Genomic Approaches to Outcome Prediction in Prostate Cancer Phillip G. Febbo, MD Duke Institute for Genome Science and Policy Department of Medicine Department of Molecular Genetics and Microbiology Duke Prostate Center Duke University Outline Outline
Prostate Cancer Oncogenesis 1. Primarily a stochastic process 2. Progressive accumulation of heritable genetic events 3. An individual s polymorphic portrait can result in the equivalent of an event or provide a hospitable environment for events 4. Single hits in some genes can be functionally deterministic for the development of cancer 5. Seldom a single path to any type of cancer Figure from The Biology of Cancer ( Garland Science 2007) Personalizing Prostate Cancer Care Normal Epithelium And Stroma Pre- Neoplastic Disease Disease States Clinically Localized Prostate cancer Biochem. relapsed Prostate cancer Non- Metastatic, Castrationresponsive Prostate cancer Metastatic, Castrationresponsive Prostate cancer Nonmet CRPC Met CRPC Docetaxel Resistant Met CRPC Molecular States GSTπ Methylation NXK3.1 Loss ETS Translocations P53 Loss PTEN Hemizygous Loss PTEN Homozygous Loss AR Amplifications? MYC Amplification Personalizing Prostate Cancer Care Personalized Medicine The ideal The right treatment for the right patient with the right diseaseat the right time The real A mechanistically rational treatment for patients with a molecular subgroup of tumors
Personalizing Prostate Cancer Care Genomic Dogma Genome Transcriptome Proteome HapMap EpiGenome microrna Splice Variation Metabolome Outline Prostate Cancer Risk No BRCA1 or BRCA2-like genes in prostate cancer. Most inherited risk probably caused by genetic variation (polymorphisms) Genome Wide Association Studies results are starting to be published Polymorphisms on Chromosome 8 found to be associated with risk Difficult to individualize recommendations based on individual DG8S737 polymorphisms: Often have relatively small impact on risk Unclear why they are associated with risk It remains unclear what to recommend for men at slight increased risk
Prostate Cancer Risk Proof of Principle: Multiple Polymorphism Risk Models Assessed 16 previously reported SNPs associated with prostate cancer risk on Chromosomes 8 and 17 Found 5 polymorphisms that could be combined to identify men at high risk # of High Risk Variants Increased Risk All Patients Increased Risk With Family History 0 1.00 1.00 >4 4.47 4.76 5 9.46 Zheng et al NEJM (2008) Prostate Cancer Risk High Risk Chemoprevention Low Risk No 1) Finasteride Treatment 2) Dutasteride 3) Life Style/Diet 4) Supplements (not Selenium or Vitamin E) Prostate Cancer Risk Polymorphisms Associated with Prostate Cancer Risk Proof of principal now established - Begins an era of genomic risk tests for prostate cancer - 5 gene test will be available for $300.00 Further refinement is required - Small percentage of men have a high predicted risk (~6%) - No discrimination between aggressive and latent cancer Each SNP found through GWAS will have the potential to improve/refine model. -Nature Genetics Feb 2008 Two GWAS identified additional SNPs on 3,6,7, 10, 11, 19, X - Additional loci are likely to be found Copy Number Variations? - There is unlikely to be a test
Prostate Cancer Diagnosis Gene Expression and Prostate Cancer Diagnosis AMACR Dhanasekaran et al (2001) Nature Luo et al (2002) Cancer Research Singh, Febbo, et al Cancer Cell (2002) Rubin et al, JAMA (2002) Prostate Cancer Diagnosis Prostate Cancer Biology ERG Translocation Unique and extrodinary expression in a subset of samples Tomlins et al Science (2005) Prostate Cancer Diagnosis Expression Analysis For Molecular Classification Breast cancer classification has been re-defined by expression analysis Hu et al BMC Genomics (2006)
-Germline Analysis Conclusions - Prostate Cancer risk best modeled as a complex trait - Models incorporating multiple polymorphisms are likely to be widely available - Unclear how to act upon knowledge of increased risk -Gene Expression focused analysis: -Identified AMACR as a marker with diagnostic value -Discovery of ETS translocations a profound and fundamental part of prostate cancer oncogenesis -Did not result in novel classification of prostate cancer -Serum Proteomics Prostate Cancer Diagnosis -Although a lot of work has been done, still relatively untested -Proteomics platforms are becoming more robust and reproducible and we are likely to see significant advances over the next 5 years Outline Prostate Cancer Recurrence Where is prostate cancer s OncotypeDX or prostaprint?
Prostate Cancer Recurrence Multiple Gene Model of Recurrence Washington University Multi-gene Model Memorial Sloan Kettering Singh, Febbo et al Cancer Cell (2002) Prostate Cancer Recurrence Multiple Independent Models Predict Recurrence Prostate Cancer Recurrence Genomic Predictors Provide Biological Insight Genes Associated with Outcome 0.00 0.20 0.40 0.60 0.80 1.00 Recurrent Non -Recurrent ITPR3 Sialyltransferase I PDGFR - beta ChromograninA HoxC6 Pathways Associated with Outcome
Prostate Cancer Recurrence Conclusions Consistent expression structure associated with biochemical recurrence in localized prostate cancer Genetic events genetic gain or loss Gene Expression multi-gene models Protein multi-protein models IHC, quantum dots, etc Multiple approaches are successful but different approaches result in different models Unclear if improved analysis will result in a model better than clinical nomograms Promise is in these approaches identifying biology associated with risk of recurrence and suggesting specific interventions Outline Prostate Cancer Treatment Single institution cohort of 529 treated with chemical castration Looked at 129 polymorphisms in 20 different androgen axis genes for association with duration of response to castration Ross et al JCO (2008)
Prostate Cancer Treatment Multi-SNP Model for Time to Progression on Hormones Prostate Cancer Treatment Current Treatment of Castration Resistant Prostate Cancer (CRPC) Secondary Hormonal Second Secondary Hormonal Third Secondary Hormonal Docetaxel Factors in deciding therapy: 1)Patient symptoms 2)PSA dynamics 3)Prior therapies 4)Prior response to therapy Treatment of CRPC is not based upon the molecular features of the disease or individual Prostate Cancer Treatment Expression Signature for Androgen Receptor Activity 0.79 Validated in cell lines, xenografts, and human tissues Mendiratta et al, JCO (2009) Probability of AR Activity
Prostate Cancer Treatment Signature for AR Activity No After TreatmentCastration No Treatment Castration Resistant 1/3 1/2 of CRPC tumors Have High AR Activity Prostate Cancer Treatment Personalized for Men with CRPC Opened 5/27/09 Prostate Cancer Treatment Radiologically- Guided Bx Path QA LCM and RNA isolation Patient RNA QA Vs. Apply AR Signature Microarray QA Microarray
Outline Conclusions Prostate cancer care is in dire need of more informative predictors of disease behavior: Who does not need treatment? Who needs aggressive treatment? What treatment should be given to specific patients? Genomic information is prognostic and predictive in prostate cancer. Germline variations associated with both risk and response to therapy Somatic genetic and expression changes associated with recurrence after definitive therapy The prognostic value of clinical and laboratory characteristics of prostate cancer are quite strong at the extremes but provide no molecular insight Unlikely that we will have a Prostaprint unless the test also suggests a specific intervention/treatment Genomic models to guide therapy in castration-resistant prostate cancer are being prospectively tested. Thank you