Abstracts - Invited Plenary Sessions

Invited Abstracts Education Session: Assay Validation Educational Session Assay Validation ED01-01 Exploiting the success and failure of cancer therapies: Investing now in accurate biomarker data to improve outcomes later. Peter S. Nelson. Fred Hutchinson Cancer Research Center, Seattle, WA. Most therapeutic clinical trials of anti-neoplastic agents are not designed to anticipate or exploit the most likely outcome: failure to meet the primary endpoint. In the vast majority of these trials, the reason for failure is poorly understood. Unfortunately, clinical studies are rarely structured for iterative analyses of cause-and-effect, a core principle of scientific investigation. However, several recent studies demonstrate the power and importance of recognizing tumor and host subtypes, and measuring drug effects at the target level. The advent of new tools for assessing biomarkers of tumor and host heterogeneity and drug response have provided unprecedented opportunities for refining concepts of success and failure in clinical trials. This presentation will highlight key advances and discuss issues in the accurate assessment of biomarkers in the context of the complexities of normal human variation and disease phenotypes. ED01-02 Accurate, reproducible, and quantitative measurement of protein analyte concentration(s) in tissue slides. David L. Rimm,1 Robert L. Camp,1 Allison Welsh,1 Christopher B. Moeder,1 Jennifer M. Giltnane,2 Sharon Pozner-Moulis,3 Marisa Dolled-Filhart,4 Harriet Kluger,1 Gina G. Chung,1 Mark D. Gustavson,4 Jason Christiansen4. 1Yale University School of Medicine, New Haven, CT; 2Vanderbilt Medical School, Nashville, TN; 3Merrimack Pharmaceuticals, Cambridge, MA; 4HistoRx, New Haven, CT. Immunohistochemistry (IHC) has been used as a tool for over a quarter of a century to assess protein concentration within the context of the spatial information that is the keystone of anatomic pathology. During that time period the chromogenic method itself has changed very little and as a result the output is still largely subjective and fraught with error as a function of uncontrolled variables. This has led to recent attempts at standardization (the 2007 ASCO/CAP guidelines for HER2 assessment) which are likely to be just the beginning of the changes that will be necessitated by the push toward evidence-based medicine and personalized medicine requiring companion diagnostics. Unfortunately, those guidelines resort to “proficiency testing” as a standard, rather than a National Institute of Standards (NIST) approved, absolute physical standard that could be based on molecular parameters and reproduced at all users’ sites. Toward the goal of generating such a standard that could be used as a future method for absolute standardization of IHC, we have invented the fluorescence-based AQUA method. This technology uses molecular colocalization to derive exact protein concentrations within subcellular compartments. It allows assessment of accuracy and reproducibility in a manner similar to an ELISA or RIA assay and has been shown to have CVs less than 5%. However, its use has revealed significant, inherent variability associated with the process of measuring protein concentrations using antibody-based methods on formalin fixed, paraffin embedded tissue. Key examples include pre-analytic variables (like time to fixation), section oxidation, lot-to-lot antibody inconsistency and other variables. These variables can be detected and standardized with a series of extrinsic and intrinsic controls. Extrinsic controls are used to show the exact concentration of both HER2 and ER using standards that can be reproduced and tested, independent of site or personnel. Intrinsic controls are more challenging, and definitive methods are still in development. Ultimately, we believe that the measurement of a protein analyte in tissue on a slide will be as accurate, objective and reproducible as measurement of a serum sodium or a blood glucose. ED01-03 Statistical issues in biomarker assay development and evaluation. Viswanath Devanarayan. Abbott Laboratories, Parsippany, NJ. This presentation will focus on calibration-curve based biomarker assays (e.g., chromatographic and ligand-binding assays) and will address topics such as the evaluation of the validity of biomarker measurements and reference standards, selection of appopriate models and weighting options for the calibration curves, assessment of analytical range, sensitivity and acceptance criteria, and the fit-for-purpose requirements for exploratory and advanced biomarker assay validation. These topics will be illustrated using data from real case studies. Some popular misconceptions on topics such as the dynamic range and linearity will be clarified, and examples of how over 5-fold improvement in analytical sensitivity can be achieved simply by the use of appropriate statistical methods will be illustrated. This will be followed by an illustration of the impact of analytical and biological variability on the identification of novel biomarkers from small and large (-omic) biomarker studies. Educational Session Non-Coding RNAs ED02-01 MicroRNA reprogramming by oncogenes and tumor suppressors. Joshua Mendell. Johns Hopkins University, Baltimore, MD. Dysregulated microRNA (miRNA) expression is a ubiquitous feature of cancer cells and select miRNAs have been demonstrated to possess oncogenic and tumor suppressing activity. Yet the mechanisms that lead to altered miRNA expression in tumors are poorly understood. One possibility is that miRNAs are under direct control of oncogenes and tumor suppressors whose gain- or loss-of-function directly contributes to abnormal miRNA expression patterns. In order to investigate this question, we have studied the miRNA components of classic oncogenic and tumor suppressor signaling pathways, such as those controlled by c-Myc and p53. Expression profiling studies in human and mouse models of B cell lymphoma have revealed that a major consequence of c-Myc activation is extensive reprogramming of the miRNA transcriptome, including the direct upregulation of oncogenic miRNAs and the direct downregulation of tumor suppressing miRNAs. We have also identified miRNAs regulated by p53 that appear to participate in the anti-proliferative and proapoptotic functions of this tumor suppressor. Together, these studies demonstrate that abnormal miRNA expression in cancer cells cannot be explained solely as an indirect consequence of the loss of cellular identity that accompanies malignant transformation. Rather, oncogenic events directly reprogram the miRNA transcriptome to favor tumorigenesis. Moreover, miRNAs appear to function as critical downstream effectors of multiple oncogenic and tumor suppressor pathways. ED02-02 miRNome integrative analysis in ovarian cancer. Lin Zhang,1 Stefano Volinia,2 Tomas Bonome,3 George Adrian Calin,4 Nuo Yang,1 Vassilis Atlamazoglou,5 Chang-Gong Liu,2 Antonis Giannakakis,1 Joel Greshock,6 Barbara L. Weber,6 Michael J. Birrer,3 Artemis Hatzigeorgiou,1 Carlo M. Croce,2 George Coukos1. 1University of Pennsylvania Medical Center, Philadelphia, PA; 2The Ohio State University Comprehensive Cancer Center, Columbus, OH; 3National Cancer Institute, Rockville, MD; 4UT M. D. Anderson Cancer Center, Houston, TX; 5Agricultural University of Athens, Athens, Greece; 6GlaxoSmithKline, King of Prussia, PA. MicroRNAs (miRNAs) are an abundant class of endogenous small noncoding RNAs that function as negative gene regulators. We present data from integrative genomic analysis of the miRNome in human epithelial ovarian cancer including miRNA microarray, array-based comparative genomic hybridization, cDNA microarray, and tissue array. miRNA expression was markedly downregulated with malignant transformation and tumor progression. A high proportion of genomic loci containing miRNA genes exhibited DNA copy number alterations in ovarian cancer (37.1%). Copy number alterations observed in >15% tumors were 58 American Association for Cancer Research Education Session: Non-Coding RNAs Invited Abstracts considered significant. We identified 41 miRNA genes with gene copy number changes that were shared among human ovarian cancer, breast cancer and melanoma (26 with gains and 15 with losses) as well as miRNA genes with copy number changes that were unique to each tumor type. miRNA gene copy changes correlate with miRNA gene expression in ovarian cancer. Thus, genomic copy number loss may account for downregulation of ~15% of miRNA genes. Additionally, epigenetic silencing may account for downregulation of at least ~36% of miRNA genes. miRNA downregulation contributes to genome-wide transcriptional deregulation. Eight miRNAs located in chromosome 14 miRNA cluster (Dlk1-Gtl2 domain) were downregulated in advanced relative to early stage EOC; two of them, mir-495 and mir-410, are potential tumor suppressor genes and were predicted to target a large part of proteincoding genes that are upregulated in the same cancers. Additional miRNAs were found to be significantly associated with clinical outcome in late stage EOC. ED02-03 MicroRNAs in control of cell proliferation. Anindya Dutta. University of Virginia Health Sciences Center, Charlottesville, VA. We are using three approaches to discover microRNAs that regulate cell proliferation. In the first approach, we focused on miRNAs induced during differentiation of C2C12 myoblasts into myotubes in vitro. One of the microRNAs induced during differentiation, miR-206, suppresses cell proliferation. By surveying mRNAs that are downregulated by miR-206 for genes predicted to be targets of the microRNA, we discovered the mRNA of DNA polymerase alpha p180 is a direct target that is destabilized upon miR-206 transfection. We now report that additional microRNAs induced during muscle differentiation target other cell-cycle regulators. After noting that targeted mRNAs are often destabilized by microRNAs, in the second approach, we tested whether global downregulation of microRNAs by knockdown of Dicer or Drosha can help identify microRNArepressed mRNAs. The principle was proved by the discovery that HMGA2 oncogene is repressed by the growth-suppressive let-7 microRNA. Indeed, chromosomal translocations in lipomas and leiomyomas de-repress HMGA2 by deleting the 3’ UTR that is normally repressed by let-7. Conversely, downregulation of let-7 has been noted in lung cancers and large leiomyomas and is correlated with over-expression of the HMGA2 oncogene. In the third approach we have cloned short RNAs from androgendependent prostate cancer cells grown in the presence or absence of androgens and subjected them to ultra-high-throughput sequencing. The frequency of a number of clones present in the libraries is changed by androgens and we have identified several microRNAs that change upon androgen-depletion. Over 30-40% of the short RNAs cloned, however, do not correspond to known microRNAs and are produced by cleavage of known mRNAs and noncoding RNAs. The diversity and abundance of the non-micro-short RNAs (nmsRNAs) contrast with how little is known of their function and suggest that much remains to be done before we understand the biological functions of many short RNAs present in the cell. In an extension of the project we have identified several microRNAs that are reproducibly altered in expression level between the androgendependent LnCAP cells and androgen-independent derivative C4-2 cells. Interestingly, the microRNAs decreased in C4-2 cells were also reported by others to be decreased as prostate cancers advance in Gleason score or metastasize. This raises the possibility that a microRNA signature may be established for following the molecular stage of prostate cancer and for predicting its predilection to recur in an androgen-independent form. ED02-04 MicroRNAs in the diagnosis and prognosis of cancer. Carlo M. Croce. The Ohio State University, Comprehensive Cancer Center, Columbus, OH Progress in understanding the biology of multiple myeloma (MM), a plasma cell malignancy, has been slow. The discovery of microRNAs (miRNAs), a class of small noncoding RNAs targeting multiple mRNAs, has revealed a new level of gene expression regulation. To determine whether miRNAs play a role in the malignant transformation of plasma cells (PCs), we have used both miRNA microarrays and quantitative real time PCR to profile miRNA expression in MM-derived cell lines (n = 49) and CD138+ bone marrow PCs from subjects with MM (n = 16), monoclonal gammopathy of undetermined significance (MGUS) (n = 6), and normal donors (n = 6). We identified overexpression of miR-21, miR-106b ~25 cluster, miR-181a and b in MM and MGUS samples with respect to healthy PCs. Selective