Abstract
Purpose: The aim was to analyze Tp53 and HDM2 in T1G3 bladder tumors and to determine the prognostic value of their alterations.
Experimental Design: Tumors (n = 119) were extracted from a prospective study of 1,356 bladder cancers. Tp53 mutations (exons 4-9) were assessed by sequencing of PCR products. HDM2 dose was assessed by quantitative PCR. p53, HDM2, and the products of p53 target genes were analyzed by immunohistochemistry. Cases were distributed in three categories. The association with prognosis was determined using Kaplan-Meier and Cox analyses.
Results: Eighty-five percent of tumors harbored alterations in Tp53 or HDM2. In group 1 (n = 77), 69 tumors had inactivating Tp53 mutations (58%), and 8 had HDM2 gains (7%). Group 2 (n = 24) comprised tumors overexpressing p53 in the absence of mutations (20%). Group 3 tumors (n = 18) had no alterations. HDM2 gains were associated to HDM2 overexpression and to wild-type Tp53. Expression of type 1 insulin-like growth factor receptor, 14-3-3 σ, and cyclooxygenase-2 was similar in groups 1 and 2 and significantly different from group 3. Survivin was expressed in the majority of tumors regardless of p53 pathway status. Taking group 3 as reference, the hazard ratios (HR) for recurrence, progression, and death were not significantly different in the other patient groups. HRs for recurrence were 1.13 for group 1 [95% confidence interval (95% CI), 0.25-5.03] and 1.40 for group 2 (95% CI, 0.27-7.20). HRs for progression were 0.50 for group 1 (95% CI, 0.18-1.40) and 0.25 for group 2 (95% CI, 0.05-1.29).
Conclusions: The p53 pathway is inactivated in most T1G3 bladder tumors. These genetic alterations do not independently predict patient's prognosis.
Tp53 is a key tumor suppressor gene involved in the maintenance of genomic stability, response to genotoxic stress, and activation of cell cycle exit/apoptosis. p53 is generally undetectable in normal cells due to its rapid degradation by ubiquitination. HDM2 codes for an E3 ubiquitin ligase responsible for p53 proteasomal degradation (1). In response to stress, p14Arf is activated leading to HDM2 sequestration, p53 stabilization, and transcriptional activation of genes involved in growth arrest, DNA repair, apoptosis, and senescence (2). HDM2 is a direct p53 transcriptional target, generating a feedback loop that limits its activity. p53 can also act through additional transcription-independent mechanisms (3, 4).
Tp53 is mutated in 50% of human tumors (5–7). Most Tp53 mutations are missense (73%): 80% occur in the DNA binding domain, 4% in the tetramerization and regulatory domains, and 1% in the transactivation domain. The five most common mutations account for 20% of all p53 core domain mutants, leading to the accumulation of inactive, dominant-negative p53 (5–7). Some p53 mutants also display gain of function properties. A moderate correlation between Tp53 mutations and p53 accumulation has been reported (8, 9).
In tumors, the activity of the p53 pathway can be modulated by other genetic alterations in addition to gene mutation, mainly through HDM2 overexpression, leading to enhanced p53 degradation. Mdm2 was isolated from amplified DNA in transformed murine cells (10), and its human counterpart (HDM2) is also amplified in some human cancers (11).
Many studies have analyzed p53 in bladder cancer: the prevalence of p53 alterations increases with stage and grade (12–15), but there is no definite evidence that p53 overexpression is an independent prognostic factor (13). We have conducted a prospective, multicenter study aimed at identifying molecular markers that are useful to predict patient's prognosis and have initially focused on T1G3 tumors because they represent a remarkable clinical challenge. They constitute 10% of all superficial tumors (and are therefore poorly represented in small studies) and are associated with a 50% risk of progression to muscle-invasive disease (16). Recently, we reported that Tp53 and FGFR3 mutations occur independently in T1G3 tumors and do not predict outcome (17). We have hypothesized that the lack of prognostic value of Tp53 mutations in this tumor subgroup might be due to the effect of other genes in this pathway, such as HDM2, and that a general readout of p53 activity (provided by the analysis of expression of p53 targets) may provide a more accurate view of the status of the pathway. Survivin and type 1 insulin-like growth factor receptor (IGF1-R) are repressed by wild-type p53, whereas 14-3-3 σ and cyclooxygenase-2 (COX-2) are induced by p53.
We report here a detailed analysis of the prognostic value of p53 pathway status in T1G3 tumors, including Tp53 mutations, nuclear p53 overexpression, HDM2 copy number changes, HDM2 overexpression, and immunohistochemical analysis of the proteins encoded by four p53 target genes.
