Citations

Abstract BACKGROUND: Incidence of glioma varies significantly by sex, and most glioma histologies occur with greater incidence in males. Previous analyses have examined the impact of estrogen exposure as a risk factor for these tumors, but have found results of varying significance and low effect size. There may be differences in effect of previously discovered risk alleles that contribute to sex differences. METHODS: Using data collected for three previous glioma GWAS in European-ancestry populations (MD Anderson Cancer Center, the San Francisco Adult Glioma Study, and the Glioma International Case Control Study) we assessed sex-specific effects for 14 previously identified and 13 newly identified glioma risk SNPs (27 total) overall and for glioblastoma (GBM) and non-GBM tumors separately. There were 3,892 male cases (59% GBM), 4,522 male controls, 2,500 female cases (52% GBM) and 4,940 female controls. Sex-specific odds ratios (ORM and ORF), 95% confidence intervals (95% CI) and p values (pM and pF) were generated using stratified logistic regression models. Data from each study were analyzed separately and combined using inverse variance weighted meta-analysis. Results were considered statistically significant at p<6.2x10-4. RESULTS: In GBM, rs11979158 (7p11.2, pM=1.01x10-10, ORM=1.43 [95% CI: 1.28-1.59]; pF=3.43x10-3, ORF=1.22 [95% CI: 1.07-1.39]) and rs2562152 (16p13.3, pM=5.59x10-4, ORM=1.23 [95% CI: 1.09-1.39]; pF=1.22x10-1, ORF=1.12 [95% CI: 0.97-1.29]) had significant effect in males only. In non-GBM gliomas, rs12076373 (1q44, pM=3.56x10-7, ORM=1.41, 95% CI: 1.23-1.61; pF=1.76x10-2, ORF=1.20, 95% CI: 1.03-1.38), rs11979158 (7p11.2, pM=1.77x10-5, ORM=1.32 [95% CI: 1.16-1.49]; pF=2.73x10-1, ORF=1.08 [95% CI: 0.94-1.24]), and rs3751667 (16p13.3, pM=9.44x10-7, ORM=1.30 [95% CI: 1.17-1.44]; pF=5.18x10-2, ORF=1.13 [95% CI: 1.00-1.28]) had significant effect in males only. Effect size for rs55705857 (8q24.21) varied significantly by sex, with ORM=2.63 (95% CI: 2.24-3.09, pM=4.42x10-32), as compared to ORF=3.95 (95% CI: 3.28-4.76, pF=1.82x10-47). A sensitivity analysis was performed due to allele frequency heterogeneity by study and results did not change. CONCLUSIONS: Sex differences and other demographic differences in cancer susceptibility can provide important clues to etiology, and these differences can be leveraged for discovery in genetic association studies. Significant differences in effect size may suggest variation in genetic effect of risk alleles or in unidentified risk factors that vary in prevalence or effect by sex. There may also be differences in the distribution of molecular subtypes within each histology by sex. Further investigation using an agnostic approach may further elucidate the relationship between effect of risk alleles and sex. Citation Format: Quinn T. Ostrom, Ben Kinnersley, Margaret Wrensch, Jeanette E. Eckel-Passow, Georgina Armstrong, Terri Rice, Yanwen Chen, John Wiencke, Lucie McCoy, Helen Hansen, Christopher Amos, Jonine L. Bernstein, Elizabeth B. Claus, Dora Il'yasova, Christoffer Johansen, Daniel Lachance, Rose Lai, Ching C. Lau, Ryan T. Merrell, Sara H. Olson, Siegal Sadetzki, Joellen Schildkraut, Sanjay Shete, Richard S. Houlston, Robert B. Jenkins, Beatrice Melin, Melissa Bondy, Jill S. Barnholtz-Sloan. Estimating sex-specific effects of genetic loci associated with glioma risk [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1315. doi:10.1158/1538-7445.AM2017-1315

