Background

Urothelial carcinoma is a significant health problem and is the fourth most common malignancy in the Western world [1]. The majority of patients present superficial disease (i.e., tumors that do not invade the muscularis mucosa), and approximately 70% of patients develop disease recurrence with 10–40% of cases ultimately progressing to muscle-invasive disease [2, 3]. Histologically, transitional cell carcinoma (TCC) is the most common subtype and represents nearly 90% of all bladder cancers[4].

Urothelial cancer has been described as a tumor type in which the TP53 gene is frequently mutated [5]. The TP53 gene plays an essential role in regulating the cell cycle and apoptosis and therefore contributes to both cellular transformation and malignancy [6]. Moreover, p53 is a DNA-binding protein that contains transcription, DNA-binding, and oligomerization activation domains; furthermore, p53 functions as a tumor suppressor [7]. However, little is known concerning the differential gene expression patterns of human tumors presenting either wild-type or mutated TP53. Sanchez-Carbayo et al. [8] have observed that bladder tumor cells assemble based on their molecular alterations in relation to the TP53 and RB signaling pathways. Moreover, they observed that cells that harbor TP53 mutations in exons 4 and 5, detectable pRB, and INK4A mutations cluster together and were distinguishable from those with TP53 mutations affecting exons 7, 10, and 11, undetectable levels of pRB, and a wild-type INK4A locus. Similarly, another study demonstrated that TP53 mutations frequently occur in the progression of bladder cancer; moreover, they identified a relationship between gelsolin and TP53 status, tumor staging, and clinical outcome [9].

In the post-genome era, high-throughput methods have enabled investigators to perform comprehensive analyses of molecular activities in different cell types, which may lead to greater understanding of carcinogenesis and tumor development. In this context, the DNA microarray has become the preferred method for the large-scale analysis of gene expression [10]. The DNA microarray is a molecular technology that enables identification of gene interactions from experimental data through computational analysis [11]. The elucidation of the molecular events involved in tumorigenesis has led directly to the discovery and application of novel biological markers, including some predictive biomarkers for bladder tumors [8, 9]. Some investigators have shown, for instance, that repression of the hTERT (human telomerase reverse transcriptase) transcript decreases the expression of the EGFR gene at the mRNA and protein levels, supporting the hypothesis that hTERT promotes the growth of tumor cells by mechanisms independent of telomere lengthening [12]. On the other hand, Kikuchi et al. [13] have also shown that expression data can easily distinguish two major histological types of non-small cell lung cancer, adenocarcinoma and squamous cell carcinoma. Subsequent analysis of the adenocarcinomas identified genes whose expression levels could separate cases with lymph-node metastasis from those without metastasis. Additionally, Staege et al. [14] have detected differences between the gene-expression profiles of chemotherapy-resistant and -sensitive Hodgkin’s lymphoma cells.

At present, TP53 sequences and gene expression profiles associated with TP53 status can be characterized in a high-throughput manner. Therefore, in the present study, we aimed to better characterize bladder cancer cell lines by gene profiling and gene networks to validate new targets involved in tumor progression using cDNA microarray and quantitative real-time PCR (qRT-PCR) technologies.

Materials and methods

Cell lines

The established cell lines RT4, 5637, and T24 from human bladder TCCs were obtained from the Cell Bank of the Federal University of Rio de Janeiro, Brazil. The 5637 cells, obtained from a moderately differentiated tumor, contain two mutations in the TP53 gene, one at codon 280 (Arg > Thr) and the other at codon 72 (Arg > Pro) [9, 15, 16]. The T24 cells, obtained from a high-grade tumor, display a TP53 allele encoding an in-frame deletion of tyrosine 126 [9, 15]. No specific mutations are found in the RT4 cells, which is a cell line established from a low-grade papillary bladder tumor [15]. The cells were cultured following standard procedures [17].

