Multi-ancestry genome-wide association study (GWAS) of pediatric acute myeloid leukemia (AML) risk identifies four risk loci

Introduction: AML is a relatively rare pediatric hematological malignancy, but its treatment outcomes trail other acute leukemias with a 5-year survival rate at ~70%. While genome-scale susceptibility studies of adult AML risk have been conducted, similar pediatric AML studies have not been published. Methods: 1854 pediatric AML cases (diagnosis age <25y) were assembled from the Children's Oncology Group (clinical trials AAML-03P1/0531/1031), Hospital for Sick Children (CA), International Berlin-Frankfurt-Münster Study Group (DE), and Royal Alexandra Hospital for Children (AUS). Cases were genotyped with the Illumina HumanOmni 2.5 BeadChip. Using ADMIXTURE-inferred global genetic ancestry, 1355 cases were grouped in African (AFR, N=95), Admixed American (AMR, N=118), East Asian (EAS, N=74) and European (EUR, N=1068) ancestry groups. Sex-/ancestry-matched publicly available adult controls from 3 external cohorts (Age-Related Eye Disease Study, Health and Retirement Study, Long Life Family Study) genotyped with the same platform were identified at a ~4:1 ratio. TOPMed-based imputation (version r3) supported ancestry-specific GWAS with 5.5-10.7 million common variants (minor allele frequency, MAF≥1%). Logistic regression models adjusted for population substructure tested variant risk associations. Multi-ancestry meta-analysis was performed using an inverse variance-weighted fixed effects model. Results: Four novel genome-wide significant (P<5x10 -8 ) pediatric AML risk loci ( PRIM2 , HERC2 , AGRN , DEFB131A ) were identified in the EUR GWAS (N=5340), with moderate per-allele odd ratios (OR range: 1.9-2.9). In silico analyses indicated lead variants at HERC2 , AGRN , and DEFB131A loci are in active chromatin regions in blood or bone marrow tissues and overlap transcription factor binding sites, including in leukemia cell lines. The HERC2 index variant (OR=1.9, 95% CI: 1.5-2.3) is associated with HERC2 expression in venous blood (GTEx). Multiple variants at known EUR adult AML risk locus KMT5B were also nominally replicated (P=8.7x10 -3 ). Additional suggestive associations (P<1x10 -6 ) in the AFR (N=475) and AMR (N=518) GWAS were seen; among these, putative pediatric AML risk locus KANK1 , which was characterized by low frequency (MAF=1-5%) AFR-specific effect alleles, i.e., nearly absent in other ancestries, is notable for its large risk effects (OR=5.6, 95% CI: 3.0-10.7). The multi-ancestry meta-analysis did not reveal additional genetic signals shared across ancestry groups. Conclusion: We report results from the first pediatric AML GWAS (N=6507, 1355 cases) using international data. We identified 4 novel EUR-specific risk loci, a plausible novel AFR-specific risk locus, and replicated KMT5B , a risk locus reported in a EUR adult AML meta-analysis. Future work includes replication and functional validation studies.