1. Tab.1: Relative transcript levels of prostate-related genes in prostate tissues and cell lines (zmol gene/zmol TBP) * Data for the prostate-related genes are given for the measured relative expression levels (zmol gene/zmol TBP). prostate tissuesprostate cell lines gene*malignant (Tu)non-malignant (Tf) LNCaP22Rv1PC-3DU145BPH-1 median (min to max) mean AibZIP24.5 (3.36 to 74.3)13.7 (0.52 to 59.2) 9.8635.9011.1080.7580.298 PCA335.4 (0.04 to 389)0.57 (0.02 to 213) 0.1000.0300.001 D-GPCR3.81 (0.04 to 136)1.55 (0.02 to 29.9) 0.1950.003 0.0010.002 EZH21.06 (0.35 to 5.86)0.53 (0.13 to 19.9) 2.1436.4231.7146.0362.405 PDEF23.3 (2.01 to 54.1)12.9 (0.17 to 59.0) 5.1683.9501.3820.0350.004 prostein16.8 (1.63 to 90.7)16.3 (0.27 to 84.9) 1.1350.8600.1010.0300.021 PSA367 (18.9 to 1350)226 (0 to 1685) 5.3841.997n.d. PSCA2.16 (0.02 to 732)1.89 (0.01 to 158) 0.0630.2040.0070.0320.097 TRPM835.9 (0.18 to 428)9.37 (0.03 to 77.7) 1.1240.0040.017n.d.0.001 Fig.1: Correlation of expression of prostein, PSA & TRPM8 with T-stage [unpaired t-test, log-transformed relative expression levels for Tf (n=106), OCD (n=59) & NOCD (n=47) samples] Conclusion mRNA levels of AibZIP, D-GPCR, EZH2, PCA3, PDEF, PSA, TRPM8 (all Tf. PCA3 is a powerful predictor of primary PCa but the inclusion of EZH2, prostein & TRPM8 adds even more to the diagnostic power (ROC analyses). The finding of a significantly higher mRNA expression of 3 genes (prostein, PSA, TRPM8) in organ-confined tumors compared to non-organ- confined tumors could be of clinical importance and should be reevaluated in prospective studies using specimens from diagnostic biopsies. Objective Aim of this study was to evaluate whether one of 9 prostate cancer (PCa)-related transcript markers or a combination of them are predictors for PCa. After careful assessment of the potential of the different prostate- associated and/or PCa-relevant candidates for comparative analyses, standardized and validated measurements of mRNA levels were performed. In addition, the power of the single markers for predicting localized disease was assessed. In order to choose a suitable reference gene for prostate tissue pairs, the mRNA expression levels of 4 housekeeping genes were determined in parallel. Material & Methods Tumor patients and cell lines - matched tissue samples (Tu & Tf) from 106 patients with primary PCa (hormone-naive, RPE cases, cM0) - patients’ median age 64 years (48 to 78) with serum levels of PSA (day -1 pre-surgery) 1.3 to 57.2 ng/ml (median 8.3) - histopathological examination according to the UICC system: 59 (56%): organ-confined disease (OCD, pT2); 47 (44%): non organ-confined disease (NOCD, pT3 and pT4) 92 (87%): patients pN0, 14 (13%) pN1 28 low grade PCa (GS 2 to 6), 5 intermediate grade PCa (GS=7; n=5) 27 high grade PCa (GS 8 to 10; n=27) 77 without PSA relapse after surgery (follow-up 32 m), 10 with PSA relapse (PSA ≥0.2ng/ml) and 29 adjuvant treated - prostate cell lines DU 145, LNCaP, 22Rv1, PC-3 and BPH-1 RNA isolation, cDNA synthesis and quantitative PCR (QPCR) - 50-60 slices of cryo-preserved tissue samples for isolation of total RNA (Spin Tissue RNA Mini Kit; Invitek) - 2 portions of 500 ng RNA for RT (Superscript II), both cDNA samples were pooled & diluted - QPCR assays with HP or TaqMan probes to quantify the mRNA of 4 housekeeping & 9 prostate-related transcripts - two independent PCRs for each cDNA sample, differences >30%: additionally QPCR round - quantity standard curves with LC capillaries coated with 10E1 - 10E7 template molecules - transcript amounts calculation by LC-software 3.5 and relative expression levels of prostate-related markers were calculated by normalization to reference genes (zmol marker/ zmol reference gene) Statistics and