Citations

Abstract Four gene expression subtypes of high-grade serous ovarian cancer (HGSC) have been previously described. In these early studies, a fraction of samples that did not fit well into the four subtype classifications were excluded. Therefore, we sought to systematically determine the concordance of transcriptomic HGSC subtypes across populations without removing any samples. We created a bioinformatics pipeline to independently cluster the five largest mRNA expression datasets using k-means and nonnegative matrix factorization (NMF). We summarized differential expression patterns to compare clusters across studies. While previous studies reported four subtypes, our cross-population comparison does not support four. Because these results contrast with previous reports, we attempted to reproduce analyses performed in those studies. Our results suggest that early results favoring four subtypes may have been driven by the inclusion of serous borderline tumors. In summary, our analysis suggests that either two or three, but not four, gene expression subtypes are most consistent across datasets.

Abstract Background: Little transcriptomic research has compared epithelial ovarian cancer (EOC) histological subtypes. We set out to characterize the transcriptomes of high-grade serous carcinomas (HGSC) and endometrioid carcinomas (EC), which make up around 70% and 20% of EOC tumors, respectively, and have some histopathological similarities. Methods: Fresh frozen tumors from EOC patients seen at the Mayo Clinic (30 EC and 62 HGSC) were used. 1ug RNA riboZero was used for library preparation using the Illumina TruSeq kit and sequenced on a HiSeq 2000 machine. Reads were aligned using TopHat2 followed by quantification of abundances using RSEM and differential expression analysis with edgeR. We analyzed transcriptomes, conducted pathway analyses, and summarized key candidate gene sets. Expressed SNVs (eSNVs) from the RNA-seq data were determined using GATK and RVboost. Results: The analysis found 699 genes with FDR < 1×10-5 for differential expression between HGSC and EC, with most genes being up-regulated in EC. The top most associated genes were TPH1, MAP2K6, KLK2, ADAM23, TESC and TRAF3IP2 (p<10-22). Pathway analysis of the genes up-regulated in EC revealed enrichment of the “basal cell carcinoma signaling pathway” (p = 1.2×10-5). Within 1 Mb of the 25 known EOC risk loci, we observed higher expression in HGSC for RSPO1 and HPSE (p <5×10-7). For genes functionally related to EOC, we observed in HGSC up-regulation (*p < 10-5, ⁁p<0.003) for FOXM1⁁, CDKN2A⁁, CCNE1*, CCND2⁁, PIK3CA*, BRCA2⁁, BIRC5⁁, MMP9⁁, FANCD2⁁, and MAML2⁁. In contrast, we found up-regulation in EC for MDM2*, KLK4*, BCL2⁁, CCND1*, ANXA4*, CDH1⁁, MMP7⁁, and MAML3⁁. We also identified 204 eSNVs (44 non-synonymous) associated with EC v HGSC subtype (p<10-4); this included an exonic TRAF3IP2 eSNV (66% EC, 13% HGSC, p = 4×10-7, chr6:111877117). Discussion: Using one of the largest sets of identically processed fresh-frozen EOC tumors, some patterns emerged among the numerous EC v HGSC transcriptomic differences. TPH1, up-expressed in EC, is regulated by SOX4 which was also up-regulated in EC. Two sets of genes related to Kallikreins serine proteases were differentially expressed, including KLK2 which is known to regulate EGFR and pro-inflammatory cytokines and is regulated by MYC. Lastly, TRAF3IP2 encodes for a protein involved in regulating cytokines through members of the NFKB pathway. Conclusions: These findings suggest important biological insights into one of the rarer EOC histologies and may aid in the development of targeted treatment options. Research is on-going to incorporate additional features (e.g., DNA methylation, copy number) into a “systems biology” framework to better understand the molecular differences between EOC histologies. Citation Format: Brooke L. Fridley, Junqiang Dai, Rama Raghavan, Chen Wang, Pengcheng Lu, Stacey Winham, Madalene Earp, Kate Lawrenson, Simon A. Gayther, Kimberly R. Kalli, Ellen L. Goode. Transcriptome characterization of high grade serous and endometrioid epithelial ovarian cancer tumors. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1833.