RNA extraction

All cells were grown and harvested at 80–90% confluence, after no more than six passages in culture, for the extraction of total RNA using the mirVana kit (Ambion, Applied Biosystems) according to the manufacturer’s instructions. The extracted RNA was stored at –80°C. The integrity of the samples was evaluated by denaturing agarose gel electrophoresis under standard conditions.

cDNA microarray and gene expression analysis

Analyses of the gene expression profiles were performed using glass slide microarrays containing 4,500 sequences (spotted in duplicate) from the human IMAGE cDNA library Consortium (http://image.llnl.gov/image/), which were kindly provided by Dr. Catherine Nguyen (INSERM-CNRS, Marseille, France) and prepared according to the protocol described by Hegde et al. [18]. The microarrays were prepared using type 7 silanized glass slides (Amersham Biosciences, England) and a Generation III Array Spotter (Amersham-Molecular Dynamics) according to the manufacturer’s instructions. The cDNA complex probes were prepared from samples of the cell lines, and reverse transcription was carried out with 10 μg of total RNA/sample in the presence the Cy3 or Cy5 fluorochromes (CyScribe Post Labeling Kit, Amersham Biosciences) and oligo dT12–18 primers. A pooled reference design was chosen. More specifically, each array was hybridized with the same reference sample (prepared from three different cell lines: U343 MG-a glioma, HeLa, and Jurkat cells) labeled with Cy5, while the experimental samples (control or treated) were labeled with Cy3. The reference sample was used in the normalization procedure. Hybridizations were performed for 15 h using an automated system (Automatic Slide Processor, Amersham Biosciences). The hybridization signals were captured using a Generation III laser scanner (Amersham Biosciences).

Image quantification was performed using the Spot software (http://www.tm4.org/spotfinder.html) [19]. Filtering, normalization, and data analysis were performed using the R Statistical Environment [20] according to the instructions and with the KTH package (http://www.biotech.kth.se/molbio/microarray/) [21]. The background for each feature was subtracted from the foreground value. Furthermore, each spot was evaluated on the basis of its circularity and the deviation between its median and the mean. Spots with irregular circularity or large differences between the mean and their median values were considered unreliable. Subsequently, the print-tip Lowess normalization method was applied, and the M values were rescaled by the MAD (median absolute deviation) value [22]. The normalized data were exported to the MEV format and analyzed using the MEV software (http://www.tm4.org/mev.html) [23], which was used to perform statistical analyses and hierarchical clustering. For the clustering procedures, normalized genes were median-centered. The information regarding gene locations and biological functions was obtained from S.O.U.R.C.E. (http://smd-www.stanford.edu/cgi-bin/source/sourceSearch) [24], NCBI (http://www.ncbi.nlm.nih.gov/) [25] and FATIGO (http://babelomics.bioinfo.cipf.es/) [26].

For data analysis, ANOVA (uncorrected P < 0.10) was used. The significance analysis of microarrays method (SAM) (FDR < 0.05) was applied for those genes identified by ANOVA, and multiclass comparisons were performed. The SAM method, which assigns a score to each gene on the basis of change in gene expression relative to the standard deviation of repeated measurements, was used for each independent experiment. The SAM method uses permutations of the repeated measurements to estimate the percentage of genes identified by chance, which is the false discovery rate (FDR), and constructs a scatter plot displaying the induced and repressed genes [27]. A complete file providing all of the genes and ESTs present in the microarrays used in this study, as well as the experimental conditions, are available at MIAME public database: Array Express accession E-MEXP-2859.

Gene network analysis

The GeneNetwork v1.2 algorithm was used to compare the median values for the different gene expressions in cases where the standard deviations did not overlap. The software implementing this algorithm can be obtained from http://idv.sinica.edu.tw/hchuang/GeneNetwork1.2Setup.exe [28]. The objective was to compute a network that established relationships between genes. Bayesian interpolation was used for the network organization. The parameters utilized were based on the probabilities for a hypothetical population of 200 individuals and a mutation rate of 0.02. All utilized networks were stabilized. To organize networks with only the genes whose expression values were significant, normalized values for the induced and repressed genes obtained from the SAM statistical analysis were used. Genes found in both tumor grades establishing interactions with TP53 and other sequences were considered.

