tyrosine residues, we transfected GFP-containing plasmids carrying either wild-type TET2 catalytic core or a TET2 construct where both tyrosine residues were replaced by alanine into an established Tet2 knockout (KO) murine cell line (C57BL6 Tet2^−/−). The Tet2 KO and Tet2 wild-type cell lines had been created from hematopoietic cells derived from Tet2 KO and control mice after stable transfection with an estrogen-regulated Hoxb8 transgene, as described previously (13). We demonstrated that the Tet2 KO cells did not express Tet2 and that both the KO and control cell lines expressed similar levels of CD34 and CD71 (Fig. 5A; Supplementary Fig. S4A and S4B). In addition, cells were at early stages in the hematopoietic differentiation program and did not display expression of myeloid markers (Supplementary Fig. S4B). We then transfected the Tet2 KO line with either wild-type TET2 catalytic core or the plasmid carrying the TET2 catalytic core with replacement of both tyrosines with alanine. Subsequently, cells were flow cytometry–sorted on the basis of GFP expression and immediately used in colony assays (Supplementary Fig. S4C). The morphologies of these colonies showed both erythroid and nonerythroid colonies (Fig. 5B). The cell line with Tet2 KO did not lead to any colony formation although there were individual viable single cells. Transfection of unmutated TET2 catalytic core in the Tet2 KO cells led to robust colony formation (Fig. 5C). The transfection of TET2 with double tyrosine to alanine mutations led to significantly reduced numbers of total colonies, suggesting a role for these tyrosines in hematopoietic colony formation. We then demonstrated that the two tyrosines are also important in promoting terminal differentiation of erythroid cells by harvesting the cells from the methylcellulose media and observing a reduction in Ter-119 expression in the mutant-containing cells compared with cells expressing the wild-type TET2 (Fig. 5D). These data show that the TET2 tyrosines 1939 and 1964 are essential for adequate hematopoietic colony formation as well as erythroid differentiation.

Myeloproliferative neoplasms (MPN) are associated with enhanced erythroid proliferation and contain activating mutations in JAK2. Because patients with the Jak2^V617F mutation show the constitutively active form of JAK2, we reasoned that TET2 enzymatic activity as well as levels of 5-hmC might be elevated in these patients based on our results demonstrating TET2 is a substrate for JAK2 kinase. We examined TET activity and 5-hmC levels in erythroid progenitors derived from stem/early progenitor cells isolated for JAK2^V617F patients and observed significantly increased active TET2 as well as increased 5-hmC levels in mutant cells (Fig. 6A and B). Having shown that JAK2^V617F mutation leads to TET2 activation and increased levels of 5-hmC, we next wanted to evaluate the functional consequences of this on genome-wide cytosine methylation. Murine hematopoietic FDCP cells were transduced with JAK2^V617F-mutant or wild-type vectors and then used for global cytosine methylation analysis with high-density HELP assay. We observed that expression of the JAK2^V617F mutation was associated with a distinct methylation profile (Fig. 6C) that was characterized by loss of cytosine methylation when compared with cells containing wild-type JAK2 (Fig. 6D). Analysis of differentially methylated loci revealed that the loss of methylation was seen to occur throughout the genome and affected all regulatory regions (Supplementary Fig. S5A). Correlation with parallel transcriptomic analysis revealed a set of genes that were associated with promoter loss of methylation and increased expression in the JAK2^V617F-containing cells (Supplementary Table S1) that included important oncogenic transcripts such as Meis1 and Hoxa9 (Fig. 6E and F).