up-regulation of miR-32 and miR-17~92 cluster was identified in MM subjects and cell lines but not in MGUS subjects or healthy PCs. Furthermore, two miRNAs, miR-19a and 19b, that are part of the miR-17~92 cluster, were shown to down regulate expression of SOCS1, a gene frequently silenced in MM that plays a critical role as inhibitor of IL-6 growth signaling. We also identified p300-CBP-associated factor, a gene involved in p53 regulation, as a bona fide target of the miR106b~25 cluster, miR-181a and b, and miR-32. Xenograft studies using human MM cell lines treated with miR-19a and b, and miR-181a and b antagonists resulted in significant suppression of tumor growth in nude mice. In summary, we have described a MM miRNA signature, which includes miRNAs that modulate the expression of proteins critical to myeloma pathogenesis. Educational Session Molecular Imaging: From Mouse to Human ED04-01 Imaging of immune cell trafficking patterns refines development of cell-based therapies. Christopher H. Contag. Stanford University School of Medicine, Stanford, CA. The unique capabilities of immune cells to traffic to tumor targets in vivo and recognize malignant cells offer significant opportunities for the development of new therapeutic strategies for treating cancer. The graft vs. tumor effect in bone marrow transplantation is well recognized and there have been advances in both understanding and utilizing this effect in control of disease progression. Despite these advances, effective use of immune cell therapy for control tumor growth has not been achieved. In vivo imaging approaches can be used to monitor cell survival, assess in vivo trafficking patterns and assess the effects of immune cell therapy on tumor targets (1, 2). Revealing immune cell behavior in the living body through imaging will lead to insights that can be used to refine immune cell therapies and develop new approaches for control of malignancy. Given the complexity of immune cell trafficking and the use of these patterns to deliver other therapies, and the requirement that delivery methods be evaluated in living animals, effective development of these approaches requires methods of study that refine animal models of cancer and accelerate their analyses. The tools being developed in the field of molecular imaging have addressed this previously unmet need and have enhanced the discovery and development of novel therapies for cancer. These imaging methods utilize molecular probes to label the cancer cell, the immune cell delivery vehicle and/or the therapeutic agent such that their numbers and location can be determined in animal models over time (3). Such imaging tools offer the power of repeated measures without sacrificing the animal yielding improved data while offering a significant saving in the numbers of animals required and the cost of the study. Among the molecular imaging methods described to date, in vivo bioluminescence imaging (BLI) has emerged as one of the most versatile and sensitive methods for in vivo studies of biology and therapeutic responses in animal models. As such BLI has been used to advance our understanding of a number of chemotherapeutic agents and biological therapies for cancer and other diseases. Expression of luciferases can be used to create to light-emitting cells, which can be studied in correlative cultures assays and then used in animal models where a low intrinsic background signal from the host animal provides significant signal to noise ratios. This relatively inexpensive and easy to use imaging modality, that doesn’t require radioactivity and uses substrates that are relatively nontoxic, is ideally suited to small animals, such as mice and rats. These laboratory rodents are small enough to allow transmission of light through Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 59 Invited Abstracts Education Session: Molecular Imaging: From Mouse to Human their tissues from luciferase-expressing cells or organs to the surface, where the photons can then be detected by sensitive camera systems that are based on charge-coupled devices. The response of the cells expressing luciferase, or the expression of luciferase by a promoter of interest, can thus observed in the complex environment of the living body. We have used this imaging approach to develop immune cell therapies and to optimize a combined biological therapy comprised of immune cell therapy and engineered viruses that were designed to be tumor selective in their replication; oncolytic viruses (4). Similar to chemotherapies, development of biological therapies is based on targeting the unique biology of cancer cells to achieve an effective therapeutic index against malignant versus normal cells. By necessity, measurements of therapeutic indices are performed in cell culture, but these assays cannot be used to develop in vivo delivery tools that effectively transport therapeutic agents to the tumor target in living subjects. It is not surprising, therefore, that despite the development of extremely potent anticancer drugs and biological therapies, their designs lack mechanisms for crossing biological barriers for directed delivery to malignant cells in vivo. Selective delivery of a therapeutic agent to a target cell can further enhance the therapeutic index in vivo due to selective accumulation of the agent within the tumor. This can both effectively reduce exposure of normal tissues to toxic agents and achieve a higher effective concentration at the tumor site and imaging of cell migration and efficacy guides development of this therapeutic strategy. The combination of cytokine induced killer (CIK) cells as the carrier vehicle and oncolytic vaccinia virus as its biological payload fall into the category of cell-based combination therapies that may offer additive or synergistic therapeutic benefits (4). CIK cells have been shown to display impressive tumor-trafficking potential following systemic delivery in preclinical models, with the majority of the cells detected in the tumor by 72h after intravenous delivery. CIK cells were developed as a tumorical population of immune cells that have characteristics of both T cells and natural killer (NK) cells. Although these cells display cytolytic ability in correlative cell culture assays and in vivo within the tumor microenvironment, large numbers of CIK cells are typically required to achieve efficient tumor clearance-effector to target ratios of 10 or 20 to one. Oncolytic vaccinia viruses have been developed as highly lytic agents with good tumor selectivity, but they are limited in their biodsitribution. By utilizing CIK cells as a carrier vehicle to deliver oncolytic vaccinia to the tumor we overcome the deficiencies of each therapy as a single agent (inefficient tumor delivery of vaccinia and limited tumor cell killing by CIK cells), while retaining their benefits. Blending an active targeting mechanism with a biologically active delivery vehicle (e.g., an immune cell) increases the therapeutic effect of the targeted agent through active concentration at the tumor site. Such approaches rely on cell types that are recruited by, or naturally traffic to, the tumor microenvironment, and these act as carrier vehicles to deliver a therapeutic agent or payload to the tumor. Tumor-targeting cells that have been used for such directed therapy include hematopoietic or immune cells, or precursor or stem cells. Since these cells have their own biology that can be modified and exploited for improved therapies, understanding and utilizing the biology of the vehicle will be essential for optimizing these approaches. The use of immune cells as carrier vehicles to deliver oncolytic virus strains to tumors has the potential to become a major weapon in the fight against cancer. References 1. Negrin RS, Contag CH. In vivo imaging using bioluminescence: implications for understanding graft-versus-host disease. Nat Rev Immunol 2006;6:484-90. 2. Shachaf CM, Kopelman A, Arvanitis C, Beer S, Mandl S, Bachmann MH, et al. MYC inactivation uncovers stem cell properties and induces the state of tumor dormancy in hepatocellular cancer. Nature 2004;431:1112-7. 3. Contag CH. In vivo pathology: Seeing with molecular sensitivity and cellular resolution in the living body. Ann Rev Pathol Mech Dis 2007;2:277-305. 4. Thorne SH, Negrin RS, Contag CH Synergistic antitumor effects of immune cell-viral biotherapy. Science. 2006;311:1780-4. ED04-02 Monitoring of tumor response to therapy by 18F-ML-10, a novel small-molecule PET tracer for apoptosis: From preclinical to clinical studies. Anat Shirvan,1 Hagit Grimberg,1 Galit Levin,1 Avi Cohen,1 Ayelet Reshef,1 Ayelet Akselrod-Balin,2 Aaron M. Allen,3 Eyal Fenig,3 Adam Steinmetz,3 David Groshar,3 Edna Inbar,3 Yulia Kundel,3 Eyal Mishani,4 Ilan Ziv1. 1Aposense, Ltd., Petach-Tikva, Israel; 2Weizmann Institute of Science, Rehovot, Israel; 3Rabin Medical Center, Petach-Tikva, Israel; 4Hadassah Medical Center, Jerusalem, Israel. Background: Apoptosis is a key mechanism of cell death associated with most anticancer treatments, including radiotherapy and chemotherapy. Non-invasive in vivo imaging of apoptosis may therefore allow early and real time assessment and optimization of treatment based on the biological tumor response in individual patients. 18F-ML-10 is a novel, compact PET tracer (α-methyl 18F-alkyl-dicarboxylic acid, MW=206) for apoptosis, that selectively accumulates in apoptotic cells from early stages of apoptosis, while being excluded from viable or necrotic cells. In preclinical studies in xenograft models in mice, upon intravenous systemic administration of the tracer in vivo, marked and selective accumulation of ML-10 was observed in tumors treated with irradiation or with various chemotherapeutic agents. Increased uptake of ML-10 was associated with tumor shrinkage, and confirmed histologically with apoptotic markers of DNA fragmentation, caspase-3 activation, and membrane phospholipid scrambling. Bio-distribution studies in vivo showed increased uptake and accumulation of ML-10 specifically in the tumor in comparison with other organs or the blood, with rapid clearance of the tracer from non-target sites. Specificity of binding was demonstrated by competitive blocking with excess of non-labeled “cold” compound. In a Phase I clinical trial in healthy volunteers, 18F-ML-10 showed favorable safety, biodistribution, dosimetry, stability and clearance profiles. Efficacy studies indicated selective accumulation of 18F-ML-10 in sites of high physiological apoptotic targets, such as the testes of healthy volunteers and in sites of pathological apoptosis, such as cerebral infarct in patients with acute cerebral stroke, creating a clear image of the target. We report here the results of the first study assessing the suitability of 18F-ML-10 for early detection of tumor response to radiation therapy in patients with brain metastases, and the respective correlation with subsequent alterations in tumor size. Methods: Patients with brain metastases, scheduled for whole brain radiation therapy (WBRT) underwent 18F-ML-10 imaging prior to (baseline), during and immediately upon completion of WBRT. Changes in 18F-ML-10 uptake relative to baseline were compared with clinical tumor response, as detected by Gd-enhanced MRI images, performed before and at 6-8 weeks after completion of treatment, and as assessed according to the RECIST criteria. Results: In 6 patients (out of 10 planned), 16 metastatic lesions were detected by Gd-MRI. Baseline 18F-ML-10 images demonstrated specific uptake in all lesions, with regions of various uptake intensities, in alignment with sporadic endogenous apoptosis within these tumors, naturally-occurring even before its augmentation by radiotherapy. Dynamic analysis showed accumulation and retention of 18F-ML-10 in the target lesions with time, while being cleared from the blood, supporting the specificity of uptake. So far 5 patients completed WBRT, and in all patients, the comparison of the baseline and post treatment 18F-ML-10 PET images 60 American Association for Cancer Research Education Session: Molecular Imaging: From Mouse to Human Invited Abstracts showed substantial treatment-related changes in uptake of 18F-ML-10 within the tumor. A Voxel-based characterization of alterations in the uptake map of 18F-ML-10 indicated changes in the intensity, in the extent of tumor volume exhibiting apoptotic response and in distribution of apoptotic areas within the target lesion. Such changes are quantitatively represented both graphically and visually. Complete follow-up is available for one patient, in whom the 18F-ML-10 uptake increased by 30%-60% after treatment, which was subsequently found to correspond to a 27-30% reduction