Abstract
Background: Urothelial carcinoma (UC) is characterized by multiple recurrent chromosomal changes on a background of increasing genomic instability. To define target genes of recurrent deletions and amplifications, we explored which gene alterations are common in UC, in two recently established cell lines, BC44 and BC61. Materials and Methods: Genes located in regions of gain or deletion in the cell lines were identified by array comparative genomic hybridization (aCGH). Six published microarray datasets were analyzed for genes differentially expressed between urothelial tumor vs. normal tissues. Gene expression and chromosomal changes were compared in BC61 cells. Results: The cell lines share homozygous deletions at 9p21 around CDKN2A and amplifications at 11q13.2 around CCND1. In both cell lines 11 genes were commonly lost and 115 gained. Across UC in general, 230 genes were expressed stronger and 349 weaker; a subset displaying corresponding genetic changes in the cell lines. The commonly affected subset contains well-investigated genes like E2F1 and CCNE1, but also several genes not yet sufficiently investigated in UC. Discussion: Our approach highlights genes involved in cell cycle regulation, apoptosis and signal transduction as commonly deregulated across UC. Nevertheless, many chromosomal regions undergoing recurrent changes harbor several commonly deregulated genes that may act jointly in UC development and progression.
Bladder cancer is a notoriously heterogeneous disease. The major histological type in most countries is urothelial carcinoma (UC), which can be grossly categorized into papillary and invasive subtypes. The second most frequent type, squamous cell carcinoma, is typically associated with chronic inflammation and is more prevalent in countries with schistosomiasis infections. Further rarer types exist, such as adenocarcinoma of the bladder. Unfortunately, the above categories are not always distinct. For instance, many UC contain morphologically evident regions with squamous differentiation or express molecular markers or epidermal differentiation. Likewise, high-grade papillary tumors may progress to invasive urothelial carcinomas (1). A major issue in bladder cancer research is therefore defining the genetic and epigenetic changes shared by all histological subtypes as well as those that are characteristic for each subtype in order to provide a basis for diagnosis, monitoring and treatment. Comprehensive studies on gene mutations, copy number alterations, gene expression changes and epigenetic disturbances in the various subtypes of bladder cancer are underway or have already been published and their evaluation may help resolve these issues (2-8).
Cell lines are an important tool for the study of bladder cancer (9). Ideally, one would like to have cell line panels at hand that represent all subtypes of bladder cancer and contain all relevant molecular alterations. There is indeed a large number of well-characterized cell lines for invasive urothelial cancers with typical genetic changes occurring in these tumors, whereas cell lines from papillary urothelial carcinoma are scarce. Assuming that the low number of cell lines from papillary tumors may be due to the inability of cancer cells to survive and grow in standard culture media, we have cultured cells from papillary urothelial cancers using conditions similar to those for normal urothelial cells, which involve low-calcium, and defined growth factors (10). Indeed, most primary tumor cultures grew for a few passages under these conditions and two novel cell lines, BC44 and BC61, were derived. BC61 was established from a non-invasive, moderately differentiated (pTaG2) papillary tumor. Upon detailed molecular analysis, the cell line revealed typical characteristics of papillary tumors such as an oncogenic FGFR3 mutation, deletion of CDKN2A and wild-type TP53. Nevertheless, karyotyping revealed chromosomal changes indicative of a progressive papillary cancer (11). BC44 was derived from a papillary protrusion of an advanced stage bladder cancer displaying mixed UC and squamous cell carcinoma morphology. Typical of invasive bladder cancers, FGFR3 was not mutated and TP53 was inactivated, albeit by an unusual mechanism. Also typical of invasive cancers, the karyotype was aneuploid (11).
We wondered which of the chromosomal alterations shared by these two bladder cancer cell lines might be relevant across bladder cancer in general. Therefore, we identified common regions of copy number gains and losses between the two cell lines and searched for genes located in these regions that show frequent expression changes in bladder cancer overall, by using published microarray expression data. Indeed, in addition to known oncogenes and tumor suppressors in bladder cancer, this analysis revealed a number of chromosomal regions and candidates that may not yet have received sufficient attention in this cancer type.