Next, we determined whether the JAK2^V617F mutation was associated with similar methylation patterns in primary patient samples. Samples from patients with idiopathic myelofibrosis (IMF) with JAK2^V617F mutation (and wild-type TET2) were compared with healthy controls [ref. 14; Gene Expression Omnibus (GEO) GSE42721; Supplementary Table S2; Fig. 6G]. JAK2^V617F samples were epigenetically distinct from controls and contained (Fig. 6G)an increased number of demethylated loci (Fig. 6H; Supplementary Tables S3 and S4). The loss of methylation in JAK2^V617F samples was abrogated when samples contained an inactivating mutation in TET2 as seen in JAK2V617F/TET2-mutant IMF sample comparison with controls (Supplementary Fig. S5B). Because half of IMF cases do not have the JAK2^V617F mutation, we also compared the methylomes of wild-type and mutant samples and again observed greater loss of methylation in the JAK2^V617F-mutant cases (Supplementary Fig. S5C). Consistently, a direct comparison of JAK2^V617F samples with and without TET2 mutations demonstrated that samples with wild-type TET2 contained a greater number of hypomethylated loci (Supplementary Fig. S5D). These methylomes demonstrated that JAK2^V617F with a wild-type functional TET2 is associated with loss of cytosine methylation in IMF samples. Next, we analyzed the gene-expression profiles of IMF cases with JAK2^V617F mutations and compared them with control samples, as obtained from an independent cohort (GEO GSE54646; ref. 15). The JAK2-mutant IMF sample transcriptomes were distinct from controls in unsupervised clustering (Fig. 6I) and revealed a greater number of overexpressed transcripts when compared with controls (Fig. 6J; Supplementary Tables S5 and S6). These data demonstrate increased hydroxymethylation, loss of cytosine methylation, and globally increased gene expression in primary MPN samples containing the JAK2^V617F mutation.

To further determine the effect of JAK2^V617F activating mutation on 5-hmC in vivo, mice with transgenic expression of a constitutively active form of JAK2 (Jak2^V617F; ref. 16) and wild-type controls were used to isolate erythroid cells (sorted CD71⁺ cells) from marrow and spleens and were used for genome-wide DNA and RNA analysis. Gene-expression profiling by RNA sequencing (RNA-seq) revealed differential expression of transcripts with greater numbers of overexpressed transcripts as well as enhanced expression levels in Jak2^V617F samples compared with wild-type samples (Fig. 7A and B). We then performed 5-hMe-Seal using genomic DNA from the two cohorts of mice, and observed increased accumulation of 5-hmC peaks in Jak2^V617F erythroid samples (Fig. 7C). To evaluate the relationship between 5-hmC and gene expression, a colocalization plot was constructed showing the coverage of the transcripts obtained from RNA-seq samples within the regions defined by the 5-hmC peaks. A positive colocalization of expressed transcripts was seen around 5-hmC peaks, and this was more pronounced in Jak2^V617F samples (Fig. 7D). The transcripts that were overexpressed and associated with gain of 5-hmC in the Jak2^V617F mice were found to be associated with important oncogenic functional pathways (Supplementary Table S7), and the differential 5-hmC sites were significantly enriched in binding sites for regulatory erythroid transcription factors Gata1, Klf1, and others (Supplementary Fig. S6). Overexpressed transcripts associated with gain of hydroxymethylation in Jak2^V617F samples included those encoding important erythroid-associated proteins Gata1, Cxcr2, Hba1, Klf1, and others (Fig. 7E and F; Supplementary Fig. S7A–S7C). Finally, we determined whether 5-hmC changes seen in murine Jak2^V617F cells were similar to gene-expression signatures seen in human MPNs that were obtained from a large transcriptomic database (GEO GSE54646). Differential 5-hmC regions in JAK2^V617F murine samples were correlated with differentially expressed genes in human JAK2^V617F samples, and a positive correlation was seen between murine and human on gene set enrichment analysis (GSEA; Fig. 7G).

In this report, we describe a novel post-translational modification (phosphorylation) of the TET2 enzyme and show that such modification results in TET2 activation leading to a global increase in 5-hmC levels in the genome. We also demonstrate that JAK2 is the kinase responsible for that phosphorylation and that association between TET2 and JAK2 is modulated by phosphorylation of the two identified residues. These changes are downstream of various cytokines that regulate hematopoiesis and provide a novel mechanistic link between cytokine-mediated signaling and epigenetic changes that have been shown to be important in hematopoiesis.

