This study was designed to define the potential clinical relevance of identifying alterations affecting p53 pathway in bladder cancer and to test a new, low-cost, high-throughput, and array-based TP53 sequencing technology. Tumor samples from 140 evaluable patients with bladder cancer were analyzed with two methods to detect TP53 gene mutations, including single-stranded conformational polymorphism followed by direct sequencing and an oligonucleotide array-based sequencing method. Immunohistochemistry was used to assess patterns of expression of p53, p21/WAF1, and mdm2. Median follow-up time was 27.6 months. Results from the above analyses were correlated with clinicopathological parameters and outcome. Combining the mutation-detection assays, 79 cases (56.4%) were found to harbor TP53 gene mutations. Direct sequencing identified 66 point mutations and five frameshift mutations. The p53 oligonucleotide array detected 65 point mutations and four splice site mutations in different exons but missed all five frameshift mutations. p53 nuclear overexpression was observed in 71 cases (50.7%), lack of p21 nuclear expression was found in 81 cases (57.9%), and mdm2 nuclear overexpression was seen in 64 cases (45.7%). In multivariate analysis, 17 patients (12.1%) had an altered p53 pathway, defined by the detection of mutant TP53 and/or p53 nuclear overexpression, loss of p21 nuclear expression, and mdm2 nuclear overexpression, and exhibited the worst clinical outcome in the observation period (P = 0.015), and it appears to be a significant prognostic factor associated with patient survival.
Alterations in the p53 pathway contribute to bladder tumor progression and are likely to provide relevant prognostic information to assist in the management of bladder cancer patients (1 , 2) . Several studies have reported that p53 nuclear accumulation and TP53 gene mutations are associated with overall survival in bladder cancer (3 , 4) . Introduction of a reliable, high-throughput sequencing method into clinical laboratories would enable clinical decisions based on the identification of inactivating p53 mutations. A DNA array-based method is a promising candidate assay for such routine use because it requires significantly fewer work hours and avoids the use of isotopes or fluorescence-based acrylamide separation of amplicons. DNA array-based sequencing has been applied previously to the BRCA1 and the cystic fibrosis transmembrane conductance regulator genes, showing promising results on relatively small numbers of cases (5 , 6) . In two recent studies, the p53 oligonucleotide array from Affymetrix, Inc. (p53 Genechip) was compared with other DNA-based, mutation-detection methods on 100 primary lung cancers (7) and 77 ovarian carcinomas (8) . Those studies showed a high detection rate of base-exchange mutations and a better sensitivity of the chip assay compared with conventional manual sequencing, where the mutated allele only represented a minor fraction of the samples, the majority being wild-type alleles.
p53 functions as a transcription factor, transactivating genes involved in identification of DNA damage (i.e., GADD45), cell cycle arrest (i.e., p21/WAF1), and apoptotic response (i.e., BAX). The critical activities of p53 require stringent, positive, and negative multilevel regulation by other factors, such as mdm2 (9 , 10) and p14/hARF (11 , 12) . Mutations or altered expression patterns affecting these regulatory components, described as the “p53 pathway,” also disrupt cellular growth and apoptosis, leading to neoplastic transformation (Fig. 1)⇓ .
In the present study, we applied different methods, including microchip-based TP53 gene sequencing, conventional manual sequencing, and IHC3 to detect the p53 status. Results from these studies were placed in the context of the p53 pathway, including assessment of mdm2 and p21 expression. We then conducted clinicopathological correlations, including outcome, with molecular parameters to further elucidate the potential clinical significance of detecting alterations affecting the p53 pathway in bladder cancer.
MATERIALS AND METHODS
Patient Characteristics and Tissues.
