Role of JAK-STAT Signaling in the Pathogenesis of Myeloproliferative Disorders

More than a half-century ago Dameshek classified polycythemia vera (PV), essential thrombocytosis (ET) and myelofibrosis (MF) as phenotypically related myeloproliferative disorders (MPD);¹ however, the molecular etiology of these disorders remained elusive for more than five decades. Our understanding of the genetic basis of these disorders was greatly enhanced in 2005, when four groups identified a single recurrent MPD-associated mutation in JAK2 (Table 1 ).^2,–5 The guanine to thymine mutation results in a substitution of valine to phenylalanine at codon 617 within the pseudokinase domain (JH2) of JAK2 (JAK2V617F). Analysis of germline DNA demonstrated that JAK2V617F is a somatic mutation in hematopoietic progenitors.³ A subset of patients, most commonly with PV, have homozygous JAK2V617F mutations, which are the result of mitotic recombination and duplication of the mutant allele.^2,–5 Since the initial reports of this mutation, these findings have been confirmed by many different groups.^6,–8 In addition, the development of more sensitive assays, including allele-specific PCR,⁵ pyrosequencing,⁶ and real-time quantitative PCR,⁹ has allowed investigators to determine more precisely the frequency of the JAK2V617F allele in PV, ET, and MF (Table 1 ). Using these more sensitive assays JAK2V617F can be detected in approximately 90–95% of patients with PV, 50–70% of patients with ET, and 40–50% of patients with MF. These data suggest rare patients with PV are JAK2V617F negative, whereas a significant proportion of patients with ET and MF lack the JAK2V617F mutation.

The JAK2V617F allele has also been observed in a small number of patients with chronic myelomonocytic leukemia (CMML), myelodysplastic syndrome (MDS), and acute myeloid leukemia (AML), though most JAK2V617F mutations in AML occur in patients with a preceding diagnosis of PV, ET, or MF.^6,10,11 Despite evidence that JAK-STAT pathway activation is common in both hematopoietic malignancies and solid tumors,^12,JAK2V617F is exclusive to disorders of the myeloid lineage and has not been observed in lymphoid neoplasms or in non-hematopoietic malignancies.^10,13 

Despite these important initial observations many questions remain regarding the role of JAK2V617F in the pathogenesis of PV, ET, and MF, and regarding the etiology of PV, ET, and MF not associated with JAK2V617F mutations. First, why does one single recurrent mutation occur at such a high frequency in myeloid disorders, but not in other malignancies? Given that other mutations, including the T875N substitution observed in an acute megakaryoblastic leukemia cell line,¹⁴ can constitutively activate JAK2, why does a single mutation activate JAK2 in the majority of patients with these disorders? Finally, how does a single mutation contribute to the pathogenesis of three phenotypically distinct MPD?

Clonality studies using X-inactivation analysis demonstrated that the majority of patients with MPD had clonal populations of myeloid and erythroid cells,^15,–17 suggesting that these disorders arise in hematopoietic progenitors. In addition, loss of heterozygosity at 9p24, now known to correspond to homozygous JAK2V617F mutations, can be identified in both myeloid and lymphoid cells in some patients with PV, further suggesting the causal mutations occur in progenitor cells with the ability to differentiate into different lineages.¹⁸ However, whether these disorders originate in hematopoietic stem cells (HSCs) or in downstream multilineage progenitors was not clarified until recently. The advent of mutiparameter flow cytometry has allowed researchers to identify and characterize specific progenitor populations,^19,20 and Jamieson et al used this approach to isolate HSCs, common myeloid progenitors (CMPs), granulocyte/macrophage progenitors (GMPs), and megakaryocytic/erythroid progenitors (MEPs) from patients with PV and then analyzed these progenitor populations for the presence of the JAK2V617F mutation.²¹ They detected JAK2V617F in HSCs, CMPs, GMPs, and MEPs from patients with PV, providing evidence that PV is a disorder that arises in HSCs and involves the myeloid, erythroid, and megakaryocytic lineages. These data are in congruence with a recent report by Ishii et al, in which JAK2V617F could be detected in different hematopoietic lineages in patients with PV,²² although the mutation could only be detected in lymphoid cells in a minority of patients. It is likely the long-term self-renewing potential of HSCs is required for disease propagation, and that JAK2V617F provides a proliferative advantage to hematopoietic cells but does not confer self-renewal properties to committed progenitors. These data parallel previous data demonstrating that BCR-ABL can transform HSCs but not committed progenitors,²³ and it will be important to determine if the same is true for JAK2V617F. It will also be important to determine if JAK2V617F is present in both HSCs and committed progenitors derived from patients with ET and MF, in order to exclude the possibility that differences in the cell of origin explain the phenotypic pleiotropy of JAK2V617F-positive MPD. In addition, transplants into NOD-SCID mice will determine whether clonal JAK2V617F-positive HSCs are capable of multi-lineage engraftment, as has been demonstrated for CD34⁺ cells from patients with MF.²⁴ 