These findings reveal a novel role of JAK2 kinase as a direct regulator of DNA methylation. JAK overactivation has been shown to globally disrupt heterochromatin to cause blood–tumor formation in Drosophila (17). This observation was independent of JAK2-mediated canonical STAT activation and could not be explained by upregulation of STAT targets. Our data demonstrate that JAK2-mediated TET activation potentially regulates gene transcription via direct epigenetic changes. TET2 serine phosphorylation has recently been reported to occur by AMPK and to result in stabilization of the protein (18). Furthermore, a report demonstrated that the mutant JAK2^V617F can be present in the nucleus of hematopoietic cells and can directly phosphorylate Tyr 41 (Y41) on histone H3, thus inhibiting the binding of Heterochromatin protein 1α (HP1α) on histones and affecting transcription via regulation of the histone code (19). The JAK2^V617F-mutant kinase has also been shown to phosphorylate the arginine methyltransferase PRMT5 and to affect oncogenic transcription and cause myeloproliferation (20). Our data now demonstrate another novel kinase activity of JAK2 that influences the DNA hydroxymethylation status of cells. Recent studies have shown a strong association of increased gene expression especially when regulatory elements within the genome are hydroxymethylated (4, 21). Our previous report demonstrated that increased global cytosine hydroxymethylation occurs during blood stem/early progenitor commitment to the erythroid lineage and is associated with upregulation of important erythroid lineage–specific genes (4). Our data now demonstrate this increased 5-hmC observed in that study is potentially due to EPO- and/or SCF-induced JAK2 activation leading to TET enzyme activation.

We also demonstrate that TET2 associates with KLF1 during EPO-mediated activation. KLF1 is a master transcription factor that is important for the erythroid differentiation program (22, 23). We had previously shown that hydroxymethylation during erythroid differentiation occurred at KLF1 transcription factor binding sites but could not provide a mechanistic link for this observation (4). The association of KLF1 with TET2 raises the possibility that TET2 recruitment to specific cis elements in the DNA may occur via lineage-specific transcription factors. Recently published reports have shown that TET2 binds transcription factors such as early B-cell factor-1 (EBF1; ref. 24) and WT1 (25). On the basis of our data, it is possible that TET2 is recruited to KLF1 binding sites through their interaction resulting in hydroxymethylation of regulatory elements leading to the activation of a specific set of genes critical for the erythroid differentiation program (4). Furthermore, this recruitment may be enhanced or dependent on cytokine stimuli, such as EPO, because we observed enhanced interaction between TET2 and KLF1 upon EPO stimulation. These results provide evidence in support of a mechanism that is CXXC domain–independent in contrast to what has been observed for TET1 and TET3.

We also demonstrate that primary samples of myeloproliferative disease with activating JAK2^V617F mutations exhibit increased TET activity and increased hydroxymethylation. Especially in samples from primary myelofibrosis, we observed that in samples with JAK2^V617F mutation, but without any TET2 mutations, there was significant decrease in cytosine methylation. These epigenetic and transcriptomic changes were lost in samples that contained a TET2 mutation in addition to the JAK2^V617F mutation. The decreased cytosine methylation seen in samples with JAK2^V617F was associated with upregulation of many oncogenic transcripts such as Meis1 and Hoxa9 that are associated with the development of myeloproliferation and myeloid malignancies (26). Furthermore, we saw a similar increase in 5-hmC in a murine model containing the Jak2^V617F mutation. Thus, taken together, the JAK2^V617F-mediated direct effects on TET activation lead to epigenetic and oncogenic changes that play potential roles in the pathogenesis of myeloproliferation.

Human primary hematopoietic cells were derived from CD34⁺ stem/early progenitor cells purified from growth-mobilized peripheral blood (purchased from Fred Hutchinson Cancer Center, Seattle, WA) from healthy donors. Cells were purified using the CliniMACS device. In experiments where lineage noncommitted early-stage cells were used, culture conditions consisted of Iscove's modified Dulbecco's medium (IMDM), 20% human serum, 1% penicillin/streptomycin, TPO (20 ng/mL), FLT3L (20 ng/mL), recombinant human IL3 (20 ng/mL), recombinant human IL6 (20 ng/mL), and SCF (100 ng/mL). Cells were cultured for 4 days prior to being used in experiments, where TET activity and phosphorylation were determined. Cells were washed with media devoid of cytokines and incubated for 4 hours in IMDM with 1% BSA prior to stimulation with FLT3L (100 ng/mL). The UT-7 cell line was obtained from ATCC and verified by microsatellite analysis.

