Comprehensive Functional Characterization of Germline ETV6 Variants Associated with Inherited Predisposition to Acute Lymphoblastic Leukemia in Children

There is increasing evidence for inherited susceptibility to acute lymphoblastic leukemia (ALL) in children. Both common and rare germline genetic variants can significantly influence the risk of developing ALL, often in an age- and/or leukemia subtype-dependent manner (e.g., ARID5B, IKZF1, GATA3, PIP4K2A, PAX5, and TP53). Recently, we and others reported a novel genetic syndrome characterized by damaging germline ETV6 mutations that co-segregate with hereditary thrombocytopenia and increased propensity for hematologic malignancies (especially ALL). Sequencing 4,405 childhood ALL cases, we subsequently identified 30 germline coding ETV6 variants that were potentially responsible for disease predisposition in 35 patients (Lancet Oncol 2015). These findings unequivocally point to ETV6 as a critical ALL risk gene; however, the exact functional consequences of the ALL-related ETV6 variants are largely unknown and the mechanism by which they contribute to leukemogenesis is unclear. To this end, we first systematically evaluated these 30 ETV6 risk variants for their transcriptional repressor activity, ability to bind to target DNA sequences, localize to the nucleus, and also to homo-dimerize. Using PF4 and MMP3 promoter-driven luciferase transcription in HEK293T cells as the model system, we showed that 18 of these 30 ETV6 variants had loss of transcriptional repression activity (median of 53.2% of wildtype ETV6 activity [range 7.1% to 83.9%]). These include 4 of the 5 frameshift variants, all 4 nonsense variants, and also 10 of 11 missense variants in the ETS/DNA binding domain. None of the missense variants outside the ETS domain had significant effects on transcriptional repression function, and surprisingly the E44fs variant showed enhanced activity. Co-expression of wildtype and each of the 18 loss-of-function variants consistently impaired the normal repressor activity of wildtype ETV6, indicating dominant-negative effects. In cells expressing both wildtype and variant ETV6, immunoprecipitation of variant protein pulled down wildtype ETV6, in line with the dominant negative mode of action of variant ETV6 via dimerization with wildtype protein. Cellular fractionation showed significantly increased cytoplasmic localization of 18 loss-of-function ETV6 variants when expressed in HEK293T cells with concomitant reduction in nuclear ETV6. Analysis using EMSA revealed that all 14 variants in the ETS domain resulted in impaired binding to the consensus sequence in ETV6 target genes in vitro, as was also true for all 4 loss-of-function frameshift variants. To examine the role of ETV6 variants on hematopoietic stem and progenitor cell function, we lentivirally transduced murine LSK cells with wildtype or variant murine Etv6 (R355X or R395C). These cells were cultured ex vivo and analyzed for myeloid and pre-B cell differentiation, using colony formation assay. We found no change in the formation of myeloid colonies, whereas the number of pre-B colonies increased slightly with mutant Etv6, although we observed no effects on cell cycle or apoptosis. Focusing on the 18 ETV6 variants with experimentally confirmed deleterious effects, we then evaluated the association of germline ETV6 status with ALL clinical features in 3,906 children from frontline ALL protocols at St. Jude Children's Research Hospital and Children's Oncology Group. Patients with these ETV6 variants were significantly older at ALL diagnosis than those without (13.3 [2.6-21.7] vs 6.8 [0.08-30.7] years, P=0.026) but were more likely to have leukocyte count <50x10⁹/L at presentation (100% vs. 68.6%, P=9.5x10^-4). 80% of patients with deleterious ETV6 variants had hyperdiploid ALL compared to 22.2% of those with normal ETV6 (P=9.3x10^-8), strongly suggesting interactions between germline and somatic variations during leukemogenesis. Finally, whole genome sequencing of matched ALL blast and germline samples from 5 familial ALLs in 2 kindreds with pathologic ETV6 variants identified somatic lesions as plausible second events to promote leukemogenesis in these cases. In conclusion, we comprehensively characterized 30 ALL risk variants in ETV6, the majority of which function by disrupting DNA binding and impairing its transcriptional repressor activity in a dominant negative fashion. ETV6 variants also affect hematopoiesis but most likely require additional somatic lesions for overt leukemogenesis.