of tumor size per Gd-MRI at 6 weeks post treatment. Conclusions: PET imaging of apoptosis with 18F-ML-10 may have an important clinical role in monitoring the effect of radiation therapy in individual patients, providing early assessment of tumor response, and thus assisting in treatment optimization for the individual patient. ED04-03 Detecting tumor responses to treatment with magnetic resonance imaging. Kevin M. Brindle. University of Cambridge, Cambridge, United Kingdom. Patients with similar tumor types frequently have markedly different responses to the same therapy. The development of novel targeted cancer therapies could benefit significantly, therefore, from the introduction of imaging methods that allow an early assessment of treatment response in individual patients. These would allow an oncologist to rapidly assess the effectiveness of a new therapy. Ineffective treatments could be abandoned at an early stage and more effective treatments selected, with attendant welfare benefits for the patient and cost benefits for the health care system (1, 2). We have been developing non-invasive and clinically applicable magnetic resonance-based methods for detecting the early responses of tumors to therapy. A primary focus has been on the development of methods for detecting tumor cell death, since the level of tumor cell death following drug treatment has been shown, in preclinical and clinical studies, to be a good prognostic indicator for treatment outcome. Thus, by monitoring tumor cell death we may get an indication of whether a particular drug is working very early during treatment, possibly within 2448 hours, and long before there is any evidence of tumor shrinkage. I will briefly describe the different approaches that we, and others, have taken in detecting and imaging cell death in tumors using magnetic resonance [reviewed in (3)]. There are several metabolic markers of cell death that can be detected using 31P MRS; the problem with these markers, however, is that they are relatively insensitive and therefore lack both temporal and spatial resolution. Cell death can also be detected from 1H MRS measurements of cytoplasmic neutral lipid accumulation. Although more sensitive, this technique is unlikely to be able to detect apoptosis in a tumor when the percentage of apoptotic cells is relatively low. A potentially much more sensitive approach is to use a targeted contrast agent that allows detection of apoptotic cells by MRI. We showed that the C2A domain of the protein synaptotagmin would bind to the phosphatidylserine that appears on the surface of apoptotic cells. When conjugated to small paramagnetic iron oxide (SPIO) nanoparticles this protein allowed MRI detection of apoptotic cells, both in vitro and in vivo, in a tumor treated with chemotherapeutic drugs (4). More recently we have developed Gd3+based versions of this agent, which give positive contrast and are therefore easier to detect, and which are smaller and thus more readily enter the tumor interstitium. The unbound material also clears more easily, thus generating better tissue contrast (5-7). However, the increased accumulation of the simplest of these Gd3+-based agents in treated tumors, which was based on analysis of the whole tumor volume, was relatively modest (1.19±0.04). We knew from the histology, however, that cell death was localized within the tumor, which suggested that we might be able to better detect contrast agent accumulation in a treated tumor by analyzing its distribution as well as its concentration in the tissue. We have shown that 2D Minkowski functionals, which have been used previously in cosmology as precise morphological and structural descriptors of the evolution and morphology of galaxies (8-10), can be used to parameterize the heterogeneity of contrast agent accumulation, and that these parameters allow us to better discriminate between treated and untreated tumors. This image analysis approach substantially increases the sensitivity and specificity of detection of the contrast agent in drug-treated tumors and could be applied in the analysis of uptake of other targeted agents. A fundamental limitation of magnetic resonance is its relatively low sensitivity due to the very low levels of nuclear spin polarization that can be achieved, even in high-field magnets. Recently a practicable method has been developed to enhance the nuclear spin polarization of nuclei such that gains in sensitivity greater than 10,000-fold can be achieved (11). This has enabled the imaging of 13C-labelled cellular metabolites in vivo and, more importantly, their enzymatic transformation into other species. We showed recently that exchange of hyperpolarized 13C label between the carboxyl groups of lactate and pyruvate, in the reaction catalyzed by the enzyme lactate dehydrogenase, could be imaged in tumors and that this flux was decreased in treated tumors undergoing drug-induced cell death (12). We have suggested that this technique could be used in the future for response monitoring in the clinic, in much the same way as 18FDG has been used with PET. We have also shown that we can monitor the conversion of hyperpolarized [5-13C]glutamine to glutamate in human hepatoma cells in vitro (13). Since glutaminase activity has been correlated with the rate of cellular proliferation, measurements of its activity in vivo could be used to detect the early responses of tumors to cytotoxic and cytostatic drugs, as has been done with 3’- deoxy-3’-[18F]fluorothymidine (FLT) and PET. Since alterations in tissue pH underlie many pathological processes, the capability to image tissue pH in the clinic could offer new ways of detecting disease and response to treatment. We have shown that tissue pH can be imaged in vivo from the ratio of the signal intensities of hyperpolarized H13CO3- and 13CO2 following intravenous injection of hyperpolarized H13CO3- [14]. The technique was demonstrated with a study on a mouse tumor model, which showed that the average tumor pH was significantly lower than the surrounding tissue. Since bicarbonate is already used intravenously in humans, we propose that this technique could be used clinically to image tumors and other pathological processes that are associated with alterations in tissue pH such as ischemia and inflammation. References 1. Neves AA, Brindle KM. Assessing responses to cancer therapy using molecular imaging. Biochim Biophys Acta 2006;1766:242-61. 2. Brindle KM. New approaches for imaging tumor responses to treatment. Nature Rev Cancer 2008;8:1-14. 3. Kettunen MI, Brindle KM. Apoptosis detection using magnetic resonance imaging and spectroscopy. Prog Nucl Mag. Reson Spectrosc. 2005;47:175-85. 4. Zhao M, Beauregard DA, Loizou L, Davletov B, Brindle KM. Non-invasive detection of apoptosis using magnetic resonance imaging and a targeted contrast agent. Nat Med 2001;7:1241-4. 5. Jung HI, Kettunen MI, Davletov B, Brindle KM. Detection of apoptosis using the C2A domain of synaptotagmin I. Bioconjug Chem 2004;15:983-7. 6. Neves AA, Krishnan AS, Kettunen MI, Hu D-E, deBacker MM, Davletov B, Brindle KM. A paramagnetic nanoprobe to detect tumor cell death using magnetic resonance imaging. Nano Lett. 2007;7:1419-23. 7. Krishnan A.S., Neves A.A., de Backer M.M., Hu D.-E., Davletov B., Kettunen M.I., and Brindle K.M. (2008). Detection of cell death in tumors using MRI and a gadolinium-based targeted contrast agent. Radiology. 246, 854-862. 8. Schmalzing J, Buchert T. Beyond genus statistics: A unifying approach to the morphology of cosmic structure. Astrophys J 1997;482:L1-L4. 9. Schmalzing J, Gorski KM. Minkowski functionals used in the morphological analysis of cosmic microwave background anisotropy maps. Monthly Notices of the Royal Astronomical Society. 1998;297:355-65. 10. Schmalzing J, Kerscher M, Buchert T. Dark matter in the universe. Proceedings of the International School of Physics Enrico Fermi 1996;132: 281-91. Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 61 Invited Abstracts Education Session: Molecular Imaging: From Mouse to Human 11. Ardenkjaer-Larsen JH, Fridlund B, Gram A, Hansson G, Hansson L, Lerche MH, et al. Increase in signal-to-noise ratio of >10,000 times in liquid-state NMR. Proc Natl Acad Sci USA 2003;100:10158-63. 12. Day SE, Kettunen MI, Gallagher FA, Hu D-E, Lerche M, Wolber J, et al. Detecting tumor response to treatment using hyperpolarized 13C magnetic resonance imaging and spectroscopy. Nature Med. 2007;13:1382-7. 13. Gallagher FA, Kettunen MI, Day SE, Lerche M, Brindle KM. 13C magnetic resonance spectroscopy measurements of glutaminase activity in human hepatocellular carcinoma cells using hyperpolarized 13C-labeled glutamine. Magn Reson Med 2008; in press. 14. Gallagher FA, Kettunen MI, Day SE, Hu D-E, Ardenkjaer-Larsen JH, in’t Zandt R, et al. Magnetic resonance imaging of pH in vivo using hyperpolarized 13C-labeled bicarbonate. Nature 2008;453:940-3. Educational Session From Discovery to Product ED05-01 Developing tissue-based predictive biomarkers in oncology. Paul M. Waring. University of Western Australia, Crawley, Australia. Over the past 30 years, the number of new drugs approved each year has steadily declined while the cost of developing each new drug has grown almost exponentially. It currently costs approximately US $900 million to bring a drug to market, mainly because the industry has such a high failure rate and because this failure occurs so late in development. A survey of the 10 largest pharmaceutical companies conducted between 1991 and 2000 showed that the overall failure rate for oncology drugs was 40% during Phase I, 70% during phase II, 60% during Phase III and of the few showing success in Phase III, 30% were not approved (1). This means that a new oncology drug entering clinical trials has a one in fourteen chance of being used in clinical practice. The pharmaceutical and biotechnology industry cannot be sustainable with the high attrition rates currently operable in drug discovery and development. To reduce this high attrition rate, and the associated costs, pharmaceutical companies need to be able to make confident decisions regarding efficacy and toxicity as early as possible during the drug development process and to ensure that those patients most likely to benefit from the drug are enrolled into the pivotal Phase III trials. Unfortunately, there are few tools available that enable “quick to kill” and “quick to proof of concept” decisions to be made during the early phases of drug development. Most oncology trials rely on CT scan measurement of tumor size or serum tumor markers to assess early clinical response; however, tumor shrinkage or falling serum tumor marker levels are often poor predictors of survival benefit and may not reveal the clinical benefit of cytostatic drugs. Randomized Phase II trials are frequently used to seek early signals of efficacy, however these are accompanied by high false positive and false negative rates, conferring considerable risk when making the critical decision to proceed, or not to proceed, to a pivotal Phase III trial. Underlying these issues are the dual realities that most cancers are heterogeneous and most drugs only benefit a subset of patients with a particular disease. The traditional “trial and error” practice of treating “all comers” in pivotal trials means that many patients have to be treated in order to identify the few that derive clinical benefit. This approach is being progressively replaced by targeted therapies that are coupled to biomarkers which help identify patients more likely to show a treatment benefit or experience adverse events. The hope is that by identifying tumors that are dependent upon the targeted pathways and by demonstrating that the drug modulates the pathway it will be possible to identify the right patients for the pivotal trials and hence significantly increase the probability of success. Although there have been a few promising examples, the successful broad implementation of this approach, however, is proving to be challenging, especially in oncology where biomarker assays are often performed on tumor samples. Once a biomarker has been identified, the technology platform selected, the assay analytically validated and its intended use defined, the challenge becomes how to incorporate the biomarker assay into the drug development program and design of the Phase II and III clinical trials. Decisions have to be made regarding whether to evaluate the biomarker prospectively or retrospectively; whether to collect and test tumor or surrogate tissue samples; whether to retrieve archival tumor samples from pathology laboratories or arrange for fresh biopsies to be taken; how best to power the trial to test both the drug