correlation of the QPCR results to clinical data - analyses by SAS software & SPSS software packages - log-transformed relative mRNA expression levels of markers (comparison Tu & Tf, paired t ‑ test) - receiver-operating characteristic (ROC) curves (to assess the diagnostic power of each separate variable univariately) & for the multivariate diagnostic rule by area under curve (AUC) of ROC curve Literature Schmidt U., Fuessel S., Koch R., Baretton G.B., Lohse A., Tomasetti S., Unversucht S., Froehner M., Wirth M.P., Meye A. Quantitative multi-gene expression profiling of primary prostate cancer. The Prostate, DOI 10.1002/pros Results Standardization of the RT-QPCR & choice of a suitable reference gene for prostate tissues - mean slopes of the regression curves 0.984 (GAPDH) to 1.149 (TBP) and of the quantity standard curves (–3.575 to –3.341) - based on log-transformed mRNA expression levels of 4 reference genes only TBP was not differentially expressed between Tu & Tf (Fig. 2), so TBP were used for normalization Differential expression of prostate-related genes (see list at the right) - median relative expression levels of prostate-related genes in Tu and Tf specimens (Tab.1) - log-transformed relative mRNA levels in Tu > Tf for AibZIP, D-GPCR, EZH2, PCA3, PDEF, PSA, TRPM8 (all p<0.001) & prostein (p=0.018) in paired t ‑ tests - highest Tu:Tf ratios for PCA3 (median 37.5-fold) & TRPM8 (median 3.7-fold) (Tab. 1, Fig. 3) Univariate and multivariate analyses for the prediction of malignant prostate tissue - ROC curves were generated and AUC were calculated for every single parameter - PCA3: highest AUC value of 0.8. As an example, choosing a sensitivity of 95%, this would result in a specificity of 46%, a positive predictive value of 64%, and a negative predictive value of 91% when using a cut off value of 0.4 zmol PCA3/zmol TBP. - EZH2 and TRPM8: AUC values of more than 0.80 thus performing better than the other single markers (Schmidt et al., 2006). - logit model: (based on EZH2, PCA3, prostein & TRPM8) AUC of 0.90 yielded (Schmidt et al., 2006) - all variables were divided into 2 to 4 classes resulting in different cut points to distinguish between Tu & Tf (Schmidt et al., 2006) - using a Wald test, the contrast between the univariate (PCA3 only) & logit model was significant (p=0.0015) indicating a better performance of the multivariate model (Fig. 4) Subgroup analysis and correlation with relevant clinico-pathological parameters - in OCD vs. NOCD significantly higher expression levels for prostein, PSA & TRPM8 (Schmidt et al., 2006) - no differences for N status Correlation of gene expression with treatment failure - no significant differences between patients without PSA relapse (n=77) & those with PSA relapse (n=10) - in contrast, statistically significant differences for AibZIP (p=0.049), PDEF (p=0.01), prostein (p=0.006) and PSA (p=0.04) in the Tu tissues of patients without a PSA relapse and patients who had received adjuvant therapy since they had NOCD at the time of surgery (n=29) Quantitative multi-gene expression profiling of primary prostate cancer Susanne Unversucht, Uta Schmidt, Axel Meye, Susanne Füssel, Rainer Koch $, Gustavo B. Baretton #, Michael Fröhner and Manfred P. Wirth Department of Urology, $ Institute of Biometry and # Institute of Pathology, Technical University of Dresden, Germany Fig.2: Boxplots of mRNA expression levels of different reference genes (distribution of log-transformed transcript levels, normalized to RNA amounts; differences of unpaired t ‑ tests) Fig.3: Ratios of expression levels (Tu:Tf) of prostate-related genes Fig.4: ROC curves for PCA3 (AUC=0.85) and the multivariate logit model comprising EZH2, PCA3, prostein and TRPM8 (AUC=0.90).