Data on 140 patients undergoing radical cystectomy for epithelial bladder cancer at Memorial Sloan-Kettering Cancer Center between October 1993 and June 1997, inclusive, were entered into a prospective database. The range of patient follow-up was from 5 to 71 months. Median follow-up time was 27.6 months. Tumor stage was assigned according to the Tumor-Node-Metastasis staging system (13) on the basis of the depth of invasion at transurethral resection. After Institutional Review Board approval and activation of the protocol, clinical information, imaging studies, and pathology parameters were reviewed, and data were entered into the database. Pathological stage was less than or equal to pT3a and greater than or equal to pT3b in 40 and 100 patients, respectively. Pathological grade was G1-G2 in 32 patients and G3 in 108 patients. Vascular invasion was present as described in 74 cases (52.8%). Nodal status was negative in 82 cases and positive in 58 cases. Neoadjuvant chemotherapy was administered to 12 patients, consisting of methotrexate, vinblastine, doxorubicin, and cisplatin in 7; ifosfamide, paclitaxel, and cisplatin in 2; and other regimens in 3. Adjuvant chemotherapy was administered in 17 patients, consisting of methotrexate, vinblastine, doxorubicin, and cisplatin in 14; ifosfamide, paclitaxel, and cisplatin in 1; and other regimens in 2. Of the 17 patients who received adjuvant chemotherapy, 3 had extravesical disease with negative nodes, 2 had positive nodes but no extravesical disease, and 12 had both extravesical disease and positive nodes. Patients were coded to ensure confidentiality, and normal and tumor tissue samples were obtained from cystectomy specimens.
Tissues were embedded in tissue cryopreservative solution (OCT Compound; Miles Laboratories, Elkhart, IN), snap frozen, and stored at −70°C. Representative H&E-stained sections of each frozen block were examined microscopically to confirm the presence of tumor, as well as to assess the percentage of tumor cells. Specific areas of normal or tumor tissues were marked on the H&E-stained slide. Unstained slides were aligned by morphology to the stained slide using operating loupes (2.5 × magnification), and corresponding areas were manually microdissected. Tumor samples were examined microscopically to confirm the specificity of dissection, and DNA was extracted from these microdissected tissue sections (see below). Additional 5-μm sections were obtained for immunohistochemical staining.
Antibodies and IHC.
Frozen sections (5 μm) were analyzed using the avidin-biotin immunoperoxidase method using a panel of well-characterized antibodies, including mouse monoclonal antihuman p53 (clone PAb1801, recognizing an epitope located between amino acids 46 and 55; Oncogene/Calbiochem Laboratories, Cambridge, MA), antihuman p21/WAF1 (clone Ab-1, recognizing an epitope mapping to amino acids 58–77; Oncogene/Calbiochem Laboratories), and antihuman mdm2 (clone 2A10, detecting an epitope in the central region of mdm2, was kindly provided by Dr. Arnold Levine, Rockefeller University, New York, NY). Sections were fixed with methanol: acetone (1:1) at 4°C for 10 min. Endogenous peroxidase activity was quenched with 0.5% hydrogen peroxide in PBS. Sections were then sequentially incubated with nonconjugated avidin and biotin (Avidin-Biotin Blocking kit; Vector Laboratories, Inc., Burlingame, CA), followed by 10% horse serum for 15 min. Sections were then incubated with appropriately diluted primary antibodies (antihuman p53, 0.2 μg/ml; antihuman p21, 5 μg/ml; antihuman mdm2, 1:500 dilution from a tissue culture supernatant solution) for 2 h at room temperature. Immunodetection was performed with biotinylated horse antimouse immonoglobulins at 1:400 dilutions for 30 min (Vector Laboratories, Inc.), followed by avidin-biotin peroxidase complexes at 1:25 dilution for 30 min (Vector Laboratories, Inc.). Diaminobenzidine was used as chromogen and hematoxylin as the nuclear counterstain.
In all staining procedures, positive and negative controls were used. Positive controls included cell lines known to express the target antigen under study. SK-UB-1 bladder cancer cells served as p53-positive control, LNCaP cells served as p21-positive control, and for mdm2 control, we used 3T3-DM cells. Two different pathologists who were blinded to the mutation analysis and clinical information separately reviewed the slides.
The cutoff points were selected based on previous publications and were as follows: p53 ≥ 20% tumor cells displaying nuclear immunoreactivities; mdm2 and p21 ≥ 10% tumor cells showing positive nuclear immunostaining (14, 15, 16) .
DNA Extraction, PCR-SSCP, and Direct Manual Sequencing.
Tissue sections (10-μm thick) were cut from each frozen block. As discussed above, tumor tissue was microdissected according to histological evaluation. DNA was extracted by using a nonorganic method (Oncor, Gaithersburg, MD), and PCR was carried out as reported previously (17) . PCR products were diluted in denaturing loading dye, heated at 95°C for 5 min, and flash cooled on ice. A total of 4 μl was loaded onto 0.5 × mutation detection enhancement gel (FMC Bioproducts, Rockland, ME) and 10% glycerol gel, respectively; they were then at 5W for 16–20 h at room temperature, as described previously (18) . After electrophoresis, the gel was dried on a vacuum gel dryer and exposed to autoradiography film for 12–20 h.
