Functional Analysis of Kinase-Activating Fusions in Ph-like Acute Lymphoblastic Leukemia

Introduction: Ph-like or BCR-ABL1-like B-progenitor acute lymphoblastic leukemia (ALL) is a high-risk subtype characterized by a gene expression profile similar to BCR-ABL1 ALL. The prevalence of Ph-like ALL rises from 10% in standard risk childhood ALL to over 25% in young adults. Next-generation sequencing of Ph-like ALL identified a variety of alterations involving kinase or cytokine receptor genes, including rearrangement, sequence mutation and copy number alterations. Chromosomal rearrangements in about one-third of Ph-like ALL cases create fusion genes of a variety of 5’ partners that involve ABL1-class genes (ABL1, ABL2, CSF1R and PDGFRB) or activate JAK family members (JAK2, TYK2, IL2RB) that are potentially amenable to treatment with ABL1-class or JAK-class tyrosine kinase inhibitors (TKIs). Notably, ABL2 (Abelson-related gene, ARG), a homolog of ABL1, has rarely been identified as a rearrangement partner in ALL. CSF1R (encoding the macrophage colony stimulating receptor) regulates the differentiation of macrophages, and is not normally expressed in lymphocytes. Likewise, rearrangements involving the JAK family member TYK2, the beta chain of the interleukin 2 cytokine receptor (IL2RB), and the neurotrophic tyrosine kinase receptor type 3 (NTRK3), have not been previously described in leukemia. The goals of this study were to assess the role of these kinase alterations in leukemogenesis, to determine the activation of signaling pathways, and to investigate the efficacy of TKIs.

Methods: Kinase fusions were expressed in interleukin-3 dependent Ba/F3 cells, and co-expressed with the dominant negative isoform of IKAROS (IK6) in interleukin-7 dependent Arf^-/- mouse pre-B cells. Xenograft models of 10 Ph-like ALL tumors - ETV6-ABL1, RANBP2-ABL1, PAG1-ABL2, RCSD1-ABL2, SSBP2-CSF1R, IGH-EPOR, ETV6-NTRK3, ATF7IP-JAK2, PAX5-JAK2 and ZEB2-PDGFRB - were generated by engrafting primary human leukemia cells into NOD-SCID IL2R gamma null (NSG) mice. Activation of kinase signaling was performed using phosphoflow cytometry analysis, and sensitivity to TKIs was assessed ex vivo and in vivo.

Results: All kinase fusions (PAG1-ABL2, MYH9-IL2RB, ATF7IP-JAK2, ETV6-NTRK3 or MYB-TYK2) induced cytokine-independent proliferation of Ba/F3 cells. Mice transplanted with Arf^-/- pre-B cells co-expressing IK6 and either RCSD1-ABL2 or SSBP2-CSF1R developed pre-B ALL (CD43+, B220+, CD19+, BP-1+ and IgM-) with a median latency of 36 and 40 days respectively, providing evidence that ABL2 and CSF1R fusions contribute to leukemogenesis.

In human leukemic cells harvested from xenograft mice we observed distinct patterns of kinase signaling activation and TKI sensitivity for the different fusions. Xenograft cells expressing ABL1-class kinase fusions showed activation of STAT5 that was inhibited with imatinib or dasatinib. Phosphorylation of CRKL, a known target of ABL1 and ABL2, was only observed in cells expressing ABL1/2 fusions. Cells harboring ATF7IP-JAK2, PAX5-JAK2 or IGH-EPOR showed phosphorylation of STAT5 that was attenuated with the JAK2 inhibitor, ruxolitinib. In contrast, cells expressing ETV6-NTRK3 signaled through the MAPK pathway with constitutive pERK1/2 that was inhibited with the ALK-inhibitor, crizotinib. This TKI response profile was confirmed by cytotoxicity assays in xenograft cells, with ABL1-class fusions being sensitive to dasatinib (IC₅₀ range 1-2nM), whilst cases harboring ATF7IP-JAK2 or EPOR rearrangement uniquely responded to ruxolitinib with IC₅₀ values of 500nM and 850nM respectively. Interestingly, in human leukemic cells harboring the ETV6-NTRK3 fusion we observed selective inhibition with both crizotinib and the FLT3 inhibitor, lestaurtinib. Pre-clinical studies on three xenograft models of Ph-like ALL - ETV6-ABL1, RCSD1-ABL2 and SSBP2-CSF1R – showed significantly reduced leukemic burden in dasatinib treated mice (20mg/kg/day p.o) compared to vehicle treated mice.

Conclusions: These data provide important insight on new targets of rearrangement in ALL and describe the first engineered mouse models of Ph-like B-ALL. Functional modeling of these alterations is essential to improve the clinical management of Ph-like ALL by identifying patients with specific genomic lesions at diagnosis and directing them to treatment with appropriate TKIs combined with chemotherapy, analogous to current treatment for BCR-ABL1 B-ALL.