Abstracts - Invited Talks

Invited Abstracts Concurrent Session 1 Concurrent Session 1 Clinical Pharmacogenetics: Assessing Global Diversity CN01-01 Unbiased Cell-Based Models to Identify GeneticVariants Associated with Chemotherapeutic Induced Toxicity. R. Stephanie Huang, Wei Zhang, Sunita Shukla, Shiwei Duan, Wasim Bleibel, Christine Hartford, M. Eileen Dolan. University of Chicago, Chicago, IL. The International HapMap Project provides a key resource of genotypic data in human lymphoblastoid cell lines (LCLs) derived from populations of European, African, Chinese and Japanese ancestry to associate with various phenotypic data to find genes affecting disease and response to drugs. These cell lines have allowed for the exploration of quantitative variation in complex traits such as gene expression and dug response in human populations. To identify genetic variants that contribute to gene expression and chemotherapy-induced cytotoxicity, we developed unbiased, whole genome cell based models. By integrating genotype, gene expression and sensitivity of HapMap cell lines to chemotherapeutic drugs, we built predictive models to identify genetic signatures for susceptibility to chemotherapy-induced cytotoxicity. Cell lines derived from 30 trios of European descent (CEU) and 30 trios of African descent (YRI) were utilized. Cell growth inhibition at increasing concentrations of chemotherapeutic agents (etoposide, daunorubicin, AraC, cisplatin or carboplatin) was evaluated. Gene expression of 176 HapMap cell lines (87 CEU and 89 YRI) was determined using the Affymetrix GeneChip® Human exon 1.0 ST Array. SNP genotype and the IC50 (concentration required to inhibit 50% cell growth) of each agent were linked through whole genome association in cell lines from combined and independent CEU and YRI populations. A second association test was performed between SNP genotype and gene expression, and linear regression was then utilized to evaluate the correlation between gene expression and drug IC50. We identified genetic variants that significantly associate with sensitivity to the cytotoxic effects of these chemotherapeutic agents through differences in gene expression in the combined CEU and YRI populations, as well as each population individually. A multivariate model indicated that these limited numbers of genetic variants explained approximately 20-65% of the observed human variation in cellular sensitivity to chemotherapy. There are advantages and limitations in the use of LCLs in pharmacogenomic discovery that will be discussed as well as the wide range of cellular phenotypes that can be studied in these cell lines to ultimately gain insight as to how genetic variation contributes to phenotypic variation. These cell lines provide a novel means to answer important pharmacogenomic questions. This Pharmacogenetics of Anticancer Agents Research (PAAR) Group (http://pharmacogenetics.org) study was supported by NIH/NIGMS grant UO1GM61393. CN01-02 Ethnicity and Pharmacogenetics of Anticancer Therapy. Boon Cher Goh1, Soo-Chin Lee2, Lee How-Sung3, Lu Fan3. 1Yong Loo Lin School of Medicine, Singapore; 2National University Hospital, Singapore; 3Yong Loo Lin School of Medicine, Singapore. The human genome is 99.6-99.8% identical; however, 0.2-0.4% variation gives up to 10 million DNA variants, of which 10% may be found between populations of different geographical groups. When interethnic variation of variant frequencies exist in important genes that determine drug disposition or effects, population differences in drug toxicity or efficacy from standard doses of drug would be apparent. Recently, clinically important differences in warfarin dosing requirements between ethnic groups has been attributed mainly to genetic polymorphisms of its drug metabolising enzyme cytochrome P-450 2C9 (CYP2C9) and its target enzyme of inhibition vitamin K epoxide reductase (VKORC1). For anticancer drugs, activating mutations of the epidermal growth factor receptor tyrosine kinase (EGFR TK) domain are more common in Asian lung cancers, resulting in higher responses to EGFR TK inhibitors gefitinib and erlotinib. Docetaxel may have higher myelosuppression in Chinese than reported in literature when given at 75mg/m2 q3w. In a phase II study, standard docetaxel/carboplatin for non-small cell lung cancer appears to more toxic but has a higher response rate in Singaporean compared to Australian patients. In a study population of 101 breast cancer patients receiving alternating cycles of single agent docetaxel and doxorubicin, Chinese patients had more myelosuppression from doxorubicin than Indian or Malay patients, and together with Indians had more myelosuppression than Malays. This interethnic variability in toxicity from chemotherapy was not explained by common polymorphisms in CYP3A5, PXR, CAR or HNF4α. Doxorubicin is metabolized intracellularly to doxorubicinol, a metabolite with 10-fold less cytotoxicity by short-chain dehydrogenases/reductases carbonyl reductases, CBR1 and CBR3. Therefore, we studied polymorphisms of CBR1 and CBR3 on doxorubicin pharmacokinetics and hematologic toxicity in this group of patients. CBR3 11G>A and 730G>A were common polymorphisms, and CBR3 11G>A (C4Y) was associated with lower conversion of doxorubicin to doxorubicinol, and consequently greater myelosuppression. Concordantly, Chinese had a lower frequency of the more active CBR11G allele than Indians, suggesting that this polymorphism may contribute to the interethnic variability of doxorubicin induced myelosuppression observed. An analysis of CBR3 RNA levels in tumour tissue showed higher expression in patients with CBR311GG genotype, suggesting increased transcriptional efficiency compared to the AA genotype. In summary, ethnogeographic factors may have significant impact on pharmacodynamic variability of anticancer drugs, and warrants more work to elucidate the genetic determinants of this variability. CN01-04 Pharmacogenetics: Views from the Pharmaceutical Industry. Eric H. Lai. GlaxoSmithKline, Research Triangle Park, NC. The completion of the Human Genome Project, the tremendous success of The SNP Consortium and the International HapMap Project have provided critical knowledge for understanding human diversity and have promised to greatly improve Healthcare. The application of genetics and genomics in the Pharmaceutical industry over the past decade has been focused on drug target discovery. In reality, the bottlenecks in the discovery and approval of new drugs are in the attrition rate in drug development (drugs with no positive efficacy) and adverse drug reactions (drugs with positive therapeutic indications but also negative side effects). This presentation will review the current status and barriers on the application of Pharmacogenetics to drug development and post-market management. Concurrent Session 2 Ultra-High Throughput Sequencing Technologies: Towards Patient Genome Sequencing CN02-03 Enabling Key Applications with the SOLiD™ System. Robert Nutter. Applied Biosystems, Foster City, CA. The decreasing cost of sequencing is driving the development of genome-wide applications based on sequence data. Instead of using different platforms for applications such as gene expression, genotyping and sequencing, it is becoming possible to use sequencing as the universal currency for most, if not all, genetic analysis. The SOLiD™ System is a ligation-based, massively parallel sequencing system that can currently generate in excess of 3 Gb of DNA sequence per run. Investigators can elect to use either fragment or mate-paired libraries and can run one or multiple samples at a time. We are taking advantage of the high throughput, accuracy, and scalability of the SOLiD™ System to collaborate with a number of scientific groups to develop a range of applications. My talk will provide current results demonstrating the power of sequencing fragment or mate paired libraries to provide data for a number of important genetic analysis applications, including transcriptome analyses, CpG methylation patterns, nucleosome phasing as well as large insertion/deletion identification and mapping. 88 AACR Centennial Conference Concurrent Session 3 Invited Abstracts Concurrent Session 3 Hereditary Cancers CN03-04 Recent Advances in Familial Colorectal Cancer Genetics and Diagnostics. Peh Yean Cheah1, Xia Cao1, Yi Hong1, Yu Hui Wong1, Lai Fun Thean1, Carol Loi1, Marian P. Kumarasinghe1, Hui Hua Li2, Yuk Ping Chau1, Kong Weng Eu1. 1Singapore General Hospital, Singapore; 2National Cancer Institute, Singapore. Familial adenomatous polyposis (FAP) is a heritable form of colorectal cancer (CRC) caused by autosomal dominant inheritance of the mutated adenomatous polyposis coli (APC) gene. It is characterized by the appearance of hundreds to thousands of adenomatous polyps in the colon and rectum of an affected individual by the second or third decade of his life. The risk of cancer is virtually 100 % if the polyps are not detected and removed in time. Coupling the protein truncation test (PTT) and direct sequencing, multiplex ligation-dependent probe amplification (MLPA) and differential allelic expression techniques, our APC mutation detection rate in FAP patients has now reached 94% which is, to our knowledge, amongst the highest detection rate in the world. South East Asian patients have some features similar to and other features distinct from Caucasian patients. The APC mutation-negative FAP variants have been screened for Axin 1, β-catenin, and MYH, three genes variously implicated in CRC tumorigenesis, and found to be mutation negative in all three genes. FAP variant patients can be classified into two groups according to the polyps and atypical clinical features. The first variant group has mixed ‘adenomatous-hyperplastic-juvenile’ polyps and the second variant group has classical polyposis (i.e. with ≥ 100 adenomatous polyps) but accelerated disease progression (statistically more CRC despite later age of diagnosis compared to FAP patients with APC mutations). The first variant group has now been renamed as Hereditary Mixed polyposis Syndrome (HMPS) patients. We have mapped the HMPS locus to chromosome 10q23 by genomewide linkage scan in a three-generation family using the GeneChip Human Mapping 10K array and subsequently confirmed it by fine mapping with microsatellite markers in two families. Candidate gene screening in the 7 Mb critical region identified an 11-bp deletion in exon 2 of the bone morphogenesis protein receptor 1A (BMPR1A) gene as the underlying germline defect in one such family. The fully penetrant mutation resulted in a truncated protein that deleted all the functional domains of BMPR1A, including the Activin receptor domain in the 5’extracellular region. Since then, we have identified BMPR1A mutations in several other families and one sporadic case, suggesting therefore that inactivating BMPR1A of the BMP pathway can initiate colorectal tumorigenesis via the ‘mixed polyposis-carcinoma’ sequence in both hereditary and sporadic CRC. For the second variant group, we have performed genomewide genotyping with the GeneChip Human Mapping SNP 5 array. We identified by copy number analysis an identical deleted region in all eight polyps from two affected members of the same family. We are currently verifying the region of interest in pursuit of the new tumor suppressor. The identification of new tumour suppressors would illuminate further the tumorigenesis process and enable presymptomatic detection, better surveillance, and management of patients. age-specific rates at older ages show several hundred-fold