Abstract BACKGROUND: Glioma accounts for ~27% of all primary brain tumors and is responsible for ~13,000 cancer-related deaths in the US each year. Glioma tumors can be broadly classified into glioblastoma (GBM) and lower-grade non-GBM. Typically gliomas have a poor prognosis irrespective of medical care, with the most common form, GBM, having a five-year survival rate of only 5%. While genome-wide association studies (GWAS) have transformed our understanding of glioma susceptibility, individual studies have had limited power to identify risk loci. METHODS: We performed the largest glioma GWAS to date, comprising a meta-analysis of six existing GWAS (6,405 cases, 14,100 controls) as well as new GWAS from the Glioma International Case Control Consortium (GICC; 4,572 cases and 3,286 controls) and University of California, San Francisco (UCSF)-Mayo (1,519 cases, 804 controls), totaling 12,496 cases (6,191 classified as GBM, 5,819 as non-GBM) and 18,190 controls. RESULTS: We identified five new risk loci for GBM at 1p31.3 (rs12752552; near JAK1, P=2.04×10-9, odds ratio (OR)=1.22), 11q14.1 (rs11233250; P=9.95×10-10, OR=1.24), 16p13.3 (rs2562152; near MPG, P=1.93x10-8, OR=1.21), 16q12.1 (rs10852606; HEATR3, P=1.29×10-11, OR=1.18), 22q13.1 (rs2235573; P=1.76×10-10, OR=1.15) and eight for non-GBM at 1q32.1 (rs4252707; MDM4, P=3.34×10-9, OR=1.19), 1q44 (rs12076373; AKT3, P=2.63×10-10, OR=1.23), 2q33.3 (rs7572263; near IDH1, P=2.18×10-10, OR=1.20), 3p14.1 (rs11706832; LRIG1, P=7.66×10-9, OR=1.15), 10q24.33 (rs11598018; OBFC1, P=3.39×10-8, OR=1.14), 11q21 (rs7107785; P=3.87×10-10, OR=1.16), 14q12 (rs10131032; P=5.07x10-11, OR=1.33) and 16p13.3 (rs3751667; P=2.61×10-9, OR=1.18). Case-only analyses confirmed the specificity of 11q14.1, 16p13.3 and 22q13.1 associations for GBM and 1q44, 2q33.3, 3p14.1, 11q21 and 14q12 for non-GBM tumors. In the combined meta-analysis, among previously published glioma risk SNPs, those for all glioma at 17p13.1 (TP53), GBM at 5p15.33 (TERT), 7p11.2 (EGFR), 9p21.3 (CDKN2B-AS1) and 20q13.33 (RTEL1) and for non-GBM at 8q24.21 (CCDC26), 11q23.2, 11q23.3 (PHLDB1) and 15q24.2 (ETFA) showed even greater evidence for association. SNPs at 10q25.2 and 12q12.1 for non-GBM tumors retained genome-wide significance (i.e. P<5.0x10-8). Associations at the previously reported loci for GBM at 3q26.2 (near TERC) and 12q23.33 (POLR3B) did not retain statistical significance. CONCLUSIONS: Our findings substantiate genetic susceptibility to GBM and non-GBM glioma being highly distinct, consistent with their distinctive molecular profiles presumably resulting from different etiological pathways. Functional analyses should lead to further insights into the biological basis of the different glioma histologies. Such information can inform gene discovery initiatives and therefore have a measurable impact on the successful development of new therapeutic agents. Citation Format: Ben Kinnersley, Beatrice S. Melin, Jill S. Barnholtz-Sloan, Margaret R. Wrensch, Christoffer Johansen, Dora Il’yasova, Quinn Ostrom, and members of GICC, Karim Labreche, Jeanette E. Eckel-Passow, Paul A. Decker, Marianne Labussière, Ahmed Idbaih, Khe Hoang-Xuan, Anna-Luisa Di Stefano, Karima Mokhtari, Jean-Yves Delattre, Pilar Galan, Konstantinos Gousias, Johannes Schramm, Minouk J. Schoemaker, Sarah J. Fleming, Stefan Herms, Stefanie Heilmann, Marcus M. Nöthen, Heinz-Erich Wichmann, Stefan Schreiber, Anthony Swerdlow, Mark Lathrop, Matthias Simon, Marc Sanson, Preetha Rajaraman, Stephen Chanock, Martha Linet, Zhaoming Wang, Meredith Yeager, Rose K. Lai, Elizabeth B. Claus, Sara H. Olson, Robert B. Jenkins, Richard S. Houlston, Melissa L. Bondy. Genome-wide association study of glioma reveals specific differences in genetic susceptibility to glioblastoma and non-glioblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1302. doi:10.1158/1538-7445.AM2017-1302

