Detection of Emerging Resistant Clones in Philadelphia-Positive Leukemia Patients Exposed to Tyrosine Kinase Inhibitors. Correlation of cDNA and Gdna Approaches

ABL1 Kinase Domain (ABL1-KD) mutations are a common resistance mechanism to tyrosine-kinase inhibitors (TKIs) in Chronic Myeloid Leukemia (CML) and Philadelphia Positive Acute Lymphoblastic Leukemia (ALL). Different ABL1-KD mutations induce different degrees of resistance to different TKIs. The early detection of these resistant mutations helps to adjust patient’s treatment. Here we present an Ultra-Deep Sequencing approach to detect and quantify acquired ABL1-KD mutations in genomic DNA (gDNA), aiming to define a robust test to detect such alterations in TKIs exposed Philadelphia-Positive Leukemia Patients with a resolution below 1E-4.

Firstly, we defined an ABL1 specific next-generation sequencing (NGS) panel designed to cover all coding regions of ABL1 exons 4-10. The 9 amplicons were designed to cover full exons where possible to detect co-occurring mutations (Figure 1A). A panel was then applied to 3 biological replicates of 3 Healthy control donors (9 NGS data points each with 220ng of gDNA). The average coverage per amplicon in all samples was at least 500,000x. The NGS data was then analyzed applying the NGS-MRD algorithm described elsewhere (Onecha, E et al. Haematologica 2019) to 25 known ABL1-KD hotspots. After applying our error correcting algorithm, we obtained an average of 135,000 (22,000-503,000) refined reads for the 25 hotspots. The limit of detection (LOD) was calculated for every position in the DNA as the mean noise (Variant Read Frequency; VRF) per position in the controls ± 3SD (standard deviation); the limit of quantification (LOQ) being defined as mean ± 10SD. For all the hotspots analyzed, the LOD was below 1E-4 and the LOQ below 3E-4 (Figure 1B), except for p.F311L (c.931T>C; LOD=2.7E-3). The high level of noise in this position, constant in the different control samples sequenced in different sequencing runs, is most likely related to the high number of homopolymers in the region.

Ten Philadelphia-Positive Leukemia patients were then screened after TKI treatment (8 CML and 2 ALL). The median BCR-ABL1 defined by quantitative PCR (ratio BCR-ABL1 vs ABL1) in these follow-up samples was 0.6% (0.034% - 95%). All patients were screened in triplicates (220ng gDNA each) and the data-points ± 1SD from the mean were considered outliers (NGS false positives) and excluded from further analysis. Five patients presented a signal above the LOD for p.T315I (c.944C>T). This position is covered by 2 different amplicons in our panel. By bioinformatically demultiplexing the signal, the detection of those five mutations in both amplicons was confirmed (Amp_4; LOD=3E-5, Amp_5; LOD=4E-5). Moreover, aiming to validate this new approach, we applied to paired RNA samples an in-house BCR-ABL1/ABL1 nested PCR + NGS approach designed to quantify those alterations in cDNA. This approach confirmed the presence of 4 out of 5 gDNA detected mutations, with a Pearson correlation of 0.92 (Pval<0.001) (Figure 1C). The only mutation not confirmed by nested PCR presented the lowest BCR-ABL1 quantification (0.034%).

Here we show an Ultra-Deep NGS based test which allows the early detection of TKI resistant emerging clones in genomic DNA samples with a resolution of 1E-4. Despite the facts that in Phi-positive leukemia patients’ other techniques such as the nested PCR are available, for most of heme- dyscrasias it is not easy to detect acquired mutations below 1% VRF. Our test can reduce this limit by at least 2 logarithms. The clinical impact of this approach is illustrated by the two LLA patients included, both under dasatinib therapy when the p.T315I mutations were detected. Those 2 patients were changed to ponatinib, reducing BCR-ABL1 levels. An extension of the cohort and the validation of our test at clinical level will be presented at the meeting.