Validation of gene expression data by quantitative real-time pcr

cDNA synthesis

The differential expression of two genes randomly selected from the microarray analysis (ANLN and S100P) was validated by real-time PCR using the TaqMan system (Applied Biosystems, USA). ANLN and S100P were selected because of their participation in cancer progression. The TP53 gene expression was also studied. Total RNA (1 μg) from the cell lines was reverse transcribed using 6 μl of random hexamer primer (10X), 6 μl of reaction buffer (10X), 2.5 μl of dNTPs (25X), and 3 μl of MultiScribe enzyme (50U/μl) (High Capacity, Applied Biosystems, USA). This mixture was incubated at 25°C for 10 min and then at 37°C for 2 h, and it was then kept at 4°C. Subsequently, each cDNA was stored at –20°C.

Quantitative real-time PCR

Each reaction tube contained 2 μl of cDNA template, 5 μl of Master Mix TaqMan 2X (Applied Biosystems), and 0.5 μl of 20× primers/probe (Assays-on-Demand Gene expression products, Applied Biosystems). GAPDH was used as the housekeeping gene. The PCR program consisted of the following steps: 2 min at 50°C, 10 min at 95°C, and 40 cycles of 15 s at 95°C, followed by 1 min at 60°C. Fluorescence data were collected during each annealing/extension step. The reactions were performed in an Applied Biosystems 7500 FAST Real-Time PCR System and were analyzed with SDS version 1.2.3 (Sequence Detection Systems 1.2.3, 7500 Real-Time PCR Systems, Applied Biosystems) software. For each PCR assay, a negative (no template) control was processed for routine quality control. Assays were replicated in four independent experiments.

Standard curve and data analysis

To generate standard curves, four serial dilutions of the mRNA derived from the RT4 cell line were used. The first dilution (i.e., highest concentration) of the RNA standard was given the relative value 100, and the three other dilutions were given the relative values of 20, 4, and 0.8. The mRNA quantity in the samples was expressed relative to the standard curve. The results were analyzed by ANOVA and the Bonferroni multiple comparison test.

Results

Gene expression analysis by cDNA microarrays

ANOVA/SAM analyses with a false discovery rate of less than 0.05 revealed 756 differentially expressed genes among the three cell lines. Unsupervised hierarchical clustering of four cDNA microarray experiments showed that the gene profiling of the RT4 cells was more similar to that of T24 than to that of the 5637 cells (Fig. 1a). Using the FATIGO software, it was determined that these genes were mainly related to level 3 biological processes, including cell communication, cell death, cell cycle, cellular component organization, cell division, cell adhesion, and cell proliferation (Fig. 1b). From these 756 differentially expressed genes, seven important clusters of genes implicated in tumor recurrence, progression, and malignancy were selected for discussion (Table 1).

Fig. 1
figure 1figure 1

Gene expression profiling from transitional carcinoma cell lines (RT4, 5637, and T24) and the associated biological processes (level 3). a Heat map and seven selected clusters of unsupervised hierarchical clustering analysis from four replicates: the red spots indicate up-regulation, green spots indicate down-regulation, black spots indicate an absence of modulation, and gray spots indicate the absence of values; b Venn diagram mainly showing genes related to biological processes that are differentially expressed among the three cell lines. (Color figure online)

Full size image
Table 1 Differentially expressed genes among the RT4, 5637 and T24 cell lines selected from seven principal clusters and their related biological processes (level 3)
Full size table

