Giles, Graham G.

AbstractBackgroundPostmenopausal hormone therapy (HT) is associated with a decreased colorectal cancer (CRC) risk. As CRC is a heterogeneous disease, we evaluated whether the association of HT and CRC differs across etiologically relevant, molecularly defined tumor subtypes and tumor location.MethodsWe pooled data on tumor subtypes (microsatellite instability status, CpG island methylator phenotype status, BRAF and KRAS mutations, pathway: adenoma-carcinoma, alternate, serrated), tumor location (proximal colon, distal colon, rectum), and HT use among 8220 postmenopausal women (3898 CRC cases and 4322 controls) from 8 observational studies. We used multinomial logistic regression to estimate odds ratios (OR) and 95% confidence intervals (CIs) for the association of ever vs never HT use with each tumor subtype compared with controls. Models were adjusted for study, age, body mass index, smoking status, and CRC family history. All statistical tests were 2-sided.ResultsAmong postmenopausal women, ever HT use was associated with a 38% reduction in overall CRC risk (OR =0.62, 95% CI = 0.56 to 0.69). This association was similar according to microsatellite instability, CpG island methylator phenotype and BRAF or KRAS status. However, the association was attenuated for tumors arising through the serrated pathway (OR = 0.81, 95% CI = 0.66 to 1.01) compared with the adenoma-carcinoma pathway (OR = 0.63, 95% CI = 0.55 to 0.73; Phet =.04) and alternate pathway (OR = 0.61, 95% CI = 0.51 to 0.72). Additionally, proximal colon tumors had a weaker association (OR = 0.71, 95% CI = 0.62 to 0.80) compared with rectal (OR = 0.54, 95% CI = 0.46 to 0.63) and distal colon (OR = 0.57, 95% CI = 0.49 to 0.66; Phet =.01) tumors.ConclusionsWe observed a strong inverse association between HT use and overall CRC risk, which may predominantly reflect a benefit of HT use for tumors arising through the adenoma-carcinoma and alternate pathways as well as distal colon and rectal tumors.

Abstract We evaluated the joint associations between a new 313-variant PRS (PRS313) and questionnaire-based breast cancer risk factors for women of European ancestry, using 72 284 cases and 80 354 controls from the Breast Cancer Association Consortium. Interactions were evaluated using standard logistic regression and a newly developed case-only method for breast cancer risk overall and by estrogen receptor status. After accounting for multiple testing, we did not find evidence that per-standard deviation PRS313 odds ratio differed across strata defined by individual risk factors. Goodness-of-fit tests did not reject the assumption of a multiplicative model between PRS313 and each risk factor. Variation in projected absolute lifetime risk of breast cancer associated with classical risk factors was greater for women with higher genetic risk (PRS313 and family history) and, on average, 17.5% higher in the highest vs lowest deciles of genetic risk. These findings have implications for risk prevention for women at increased risk of breast cancer.

AbstractBackgroundBody mass index (BMI) and diabetes are established risk factors for colorectal cancer (CRC), likely through perturbations in metabolic traits (e.g. insulin resistance and glucose homeostasis). Identification of interactions between variation in genes and these metabolic risk factors may identify novel biologic insights into CRC etiology.MethodsTo improve statistical power and interpretation for gene‐environment interaction (G × E) testing, we tested genetic variants that regulate expression of a gene together for interaction with BMI (kg/m2) and diabetes on CRC risk among 26 017 cases and 20 692 controls. Each variant was weighted based on PrediXcan analysis of gene expression data from colon tissue generated in the Genotype‐Tissue Expression Project for all genes with heritability ≥1%. We used a mixed‐effects model to jointly measure the G × E interaction in a gene by partitioning the interactions into the predicted gene expression levels (fixed effects), and residual G × E effects (random effects). G × BMI analyses were stratified by sex as BMI‐CRC associations differ by sex. We used false discovery rates to account for multiple comparisons and reported all results with FDR <0.2.ResultsAmong 4839 genes tested, genetically predicted expressions of FOXA1 (P = 3.15 × 10−5), PSMC5 (P = 4.51 × 10−4) and CD33 (P = 2.71 × 10−4) modified the association of BMI on CRC risk for men; KIAA0753 (P = 2.29 × 10−5) and SCN1B (P = 2.76 × 10−4) modified the association of BMI on CRC risk for women; and PTPN2 modified the association between diabetes and CRC risk in both sexes (P = 2.31 × 10−5).ConclusionsAggregating G × E interactions and incorporating functional information, we discovered novel genes that may interact with BMI and diabetes on CRC risk.

