Prognostic significance of TP53 alterations in breast carcinoma.

Constant denaturant gel electrophoresis (CDGE) was used to screen 179 breast carcinomas for mutations in the conserved regions of the TP53 gene (exons 5 through 8). Mutations were found in 35 of 163 primary tumours (21%) and in 5 of 16 metastases (31%) and resided predominantly in exon 7. The majority of the mutations were G:C-->A:T transitions. Immunohistochemistry demonstrated nuclear accumulation of p53 protein in 35 of 162 primary tumours (22%) and in four of 15 metastases (27%). TP53 mutation was strongly associated with nuclear accumulation of p53 protein. In total 42 of 163 primary tumours (26%) and 5 of 16 metastases (31%) were demonstrated to contain TP53 alterations (mutation and/or nuclear protein accumulation). TP53 alteration in primary tumour was significantly associated with the following parameters: positive node status, T status > 1, negative oestrogen receptor status, negative progesterone receptor status, presence of ERBB2 gene amplification, and invasive ductal histology. Furthermore, there were statistically significant associations, independent of other prognostic factors, between TP53 alterations in primary tumour and disease-free and overall survival.

Prognostic significance of TP53 alterations in breast carcinoma T.I. Andersen', R. Holm2, J.M. Nesland2, K.R. Heimdall, L. Ottestad3 & A.-L. B0rresen' 'Departments of Genetics and 2Pathology, Institute for Cancer Research, 3Department of Oncology, The Norwegian Radium Hospital, N-0310 Oslo, Norway. Summary Constant denaturant gel electrophoresis (CDGE) was used to screen 179 breast carcinomas for mutations in the conserved regions of the TP53 gene (exons 5 through 8). Mutations were found in 35 of 163 primary tumours (21%) and in 5 of 16 metastases (31%) and resided predominantly in exon 7. The majority of the mutations were G:C+A:T transitions. Immunohistochemistry demonstrated nuclear accumulation of p53 protein in 35 of 162 primary tumours (22%) and in four of 15 metastases (27%). TP53 mutation was strongly associated with nuclear accumulation of p53 protein. In total 42 of 163 primary tumours (26%) and 5 of 16 metastases (31%) were demonstrated to contain TP53 alterations (mutation and/or nuclear protein accumulation). TP53 alteration in primary tumour was significantly associated with the following parameters: positive node status, T status > 1, negative oestrogen receptor status, negative progesterone receptor status, presence of ERBB2 gene amplification, and invasive ductal histology. Furthermore, there were statistically significant associations, independent of other prognostic factors, between TP53 alterations in primary tumour and disease-free and overall survival.
Mutations in the TP53 gene have been found in a variety of human malignancies, and are considered to represent the most common genetic alterations in human cancer (Levine et al., 1991). Inherited mutations in the heterozygous state have been described in noncancerous cells from members of Li-Fraumeni syndrome families (Malkin et al., 1990;Srivastava et al., 1990), and have recently been identified in breast cancer patients outside such families as well (B0rresen et al., 1992;Malkin et al., 1992;Prosser et al., 1992;Sideransky et al., 1992).
The TP53 tumour suppressor gene is located at chromosome 17pl 3.1 and encodes a 53 kDa cell cycle regulatory nuclear phosphoprotein (Levine et al., 1991;Lane, 1992). Most of the TP53 mutations reported in human cancers are located within the evolutionary conserved regions of the gene (codons 110-307 of totally 393) (Hollstein et al., 1991;deFromentel & Soussi, 1992). The mutations usually are missense, giving rise to altered proteins (Levine et al., 1991). An altered conformation enables most of the mutant proteins to inactivate wild-type protein by forming inactive oligomeric complexes (Nigro et al., 1989). Most of the mutant p53 proteins have considerable longer half-life than wild-type protein, resulting in accumulation of the mutant protein in the transfected or neoplastic cells (Hinds et al., 1990;Iggo et al., 1990). The negative r:egulatory effects of TP53 upon cell proliferation has been demonstrated to be inactivated by mutations and by the presence of DNA virus proteins (Lane, 1992;. Furthermore, increased levels of MDM2 protein and sequestering of p53 protein in the cytoplasm have been suggested to be associated with inactivation of TP53 tumour suppressor function (Moll et al., 1992;Momand et al., 1992;Oliner et al., 1992;. Several findings indicate that inactivation of TP53 is associated with a growth advantage in breast carcinoma: Accumulation of p53 protein has been reported to be associated with high grade tumours, increased levels of epidermal growth factor receptor (EGFR), presence of the proliferation associated antigen Ki67, advanced stage, metastatic spread, and low concentrations of oestrogen and progesterone receptors (Cattoretti et al., 1988;Thompson et al., 1990;Davidoff et al., 1991a;Varley et al., 1991;Mazars et al., 1992). Recent reports conclude that positive p53 immunostaining in primary tumour represents a prognostic parameter, suggesting that p53 protein accumulation might become clinically useful as an indicator of breast cancer aggressiveness (Iwaya et al., 1991;Ostrowsky et al., 1991;Varley et al., 1991;Isola et al., 1992;Mazars et al., 1992;Thor et al., 1992).
The aims of the present study were (a) to determine the nature and frequency of both TP53 mutations and p53 protein accumulation in breast carcinomas, (b) to study the associations between TP53 alterations, ERBB2 gene amplification, histopathological and clinical parameters, and (c) to evaluate if TP53 alterations provide prognostic information in primary breast carcinoma.

