High throughput comparative genomic hybridization array analysis of multifocal urothelial cancers

Abstract

The purpose of this study was to examine genetic alterations occur during synchronous or metachronous multifocal development of urothelial cancers on the whole genome using a comparative genomic hybridization (CGH) array. We used 10 tumor pairs (2 tumors for each patient), in which we had previously defined a clonal relationship by microsatellite analysis. For CGH array analysis, Vysis GenoSensor Array 300 kit was used. An unsupervised hierarchical cluster analysis revealed that the tumors from one patient were clustered together independent of the tumors of all other patients. On the other hand, many genetic divergences among multifocal urothelial cancers were newly found by a CGH array analysis. The concordant genetic alteration patterns of the chromosomal arm in tumor pairs were most frequently observed in 9p, 9q, 8p, 7p, 7q and 11q, while discordant patterns were most frequently found in 15q, 20q, 2q, 10p and 11q. Investigation using a CGH array showed that genetically stable multifocal tumors were less frequent, and that a large percentage of urothelial cancers accumulate genetic alterations during multifocal development by clonal evolution. We might have to consider these genetic accumulations during multifocal development when designing strategies for prevention and detection of recurrent multifocal urothelial cancers. CGH array can be a powerful tool for genetic analysis of multifocal urothelial cancer. (Cancer Sci 2006; 97: 746–752)

Abbreviations:

BAC

bacterial artificial chromosome

CGH

comparative genomic hybridization

DAPI

4,6,‐diamidino‐2‐phenylindole

dCTP

2′‐deoxycytidine 5′‐triphosphate

FISH

fluorescence in situ hybridization

LOH

loss of heterozygosity

PAC

P1‐derived artificial chromosome

PCR

Polymerase Chain Reaction

SNP

single nucleotide polymorphisms

WHO

world health organization

SSC

sodium chrolide‐sodium citrate.

Urothelial cancers have two clinically important features, multifocality and recurrence. Around 30% of urothelial cancers are found as multiple tumors at the time of diagnosis.^{( 1 )} Urothelial cancers occur most often (70%) as superficial cancer that can be treated by endoscopic treatment, and 60–80% of patients present one or more recurrences after initial treatment. This multifocal nature of superficial urothelial cancer has been a good material to trace accumulation of genetic alterations in multifocal development of tumor and provided insights into relationship between genetic alterations and carcinogenesis.

Many studies using molecular analysis have suggested a monoclonal origin for multifocal urothelial cancer as appears from X chromosome inactivation studies and genetic and cytogenetic analyzes,^{( 2} , ³ , ⁴ , ^{5 )} while other studies have shown an independent origin.^{( 6} , ^{7 )} In our previous LOH analysis of multifocal urothelial cancer using 20 microsatellite markers, we demonstrated that genetic alterations detected were stable in 9 (64%) of the 13 patients with multiple metachronous tumors with a possible identical clonal origin.^{( 5 )} Because urothelial cancer is characterized by highly complex chromosomal changes affecting numerous chromosomal loci, genome‐wide screening may identify genetic divergence among multifocal urothelial cancers that have been missed by the methods applied to date.

Detection of recurrent tumors by non‐invasive technique has been an important issue in clinical management of superficial urothelial cancer. Genetic analysis of urine sediments in urothelial cancer patients has been used for diagnosis and follow‐up of urothelial cancers, such as FISH, microsatellite analysis and SNP array.^{( 8} , ⁹ , ¹⁰ , ^{11 )} However, there have been few reports describing the genetic alterations of metachronous multifocal tumors in increased resolution of current technique of DNA analysis.

In this study, to investigate genetic alterations occur during multifocal development of superficial urothelial cancers throughout the genome, we used a Vysis Genosensor CGH array kit with 287 target clone DNAs to analyze 20 tumor specimens from 10 patients in which a clonal relationship had been defined in the previous study using microsatellite analysis.