AbstractPrevious transcriptome‐wide association studies (TWAS) have identified breast cancer risk genes by integrating data from expression quantitative loci and genome‐wide association studies (GWAS), but analyses of breast cancer subtype‐specific associations have been limited. In this study, we conducted a TWAS using gene expression data from GTEx and summary statistics from the hitherto largest GWAS meta‐analysis conducted for breast cancer overall, and by estrogen receptor subtypes (ER+ and ER−). We further compared associations with ER+ and ER− subtypes, using a case‐only TWAS approach. We also conducted multigene conditional analyses in regions with multiple TWAS associations. Two genes, STXBP4 and HIST2H2BA, were specifically associated with ER+ but not with ER– breast cancer. We further identified 30 TWAS‐significant genes associated with overall breast cancer risk, including four that were not identified in previous studies. Conditional analyses identified single independent breast‐cancer gene in three of six regions harboring multiple TWAS‐significant genes. Our study provides new information on breast cancer genetics and biology, particularly about genomic differences between ER+ and ER− breast cancer.

Abstract Background: Lipid traits have been inconsistently linked to risk of non-Hodgkin lymphoma (NHL). We examined the association of genetically predicted lipid traits with risk of diffuse large B-cell lymphoma (DLBCL), chronic lymphocytic leukemia (CLL), follicular lymphoma (FL), and marginal zone lymphoma (MZL) using Mendelian randomization (MR) analysis. Methods: Genome-wide association study data from the InterLymph Consortium were available for 2,661 DLBCLs, 2,179 CLLs, 2,142 FLs, 824 MZLs, and 6,221 controls. SNPs associated (P &lt; 5 × 10−8) with high-density lipoprotein (HDL, n = 164), low-density lipoprotein (LDL, n = 137), total cholesterol (TC, n = 161), and triglycerides (TG, n = 123) were used as instrumental variables (IV), explaining 14.6%, 27.7%, 16.8%, and 12.8% of phenotypic variation, respectively. Associations between each lipid trait and NHL subtype were calculated using the MR inverse variance–weighted method, estimating odds ratios (OR) per standard deviation and 95% confidence intervals (CI). Results: HDL was positively associated with DLBCL (OR = 1.14; 95% CI, 1.00–1.30) and MZL (OR = 1.09; 95% CI, 1.01–1.18), while TG was inversely associated with MZL risk (OR = 0.90; 95% CI, 0.83–0.99), all at nominal significance (P &lt; 0.05). A positive trend was observed for HDL with FL risk (OR = 1.08; 95% CI, 0.99–1.19; P = 0.087). No associations were noteworthy after adjusting for multiple testing. Conclusions: We did not find evidence of a clear or strong association of these lipid traits with the most common NHL subtypes. While these IVs have been previously linked to other cancers, our findings do not support any causal associations with these NHL subtypes. Impact: Our results suggest that prior reported inverse associations of lipid traits are not likely to be causal and could represent reverse causality or confounding.