Materials and Methods
Patients and tumor samples. Cases were drawn from the EPICURO study, which comprises 1,356 consecutive incident bladder cancer cases recruited prospectively, between 1997 and 2001, in 18 general hospitals in Spain. Sociodemographic and clinical information was retrieved from hospital records. Tumors were staged and graded according to the tumor-node-metastasis classification and the WHO-International Society of Urological Pathology as described in detail elsewhere (18). All blocks from the initial tumor were sent to the coordinating center where expert pathologists reviewed a section from each block and classified the tumor following strict criteria (J.L.); 119 T1G3 cases identified are the subject of this report. Mean age was 67 years (range, 41-79 years), and median age was 68 years. Eighty-six percent were men; 59% had a single tumor, and 30% had more than one tumor (this information was missing for 11% of cases). In 41% of cases, tumors were located in more than one region of the bladder; 9% of tumors were located in the trigone; and in 47% of cases, the tumor was at other sites (this information was missing for 2% of cases). Primary treatment was as follows: transurethral resection alone (23%), transurethral resection and intravesical Bacillus Calmette-Guerin (46%), transurethral resection and intravesical chemotherapy (12%), and other treatments (19%).
Cases were prospectively followed-up both through hospital records and by telephone interviews to the patients or a next-of-kin when the former was not reachable or was deceased. Outcome was defined as follows: recurrence, appearance of a new non–muscle-invasive tumor after primary treatment; progression, appearance of a new muscle-invasive tumor. All deaths were recorded but only bladder cancer–related deaths were considered for survival analyses. As of November 2004, 48 (40%) subjects were alive and free of disease with a median follow-up of 54 months (range, 38-72 months). Of 40 deaths, 16 were due to bladder cancer; the latter cases were censored at the time of death for the analysis. Survival was computed as the period comprised between cancer diagnosis and death or last control. There were no cases lost to follow-up.
Written informed consent was obtained from all patients. The study was approved by the Ethics Committees of all participating institutions.
Immunohistochemistry. All immunohistochemical assays were done using formalin-fixed, paraffin-embedded tissue. The following antibodies were used: mouse monoclonal antibody DO7 (Novocastra, Newcastle upon Tyne, United Kingdom) detecting p53 was used at 1:50; monoclonal antibody 2A10 (kindly provided by Dr. C. Cordón-Cardó, Memorial Sloan-Kettering Cancer Center, New York, NY) detecting HDM2 was used at 1:500 (19); rabbit antiserum RB-9245 (Lab Vision, Freemont, CA) detecting survivin was used at 1:100; mouse monoclonal antibody 24-31 (Neomarkers-Lab Vision, Freemont, CA) detecting IGF1-R was used at 1:100; monoclonal antibody 1433S01 (Neomarkers-Lab Vision) detecting 14-3-3 σ was used at 1:100; rabbit antiserum RB-9072 (Lab Vision) detecting COX-2 was used at 1:100.
p53 and HDM2 expression were assessed using sections from standard tumor blocks. For p53, antigen retrieval was done using 10 mmol/L citrate (pH 7.3) at 120°C for 1 minute in an autoclave. HDM2 was retrieved using 10 mmol/L citrate (pH 7.3) for 15 minutes in a microwave. A Tech-Mate 500 instrument (Ventana Medical System, Tucson, AZ) was used for immunostaining. As secondary antibody, the Envision+ anti-mouse reagent was applied (DAKO, Copenhagen, Denmark). Reactions were developed using diaminobenzidine. Sections were counterstained, dehydrated, and mounted. Controls included the use of a paraffin-embedded pellet containing a mixture of variable proportions (5-100%) of HT-29 cells (harboring a Tp53 missense mutation) with KATO III cells lacking Tp53. HDM2 controls included a panel of previously tested tissues kindly provided by Dr. C. Cordón-Cardó. All tissue areas were examined for staining; both intensity (1+ to 3+) and proportion of reactive cells were assessed. p53 nuclear overexpression was defined as ≥20% reactive cells; HDM2 positivity was defined as any nuclear immunostaining. To assess whether the cutoff value used to define p53 positivity influenced the results, other cutoffs were considered in the analyses, but the findings were similar to those reported using the 20% threshold (data not shown). A histoscore was calculated by multiplying the percentage of cells in each category by the intensity; the products were added to each other, as described (20). Histoscores were classed in 10 unit categories.
Expression of p53 target genes was assessed using a tissue microarray containing two cores of each tumor sample. Antigen retrieval was carried out with 0.1 mol/L citrate (pH 6) and pressure cooker for 20 minutes for survivin, 14-3-3 σ, and COX-2; for IGF1-R, 50 mmol/L EDTA (pH 8) was used. Immunohistochemical assays were done as described above. Scoring was done by two independent investigators; in cases of discrepancy, a consensus was reached. Tumors were considered to express a given protein when immunoreactivity was detected in at least one of the two tumor cores. Results of tissue microarray experiments were scored as a dichotomous variable (positive or negative).
Tp53 and HDM2 molecular analyses. The most representative tumor block was selected for manual microdissection and DNA isolation. Tp53 exon 4 to 9 sequencing was done as described (17).
