p53 mutation, but not p53 overexpression, correlates with survival in head and neck squamous cell carcinoma.

Survival in squamous cell carcinoma of the head and neck (HNSCC) was compared with overexpression and mutation of the p53 gene. Archival tissue from 77 tumours was analysed for protein expression using immunohistochemistry (IHC) with the monoclonal antibody Do-7, and for the presence of mutation in exons 5-8 using single-stranded conformation polymorphism (SSCP), followed by DNA sequencing in SSCP-positive cases. p53 expression was scored as high (>70% nuclei stained) in 25 (32%) tumours, as intermediate (10-70% nuclei stained) in 19 (25%) tumours and as low (<10% nuclei stained) in 33 (43%) tumours. Twelve (18%) tumours exhibited gene mutation (ten missense and two nonsense mutations) and an additional five tumours contained changes that could not result in amino acid substitution or protein truncation. There was no correlation between gene expression and mutation, mutations being equally frequent in tumours with either high (4/25), intermediate (4/19) or low protein expression (4/33). Fifty-eight patients were eligible for survival analysis. There was a strong correlation between p53 mutation and cause-specific survival; median survival among mutated cases was 12.5 months compared with >160 months among non-mutated patients (P < 0.005). There was no correlation between p53 overexpression and survival. The results suggest that p53 mutation status is an important prognostic factor in HNSCC, and that IHC analysis of protein overexpression is an inadequate measure of gene mutation in these tumours.

Advances in molecular biologv have provided clues to the pathogenesis of cancer and show-n the involvement of oncogene activation and tumour-suppressor gene inactivation (Chang et al. 1995: Greenblatt et al. 1994. In the investigation of the carcinogenic process in head and neck squamous cell carcinoma (HNSCC) much interest has focused on the p53 tumour-suppressor gene. its inactivation and the possible prognostic implications of p53 overexpression and mutation.
Mutation in the p53 tumour-suppressor gene is the most frequently found genetic aberration in human cancer (Harris and Hollstein. 1993). The normal p53 protein functions as a cell cycle checkpoint and sensor of DNA damage in the cell. and modulates such important events as G1-arrest. DNA repair and apoptosis (Levine et al. 1991: Harris andHollstein. 1993). Cells xwith mutated p53 are predisposed to further genetic alterations bv means of inadequate DNA repair. escape from apoptosis and manifestation of the DNA damage in subsequent cell cycles. Other mechanisms of p53 inactivation include binding to DNA tumour mirus proteins such as papilloma virus E6. or to overexpressed cellular genes such as the MDM2 oncoaene (Lex ine et al. 1994).
The association betmeen head and neck cancer development and carcinogenic factors such as alcohol or tobacco use. as w-ell as exposure to other enxvironmental and occupational factors. is well documented (Landrinan and Baker. 1991). Several reports reveal an association between p53 overexpression and p53 mutation in head and neck carcinogenesis (Field et al. 1991(Field et al. . 1992: Brennan et al. 1995). Immunohistochemical studies have shown p53 overexpression to be an early event in HNSCC carcinogenesis. bein, found in dvsplastic lesions and CIS before the development of invasixe carcinoma. It is not known if this exent reflects p53 protein accumulation as a result of gene mutation. or merelv a normal p53 response to DNA damage because of the activity of a carcinogaen (Bovle et al. 1993: Nees et al. 1993: Pavelic et al. 1994: Shin et al. 1994: Wang et al. 1994: el-Naggar et al. 1995. Generally an association between mutation and oxerexpression is assumed. Howxever. in HNSCC there is emerging exidence of a discrepancy between the results achieved with molecular analx sis and those usincg immunohistochemical methods (Mineta et al. 1995: Nylander et al. 1995.
The aim of the present inxestigation was txofold: to study the concordance betx een p53 mutation and immunohistochemical oxverexpression. and to evaluate the prognostic implications of p53 mutation/ox erexpression.

Patients and tumours
The files for the period Januar-1987 to IMay 1991 in the Department of Pathology. Unixersity Hospital. Lund. Sxweden.
were examined and biopsies classified as oral or orophary ngeal squamous cell carcinoma retrieved and re-examined. Seventyseven head and neck cancer specimens were thus identified (Table  1). The corresponding patient charts in the Department of Otorhinolarvngology/Head & Neck Surgery (tertiarv referral centre) were reviewed as to the clinical course of disease and the tumour classification according to UICC criteria (Hermanek and Sobin. 1987) was reconfirmed. Five tumours were reclassified. four as hypopharyngeal and one as supraglottic laryngeal. The male to female ratio was 2.2:1. The initial staging was based on clinical examination. computerized tomography (CT) or magnetic resonance (MR) imaging and palpation or endoscopy under anaesthesia. Thirty-nine per cent were stage I-II and 61%7c stage III-IV. Seven patients had recurrent tumours. of the remaining 70 cases. 59% were stage T1-2 and 41% T3-4. 65% were classified NO and 35%c N+. Only two patients had distant metastases at the time of diagnosis. Thir-t-four per cent of tumours were classified as well differentiated. 44% as moderately differentiated and 22% as poorly differentiated.

