Purpose: Bladder carcinogenesis is believed to follow alternative pathways of disease progression driven by an accumulation of genetic alterations. The purpose of this study was to evaluate associations between measures of genomic instability and bladder cancer clinical phenotype.

Experimental Design: Genome-wide copy number profiles were obtained for 98 bladder tumors of diverse stages (29 pTa, 14 pT1, 55 pT2-4) and grades (21 low-grade and 8 high-grade superficial tumors) by array-based comparative genomic hybridization (CGH). Each array contained 2,464 bacterial artificial chromosome and P1 clones, providing an average resolution of 1.5 Mb across the genome. A total of 54 muscle-invasive cases had follow-up information available. Overall outcome analysis was done for patients with muscle-invasive tumors having “good” (alive >2 years) versus “bad” (dead in <2 years) prognosis.

Results: Array CGH analysis showed significant increases in copy number alterations and genomic instability with increasing stage and with outcome. The fraction of genome altered (FGA) was significantly different between tumors of different stages (pTa versus pT1, P = 0.0003; pTa versus pT2-4, P = 0.02; and pT1 versus pT2-4, P = 0.03). Individual clones that differed significantly between different tumor stages were identified after adjustment for multiple comparisons (false discovery rate < 0.05). For muscle-invasive tumors, the FGA was associated with patient outcome (bad versus good prognosis patients, P = 0.002) and was identified as the only independent predictor of overall outcome based on a multivariate Cox proportional hazards method. Unsupervised hierarchical clustering separated “good” and “bad” prognosis muscle-invasive tumors into clusters that showed significant association with FGA and survival (Kaplan-Meier, P = 0.019). Supervised tumor classification (prediction analysis for microarrays) had a 71% classification success rate based on 102 unique clones.

Conclusions: Array-based CGH identified quantitative and qualitative differences in DNA copy number alterations at high resolution according to tumor stage and grade. Fraction genome altered was associated with worse outcome in muscle-invasive tumors, independent of other clinicopathologic parameters. Measures of genomic instability add independent power to outcome prediction of bladder tumors.

Bladder cancer comprises of highly heterogeneous tumors. Approximately 70% of tumors are superficial (stages pTa, pT1, pTis), whereas the remaining are muscle infiltrating (stages pT2-pT4) at the time of initial presentation. Muscle invasive tumors are thought to arise de novo or within an area of carcinoma in situ, as well as occasionally developing from previously superficial tumors. Patients with superficial and invasive tumors have remarkably different 5-year survival rates (1) reflecting their biological differences.

Like other solid tumors, it has been suggested that this morphologic heterogeneity originates from underlying genetics leading to diverse pathways of tumor development and progression. There are a number of studies indicating that there is considerable variability in the degree to which bladder tumors are altered at the chromosomal level and the spectrum of alterations can vary depending on the grade of differentiation and the tumor stage.

Cytogenetic and loss of heterozygosity studies of bladder carcinomas and cell lines have revealed a number of recurrent genetic aberrations including amplifications or gains on 8q22-24, 11q13, 12q14-15, 17q21, and losses on chromosomes 9, 8p22-23, 13q, and 17p (2, 3). Some of these aberrations have been associated with the pathologic stage and/or outcome of bladder cancer. The advent of array-based comparative genomic hybridization (CGH) has allowed the high-throughput mapping of DNA copy number alterations in bladder cancer (4, 5) and in other tumors (6, 7). The patterns of copy number alterations identified by array CGH have been reported to aid in differentiating tumors into more biologically and clinically relevant subtypes and the higher resolution has led to precise mapping of the boundaries of amplified and deleted regions indicating candidate genes relevant to cellular control pathways.

Array CGH allows a unique view of the genomic instability that a tumor has undergone before diagnosis. Both the amount of genomic copy number alteration and the specific loci involved are defined in one analysis. Different representations of instability, such as whole chromosome gain or loss, partial or complete chromosome arm changes, high-level amplification, and homozygous deletions, are easily quantified. The type, degree, and locations of these changes may have prognostic and therapeutic implications.

In this study, we report array-based CGH analysis in a set of 98 transitional cell carcinomas of the bladder of different stage, grade, and outcome. A high frequency of DNA copy number alterations was detected involving regions previously documented in bladder cancer. Most importantly both the overall frequency of alterations and the frequency of individual clones were significantly different between tumors with different clinicopathologic characteristics. The fraction of genome altered (FGA) was significantly associated with overall outcome in patients with muscle-invasive disease.

Samples and DNA preparation. Freshly frozen tissue was collected from 98 patients with bladder cancer from the University of California San Francisco Medical center from 1991 to 2003. Patients provided signed informed consent for participation in this study that was approved by the University of California San Francisco institutional review board. For each case, all blocks from the surgical specimen were reviewed to determine pathologic grade, stage, and histologic tumor type. Tumors were staged according to the American Joint Committee on Cancer (8) and graded according to the WHO and International Society for Urological Pathology classification system (9).

Before DNA extraction, an initial H&E-stained frozen section was reviewed to assess tumor quality and content. Normal and necrotic tissue were excluded by trimming of the frozen block. A tumor sample was considered suitable for study if the proportion of tumor cells was higher than 70%. DNA was extracted from tissue using standard proteinase K digestion followed by phenol/chloroform extraction (protocol available at http://cc.ucsf.edu/people/waldman/Protocols). Normal genomic DNA was obtained from Promega (Madison, WI) and was used for reference.

Array comparative genomic hybridization. Human array 2.0 chromium surface arrays were provided by the University of California San Francisco Cancer Center Array Core. Each array consisted of 2,464 bacterial artificial chromosome and P1 clones printed in triplicate, providing representation of the entire genome at a 1.5 Mb resolution (10).

A total of 500 ng of each sample DNA was random prime labeled using the BioPrime kit (Invitrogen, Carlsbad, CA; protocol available at http://cc.ucsf.edu/people/waldman/Protocols). Tumor DNA was labeled with FluoroLink Cy3-conjugated dUTP and normal genomic DNA was labeled with FluoroLink Cy5-conjugated dUTP (Amersham Pharmacia, Piscataway, NJ). Excess primers and nucleotides were removed using Sephadex G50 columns (Amersham Pharmacia). Labeled tumor and reference DNA were combined with 100 μg Cot1 DNA (Invitrogen) and the mixture was precipitated using 3 mol/L sodium acetate. Precipitated sample was dissolved in a solution containing 600 μg yeast tRNA, 9% SDS, 50% formamide, 10% dextran sulfate, and 2× SSC. The DNA probe was then denatured at 73°C for 15 minutes, cooled at 37°C for 30 minutes, and applied to the array. The slide was finally placed at 37°C for 48 hours in a sealed hybridization chamber containing 50% formamide/2× SSC for humidity. After hybridization, each slide was washed with 50% formamide and 2× SSC twice at 45°C for 12 minutes, followed by two washes in phosphate buffer with 0.1% NP40 (pH 8.0). Slides were mounted in 90% glycerol in phosphate buffer containing 4′,6-diamidino-2-phenylindole at a concentration of 0.3 μg/mL and coverslipped.

A number of normal male DNA versus normal female DNA hybridizations were done to identify outliers and polymorphic clones, which were excluded from further analysis.