In experiments where cells were cultured to generate erythroid progenitors, CD34⁺ cells (day 0) were cultured until day 6 according to methods described previously (27). Erythroid progenitors were then washed with media devoid of cytokines and incubated for 4.5 hours in IMDM with 1% BSA prior to stimulation with either EPO or SCF. In experiments where inhibitors were used, cells were exposed to the JAK2 kinase inhibitor ruxolitinib (INCB018424) at 100 nmol/L or treated with the FLT3 inhibitors sorafenib and AC2220 (10 nmol/L) prior to stimulation with cytokines. UT-7 EPO-dependent subline cells (28) were cultured in 10% FBS, IMDM, 1 U/mL EPO. In experiments where cells were deprived of EPO prior to stimulation, the time period of EPO deprivation was 18 hours followed by EPO stimulation for 30 minutes.

Total TET activity was estimated by fluorescence-based ELISA kit according to the instructions provided by the manufacturer (Epigentek). This assay involves the conversion of 5-mC substrates coated onto the microplate wells by the TET enzymes in the sample and results in the conversion of 5-mC to 5-hmC, which is then detected fluorometrically using a specific antibody against 5-hmC. Nuclear lysates were prepared using a Nuclear Extraction Kit (Novagen) and protein concentrations were measured using the Bradford assay. The amount of hydroxymethylated product in the samples was measured using a standard curve (0–2 ng) constructed using the regents provided in the ELISA, and TET activity was calculated as ng/min/mg.

The 5-hmC levels were assessed by antibody-dependent colorimetric assay using the MethylFlash Hydroxymethylated DNA Quantification Kit (Epigentek). DNA was extracted from primary erythroid progenitor cells (day 6 of culture) derived from in vitro culture of CD34⁺ cells purified from healthy donors used in the assay following the manufacturer's recommendation. A total of 200 ng of DNA was used in each assay. The 5-hmC content was expressed as percentages of total DNA.

Cell lysis, immunoprecipitation, and Western blot analysis were performed according to the protocols from Pierce. Cell lysis and immunoprecipitation were carried out in a buffer containing 25 mmol/L Tris•HCl pH 7.4, 150 mmol/L NaCl, 1% NP-40, 1 mmol/L EDTA, 5% glycerol, and the protease inhibitor cocktail (Thermo Fisher Scientific). The following antibodies were used for immunoblotting and immunoprecipitations: mouse monoclonal anti-TET2 (Diagenode), rabbit polyclonal anti-TET2 (Abiocode), rabbit polyclonal anti-JAK2 (Abcam), mouse monoclonal anti-EKLF 6B3 (N; ref. 23), anti-FLAG M2 Magnetic Beads (Sigma), mouse monoclonal anti-FLAG (Sigma), mouse monoclonal anti-GAPDH (Santa Cruz Biotechnology), mouse monoclonal anti-Tubulin (Santa Cruz Biotechnology), and mouse monoclonal anti-Lamin (Santa Cruz Biotechnology). The specificity of mouse monoclonal anti-TET2 antibody (Diagenode) used in immunoprecipitation studies was confirmed by mass spectrometry by using an aliquot from the immunoprecipitated protein complex.

UT-7 cells were transduced with GFP/FLAG-TET2-Catalytic cDNA plasmid, and 48 hours after transduction, cells were harvested and lysed as described previously, and the protein concentration was determined by Bradford (Bio-Rad). Lysates were immunoprecipitated using the anti-FLAG magnetic beads. Following SDS-PAGE and blotting, the membranes were incubated with either an anti-JAK2 antibody or an anti-KLF1 antibody. Infrared dye–labeled secondary antibodies and an Odyssey Infrared Imaging System (LI-COR) were used to detect and quantitate the JAK2 and KLF1 bands.

Recombinant wild-type cDNA constructs containing the TET2 catalytic domain (aa 1129–2002) tagged with and without the GFP epitope at the N-terminus or with and without the GFP epitope–tagged TET2 catalytic domain carrying Tyr-1939, Tyr-1964, or double mutations were generated using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs). These DNA constructs were subcloned into pCMV6-ENTRY plasmid (Origene). These plasmids were then transfected into HEK 293 and recombinant TET2 proteins were produced and used in in vitro kinase assays. The JAK2 and KLF1 (N-terminal 10× His- tag) were subcloned into the pET15b vector (Novagen).