Abstract Background: Results from epidemiologic studies examining polyunsaturated fatty acids (PUFA) and colorectal cancer risk are inconsistent. Mendelian randomization may strengthen causal inference from observational studies. Given their shared metabolic pathway, examining the combined effects of aspirin/NSAID use with PUFAs could help elucidate an association between PUFAs and colorectal cancer risk. Methods: Information was leveraged from genome-wide association studies (GWAS) regarding PUFA-associated SNPs to create weighted genetic scores (wGS) representing genetically predicted circulating blood PUFAs for 11,016 non-Hispanic white colorectal cancer cases and 13,732 controls in the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO). Associations per SD increase in the wGS were estimated using unconditional logistic regression. Interactions between PUFA wGSs and aspirin/NSAID use on colorectal cancer risk were also examined. Results: Modest colorectal cancer risk reductions were observed per SD increase in circulating linoleic acid [ORLA = 0.96; 95% confidence interval (CI) = 0.93–0.98; P = 5.2 × 10−4] and α-linolenic acid (ORALA = 0.95; 95% CI = 0.92–0.97; P = 5.4 × 10−5), whereas modest increased risks were observed for arachidonic (ORAA = 1.06; 95% CI = 1.03–1.08; P = 3.3 × 10−5), eicosapentaenoic (OREPA = 1.04; 95% CI = 1.01–1.07; P = 2.5 × 10−3), and docosapentaenoic acids (ORDPA = 1.03; 95% CI = 1.01–1.06; P = 1.2 × 10−2). Each of these effects was stronger among aspirin/NSAID nonusers in the stratified analyses. Conclusions: Our study suggests that higher circulating shorter-chain PUFAs (i.e., LA and ALA) were associated with reduced colorectal cancer risk, whereas longer-chain PUFAs (i.e., AA, EPA, and DPA) were associated with an increased colorectal cancer risk. Impact: The interaction of PUFAs with aspirin/NSAID use indicates a shared colorectal cancer inflammatory pathway. Future research should continue to improve PUFA genetic instruments to elucidate the independent effects of PUFAs on colorectal cancer.

Abstract Background Although 20 pancreatic cancer susceptibility loci have been identified through genome-wide association studies in individuals of European ancestry, much of its heritability remains unexplained and the genes responsible largely unknown. Methods To discover novel pancreatic cancer risk loci and possible causal genes, we performed a pancreatic cancer transcriptome-wide association study in Europeans using three approaches: FUSION, MetaXcan, and Summary-MulTiXcan. We integrated genome-wide association studies summary statistics from 9040 pancreatic cancer cases and 12 496 controls, with gene expression prediction models built using transcriptome data from histologically normal pancreatic tissue samples (NCI Laboratory of Translational Genomics [n = 95] and Genotype-Tissue Expression v7 [n = 174] datasets) and data from 48 different tissues (Genotype-Tissue Expression v7, n = 74–421 samples). Results We identified 25 genes whose genetically predicted expression was statistically significantly associated with pancreatic cancer risk (false discovery rate &lt; .05), including 14 candidate genes at 11 novel loci (1p36.12: CELA3B; 9q31.1: SMC2, SMC2-AS1; 10q23.31: RP11-80H5.9; 12q13.13: SMUG1; 14q32.33: BTBD6; 15q23: HEXA; 15q26.1: RCCD1; 17q12: PNMT, CDK12, PGAP3; 17q22: SUPT4H1; 18q11.22: RP11-888D10.3; and 19p13.11: PGPEP1) and 11 at six known risk loci (5p15.33: TERT, CLPTM1L, ZDHHC11B; 7p14.1: INHBA; 9q34.2: ABO; 13q12.2: PDX1; 13q22.1: KLF5; and 16q23.1: WDR59, CFDP1, BCAR1, TMEM170A). The association for 12 of these genes (CELA3B, SMC2, and PNMT at novel risk loci and TERT, CLPTM1L, INHBA, ABO, PDX1, KLF5, WDR59, CFDP1, and BCAR1 at known loci) remained statistically significant after Bonferroni correction. Conclusions By integrating gene expression and genotype data, we identified novel pancreatic cancer risk loci and candidate functional genes that warrant further investigation.

