p53 Null Mutations Undetected by Immunohistochemical Staining Predict a Poor Outcome with Early-Stage Non-Small Cell Lung Carcinomas

INTRODUCTION

Lung cancer is one of the leading causes of cancer mortality in the world (1 , 2) , and its incidence is increasing in Japan (3) . Histologically, lung cancer is classified into two main types: small cell lung carcinomas and the NSCLCs, 3 consisting of adenocarcinomas, squamous cell carcinomas, large cell carcinomas, and other rare types (2 , 4) . The long-term survival rate of NSCLC patients remains unsatisfactory, even when they undergo complete and potentially curative surgery (5) . Therefore, there is an urgent need for new strategies aimed to improve lung cancer management. However, it is often difficult to distinguish unfavorable patients from others with a better prognosis based on the existing diagnostic and management tools. Recent advances of molecular biology and genetics have raised the possibility of new diagnostic techniques and treatment for application in clinical oncology. If individuals with a poor prognosis could be classified at surgery by use of molecular biological techniques, there is a possibility that their survival might be improved by more aggressive adjuvant chemotherapy or innovative gene therapy.

The normal p53 protein has important functions in cell-cycle checkpoints and modulates such important events as G₁ arrest, DNA repair, and apoptosis (6, 7, 8) . Alterations of the p53 gene could play crucial roles in the genesis of carcinomas and have been shown to be one of the most common molecular biological changes in various human neoplasms, including NSCLCs (7, 8, 9, 10) . More than 85% of the mutations have been reported in exons 5–8, most being missense mutations (7 , 8) . Most mutant p53 proteins produced by missense mutations are resistant to degradation and thus have prolonged half-lives, allowing their detection by immunohistochemical staining. Accordingly, many investigators have used this approach as a screening method to identify p53 alterations in tumor samples (7 , 8 , 11) . On the other hand, null mutations (nonsense, deletions, insertions, and splicing junction mutations) are less common in exons 5–8 but predominate outside these exons, resulting in premature stop codons and shortened protein products that often may not be detectable by immunohistochemical staining (8 , 12, 13, 14) .

In NSCLCs, there have been many studies of the relationship between p53 alterations and prognosis (15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28) . However, those have yielded conflicting results, and thus, the clinical importance of p53 alterations as a prognostic factor remains controversial. This may be partially due to methodological differences in examination for p53 gene alterations; many authors used immunohistochemistry (15, 16, 17 , 21, 22, 23, 24, 25) , but others used DNA analysis, such as SSCP, with or without sequencing, and discordance among findings with these methods has been reported (11, 12, 13, 14 , 26 , 29 , 30) .

To clarify the relationship between p53 mutations, especially the type of mutation, and immunohistochemical staining as well as prognosis, we have investigated mutations in exons 4–8 and 10 by DNA sequencing and p53 protein accumulation using immunohistochemistry with a series of surgically treated NSCLC specimens. Correlations of the results with patients’ survival rates were also examined.

MATERIALS AND METHODS

Tumor Samples and Clinicopathological Data.

p53 alterations were examined in a series of 151 NSCLCs (113 adenocarcinomas and 38 squamous cell carcinomas) that were not accompanied by other primary malignancies and were resected consecutively from 1989 to 1993 at the Cancer Institute Hospital, Tokyo, Japan. Of the patients, seven were excluded from the analyses of survival: three did not have a sufficiently detailed follow-up, two suffered operation-related deaths within 30 days, and two were other hospital mortalities. The remaining 144 patients were eligible for evaluation of survival. All patients analyzed had undergone a potentially curative resection with lobectomy or pneumonectomy, combined with pulmonary hilar and mediastinal lymph node dissection. None had received chemotherapy or radiotherapy before surgery, but 76 (53%) patients had received postoperative adjuvant therapy. Survival duration was calculated from the date of operation until the date of the last follow-up or the date of death. The overall survival was used as an end point. Complete follow-up information was available on all of these 144 patients, with median follow-up periods of 61 months (range, 2–84 months).