Treatment
The standard therapy for SCCHN at our department has been described previously (ZaItterstrom et al. 1991 ) Briefly. the general principles were as follows: patients with TI tumours underwent primary surgery: for T2 and resectable T3 and T4 tumours. preoperative radiotherapy was gisven. with a target absorbed dose of 50 Gy. or in some cases chemotherapy was administered. followed by surgery: laryngeal TI-3 carcinomas and all non-resectable T3 and T4 tumours were gilen full-dose radiotherapy (64)(65)(66)(67)(68)(69)(70) Gy.) in some cases followed by salvage surgery. those with regional metastasis at the time of diagnosis were treated with radiotherapy followed by neck dissection.

METHODS Immunohistochemistry
Sections (4 gm thick) %vere dewaxed v-ith xylene. hydrated through graded alcohols and rehydrated in w-ater. Sections were heated three times in a micro% a% e oven in citrate buffer (pH 6.0) : c-a-ce, 7OM P3 for 5 mmn each time. allowed to cool to the room temperature gradually and finally rinsed in distilled water. Endogrenous peroxidase activity was blocked using 0.5% hydrogen peroxide in methanol at room temperature for 30 min. after which sections were rinsed in distilled water and PBS (phosphate-buffered saline). Twenty per cent rabbit serum was applied to the sections for 10 mn as a blockina reagyent to reduce non-specific bindingy. A 1:1000 dilution of the monoclonal antibody to p53 protein (Do7: Dako. Copenhagen. Denmark). w-hich recognizes both wild-type and mutant protein was used as the primary antibody. DO-7 recognizes an epitope in the N-terminus of the human p53 protein residing between amino acids 35 and 45. Sections were incubated at 4ZC overmight. After return to room temperature for 30 mmn. sections were rinsed in PBS and incubated with the biotinylated secondary antibody for 30 min. followed by streptavidin peroxidase reagents (StreptABComplex. Dako) for 30 mim. After washing in PBS. sections were incubated in diaminobenzidine solution for 5 mmn. washed in tap water for 10 iun and then counterstained with Mayer's haematoxylin.
DNA extraction from the archival material We extracted DNA from paraffin blocks using the method of Lungu et al (1992). Briefly. three 10 jm sections were taken from the paraffin block. placed in a microfuge tube with 150 ji of digestion buffer containinc 50 mM Tris (pH 8.5). 1 mMr EDTA. 0.5% Tween 20 and 200 jg ml-' Proteinase K. Sections were incubated at 65°C for 2 h. then heated to 95°C for 10 min to destroy the proteinase. The samples were then centrifuged for S min at full speed. after which the aqueous phase containing, DNA from the archival materials was removed and stored at -70'C.

RESULTS
The cancer samples generally had little stromal cell contamination and only a few samples contained a minor fraction of tumour cells. In general. two-thirds or more of the slides consisted of tumour cells. Onlv 3 out of the 77 biopsy specimens studied were considered to hax e scant tumour. The first had no detectable p53 mutation and high (>70%c) nuclear staining. The second had no detectable p53 mutation. and less than 10% nuclear staining. The third had a p53 mutation and less than 10% nuclear staining. Only the second patient was eligible for survixval analysis (treated for cure. no previous malignancy).

Immunohistochemistry
Fomr-four cases (57.1%) of the 77 analysed tumours revealed immunohistochemical positivity for p53 (> 10% staining nuclei). staining (10-70%) and 26% high staining (>70%). Neither for p53 mutation (missense. nonsense and silent) nor for p53 immunohistochemical overexpression (data not shown) were there any differ-10-70%no (n = 15) ences in the distribution of T or N status. stage. age or sex between the groups (Table 4). <10% (n = 28) Univariate analysis revealed both N status (P = 0.01. log rank) and p53 mutation (P = 0.001. log rank) (Figure 2) to be associated with survival. This was. however. not the case for immunohistochemical expression of p53 (Figure 3). Median survival for >70% (n = 15) patients with missense or nonsense p53 mutation was 12.5 months and for patients without such mutations median survival extended beyond 90 months. In a further subgroup. survival analysis of p53-80 10 mutated and non-mutated cases was subdivided into groups with or without increased immunohistochemical p53 expression. ly death for Whether or not the cut-off limit was between <10% and > 10% (10-70% staining staining nuclei. or between <70% and >70% staining nuclei (data (n = 58; P= NS) not shown), no further prognostic information was gained. Using this subdivision some of the groups, however, were small. In a multivariate analysis, p53 mutation was still a strong risk Among them 25 cases were scored 3 (more than 70% staining nuclei) and 19 cases were scored 2 (between 10% and 70% staining nuclei). Thirty-three cases (42.9%) showed score 1 (less than 10% staining nuclei) ( Table 2).

