Identification Of a Novel Mode Of Kinase Inhibitor Resistance In JAK2: JAK2 Inhibitor Resistance Is Mediated By The Generation Of 45-Kda JAK2 Variant Which Alters The Kinase Domain Structure

JAK2 V617F can be identified in the majority of cases of polycythemia vera (PV), and in 50% of cases in essential thrombocythemia (ET) and idiopathic myelofibrosis (IMF). JAK2 inhibitors including ruxolitinib (Jakavi, INCB018424), TG101348 and lestaurtinib (CEP-701) display clinical activity in clinical trials in PV, ET and IMF, and ruxolitinib has recently been approved for the treatment of primary and secondary myelofibrosis. In other malignancies it has been demonstrated that acquired resistance to kinase inhibitors is due to emergence of secondary resistance mutations in the target kinase. So far no JAK2 drug resistant mutations have been described in patients. To identify JAK2 point mutations mediating ruxolitinib resistance, we exposed JAK2 V617F expressing Ba/F3 cells to ruxolitinib. Surprisingly, sublines resistant to ruxolitinib at 1000nM, 2000nM and even 4000nM did not harbor point mutations neither in the kinase nor in the pseudokinase domain of JAK2. However, western blot analysis of sublines resistant to ruxolitinib revealed a 45-kDa JAK2 variant together with full length JAK2V617F protein in 87% of the cases. Sequencing of the short form in drug resistant clones revealed a novel JAK2 variant missing amino acids 76 to 820 resulting in the N-terminal FERM domain directly fused to the kinase domain of JAK2 (FERM-JAK2). FERM-JAK2 was highly resistant to the ATP-competitive JAK2 inhibitors ruxolitinib and TG101348. JAK2 exists in an either active or inactive state, which is largely regulated by phosphorylation of tandem tyrosines located within the activation loop. Substitution of phenylalanine at tyr1007, tyr1008 or both leads to complete inactivation of V617FJAK2 and WT-JAK2.In contrast, Y to F mutation of tyr1007, tyr1008 or both in FERM-JAK2 does not prevent activation of STAT5 and transformation of Ba/F3 cells, suggesting that FERM-JAK2 preferentially exist in an inactive state. This would impede drug binding and explain the resistant phenotype of FERM-JAK2.

Phospho-deficient mutant studies further provided evidence that tyr866, tyr966, and tyr972 are crucial for the activation of FERM-JAK2, whereas in V617FJAK2 and WT-JAK2 tyr972, tyr1007 and tyr1008 were critical autophosphorylation sites. Ectopic expression of FERM-JAK2 in HEK293T, NIH3T3 and Ba/F3 cells showed activation of STAT5 and transformation of Ba/F3 cells to factor independence. Interestingly, co-immunoprecipitation studies revealed that FERM-JAK2 does not bind or activate the IL-3R in contrast to V617FJAK2. Ectopic expression of FERM-JAK2, in contrast to V617FJAK2, was able to activate STAT5 in a cytokine receptor deficient cell line. In vitro binding studies revealed that FERM-JAK2 directly binds and activates STAT5. Using flag- and myc-tagged FERM-JAK2 we show that the FERM domain is sufficient for an efficient dimerization and activation of FERM-JAK2 in the absence of an cytokine receptor. Transplantation of retrovirally transduced murine bone marrow with FERM-JAK2 compared to JAK2 V617F showed an accelerated phenotype with marked increases in WBC, HCT, RBC, HB, reticulocytes, pronounced splenomegaly and loss of body weight in mice. In contrast to V617FJAK2 expressing mice, FERM-JAK2 expressing mice displayed a lethal phenotype. Taken together, we could identify a novel JAK2 variant which alters the kinase domain structure, leads to direct, cytokine receptor independent STAT5 activation, and gives rise to an accelerated, PV-like disease in mice.