and the biomarker; and how to design the study to determine whether the biomarker has prognostic as well as predictive value. Prospective clinical validation of a biomarker to be used to select or stratify patients being enrolled in a pivotal trial is not a trivial undertaking. Most importantly, it requires a high level of a priori confidence in the biomarker’s predictive value. Furthermore, the assay has to be in a commercializable form and the testing laboratories shown to be proficient in performing the assay prior to enrolling the first patient. For a novel biomarker or assay platform, this could take at least three years, so assay development needs to begin as early as possible. If archival tumor samples are to be tested, efficient processes have to be in place to quickly retrieve the patients’ samples, often from hundreds of pathology laboratories, otherwise there will be unacceptable enrollment delays. With fresh tumor biopsies, ethical and logistical challenges invariably restrict their use to certain indications and a few carefully selected Phase II trial sites. Surrogate samples, such as blood and serum, while being much more accessible carry more risk because the biomarker may not be representative of the tumor (e.g., serum proteomics) or may only be informative in a small subset of patients (e.g., circulating tumor cells or DNA). Accordingly, few surrogate biomarkers have been adequately validated to enable decision making in clinical trials. Retrospective biomarker analysis has the advantage that patient enrollment is not compromised; the assay does not need to be ready prior to study commencement; and multiple hypotheses, biomarkers, and assay cut points can be evaluated. However, this approach is only suitable for exploratory biomarker studies and is rarely a path to regulatory approval. Furthermore, it may not be possible to perform a follow-on prospective biomarker validation study because it would be ethically unacceptable to deny the control group a drug with now proven efficacy. Retrospective biomarker studies performed on retrieved archival tumor samples are often inconclusive because samples can only be retrieved on a proportion of the trial subjects and the quality and volume of tissue limits the type and power of biomarker analyses. This combined therapeutic and diagnostic paradigm holds great promise but also creates new challenges. In this talk, I shall explore several issues that arose during the development of biomarker assays that highlight the need for greater awareness among patients, patient advocates, institutional review boards, clinicians, and pathologists of the critical need to obtain tumor samples in order to identify which patients are most likely to benefit. References 1. Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Discovery 2004;3:711-5. 62 American Association for Cancer Research Education Session: From Discovery to Product Invited Abstracts ED05-02 Personalized healthcare strategies and challenges in oncology. Walter Koch. Roche Molecular Systems, Pleasanton, CA. The increasing discovery and development of targeted cancer therapeutics is giving rise, in some cases, to the need for companion diagnostic assays to select which patients are candidates for a particular therapy. The most straightforward application for these assays is to detect the presence of the drug target itself—for example, HER2 overexpression—to select breast cancer patients most likely to benefit from traztuzumab treatment. Because the therapeutic target is known even before clinical trials are initiated, it is in principle easier to develop a test method for selection of patients that is close to the final format of the in vitro diagnostic (IVD) companion assay that may eventually be coregistered with the therapeutic, should it prove safe and effective. In other cases, predictive biomarkers are discovered during drug development, or even after a therapeutic has been approved for clinical use. A recent example of such a biomarker is the link between non-mutated, wild-type KRAS in tumors, and the efficacy of treatment with anti-EGFR therapies such as cetuximab and panitumumab, in patients with metastatic colorectal cancer. Clearly this latter scenario poses several challenges for rapid development of an IVD test that might be required by global regulatory agencies. Aligning diagnostic test development under Design Control with the pharmaceutical product development process is non trivial for various reasons. Therefore, we have implemented an integrated organizational structure and processes that facilitate use of scientific insights from diagnostics in drug development, and optimize in parallel the various stages of diagnostic and drug development. This approach is improving understanding of disease heterogeneity at the molecular and individual level, and identification of biomarkers to characterize or select patient populations for clinical trials that are most likely to benefit. For situations where nucleic acid analytes are expected to be informative, standardized procedures and kits have been developed for DNA or RNA extraction from formalin-fixed paraffin-embedded tissue (FFPET) sections. These sample preparation kits have been coupled with PCR-based assays developed for panels of biomarkers, including common oncogene mutations, as well as mRNA quantification assays for genes whose differential expression is associated with disease pathology and/or drug response. The PCR assays use generic, standardized master mixes manufactured in a GMP facility, for ultimate use on IVD instruments. The advantage of this approach is more rapid translation from biomarker discovery to standardized assays for use in translation research, clinical validation studies, and when necessary, approvable IVD companion diagnostic assays. A current example to illustrate the benefits of this strategy is the codevelopment of PLX4032, a novel small molecular inhibitor designed to selectively inhibit the cancer causing form of the B-Raf kinase, together with a PCR assay designed to detect the BRAFV600E mutation in FFPET samples. Activating mutations in the BRAF oncogene have been reported in more than 65% of cases of malignant melanoma and a variety of other malignancies, including colorectal cancer. The BRAFV600E missense mutation is observed in over 80% of the oncogenic BRAF alleles to date and is associated with constitutive activation of B-Raf kinase. For these cancers, selective inhibition of oncogenic B-Raf kinase represents a rational therapeutic strategy. A Phase I safety study of PLX4032 is underway including a secondary objective to assess pharmacodynamic activity in patients with malignant melanoma who have the V600E BRAF oncogenic mutation. Additional examples of diagnostic and therapeutic codevelopment programs will be discussed. Educational Session Molecular Marker Driven Trial Design ED06-01 Statistical issues in identifying, validating, and using molecular markers in clinical trials. Marc E. Buyse. International Drug Development Institute, Ottignies, Belgium. Definitions: A biomarker can be formally defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention”. Biomarkers can include biochemical markers, cellular markers, cytokines, genetic markers, gene expression profiles, imaging markers, physiological markers, or any other measurements made on the tumor. They can be measured once before a treatment is administered, or repeatedly before, during and after the treatment is administered, in which case interest focuses on changes in the biomarkers over time. Biomarkers can be classified into three broad groups: • Prognostic biomarkers, which affect the outcome of patients in terms of a clinical endpoint. • Predictive biomarkers, which affect the effect of a specific treatment on a clinical endpoint. • Surrogate biomarkers, which may replace a clinical endpoint in clinical trials carried out to evaluate the effect of a specific treatment. Identification and validation of biomarkers: Prognostic biomarkers. A prognostic biomarker affects the clinical outcome regardless of treatment. Hence the baseline value of the biomarker, or changes in the biomarker over time, should be correlated with the clinical endpoint in untreated or in treated patients. This condition is straightforward to establish: the prognostic impact of a biomarker can be investigated retrospectively in a series of patients providing a sufficient number of clinical events are available. In addition, a prognostic biomarker will be of clinical interest only if its impact on the clinical endpoint of interest is large enough, hence the number of patients required will often be smaller than what is required to establish or confirm a treatment benefit. The ideal setting to identify and validate a prognostic biomarker is a randomized therapeutic trial, in which all procedures to measure the biomarker are specified in the protocol (biological material collected, methods of preservation, assays and quantitation methods used, quality control procedures, reproducibility assessments, etc.) and in which all patients fulfill well-defined characteristics and are treated and followed according to a pre-determined schedule. One key issue is the choice of appropriate statistic(s) to quantify the impact of a prognostic biomarker on the clinical outcome of interest. The p-value of a test comparing the clinical outcome in two groups of patients, one with and another without the biomarker, provides insufficient evidence that the biomarker is of any clinical use. Measures that quantify the magnitude of effect of the biomarker on the clinical endpoint, such as the odds ratio (for tumor response and other dichotomous endpoints) or the hazard ratio (for survival time and other time-related endpoints), along with their confidence limits, are far more informative. However these statistics are not sufficient to gauge the performance of the prognostic biomarker. Measures of predictive accuracy are needed as well, and these are usually provided by the sensitivity and specificity of the biomarker, or equivalently by its positive and negative predictive value. If the biomarker is a continuous measurement, ROC (Receiver Operating Characteristics) curves are informative. Last but not least, a biomarker is of interest only if it provides additional prognostic value, over and above that of all easily measured clinical and pathological characteristics of the patients. Predictive biomarkers. A predictive biomarker modifies the effect of treatment on the clinical outcome. Hence, the baseline value of the biomarker, or changes in the biomarker over time, should be correlated with the effect of treatment on the clinical endpoint of interest. This condition is far more difficult to establish than a mere prognostic impact of the biomarker on the endpoint. It is a common misconception that a Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 63 Invited Abstracts Education Session: Molecular Marker Driven Trial Design biomarker having prognostic value in a group of treated patients is predictive, even if the biomarker does not have prognostic value in another group of untreated patients. Indeed, the lack of impact in one group and the impact in another may be due to different selection criteria and other confounding factors that make such an indirect comparison untenable. The most reliable way to formally identify and validate a predictive factor is through a comparative trial, where patients are randomized either to a standard treatment or to an experimental treatment for which a predictive biomarker is thought to exist, since such a design makes treated and untreated patients comparable. The statistical evidence required to establish that a biomarker is truly predictive is an open question. The most convincing situation would be one in which an interaction test between the effect of treatment and the biomarker status reaches statistical significance. The major problem of interaction tests is that they lack power, so that a very large trial would be required for the test to reach significance. Often biological considerations will alleviate the need for definite statistical evidence. For instance, it is acceptable to consider HER2-neu status as a predictive biomarker for herceptin, even though no study was carried out to formally test for an interaction between the biomarker and the targeted agent. Surrogate biomarkers. A surrogate biomarker must be able to replace a clinical endpoint for the purposes of evaluating the effect of an experimental treatment as compared with a standard treatment. The requirements for a biomarker to be considered a valid surrogate have been a theme of intense debate in the statistical literature over the