Genomic Architecture and Clonal Dynamics of Early Relapsed BCP-ALL

Relapse represents the most common cause of therapy failure in B-cell precursor ALL acute lymphoblastic leukemia (BCP-ALL), and is caused by selective outgrowth of therapy-resistant leukemic cells. Two-third of BCP-ALL relapses present after treatment, i.e. after two years. These relapses may originate from leukemic (sub)clones that remained in a quiescent state during treatment or that could not be reached by the chemotherapeutics. Relapses that occur during treatment are different in that they display clonal outgrowth in the presence of chemotherapeutics, and these patients have poorer outcomes. The aim of this study is to explore the genomic abnormalities in leukemia with early relapse, and investigate the clonal dynamics of relapses that arise during treatment.

We included 17 BCP-ALL cases which relapse during treatment (<2yrs) according to DCOG protocols ALL9, ALL10 or ALL11. Median remission time was 1.08 yrs (range 0.48-1.95). Whole exome sequencing was performed on DNA isolated at time of first diagnosis, complete remission and relapse from bone marrow or peripheral blood, with an average read depth on target of 108x. After mapping of the reads, variants were called using HaplotypeCaller. In total, we identified 1771 somatic mutations in 1562 genes. Per case, a median of 21 mutations were detected at diagnosis (range 10-630) and 31 at relapse (range 10-652). A hypermutation profile was observed in one diagnosis-relapse pair, and two additional relapses.

All cases harbored mutations shared between diagnosis and relapse, which were mostly part of the major clone at both time points. However, the fraction of shared mutations varied considerably between cases, ranging from <10% in 3 cases to >80% in 4 cases. Based on the clonal dynamics, 3 distinct groups were recognized. Group I includes two cases, with relapses within 6 months, in which the (sub)clonal mutation spectrum between diagnosis and relapse was identical. Group II (n=10) presented with a relapse closely resembling the major clone at diagnosis. Mostly, these relapses acquired new mutations and they often branched off from the major clone already before the time of diagnosis. Finally, Group III (n=5) consists of cases in which the relapse originates from a minor subclone at diagnosis that hardly resembled the major clone, suggesting a clonal switch during treatment.

Next, we analyzed the genes with mutations that were predicted to be damaging (truncating and non-synonymous conserved missense variants). We performed pathway analysis for these genes and identified RAS pathway genes to be frequently mutated among shared mutations, while mutations in genes involved in epigenetic regulation, chromatin condensation and regulation of transcription were acquired. In total, 7 of the genes with (predicted) pathogenic mutations in relapse were affected in at least two cases, including known genes like KRAS, CREBBP, and WHSC1 (NSD2). CREBBP mutations were never part of the major clone at diagnosis and were present in cases with numerical chromosomal aberrations. Both cases with hotspot E1099K mutation in WHSC1 were t(1;19) translocation-positive.

Recent studies showed that somatic mutational signatures, composed of the six substitution subtypes in a 3-nucleotide context, expose specific biological processes underlying tumor development, including defects in genomic maintenance and repair. Currently 30 mutational signatures have been described [http://cancer.sanger.ac.uk/cosmic/signatures]. Despite the low number of mutations in most samples, we identified at least 5 of these signatures, including the most common signature 1, a signature associated with aberrant AID/APOBEC activity (signature 2), and three signatures associated with mismatch repair deficiency (6, 15, 26). Most cases carried multiple signatures, but signature 2 was very prominent is one relapse and one diagnosis-relapse pair. Most signatures were preserved from diagnosis to relapse suggesting that the same mutational processes remained active.

Taken together, our results show considerable heterogeneity in the group of children with early relapse of BCP-ALL. Two cases with the shortest remission times relapsed without notifiable somatic changes, whereas most other early relapses appeared to arise from minor or newly appearing subclones. These findings demonstrate the strong clonal selection that occurs during treatment in cases with very early relapse.