To monitor HDM2 copy number, real-time PCR was done using an ABI Prism 7900 Sequence Detection System (Applied Biosystems, Foster City, CA). Taqman probes were labeled with 6-carboxyfluorescein. Sequences were CDK5 forward, GGAACTGTGTTCAAGGCCAAA; CDK5 reverse, CCCGTTTCAGAGCCACGAT; CDK5 probe, ACCGGGAGACTCATG; HDM2 forward, CGGGAGTTCAGGGTAAAGGT; HDM2 reverse, GCGCAGCGTTCACACTA; HDM2 probe, ATCATCCGGACCTCCC. PCR reaction for CDK5 consisted of 900 nmol/L primers, 100 nmol/L Taqman probe, and DNA (5 ng) in 1× Taqman Universal Master Mix (Applied Biosystems). PCR reaction for HDM2 consisted of 1 μL primer and probe mix and DNA in 1× Taqman Universal Master Mix. All reactions were done in duplicate. PCR variables were 50°C for 2 minutes and 95°C for 10 minutes and 40 cycles at 95°C for 10 seconds and 1 minute at 60°C. Taqman software was used to calculate Ct, where Ct corresponds to the point at which the product is distinguishable from background. To quantify gene dosage and fold changes, a calibrator consisting of a pool of DNA from paraffin-embedded normal tonsil, liver, kidney, and spleen was used. Dose changes were defined as HDM2/CDK5 ≥1.3 (gain), ≥5 (amplification), and ≤0.4 (loss).
Statistical analyses. Mutational status (wild type versus mutant), type of mutation (transition versus transversion), location (exon versus intron), amino acid change (substitution versus premature stop codon), and number of mutations were considered as variables. The association between Tp53 mutational status and clinical and pathologic characteristics of tumors was assessed by applying χ2, Fisher's exact, Student's, or Mann-Whitney U tests, as appropriate. Kaplan-Meier survival curves were computed by each category of the potential prognostic factors, and the log-rank and Breslow tests were applied to compare curves. Cox proportional hazard ratios (HR) were estimated to obtain risks of recurrence, progression, and death for cases in each molecular category, after adjusting for other confounder variables. The assumption of proportional hazards was checked for each variable. The final predictive models were fitted after forcing the p53 pathway status variable at all steps; in the adjustment, classic prognostic factors (age, sex, tumor size, growth pattern, multiplicity, multifocality, and treatment) and region of catchment were included. Results were considered significant at the two-sided P of 0.05 level. SPSS 2003 version 12.0 statistical package was used (SPSS, Inc., Chicago, IL).
To compare the expression of the products of p53 target genes in the different categories according to p53 pathway status, χ2 test was used.
Results
Tp53 mutational status. A total of 100 mutations were identified in 78 tumors (65%; missense = 70; silent = 9; nonsense = 9; frameshift = 7). In addition, there were four mutations at intron-exon boundaries and a 30-bp in-frame deletion. Multiple alterations were present in 19 tumors (16%). The mutation distribution was as follows: exon 4 (n = 13), exon 5 (n = 27), exon 6 (n = 8), exon 7 (n = 14), exon 8 (n = 21), exon 9 (n = 3). The two most commonly mutated codons were 280 and 245 (Fig. 1). There were 8% single nucleotide deletions and 92% single base substitutions with 71% transitions (88% G:C > A:T; 12% A:T > G:C) and 29% transversions (55% C:G > G:C; 26% G:C > T:A; 11% A:T > C:G; 8% A:T > T:A). Twenty-three percent (21 of 92) of single base substitutions and 26% of G:C > A:T transitions occurred at CpG sites. There were nine silent mutations (absent from germ line DNA) in exons 4 (n = 3), 8 (n = 2), and 9 (n = 4). Cases harboring exclusively silent (n = 6) or intronic mutations outside splice donor/acceptor sites (n = 2) and the case with the 30-bp in-frame deletion were considered as “wild type” for analysis. Overall, 69 tumors (58%) harbored inactivating mutations. In 30% of them, the wild-type allele was undetected by sequencing, suggesting that the remaining allele had been lost. Detailed information is provided in Supplementary Table S1.
Comparison of Tp53 sequence and p53 accumulation. Cases were classified according to the following immunohistochemical findings: (i) >20% cells showing nuclear overexpression (n = 78, 66%); (ii) 1% to 20% of cells showing nuclear staining (n = 23, 20%); and (iii) negative, p53 being undetectable in all tumor cells (n = 18, 14%). Table 1 shows the comparison of molecular and immunohistochemical analyses. Representative results are shown in Supplementary Fig. S1. Fifty-two of 78 (67%) type i tumors were mutant; 98% of them had a missense change, and five cases had a mutation predicted to yield a truncated protein. Importantly, no mutation could be identified in 26 of 78 (33%) tumors. Eight of 23 type ii tumors (35%) were mutant with a predominance of missense changes. Nine of 18 (50%) type iii tumors harbored a mutation yielding a truncated protein; two cases had a second missense mutation.