last years. In early discussions about surrogate biomarkers, a common misconception was that it was sufficient for a biomarker to be prognostic for the clinical endpoint to establish surrogacy. It was later argued that the presence of an association between the biomarker and the clinical endpoint, no matter how strong, does not imply surrogacy. An additional key requirement is that there is also an association between treatment-induced changes in the biomarker and corresponding changes in the risk of the clinical endpoint. The best theoretical setting to validate a potential surrogate biomarker consists of several large-scale randomized clinical trials in the context of a meta-analysis, or a large randomized trial that can be broken down in smaller, clinically meaningful units, such as countries or clinical sites. In order for a surrogate biomarker to be used in the clinical development of a new drug, surrogacy would have to be demonstrated across a range of treatments testing multiple new drugs, or one new drug compared with multiple comparators. This again suggests recourse to a meta-analysis of several randomized trials. Uses of biomarkers: Biomarkers are already in common use in patient management: PSA is used to monitor progression of disease in prostate cancer, CEA in colorectal cancer, etc. This does not automatically imply that these biomarkers are useful for clinical research purposes. In terms of clinical development of new treatments, biomarkers can be used to select patients eligible for clinical trials, to stratify patients at entry in clinical trials, to monitor patients and guide treatment decisions, or to substitute for a clinical endpoint in the evaluation of the effects of new treatments. Prognostic and predictive biomarkers. Prognostic biomarkers are useful primarily to adjust the therapy for individual patients, with patients of poor prognosis being treated more aggressively than patients of good prognosis. Although prognostic biomarkers can be useful, it is mostly predictive biomarkers that have the potential of changing oncology practice in the near future. Indeed, the presence of unknown predictive factors in a patient population can profoundly affect the statistical power of trials aimed at showing the benefit of new treatments for these patients. Conversely, if a biomarker could reliably identify a subset of patients who derive the most benefit (or the least toxicity) from a new treatment, then clinical trials could be restricted to this subset, and the treatment of future patients could be targeted at this subpopulation only. Surrogate biomarkers. The search for prognostic or predictive biomarkers has been intense over the last years, but the real long-term hope is to identify a surrogate biomarker, i.e. a biomarker capable, in and of itself, to predict that a new treatment will have a desired effect on the ultimate endpoint of interest. Currently, no such biomarkers have been identified in oncology. One of the most intensely studied cancer biomarker is prostate-specific antigen (PSA). Changes in PSA often antedate changes in bone scan, and they have long been used as an indicator of response in patients with androgen-independent prostate cancer. Notwithstanding the usefulness of PSA for patient management, it was recently shown that PSA could not be considered a surrogate for long-term clinical endpoints. Similarly, tumor shrinkage was shown not to be an acceptable surrogate endpoint for overall survival in advanced colorectal cancer. Even so, biomarkers (such as tumor response) have often been used for drug approval in situations where the number of patients and / or the follow-up required to show the treatment effect on the long-term clinical endpoints would have led to unrealistic trials. ED06-02 Neoadjuvant trials in breast cancer: Integrating biology, imaging, and response to treatment. Angela DeMichele. Abramson Cancer Center, University of Pennyslvania, Philadelphia, PA The neoadjuvant setting provides a rich translational opportunity to better understand response of breast tumors to standard and novel therapies. Numerous trials have utilized this approach to discover and validate molecular profiles, genetic and protein biomarkers and imaging modalities. The goal of such studies is to refine optimal treatment selection, identify appropriate patients for targeted therapeutics and noninvasively monitor response. However, such trials pose a variety of scientific, logistical and clinical challenges to their conduct and the integration of these approaches to improve clinical practice has yet to be fully realized. The ISPY Trial is a multi-institutional correlative science trial of locally advanced breast cancer integrating serial biopsy to characterize biology and MRI/MRS to measure response of tumors to neoadjuvant therapy. In this educational session, results of ISPY and similar neoadjuvant trials will be used to illustrate the major issues surrounding the design and conduct of such trials. Screening and recruitment, sample procurement and processing, patient consent issues and data analysis approaches will be examined. Finally, the evidence supporting individualized treatment selection and applications of these trial results will be reviewed, with a focus on how such trials will shape both therapeutic development and clinical care of breast cancer in the future. 64 American Association for Cancer Research Opening Session: Keynote Address Invited Abstracts Opening Session Keynote Address KN01 Models of molecular diversity to facilitate marker guided therapy. Joe W. Gray, Debopriya Das. Lawrence Berkeley National Laboratory, Berkeley, CA. Several hundred drugs targeted against molecular defects in cancer are now being tested clinically. This number will grow substantially as comprehensive analyses of cancer genomes are completed. Early development of biomarkers that predict clinical response will allow these drugs to be tested in tumors most likely to respond. This will increase the probability that drugs effective in tumor subpopulations will not be missed and will decrease the cost of early trials by allowing them to be tested in small numbers of patients whose tumors are most likely to respond. It also will allow predictive biomarkers to be developed in parallel with experimental drugs so that successful drugs enter the clinical market with fully validated predictive biomarkers. We are pursuing the hypothesis that selection of predictive markers for breast cancers can be accomplished by assessing responses to new drugs in an in vitro system comprised of 50 breast cancer cell lines and correlating these with genomic, transcriptional and proteomic features measured in order to identify molecular features associated with response. This in vitro systems approach is based on our observations that the cell line collection models much of the molecular diversity and recurrent aberrations found in primary tumors and that predictive markers can be developed using information about molecular features associated with response in vitro. In order to test the in vitro approach, we have developed novel computational and experimental approaches for identification of non-linear signatures of response and we have applied these in assessing responses to ~40 approved and experimental therapeutic agents. We have tested the clinical utility of one of these— a 6-gene predictive assay for lapatinib developed using this approach, in two independent clinical trials. Plenary Session Pharmacogenomics PL02-01 Personalized medicine: Advantages and shortcomings of genomics versus metabonomics. Daniel W. Nebert. University of Cincinnati Medical Center, Cincinnati, OH. Sequencing of the entire human genome, the mapping of common haplotypes of single-nucleotide polymorphisms (SNPs) by HapMap Phases I and II, the Encyclopaedia of DNA Elements (ENCODE) Project, and costeffective genotyping technologies leading to genome-wide association studies (GWAS) have led recently to incredible advances in the fields of human genomics and pharmacogenomics. These breakthroughs have combined convincingly in the past several years to demonstrate the requirements needed for all genotype-phenotype association studies, and to separate true associations from the plethora of false positives. While research in human genetics has moved from monogenic to oligogenic to complex diseases, its pharmacogenetics branch has followed, usually a few years behind. The continuous discoveries, even today, of new surprises about our genome cause us to question the large number of reviews declaring that “personalized medicine is almost here” or that “individualized drug therapy will soon be a reality.” After the 2007 publication of the ENCODE data—in which ~1% of selected portions of the human genome was characterized and studied in-depth including comparative genomics with the chimpanzee, mouse, and pufferfish genomes—it has now become apparent that we no longer know what “a gene” is. In recent years, the definition of “a gene” had been: “A segment of DNA, including all regulatory sequences for that DNA, which encodes a functional product— whether it is a protein, small peptide, or one of the many classes of regulatory RNA.” Among the massive amounts of data discovered by ENCODE was the finding that regulatory sequences for a transcript on one chromosome can be located on a second and/or third chromosome. Moreover, genes that are located on chromosomes geographically next to the nuclear membrane are more likely to exhibit high expression levels than genes located on chromosomes further inside the nucleus. It has also become increasingly apparent that complex diseases reflect the contributions of hundreds of genes, wherein each gene usually contributes no more than 1% to the overall phenotype (1). Thus, the current proposed definition of “a gene” is: “A union of genomic sequences encoding a coherent set of potentially overlapping functional products.” This definition sidesteps all the complexities of regulation and transcription (2). The exploding fields of epigenomics and copy number variants (CNVs) have also thrown a monkey wrench into our hopes for simplifying personalized medicine and/or individualized drug therapy. Even monozygotic (“identical”) twins show differences in methylated DNA and carry varying numbers of CNVs, perhaps explaining why one monozygotic twin may develop, e.g., leukemia at age 55 while the other might never be affected by this disease. In summary, dozens of reasons exist to show that an “unequivocal genotype” or even an “unequivocal phenotype” is virtually impossible to achieve in current limited-size studies of human populations. This problem (of insufficiently stringent criteria) leads to a decrease in statistical power and, consequently, equivocal interpretation of most genotype-phenotype association studies. What is the best approach for the future? GWAS, using populations of one thousand or more per group, is a possible portion of the answer. It remains unclear, however, whether personalized medicine or individualized drug therapy will ever be achievable by means of DNA testing alone. It is therefore proposed that the future solution might lie in some combination of genomics, transcriptomics, proteomics, and especially metabonomics. This latter developing field, i.e., studying urinary metabolite profiles, seems very promising. If there is any drawback, it is because metabonomics is extremely sensitive: metabolite profiles can be altered easily by modest exercise, menstrual cycle, diet, and even use of an antibiotic that disturbs the intestinal flora. However, by examining and comparing urinary metabolites in the same patient—year after year, decade after decade— one should be able to capture not only changes that reflect epigenomics and differences in CNV number, but also any other forms of non-Mendelian traits caused by the environment (changes in life-style, occupation, diet, chronic disease, etc.). References 1. Nebert DW, Zhang G, Vesell ES. From human genetics and genomics to pharmacogenetics and pharmacogenomics: Past lessons, future directions. Drug Metab Rev 2008;40:187-224. 2. Gerstein MB, Bruce C, Rozowsky JS, Zheng D, Du J, Korbel JO, Emanuelsson O, Zhang ZD, Weissman S, Snyder M. What is a gene, post-ENCODE? History and updated definition. Genome Res 2007;17:669-81. Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 65 Invited Abstracts Plenary Session: Molecular Classification and Prognostic Markers Plenary Session Molecular Classification and Prognostic Markers PL03-01 Ovarian cancer classification: Lessons from morphology, molecules, and mice. Kathleen R. Cho. University of Michigan Medical School, Ann Arbor, MI. Ovarian carcinomas are a heterogeneous group of neoplasms traditionally sub-classified based on type and degree of differentiation as assessed by light microscopy. Although the current clinical management of ovarian carcinoma largely fails to take this heterogeneity into account, it is becoming evident that each major histological type has characteristic genetic defects that deregulate specific signaling pathways in the tumor cells. Moreover, within the most common histological types, the molecular pathogenesis of low-grade versus high-grade tumors appears to be largely distinct. For example, low-grade serous carcinomas (characterized by frequent mutations of KRAS and BRAF), likely arise from serous borderline tumors, which in turn develop from benign serous cystadenomas. This stepwise tumor progression in the low-grade pathway contrasts with the rapid progression pathway of high-grade serous carcinomas (characterized by chromosomal instability and frequent TP53 mutations), for which precursor lesions are not well recognized. High-grade serous carcinomas may arise from ovarian surface inclusions, peritoneal mesothelium, or the distal portion (fimbriae) of the fallopian tube. These carcinomas disseminate to pelvic and peritoneal organs early in their progression. Like their low-grade serous counterparts, low-grade endometrioid carcinomas are often associated with recognizable precursor lesions, such as endometriosis and/or benign endometrioid neoplasms. Low-grade endometrioid carcinomas frequently harbor mutations that constitutively activate PI3K/Pten and Wnt/β-catenin-dependent signaling. In contrast, high-grade endometrioid carcinomas typically lack PI3K/Pten and Wnt/βcatenin pathway defects and have frequent TP53 mutations. The global gene expression profiles of high-grade endometrioid and serous carcinomas are quite similar to each other and largely distinct from the profiles of low-grade endometrioid, clear cell, and mucinous carcinomas. Genetically engineered mouse models of each major subtype of ovarian cancer will undoubtedly prove useful for improving knowledge of ovarian cancer biology and for preclinical testing of novel therapeutic strategies. Significant progress has been made in developing histologic-type specific models of ovarian cancer. As noted above, defects in the PI3K/Pten and Wnt/β-cat signaling pathways often co-occur and likely cooperate in the development of some human endometrioid adenocarcinomas. Deregulation of these two pathways in the murine ovarian surface epithelium by conditional inactivation of the Pten and Apc tumor suppressor genes results in the formation of adenocarcinomas morphologically similar to human endometrioid adenocarcinomas with 100% penetrance, short latency, and rapid progression to metastatic disease. The biological behavior and gene expression patterns of the murine cancers resemble those of human ovarian cancers with comparable signaling pathway defects. Studies are underway to further develop the model as a tool for pre-clinical testing of therapies that target deregulated Wnt/β-catenin and PI3K/Pten signaling, which may ultimately lead to better clinical outcomes for women with ovarian cancer. PL03-02 Pharmacogenetics of lung cancer: An integrative epidemiologic approach. Margaret R. Spitz, Xifeng Wu. UT M. D. Anderson Cancer Center, Houston, TX. Overview: Treatments for lung cancer have improved survival only marginally over the past few decades. Patients with non-small cell lung cancer (NSCLC) are commonly treated with platinum-based chemoradiotherapy with response rates being unpredictable, even controlling for the known prognostic factors of stage, histology, grade and performance status. Therefore, the ability to predict therapeutic response in these patients is of immense clinical benefit. Currently only clinical variables are used to guide treatment decisions with modest ability to predict overall survival. Molecular signatures derived from global gene expression profiling have shown promise in predicting clinical outcome, as have pathway-based or genome-wide identification of somatic aberrations using high-density comparative genomic hybridization in tumor tissues. Integrative epidemiology. The annotation of the human genome provides an opportunity to explore the impact of germline genetic variation in determining survival differences in non-small cell lung cancer. Evaluation of genetic variants such as single nucleotide polymorphisms (SNPs) is being recognized as an alternative and complementary approach to tumor molecular profiling and is producing promising results. A unifying premise to applying a molecular epidemiologic approach to outcome prediction is the concept of Integrative Epidemiology, i.e., that the same genes that are implicated in cancer risk, are also involved in the prediction of patient outcome. This implies that risk prediction and prediction of outcome are part of a continuum. Therefore, by analyzing the genetic make-up of both the tumor and the patient, we can generate a “personalized molecular medicine” profile, or pharmacogenetic profile that can be used to individualize therapy and help understand the functional consequences of chemoprevention, chemotherapy, or radiotherapy response and toxicity. This approach will also enable clinicians to select drugs likely to be most effective and least toxic for each patient. This is the promise and potential of pharmacogenetics. However, it is unlikely that a single polymorphism in a single gene would have a strong effect on treatment response. We therefore need to move beyond the candidate gene approach to a pathway-based polygenic approach, in order to identify clinically relevant combination of genetic markers. In this way, we can amplify the effects of individual polymorphisms and enhance the predictive power of the genetic variants. In this presentation we provide two examples of employing a pathway based genotyping approach, to assess the combined effects of a panel of polymorphisms that interact in the same pathway, rather than focusing on the candidate gene approach. DNA repair. DNA repair may be considered a double-edged sword. Nucleotide excision repair (NER) is the primary DNA repair pathway responsible for the removal of tobacco-related polycyclic aromatic hydrocarbon-induced DNA adducts. Suboptimal DNA repair capacity, as assessed by the in vitro host cell reactivation assay, is an independent risk factor for cancer. Likewise, cisplatin and other platinum agents bind preferentially to DNA, and the level of tissue platinum-DNA adducts is correlated with clinical outcome. Resistance to platinum agents has been linked to enhanced tolerance and to efficient capacity to repair the resultant DNA damage. On the other hand, reduced capacity for nucleotide excision repair has been shown to be associated with improved response to chemotherapy (due to decreased removal of tissue platinum adducts). As proof of principle and to gain further insight into the prognostic value of genetic polymorphisms in cisplatin response, we analyzed the associations of 12 functional polymorphisms in 7 key genes involved in NER in 229 patients with advanced NSCLC patients receiving first-line cisplatin-based chemotherapy. Joint analysis was performed to evaluate the cumulative effects of specific pathways. Survival tree analysis was conducted to identify patient subsets with distinctive survival patterns. Several biologically plausible main effects were identified in individual analysis. Specifically, patients carrying the RAD23 and XPA_23 variant-containing genotypes exhibited significantly poorer survival, while the XPD_751 and XPC_PAT and ERCC1 genotypes were associated with improved survival. We also assumed an additive model and summed the unfavorable genotypes (based on hazard ratios from the main effect analysis that had P values <0.1). We found a significant gene dosage effect for increasing hazard of death associated with increasing numbers of potential high-risk alleles in the nucleotide repair pathway combined. Likewise, there was a significant trend of reduced risk of death with decreasing number of putative unfavorable genotypes (P for trend <0.001 and log-rank p<0.001). The median survival times were 10.9, 12.9 and 26.9 months for patients with ≥5, 3 to 4, and <3 unfavorable genotypes respectively (log-rank p<0.001). 66 American Association for Cancer Research Plenary Session: Molecular Classification and Prognostic Markers Invited Abstracts Survival tree analysis revealed potential higher-order gene-gene interactions and categorized subgroups with dramatically different survival experiences, based on distinct genotype profiles. We also constructed a prediction hazard model. The area under the curve (AUC) increased from 0.71 (using clinical variables only) to 0.84 (incorporating clinical, epidemiological, and genetic variations). These results highlight the potential of taking a pathway-based approach and using survival tree analytic approaches to identify subgroups of patients with distinctly differing outcomes. Inflammation pathway. Another proof of principle example lies in the inflammation pathway. A sustained inflammatory response that is either not restrained or is aberrantly regulated, creates a microenvironment abundant in inflammatory cells, growth factors, activated stroma, enhanced angiogenesis, and oxidative stress, conditions that predispose to tumorigenesis. Likewise interaction of lung cancer cells with macrophages promotes both invasiveness and matrix-degrading activity. Therefore variants in the inflammation pathway could be markers both of risk and outcome. We have evaluated a panel of 59 single nucleotide polymorphisms (SNPs) in 37 inflammation-related genes in both risk of lung cancer and in outcome in patients with advanced stages of NSCLC. The interleukin 1B (IL1B) C3954T variant genotype was significantly associated with lung cancer risk, after controlling for false positives. This association was stronger in heavy smokers and in patients with emphysema, suggesting that, in the setting of heavy or prolonged exposure to tobacco or other agents, an aberrant inflammatory response may promote lung damage, eventually leading to lung cancer. Likewise, we have evaluated the same panel of genes as prognostic factors, and as predictive of radiation toxicity. Specifically we noted that the risk of radiation -related pneumonitis was 3.8 for patients with one or two adverse inflammation genotypes relative to those with no adverse genoytpes, and 9.5 for those with 3 or more adverse genotypes, with a significant gene dosage effect. Genome wide association studies. Sequencing of the entire human genome, the mapping of common haplotypes of single-nucleotide polymorphisms (SNPs), and cost-effective genotyping technologies have all made genome-wide association studies feasible. Most genome-wide association studies (GWAS) use dense maps of germline genetic variants such as single nucleotide polymorphisms (SNPs) that cover the whole genome to identify genetic markers of risk in a hypothesis generating mode. While the case-control approach is the study design of choice for such association studies, we propose a complementary approach to exploit the depth and diversity of data derived from genome-wide association (GWA) studies for pharmacogenetic research. Specifically we envision a case series design that follows clinically homogenous subsets of patients enrolled in GWA studies longitudinally for data on response to therapy, toxicity, and survival. It is recognized that identifying such patients receiving similar therapy and follow-up is a challenge and the usable sample size will decrease, but the GWAs have large numbers of case from which to select these usable subsets. The objective is to construct comprehensive prognostic/prediction profiles of epidemiologic, clinical and genetic data to improve patient outcomes. These data can also be used for replication of findings from other case series analyses. In the longer-term, complementary technologies, such as tissue arrays and tissue lysate arrays from patient tumors, as well as functional genomics to demonstrate the functional effects of the relevant SNP’s (siRNA, miRNA and chemical genomics screens) will help to validate which molecular markers or groups of molecular markers best predict risk, early diagnosis, and response to therapy. Consortial studies can be combined for pooled analyses, and studies across different ethnic groups can provide valuable LD information in the regions of interest. The need for collaborative team science is critical. Supported by NCI grants CA55769, CA127219, CA70907, CA111646, Flight Attendant Medical Research Institute Plenary Session Molecular Response Prediction Markers PL04-01 Molecular