Allelic Deletions of Cell Growth Regulators during Progression of Bladder Cancer¹

One of the most important features of urothelial cancers of the bladder and upper urinary tract is metachronous or synchronous multifocal occurrence with high frequency. Between 70% and 80% of patients with bladder cancer have only noninvasive disease (Ta) or tumors with invasion no deeper than the lamina propria (T1) on initial presentation, and the remainder have muscle-infiltrating or deeper cancers (T2-T4) (1). The risk of developing a muscle-invasive disease is only 10% in a patient with a noninvasive bladder tumor (Ta), whereas the majority of patients diagnosed with concomitant carcinoma in situ (flat lesion grade 3–4) will develop a muscle-invasive tumor. At present, one cannot predict which patient with a noninvasive tumor will experience progression to invasive disease and which one will not. The search for predictive markers has been the aim of many studies and has included, among others, studies of nuclear volume (2), blood group antigens (3, 4), adhesion potential (5, 6),epidermal growth factor receptors, (7) and matrix metalloproteinases (8). However, these markers have only shown a general relation to prognosis and provide no definitive information for the specific patient. In recent years, a large body of information has been accumulated on the growth-regulatory pathway through which the p53 and Rb proteins are working, including proteins encoded by genes like MDM2, CDKN1A (p21), CDKN2A(p16, p14arf), RB1, E2F, and MYCL (Fig. 1). In the present study, we investigated allelic losses of these genes as predictors of disease course. Alterations of the TP53gene seem to be of importance in most cancers. It is known that LOH³of the TP53 gene is correlated to high grade and stage of bladder tumors (9, 10), and LOH of 17p (the TP53 locus) is associated with an invasive phenotype(11). Furthermore, positive immunostaining for p53 protein correlates with disease progression (12). Other components of the p53 and Rb pathways have also been investigated in bladder cancers, including MDM2 (13, 14), p21(15, 16) CDKN2A (17), E2F (18), and RB1 (19, 20). Despite the many studies of these components in bladder cancer (9, 10, 17, 21, 22), very few, if any, have analyzed the various components of this pathway in single individual tumors to verify the number of different alterations. This could be of importance because losses of some gene products may promote cancer,whereas others may inhibit it. Furthermore, alterations of specific genes should be interpreted in the context of the presence or absence of downstream effector proteins. Based on this, we examined all of the shown growth regulators (Fig. 1) related to the p53 and Rb pathways in each tumor from three groups of patients: (a)patients with primary muscle-invasive tumors; (b) patients with recurrent noninvasive tumors; and (c) patients with progressing tumors. A number of novel findings relating allelic deletions to clinicopathological data were made. We detected a significant difference in the numbers of cell growth-regulatory gene loci affected by allelic deletions in low-stage versushigh-stage tumors, frequent deletions at the MYCL locus, a difference in the pattern of allelic deletions in high-stage tumors with and without field disease, a more pronounced deletion of TP53 and RB1 in the progressing tumor group, and the presence of combined deletions of TP53 and RB1 only in the progressing tumor or muscle-invasive groups.

Three groups of patients were selected from a clinical data and tissue bank from approximately 1000 patients followed for more than 3 years on average. The noninvasive group consisted of 23 patients (18 males and 5 females; median age, 73 years; age range, 42–83 years) who had at least three metachronous stage Ta tumors and did not have a tumor of higher stage during a median follow-up of 205 weeks (range, 72–218 weeks). The muscle-invasive tumor group consisted of 22 patients (20 males and 2 females; median age, 67 years; age range, 46–84 years) who had a stage T2 or higher stage tumor as their first bladder tumor ever. The progressing group consisted of 23 patients (17 males and 6 females;median age, 71 years; age range, 50–83 years) who had a stage Ta or T1 tumor as their first tumor and whose disease later progressed to a higher stage. We included both Ta and T1 to get a reasonable number of tumors. From each patient, tumor tissue and blood were collected if informed consent was obtained.

Tumors were obtained fresh from surgery, frozen immediately, and stored at −80°C. DNA was extracted from tumor tissue and blood by using a Puregene DNA extraction kit (Gentra Systems, Minneapolis, MN) following the manufacturer’s instructions. Because of possible normal tissue contamination, all invasive tumors without evident allelic losses were reanalyzed using microdissected tumor tissue. Microdissection was performed using ×100 magnification in a microscope and 4–10-μm-thick serial sections of paraffin-embedded tumors. Microdissected tumor areas were deparaffinized by heating to 70°C,followed by rinsing in Xylol. DNA was extracted as described above.