The 144 patients included 92 men and 52 women, and the mean patient age at surgery was 62 years (range, 26–84 years; Table 1 ⇓ ). Histopathological classification and differentiation of the tumors was determined by two of the authors (E. T., Y. I.) according to the 1981 WHO classification of lung tumors (4) . There were 107 adenocarcinoma and 37 squamous cell carcinoma patients, with 43 well, 76 moderate, and 25 poorly differentiated carcinomas. According to the TNM staging system of the Union International Contre le Cancer (31) , 68 patients were p stage I, 16 were p stage II, 39 were p stage IIIA, and 21 were p stage IIIB. Finally, p stages were divided into two groups: p stage I, and p stage II-IIIB (68 and 76 patients, respectively). The patient’s smoking history (number of cigarettes per day, starting age, and duration of smoking) was obtained from preoperative personal interviews, with division into nonsmokers without any past history of smoking and smokers, including both patients with a past history of the habit and current smokers (Table 1) ⇓ .

DNA Preparation, p53 Mutation Analysis, and DNA Sequencing.

Fresh tumor samples paired with corresponding noncancerous tissue were obtained from all patients, and then quickly frozen in liquid nitrogen and stored at −80°C until analysis, as described previously (32) . Genomic DNAs were prepared. Exons 4–8 and 10 of p53 were analyzed by the PCR-SSCP method (33) . Coding sequences including exon-intron boundaries were amplified by PCR with the primers and PCR conditions described previously (34) . The 5′ ends of each sense and antisense primer were labeled with 6-carboxyfluorescein and 4,7,2′,7′-tetrachloro-6-carboxyfluorescein (Japan Bio Service Corp., Asaka, Japan), respectively. SSCP was performed at 22°C, with loading onto nondenaturing 4% polyacrylamide gels with 10% glycerol, and analyzed by ABI PRISM 377 (Perkin-Elmer Corp.). When genomic DNA extracted from tumors showed a different SSCP pattern from corresponding normal lung tissues, both genomic DNAs were amplified with the primers in the presence of [α-³²P]dCTP to elute the shifted DNA fragment for sequence analysis. After PCR under the same cycle conditions, PCR products were electrophoresed in a nondenaturing 5% polyacrylamide gel with 10% glycerol at the most suitable temperature and 35 W constant power for 2–3 h. Gels were subjected to drying at 80°C for 1 h and autoradiographed at room temperature overnight. Both normal and abnormal DNA fragments were eluted from the dried gels and reamplified by PCR with the same primers and PCR conditions. To characterize p53 gene mutations, the reamplified DNAs were sequenced using a dRhodamine terminator cycle sequencing kit (Applied Biosystems Inc.) and ABI PRISM 377.

Immunohistochemical Staining.

Sections (4 μm thick) of formalin-fixed, paraffin-embedded tissue, including large cut surfaces of adenocarcinomas, were immunohistochemically stained by the avidin-biotin peroxidase complex method. RSP53 (Nichirei Corp., Tokyo, Japan), a rabbit polyclonal antibody recognizing a linear epitope in human p53 located between amino acids 54–69 and generally used in Japan (35, 36, 37) , was used as the first antibody. It recognizes both wild-type and mutant p53 protein.

After deparaffinization, the sections were immersed in 0.3% hydrogen peroxide in methanol for 30 min and incubated overnight at 4°C with the primary antibody RSP53. The slides were then washed in PBS and exposed to the secondary antibody, biotinylated swine antirabbit immunoglobulin (DAKO, Glostrup, Denmark), for 1 h at room temperature. Finally, the slides were again washed in PBS and incubated for 30 min with avidin-biotin-peroxidase complex (DAKO), and the peroxidase color reaction was developed. For microscopic examination, the slides were counterstained with Harris’ hematoxylin. Normal rabbit IgG at the same concentration as the primary antibodies served as a negative control. Immunohistochemical reactions were considered positive if more than 15% of the nuclei were stained.

Statistical Analysis.