markers predictive of response to EGFR inhibitors. Bruce E. Johnson, Pasi Janne. Dana-Farber Cancer Institute, Boston, MA. Four years have passed since mutations of the tyrosine kinase domain of the epidermal growth factor receptor (EGFR) were discovered in patients with lung cancer who had dramatic clinical responses to treatment with the epidermal growth factor receptor tyrosine kinase inhibitors (EGFRTKIs), gefitinib and erlotinib. Additional laboratory and clinical studies have provided further insights into the biological impact of EGFR mutations. The other common genomic changes that arise in lung cancer that have an impact on EGFR-TKI sensitivity include KRAS and PTEN mutations, secondary T790M mutations in EGFR, and MET amplification. The retrospective and prospective studies have shown that EGFR mutations are closely associated with an increased response rate, prolongation in time to progression and survival compared to those with a wild type EGFR. Patients with EGFR mutations treated with gefitinib or erlotinib have a response rate of approximately 80%, a median time to progression in excess of approximately one year, and a median survival in excess of two years. The prospective randomized studies of patients with non-small cell lung cancer and sensitizing mutations of EGFR treated with chemotherapy versus erlotinib or gefitinib are ongoing outside of the United States in Europe and Japan. The genomic changes associated with resistance to treatment with gefitinib and erlotinib are a DNA mutation which changes the threonine to methionine at the 790th amino acid of EGFR known as the (T790M) mutation as well as amplification of the MET oncogene. The T790M mutation in EGFR is responsible for approximately half of the acquired resistance while MET amplification is responsible for about 30%. Irreversible inhibitors including HKI-272 and PF-299804 can cause growth inhibition in a non-small cell lung cancer with both the resistance and sensitizing mutations, while gefitinib and erlotinib do not. HKI-272 has been tested in a randomized Phase II trial and PF-299804 has completed Phase I testing and the Phase II trials have started. The Phase II trial of HKI-272 enrolled patients previously treated with gefitinib and erlotinib and EGFR mutation testing was prospectively incorporated into the trials. The results of these trials will be reported in 2009. The Phase I trial of PF299804 has been completed and the Phase II trials are beginning. Patients with non-small cell lung cancer treated with chemotherapy plus either erlotinib or gefitinib have had partial responses to subsequent treatment with PF-299804. Preclinical work has documented the non-small cell lung cancer cell line, HCC827, can be made to be resistant to gefitinib and is referred to as HCC827 GR. The mechanism of resistance is caused MET amplification. The in vitro and in vivo experiment have shown that joint inhibition of MET and EGFR with gefitinib plus a MET inhibitor can slow the growth of the HCC827 GR, the lung cancer cell line that developed resistance to gefitinib. Clinical trials with an EGFR inhibitor, erlotinib, with MET inhibitors are to be undertaken. Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 67 Invited Abstracts Plenary Session: Molecular Response Prediction Markers PL04-02 Targeting the hedgehog pathway: From bench to clinic. Frederic J. de Sauvage. Genentech, Inc., South San Francisco, CA. The Hedgehog (Hh) pathway is an ancient signaling cascade that directs patterning in most animals and is crucial for proper development. While Hh signaling is very active during most stages in embryogenesis, it remains relatively quiet in adult life. However, aberrant reactivation of the pathway in adult tissue can lead to the development of cancer. Hh pathway activation in tumors such as basal cell carcinoma (BCC) and medulloblastoma is the result of inactivating mutations in PATCHED (PTCH) or activating SMOOTHENED (SMO) mutations. Furthermore, in other solid tumors, such as colon and pancreas, Hh ligand expression is upregulated in tumor cells and acts in a paracrine manner to activate the pathway in the surrounding tumor stroma. A consequence of Hh pathway activation is the transcriptional upregulation of unique target genes and the expression levels of these transcripts that could serve as potential pharmacodynamic and predictive biomarkers of pathway activity. Targeting the Hh pathway with small molecule antagonists therefore provides new therapeutic opportunities for the treatment of both tumor types. GDC0449, a systemic Hedgehog (Hh) pathway antagonist, was tested in a firstin-human, first-in class, Phase I study with locally advanced, multifocal or metastatic BCC patients. Antitumor activity was observed in BCC patients enrolled, thereby confirming the importance of inhibiting aberrant Hh signaling in tumors with mutations in the Hh pathway. Plenary Session Pharmacodynamics of Markers PL06-01 Celecoxib for prevention of sporadic colorectal adenomas: Patient selection to optimize efficacy and safety. Monica Marie Bertagnolli. Brigham and Women’s Hospital, Boston, MA. Celecoxib is a selective cyclooxygenase-2 inhibitor used for treatment of severe arthritis and for reduction of colorectal adenoma burden in patients with familial adenomatous polyposis. Recent randomized trials also showed that celecoxib reduced formation of sporadic adenomas in patients at risk for colorectal cancer. The Adenoma Prevention with Celecoxib (APC) Trial randomized 2035 patients to receive either placebo, celecoxib 200 mg bid or celecoxib 400 mg bid for 3 years, with a total observation period of 5 years following randomization. A dose-dependent decrease in sporadic adenoma incidence was observed over the 3-year treatment interval. Celecoxib treatment was also associated with an increased risk of serious cardiovascular events. The APC Trial prospectively collected data on patient family history, comorbid conditions, and concomitant medication use. In addition, participants provided blood and tissue specimens for analyses of factors influencing treatment response and toxicity. We hypothesized (prospectively) that celecoxib efficacy would be influenced by patient medical history, with greatest efficacy observed in those at high cancer risk due to the presence of advanced adenomas at baseline. We also predicted (retrospectively) that patients with a baseline history of cardiovascular disease would be more likely to experience cardiovascular adverse events while taking celecoxib. Genetic variants in the cytochrome p450 2C9 (CYP2C9) enzyme, particularly the I359L allele, are associated with impaired metabolism of celecoxib. Because both efficacy and toxicity of celecoxib were associated with dose in the APC Trial, we hypothesized that these outcomes would also be influenced by CYP2C9 genotype. Specifically, we predicted that APC Trial patients bearing the I359L allele would demonstrate both greater reduction in adenoma incidence and greater risk for cardiovascular toxicity. Our results showed that treatment with celecoxib is associated with a significant reduction in advanced adenoma incidence during a 3 year treatment interval, with efficacy that persisted over a 5 year total observation period. We also found that dose-related benefits and cardiovascular toxicity of celecoxib appear to be modified by genetic variation in its metabolism. References 1. Arber N, Kuwada S, Leshno M, Sjodahl R, Hultcrantz R, Rex D. Sporadic adenomatous polyp regression with exisulind is effective but toxic: A randomized, double-blind, placebo controlled, dose-response study. Gut 2006;55:367-73. 2. Bertagnolli MM, Eagle CJ, Zauber AG, et al. A randomized trial of celecoxib to prevent sporadic colorectal adenomas. N Engl J Med 2006;355:873-84. 3. Solomon SD, McMurray JJ, Pfeffer MA, et al. Cardiovascular risk associated with celecoxib in a clinical trial for colorectal adenoma prevention. N Engl J Med 2005;352:1071-80. Plenary Session Molecular Markers and Cancer Stem Cells PL05-01 Gene signature of cancer stem cells is manifested within an intrinsic subgroup of breast cancers with mesenchymal properties. Melissa Landis,1 Chad J. Creighton,1 Michael T. Lewis,1 Xiaoxian Li,1 Helen Wong,1 Anna Tsimelzon,1 Jason Herschko,2 Cheng Fan,2 Xiaping He,2 C. Kent Osborne,1 Anne Pavlick,1 J. Michael Dixon,3 Susan G. Hilsenbeck,1 Charles M. Perou,2 Jeffrey Rosen,1 Jenny Chang1. 1Baylor College of Medicine, Houston, TX; 2Lineberger Comprehensive Cancer Center, Chapel Hill, NC; 3Western General Hospital, Edinburgh, United Kingdom. Background: Breast cancer stem cells characterized by CD44+/CD24-/low may be resistant to therapy and responsible for relapse. Mmammospheres (MSs) can be propogated as an in vitro surrogate assay for increased self-renewal. We set out to define a “signature” expression pattern associated with CD44+/CD24-/low, mammosphere-forming cells. Methods: Breast cancer biopsies (n=19) were digested, stained with CD24, CD44, and lineage antibodies, and analyzed by flow cytometry. A portion of the unsorted cells were plated under serum-free conditions to form MSs (n=16). Gene expression, using the Affymetrix U133 GeneChip platform, of cancer cells bearing CD44+/CD24-/low vs. other sorted cells, and between cancer MS vs. the primary invasive cancers were analyzed. Gene expression from two trials (neoadjuvant letrozole N=18, and docetaxel, N=12) were used as validation studies. Results: In the flow-sorted CD44+/CD24-/low vs. other cells, 1,424 named genes were elevated (p<0.01, fold change>1.5, FDR=0.20). The comparison between MSs vs. primary cancers yielded 1,890 elevated genes (FDR=0.25). Between the two sets, 380 genes were in common, a highly significant overlap (p=1E-5, one-sided Fisher’s exact). This overlap was ~40% greater than what would be expected (n=110) if the two sets had no biological relevance. Differential pathways included genes in PI3K/AKT signaling (PI3K3R3, ErbB3, FGFR2, PRLR), and the Notch pathway - a known regulator of normal and malignant stem/progenitor cells (Jagged-2, MAML2, Deltex, HES1). This signature was found exclusively activated in tumors of the recently identified “claudin-low” subtype, characterized by overexpression of many mesenchymal genes. Both signatures were validated in two independent data sets comparing the expression profiles of paired breast cancer core biopsies before vs. after letrozole or docetaxel chemotherapy. Conclusion: The mesenchymal association provides a possible explanation for the intrinsic resistance of breast cancer stem cells to therapy. 68 American Association for Cancer Research Plenary Session: Noninvasive Markers Invited Abstracts Plenary Session Noninvasive Markers PL07-01 Inference engines for phospho-signaling networks, disease mechanism, and clinical outcome at the single cell level. Garry P. Nolan. Stanford School of Medicine, Stanford, CA. Intracellular assays of signaling systems has been limited by an inability to correlate functional subsets of cells in complex populations based on active kinase states or other nodal signaling junctions. Such correlations could be important to distinguish changes in signaling status that arise in rare cell subsets during functional activation or in disease manifestation. We have demonstrated the ability to simultaneously detect activated kinases and phosphoproteins in simultaneous pathways in subpopulations of complex cell populations by multi-parameter flow cytometric analysis. We have applied this technology to the study of normal human cell populations as well as several disease states including Acute Myelogenous Leukemia, and Follicular Lymphoma, colon cancer and infiltrating immune cells of cancers among others. The tremendous amounts of correlated data generated via phospho-flow allows amalgamation that automates signaling network determination using Bayesian analysis (and a unique computational approach using a new electronic architecture for a “statistics supercomputer”). Bringing these together, we have initiated the generation of a comprehensive network topology map of signaling in all primary immune subsets. Thus, we are bringing single cell analysis of multiple kinase pathways together with novel computational and electronic approaches to increase the depth of drug screening to directly reach screening in primary cells. PL07-03 Targeting cell signaling in lung cancer: Opportunities for novel drug development. Fadlo R. Khuri, Haian Fu, Suresh Ramalingam, and Shi-Yong Sun, Departments of Hematology and Medical Oncology and Pharmacology, Winship Cancer Institute, Emory University, Atlanta, GA Lung cancer is by far the leading cause of cancer-related death worldwide, with over 1.6 million lung cancer cases expected in 