Image and data processing. Three, 16-bit fluorescence single-color intensity images (4′,6-diamidino-2-phenylindole, Cy3, and Cy5) were collected from each array using a charge coupled device camera (Sensys, Photometric, equipped with a Kodak KAF 1400 chip) coupled to a 1× optical system, as previously described (11). The Spot 2.0 software program (available at http://www.jainlab.org/downloads.html) was used for image analysis (12). Intensities were determined for each spot by subtracting local background from foreground intensities. The following spots were excluded for further analysis: (a) spots that were <15 pixels in size, (b) spots with a correlation of <0.9 of the two fluorescent dyes or ones that were in the bottom 10th percentile, (c) spots that had a ratio of (reference intensity) / (sum of reference and test intensity) <0.1, and (d) spots with a test plus reference intensity of <200 or ones that were in the bottom 20th percentile. For each clone, the average ratio of test over reference intensity was calculated from the three replicate clones on the array. Clones whose ratio was derived from only one of three replicates and clones whose replicate ratios had a log2 SD > 0.33 were discarded. The human DNA sequence draft at http://genome.ucsc.edu (July 2003 freeze) was used to map clones. The log2 ratios for each case were median centered to zero. Thresholds for determining chromosome gain or loss were defined separately for each individual sample using a discrete time hidden Markov model (13) as implemented in the aCGH package of BioConductor open source software (http://www.bioconductor.org), correcting for varying signal-to-noise in each sample. A clone was declared aberrant if its absolute value exceeded the tumor-specific threshold computed as 2.5 times the estimate of the SD of the experimental noise for a given profile. A clone was considered amplified if it was part of a region (no larger than 10 Mb) whose absolute log2 ratio exceeded 0.9 and also exceeded the log2 ratio of its immediately flanking regions by at least 0.5. The algorithm also identified regions (also narrower than 10 Mb) whose log2 ratios exceeded immediately flanking segments by 0.9. The latter rule permitted the detection of amplifications originating out of regions present at decreased copy number. Clones with log2 ratios <−0.75 were considered high-level losses. These levels were chosen empirically due to the expected presence of normal cells in the specimens and tumor heterogeneity and are concordant with the thresholds used in the previously published array CGH articles (14).

The frequency of gains and losses for a given clone in a group of interest was calculated as the proportion of samples in which a clone was gained or lost in that group. As our CGH arrays do not cover the entire genome, each clone was assigned a genomic distance equal to the sum of half of the distance between its center and that of its two adjacent clones to quantitate the fraction of the genome altered.

Statistical analysis. Clones with significantly different copy number between different tumor groups were identified using a t statistic with pooled variance. Adjustment for multiple comparisons was made using false discovery rate.

Kruskal-Wallis nonparametric test was used to assess significance of the different measures of genomic instability (FGA, number of whole chromosome changes, number of copy number transitions within a chromosome, total number of chromosomes containing transitions, amplifications, and deletions) between tumors of different stage, grade, and outcome.

Survival probabilities were estimated using Kaplan-Meier or Cox proportional hazards analysis in R. Variables were defined in either continuous or dichotomous fashion.

Prediction analysis for microarrays (15) was used to identify clones that could best predict each bladder tumor class. Prediction analysis for microarrays uses a modified version of the nearest centroids classification method, which “shrinks” the centroids by means of soft thresholding. Ten-fold cross-validation was used to choose a threshold, which minimized classification errors and selected a list of predictive genes. Missing values were estimated across the data set, using a 10 nearest-neighbor impute engine.

One-way unsupervised hierarchical clustering was done for tumors using Ward's method of linkage and Euclidean distance as a metric.

Clinicopathologic characteristics of the samples. Ninety-eight primary bladder tumors of transitional histology were studied. These included 29 stage pTa (21 low grade and 8 high grade), 14 pT1, 15 pT2, 25 pT3, and 15 pT4 tumors (all pT1-pT4 tumors were high grade; Table 1). The median age for all patients was 69 years. Transurethral resection was used to sample 27 of the 29 stage pTa tumors and 11 of the 14 pT1 tumors, whereas cystectomy was done for 51 of the 55 stage pT2 or higher tumors. Outcome information was available for 54 patients with muscle-invasive disease (stage pT2 or higher). Fifteen of these patients survived with a median follow-up of 42.8 months, whereas 39 patients died with a median survival of 8.8 months. A listing of clinicopathologic information can be found in Supplementary Data (Clinical.xls).

Table 1.

Clinicopathologic characteristics

n (%)
Stage  
    pTa 29 (29.6) 
    pT1 14 (14.3) 
    pT2 15 (15.3) 
    pT3 25 (25.5) 
    pT4 15 (15.3) 
Gender  
    Male 70 (71.4) 
    Female 28 (28.6) 
Lymph node status  
    Negative 27 (49) 
    Positive 16 (29) 
    Unknown 12 (22) 
    Clinical outcome*  
    Alive (≥24 mo) 10 (23) 
    Dead (<24 mo) 34 (77) 
n (%)
Stage  
    pTa 29 (29.6) 
    pT1 14 (14.3) 
    pT2 15 (15.3) 
    pT3 25 (25.5) 
    pT4 15 (15.3) 
Gender  
    Male 70 (71.4) 
    Female 28 (28.6) 
Lymph node status  
    Negative 27 (49) 
    Positive 16 (29) 
    Unknown 12 (22) 
    Clinical outcome*  
    Alive (≥24 mo) 10 (23) 
    Dead (<24 mo) 34 (77) 
*

Muscle invasive tumors only.

Stage- and grade-specific genomic profiles. Array-based CGH identifies copy number gains and losses on a clone-by-clone basis. Such copy number gains and losses were defined using tumor-specific thresholds for each sample (median threshold for all samples ± 0.25). A representative profile of array CGH of a primary bladder tumor is shown in Fig. 1. A file containing all primary data log2 ratios can be found in Supplementary Data (log2ratios.xls).

Fig. 1.

Whole-genome array CGH profile from a primary bladder cancer (stage pTa, low grade). Copy number changes relative to normal, sex-matched DNA are shown for each clone. Horizontal solid lines, tumor-specific log2 ratio threshold of 0.2 that was used in this case to determine clones gained (+0.2) or lost (−0.2). Clones are ordered from chromosomes 1 to 22 and within each chromosome on the basis of their map position (http://genome.ucsc.edu/version July 2003). Two of the most frequent changes in bladder cancer are seen, high-level gains (amplifications) in the region of cyclin D1 (11q13), and loss of clones across the entire chromosome 9.

Fig. 1.

Whole-genome array CGH profile from a primary bladder cancer (stage pTa, low grade). Copy number changes relative to normal, sex-matched DNA are shown for each clone. Horizontal solid lines, tumor-specific log2 ratio threshold of 0.2 that was used in this case to determine clones gained (+0.2) or lost (−0.2). Clones are ordered from chromosomes 1 to 22 and within each chromosome on the basis of their map position (http://genome.ucsc.edu/version July 2003). Two of the most frequent changes in bladder cancer are seen, high-level gains (amplifications) in the region of cyclin D1 (11q13), and loss of clones across the entire chromosome 9.