Abstract Purpose: Ovarian cancer (OC) patients received maximum benefits when complete dissection with no macroscopic disease is achieved. It has been suggested that surgical outcome is associated with tumor biology and therefore can be predicted from tumor molecular characteristics such as gene expression levels. Noticeably, several debulking-associated genes have been independently reported based on meta-data analysis and further validations in individual center's cohorts. Furthermore, a recent OC study of the cancer genome atlas (TCGA) project revealed the presence of four expression subtypes in high-grade serous ovarian cancer (HGSOC). We have independently confirmed this association with prognosis in our OC cohort. With the awareness that surgical practices and rates of complete resection vary among institutions, we sought to evaluate whether expression changes of previously reported debulking-associated genes are associated with complete dissection in a HGSOC cohort in a center with an aggressive surgical approach. Further, we also tested if the HGSOC four expression subtypes are associated with complete dissection. Experimental Design: We constructed a cohort of 262 stage III/IV HGSOC patient cases diagnosed in Mayo Clinic between 1993 and 2011. Cases were categorized into two groups based upon debulking outcome: 84 complete resection cases with no gross visible residual disease (RD0), 178 cases with residual disease greater than 0 cm (RD1). Tumor mRNA expression was previously measured using Agilent 4x44k platforms. Eight previously published debulking-associated genes were evaluated for their expression differences between RD0 and RD1 patient groups using t-test: FABP4, ADH1B, POSTN, CXCL14, FAP, NUAK1, PTCH1 and TGFBR2. For testing subtype-debulking associations, we first quantified molecular subtype activation scores of individual tumors using single-sample gene set enrichment analysis (ssGSEA), and computed standardized scores after permutation-based scaling adjustments to account for gene set size difference. Then, standardized subtype scores are associated with complete dissection (RD0) vs. other surgical outcome (RD1) using logistic regression. The independent subtype-associations were further evaluated by adjusting known debulking outcome factors, including age, ascites, tumor grade and stage. Results: For eight previously reported debulking-associated genes, only one demonstrated a significant expression difference between RD0 and RD1 groups: CXCL14, t-test p-value = 9.2x10-3. When comparing debulking vs. HGSOC subtypes, we observed a significantly reduced probability of complete resection for the mesenchymal subtype (p-value = 1.1x10-2, OR = 0.56, 95% CI [0.35, 0.87]). Consistent with this observation, we have previously described the mesenchymal subtype as a poor prognosis group. On the other hand, the immunoreactive subtype was associated with a higher likelihood of complete resection (p-value = 9.4 x 10-3, OR=2.19, 95% CI = [1.22, 4.01]). The significant associations between these two subtype activities and complete resection in advanced stage disease remained after adjusting for age, ascites, tumor grade and sub-stage: mesenchymal (p-value = 1.5 x 10-2 ,OR = 0.48, 95% CI = [0.26, 0.85]); immunoreactive subtype (p-value = 4.8 x 10-2 , OR = 2.11, 95% CI = [1.04, 4.53]). Conclusion: This single center HGSOC analysis demonstrates the important of validation of debulking signatures in centers with diverse rates of complete resection. We were only able to validate an association with complete resection for one of eight previously reported debulking-associated genes. We also showed that HGSOC subtypes have significant associations with likelihood of complete dissection. Further studies are warranted for confirming subtype associations in other institutional HGSOC cohorts. Citation Format: Chen Wang, Jamie Bakkum-Gamez, Andrea Mariani, William Cliby, Ellen Goode. Tumor expression subtypes of high-grade serous ovarian cancer demonstrate associations with likelihood of complete resection: a single-center study. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research: Exploiting Vulnerabilities; Oct 17-20, 2015; Orlando, FL. Philadelphia (PA): AACR; Clin Cancer Res 2016;22(2 Suppl):Abstract nr B80.

Abstract Epithelial ovarian cancer (EOC) is the fifth leading cause of cancer death among women in the United States (5% of cancer deaths). The standard treatment for patients with advanced EOC is initial debulking surgery followed by carboplatin-paclitaxel combination chemotherapy. Unfortunately, even with modern chemotherapy, most patients relapse and die with the five-year overall survival around 45%. In addition, those patients that initially respond to taxane-platinum therapy eventually develop platinum-resistant tumors and relapse. Thus, finding novel therapeutics for treating EOC is essential. One approach that has been used widely in cancer drug discovery is Connectivity Mapping (CMAP) using gene expression data and drug phenotype data assessed on cancer cell lines (CCL) (i.e. “sensitive” or “resistant” based on dose-response drug data). However, recent publications have highlighted some issues with the use of these standard CCL sets that are often used in these drug screening studies. Therefore, instead of CMAP analysis based on gene expression signatures developed on drug response data collected on a small set of EOC cell lines, we took a novel CMAP approach based on information collected on EOC patients and clinical endpoints. That is, we determined the gene expression signature (e.g., set of genes) that were associated with time to recurrence in data collected on EOC patients within TCGA (n = 518) and a Mayo Clinic study (n = 474). The set of genes associated with time to recurrence (with and without adjustment for additional clinical covariates) was completed separately for TCGA and Mayo Clinic studies, restricting all analyses to EOC tumor of high-grade serous histology. Each of these 4 sets of genes, where genes with hazard ratios (HR) > 1 coded as “positively” associated and genes with HR < 1 coded as “negatively” associated with the phenotype, were inputted into CMAP software (Broad Institute) to determine a set of drugs for which our signature “matches” the “reference” signature. The drugs that overlapped between the CMAP analyses on the two studies were carried forward for validation studies involving drug screens on a set of 10 EOC cell lines. Of the 11 drugs carried forward, we a priori hypothesized that 5 of these drugs would kill EOC cells (mitoxantrone, podophyllotoxin, wortmannin, doxorubicin, and 17-AAG), for which all 5 showed an ability to kill EOC cells in vitro (cell lines were treated using serial dilutions of the drugs (20-0 μM) for 72 h followed by measuring cell viability using the CellTiter-Blue assay). Further research is on-going to understand the functional mechanism by which these candidate drugs might treat EOC and to apply this bioinformatics approach to other cancer drug screening studies. Note: This abstract was not presented at the meeting. Citation Format: Brooke L. Fridley, Rama Raghavan, Gottfried E. Konecny, Chen Wang, Ellen L. Goode, Harsh B. Pathak, Stephen Hyter. A bioinformatics approach to drug discovery: patient based connectivity mapping. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4853. doi:10.1158/1538-7445.AM2015-4853