variation, and incidence rates can change very rapidly over time. The highest rates of CRC in the world are in Japan. The degree of difference from place to place and the rapidity of change over time are explainable, not by genes, but by changes in the environment – or perhaps by their interaction. Rare familial forms of CRC are due to deleterious mutations in known and unknown genes; nonetheless, the large majority of CRC is not associated with a known inherited genetic lesion or even a strong family history. Known environmental factors increase risk (e.g., meat, smoking) and others reduce risk (e.g., vegetables, folate, aspirin, post-menopausal hormones, calcium and vitamin D). Many less dramatic genetic changes (polymorphisms) are relevant only in the presence of environmental factors. In 1980, we proposed that estrogens would reduce risk of colon cancer and subsequently showed that, indeed, oral contraceptives and higher parity were associated with lower risk of colon cancer. The results of the Women’s Health Initiative trial show that PMH prevents CRC: more than 2 decades from original hypothesis to confirmation in a clinical trial. The role of aspirin and other NSAIDs in lowering risk of CRC is well established, beginning with observations in animals, through the study showing that NSAIDs were effective even in FAP patients, to the results of randomized clinical trials of prevention of metachronous polyps. NSAIDs consistently lower risk of polyps and cancer in average-, moderate-, and high-risk individuals; longer duration of use predicts lower risk; and risk resumes after use is stopped. Despite the emergence of clear evidence for the chemopreventive capacity of specific agents, the implementation of preventive regimens is problematic because of: differences in risk in the population, determined by age, genetic inheritance, lifestyle, and history; differences between the sexes; differences in effectiveness of agents; differences in the timing of use across the lifespan; and differences in the acceptable degree of possible harm. Primary prevention by weight control and physical activity and secondary prevention by screening, early detection, and polypectomy contribute significantly to the ability of populations and individuals to lower their risk of CRC; the duration of protection following screening by sigmoidoscopy may be as long as 15 years or more. CN04-03 Cyclin D1 as a Target for Cancer Chemoprevention. Ethan Dmitrovsky, Konstantin H. Dragnev, Qing Feng, Xi Liu, Fabrizio Galimberti, David Sekula, Steven Blumen, Candice Black, Vincent Memoli, David Johnstone, William Nugent, James R. Rigas, Sutisak Kitareewan, Sarah J. Freemantle. Dartmouth Medical School, Hanover, NH. Tobacco carcinogen-treatment of immortalized human bronchial epithelial (HBE) cells uncovered novel cancer chemoprevention targets. Treatment of these cells with classical and non-classical retinoid receptor agonists highlighted induced cyclin D1 proteasomal degradation as a molecular pharmacologic target for cancer chemoprevention. We previously reported cyclin D1 was often aberrantly expressed in human pre-malignant and malignant lung tissues. This helped provide a rationale to target cyclin D1 in clinical trials. To understand engaged mechanisms, the role of cyclin D1 in chemoprevention was studied in relevant preclinical models. Chemoprevention of tobacco-carcinogen exposed HBE cells by retinoic acid receptor (RAR) and retinoid X receptor (RXR) agonists, among other agents, was linked to cyclin D1 proteasomal degradation. This was proposed as a chemopreventive mechanism since induced cyclin D1 proteolysis conferred cell cycle arrest at G1 and in turn permitted repair of genomic DNA damage by carcinogens. Threonine 286 mutation stabilized cyclin D1 protein, implicating a phosphorylation event in this regulation. A phospho-specific anti-cyclin D1 antibody that can recognize phosphorylation changes at threonine 286 was used to confirm in immunoblot analysis that phosphorylation occurred. Glycogen synthase kinase (GSK) inhibitors revealed this kinase regulated post-translational regulation of cyclin D1, but not other D-type cyclins. To elucidate which of the 18 lysines present in cyclin D1 mediated ubiquitin-dependent degradation, these residues were engineered with individual or multiple Concurrent Session 4 Cancer Prevention CN04-01 Preventing Colorectal Neoplasia Methods: Means and Myths. John D. Potter. Fred Hutchinson Cancer Research Center, Seattle, WA. Colorectal cancer (CRC) is the third most common cause of cancer death after lung and stomach cancer, with approximately 1 million new cases and over half a million deaths, worldwide, each year. This cancer shows 20- to 30-fold variation in age-adjusted rates internationally, the Translational Cancer Medicine: Technologies to Treatment • November 4-8, 2007 • Singapore 89 Invited Abstracts Concurrent Session 4 mutations. Domains responsible for triggering cyclin D1 degradation were identified and these results will be presented. These stabilizing mutations shared an ability to preferentially localize cyclin D1 to the nucleus likely sequestering this protein from cytosolic degradation enzymes. An important role for cyclin D1 in chemoprevention was independently shown using small interfering RNAs (siRNAs) that targeted this species for repression. To uncover other involved degradation programs, gene profiling experiments were performed using carcinogen-transformed and chemoprevented HBE cells. Those studies independently identified the E1like ubiquitin-activating enzyme (UBE1L) and its physical partner, ISG15, as engaged in regulating cyclin D1 stability. Since treatment with epidermal growth factor (EGF) augmented HBE cell growth and cyclin D1 expression, an epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) was used to block these effects. Together, these in vitro experiments provided a strong rationale to conduct proof of principle trials using agents that targeted cyclin D1. Proof of principle trials monitor changes in pharmacodynamic targets in pre-treatment versus post-treatment tumor biopsies. Changes are then related to pharmacokinetic measurements of a drug in plasma and tumor. Results of proof of principle trials will be presented that use a rexinoid (bexarotene), an EGFR-TKI (erlotinib), or a bexarotene and erlotinib combination regimen to target effectively cyclin D1 expression in lung cancer cases. A phase I clinical trial using bexarotene and erlotinib to target lung and aerodigestive tract cancers showed encouraging anti-tumor activity against lung cancer. This set the stage for an ongoing phase II combination therapy trial with bexarotene and erlotinib for relapsed lung cancer, as will be discussed. This led to translating this work back to the laboratory using useful animal models for lung cancer chemoprevention studies. One that will be presented is a human surfactant C promoter-driven cyclin E mouse transgenic model that targets expression to the lung and spontaneously activated cyclin D1 expression. Notably, these mice develop pre-malignant and malignant (adenocarcinoma) lung lesions with features highly reminiscent of those found in clinical lung carcinogenesis. Stable cell lines have been derived from these murine lung cancers. This makes possible the study of antineoplastic effects of chemopreventive agents ex vivo. These cells are able to form lung tumors following tail vein injections into syngeneic mice. This opens up the prospect for rapid assessment of chemopreventive agent effects in vivo. Taken together, these and other findings that will be presented strongly implicate cyclin D1 as a novel molecular pharmacological target for cancer chemoprevention. Encouragingly, the various prognostic classifiers have proven to be highly concordant with each other in retrospective validation studies, despite having only a few overlapping genes. Work by our group has shown that one reason for this may be that these multigene predictors reflect similar oncogenic pathways where genes involved in proliferation and the cellcycle represent the common driving force. The potential for these molecular signatures is vast but their entry into the clinic can only take place after prospective validation for robustness and reproducibility. Many of the studies to date, in fact, have been small retrospective series using different patient populations, different microarray platforms and different statistical models. The results of large prospective, randomized trials such as MINDACT (Microarray In Node negative Disease may Avoid ChemoTherapy) and TAILORx (Trial Assigning IndividuaLized Options for Treatment, will provide the much needed Level 1 evidence for at least two of the multigene predictors, facilitating the shift from “bench-to-bedside”. However, validation strategies for “predictive” signatures still lag behind. To help address this issue, we will present a potential new model for accelerating that “bench-to-bedside” shift. Once validated, the ultimate goal for these molecular signatures is to be part of an integrative decision-making algorithm based on multiple levels and sources of prognostic and predictive data. Alongside traditional clinico-pathological risk parameters, molecular tools such as gene expression profiles and developments in functional proteomics will enable us to treat patients with the therapies that are most likely to confer individual benefit and, equally important, to spare patients from unnecessary treatments and associated toxicities. Concurrent Session 6 Systems Medicine CN06-02 A PK/PD and Systems Biology Approach for Antiangiogenic Drug Discovery. Jonathan M. Yingling. Eli Lilly and Company, Indianapolis, IN. Angiogenesis is one of the key biological processes in tumor growth and metastasis. Clinical validation of angiogenesis as a therapeutic target fueled the burgeoning pipeline of small molecule and biologics which target this mechanism. Yet, we are still at the beginning of understanding the diverse ligands and receptors involved in this process. We are interested both in developing novel therapeutics based on our targeted PK/PD strategy and novel systems biology approaches to better understand the angiogenic process and its role in tumor progression. We have developed a systems biology approach involving a co-culture assay of endothelial cells and fibroblasts. High content image analysis allows discrimination of complex morphological features during in vitro angiogenesis. ECs form cord like capillary structure when co-cultured with fibroblasts in vitro over 9 to 12 days. The EC competence and the fibroblast supportiveness of cord formation are highly cell passage dependent. We performed complementary gene expression analysis on both EC and fibroblasts under optimal and suboptimal conditions to model angiogenesis control in vitro. We established an interactome from differentially expressed genes (DEGs) in this co-culture system. The interactome consists of 26000 interactions and represents a scale free network. Major signaling pathways involved in angiogenesis regulation are mapped in this network along with feedback and resistance mechanisms and oncogene and tumor suppressor genes. A phenotypic drug discovery effort has been initiated to identify novel small molecule inhibitors of angiogenesis utilizing the co-culture system and efforts are underway to enable in vivo characterization of the molecular mechanisms of antiangiogenesis of compounds identified in this screen. Details of our chemoand bio-informatic approach to studying angiogenesis will be presented. Concurrent Session 5 Advances in Cancer Therapeutics CN05-03 When will Molecular Fingerprints Dominate Cancer Treatment Selection? Martine J. Piccart-Gebhart, Phuong Dinh, Christos Sotiriou. Jules Bordet Institute - Université Libre de Bruxelles (ULB), Brussels, Belgium. Historically, the selection of adjuvant systemic therapy in early breast cancer has relied on risk assessment incorporating both patient-related and tumor-related prognostic factors. These prognostic factors serve to characterize the background level of risk of relapse against which the benefits and burdens of adjuvant therapies are weighed. Tumor-related factors include lymph node involvement, tumor size, tumor grade and estrogen-receptor (ER) and HER-2 status. With the advent of microarray technology, unprecedented opportunities have become available for performing comprehensive molecular and genetic profiling of breast cancer. This has resulted in a) a new molecular classification of breast cancer into clinically relevant subtypes, b) various molecular “prognostic” signatures that have performed, as well as, if not better than, traditional clinico-pathological prognostic factors in predicting clinical outcome, and c) a number of molecular signatures that bear the exciting promise of being able to identify subsets of patients who are particularly sensitive to a given drug or regimen of drugs (“predictive” signatures). 