The in vitro kinase assays contained 1.5 mg of protein containing GFP/FLAG-tagged TET2 catalytic domain (130 Da) immunoprecipitated with an anti-FLAG M2 magnetic beads, 200 ng of active recombinant JAK2 (Abcam), in vitro kinase buffer (25 mmol/L HEPES pH 7.5, 10 mmol/L MgCl₂, 0.5 mmol/L EGTA, 0.01% Triton X-100, 2.5 mmol/L DTT, 0.5 mmol/L sodium orthovanadate, 5 mmol/L glycerophosphate), and 200 μmol/L ATP in a volume of 100 μL. The kinase assay was initiated by incubating the kinase assay reactants for 1 hour at 30°C followed by diluting the reaction mix with 500 μL of PBS prior to collecting the immune precipitate using a magnetic selection method. Then, the immunoprecipitates were resolved by SDS-PAGE on a 4% to 15% SDS-polyacrylamide gel, transferred onto a polyvinylidene difluoride membrane, and then probed with an anti–phosphotyrosine-specific mouse mAb (Millipore) or a phosphoserine-specific mouse mAb (Qiagen). A similar approach was taken in in vitro kinase assays that involved TET2 catalytic domain carrying specific tyrosine mutations.

HEK293 cells were transfected with TET2-Catalytic-FLAG. Immunoprecipitated TET2 was phosphorylated by active recombinant JAK2 for 1 hour. The TET2-Catalytic-FLAG protein was captured by anti-FLAG magnetic beads and separated by SDS-PAGE. A band corresponding to the phosphorylated TET2 was excised from the gel and subjected to further separation on a 4% to 15% SDS-PAGE gel followed by staining with Coomassie Blue Protein Stain. The TET2 band was identified by size, and was excised from the gel and submitted to MS Bioworks, LLC, for PTM profiling, which was done by nano-LC/MS-MS with a Waters NanoAcquity High-Pressure Liquid Chromatography system (Thermo Fisher Scientific). Peptide identifications were performed using Scaffold software together with the MASCOT server for searching against the Swiss-Prot human database. The Mascot was searched with a fragment ion mass tolerance of 0.02 Da and a parent ion tolerance of 10 ppm. Carbamidomethyl of cysteine was specified in Mascot as a fixed modification. Deamination of asparagine and glutamine, oxidation of methionine, acetylation of the N-terminus, and phosphorylation of serine, threonine, and tyrosine were specified in Mascot as variable modifications.

Patient and healthy human samples were collected under an approved Institutional Review Board (IRB) protocol and written informed consent at University of Chicago (Chicago, IL) and Albert Einstein College of Medicine (Bronx, NY). In experiments where 5-hmC levels were quantified in JAK2^V617F patient samples or healthy counterparts, a mass spectrometry approach was used. Briefly, DNA hydrolysis was performed as previously described by Gan and colleagues (29) with minor modifications. Two micrograms of genomic DNA was first denatured by heating at 100°C for 3 minutes. Five units of Nuclease P1 (Sigma) were added and the mixture incubated at 37°C for 2 hours. A 1/10 volume of 2 mol/L ammonium bicarbonate and 0.002 U of venom phosphodiesterase 1 (Sigma) were added to the mixture and the incubation continued for 2 hours at 37°C. Next, 0.5 U of calf intestine alkaline phosphatase (Invitrogen) was added and the mixture was incubated for 1 hour at 37°C. Samples were then diluted to 200 μL with water and filtered through 0.22-μm syringe filters (Millipore). Samples were run on an Agilent 1290 Series liquid chromatography machine in tandem with the Agilent 6460 Triple Quad Mass Spectrometer. LC separation was performed at a flow rate of 300 μL/minute through ACQUITY UPLC Oligonucleotide BEH C18 Column (Waters) with a mixture of 100% methanol and 0.1% formic acid in water. The effluent from the column was directed to an electrospray ion source (Agilent Jet Stream) connected to a triple quadrupole mass spectrometer (Agilent 6460 QQQ) operating in the positive ion multiple reaction monitoring (MRM) mode using previously optimized conditions, and the intensity of specific MH⁺→fragment ion transitions was recorded [dC (228.1/112.1); 5-mC (242.1/126.1); 5-hmC (258.1/142.1)]. The measured percentage of 5-mC and 5-hmC in each experimental sample was calculated from the MRM peak area divided by the combined peak areas for 5-mC plus 5-hmC plus C (total cytosine pool). To correct for the different detection sensitivity of different nucleosides, observed percentages were then normalized by the standard curves, which were constructed from measuring standard samples with known 5-mC and 5-hmC percentages.