. | Negative (n = 18) . | ≤20% (n = 23) . | >20% (n = 78) . |
---|---|---|---|
Wild type* | 9 (50%) | 15 (65%) | 26 (33%) |
Mutant | 9 (50%) | 8 (35%) | 52 (67%) |
No. alterations | 13 | 16 | 71 |
Nonsense or frameshift† | 9 (69%) | 2 (13%) | 5 (7%) |
Missense mutations | 2 (15%) | 9 (56%) | 59 (83%) |
Intronic mutations | 1 (8%) | 1 (6%) | 2 (3%) |
Silent mutations | 0 | 4 (25%) | 5 (7%) |
Deletion (30 bp) | 1 (8%) | 0 | 0 |
. | Negative (n = 18) . | ≤20% (n = 23) . | >20% (n = 78) . |
---|---|---|---|
Wild type* | 9 (50%) | 15 (65%) | 26 (33%) |
Mutant | 9 (50%) | 8 (35%) | 52 (67%) |
No. alterations | 13 | 16 | 71 |
Nonsense or frameshift† | 9 (69%) | 2 (13%) | 5 (7%) |
Missense mutations | 2 (15%) | 9 (56%) | 59 (83%) |
Intronic mutations | 1 (8%) | 1 (6%) | 2 (3%) |
Silent mutations | 0 | 4 (25%) | 5 (7%) |
Deletion (30 bp) | 1 (8%) | 0 | 0 |
Comparison of immunohistochemical findings according to the Tp53 status: wild type versus mutant (Fisher's exact test, P = 0.019).
Comparison of immunohistochemical findings according to the nature of the Tp53 mutation: nonsense or frameshift versus missense (Fisher's exact test, P > 0.001).
The proportion of mutant cases according to histocore was 0 (10 of 23, 43%), 1 to 50 (7 of 26, 27%), 51 to 250 (17 of 28, 61%), and 251 to 300 (35 of 42, 83%). Overall, immunohistochemistry did not accurately predict Tp53 mutational status (Supplementary Fig. S2).
HDM2 dosage and HDM2 immunohistochemical analysis. Using Taqman and immunohistochemical assays, 112 (94%) and 113 (95%) tumors were successfully analyzed, respectively. Ten tumors (9%) showed HDM2 gain/amplification. Seven (6%) samples contained ≥5% immunoreactive tumor cells, and additional 11 (10%) samples contained <5% positive tumor cells. In the remaining cases, HDM2 was undetectable.
The summary of results in 104 tumors analyzed by both techniques is shown in Table 2A. In 8 of 10 (80%) cases with HDM2 gain/amplification, the protein was overexpressed; in six of them, there were ≥5% of positive cells. HDM2 was detected in 8 of 94 (9%) tumors without HDM2 gain; in seven of them, there were <5% of positive cells present. HDM2 gains were significantly associated with protein overexpression (P > 0.001, Fischer's exact test).
A. HDM2 gene copy number and HDM2 immunohistochemical analysis* . | . | . | ||||
---|---|---|---|---|---|---|
Immunohistochemistry . | Gene copy number . | . | ||||
. | Gain . | No change/loss . | ||||
Negative | 2 (2%) | 86 (82%) | ||||
<5% | 2 (2%) | 7 (7%) | ||||
>5% | 6 (6%) | 1 (1%) | ||||
B. Tp53 mutational status in relationship to HDM2 immunohistochemistry† | ||||||
Negative | <5% + cells | ≥5% + cells | ||||
Mutant Tp53 | 57 (88%) | 8 (12%) | 0 | |||
Wild-type Tp53 | 38 (79%) | 4 (8%) | 6 (13%) |
A. HDM2 gene copy number and HDM2 immunohistochemical analysis* . | . | . | ||||
---|---|---|---|---|---|---|
Immunohistochemistry . | Gene copy number . | . | ||||
. | Gain . | No change/loss . | ||||
Negative | 2 (2%) | 86 (82%) | ||||
<5% | 2 (2%) | 7 (7%) | ||||
>5% | 6 (6%) | 1 (1%) | ||||
B. Tp53 mutational status in relationship to HDM2 immunohistochemistry† | ||||||
Negative | <5% + cells | ≥5% + cells | ||||
Mutant Tp53 | 57 (88%) | 8 (12%) | 0 | |||
Wild-type Tp53 | 38 (79%) | 4 (8%) | 6 (13%) |
Fisher's exact test, P < 0.001.
P = 0.011.
p53 pathway status in T1G3 tumors.HDM2 gains were detected in 2 of 67 (3%) Tp53 mutant tumors and in 8 of 45 (18%) Tp53 wild-type tumors (Supplementary Fig. S3; P = 0.06, Fischer's exact test).
