Abstract C28: Discovery of new protein biomarkers for prostate cancer diagnosis using a Pten-dependent mouse model and their validation in human patients

Abstract

Abstract Purpose: The identification of novel biomarkers for the detection of localized cancers will have a profound impact on human health, especially for tumors with silent progression such as prostate cancer. Current prostate cancer biomarkers such as Prostate Specific Antigen (PSA) have limited accuracy and there is a need for improved non-invasive diagnostic tests. Material and Methods: Blood is thought to pick up molecular cues as it circulates through the various tissues, collectively reflecting molecular processes active in those different tissues. To increase the proteomic detection sensitivity, we have developed a method for the selective analysis of N-glycosites from tissue and serum. The analysis of glycoproteins is ideally suited for the detection of such signatures because they are preferentially released from tissue and deposited in the bloodstream where they can be detected in a non-invasive fashion. In our study we initially used the Pten conditional knock-out mouse model for prostate cancer progression. This model in which the different stages of the disease can be deliberately and reproductively induced and followed, allowed us to simulate with very high fidelity the early steps of the human disease. Under the assumption that proteins of interest are enriched in the tissue, we performed label-free comparative proteomics of Pten−/− and control tissue at different time points. Interesting candidates were then monitored in the corresponding mice sera by targeted mass spectrometry using Multiple Reaction Monitoring (MRM) at high sensitivity and selectivity. Results: This approach enabled us to identify 785 N-linked glycoproteins, mostly located to relevant cellular compartments such as the extracellular space, the plasma membrane or the secretory pathway. Quantitative proteomic comparison of mice with homozygous Pten deletion and constitutive PKB activation with the corresponding control animals led to the identification of 153 candidate biomarkers differentially regulated in a Pten dependent manner. The validation phase was carried out using 76 sera collected from patients with either benign prostatic hyperplasia (n=35) or localized prostate cancer (n=41). Candidate biomarkers were monitored by immunoassays and/or targeted mass spectrometry. Using a cross-validated logistic regression approach in our ongoing study, we have identified a 4-biomarker signature that correctly predicts 71% of cases (sens. 78%/spec. 63%). In comparison, PSA correctly predicted 67% (sens. 89%/spec. 32%) at a cutoff of 2 ng/ml. By combining PSA with the 4-biomarker signature, accuracy was greatly improved and 87% of the cases were correctly predicted (sens. 90/spec. 83%). We further extended our analysis to the discovery of novel candidate serum biomarkers for the discrimination of aggressive cancers (Gleason score>7) from slow progressing cancers (Gleason score<=6) that may be considered for active surveillance instead of immediate active treatment. Conclusion: The use of a translational approach which relies on a genetically defined mouse model together with novel proteomics techniques and validation in human sera is thus an ideal starting point for the discovery of novel biomarkers. Citation Information: Cancer Res 2009;69(23 Suppl):C28.