Valine 617 is within the pseudokinase domain (JH2) of JAK2; this domain has significant homology to the kinase domain of JAK2 (JH1) but lacks catalytic activity. As deletion of JAK2 JH2 leads to increased JAK2 kinase activity,²⁵ it is thought that the pseudokinase domain serves an autoinhibitory role similar to the juxtamembrane domain of receptor tyrosine kinases such as FLT3.²⁶ Though it is likely that substitution of valine to phenylalanine results in a loss of autoinhibition, structural insight into how the JAK2 kinase and pseudokinase domains interact will be required to understand how the valine to phenylalanine substitution leads to constitutive kinase activity. Biochemical evidence also supports the notion that JAK2V617F is a constitutively active tyrosine kinase. When expressed in 293T cells, JAK2V617F, but not wild-type JAK2, is strongly autophosphorylated even when expressed at low levels, consistent with constitutive kinase activity.³ In addition, in vitro kinase assays with JAK2V617F and wild-type JAK2 showed that JAK2V617F has greatly increased kinase activity, as assessed by autophosphorylation and by substrate phosphorylation.⁸ 

When expressed in Ba/F3 cells co-expressing the erythropoietin (EPO) receptor (Ba/F3-EPOR), expression of JAK2V617F resulted in EPO independent growth and in EPO hypersensitivity,³ similar to what is observed for erythroid colonies from patients with PV.²⁷ Given that EPOR is a homodimeric type I cytokine receptor expressed on cells of the erythroid lineage, it was logical to test whether JAK2V617F could signal when co-expressed with other homodimeric type I cytokine receptors important in myeloid differentiation. Co-expression of JAK2V617F with the thrombopoietin receptor (MPL) or the granulocyte colony-stimulating factor receptor (GCSFR) transformed cells to factor independent growth and led to constitutive phosphorylation of STAT5, an important downstream effector of JAK2 signaling.²⁸ Although this does not rule out the possibility that JAK2V617F can interact with non-homodimeric type I cytokine receptors, it suggests that the cytokine receptors relevant for red blood cell (EPOR), megakaryocytic (MPL), and granulocytic (GCSFR) differentiation are important for JAK2V617F-mediated transformation in PV, ET, and MF.

Many patients are heterozygous for JAK2V617F as assessed by DNA sequencing, suggesting that there is either a subpopulation of cells which are homozygous for JAK2V617F mixed with wild-type cells, or a clonal population of cells with one wild-type copy of JAK2 and one mutant copy of JAK2. In PV, data from clonality/quantitative JAK2V617F assessment⁹ and from colony assays²⁹ suggest that most PV patients have a subpopulation of cells homozygous for JAK2V617F, whereas in ET clonal cells are heterozygous for JAK2V617F. It is therefore important to determine whether the wild-type allele can interfere with the ability of JAK2V617F to constitutively signal in the heterozygous state, and whether there is an effect of gene dosage on activation of signal transduction. Transient co-expression of wild-type JAK2 does not interfere with the ability of JAK2V617F to autophosphorylate, even when wild-type JAK2 is expressed at a higher level than the mutant kinase.³ This suggests that JAK2V617F kinase activity is unaffected by co-expression of wild-type JAK2. In contrast, when JAK2V617F and wild-type JAK2 were co-expressed in Ba/F3 cells, cytokine-independent growth was attenuated, suggesting in this cellular context that wild-type JAK2 is able to interfere with JAK2V617F-mediated transformation.² Taken together, these experiments do not conclusively demonstrate whether heterozygous JAK2V617F can activate signal transduction, as it is hard to precisely quantify the relative expression of both alleles when they are overexpressed in cell culture systems. Ultimately murine models using knock-in strategies will allow delineation of the effects of JAK2V617F gene dosage on signal transduction and on phenotype.