90 AACR Centennial Conference Concurrent Session 6 Invited Abstracts CN06-03 A Robustness-based Approach to Systems-oriented Drug Design. Hiroaki Kitano. The Systems Biology Institute, Tokyo, Japan. Many potential drugs that specifically target a particular protein considered to underlie a given disease have been found to be less effective than hoped, or to cause significant side effects. The intrinsic robustness of living systems against various perturbations is a key factor that prevents such compounds from being successful. By studying complex network systems and reformulating control and communication theories that are well established in engineering, a theoretical foundation for a systemsoriented approach to more effectively control the robustness of living systems, particularly at the cellular level, could be developed. Here, I use examples that are based on existing drugs to illustrate the concept of robustness, and then discuss how a greater consideration of the importance of robustness could influence the design of new drugs that will be intended to control complex systems. CN06-04 Genomic and epigenomic profiling in cancer. Hiroyuki Aburatani. University of Tokyo, Tokyo, Japan. Any individual cancer is not identical. Recent advances in genomic and epigenomic analysis are leading to a molecularly based reclassification of cancer, and hopefully will create many opportunities for improved cancer treatment, including accelerating development of novel therapeutics. Information on genome, epigenome, proteome and gene networks (cistrome) must be integrated to elucidate the mechanism of cancer. In genomic analysis, we need to characterize genetic changes in tumor cells, as well as genetic variation. Human genome has more genetic variation than previously expected, including SNPs as well as non-SNP DNA variation, like copy number variation (CNV) (1). Meanwhile, DNA methylation is an epigenetic mark crucial in regulation of gene expression and lineage commitment. Epigenetics can explain reversible heritable changes in gene function that occur without a change in the DNA sequence, and must be involved in cancer. Recent progress in epigenetic analysis will be presented. Epigenomic Profiling Most of previous studies on DNA methylation detection were not comprehensive, because only limited genomic regions, such as selected genes or promoter CpG islands, were analyzed, mostly with the use of methylation-sensitive restriction enzymes. Recently we have established a method for detection of DNA methylation using oligonucleotide tiling arrays, which, coupled with methylated DNA immunoprecipitation (MeDIPchip), has enabled us to analyze comprehensive methylation profiles (2). Genomic landscape of DNA methylation and histone modification. DNA methylation patterns in the HOXA gene cluster region were integrated with histone H3 and H4 acetylation patterns in HCT116 colon cancer cell line. Little histone acetylation was observed in the hypermethylated regions, demonstrating a reciprocal relationship between DNA methylation and histone H3 and H4 acetylation, i.e. active chromatin structure. de novo DNA methylation in ES cell differentiation. To detect de novo DNA methylation in the establishment of chromatin structure during development, we measured the global methylation patterns during ES-cell differentiation to three different lineages, ecto-, meso- and endoderm. Methylation patterns were distinct among different lineages and will change dynamically during early embryogenesis. Methylation marker. Using a tiling array covering 25,500 human promoter regions, MeDIP-chip analysis showed that methylation was not restricted in CpG islands, implicating the importance of unbiased analysis, and that its patterns were distinct among various tissues and between cancer versus normal tissues. It will be applied to efficiently search for epigenomic biomarkers for early cancer detection or cancer risk assessment, such as methylation in the promoter regions of SFRP genes in colon cancer. Gene silencing. As an integrated approach to better elucidate a molecular basis of cancer, we have combined genomic and epigenomic profiles. Genotyping arrays were used to detect chromosomal aberrations in an allelic manner using Genome Imbalance Map algorithm or GEMCA (3-5). We observed that uniparental disomy (UPD) is not a rare event and gives rise to loss of heterozygosity (LOH) in various cancers, which cannot be identified by the conventional array CGH analysis because UPD regions do not show any total copy number changes. While homozygous deletions detected within LOH regions in several tumors will be relevant in tumor progression, we also examined epigenetic silencing as described above. Currently, genome-wide screening in gastroenterological cancers for aberrant methylation in the LOH regions is under way to identify novel candidates for tumor suppressors. References 1. Redon R, Ishikawa S, Fitch KR, et al. Global variation in copy number in the human genome. Nature 2006;444(7118):444-54. 2. Hayashi H, Nagae G, Tsutsumi S, et al. High-resolution mapping of DNA methylation in human genome using oligonucleotide tiling array. Hum Genet 2007;120(5):701-11. [Sep 26, 2006; Epub ahead of print] 3. Ishikawa S, Komura D, Tsuji S, et al. Allelic dosage analysis with genotyping microarrays. Biochem Biophys Res Commun 2005;333(4):1309-14. 4. Komura D, Shen F, Ishikawa S, et al. Genome-wide detection of human copy number variations using high-density DNA oligonucleotide arrays. Genome Res 2006;16(12):1575-84. 5. Midorikawa Y, Yamamoto S, Ishikawa S, et al. Molecular karyotyping of human hepatocellular carcinoma using single-nucleotide polymorphism arrays. Oncogene 2006;25(40):5581-90. Opening Session Keynote Addresses KN01-03 New Results in Susceptibility to Breast Cancer and their Implications. Bruce A. J. Ponder1, Paul Pharoah2, Dennis Ballinger3, David Cox3, Alison Dunning2, Douglas Easton2. 1Cancer Research UK Cambridge Research Institute, Cambridge, United Kingdom; 2Strangeways Research Laboratories, Cambridge, United Kingdom; 3Perlegen Sciences Inc, CA 94043, CA. The extent of familial clustering of breast cancer provides an estimate of the total inherited contribution to susceptibility. Of this, roughly 25% has been explained by moderately or strongly predisposing mutations in genes such as BRCA1 and 2, CHK2, TP53 and others. The pattern of familial clustering outside the known effects of these mutations suggests that the remaining 75% of predisposition is mostly explained by polygenic effects that is, the combined effects of many genetic variants of individually small effect. No inference can be made, however, about the proportions in which the predisposing variants are common or rare, strong or weak. Because the supposed polygenic variants are individually of weak effect, they do not cause multiple case families and cannot be sought by classical linkage analysis. Instead a case control strategy is used, called an “association study.” In this, the frequency of occurrence of individual genetic variants (measured as single nucleotide polymorphisms, or SNPs) is compared between a large set of breast cancer cases, and healthy controls of the same genetic background. A reproducible difference in SNP frequency between cases and controls that implies that that SNP, or one closely associated with it in the genome (in linkage disequilibrium, within the same haplotype block) confirms susceptibility. This method generally has good statistical power to detect variants only with a frequency in the population of 5 to 10% or greater. A given gene is completely tested if a set of variants is identified that are able to report, through linkage disequilibrium, on all the other common genetic variants in that gene (in this case, the gene is said to be “completely tagged”). The power of such studies can be expressed in terms of the probability of detection, at a given significance level, of any common variant that accounts for a specified proportion — say 1% of the total genetic contribution to susceptibility. A high level of statistical significance is required before a putative association can be accepted. This is primarily because of the low prior probability that a given variant, out of the Translational Cancer Medicine: Technologies to Treatment • November 4-8, 2007 • Singapore 91 Invited Abstracts Opening Session: Keynote Addresses millions in the genome, would in fact be truly associated with susceptibility. p values of 10-2 and less are not uncommon, but cannot be accepted until they have been shown to be replicable in a number of independent studies using different sample sets. To address this, we and others set up an International Breast Cancer Association Consortium (BCAC), which currently has over 20,000 cases and 20,000 controls, for replication studies. The first attempts at association studies were directed at single candidate genes. These studies were largely unproductive. In order to speed up the rate of discovery, we designed in 2003 the first genome-wide association study for breast cancer. In this we were able to test simultaneously a set of 266,000 variants, distributed across the genome and chosen to give the best compromise between cost and completeness of coverage. We used a three-stage design. In the first stage, 400 cases and controls were compared using all 266,000 SNPs. The cases were chosen for having strong family histories and were, therefore, presumably “genetically enriched.” In Stage 2, the top ranked 5% of SNPs from Stage 1 were tested in a population-based set of 4,000 cases and controls. In Stage 3, so far the top 30 SNPs from Stage 2 have been further tested in the 20,000 cases and controls of the Breast Cancer Consortium. We estimate that this study had roughly 65% coverage of the genome, and 90% power to detect any common genetic variant that accounts for 1% of the total genetic effect. Of the 30 top ranked SNPs from Stage 2, five were clearly reproducible in the BCAC case control set in Stage 3; three gave borderline results in the same direction as had been observed in Stage 2; and the remainder showed no effect and were presumably therefore technical artefacts or false positives. The SNP