Materials and Methods
Cell lines and cell culture. The BC44 and BC61 cell lines were derived from primary cultures established as described (10). Cellular and molecular characteristics of these cell lines have been described (11). In brief, both cell lines were routinely maintained in Epilife medium supplemented with 0.25 ng/ml epidermal growth factor (EGF), 12.5 μg/ml bovine pituitary extract (all purchased from Cascade Biologics, Darmstadt, Germany), 1% non-essential amino acids (Invitrogen, Darmstadt, Germany), 1% ITS mix (Invitrogen, Darmstadt, Germany), 3 mM glycine and 10% fibroblast-conditioned medium on a collagen IV matrix (Sigma, Munich, Germany). Cells were passaged twice weekly at a 1:3 ratio using accutase (PAA, Cölbe, Germany) for detachment. Primary urothelial cells (UP) were prepared from ureters after nephrectomy and were routinely maintained in keratinocyte serum-free medium (KSFM, Gibco, Darmstadt, Germany) supplemented with 25 μg/ml bovine pituitary extract and 2.5 ng/ml EGF (12).
DNA and RNA extraction. Total genomic DNA was isolated from sub-confluent cell cultures using the Blood and Cell Culture DNA Midi Kit (Qiagen, Hilden, Germany). Total RNA was isolated from sub-confluent cell cultures using the RNA Mini or Micro Kit (Qiagen).
CGH-array and expression microarray hybridization and evaluation. Array comparative genomic hybridization (aCGH) was performed using an array of 60mer DNA-oligonucleotide probes with a median overall probe spacing of 13 kb (SurePrint G3 Human CGH Microarray 180k, Agilent Technologies, Palo Alto, CA, USA) as described (11) essentially according to the protocol of the manufacturer (Protocol 6.2.1, December 2010, Agilent Technologies). The hybridization patterns were analyzed using a microarray scanner (DNA Microarray Scanner with SureScan High-Resolution Technology, Agilent Technologies, an array feature extraction program (Feature Extraction Software 10.7.1.1, Agilent Technologies, and a data evaluation software package (Genomic workbench 6.5, Agilent Technologies). Deletions and gains were recorded if the Cy5/Cy3-fluorescence ratio exceeded ±0.5 log2 in one probe. Deletions were considered as homozygous if the Cy5/Cy3 fluorescence ratio exceeding -1 log2 in at least five consecutive probes. Gains were defined as amplifications if the Cy5/Cy3 fluorescence ratio exceeded +2 log2 in at least five consecutive probes. As basic adjustment, the ADM-2 algorithm of the data evaluation software was used with a threshold of 6.0.
For gene expression analysis by microarray, total RNA preparations were checked for RNA integrity using the Agilent 2100 Bioanalyzer. All samples in this study showed high quality RNA Integrity Numbers (RIN 10). RNA was quantified by photometric Nanodrop measurement. Synthesis of cDNA and subsequent biotin labeling of cRNA was performed according to the manufacturer’s protocol (3’ IVT Express Kit; Affymetrix, Inc., City, Country). Briefly, 100 ng of total RNA were converted to cDNA, followed by in vitro transcription and biotin labeling of aRNA. Fragmentation labeled aRNA was hybridized to Affymetrix PrimeViewTM Human Gene Expression Microarrays for 16 h at 45°C, stained by a streptavidin/phycoerythrin conjugate and scanned as described in the manufacturer’s protocol.