Close modal

Specific loci harboring recurrent alterations across tumors were seen. There was a high frequency of aberrations involving whole chromosomes, whole chromosome arms, and consecutive clones on chromosome arms. The frequency of copy number changes showed characteristic patterns according to stage (Fig. 2) and grade (Supplementary Data: GradeFrequency). Stage pTa tumors had a 12% average frequency of alterations. The most frequent alteration in stage pTa tumors was loss of clones across the entire chromosome 9 occurring with an average frequency of 47% for 9p and 46% for 9q. The clone containing CDKN2A/p16 (CTB-65D18) had the highest frequency of loss in pTa tumors at 65%. Stage pT1 tumors had a higher average frequency of alterations than pTa tumors (27%). The highest average frequency losses in this tumor set were on 11p (54%), whereas the highest average frequency gains were on 8q (55%). These tumors had a similar frequency of loss of clones on chromosome 9p (51%) and 9q (43%) as pTa tumors. Finally, muscle-invasive tumors had an average frequency of alterations of 18%. Although these tumors retained the same spectrum of aberrations as pT1 tumors, they generally showed lower frequencies. Clones with the highest frequency of loss in these tumors were on 8p (29%), whereas clones with the highest frequency gains were on 20q (26%). A list of the most frequently recurrent aberrations with their corresponding frequencies by chromosome arm can be seen in Table 2.

Fig. 2.

Array CGH according to tumor stage. Copy number frequency plots for 29 pTa (21 low grade and 8 high grade), 14 pT1, and 55 muscle-invasive (pT2-4) tumors. Gains are shown above and losses below the 0 horizontal. Vertical dashed lines, centromeres.

Fig. 2.

Array CGH according to tumor stage. Copy number frequency plots for 29 pTa (21 low grade and 8 high grade), 14 pT1, and 55 muscle-invasive (pT2-4) tumors. Gains are shown above and losses below the 0 horizontal. Vertical dashed lines, centromeres.

Close modal
Table 2.

Most frequent genomic alterations by chromosome arm

StageChromosome armFrequency (%)Chromosome armFrequency (%)
pTa Gain 20q 17 Loss 9p 47 
     9q 46 
     11p 24 
     10q 20 
     13q 17 
     8p 16 
     17p 15 
       
pT1 Gain 8q 55 Loss 11p 54 
  20q 47  9p 51 
  19q 36  9q 43 
  1q 34  8p 38 
  20p 33  18q 32 
  17q 32  2q 32 
  19p 27  10q 29 
  5p 25  6q 23 
  2p 22  18p 21 
  3q 18  17p 18 
     13q 17 
     5q 16 
     14q 15 
     16q 15 
       
pT2-4 Gain 5p 27 Loss 8p 29 
  20q 26  9p 21 
  8q 23  11p 18 
  10p 21  17p 17 
  20p 21  18q 17 
  7p 20  10q 16 
  3q 18  5q 15 
  1q 17  6q 15 
     16q 15 
StageChromosome armFrequency (%)Chromosome armFrequency (%)
pTa Gain 20q 17 Loss 9p 47 
     9q 46 
     11p 24 
     10q 20 
     13q 17 
     8p 16 
     17p 15 
       
pT1 Gain 8q 55 Loss 11p 54 
  20q 47  9p 51 
  19q 36  9q 43 
  1q 34  8p 38 
  20p 33  18q 32 
  17q 32  2q 32 
  19p 27  10q 29 
  5p 25  6q 23 
  2p 22  18p 21 
  3q 18  17p 18 
     13q 17 
     5q 16 
     14q 15 
     16q 15 
       
pT2-4 Gain 5p 27 Loss 8p 29 
  20q 26  9p 21 
  8q 23  11p 18 
  10p 21  17p 17 
  20p 21  18q 17 
  7p 20  10q 16 
  3q 18  5q 15 
  1q 17  6q 15 
     16q 15 

NOTE: The frequencies of gains and losses on each chromosome arm were averaged according to stage. Only average frequencies of ≥15% are shown.

A direct comparison of tumors of different stage and grade was made to identify individual clones significantly altered between the groups. Robustness of the findings was assured by performing adjustment for multiple comparisons (false discovery rate ≤ 0.05). A list of the differentially altered regions between the different stages can be seen in Supplementary Data (stage.comparison.regions), whereas a comprehensive list of all these clones is in Supplementary Data (pTa.vs.pT1.signclones, pTa.vs.pT24.signclones, and pT1.vs.pT24.signclones).

Genomic instability. Differences in the extent of the genome affected were identified among the tumors of different stage and grade. The extent of the genome affected was defined as the FGA (as represented by the clones on our array). The median FGA for all the tumors studied was 17%. Stage pTa tumors had the lowest median FGA at 9%, whereas pT1 tumors had the highest median FGA at 27% (Kruskal-Wallis, P = 0.0003). Interestingly, muscle-invasive tumors had a median FGA of 16%, which was lower than that of pT1 tumors (Kruskal-Wallis, P = 0.03; Fig. 3). When pTa tumors were separated into low and high grade, low-grade tumors had a much lower FGA (median FGA = 8%) than that of high-grade tumors (median FGA = 20%; Kruskal-Wallis, P = 0.097; Supplementary Data: GradeFGA). A plot of the FGA distribution according to stage and grade can be seen in Supplementary Data (FGArange). Interestingly, a total number of seven pTa and six muscle-invasive tumors with the lowest FGA (<3%) had no whole chromosome or chromosome arm changes, but showed only gains and losses of individual clones scattered over the genome. These clones do not seem to be due to polymorphisms or outliers based on the information from our normal-normal control hybridizations.

Fig. 3.

FGA by tumor stage. Black column, FGA gained; gray column, FGA lost. The FGA for each tumor was quantitated by assigning for each clone a distance equal to the sum of one half of the distance between its own center and that of its neighboring clones. Stage pTa tumors had the lowest median FGA (9%), whereas pT1 tumors had the highest (27%). Muscle-invasive tumors had a median FGA of 16%.

Fig. 3.

FGA by tumor stage. Black column, FGA gained; gray column, FGA lost. The FGA for each tumor was quantitated by assigning for each clone a distance equal to the sum of one half of the distance between its own center and that of its neighboring clones. Stage pTa tumors had the lowest median FGA (9%), whereas pT1 tumors had the highest (27%). Muscle-invasive tumors had a median FGA of 16%.

Close modal

FGA is a measure of genomic instability, representing the overall burden of gains and losses in each tumor. Other measures representing more specific mechanisms of chromosome alterations can also be defined, including (a) the number of whole chromosome changes, (b) the number of copy number transitions within a chromosome (representing unbalanced translocations or deletions), and the total number of chromosomes containing such transitions, (c) amplifications, and (d) deletions.

For all tumors, the number of copy number alterations involving whole chromosomes (median = 1) was significantly less frequent than alterations affecting entire chromosome arms or portions of arms (median = 8; Kruskal-Wallis, P < 0.0001).

Superficial tumors showed significant differences in all of the measures of genomic instability assessed according to both stage and grade. Stage pT1 tumors had a greater number of whole chromosome changes (P = 0.012), number of transitions (P = 0.044), number of chromosomes containing transitions (P = 0.004), and number of amplifications (P = 0.035) than stage pTa tumors. The number of high-level losses did not differ significantly between these two stages. When pTa tumors were compared according to their grade, higher-grade tumors showed significantly increased number of transitions (P = 0.019) and number of chromosomes containing transitions (P = 0.012) compared with low-grade tumors. However, the number of whole chromosome changes (P = 0.584), number of amplifications (P = 0.535), and number of high level losses (P = 0.203) did not reach significance, probably due to the small number of tumors in the high-grade pTa tumor group (n = 8).