Abstract Protein expression microarrays, also called antibody arrays, represent a new technology that allows the expression level of proteins to be assessed directly. As is also the case with gene expression microarrays, it is hoped that protein expression microarrays will aid in biomarker discovery, predicting disease outcomes and response to treatments, and detecting molecular mechanisms and/or pathways associated with a particular disease state. However, accurately achieving these aims is dependent upon suitable experimental designs, normalization procedures that eliminate systematic bias, and appropriate statistical analyses to assess differential expression or expose expression patterns. In the last five years, a large amount of research has been devoted to two-color cDNA arrays to improve experimental design, normalization and statistical analyses to assess differential expression and classification. These methods are directly applicable to two-color antibody arrays. The objective of this article is to discuss statistical methods that have been developed for cDNA arrays and describe how the methods can be directly applied to antibody arrays.

414 Background: While mutations in PBRM1 (~40%) and BAP1(~10%) are associated with clinical outcomes and pathologic features in clear cell renal cell carcinoma (ccRCC), the impact of protein expression of these genes remains unknown. Herein, we quantify PBRM1/BAP1 protein expression in a large cohort of patients with localized ccRCC and associate expression with cancer-specific survival (CSS) and pathologic features. Methods: We utilized the Mayo Clinic Renal Registry and identified 1,416 patients who underwent nephrectomy to treat clinically localized ccRCC between 1/3/1990 and 4/14/2009. We used immunohistochemistry (IHC) to detect PBRM1/BAP1 expression, and a central pathologist blinded to the outcomes scored tumors as either positive or negative. Tumors with heterogeneous or equivocal staining were excluded from this analysis. We generated Cox proportional hazard regression models for associations with ccRCC-SS, and we employed Mann-Whitney U tests for associations with pathologic features. Results: Of the 1,416 samples, 1,232 (87%) were PBRM1/BAP1 positive or negative, 163 (11%) had heterogeneous staining, and 21 (1%) could not be assessed. The distribution and association of PBRM1/BAP1 phenotypes with clinical outcomes are listed in the table below. PBRM1+/BAP1+ tumors have the best CSS, and PBRM1-/BAP1- have the worst. In addition, PBRM1/BAP1 expression strongly associated with the tumor size, stage, grade, and tumor necrosis (p<0.0001). Conclusions: This study is the first and largest to quantify PRBM1/BAP1 protein expression in ccRCC tumors. We were able to quantify PBRM1/BAP1 through IHC in the vast majority of tumors (87%), and PRBM1/BAP1 expression strongly associates with both CSS and pathologic tumor characteristics. Our data confirms our previous findings of the importance of PRBM1/BAP1 in the molecular pathogenesis of ccRCC. [Table: see text]