Data analyses on Affymetrix CEL files were conducted with GeneSpring GX software (Vers. 12.5; Agilent Technologies). Probes within each probeset were summarized by RMA after quantile normalization of probe level signal intensities across all samples to reduce inter-array variability (13). Input data pre-processing was concluded by baseline transformation to the median of all samples. After grouping of samples according to their respective experimental condition (UP236 vs BC61, four replicates each) a given probeset had to be expressed above background (i.e. fluorescence signal of a probeset was detected within the 20th and 100th percentiles of the raw signal distribution of a given array) in all four replicates in at least one of two conditions to be further analyzed in pairwise comparisons. In keeping with the analysis below, differential gene expression was statistically determined by moderated t-test, using RStudio Bioconductor package limma (14, 15). Resulting p-values were corrected for multiple testing by the Bonferroni method (16). The significance threshold was set to p(corr)=0.05. All genes are named according to the official gene symbols.
Gene expression analyses of public datasets. The six datasets used were downloaded from Gene Expression Omnibus (GEO) and European Bioinformatics Institute (EBI). Information about these datasets (17-23) is listed in Table I. All genes were investigated for differential expression of urothelial tumor vs. normal tissues by a moderated t-test, using the R package limma (14, 15). Genes showing nominal p-values <0.05 in at least four datasets were considered as “commonly altered in UC”. Comparisons with pTa and ≥pT2 tumor tissues were performed analogously. All genes are named according to the official gene symbols.
Public microarray expression datasets used in our analysis. References are given in the first row. The Geo Accession codes in the second row identify the datasets in the Gene Expression Omnibus (GEO) and European Bioinformatics Institute (EBI) databases. Total sample numbers with the distribution between normal and cancer tissues are listed in row three. The last row gives the microarray platforms used.
Homozygous deletions and amplifications in BC44. Homozygous deletions (negative values) and amplifications (positive values) in the BC44 genome identified by aCGH. Chromosomes, cytobands and location of the starting base of the first and last affected probe (hg version 19) are given as well as the number of affected probes. The last column gives the average log ratio of the copy number changes.
Results
Detailed analysis of chromosomal changes in BC44 and BC61. The karyotypes and gross copy number changes of the BC44 and BC61 cell lines have been previously described (11). In brief, BC44 is aneuploid with a pseudotriploid (female) genome displaying major rearrangements of several chromosomes including chromosomes 1, 3, 4, 6, 7, 8, 16, 17, 18 and 19. BC61 has a tetraploid (male) karyotype with a more limited number of chromosomal changes, most of which are typical of progressive papillary urothelial carcinomas, such as loss of chromosome 9 and chromosome 11, isochromosome 5p and gain of chromosome 20. We have now performed a detailed analysis of the aCGH data for these cell lines in order to define the genes affected by amplifications or deletions in either cell line as well as in both.
Applying stringent criteria, presumable homozygous deletions in BC44 are detected at 3p12.3, 7q31.1, 8q11.21, 8p23.3 - p23.2, 9p21.3, 15q11.1 - q11.2, Xq28 and Xq21.33 (Table II). As commonly observed in cancer, many homozygous deletions contain no or few protein-coding genes (24). The prime exception is the 9p21 deletion (Figure 1A) including the tumor suppressor genes CDKN2A and CDKN2B, the most frequent target of homozygous deletions across all human cancers (25). In BC44 this deletion extends across 1.94 Mbp. A large deletion of the terminal 8p region encompassing a large number of genes, including several tumor suppressor candidates, is actually rather monosomic in the approximately triploid genome. The most clear-cut amplification unit in BC44 involves a region at 11q13.2 - q13.3 of approximately 2.03 Mbp centered around the CCND1 gene encoding Cyclin D1 (Figure 1B, Table II).
Detailed structure of homozygous deletions and amplifications on chromosome 9 and 11 in BC44 and BC66. Affected genes are indicated within the blue boxes and the next unaffected adjacent genes are marked outside each box. A: Chromosome 9 in BC44. B: Chromosome 9 in BC61. C: Chromosome 11 in BC44. D: Chromosome 11 in BC61.
Homozygous deletions and amplifications in BC61. Homozygous deletions (negative values) and amplifications (positive values) in the BC44 genome identified by aCGH. Chromosomes, cytobands and location of the starting base of the first and last affected probe (hg version19) are given as well as the number of affected probes. The last column gives the average log ratio of the copy number changes.
