Abstract PR-2: Characterization of germline genetic variants underlying racial disparities of childhood leukemia outcome

Whether treatment outcome of childhood acute lymphoblastic leukemia (ALL) differs by race or ethnicity is controversial. Some, but not all, clinical studies demonstrated that black, Hispanic, and American Indian children with ALL fared worse as compared to whites and Asians. Factors contributing to these racial differences have been largely elusive. In this study, we sought to use genome-wide germline single nucleotide polymorphism (SNP) genotypes to 1) determine racial composition in children with ALL, 2) to examine relationships between ancestry and relapse risk, and 3) to identify SNPs related to racial difference in relapse risk. Applying principal components analysis (PCA) to genotypes at 588,920 germline SNPs, we identified 3 major axes of variation (principal components, PCs) in 998 individuals: 683 patients with ALL (450 from St. Jude Children's Research Hospital [St. Jude] Total XIIIB and Total XV trials and 233 from Children's Oncology Group [COG] P9906 protocol), and 315 reference population samples (60 HapMap CEU, 60 HapMap YRI, 90 HapMap CHB/JPT, and 105 American Indians, serving as references for white, black, Asian, and American Indian populations, respectively). While PC1 and PC2 were informative of the black and East Asian ancestries, respectively, PC3 primarily captured genetic variation characteristic of American Indian ancestry. Self-declared Hispanic patients with ALL (n=75) formed a continuous cline in PC3 between the American Indians and whites. When assessed for relationships with outcome (after accounting for treatment differences), only PC3 (not PC1 or PC2) exhibited a positive correlation with risk of relapse in the St. Jude cohort (P=0.038). The association of higher proportion of American Indian ancestry (PC3) with increased relapse risk was replicated in the COG cohort (P=0.003). Importantly, PC3 remained significantly associated with outcome even within the group of patients of self-declared white race (P=0.006), after accounting for known prognostic factors (P=0.041), or adjusting for self-declared race (P=0.044). Of 588,920 germline SNPs interrogated, 93,576, 1,985, and 95 were exclusively associated with PC1, PC2, and PC3 at a genome-wide level (P<10-7), respectively. This is consistent with the notion that blacks accumulated the highest degree of genetic variation, with other ancestral groups displaying less variation, in agreement with the timeline for migration out of Africa. Of 95 PC3-specific SNPs, 52 were in genes, most of which were intronic. Also, 11 of the 95 PC3-related SNPs were associated with risk of relapse (P<0.05). For all 11 SNPs, genotypes associated with higher PC3 were associated with higher risk of relapse, consistent with the positive correlation between PC3 and relapse risk, significantly more than what would be expected by chance (P=0.018). Finally, 5 of the 11 SNPs that were associated with both PC3 and relapse risk reside in a single linkage disequilibrium block within the CORIN gene on chromosome 4p12. In conclusion, we demonstrated that patients with a high proportion of American Indian ancestry (greatest in Hispanics) are at an increased risk of leukemia relapse, and identified a small number of germline genetic variations that potentially contribute to the racial disparities in leukemia outcome.