The prognostic significance of the accumulation of p53 tumour-suppressor gene protein in gastric adenocarcinoma.

We have studied the expression of p53 in 206 patients with gastric adenocarcinomas. A standard immunohistochemical technique employing the CM-1 anti-p53 polyclonal antibody was applied to the routinely fixed and paraffin-embedded material from these tumours; overexpression of p53 was defined as positive nuclear staining: 46% (94/206) of gastric carcinomas expressed high levels of p53. There was no significant correlation between p53 positivity and the tumour grade, growth pattern, the Lauren type or lymph node metastases. Correlation with disease stage was only marginally significant (P = 0.05). Life table analysis revealed a highly significant association between p53 expression and survival (P = 0.0062), the odds ratio of death being 1.89 (95% confidence interval 1.33-2.69). The overall 5-year survival of patients with p53-positive tumours was 3% compared with 16% for those with p53-negative tumours (median survival time being 5.6 and 11.4 months respectively). These data suggest that overexpression of the p53 oncoprotein is an independent marker of shortened survival in gastric cancer patients.

p53 is a nuclear phosphoprotein first discovered as a result of its binding to large T antigen, the dominant transforming oncogene of the DNA tumour virus SV40 (Lane & Crawford, 1979). Wild-type p53 protein is believed to be involved in the regulation of cellular proliferation, acting by inhibition of progression from the GI to the S-phase of the cell cycle (Levine et al., 1991).
The human p53 gene maps to the short arm of chromosome 17 (Isobe et al., 1986) and is frequently mutated in a wide variety of human cancers (Nigro et al., 1989;Lane & Benchimol, 1990; Bartek et al., 1991;Hollstein et al., 1991). These mutations (mostly missense) are clustered in the highly conserved regions of the p53 genome (Nigro et al., 1989;Hollstein et al., 1991;Varley et al., 1991) and, in addition, there is often accompanying loss of heterozygosity at l7pl3.1 (Nigro et al., 1989;Sano et al., 1991;Varley et al., 1991). Causal involvement of p53 abnormalities in many cancers is indicated by the high predisposition for the development of cancer of various types in transgenic mice with homozygous null mutations of the gene (Donehower et al., 1992), as well as in patients with germline mutations (Malkin et al., 1990). Transfection of wild-type p53 into cultured malignant cells has been shown to result in the inhibition of cellular proliferation as well as morphological transformation Mercer et al., 1990). Mutant p53 proteins may contribute to the malignant transformation not only by the inactivation of this suppressor activity but also by acting, in their own right, as dominant transforming oncogenes (Jenkins et al., 1985;Hinds et al., 1989).
As a consequence of its short half-life, wild-type p53 is present in cells in minute amounts and does not accumulate to detectable levels Reihsaus et al., 1990). In contrast, mutations are associated with a prolongation of the protein half-life; the mutant p53 gene products are more stable and accumulate to levels that can be detected by an immunohistochemical technique Lane & Benchimol, 1990). Antibodies specific for human p53 have been generated (Banks et al., 1986;Gannon et al., 1990;Bartek et al., 1991;Midgley et al., 1992;Vojtesek et al., 1992), and the findings of a large number of immunohistochemical studies have confirmed p53 accumulation to be a very common feature of human malignancy (Bartek et al., 1991;Scott et al., 1991;McLaren et al., 1992;Thor et  1992). Furthermore, detection of high cellular levels of p53 has been shown to correlate in most, but not all, cases with mutation of the gene (Bartek et al., 1990;Iggo et al., 1990;Rodrigues et al., 1990;Tamura et al., 1991).
In breast cancer (Thor et al., 1992), p53 overexpression has been shown to be of prognostic significance. However, in lung cancer, there is no such positive correlation (McLaren et al., 1992). Studies in colorectal cancer have so far produced conflicting results: some (Sun et al., 1992) link p53 overexpression with survival, whereas others (Scott et al., 1991) did not find any statistical correlation between high cellular levels of p53 and survival.
Gastric cancer is one of the most common malignancies and carries a poor prognosis (Breaux et al., 1990). Despite the decline in its incidence, the disease is still responsible for about 11,000 deaths per annum in the United Kingdom (Cancer Research Campaign, 1990). The molecular events leading to the development of gastric cancer are largely unknown, but there is now enough evidence to suggest that the functional inactivation of the p53 gene through allelic loss and mutation plays an important part (Kim et al., 1991;Sano et al., 1991;Tamura et al., 1991;Matozaki et al., 1992).
The aim of the present study was to determine whether p53 status in gastric cancer is correlated statistically with various histopathological indicators of poor prognosis and long-term survival. We have used the polyclonal antibody CM-1 in an immunohistochemical assay for p53 in formalinfixed, paraffin-embedded archival tissues from 206 patients with primary gastric adenocarcinomas.

