Urothelial carcinoma (UC) is characterized by nonrandom chromosomal aberrations, varying from one or a few changes in early-stage and low-grade tumors, to massively rearranged karyotypes in muscle-invasive lesions 1. Rearrangements of chromosome 9, resulting in loss of chromosome 9 material, are the most common cytogenetic alterations and are seen in close to half of the cases. Loss of material from chromosome arm 11p, and gains of chromosome 7 and chromosome arm 1q seem to be early but secondary changes. UC is also characterized by a strong association between tumor stage/grade and karyotypic complexity, indicating that progressive accumulation of genetic alterations is the driving force behind multi-step bladder carcinogenesis 2. Cytogenetic data have been corroborated by comparative genomic hybridization (CGH) investigations 345, which in addition have been valuable for identification of high-level amplifications. Recent array CGH analyses of bladder tumors 67 and cell lines 8, based on non-tiling bacterial artificial chromosome (BAC) arrays have shed further light on the genomic changes underlying neoplastic development. These studies have facilitated the molecular delineation of additional amplified and deleted regions to the level of specific candidate genes. Recurrent high-level local amplifications have so far been reported in many genomic regions, including 1q21-24, 3p22-24, 6p22, 8q21-22, 10p13-14, 11q13, 12q13-15, and 17q21 345678. To provide a more complete and detailed map of previously observed amplified and deleted segments, and to identify new copy number aberrations (CNAs) of importance to urothelial tumorigenesis, we used whole-genome tiling-resolution array CGH based on a 32 k BAC clone set to analyze 38 primary urothelial carcinomas. In order to allow a comprehensive analysis also of transcriptional effects of the identified CNAs, the copy number analysis was complemented with expression profiling using high-density (36 k) oligonucleotide micro arrays. With this approach we present genomic borders and putative target genes for 9 recurrent amplifications, some of which specific for advanced tumors, and for 7 previously not described homozygous deletions, of which one is recurrent. We also studied the pattern of CNA occurrence, and show that 6p22 amplifications are not independent of homozygous deletions in 9p21, and that amplification of paralogous segments in 8q22 and 2p25 appear to act complementary and may have equivalent biological outcomes as they both result in increased expression of members of the YWHA genes, encoding 14-3-3 proteins.

Results from the genome-wide DNA copy number analyses of the 38 tumors are shown in Figure 1A (the complete data set is available in Additional file 2). Low and medium grade (G1/G2) tumors generally contained few changes (Figure 1B), paralleled by similar low level of complexity in Ta tumors (Table 1). T1 tumors showed increased frequencies of genomic imbalances, and T2-T4 tumors showed even more complex aberrations with as many as 24 different frequent (> 30%) imbalances. Losses and homozygous deletions were seen at similar frequencies in low/medium and high grade tumors, whereas gains and particularly amplifications were considerably more frequent in G3 tumors (Figure 2); high grade tumors on average showed more than 10 times the number of amplifications as low/medium grade tumors (4.2 vs. 0.3). The most frequent amplifications were localized to 6p22 (8 cases) and 8q22 (5 cases), respectively, both observed exclusively in high-grade tumors. Because only one of the tumors showed co-occurrence of the two events, at least one of these aberrations were present in more than 50% of the G3 tumors (12 of 23 cases).

Apart from the two major targets mentioned above, seven additional segments showed recurrent (≥ 3 cases) overlapping amplifications; 1p34, 1q23, 2p25, 3p25, 10q22, 11q13, and 16p13 (Table 2). The most common of these, 1q23, spanned a 5.4 Mb region with a commonly amplified segment of 0.8 Mb. The genomic profiles of the individual cases were very diverse e.g., case 27, with the largest amplification showed the highest amplifications levels in a region 1.2 Mb proximal to the commonly amplified segment, whereas case 28 showed peak log₂ ratios on the distal side. These findings indicate that 1q23 amplifications are heterogeneous and may represent different genomic events with no common target region. Putative target genes in the other recurrent core amplicons are listed in Table 2. The 11q13 amplification, including CCND1, was the only genomic amplification that was recurrent in G1/G2 tumors. A statistical analysis (SAM) confirmed that local amplification of 1q23, 6p22, and 8q22 were significant discriminators between G1/G2 and G3 tumors (data not shown).

We performed whole-genome tiling-resolution array CGH analyses to characterize gene copy number aberrations in 38 cases of urothelial carcinomas. To make a comprehensive evaluation of the consequences at the expression level the array CGH data was compared with expression data obtained by high density expression profiling. Low grade and early stage tumors showed a limited number of genomic imbalances with gain of 1q and loss of chromosome 9q as the most frequent changes. T1 tumors harbored additional changes and frequently showed losses of 6q, 8p, 9q, 11p, 17p, 19p, and 22q, and gains of 1q, 5p, 8q, 10p, and 13, and 5p. Late-stage tumors, T2-T4, showed several additional genomic imbalances and showed complex genomic changes. A major difference was seen between G1/G2 and G3 tumors with regard to gains and amplifications. Gains were more frequent in G3 tumors and amplifications more than 10 times as frequent in high grade than in low/medium grade tumors. This is in line with the recent finding that high grade tumors show frequent anaphase bridges that ultimately may lead to breakage-fusion-bridge cycles producing genomic amplifications 19. In contrast to Blaveri et al. 7, we did not see a reduction in the number of affected regions in late-stage tumors. Instead T2-T4 tumors showed considerably more complex changes than Ta/T1 tumors, corresponding to previous cytogenetic and conventional CGH results that have shown a larger number of genomic alterations in high-grade and late-stage tumors 23420.

