Ultra-Deep Next-Generation Sequencing (NGS) Detects BCR-ABL1 Kinase Domain Mutations with High Sensitivity and Allows to Monitor the Composition of Distinct Subclones During Tyrosine Kinase Inhibitor Treatment

Abstract 2268

Since the introduction of imatinib mesylate for treating patients with chronic myeloid leukemia (CML) data has shown that a small proportion of patients in chronic phase experience refractoriness to kinase inhibitor treatment due to mutations in the ABL1 domain. A variety of molecular methods provide a clinically feasible workflow to detect these mutations but have both strengths and weaknesses: e.g. direct sequencing of nested PCR assays amplifying the ABL1 part of BCR-ABL1 fusion transcript has a low diagnostic detection sensitivity of 10%; and more sensitive methods such as allele-specific PCRs are only available for known mutations. Here, we investigated the utility of ultra-deep next-generation sequencing (NGS) to detect and monitor the composition of ABL1 kinase mutations at diagnosis and during inhibitor treatment. We included 62 samples from 13 CML patients, diagnosed between 10/2005 - 07/2009. The chimeric BCR-ABL1 fusion transcript was amplified from cDNA. Subsequently, 2 overlapping amplicons were designed to amplify a 740 bp stretch of the ABL1 kinase domain to be processed by 454 Titanium amplicon chemistry protocol (454 Life Sciences, Branford, CT). In median, 454 sequencing data was generated for each patient across 4 time points (range 3–11) with a median time span of 11 months (range 5–35) from state of first diagnosis to the most recent investigation. For each investigation a median of 2234 reads specific for the ABL1 kinase domain was obtained. In this selected cohort, in 12/13 patients, previous Sanger sequencing had detected a spectrum of 10 different kinase domain mutations during course of inhibitor treatment. In all 12 patients deep-sequencing enabled the quantitative detection of these mutations with 100% concordance. In 1/13 cases carrying a t(2;3)(q31,q27), neither NGS nor routine methods detected any mutation over four distinct time points, although the %BCR-ABL1/ABL1 was persistently high (ratios of 108, 84, 80, and 108, respectively). According to European LeukemiaNet guidelines (Baccarani et al., JCO 2009), kinase domain mutation screening is only recommended if BCR-ABL1 transcript levels are increasing at consecutive time points. We therefore investigated at which point in time NGS would allow to detect resistant clones subsequent to inhibitor therapy even though alternative testings had not been performed. In 11/12 cases resistant clones were already early detectable after start of treatment with tyrosine kinase inhibitors, with a median time span of 3.9 months from first diagnosis (range 1.1–19.5). We here demonstrate in 6 patients, treated with Imatinib or Dasatinib that even at transcript levels of BCR-ABL1/ABL1, ranging in our cohort from 1.1%-13%, mutations were detectable (range of mutations: 4%-97% of sequencing reads). These points to a very fast selection process of resistance-defining clones. More importantly, NGS was able to quantitate in 4 patients an increasing mutational burden of resistant clones while the %BCR-ABL1/ABL1 stayed consistently low. For example, one patient harbored the L387M mutation, detected with 54% sequencing reads harboring the mutation (1.1 %BCR-ABL1/ABL1). After 8 months, the L387M mutation was detectable in 97% NGS reads, while the %BCR-ABL1/ABL1 was 3.7. Moreover, it was possible to monitor in detail the molecular composition in 3 patients harboring concomitantly more than one mutation. NGS demonstrated that one of these mutations or even a novel clone was dominant in a subsequent time point, in particular after changing the inhibitor. For example, one patient harbored concomitantly the mutations L284V (30%), Y253H (17%), and T315I (4%). The L284V and Y253H mutations disappeared during Dasatinib treatment, and T315I was thereafter persistent -within 38 days- with a high mutation burden of 98%. Importantly, 454 NGS never detected more than 3 resistant clones concomitantly per patient and thus in no case indicated the existence of a variety of coexisting low-level mutations. However, in all 12 patients at the last time point of our investigation there was only one resistance mutation detectable (in 6/12 cases: T315I). In conclusion, as investigated in this work, NGS is technically feasible for high-throughput serial monitoring of CML patients and, moreover, provides an accurate and highly-sensitive quantitative assessment of mutations leading to therapy resistance but also allows uncovering novel mutations.