Materials and methods
Patient population A total of 206 patients (93 males, 113 females; mean age 67 years, ranging from 38 to 83 years) with primary gastric adenocarcinomas diagnosed in Tayside between 1984 and 1987 were studied. All patients had undergone surgery for the primary disease and none had received preoperative or post-operative adjuvant therapy (radiotherapy or chemotherapy). The tumour histology was reviewed independently by one of us (D.H.) and the lesions were classified in relation to the growth pattern (Ming, 1977), nuclear grading (Watanabe et al., 1990), histologic type (Lauren, 1965), disease stage (Miwa, 1984) and lymphatic invasion (Table I). Representative formalin-fixed, paraffinembedded tissue blocks from each tumour were then used for the immunohistochemical assay for p53.
Statistical analysis Survival data were available for 160 patients. Life table analysis was performed using the SPSS-PC statistical package, which was also used to compare survival between subgroups using the Lee-Desu statistic (Lee, 1980). To assess the effects of risk factors while allowing for the presence of other influences on survival, a proportional hazards model (Cox, 1972) was fitted using the Nanostat program, which was also used for plotting the unadjusted survival curves.

Immunohistology
Four tissue blocks were examined from each patient. Four micron paraffin sections were cut, mounted and air dried overnight at 37°C. The sections were then stained for p53 using the indirect immunoperoxidase technique described previously (Joypaul et al., 1993a). The antibody used was CM-1, a rabbit polyclonal antiserum raised against recombinant human p53 protein (Midgley et al., 1992). Briefly, CM-1 was applied at a dilution of 1:500 after blocking with 2% goat serum. Endogenous peroxidase was then blocked using 3% hydrogen peroxide. Biotinylated anti-rabbit immunoglobin antiserum was then applied at 1:25 dilution followed by horseradish peroxidase-labelled steptavidin (StrAviGen Immunodetection System, BioGenex Labs, UK). The reaction was developed using diaminobenzidine as the chromogen and the sections were then counterstained in light haematoxylin.
We used Tris-buffered saline (TBS), instead of the primary antibody, as negative controls in each case, and the positive control sections were from a breast adenocarcinoma known to express high levels of p53.
The sections were then assessed independently by two pathologists (D.H. and S.Q.) without knowledge of the clinical outcome of the patients; nuclear staining of cells was considered positive for abnormal p53 expression.

Results
Positive p53 staining was observed in 94 of the 206 tumours examined (46%). In no case studied was there staining of the normal gastric epithelial cells or of stromal cells. The staining was nuclear in all instances (Figures 1 and 2) with background or cytoplasmic staining being absent or minimal. In the majority of the p53-positive tumours (> 70%), a diffuse staining pattern in which all or nearly all of the cancer cells exhibited nuclear positivity was observed (Figure 1). In contrast to this homogeneous pattern, the nuclear staining in the remaining positive tumours was more varied. Here, as shown in Figure 2, only a fraction (less than 25%) of the carcinoma cells contained immunoreactive p53 protein. Table I summarises the histological classification and staging of the 206 gastric adenocarcinomas studied. The numbers of patients in each group together with the corresponding immunohistochemical results for p53 expression are also shown. Our findings indicate that there was no significant association between p53 nuclear positivity and the tumour growth pattern (expansile or infiltrative) or the histological type (Lauren classification). Similarly, no relationship was found between p53 expression and the nuclear grade of the tumours. As shown in Table I, 128 of the gastric cancers had metastasised to the perigastric lymph nodes; 52 had not and there were no data available for the remaining 26 cases. There was no significant relationship between p53 positivity and nodal status (48% versus 39% respectively for the cases with and without lymph node metastases).
Of the 206 gastric adenocarcinomas, ten were confined to the mucosa (stage I), 30 reached the submucosa (stage II), 73 involved the muscularis propria (stage III) and the remainder (n = 93) reached the serosa (stage IV). Our results show a trend towards an increase in p53 expression with increasing   (Table I); this was found to be marginally significant (P = 0.05).
Follow-up data were available for 160 of the 206 patients (the prevalence of p53 positivity was 43%, not significantly different from the total group) and, using the method of Kaplan and Meier (Kaplan & Meier, 1958), survival curves were plotted by p53 status (Figure 3). Analysis based on the Cox's proportional hazards model (Cox, 1972) revealed a significantly shortened survival time (P = 0.0062) in the p53positive cohort (n = 69) compared with the p53-negative group (n = 91). The odds ratio of death, after allowing for the effects of the other histopathological parameters in this multifactorial analysis, was 1.89 (95% confidence interval 1.33-2.69). The effects of p53 expression on survival was not dependent on tumour staging. Also, no difference in survival was detected between the group of patients with homogeneous p53 staining and those with a heterogeneous pattern.
The median survival time of the patients with and without p53 expression was 5.6 and 11.4 months respectively ( Figure  3). The overall 5-year survival for the p53 negative group was 16% compared with 3% in the p53-positive group, thus confirming that in gastric cancer patients p53 expression is associated with a poor prognosis.