which showed the strongest association lay in an intron of the gene FGFR2. The assignment of the other variants to specific genes is still provisional, because in each case the variant lies within a large linkage disequilibirium region in the genome. In total, the five SNPs for which there is unequivocal association account for just under 4% of the total genetic variance. Studies are underway to identify the causative SNPs from within the linkage disequilibrium blocks, and to elucidate their function. Individually, these SNPs confer small relative risks — of the order of 1.1 to 1.3 fold as heterozygotes, and a maximum of 1.6 fold in the FGFR2 homozygote. They are therefore, taken singly, unlikely to be of great clinical value in refining estimates of individual risk. In combination, however, polygenic variants may have quite striking effects. Modelling by our group suggests that, if all the putative genetic variants could be identified and used to construct a risk profile, it would be found that approximately half of all breast cancer occurs in the 12% of women most at risk. However, we are some way from identifying all of those variants. The variants that have been identified to date might allow one to identify a predisposed 10% of women who have 15% of breast cancer incidence. On the other hand, because the FGFR2 variant, for example, is common (allele frequency 38%), despite its relatively modest effect on risk, it has a high “attributable fraction.” That is to say, if the effect of that variant could be neutralised in the population, total breast cancer incidence might be reduced by approximately 16%. In principle, therefore, an intervention that would neutralise the effect of the FGFR2 SNP might have significant public health application. But of course such a proposed intervention also raises a myriad of problems around risks and benefits, how these could be assessed, public understanding, etc. An interesting but speculative possibility is to ask whether all of the individual genetic variants that in combination would place a woman at high risk are independent of one another; or whether perhaps in any given individual there is an association of variants that target different part of the same physiological process or network. It the latter were true, it might conceivably be that one could identify a disturbed phenotype related to that genetic network, modulation of which in the population by acceptable methods might significantly reduce the overall breast cancer burden. This would be analogous to “cholesterol” or “blood pressure” for cancer. Plenary Session 1 Frontier Technologies PL01-01 Whole Genome Chromatin Interaction Analysis of Transcriptional Regulation Networks in Human Cancer Cells. Yijun Ruan. Genome Institute of Singapore, Singapore. Modern biology research is largely and will be continuously relying on our ability to sequence and characterize the human genome and other model genomes. Rapid developments of next-generation sequencing technology and novel application strategies are evident in recent years. Our approach of making contribution is to improve tag-based sequencing strategy. We first developed the paired-end ditagging (PET) approach and devised the Gene Identification Signature (GIS-PET) analysis for accurate demarcation of boundaries of full-length transcripts (1) and highly efficient identification of unconventional fusion transcripts in human cancer cells (2). PET-based transcriptome/genome analysis is further enhanced by “short read” but “highly paralleled” next-generation sequencing instruments (454/Roche, Solexa/Illumina, and SOLiD/ABI) for multiplex sequencing of paired-end ditags (MS-PET) (3). To identify and study promoters and regulatory elements that mediate gene transcription, we invented the ChIP-PET analysis (chromatin immuno-precipitation coupled with paired-end ditagging) for highly accurate, robust and unbiased genome-wide identification of transcription factor binding sites (4-8). A growing body of data has shown that a large portion of regulatory elements are localized far away from gene coding regions. This observation cannot be explained by a simple linear relationship along the genome, and our current technologies are inefficient to address non-linear interactions between functional DNA elements. To elucidate potential 3-dimentional interaction relationship of functional elements, we have developed a new strategy for whole genome Chromatin Interaction Analysis using Paired End diTagging (ChIA-PET). In this approach, long range chromatin and inter-chromosomal interactions mediated by protein factors are fixed and the fragmented DNA-protein complexes are enriched by chromatin immuno-precipitation. The tethered DNA fragments in each complex are jointed with linker sequence by ligation, and the ligation products are further analyzed by MS-PET sequencing. Consequently, the ChIA-PET sequences are mapped to the reference genome to reveal the relationship of remote chromosomal regions brought in close proximity by protein factors. Our results suggest that long-range and multiplex looping interaction is an extensive mechanism in regulating gene transcription. Together, PET-based DNA sequencing analyses have been enabling us to generate large volume data efficiently from various dimensions around transcription activity. The accumulation and integration of these datasets are starting to emerge a comprehensive picture of transcription regulation networks. References 1. Ng et al. Nature Methods 2005;2:105. 2. Ruan et al. Genome Res 2007;17:828. 3. Ng et al. Nucleic Acid Research 2006;34:e84. 4. Wei et al. Cell 2006;124:207. 5. Loh et al. Nature Genetics 2006;38:431. 6. Zeller et al. PNAS 2006;103:17834. 7. Lin et al. PLoS Genetics 2007;3:e87. 8. Lim et al. Mol Cell 2007;27:622. PL01-03 Nanostructure Processing of Advanced Biomaterials and Biosystems. Jackie Yi-Ru Ying. Institute for Bioengineering and Nanotechnology, Singapore. Nanostructured materials have been designed for the controlled delivery of drugs, proteins and genes. These materials can be engineered for targeting specific cell types, and with responsiveness to selected stimuli. The intelligent delivery systems provide for greater efficiency and reduced side-effects for treatment of various diseases, including cancer. Fluorescent quantum dots have been subjected to thin glutathione or 92 AACR Centennial Conference Plenary Session 1 Invited Abstracts silane coating to render them water-soluble, buffer-stable and noncytotoxic for biological applications. They are further bioconjugated to target specific cell types and compartments for biolabeling and bioimaging applications. We have also derived novel nanocomposites of quantum dots and magnetic nanoparticles to create multifunctional systems for imaging, bioseparation and targeted delivery applications. These materials are of particular interest for cancer and stem cell research. Miniaturized biological and medical devices have been fabricated with microfluidic control for drug screening and medical diagnostic applications. They allow for accurate and high-speed studies and detections involving small sample volumes. Such systems are being developed specifically cancer and infectious disease research. infrequently. These findings may have implications for future anticancer drug development. Plenary Session 3 Animal Systems and Imaging in Cancer Research PL03-03 Role of Polycomb Repressors in Stem Cells, Cancer, and Development. Maarten van Lohuizen. The Netherlands Cancer Institute, Amsterdam, The Netherlands. Repressive Polycomb-group (Pc-G) protein complexes and the counteracting Trithorax-group (Trx-G) of nucleosome remodeling factors are involved in the dynamic maintenance of proper gene expression patterns during development, acting at the level of chromatin structure. As such, they are important controllers of cell fate. When deregulated, these master switches of gene expression are strongly implicated in formation of a diverse set of cancers. An example is the Pc-G gene Bmi1 which is overexpressed in medulloblastoma, Non small cell lung cancer, hepatocellular carcinoma and breast cancer and Glioma and is causally implicated in leukemia. We and others have recently implicated Bmi1/Pc-G as a critical regulator of stem cell fate in hemapoietic stem cells, neural stem cells, mammary epithelial precursor cells and ES cells. In addition, we have shown that Bmi1 is regulated by the Shh pathway and that the Ink4a/ARF tumors suppressors are critical Bmi1 target genes in stem cells and in cancer formation. However our recent work on brain cancer (Glioma) points to important ink4a/ARF-independent Bmi1 targets involved in adhesion and motility. Comprehensive profiling of Polycomb target genes in Drosophila revealed its crucial conserved role in repressing lineage differentiation pathways and morphogens, including Wg, Hh, Delta and Notch. Furthermore, we have characterized in detail an essential E3ubiquitin ligase activity in the PRC1 Polycomb complex that consists of a functional Ring1B-Bmi1 heterodimer. This E3 ligase activity is required for maintenance of Polycomb repression in normal - and cancer stem cells and hence offers potential novel ways to target cancer stem cells or tumor reinitiating cells in which the activity of this E3 ligase is hyperactivated. This is further substantiated by a novel way by which the activity of the Ring1B.Bmi1 E3 ligase is controlled. The implications of these findings for stem cell biology, development and cancer will be discussed. PL03-04 Targeting the Notch Pathway in Cancer. A. Thomas Look. Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA. Chromosomal translocations that produce a truncated and activated form of the NOTCH1 receptor have been identified in rare cases of human T-cell acute lymphoblastic leukemia (T-ALL). To uncover more frequent types of NOTCH1-activating mutations, we analyzed primary T-ALL samples and identified NOTCH1 mutations within the heterodimerization domain (HD) in 27% and truncating mutations that deleted the PEST destruction box (DPEST) in 15% of childhood T-ALL blasts. Both of these regions were simultaneously mutated in the same NOTCH1 gene of 16% of cases, providing evidence for multi-hit mutagenesis affecting a single oncogene in primary T-ALL samples at diagnosis (1). These mutations were shown to occur within each of the multistep molecular pathways that can lead to the transformation of T-cell progenitors during development, suggesting that some form of NOTCH pathway disruption may be required as a first step in the induction of most cases of T-ALL. These findings greatly expand the role of activated NOTCH1 in the molecular pathogenesis of human TALL, and provide a strong rationale in for targeted therapies of this disease that interfere with NOTCH signaling, because mutationally activated forms of NOTCH1 are still dependent on enzymatic cleavage for activity. To capitalize on this opportunity, a novel Phase I clinical trial of a potent NOTCH pathway inhibitor was recently conducted at the Dana-Farber Cancer Institute to specifically target the NOTCH pathway in adults with relapsed or refractory T-ALL. These results in T-ALL suggested that the HD-N, HD-C and Pest domains of the human NOTCH 1, 2, 3 and 4 receptors might also be the target of Plenary Session 2 Proteomics, Genomics, and Biomarker Discovery PL02-02 Drugging “Undruggable” Targets in Cancer. Gregory L. Verdine. Harvard University, Cambridge, MA. One of the most vexing problems in life science is that of “undruggability,” the difficulty of targeting certain biological macromolecules in vivo using existing drug or ligand discovery technologies. It has been estimated that as many as 80-90% of all potential targets, including many that have been extensively validated in humans and in animal models, are undruggable. The Verdine laboratory is developing powerful new chemistry-based platform technologies to address these undruggable targets. Specifically, the lab is developing “synthetic biologics,” molecules that, like biologics, possess the ability to target large flat surfaces, but that, like small molecules, are fully synthetic and hence can be modified at will. Progress on the development of synthetic biologics will be reviewed in this talk. References 1. Walensky LD, Kung AL, Escher I, et al. Science 2004;305:1466-70. 