Materials and Methods

Patients and tumor samples

Topologically distinct urothelial cancers of the bladder, ureter, and renal pelvis in 10 patients were included in this study (2 tumors for each patient). Tumor samples were snap‐frozen and stored −80°C until DNA extraction. The present study did not include tumor materials which contaminated more than 20% of normal interstitial cells in hematoxylin and eosin (HE) staining. Normal reference DNA was obtained from the peripheral blood of each patient. Tumor and reference DNAs were prepared by proteinase K digestion and phenol/chloroform extraction. The tumor stage and grade were classified according to the tumor‐node‐metastasis system and WHO criteria, respectively,^{( 12} , ^{13 )} by pathologists who were unaware of the aims of this study. All tumors were urothelial carcinomas, and brief clinical and pathological data are presented in Table 1.

Table 1.

Characteristics of 20 multifocal urothelial tumors of 10 patients

Patient		Tumor	Interval (month)	Site †	Grade	pT
Group I	9	2		B2	1	a
		8	38	B4	1	a
	26	1		B5	2	1
		2	0	B2	2	1
	2	2		B3	1 > 2	a
		8	3	P	1,2	a
	18	1		B2	1	a
		2	12	B2	1	a
	10	1		P	2	1
		2	7	B2	2	1
	15	1		B2	2	1
		2	37	B2	2	1
Group II	21	1		B1	2	1
		5	8	B2	2 > 1	a
	16	1		U	2	a
		4	7	B2	2	NE ‡
	23	1		P	2	3
		2	18	B4	2	a
	30	1		B2	2	a
		2	8	B5	2	a

Patients were divided into two groups by the previous microsatellite analysis.^{( 5 )} Group I patients had concordant LOH patterns alone. Group II patients had both discordant and concordant LOH patterns. ^†U, ureteral tumor; P, renal pelvic tumor; B, bladder tumor [the locations of the bladder tumor were as follows: B1, trigone; B2, posterior wall; B3, right wall; B4, left wall, B5, dome]; ^‡NE, not examined.

These specimens were analyzed in our previous LOH study and a clonal relationship had already been defined.^{( 5 )} Using microsatellite markers, we examined genetic alterations at 20 loci on eight chromosome arms (2q, 4p, 4q, 8p, 9p, 9q, 11p, and 17p). The markers we used D2S206 and D2S336 on chromosome 2q; D4S404 and D4S1546 on 4p; D4S426 and D4S171 on 4q; D8S261 and D8S520 on 8p, D9S171, D9S126, D9S1749 and D9S736 on 9p; D9S66, D9S1848, D9S1793 and GSN on 9q, D11S907 and D11S922 on 11p; and D17S796 and D17S1176 on 17p. Completely identical LOH patterns were detected in six patients (No. 9, 26, 2, 18, 10, and 15), while discordant patterns together with concordant patterns were observed in the remaining four patients (No. 21, 16, 23, and 30). In patient 30, no clonal relationships could be defined in the previous microsatellite study. All tumor pairs in other patients were considered to be monoclonal based on the results of microsatellite analysis. Written informed consent was obtained from each patient before surgery, according to the ethical guidelines of our university.

Array‐based CGH (Array CGH)

Array‐based CGH was performed using a GenoSensor Array 300 kit (Vysis, Downers Grove, IL), which contained 287 target clone DNAs (P1, PAC or BAC clones) representing regions that are important in cytogenetics and oncology. Each DNA clone was arrayed on the slide in three spots. Tumor and reference DNAs (100 ng) were labeled with Cy3‐dCTP or Cy5‐dCTP (Perkin Elmer, Wellesley, MA), respectively, by random priming reaction. Labeled DNAs were mixed with a microarray hybridization buffer that contained a high concentration of Cot‐1 DNA (Vysis). The probe mixture was denatured at 80°C for 10 min and incubated at 37°C for 1 h before being transferred to the microarray. After an overnight hybridization at 37°C, the microarrays were washed three times with 50% formamide/2XSSC at 40°C for 10 min each. Thereafter, the microarrays were subjected to four washes with 1XSSC at room temperature for 5 min each. The microarrays were counterstained with the DAPI IV mounting solution (Vysis).

Images of fluorescence signals were captured and analyzed with the GenoSensor Reader System and GenoSensor Reader software (Vysis) according to the manufacturer's instructions. The mean normalized test/reference (T/R) ratio was calculated from each set of T/R ratio of three spots. Gain and loss of DNA copy number were judged by T/R ratios >1.2 and <0.8, respectively. Two different sets of labeling and hybridization tests were performed on two tumors from one patient (Patient No. 15) to validate the reproducibility of the GenoSensor Array 300 kit. Spots on which the difference in the ratios was greater 0.2 were 3 (1%) of 286 evaluated spots in tumor 1 and 1 of 286 (0.3%) in tumor 2.