Materials and methods
Patient material Material for this study was obtained from 163 patients with primary breast carcinomas and 16 patients with breast carcinoma metastases admitted to The Norwegian Radium Hospital during the period from 1984 to 1989. The 163 patients with primary carcinoma that were included in the survival studies had a mean observation time of 45.8 months (range 0.1 to 94.1 months, median 38.3 months) and a mean age at diagnosis of 56.6 years (range 31.1 to 85.6 years, median 56.5 years). Lymph node dissection was performed on all but four of the patients with primary carcinoma. Tumour and node status was decided based on the pathologists reports, according to the 1988 TNM classification (Beahrs et al., 1988). Formalin-fixed, paraffin embedded tumour tissue from each case was processed for light microscopy and examined by the pathologist (JMN) according to the WHO criteria.
Tumour tissue was obtained from each patient at surgery. The samples included 163 primary tumours, 12 locoregional recurrences, two supraclavicular, one bronchial, and one skin metastasis. The tumour tissue was immediately frozen and stored in liquid nitrogen for DNA and hormone receptor analyses. Frozen tissue for immunohistochemistry was available from 51 of the primary tumours and nine of the metastases, whereas paraffin embedded tumour tissue for such analyses was available in all but two (one primary tumour and one metastasis) of the remaining cases.
DNA analyses DNA was isolated from tumour tissue by phenol/chlorophorm extraction followed by ethanol precipitation (Kunkel et al., 1977). Amplification of exons 5-8 of the TP53 gene was performed by the polymerase chain reaction (PCR) using primers and conditions as previously described (B0rresen et al., 1991;Smith-S0rensen et al., in press). The five PCR fragments covered the evolutionary conserved regions of the gene, in which more than 80% of the reported mutations have been found; codons 128-153 (exon 5), codons 155-185 (exon 5), codons 189-215 (exon 6), codons 237-253 (exon 7), and codons 265-301 (exon 8). The five fragments from each tumour were screened for TP53 mutations using constant denaturant gel electrophoresis (CDGE) as previously described (B0rresen et al., 1991;Condie et al., in press;Smith-S0rensen et al., in press). Samples that had abberantly migrating bands in the CDGE analyses were reamplified and the abberant bands were confirmed by perpendicular denaturing gradient gel electrophoresis (DGGE). Exon 6 fragments that gave abberant bands were digested with TaqI and analysed on a polyacrylamide gel to identify samples with the codon 213 A-*G polymorphism (Serra et al., 1992). All PCR fragments that had a mobility different from normal DNA in CDGE and DGGE analyses were considered to be mutant, except the exon 6 fragments demonstrated to contain the codon 213 polymorphism. Mutant samples were submitted to PCR with one biotinylated primer, and were sequenced directly with standard dideoxy sequencing reactions, using Dynabeads M280-Streptavidin (Dynal AS, Norway) as solid support.
Mutation analysis (exons 5, 7 and 8) of 32 of the samples and ERBB2 gene amplification studies of 121 of the primary tumours have previously been published (B0rresen et al., 1990;B0rresen et al., 1991;Ottestad et al., in press). Immunohistochemistry Frozen sections and paraffin embedded tissue were immunostained using the avidin-biotin-peroxidase complex (ABC) method (Hsu et al., 1981). The frozen sections were air-dried overnight, fixed in cold acetone at 4°C for 10 min and washed in phosphate buffered saline (PBS), pH 7.4. The paraffin embedded tissue was fixed in 4% buffered formaldehyde at room temperature. Four to six gm thick sections from paraffin embedded blocks were mounted on coated slides and incubated for 30 min at 56°C and overnight at 37°C. The three monoclonal antibodies PAb 421, PAb 1801, and PAb 240 (Oncogene Science) were applied on frozen sections, whereas PAb 1801 and the polyclonal NCL-CM1 (Novacastra Laboratory Ltd.) were applied on paraffin embedded material. Twenty-two tumours were tested with PAb 1801 on frozen sections as well as paraffin embedded tissue to compare the quality of the stainings. PAb 421, PAb 1801, and NCL-CM1 detect mutant and wild-type protein, whereas PAb 240 recognises mutant and denatured wild-type protein. Briefly, the sections were incubated for 18-22 h at 4°C with specific primary antibodies (PAb 421 diluted 1:100 (1 fig IgG2U mI'), PAb 1801 diluted 1:300 (0.33 gtg IgG, ml-), PAb 240 diluted 1:100 (1 jg IgG, ml-'), and NCL-CM1 diluted 1:300). The reactions were then incubated with biotin labelled secondary antibody and avidin-biotinperoxidase complex. The peroxidase reaction was developed using diamino-benzidine as a chromogen. All series included positive controls. Negative controls included substitution of polyclonal primary antiserum with rabbit serum diluted 1:300, whereas negative controls for the monoclonal antibodies were performed using mouse myeloma protein of the same subclass and concentration as the monoclonal antibody. All controls gave satisfactory results. Only cells with nuclear staining were scored as p53 protein immunopositive. The amount of immuno-positive cells was semiquantitatively estimated.
Oestrogen and progesterone receptor determinations Oestrogen and progesterone receptors were determined using monoclonal antibodies in an enzyme immunoassay for quantitative measurement (Abbott ER and PgR-EIA monoclonal). Samples with a receptor concentration exceeding 9 pmol g-' protein were considered to be receptor positive.