When expressed in vitro, JAK2V617F constitutively activates multiple signaling pathways known to be downstream of wild-type JAK2. These include STAT5, a signaling molecule which is recruited and then phosphorylated by the cytokine receptor-JAK2 dimeric complex; phosphorylated STAT5 then translocates to the nucleus and acts as a transcription factor. STAT5 is known to activate transcription of many genes important in cell proliferation and survival, including Bcl-X, an anti-apoptotic protein that is overexpressed in PV erythroid progenitors.³⁰ In addition, cells expressing JAK2V617F display constitutive activation of the MAP kinase pathway (as assessed by phosphorylation of ERK), and of the PI3K pathway (as assessed by phosphorylation of AKT). It is important to note that many other oncogenic tyrosine kinases activate the same signal transduction pathways, and the role and requirement for each of these signaling pathways in the transformation of hematopoietic cells by JAK2V617F is not known. Subsequent experiments using genetic models (e.g., mice deficient in a specific signaling molecule) or biochemical assessment (e.g., using inhibitors of specific downstream pathways) will be required to elucidate the role of the STAT, MAP kinase, and PI3K signaling pathways in the pathogenesis of PV, ET, and MF.

Although the discovery of the JAK2V617F allele was reported less than two years ago, significant insight regarding the in vivo effects of JAK2V617F has emerged from retroviral bone marrow transplant models similar to those used to prove that expression of BCR-ABL is necessary and sufficient for the development of chronic myelogenous leukemia (CML).³¹ In their initial report of the JAK2V617F allele, James et al reported that expression of JAK2V617F, but not wild-type JAK2, results in polycythemia and splenomegaly in recipient mice 28 days after transplantation, suggesting that JAK2V617F is sufficient for the development of polycythemia.² Subsequent reports have provided additional insight into the in vivo effects of JAK2V617F.^32,33 

Our group used the murine bone marrow transplant assay to assess the effects of wild-type JAK2 and JAK2V617F expression in C57Bl/6 and Balb/c recipient mice.³² Expression of JAK2V617F, but not wild-type JAK2, resulted in an oligoclonal MPD characterized by marked polycythemia (Hct > 70%) and splenomegaly in both genetic backgrounds. In some recipient mice, marked polycythemia was maintained for an extended observation period of 9 months, whereas in other recipient mice a reduction in hematocrit was observed due to progressive bone marrow fibrosis (G.W. and D.G. Gilliland, unpublished observation). Interestingly, there were differences in the effect of JAK2V617F in the different genetic backgrounds, as JAK2V617F expression resulted in marked leukocytosis and bone marrow reticulin fibrosis in Balb/c mice but not in C57Bl/6 mice, in which reticulin fibrosis was only noted in the spleen. This suggests that there are host modifiers that affect the phenotype induced by JAK2V617F expression. Although expression of JAK2V617F in murine bone marrow caused megakaryocytic hyperplasia, recipient mice did not develop thrombocytosis, likely due to defects in megakaryocytic maturation and platelet production. Activation of JAK2 and STAT5 was observed in JAK2V617F-expressing cells but not in cells expressing wild-type JAK2, demonstrating activation of signal transduction by JAK2V617F in vivo. In addition, erythroid colonies from mice transduced with JAK2V617F demonstrated EPO-independent and EPO-hypersensitive erythroid colony growth comparable to the endogenous erythroid colony formation observed with human PV blood and bone marrow cells.

Lacout et al used a retroviral bone marrow transplant assay to assess the effects of JAK2V617F/wild-type JAK2 expression on C57Bl/6 mice and then followed these mice for 6 months after transplantation.³³ Mice expressing JAK2V617F developed polycythemia and leukocytosis, peaking at 3 months post transplant, followed by a progressive reduction in cell counts that was associated with progressive fibrosis. The degree of bone marrow fibrosis was variable, and interestingly, mice with high-grade fibrosis developed anemia, thrombocytopenia, and neutrophilia consistent with post-PV MF. Again, with the exception of one mouse with transient thrombocytosis, elevated platelet counts were not observed in mice transduced with JAK2V617F.

The data from these bone marrow transplant models of JAK2V617F-mediated disease are important, as they suggest that JAK2V617F is necessary and sufficient for the development of PV. Despite these observations important questions remain regarding the in vivo effects of JAK2V617F. First, although mice transduced with JAK2V617F display certain features of MF (reticulin fibrosis) and ET (megakaryocytic hyperplasia), the phenotype of JAK2V617F in this assay is most similar to human PV. The dominant erythroid phenotype of JAK2V617F may reflect differences in the pathogenesis of the different MPD. One possibility is that PV may result from acquisition and subsequent homozygosity of JAK2V617F, whereas a second cooperating genetic event in addition to JAK2V617F is required for the development of ET and/or MF (Figure 1; see Color Figures, page 511). This hypothesis is supported by the observation that some patients with ET have monoclonal granulopoiesis even when a minority of cells are JAK2V617F positive.⁹ Alternatively, as suggested by Lacout et al,³³ gene dosage and level of expression of JAK2V617F might dictate MPD phenotype; this is supported by the observation that patients with PV, but not ET, frequently have homozygous JAK2V617F mutations. In addition, the genetic cause of the strain-specific differences in phenotype remains unknown. Finally, the effect of physiologic expression of JAK2V617F in the heterozygous and homozygous state cannot be assessed using these models, and will instead require analysis of JAK2V617F knock-in mice. Despite these limitations, these models provide a powerful model of human PV and provide an in vivo platform for preclinical testing of JAK2 inhibitors as they become available.