TP53 gene networks

For the TP53 network analysis, 195 genes from the seven clusters were selected. Bayesian interpolation analyses of RT4 × 5637 (Fig. 2) and RT4 × T24 (Fig. 3) were performed to understand the mechanisms involved in carcinogenesis in the two mutated cell lines. The RT4 × 5637 network showed 183 differently expressed genes. Of these genes, 105 most likely modulate TP53, and 77 are likely modulated by the TP53 gene. About one-third (66/183) of the genes are characteristic of and appear only in the RT4 × 5637 network, including CDC40, OAS2, ATP1B3, PSMA1, MAP4K2, and TBXAS1. The second network from the RT4 × T24 analysis showed 163 genes interacting with TP53, of which 61 most likely modulate TP53 and 102 genes that are likely modulated by TP53. About one-fifth (33/162) of the genes are characteristic of the RT4 × T24 network, including ITGA5, PDIA3, and GPR34. Moreover, the gene network analyses showed that TP53 modulated SP100 in the RT4 × T24 network and was modulated by SP100 in the RT4 × 5637 analysis. Differential interactions were also found between TP53 and ANLN in the two gene networks. However, the interaction between ANLN and TP53 was likely indirect (i.e., through other genes) (data not shown).

Fig. 2
figure 2

Bayesian network of the RT4 × 5637 cell lines. The TP53 transcript is in the center, and the probabilities of interactions with 182 other genes are shown

Full size image
Fig. 3
figure 3

Bayesian network of the RT4 × T24 cell lines. The TP53 transcript is in the center, and the probabilities of interactions with 163 other genes are shown

Full size image

S100P, ANLN, and TP53 gene expression

To validate the data obtained from the cDNA microarrays, two differentially expressed genes, S100P and ANLN, were tested by real-time qPCR. Additionally, the TP53 expression was confirmed. The PCR efficiency of all genes was better than 83%.

The results showed an up-regulation of the S100P gene in RT4 when compared to the 5637 and T24 cell lines. The relative expression of the S100P gene in RT4, 5637, and T24 was 1, 0.06925, and 0.00172, respectively. The expression of the ANLN gene was down-regulated in RT4 and T24 compared to 5637, with relative expression levels of 1, 50.3575, and 3.71 in RT4, 5637, and T24, respectively. These data were similar to those obtained with the cDNA microarrays. The expression of TP53 was significantly lower in T24 and RT4 than in the 5637 cells. The relative expression levels of TP53 were 1, 28.985, and 0.4815 in RT4, 5637, and T24 cells, respectively (Fig. 4).

Fig. 4
figure 4

Relative expression of the ANLN, S100P, and TP53 genes. The RT4 cells had a relative expression value of 100. The data represent the mean and standard deviation for four independent experiments. *P < 0.05 when compared to the other two cell lines

Full size image

Discussion

Bladder cancer consists of a heterogeneous group of tumors that follow different pathways of development and progression. Further elucidation of the course of tumor progression and classification into distinct groups will likely have a significant effect on the management of this disease [29]. As a result, we aimed to characterize the gene expression profiles of three bladder cancer cell lines from different tumor grades and containing different TP53 alleles using cDNA microarrays technique validated by real-time qPCR.

Unsupervised analysis showed no influence of TP53 gene status on the hierarchical clustering for the three cell lines. From this analysis, RT4, a cell line derived from a superficial tumor with wild-type TP53, had a similar gene expression profile to that of T24, a cell line derived from an invasive tumor with mutated TP53; on the other hand, the 5637 cells, which were also derived from an invasive tumor moderately differentiated tumor and contain mutated TP53, had different gene expression profiles.

Several authors have shown that most of the mutations observed in human tumors abrogate or attenuate the binding of p53 to its consensus DNA sequence and, consequently, avoid the transcriptional activation of the p53 target genes. These mutations can result in partial or complete loss of the p53 [30] function. Several mutations were shown to confer new functions for mutant p53 that are independent of wild-type p53 [31]. These observations have several implications, including a possible heterogeneous clinical phenotype depending on whether p53 itself is mutated and on the site of the mutations or whether the p53 function is modified indirectly [32]. Moreover, some authors have suggested different responses to chemotherapy based on TP53 status. As2O3, an ancient anticancer drug, suppresses the growth of the wild-type p53U87MG human glioma cells by regulating expression of genes involved in cell cycle arrest, stress and toxicity. On the other hand, this drug acts on mutant p53T98G cells by regulating genes related to Bcl-2, tumor necrotic factor receptor and ligand families [33].