Discussion
The placement of oncogenes in the aetiology of human malignancy has resulted in the evaluation of an increasing number of molecular markers as useful diagnostic or prognostic indicators. p53 is the most widely studied marker and we have previously reported that its overexpression occurs as a late event in the gastric carcinogenic pathway (Joypaul et al., 1993a), suggesting that p53 mutation is an important step in the pathogenesis of gastric cancer. In the present report, we have extended the study population to 206 patients with primary gastric adenocarcinomas in order to determine the association between p53 overexpression and long-term survival.
We detected increased expression of the p53 protein in 46% of the gastric tumours, and this compares favourably with the results of a smaller study on gastric cancer (Martin et al., 1992). In addition, the data are consistent with the findings in other malignancies (Nigro et al., 1989;Bartek et al., 1991;Scott et al., 1991;McLaren et al., 1992;Thor et al., 1992). In the majority of the positive carcinomas, staining of the p53 protein was intense and confined to all or nearly all of the tumour cells ( Figure 1). In contrast, in the remaining positive cases, p53 staining was still nuclear but limited to a variable fraction of the malignant cells (usually fewer than 25%, as illustrated in Figure 2). Although this heterogeneity in staining has been attributed to causes such as the specific type of p53 mutation (Varley et al., 1991) or cell cycle variation in p53 levels (Morkve & Laerum, 1991;Purdie et al., 1991;Varley et al., 1991) there is no doubt that it represents abnormal p53 expression. It is also important to note that this heterogeneity in staining may lead to problems of sampling, and in this context the recently described methods to detect p53 in homogenised tumour samples may prove to be a useful alternative (Joypaul et al., 1993b).
In the present study, comparison of p53 nuclear staining with various markers of high malignant potential (tumour grade, lymph node metastases and depth of tumour penetration) did not identify any significant correlation. In contrast, analysis based on the Cox test (Figure 3)  expression to be an independent marker of shortened survival time (P = 0.0062). This association is particularly informative since our study population was not biased by different treatment protocols as none of the patients received adjuvant therapy (chemotherapy and/or radiotherapy). To our knowledge, only one small study of 75 patients with gastric cancers has examined p53 expression in relation to survival, and the results are similar to ours (Martin et al., 1992). Furthermore, correlation between accumulation of p53 and survival has been reported in breast cancer (Thor et al., 1992). Studies in colorectal and lung cancers have, however, produced conflicting data (Scott et al., 1991;McLaren et al., 1992;Sun et al., 1992). The cause of this possibly tissuespecific variation is at present unknown. While the use of p53 as a prognostic marker in gastric cancer would appear to be justified, it is also important to consider the molecular basis of its accumulation and its possible role as a target of novel therapeutic agents. Until recently, molecular studies have confirmed point mutations in the gene to be the most usual cause of p53 accumulation (Bartek et al., 1990;Iggo et al., 1990;Rodrigues et al., 1990;Tamura et al., 1991), but there is now increasing evidence to suggest that other events can also result in overexpression of the protein. These include changes in the cellular environment resulting from DNAdamaging events (Hall et al., 1992) or as part of the normal apoptotic pathway (programmed cell death) (Lane, 1993). In addition to its general correlation with expression of mutant p53, high levels of p53 have also been linked with an increase in cellular proliferation (Scott et al., 1991). Deficiency of wild-type p53 could conceivably lead to a more aggressive phenotype as a result of an excessive number of cells being non-quiescent. Agents which could convert mutant p53 protein into a more 'wild type' conformation might therefore be expected to inhibit proliferation and possibly induce apoptosis in cells with a p53-positive phenotype (Hupp et al., 1993). In conclusion, our study has demonstrated immunodetection of p53 in 46% of gastric adenocarcinomas, associated with a significantly shortened survival. The hope therefore remains that novel therapeutic agents that can target abnormal or mutant p53 (thus nullifying its growth-promoting effects) will be developed and become available in the near future (Hupp et al., 1993).
We thank Stewart McPherson and George Coghill for the excellent technical assistance given.