Stage pTa tumors showed a significantly lower level of all the measures of chromosomal instability assessed when compared with muscle-invasive tumors, including the number of whole chromosome changes (P = 0.015), number of transitions (P = 0.006), number of chromosomes containing transitions (0.0002), and number of amplifications (P = 0.003). Interestingly, there were significantly more high-level losses in pTa tumors than in muscle-invasive tumors (P = 0.02). When pT1 tumors were compared with muscle-invasive tumors, pT1 tumors showed overall a greater level of all the measures of genomic instability calculated but they did not reach statistical significance. The only exception was the number of high-level losses, which was higher in pT1 than in muscle-invasive tumors (P = 0.05).

Regions of high-level amplification. High-level amplifications were identified in 45 of the 98 patients (6 pTa, 8 pT1, and 31 pT2-4) and affected 233 clones. Twenty-nine of these clones showed amplification in three or more cases. A list of these amplified clones with their chromosomal location is shown in Table 3.

Table 3.

Amplifications in the 98 bladder tumors

CloneChromosome bandStart position*GenesNo. tumors
RP11-193J5 1q24 163,240,622  
RMC01P52 1p34 39,779,782 MYCL1 
RP11-43B4 6p22 20,117,736 ID4 
RP11-159C8 6p22 20,442,777 E2F3 
CTD-2018P8 6p22 20,622,576 CDKAL1 
RP11-3D15 6p22 21,715,615 SOX4 
RP11-273J1 6p22 22,030,457 LOC401237 
RP11-10G10 8q22 101,112,122 SPAG1, RNF19 
RP11-145G10 8q24 128,499,737  
RP11-85M7 10p14 6,649,014 SFMBT2 
RP11-33J8 10p14 7,199,612  
RP11-35I11 10p14 8,922,795  
RP1-88B16 11q13 69,133,863 CCND1prox 
CTD-2192B11 11q13 69,133,897 CCND1 
RP1-4E16 11q13 69,134,005 CCND1 
RP1-128I8 11q13 69,134,005 CCND1 
RP1-162F2 11q13 69,134,063 CCND1end 
RP1-17L4 11q13 69,134,124 FGF4 
CTB-36F16 11q13 69,134,163 FGF3 
CTC-437H15 11q13 69,134,772 EMS1, CTN1 
RP11-120P20 11q13 70,129,383 SHANK2 
GS-7N12 11q13-14 74,760,360 PAK1 
CTD-2222B22 11q23 116,613,723 PCSK7 
CTB-136O14 12q14 67,488,237 MDM2 
RP11-15L3 12q15 68,217,525 FRS2, CCT2 LRRC10, VMD2L3 
CTB-82N15 12q14 68,366,100  
RP11-183A20 13q33 106,617,308 TNFSF13B 
RP11-140E1 19q13 43,372,950 DPF1, PPP1R14A, SPINT2, KCNK6 
RP11-118P21 19q13 43,861,826 ACTN4 
CloneChromosome bandStart position*GenesNo. tumors
RP11-193J5 1q24 163,240,622  
RMC01P52 1p34 39,779,782 MYCL1 
RP11-43B4 6p22 20,117,736 ID4 
RP11-159C8 6p22 20,442,777 E2F3 
CTD-2018P8 6p22 20,622,576 CDKAL1 
RP11-3D15 6p22 21,715,615 SOX4 
RP11-273J1 6p22 22,030,457 LOC401237 
RP11-10G10 8q22 101,112,122 SPAG1, RNF19 
RP11-145G10 8q24 128,499,737  
RP11-85M7 10p14 6,649,014 SFMBT2 
RP11-33J8 10p14 7,199,612  
RP11-35I11 10p14 8,922,795  
RP1-88B16 11q13 69,133,863 CCND1prox 
CTD-2192B11 11q13 69,133,897 CCND1 
RP1-4E16 11q13 69,134,005 CCND1 
RP1-128I8 11q13 69,134,005 CCND1 
RP1-162F2 11q13 69,134,063 CCND1end 
RP1-17L4 11q13 69,134,124 FGF4 
CTB-36F16 11q13 69,134,163 FGF3 
CTC-437H15 11q13 69,134,772 EMS1, CTN1 
RP11-120P20 11q13 70,129,383 SHANK2 
GS-7N12 11q13-14 74,760,360 PAK1 
CTD-2222B22 11q23 116,613,723 PCSK7 
CTB-136O14 12q14 67,488,237 MDM2 
RP11-15L3 12q15 68,217,525 FRS2, CCT2 LRRC10, VMD2L3 
CTB-82N15 12q14 68,366,100  
RP11-183A20 13q33 106,617,308 TNFSF13B 
RP11-140E1 19q13 43,372,950 DPF1, PPP1R14A, SPINT2, KCNK6 
RP11-118P21 19q13 43,861,826 ACTN4 
*

Based on University of California Santa Cruz human DNA sequence draft of July 2003 freeze.

Clones with amplifications in at least three tumors are shown.

The six most frequently amplified clones on 6p22.3 span a region of ∼2 Mb between RP11-43B4 and RP11-273J1. These clones were present in significantly higher frequencies in the muscle-invasive tumors (P= 0.002). Genes in this region include E2F3 and CDKAL1. This area is a recurrent area of amplification in bladder cancer and four of the six clones that were found most frequently amplified in this study were the identical clones identified in our previous bladder CGH study on a different tumor set (4). The chromosome 6p22.3 CGH profile of tumors with at least three clones amplified in the critical region of amplification is available in Supplementary Data (Amplifications).

Eight of the clones on chromosome 8q22-24 were amplified in at least four cases. Two of these eight clones were also reported as amplified in our previous array CGH study. The most frequently amplified clones on the 8q region were RP11-128P9 (133,448,746-133,448,897 kb) and RP11-45B19 (135,479,232-135,641,838 kb), including the KCNQ3 and ZNF406 genes, respectively.

The 12q15 region of amplification contained three clones, which were amplified in at least four cases. All but one of the cases that contained amplifications of these three clones were superficial tumors. The most frequently amplified clones were CTB-136O14 (67,488,237-67,674,625) and RP11-15L3 (68,217,525-68,403,913 kb) and contained the FRS2, CCT2, VMD2L3, and MDM2 genes.

Deletions. High-level losses were identified in 59 of the 98 patients (19 pTa, 10 pT1, 30 pT2-4) affecting 349 clones, of which 59 clones showed deletions in three or more cases. A list of the deleted clones is shown in Table 4.

Table 4.

