Search

Quantitative metrics to objectively assess the fidelity of cancer models, such as cell lines, organoids, or patient-derived xenografts (PDXs), remain elusive, with histological criteria or the presence of specific mutations often used as driving principles. However, focusing on individual mutations inevitably ignores the effect of a large complement of additional, model-specific genetic and epigenetic events. As a result, effective model fidelity assessment is best performed a posteriori, for instance, by determining whether dependencies identified in a specific tumor model (e.g., a cell line or organoid) are recapitulated in vivo in PDXs or in patients, via pre-clinical or clinical trials. Unfortunately, such an approach is inefficient and time-consuming, creating an urgent need for methodologies capable of effectively and quantitatively assessing model fidelity a priori. This unmet need motivated us to develop and test a quantitative, molecular-level framework (OncoMatch), to assess the fidelity of a given tumor model in the context of a specific biological question, and in particular for addressing issues of drug sensitivity. We addressed this challenge by integrating two independent computational metrics to assess: (a) conservation of regulatory networks inferred from patient-derived samples in a model of interest, and; (b) overlap of master regulator (MR) proteins–i.e., proteins representing the mechanistic determinants of the transcriptional state associated with the phenotype of interest–as inferred from patient-derived and model-derived samples. We show that these molecular-level criteria can effectively identify cell lines that recapitulate patient-specific drug mechanism of action and drug sensitivity, independent of histological consideration. By leveraging gene expression profiles of drug-perturbations in primary cells and explants derived from gastroenteropancreatic-neuroendocrine tumor (GEP-NET) patients, we show that H-STS, an EBV-immortalized lymphoblastoid cell line, represents a high-fidelity model for the assessment of drug mechanism of action and drug sensitivity in these tumors. In particular, our analysis shows highly significant conservation of drug mechanism of action for 60 of 95 profiled drugs (63%, p < 10-10, Bonferroni's corrected), and higher conservation among drugs exhibiting greater bioactivity in this context. This rate is comparable and, in fact higher than what is achieved using tumor-type-matched cell line pairs representative of glioma, pancreatic and prostate carcinoma, and dramatically higher than the conservation observed between unrelated models, which we used as negative controls. Based on this systematic analysis, OncoMatch represents a valuable addition to our repertoire of tools for prioritizing cell lines, organoids and patient-derived xenograft models as high-fidelity human cancer models. We provide comprehensive prioritization of 921 cell lines as potential high-fidelity models for 10,024 human tumor samples in TCGA. This represents an actionable resource to guide selection of cell line models for specific drug mechanism of action and drug sensitivity studies. Citation Format: Mariano J. Alvarez, Yan Pengrong, Mary L. Alpaugh, Michaela Bowden, Ewa T. Sicinska, Chensheng W. Zhou, Charles Karan, Ronald B. Realubit, Prabhjot S. Mundi, Adina Grunn, Jager Dirk, John A. Chabot, Antonio T. Fojo, Paul E. Oberstein, Hanina Hibshoosh, Jeffrey W. Milsom, Matthew H. Kulke, Massimo Loda, Gabriela Chiosis, Diane L. Reidy-Lagunes, Andrea Califano. OncoMatch: Unbiased, quantitative assessment of cancer model fidelity for drug sensitivity and mechanism of action elucidation [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr LB-003.

