Rizvi, Abbas

Abstract The first two authors and last two authors contributed equally. Genome-wide association studies (GWAS) have identified risk loci for Acute Lymphoblastic Leukemia (ALL), Chronic Lymphoblastic Leukemia (CLL) and Non-Hodgkin Lymphoma, however an Acute Myeloid Leukemia (AML) GWAS has not been published to date. We performed a GWAS to identify AML and Myelodysplastic Syndrome (MDS) risk loci using a nested case-control study design in the DISCOVeRY-BMT cohorts which includes almost 2000 AML and MDS patients as cases and 2813 unrelated donors as controls. Genotyping was performed using the Illumina Human OmniExpress BeadChip and imputed using the Haplotype Reference Consortium, yielding > 8 million high-quality variants for statistical analysis. Logistic regression models with AML (de novo AML with normal cytogenetics, de novo AML with abnormal cytogenetics and therapy-related AML) and MDS cases and European American healthy donor controls adjusted for age and sex were used to test the association of each SNP with disease status. To identify the strongest association signal with disease we conducted a summary statistic SNP-based association analysis (ASSET) using non-overlapping AML and MDS cases implemented in R statistical software. ASSET uses an exhaustive search for SNPs with small but common pleiotropic effects across groups of traits while accounting for the multiple tests required by the subset search, as well as any shared controls between groups. This approach allowed us to further investigate the heterogeneity within AML subtypes and to gain increased power by pooling subtypes that show pleiotropic effects. ASSET genome wide (GW) significance is defined as P<5.0x10-8, however, we present results of each subset analysis. ASSET analyses identified an association of T allele at rs12203592, Interferon Regulatory Factor 4 (IRF4), with increased risk of de novo AML and MDS (Figure 1) (OR=1.37; 95% CI, 1.25-1.51, Pmeta=1.59x10-10). The variant has been reported significant in GWAS of skin pigmentation, hair color, non-melanoma skin cancer, squamous cell carcinoma, actinic keratosis, and childhood ALL; GW case-control studies of neuroblastoma and breast cancer show suggestive association signals (P<5 x 10-6). IRF4 belongs to the IRF family of transcription factors and is a key regulator of differentiation stages in hematopoiesis. The T allele at rs12203592 is associated with significantly increased expression of IRF4 in whole blood and lung tissue and in vitro mouse studies have shown the T allele to be associated with higher levels of IRF4 transcription. rs12203592 is <80bp from an IRF4 transcription start site and in an important position within NF-κB motifs in multiple blood and immune cell lines, supporting the hypothesis that this SNP modulates NF-κB repression of IRF4 expression. Another variant in the IRF4 regulatory region, rs62389423 (A allele), showed a putative association with subsets of de novo AML and MDS (OR=1.36; 95% CI,1.21-1.52, Pmeta=1.2x10-7). Although not in linkage disequilibrium with rs12203592, the A allele has also been previously associated with susceptibility to skin cancer and melanoma in multiple GWAS. Analysis of 596 UKbiobank CLL cases and >300,000 controls, shows the A allele correlates with a 50% increased risk of CLL (95% CI, 1.32-1.73, P= 2.4x10-9).Several other GWAS also show an association between additional IRF4 variants and CLL. Analyses in individual subtypes also revealed an intronic variant, rs10098598, in RAD21 to be associated with de novo AML (OR=2.18; 95% CI, 1.60-2.97; Pmeta=8.9x10-7). RAD21, on8q24, a region known to be associated with multiple cancer types, is involved in DNA double-stranded break repair and in chromatid cohesion in mitotic cells. Loss of function would theoretically lead to chromosomal instability and tumorigenesis. Components of the cohesion complex (including RAD21) are somatically mutated in ~12% of de novo AML and MDS patients, where it is frequently an early event. We provide the first GW evidence of association between a common variant and AML susceptibility. This SNP has been shown to be associated with multiple phenotypes, suggesting that there are pleiotropic effects at work. Our RAD21 finding is consistent with the role of cohesion in leukemogenesis and provides some evidence of its role in de novo AML susceptibility. Replication and further subset analyses with genome-wide data in >2000 AML cases are ongoing. Disclosures Griffiths: Alexion Inc.: Honoraria, Research Funding; Astex/Otsuka Pharmaceuticals: Honoraria, Research Funding; Celgene, Inc: Honoraria, Research Funding; Novartis, Inc.: Research Funding; Pfizer, Inc.: Research Funding. McCarthy:Bristol Myers Squibb: Consultancy, Honoraria; Celgene: Consultancy, Honoraria, Research Funding; Karyopharm: Consultancy, Honoraria; Takeda: Consultancy, Honoraria; Janssen: Consultancy, Honoraria.