By the same criteria, homozygous deletions in BC61 are detected at 1q31.3 - q32.1, 2q21.2, 3q13.31, 3p26.1, 4q13.3, 9p21.3, 11p13 and Xq26.2; amplifications are found at 11q13.2 - q13.3, 11q14.1 and 12q15 (Table III). Notably, despite its overall less aberrant karyotype, the number of focal deletions and amplifications is not appreciably lower in BC61 than in BC44 (Table III). In addition, all regions with homozygous deletions actually contain protein-coding genes. Notably, the deletion at 9p21.3 of about 1.19 Mbp and the amplification at 11q13.1-13.2 of about 2.53 Mbp are common to both cell lines, albeit differing in extension (Figure 1C, D). The additional amplification unit at 12q15 in BC61 is particularly interesting as it contains mouse double minute 2 homolog (MDM2).
We then applied less stringent criteria to identify genes affected by chromosomal gains and losses in the two cell lines. This procedure yielded 1075 (BC61) and 305 (BC44) genes, respectively, subject to loss and 551 (BC61) and 2001 (BC44) genes, respectively, subject to gains. Of these, 11 genes were lost and 115 were gained in both cell lines (Table IV).
Comparison of chromosomal changes in BC44 and BC61 to common expression changes in UC. In order to investigate whether genes affected by copy number changes in the two cell lines are regularly deregulated in bladder cancers, 6 published expression microarray datasets were analyzed (Table I). In the first step, all genes were selected from each individual array that showed significantly stronger or lower expression in tumors compared to normal tissues, without adjustment for multiple testing. Instead, in the second step we considered those genes that were significantly changed (p<0.05) in at least 4/6 arrays as “commonly deregulated in bladder cancer”. Based on all platforms, 349 genes were commonly down-regulated and 230 genes were commonly up-regulated. Of note, the Lindgren et al. (50) study used a different array platform. Omitting this dataset therefore leads to a larger set of commonly deregulated genes.
According to an analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.7 bioinformatics tool (26, 27), the genes commonly down-regulated in UC were particularly over-represented in the Gene Ontology groups “response to endogenous stimulus” (padj<10-5) and in its subgroup “response to steroid hormone stimulus” (padj<10-5). A disproportionate fraction of the down-regulated genes encodes plasma membrane proteins (padj=10-6). In addition, several under-expressed genes regulate cell death or apoptosis (padj=10-2). The functions of the commonly up-regulated genes were less diverse. Interestingly, however, almost all of the significantly enriched GO groups related to cell cycle, mitosis, DNA replication and repair (padj=10-6).
Genes commonly affected by deletions or gains in BC44 and BC61. Genes commonly affected by deletions or gains in BC61 and BC44 identified by aCGH. The columns list the genes exhibiting concordant changes in both cell lines.
Comparing the genes “commonly deregulated in bladder cancer” to those accordingly lost or gained in BC44 or BC61 gave sizable sets of overlap, shown in Table V.
We then performed the same analysis for the microarray datasets considering only differences between muscle-invasive cancers (≥pT2) and normal tissues or considering only differences between non-invasive papillary tumors (pTa) and normal tissues. Twenty-three genes remained significant across all analyses (Table VI). These genes are predicted to become gradually up- or down-regulated during UC progression.
Comparison of genomic and expression changes in BC61. In a final series of analyses, we addressed the question of how numerical chromosomal changes in BC61 relate to changes in gene expression. For that purpose, an expression microarray analysis was performed comparing BC61 with normal urothelial cells in primary culture (UP), which have a normal genome. Of note, these cells proliferate at a similar rate as UC cell lines (28). Four independent RNA preparations from each cell type were used and applied to Affymetrix PrimeView™ Human Gene Expression microarrays.
