Overexpression of p53 in Tumor-distant Epithelia of Head and Neck Cancer Patients Is Associated with an Increased Incidence of Second Primary Carcinoma¹

SCC³of the head and neck is the sixth most common malignancy in the world,and the 5-year survival rate is one of the lowest among all cancers (1). Accumulation of mutations in oncogenes and tumor suppressor genes is a major molecular mechanism of tumor development (2). Among the tumor suppressor genes, inactivation of the p53 gene is one of the most frequent events (3). Overexpression of the p53 protein is frequently but not always associated with gene mutation and is regarded to be a promising candidate that might predict patient prognosis (4). Taking recent publications together, there is now good evidence for a prognostic impact of p53 mutations in head and neck cancer, i.e., they seem to be associated with an increased risk of locoregional treatment failure (5, 6). p53 DNA contact mutations appear to harbor a distinctly high malignant potential and to predict a poorer survival (7). Regarding p53 overexpression, Shin et al. (8), in their cohort of 69 patients with definitive local therapy (46% were stage I and II tumors), demonstrated a predictive value of p53 overexpression for shorter survival. However, other studies failed to confirm these results or revealed contradictory results (9, 10, 11, 12, 13).

p53 overexpression not only occurs in 23–70% of all head and neck tumors (4) but also has commonly been found in dysplastic and nondysplastic premalignant epithelial lesions (9, 14, 15, 16, 17, 18). Even less is known about the clinical value of the observed p53 overexpression in these epithelia. Uhlmann et al. (15) performed a retrospective study on laryngeal cancer patients who previously underwent laryngeal biopsies without the diagnosis of cancer. Linear basal overexpression of p53 in premalignant lesions was found to be significantly associated with an increased risk of progression to cancer (15). In a prospective study,Gallo et al. (16) have reported that simultaneous p53 and p16^INK4a alterations in premalignant laryngeal lesions have prognostic relevance in the progression to a malignant tumor. Thus, in the case of laryngeal precursor lesions, p53-overexpressing cells seem to have some malignant potential. In oral premalignant lesions, the predictive power of p53 overexpression appeared even less pronounced than in laryngeal lesions. Ogden et al. (17) have noted that p53 overexpression does not necessarily predict further malignant disease. Taking these studies together, it is clear that p53 overexpression will not turn out to be a definitive biomarker (indicating 100% risk) for second primary carcinoma (reviewed in Ref. 18). Whether p53 overexpression has reasonable predictive power at all has not been rigorously answered. For answering this question, long-term follow-up studies are required. For instance, on the basis of an actuarial rate of second primary cancer of 10% and on an 3-fold increase of this rate in a patient group with p53-positive, tumor-distant epithelia, a cohort size of 50 patients will only provide a study power of 50%. Therefore,the literature does not contain a study with a sufficiently large number of patients to answer this question.

We have shown previously that in multiple anatomical sites within one cancer patient, histologically normal appearing tumor-distant epithelial cells can be found that express a mutated p53 protein (19). This finding, which was confirmed by Waridel et al. (20), has provided a possible molecular basis for the development of second (multiple) primary carcinomas.

Here, we present a prospective long-term follow-up study involving 105 patients (including 15 patients from our previous work) with malignant head and neck tumors. We have focused on the analysis of p53 protein expression in tumor-distant epithelia at the time of surgery of the primary tumor and have compared this with clinicopathological parameters, including the occurrence of second primary carcinoma. Immunohistochemical staining of tissue sections is a simple, rapid, and reliable method to assess p53 protein expression and can be performed according to standard protocols in many clinical laboratories. The size of our patient cohort was estimated to be sufficiently large to answer whether p53 overexpression is associated with second primary carcinomas.

Our results show that it is indeed possible, using this simple assay,to identify patients with higher and lower risk for second primary SCCs of the head and neck.

Clinicopathological parameters from patients with head and neck cancer were obtained from patient records and the statements of the pathology department. Only patients from whom fresh-frozen biopsies could be obtained and who were regularly seen at follow-up examinations were recruited. The enrollment period started from May 1990 and ended in December 1998. Inclusion criteria were SCC only (excluding skin cancer)and patients presenting with primary tumors (excluding patients with recurrent tumors and patients with secondary tumors who had been treated with radiation and/or chemotherapy) at the Department of Oto-Rhino-Laryngology of the University of Heidelberg. Accuracy of the clinical data was validated by two independently reviewing investigators, who were unaware of the results of p53 staining. The data collected included age, gender, primary tumor site (oral cavity,oropharynx, larynx, hypopharynx, or other), tumor size(T_1–4), lymph nodal status(N_0–3), American Joint Committee on Cancer Stage(stages I–IV), histological grading (well, moderately, poorly, and undifferentiated), and cause of death. In follow-up, recurrences of primary tumor and occurrences of lymph node metastases, distant metastases, and second primary tumors were recorded. For the diagnosis of second primary tumors, a modification of the criteria of Warren and Gates was used (see Ref. 21). Only second primary tumors occurring in the aerodigestive tract were enrolled in this study.