2. Walensky LD, Pitter K, Morash J, et al. Mol Cell 2006;24:199-210. 3. Bernal F, Tyler AF, Korsmeyer, SJ, et al. J Am Chem Soc 2007;2456-7. PL02-05 Patterns of Somatic Mutation in Human Cancer Genomes. Michael R. Stratton. Wellcome Trust Sanger Institute, Cambridge, United Kingdom. All cancers are due to abnormalities in DNA. The availability of the human genome sequence has led to the proposal that resequencing of cancer genomes will reveal the full complement of somatic mutations and hence all the cancer genes. To explore the nature of the information that will be derived from cancer genome sequencing we have sequenced the coding exons of the family of 518 protein kinases, ~1.3Mb DNA per cancer sample, in 210 cancers of diverse histological types. Despite the screen being directed toward the coding regions of a gene family that has previously been strongly implicated in oncogenesis, the results indicate that the majority of somatic mutations detected are “passengers.” There is considerable variation in the number and pattern of these mutations between individual cancers, indicating substantial diversity of processes of molecular evolution between cancers. The imprints of exogenous mutagenic exposures, mutagenic treatment regimes and DNA repair defects can all be seen in the distinctive mutational signatures of individual cancers. This systematic mutation screen and others have previously yielded a number of cancer genes that are frequently mutated in one or more cancer types and which are now anticancer drug targets (for example BRAF, PIK3CA, and EGFR). However, detailed analyses of the data from our screen additionally suggest that there exist a large number of additional “driver” mutations which are distributed across a substantial number of genes. It therefore appears that cells may be able to utilise mutations in a large repertoire of potential cancer genes to acquire the neoplastic phenotype. However, many of these genes are employed only Translational Cancer Medicine: Technologies to Treatment • November 4-8, 2007 • Singapore 93 Invited Abstracts Plenary Session 3 mutational activation in other types of human cancers. This rationale is particularly compelling in tumor types for which functional data implicate NOTCH pathway activation (e.g., breast cancer, glioblastoma, CLL, pancreatic cancer, and colon cancer). However, we sequenced the HD-N, HD-C and Pest coding regions of all 4 NOTCH genes in >20 primary tumors or tumor cell lines of breast, pancreas, prostate, kidney, and lung cancers, glioblastoma, melanoma, multiple myeloma, CLL, and neuroblastoma and did not identify any activating mutations similar to the ones that we found in T ALL. Nevertheless, very low Numb levels have been documented in ~50% of human breast cancers, which should accentuate trafficking of the Notch signal from the cell surface to the nucleus, and such cases have been shown to have inhibited colony-forming potential in the presence of a NOTCH pathway inhibitor (2). In addition, gene amplification and overexpression of the NOTCH3 receptor has been reported in a high percentage of high-risk ovarian carcinomas (3). Accordingly, phase I/II trials of a potent NOTCH pathway inhibitor are now underway in high risk relapsed or metastatic adult solid tumors. References 1. Weng AP, Ferrando AA, Lee W, et al. Science 2004;306:269-71. 2. Pece S, Serresi M, Santolini E, et al. J Cell Biol 2004;167:215-21. 3. Park JT, Li M, Nakayama K, et al. Cancer Res 2006;66:6312-8. PL03-05 Dissecting Signal Transduction Pathways In vivo with Dynamic Real-time Molecular Imaging. David Piwnica-Worms. Washington University School of Medicine, St. Louis, MO. Genetically-encoded imaging reporters introduced into cells and transgenic animals enable noninvasive, longitudinal studies of dynamic biological processes in intact cells and living animals. The most common reporters include firefly luciferase (bioluminescence imaging), green fluorescence protein (fluorescence imaging), Herpes Simplex Virus-1 thymidine kinase (positron emission tomography) and variants with enhanced spectral and kinetic properties optimized for use in vivo. When cloned into promoter/enhancer sequences or engineered into fusion proteins, molecular imaging reporters enable fundamental processes such as transcriptional regulation, signal transduction cascades, protein-protein interactions, oncogenic transformation, cell trafficking and targeted drug action to be temporally and spatially registered in vivo. Dynamic analysis of cancer with genetically-encoded imaging reporters provides new insight into cancer-specific molecular and regulatory machinery within the contextual environment of the whole animal. Plenary Session 4 New Trials and Genomic-Based Design PL04-02 Design of Clinical Trials of Targeted Therapy. Stephen L. George. Duke University Medical Center, Durham, NC. The development of the randomized clinical trial (RCT) is arguably the most important contribution to scientific medicine of the 20th century. Beginning in the late 1940s, clinical trials rapidly became the gold standard method throughout the world for assessing the relative effects of therapies for diseases and medical conditions. By the end of the century and continuing into the 21st century, thousands of trials have been and are being conducted world-wide in virtually all disease areas. Regulatory agencies now require properly designed and executed RCTs as the primary source of information on the safety and efficacy of therapies proposed for marketing approval in specified indications. The design of RCTs utilize sound scientific principles for the reduction of bias and for increasing the precision and controlling error rates in the testing of pre-specified statistical hypotheses. Powerful tools for controlling bias, including randomization of treatment assignment, blinded treatment assignments, blinded evaluation of outcome, and intent-to-treat analyses have become so widely accepted that they now seem almost self-evident requirements, but it is important to remember that these are relatively new developments in scientific medicine. The hallmark of modern RCTs, in contrast to studies prior to the time of wide-spread application of such trials, is that sufficient numbers of similar patients (those meeting prespecified eligibility criteria) need to be treated alike following a written protocol covering all aspects of the treatment and the details of the trial. Assuming that the trial is well-designed and executed, the number of patients so treated is the primary determinant of the precision of the results and the reliability of the conclusions from the trial. This concept of moving from individualized treatment regimens given idiosyncratically to each patient based primarily on the prior experience of the treating physician to treating large numbers of patients in a similar fashion is central to the success of RCTs. Although the assumptions underlying the design of a RCT do not require that the patients are completely homogeneous, the primary conclusions are usually stated on a population basis. For example, the conclusion might be that the overall survival distribution of patients entered on study was significantly better for patients randomized to receive treatment A than for those randomized to receive treatment B. The difference is often characterized, sometimes misleadingly, by some simple summary of the overall distribution (e.g., the estimated median OS on treatment A was 18 months compared to 13 months on treatment B) although the commonly used test statistics compare the overall distributions (e.g., a log rank test of the OS distributions). Secondary exploratory analyses might be conducted to see if the treatment effect differs by certain patient characteristics, but the overall comparison is properly understood to represent the primary result of the trial. Recently, the efforts to develop therapies targeted to some specific biologic target has raised questions about the proper way to evaluate such therapies in RCTs. The oft-stated ultimate goal of such targeted therapy is individualized therapy based on a patient’s demographic (eg, age, race, gender, etc) and biologic characteristics (eg, biologic markers, genomic data, specific biologic pathway targets, etc). That is, the specific therapy given to each patient would be tailored to the particular patient being treated, designed to achieve the best possible outcome for that patient. Although this concept has strong intuitive appeal, the irony is that this brings us full circle back to the pre-clinical trials era of individualized therapy. A major difference, of course, is that in the new setting the individualization is based on solid scientific evidence of how to tailor the therapy. In the old setting, the basis for the individualization was at best not clear. Nevertheless, there are challenges in the design of clinical trials to examine the effectiveness of individualized, or at least more tailored, therapies. In this paper, some issues around the design of clinical trials of targeted therapies will be explored. Suppose we wish to compare, in a RCT, a regimen (T) involving some targeted therapy with a standard regimen (S) not involving the targeted therapy. Patients with the target in question, assumed in the simplest case to be a binary characteristic (i.e., either the target is present or not), are identified via a standardized assay applied to all patients prior to entering the trial. We are interested in the relative efficiency of different designs in this setting, where efficiency is measured in terms of the numbers of patients required on the trial, the duration of the trial, and the cost of the trial, including the costs of eligibility screening. The two most obvious choices of design are the traditional design in which all patients are entered regardless of the assay results - or perhaps the design in which the assay is not conducted at all (thus eliminating the assay costs), and a targeted design in which only patients with a positive assay are entered in the trial. In order to determine the relative efficiency of competing designs, certain assumptions or prior results are required. These assumptions are critical in determining the relative efficiency of the designs. Key assumptions include the accuracy of the assay, in particular the rates of false positives and false negatives; the effect of the targeted therapy both in patients with the target and in patients without the target; the prevalence of patients with the target; and, particularly for comparison with the design omitting the assay, the cost of the assay. In the simplest setting, in which the assay is perfect, yielding no false positives and no 94 AACR Centennial Conference Plenary Session 4 Invited Abstracts false negatives, and in which the favorable treatment effect is