Background Genome-wide association studies (GWAS) conducted in populations of European ancestry (EA) have identified and confirmed 23 germline susceptibility loci for multiple myeloma (MM). The effect sizes of single nucleotide polymorphisms (SNPs) at these loci are small, therefore combining them into a single summary measure, known as a polygenic risk score (PRS), may provide a more meaningful risk factor. We have previously shown a PRS comprised of the 23 SNPs for MM contributes to increased risk of MM, with a 2.7-fold increase for highest vs. lowest PRS quintiles. Whether the MM-PRS is also associated with overall survival (OS) in MM cases has not been evaluated. We examined the association between MM-PRS and OS in two EA studies. Methods The first study consisted of 2,179 EA MM cases from ten studies included in the Multiple Myeloma Working Group within the International Lymphoma Consortium (InterLymph). Cases were diagnosed between 1970 and 2015 and genotyped using multiple platforms (Oncoarray, Affymetrix, Human660W-quad Beadchip, and Illumina arrays); 885 cases also had stage [based on International Staging System (ISS)] available. Each of the GWAS was subjected to rigorous standard quality control independently (prior to imputation via the Michigan imputation server based on the Haplotype Reference Consortium (HRC). The second study consisted of 515 newly diagnosed EA MM cases from CoMMpass (Relating Clinical Outcomes in Multiple Myeloma to Personal Assessment of Genetic Profile), diagnosed from 2011-2013, who had whole genome sequencing (WGS) performed on germline DNA. The WGS data was used to call common germline genetic variants through the Mayo Clinic bioinformatics pipeline. Briefly, genetic variants were detected with GenomeGPS, aligned to the hg19 reference genome, called using the GATK (V3.6) Haplotype Caller, and merged for multiple-sample joint calling. To reduce the false positive variants, variant quality score recalibration (VQSR) was applied for both SNPs and INDELs. After quality control, 458 EA samples remained. Follow-up was available for both studies and consisted of time from MM diagnosis date until death or date of last known follow-up. The PRS was constructed from the 23 MM SNPs using the published per allele odds ratio associated with MM risk. The published log odds ratios for each SNP were multiplied by the number of risk alleles (0, 1, 2) for the corresponding SNP, and summed, resulting in a unique score per person. Kaplan-Meier curves and Cox proportional hazard models were used to assess the association between PRS with MM OS considering two models: 1) adjusted for age, sex, study and 2) additional adjustment by stage (ISS). Hazard ratios (OR) and 95% confidence intervals (CI) were estimated. The PRS was evaluated both as a continuous variable, per standard deviation (SD), and as a categorical variable (quintiles). Results MM cases (N=2,179) in the InterLymph study were 59% male and 41% female and the median age was 61.0 years (26-90 years). Median follow-up time was 57.2 months (1.0-509.0 months) with 868 reported deaths. MM cases with stage information available consisted of 20% stage I (n=178), 53% stage II (n=466), and 27% stage III (n=241). No association was observed between PRS and OS in MM patients regardless of adjustment for stage (continuous PRS (HR: 1.03, 95% CI: 0.83-1.28, P=0.80) or by quintile PRS (p>0.05)) (Table). The CoMMpass EA MM cases (n=458) had similar distributions for sex (61% male and 39% females) but were slightly older 65 years (27-93 years) and had shorter follow-up time (median=39.75 months (0.13-77.2)) with 117 deaths. Stage was available for 96% of CoMMpass cases including 36% stage I (n=159), 33% stage II (n=146), and 31% stage III (n=134). We also observed no association of PRS and OS in the CoMMpass study (HR=1.02, 95% CI: 0.72 -1.46, P= 0.89), adjusted for age, sex, and stage (Table). Discussion A PRS score for MM risk is not associated with OS for MM cases in two EA populations. Given that prior studies have shown association of genetic variation with MM survival, efforts to identify additional loci associated with OS or MM specific survival are warranted. Future studies should also consider germline variants impact on molecular subtypes, specific therapies, and outcomes. Disclosures Kumar: Celgene: Consultancy, Research Funding; Janssen: Consultancy, Research Funding; Takeda: Research Funding.

