Patient-Tailored Deep Sequencing of Peripheral Blood Enables Early Detection of Relapse in Childhood Acute Myeloid Leukemia

Relapse remains a major therapeutic challenge in children with acute myeloid leukemia (AML). Outcome after relapse may improve if preemptive therapy is initiated at first evidence of leukemia regrowth. Early detection of imminent relapse requires molecular measurable residual disease (MRD) monitoring after therapy completion. Today, this is possible only in about 40% of children with AML that harbor genetic abnormalities applicable for quantification using standardized qPCR assays. To enable disease surveillance for all patients, we developed patient-tailored deep sequencing (DS) MRD analysis, which provides highly sensitive detection of leukemia-specific mutations. We investigated the potential of this method for early relapse detection in peripheral blood (PB), the only easily accessible source for MRD sampling in children.

PB samples were collected at monthly intervals during follow-up from 45 children diagnosed with AML and treated according to The Nordic Society of Pediatric Haematology and Oncology (NOPHO)-DBH AML 2012 protocol between January 2013 and May 2016 in Denmark, Norway, Sweden and Finland (508 samples, median 11 samples/patient, range 3-27). Nine patients with relapse (median age 5 years, range 0-8) had available diagnostic and relapse material and were included in this study. The patients displayed core binding factor abnormalities (n=3), KMT2A-rearrangements (n=3), monosomy 7 (n=1) or normal karyotype (n=2) at AML diagnosis. Leukemia-specific single nucleotide variants (SNVs) were identified with exome sequencing (ES) of sorted leukemic cells with lymphocytes or remission PB as constitutive DNA template. A variant allele frequency (VAF) with 95% confidence interval including 50% indicates presence of the mutation in all leukemic cells at diagnosis. With the exception of 2 cases with only subclonal mutations at diagnosis, leukemia-specific SNVs with VAF of 50% at diagnosis and persistence at relapse were selected as MRD targets. MRD target mutations were quantified in PB samples preceding overt relapse using patient-tailored DS assays with sensitivity of VAF 0.02%. In diagnostic samples, ES identified 53 leukemia-specific SNVs (median 4 SNVs/patient, range 2-12) of which 33 were also present at relapse (median 2 SNVs/patient, range 1-9). The number of mutations identified at diagnosis increased with age (R_s 0.83, p=0.006). All patients had at least one leukemia-specific SNV detected at both diagnosis and relapse. Twenty-one MRD target mutations (median 2 SNVs/patient, range 1-3) were quantified in PB (55 samples, median sampling interval 28 days, range 11-80) using DS. In 8/9 patients, at least one SNV was detected in PB before overt relapse occurred. The first PB sample showing MRD positivity (median VAF 0.14%, range 0.03-0.44) preceded hematological relapse at a median interval of 3 months (range 0-7.9). In 6 patients not preemptively treated, the median doubling time based on VAF increments was 7 days, with great variability between individuals and genotypes (range 4-28 days). Three patients had molecular relapse diagnosed by qPCR used in clinical diagnostics and received individualized preemptive treatment. In these 3 patients, DS detected mutations in PB for >100 days preceding overt relapse and the doubling times were 14, 25 and 36 days.

In conclusion, DS of leukemia-specific mutations at frequent intervals in PB enables early detection of relapse and ES at diagnosis may identify SNVs applicable for such longitudinal MRD monitoring. This approach facilitates molecular disease surveillance and initiation of preemptive therapy in AML patients without established qPCR targets.