383 HYALURONIC ACID AND HYAL1 HYALURONIDASE: PROGNOSTIC MARKERS FOR BLADDER CANCER

381



382

Improving gene array analysis: The application of Artificial Intelligence identifies novel biomarkers of superficial bladder cancer progression

EZH2 polycomb transcriptional repressor expression correlates with methylation of the APAF-1 gene in superficial transitional cell carcinoma of the bladder

Catto J.1, Abbod M.2, Linkens D.3, Wild P.4, Herr A.5, Wissman C.6, Pilarsky C.7, Hartmann A.4, Hamdy F.1

Christoph F., Hinz S., Weikert S., Krause H., Schrader M., Miller K.

University of Sheffield, Academic Urology Unit, Sheffield, United Kingdom, 2Brunel University, 3School of Engineering and Design, London, United Kingdom, 3University of Sheffield, 5Department of Automatic Control and Systems Engineering, Sheffield, United Kingdom, 4Regensburg University, Pathology Department, Regensburg, Germany, 5Dresden University, Institute for Clincal Genetics, Dresden, Germany, 6Humboldt University, Department of Internal Medicine, Berlin, Germany, 7 Dresden University, Department of Surgery, Dresden, Germany

Charite - Universitaetsmedizin Berlin, Urology, Berlin, Germany

1

Introduction & Objectives: New methods are needed to accurately predict the behaviour of an individual patient’s tumour. The global analysis of gene expression using array technology appears a promising method for this purpose. However, interpretation of array data is difficult due to its volume and the reliance upon linear statistical methods. Here we analyse a well defined superficial bladder cancer (SBC) gene array using traditional and novel artificial intelligence (AI) methods. Material & Methods: Genes from a 6,117 probe (2,800 gene) array dataset (Wild et al Cl Can Res 2005) were ranked (Pearsons coefficient) according to their association with disease progression. We studied the top 200 genes, obtained from 75 microdissected SBC specimens and controls, to identify new markers of progression. We analysed the data using logistic regression (LR), artificial neural networks (ANN) and neurofuzzy modelling (NFM) to classify tumour grade, stage and progression. We constructed various repetitive algorithms to test effect of individually removing/ averaging/ minimising/ maximising each probe upon the predictive prognostic accuracy of the data. We ranked each probe accordingly. We then designed predictive models to test the accuracy of panels of the top ranked probes. Results: Principle component analysis was used to reduce the 200 genes and correctly separate progressing from non-progressing tumours. ANN and NFM were used to rank the 200 genes (using model classification error). Panel’s of the top scoring ANN, NFM and the original genes found by Wild et al. were used to predict tumour behaviour. The AI panel’s were more accurate than those derived using statistical regression (original genes; scatter plots in Figure A, B, C). Using the new AI panel of 6 genes we were able to accurately identify tumour progression (KM Curve, Figure D) in the 67 tumours. Conclusions: AI allows interrogation of large array datasets without reliance upon statistical or biological linearity. This explorative method has identified an accurate panel of 6 genes that predicts progression of SBC. Immunohistochemistry, using a large tissue microarray of new SBC, for these genes is ongoing to investigate the robustness of this panel.



383

Hyaluronic acid and hyal1 hyaluronidase: prognostic markers for bladder cancer

Kramer M.W.1, Golshani R.2, Merseburger A.S.3, Hennenlotter J.3, Soloway M.4, Kuczyk M.3, Lokeshwar V.B.5 University of Miami Miller School of Medicine & University of Tuebingen, Departments of Urology, Tuebingen, Germany, 2University of Miami Miller School of Medicine, Department of Cell Biology and Anatomy, Miami, United States of America, 3University of Tuebingen, Department of Urology, Tuebingen, Germany, 4University of Miami Miller School of Medicine, Department of Urology, Miami, United States of America, 5University of Miami Miller School of Medicine, Department of Urology & Cell Biology and Anatomy, Miami, United States of America 1