Variant and normal SSCP bands were cut out from the gels after alignment with the autoradiograph, and the DNA was eluted in 100 μl of double-distilled H2O at room temperature for 24 h and amplified by PCR. The PCR products were sequenced by using the standard dideoxy chain termination approach, as recommended by the manufacturer (United States Biochemical Corp., Cleveland, OH). Samples were electrophoresed on an 8% sequencing gel at 75 W for 2–3 h. The gel was dried and exposed overnight at room temperature.
p53 Oligonucleotide Array Assay (Genechip p53).
Primers, reference DNA, control oligonucleotide F1, and fragmentation reagents were supplied by Affymetrix, Inc. (Santa Clara, CA). The assay was performed as reported previously (19) .
Overall survival was calculated from date of cystectomy to date of last follow-up or the end point of failure. Patients who died of all causes were considered treatment failures. The survival probability was estimated by the Kaplan-Meier method (20) , and statistical significance of observed difference in complete survival curves for the alteration of p53, p21, and mdm2 biomarkers was evaluated by the Log-rank test (21) . To further investigate the association between patient survival and biomarkers, Cox regression model was used to estimate RR and 95% CIs through PHGLM procedure in SAS (22 , 23) . RRs were adjusted for potential confounding factors, such as tumor stage, lymph node, vascular invasion, and histological grade (21) , when analyzing possible effects of each molecular variable.
TP53 Mutations and Altered Patterns of p53 Expression: Detected by Different Methods in Bladder Cancer.
Figs. 2⇓ and 3⇓ illustrate results obtained using these methods. The PCR-SSCP/sequencing strategy used for the present study only assessed exons 4–9. It detected 66 point mutations and five frameshift mutations. The oligonucleotide array assay allowed the analysis of all coding TP53 exons (exons 2–11) with the identification of 65 point mutations and four splice site mutations in different exons-intron boundaries. Taking data from both approaches, a total of 76 cases showed to harbor TP53 mutations. Five of these mutations were defined as silent mutations (no amino acid change), eight generated stop codons, four were deletions (loss of 1–13 bp), and one case showed a 3-bp insertion. In addition, four cases had splice mutations affecting intronic sequences. The remaining 65 mutant cases displayed missense mutations (Fig. 4)⇓ .
Of 140 cases, 123 showed identical results; 62 cases harbored TP53 mutations, and 61 cases were wild type. Giving that the microchip algorithm is not optimized for the identification of intragenic deletions or insertions, it failed to detect the five intragenic deletions or insertions. Meanwhile, the exon 10 mutation and the four intronic splice mutations were out of the region that the direct sequencing screened. We observed that five mutations, which gave weak or hardly visible mutated bands on conventional slab gels, showed a clear signal on the microchip, as it holds a probe specific for each particular mutated variant (Fig. 2)⇓ . In four cases, mutations were detected only by direct manual sequencing and gave no signal at all on the microchip. However, another four cases revealed a strong signal in the microchip, but no tumor-specific bands could be detected in the sequencing gel (Table 1)⇓
p53 nuclear overexpression was observed in 71 cases (50.7%). Because IHC could only detect mutant p53 proteins based on epitope recognition, it missed nine cases with nonsense TP53 mutations, including three frameshift mutations and six stop codons, which can generate truncated proteins undetectable by IHC. We also observed that 9 patients with wild-type DNA sequences had a positive p53 phenotype based on the detection of p53 nuclear immunoreactivity (Fig. 3)⇓ .
p53 Status in Bladder Cancer: Association with Patient Survival.
Table 2⇓ summarizes results obtained in the laboratory and correlations with patient survival. Of the 140 analyzed bladder cancer patients, 94 cases (67%) displayed TP53 mutations and/or p53 nuclear overexpression by at least one of the methods used. Moreover, 67 cases had positive findings with at least two methods, and 60 cases showed concomitant positive results with all three methods used. Although an increased RR of death (RR = 1.65, P < 0.05) was observed in the group of p53 (one of three) in univariate analysis, the risk became insignificant after adjusting for confounding factors, including tumor stage, tumor grade, lymph node involvement, and vascular invasion (Table 2)⇓ . Neither TP53 mutations nor p53 nuclear overexpression detected by different methods showed a statistically significant association with patient survival, tumor stage, tumor grade, lymph node involvement, and vascular invasion (data not shown).