When considering HDM2 immunohistochemical findings, 8 of 65 (12%) Tp53 mutant cases expressed HDM in 1% to 5% of cells. Among Tp53 wild-type cases, 10 of 48 (21%) tumors expressed HDM2: in six of them, ≥5% cells were detected; in the remaining four cases, only scattered positive cells were observed (Table 2B; P = 0.011, Fisher's exact test).
Of 119 tumors in which Tp53 and HDM2 were analyzed, 101 (85%) showed alterations in one of them. We classified subjects according to the strength of the evidence of p53 pathway inactivation: group 1 comprises 69 (58%) patients with tumors harboring Tp53 pathogenic mutations and 8 (7%) displaying HDM2 gains/amplifications; group 2 comprises 24 (20%) patients whose tumors showed nuclear p53 overexpression in >20% cells in the absence of detectable Tp53 mutation or HDM2 alteration at the DNA or protein level; group 3 comprises 18 (15%) patients whose tumors displayed no detectable alterations in Tp53 or HDM2. Cases in the three groups did not show significant differences regarding clinicopathologic characteristics (Table 3), nor age, sex, recruitment area, tobacco consumption, maximum level of education, and body mass index (data not shown).
. | Group 1, n (%) . | Group 2, n (%) . | Group 3, n (%) . | P . | ||||
---|---|---|---|---|---|---|---|---|
No. cases | 77 (64.7) | 24 (20.2) | 18 (15.1) | |||||
Tumor size (cm) | ||||||||
≤3 | 35 (45.5) | 10 (41.7) | 11 (61.1) | 0.253 | ||||
>3 | 17 (22.1) | 2 (8.3) | 2 (11.1) | |||||
Missing | 25 (32.5) | 12 (50.0) | 5 (27.8) | |||||
Localization | ||||||||
>1 sites | 34 (45.3) | 10 (41.7) | 6 (37.5) | 0.573 | ||||
Trigone | 8 (10.7) | 2 (8.3) | 0 (0.0) | |||||
Other | 33 (44.0) | 12 (50.0) | 10 (62.5) | |||||
Multiplicity | ||||||||
1 tumor | 45 (65.2) | 15 (65.2) | 10 (71.4) | 0.901 | ||||
>1 tumor | 24 (34.8) | 8 (34.8) | 4 (28.6) | |||||
Growth pattern | ||||||||
Papillary | 46 (68.7) | 13 (61.9) | 14 (87.5) | 0.367 | ||||
Solid | 15 (22.4) | 7 (33.3) | 1 (6.3) | |||||
Mixed | 6 (9.0) | 1 (4.8) | 1 (6.3) | |||||
Treatment | ||||||||
TUR | 14 (18.4) | 9 (39.1) | 5 (29.4) | 0.667 | ||||
TUR + BCG | 36 (47.4) | 10 (43.5) | 9 (52.9) | |||||
TUR + chemotherapy | 11 (14.5) | 2 (8.7) | 2 (11.8) | |||||
TUR + BCG + chemotherapy | 5 (6.6) | 1 (4.3) | 0 (0.0) | |||||
Other | 10 (13.2) | 1 (4.3) | 1 (5.9) | |||||
Outcome | ||||||||
Disease-free | 46 (59.7) | 14 (58.3) | 10 (55.6) | 0.947 | ||||
Recurrence | 14 (18.2) | 5 (20.8) | 3 (16.7) | |||||
Progression | 15 (19.5) | 2 (8.3) | 5 (27.8) | |||||
Death from cancer | 8 (10.4) | 4 (16.7) | 4 (22.2) |
. | Group 1, n (%) . | Group 2, n (%) . | Group 3, n (%) . | P . | ||||
---|---|---|---|---|---|---|---|---|
No. cases | 77 (64.7) | 24 (20.2) | 18 (15.1) | |||||
Tumor size (cm) | ||||||||
≤3 | 35 (45.5) | 10 (41.7) | 11 (61.1) | 0.253 | ||||
>3 | 17 (22.1) | 2 (8.3) | 2 (11.1) | |||||
Missing | 25 (32.5) | 12 (50.0) | 5 (27.8) | |||||
Localization | ||||||||
>1 sites | 34 (45.3) | 10 (41.7) | 6 (37.5) | 0.573 | ||||
Trigone | 8 (10.7) | 2 (8.3) | 0 (0.0) | |||||
Other | 33 (44.0) | 12 (50.0) | 10 (62.5) | |||||
Multiplicity | ||||||||
1 tumor | 45 (65.2) | 15 (65.2) | 10 (71.4) | 0.901 | ||||
>1 tumor | 24 (34.8) | 8 (34.8) | 4 (28.6) | |||||
Growth pattern | ||||||||
Papillary | 46 (68.7) | 13 (61.9) | 14 (87.5) | 0.367 | ||||
Solid | 15 (22.4) | 7 (33.3) | 1 (6.3) | |||||
Mixed | 6 (9.0) | 1 (4.8) | 1 (6.3) | |||||
Treatment | ||||||||
TUR | 14 (18.4) | 9 (39.1) | 5 (29.4) | 0.667 | ||||
TUR + BCG | 36 (47.4) | 10 (43.5) | 9 (52.9) | |||||
TUR + chemotherapy | 11 (14.5) | 2 (8.7) | 2 (11.8) | |||||
TUR + BCG + chemotherapy | 5 (6.6) | 1 (4.3) | 0 (0.0) | |||||
Other | 10 (13.2) | 1 (4.3) | 1 (5.9) | |||||
Outcome | ||||||||
Disease-free | 46 (59.7) | 14 (58.3) | 10 (55.6) | 0.947 | ||||
Recurrence | 14 (18.2) | 5 (20.8) | 3 (16.7) | |||||
Progression | 15 (19.5) | 2 (8.3) | 5 (27.8) | |||||
Death from cancer | 8 (10.4) | 4 (16.7) | 4 (22.2) |
NOTE: Group 1: cases with pathogenic Tp53 mutations or HDM2 gain/amplification; group 2: cases with abnormal p53 immunohistochemical findings but no evidence of Tp53 mutation or HDM2 gain; group 3: cases for whom Tp53 mutational analysis, HDM2 copy number determination, and p53 and HDM2 immunohistochemistry failed to disclose any abnormality.