The binding analyses were performed on a BiacoreT200 instrument (GE Healthcare) according to the manufacturer's instructions andpreviously published methods. Amino acid sequences of the peptides immobilized on a Series S Sensor Chip SA (streptavidin; GE Healthcare) consisted of Y1939Y1964;Biotin-¹⁹³⁴KYGPDYVPQKSHGKKVKREPAEPHETSEPTYLRFIK¹⁹⁶⁹; pY1939pY1964;Biotin-¹⁹³⁴KYGPD(pY)VPQKSHGKKVKREPAEPHETSEPT(pY)LRFIK¹⁹⁶⁹, pY1939; Biotin-¹⁹³⁹KYGPD(pY)VPQK¹⁹⁴³, pY1964; Biotin-¹⁹⁵⁸SEPT(pY)LRFIK¹⁹⁶⁹. The peptides were initially prepared as 1 mmol/L stocks in 100% DMSO and diluted to 25 μg/mL with HBS-EP (10 mmol/L HEPES, pH 7.4, 150 mmol/L NaCl, 3 mmol/L EDTA, and 0.05% surfactant P20) for immobilization. Three analyte proteins (JAK2, KLF1, and p38α) were first buffer exchanged with SPR binding buffer (HBS-EP + 10% glycerol). A series of increasing concentrations (0–1 μmol/L at 2-fold dilution) of each analyte were applied to all four channels in SPR binding buffer at a 30 μL/minute flow rate at 25°C. Sensorgrams were referenced with zero concentration signals, and analyzed using the Biacore T200 evaluation software v3.0. Response units at each concentration were measured during the equilibration phase for steady-state affinity fittings, and the K_D values were determined by fitting the data to the equation as shown below, where y is the response, y_max is the maximum response, and x is the analyte protein concentration.

Total RNA was extracted from C57BL6 Tet2 wild-type and C57BL6 Tet2 KO cells using the RNeasy Mini Kit (Qiagen). First-strand cDNA was synthesized by the High Capacity cDNA ReverseTranscription Kit (Applied Biosystems). Real-time PCR (ViiA7, Applied Biosystems) was performed using the Power SYBR Green PCR Kit (Applied Biosystems). PCR amplifications were performed in triplicate for TET2 along with parallel measurements of β-actin cDNA (an internal control). To confirm specific amplification of the desired PCR product, melting curves were analyzed and PCR products were separated on a 1.5% agarose gel. The following primers were used:Tet2_F:5′-CCAAGACCAAGAAAGCAGCTCG-3′, Tet2_R:5′-CCGAAAGCTGCGGTTGTGC-3′, Actin_F:5′-TGAACCCTAAGGCCAACCGTGAAA-3′, Actin_R:5′-CAGGATGGCGTGAGGGAGAGCATAG-3′. cDNA was also used for RT-PCR assay.