All tissues were collected during the operation and immediately snap-frozen in 2-methyl-butane, precooled in liquid nitrogen, and stored until use at −80 C. One hundred fifty-one snap-frozen,tumor-distant epithelial tissues were available for study. Precise information on the anatomical location was available in 113 cases (the other biopsies were merely classified as tumor-distant). The tumor-distant epithelia were derived from the site adjacent to the tumor (e.g., oropharyngeal mucosa in case of a larynx carcinoma and hypopharyngeal mucosa in case of an oropharynx carcinoma)or from the contralateral subsite in case of oral cavity and larynx tumors (see Table 2). These small biopsies had a minimum distance of 4 cm to the margin of the tumor and were histopathologically tumor free. Epithelia of 63 patients who underwent operation for other reasons than malignancy were collected for comparison.

Immunohistochemical staining of the frozen sections with p53-antibody Bp53-12 (Progen, Heidelberg, Germany) was performed as described earlier (19). Tissues were ranked as p53 positive if parabasal cell clusters or a continuous cell nest of more than five cells with a strong nuclear staining were seen. Faint cytoplasmic or nuclear staining not meeting the former criteria and all other tissues were ranked as p53 negative. In case of availability of several tissues from one patient, one positive epithelium was enough to rank the patient as p53 positive for tumor-distant epithelia.

The major statistical endpoints of this study were the incidence and the time to diagnosis of posttherapeutic events during follow-up, in particular local recurrence from primary carcinoma or second primary carcinoma, in relation to the p53 expression status in tumor-distant epithelia or primary tumors, respectively. Events of tumor relapse were calculated from the date of initial therapy. Calculations were performed by using Fisher’s exact test or unpaired Students t test (GraphPad version 2.02; InStat Software) for univariate analyses. P < 0.05 was regarded as significant. All Ps are expressed as mean ± 1 SD. The Kaplan-Meier method was used to compare the incidence and time course of tumor relapse events between groups with p53 positivity and p53 negativity. Differences were quantitated and evaluated for significance using the log-rank test stratified for tumor site.

The study was approved in accordance with the declarations of Helsinki.

One hundred five patients met the inclusion criteria. The p53 expression status in the primary tumors and in the tumor-distant epithelia of these patients and the main clinical characteristics are summarized in Table 1. Forty-nine cancer patients (46.7%) revealed p53-positive epithelia,whereas no p53 positivity could be detected in epithelia obtained from 63 noncancer patients. Examples of p53 positivity in frozen sections of tumor-distant epithelia are presented in Fig. 1. The pattern seen in Fig. 1,a, i.e., two parabasally located small clusters of stained cells, represents the threshold positivity. Fig. 1,b shows the staining pattern most often encountered in tumor-distant mucosa, i.e.,staining of basal and parabasal cells. The positive staining in Fig. 1,c extended from the basal-most layer into the first suprabasal layers. This pattern was reported to be predictive for second primary cancer (22). For comparison, Fig. 1,d shows the transition from negative to positive in close proximity of a SCC. p53 expression in the epithelia did not correlate with the p53 protein status of the primary tumor; neither did it correlate with tumor size, tumor stage, nodal status, or histological grading. Regarding smoking habits, there was an insignificant difference between the two groups. Smoking habits remained balanced between the groups during follow-up. Because this study focused primarily on the question of a predictive role of the p53 protein status in tumor-distant mucosa at the time of presentation with the primary tumor, we did not attempt to obtain detailed information about tobacco consumption. The follow-up period ranged from 5 to 106 months,with a mean time of 55 months. The median occurrence of tumor-dependent events was 9.9 ± 6.2 months (mean ± 1 SD) for primary recurrences (n = 20), 42.8 ± 26 months for second primary tumors (n = 15), 25.5 ± 25.3 months for lymph node metastases (n = 20), and 15.5 ± 11.7 months for distant metastases (n = 24). Forty-two patients died during the follow-up period. Death rate, the occurrence of new lymph node and distant metastases, and recurrences of the primary tumor were not associated with p53 overexpression in the epithelia. However, the rate of second primary tumors in the aerodigestive tract was significantly higher in patients whose epithelia showed p53 overexpression (14 versus 1 in the p53-negative group). Table 2 demonstrates that p53-positive cell clusters were seen more frequently in epithelia of the hypopharynx than in the oropharynx, oral cavity, or larynx; this difference was, however, not significant. Within anatomical subgroups of the larynx, p53 overexpression was significantly more frequent in supraglottic and glottic regions than in the subglottis. Fig. 2,A shows that Kaplan-Meier analyses and log-rank tests confirmed the significant association between p53 overexpression specifically in the epithelia and second primary malignancy. The same analysis confirmed the lack of such a correlation of p53 overexpression in the epithelia with local recurrences from the primary tumors (Fig. 2,B). There was also no significant influence of mucosal p53 overexpression with disease-free and overall survival (not shown). Importantly, we could not find a correlation between the p53 status in the primary tumors and the incidence of second primary tumors (Fig. 2 C). This latter result was confirmed either when a larger cohort of 250 patients including the 105 patients of this study was analyzed or when those patients were analyzed separately, of whom no tumor-distant mucosal biopsies were available (not shown). The anatomical site of the second primary tumors did not significantly correlate with the sites from where the mucosal biopsies analyzed had been taken (not shown).