To assess any correlations between frequencies or types of mutation (missense mutation or null mutation) and clinicopathological data, the χ² test, Fisher’s exact probability test, and Student’s t test were used, with P < 0.05 indicating a significant difference. Survival curves were created by the Kaplan-Meier method, and the statistical significance of differences was calculated by the log-rank test. Multivariate analyses were performed to identify independent prognostic factors and to assess relative risk using the Cox proportional hazards model with the Statistical Package for Social Science (SPSS). In this model, seven factors potentially related to survival (age at surgery, sex, histology, differentiation of the tumor, p stage, smoking status, and p53 status) were included, and the model selection for identifying the subset of significant variables was based on the stepwise method for backward selection.

RESULTS

p53 Mutations and Relationship with Clinicopathological Parameters.

Of the 144 NSCLCs examined, 65 samples (45%) had a mutation of the p53 gene in exons 4–8 or 10: 3 (5%) located in exon 4; 16 (25%) in exon 5; 9 (14%) in exon 6; 15 (23%) in exon 7; 14 (22%) in exon 8; 4 (6%) in exon 10; and 4 (6%) in splicing junctions (Table 2) ⇓ . No mutations were found in normal lung tissue samples, except with patients carrying a polymorphism in exon 4, codon 72 (data not shown; Ref 38 ). p53 gene mutations were more frequent in squamous cell carcinomas than in adenocarcinomas, as reported previously (Table 1 ⇓ ; Refs. 8 , 39 ). There was no difference in mean age between p53-positive and -negative status. Tendencies toward more frequent mutations in males than in females and in smokers than in nonsmokers were observed. Frequencies of mutations did not differ with differentiation status, even when tumors were divided into the two histological types (data not shown). Mutations were observed significantly more frequently in p stage II-IIIB than in p stage I patients (P = 0.01).

When mutations were classified into two types, missense and null, there were 44 (68%) and 21 (32%), respectively, the latter including 7 nonsense mutations, 8 deletions, 2 insertions, and 4 mutations at splicing junctions (Table 1) ⇓ . There was no mean age difference between the two types. In addition, proportions did not differ significantly with sex, histology, differentiation status, or p stage. A higher proportion of null mutations was observed in smokers than in nonsmokers with statistical significance (P = 0.04).

Immunohistochemical Staining and the Relation to p53 Mutations.

To clarify the relationships between p53 mutations and type and immunohistochemical staining, adenocarcinomas, which were the major cohort in this study, were stained immunohistochemically (Table 3) ⇓ . Forty (37%) of 107 adenocarcinomas were positive. Although a statistically significant correlation was found between p53 mutations and immunohistochemical staining, 20 (47%) of the 43 mutated tumors were not positive and the concordance rate between immunohistochemical staining and mutation was only 65%. When the p53 gene abnormality was confined to exons 5–8, it increased to 68%. Immunohistochemical staining was detected in only 7% of cases with null mutations but in 76% of those with missense mutations. No relationship was found between immunohistochemical results and clinicopathological parameters (data not shown).

Survival and p53 Status.

Examination of the relationships between p53 alterations and overall survival, using the Kaplan-Meier method and the log-rank test, revealed a shorter survival period for the 144 NSCLC patients with than those without p53 mutations (P = 0.042; Fig. 1A ⇓ ). However, considering the strong correlation between frequency of the p53 gene mutation and the tumor stage, we separated the patients into two groups, p stage I and p stage II-IIIB patients. In the 68 p stage I NSCLC patients, there was a trend toward poorer prognosis for those with p53 mutations (P = 0.117; Fig. 1B ⇓ ). In p stage II-IIIB patients, there was no difference in survival (P = 0.845). There was no difference in survival between immunohistochemically positive and negative adenocarcinomas in all cases, in p stage I and in p stage II-IIIB (P = 0.364, 0.230, and 0.667, respectively; Fig. 1C ⇓ ).

Finally, we studied the relationship between the type of p53 mutations and survival by stage. Interestingly, in stage I, patients with null mutations showed a shorter survival than those without mutations (P = 0.008) and than those with missense mutations, although this latter difference did not reach statistical significance (P = 0.079; Fig. 1D ⇓ ). There was no difference in survival between patients with missense mutations and those without mutations (P = 0.802). In p-stage II-IIIB patients, there were no correlations between types of mutations and survival period (data not shown). We also investigated which exon or domain had the most important influence on prognosis, but no associations could be found.

