Substitution of Pseudokinase Domain Residue Val-617 by Large Non-polar Amino Acids Causes Activation of JAK2*

Explaining the uniqueness of the acquired somatic JAK2 V617F mutation, which is present in more than 95% of polycythemia vera patients, has been a challenge. The V617F mutation in the pseudokinase domain of JAK2 renders the unmutated kinase domain constitutively active. We have performed random mutagenesis at position 617 of JAK2 and tested each of the 20 possible amino acids for ability to induce constitutive signaling in Ba/F3 cells expressing the erythropoietin receptor. Four JAK2 mutants, V617W, V617M, V617I, and V617L, were able to induce cytokine independence and constitutive downstream signaling. Only V617W induced a level of constitutive activation comparable with V617F. Also, only V617W stabilized tyrosine-phosphorylated suppressor of cytokine signaling 3 (SOCS3), a mechanism by which JAK2 V617F overcomes inhibition by SOCS3. The V617W mutant induced a myeloproliferative disease in mice, mainly characterized by erythrocytosis and megakaryocytic proliferation. Although JAK2 V617W would predictably be pathogenic in humans, the substitution of the Val codon, GTC, by TTG, the codon for Trp, would require three base pair changes, and thus it is unlikely to occur. We discuss how the predicted conformations of the activated JAK2 mutants can lead to better screening assays for novel small molecule inhibitors.

Janus kinases (JAKs) 6 are a family of receptor-associated protein tyrosine kinases involved in a variety of downstream cellular signaling pathways including the JAK/STAT pathway, crucial in cytokine signaling. There are four family members in mammals, JAK1, JAK2, JAK3, and TYK2, as well as homologues in birds, fish, and insects (1).
Based on their primary structure, members of the JAK family are characterized by the presence of seven regions of conserved homology, denoted JAK homology (JH) domains JH1-JH7 (2). The amino-terminal JH7-JH5 domains make up the FERM domain, critical for the association of JAKs to their receptors and in some cases for receptor cell-surface localization (3)(4)(5). The JH4 -JH3 domains display some homology to classical SH2 domains but their function is unclear (6). The carboxyl terminus contains the pseudokinase and kinase domains, JH2 and JH1, respectively. The kinase domain of JAK2 consists of a nucleotide-binding loop, a catalytic loop, and an activation loop that houses the two autophosphorylation sites, Tyr-1007 and Tyr-1008, of which Tyr-1007 is crucial for activation (7). The JAK2 JH2 domain displays a similar architecture to JH1 but lacks several critical residues characteristic of a functional kinase, such as the conserved aspartic acid residue involved in the phosphotransfer reaction and the third glycine in the GXGXXG motif of the nucleotide-binding loop (8).
The JAK2 V617F mutation in the JH2 domain is present in Ͼ95% of polycythemia vera (PV) patients and in 50% of essential thrombocythemia and idiopathic myelofibrosis patients (9 -12). The mutation leads to constitutive/dysregulated kinase activity (13), which activates the JAK2 signaling pathway (9). Efforts to understand the details of JAK2 V617F activation have been hampered by the lack of a full three-dimensional structure for any Janus kinase. Functional studies had previously suggested an inhibitory effect of the pseudokinase domain on the kinase domain of JAK2 (14). Structural predictions pointed to the ␤4 -␤5 loop of JH2 as a requirement for inhibition (15). Consequently, it was proposed that the V617F mutation diminishes some of this inhibition and induces constitutive JAK2 V617F activity (16). It is noteworthy that the homologous V617F mutations also activate JAK1 and TYK2, the other Janus kinases with conserved valine residues at this position (16).
Physiologically, the signaling cascade transduced by JAK2 upon ligand binding to cytokine receptors, such as the erythropoietin receptor (EpoR), is kept in check by activation of suppressors of cytokine signaling (SOCS) proteins. For example, SOCS1 and SOCS3 bind to the catalytic groove of JAK2 through their kinase inhibitory region, thus inhibiting catalytic activity and marking the kinase for ubiquitination and degradation (17). This physiologic negative regulation was shown to be impaired for the JAK2 V617F mutant. Instead of down-modulating JAK2 V617F, SOCS3 protein actually enhanced its activity, leading to enhanced tyrosine phosphorylation of SOCS3 itself, which impairs its ability to target the kinase for degradation (18).