Abstract Over the past decade, there have been numerous published candidate gene association studies of non-HLA variants, genes and/or pathways in relation to BMT survival outcomes. Most of these studies investigated a small number of candidate single nucleotide polymorphisms (SNPs), chosen based on known gene function, in a limited number of cases without replication. Here we test the association of previously reported SNPs with survival outcomes (overall survival (OS), progression-free survival (PFS), death due to disease (DD) or transplant-related mortality (TRM)) after URD BMT using data from the DISCOVeRY-BMT (Determining the Influence of Susceptibility Conveying Variants Related to one-Year mortality after BMT) study which included 2,052 European descent recipient-unrelated donor pairs (Cohort 1) and 763 recipient-unrelated donor pairs (Cohort 1) reported to the Center for International Blood and Marrow Transplant Research (CIBMTR) between 2000-2009 and 2009-2011, respectively. DISCOVeRY-BMT is a Genome Wide Association Study (GWAS) that includes over 9 million typed or imputed SNPs. This represents the largest study to date and is well powered to test both SNP and gene-based association hypotheses. We conducted an extensive literature review of PubMed using MESH terms related to transplant and hematologic disease which yielded 76 candidate gene association studies that investigated a total of 46 genes with any survival outcome. Of the 46 genes, 11 genes (CTLA4, CYP2B6, CYP2C19, GSTM1, GSTT1, IL23R, IL6, MTHFR, NOD2/CARD15, TNFSRF1B, and VDR) were studied in at least 2 publications. A total of 16 SNPs in 8 of these genes reported a significant (P< 0.05) association with at least one survival outcome (OS, PFS, DD or TRM). We tested these 16 SNPs for an association with the survival outcome originally reported in each article (Table 1). All association models included age at BMT, diagnosis (AML, ALL, MDS), disease status at BMT, cell source (blood, bone marrow) and year of BMT. P-values from Cohorts 1 and 2 were combined using a fixed-effects model in METAL software. Since the candidate gene studies have a gene-based hypothesis, we also performed gene-level testing using the VErsatile Gene-Based Association Study (VEGAS2) software to measure the aggregate effect of all available typed and imputed SNPs from our GWAS data in these 8 genes. The candidate gene association studies included heterogeneous patient populations of autologous and/or allogeneic (related or unrelated donors) BMT for a variety of hematologic diseases. Most candidate gene studies performed genotyping on both donors and recipients, and SNPs were tested either independently for donor or recipient genomes, or considering the total number of variant alleles the recipient-donor pair had at a given locus (0-4) (Table 1). The 16 SNPs from the previous candidate gene association studies were not significantly associated at a nominal P<0.05 to either the survival outcome reported in the original published paper, or to any survival outcome (Table 1). Likewise, the VEGAS2 gene-based analysis did not reveal any statistically significant associations for any of the 8 genes at P<0.05. Our study also corroborated prior reports of SNPs and genes that were not associated with survival outcomes. Candidate gene association studies use a priori biological knowledge to select genes that are likely associated with outcome and subsequently SNPs within the gene thought likely to affect function based on knowledge at the time of the investigation. Our findings suggest that neither the genetic variation tested in these candidate genes, nor other individual SNPs in the gene independently or collectively are not associated with survival outcomes in recipients of URD BMT. Our failure to replicate the prior candidate gene studies may be due to these SNPs being dependent on the underlying disease (ie. AML, ALL, etc), the donor relation (related, unrelated, autologous), or that there is no true association of these SNPs with the outcomes studied. #These authors contributed equally Table 1. Table 1. Disclosures McCarthy: Gamida Cell: Honoraria, Membership on an entity's Board of Directors or advisory committees; Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Karyopharm: Honoraria, Membership on an entity's Board of Directors or advisory committees; Onyx: Honoraria, Membership on an entity's Board of Directors or advisory committees; Bristol Myers Squibb: Honoraria, Membership on an entity's Board of Directors or advisory committees; Janssen: Honoraria, Membership on an entity's Board of Directors or advisory committees; The Binding Site: Honoraria, Membership on an entity's Board of Directors or advisory committees; Sanofi: Honoraria, Membership on an entity's Board of Directors or advisory committees. Hahn:Novartis: Equity Ownership; NIH: Research Funding. Sucheston-Campbell:NIH/NCI: Research Funding.

Key PointsIKZF1 associations with high-risk B-ALL may differ by age and sex. A novel variant on chromosome 14, rs189434316, is associated with over a 3.5-fold risk of normal cytogenetic B-ALL.

Key PointsCandidate SNP associations with survival outcomes after URD transplant are most likely false-positive findings. Over 85% of candidate SNPs are not linked to a biochemical function; of those that are, about half are not linked to the candidate gene.