Acquired Mutations within the JAK2 Kinase Domain Confer Resistance to JAK Inhibitors in an in Vitro model of a High-Risk Acute Lymphoblastic Leukemia

Introduction

Philadelphia chromosome-like acute lymphoblastic leukemia (Ph-like ALL) is a high-risk subtype of ALL associated with high relapse rates and poor survival. Rearrangements of Janus kinase 2 (JAK2r) are present in approximately 5% and 14% of pediatric and young adult Ph-like ALL cases respectively. The resultant JAK2 gene fusions drive leukemogenesis through constitutive activation of the JAK/STAT signaling pathway and are associated with very poor outcomes in patients with Ph-like ALL. All JAK inhibitors in clinical development are type I inhibitors, which bind in the ATP-binding site of JAK2. A phase II clinical trial is currently assessing the only FDA-approved JAK1/2 inhibitor, ruxolitinib in high-risk B-cell ALL cases harboring JAK2 alterations. The development of treatment resistance to targeted inhibitors in other diseases is well documented and often results in disease relapse. Elucidating mechanisms of ruxolitinib resistance in JAK2r ALL will inform approaches to monitor the emergence of resistance in ongoing clinical trials and enable the development of therapeutic strategies to overcome or avert resistance.

Methods

JAK2r B-ALL was modelled in the pro-B cell line, Ba/F3, by expressing the high-risk B-ALL fusion, ATF7IP-JAK2. Ruxolitinib resistance in three independent ATF7IP-JAK2 Ba/F3 cell lines was achieved following dose escalation to a clinically relevant dose of 1 μM ruxolitinib. Sanger sequencing of the RT-PCR amplified JAK2 fusion revealed each resistant line had acquired a different mutation within the JAK2 kinase domain. Therapeutic sensitives were assessed by staining with Fixable Aqua Dead Cell Stain (Invitrogen) and Annexin V, and analysis by flow cytometry. Alterations in signaling pathways were determined using phosphoflow cytometry and western blot analysis. Computational modelling of acquired JAK2 mutations and subsequent influence on ruxolitinib binding was performed using ICM-Pro (Molsoft L.C.C.).

Results

In addition to the identification of two known ruxolitinib resistant mutations, JAK2 p.Y931C and p.L983F, a novel p.G993A mutation was identified. All mutations localized to the ATP/ruxolitinib binding site and conferred resistance to multiple type-I JAK inhibitors, including ruxolitinib, BMS-911543, and AZD-1480 (Table 1). JAK2 p.G993A ATF7IP-JAK2 Ba/F3 cells were also resistant to the type-II JAK inhibitor, CHZ-868, which binds in an allosteric site of JAK2 in addition to the ATP-binding site. Ruxolitinib resistance correlated with sustained downstream STAT5 activation in the presence of 1 μM ruxolitinib compared with non-mutant ATF7IP-JAK2 Ba/F3 cells. Intracellular phosphoflow cytometry of ruxolitinib-resistant ATF7IP-JAK2 Ba/F3 cells confirmed constitutive activation of JAK/STAT signaling in the presence of 50 nM ruxolitinib, in contrast to non-mutant ATF7IP-JAK2 Ba/F3 cells. Computational modelling suggested that JAK2 p.L983F (Fig. 1D) sterically hinders ruxolitinib binding, while JAK2 p.Y931C may reduce ruxolitinib binding affinity by disruption of a critical hydrogen-bond (Fig. 1B). The novel JAK2 p.G993A mutation is predicted to alter DFG-loop dynamics by stabilizing the JAK2 activation loop (Fig1C).

Conclusions

This study demonstrates that the JAK2 ATP-binding site is susceptible to JAK inhibitor resistant mutations following ruxolitinib exposure in the setting of JAK2r ALL, highlighting the importance of monitoring the emergence of mutations within this region. In addition to previously described mutations we identified a novel JAK2 p.G993A mutation that conferred resistance to both type-I and type-II JAK inhibitors. The JAK2 p.G993A mutation was postulated to modulate the stability of a conserved domain. Understanding mechanisms of ruxolitinib resistance, as modelled here, has the potential to inform future drug design and the development therapeutic strategies for this high-risk cohort.