The prognosis of head and neck cancer patients with definitive locoregional tumor treatment is influenced by a peculiar characteristic of this cancer type–the frequent incidence of second primary tumors (21). In the present study, we have used immunohistochemistry and long-term follow-up on a sufficiently large number of cases to examine whether the frequently observed overexpression of p53 in tumor-distant epithelia of head and neck cancer patients (14, 15, 16, 17, 18) could be a biomarker with predictive value for second primary cancer. We show that overexpression of p53 in tumor-distant epithelia was associated with an increased incidence of second primary carcinoma (Table 1; Fig. 2). At the same time, as shown in Table 1 and Fig. 2 C, the data provide strong evidence against a correlation between the p53 protein status in the primary tumor and the occurrence of second primary tumors. This is in contrast to the study by Shin et al. (8),who have described a weak association between p53 overexpression in the tumor and the rate of second primary carcinoma. In another large cohort of ∼145 head and neck cancer patients in which no tumor-distant mucosal biopsies were available, we also did not obtain any evidence for an association of p53 overexpression in the primary SCCs and the rate of second primary carcinoma (data not shown).

This is, to our knowledge, the first study of a sufficiently large-scale providing evidence that p53 overexpression in tumor-distant mucosa of head and neck cancer patients indicates an increased predisposition or susceptibility of the mucosa of these patients to progression to a second malignant tumor. We have gathered several arguments to support this notion: (a) as a consequence of using frozen sections rather than paraffin sections, we noted that p53 positivity was extremely specific for tumor-distant epithelia of cancer patients (Table 2). In contrast, staining of paraffin sections after antigen retrieval has a much lower specificity for malignancy or premalignancy because p53 positivity is also found in benign epithelial lesions (23) and in mucosal biopsies from healthy individuals (24). In the latter study, van Oijen et al. (26) found that smoking increased the frequency of p53 overexpression among cancer patients but not among healthy individuals. In our cohort, the incidence of p53 overexpression was also higher in the smoking group, but the difference was not statistically significant (Table 1); (b) because these epithelia were picked at random, either from the adjacent anatomical site or from the contralateral subsite, they represent an undefined entity (in contrast to leukoplakias, for example), in line with a prospective study design. This makes confounders unlikely to occur;(c) the second primary tumors occurred much later than the local recurrences from primary tumors (9.9 ± 6.2 months for primary recurrences and 42.8 ± 26 months for second primary tumors). This difference in the time interval to presentation with secondary cancer is regarded to be crucial for the discrimination between second primary and local recurrence, although in individual cases discrimination may be difficult. Jones et al.(21) in a large retrospective study found an actuarial second primary rate of 9.1% and a median time to presentation of 36 months. Our data therefore indicate that the second primaries have originated independently from the local recurrences. This is in line with previous findings by Chung et al. (25) and van Oijen et al. (26), who showed discordant p53 gene mutations in primary and second primary cancers and with discordant p53 gene mutations in primary tumor and tumor-distant mucosa (19, 20); (d) it is noteworthy that the anatomical distribution of p53 positivity correlated well with the tumor site distribution. In the larynx, only supraglottic and glottic regions but not subglottic regions showed p53 overexpression. This is interesting, because subglottic tumors are very seldom seen; and(e) most importantly, we have demonstrated a predictive value for second primary carcinoma only for the p53 status of the epithelia but not for the p53 status of the primary tumors. This last argument supports the notion that the second primary tumors can result from events in the epithelia, as postulated by the “field cancerization” concept (27). Our failure to find a correlation between p53 positivity in tumor-distant epithelia and tumor size and stage, as well as between second primary malignancy and primary tumor size and stage, also supports the alternative concept of field cancerization as a major basis of second primary malignancy in head and neck cancer.