Introduction & Objectives: Due to heterogeneity in tumour progression and recurrence and variability in treatment response, the clinical management of patients with bladder cancer is both challenging and costly. Hyaluronic acid (HA) and its degrading enzyme, hyaluronidase (HAase) are associated with tumor growth, invasion and angiogenesis. HYAL1 is a HAase expressed by tumor cells. HYAL1-v1 is an enzymatically inactive variant of HYAL1 that negatively regulates BCa growth and invasion. Urinary HA and HAase levels, measured using the HA-HAase test, are highly accurate markers for detecting BCa. In this study, we analyzed the prognostic potential of these markers to predict tumor recurrence, progression and response to treatment. Material & Methods: A cohort of 179 BCa specimens from patients with muscle and non-muscle invasive disease on whom there is a minimum 3-year follow-up was included in the study. The specimens were stained for HA, HYAL1 and HYAL1-v1 using tissue microarrays (T.M.A.) and immunohistochemistry techniques. Each tissue specimen was assessed on a staining scale ranging between 0 (no staining) to 300 (strong staining). Results: Both HA and HYAL1 staining was higher in muscle invasive tumors than in non-muscle invasive tumours (P < 0.001). With 175 as the cutoff limit, HA staining predicted progression with 75% sensitivity and 51% specificity (P < 0.05). Using 195 as the cutoff limit, HYAL1 predicted progression with 65% sensitivity and 62% specificity (P< 0.01). The combined HA-HYAL1 staining inferences had 81% sensitivity and 70% specificity (cutoff 235) to predict disease progression. The chi-square analysis showed that high combined HA-HYAL1 staining indicates a 10-fold higher risk for progression (P < 0.0001; chi-sqaure:17.7; odds ratio = 10). In addition both HYAL1 and combined HA-HYAL1 predicted failure of intravesical treatment, measured as disease progression (P < 0.01). The combined HA-Hyal1 staining had 82% sensitivity and 65% specificity to predict treatment failure (P < 0.001 chi-sqaure: 9.4; odds ratio = 8.8). Contrary to HA and HYAL1 the HYAL1-v1 expression decreased with tumor grade and stage (P < 0.01). Conclusions: The expression of HYAL1-v1, a negative regulator of BCa is down-regulated in BCa. HA and HYAL1 markers may have prognostic potential to predict progression to muscle invasion and response to treatment in patients with BCa. Furthermore, combined HA-HYAL1 has better prognostic potential than individual HA and HYAL1 markers. Grant Support: 5RO1 CA72821-09 (VBL), Int. Acad. Life Sci BMEP (MK)

Eur Urol Suppl 2007;6(2):118

Introduction & Objectives: The EZH2 gene controls methylation of various EZH2 target promoters. The APAF-1, DAPK-1 und IGFBP-3 genes are frequently methylated in bladder cancer and methylation of these genes is found in more aggressive tumor types. We investigated a potential link between EZH2 mRNA expression and the extent of APAF-1, DAPK-1 and IGFBP-3 methylation in urothelial transitional cell carcinoma (TCC) and correlated the data with histopathological parameters and follow-up. Material & Methods: EZH2 mRNA expression was measured by real-time RTPCR and the methylation analysis was performed using methylation specific real-time PCR. Tissue specimens were obtained from 35 patients with TCC. Results: EZH2 mRNA expression was detected in all tumor specimens investigated. The EZH2 expression levels correlated well with the differentiation grade of the tumor specimens (p=0.03) and the APAF-1 methylation correlated with tumor stage (p=0.0001) and grade (p=0.004). Matched pair analysis demonstrated a statistically significant correlation between elevated EZH2 mRNA expression and higher methylation levels of APAF-1 in superficial (p=0.024) and well differentiated (p=0.04) TCC. In patients with recurrent TCC, APAF-1 and IGFBP-3 methylation levels were significantly higher (p=0.03; p=0.01), which was not observed when EZH2 mRNA expression or DAPK-1 methylation levels were related to the clinical outcome. Conclusions: In conclusion, our data show that EZH2 expression and APAF1 methylation are related to tumor progression and invasiveness. Moreover, these data present first evidence that APAF-1 methylation is related to transcriptional activity of EZH2 expression in early-stage tumor disease of the bladder.



384

Screening of telomerase related genes in bladder cancer using complementary deoxyribonucleic acid microarray: preliminary results Wang W.1, Chen S.1, Liu Y.1, Zhang G.1, Bi W.1, Liu X.2

Beijing Tongren Hospital, Captial Medical University, Urology, Bei Jing, China, National Engineer Research Center for Beijing Biochip Technology, Biochip, Bei Jing, China 1 2

Introduction & Objectives: In the post-genome era, and in view of the advent of high-throughput methods of molecular analysis, it is expected that specific tumor types with or without telomerase activity (TA) will have distinct gene expression profiles. The elucidation of the molecular events involved in activation of telomerase is leading directly to the understanding of development of bladder cancer. In the present study, We try to investigate telomerase related genes in bladder cancer using complementary deoxyribonucleic acid microarray. To our knowledge, this was the first attempt to screen the telomerase-related genes in bladder cancer. Material & Methods: Tissue samples were obtained by cystectomy or transurethral resection from 18 patients, 38 to 73 years old (mean age 58) with transitional cell carcinoma (TCC) of the bladder. 18 samples were subjected to Telomeric repeat amplification protocol (TRAP) to analyze TA. Then, complementary deoxyribonucleic acid microarrays containing 21,329 different genes was used to analyze gene expression among 18 Specimens with TRAP confirmed TA positive samples (n = 12) or TA negative samples (n = 6). Quantitative real-time polymerase chain reaction was performed for selected genes to validate the microarray data. Results: Significant up-regulation of 76 genes was associated with TA positive samples, but not in TA negative samples. This effect included genes involved in genes related to cell cycle, cell adhesion, angiogenesis, and apoptosis. Conversely, significant down-regulation of 71 genes was associated with TA negative samples, but not in TA positive samples, including genes related to immune responses, cell cycling, and transcription. Conclusions: This study shows the usefulness of complementary deoxyribonucleic acid microarray technology for screening telomerase-related genes in TCC of bladder. The detailed clarification of these telomerase-related genes will shed light on the question, which genes might involved in activation of telomerase in TCC of bladder.