Early Detection and Molecular Characterization of Therapy-Related Leukemia in Children Reveals Patterns of Disease Transformation and Guides Future Surveillance Protocols

Background

Therapy-related myelodysplastic syndrome (MDS) and acute leukemias represent a major cause of non-relapse morbidity and mortality in childhood cancer survivors, and have been associated with exposure to cytotoxic therapies (e.g. radiation, alkylators, topoisomerase inhibitors). Neuroblastoma (NB) patients receive multimodality therapy with intensive chemotherapy, radiation, and immunotherapy, and have had high rates of treatment-related leukemias (Kushner, Cheung et al. 1998). Whilst specific therapeutic modalities have been associated with distinct cytogenetic and molecular abnormalities, our understanding of the relationships between timing of mutation acquisition, dynamics of clonal selection in relation to specific therapeutic modalities, and how these in unison result in overt leukemia, remains limited. Motivated to study these relationships and inform future screening guidelines, we characterized serial bone marrow (BM) samples obtained during surveillance for NB recurrence and therapy-related leukemias.

Methods

We studied a total of 219 serial samples from 55 NB patients treated at MSKCC over a 21-year period. These included 19 patients with MDS or leukemic transformation (median time following NB diagnosis 4.4 years), 15 with transient cytogenetic abnormalities (median time to abnormality 3.1 years), and 21 matched controls (median disease-free follow-up 8.1 years). On average, we analyzed 4 samples per transformation patient, representative of pre-treatment timepoints at NB diagnosis, during NB treatment, and throughout follow-up, with a lead time of 18 years - 1 month prior to transformation, and at time of leukemic transformation. Comprehensive genomic profiling with targeted gene sequencing (MSK-IMPACT Heme), RNA-seq, and Archer FusionPlex was performed to capture acquired gene mutations, chromosome-level copy number alterations (CNA), and fusion genes at the time of diagnosis. Backtracking studies were performed in longitudinal samples with complete molecular and clonal characterization.

Results

We detected at least one disease-defining alteration in all cases with MDS or leukemic transformation at time of diagnosis, with a total of 61 putative oncogenic events across all patients (median 3 alterations per patient, range 1-12). As expected, the most frequent events were MLL fusions (n=6 patients), and mutations in TP53 (n=5 patients). The remaining cases harbored chromosomal aneuploidies or acquired gene mutations in NPM1, IDH1, PTPN11, NRAS, BCOR, CUX1, STAG2, WT1, amongst other genes, at a median variant allele frequency 24% (range 4-68%). Backtracking studies identified at least one of these mutations in 81% of patients at a sampling time point prior to diagnosis. In contrast, only two patients (2/15) from the cohort with transient cytogenetic abnormalities had acquired somatic mutations detected at a median VAF of 3%, with resolution of the molecular alterations in subsequent samples. One of the control cases also had an identified mutation, though this patient died of NB with limited hematologic follow-up.

The median time of detection of a putative driver alteration was 6 months prior to leukemic transformation, with the earliest identified at 2.8 years prior to disease transformation. At least 3 cases of MLL fusions were detected 2.2, 14.5, and 20.9 months prior to diagnosis. Mutations in TP53 co-occurred with CNAs in all patients in our cohort, and has been shown to be predictive of chemoresistant disease. Mutations in TP53 were also identified in at least 2 pre-leukemic samples per patient in 4 of 5 cases, at a median VAF 5% (range 5-20%). In all of these cases, TP53 mutations preceded clinically detectable CNA. Genetic evolution led clonal dominance, which, intriguingly, often preceded disease presentation in the context of normal hematopoeisis. We also found evidence of clonal drifts, possibly as a consequence of treatment effects.

Conclusions

Our preliminary data demonstrate that NB patients at risk of developing secondary leukemia can be identified by molecular profiling of BM aspirates obtained during routine disease surveillance for NB. These findings present an opportunity for the development of early detection studies for patients with pediatric malignancies undergoing intensive therapy and importantly inform studies into mechanisms of leukemic transformation and specific gene-treatment effects.