446 Background: Despite the development of prognostic algorithms based on clinico-pathologic features, the ability to identify aggressive forms of clear cell renal cell carcinoma (ccRCC) remains suboptimal. Topoisomerase IIA (TOP2a) is a biomarker of DNA replication and a target for antineoplastic agents. Herein, we evaluate the association of TOP2a expression in ccRCC tumors with pathologic features of aggressiveness and risk of cancer-specific death. Methods: We identified 947 patients who underwent nephrectomy to treat clinically localized ccRCC between January 16, 1990, and September 27, 2006. TOP2a expression was assessed using IHC and scored as number of positive cells per mm2. We evaluated TOP2a expression using a continuous variable and tertile categories. For associations with pathologic features, we employed Kruskal-Wallis tests and for associations with cancer-specific survival, we generated Cox proportional hazard regression models. Results: HigherTOP2a expression is associated with later stage, higher grade and higher Mayo SSIGN score (all p < 0.001). The risk of death from RCC increases with increasing TOP2a expression (p trend < 0.0001). Compared to patients in the lowest tertile, those patients with tumors in the highest tertile of TOP2a expression were at increased risk of RCC death (HR=2.31 95% CI 1.64-3.25; p < 0.0001). Interestingly, among those patients with low risk disease (SSIGN score 0-3; ~95% 10 year survival), those with high TOP2a were at increased risk of RCC death (HR=3.09 95% CI 1.29-7.40; p = 0.01). Conclusions: Higher TOP2a expression is associated with more aggressive pathologic features and increased risk of cancer-specific death among patients undergoing surgery for localized ccRCC. If confirmed, these data support further inquiry for TOP2a as a prognostic and predictive biomarker for ccRCC patients.

507 Background: Clear cell renal cell carcinoma (ccRCC) is a well-described molecularly heterogeneous tumor. Herein, we assessed the concordance of two of the most commonly mutated genes in ccRCC, PBRM1 (~50%), and BAP1 (~15%), in patient-matched primary and metastatic tumors. Methods: One pathologist (PK) assessed PBRM1 and BAP1 protein expression using immunohistochemistry (IHC) in 99 patients with a primary and at least one metastatic ccRCC tumor available for analysis. All available metastatic tumors were analyzed. Results: A total of 99 patients (48 M0 and 51 M1) had both a primary tumor and at least one metastatic tumor available for analysis. There were a total of 158 metastases with one patient having up to 7 metastases available for analysis. The concordance between primary and patient-matched metastasis was 87% for PBRM1 and 99% for BAP1. We observed a similar concordance between patients with M0 versus M1 disease. Conclusions: While ccRCC is molecularly heterogeneous, PRBM1, and BAP1 are largely concordant between primary and metastatic lesions suggesting that PBRM1 and BAP1 are genetically truncal events in the molecular pathogenesis of ccRCC.

447 Background: Components of the peripheral blood counts (PBC) are associated with poor prognosis in patients diagnosed with non-metastatic clear cell renal cell carcinoma (ccRCC). Related to this, higher tumor expression of programmed death ligand one (PDL1) is also associated with poor ccRCC prognosis. Herein, we utilize a large cohort of patients undergoing nephrectomy for localized ccRCC to evaluate the association of PBC with tumor expression of PDL1. Secondarily, we examine the association of PBC and relapse free survival (RFS) after adjusting for age and PDL1 expression. Methods: Through chart review, we obtained PBC values at time of nephrectomy for patients enrolled in the Mayo Clinic Renal Registry from 1990-2009 who were treated for localized (M0) ccRCC including the following: hemoglobin (HGB), white blood cell (WBC), absolute neutrophil count (ANC), absolute lymphocyte count (ALC), absolute eosinophil count (AEC), absolute monocyte count (AMC), and platelets (PLT). We determined tumor PDL1 expression as a continuous variable (% of positive PDL1 tumor cells) using the 5H1 antibody on archived tissues. We analyzed the correlation of PBC values with PDL1 expression as a continuous variable using a linear estimate. To evaluate the association of PBC values with RFS after adjusting for PDL1 expression, we employed multivariate Cox regression models. Results: A total of 706 ccRCC patients had available PDL1 expression and PBC. PDL1 expression was inversely associated with HGB (linear estimate of -0.02, p<0.001) and positively associated with platelets (linear estimate 1.73, p<0.001) whereas there was no association with PDL1 with WBC, ANC, AEC, or AMC. After adjusting for age and PDL1 expression, a higher HGB (HR=0.87, p<0.001), higher ALC (HR 0.68, p=0.003), and higher AEC (HR 0.12, p=0.002) were all associated with improved relapse free survival. Conversely, higher PLT was associated with decreased RFS (HR 1.002, p<0.001). Conclusions: PDL1 expression is associated with decreased HGB and increased PLT at the time of nephrectomy. After adjusting for PDL1 expression and age, higher HGB, ALC, and AEC are associated with improved RFS, while higher platelets are associated with decreased RFS.

