Evidence for Heterogeneous Genetic Associations with Acute Lymphoblastic Leukemia (ALL) By Cytogenetics and Sex in High-Risk Patients Treated with Matched Unrelated Donor Allogeneic Blood or Marrow Transplant (URD-BMT)

Inherited genetic polymorphisms in ARID5B, GATA3, and IKZF1 have been reported to be associated with ALL and vary by cytogenetic subgroups. For example, ARID5B variants are associated with hyperdiploid B-ALL in pediatric patients, which occurs in approximately 30% of ALL patients and is a marker of good prognosis. In contrast, a variant in GATA3 (rs3824662) has been consistently associated with ALL regardless of cytogenetics. While associations by subtype and across age have been explored in ALL patients, the effect of sex on the association of germline variants with ALL has not been tested, despite the higher incidence and worse prognosis of ALL in males than females. We tested the association of previously reported ALL variants in a high-risk ALL population of URD-BMT recipients, assessing the association of these variants with different cytogenetic subgroups and for heterogeneity of effect by sex.

Genotyping was performed using the Illumina Omni-Express BeadChip containing approximately 730,000 single nucleotide polymorphisms (SNPs). After quality control, cohort 1 (C1) and cohort 2 (C2) included 364 (C1) and 82 (C2) B-ALL cases of continental European ancestry and 2,229 (C1) and 809 (C2) controls (unrelated healthy donors). The sample size of non-European cases was not adequate for genetic analyses. Cases were a subset of the genome wide association study (GWAS) named Determining the Influence of Susceptibility COnveying Variants Related to one-Year mortality after BMT (DISCOVeRY-BMT)  a parent GWAS study in collaboration with the Center for International Blood and Marrow Transplant Research.  We tested SNPs within GATA3, IKZF1 and ARID5B for associations with disease status using logistic regression assuming an additive model adjusted for age. Analyses were done separately for the following B-ALL subgroups: hyperdiploid negative ALL, Ph-negative ALL,  Abnormal cytogenetics and Normal cytogenetics. P-values for each cohort were combined using METAL software with weights proportional to the square root of the sample size(P_meta).  Sex-stratified analyses were performed using all B-ALL cases and by subgroup. Tests for heterogeneity using Cochran’s Q method and I² statistic were used to calculate the percentage of variation across sex and cytogenetic group, abnormal versus normal).

The study population consisted mostly of non-hyperdiploid and Ph negative B-ALL. The GATA3 variant conferred a 78% (P=2.9x10^-10) and 82% (P=0.001) increased odds of B-ALL in C1 and C2, respectively, P_meta =1.7x10^-12 (Table 1). The GATA3 SNP was significantly associated, P_meta < 5x10^-8, with all subgroups except abnormal cytogenetics ALL. The SNP in IKZF1 was associated (P_meta < 5x10^-8) with B-ALL overall, Ph negative and normal cytogenetics ALL, with some evidence of heterogeneity in odds ratios (ORs) between normal and abnormal cytogenetics B-ALL in both cohorts. The ARID5B variant, previously shown to be strongly associated with pediatric ALL, showed weak or no evidence of association (Table 1). IKZF1 (rs11980379) associations were stronger across all associated subgroups in males (Table 2). When comparing ORs calculated for males and females in there was evidence of significant heterogeneity (Table 2) for B-ALL overall (P=0.02), non-hyperdiploid (P=0.02) and normal cytogenetics (P=0.004).

GATA3 showed consistent association, with replication, across the cytogenetic subgroups analyzed. IKZF1 variant demonstrated differences by sex, with significant associations between this SNP and B-ALL, observed primarily in males, regardless of cytogenetics. Thus, sex-specific associations were not an artifact of underlying cytogenetic distributions between males and females. The variation between males and females indicates the IKZF1 variant may impact ALL risk differently in males and females, potentially explaining both the overall greater risk of ALL in males and the risk of ALL with worse outcomes. ARID5B associations have previously been associated with hyperdiploid ALL, however the majority of our high-risk study population is non-hyperdiploid, so ARID5B may be a marker of good risk ALL which is rarely treated with URD-BMT.