Altered Patterns of p21 and mdm2 Expression: Association with Patient Outcome.
Both p21/WAF1 and hMDM2 genes are transactivated by p53 and are an integral part of the p53 pathway. We analyzed patterns of p21 and mdm2 expression at the protein level using specific antibodies for IHC (Fig. 3)⇓ . Table 3⇓ summarizes these results and correlates them with patient survival.
We observed that 59 cases (42.1%) had a p21-positive phenotype, which we associated with the detection of p21 wild-type pattern (Fig. 3⇓ ; Table 3⇓ ). In this context, normal urothelium also showed p21 expression (data not shown). The remaining 81 cases showed low detectable levels of p21 nuclear staining and were considered to have a p21-negative phenotype (Fig. 3⇓ ; Table 3⇓ ). Nevertheless, p21 phenotype by itself was not significantly associated with patient outcome (Table 3)⇓ .
We observed that 76 cases (54.2%) had an mdm2-negative phenotype, which we associated with the detection of a wild-type pattern (Fig. 3⇓ ; Table 3⇓ ). The remaining 64 cases showed nuclear overexpression of mdm2 proteins and were considered to have an altered positive phenotype (Fig. 3⇓ ; Table 3⇓ ). We found that mdm2-positive phenotype was significantly associated with early tumor stages and with poor survival (P < 0.05; data not shown).
Impact of p53 Pathway on Patient Outcome.
Considering that the Genechip algorithm is not optimized for identification of intragenic deletions or insertions (19) , manual sequencing was used to screen the region defined by exons 4–9, and IHC was applied to detect abnormal p53 product. In the context of this study, we defined abnormal p53 status as the identification of p53 alterations by at least two of the three methods used, namely, p53 Genechip, direct sequencing, and/or IHC, except for the cases displaying frameshift mutations. On the basis of this definition, a group of 67 bladder cancer patients (47.8%) was considered to have a p53 abnormal geno/phenotype, whereas the other group included the 73 remaining cases.
We defined the “normal p53 pathway” when: (a) p53 was wild type (no mutations identified by sequencing methods and negative phenotype; n = 17), or only one of the three methods revealed a potential abnormality (n = 2, showing either microchips or manual sequencing abnormal, but none of them displayed the positive IHC staining); (b) p21 was identified as a positive phenotype; and (c) mdm2 was found to be negative (n = 19 cases; Table 3⇓ ). We observed that patients with this normal p53 pathway (n = 19 cases) had a very low death rate and were stratified under a “low-risk” category. We defined an “abnormal p53 pathway” when all three molecular markers under study showed altered geno/phenotypes (n = 17 cases; Table 3⇓ ). We observed that these patients had an aggressive course of their disease and a high death rate. We stratified these patients under a “high-risk” category. Finally, we observed that all of the other geno/phenotypes fall into an “intermediate risk” group (n = 104; Table 3⇓ ).
To establish the relevance of detecting alterations of the p53 pathway in the context of bladder cancer progression, we analyzed the impact of alterations affecting the pathway. A significant higher RR of death was observed for patients with an abnormal p53 pathway compared with those with the normal p53 pathway (P < 0.05). This association was still significant in the multivariate analysis, after adjusting for the confounding variables listed above (P < 0.05; Table 3⇓ ). Moreover, detection of the abnormal p53 pathway also had a significant association with overall survival (P = 0.023; Fig. 5⇓ ).
Data from the present study revealed that there is a good agreement between direct sequencing and microchip-based mutation detection for the TP53 gene in bladder cancer. Both analyses were conducted on 140 bladder tumors and showed a 94% correlation (19) . Five cases harboring deletions (n = 4) and insertions (n = 1) were not recognized as mutants by the microchip assay (Table 1)⇓ . This is because of the fact that the mutation-detection algorithm has not been optimized in the p53 oligonucleotide array for the identification of such structural abnormalities. Similar results were recently reported by Ahrendt et al. (7) in an analysis of 100 primary lung cancers and by Wen et al. (8) in 77 ovarian carcinomas. In those studies, none of the frameshift mutations was detected by the p53 oligonucleotide array used, which was the same p53 Genechip method used in our study. It appears that in bladder cancer, this kind of genetic abnormality is rare, accounting in the present study for only ∼3.5%. Whether an optimized algorithm would improve the detection of such changes is at present unknown.