The role of MSI in colorectal cancer (CRC) is well characterized, and colon tumors are classified as MSI-High or MSI-Stable by screening specific microsatellites. MSI-H compared to MSI-S is predictive for therapy outcome in chemotherapy and immunotherapy and has been associated with distinct characteristics and favorable results including better prognosis, a higher 5-year survival, and lesser metastasis. Several other tumors also present MSI changes reflecting mismatch repair deficiency. While tumor testing is the gold standard, a convenient approach to screen for MSI before and during cancer treatment is screening circulating DNA (liquid biopsy) using a blood draw, thereby interrogating ‘systemic' MSI reflecting primary or secondary (occult) tumor status at the time of blood collection. However, presence of excess unaltered, wild-type DNA often masks alterations such as MSI. Using capillary electrophoresis or next generation sequencing for MSI detection presents challenges at low levels of MSI due to polymerase slippage (‘stutter') that generate high false positive rates at positions of homo-polymers. We present a new approach for enrichment of altered micro-satellites prior to DNA-amplification thereby facilitating their detection. We recently developed nuclease-assisted minor-allele enrichment with probe-overlap, a single-step approach that removes WT-DNA and enriches mutation-containing alleles. Here we adapted NaME-PrO for detection of homopolymer indels, for MSI detection. The method employs a double-strand-DNA-specific nuclease and overlapping oligonucleotide-probes interrogating multiple micro-satellite targets. Following DNA denaturation, the probes form double-stranded regions with their targets, thereby guiding nuclease digestion to selected sites. Microsatellite indels create ‘bulges' that inhibit digestion, thus subsequent amplification yields DNA with microsatellite alterations enhanced at multiple targets. The assay is applied at the genomic or circulating-DNA level prior to amplification, thereby avoiding polymerase-introduced ‘stutter peaks' arising from WT DNA. Inclusion of organic solvents allows homogeneous application of the method in closed tube reactions. We validated the method by evaluating 5 MSI targets simultaneously, NR27, NR21, NR24, BAT25 and BAT26 using DNA from tumor biopsies and circulating-DNA from colorectal cancer patients. The technique enriched all altered targets and detected microsatellite alterations down to 0.01% altered allele frequency, thus improving detection sensitivity by >100-fold relative to current approaches. We anticipate application of this highly- multiplex-able method either with standard 5-plex reactions in conjunction with capillary electrophoresis or with NGS-based detection of MSI on thousands of targets to enable sensitive detection in tumors and liquid biopsies. Citation Format: Ioannis Ladas, Chen Song, Fangyan Yu, Ka Wai Leong, Katerina Troullinou, Matthew Kulke, Harvey Mamon, Mike G. Makrigiorgos. Sensitive detection of microsatellite instability (MSI) in tumors and liquid biopsies using nuclease-based enrichment [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 942.

Pancreatic ductal adenocarcinoma (PDAC) is currently the fourth-leading cause of cancer-related death in the United States and is projected to become the second leading cause by 2030. Most patients present with advanced disease and die within 12 months of diagnosis. Recent genomic studies of primary pancreatic cancer resection specimens have identified several molecular alterations and genomic subtypes of the disease that may guide precision medicine approaches to clinical management. However, the molecular landscape of metastatic PDAC has been less well characterized. Moreover, biopsy-driven studies in metastatic PDAC have been historically very challenging due to the aggressive course of this disease as well as the low-volume and heterogeneous nature of biopsies that makes deep molecular characterization difficult. Insufficient genomic analysis of a patient’s tumor early in their disease course is a major barrier to enrollment on clinical trials of targeted therapies. To address these limitations, we have implemented a multi-disciplinary clinical and research biopsy protocol to enable real time comprehensive molecular characterization of metastatic PDAC biopsy specimens. We have performed core needle biopsies of metastatic lesions in the liver or peritoneal cavity in 42 patients at the time of initial presentation. A low rate of complications was observed, with only a single patient having a self-limited hemorrhagic complication after liver biopsy. On average, 4-6 separate biopsy specimens were collected from each patient for histopathology and genomic analysis. Whole exome sequencing (WES) was performed in a CLIA-certified laboratory and a comprehensive molecular report of somatic alterations and selected pathogenic germline variants was returned to the referring clinician with a typical turn-around time of 3-5 weeks. We observed a striking incidence of recurrent germline and somatic alterations in DNA-damage repair genes, such as BRCA2, ATM and CHEK2. We also observed alterations in genes with known therapeutic implications, such as BRAF, RNF43, STK11 and ROS, and in select cases, these results guided choice of second or third line therapy. In parallel to WES, we performed RNA sequencing on bulk tumor tissue and readily identified expression signatures defining multiple subtypes of tumor and stroma that may have prognostic or therapeutic implications for tumor- or stroma-directed therapies. Collectively, these results demonstrate the feasibility and value of real-time genomic characterization of metastatic PDAC and provide a path forward for improved stratification and enrollment of PDAC patients on molecularly defined clinical trials. Citation Format: Andrew J. Aguirre, Scott Carter, Nicholas Camarda, Arezou Ghazani, Jonathan Nowak, Annacarolina Da Silva, Lauren Brais, Dorisanne Ragon, Devin McCabe, Lori Marini, Kristin Anderka, Karla Helvie, Nelly Oliver, Ana Babic, Paul Shyn, Douglas Rubinson, Anuj Patel, James Cleary, Nadine McCleary, Matthew Kulke, Thomas Clancy, Leona Doyle, Jason Hornick, Christine Ardito-Abraham, Ruth Yu, Michael Downes, Ronald Evans, Richard A. Moffitt, Jen Jen Yeh, William C. Hahn, Charles Fuchs, Robert Mayer, Nikhil Wagle, David Tuveson, Levi A. Garraway, Brian M. Wolpin. Real-time genomic characterization of metastatic pancreatic cancer to enable precision medicine [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3036. doi:10.1158/1538-7445.AM2017-3036