Deletions in the 98 bladder tumors

CloneChromosome bandStart position*GenesNumber of tumors
RP11-98C1 2q14 118,783,032 FLJ39081, INSIG2 
RP11-504L12 2q35 217,012,575 PECR 
RP11-247E23 2q36 223,880,100  
RP11-188B21 2q37 233,910,314 TNRC15, KCNJ13 
RP11-21K1 2q37 235,977,579  
RP11-116M19 2q37 236,916,706 CENTG2 
CTB-172I13 2 q tel Telomere  
RP11-79N22 4p15 18,622,376  
RP11-194B9 4p15 28,103,518  
RP11-6L19 4q32 167,513,183 TLL1 
RP11-252I13 5q21 107,005,165 EFNA5 
RP11-81C5 5q22-23 115,231,364 APG12L, AP3S1 
RP11-58F7 7q36 156,989,618 PTPRN2 
RP11-117P11 8p23 2,057,936 MYOM2 
RP11-82K8 8p23 2,080,313 MYOM2 
RP11-246G24 8p23 2,355,758  
RP11-112G9 8p23 10,038,625 MSRA 
RP11-254E10 8p23 11,164,156  
RP11-165O14 9p24 5,873,409 MLANA, KIAA2026, RANBP6 
CTD-2006L14 9p24 5,892,906 MLANA, KIAA2026, RANBP6 
RP11-264O11 9p23 10,578,156  
RP11-109M15 9p22 16,141,129  
CTB-65D18 9p21 21,958,037 CDKN2A 20 
RP11-33O15 9p21 22,823,087  
RP11-17J8 9p13 34,511,041 UNQ470 
RP11-165H19 9p13 34,589,621 C9orf23, DCTN3, ARID3C, OPRS1, GALT, IL11RA, CCL27, CCL19, CCL21 
RP11-61G7 9p13 35,799,611 SPAG8, HINT2 
RP11-19O14 9q21 68,683,031 TRPM3 
RP11-8L13 9q21 16,996,818  
RP11-14J9 9q21 73,221,555  
RP11-57N18 9q21 73,439,444 LOC138932 
RP11-8D10 9q21 80,348,013  
RP11-9I18 9q22 88,624,086 LOC340515 
RP11-128O12 9q22 91,549,881 C9orf10OS 
RP11-54O15 9q22 93,006,894 C9orf3 
RP11-5K11 9q31 98,443,107 TEX10, MGC17337 
RP11-4O1 9q31-32 110,119,611 SUSD1 
RP11-9M16 9q32 112,531,818 AKNA, DFNB31 10 
RP11-229N14 9q33 115,037,234 ASTN2 
GS1-135I17 9q34-tel 133,886,959 VAV2 
RP11-17O5 10q22.2 78,595,996 KCNMA1 
CTB-46B12 10q23 89,287,772 PTEN 
RP11-129G17 10q23 89,676,027 FLJ11218 
RP11-8D20 10q24 97,762,569 DNTT 
RP11-19K9 10q24 99,294,901 CRTAC1 
RP4-693L23 11p15 2,852,537 CDKN1C 
RP11-47D7 11p15 11,610,583 AF116621 
RP11-62G18 11p15 19,596,327 NAV2 
RP11-11A11 11p15 19,912,236 NAV2 
RP11-72A10 11p12 36,741,042  
RP11-34F8 11p12 40,886,858  
CTD-2012D15 11q22 104,333,213  
RP11-36M5 11q23 113,531,008 ZNF145, NNMT 
RP11-8K10 11q23 119,094,962 PVRL1 
RP11-15J15 11q24 125,423,862 CDON 
RP11-20M1 11q24 125,859,929 KIRREL3 
RP11-52B21 13q14 45,018,239 CHDC1 
RP11-4B17 18q23 72,858,573 MBP 
RP11-7H17 18q23 75,214,350 ATP9B, NFATC1 
CloneChromosome bandStart position*GenesNumber of tumors
RP11-98C1 2q14 118,783,032 FLJ39081, INSIG2 
RP11-504L12 2q35 217,012,575 PECR 
RP11-247E23 2q36 223,880,100  
RP11-188B21 2q37 233,910,314 TNRC15, KCNJ13 
RP11-21K1 2q37 235,977,579  
RP11-116M19 2q37 236,916,706 CENTG2 
CTB-172I13 2 q tel Telomere  
RP11-79N22 4p15 18,622,376  
RP11-194B9 4p15 28,103,518  
RP11-6L19 4q32 167,513,183 TLL1 
RP11-252I13 5q21 107,005,165 EFNA5 
RP11-81C5 5q22-23 115,231,364 APG12L, AP3S1 
RP11-58F7 7q36 156,989,618 PTPRN2 
RP11-117P11 8p23 2,057,936 MYOM2 
RP11-82K8 8p23 2,080,313 MYOM2 
RP11-246G24 8p23 2,355,758  
RP11-112G9 8p23 10,038,625 MSRA 
RP11-254E10 8p23 11,164,156  
RP11-165O14 9p24 5,873,409 MLANA, KIAA2026, RANBP6 
CTD-2006L14 9p24 5,892,906 MLANA, KIAA2026, RANBP6 
RP11-264O11 9p23 10,578,156  
RP11-109M15 9p22 16,141,129  
CTB-65D18 9p21 21,958,037 CDKN2A 20 
RP11-33O15 9p21 22,823,087  
RP11-17J8 9p13 34,511,041 UNQ470 
RP11-165H19 9p13 34,589,621 C9orf23, DCTN3, ARID3C, OPRS1, GALT, IL11RA, CCL27, CCL19, CCL21 
RP11-61G7 9p13 35,799,611 SPAG8, HINT2 
RP11-19O14 9q21 68,683,031 TRPM3 
RP11-8L13 9q21 16,996,818  
RP11-14J9 9q21 73,221,555  
RP11-57N18 9q21 73,439,444 LOC138932 
RP11-8D10 9q21 80,348,013  
RP11-9I18 9q22 88,624,086 LOC340515 
RP11-128O12 9q22 91,549,881 C9orf10OS 
RP11-54O15 9q22 93,006,894 C9orf3 
RP11-5K11 9q31 98,443,107 TEX10, MGC17337 
RP11-4O1 9q31-32 110,119,611 SUSD1 
RP11-9M16 9q32 112,531,818 AKNA, DFNB31 10 
RP11-229N14 9q33 115,037,234 ASTN2 
GS1-135I17 9q34-tel 133,886,959 VAV2 
RP11-17O5 10q22.2 78,595,996 KCNMA1 
CTB-46B12 10q23 89,287,772 PTEN 
RP11-129G17 10q23 89,676,027 FLJ11218 
RP11-8D20 10q24 97,762,569 DNTT 
RP11-19K9 10q24 99,294,901 CRTAC1 
RP4-693L23 11p15 2,852,537 CDKN1C 
RP11-47D7 11p15 11,610,583 AF116621 
RP11-62G18 11p15 19,596,327 NAV2 
RP11-11A11 11p15 19,912,236 NAV2 
RP11-72A10 11p12 36,741,042  
RP11-34F8 11p12 40,886,858  
CTD-2012D15 11q22 104,333,213  
RP11-36M5 11q23 113,531,008 ZNF145, NNMT 
RP11-8K10 11q23 119,094,962 PVRL1 
RP11-15J15 11q24 125,423,862 CDON 
RP11-20M1 11q24 125,859,929 KIRREL3 
RP11-52B21 13q14 45,018,239 CHDC1 
RP11-4B17 18q23 72,858,573 MBP 
RP11-7H17 18q23 75,214,350 ATP9B, NFATC1 
*

Based on University of California Santa Cruz human DNA sequence draft of July 2003 freeze.

Clones with deletions in at least three tumors are shown.