Data analysis of array CGH Results

We performed an unsupervised hierarchical cluster analysis on log base 2 transformed data obtained for 287 target clones to weight how the prevalence of genomic changes is similar or different. The Ward linkage and cosine coefficient metric were used. The results were visualized in the software program R.

In order to assess global chromosomal aberrations in multifocal urothelial cancers, our subsequent analysis focused to ‘the pattern of chromosomal aberrations between tumor pairs in each patient’. On one clone of the array, an identical genetic alteration of two tumors in one patient was defined when a loss or gain was detected and the result was identical. Different genetic alterations were defined as follows: 1. Loss or gain was detected; 2. The result was different in the two tumors; 3. The difference in the ratio of the two tumors was greater than 0.2. Then, the pattern of chromosomal aberrations compared between tumor pairs was evaluated on each chromosomal arm. The pattern of chromosomal aberrations was defined as concordant or discordant if identical or different gains or losses were detected in 2 or more clones on the chromosomal arm. If both discordant and concordant patterns coexisted, we defined the chromosomal arm as discordant. A chromosomal arm which had only one clone with a gain or loss was neglected.

Quantitative real‐time PCR

Quantitative real‐time PCR analyzes were performed on the same set of urothelial cancers to detect copy number changes for four genes, BIN1 (2q14), CDKN2A (p16) (9p21), CCND1 (11q13), and MAP2K5 (15q23), which were included in the CGH array target DNAs. The relative gene copy number was evaluated by the comparative C_T Method as described by Ginzinger et al.^{( 14 )} on the ABI Prism 5700 Sequence Detection System (PE Applied Biosystems, Foster City, CA). The PFKL gene served as an internal control, located on 21q22.3, which is rarely altered numerically in urothelial cancers.^{( 15 )} Twenty normal peripheral blood samples were examined to determine the reference range.

Results

Overall results

Of 287 clones examined in 20 tumors (5740 in total) gain and loss were observed in 437 (7.6%) and 512 (8.9%), respectively. On average, one tumor had gains in 21.9 (4–42) clones and losses in 25.6 (1–53) clones. The identical alterations between tumor pairs were most frequently observed in CDKN2A (p16) (9p21), MTAP (9p21.3), and AFM137XA11 (9p11.2) (6 patients), CTSB (8p22), D8S596 (8p Tel.), D8S504 (8p Tel.), AF170276 (9p Tel.) and D9S913 (9p Tel.) (5 patients). On the other hand, the different alterations between tumor pairs were frequently identified in SNRPN (15q12), UBE3A, D15S10 (15q11‐q13) and STK6 (20q13.2‐q13.3) (4 patients), MYBL2 (20q13.1) and PTPN1 (20q13.1‐q13.2) (3 patients). These genes/loci may be susceptible to genetic alterations during multifocal development of urothelial cancers.

Hierarchical clustering analysis

Fig. 1 shows the cluster dendrogram of 20 tumors of 10 patients based on the similarity of genetic alterations detected by CGH array analysis. Tumor pairs from one patient were clustered together. One tumor was more related to the second tumor from the same patient than tumors from all other patients, suggesting that these tumor pairs were clonally related.

Figure 1.

Hierarchical clustering of data from 287 clones on the GenoSensor array 300 in 20 tumors. The data are presented in a matrix format. A column corresponds to a single tumor, and each row corresponds to a single clone ordered by mapping position. Gain or loss of a clone is represented by the color of the corresponding cell in the matrix. Green indicates gain; black, no change; and red, loss. Color saturation is proportional to the magnitude of the difference. The sidebars to the left of the matrix format represent chromosome cluster, ordered from chromosome 1 to Y. The horizontal dendrogram shows the association between the tumors and the length of the branches of the dendrogram reflects the similarity between tumors. Note that one tumor is more related to the second tumor from the same patient than tumors from all other patients.

The pattern of chromosomal aberrations on the CGH array

On average, concordant patterns of chromosomal aberrations were detected in 5.8 chromosomal arms (range 1–13) in one patient; while discordant patterns were detected in 4.2 chromosomal arms (range 0–12). Fig. 2 summarizes the patterns of chromosomal aberrations.

