Characterization of the Genomic Landscape of BCR-ABL1 Kinase-Independent Mechanisms of Resistance to ABL1 Tyrosine Kinase Inhibitors in Chronic Myeloid Leukemia

Despite the well-established success of ABL1 tyrosine kinase inhibitors (TKIs) in the treatment of patients with chronic myeloid leukemia (CML), approximately 20% of patients treated with frontline imatinib develop resistance by 5 years on therapy. The majority (~60%) of such resistant cases are explained by acquired mutations within the BCR-ABL1 kinase domain that compromise inhibitor binding, and nearly all of these mutations are effectively targeted by one or more of the 2^nd and 3^rd generation ABL1 kinase inhibitors. In contrast, the remaining ~40% of imatinib-resistant cases harbor no explanatory BCR-ABL1 kinase domain mutation, presumably attributable to BCR-ABL1 kinase-independent mechanisms. We hypothesized that resistance in these patients results from acquired auxiliary molecular aberrations which persistently activate signaling pathways downstream despite inhibition of BCR-ABL1 kinase activity. To identify such mechanisms, we performed whole exome sequencing and RNA sequencing on a cohort of 135 CML patients comprising the following subgroups: newly diagnosed/TKI naïve (n=28), BCR-ABL1 kinase-dependent resistance (n=31), and BCR-ABL1 kinase-independent resistance (n=65), and TKI-induced remission (n=7). Resistant patients were required to have demonstrated clinical resistance to one or more ABL1 kinase inhibitors in the form of suboptimal response or loss of cytogenetic response; the subtype of resistance was defined based on the presence or not of an explanatory BCR-ABL1 kinase domain mutation at the time of resistance. The majority of samples collected were from patients with chronic phase CML (n=97), although smaller cohorts of accelerated phase CML, blast crisis CML, and Ph+ ALL were also profiled (n=20, 19, and 9, respectively). Among the 44,413 protein-altering and 902 splice site variants detected across the 120 WES samples, there were on average 908 missense, 146 truncation and 69 splice variants per sample. Genes with truncation and missense variants were compared between BCR-ABL1 kinase-independent and -dependent resistant chronic phase samples. A total of 44 genes were seen with a frequency difference of at least 10%, including PLEKHG5 and NKD2 (30% and 28% difference, respectively), which are involved in regulation of NF-kB and Wnt signaling. Consistent with previous reports, we also detected EZH2 and TET2 as exclusively mutated in the BCR-ABL1 kinase-independent resistance patients (6% and 3%, respectively). Further analyses stratifying variants among resistant patients according to specific ABL1 kinase inhibitor therapy failed and comparing, where available, serial samples from pre- and post-treatment for clonal expansion are underway. Additionally, sufficient material was available to perform ex vivo small-molecule inhibitor screening for 48 patient specimens, the resultant data of which was used to generate putative effective drug target profiles and integrated with exome sequencing variants to prioritize variants of functional relevance (HitWalker; Bottomly et al., Bioinformatics 2013). Among 23 patient samples exhibiting BCR-ABL1 kinase-independent resistance, the mutated genes most frequently ranked in the top 10 functional-prioritized variants were: ABL1 (which included non-kinase domain variants; 34.7%), MAP3K1, MUC4, FGF20 (each 17.4%), ARHGEF15, MEF2A, EPHA8, TYRO3, BMP2K, and IRS1 (each 13.0%). Notably, the top six candidates are members of the neutrophin (ABL1, MAP3K1, and IRS1), EPHA forward (EPHA8, ARHGEF15), and p38 MAPK signaling pathways (MAP3K1 and MEF2A). Taken together, these findings suggest that several of the same pathogenic molecular abnormalities seen in other myeloid malignancies are also present in CML patients with BCR-ABL1 kinase-independent resistance, including a subset which align to persistent re-activation of signaling pathways involved in CML disease pathogenesis and progression. As such, genetic and/or functional profiling of these patients in the clinic may translate to actionable candidates for combination therapy to maximize disease control and improve patient outcomes.