2007, and over 1.3 million deaths (1). Renewed efforts in drug development have led to the approval of eight agents (vinorelbine, paclitaxel, gemcitabine, docetaxel, erlotinib, bevacizumab, pemetrexed, topotecan) in the last decade alone for the treatment of lung cancer. Median survival for stage IV disease has improved from four months with best supportive care to over one year with chemotherapy plus bevacizumab (2), an antiangiogenic agent. Thus, novel opportunities have increasingly presented themselves for the development of new agents, a significant impediment to which has been the better understanding of these novel agents and the signaling pathways that they potentially disrupt. Further development of novel agents thus necessitates a greater understanding of signaling pathways in lung cancer, the ability to temporally measure changes in these pathways and the pathways that mediate sensitivity and resistance to novel agents. Insight into carcinogenesis has led to evidence that cancer cells become increasingly dependent on selected signaling pathways. Some pathways that lung cancer cells become dependent on are those that signal through the epidermal growth factor receptor (EGFR), and the insulin-like growth factor receptor (IGFR) pathway; upregulation of these pathways is apparent in a significant majority of lung cancers (3). Increasing evidence suggests that blockade of the EGFR pathway by small molecule tyrosine kinase inhibitors or by monoclonal antibodies can disrupt pathways that are important for lung cancer survival. For example, erlotinib is shown to prolong survival in patients previously treated with platinums and taxanes (4), and cetuximab has been shown to improve survival when combined with cytotoxic platinum-based chemotherapy in the front-line chemotherapy setting (5, 6). While the improvements in most cases have been relatively modest, it may well be that better temporal understanding of which pathways are blocked and the extent of the blockade can be enhanced by improvements in functional imaging. Thus, ongoing studies seek to assess whether disruption of critical signaling pathways that enable tumor progression can be accomplished and be assayed by either FDG-PET imaging or functional MRI. Another important therapeutic approach is that targeting angiogenesis. Bevacizumab, a potent monoclonal antibody specific for vascular endothelial growth factor receptor (VEGFR) has been shown to enhance the efficacy of cytotoxic chemotherapy against lung cancer. In a phase III trial in which this potent agent was combined with carboplatin and paclitaxel, survival was improved when bevacizumab was added to chemotherapy in a population of patients with advanced non-small cell lung cancer (2). Critically, the percentage of patients who had major clinical responses was more than doubled in the arm that added the monoclonal antibody. To understand the reasons behind the efficacy of this combination, investigators have utilized functional imaging, which has shown that in patients who received the combination of chemotherapy with an anti-vascular agent, significant necroses of the primary tumor was seen on the CT scan without overt tumor shrinkage early on in the process. This occurs as the tumor’s neovasculature supports growth whereby disrupting tumor blood flow, a process that is best captured on dynamic contrast MRI. Here, even subtle changes in tumor vasculature can be demonstrated, with the more impressive disruptions tending to correlate with more impressive response rates. Whether these changes end up correlating with enhanced overall survival is yet to be seen. There is growing evidence that some of the most critical pathways that lung cancers are “addicted” to are those relevant to the Akt/mTOR axis, a critical pathway that supports cancer survival and inhibits those that lead to cell death. Findings from several groups have shown that several pathways are often upregulated in lung cancer, and a phase III trial of high risk renal cell cancer patients treated with mTOR inhibitors or rapamycin analogs meaningfully improved survival in patients with poor prognosis renal cancer (7, 8). Our group was the first to show that treatment with rapamycin or its analogs leads to a transient and unexpected upregulation of phospho-Akt, thereby increasing the cancer cell’s dependence on this pathway (9). While several groups have shown that rapamycin analogs, or rapalogs, have activity in lung cancer and neuroendocrine tumors as well as in renal cell cancer, our continued studies suggest that initial treatment with rapamycin analogs enhances dependence on the PI3-kinase/Akt/mTOR pathway, therefore making lung tumors more susceptible to a second hit along this pathway. Our data have suggested that second hits with Akt inhibitors lead to accelerated cell death in tumors after initial treatment with rapamycin. Furthermore, a phase Ib trial conducted by our group has shown initially that those patients able to take three weeks of preoperative everolimus or RAD001 have moderate downregulation of FDG-PET uptake, and work continues to assess whether this correlates with modulation of Akt. In conclusion, the development of novel agents targeting cell signaling in lung cancer proceeds at a brisk pace. It is critical that functional imaging be accurately deployed to better understand temporal changes that are accelerated by the tumor, and those mediated by prior therapy, so that we can intervene in a temporally precise fashion. References 1. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T, Thun MJ. Cancer statistics, 2008. CA Cancer J Clin. 2008;58(2):71-96. 2. Sandler A, Gray R, Perry MC, Brahmer J, Schiller JH, Dowlati A, Lilenbaum R, Johnson DH. Paclitaxel-carboplatin alone or with bevacizumab for nonsmall-cell lung cancer. N Engl J Med. 2006;355(24):2542-50. 3. Hirsch FR, Varella-Garcia M, Bunn PA Jr, Di Maria MV, Veve R, Bremmes RM, Barón AE, Zeng C, Franklin WA. Epidermal growth factor receptor in non-small-cell lung carcinomas: correlation between gene copy number and protein expression and impact on prognosis. J Clin Oncol. 2003;21(20):3798-807. Molecular Diagnostics in Cancer Therapeutic Development • September 22-25, 2008 • Philadelphia, PA 69 Invited Abstracts Plenary Session: Noninvasive Markers 4. Shepherd FA, Rodrigues Pereira J, Ciuleanu T, Tan EH, Hirsh V, Thongprasert S, Campos D, Maoleekoonpiroj S, Smylie M, Martins R, van Kooten M, Dediu M, Findlay B, Tu D, Johnston D, Bezjak A, Clark G, Santabárbara P, Seymour L; National Cancer Institute of Canada Clinical Trials Group. Erlotinib in previously treated non-small-cell lung cancer. N Engl J Med. 2005;353(2):123-32. 5. Rosell R, Robinet G, Szczesna A, Ramlau R, Constenla M, Mennecier BC, Pfeifer W, O’Byrne KJ, Welte T, Kolb R, Pirker R, Chemaissani A, Perol M, Ranson MR, Ellis PA, Pilz K, Reck M. Randomized phase II study of cetuximab plus cisplatin/vinorelbine compared with cisplatin/vinorelbine alone as first-line therapy in EGFR-expressing advanced non-small-cell lung cancer. Ann Oncol. 2008;19(2):362-9. 6. Pirker R, Szczesna A, von Pawel J, Krzakowski M, Ramlau R, Park K, Gatzemeier U, Bajeta E, Emig M, Pereira JR. FLEX: A randomized, multicenter, phase III study of cetuximab in combination with cisplatin/vinorelbine (CV) versus CV alone in the first-line treatment of patients with advanced non-small cell lung cancer (NSCLC). J Clin Oncol 26: 2008 (May 20 suppl; abstr 3) 7. Hudes G, Carducci M, Tomczak P, Dutcher J, Figlin R, Kapoor A, Staroslawska E, Sosman J, McDermott D, Bodrogi I, Kovacevic Z, Lesovoy V, Schmidt-Wolf IG, Barbarash O, Gokmen E, O’Toole T, Lustgarten S, Moore L, Motzer RJ; Global ARCC Trial. Temsirolimus, interferon alfa, or both for advanced renal-cell carcinoma. N Engl J Med. 2007 356(22):2271-81. 8. Motzer RJ, Escudier B, Oudard S, Hutson TE, Porta C, Bracarda S, Grünwald V, Thompson JA, Figlin RA, Hollaender N, Urbanowitz G, Berg WJ, Kay A, Lebwohl D, Ravaud A; RECORD-1 Study Group. Efficacy of everolimus in advanced renal cell carcinoma: a double-blind, randomised, placebo-controlled phase III trial. Lancet. 2008;372(9637):449-56. 9. Sun SY, Rosenberg LM, Wang X, Zhou Z, Yue P, Fu H, Khuri FR. Activation of Akt and eIF4E survival pathways by rapamycin-mediated mammalian target of rapamycin inhibition. Cancer Res. 2005;65(16):7052-8. 10. Wang X, Yue P, Chan CB, Ye K, Ueda T, Watanabe-Fukunaga R, Fukunaga R, Fu H, Khuri FR, Sun SY. Inhibition of Mammalian Target of Rapamycin Induces Phosphatidylinositol 3-Kinase-Dependent and MnkMediated Eukaryotic Translation Initiation Factor 4E Phosphorylation. Mol Cell Biol. 2007; 27(21):7405-13. The Centers for Medicare & Medicaid Services (CMS) is charged with the implementation of CLIA, including laboratory registration, fee collection, surveys, surveyor guidelines and training, enforcement, approvals of PT providers, accrediting organizations and exempt states. The Centers for Disease Control and Prevention (CDC) is responsible for the CLIA studies, convening the Clinical Laboratory Improvement Amendments Committee (CLIAC) and providing scientific and technical support/consultation to DHHS/CMS. The Food and Drug Administration (FDA) is responsible for test categorization. To enroll in the CLIA program, laboratories must first register by completing an application, pay fees, be surveyed, if applicable, and become certified. CLIA fees are based on the certificate requested by the laboratory (that is, waived, PPM, accreditation, or compliance) and, for moderate and high complexity laboratories, the annual volume and types of testing performed. Waived and PPM laboratories may apply directly for their certificate as they aren’t subject to routine inspections. Those laboratories which must be surveyed routinely; i.e., those performing moderate and/or high complexity testing, can choose whether they wish to be surveyed by CMS or by a private accrediting organization. The CMS survey process is outcome oriented and utilizes a quality assurance focus and an educational approach to assess compliance. PL08-02 Medicare coverage of genetic testing. Jeffrey Roche. Center for Medicare and Medicaid Services, Baltimore, MD. The process leading to a national coverage determination for Medicare at the Center for Medicare and Medicaid Services (CMS) is outlined. In particular, the use of reasonable and necessary criteria for Medicare coverage (as required by the Social Security Act) is explained. Specifically, coverage for molecular and genomic testing at the local Medicare contractor level is distinguished from national coverage decisions in terms of process and scope. Examples of coverage decision thinking are provided. Some of the preferred features of evidence to support a coverage determination are explained. The overarching principle of any Medicare coverage determination—to enhance health outcomes for Medicare beneficiaries—is illustrated by an example in progress at CMS. References for obtaining further information about the coverage determination process at CMS are provided. Plenary Session Regulatory Issues and Science Policy PL08-01 Test method validation requirements under CLIA. Penny Keller. Centers for Medicare and Medicaid Services, Baltimore, MD. CLIA Program. Congress passed the Clinical Laboratory Improvement Amendments (CLIA) in 1988 establishing quality standards for all laboratory testing to ensure the accuracy, reliability and timeliness of patient test results regardless of where the test was performed. A laboratory is defined as any facility which performs laboratory testing on specimens derived from humans for the purpose of providing information for the diagnosis, prevention, treatment of disease, or impairment of, or assessment of health. CLIA is user fee funded; therefore, all costs of administering the program must be covered by the regulated facilities, including certificate and survey costs. The final CLIA regulations were published on February 28, 1992 and are based on the complexity of the test method; thus, the more complicated the test, the more stringent the requirements. Three categories of tests have been established: waived complexity, moderate complexity, including the subcategory of provider-performed microscopy (PPM), and high complexity. CLIA specifies quality standards for proficiency testing (PT), patient test management, quality control, personnel qualifications and quality assurance for laboratories performing moderate and/or high complexity tests. Waived laboratories must enroll in CLIA, pay the applicable fee and follow manufacturers’ instructions. Because problems in cytology laboratories were the impetus for CLIA, there are also specific cytology requirements. 70 American Association for Cancer Research