DNA from tumor and blood was analyzed for allelic deletions and MIN by using microsatellite markers. Microsatellites were chosen near growth-regulatory genes using information from the National Center for Biotechnology Information database Genemap 98 or based on published data (23, 24). Sequences of the primers used are listed in Table 1. Fluorescence-labeled primers were purchased from Hobolt DNA Synthesis(Hilleroed, Denmark) and DNA Technology (Aarhus, Denmark).

PCR reactions were carried out in a 19-μl volume containing 100–200 ng of purified genomic DNA; 2 pmol of each primer; 50 mmKCl; 10 mm Tris (pH 9.0); 1.5 mmMgCl₂; dATP, dTTP, dCTP, and dGTP (132 mm each); and 0.5 unit of Taq DNA polymerase. The reaction mixture was subjected to 5 min of denaturing at 95°C and 30–40 cycles of 95°C for 0.5–1.5 min, 50°C–63°C for 45 s to 1.5 min, and 72°C for 1.5 min. The final cycle was followed by a step at 72°C for 5 min. PCR conditions were optimized according to the sequence of the primers, and PCR was carried out in a MJ Research Thermocycler (PTC 200/PTC 225).

PCR products were analyzed by electrophoresis on 4.75%polyacrylamide/6 m urea denaturing gels using an ABI Prism 377 DNA sequencer and Genescan software (3 h, 51°C, 2.96 V).

MIN was defined as the presence of significant new bands, following PCR amplification of tumor DNA, that were not present in corresponding normal DNA. Tumors with MIN were excluded from allelic deletion examination. Allelic deletion was defined as loss of a significant part of one allele in tumor DNA compared with the corresponding normal DNA. This was determined by comparing the area under the curve for the two alleles from tumor and normal DNA in the curves obtained from the Genescan program. To determine the between-run variation, we repeated the entire procedure on 10 samples of corresponding tumor and normal DNA for the first six primer sets. We normalized the areas by dividing the ratio between tumor alleles with the ratio between normal alleles [(tumor allele 1:tumor allele 2)/(blood allele 1:blood allele 2)]. Based on these data, the variation of each amplicon was defined,and a significant loss was defined as a difference of more than 3 SDs between the allelic ratio in tumor and the allelic ratio in blood. Determinations of area under the curve in the first six primer sets were compared with a visual scoring by two independent observers and showed a high degree of correlation (93%). Therefore, we decided to use visual scoring by two independent and experienced observers on the rest of the amplicons. All cases with allelic deletions or MIN in primary analysis were confirmed by repetition of the complete assay and rescoring.

To analyze the importance of the number of genomic deletions in a single tumor, we divided the material into tumors having only one or no deletion, (low frequency) and tumors having two or more deletions (high frequency).

Two different tests were used. For samples larger than 50 and figures expected to be larger than 5, we used the χ²test. For smaller samples and figures, we used the two-sided Fisher’s exact test. When one locus was analyzed with more than one primer, the combined result was counted as one event.

The deletion of alleles was examined at the following loci in each tumor: TP53 (two markers); MDM2; CDKN1A(p21); MYCL; CDKN2A (p16 and p14arf); RB1; and E2F (Fig. 2). A pronounced difference in the frequency of allelic deletion was found between the noninvasive group of tumors and the muscle-invasive group of tumors. Seven of the 23 noninvasive tumors showed deletion of at least one locus compared with 19 of 22 muscle-invasive tumors (Table 2). In the muscle-invasive tumors, an average of 1.6 loci were deleted compared with 0.3 loci in the tumors of the noninvasive group(P = 0.0000002, χ²test). If a value of ≥2 deletions was used to define tumors with frequent deletions, then the number of tumors with frequent deletions amounted to 11 of 22 muscle-invasive tumors and only 1 of 23 noninvasive tumors (P = 0.001; two-sided Fisher’s exact test; Table 2) because tumor 4 from patient 60 was the only one that lost both a TP53 and an E2F allele in the latter group. It was remarkable, however, that half of the muscle-invasive tumors had only one deletion (eight tumors) or no deletions (three tumors; Table 2).

In the muscle-invasive tumors, different combinations of deletions were detected from deletion of one RB1 allele to deletion of TP53, CDKN1A, CDKN2A, and RB1 alleles in a single tumor. The RB1 and MYCL loci showed significantly more deletions in muscle-invasive tumors than in noninvasive ones [RB1, P = 0.001; MYCL, P = 0.003 (two-sided Fisher’s exact test)]. CDKN2A, which is located at chromosome 9p, was deleted with a similar frequency in noninvasive (3 of 13 tumors) and muscle-invasive(6 of 17 tumors) tumors, as expected.

