Volume 64, Issue 6, Pages 941-950 (December 2013) A 36-gene Signature Predicts Clinical Progression in a Subgroup of ERG-positive Prostate Cancers  Delila Gasi Tandefelt, Joost L. Boormans, Hetty A. van der Korput, Guido W. Jenster, Jan Trapman  European Urology  Volume 64, Issue 6, Pages 941-950 (December 2013) DOI: 10.1016/j.eururo.2013.02.039 Copyright © 2013 European Association of Urology Terms and Conditions

Fig. 1 (A) Gene expression and clinical progression characteristics of primary prostate cancer patients. (a1) Gene Cluster and TreeView were used for unsupervised hierarchical average linkage clustering of microarray data determined by Affymetrix GeneChip Human Exon 1.0 ST. The data of 48 primary prostate cancer samples were normalized to geometric mean and log2 ratios for the samples represented in the heat-map. Two main clusters, A and B, are separated; furthermore, cluster B contains the distinct subgroups B1 and B2. (a2) Significance of microarrays (SAM) was used to identify differentially expressed genes between clusters A and B with false-discovery rate set to zero. A heat map of the 25 top differentially expressed genes was generated in TreeView and Gene Cluster.ERG was the most significantly differentiating gene between clusters A and B. (a3) A zoomed-in view of the ERG cluster. The top 25 genes calculated by SAM are marked in yellow. (B) Kaplan-Meier curves assessing the probability of prostate-specific antigen recurrence-free survival, overall survival, or cancer-specific survival of clusters A, B1, and B2 as a function of time after surgery. European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

Fig. 2 Gene expression and clinical progression characteristics of patients with prostate cancer in which ERG is overexpressed. (A) Unsupervised hierarchical average linkage clustering of all 33 ERG-positive prostate cancer samples separated into clusters BI and BII. The four ERG-positive samples from cluster A (Fig. 1) are in group BI (G006) and group BII (G157, G177, and G288), respectively. (B) Kaplan-Meier curves assessing the probability of prostate-specific antigen (PSA) recurrence-free survival, overall survival, and cancer-specific survival as a function of time after surgery between groups BI and BII. BI is correlated with good prognosis (GP) and BII with poor prognosis (PP). Numbers of patients included at different time points are indicated below the Kaplan-Meier curves. (C) BRB-array tool, class-prediction analysis was used for selecting the genes that predict groups GP and PP. Expression of the 36 genes in the resulting classifier is presented in a heat map. European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

Fig. 3 Validation of the classifier using primary prostate cancer samples of a completely independent patient cohort [27]. Using the 36-gene classifier, BRB-array tool was applied to predict the good prognosis (GP) and poor prognosis (PP) groups. Kaplan-Meier curves assessing prostate-specific antigen (PSA) recurrence-free survival were generated for samples positive or negative for ERG. European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

Fig. 4 Box plots showing median expression of INHBA, CDH11, and GREB1 messenger RNA in normal prostate (NP; n=12) and the poor prognosis (PP) and good prognosis (GP) groups. For all three genes, p<0.001 by the Mann-Whitney U test when comparing GP with PP or NP with PP. Outliers are depicted by an open circle. European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

Fig. 5 The effect of modulating cadherin 11, type 2, OB-cadherin (osteoblast) (CDH11) and inhibin, beta A (INHBA) expression on in vitro biologic properties of prostate cells. (A) Western blot analysis of CDH11 protein expression. PC3 cells that express CDH11 were infected with short-hairpin 5700 (CDH11) lentivirus for CDH11 downregulation or pLKO-control virus. Immortalized, normal prostate PNT2C2 cells were infected with pLenti-CMV-h-CDH11 lentivirus that expresses CDH11 or control pWPXLd-GFP virus. Protein detection was performed with the antibodies CDH11 (Life Technologies Corp, Carlsbad, CA, USA) and β-actin (Sigma-Aldrich Co, St. Louis, MO, USA) as loading control and visualized by chemiluminescence. (B) Western blot analysis of INHBA protein expression. PC3 cells were infected with short hairpin (sh) 5704 (INHBA) or pLKO-control virus. PNT2C2 cells were infected with pLenti-CMV-h-INHBA lentivirus or control pWPXLd-GFP virus. Protein detection was performed with the antibodies INHBA (Novus Biologicals, Littleton, CO, USA) and β-actin as loading control. (C) PC3-shCDH11, PC3-shINHBA, and control PC3-pLKO cells were assayed for migration over 8-μm pore membranes or in soft agar in an anchorage-independent growth assay. (D) PNT2C2 cells infected with CDH11- or INHBA-expressing lentiviruses or control PNT2C2-GFP cells were grown in soft agar for anchorage-independent growth. European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions

European Urology 2013 64, 941-950DOI: (10.1016/j.eururo.2013.02.039) Copyright © 2013 European Association of Urology Terms and Conditions