Following normalization and centering of RNA expression data, 39,367 out of 49,394 probes on the array were evaluable. After Bonferroni-adjustment for multiple testing, 8,755 probes were differentially expressed at p<0.05 by moderated t-test. Of these, 4,558 probes corresponding to 2,906 genes were more strongly expressed in BC61 and 4,558 probes corresponding to 2,614 genes were more strongly expressed in UP. The 100 genes with the highest fold changes are listed in table VII.
The differentially expressed genes were analysed for systematic differences using the DAVID bioinformatics platform (26, 27). The genes with lower expression in BC61 were highly significantly localized on chromosomes 9 (230 genes), 5 (171 genes), 10 (184 genes) and 16p13.3 (38 genes). In functional terms, they were particularly enriched in genes involved in RNA biogenesis and metabolism, including cellular and mitochondrial ribosomal biogenesis. This enrichment could partly, but not completely, be explained by the localization of the relevant genes in chromosomal regions lost in BC61. Among the up-regulated genes, 187 were localized on chromosome 20, making localization on this chromosome a striking characteristic. In functional terms, transcription factors, activators and repressors, and especially zinc finger proteins were significantly over-represented.
Genes within chromosomal regions of deletions or gains in BC44 or BC61 with differential expression between tumor and normal tissues in six publicly available RNA expression datasets. Deletions and gains were identified by aCGH. RNA expression: up- or down-expressed genes in urothelial cancer tissue, with a p-value <0.05 (moderated T-test) in the comparison of urothelial cancer of all stages with normal urothelial tissue. The underlined genes are also differentially expressed in BC61 compared to normal urothelial cells.
This analysis indicates that differences in gene expression between BC61 and normal urothelial cells are dominated, albeit not fully determined by chromosomal gains and losses. For a systematic analysis of the relation between chromosomal changes and gene expression changes, the gene lists obtained by aCGH and expression microarray analysis were compared. Indeed, whereas 310 genes with lower expression in BC61 had lower copy numbers, only 30 had increased copy numbers. Conversely, 237 genes with higher expression in BC61 had increased copy numbers, but only 48 had lower copy numbers. However, it should be noted that a large number of genes showed expression changes without copy number changes and conversely, a large number of genes with copy number changes remained steady in expression.
Genes affected by deletions and gains in BC44 or BC61 with differential expression across all tumor stages in the six public RNA expression files. These genes are a subset of those in Table IV. Their expression differs significantly not only between normal and cancer tissues, but also between normal tissue and pTa tumors as well as between normal tissue and muscle-invasive carcinomas. The underlined genes are additionally differentially expressed in BC61.
Discussion
The technical progress in next-generation sequencing has allowed large-scale analyses that are expected to lead to a comprehensive catalogue of genomic changes in all human cancers in the near future (8). In the context of these projects, several consortia are addressing urothelial cancer. Among others, two papers by a Chinese consortium (5, 6) delineate mutations and copy number changes in a large (99 cases) series of UC. A comprehensive analysis of invasive UC in US-American patients has recently been published by the TCGA consortium(8). In addition, several European groups have published large sets of expression data and copy number changes (3, 4), some of which have been used in the present analysis.
The task at hand in bladder cancer research is now the interpretation of this huge amount of data, with the ultimate aims of understanding the biological mechanisms driving the development and progression of this common cancer, and of translating the acquired knowledge into improved diagnostic and therapeutic procedures. One important step towards this ultimate goal, is the identification of relevant genes targeted by copy number changes, which often encompass large regions of the genome. Our findings in two newly established bladder cancer cell lines can serve as an example to this end.
Despite their quite different origin, both cell lines had in common one amplification at 11q13 and a homozygous deletion at 9p21 that is thought to primarily target the established tumor suppressor genes CDKN2A and CDKN2B. Of note, other genes in the region may also be relevant, e.g. ELAVL2 [3]. The amplification unit at 11q around the CCND1 gene, likewise, is actually quite large and affects several genes, some of which are indeed overexpressed in BC61. Regardless, the coexistence of these two alterations is a bit puzzling since the products of CDKN2A, CDKN2B (p16INK4A and p15INK4B) and of CCND1 (Cyclin D1) act in the same pathway, namely regulation of CDK4/6. Both inactivation of CDKN2A, most frequently by homozygous deletion, and amplification of CCND1 are common alterations in UC. Further common changes in this pathway include amplification and overexpression of E2F1 and E2F3, and loss of pRB1 function through gene mutation and deletion (4, 6, 8, 19). RB1 mutations are complementary to CDKN2A, but are often accompanied by amplification and overexpression of E2F3, another apparent redundancy in need of explanation (4, 29). Accordingly, only E2F1, but not E2F3 or RB1, showed conspicuous changes in the two new cell lines.