Here we utilized a random mutagenesis approach to screen all other possible 18 mutations at position 617 of JAK2 and address the question of why we only detect the Val to Phe mutation in patients. We identified four novel mutants that were able to support cytokine-independent growth of Ba/F3 cells expressing the erythropoietin (Epo) receptor. Only one mutant, JAK2 V617W, was comparable in signaling strength with JAK2 V617F, and only this mutant could phosphorylate and stabilize SOCS3, in a similar manner to JAK2 V617F. Furthermore, mice reconstituted with bone marrow cells transduced with JAK2 V617W displayed a myeloproliferative disease predominating on both erythroid and megakaryocytic lineages. Based on our results on the novel active JAK2 mutants, we propose a model for the pathologic activation of JAK2 by the V617F mutation. We suggest that although JAK2 V617W would most probably be pathogenic in patients, it may not occur naturally due to the necessity of three base pair changes to obtain the substitution of valine to tryptophan.

EXPERIMENTAL PROCEDURES
Generation of JAK Mutants-Several mutagenesis reactions with degenerate primers were carried out on murine JAK2 WT with the QuikChange site-directed mutagenesis kit (Stratagene) until the 19 V617X mutants (where X is any amino acid) were obtained. All constructs were cloned into the bicistronic retroviral vector pMX-IRES-GFP and verified by sequencing. Human JAK2 WT-pMEGIX was a kind gift of Dr. William Vainchenker, Institut Gustave Roussy, Villejuif, France. Murine JAK1 and human JAK2 mutants were made with complementary forward and reverse primers using the same kit. They were cloned into the vectors pMX-IRES-GFP and pMEGIX-IRES-GFP, respectively, and verified by sequencing.
Cell Lines and Retroviral Transductions-␥-2A are JAK2-deficient human fibrosarcoma cells. The JAK1-deficient U4C human fibrosarcoma cells (19) were a kind gift of Dr. Ian Kerr, Imperial Cancer Research Fund, London. Ba/F3-EpoR cells are murine interleukin-3-dependent cells that are expressing the murine erythropoietin receptor. WT and mutant JAKs were transfected into BOSC packaging cells to produce retroviruses that were subsequently used to infect Ba/F3-EpoR cells as described (20). GFP-positive cells were sorted 72 h after infection, washed, and cultured in the absence of cytokines. Cell numbers were recorded with a Coulter cell counter.
Dual Luciferase Assays-The STAT5 transcriptional activity of the various mutants was measured in ␥-2A cells (fibrosarcoma cells deficient in JAK2) and U4C cells (fibrosarcoma cells deficient in JAK1) by Dual-Luciferase assays with the STAT reporter, pGRR5-Luc (21). Cells were seeded in 24-well plates overnight and transfected using Lipofectamine, with pGRR5, STAT5 (or STAT3 for U4C cells), the cDNA coding for each individual JAK mutant, and pRLTK-Luc as an internal control. Medium was changed 4 and 24 h after transfection. The cells were lysed 48 h after transfection, and luminescence was recorded on a TD-20/20 or Glomax 96-well plate luminometer. When performing the assay on stably transduced sorted or selected the Ba/F3-EpoR cells (expressing each JAK2 mutant), cells were starved overnight in RPMI medium with 1 mg/ml bovine serum albumin and electroporated the next day with the pGRR5 and pRLTK luciferase reporters. The cells were subsequently cultured for 2 h and lysed in 100 l 1ϫ passive lysis buffer, and their luminescence was recorded.
In Vivo Reconstitution of Mice-Murine WT or V617W JAK2 cDNAs were cloned in pMEGIX. Virus was produced in 293 Epstein-Barr virus nuclear antigen cells. Bone marrow cells were collected from C57BL/6J mice 4 days after 5-fluorouracil treatment, infected with virus, and injected intravenously into lethally irradiated mice. Hematocrit levels were determined 4 weeks after transplantation. Peripheral blood cell counts were recorded using a Melet-Schloesing Laboratories cell counter, and blood smears were analyzed after May-Grünewald-Giemsa staining. Histopathological analyses were performed after 2-4 months. Paraffin-embedded sternums and spleens were stained with hematoxylin-eosin and Giemsa. Reticulin fibers were revealed by silver staining according to the Gordon Sweet method (32), and collagen was revealed by trichrome staining.