Abbreviations: TUR, transurethral resection; BCG, Bacillus Calmette-Guerin.
Expression of p53 target genes in relationship to p53 pathway status. To assess the functional status of the p53 pathway, tissue microarrays were used to examine the expression of the proteins encoded by four p53 target genes. Expression, either at the single protein level or combined, was compared in tumors from the three groups described above.
Survivin was expressed by the majority of tumors regardless of their p53 status, and the differences among groups were not statistically significant (P = 0.922). By contrast, there were significant differences in the expression of IGF1-R (P = 0.045), 14-3-3 σ (P = 0.016), and COX-2 (P < 0.001) in relationship to the p53 status of tumors. The expression of these three proteins was very similar in groups 1 and 2 (comprising tumors with shown and suspected inactivation of p53, respectively). By contrast, expression of each of them was significantly different among tumors in group 3. IGF1-R and 14-3-3 σ were more commonly expressed in tumors without alterations in p53 or HDM2, whereas COX-2 was more commonly expressed in tumors displaying p53 pathway alterations (Fig. 2; Supplementary Table S2).
p53 pathway status and patient's prognosis. To assess the association of p53 pathway status with patient's prognosis, outcome of patients included in the three groups described above was compared.
Overall, 70 of 119 (59%) patients were free of tumor at the last follow-up (median follow-up for patients who were free of disease = 54 months). Cox models were applied to analyze differences in outcome, taking group 3 (in which no alterations were detected) as reference, after adjustment for classic prognostic markers. Disease-free survival was similar for patients in the three groups (Fig. 3A); the HR for cases in groups 1 and 2 was 0.92 [95% confidence interval (95% CI), 0.35-2.38] and 1 (95% CI, 0.34-2.92), respectively.
Regarding recurrence, no differences were observed among the three groups (Fig. 3B). The multivariable model yielded HR of 1.13 (95% CI, 0.25-5.03) and 1.40 (95% CI, 0.27-7.20) for groups 1 and 2, respectively. As of progression, no differences were found (Fig. 3C): both group 1 (HR, 0.50; 95% CI, 0.18-1.40) and group 2 (HR, 0.25; 95% CI, 0.05-1.29) displayed lower risks than the reference group, but the differences failed to reach significance. Death from bladder cancer was also similar in the three groups (Fig. 3D). Again, groups 1 (HR, 0.47; 95% CI, 0.12-1.84) and 2 (HR, 0.53; 95% CI, 0.11-2.69) showed lower risks than the reference group, but differences were not significant.
Our study is not a clinical trial, and it is conceivable that differences in patient management might contribute to the results. Therefore, we did outcome analysis in the subgroup of 55 patients whose primary treatment was transurethral resection and intravesical Bacillus Calmette-Guerin. In a univariate analysis, p53 status was not associated with disease-free survival, recurrence, progression, or death from bladder cancer. In the multivariate analysis, patients in group 3 fared worse for all outcomes, but the differences among the three groups were not statistically significant. These results are shown in Supplementary Table S3.
Discussion
It is becoming increasingly clear that the assessment of multiple genes involved in a pathway is more appropriate for the biological analysis of tumors than assessment of individual gene alterations (21). p53 alterations have been proposed as prognostic markers in bladder cancer patients (15), but the evidence that they can independently contribute to the prediction of outcome is scarce (13). Lianes et al. were first to examine the combined effect of p53 and MDM2 alterations in a relatively small panel of human bladder cancers (22). Three major problems of many published studies of biomarkers as prognostic factors are their retrospective nature, small sample size, and the heterogeneity of the patients included (23). In bladder cancer, few reports have focused on T1 tumors, and the findings are not consistent: Vatne et al. reported no association of p53 alterations with prognosis (24), whereas others have found them to be associated with time progression (25) or survival (15).