A strong argument can be made that the establishment of bladder cancer cell lines requires the inactivation of the RB1/p16INK4A/CDK4 circuit that controls the cell cycle in G1 and may prevent immortalization (30). In a similar fashion, almost all bladder cancer cell lines have defects in p53, mostly by mutations coupled with loss of one allele of the TP53 gene located at 17p. The concomitant inactivation of both control systems seems to be characteristic of invasive bladder cancers (1, 31). Alternative to mutations of the TP53 gene itself, p53 function may become compromised by loss of p14ARF1, encoded by CDKN2A in an alternative reading frame, or by MDM2 amplification. In the BC44 cell line, p53 is inactivated by an unusual mechanism, present in a small fraction of bladder cancers, which leads to the sole expression of protein p53Δ133 isoforms lacking the amino-terminal transactivating domain (11). The discovery of a MDM2 amplification in BC61 suggests an explanation of how this cell line copes with the presence of wild-type p53. Interestingly, a similar constellation has been reported in the cell line BFTC905, which was derived from a pTa high grade UC (32). Indeed, in the analysis of public microarray datasets, MDM2 overexpression is only nominally significant in the comparison of pTa tumors with normal tissues. Several dedicated studies confirm that MDM2 overexpression is more often associated with pTa and pT1 tumors (4, 31, 33, 34).
Top 100 genes expressed more strongly in BC61 or in normal uroepithelial cells (UP) by fold change. All genes were significantly differentially expressed according to Bonferroni-corrected moderate T-test.
Of note, the amplification unit at 12q contains a number of genes in addition to MDM2 that are overexpressed in BC61. The same is true for the 11q amplification unit. For instance, our analyses revealed PPFIA1 as a gene with overexpression gradually increasing with UC progression. Coamplification of PPFIA1 with CCND1 has previously been noted in other cancer types (35, 36). In addition, the amplified region encodes the MiR-548k gene. MicroRNAs are now recognized as important factors in tumor development (37). Our study highlights the genes encoding miR-644 and miR-548k. MiR-548k at 11q13 is amplified in both cell lines and its product interacts with many genes (targetScanHuman) that are important for cell metabolism and survival, whereas miR-644 may regulate housekeeping genes like GAPDH and could be a confounder of measurements using these as a reference (38). To our knowledge, neither miRNA has been studied in detail in UC to date.
In addition to the 9p21 region, several further homozygous deletions were detected in either BC44 or BC61. In principle, homozygous deletions in a cancer could signal the location of a potent tumor suppressor, but - paradoxically - could alternatively occur at regions that are completely dispensable for tumor cell survival, such as “gene deserts”. The latter seems to be the case in BC44, whereas true homozygous deletions in BC61 do contain defined genes. However, none of these was recurrently down-regulated in the microarray analyses and they are, therefore, unlikely to represent a major tumor suppresor gene. The issue is different for the distal part of chromosome 8p, which is strongly under-represented in BC44, but not truly homozygously deleted. This region has long been implicated in UC progression (39). It contains a large number of plausible tumor suppressor candidates, but none has been identified as individually critical in UC. According to our analysis, expression of a number of genes located in 8p21-23.2 (ADAM28, HR, PHYHIP, BMP1, PPP3CC, SORBS3, PCM1, SLC7A2, KBTBD11) is commonly diminished. Collectively, our analysis supports the proposition of Williams et al. (39) that loss of several genes at 8p may contribute to urothelial carcinogenesis. Accordingly, collaboration of several tumor suppressors at 8p22 has recently been reported to contribute to hepatocarcinogenesis (40).