Using bootstrapping methods, Sanchez-Carbayo et al. [8] have grouped bladder cancer cell lines based on their p53/pRB/INK4A pathways status. However, these authors have also observed that cells known to harbor mutations in exon 8 in the core domain (for instance, the 5637 cell line) were not included in the cluster described by the bootstrap technique. Thus, they have suggested that although the p53/pRB/INK4A pathways are important in the development and progression of bladder cancer, further investigations are necessary to identify other involved networks. On the other hand, when analyzing genome-wide expression profiles across different in vitro studies, it is difficult to find common signatures associated with the TP53 mutations, which is likely due to variations in the type of mutant analyzed, the cellular system, and other technical variables [31]. In our study, the gene networks determined by Bayesian interpolation showed interactions among TP53 and several other genes, suggesting that other pathways might be associated with the malignant phenotype. These interactions were different in the two gene networks, but TP53 was not a regulatory gene node in either network. For instance, in the 5637 gene network, the S100P gene was shown to likely stimulate TP53, while in the T24 gene network, the S100P gene seemed to be stimulated by the TP53 gene.

When the results from the microarrays were classified according to gene functions, 406 of the 756 genes were those with no functional annotation. Of the remaining 350 genes, many were mainly involved in cellular communication, cell cycle, cell division, cell death, cellular component organization, cell adhesion, and cell proliferation, among other functions. These functions are essential for normal cellular processes, and when these pathways are altered, they can be implicated in the development and progression of bladder cancer.

Our results showed decreased expression of genes related to apoptosis in the RT4 and T24 cell lines compared to the 5637 cell line. The down-regulation of the CRADD and HTRA2 genes indicates that apoptosis regulated by the activation of caspases could be compromised in these cells. The HTRA2 gene acts as an inhibitor of IAP (inhibitor of apoptosis protein) [34], and CRADD promotes the interaction between the adaptor molecules and the prodomain of the initiator caspase 8 [35]. Additionally, differences in several signaling pathways, such as AMP kinase, JAK/STAT3, and MAP kinase (p38 MAPK, ERK, JNK), were observed among the three cell lines. We detected down-regulation of the adipoR1 (involved in the AMP kinase pathway), ABCA7 (involved in the ERK phosphorylation pathway), DUSP22 (involved in the ERK and MAPK pathways), and AKAP7 (involved in the second messenger-mediated signaling events) genes in the RT4 and T24 cells. Alterations in these pathways could trigger specific signals that might lead to uncontrolled cell growth, enhanced angiogenesis, inhibition of apoptosis, and genetic instability, thereby resulting in tumor development. On the other hand, the induction of genes related to cellular proliferation can also contribute to tumor progression. The CALU gene, for example, was up-regulated in the RT4 and T24 cell lines compared to the 5637 cell line. The increase in CALU expression suggests the transmission of accelerated cellular information [36], contributing to the uncontrolled cell proliferation observed in tumors.

Because the 5637 cell line presented different gene expression profiles than the other two cell lines, other genes and mechanisms involved in apoptosis and cell proliferation may be associated with bladder tumors. In agreement with the results of Hinata et al. [37], our qRT-PCR experiments showed that the TP53 gene was highly expressed in the 5637 cells. Several authors have demonstrated that the nuclear accumulation of p53 shows a high correlation with gene mutations [38]. Genes related to transcription, replication, and DNA synthesis were also observed to be differently expressed among the three cell lines. The up-regulation of TRRAP, SUMO1, ZMIZ1, and RBMS1 in the 5637 cells might indicate stimulation of cell proliferation pathways that contribute to tumor development. TRRAP and SUMO1 are important genes in the regulation of transcription [39, 40], while ZMIZ1 acts in transcription and DNA replication and RBMS1 is a DNA replication factor [41]. Moreover, death mechanisms were inhibited in the 5637 cells. The up-regulation of the GSTP1, SMAD4, and CHAF1B genes seems to block apoptosis and induce cell proliferation. Conversely, genes related to cell-cycle control were altered in the three cell lines. The CHEK1 and CCT4 genes were down- and up-regulated, respectively, in the 5637 cells. CHEK1 is important in S phase and for the G2/M checkpoint [42]; meanwhile, CCT4 is a positive regulator of the G1/S transition in the cell cycle [43]. Together, the data indicate that these checkpoints might be altered in tumoral cells.