As expected, the clones with the highest frequency of high-level losses were located on chromosomes 9p24-13 (5,873,409-35,799,611 kb) and 9q21-34 (68,683,031-1,333,886,959 kb). Clone CTB-65D18 containing CDKN2A was the most commonly deleted clone on 9p (20 of 98 tumors), whereas clone RP11-9M16 containing AKNA and DFNB31 was the most common deleted clone on 9q (10 of 98 tumors). Quantitative real-time PCR was used as an independent method to confirm the CDKN2A homozygous deletion. Taqman analysis showed single-copy loss of multiple chromosome 9 regions, with further loss of the CDKN2A sequence in a subset of cell lines and tumors, agreeing with our CGH analyses.5

5

In preparation.

Other deleted clones were located on 2q14-37 (118,783,032 kb), 8p23 (2,057,936 kb), 10q22-24 (78,595,996-99,294,901 kb), and 11p15-12 (2,852,537-40,886,858 kb).

Outcome. A total of 54 muscle-invasive cases had follow-up information available (median survival 11 months; Supplementary Data: Clinical.xls). For analysis of associations with outcome, subsets of contrasting patients were selected with good prognosis (alive >2 years, n = 10) and bad prognosis (dead in <2 years, n = 34). There were no significant differences in the distribution of stage or nodal status between the two prognosis groups.

The FGA was significantly lower in the muscle-invasive patients with good prognosis (median FGA = 3.5%) than patients with bad prognosis (median FGA = 22.4%; P = 0.002). Other measures of genomic instability were also significantly lower in the patients with good versus bad prognosis, including whole chromosome changes (P = 0.01), number of transitions (P = 0.05), number of chromosomes with transitions (P = 0.016), and the number of high-level losses (P = 0.015). However, the difference in the number of amplifications did not reach significance (P = 0.09).

Univariate Cox proportional hazards analysis showed significant association between overall survival and FGA as a continuous (P = 0.013) or dichotomized variable (P = 0.0049) in muscle-invasive patients, with low FGA associated with longer overall survival. For dichotomized analysis, a threshold of FGA (10%) was chosen as the optimum balance between sensitivity (88%) and specificity (70%) as determined by receiver operator characteristic analysis. FGA was not associated with any other clinicopathologic variables tested. Univariate analysis showed association between overall survival and the presence of amplified clones (P = 0.033), deleted clones (P = 0.052), chromosomes with transitions (P = 0.055), and whole chromosome changes (P = 0.07). Extent of invasion (stage pT2, pT3, or pT4), nodal status, sex, and age were not significantly associated with outcome in these muscle-invasive patients.

Multivariate analysis using a Cox proportional hazards model was done to identify statistically independent factors in overall survival. When variables considered above (in the univariate model) were chosen in a stepwise reduction, FGA was the only independent predictor of overall survival when analyzed as a continuous or dichotomous variable.

Cox proportional hazards analysis was also used to identify individual clones that were associated with prognosis. This analysis identified 137 individual clones, the majority of which were gained on chromosomes 2p25-11 (8,710,225-88,494,982 kb), 5p13-15 (3,055,081-31,667,201 kb), 8q22 (99,202,014-101,125,862 kb), and lost on 5q11-34 (54,222,747-160,719,699 kb), 8p23-21 (2,080,313-19,617,041 kb), 8p12 (28,889,000-31,264,731 kb), 10q21-26 (64,541,582-128,795,405 kb), and 17q21-25 (30,082,000-80,815,074 kb).

The prediction analysis for microarrays method of supervised classification was used for outcome prediction in these muscle-invasive patients with good versus bad prognosis. Using cross-validation to reduce bias due to overfitting, a classification success rate of 71% was obtained based on 102 unique clones (55 of these clones were also identified in the Cox proportional hazards analysis described above). The majority of prediction analysis for microarrays predictive clones were located on chromosomes 8 (23 clones on 8p and 18 clones on 8q) and 5 (7 clones on 5p and 17 clones on 5q). The combination of clones on chromosomes 5q and 8p did not add significantly to prediction of overall outcome.

Unsupervised classification of muscle-invasive patients grouped the tumors into three clusters (Fig. 4). Interestingly, the tumors in these three clusters had different median FGA (Kruskal-Wallis rank sum, P < 0.0001). The tumors in the left cluster had FGA = 3.8%, in the middle cluster FGA = 13.5%, and the patients in the right cluster had FGA = 35.6%. Kaplan-Meier survival analysis showed significant association with outcome when done based on these tumor clusters (P = 0.019; Fig. 4). Patients in the left cluster were associated with the best survival and also had the lowest FGA value.

Fig. 4.

Unsupervised hierarchical clustering of muscle-invasive bladder tumors (n = 44). Left, clustering of tumors shows three large clusters (top: green, red, blue). The horizontal band at top shows tumor columns with good prognosis (alive >2 years, n = 10) in purple, and bad prognosis (dead in ≤2 years, n = 34) in green. Rows represent individual clones on chromosomes in ascending order (light and dark blue for the p and q arms of odd chromosomes and green and yellow for the p and q arms of even chromosomes). Decreased copy numbers are in red and increased copy numbers are in green. Right, Kaplan-Meier survival analysis showing significant association with outcome (P = 0.019) when done based on tumor groups defined by unsupervised clustering (on the left): green, left cluster; red, middle cluster, and blue, right cluster.

Fig. 4.

Unsupervised hierarchical clustering of muscle-invasive bladder tumors (n = 44). Left, clustering of tumors shows three large clusters (top: green, red, blue). The horizontal band at top shows tumor columns with good prognosis (alive >2 years, n = 10) in purple, and bad prognosis (dead in ≤2 years, n = 34) in green. Rows represent individual clones on chromosomes in ascending order (light and dark blue for the p and q arms of odd chromosomes and green and yellow for the p and q arms of even chromosomes). Decreased copy numbers are in red and increased copy numbers are in green. Right, Kaplan-Meier survival analysis showing significant association with outcome (P = 0.019) when done based on tumor groups defined by unsupervised clustering (on the left): green, left cluster; red, middle cluster, and blue, right cluster.

Close modal

Bladder cancer consists of a heterogeneous group of tumors that differ in their types and range of genetic alterations (16, 17). As it has been reported for other solid tumors, one hallmark of the pathway of tumor development and progression is an increase in genome complexity. In this study, array CGH was used to dissect the spectrum of alterations in bladder cancer and to identify recurrent aberrations that may contain cancer-related genes. Different types of genomic instability were also assessed to better understand the underlying mechanisms of genomic instability during tumor development. Such measurements are helpful in further stratifying tumors into more homogeneous and clinically relevant subgroups and can be used to develop markers to better predict clinical outcome.

This study has confirmed and extended at high resolution previously reported loci showing alterations in bladder cancer. Many of the same regions were previously reported by other investigators using loss of heterozygosity (18, 19), chromosomal CGH (2, 3, 2026), and/or array-based CGH (4, 5). Apart from detecting regions of recurrent aberrations, individual clones that were significantly different between tumors of different stage and grade were identified. Overall, there was an increase in the levels of chromosomal alterations for both increasing grade and stage and strong similarities were found in the levels of chromosomal aberrations between pT1 and muscle-invasive tumors similar to previous reports (21). Loss of clones across the entire chromosome 9 was the most frequent alteration in stage pTa tumors, and in some tumors it was the only alteration identified. The frequency of loss for 9p versus 9q was balanced in tumors from all pathologic stages.