The published large-scale mutational analyses (6, 8) suggest particular oncogenes or tumor suppressors apparently targeted by other chromosomal changes common to BC44 and BC61. However, many chromosomal changes encompass large regions, as seen prominently in the two cell lines. In addition, point mutations are not the only indicators of genes involved in carcinogenesis and are, in fact, sometimes difficult to interpret. In general, however, one would expect oncogenes located in a region undergoing gain or amplification to be overexpressed and conversely, tumor suppressors affected by deletions to be down-regulated. The comparison of gene expression and copy number changes in BC61 illustrates this point. Thus, a much higher fraction of genes located in gained chromosomal regions were up-regulated than down-regulated and conversely, a much lower fraction located in regions undergoing loss were down-regulated than up-regulated. However, the majority of genes located in such regions did not show significant or large changes in gene expression. Thus, filtering for genes undergoing expression changes in the expected direction should restrict the number of potential pro- and anti-oncogenic genes within a particular chromosomal region.
Indeed, considering the genes located in regions of gain or loss in the two cell lines (beyond those affected by true amplifications or homozygous deletions) and searching for those that were deregulated in several published microarray datasets, yielded a list of candidates in particular for chromosomal regions that are not as well-explored in UC, albeit commonly affected by copy number changes (Table V). Such regions include 8p for deletions and 1q, 5p, 19q and 20q for gains. In addition, the comparison of published expression data sets and copy number changes in BC44 and BC61 highlighted 23 genes (Table VI) that appeared to become up- or down-regulated in a gradual fashion during UC progression.
Despite many provisions from biological and statistical arguments applying to this sort of analysis, several genes identified are good candidates for further investigation in UC. First, the products of CDC2 (CDK1) and SKP2 are both involved in the regulation of cell cycle transition from late G1 to S-phase, like E2F1 discussed above. Indeed, CDC2 has already been implicated in the progression of UC (41). In contrast, the F-Box protein SKP2, which is involved in the degradation of several important proteins including p27KIP1, a crucial antagonist of Cyclin E1, is poorly studied in this cancer type, despite its importance in others (42). Collectively, these findings indicate that the disturbances in the regulation of cell cycle in UC are not limited to the early G1 phase, i.e. by changes in Cyclin D1 and p16INK4A. Interestingly, the SKP2 gene is located on chromosome 5p, which is frequently gained in UC, often by formation of i5p with concomitant loss of the long arm. This chromosome arm contains other genes known to contribute to UC, prominently the telomerase gene TERT, which may explain why i5p is prevalent in papillary tumors. Conceivably this common gain may target both TERT and SKP2. The nucleoporin gene NUP155 at 5p counts among the gradually up-regulated genes.
Secondly, the list contains several genes deleted in common regions of deletion. At 8p21-22, SORBS3 and PPP3CC are particularly notable. SORBS3 encodes Vinexin, which is involved in focal and cell-cell adhesion. So far, it has only been studied in hepatocellular carcinoma (43). PPP3CC encodes a regulatory subunit of the calcium-dependent protein phosphatase calcineurin. Calcineurin limits stem cell potential in the epidermis (44), but it is unknown whether it functions in the same manner in the urothelium. Interestingly, down-regulation of calcineurin occurs during bladder hypertrophy (45). Interestingly, as a regulator of non-canonical WNT-signaling, the product of the ROR2 gene at 9q, commonly and progressively down-regulated in UC, may act upstream of calcineurin (46)
Further genes highlighted on chromosome 9q include SPTAN1, ZFAND5, GNAQ, TLE4, CDC14B, TMOD1, GSN, CDK5RAP2, TRAF1 and USP20, which have not yet been explored in the context of UC. A few, like NTRK2 and TGFBR1, have already been investigated because of their function in regulating GSK3 and TGFβ signaling, respectively (47, 48). Deletions at 9q are among the most frequent chromosomal changes in UC, but neither their targets nor their impact is fully understood despite intense research (31). Overall, it is likely that the deletions target several genes in a synergistic manner. The same argument may even hold for chromosome 10q, where PTEN is an established target for deletions that are typically associated with tumor progression (31). Our present analysis highlights FAS (also known as TNFRSF6), a death receptor known to become progressively dysfunctional in UC and contribute to apoptosis resistance (49, 50) as a further target.
