Genome-wide association study of neurocognitive outcomes among childhood cancer survivors.

10063 Background: Long-term survivors of childhood cancer experience heightened risk of late effects, with about 40% developing cognitive impairment. Established risk factors include cranial radiation and specific chemotherapies. Variability in treatment-related outcomes has been associated with genetic factors, though prior studies examined targeted pathways and not whole genome approaches. Methods: Participants included 4,077 childhood cancer survivors with whole genome sequencing and direct neurocognitive testing from the St. Jude Lifetime Cohort Study (SJLIFE). The mean (standard deviation) age at primary cancer diagnosis was 7.8 (5.7) years, and 28.5 (10.3) years at neurocognitive testing; 52.3% were male; 78.9% were Non-Hispanic White. Linear regression evaluated associations between common genetic variants (minor allele frequency, MAF≥1%) and 20 neurocognitive measures (as age-adjusted Z-scores). Analyses were adjusted for sex, age at primary childhood cancer diagnosis, age at neurocognitive testing, cumulative doses of high-dose methotrexate, intrathecal methotrexate, anthracyclines and cranial radiation, and genetic ancestry . Loci with P ≤5x10 -8 were considered genome-wide significant and evaluated in stratified analysis by genetic ancestry and treatment exposures. Results: 21 SNPs met genome-wide significance for associations with ≥1 neurocognitive measure, 9 SNPs with a MAF ≥1% in EUR (n = 3,312) and AFR (n = 636) survivors. All 9 SNPs had 1.1-3.6 times larger effect sizes in AFR compared to EUR. The biggest differences in ancestry groups were seen when stratifying analyses by anthracycline exposure. An intronic variant in ERG , rs1309269486, had a 2.0-times larger effect on motor speed in exposed EUR (β = -0.70; P = 2.9x10 -6 ) compared to unexposed EUR (β = -0.35; P = 0.039). In AFR, rs1309269486 was only associated with motor speed in unexposed survivors (β = -1.33; P = 2.9x10 -3 ), with a 3.8-times greater effect size compared to unexposed EUR. ERG has been shown to play a role in neurogenesis. Another intronic variant in P2RY12 , rs1755678683, also showed a 1.8-times larger effect on attention span in exposed EUR (β = -0.68; P = 5.9x10 -4 ) than unexposed EUR (β = -0.38; P = 0.035). However, it was associated with attention span in only unexposed AFR (β = -1.28; P = 2.1x10 -3 ), showing 3.4-times greater effect than in unexposed EUR. P2RY12 plays a role in microglia function, neuroinflammation, and neurodegeneration. An intronic variant in COL15A1 , rs1837227843, was associated with working memory in exposed EUR (β = -0.47; P = 5.3x10 -7 ) and unexposed AFR (β = -0.58; P = 4.9x10 -3 ). While COL15A1 is highly expressed in many brain tissues, previous associations with neurocognition have not been established. Conclusions: These findings highlight the independent and combined role (with treatment) of genetics and genetic ancestry in adverse neurocognitive impairment among survivors of childhood cancer.