DNA methylation markers assessed at birth are associated with future development of pediatric acute lymphoblastic leukemia

Abstract The underlying etiologic development of pediatric acute lymphoblastic leukemias (ALL) is not fully understood.While clinical outcomes in ALL patients have improved significantly, cytotoxic chemotherapy may result in both acute toxicity and long-term adverse effects of treatment. For this reason, the identification of individuals at greatest risk of developing ALL and efforts toward disease prevention are critically important to reduce pediatric cancer-related morbidity. Evidence suggests most childhood ALL is initiated in the intrauterine period, a time in which DNA methylation patterns are also established in the developing embryo. DNA methylation may therefore serve as a useful sentinel marker of pre-disease and telegraph modifiable risk factors in pediatric ALL. We investigated epigenome-wide DNA methylation status of archived neonatal dried blood spots (DBS), prior to overt leukemia onset, in a meta analysis of ALL case and matched healthy control participants in the California Childhood Leukemia Study and California Childhood Records Linkage Project to identify early epigenetic markers associated with ALL development. Participant DBS samples underwent array-based epigenome-wide DNA methylation assessment using the Illumina EPIC and 450k platforms which were distributed into Set A, B and C based on the timing of enrollment. A total of 1702 participants, including 793 pediatric-ALL cases and 909 matched, healthy controls, were assessed (Set A: 225 cases, 228 controls; Set B:137 cases, 442 controls; Set C: 431 cases, 239 controls). Raw DNA methylation data underwent quality control (QC) and data normalization using Minfi and SeSAme pipelines. A total of 381,823 CpGs passed QC measures for Set A, 722,590 for Set B and 702,265 for Set C. DNA methylation beta values were coverted to M values for analysis. An epigenome-wide association study (EWAS) assessing DNA methylation M-values by ALL case status was conducted using multivariate linear regression, controlling for ALL subtype, genetic ancestral structure, nucleated cell proportions and batch effect for each individual dataset. Nucleated cell proportions were derived from cell reference-based deconvolution (IDOL) followed by isometric log-ratio transformation and robust PCA adjustment. Meta analysis was conducted for the 352,128 shared probes in the three datasets using METAL. Significance for differentially methylated probes (DMP) was set at a false discovery rate (FDR) 0.05 and differentially methylated regions (DMRs) at Šidák P0.05. Meta analyzed regression results identified a single significant DMP (cg13344587) on chromosome (Chr) 10 associated with the gene body region of AT-rich interaction domain 5B (ARID5B, FDR=2.60x10-05). A total of 108 DMRs were identified, with the most significant regions identified on Chr 1 at lymphocyte transmembrane adaptor 1 (LAX1, Šidák P=9.86x10-15) and Chr 11 at Fli-1 proto-oncogene, ETS transcription factor (FLI1, Šidák P=4.51x10-10). Other genes associated with the top 20 most significant DMRs included CYP2E1, GAS7, VTRNA2-1, HLA-F, PRSSL1, AFG3L1,B3GALT4, PSMG4, WT1, ACY3. TP53INP, GPR97, RNF144A, CUX2, ZNF22-AS1, and RARA. While only nominally significant, 'Regulation of cell death,' was the top gene ontology term (P=1.81x10-5) and 'Th17 cell differentiation,' the top KEGG pathway term (P=2.70x10-4) associated with top DMRs. Here, we present one of the largest EWAS assessments to date of California-based pre-diagnostic archived DBS samples in which we identified 1 DMP and 108 DMRs associated with the future development of pediatric ALL. This provides evidence that early changes in DNA methylation may be reflective of the presence of pre-leukemic precursors and serve as risk markers for the development of pediatric ALL. ARID5B contains a known genetic polymorphism associated with ALL. FLI1, a regulator of megakaryopoiesis, was previously identified as a DMR associated with trisomy 21 and suggested to b a potential contributor to the increased risk of leukemia in this population. Our results show alterations in FLI1 methylation also associate with ALL in non-trisomy 21 individuals. The methylation status of LAX1 is known to vary with season of birth, suggesting potential to act as a modifiable contributor to the risk of pediatric ALL. Together, these results give insight into early alterations in epigenetic regulation contributing to leukemogenesis and may serve as important early molecular markers of pediatric ALL.