Statistical analyses
Comparisons between groups were performed using Chi square tests with Yates' correction. Disease free and overall survival were calculated using the life-table method, and differences between survival curves were tested using the log rank test. To simultaneously analyse the importance of several prognostic factors, the Cox proportional hazard model (Cox, 1972) with a stepwise procedure, was used. The following variables were included in the analyses: TP53 mutation, nuclear p53 protein accumulation, TP53 alterations (mutation and/or nuclear protein accumulation), node status (node negative vs node positive), T-status (T = 1 vs T> 1), ERBB2 gene amplification, oestrogen and progesterone receptor status. ERBB2 gene amplification was not included in the multivariate analyses since 42 of the primary tumours were lacking ERBB2 data. P-values < 0.05 were considered statistically significant. All statistical analyses were performed using the BMDP statistical software package (Dixon et al., 1990).

TP53 mutation analyses
The CDGE analyses demonstrated that 35 of the 163 primary tumours (22%) and 5 of the 16 metastases (31%) contained mutations in exons 5 through 8. One primary tumour contained two independent mutations (tumour no. 29, Table I). A representative CDGE gel is shown in Figure la. Samples that were suspected to harbour mutations were reamplified and submitted to DGGE (Figure lb). PCR fragments that had a mobility different from normal DNA in the CDGE and DGGE were considered to be mutant. However, exon 6 fragments that had aberrant bands were not considered to be mutant if they contained the polymorphic TaqI site in codon 213. The approximate position and nature of each mutation could be predicted from these gels. For instance, fragments that migrated more slowly than the wild-type were considered to have undergone G:C-*A:T transitions, since such mutations result in the destabilizing loss of a hydrogen bond. The exon in which each mutation resided, is given in Table I. Thirteen mutations were found in exon 5, three in exon 6, twelve in exon 7, and thirteen in exon 8. Accordingly, mutations were most frequently observed in exon 7 (0.235 mutations/base pair screened in exon 7 vs 0.117, 0.076, and 0.037 mutations/base pair screened in exons 8, 5 and 6, respectively, P = 0.0014).
Samples that were mutant according to the melting gel analyses (CDGE and DGGE) were submitted to direct sequencing ( Figure lc and Table II). The sequence analyses failed to confirm mutations in six samples that were mutant according to the melting gels. These results were confirmed by reamplification and resubmission to CDGE, DGGE, and sequence analysis. It should be noted that the mutant bands were faint in three of the six tumours (tumours no. 118, 228 and 416), but that the heteroduplexes convincingly indicated mutations. Furthermore, four of the six samples had nuclear accumulation of p53 protein (tumours no. 406, 416, 440, and 449, Tables I and II). Table II shows that 29 point mutations and six deletions were found by sequencing. Five point mutations were found in codon 248, three in codon 273, and two in codons 128, 281, and 282. The majority of the point mutations appeared at G:C base pairs (22 of 29, 76%). G:C-+A:T transitions accounted for 17, G:C-*T:A transversions for 4, and G:C-*C:G transversions for 1 mutation (59, 14, and 4% of the point mutations, respectively). Four of the point mutations at T:A base pairs were T:A-*G:C transversions and three were T:A-*C:G transitions. The deletions consisted of three single base pair losses (tumours no. 22, 71, and 418), one 8 base pair loss (tumour no. 135), and two identical 17 base pair losses (tumours no. 20 and 193, involving codons 174-180).