Here, we focus on Tp53 and HDM2, as well as in the proteins encoded by four p53 target genes, to assess the status of the p53 pathway and its relationship with outcome in the largest series of prospectively collected and followed patients with T1G3 tumors. We find that 58% of cases had pathogenic alterations in Tp53, a proportion that is similar to that reported in muscle-invasive tumors (26). The spectrum, distribution, and type of mutations were similar to those of the IARC database (Supplementary Tables S4-S7). Amino acid substitutions were the most frequent alteration found. The most commonly mutated codons in our study were 280 and 245: these findings support the notion that codon 280 is a hotspot in bladder cancer because it is mutated in 4% of bladder tumors versus 1% of tumors from all sites (5, 27).
p53 immunostaining has extensively been used as a surrogate marker of Tp53 mutation detection despite the fact that there is extensive evidence of discordance between the two types of assays (9, 28–30). The proportion of tumors overexpressing nuclear p53 in our study is similar to the 55% reported in a previous, smaller study of T1G3 tumors (n = 22; ref. 25). Using either a standard threshold or a Histoscore (20), we find that p53 immunohistochemical results predict poorly the molecular status of Tp53. Approximately half of the tumors scoring negative by immunohistochemistry harbored truncating Tp53 mutations. As in other recent studies (28, 30), approximately one third of tumors displaying strong p53 immunoreactivity lacked Tp53 mutations, suggesting additional mechanisms leading to altered p53 expression and function. To assess the significance of strong nuclear p53 accumulation in these tumors, we examined the expression of the products of four well-characterized p53 target genes using immunohistochemistry and tissue microarrays. Survivin expression was similar in all groups. By contrast, IGF1-R, 14-3-3 σ, and COX-2 showed a remarkably similar pattern of expression in tumors belonging to groups 1 and 2 and a very different pattern in group 3 tumors, supporting the contention that group 2 tumors harbor an inactive p53 pathway. IGF1-R has been shown to be repressed by p53 (31), and our finding that it is more commonly expressed in group 3 tumors is at odds with this evidence. However, the difference in expression was marginally significant. By contrast, the differences in expression were much stronger for 14-3-3 σ and COX-2. The former is induced by p53 (32), and our findings are in agreement with this notion because this protein was detected in 10 of 15 tumors from group 3 versus 21 of 70 and 6 of 23 tumors from groups 1 and 2, respectively. COX-2 is also induced by p53, although this effect requires intact nuclear factor-κB activity (33) and is more complex given that COX-2 can, in turn, inhibit the proapoptotic effect of wild-type p53 (34). We find higher levels of expression of COX-2 among tumors in groups 1 and 2 than those in group 3, suggesting that other factors (in addition to p53) influence COX-2 expression in tumors. Some p53 mutant proteins display gain-of-function properties (35, 36), possibly accounting for these observations. In this regard, a similar association between p53 alterations and COX-2 overexpression has been reported in ovarian adenocarcinomas (37) and gastric lymphomas (38).
We find that 16% of tumors harbor alterations in HDM2. This value is lower than that reported by Lianes et al. but it is similar to that reported in other studies (39). HDM2 alterations were significantly more common among Tp53 wild-type tumors, as expectable if inactivation of either gene leads to similar molecular defects. Overall, we find evidence for inactivation of the p53 pathway in 85% of T1G3 tumors.
By analyzing two crucial members of this pathway and using complementary techniques, including immunohistochemistry and sequencing for Tp53 and immunohistochemistry and quantitative PCR for HDM2, we have aimed at better assessing the role of the p53 pathway in a large group of T1G3 tumors. This strategy is more appropriate than the use of retrospective studies, including more heterogeneous patient populations (40–45). Because our study is not a clinical trial, patients were not treated following a standardized clinical protocol, and we followed patients prospectively and collected detailed information on treatment, allowing to adjust for this variable in the analyses. We compared the outcome of patients whose tumors had conclusive (group 1), suggestive (group 2), or no evidence (group 3) of p53 pathway inactivation and found that p53 alterations are not independent predictors of outcome (recurrence, progression, or bladder cancer–related death). The results were similar when the analysis was restricted to the group of patients treated with transurethral resection and Bacillus Calmette-Guerin, further supporting the validity of our conclusions on patient outcome. Due to the small number of events, a longer follow-up is required to conclusively establish a lack of association with death from bladder cancer. Importantly, we have also found that p53 nuclear overexpression does not predict outcome in the complete series of 995 non–muscle-invasive bladder cancers included in our study (46).
