Single Cell Sequencing Reveals Evolution of Tumor Heterogeneity of Acute Myeloid Leukemia on Quizartinib

Introduction: Clonal selection occurring under the selective pressure of Fms-like tyrosine kinase 3 (FLT3) tyrosine kinase inhibitor (TKI) therapy is not fully understood. To date, the most common mutations found at the time of resistance to potent and selective type II FLT3 TKIs, like quizartinib, are on-target kinase domain (KD) mutations. Activating RAS pathway mutations are a common mechanism of resistance to type I FLT3 TKIs (McMahon et al., 2019 and Zhang et al., 2019). Changes in the acute myeloid leukemia (AML) genetic landscape at relapse on single agent or combination type II FLT3 TKI therapy have not been well-characterized at the single cell (SC) level.

Methods: We performed SC sequencing with a novel high-throughput SC DNA sequencing platform, Mission Bio Tapestri, on bone marrow or peripheral blood samples from 11 patients with FLT3-mutant relapsed/refractory (R/R) AML treated with quizartinib. We analyzed 13 samples from 7 patients treated with quizartinib monotherapy and 12 samples from 4 patients treated with a combination of quizartinib and omacetaxine.

Results: We analyzed a mean of 6,636 cells/sample (range 2,437-13,538). Each leukemia was made up of an average of 5 clonal sub-populations (range 2-7). 5/10 patients with serial samples demonstrated increased polyclonality at relapse, with a mean of 4 clones pretreatment (range 2-6) and 5 at relapse (range 2-7). All patients had an internal tandem duplication in FLT3 (FLT3-ITD) in at least 1 clone. Using SC sequencing, we visualized distinct clones with both hetero -and homo-zygous FLT3-ITD mutations. At the time of relapse after quizartinib, 7 patients (6 on monotherapy and 1 with combination therapy) developed at least one additional FLT3 KD mutation. These on-target mutations were found either as co-mutations within the FLT3-ITD clone or in a native FLT3 clone without an ITD. 4 of these 7 patients had polyclonal KD mutations (mean 2 distinct KD mutations; range 1-3), most commonly at the D835 locus. In 6/7 patients, each KD mutation was found in separate cellular subclones. However, 1 patient had multiple KD mutations within the same subclone (at the D835 and S838 loci). In 6/7 patients who relapsed with a KD mutation, the KD mutation was the dominant clone at relapse. No clones with on-target KD mutations were present prior to therapy. However, all relapse clones containing off-target mutations, found in 3/11 patients, were present in small subclones prior to clinical relapse. As an example, 1 patient who relapsed with 2 NRAS mutations had 2 distinct pre-existing clonal NRAS mutant populations, without FLT3-ITD co-mutations, detectable at <1% frequency pretreatment. These NRAS mutant, FLT3 wild type populations expanded to make up the dominant population at the time of relapse. One patient relapsed with both on- and off-target mutations, in WT1 and FLT3-D835, which co-occurred and were also found alone different subclones. Importantly, of the 4 patients for whom samples were analyzed at the time of morphologic remission, all 4 had leukemia cells detectable by SC sequencing.

Conclusions: Single cell (SC) sequencing allows for direct measurement of clonality, co-mutational status, zygosity, and clonal evolution, which can only be inferred through traditional bulk NGS. This detail reveals that individual mutations may be present in a number of different combinations and confirms that polyclonality is common even within a single leukemia. From time of diagnosis, knowledge of which mutations co-exist in the same clone can suggest potential mechanisms of resistance and help direct treatment. On therapy, SC sequencing can conclusively demonstrate clonal selection. SC sequencing allows for detection of clones not detectable by bulk NGS. The fact that all 4 patients sequenced at the time of morphologic remission had measurable residual disease (MRD) by SC sequencing suggests the benefit of SC sequencing as a strategy for MRD detection. At the time of relapse, identifying the dominant as well as smaller clonal subsets can help direct appropriate combination targeted therapy. Despite the fact that highly selective type II agents, like quizartinib, have historically been thought to select for on-target resistance, SC sequencing highlights more diverse mechanisms of resistance including off-target mutations. Our data supports the need for effective combination therapy to delay or suppress outgrowth of resistant clones.