Pancreatic malignancies are the fourth leading cause of all cancer-related deaths of both men and women in the United States. A majority of patients with pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors (pNET) present with advanced disease due to a lack of specific symptoms. The high mortality rate associated with pancreatic cancer can be attributed to both a lack of clinical diagnostic tests for early detection as well as an inadequate understanding of the underlying molecular mechanisms of its aggressive pathogenesis. We have recently become interested in the role that ADAM12 (a disintegrin and metalloprotease 12), a member of the disintegrin metalloprotease family, may play in the development and progression of pancreatic cancer. We have found that ADAM12 protein levels are significantly upregulated in human pancreatic cancer and in mouse models of PDAC and correlate with disease progression. ADAM12 levels were also upregulated in precursor PanIN lesions in a transgenic mouse model. ADAM12 transcript and protein expression is higher in poorly differentiated/quasi-mesenchymal pancreatic cancer cell lines such as Panc1 and MiaPaca2 compared to well-differentiated/classical cell lines such as AspC1 and BxPC3. Downregulation of ADAM12 in pancreatic tumor cells resulted in reduced cell migration, invasion and proliferation whereas apoptotic rates were significantly higher. In agreement with these findings, activation of common signaling pathways including pEGFR, pSTAT3 and pErk were also downregulated in response to ADAM12 silencing in these cells. We have also determined whether ADAM12 could be detected in the urine of patients with pancreatic malignancies and whether ADAM12 levels might serve as an independent predictor of disease status. Retrospective analyses of urine samples (n=130) from PDAC and pNET patients as well as age- and sex-matched controls were conducted. Urinary ADAM12 levels were determined using a monospecific ELISA system. Multivariable logistic regression analyses indicated that, when controlling for age and sex, urinary ADAM12 could serve as significant independent predictor for distinguishing PDAC (P Citation Format: Roopali Roy, Adelle Dagher, David Zurakowski, Matthew Kulke, Marsha A. Moses.{Authors}. ADAM12 contributes to the malignant potential of pancreatic cancer and may serve as a non-invasive biomarker for its detection. [abstract]. In: Proceedings of the AACR Special Conference on Pancreatic Cancer: Advances in Science and Clinical Care; 2016 May 12-15; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2016;76(24 Suppl):Abstract nr A53.

Pancreatic malignancies are the fourth leading cause of all cancer-related deaths of both men and women in the United States. Each year, approximately 32,000 new patients are diagnosed with this disease and nearly the same number die from it. A majority of patients with pancreatic malignancies, including both pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors (pNET) present with advanced disease due to a lack of specific symptoms and current diagnostic limitations, making this disease extremely difficult to detect. The combination of poor prognosis and late presentation of pancreatic cancer patients highlights the urgent need for the development of effective, early detection strategies for this disease. Our laboratory has established a comprehensive biomarker discovery initiative whose objective is to identify proteins present in urine of cancer patients, to determine whether their presence might be relevant to disease status and stage and to validate their diagnostic and prognostic efficacy in large scale clinical trials. In the current study, our goal was to determine whether ADAM12 (a disintegrin and metalloprotease 12) could be detected in the urine of patients with pancreatic malignancies and whether ADAM12 levels might serve as an independent predictor of disease status. Retrospective analyses of urine samples (n=130) from PDAC and pNET patients as well as age- and sex-matched controls were conducted. Urinary ADAM12 levels were determined using a monospecific ELISA system. In addition, ADAM12 protein expression in tumor and normal pancreatic tissues was analyzed via immunohistochemistry (IHC). Multivariable logistic regression analyses indicated that, when controlling for age and sex, urinary ADAM12 could serve as significant independent predictor for distinguishing PDAC (P Citation Format: Roopali Roy, David Zurakowski, Matthew Kulke, Marsha A. Moses. Urinary ADAM12 levels detect the presence of pancreatic cancer. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 891. doi:10.1158/1538-7445.AM2014-891