Abstract Introduction The 2-3 fold excess risk of multiple myeloma (MM) among family members of cases suggests a heritable contribution to risk. Recently, a genome-wide association study (GWAS) identified two genome-wide significant and one promising novel loci associated with multiple myeloma risk. To confirm these associations and identify additional novel risk loci, we performed a four-center, genome-wide association meta-analysis. Methods A fixed effects model was used for the meta-analysis which included a total of 1248 cases and 1485 controls, all of European descent, genotyped and analyzed at four separate centers with samples contributed by 10 studies. After quality control and imputation using the 1000 Genomes Project, the analysis included ∼9.5 million variants (λ=1.024). Associations between (single nucleotide polymorphisms) SNP genotypes and MM risk were evaluated under a log-additive model of inheritance, with each study adjusting for age, sex, and up to 10 principal components to control for population stratification. Promising results were replicated in an independent set of 1587 cases and 1770 controls using TaqMan, for a total of 2835 and 3255 cases and controls, respectively, in a combined meta-analysis. Results The discovery meta-analysis did not reveal any genome-wide significant associations (defined as p<5 x 10-8). We used a novel pruning algorithm to identify the top 35 most promising single nucleotide polymorphisms (SNPs) to advance to replication. We successfully genotyped 22 SNPs in the replication set. In the combined discovery and replication meta-analysis, rs1345359 at 12q23.1 was the most strongly associated SNP (P=9 x 10-8, Table 1). The variant allele C was associated with reduced risk (odds ratio discovery set [OR]= 0.69, OR replication set = 0.78, OR combined = 0.74). A second locus at 20q13.2 (rs150220835), was associated with a two-fold increased risk (P=1.22 x 10-6), a borderline increased risk (P=0.0900) and 45% increased risk (P=2.44 X 10-5) in the discovery, replication, and combined analysis sets respectively (Table 1). We also confirmed the association between MM risk and two previously published SNPs (rs4487645, p=0.0007and rs105251, p=0.0044) (Broderick et al., Nat. Genet., 2011). The third previously suggested SNP (rs6746082) was of nominal significance (p=0.0517) in the meta-analysis. Discussion We confirmed the association between MM risk and two previously published SNPs and identified a possible association with a novel SNP in chromosome 12q23.1 (rs1345359). This SNP is not located in a gene nor associated with biofeatures in ENCODE, thus further examination of correlated SNPs is necessary to identify a functional SNP linked to this locus. We also found suggestive evidence for a second locus at 20q13.2 requiring additional replication. Larger studies would improve risk variant discovery for this rare hematologic malignancy. Disclosures: Wolf: Celgene: Honoraria, Research Funding; Millenium: Honoraria; Onyx: Honoraria. Anderson:Celgene, Millennium, BMS, Onyx: Membership on an entity’s Board of Directors or advisory committees; Acetylon, Oncopep: Scientific Founder , Scientific Founder Other.

Abstract We previously identified five single nucleotide polymorphisms (SNPs) at four susceptibility loci for diffuse large B-cell lymphoma (DLBCL) in individuals of European ancestry through a large genome-wide association study (GWAS). To further elucidate genetic susceptibility to DLBCL, we sought to validate two loci at 3q13.33 and 3p24.1 that were suggestive in the original GWAS with additional genotyping. In the meta-analysis (5662 cases and 9237 controls) of the four original GWAS discovery scans and three replication studies, the 3q13.33 locus (rs9831894; minor allele frequency [MAF] = 0.40) was associated with DLBCL risk [odds ratio (OR) = 0.83, P = 3.62 × 10−13]. rs9831894 is in linkage disequilibrium (LD) with additional variants that are part of a super-enhancer that physically interacts with promoters of CD86 and ILDR1. In the meta-analysis (5510 cases and 12 817 controls) of the four GWAS discovery scans and four replication studies, the 3p24.1 locus (rs6773363; MAF = 0.45) was also associated with DLBCL risk (OR = 1.20, P = 2.31 × 10−12). This SNP is 29 426-bp upstream of the nearest gene EOMES and in LD with additional SNPs that are part of a highly lineage-specific and tumor-acquired super-enhancer that shows long-range interaction with AZI2 promoter. These loci provide additional evidence for the role of immune function in the etiology of DLBCL, the most common lymphoma subtype.