Mutational Mechanisms in Pediatric Acute Lymphoblastic Leukemia with Multiple Relapses

Pediatric acute lymphoblastic leukemia (ALL) is generally marked by a low mutational load at initial diagnosis, which increases in subsequent relapses, partly because of hypermutation in a subset of cases (Waanders et al., 2020, Blood Canc. Disc., 1:96). This increased mutational load can be driven by intrinsic mutational processes or could be therapy related. In this study, we have investigated these mutational processes in pediatric ALL patients who suffered multiple relapses, to determine which processes are active in (relapsed) ALL and how they behave during disease progression before and after therapy.

We identified 63 historical cases of ALL with multiple relapses, and selected patients by availability of material and blast percentage above 70% in multiple leukemic samples. If viable material with less than 70% blasts was available, mononuclear cells were enriched by sorting for CD10 positivity. We collected and whole genome sequenced 97 tumor samples from 28 ALL patients (27 precursor B-ALL and 1 T-ALL) with 2 (n=17) or more (n=11) relapses, and 44 remission samples. Somatic single nucleotide variants (SNVs) were called by Mutect2. SNVs were clustered per sample based on their dynamics in time and all clusters in all patients were included for de novo signature extraction using the R package MutationalPatterns. A mutational process was deemed active in a sample if >150 SNVs and >10% of SNVs in a cluster were attributed to a mutational process by de novo signature extraction.

The majority of ALL patients with multiple relapses had a mutational load above average, compared to an unselected diagnosis cohort, which often increased during disease progression. Six cases (21%) passed the hypermutation threshold of 1.3 mutations/Mb as previously defined (Waanders et al., 2020). However, mutational signature analysis revealed that mutational processes were active in a much larger proportion of cases. We extracted 7 single base substitution (SBS) signatures, of which 4 were known COSMIC signatures (cosine similarity >0.95) and 2 were similar to known COSMIC signatures (cosine similarity >0.8). At time of diagnosis 4 distinct mutational processes were found to be active alone or in combination in 10 cases (36%), which is much higher than an unselected diagnosis cohort (13%). These processes included APOBEC-associated mutagenesis (SBS2 & SBS13), a mutational pattern resembling UV-like damage (SBS7a), a process that may be caused by reactive oxygen species (SBS18-like) and an unknown process (Fig. 1). By clustering SNVs per patient in time we followed the activity of mutational processes and identified rising and falling (sub)clones. APOBEC mutagenesis was found to be an ongoing process, sometimes appearing for the first time at relapse (Fig. 1B), although when xenografted in mice (n=2) no continued mutagenesis was observed. SBS7a or SBS18-like mutagenesis always occurred already at diagnosis. One patient, who relapsed 5 times, acquired at each presentation between 500-1500 new SNVs with a specific pattern that did not match any known cosmic mutational signatures, and may represent a novel intrinsic mutational mechanism.

At time of relapse, additional (mostly treatment-related) mutational processes became active, eventually affecting 17/28 (61%) patients (Fig. 1A), much higher than observed in cases with single relapses (~40%; Li et al., 2020, Blood, 135:41). We observed the thiopurine-associated SBS87 in 9 (32%) patients, and an unexplained SBS86-like mutational signature in 2 patients (Fig. 1B). Furthermore, we noticed SBS17-like mutagenesis in one patient. The etiology of SBS17 is as yet not established, but it has been proposed to be caused by reactive oxygen species.

To conclude, ALL patients with multiple relapses showed an increased prevalence and large variety of interplaying mutational processes at diagnosis (36%) and final relapse (61%), with intrinsic or therapy induced origins.

Lelieveld:ENPICOM: Current Employment.