Elucidation Of Tumor Clonal Diversity in An AML Drug Resistance Model Using A High Throughput Single Cell Genome Amplification Method.

Swetha D. Velivela1, Jay A.A. West1, Joe Dahl1, Jon S. Zawistowski1

1: BioSkryb Genomics, Durham, NC, USA

Introduction: Genomic plasticity within tumors contributes to the cellular heterogeneity, which can drive treatment resistance. However, detection of rare, treatment-resistant progenitors is extremely difficult using either conventional bulk population analysis or when analyzing a few individual cells. The goal of our study was to develop a high-throughput automated workflow that can detect intrinsic and acquired mechanisms of resistance to quizartinib, an FLT-3 inhibitor, in an acute myeloid leukemia (AML) cell line, MOLM-13.

Methods: A “continual” drug dosage resistance model was created by 2-month duration of 2nM quizartinib, replenished every 3 days with fresh media, while a “dose-escalation” model was created by increasing the quizartinib dose weekly by 100 pM increments up to 2 nM. The single-cell whole genome amplification was performed using ResolveDNA (BioSkryb Genomics) with digital cell dispensing (HP D100), digital liquid dispensing (HP D300), and automated library preparation (Agilent Bravo). 184 resistant and 184 parental cells were sequenced by lowpass sequencing (Illumina NextSeq2000). 1,104 were processed at 272plex for whole exome sequencing. Bioinformatics analysis was performed using BaseJumper (BioSkryb Genomics).

Results: The degree of genomic plasticity observed in the parental and two drug dosage models is significant, finding both loss (5p) and gain (19q) of chromosomal regions as a direct result of the drug regime. Copy number variation (CNV) within these groups shows dynamic genomic rearrangements within a single drug treatment model. Specifically, in the continual dosage model CNV analysis clusters into two predominant models with genomic rearrangements in the pathway for drug resistance of FLT-3 and or downstream MAPK activation circumventing the primary drug target. Principal component analysis (PCA) of single cells reveals 6,444 genomic variants that stratify the treatment groups, including a FLT3 N841K secondary mutation predominant in the resistant population but identified at low frequency in the parental population.

Conclusions: Detection of a FLT3 secondary mutation in the treatment-naïve cell population demonstrates the diversity within treatment naïve cells and highlights the role of natural selection during drug treatment driving resistance. Similarly, the distinct mechanisms of acquired therapy resistance between models shed light on the marked genomic plasticity resulting from varying modes of selection pressure.