174 Background: AA/P is an FDA approved treatment for mCRPC. Since markers of early resistance to AA/P are unknown, we report initial findings of androgen receptor (AR) based associations with short term (12 week) progression on AA/P. Methods: mCRPC stage patients (pts) initiating pre-chemotherapy AA/P underwent metastatic site biopsies at baseline (pre AA/P) and after 12 weeks. Baseline somatic whole exome DNA, tumor gene expression for AR full length (ARFL), AR splice variant 7 (ARV7) and ARV7/ARFL ratios were compared in progessors versus non-progressors. Progression at or within 12 weeks of AA/P therapy was defined as death or disease progression by PCWG2 “composite progression (CP)” and/or “radiographic progression” endpoints. Wilcoxon rank-sum tests were used to test for differences in the two groups for comparing ARFL, ARV7 expressions and ARV7/ARFL ratios and chi square tests were used for differences in copy number variation. Results: Between 1/2013 and 6/2014, 59 pts were enrolled of which 44 have disease assessment data at the12-week time point. CP was observed in 17/44 patients. DNA seq and clinical data was available for 42/44 pts. Using radiographic progression at 12 weeks, AR Amplification/Gain was observed in 20/26 non-progressors (13 with Amplification) and in 9/16 progressors (P-value = 0.19; OR 2.5). ARFL and ARV7 gene expressions in both groups is provided in table. Conclusions: A trend towards higher ARV7/ARFL gene expression ratio in metastases was observed with early progression on AA/P. AR gain/amplification is observed less often in pts with early progression. Validation of these findings is on-going in this prospective trial. Clinical trial information: NCT# 01953640. [Table: see text]

AbstractPleomorphic xanthoastrocytoma (PXA) is a rare localized glioma characterized by frequent BRAF V600E mutation and CDKN2A/B deletion. We explored the association of copy‐number variants (CNVs) with BRAF mutations, tumor grade, and patient survival in a cohort of 41 PXA patients using OncoScan chromosomal microarray. Primary resection specimens were available in 38 cases, including 24 PXA and 14 anaplastic PXA (A‐PXA), 23 BRAF V600E mutant tumors (61%). CNVs were identified in all cases and most frequently involved chromosome 9 with homozygous CDKN2A/B deletion (n = 33, 87%), a higher proportion than previously detected by comparative genomic hybridization (50%–60%) (37). CDKN2A/B deletion was present in similar proportion of PXA (83%), A‐PXA (93%), BRAF V600E (87%), and wild‐type (87%) tumors. Whole chromosome gains/losses were frequent, including gains +7 (n = 15), +2 (n = 11), +5 (n = 10), +21 (n = 10), +20 (n = 9), +12 (n = 8), +15 (n = 8), and losses −22 (n = 11), −14 (n = 7), −13 (n = 5). Losses and copy‐neutral loss of heterozygosity were significantly more common in A‐PXA, involving chromosomes 22 (P = 0.009) and 14 (P = 0.03). Amplification of 8p and 12q was identified in a single tumor. Histologic grade was a robust predictor of overall survival (P = 0.003), while other copy‐number changes, including CDKN2A/B deletion, did not show significant association with survival. Distinct histologic patterns of anaplasia included increased mitotic activity in an otherwise classic PXA or associated with small cell, fibrillary, or epithelioid morphology, with loss of SMARCB1 expression in one case. In 10 cases, matched specimens were compared, including A‐PXA with areas of distinct low‐ and high‐grade morphology (n = 2), matched primary/tumor recurrence (n = 7), or both (n = 1). Copy‐number changes on recurrence/anaplastic transformation were complex and highly variable, from nearly identical profiles to numerous copy‐number changes. Overall, we confirm CDKN2A/B deletion as key a feature of PXA not associated with tumor grade or BRAF mutation, but central to the underlying genetics of PXA.