Tet2 KO cell lines were created using C57BL/6J wild-type Tet2 mice (stock# 0006641) and the Tet2 KO (stock # 0233591) mice purchased from Jackson Laboratories using published methodology in A.F. List's laboratory (Moffitt Cancer Center, Tampa, FL; ref. 13). Briefly, C57BL6 Tet2^−/− bone marrow hematopoietic stem/progenitor cells (HSPC) were stably transfected with an estrogen-regulated Hoxb8 transgene to immortalize both wild-type and KO HSPCs. Flow cytometry on these established cell lines was performed using fluorochrome-conjugated antibodies against Cd34-PE, Cd71-PE, and Ter-119-PE to access for surface protein expression levels. C57BL6 Tet2 KO cells were transfected with 5 μg of wild-type or double mutation (Y1939A+Y1964A) GFP/FLAG-tagged TET2 catalytic core plasmid in 5 × 10⁵ cells per electroporation using Neon Transfection System 10 μL Kit (Invitrogen). After transfection, cells expressing wild-type or phospho-mutant TET2 (substituting Tyr-1939 and Tyr-1964 with alanine) catalytic core were purified by FACS sorting for GFP⁺ cells. These purified cells were used immediately for performing colony assays. GFP⁺ population cells (GFP/FLAG-TET2-Catalytic) were plated at 5,000 cells/plate in MethoCult SF M3436 (StemCell Technologies) containing methylcellulose in IMDM, recombinant human insulin, BSA, 2-mercaptoethanol, human transferrin, murine SCF, β-estradiol, and erythropoietin. The colony assay medium was optimal for growing BFU-Es and not myeloid colonies, which were maintained in culture for 19 days, prior to analyses. Cells were harvested for evaluation of morphology and flow cytometry analysis for Ter-119 expression as a marker of terminal erythroid differentiation.

Patient samples with MPN were collected at either Johns Hopkins University (Baltimore, MD) or Albert Einstein School of Medicine (Bronx, NY) under IRB-approved protocol and written consent. Genomic DNA from neutrophils was isolated from peripheral blood via Ficoll gradient density separation and used for methylation analysis by the HELP assays (HD17). HELP microarray data have been submitted to the GEO database for public access (GSE42721; ref. 14). FDCP cells were stably transfected with JAK2^V617F or JAK2 WT retroviral plasmids and used for genome-wide methylation analysis with HELP assay (HD18; GEO accession GSE116537).

MPN patient neutrophil gene expression data was obtained from GSE54646 (15). Patients with IMF with JAK2V617F were compared with controls.

Mice with transgenic expression of a constitutively active form of JAK2 (Jak2^V617F; ref. 16) and wild-type controls were used to isolate erythroid lineage cells (sorted CD71⁺ cells) from marrow and spleen. All animal experiments were approved by the Institutional Animal Care and Use Committee of Northwestern University (Evanston, IL). One hundred nanograms of genomic DNA per sample was fragmented in 50 μL Tagmentation buffer at 55°C. Fragmented DNA was purified by Zymo DNA Clean & Concentrator Kit (Zymo Research). Then, the selective 5-hmC chemical labeling was performed in 25 μL glucosylation buffer (50 mmol/L HEPES buffer pH 8.0, 25 mmol/L MgCl₂) containing the fragmented DNA, 100 μmol/L N₃-UDP-Glc, and 1 μmol/L β-GT, and incubated at 37°C for 2 hours. After purifying in 45 μL nuclease-free water, 1.5 μL DBCO-PEG₄-Biotin (Click Chemistry Tools, 4.5 mmol/L stored in DMSO) was added and incubated at 37°C for 2 hours. The biotin-labeled DNA was pulled down by 5 μL C1 Streptavidin beads (Life Technologies) for 15 minutes at room temperature. Next, the C1 bead-captured DNA fragments were subjected to 12 cycles of PCR using Nextera DNA Sample Preparation Kit (Illumina). The resulting amplified product was purified by 0.9X AMPure XP beads (Beckman Coulter). Input library was made by direct PCR (4 cycles) from fragmented DNA without chemical labeling and pull-down. All these libraries were sequenced using Illumina NextSeq 500 PE42 platform. Peak calling for each sample was performed using MACS (version 1.4.2) with default parameters as performed previously (21).