Other recurrent alterations present at high frequency across different stages were the loss of chromosome 8p accompanied by a gain of 8q. This characteristic transition pattern of chromosome 8 has also been reported in other types of tumors, including colon, prostate, and breast. In all these tumors, extensive areas of distal 8p are lost. Interestingly, five CGH studies (including ours) using different tumor types (bladder, colon, prostate, and fallopian tube) identified the same clone (bacterial artificial chromosome RP11-121F7) as homozygously lost on 8p23.2. This clone encompasses the CSMD1 gene (CUB and Sushi multiple domains 1), the function of which is currently unknown. The identification of a functionally significant gene in this region has the potential as a diagnostic or therapeutic target with wide application to solid tumors. Muscle-invasive tumors showed roughly the same spectrum of changes as pT1 tumors. However, gains on chromosome 7p and 7q were seen at significantly higher frequencies in the muscle-invasive tumors. Chromosome 7 aneusomy has been previously shown by fluorescence in situ hybridization to be significantly different between pT1 and pT2-3 tumors (27) and has been associated with tumor aggressiveness (28).

In this study, significant differences in the FGA were identified among the tumors of different stage, grade, and patient outcome. Stage pTa tumors had the lowest levels of alteration, and stage pT1 showed a much higher overall FGA. FGA seemed to stabilize in the muscle-invasive cases, with a slight decrease relative to pT1 tumors. Low-grade superficial tumors had a significantly lower FGA than did high-grade tumors. Interestingly, some of the low-grade pTa tumors and even some of the muscle-invasive tumors had a very low FGA (<3%). These tumors had no gross chromosomal alterations whatsoever by array CGH, but only gains or losses of a few individual clones. The alterations of these individual clones may represent true genomic changes, although hybridization noise or even polymorphisms affecting specific clones in individual patients might also be suspected. This observation of very low FGA in certain tumors suggests that some bladder tumors evolve by mechanisms that result in little chromosomal change, whereas others undergo marked chromosomal rearrangements. This variability may be due to differences in tumor initiation, processes of genomic instability, or the individual genotypic background of the tumor. It is possible, for example, that the tumors with few chromosomal changes arise in epithelial stem cells already having an active telomerase so that telomere crisis with associated extensive genomic rearrangement is avoided. Similarly, tumor initiation and progression associated with defects in mismatch repair might lead to tumors with few chromosomal abnormalities, although this is reported to be rare in bladder cancer. Overall, in our tumor set, copy number alterations affecting chromosome arms or portions of the arms (transitions) were more frequent than whole chromosome changes. This was most prominent in low-grade pTa tumors where 5 of 21 (24%) of the tumors had transitions and no whole chromosome changes. These observations are consistent with a model in which telomeric dysfunctions and breakage-fusion-bridge events occur at an earlier stage in tumor development than errors in whole chromosome segregation. It is interesting to note that most of the tumors analyzed in this study exhibited both numerical aberrations and highly complex structural aberrations, increasing from pTa to pT1 and muscle-invasive tumors.

One surprising finding in this study was the significantly higher FGA in stage pT1 tumors compared with muscle-invasive tumors and the absence of significant differences between these stages in the other measures of genomic instability assessed. Perhaps, over time, tumors acquire a more “stable” genome, already optimized for growth, invasion, and dissemination, making it less probable that additional lesions will confer further advantage. Direct evaluation of the rate of genomic instability during tumor evolution might best be measured using matched pairs of samples from the same patient at different stages of progression (e.g., pT1 versus pT3). The decrease in genomic alterations in muscle-invasive tumors might also have been due to increased normal cell contamination compared with superficial tumors, with a larger component of stroma and inflammation within the tumor. However, tumor blocks were grossly dissected to assure that at least 70% of tumor cells was present. Previous cell line mixing studies have shown that this degree of admixture does not affect detection of copy number alterations. Further, comparison of different dissection procedures (gross block dissection versus fine microdissection yielding 95% tumor cells) showed that the frequency of alterations did not significantly increase with purity >70%.6

6

Unpublished data.

Higher-stage tumors might be expected to show increased tumor heterogeneity and aneuploidy. Array CGH detects DNA alterations present in common within the tumor cell population, and so it is not able to account for tumor heterogeneity. Similarly, the complexity of DNA copy number alterations present in an aneuploid tumor may lead to decreased detection of such alterations due to the difficulty during array CGH data analysis of defining copy number gains and losses in these tumors. Different algorithms used for analysis of array CGH data may also affect the detection of genomic alterations in an aneuploid background.

There is wide agreement that recurrent genomic aberrations may highlight chromosomal regions that are important for tumor development. The importance of recurrent aberrations involving gene dosage is particularly clear. In many cases, these aberrations contain known oncogenes and tumor suppressor genes whose expression levels are altered by genomic changes. The large size of most aberrations detected by chromosomal CGH or cytogenetics makes it difficult to narrow down candidate genes. However, using array CGH, the regions of high-level aberrations (amplifications and homozygous deletions) can be studied at much greater resolution and their boundaries can be identified with greater precision. However, identifying the relevant cancer-related genes is still difficult because most aberrations are large, containing multiple genes. For example, the region of recurrent amplification on chromosome 11q13 containing CCND1 also contains growth factors FGF19, FGF4, and FGF3, and actin-binding oncogene EMS1. It is possible that the concurrent impaired expression of all these genes is necessary for a particular tumor phenotype and may drive the appearance of high-level recurrent rearrangements in this region. In these cases, array CGH can help refine the number of candidate genes that would need to be further characterized.

Clinical outcome for patients with muscle-invasive bladder cancer is only partially predicted by tumor stage. Grade is not a useful prognostic marker in this tumor group because they are all high grade. Although specific genetic changes have been described, which are associated with higher disease stage (at 1p, 3p, 13q, 17p, 20p), they did not predict outcome independent of standard clinicopathologic parameters (19, 22). A large number of additional molecular markers have also been nominated as predictors of outcome in muscle-invasive bladder cancer. However, conclusions from these studies have been inconsistent, and few if any of the markers is able to discriminate good versus poor prognosis groups sufficiently to have clinical utility (29). In this study, FGA was significantly higher in patients with bad clinical outcome (died in <2 years) versus in patients with good outcome (alive in >2 years). This was independent of stage, nodal status, and age, although these other clinical variables (stage, age, sex, nodal status) were not themselves significantly associated with prognosis in this group of 54 patients. Multivariate Cox proportional hazards analysis identified FGA as the only independent prognosticator of overall outcome.

Unsupervised classification separated muscle-invasive patients into three groups. These groups were significantly associated with overall outcome. Further, the tumor cluster with the worst outcome had the highest FGA, providing additional evidence for the correlation of FGA with outcome in patients with muscle-invasive disease. However, for FGA to be applied in a clinical setting, a cutoff would be necessary. When, in this study, FGA was used as a dichotomous variable, a threshold of <10% was best for identifying patients with poor prognosis. This threshold was chosen after performing receiver operator characteristic analysis to determine the optimum balance between sensitivity and specificity of FGA in predicting patients with worse outcome. Further studies in independent data sets would be necessary to fine tune such a threshold and make it applicable in a wider clinical setting. In the future, this could be applied for choice of more aggressive therapies, and in some cases sparing therapy in patients with better outcome. The quantitative nature of FGA is also appealing for its development as a clinical marker. FGA can be used as a means to determine the “genomic grade” of a tumor in a similar way to the histologic stage or grade, and together with a panel of other molecular markers can be used to predict outcome of patients with muscle-invasive disease.