A lack of cell lines and animal models has been an obstacle in developing novel therapeutic agents for advanced carcinoid tumors. In the past two years we have developed a collection of short-term carcinoid cell lines. Carcinoid cells in culture grow slowly and are importantly attached to, much faster, growing stromal fibroblasts, referred to as CAF's (Carcinoid Associated Fibroblasts). This phenotype is conserved in vivo and closely mimics the status of these tumors in humans. CAF's are known to have pro-tumorigenic effects; are attractive targets for cancer therapy; as well as valuable clinical biomarkers of prognosis. With the availability of an in-house panel of primary cell lines and control cultured fibroblasts, teasing apart the tumor-fibroblast relationship is feasible through collection of cell culture supernatants and in vitro screening of secreted proteins. We present here our initial screening studies using three cell culture environments, namely normal human fibroblasts, carcinoid-associated fibroblasts (CAF's) and the tumor microenvironment co-culture of stromal cells and CAF's. We screened 55 secreted proteins with known carcinogenic associations by multiplexed bead-based (BioPlex, BioRad) analysis. We have identified a potential 5-marker CAF signature encompassing elevated secretion of IL-6 and MCP-1; and the suppression of IL-8 and FGF basic, coupled with VEGF expression. As part of the next phase we plan to utilize our collection of banked whole blood, serum, and plasma specimens comprising over 1000 cases of patients with neuroendocrine tumors, to validate our secreted protein signature in early diagnosis and / or assessing progression and prognosis in patients with neuroendocrine tumors based on stage and tumor subtype. Citation Format: Michaela Bowden, Ewa Sicinska, Matthew Kulke, Massimo Loda. Understanding the role of the carcinoid associated fibroblasts in the neuroendocrine tumor microenvironment. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 168. doi:10.1158/1538-7445.AM2014-168

Background: Neuroendocrine tumors (NET) have been shown to be responsive to treatment with VEGF pathway inhibitors. We hypothesized that genetic variation in VEGFA, VEGFR1 or VEGFR2 is associated with progression free survival (PFS) and/or overall survival (OS) in patients with metastatic NET treated with bevacizumab, a monoclonal antibody targeting VEGFA. Methods: We selected 61 functional and tagging SNPs with a minor allele frequency of 5% or greater, with full gene coverage of VEGFA, VEGFR1 and VEGFR2. We pooled data from two genotyped datasets, selecting 82 patients with metastatic NET who received bevacizumab administered either as a single agent or as part of a combination regimen. Using a Cox proportional hazards regression analysis, and an additive genetic model, we tested the SNPs’ associations with PFS and OS adjusting for age and stage at diagnosis, gender, tumor origin, grade of differentiation, and genotyping dataset. Results: We found that 4 intronic tagging SNPs in VEGFR1, rs7987649, rs9508021, rs9513095 and rs2104330 were associated with PFS at p Conclusions: In patients with neuroendocrine tumors treated with bevacizumab, the VEGFR1 SNP rs7987649 appears to be associated with progression-free survival. We note that three SNPs identified in our analysis, rs9508021, rs9513095 and rs2104330, are in strong linkage disequilibrium with SNPs rs7993418 and rs9582036, which have been associated with increased VEGFR1 expression and with PFS and OS in pancreatic adenocarcinoma (Lambrechts, 2012). Further study of loci within this region of VEGFR1 is warranted to discover potential causal variants affecting NET survival. Citation Format: Monica Ter-Minassian, ZhiRong R. Qian, Jennifer A. Chan, Susanne M. Hooshmand, Lauren K. Brais, Rachel Heafield, Xihong Lin, Geoffrey Liu, David C. Christiani, Matthew H. Kulke. Associations of genetic polymorphisms in VEGFR1 with progression-free and overall survival in patients with neuroendocrine tumors treated with the VEGFA inhibitor bevacizumab. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 2199. doi:10.1158/1538-7445.AM2013-2199

BACKGROUND: Serum chromogranin A (CgA) or alkaline phosphatase (ALP) are often assessed in patients with advanced neuroendocrine tumor (NET) patients but their prognostic significance continues to be debated. We evaluated whether elevated CgA or ALP levels were associated with shorter survival in a large, prospectively collected cohort of NET patients. METHODS: We identified patients with metastatic NET enrolled in an institutional database between 2003-10. We used the single measurement of serum CgA or ALP levels collected with shortest time after metastatic diagnosis and classified them as binary variables (elevated/normal) based on clinical laboratory specification of elevated value. Elevated biomarker levels were correlated with overall survival using the log-rank test. Hazard ratios for each biomarker were calculated using multivariate Cox regression, adjusting for baseline characteristics of age, gender, tumor histology, and tumor subtype. Each biomarker was modeled as a time-varying covariate to account for variability in time from initial metastatic diagnosis to time of biomarker test. RESULTS: We identified 526 patients who presented at our institution with metastatic NET. Of these, 349 had available CgA values and 350 had available ALP values. There were 146 and 153 deaths among those tested for CgA and ALP, respectively. The median survival time from time of metastatic diagnosis (yrs, (95%CI)) for patients with elevated CgA was 5.98 (5.04, 7.3), as compared to 13.27 (9.51, -) for those at normal CgA limits (p Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 3633. doi:1538-7445.AM2012-3633