The muscle-invasive tumors could be separated into two groups with or without concomitant carcinoma in situ. The patients with concomitant lesions have a field disease with grade 2 dysplasia or carcinoma in situ/invasive carcinoma in selected site biopsies. As a novel finding, we observed that the deletions of MYCL and RB1 alleles were more pronounced in patients having field disease because 11 of 14 informative cases showed deletions compared with 2 of 8 cases without field disease(P = 0.04, two-sided Fischer’s exact test). All tumors with more than two deletions had concomitant field disease(Table 2; Fig. 3).

Because of the significant difference between the noninvasive and muscle-invasive tumors, we decided to examine losses of TP53, CDKN2A, MYCL, and RB1in a group of patients with progressing tumors and in patients with recurrent noninvasive tumors (Table 3). As a novel finding, we demonstrate that the number of patients with allelic deletions of RB1 in any tumor was significantly higher in the progressing group (7 out of 21 informative cases) than in the recurrent Ta group (0 out of 17 informative cases; P = 0.02, two-sided Fisher’s exact test). For the TP53 locus, 12 of 17 versus 3 of 19 informative patients had deletions (P = 0.002, two-sided Fisher’s exact test). We compared all Ta tumors later developing into invasive tumors with tumors from patients with recurrent Ta tumors and found allelic deletion of RB1 in 2 of 11 tumors in the progressing group versus 0 of 43 tumors in the recurrent noninvasive group(P = 0.076, two-sided Fisher’s exact test). For TP53, the numbers were 4 of 9 versus 4 of 46(P = 0.037, two-sided Fisher’s exact test). The most striking difference was that RB1 never showed deletions in the group with recurrent noninvasive tumors. The CDKN2A locus and MYCL were deleted to the same extent in both groups. The deletions of RB1 were present from the beginning of the progressive disease course in all but one case, as was TP53 except in three cases. Deletions of alleles at these two loci could be important predictors for the disease course. Deletions were not detected at all visits in some cases (Table 3). This might indicate different tumor cell populations. As shown by case 679 in the recurrent noninvasive group, different populations of tumor cells seemed to be present because the CDKN2A locus showed deletion in tumor 1 and a normal pattern in tumor 3, whereas the MYCL gene locus showed the opposite pattern (Fig. 2 C). Seven of the 23 patients that showed progression had no deletions at all of the examined loci. Intervals between recurrences were similar in the two groups. An average of 236 days passed between the recurrence of noninvasive tumors in the group that did not progress, and an average of 217 days passed from the last Ta or T1 tumor to the first tumor of a higher stage in the progressing group.

If we define the simultaneous deletion of two or more gene loci in TP53, MYCL, RB1, and CDKN2Aas a high frequency of deletions, then 10 patients showed a high frequency of deletion in the progressing group, and only 1 patient showed such a high frequency in the recurrent noninvasive group. The number of tumors with a high frequency of deletion differed markedly between the groups examined (Fig. 3) but was always 50% or less of examined tumors.

We have studied the allelic deletion of genes involved in cell cycle regulation in a well-characterized clinical material consisting of bladder cancer patients with stable or progressing tumors. We found a significantly higher frequency of allelic deletions in high-stage versus low-stage tumors and, as a novel finding, in progressing versus recurrent noninvasive disease. The genes most often affected by allelic deletions in muscle-invasive tumors were TP53, RB1 and, as a novel finding, MYCL. Dichotomizing the patients into those with tumors having a high frequency of deletions (two or more deletions) and those with a low frequency of deletions showed, for the first time, a remarkable overrepresentation of high-frequency deletions in progressing tumors. Although allelic deletions were frequent in certain patient groups, some patients in each group showed a complete absence of deletions, indicating that the muscle-invasive or progressing tumors might consist of two different groups, one with allelic deletions and one with other characteristics.

The higher frequency of allelic deletions in muscle-invasive tumors corresponds well with the findings by other authors that mutations,allelic loss, and abnormal expression of the p53 protein are more pronounced in high-stage tumors than in low-stage tumors (9, 10, 12). Allelic loss was found mainly in RB1, MYCL, CDKN2A, and TP53 and was rare for the genes MDM2 and E2F.

