Abstract

Background and aim: The importance of sensitive minimal residual disease (MRD) analysis for determination of response to treatment in acute myeloid leukemia (AML) is becoming increasingly evident. Routinely, this analysis is performed using multiparameter flow cytometry, and in select cases with fusion transcripts using reverse transcription polymerase chain reaction. The drawback with flow cytometry is that it is associated with false negativity due to immunophenotypic shifts during treatment and in pending relapse. In addition, leukemia immunophenotypes often overlap with the normal regenerating bone marrow cell populations. Therefore, other means of identifying remaining leukemic cells are warranted. Leukemic cells in AML are characterized by somatic mutations in recurrently mutated genes as well as in random genes, in most cases as single nucleotide variations (SNVs). We have previously reported that leukemia-specific mutations can be readily identified at the time of diagnosis of AML using exome sequencing of high purity sorted leukemic cells and lymphocytes. The aim here was to show that leukemia-specific mutations identified with exome sequencing at diagnosis can serve as markers for MRD, quantified with targeted deep sequencing, during follow-up.

Method: Seventeen cases of AML, age 2-71 years old, were included in the study. Leukemic cells and lymphocytes were sorted using fluorescence activated cell sorting (FACS), from blood or bone marrow at diagnosis of AML. Exome sequencing of sorted cell populations was performed on the Illumina platform. Variant calling was performed with Mutect for SNVs and with Strelka and Varscan for short insertions/deletions. The data was subjected to an in-house statistical algorithm to identify variants present in all leukemic cells and thus suitable for MRD analysis. For targeted deep sequencing, the Truseq-library system was used for in-house PCR and sequencing on the Illumina Miseq platform (2x150 bp). The acquired reads were stitched using PEAR, aligned to the human reference genome and the resulting alignments were analyzed with in-house scripts with respect to specific SNVs and NPM1 insertion.

Results: Exome sequencing of the paired leukemia/lymphocyte samples identified 240 leukemia-specific SNVs (14 (0-29) per case (median, range) and 22 small insertions and deletions (1 (0-5) per case). The most common type of mutation was, as expected, substitution of cytosine to thymine (CàT). The number of leukemia specific SNVs correlated with age (r=0.76, p<0.001). Mutations suitable for MRD analysis were identified in all but one of the investigated AML cases. Targeted deep sequencing of leukemic cells in serial dilutions established linearity down to a determined variant allele frequency (VAF) of 0.025% for SNVs and of 0.016% for insertion in NPM1. The level of detection (mean+3SD of normal samples) was VAF 0.025% for SNVs and VAF 0.007% for insertion in NPM1. Targeted deep sequencing was then performed on DNA prepared from follow-up bone marrow slides from a patient with AML with mutations suitable for MRD analysis according to our algorithm. Targeted deep sequencing of three SNVs (in the genes CPS1, ITGB7 and FAM193A) and NPM1 type A mutation could detect mutations at all eight time points tested. There were strong correlations between the detected mutation load of the SNVs and the NPM1 type A mutation and all four mutations were present at relapse 10 months after diagnosis. Targeted deep sequencing of SNVs was in this case more sensitive and robust than multiparameter flow cytometry, which could not detect leukemic cells (<0.1% of all cells) at two of the tested time points (5 and 8 months after diagnosis) and showed a completely switched immunophenotype of leukemic cells at relapse.

Conclusions: Exome sequencing of high purity sorted leukemic cells and lymphocytes at the time of diagnosis of AML can identify leukemia-specific mutations suitable for MRD analysis. With targeted deep sequencing of leukemia-specific SNVs identified in this manner, leukemic cell burden can be estimated with high sensitivity during follow-up. The method could be used for patient-tailored MRD analysis in AML.

Disclosures

No relevant conflicts of interest to declare.

Author notes

*

Asterisk with author names denotes non-ASH members.