Interestingly, the TBXAS1 and SLC2A1 genes were up-regulated in the RT4 cells. Several authors have shown that the TBXAS1 up-regulation seems to be associated with advanced stages, high tumor grades and poor prognosis for patients with bladder cancer [44]. Additionally, high expression of SLC2A1 in tumors has been associated with poor prognosis [45]. Based on our data, we believe that the use of these genes as potential tumor markers should be viewed carefully because high expression of these genes was also observed in the superficial tumor cell line. Conversely, SERPINH1, which regulates collagen production, was down-regulated in the TP53-mutated cells relative to the wild-type TP53 cells. Some authors have found that SERPINH1 is silenced in neuroblastoma cell lines [46]; in our study, this gene seemed be associated with the tumoral grade. The low-grade bladder tumor presented up-regulation of the SERPINH1 expression, which could negatively influence the malignant potential of the RT4 cells.

In addition to the differences observed between the 5637 cells and the RT4 and T24 cells, it is important to note that the balance between cell proliferation and cell death involves several mechanisms and that, sometimes, the signal for cell survival surpasses signals for apoptosis. We observed different gene expression profiling in the three cell lines. Different genes associated with the inhibition of apoptosis and stimulation of cell proliferation were observed to be down- or up-regulated in the three lines, indicating that different mechanisms are involved in the development and progression of bladder cancer. Moreover, the divergence between the gene expression profiles of the 5637 and T24 cell lines, both of which have TP53 mutations, may reflect the complex biology of TP53 in the regulation of apoptosis and cell proliferation. Because the TP53 network is closely linked to many other cellular pathways, it is possible that defects in some of these pathways, either inherited or acquired somatically, could qualitatively or quantitatively interfere with the p53 function. Furthermore, p53 is only one component of the giant surveillance network whose efficiency is modulated by many other elements, including negative (Mdm2, Mdmx, Pirh2 and COP1) [47] and positive (DERP6) [48] regulators of p53, other members of the p53 family and several other signaling pathways [49].

Our data showed that the expression levels of ANLN and S100P as detected by qRT-PCR were significantly correlated with the mRNA expression levels observed in the cDNA microarray. The S100P gene encodes a calcium-binding protein expressed in different tumors that is functionally implicated in the malignant phenotype [50]. On the other hand, ANLN encodes the human homologue of anillin and plays an important role in cytokinesis [51]. We observed that the expression levels of S100P and ANLN were not associated with the tumor grade: S100P was significantly up-regulated in RT4 (low-grade superficial bladder tumor) compared to the 5637 and T24 cells, while ANLN was over-expressed in the cells from the grade 1 (RT4) and grade 3 (T24) tumors. Although the up-regulations of these two genes are important biomarkers for some tumors, our data showed lower expression of S100P in higher grade cells and ANLN in the RT4 and T24 cells. Therefore, S100P and ANLN expression levels should be carefully considered when planning therapeutic strategies for bladder cancer.

Conclusions

In conclusion, the molecular profiling using cDNA microarrays did not cluster the bladder cancer cell lines based on their TP53 gene status. Our comprehensive expression profiling data can provide new insights into the molecular biology of bladder cancer. The identification of novel genetic susceptibility markers will help to understand the biology of urothelial carcinogenesis and may be integrated with clinical data and epidemiological and genetic risk factors for identifying individuals at high risk for developing bladder cancer.