Our finding that patients in the group lacking Tp53 or HDM2 alterations fared worst of all, although the differences were not statistically significant, merits further analysis. Bartkova et al. have shown a role of ATM-Chk2-p53 activation early in tumorigenesis, and this response may be lost in more advanced tumors (47).
Altogether, our findings indicate that inactivation of the p53 pathway is an almost universal requirement for the progression of bladder tumors. PCR analysis of DNA from laser-microdissected tumor cells from group 3 cases failed to show homozygous Tp53 deletions (data not shown). Alternatively, other genetic pathways may lead to tumor progression. Indeed, if p53 or a related pathway is inactive in most, if not all, T1G3 tumors, it may come as no surprise that there are no significant differences in outcome among the various tumor subgroups. The expression analysis of the product of p53 target genes (Fig. 2; Supplementary Table S2) supports the idea that, despite a similarly poor prognosis, group 3 tumors are molecularly distinct.
Overall, we conclude that alterations in the p53 pathway are an essential component of bladder cancer progression, that T1G3 tumors resemble muscle-invasive tumors at the genetic level and may in fact be “invasive tumors that have not yet invaded muscle,” and that molecular analysis of Tp53 and HDM2 and their products does not independently predict patient outcome.
Appendix: Members of the Participating Centers
Institut Municipal d'Investigació Mèdica, Universitat Pompeu Fabra, Barcelona (coordinating center): M. Kogevinas, N. Malats, F.X. Real, M. Sala, G. Castaño, M. Torà, D. Puente, C. Villanueva, C. Murta, J. Fortuny, E. López, S. Hernández, R. Jaramillo; Hospital del Mar, Universitat Autònoma de Barcelona, Barcelona: J. Lloreta, S. Serrano, L. Ferrer, A. Gelabert, J. Carles, O. Bielsa, K. Villadiego; Hospital Germans Tries i Pujol, Badalona, Barcelona: L. Cecchini, J.M. Saladié, L. Ibarz; Hospital de Sant Boi, Sant Boi, Barcelona: M. Céspedes; Centre Hospitalari Parc Taulí, Sabadell, Barcelona: C. Serra, D. García, J. Pujadas, R. Hernando, A. Cabezuelo, C. Abad, A. Prera, J. Prat; Centre Hospitalari i Cardiològic, Manresa, Barcelona: M. Domènech, J. Badal, J. Malet; Hospital Universitario, La Laguna, Tenerife: R. García-Closas, J. Rodríguez de Vera, A.I. Martín; Hospital La Candelaria, Santa Cruz, Tenerife: J. Taño, F. Cáceres; Hospital General Universitario de Elche, Universidad Miguel Hernández, Elche, Alicante: A. Carrato, F. García-López, M. Ull, A. Teruel, E. Andrada, A. Bustos, A. Castillejo, J.L. Soto; Universidad de Oviedo, Oviedo, Asturias: A. Tardón; Hospital San Agustín, Avilés, Asturias: J.L. Guate, J.M. Lanzas, J. Velasco; Hospital Central Covadonga, Oviedo, Asturias: J.M. Fernández, J.J. Rodríguez, A. Herrero; Hospital Central General, Oviedo, Asturias: R. Abascal, C. Manzano, T. Miralles; Hospital de Cabueñes, Gijón, Asturias: M. Rivas, M. Arguelles; Hospital de Jove, Gijón, Asturias: M. Díaz, J. Sánchez, O. González; Hospital de Cruz Roja, Gijón, Asturias: A. Mateos, V. Frade; Hospital Alvarez-Buylla, Mieres, Asturias: P. Muntañola, C. Pravia; Hospital Jarrio, Coaña, Asturias: A.M. Huescar, F. Huergo; Hospital Carmen y Severo Ochoa, Cangas, Asturias: J. Mosquera; Centro Nacional de Investigaciones Oncológicas, Madrid: M. Esteller; Universitat Autònoma de Barcelona, Bellaterra: R. Miró, R. Marcos; Progenika, Derio, Bizkaia: A. Martínez.
Grant support: Instituto de Salud Carlos III, Ministerio de Sanidad, Madrid grants FIS 98/1274, 00/0745, C03/009, C03/010, G03/160, and G03/174 (EPICURO study) and Ramón Areces Foundation, Madrid predoctoral fellowship (E. López-Knowles). S. Kishore was a recipient of a Fullbright Scholarship.
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 are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
Acknowledgments
We thank S. Mancilla, A. Alfaro, G. Carretero, F. Fernández, I. López, M. Torà, S. Serrano, D. Ferrer, and the many clinicians, investigators, nurses, technicians, and patients participating in the study, as well as the members of the EPICURO project without whose continued support the study would not have been possible; Dr. C. Cordón-Cardó for providing reagents; and L. Pérez Jurado, K. Kelsey, and M. Karagas for valuable discussions.