Figure 2.

Summary of genetic alteration patterns of chromosomal arms.

In 8 of 10 patients, both concordant and discordant patterns were observed. In patient No. 9, concordant patterns were found in chromosomal arms 8q (gain), 9p (loss), 11p (partial loss), 12q (partial gain) and 22q (partial loss), while discordant patterns were found in 1p, 2p, 8p, 9q, 18p and 18q (Fig. 3a,b). In 2 of 10 patients, the patterns of chromosomal aberrations were identical. The results of patient No. 15 are shown in Fig. 3c,d. Of note, these two tumors developed with a 3 years interval and the patient had 4 tumor recurrences between the two tumors. These data suggest that in part of superficial bladder cancers, the whole genome is genetically stable during development of metachronous cancers.

Figure 3.

(a,c) Graphical representation of CGH array analysis of patient No. 9 and 15, whole genome. Vertical lines indicate the boundaries of chromosomes. Both concordant and discordant patterns of chromosomal aberrations were observed in patient No. 9, while the patterns of chromosomal aberrations were completely identical in patient No. 15. (b,d) Representative chromosome arms. In patient No. 9, a concordant pattern was found in 9p (loss), while discordant patterns were found in 9q, 18p and 18q. In patient No. 15, loss of 8q, gain of 8q, and a partial loss and gain of 11q were observed in both tumors. Vertical lines indicate boundaries of chromosome arms. Average log₂ ratios were plotted for all clones on chromosome position. Thresholds for gain or loss are shown within log₂ ratios of 0.2 and −0.2, respectively. T; tumor, Chr; chromosome.

Concordant patterns of chromosomal aberrations between tumor pairs were most frequently observed in 9p (7 patients), 8p (5 patients), 9q, 22q (4 patients), 7p, 7q, 11q, 17p and 17q (3 patients). Among concordant patterns, aberrations on 9p, 9q, 8p and 17p were losses of the chromosomal arm, and on 7p, 7q and 17q were gains. Discordant patterns were most frequently found in 15q (5 patients), 20q (4 patients), 2q, 10p, and 11q (3 patients). Among the total 42 discordant patterns, simple addition of genetic alterations to latter tumor was found in only 7, which suggests that the accumulation of genetic alterations does not parallel the chronological order of tumor occurrence.

Comparison of CGH array and PCR‐based microsatellite analysis

Exactly same loci to 20 microsatellite markers used in the previous study were not included in the CGH array target DNAs. Of 20 microsatellite markers used in the previous study, 6 markers for which the distance to the nearest spot on the CGH array was within 1 Mb length were selected and the results of the 2 methods were compared. Of 104 informative microsatellite analyzes using 6 markers, D8S520, D9S126, D9S1749, D9S1848, D9S66 and D11S922, 61 showed LOH. Overall, of 104 informative microsatellite analyzes, 59 (56.7%) matched the results of the CGH array. Among the 61 LOH, 36 (59%) were detected as gains or losses by the CGH array. The results of the CGH array analysis and microsatellite analysis were not closely matched, probably it is because the loci of most microsatellite markers were not exactly same as those of the spot on the CGH array. In addition, the difference in the result may be caused by the difference of the assay platform; microsatellite involves amplification step while the CGH array does not.

Validation of CGH array analysis by quantitative Real‐time PCR

Quantitative real‐time PCR was performed on all tumor samples using four genes. We calculated the Pearson correlation coefficient (r) between the results of the CGH analysis and quantitative real‐time PCR analysis. We found a strong correlation BINI (r = 0.75), CDKN2A (r = 0.70), CCND1 (r = 0.96), and MAP2K5 (r = 0.71) between them. Both methods matched in 87.5% (70 of 80) when the copy numbers of each gene/locus were decided by using their respective cutoff criteria. The discordant results between them were found only in the case which the genes/loci had the subtle copy number gain/loss. The discordance of the results using both methods may be inevitable to some degree in cases of subtle copy number gain/loss. In addition, the difference in the result is probably caused by the difference of assay methodology, or the low resolution of the CGH array. Among 7 tumor pairs showing a discordant pattern of chromosomal aberrations in the CGH array, we also obtained matched results in 4 pairs in quantitative real‐time PCR analysis.