mdm2 up-regulation is expected to be unfavorable because the protein abrogates the growth suppression function of p53. In human breast cancer, mdm2 overexpression was seen in 24 of 33 cases(25). In bladder cancer, a positive correlation between p53 accumulation and mdm2 overexpression was shown, but mdm2 overexpression alone had no prognostic significance (13). The CDKN2A gene encodes cell cycle-regulatory proteins p16 and p14arf, which share one exon with different reading frames. Both are tumor inhibitors; p14 binds to mdm2 and inhibits p53 degradation,and p16 binds to cyclin-dependent kinase 4 (Fig. 1; Ref.26). A microsatellite at the IFNA locus is commonly used to investigate the CDKN2A locus (27, 28); however,deletions may occur that affect only the CDKN2A gene and not IFNA. Based on this, our findings concerning CDKN2A may underestimate the frequency of deletions of the CDKN2A gene. The tumor inhibitor p21, like p16, works through inhibition of cyclin-dependent kinases, abrogating the phosphorylation of rb1 and arresting the cell cycle at the G₁ checkpoint. The CDKN1A gene is directly transcribed by p53 and is related to morphological indicators of cell proliferation. Its relation to prognosis in bladder cancer is ambiguous (15, 16). In a previous publication (17), the CDKN2A gene was deleted in 13 of 140 bladder tumors, all of which had small defined deletions at 9p21.

The E2F gene product is a transcription factor that promotes G₁ progression. Loss or inactivation of this protein could, in theory, arrest the cell cycle, but in a recent publication (18), lack of the e2f protein was related to disease progression, and e2f inactivation may be related to replication errors (29). The rb1 protein binds to e2f and thus causes cell cycle arrest. Individuals with germ-line mutations in the RB1 gene are at high risk of developing retinoblastoma and other cancers (30). Altered expression of rb1 is associated with invasive bladder tumors (31) and bladder cancer recurrence (19), and the alteration of p53 and rb1 in the same tumor is associated with tumor progression(20). The role of MYCL is difficult to interpret. The MYCL gene is a proto-oncogene. Among other mechanisms, it might work through inhibiting the p21-mediated inhibition of rb1 phosphorylation (32). Another member of the myc family, C-MYC, plays a role in p53-dependent apoptosis(33). The role of MYCL is less clear because it accelerates apoptosis after interleukin 3 withdrawal, whereas overexpression produces resistance to cytotoxic drugs(34). We found one previous publication (35)on MYCL and bladder cancer in which the authors concluded that the MYCL genotype was not a prognostic factor in bladder cancer. In our material, allelic deletion of MYCLwas associated with high-stage tumors, possibly indicating a growth advantage for cells with only one MYCL allele.

Lack of p53 protein is supposed to lead to increased cell division because the CDKN1A gene will not be encoding p21 protein,and there will be increased action of cyclin-dependent kinases and release of e2f, leading to S-phase progression and possible tumor growth. This will be enhanced by the lack of p16 protein, which is needed to stop the cyclin-dependent kinase-dependent phosphorylation of rb1 and release of e2f. As a result, loss of TP53, CDKN1A, CDKN2A, and RB1 may give the cell a growth advantage, whereas loss of E2F and MDM2 may stop S-phase progression and possible tumor growth(32). It was characteristic in our material that E2F and MDM2 were rarely hit by deletions compared with the other genes, indicating that almost only unfavorable deletions leading to increased S-phase progression were present.

However, if the deletions we observed have any biological importance,they should influence the amount or quality of the gene products. This could be due to either a gene dose effect or inactivation of the remaining allele by a mutation or by methylation of CpG islands in the promoter region. The latter is a known mechanism that reduces the transcription of genes (36). There are some possibilities for underscoring deletions. We tried to rule out normal tissue contamination by microdissecting paraffin-embedded tissue in muscle-invasive tumors that showed no deletions when frozen tumor tissue was examined. Another possibility is too much template. Because almost all tumors showed some degree of deletion, it seems that saturation of the PCR process is no problem. Other authors have used 5–300 ng of template (37, 38). We conclude that allelic deletions of genes involved in cell growth regulation are of importance in the progression of bladder cancer. We detected a pronounced difference in the frequencies of deletions in noninvasive versus muscle-invasive tumors. If we define the simultaneous loss of two or more gene loci in TP53, MYCL, RB1, and CDKN2A as a high frequency of deletions,then 11 of 23 muscle-invasive tumors and only 1 of 23 noninvasive tumors showed this pattern. A striking new finding was that 10 patients in the progressing group and only 1 patient in the recurrent noninvasive group showed a high frequency of deletions. The differences were more pronounced for RB1, MYCL,and TP53. The number of tumors showing a high frequency of deletions was markedly different between the groups examined but was always 50% or less of examined tumors. Based on this, one might suggest that two different pathways may lead to bladder tumor progression, one in which deletion of alleles in cell cycle regulators is common and one in which such deletions do not occur.

We thank Ingelis Thorsen, Lotte Gernyx, and Bente Hein for skillful technical assistance; Alexander Iovanowich for software programming; and Flemming Brandt Sørensen for assistance with tissue material.