Supervised classification and Cox proportional hazards analysis applied to individual clones identified clones associated with outcome that were localized predominantly on chromosomes 8p, 8q, 5p, and 5q. Chromosome 8p deletions have been found frequently in several tumor types, associated with invasive tumor growth. Chromosome 5 aberrations have also been associated with increasing bladder cancer stage (30). In this study, alterations on either 8p or 5q were able to successfully predict overall outcome in patients with muscle-invasive disease. However, their combination did not have the same success.

This study shows at high resolution the range of specific genomic changes present in bladder cancer. Tumors also showed a wide range in the pattern of alterations, suggesting that alternative mechanisms of genomic instability may play a role in this tumor type. The FGA was independently associated with outcome in this tumor set, and may reflect a “genetic grade” for development more informed choices for therapeutic decision making. These studies should be validated, especially using independent tumor sets from multiple institutions.

Grant support: NIH CA089715.

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 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article can be found at http://cc.ucsf.edu/people/waldman/bladder/blaveri.cgh.

1
Lee R, Droller MJ. The natural history of bladder cancer. Implications for therapy.
Urol Clin North Am
2000
;
27
:
1
–13, vii.
2
Kallioniemi A, Kallioniemi OP, Sudar D, et al. Comparative genomic hybridization for molecular cytogenetic analysis of solid tumors.
Science
1992
;
258
:
818
–21.
3
Kallioniemi A, Kallioniemi OP, Citro G, et al. Identification of gains and losses of DNA sequences in primary bladder cancer by comparative genomic hybridization.
Genes Chromosomes Cancer
1995
;
12
:
213
–9.
4
Veltman JA, Fridlyand J, Pejavar S, et al. Array-based comparative genomic hybridization for genome-wide screening of DNA copy number in bladder tumors.
Cancer Res
2003
;
63
:
2872
–80.
5
Hurst CD, Fiegler H, Carr P, Williams S, Carter NP, Knowles MA. High-resolution analysis of genomic copy number alterations in bladder cancer by microarray-based comparative genomic hybridization.
Oncogene
2004
;
23
:
2250
–63.
6
Albertson DG, Ylstra B, Segraves R, et al. Quantitative mapping of amplicon structure by array CGH identifies CYP24 as a candidate oncogene.
Nat Genet
2000
;
25
:
144
–6.
7
Wilhelm M, Veltman JA, Olshen AB, et al. Array-based comparative genomic hybridization for the differential diagnosis of renal cell cancer.
Cancer Res
2002
;
62
:
957
–60.
8
Frederick L, Greene DLP, Irvin D, et al. editors. AJCC cancer staging manual. 6th ed. New York: Springer-Verlag; 2002. p. 367.
9
Lopez-Beltran ASG, Gasser T, Hartmann A, et al. Infiltrating urothelial carcinoma. In: Eble JNSG, Epstein JI, Sesterhenn IA, editors. WHO pathology and genetics tumors of the urinary system and male genital organs; 2004. p. 93.
10
Snijders AM, Nowak N, Segraves R, et al. Assembly of microarrays for genome-wide measurement of DNA copy number.
Nat Genet
2001
;
29
:
263
–4.
11
Pinkel D, Segraves R, Sudar D, et al. High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays.
Nat Genet
1998
;
20
:
207
–11.
12
Jain AN, Tokuyasu TA, Snijders AM, Segraves R, Albertson DG, Pinkel D. Fully automatic quantification of microarray image data.
Genome Res
2002
;
12
:
325
–32.
13
Fridlyand J, Snijders A, Pinkel D, Albertson DG, Jain AN. Application of hidden Markov models to the analysis of the array CGH data.
J Multivariate Analysis
2004
;
90
:
132
–53.
14
Nakao K, Mehta KR, Fridlyand J, et al. High-resolution analysis of DNA copy number alterations in colorectal cancer by array-based comparative genomic hybridization.
Carcinogenesis
2004
;
25
:
1345
–57.
15
Tibshirani R, Hastie T, Narasimhan B, Chu G. Diagnosis of multiple cancer types by shrunken centroids of gene expression.
Proc Natl Acad Sci U S A
2002
;
99
:
6567
–72.
16
Orntoft TF, Wolf H. Molecular alterations in bladder cancer.
Urol Res
1998
;
26
:
223
–33.
17
Knowles MA. Molecular genetics of bladder cancer: pathways of development and progression.
Cancer Surv
1998
;
31
:
49
–76.
18
Tzai TS, Chen HH, Chan SH, et al. Clinical significance of allelotype profiling for urothelial carcinoma.
Urology
2003
;
62
:
378
–84.
19
Sengelov L, Christensen M, von der Maase HD, et al. Loss of heterozygosity at 1p, 8p, 10p, 13q, and 17p in advanced urothelial cancer and lack of relation to chemotherapy response and outcome.
Cancer Genet Cytogenet
2000
;
123
:
109
–13.
20
Simon R, Burger H, Semjonow A, Hertle L, Terpe HJ, Bocker W. Patterns of chromosomal imbalances in muscle invasive bladder cancer.
Int J Oncol
2000
;
17
:
1025
–9.
21
Zhao J, Richter J, Wagner U, et al. Chromosomal imbalances in noninvasive papillary bladder neoplasms (pTa).
Cancer Res
1999
;
59
:
4658
–61.
22
Richter J, Wagner U, Schraml P, et al. Chromosomal imbalances are associated with a high risk of progression in early invasive (pT1) urinary bladder cancer.
Cancer Res
1999
;
59
:
5687
–91.
23
Koo SH, Kwon KC, Ihm CH, Jeon YM, Park JW, Sul CK. Detection of genetic alterations in bladder tumors by comparative genomic hybridization and cytogenetic analysis.
Cancer Genet Cytogenet
1999
;
110
:
87
–93.
24
Simon R, Burger H, Brinkschmidt C, Bocker W, Hertle L, Terpe HJ. Chromosomal aberrations associated with invasion in papillary superficial bladder cancer.
J Pathol
1998
;
185
:
345
–51.
25
Hovey RM, Chu L, Balazs M, et al. Genetic alterations in primary bladder cancers and their metastases.
Cancer Res
1998
;
58
:
3555
–60.
26
Muscheck M, Abol-Enein H, Chew K, et al. Comparison of genetic changes in schistosome-related transitional and squamous bladder cancers using comparative genomic hybridization.
Carcinogenesis
2000
;
21
:
1721
–6.
27
Cianciulli AM, Bovani R, Leonardo C, et al. DNA aberrations in urinary bladder cancer detected by flow cytometry and FISH: prognostic implications.
Eur J Histochem
2001
;
45
:
65
–71.
28
Waldman FM, Carroll PR, Kerschmann R, Cohen MB, Field FG, Mayall BH. Centromeric copy number of chromosome 7 is strongly correlated with tumor grade and labeling index in human bladder cancer.
Cancer Res
1991
;
51
:
3807
–13.
29
Erill N, Colomer A, Verdu M, et al. Genetic and immunophenotype analyses of TP53 in bladder cancer: TP53 alterations are associated with tumor progression.
Diagn Mol Pathol
2004
;
13
:
217
–23.
30
von Knobloch R, Bugert P, Jauch A, Kalble T, Kovacs G. Allelic changes at multiple regions of chromosome 5 are associated with progression of urinary bladder cancer.
J Pathol
2000
;
190
:
163
–8.