Our results confirm earlier studies that had already revealed some predictive impact of p53 overexpression for malignant transformation (15, 16, 17). It should be stressed again, however, that our results also agree with the conclusion of another recent study on 21 patients with oral cancer, that p53 overexpression does not necessarily predict further malignant disease (16), because the positive predictive value of the p53 expression status in tumor-distant epithelia for second primary tumors is moderate. Although the actuarial rate of second primary cancer in our total patient cohort was 14.3%,this rate increased to 28.6% in the p53-positive group (14 of 49 patients were affected), whereas in the p53-negative group, the rate of second primary cancer was only 1.8%. Hence, the difference between the groups regarding risk of second primary cancer was large (odds ratio,22.0). Because of the very high specificity of p53 overexpression for the mucosa in patients with a malignant tumor and the good sensitivity of mucosal p53 overexpression in relation to second primary cancer, it is a good biomarker for the development of second primary tumors(irrespective of whether this results from field cancerization or by intraepithelial spread).

What is the relationship between p53 overexpression and p53gene mutation in tumor-distant mucosa, and would the identification of p53 gene mutation in tumor-distant epithelia provide a better predictor for second primary tumor development? Previously, we found p53 gene mutations only in a fraction of the tumor-surrounding epithelia with p53 overexpression (19). We have preliminary data from DNA sequencing of p53-positive mucosal biopsies from 11 patients, 4 of whom had second primary cancer in the upper aerodigestive tract. Among the latter group, 18 microdissected areas were sequenced. In 5 lesions, including those presented in Fig. 1, b and c, mutations were found, attributable to three of the four patients with second primary cancer(mutation:overexpression ratio, 0.28). The mutational status differed from the primary tumors. In the group of 7 patients without second primary tumor in the upper aerodigestive tract, sequencing 11 p53-positive microdissected areas revealed one mutation(mutation:overexpression ratio, 0.01). It seems notable that this patient experienced a prostate carcinoma as second primary malignancy(data not shown). On the basis of this limited experience, the assumption seems justified that the correlation between p53 mutation and overexpression is generally low in tumor-distant mucosal lesions,and the benefit of mutational analysis would be uncertain. Large-scale mutational analysis of tumor-distant mucosal biopsies as it was performed in the basic studies (19, 20), which is very expensive and time consuming, is not feasible for most clinical laboratories and for a large number of cases, whereas immunohistochemical staining using established protocols is easy to set up and gives fast and reliable information. Furthermore, mutational analysis is confronted with the problem that negative results are not conclusive. This is because these lesions are mostly very small (note the PCR-SSCP analyses in Ref. 19),⁴and it is not clear whether microdissection of undefined mucosal areas will actually catch cells with p53 mutations. Possibly, the new DNA chip technology can be adapted to identify mutations despite a high background of wild-type sequence (28).

It is also possible that other markers might improve the positive predictive value of p53 overexpression for the individual epithelial lesions. Because cytogenetic alterations are another hallmark of premalignant and malignant transformation and have been found to be correlated with p53 overexpression and mutation (Ref. 29and references therein), allelic imbalances assessed by microsatellite marker analysis (30) and numerical chromosomal aberrations assessed by fluorescence in situ hybridization (25, 30, 31, 32, 33) appear particularly promising.

In conclusion, p53 overexpression in tumor-distant epithelia should serve as a biomarker to identify those patients who are at high risk to develop second primary carcinomas. Because patients with locally cured malignant tumors are more likely to die from a second primary carcinoma rather than from the recurrence of the primary tumor itself, this potential biomarker may allow new intervention strategies for those patients.

We are greatly indebted to Antje Schuhmann and Wolfgang Klein-Kühne for excellent technical support. We specifically thank Dr. Christa Flechtenmacher, Institute of Pathology, University of Heidelberg, for histopathological assessment and helpful discussions.