The V617F Mutation Is Only One of Several Mutations at Position Val-617 That Can Lead to Constitutive
Signaling by JAK2-The V617F mutation in JAK2 found in PV and other myeloproliferative disorders replaces an aliphatic Val residue with a bulky aromatic Phe residue. This substitution in the pseudokinase domain induces constitutive activity of the kinase presumably by eliminating the negative regulation of the JH2 domain on the JH1 (15). The question we asked was whether any of the other 18 natural amino acids could induce a similar effect as V617F when substituted at position 617.
Using random mutagenesis with degenerate primers, we generated constructs containing each of the 20 natural amino acids at position 617 of murine JAK2 and cloned them into the bicistronic vector, pMX-IRES-GFP. 293T-derived BOSC cells were used to produce retroviral supernatants that were subsequently utilized to infect murine Ba/F3 cells expressing the Epo receptor. The cells were sorted for similar GFP levels 72 h later.
Proliferation of sorted Ba/F3-EpoR cells expressing the 20 JAK2 mutants was monitored over a period of 7 days, and the mutants able to grow in the absence of cytokines were selected and counted. JAK2 mutants V617I, V617L, V617M, and V617W, in addition to the V617F positive control, were able to render Ba/F3-EpoR cells autonomous (selected) under these conditions (Fig. 1A). V617C could also eventually induce autonomous growth, but this required a longer delay, and growth was established slowly, suggesting that further selection was required to obtain complete cytokine independence, relative to the other four mutants. The cells expressing the other 13 mutants and the WT JAK2 were unable to proliferate in the absence of growth factors (data not shown), suggesting that the selected mutants were constitutively active. The V617W mutant reproducibly exhibited the shortest growth lag time of all the mutants, proliferating faster than the V617F mutant (Fig.  1A). Lysates were prepared after cell sorting and subsequently after selection of these sorted cells for autonomous growth, and the levels of JAK2 protein were compared for each mutated JAK2 in these two conditions. In all cases except V617W, and to a lesser extent V617F, selection for autonomous growth after cell sorting led to an increase in protein levels of the respective JAK2 V617X mutant (Fig. 1B). The JAK2 V617X mutants that could not support cytokine-independent growth were stably expressed in sorted Ba/F3-EpoR cells at levels similar to WT (Fig. 1C).
We quantified the STAT5 transcriptional activity of the 20 mutants by dual luciferase assays in JAK2-and STAT5-deficient ␥-2A fibrosarcoma cells. For each sample, we transfected cDNAs coding for the individual JAK2 mutant, STAT5A, the general STAT reporter pGRR5-Luc (21), and the luciferase reporter pRLTK-Luc as an internal control. Generally, the mutants that could proliferate in the absence of cytokines were able to induce much higher STAT5 transcriptional activity than WT ( Fig. 2A). V617I, V617L, and V617M induced a STAT5 activity five times higher than WT, whereas V617W enhanced STAT5 activity almost 10 times and approximately double that of V617F. On the other hand, the V617C mutant induced a STAT5 transcriptional activity comparable with the level induced by overexpressed WT JAK2. This mutant was also quite weak in promoting autonomous growth, which required a long lag period (Fig. 1A). Additionally, we have performed the luciferase assays by transfecting the cDNA coding for STAT5B instead of STAT5A and noticed similar results (supplemental Fig. 1). We also measured the activity of V617C with pLHRE-Luc, a luciferase reporter that responds preferentially to STAT5 (24), and pGL3bPpr2-Luc (25), a reporter that responds to STAT3, and still could not detect a transcriptional level higher than WT (data not shown), suggesting either a very low or a very transient activity of V617C.
We also carried out luciferase assays to study STAT transcriptional activation directly in the Ba/F3-EpoR cells stably expressing each mutant, which were growing autonomously (without cytokines), which we denote as "selected." Cells were washed, incubated without serum overnight, and electroporated with the pGRR5 and pRLTK reporters. As a negative control, we used sorted Ba/F3-EpoR cells expressing WT JAK2. Selected cells expressing the V617I, V617L, V617M, and V617W mutants all displayed STAT5 transcriptional activity well over that of WT (Fig. 2B). We then performed a Western blot on these cells and probed them with antiphospho-STAT5 (STAT5 phosphorylated at tyrosine 694) and anti-STAT5 antibodies. The STAT5 activity observed in the luciferase assay was confirmed in the Western blot as the V617W mutant exhibited a high pSTAT5 level, comparable with V617F. V617I, V617L, and V617M all demonstrated a phospho-STAT5 level higher than WT but lower than V617F and V617W (Fig. 2C). The same samples were also probed with anti-JAK2 antibodies. The lower levels of JAK2 V617F and especially of V617W may reflect their strength of signaling as these mutants do not require high protein levels for activity (Fig. 2C). This is consistent with the levels of V617F and V617W we observe in Fig. 1B.