Despite in vitro and in vivo data suggesting JAK2V617F may be sufficient for the development of PV, there is evidence that additional genetic events may contribute to the pathogenesis of PV, ET, and/or MF. Given that the identical JAK2 allele is identified in patients with three phenotypically related, but clinically distinct, MPD, additional genetic and epigenetic events must contribute to the phenotypic pleiotropy of these disorders. In addition, recent genetic data suggest inherited alleles may contribute to the pathogenesis of MPD regardless of JAK2 mutational status. In a study by Bellane-Chantelot et al families with more than one person diagnosed with a MPD were analyzed for JAK2V617F mutations.³⁴ Somatic JAK2V617F mutations were identified in some, but not all MPD patients with a positive family history of MPD. Interestingly, in some families there were both JAK2V617F-positive and JAK2V617F-negative affected members, and a small number of relatives without a diagnosis of PV, ET or MF had endogenous erythroid colony formation in the absence of the JAK2V617F allele. These results suggest that there are unidentified mutations that contribute to the pathogenesis of PV, ET and MF, regardless of JAK2 mutational status, and there may be “initiating events” which precede the acquisition of JAK2V617F in these disorders.³⁵ 

The discovery of JAK2V617F provided important insight into the pathogenesis of PV, ET, and MF, but the etiology of JAK2V617F-negative ET and MF remained unknown. The majority of JAK2V617F-negative ET and MF have clonal granulopoiesis,^8,9 so the inability to detect mutant JAK2 in these patients is not due to a lack of granulocyte involvement. Given that co-expression of the homodimeric type I cytokine receptors EPOR, MPL, and GCSF was shown to be important for JAK2V617F-mediated transformation,²⁸ we asked if mutations in these cytokine receptors could activate JAK2 in JAK2V617F-negative MPD. There is precedent for cytokine receptor abnormalities in PV, ET, and MF, as previous studies have identified inherited EPOR and MPL mutations in familial erythrocytosis and thromb-ocytosis,^36,–40 and impaired expression of MPL has been observed in platelets from patients with MPDs.⁴¹ Sequence analysis of EPOR, TPOR and GCSFR in JAK2-negative PV, ET and MF led to the discovery of a somatic tryptophan to leucine substitution mutation at the transmembrane-juxtamembrane junction of MPL (MPLW515L).⁴² The MPLW515L allele occurs in approximately 10% of patients with JAK2V617F-negative MF, and in a smaller proportion of patients with ET; a smaller number of patients have an alternate mutation at codon 515, which results in a tryptophan to lysine substitution (MPLW515K).⁴³ This allele is not observed in PV or other myeloid malignancies, suggesting that activation of JAK-STAT signaling by MPLW515L is specific to ET and MF. Expression of MPLW515L, but not wild-type MPL, results in factor-independent growth and constitutive activation of downstream signaling proteins including the STAT3, MAP kinase, and PI3K signal transduction pathways. Expression of MPLW515L in recipient mice resulted in a myeloproliferative disease with similarities to human MF, including reticulin fibrosis, megakaryocytic hyperplasia, splenomegaly, and extramedullary hematopoiesis. In contrast to JAK2V617F, expression of MPLW515L in recipient mice resulted in marked thrombocytosis. Although both JAK2V617F and MPLW515L occur in human MF and activate similar signaling pathways, their expression in the bone marrow transplant assay results in distinct phenotypes. Subsequent work should elucidate the basis for the difference between the in vivo effects of MPLW515L and JAK2V617F. A small number of patients had both MPLW515 and JAK2V617F mutations; the significance of this observation and the potential cooperativity of MPLW515L and JAK2V617F in vitro and in vivo are under investigation. Most importantly, there are patients with ET and MF who are JAK2/MPL negative, and we predict that genomic analysis of other genes in the JAK-STAT signaling pathway will lead to the discovery of additional disease alleles in ET and MF.

