[CANCER RESEARCH 62, 6973–6980, December 1, 2002] Molecular Profiling of Bladder Cancer Using cDNA Microarrays: Defining Histogenesis and Biological Phenotypes1 Marta Sanchez-Carbayo,2 Nicholas D. Socci, Elizabeth Charytonowicz, Minglan Lu, Michael Prystowsky, Geoffrey Childs, and Carlos Cordon-Cardo Division of Molecular Pathology, Memorial Sloan-Kettering Cancer Center, New York, New York 10021 [M. S-C., E. C., M. L., C. C-C.], and Department of Pathology, Seaver Foundation for Bioinformatics [N. D. S.] and Department of Molecular Genetics [M. P., C. C-C.], Albert Einstein College of Medicine, Bronx, New York 10461 ABSTRACT herin-catenin complexes, have been described to be involved in blad- der cancer progression (9–12). This study was designed to characterize the expression profiles of nine In the post-genome era, and in view of the advent of high-through- bladder cancer cell lines (T24, J82, 5637, HT1376, RT4, SCaBER, put methods of molecular analysis, it is expected that specific tumor TCCSUP, UMUC-3, and HT1197) using cDNA microarrays (8976 genes types will have distinct gene expression profiles (13, 14). The eluci- and expressed sequence tags). Novel targets involved in bladder cancer progression of potential clinical relevance were validated by immunohis- dation of the molecular events involved in tumorigenesis and tumor progression is leading directly to the discovery and application of 193 ؍ tochemistry using tissue microarrays of primary bladder tumors (n cases). Hierarchical clustering classified uroepithelial cells based on their novel biological markers. The diagnosis and prognosis of certain histopathogenesis and cell cycle alterations. Keratin 10 and caveolin-1 neoplasms are in many cases enhanced by the use of such markers, transcripts were more abundant in tumor cells from squamous and inva- and the marker itself may constitute a therapeutic target. In the present sive origin. Their combined expression was shown to stratify bladder study, we have attempted to further characterize bladder cancer and to tumors and define squamous differentiation. To assess the robustness of validate new targets involved in bladder tumor progression using a the clustering analysis, a bootstrap resampling technique was used. This combination of cDNA and tissue microarray technologies. grouped tumor cell lines based on their biological properties, including cell cycle and cell adhesion features. E-cadherin, zyxin, and moesin were identified as genes differentially expressed in these clusters and related to MATERIALS AND METHODS the p53, RB, and INK4A status of the cell lines. Loss of these adhesion Cell Culture and RNA Extraction molecules was associated with stage and grade in primary tumors (P < 0.05), and moesin expression was also associated with survival Nine bladder cancer cell lines (T24, J82, 5637, HT1376, RT4, SCaBER, -Deregulation of cell cycle and apoptotic pathways, such as TCCSUP, UMUC-3, and HT1197) were obtained from American Type Cul .(0.01 ؍ P) mutations or altered expression of p53, pRB, and INK4A (p16), is neces- ture Collection (Manassas, VA) and cultured under identical conditions fol- sary for uroepithelial transformation. However, it appears that deregula- lowing standard procedures. All cells were grown and harvested at 75%–90% tion of cell adhesion is a common event associated with tumor progression confluence no longer than 4–6 passages in culture for the extraction of total in uroepithelial neoplasms. RNA using the RNeasy protocol (Qiagen, Valencia, CA). Cytospins were also prepared and later used for target validation. Preparation of cDNA Microarrays INTRODUCTION A set of 8976 sequence-verified human IMAGE cDNA clones, representing Bladder cancer is one of the most common malignancies in devel- both known genes and expressed sequence tags, were PCR-amplified and oped countries, ranking as the sixth most frequent neoplasm (1). spotted onto polylysine-coated microscope slides using a custom robot de- BBC3 is the most common malignant neoplasm in Egypt and also signed and built at Albert Einstein College of Medicine4 (15). occurrs with a high incidence in other regions of the Middle East and East Africa (2). Certain clinical and pathological features of BBC are Labeling of cDNA, Hybridization to Arrays, and Image Acquisition different from those described for cTCC, such as the high incidence of Ten ␮g of total RNA of each cell line were labeled with Cy5 (red) and detecting squamous metaplasia and the development of SCC (2). TCC hybridized against 10 ␮g of total RNA of a pool containing equal RNA has been classified into two groups with distinct behavior and differ- quantities of all of these cell lines labeled with Cy3 (green). Labeling and ent molecular profiles: (a) low-grade tumors (which are always pap- hybridization of cDNA to arrays were carried out as described previously (16). illary and usually superficial); and (b) high-grade tumors (which are We carried out one duplicate or one reverse-labeling experiment for validation either papillary or nonpapillary and are often invasive; Ref. 3). The of expression changes of the hybridization of the cell lines. After hybridization, inactivation of both RB and p53 pathways has been shown to be slides were washed, dried, and scanned by a custom-built laser scanner (15). Intensity data were integrated with 8ϫ oversampling (15). Scanalyse software required for the transformation and immortalization of uroepithelial was used for gridding and calculation of red (R) and green (G) signal inten- cells (4–7), and their alterations are common and of predictive nature sities (17). in clinical studies of bladder cancer (5, 8). Cross-talk between these pathways and adhesion signaling, such as those generated by cad- Collection and Analysis of the Data of the cDNA Microarrays Normalization. Before any analysis, plots of the fold change versus the Received 8/9/02; accepted 9/27/02. average intensity were examined to look for abnormalities in single-array data. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with It is common to plot a red versus green channel scatter plot to examine 18 U.S.C. Section 1734 solely to indicate this fact. distribution of intensities; however, we found that transforming to fold change 1 Supported in part by National Cancer Institute Grant CA-47538 (to C. C-C.) and by versus average intensity displayed the data in a more easily viewed form. If grants from the NINDS NIH (Grant NS39662) and the Seaver Foundation in Bioinfor- Ired is the background subtracted red channel intensity, and Igreen is the matics (to N. D. S.). 2 To whom requests for reprints should be addressed, at Division of Molecular background subtracted green intensity, then the following variables were Pathology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY created: R ϭ Ired/Igreen and A ϭ͌(Ired ϫ Igreen), where R is simply the fold 10021. Phone: (212) 639-7746; Fax: (212) 794-3186; E-mail: [email protected]. change ratio, and A is the average intensity (the geometric mean which is 3 The abbreviations used are: BBC, bilharzial-related bladder carcinoma; TCC, tran- sitional cell carcinoma; cTCC, conventional TCC; SCC, squamous cell carcinoma; S- BBC, squamous BBC; T-BBC, transitional BBC; CI, confidence interval. 4 sequence.aecom.yu.edu/bioinf/funcgenomic.html. 6973 Downloaded from cancerres.aacrjournals.org on September 29, 2021. © 2002 American Association for Cancer Research. EXPRESSION PROFILING OF BLADDER CANCER equivalent to averaging the log intensity). The curvature in the scatter plot were used for immunohistochemistry (19). We used the following antibodies indicated a dependence on the ratio R on the overall intensity. This curve is at certain conditions: (a) caveolin-1, mouse monoclonal IgG1 at 1:1000 dilu- then used to normalize the data: logIred/Igreen ϪϾlog (Ired/Igreen) Ϫ c(A) tion (2.5 ␮g/ml; BD Transduction Laboratories, Lexington, KY); (b) keratin where c(A) is the fit. This is equivalent to multiplying the green channel 10, mouse monoclonal clone DC-K10 at 1:2000 dilution (1.0 ␮g/ml; Neo- intensity (or dividing the red) by an intensity dependent normalization constant markers, Fremont, CA); (c) E-cadherin, mouse monoclonal clone 36 at 1:1000 k(A) where log[(k(A)] ϭ c(A). Optimal normalized data should be horizontal dilution (2.5 ␮g/ml; BD Transduction Laboratories); (d) moesin, mouse mono- and centered at zero. Samples were normalized using this intensity-dependent clonal clone 38/87 at 1:50 dilution (4 ␮g/ml) with microwave pretreatment of normalization using the Splus function lowess (18). Normalized fold changes the slides (Neomarkers); (e) zyxin, mouse monoclonal clone 21 at 1:25 dilution in gene expression were then used to further analyze and cluster the various (10 ␮g/ml) with microwave pretreatment of the slides (BD Transduction cell lines. Laboratories); (f) total RB, mouse monoclonal clone 3C8 at a final concen- Cutoffs. The following filter values were used: the absolute value of the tration of 1.2 ␮g/ml (QED Bioscience, San Diego, CA); (g) p16, mouse fold change (R/G and G/R) had to be greater than 2.0 in at least one experi- monoclonal clone clone DCS-50.1/H4 at 2.5 ␮g/ml (Calbiochem, Cambridge, ment, and the average intensity (A) had to be greater than 300. This filter MA); and (h) a mouse antihuman monoclonal antibody against p53 (1:500 reduced the number of genes from 8976 to 234. Data were filtered to select dilution; Ab-2; clone 1801; Calbiochem). Staining conditions were optimized genes that had both a fold change to remove the background of mostly on sections from formalin-fixed, paraffin-embedded tissue controls for each unchanging genes and an average intensity distinguishable from the noise of antibody as specified by the manufacturers. Antibody reactivity was detected the microchip hybridization.