restricted to patients with the target, the relative efficiency of the targeted design in terms of the numbers of patients can be quite large. However, in cases where these assumptions are not met, the efficiency gains may be more modest; indeed, the targeted design can be less efficient in certain situations. If the assumptions required to achieve a major increase in efficiency are of uncertain validity, one alternative is to design trials which test for overall treatment effects as in the traditional designs while also including valid tests of effects limited to patients with the target. These trials require careful pre-specification of objectives and procedures which control the error rates, but provide a practical alternative design in this situation. Some possibilities along these lines will be presented. PL04-03 Genomic Strategies for Personalized Cancer Therapy. Joseph R. Nevins. Duke Institute for Genome Sciences and Policy, Durham, NC. The ability to tailor cancer therapy to characteristics of the individual patient is key in achieving a successful outcome. We have made use of genomic data to develop a capacity to predict risk for individual patients and then combine this with a capacity to identify potential therapeutic options. An example can be seen in early stage lung cancer where the current standard of care for Stage 1A patients is surgery and observation with no indication of benefit from chemotherapy. Nevertheless, 25% of Stage IA patients will have a disease recurrence suggesting the need to identify individuals in this subgroup for more effective therapy. We developed gene expression profiles that could predict risk of recurrence, performing significantly better than previously described clinical prognostic factors. This predictive model was validated in two independent patient cohorts. Importantly, we show that this provides an opportunity to reclassify Stage IA patients to identify a subset of higher risk patients that might be appropriate for adjuvant chemotherapy. We have also developed a series of gene expression signatures that have the capacity to predict sensitivity to cytotoxic chemotherapeutics, including those used in the treatment of lung cancer. Making use of in vitro drug response data generated with the NCI-60 panel of cancer cell lines as well as other sources, coupled with baseline Affymetrix gene expression data, we developed genomic predictors of response and resistance. Many of these predictors have been validated with chemotherapeutic response data from patient samples, including a prospective breast neoadjuvant study, demonstrating a capacity to identify those patients most likely to respond to a given drug. Importantly, these predictions also accurately identify those patients likely to be resistant to current standard of care regimens emphasizing the importance of developing strategies for effective therapy of these individuals. One approach has made use of predictions of pathway deregulation in cancer cell lines that are shown to coincide with sensitivity to therapeutic agents that target components of the pathway, underscoring the potential for such pathway prediction to guide the use of targeted therapeutics. Finally, we have combined the chemotherapy predictive signatures with the signatures of oncogenic pathway deregulation to identify therapeutic strategies that make use of all available drugs, matched to the characteristics of the individual patient. We suggest that this approach provides a strategy towards the development of personalized treatment options for the individual patient. PL04-04 Recurrent Gene Fusions in Prostate Cancer. Arul M. Chinnaiyan. University of Michigan Medical School, Ann Arbor, MI. To date, the great majority of disease-specific, recurrent chromosomal rearrangements have been characterized in hematological malignancies and mesenchymal tumors and not in common epithelial tumors such as breast, lung, colon, or prostate cancer. Here, we employed a bioinformatics approach on a compendium of cancer gene expression data to discover candidate oncogenic chromosomal aberrations based on outlier gene expression. In addition to identifying many gene partners of characteristic rearrangements in human malignancies, this approach identified two members of the ETS family of transcription factors, ERG and ETV1, as outliers in prostate cancer. Either ERG or ETV1 was over-expressed in the majority of prostate cancers (50-70%) and were mutually exclusive across several independent gene expression datasets, suggesting that they may be functionally redundant in prostate cancer development. By RNA ligase-mediated rapid amplification of cDNA ends (RACE), we identified a recurring gene fusion of the 5’ untranslated region of a prostate-specific, androgen-regulated gene TMPRSS2 to ERG or ETV1 in prostate cancer cases which over-expressed the respective ETS family member. These gene fusions were confirmed using quantitative PCR (QPCR) and sequencing of reverse transcription PCR products. In addition, using fluorescence in situ hybridization (FISH), we demonstrated that 23 of 29 (79%) prostate cancer samples harbor rearrangements in ERG or ETV1. Furthermore, in vitro cell line studies suggest that the androgen-responsive promoter elements of TMPRSS2 mediate the aberrant over-expression of ETS family members in prostate cancer. Subsequently, we interrogated the expression of all ETS family members in prostate cancer profiling studies and identified outlier expression of ETV4 in two of 98 cases. In one such case, we confirmed the over-expression of ETV4, and by RACE, QPCR and FISH, we identified fusion of the TMPRSS2 and ETV4 loci. Together, these results suggest a pathogenetically important role for recurrent chromosomal rearrangements in common epithelial tumors and have implications in the molecular diagnosis and treatment of prostate cancer. Importantly, these results identify three molecular subtypes of prostate cancer, TMPRSS2:ERG, TMPRSS2:ETV1 and TMPRSS2:ETV4, and suggest that dysregulation of ETS family member expression through gene fusions with TMPRSS2 may be a generalized mechanism for prostate cancer development. In our most recent work, we explored the mechanism of ETS family over-expression in prostate tumors. Remarkably, we identified novel 5’ fusion partners in prostate tumors with outlier expression of ETS family members, including untranslated regions from a prostate-specific androgen-induced gene and endogenous retroviral element, a prostatespecific androgen-repressed gene, and a strongly expressed housekeeping gene. As the commonality of these rearrangements is the aberrant overexpression of ETS genes, we recapitulated this event in vitro. We demonstrate that ETS over-expression in multiple benign prostate cells induces a marked increase in invasion, confirming the role of ETS gene rearrangements in prostate cancer development. Identification of distinct classes of ETS gene rearrangements demonstrates that dormant oncogenes can be activated in prostate cancer by juxtaposition to tissue-specific or ubiquitously active genomic loci. Subversion of active genomic regulatory elements may serve as a more generalized mechanism for carcinoma development. Furthermore, the identification of androgen-repressed and insensitive 5’ fusion partners has important implications for the antiandrogen treatment of advanced prostate cancer. Plenary Session 5 Cancer Stem Cells PL05-02 Breast Stem Cells and Their Implications for Cancer. Jane Visvader1, Marie-Liesse Asselin-Labat1, Francois Vaillant1, Mark Shackleton1, Kate Sutherland1, Toula Bouras1, Jacqueline van der Wees2, Frank Grosveld2, Geoffrey Lindeman1. 1Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; 2Erasmus University, Rotterdam, Netherlands Antilles. Interest in mammary stem cells (MaSCs) has been stimulated by their potential role in breast tumorigenesis. We have isolated discrete populations of mouse mammary epithelial cells on the basis of expression of the cell surface markers (CD29 and CD24) and transplanted these into cleared fat pads of recipient mice to identify the mammary stem cell. A population that expresses ‘basal’ markers and is highly enriched for mammary stem cells was defined. An absolute expansion in the number of mammary stem cells was observed in the pre-neoplastic phase in the MMTV-wnt-1 mammary tumor model, suggesting that these tumors originate within mammary stem cells. Interestingly, the basal stem cell- Translational Cancer Medicine: Technologies to Treatment • November 4-8, 2007 • Singapore 95 Invited Abstracts Plenary Session 5 enriched population was found to be ‘triple negative’ for ER, PR and ErbB2. This phenotype is reminiscent of human basal tumors, inferring that the mammary stem cell may be the ‘cell of origin’ for this poor prognosis subset. We have recently further subdivided the luminal epithelial cellcontaining population on the basis of CD61 ( 3-integrin) expression. The CD61+ subset contained luminal progenitors, as defined by colony forming assays, and is the earliest identified cell-type to express ER. Since the GATA-3 transcription factor is a key marker of ‘luminal’ breast tumors, we studied its role in normal mammary gland development. Utilising conditionally targeted mice, Gata-3 was found to be essential for mammary gland morphogenesis in both the embryo and adult. Gata-3 deficiency resulted in an expansion of CD61+ luminal progenitors and a concomitant block in epithelial differentiation in both virgin and pregnant glands. These findings provide evidence for the existence of an epithelial hierarchy within the mammary gland and reveal that GATA-3 is a critical regulator of luminal cell differentiation. It is likely that high expression of GATA-3 in luminal tumors confers a favourable prognosis due to the promotion of differentiation. Conversely, GATA-3 loss could result in a more clinically aggressive phenotype. PL05-03 Leukemia Stem Cells in CML: Characterization and Elimination. Joanne C. Mountford, Heather G. Jorgensen, Mhairi Copland, Niove Jordanides, Francesca Pellicano, Tessa L. Holyoake. University of Glasgow, Glasgow, United Kingdom. Chronic Myeloid Leukaemia (CML) is a myeloproliferative disease that originates in a haemopoietic stem cell (HSC) as a result of the t(9;22) translocation, giving rise to the Philadelphia chromosome and bcr-abl oncoprotein. The disease starts in chronic phase, but as a result of genomic instability, it progresses over time to accelerated phase and then to blast crisis, becoming increasingly resistant to therapy. Bcr-abl is a constitutively active tyrosine kinase that has been targeted by tyrosine kinase inhibitors (TKIs), including imatinib (IM), nilotinib and dasatinib. We have developed various flow cytometry techniques to enable us to isolate candidate CML stem cells from chronic phase patients at diagnosis. These cells efflux Hoechst dye, express CD34, lack CD38 and are cytokine non-responsive in culture over periods of up to 12 days in growth factors. These stem cells have also been shown to regenerate bcr-abl positive haemopoiesis in immuno-compromised mice upon transplantation. We previously demonstrated that IM had an anti-proliferative effect on CML stem cells but did not induce apoptosis. Clinical experience now confirms that IM may not target CML stem cells in vivo with few patients achieving complete molecular remission and relapse occurring rapidly upon drug withdrawal. Our recent efforts have focused on understanding why CML stem cells are resistant to IM and on trying to find novel ways to induce apoptosis of this population. We have