We were interested in determining whether the activating mutations at position 617 of murine JAK2 would have similar effects in human JAK2. We generated human JAK2 V617F, V617I, V617L, V617M, V617W, and V617C mutants through site-directed mutagenesis. A luciferase assay in selected Ba/F3-EpoR cells expressing the WT and mutant human JAK2 constructs revealed a similar activity as in the luciferase assay of selected murine cells, with notably low STAT5-dependent transcriptional activity induced by the human JAK2 V617C, which was also much weaker in inducing autonomous growth (data not shown).
To further compare the activities of the human JAK2 mutants, we performed a Western blot on Ba/F3-EpoR cells expressing each human JAK2 Val-617 mutant, previously selected for autonomous growth. To monitor activation, we used antibodies that specifically detect the phospho-tyrosines that are required for activation of JAK2 (Tyr-1007). The phospho-JAK2 levels were also significantly higher in V617W, similar to V617F. The pJAK2 levels of V617C, V617I, and V617L mirrored the luciferase assay results, pointing to these mutants being weaker than V617F and V617W (Fig. 3A). Western blot analysis with antiphospho-Tyr-694 of STAT5 confirmed the luciferase results as the active mutations displayed a level of STAT5 phosphorylation higher than WT (data not shown).
JAK2 V617W Induces Strong Tyrosine Phosphorylation of SOCS3-We have previously shown that the V617F mutant not only escapes negative regulation by SOCS3 but that SOCS3 may even enhance its phosphorylation, constitutive activity, and ability to promote proliferation. Rather than being degraded, SOCS3 is phosphorylated and stabilized in the presence of JAK2 V617F (18). We wanted to test the effect of SOCS3 on the other active murine JAK2 V617X mutants. In 293T cells transiently transfected with each mutant, we determined that only JAK2 V617F and JAK2 V617W strongly induced tyrosine phosphorylation of SOCS3 ( Fig. 3B and supplemental Fig. 2). SOCS3 actually stabilized the phosphorylated form of JAK2 V617F, and to a lesser extent that of JAK2 V617W, but not that of the other mutants. These results are in agreement with those previously reported for JAK2 V617F (18). Taken together, these data suggest that the one mutant that appears to mimic the effects of JAK2 V617F is V617W as it is the strongest in various biologic or signaling assays and overcomes inhibition by SOCS3. That is the reason why we examined the effects of this mutant in vivo.

JAK2 V617W Induces a Myeloproliferative Disease Predominating on Erythroid and Megakaryocytic
Lineages in Mice-To examine the effect of the V617W mutation in vivo, we reconstituted mice with this mutant and assessed their phenotype 4 weeks after transplantation. The mice expressing JAK2 V617W developed erythrocytosis, with a hematocrit Ͼ50% (Fig. 4A). This phenotype, reminiscent of the one present in PV, was similar to the one obtained in mice with the V617F mutation (9,26,27). The V617W mice also had a secondary excess of platelets, and to a lesser extent, of granulocytes in the blood. The bone marrow was characterized by a significant increase in the number of megakaryocytes and an increased ratio of myeloid to erythroid precursors. The enlarged spleen was associated with an alteration of the normal architecture due to an expansion of red pulp with maturing myeloid elements, megakaryocytes, and erythroblast progenitors. In both spleen and bone marrow sections, megakaryocytes appeared larger, atypical, with emperipolesis in megakaryocyte cyto-plasm and frequently clustered. Silver staining showed a moderate increase in reticulin fibers in bone marrow as well as in spleen (Fig. 4B). No extramedullary hematopoiesis was detected in liver sections after 4 months.
The Homologous V617W Mutation in JAK1 (V658W) Induces Constitutive Signaling-Finally, in a previous report (16), we have shown that the JAK2 V617F homologous mutations, JAK1 V658F and TYK2 V678F, induce constitutive activation of their respective kinases. Our data pointed to a potential similar mechanism of activation of these three mutated kinases (16).