Mutations in four additional cases were only detected by direct sequencing but not by the microchip (Table 1)⇓ . The chip algorithm handles signals from various probes interrogating each mutation. In case of low or absent background hybridization and a high positive signal, detection of mutation is unambiguous. However, when some probes are noise filled and the specific mutant signal is low, the algorithm may lead to a false base call. This is a caveat that could be improved in a new chip generation based on synthetic structures mimicking nucleotides and possessing more homogeneous hybridization stringency.
Microchip-based mutation detection identified four TP53 mutations that were not detected in the slab gels (Table 1)⇓ . This could be explained by the high sensitivity of the microchip method and the advantage of seeing different signals in tiles from a given nucleotide position, which might be easier than judging the presence of a very weak mutant band in the sequencing gel. Because of the fact that we used manual microdissection in all cases, tumor cells were not chosen as strictly as they could have been by the use of laser microdissection. In addition, a similar situation may occur in the heterozygous samples. This specific circumstance could happen when the mutant population is represented by a “sub-clone” of tumor cells, in which case, the amplification of wild-type alleles will definitively obscure the mutant band in the X-ray film used for final evaluation. The reason for the sensitive detection of small mutant fractions of the samples is probably attributable to the presence of probes on the chip with perfect homology with the mutated sequence, giving a clear and readily detectable signal.
The structure-function relationship of certain critical regions of the p53 protein has been solved at the molecular level (24 , 25) . In general, three main domains are defined on the p53 protein, including an NH2-terminal transactivation region, a middle-portion core DNA sequence-specific binding domain, and a COOH-terminal region that includes the nuclear localization signal and the tetramerization domain (26) . Most mutations cluster in the core DNA-binding domain (27) . More specifically, they affect six residues, known as the “hot spots” (five arginines and a glycine), that are critical for the p53-DNA interaction and structural folding of the molecule (26) . We found that the majority of mutations identified in the 140 bladder tumors analyzed in the present study also clustered in this crucial functional region (Fig. 4)⇓ . Only two silent mutations were identified in the transactivation domain. Regarding the COOH-terminal region, only one mutation was found by the microchip on exon 10 (Fig. 4)⇓ .
The concept of alterations affecting “genetic pathways” is becoming more than just a molecular biology exercise. However, to date, this is not being systematically applied to the study of primary clinical tumor material. There have been a few published studies dealing with p53, p21, and mdm2 in bladder cancer. However, most studies were specifically conducted in superficial, low-grade bladder tumors (28, 29, 30) or explored the relationship between the p53 and radiotherapy (31) . The more comprehensive approach we undertook in this study was aimed at better understanding this critical pathway, as well as at conducting correlative clinicopathological analyses to disclose its potential clinical relevance.
The mdm2 proto-oncogene binds to p53 and acts as a negative regulator, inhibiting its transcriptional transactivation activity (32 , 33) . Overexpression of mdm2 may overcome wild-type p53-mediated suppression of transformed cell growth, and it is one of the mechanisms inactivating p53 function (34) . In this study, although mdm2 nuclear overexpression was not found to be of clinical significance by itself, identification of a combined positive phenotype for mdm2 and an altered p53 status significantly increased the RR of death (P < 0.05; Table 3⇓ ), thus, suggesting that mdm2 and p53 have a negative cooperative effect impacting on bladder tumor progression and survival. We had reported previously similar findings for patients with soft tissue sarcoma (35) . In that study, 22 of 211 cases had abnormally high levels of both mdm2 and p53, and there was a striking significant correlation between detection of this specific phenotype and poor survival. Because p53 transactivates mdm2, overexpression of mdm2 products in the context of an altered p53 could be explained by mutations acquiring “gain of function.” Such paradigm for p53 mutations has been reported to enhance tumorigenicity, metastatic potential, and resistance to certain therapeutic agents (36 , 37) . Alternatively, mdm2 overexpression could be produced by the transactivating functions of the recently described p53 homologues, p63 and p73 (38 , 39) . In that regard, Chi et al. (41) reported elevated expression of p73 in 18 of 45 (40%) of bladder tumors and that this increase was related to tumor-specific biallelic expression. Moreover, p73 overexpression was associated with tumor progression, suggesting an oncogenic rather than a suppressor role for p73 in bladder cancer (40 , 41) . Of interest was that mdm2-positive phenotype was higher for patients presenting with early tumor stage. Lianes et al. (43) reported similar findings in an independent study, revealing a striking association between mdm2 overexpression and low-stage tumors (42) . In addition, seven of nine cases with p53 overexpression that displayed normal reading in DNA sequence were observed with increased mdm2 levels in this study. This supports previous hypotheses that the accumulated p53 protein could be attributed to increased levels of mdm2, known to stabilize p53 (34) .