The incidence of esophageal adenocarcinoma (EA) has increased approximately 500% in Western countries over the last four decades. The rapidly increasing incidence and sporadically development feature of EA suggest that gene-environment (G-E) interactions dominate the etiology. However, how G-E contributes to EA carcinogenesis is still poorly understood. We used a case-only approach to test the effect of G-E interactions in the occurrence of EA. 1330 functional and/or tagging SNPs selected from 354 cancer-associated genes were genotyped in 335 Caucasian EA cases. G-E (gastroesophageal reflux symptoms, GERD; BMI; and smoking) interactions were assessed by a two-step approach. First, we applied random forest (RF) to screen for important SNPs that had main effects as well as interaction effects. Second, we used case-only logistic regression (LR) models to estimate the G-E interaction effect, adjusting for covariates and false-discovery rate (FDR). RF analyses identified three sets of SNPs (94 SNPs-GERD, 38 SNPs-smoking, and 44 SNPs-BMI, respectively) that had the highest importance scores and lowest classification error rates. Further case-only LR analysis revealed that multiple interaction markers were significantly associated with EA risk: GERD*rs2237051 of EGF(OR = 1.39, P = 3.50E-07), GERD*rs2440 of XRCC5 (OR = 1,80, P = 0.0008), GERD*rs2237051*rs2440 (OR = 1.85, P = 0.004), smoking*rs10842514 of KRAS (OR = 0.85, P = 0.012), BMI*rs2305742 of IL12RB1 (OR = 0.84, P = 0.003), BMI*rs11568820 of VDR (OR = 0.97, P = 0.0005), BMI*rs11244142 of ABL1 (OR = 1.47, P = 2.09E-05), BMI*rs2800975 of XPA*rs743572 of CYP17A (OR = 3.62, P = 0.0009), and BMI*rs11244142*rs20547 of IL13 (OR = 1.89, P = 0.023). These results indicate that genetic variability may contribute to EA development through interactions with environmental factors. Our data also suggest that G-E interactions may be G- and E-specific in modulating the risk of EA. Supported by grants: NIH grants CA92824, CA74386, CA90578, and CA119650); Flight Attendant Medical Research Institute (FAMRI) grant 062459_YCSA; the Kevin Jackson Memorial Fund and Alan Brown Chair of Molecular Genomics. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 3742. doi:10.1158/1538-7445.AM2011-3742

BACKGROUND: Genetic risk factors for sporadic neuroendocrine tumors remain poorly understood. To identify genetic variants associated with neuroendocrine tumor risk, we first performed a large-scale assessment of common genetic variants in candidate cancer genes in a discovery set of cases and controls. We then sought to replicate these associations in an independent cohort of neuroendocrine tumor cases and controls. METHODS: We designed an Illumina GoldenGate custom 1536 loci panel on a BeadArray chip (San Diego, CA) containing tagging and functional single nucleotide polymorphisms (SNPs) in 355 genes implicated in neuroendocrine tumors or in broader, established cancer-related signaling pathways. Using this panel, we evaluated potential risk associations in a discovery set of 274 cases and 329 controls. We then evaluated the top associated SNPs in an independent replication set of 235 neuroendocrine tumor cases and 113 controls using Sequenom iPlex. Subjects were all Caucasian in both sets. We evaluated potential risk associations using multiple logistic regression, after adjusting for the significant confounders of age and gender with the dominant model and test for trend. RESULTS: Evaluation of 1536 SNPs in our discovery cohort revealed 18 SNPs with risk associations at a p-value CONCLUSION: Our findings suggest that genetic variation in IL12A, DAD1, and TSC2 is associated with neuroendocrine tumor risk as shown in two independent cohorts. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 929.

Esophageal adenocarcinoma (EA) is a complex malignancy that involves multiple genetic and environmental risk factors. Genetic polymorphisms in angiogenesis pathway, gastroesophageal reflux symptoms (reflux), higher body mass index (BMI), and tobacco smoking have been individually associated with EA development. However, interactions among multiple factors in EA risk have not been well characterized. We conducted a case-only study to investigate gene-environment interactions among 145 functional/tagging SNPs of angiogenesis pathway genes, reflux, smoking, and BMI in 335 patients with EA. We used a two-stage approach to detect interactions: first, we applied random forest (RF) approach to select important interacting markers based on mean decrease in accuracy and mean decrease in GINI index; then we estimated the interaction odds ratios (ORinter) of RF selected interaction markers by case-only logistic regression, adjusting for covariates and false discovery rate (FDR). We identified twelve interaction markers that were significantly associated with EA risk (all FDR_Ps Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 4739.