Here we wanted to determine in JAK1 the effect of our weakest and strongest JAK2 mutants, V617C and V617W, respectively. We transiently transfected JAK1-deficient U4C fibrosarcoma cells with either the cDNA coding for JAK1 V658C or the cDNA FIGURE 2. The selected cytokine-independent murine JAK2 (mJAK2) V617X mutants display high levels of STAT5 transcriptional activity. A, JAK2-deficient ␥-2A cells, which are also deficient in STAT5, were co-transfected with cDNAs coding either for JAK2 WT or for each V617X mutant individually, along with STAT5A, and the pGRR5-Luc and pRLTK-Luc reporters. The mutants that were able to proliferate in the absence of cytokines when expressed in Ba/F3-EpoR cells displayed a STAT5-dependent transcriptional activity higher than that of the WT (average of three replicates Ϯ S.D.). Shown is one representative experiment out of four independent experiments. B, selected Ba/F3-EpoR cells expressing each murine active mutant were starved overnight in RPMI medium with 1 mg/ml bovine serum albumin and electroporated with pGRR5-Luc and pRLTK-Luc the next day. The cells were lysed, and their luminescence was recorded. All five mutants, although to a lesser degree V617C, generated a high STAT5 activity. One of three independent experiments is depicted. C, tyrosine phosphorylation was examined in selected Ba/F3-EpoR cells expressing each murine active JAK2 mutant, which were lysed and probed with anti-pSTAT5 and anti-STAT5 antibodies, respectively. pSTAT5 levels were high for the selected mutants, whereas the levels for WT JAK2 and V617C were undetectably small. The JAK2 levels for these mutants were in agreement with the strength of their activity. JAK2 V617F and V617W displayed lower JAK2 levels as they are the mutants with the highest activity. coding for V658W and examined STAT3 transcriptional activity of these mutants. JAK1 V658W mutant displayed a high STAT3 activity, whereas V658C had a STAT3 activity comparable with WT JAK1 (Fig. 5). These findings support the notion that that JAK2 and JAK1 share common conformations that allow them to be regulated similarly by mutants at the homologous 617 position.

DISCUSSION
We set out to identify other mutations at position 617 of JAK2 that would constitutively activate signaling in a manner similar to the V617F mutation present in myeloproliferative diseases. We hypothesized that discovery and characterization of other activating mutations at this position would explain the uniqueness of the Val to Phe mutation in this disease and provide some structural clues to its mechanism of activation.
Our main finding is that, in addition to V617F, several other mutations at this position can, to various extents, induce constitutive signaling. These are V617I, V617L, V617M, and V617W. Among them, only V617W appears to be strong enough to induce a level of STAT5 activation comparable with V617F. In addition, only V617W (and V617F) is resistant to the physiologic down-modulating action of SOCS3 and stabilizes phosphorylated SOCS3 protein. In contrast, V617C appears to induce a partial activation that requires long delays and presumably other events before establishing cytokine-independent growth. Results we present are obtained in Ba/F3-EpoR cells, where JAK2 V617F expression rapidly induces autonomous growth. In addition, similar qualitative results were obtained by expression of the mutated JAK2 proteins in parental Ba/F3 cells (data not shown), except that longer lag periods were necessary to obtain autonomous growth. These differences have recently been addressed for JAK2 V617F and Ba/F3 cells and were shown to be due to limiting amounts of endogenous receptors required to recruit JAK2 to the plasma membrane and promote signaling (28). That is the reason why we decided that a study of signaling by JAK2 mutated proteins would be more informative in Ba/F3-EpoR cells.
The only available structure for the combined JH1 and JH2 domains of JAK2 is a homology model in which an interface of interaction between the JH1 and JH2 is proposed, involving residues in the activation loop of JH1 and a loop connecting ␤-strands 4 and 5 of JH2 (15). The first residue of this loop is Val-617. The authors suggest two different types of interaction occurring between JH1 and JH2 in this particular interface. In its inactive form, the activation loop of the kinase domain is pulled back, blocking ATP and substrate access. This conformation is held together by stabilizing interactions between residues at the base of the activation loop and residues around Val-617. Upon ligand binding, receptors undergo a conformational change, such as a rotation of the juxtamembrane domains that bind JAK2 (20), and as a result, the two JAKs appended to each receptor monomer can transphosphorylate. This conformational change forces the activation loop out,  freeing up access to the nucleotide-binding and catalytic loops, and initiates the signal transduction cascade. In this active conformation, the base of the JH1 activation loop is predicted to make unfavorable interactions with residues in the vicinity of Val-617 of JH2 (15).