Discussion

In the present study, using commercially available CGH array kits, we investigated global chromosomal aberrations in multifocal urothelial cancers. The multiloci survey by the CGH array of the whole genome enabled us to detect genetic alterations on the loci that we did not detect in the previous study. In the previous report, 64% of tumor pairs showed completely identical LOH patterns and we concluded that a high percentage of superficial urothelial cancers are genetically stable. However, among tumor pairs of 6 patients with completely concordant LOH patterns on microsatellite analysis, discordant patterns of chromosomal aberrations between tumor pairs were newly found in 4 patients using the CGH array. The current results showed that such genetically stable multifocal cancers are rather infrequent and that a large number of superficial urothelial cancers accumulate genetic alterations during multifocal development. On the other hand, in patients 18 and 15, only concordant chromosomal aberration patterns were observed even by genome‐wide CGH array analysis, which indicates that some multifocal urothelial cancers are genetically stable during multifocal development.

Since the introduction of molecular analysis into clonal analysis, the majority of multifocal urothelial cancers have been suggested to be monoclonal or have a common clonal origin.^{( 4} , ¹⁶ , ^{17 )} In multistep carcinogenesis model, an original transformed cell grows out and sheds cells into the lumen of the bladder. Some of these cells will have required additional genetic alterations. Interestingly, among the 42 discordant patterns, simple addition of genetic alterations to latter tumor was found in only 7. In these tumor pairs, the latter tumor is not directly derived from the former tumor. van Tilborg et al. performed the LOH analysis of metachronous multifocal bladder cancer using 48 microsatellite markers,^{( 18 )} and reported similar findings to our data. To develop a strategy to prevent and detect recurrent multifocal urothelial cancers, it should be remembered that the chronology of tumor appearance does not always run in parallel with the accumulation of genetic alterations during the process of intraepithelial spread or intraluminal seeding. The commercially available CGH array used in this study is a relatively low‐resolution scan, and much higher resolution scans of the bladder cancer genome have been reported.^{( 19 )} By using a CGH array with high resolution and high throughput for analysis of a larger number of samples, we might be able to draw more detailed genetic trees and pedigrees of multifocal urothelial cancers and trace genetic alterations accumulated during multifocal development.

The development and progression of urothelial cancer is the result of a series of genome instability occurring over the lifetime of a tumor. The number of chromosomal alterations was reported to be higher in high grade, advanced stage tumors.^{( 20} , ^{21 )} Discordant chromosomal aberrations between multifocal tumors might reflect genome instability as well. It would be useful to know whether the degree of discordant chromosomal aberrations has an association with recurrence, progression, tumor grade and stage. Considering that most tumors we analyzed were low grade superficial type, it is likely that many of the chromosomal aberrations we detected were not related to differences in grade or stage. With accumulation of data on the chromosomal aberrations of multifocal tumors, it might become possible to identify target chromosomal regions and target genes which play an important role in recurrence, progression and invasion of urothelial cancers.

A number of studies have reported the detection of tumor cells in urine. Halling et al. reported high sensitivity (81%) of multicolor FISH analysis consisting of probes for chromosome 3, 7 and 17, and the 9p21 band in detecting urothelial cancer.^{( 8 )} Several investigators showed that 85% or greater sensitivity was achieved using 13–20 highly informative microsatellite markers to detect genetic alterations in urinary exfoliated cells.^{( 9} , ^{10 )} Furthermore, Hoque et al. used a SNP array to detect urothelial cancer and reported a 100% detection rate.^{( 11 )} Our data suggest that we should be careful when we use these genetic methods for monitoring urothelial cancer recurrence, because the genetic alterations we detect in primary tumors may not be present in recurrent tumors. In the current study, concordant patterns of chromosomal aberrations were frequently detected on 9p, 8p, 9q, 22q, 7p, 7q, 11q, 17p and 17q. The loci on these chromosomal arms may be good candidates for probes for efficient monitoring of urothelial cancer recurrence. Furthermore, our data is consistent with the view that genetic alterations on chromosome 9 occur early in the development of multifocal urothelial cancer and remain stable during clonal progression. Accumulation of data regarding chromosomal aberrations among multifocal urothelial cancers would be necessary to improve efficiency and reduce the cost of the genetic analysis to detect tumor cells in urinary exfoliated cells.