Given the remarkable clinical success of imatinib for the treatment of BCR-ABL positive CML,^44,FIP1L1-PDGFRα positive hypereosinophilic syndrome (HES),⁴⁵ and CMML associated with PDGFRβ rearrangements,⁴⁶ and the rapid development of second generation BCR-ABL inhibitors for imatinib refractory CML,^47,48 it is hoped that the discovery of JAK2V617F will facilitate the development of molecularly targeted therapy for PV, ET, and MF. Although JAK kinase inhibitors with specific activity against JAK2 have previously been identified and characterized,⁴⁹ most of these compounds do not have sufficient potency (e.g., cellular IC₅₀ < 1–2 μM) to be used in preclinical and clinical testing. More recently, compounds with sufficient potency against JAK2 have been identified,⁵⁰ and we have shown that a JAK kinase inhibitor with nanomolar activity against JAK2 can potently inhibit growth and signaling in cells transformed by JAK2V617F.³ Multiple groups have begun work on JAK2 inhibitors with favorable pharmacokinetic and pharmacodynamic properties, and given that JAK3-specific inhibitors have demonstrated efficacy in preclinical models of organ allograft rejection⁵¹ it is likely that preclinical testing will identify compounds with JAK2-specific activity. As loss of function mutations in JAK3 are associated with severe combined immunodeficiency,⁵² it will be important for JAK2 inhibitors to have minimal inhibitory activity against JAK3. JAK2 inhibitors with significant inhibitory activity against JAK3 at pharmacologic doses might have suppressive effects on cell-mediated immunity, which would be undesirable given that patients with PV, ET, and MF will likely require long-term therapy with a JAK2 inhibitor.

Although JAK2 has many important functions in cell signaling, JAK2 function is indispensible for definitive hematopoiesis, and specifically for erythropoiesis.⁵³ Thus inhibition of JAK2 may cause dose-dependent cytopenias, particularly anemia, but the goal of therapy for PV is a dose-dependent reduction in hematocrit. In the case of MF, given that bone marrow transplantation that demonstrates eradication of the malignant clone results in resolution of marrow fibrosis,⁵⁴ it is hoped that specific therapy against the mutant allele(s) will ameliorate the fibrosis, extramedullary hematopoiesis, and cytopenias observed in MF. In addition, given the important role of JAK2 signaling in prolactin- and growth hormone–mediated signal transduction, it is possible that endocrine side effects might occur in patients treated with JAK2 inhibitors.

It is also important to consider whether JAK2 inhibitors should be tested against all patients with PV, ET and MF or should be limited to patients with JAK2V617F mutations. As JAK kinase inhibition has similar inhibitory effects against cells expressing MPLW515L,²⁹ it would be appropriate to offer therapy with novel JAK2 inhibitors to ET and MF patients with MPLW515 mutations. Moreover, clinical activity of JAK2 inhibitors in MPD patients without known JAK2 or MPL mutations might provide insight into the genetic basis in this subset of patients; responses to empiric imatinib therapy led to the discovery of the FIP1L1-PDGFRα fusion gene in HES.⁴⁵ 

Finally, as has been demonstrated for BCR-ABL positive CML,^55,–57 the presence of a single mutation in many patients with PV, ET, and MF will allow for assessment of molecular response to therapy. Allele-specific assays such as allele-specific PCR,⁵ pyrosequencing,⁶ and quantitative real-time PCR⁹ can precisely quantitate JAK2V617F mutational burden, and quantitative assessment of JAK2V617F mutational burden has been used to demonstrate that the JAK2V617F mutant clone persists in PV patients treated with imatinib or interferon-alpha.⁵⁸ Given the need to assess response to therapy with molecularly targeted agents, quantitative assays for JAK2V617F should be used to follow response to therapy in trials of JAK2 inhibitors in PV, ET and MF.

Although recent discoveries have provided important insight into the molecular etiology of PV, ET, and MF, many questions remain regarding the role of JAK2V617F, MPLW515L, and unidentified disease alleles in the pathogenesis of these disorders. First and foremost, the role of a single disease allele in three distinct disorders of the myeloid lineage remains elusive. Genetic models using knock-in strategies will allow an assessment of the effects of gene dosage on phenotype, and insight into the crystal structure of JAK2V617F and MPLW515L will allow us to better understand the mechanism of activation of these alleles. The genetic basis of MPD not associated with JAK2V617F or MPLW515L mutations has not been discovered, though analysis of JAK-STAT signaling components may lead to the discovery of additional mutant alleles. Most importantly, the effort being expended toward the development of JAK2 inhibitors for PV, ET, and MF has the potential to offer specific, highly effective molecularly targeted therapies for patients with these disorders.