Dissecting Childhood B-Other Acute Lymphoblastic Leukemia: Lower Frequency of Druggable Gene Fusions in European Population?

Novel biological subtypes within B-other (B-o) acute lymphoblastic leukemia (ALL) have been described recently, based on the presence of genomic rearrangements (r.) of specific genes ( DUX4 , ZNF384 and MEF2D ) or gene expression signatures (Ph-like and ETV6-RUNX1 -like ALL). Genomic profiling not only helped to dissect heterogeneity of B-o phenotype and outcome, but it also identified novel druggable aberrations such as kinase and cytokine receptor gene fusions, that were shown to occur mainly in Ph-like ALL. Unfortunately, only very limited population-based data are available to estimate an unbiased frequency of these novel genetic aberrations and subtypes, especially in European countries.

We have analyzed a consecutive cohort of pediatric B-o ALL (negative for ETV6-RUNX1 , TCF3-PBX1 , BCR-ABL1 and KMT2A -involving fusions, hyper- and hypo-diploidy) diagnosed in the Czech Republic between 11/2012 and 5/2017. Transcriptome sequencing (RNAseq) was performed in 69/75 cases. High-density SNP array and whole exome sequencing were performed in 65 and 48 of those, respectively. Fusion transcripts, SNVs and indels were called from sequencing data using in-house pipeline based on publicly available software tools. Targeted analysis was performed to identify DUX4- r. Hierarchical clustering based on published gene sets defining DUX4 -r., Ph-like and ETV6-RUNX1 -like ALL was performed using RNAseq data. Somatic copy number aberrations and regions of uniparental disomy were called from SNP array data.

Seventy five B-o cases represented 24% of all pediatric ALL (1-18 years) diagnosed within this period. According to hierarchical clustering, 23/69 cases were classified as DUX4 -r. ALL; all harbored DUX4 -r. They accounted for 33% of B-o ALL, representing a larger proportion than reported by others (Liu et al., EBioMedicine 2016; Lilljebjörn et al., Nat Commun 2016). Three cases (4%) and one case (1%) were positive for ZNF384-r. and MEF2D-r., respectively. Eleven cases (16%) were classified as Ph-like ALL; this subgroup included 7/9 cases with CRLF2 -r. and a single case with ETV6-ABL1 fusion. No other kinase/cytokine receptor gene fusion was detected. Four patients (6%) were classified as ETV6-RUNX1 -like ALL; all harbored aberrations of ETV6 . Strikingly, all 4 ETV6-RUNX1 -like cases displayed concurrently also Ph-like signature. The remaining 27 cases (39%) were not assigned to any novel subtype. Other lesions activating kinase signaling [mutations of NRAS (n=14), KRAS (n=7), FLT3 (n=4), PTPN11 (n=3), JAK2 (n=3), JAK1 (n=1), NF1 (n=1) and BRAF (n=1)] were found in 36 cases and were not enriched in Ph-like or other subtype. Of note, 9/27 of "unassigned" (but 0/42 of "assigned") B-o harbored in-frame "non-kinase" PAX5 gene-involving fusions, both already described and novel, with largely unknown biological and clinical significance.

Our data show that frequency of ZNF384 -r., MEF2D -r. and, importantly, also of kinase/cytokine receptor gene fusions (apart from CRLF2 -r.) might be lower in European population than estimated from data based on non-consecutive non-European cohorts (Liu et al.; Gu et al., Nat Commun 2016). To that extent our findings correspond with the recently published Swedish study (Lilljebjörn et al.); however, higher frequency of DUX4-r. ALL in our cohort indicates that differences in genetic composition of pediatric ALL may still occur even within Europe. Our findings highlight one of the flaws of Ph-like testing - non-specificity of Ph-like signature. Of note, the overlap of Ph-like and ETV6-RUNX1 -like signatures is also evidenced by positivity of Ph-like signature in ETV6-RUNX1 + ALL in study by Reshmi et al. (Blood 2017). Very low frequency of "kinase fusions" (apart from CRLF2 -r. which can be indirectly, yet reliably, diagnosed by flow cytometry) questions the use of Ph-like testing as a tool to pre-select ALL with high probability of having these lesions. Despite the use of complex genomics techniques a significant proportion of B-o ALL remains unclassified; interestingly, PAX5 fusions seem to be present in a significant proportion of them.

To conclude, our population-based data confirm and specify novel genetic subtypes of pediatric B-o ALL; however, the low incidence of druggable kinase-activating fusions questions the extent of applicability of recently promoted individually "tailored" treatment.

Support: AZV 15-30626A; AZV 16-30186A; GACR 15-06049Y; Primus/MED/28