Targeted Deep Sequencing of Genetic Alterations Identified By Whole Exome Sequencing Reveals Clonal Evolution in Pediatric T-Lymphoblastic Leukemia

Precursor T-cell acute lymphoblastic leukemia (T-ALL) represents one of the major challenges of pediatric oncology, because relapses are frequently refractory to treatment and fatal. The molecular understanding of progression to relapse in T-ALL is limited. We aimed at identifying patterns of clonal evolution and at describing mechanisms of relapse by comparing the genetic and epigenetic alterations in primary and in relapsed pediatric T-ALL.

DNA from bone marrow of 13 patients with T-ALL at primary disease, remission and relapse was analyzed by a combination of multiplex ligation- dependent probe amplification (MLPA), Illumina 450k array, whole exome sequencing (WES) and targeted deep sequencing. Targeted deep sequencing was performed after target capture with Agilent HaloPlex. In the target capture design, all loci that showed somatic mutations in WES were included. Deep sequencing was done on all primary disease and relapse samples and on a subset of remission samples. Allele frequencies by HaloPlex were highly reproducible, corresponded well to allele frequencies of loci that were well covered in WES and were consistent after serial dilutions.

Analysis of DNA methylation using the Illumina 450k array showed that methylation of relapse samples does not differ significantly from the methylation of the matching primary disease samples, with the variability between different patients being much larger than the variability within samples from the same patient.

WES identified on average 10 single nucleotide variants (SNVs) and 1.8 small insertions and deletions (indels) in primary T-ALL and 23.2 SNVs and 2.6 indels in the corresponding relapse samples. Only about 30% of SNVs and indels identified in relapse were already detected in primary disease by WES, while most amplifications and deletions that had been detected by the combination of MLPA and read depth analysis of WES data were conserved from primary disease to relapse. Recurrently, we identified known and novel drivers of T-ALL (NOTCH1, FBXW7, PHF6, WT1, PTEN, NRAS, STAT5B).

Targeted resequencing of mutated genes at high depth (median coverage 6233, 90% of targets covered >1000x) identified rare subclonal alleles with a sensitivity in the range of 10^-2 to 10^-4, depending on the coverage of each individual locus. This allowed us to distinguish de novo mutations that were acquired during treatment from mutations that had already been present at initial diagnosis and were selected for in relapse. Depending on the contribution of clonal selection or de novo mutations, at least two different patterns of relapse could be identified: In a smaller proportion of leukemias, all mutations present at first diagnosis were again detected in relapse, with some additional mutations that were specific for relapse. In most leukemias, the major clone at relapse had arisen from a minor subclone at primary disease and has acquired additional mutations, indicating that clonal selection was the main contributor to the evolution of relapse. In all cases, at least one genetic alteration was detected in samples from both time points.

The example of activating mutations in the nucleotidase NT5C2, which have previously been proposed to contribute to resistance against nucleoside analogues, illustrates the genetic plasticity of T-ALL: Activating NT5C2 mutations were identified in 4 out of 13 relapse samples. The only activating NT5C2 mutation that was already detected in a primary disease sample at low allele frequency was not present in the corresponding relapse sample but was replaced by another activating NT5C2 mutation. This indicates that mutations acquired during treatment may outcompete subclonal mutations that were present in the primary leukemia. In at least two relapse samples, subclonal NT5C2 mutations were detected, compatible with the notion that acquisition of resistance towards chemotherapy by mutation of NT5C2 is a late event on the way to relapse.

Conclusion: The acquisition of novel genetic alterations and selection of treatment resistant subclones are main contributors to T-ALL relapse. We now aim at identifying molecular signatures that characterize treatment resistant subclones, which may be included in risk stratification algorithms of primary T-ALL.