p53 mutations
Nucleotide sequence alterations were found in 17 (22%) of the 77 analysed tumours. Ten of these were missense mutations resulting in amino acid substitutions. two were nonsense mutations resulting in protein truncation, whereas five alterations would not lead to amino acid changes (Table 3). Thus, only 12 (16% of 77) tumours exhibited mutations that could alter the function of the protein.
Seven of these were transversion mutations (four G -e T. one C -e A. one C -* G. one G -* G). and five were transitions (two G -* A and three C -e T. none at CpG dinucleotides). In the following computations. only cases exhibiting missense or nonsense mutation were included.

PCR-SSCP vs IHC analysis
The concordance between p53 mutation, resulting in a missense or nonsense mutation, and increased p53 expression in immunohistochemistry was poor (Table 2). If only cases with either high or low IHC expression (n = 58) are considered, IHC and PCR-SSCP analysis were discordant in 25 cases (43%). A SSCP gel, sequencing gel and immunohistochemistry staining of a tumour with concordant findings are shown in Figure 1.

Survival vs p53 mutation/overexpression
Eight of 58 evaluable cases exhibited a p53 mutation resulting in a missense or nonsense mutation (Table 3). Of these 58 tumours. 48% exhibited low p53 IHC (<10% staining nuclei). 26% intermediate tactor and the impact ot N status was reduced below sigmticance (Table 5). DISCUSSION p53 mutations that lead to altered protein conformation can make the protein more stable and prolong its half-life (Lane and Benchimol. 1990). It is therefore possible to detect an accumulation of mutated p53 protein in head and neck cancer using immunohistochemistry. The frequency of p53 immunohistochemical overexpression in the present material (57% with > 10% staining nuclei) is in accordance with findings in previous studies of HNSCC (Field et al. 1991: Ogden et al. 1992: Watling et al. 1992: Dowell and Hall. 1994: Xu et al, 1994: Nylander et al. 1995.
As we previously reported in preliminary form (Mineta et al. 1995). there is a pronounced discordance between p53 mutation and p53 immunohistochemical overexpression. This confinns some earlier findings in HNSCC (Xu et al. 1994;Nylander et al. 1995) and in skin cancers (Kubo et al, 1994). but contradicts others (Ahomadegbe et al, 1995). The difference in findings is not likely to be attributed to the antibody used as DO-7 was applied in the present as well as in two (Ahomadegbe et al. 1995: Nylander et al. 1995 of the four other studies. False-negative findings (mutation without overexpression) can be attributed to p53 mutations at splice sites, frame shifts or nonsense mutations, which would be predicted to encode for truncated p53 proteins not detected by iinmmmunohistochemistry. Studies of the crystal structure of the p53 tumour suppressor-DNA complex (Cho et al, 1994;Milner, 1995) have made it clear that the majority of p53 point mutations affect the residues within the core domain and inactivate the function by abolishing its sequencespecific DNA-binding capacity. However, these mutations do not significantly affect the structure of the protein. Therefore, the discrepancy between the PCR-SSCP and the immunohistochemical results may depend on the site of p53 mutation or the anti-p53 antibody employed. Negative staining may also be due to overfixation. delayed fixation or inadequate tissue processing of the tumour sample, which allows degradation of the p53 protein.
The p53 antibody (Do-7) used in this study reacts with both the wild type and mutant type of the p53 protein. Our data, which p53 mutation and survival in H&N cancer 1089 demonstrate three cases with p53 missense mutations not detected immnunohistochemicallv. suggest that the mutational change in the p53 gene does not necessarily result in increased stability of the p53 protein.
Discordant results (i.e. nuclear staining without pS3 mutation) can depend on the sequencing strategy. For example. mutations may occur outside the studied exons 5. 6. 7 and 8. For HNSCC. however. approximately 98%e of mutations are found w-ithin exons 5-8 (Greenblatt et al. 1994). Other reasons for the failure of detection of p53 mutations include insufficient tumour in the sample or insufficient quality of DNA from archival materials.
Overexpression is not only seen in gene mutation. but also in cases with retention of the wild-type target protein in the tumour cell. Both mutational stabilization of the p53 protein and elevated levels of wild-type p53 protein allow detection by immunohistochemistry. Thus. if the secondary stabilization of p53 occurs by some mechanism other than gene mutation. overexpression can be demonstrated.