Mice with transgenic expression of a constitutively active form of Jak2 (Jak2^V617F; ref. 16) and wild-type controls were used to isolate erythroid lineage cells (sorted CD71⁺ cells) from marrow and spleen. Fifty nanograms of mRNA per sample was used for RNA-seq library preparation (KAPA BIOSYSTEMS, KAPA RNA HyperPrep kit, KK8540). Following the manufacturer's protocol, RNA in 10 μL nuclease-free water was mixed with 10 μL 2× Fragment, Prime, and Elute Buffer. The reaction was heated at 94°C for 8 minutes and quenched on an ice-water bath. The first strand synthesis was performed by adding 10 μL Master Mix (KAPA Script enzyme and buffer) and incubated at 25°C for 10 minutes, 42°C for 15 minutes, 70°C for 15 minutes, and 4°C hold. Then, 30 μL second strand synthesis and A-tailing mixture was added to the reaction and incubated at 16°C for 30 minutes, 62°C for 10 minutes, and 4°C hold. Next, Illumina adaptors were added and ligated at 20°C for 15 minutes. After two rounds of postligation clean-up using KAPA Pure Beads, the elution was PCR amplified and purified by 1× KAPA Pure Beads. Finally, the library was subjected to Illumina NextSeq 500 PE42 platform. Sequence reads from RNA-seq were aligned to genomic sequences using TopHat. Gene expression analysis was performed using Cufflinks as described previously (21).

P. Ji is a consultant/advisory board member for Longbiopharma. R.L. Levine is a member of the supervisory board at Qiagen; a member of the scientific advisory board at Loxo (Lilly), C4, and Isoplexis; a consultant at Novartis, Janssen, Roche, and Astellas; reports receiving commercial research grants from Roche and Prelide, and has ownership interest (including stock, patents, etc.) in Qiagen, Isoplexis, and Auron. U. Steidl reports receiving commercial research grants from GlaxoSmithKline, BayerHealthcare, and Aileron Therapeutics, has ownership interest (including stock, patents, etc.) in Stelexis Therapeutics, and is a consultant/advisory board member for Celgene, Aileron Therapeutics, Stelexis Therapeutics, and Pieris Pharmaceuticals. Y. Sauntharajah has ownership interest (including stock, patents, etc.) in EpiDestiny, is a consultant/advisory board member for EpiDestiny, and has received royalties paid by the University of Illinois for patents related to tetrahydrouridine and decitabine. C. He has ownership interest (including stock, patents, etc.) in Accent Therapeutics, Epican, and Wisegene and is a consultant/advisory board member for Accent Therapeutics. A. Verma reports receiving commercial research grants from GlaxoSmithKline, Eli Lilly, Medpacto, Incyte, and Bristol-Myers Squibb, has ownership interest (including stock, patents, etc.) in Stelexis Therapeutics, and is a consultant/advisory board member for Celgene, Acceleron, Stelexis Therapeutics, and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.J. Jeong, L.A. Godley, A. Verma, A. Wickrema

Development of methodology: J.J. Jeong, J. Nie, T.D. Bhagat, A.F. List, A. Verma, A. Wickrema

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.J. Jeong, X. Gu, S. Sundaravel, W.-L. Kuo, T.D. Bhagat, J. Cao, S. Nischal, K.L. McGraw, A. Artz, M.J. Thirman, A. Moliterno, P. Ji, L.A. Godley, Y. Sauntharajah, C. He, A. Verma

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.J. Jeong, S. Sundaravel, W.-L. Kuo,T.D. Bhagat, K. Pradhan, J. Cao, R.L. Levine, L.A. Godley, U. Steidl, Y. Sauntharajah, C. He, A. Verma, A. Wickrema

Writing, review, and/or revision of the manuscript: W.-L. Kuo, T.D. Bhagat, S. Bhattacharyya, M.R. Bishop, A. Moliterno, R.L. Levine,L.A. Godley, J.J. Bieker, Y. Sauntharajah, A. Verma, A. Wickrema

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Liu, W.-L. Kuo, T.D. Bhagat, A. Verma

Study supervision: T.D. Bhagat, A. Verma, A. Wickrema

Other (reagent support): J.J. Bieker

We thank Dr. Michael Ford at MS Bioworks for providing excellent technical support and helpful discussions. We also wish to thank Dr. Subhradip Karmakar for technical support with immunoprecipitation studies, Dr. Hyun Lee at the Biophysics Core Facility at University of Illinois, and Dr. Michael Thirman for providing patient samples. S. Sundaravel is supported by NCI F99/K00 (CA223044) award. This work was supported in part by MSKCC Support Grant/Core Grant P30 CA008748 (to R.L. Levine). This work was supported in part by the Evans Foundation, Harith Foundation, Suhith Wickrema Memorial Fund, and the Leinbach Family Fund.