The p21/WAF1 gene is a transcriptional target of p53, required for cell cycle arrest in response to DNA damage and cellular stress, because p21 functions as a cyclin-dependent kinase inhibitor. In the present study, p21 nuclear expression was considered as the wild-type phenotype, whereas low levels or lack of p21 defined the abnormal phenotype. We did not find p21 to be of clinical significance by itself; however, p21 was able to stratify patients with mutant p53. We observed that patients displaying p53 alterations and p21-negative phenotype had a reduced survival (P < 0.05). Stein et al. (42) reported similar findings, studying a cohort of 242 bladder cancer patients (43) . In that study, patients with p53-altered and p21-negative tumors had a higher rate of tumor recurrence and worse survival compared with those with p53-altered but p21-positive tumors. Because the p21/WAF1 gene itself is not a target for mutations (44) , identification of the p21-negative phenotype most likely reflects the presence in the tumor of a nonfunctional, mutant p53. Several other studies in other tumor types, such as colon cancer and hepatocellular carcinoma, have also revealed that detection of a p21-negative phenotype is associated with the presence of p53 mutations and an aggressive clinical course (45) .
Finally, the combination of altered p53 status, overexpression of mdm2, and loss of p21 was found to be significantly associated with poor survival in comparison with the normal p53 pathway. Taken together, this abnormal p53 pathway was found to be an independent predictor of survival (P < 0.05). Similar results regarding deregulation of the pathway, namely p53 alterations and detection of mdm2-positive/p21-negative phenotypes, were reported by Stefanaki et al. in lymphomas of the mucosa-associated lymphoid tissue. Moreover, they found that the expression of p53 and mdm2, in the absence of p21, was associated with high proliferative activity and a more aggressive subset of lymphomas. The genes involved in the p53 pathway govern the programs of cell growth and death (46 , 47) . The roles of the molecules involved in these regulatory systems are considered crucial in the transition imposed at the G1-S checkpoint and in the response to different cellular stress situations. Issues of unchecked proliferation and lack of apoptosis relate to the clinical phenomena of selective growth advantage and resistance to treatment (48 , 49) . In the present study, we observed that patients harboring a normal p53 pathway had a low death rate and could be considered into a low-risk category. However, patients displaying an abnormal p53 pathway were found to have an aggressive course of their disease, a high death rate, and could be considered as high-risk cases. Moreover, detection of this abnormal p53 geno/phenotype was found to be significantly associated with decreased overall survival. Thus, it is our working hypothesis that complete deregulation of the p53 pathway is associated with bladder cancer progression and with the development of tumors exhibiting an aggressive biological behavior.
In summary, we believe that microarray-based sequencing technologies could be of value in routine molecular diagnostics, more specifically, for detecting TP53 base-exchange point mutations and that clinically oriented molecular assays need to be “modular” and incorporate multiple predictive markers. It appears that alterations of p53 and mdm2 impact on patients’ outcome during the early course of the disease, probably attributable to the imposed negative effects on apoptosis and cell growth. However, it is the accumulation of other disorders in the p53 pathway, such as lack of p21 expression, that is associated with bladder cancer progression and that represents a significant predictive factor.
We thank Drs. Nathan Ellis, William Gerald, Marc Ladanyi, Arnold Levine, and Carol Prives for their suggestions and critical review of the manuscript
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.
↵1 Supported by NIH Grants CA-47538 (to C. C-C.) and ES-06718 (to Z. F. Z).
↵2 To whom requests for reprints should be addressed, at Division of Molecular Pathology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021. Phone: (212) 639-7746; Fax: (212) 794-3186; E-mail:
↵3 The abbreviations used are: IHC, immunohistochemistry; SSCP, single-strand conformational polymorphism; RR, relative risk; CI, confidence interval.
- Received July 9, 2001.
- Revision received October 8, 2001.
- Accepted October 8, 2001.