shown that CML stem cells express very high levels of functional wild type bcr-abl, these findings have been confirmed by other labs. However, we did not detect gene amplification or kinase domain mutations within the stem cell population, thus the cells express very high levels of wild type bcr-abl protein. One possible explanation for the relative insensitivity of CML stem cells to TKI-induced killing is that the level of protein is too high to be efficiently inhibited. This may be exacerbated by the presence of drug transporters on CML stem cells that may actively reduce the intracellular concentration of TKI. In order to investigate this possibility we have examined the expression and function of ATP-Binding Cassette (ABC) proteins on these cells. Members of the ABC family have previously been shown to be responsible for classic multidrug resistance in many malignancies and have been found to be highly expressed on normal HSC and in AML. We looked at the relative expression between normal CD34+ cells and CML CD34+ cells of the transporters ABCB1 (MDR1 or p-gp), ABCC1 (MRP1) and ABCG2 (BCRP or MXR). We also investigated whether TKI are effluxed by these proteins and finally whether inhibition of the transporter function enhanced TKI-induced killing of the resistant CML stem cell population. We found that ABCG2 is expressed more highly in CML whereas ABCB1 is less abundant, however, we found that both IM and nilotinib inhibited transport via these proteins but were not effluxed themselves. Additionally, when we inhibited transporter activity we did not see any increase in the effectiveness of the TKIs. Therefore, we conclude that it is unlikely that ABC transporter mediated drug resistance is solely responsible for the insensitivity of CML stem cells to TKI-induced apoptosis. Dasatinib, a is a newer TKI which has multiple targets including src-family kinases and bcr-abl, and has greater potency than IM. We have found that dasatinib inhibits bcr-abl activity more efficiently than IM but still does not induce apoptosis of the stem cell population. Thus it seems that single agent TKI may not be the way to eradicate CML stem cells. As we do not see complete inhibition of bcr-abl activity we decided to combine TKI with agents rationally chosen to target downstream pathways that are activated by bcr-abl including PI3K, Ras and MAPK signalling. We found that most of these combinations are no more effective than TKI alone in killing CML stem cells however, farnesyltransferase inhibitors were found to have a significant effect. BMS-214662 is the most effective of these and induces apoptosis of phenotypically and functionally defined CML stem cells in vitro, as a single agent and in combination with IM or dasatinib. The effect against CML stem cells is selective with less effect on normal stem cells. The drug is also effective against blast crisis CML stem cells and equally effective against wild type and mutant bcr-abl, including the most resistant mutant T315I. In association with apoptosis there is activation of caspase-8 and caspase-3, inhibition of the MAPK pathway and other signalling pathways. Furthermore, BMS-214662 synergises with MEK1/2 inhibitors, suggesting a second mechanism other that RAS inhibition that is responsible for induction of apoptosis. This is supported by our finding that a highly related agent BMS225975 has very similar farnesyltransferase inhibitory activity but very much lower capacity to kill CML stem cells. BMS214662 is the first agent we have tested that has significant activity against CML stem cells and offers the possibility of stem cell directed therapy that we have been seeking. We are exploring further the mechanism of action of BMS-214662 and extending these findings to other cancer stem cell disorders and hope to move this pre-clinical work to a clinical trial combining dasatinib with BMS-214662 in CML. Plenary Session 6 Clinical Intervention and Populations Outcomes: Global Strategies PL06-01 Normalization of Tumor Vasculature and Microenvironment by Antiangiogenic Therapies: From the Bench to Bedside and Back. Rakesh K. Jain. Massachusetts General Hospital, Boston, MA. Solid tumors require blood vessels for growth, and many new cancer therapies are targeted against the tumor vasculature. The widely held view is that these antiangiogenic therapies destroy the tumor vasculature, thereby depriving the tumor of oxygen and nutrients. Indeed that is the ultimate goal of antiangiogenic therapies. However, emerging preclinical and clinical evidence support an alternative hypothesis — that judicious application of agents that block angiogenesis directly (e.g., Avastin, Recentin) and indirectly (e.g., Herceptin) can also transiently “normalize” the abnormal structure and function of tumor vasculature. In addition to being more efficient for oxygen and drug delivery, the normalized vessels are fortified with pericytes, which can hinder intravasation of cancer cells — a necessary step in hematogenous metastasis. Drugs that induce vascular normalization can also normalize the tumor microenvironment — reduce hypoxia and interstitial fluid pressure — and thus increase the efficacy of many conventional therapies if both are carefully scheduled. Reduced interstitial fluid pressure can decrease tumor-associated edema as well as the probability of lymphatic dissemination. Independent of these effects, alleviation of hypoxia can decrease the selection pressure for a more malignant phenotype. Finally, the increase in proliferation of cancer cells during the “vascular normalization window” can potentially sensitize tumors to cytotoxic agents. Our recent Phase II clinical trials in patients with rectal carcinomas and glioblastomas (GBM) support our pre-clinical 96 AACR Centennial Conference Plenary Session 6 Invited Abstracts findings on vascular normalization. Moreover, in GBM patients, the normalization window — identified using advanced MRI techniques — can last one to four months, and the resulting changes in tumor vasculature correlate with circulating molecular and cellular biomarkers in these patients. The encouraging results on patients survival from the GBM Phase II trial has led to an international multi-center randomized Phase III clinical trial in GBM patients. PL06-02 Role of Anti-EGFR Therapy in Asian Cancer Patients. YungJue Bang, Sae-Won Han, Tae-You Kim. Seoul National University College of Medicine, Seoul, Republic of Korea. Epidermal growth factor receptor (EGFR) is frequently overexpressed in various malignancies which is often associated with poor prognosis. Its activation has important role in various steps of carcinogenesis and progression. Recently, efforts have been made to develop anti-cancer agents inhibiting EGFR activities which include small molecular tyrosine kinase inhibitor (TKI) and monoclonal antibody. EGFR inhibition using small molecular TKI, gefitinib and erlotinib, has been most promising in non-small-cell lung cancer (NSCLC). EGFR TKI showed modest but meaningful activity with tolerable toxicity in pretreated NSCLC. Interestingly, phase II studies and data from expanded access program of gefitinib showed that response rates were higher in patients with East-Asian ethnicity, female sex, adenocarcinoma or bronchioloalveolar carcinoma (BAC) histology, and never-smoking history. In the phase III trial of erlotinib (BR.21) which compared erlotinib with best supportive care in previously treated NSCLC, erlotinib treatment significantly prolonged survival. Response rates were also higher in females, adenocarcinomas, never-smokers and Asians although the survival benefit of erlotinib treatment was not limited to these subgroups. Although gefitinib failed to improve survival in the overall study population in the phase III gefitinib trial, Iressa Survival Evaluation in Lung Cancer (ISEL), planned subgroup analysis of ISEL showed significant survival benefit in never-smokers and Asians. Again, higher objective responses were seen in females, adenocarcinomas, never-smokers, and Asians. These findings suggest that patient selection is of great importance for optimized efficacy of EGFR TKI. The mysterious link between the clinical predictors and the higher response to EGFR TKI can be partly explained by the presence of the activating EGFR tyrosine kinase domain mutations. EGFR mutations are more frequently found in Eastern Asians, adenocarcinomas, never-smokers, and females. EGFR mutant tumor comprises 30-50% of NSCLC from Asian patients in contrast to those from Caucasian patients of which around 10% harbor EGFR mutations. It is unclear what leads to the difference in mutational frequency, whether it be genetic or environmental. EGFR mutation confers constitutive autophosphorylation and activation of the receptor and its downstream signal cascades. Moreover, the classic mutations (L858R and deletion in exon 19) are extremely sensitive to EGFR TKI. The high response rates to EGFR TKI have been confirmed in prospective studies of EGFR TKI as first-line treatment in NSCLC harboring the classic mutations. EGFR mutant patients treated with gefitinib show longer survival compared to gefitinib-treated wild-type patients. The longer survival seen in mutant patients are due to the enhanced sensitivity of the mutant tumors to EGFR TKI or the innate indolent biology needs to be answered in randomized studies. Although the biomarker analysis of BR.21 suggested that EGFR mutation does not have role in predicting benefit from erlotinib treatment, the number of EGFR mutation positive patients included in the analysis (34 patients) is too small to draw any definitive conclusion about the role of EGFR mutation in response and survival prediction of EGFR TKI treatment. Another molecular predictor of EGFR TKI sensitivity is increased EGFR gene copy number determined by FISH which include high polysomy and gene amplification. Single-arm studies from Western populations suggested that EGFR FISH has stronger association with response and survival compared to mutation in gefitinib treated patients. In contrast, Asian studies repeatedly reported that EGFR mutation has stronger association with response and survival in gefitinib-treated NSCLC compared to gene copy number status. Potential explanation for these discrepancies according to ethnicities is the possible ethnic difference in the mechanism of EGFR ‘addiction’ and EGFR TKI sensitivity. EGFR mutation may have major role in Asians compared to high EGFR gene copy number having predominant role in Caucasians. Given the significant association between EGFR mutation and high gene copy number, what leads to the possible difference among the ethnicities warrants further exploration. The results from the biomarker studies of the randomized phase III trials (BR.21 and ISEL) suggest that EGFR TKI leads to survival prolongation only in the FISH+ patients. The important role of EGFR gene copy number in EGFR TKI treatment among Western population is evident from these studies. However, caution should be taken in directly applying these data to Asian patients. Less than 10% of patients included in the molecular analyses of the phase III trials were Asian origin and the mutational frequencies of EGFR in the studies were lower than those reported in Asian studies. Randomized phase III study having appropriate power for biomarker analysis needs to be conducted in Asian patients who have high mutational frequency to identify the exact roles of EGFR gene copy number and mutation in EGFR TKI treatment among Asians. In addition to the mutations, polymorphisms of EGFR also show different