The analysis showed that amplifications of 1q23 were highly heterogeneous. Individual cases showed extended amplicons with peak copy number ratios in different areas and with no obvious commonly amplified segment. Gain of chromosome arm 1q is frequent in UC 12 and local amplifications in 1q23 have previously been reported in studies using conventional CGH 35. However, due to limited resolution in traditional CGH analysis diverse amplifications within a 5 Mb genomic region may have been interpreted as a common genomic event and hence assumed to target the same genomic segment. Interestingly, the present tiling resolution analysis indicates that this is most likely not the case. There are at least two possible explanations for these findings; either the 1q23 region harbors more than one target segment, or the 1q23 region show local genomic instability resulting in frequent gains or high level amplifications. Both situations would result in a heterogeneous pattern of changes.

The most frequently amplified chromosomal segment was 6p22. This segment has been repeatedly shown to be amplified in UC and to cover the genes E2F3, CDKAL1, and SOX4 682122232425. In the present investigation a 1.8 Mb core region, included in 88-100% of the cases with 6p22 amplifications, was defined. This region contains MBOAT1, E2F3, CDKAL1, and SOX4. The core segment was flanked by two regions included in 38-75% of the amplifications that on the distal side contained ID4, and on the proximal side PRL and HDGFL1. A further distally located region, present in three (38%) of the cases with amplification, contained ten additional genes. In total a more than 6 Mb region was found to be repeatedly included in the 6p22 amplifications. To identify possible targets genes within this segment we used different criteria based on available expression data. Six genes; NUP153, AOF1, DEK, SOX4, FAM8A1, and MBOAT1 showed top ranking expression levels when amplified. By Q-PCR we also established a strong association between E2F3 and CDKAL1 gene expression and gene copy numbers. This identifies eight possible candidate genes in 6p22 that may be ranked according the frequency by which they take part in 6p22 amplifications as CDKAL1, E2F3, and SOX4 > MBOAT1 > DEK, NUP153, FAM8A1, and AOF1.

Several investigations have shown a strong association between gene copy numbers and gene expression for E2F3 and CDKAL1 82122. In two recent studies knock down experiments of E2F3 in bladder cancer cells lines with 6p amplification was shown to strongly reduce cellular proliferation whereas a similar knock down of CDKAL1 did not have this effect 2627. This indicates E2F3 as a major target for the 6p22 amplification. The E2F3 locus produces two alternative proteins, E2F3a and E2F3b 28. E2F3a is an inducible activator of cell cycle progression whereas E2F3b is constitutively expressed and functions as a negative regulator of p14^ARF expression, thereby compromising TP53 activity 29. Genomic amplification of the E2F3 locus may consequently contribute to tumor development both by accelerating the cell cycle and by raising the threshold for apoptosis via inactivation of p14^ARF. In this context it is worth noting that none of the cases with 6p amplification showed homozygous deletions leading to loss of p14^ARF. Hence, amplification of 6p and homozygous deletion of the CDKN2A/CDKN2B locus behaves as complementary events; one leading to the simultaneous loss of CDKN2A, a cell cycle inhibitor, and p14ARF activity, and a second to increased E2F3, a cell cycle activator, and reduced p14^ARF activity. As a consequence 65% of the G3 tumors may show impaired p14^ARF function.

SOX4 belongs to the SRY-related HMG-box (SOX) family of transcription factors involved in the regulation of embryonic development and in cell fate determination. The protein acts as a transcriptional regulator in the apoptosis pathway as well as in pathways leading to tumorigenesis. SOX4 is over expressed in several tumor types 30 and was found to be significantly over expressed in bladder tumors in a recent investigation 31, suggesting a role for SOX4 in bladder cancer development. Intriguingly, both experimentally increased 31 and decreased 32SOX4 expression induces an apoptotic response. This may indicate that SOX4 expression has to be finely tuned not to induce deleterious cellular responses and may, partly, explain why no consistent correlation between SOX4 amplification and expression has been seen in previous investigations 82131.

The MBOAT1, DEK, FAM8A, AOF1, and NUP153 genes were amplified at lower frequencies than E2F3. Of these DEK, a chromatin-associated protein could provide a selective advantage when amplified as DEK function as an inhibitor of senescence 33. Furthermore, as DEK is under the transcriptional control of E2F3 34 co-amplification of DEK with E2F3 may enhance DEK expression further. NUP153 plays an important role in nuclear pore function and have key functions in both import and export to the nucleus. NUP153 is known to affect the sub cellular distribution of transcription factor such as SMAD2, STAT1, and PU.1 35, hence, altered NUP153 expression may modify the accessibility of key regulatory proteins. The possible roles of MBOAT1, FAM8A1, and AOF1 in carcinogenesis are, however, less clear. Taken together, the present comprehensive analysis of 6p22 amplification in UC indicates that SOX4, DEK and NUP153 may contribute to cellular transformation when co-amplified with E2F3, and hence that 6p22 may represent a composite amplicon with more than one potential target gene.