Accumulation of wild-type p53 protein has been found in an inherited cancer (Barnes et al. 1992) or cancer treated w-ith chemotherapeutic drugs or radiation (Kastan et al. 1991: Vogelstein andKinzler. 1992).
Such non-mutational stabilization of the protein is most probably the result of interruption of the normal degradative pathway of p53. Other proteins such as the products of cellular oncogene mdm-2 (Monaud et al. 1992: Meltzer. 1994. or the products of DNA v-iruses. including SV-40 large T antigen. E lb of adenovirus (Gannon et al. 1990: Cesarman et al. 1993) and E6 of HPV (Scheffner et al. 1990: Wemess et al. 1990) can bind to the p53 gene and inacti-ate the abilitv to act as a transcription factor. resulting in p53 stabilization. Phosphorylation due to cdc2-like kinase could also alter the p53 protein. which can then be detected bv immunohistochemistrv (Moll et al. 1992).
In the present study p53 mutations were found at a frequency of 16%c (12177). which is within the lower range of the percentage of abnormal findings reported in the literature in HNSCC (Boyle et al. 1993: Brennan et al. 1995: Dowell and Hall. 1994: Greenblatt et al. 1994: Xu et al. 1994: Nvlander et al. 1995. The low frequency may reflect the small sample size currentlv available. or it mav be due to inconsistent amplification of DNA from archi-al materials. or to lower mutated DNA concentration in samples. although mutations are reported to be still detectable when constituting only 15%7c of total DNA (Wu and Darras. 1993).
The issue of decreased sensitivitv caused by contaminatinc, benign cells is of greatest importance in PCR studies involving detection of loss of heterozygositv. for example. In studies such as ours. it is unclear how much contaminating cells lower the threshold of detection of p53 mutation in formalin-fixed material. if at all. and a sample judged insufficient in the amount of tumour for immunohistochemical assessment produced evidence of p53 mutation in the PCR. The proportion of tumour cells in all samples vastly exceeded. in any case. the generally accepted sensitivity of PCR in detecting 1 mutated cell amongy 100 000 normal cells.
The simple possibility that a dissimilarity exists between cohorts of patients with respect to carcinogen exposure may influence the findings. It also appears as though the frequency of detected mutations max be lower when DNA is retrieved from archival rather than from fresh tissue.
Studies have implicated p53 protein expression as an independent prognostic factor in carcinomas of the breast. stomach. colonlrectum, bladder and NSCLC [review-ed in Changa et al (1995). Dowell and Hall (1994)]. The clinical relevance of p53 overexpression in HNSCC has been under debate. We could not find any correlation between p53 immunohistochemical overexpression and survival (Figure 3). There are studies indicatinc a correlation between p53 overexpression and survival. some reporting better survival in patients wvith overexpression (Sauter et al. 1992). Overexpression has also been reported to show strong association with a histoloaical malignancy grading scale with prognostic capability (Watling et al. 1992). However. the lack of correlation betmeen p53 expression and clinicopathological parameters or survival as originally reported by Field et al (1991) has subsequently been substantiated by many reports [reviewed in Field et al (1993)].
Studies on the relation betmeen p53 mutation in HNSCC and clinicopathological parameters or survival are sparse. Koch et al (1996) found an association with recurrence but not survival. The present finding of p53 mutation as a strong and independent prognostic variable contrasts with the results of Ahomadeabe et al (1995). who did not find any correlation between mutation and clinical staae or 5-year survival. However. they studied fresh tissue from both metastases (n = 50) and primary tumours (n = 28). 13 of which were matched specimens. They also found a good correlation between mutation and overexpression. Nvlander et al (1995) in a studv of 80 HNSCCs of the oral cavity using archival specimens could not find any relation between p53 mutation and survival. Their material. however. deviated from the general characteristics of HNSCC. w-ith a male to female ratio of 0.86:1 and a high frequency of a novel non-random 14-bp deletion in exon 8 (Nylander et al. 1996). differences that might explain the discordant findings w-ith respect to survival.
In conclusion. we verified previous findings of a lack of concordance between immunohistochemical overexpression of nuclear p53 and mutation of the p53 gene. as well as the absence of prognostic information with respect to survival in p53 overexpression. On the other hand. p53 mutation seems to be a strong and independent variable for survival prognosis.