Multivariate analysis of age, sex, histology, tumor differentiation, p stage, and the presence or the types of the p53 gene mutation was performed to examine the interrelationship of possible prognostic factors and survival (Table 4) ⇓ . In all cases, p53 mutation and its types were not independent prognostic factors. However, in p stage I patients, the presence of a p53 mutation was an independent prognostic factor with a statistical trend (P = 0.14). Furthermore, a null mutation independently predicted poor survival for p stage I patients with a relative risk of 4.30 (P = 0.03).

DISCUSSION

Many p53 mutational analyses have been performed for various tumors, and it has been reported that most mutations occur in exons 5–8, in so-called “hot-spots” (7 , 8) . However, we added exons 4 and 10, which were mutated most frequently after exons 5–8 in the present study. The observed frequencies of p53 mutations were similar to those of Japanese reports, in which mutations on exons 2–11 were studied extensively in NSCLC patients (28 , 39) . Therefore, we consider that almost all mutations presenting in lung cancers could be found in the present study and that our examined cases are representative for p53 mutations in NSCLCs in Japan. When the proportions of each mutational type were analyzed, null mutations constituted 32%, in line with previous studies (20–42%; Refs. 12, 13, 14 , 27 , 28 ).

The significantly poor clinical evolution observed for patients with a null mutation, but not with a missense mutation, in p stage I might be associated with the following. Null mutations generally cause loss of the COOH-terminal domain of the p53 protein, which plays an important role in tumor genesis. Thus, it has been reported that a synthetic peptide derived from the p53 COOH-terminal domain can restore the transactivation function of at least some mutant p53 proteins in living cells and further cause growth inhibition and apoptosis in a mutant p53-carrying human tumor cell line (40) . Another study demonstrated that COOH-terminal domain mutants, especially oligomerization-defective mutant 342-stop, did not exhibit sequence-specific DNA binding or failed to transactivate p21^Waf1/Cip1, Bax, and IGF-BP3 transcriptionally (41) . Thus, loss of COOH-terminal domain may equal loss of p53 gene function. On the other hand, with a significant number of missense mutations, the ability to bind DNA and form tetramers is retained, so that a part of the gene function may remain.

One earlier report suggested that null mutations predict a poor outcome in NSCLC patients, but the result was negated by multivariate analysis (27) . Thus, this is the first clear demonstration by multivariate analysis in a large cohort of patients with adequate staging and follow-up that null mutations can predict a poor outcome. There has been a report that missense mutations predict poor outcome in stage I NSCLC patients (28) , but this is not in conformity with the functional properties of p53 mutant proteins described above. Further analysis of a larger number of patients, possibly in a prospective fashion, is now required to elucidate the prognostic impact of p53 null mutations and missense mutations in NSCLC patients.

In the present study, p53 mutations were related to a poor prognosis only in p stage I NSCLC patients but not in those with p stage II or more advanced stage disease. This may be explained as follows. Malignancy in a tumor increases with successive alterations of oncogenes and tumor suppressor genes. In the early stages of NSCLC, the numbers of such altered genes are still small, and accordingly, loss of p53 function exerts a large effect. However, in late or advanced stage malignancies, the presence of large numbers of accumulated alterations means that the effect of p53 dysfunction is relatively minor.

The frequency of p53 immunohistochemical staining in this study was in accordance with previous findings for lung adenocarcinomas (24 , 26 , 37) , and the concordance rate with p53 mutations was also similar to those in the literature (20 , 26 , 28) . All null mutations except one were immunonegative, perhaps explaining controversial results regarding prognosis related to p53 alteration in NSCLC patients.

In summary, null mutation were an independent significant prognostic factor in p stage I NSCLC patients. The mutation accounted for a significant proportion of p53 mutations and could not be detected by immunohistochemistry. Analysis of p53 mutational type is warranted to predict the clinical course of stage I NSCLC patients undergoing radical surgery; it also allows optional choices of innovative therapies.