The third group of genes deserving explicit discussion is that on chromosome 20. Gain of chromosome 20 is common in UC, but it is only partly understood which genes are targeted. One interesting candidate is the anti-apoptotic gene BCL2L1, for which focal amplifications have been reported recently (8). Two additional chromosome 20 genes identified here are ERGIC3 and TPX2. ERGIC3 is an ER protein implicated in the development and progression of lung cancer (51). TPX2 (TPX2, microtubule-associated) is commonly overexpressed in UC and the gene is gained in both cell lines. TPX2 is an oncogene candidate in several cancers, likely acting through its effects on mitosis and the spindle apparatus. There is direct evidence that it contributes to bladder cancer progression (52).
Table VI lists several genes that are becoming progressively deregulated in UC, which have already been linked to phenotypes involved in tumor progression, KCNH2 (53), GSN (54), LSR (55), and FAS (49), with specific data in UC for the latter three. Furthermore, NCSTN and PARP1 are gained in BC44 and commonly overexpressed in UC. They are located at chromosome 1q, where, in UC 1, q22-24 is often affected by gains or amplifications (4, 56, 57). NCSTN encodes nicastrin, a component of γ-secretase, which processes the amyloid precursor protein, but also, presumably more relevant to cancer, notch proteins. Notch activation is pro-tumorigenic or tumor-suppressive in a cell-type dependent fashion (58); in UC, there is insufficient data to date (59, 60). PARP1 encodes the poly (ADP-ribose) polymerase I which influences many cellular processes, but is best known for its function as a DNA repair cofactor and as a therapeutic target (61, 62). Its function in UC certainly deserves detailed investigation. Another such candidate is WWOX (WW domain-containing oxidoreductase), which is deleted in BC44 and commonly down-regulated in UC. The gene is located at a fragile site in 16q23.3-24.1 and likely acts as a tumor suppressor in many cancers (63).
Finally, a number of genes at chromosome 19q13 gained in BC44 are up-regulated in UC. Of these, LSR and FXYD3 have been studied. The FXYD3 protein, a regulator of ion transport across cellular membranes, is overexpressed in the majority of UC of the upper genitourinary tract and in high grade bladder cancers (64). The “lipolysis stimulated lipoprotein receptor” encoded by LSR is often up-regulated in UC. Its knockdown by siRNA significantly increased the motility and invasion capacity of bladder cancer cells (55). Notably, in breast cancer LSR displayed opposite properties (65). These discrepant findings along with the results of our analysis indicate a need for further investigations of LSR in cancer.
In conclusion, our study provides several novel candidate genes in chromosomal regions with recurrent gains or losses in UC in addition to already well-studied genes. In particular, it highlights several regions that have not been subjected to close scrutiny yet, such as 5p or 20q, that obviously deserve even more attention, like 8p. Moreover, our analysis supports the idea that many chromosomal changes in UC target several genes with additive or synergistic effects on tumor development and progression.
- Received April 15, 2014.
- Revision received May 27, 2014.
- Accepted May 28, 2014.
- Copyright© 2014, International Institute of Anticancer Research (Dr. John G. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- Association Between BMI and DNA Methylation in Blood or Normal Adult Breast Tissue: A Systematic Review
- miR-187* Enhances SiHa Cervical Cancer Cell Oncogenicity Via Suppression of WWOX
- Genetic Variations in the 3-untranslated Regions of Genes Involved in the Cell Cycle and Apoptosis Pathways Affect Bladder Cancer Risk