We previously proposed that the V617F mutant diminishes the putative inhibition of the JH1 kinase domain by the JH2 pseudokinase domain (16). Now we show that such an effect can be obtained by bulky non-polar residues. More specifically, the nature of the four Val-617 mutants that constitutively activated JAK2 points to a hydrophobic, side-chain length dependence for activation (Fig. 6, A and B), presumably by diminishing the favorable interactions between the ␤4 -␤5 loop (in JH2) and the base of the activation loop (in JH1), which were suggested to maintain JH1 inactive (15). The inability of "shorter" hydrophobic non-polar residues with a one-carbon-long side chain, such as Gly, Ala, or Pro, to constitutively activate JAK2 supports this. The fact that both Phe and Trp strongly activate the kinase, whereas Tyr and His do not, suggests that a hydrophobic interaction is necessary for activation of JAK2.
Although JAK2 V617F is present in Ͼ95% of PV patients, the rest of the PV patients appear to harbor a number of exon 12 mutations that clustered around Lys-539, in the region that links the SH2 and JH2 domains (29). Molecular modeling suggests that the K539L mutations would be placed in a loop very close to the loop that contains the Val-617 residue, thus implicating a similar conformation as that of V617F as a mechanism of activation (29).
Taken together, these results suggest that the pseudokinase domain mutants stabilize an opened conformation of the activation loop. Since JAK2 V617F is active in the absence of ligand stimulation of receptors, which would be predicted to modify the relative positions of the JAK molecules, it is possible that the conformation of the activation loop in the JAK2 V617F differs from that of activated JAK2 WT. The implication of such a difference is that appropriate screening systems that discriminate between ligand-activated WT JAK2 and JAK2 V617F might lead to small molecules that specifically target JAK2 V617F. This would be beneficial for patients since inhibiting the wild type JAK2 might induce unwanted effects, such as blocking red blood cell formation or the functions of several other cytokines, such as interferon-␥.
Our results also provide a possible explanation for the exclusivity of the phenylalanine substitution of Val-617 in PV patients. Out of the five Val-617 mutants that constitutively activated JAK2, only V617W was comparable with V617F, as measured by proliferation assays, JAK2 phosphorylation levels, and transcriptional assays. However, the sole codon that codes for Trp, TGG, can only be obtained by substitution of three nucleotides from the valine codon, GTC. This event is clearly less likely than the one-nucleotide substitution required to obtain Phe, TTC. Moreover, we could show that also similar to V617F, the Trp mutant escapes negative regulation by SOCS3, whereas the other activating mutants were not able to do so. Tyrosine phosphorylation of the SOCS3 SOCS box impairs its ability to recruit ubiquitin ligase (E3) and thus to down-modulate JAK2 and itself (30). It has also been proposed that one mechanism by which SOCS3 inhibits activity of JAK2 is through binding with its kinase inhibitory region to the catalytic groove of JAK2 and acting as a pseudosubstrate (31). It is possible that having Trp or Phe at position 617 of JAK2 changes the conformation of the structure immediately around it and diminishes the inhibitory effect of the kinase inhibitory region.
In conclusion, through the identification and characterization of several novel JAK2 V617X activating mutants, we propose that the mechanism of activation of JAK2 V617F involves a hydrophobic, side-chain length-dependent effect that would destabilize the physiologic inhibition exerted by JH2 on JH1. This type of activation mechanism may also function in JAK1 and TYK2, as our results with JAK1 V658W suggest, and as TYK2 with a homologous V617F mutation is constitutively active (16). The corroboration of these genetic and functional data on various JAK mutants with structural models may facilitate the isolation of specific small molecule inhibitors that would target pathogenic JAK mutants and spare the wild type JAKs. Such inhibitors would likely have to be non-ATP competitors, targeting hydrophobic pockets other than that of ATP. The results presented in this report describing novel mutants of JAK2 indicate that hydrophobic, side-chain length-dependent activation of JAK2 may be a useful target for inhibitor screening.