The core amplicon in 8q22 extended over a 1.8 Mb region and included 8 annotated genes of which YWHAZ (14-3-3-zeta) and POLR2K showed a strong association between gene copy number and gene expression. Interestingly, three tumors with no amplification of 8q22 showed amplification of a 1 Mb paralogous region in 2p25. This segment included several genes similar to genes within the 8q22 segment, such as YWHAQ, GRLH1,HCAL1, RRM2, and KLF11. Two of the paralogous genes within the 2p25 amplicon showed a strong association between gene copy numbers and gene expression, YWHAQ (14-3-3-theta) and RRM2. Hence, the common denominator of the 8q22 and 2p25 amplicons is the amplification and over expression of the paralogous 14-3-3-zeta and 14-3-3-theta genes. Furthermore, if amplification of the YWHA genes may substitute for each other, amplification of these genes is as common as amplification of E2F3.

There are seven 14-3-3 isoforms described in mammals 36. The proteins are widely expressed and bind as dimers to client proteins thereby modulating their enzymatic activity, sub cellular localization, or potential to form protein complexes; more than 200 proteins have been reported to associate with these proteins. It is believed that 14-3-3 proteins functions as general survival factors by enhancing pro-survival signaling and suppressing pro-apoptotic proteins 36. The findings that 14-3-3-zeta and -theta show increased expression in oral squamous cell carcinomas and lung, stomach, and breast cancers 37383940, and that mice transgenic for 14-3-3-zeta develop various types of tumours at young age 41 favors the conclusion that 14-3-3-zeta and -theta may act as oncogenes also in urothelial carcinomas.

Homozygous deletions were detected in six locations on chromosome 9 and in four locations in the remaining genome; 4q35, 10q23, 13q14, and 21q21. None of the observed deletions coincides with deletions frequently seen in cell lines 42 and only one, 9p21, has previously been reported for urothelial carcinomas. Among the deletions not occurring on chromosome 9 three, 4q35, 10q23, and 21q21 respectively, extended over genes not previously associated with tumor development and may merit further analyses. The 13q14 deletion, on the other hand, affected the known tumor suppressor gene RB1, and hence most likely has a pathological effect. As expected the most frequent homozygous deletions occurred at 9p21, in a 4 Mb segment that consistently included CDKN2A and CDKN2B. The observed frequency of CDKN2A homozygous deletions reached 32%, which is close to the frequency found by Q-PCR of micro dissected tumor biopsies 43. The deletions were more extensive on the proximal than on the distal side of CDKN2A which may be caused by the fact that the proximal region is less dense with genes. An intriguing finding was the clustering of homozygous deletions to chromosome 9. Homozygous deletions were seen in five chromosome 9 locations, excluding the high-frequency CDKN2A region, compared with a total of four homozygous deletions on other chromosomes. Furthermore, one case showed two homozygous deletions, and another case one homozygous deletion in addition to a 9p21 deletion. This high frequency of homozygous deletions is astonishing but is in agreement with the high frequency of LOH and chromosomal 9 losses seen in UC 44. The accumulated data reveals no specific pattern of LOH, which has led to the suggestion that most of the LOH seen in chromosome 9 may be caused by unspecific mitotic recombination 45. A possible mechanism for the frequent homozygous deletions could consequently be unequal mitotic recombination events. Such events would produce homozygous deletions after segregation of the recombined homologues, as well as segmental duplications. Indeed, possible segmental duplications on chromosome 9 were seen in a limited number of cases. Irrespective of the mechanisms causing the deletions, the high frequency of homozygous deletions at various locations on chromosome 9 point to the possibility that these deletions may have a different origin and cellular consequence than the more frequent 9p21 homozygous deletions.

The presented data indicates 6p22 as a composite amplicon with more than one possible target gene and suggests that amplification of 6p22 and homozygous deletions of 9p21 may have complementary roles. The combined data for amplified 8q22 and 2p25 paralogous genomic segments indicated that alterations of YWHA (14-3-3) genes may be important in the development of urothelial carcinomas. Furthermore, chromosome 9 show an exceptionally high frequency of homozygous deletions compared to other chromosomes.

The author(s) declare that they have no competing interests.

MHe performed the array-CGH analyses and took active part in the bioinformatical and statistical analyses, and in preparing the manuscript; DL performed the expression profiling and took active part in the bioinformatical and statistical analyses as well as preparing the manuscript, TJ performed the Q-PCR analyses; SV performed bioinformatic analyses; FL, SG, and WM provided tumor samples and clinical data; GC performed the pathological evaluations; ÅB provided the array-CGH platform and technical assistance; MHo conceived the investigation and took active part in the bioinformatical and statistical analyses as well as preparing the manuscript. All authors read and approved the manuscript.