The Heterogeneous Fusion Gene Landscape in Pediatric Acute Lymphoblastic Leukemia

Background: Next generation sequencing allows for the detection of expressed fusion transcripts across the transcriptome and has spurred the discovery of many novel chimeric transcripts in various cancers. Structural chromosomal rearrangements that lead to fusion transcripts are a hallmark of acute lymphoblastic leukemia (ALL) and serve as markers for diagnosis and stratification of pediatric ALL patients into prognostically relevant subgroups. Improved delineation of structural alterations in ALL could provide additional information for prognosis in ALL and for improved stratification of patients into treatment groups.

Methods: To identify novel fusion transcripts in primary pediatric ALL cells we performed whole transcriptome sequencing of 134 BCP and T-ALL patient samples collected at diagnosis. Our study include samples from patients with the well-known ALL subtypes t(12;21)ETV6-RUNX1, high hyperdiploid (51-67 chromosomes), t(9;22)BCR-ABL1, 11q23/MLL and dic(9;20), in addition to patients with undefined karyotype or non-recurrent cytogenetic aberrations ("undefined" and "other") (n=58). FusionCatcher was used for the detection of somatic fusion genes, followed by a stringent filtering pipeline including gene fusion validation by Sanger sequencing in order to reduce the number of false positives. Principal component analysis (PCA) of patients with fusion genes was performed using genome wide gene expression levels and DNA methylation levels (Infinium HumanMethylation450 bead array).

Results: We identified and validated 60 unique fusion events in almost half of the analyzed patients (n=69). Of the identified fusion genes, 60% have not previously been reported in ALL or other forms of cancer. The majority of the fusion genes were found in a single patient, but 23% were recurrent, including known ALL fusion genes (n=10) and novel fusion genes (n=7). We found that BCP-ALL samples displayed a higher number of validated fusion genes (54%) compared to the T-ALL samples (28%) moreover in BCP-ALL patients with "other" and "undefined" karyotypes, we detected fusion genes in 71% and 61% of the samples, respectively. High hyperdiploid patients had the lowest rate of validated fusion genes (24%) compared to the other well-known subtypes, where we detected subtype-associated fusion genes in 97% of cases. We also identified promiscuous fusion gene partners, such as ETV6, RUNX1, PAX5 and ZNF384 that fused with up to five different genes. Interestingly, PCA revealed molecularly distinct gene expression and DNA methylation signatures associated with these fusion partners.

Conclusion: RNA-sequencing of pediatric ALL cells revealed a detailed view of the heterogeneous fusion gene landscape, identifying both known and novel fusion genes. By grouping samples based on recurrent gene fusion partners we are able to find shared gene expression and DNA methylation patterns compared to other subtypes of ALL, suggesting a shared molecular etiology within these distinct subgroups, offering novel insights into the delineation of fusion genes in ALL.