distribution according to ethnicity. CA dinucleotide repeat polymorphism in intron 1 of EGFR, which can modulate transcription of EGFR, shows higher repeat numbers in Asians compared to other ethnicities. Lower repeat number tends to be associated with better treatment outcome of gefitinib. Two other common polymorphisms (-216G/T and -191C/A) located in the promoter region of EGFR associated with altered promoter activity and gene expression also has ethnic difference in distribution. Asians almost uniformly carry -216G/G and 191C/C, whereas -216G/T or T/T and -191C/A or A/A are also commonly found in other ethnicities. Collectively, genetic make-up of EGFR in NSCLC seems to be different in Asians and Caucasians. As our understanding of genetic changes in cancer enriches, more ethnic differences are expected to be uncovered. Therefore, different approach in biomarker analysis and customized treatment strategy may be necessary according to ethnicity in the future. Moreover, ethnic background should always be taken into account when performing translational research. Gastric cancer is one of the leading causes of cancer-related mortality throughout the world, especially in Asian countries. Treatment outcomes are still disappointing although therapeutic improvements have been seen in the past decades. To further improve treatment results of gastric cancer, EGFR or HER2 targeted agents are being evaluated in combination with conventional chemotherapy regimens. EGFR expression is detected in 30-50% of advanced gastric cancer and is associated with advanced disease. EGFR gene is amplified in 5-10% of gastric cancer. A recent phase II study of cetuximab in combination with FOLFIRI in untreated gastric or gastroesophageal junction adenocarcinoma showed encouraging results. We are conducting a phase II study of cetuximab in combination with modified FOLFOX6 in recurrent or metastatic gastric cancer which have recently completed patient accrual. Comprehensive analyses of biomarkers are also planned. HER2 is reported to be expressed in 10-90% of gastric cancer and gene amplification is found in 5-25%. HER2 amplification is more frequently found in intestinal type compared to diffuse type cancers. HER2 expression or amplification is associated with advanced stage and poor prognosis. Preclinical studies showed that trastuzumab potentiated antitumor activity of cytotoxic chemotherapies. With these backgrounds, phase III study comparing trastuzumab in combination with fluoropyrimidine and cisplatin versus chemotherapy alone as first-line therapy in HER2 positive advanced gastric cancer is currently underway. Translational Cancer Medicine: Technologies to Treatment • November 4-8, 2007 • Singapore 97 Invited Abstracts Plenary Session 7 Plenary Session 7 Target to Lead: New Strategies in Drug Development PL07-02 From Cancer Genomics to Cancer Treatment. Yusuke Nakamura. University of Tokyo, Tokyo, Japan. cDNA microarray technologies have enabled us to obtain comprehensive data for gene expression profiles of human cancers. To isolate novel targets for diagnosis (predictive marker for the efficacy of treatment as well as tumor marker) and for treatment of cancer (molecular-targeting drug, cancer vaccine, antibody), we have been comparing expression profiles of cancer cells originated from various organs with their corresponding non-cancerous tissues using a cDNA microarray that consists of more than 30,000 cDNAs/ESTs. These experiments disclosed a number of genes that appeared to be involved in development and/or progression of cancers in those tissues. So far, we have analyzed more than 1,200 cases of clinical cancer samples of the liver, pancreas, stomach, colon, esophagus, bile duct, uterus, lung, ovary, kidney, urinary bladder, testis, prostate, breast, and soft tissues as well as acute and chronic myeloid leukemias. We have selected hundreds of candidate genes by the criteria as follows; (1) genes whose expression was transactivated in a large proportion of cancer tissues in comparison with their corresponding normal tissues and (2) genes whose expression was not observed or hardly detectable in any important vital organs. The further functional analysis identified dozens of genes that are likely to function as oncogenes in various cancers. The suppression of expression of such genes with small-interfering RNAs (siRNAs) induced cell cycle arrest, apoptosis, or suppression of anchoring-dependent cell growth. For example, through the genome-wide expression profiles of renal cell carcinomas (RCCs), we identified that hypoxia-inducible protein-2 (HIG2) was expressed exclusively in RCCs and fetal kidney. Induction of HIG2 cDNA into mammalian cells led to secretion of the gene product into culture media and resulted in enhancement of cell growth. Small interfering RNA (siRNA) effectively inhibited expression of HIG2 in human RCC cells that endogenously expressed high levels of the protein, and significantly suppressed cell growth. Moreover, addition of polyclonal antiHIG2 antibody into culture media induced apoptosis in RCC-derived cell lines. ELISA analysis of clinical samples identified secretion of HIG2 protein into the plasma of RCC patients even at an early stage of tumor development, whereas it was detected at significantly lower levels in healthy volunteers or patients with chronic glomerulonephritis. The combined evidence suggests that this molecule represents a promising candidate for development of molecular-targeting therapy and could serve as a prominent diagnostic tumor-marker for patients with renal carcinomas. We also identified overexpression of Dickkopf-1 (DKK1) in the majority of human cancers in several organs and found a significant association of DKK1 expression with poor prognosis of lung and esophageal cancers. We established an ELISA system to measure serum DKK1 and found that the proportion of the DKK1-positive cases was 185 (69.8%) of 265 lung cancer, 51 (63.0%) of 81 esophageal cancer, 108 (59.3%) of 182 cervical cancer, 15 (55.6%) of 27 prostate cancer, 110 (65.1%) of 169 breast cancer, 89 (53.0%) of 168 hepatocellular carcinoma, 32 (29.9%) of 107 bile duct cancer, 14 (34.1%) of 41 pancreatic cancer, and 39 (38.6%) of 101 gastric cancer patients, while only 10 (4.8%) of 207 healthy volunteers were falsely diagnosed. In addition, the invasive or growth activity of the cancer cells which highly expressed DKK1 were suppressed by addition of anti-DKK1 antibody into their culture media. Administration of anti-DKK1 antibody to tumor-implanted mice significantly suppressed tumor growth without any obvious adverse events. The data suggest usefulness of DKK1 as a serum biomarker in clinic and as a target for the development of therapeutic antibodies. In addition, we reported previously Frizzled homologue 10 (FZD10), a member of Frizzled family, to be a promising therapeutic target for synovial sarcomas (SS). We established a murine monoclonal antibody (MAb), namely, 92-13 for FZD10 product that had specific binding activity against native FZD10 product expressed in SS cell lines. Subsequent immunohistochemical analyses with the MAb 92-13 confirmed an absence or hardly-detectable level of FZD10 protein in any normal human organs except the placenta. Moreover, we validated the specific binding activity of this MAb in vivo after injection of fluorescent-labeled mAb into the mice carrying SS xenografts by the use of the in vivo fluorescent imaging system as well as radioisotopes. Since the MAb 92-13 was effectively internalized into the SS cells after its binding to FZD10 on the cell surface, we attempted radioimmunotherapy using an 90Yttrium-labeled MAb 92-13 (90Y-MAb 92-13) to a mouse SS-xenograft model. Expectedly, a single intravenous injection of 90Y-MAb 92-13 drastically inhibited tumor growth of SS in mice; in four of eleven mice treated, tumors were completely disappeared at 60 days after the treatment. This result indicates that MAb 92-13 could be utilized as the novel treatment modality for SS. These results indicated that systematic expression analysis should be a very effective approach for identification of molecules that are potential targets for development of novel therapeutic drugs and diagnostic tools. PL07-03 Targeting Epigenetic Process for Therapeutic Control of Wnt/beta-catenin Signaling in Colorectal Cancer. Qiang Yu. Genome Institute of Singapore, Singapore. Transcriptional inactivation of tumor suppressors via aberrant epigenetic events plays an important role in cancer development. Cancer epigenetic therapy aims at reversing the epigenetic process to restore the expression of silenced tumor suppressors. Given the complexity and coordination of multiple gene inactivation mechanisms, a drug combination approach that more effectively modulate epigenetic processes needs to be developed. We have developed a novel pharmacological approach that effectively reverses histone modifications in cancer cells by coordinately modulating both histone methylation and deacetylation. This pharmacological combination approach induces strong apoptosis induction through robust reactivation of many silenced tumor suppressors. In particular, we have identified a novel epigenetic regulator of Wnt/beta-catenin signaling that is silenced in colon cancer. Reactivation by gene by our combination treatment leads to strong inhibition of Wnt/β-catenin signaling and massive apoptosis in colorectal cancer cells. Our study points to a novel strategy for cancer epigenetic drug treatment and in particular for therapeutic control of the aberrant Wnt signaling pathway in colorectal cancer. PL07-04 Functional and Prognostic Insights into the Altered Cancer Genome. Jinqiu Zhang, Xuejing Liu, Arpita Datta, Bing Lim, Krishna Karuturi, Patrick Tan, Philippe Broet, Sophie Broet, Lance D. Miller. Genome Institute of Singapore, Republic of Singapore, Singapore. Gene expression profiling of primary human tumors has proven valuable in elucidating gene expression signatures prognostic of patient outcomes. Recently, two such signatures predictive of distant metastasisfree survival in node-negative breast cancer patients have been discovered and validated in a large multi-center study. Both signatures (the Agendia 70-gene classifier and the Veridex 76-gene classifier) reproducibly identified low-risk groups of patients with >95% relapse-free survival at 10 years - representing a substantial reduction in the number of patients who would unnecessarily receive systemic therapy according to conventional treatment guidelines. However, in the high-risk groups, 6070% of patients (representing about half of all patients in the cohort) also remained disease-free, indicating that the over-treatment problem is far from solved. In this context, we have begun to investigate the prognostic potential of gene-gene interactions, in the form of gene expression ratios that might offer greater prognostic resolution for the high-risk nodenegative breast cancer patients. Our systematic approach is, by necessity, highly computational, involving the evaluation of hundreds of millions of possible gene-gene combinations that might form “interaction signatures” with practical advantages over conventional gene signatures. In this presentation, I will describe our strategy and some preliminary findings that relate to the prognostic cross-cohort transportability of both individual genes and gene ratios, as well as a decision tree-based classification method that applies different gene signatures to different patients depending on predefined classification probability thresholds. 98 AACR Centennial Conference