Neither of the two tumours with stop codon creating mutations (Table II, no. 78 at codon 204 and no. 434 at codon 271) contained cells with nuclear p53 protein accumulation, whereas the only tumour with a sense mutation (Table II,  Immunohistochemistry demonstrated that 35 of 162 primary carcinomas (22%) and four of 15 metastases (27%) contained cells with nuclear p53 protein accumulation (Table 1). The p53 protein immunopositive tumours exhibited diffuse or granular nuclear staining. Representative positive stainings are shown in Figure 2a-c. No immunoreactivity was seen in the normal breast tissue surrounding the tumours, but in a number of cases PAb 240 gave nonspecific binding in connective tissue. These samples were scored as negative.
Twenty-two tumours were analysed with PAbl801 on frozen as well as on paraffin embedded tissue. Ten of these samples were scored as positive and nine as negative on both frozen and paraffin embedded tissue, whereas three samples gave positive staining only on paraffin embedded tissue.
The association between TP53 mutation and p53 protein accumulation was highly significant. Among the 177 tumours that were analysed for both kind of alterations, nuclear p53 protein accumulation was detected in 32 of 39 mutated and seven of 138 unmutated samples (P<0.00001). In total 42 of 163 primary tumours (26%) and five of 16 metastases (31%) had some kind of TP53 alteration. Table III gives the relationships between TP53 alterations in primary tumour and several clinical and histopathological parameters, including ERBB2 gene amplification. TP53 alterations were significantly more often found in ductal tumours, tumours with T-status larger than 1, node positive tumours, oestrogen receptor negative tumours, progesterone receptor negative tumours, tumours with ERBB2 gene amplification, and invasive ductal carcinomas.

Relationship to clinical and histopathological parameters
Univariate survival.analyses Univariate analyses (Mantel Cox tests) demonstrated significantly shorter overall and disease-free survival for patients who had primary tumours with TP53 alterations (Table IV). TP53 mutation was a somewhat weaker, but  significant predictor of both overall and disease-free survival, towards reduced overall and and so was nuclear p53 protein accumulation. Node status TP53 alterations ( Figure 3a represented the most powerful prognostic factor in these analyses. The node positive patients (n = 59) showed a trend towards reduced overall survival and had borderline Multivariate survival analyses significantly reduced disease-free survival if their primary The multivariate analyses s tumours harboured TP53 alterations, whereas the node alterations as significant vari negative patients (n = 100) showed non-significant trends vival, whereas node status, disease-free survival if they had and b).
;elected node status and TP53 iables for predicting overall sur-TP53 alterations, and T-status were selected as significant variables for predicting diseasefree survival (Table Va and b). When substituting TP53 mutation or nuclear p53 protein accumulation for TP53 alterations in the models, these factors were weaker, but still statistically significant independent prognostic parameters for predicting disease-free survival (P = 0.023 and P = 0.036, respectively). In the overall survival analysis TP53 mutations were of borderline significance (P = 0.051), whereas p53 protein accumulation not turned out to be of independent prognostic relevance (P = 0.200).

Discussion
The present work demonstrates TP53 mutations in exons 5 through 8 in 21% of the primary breast carcinomas. The frequency has previously been reported to be 17-46% (Kovach et     al., 1993). The varying frequencies might be due to differences in the proportion of early stage patients in the populations studied. Furthermore, it might reflect that different regions of the gene have been screened for mutations. In the present work mutations were observed more frequently in exon 7 (codons 237-253) than in exons 5, 6 and 8 (mutations/base pair screened). This observation is particularly interesting considering the fact that parts of exon 7 are extremely well conserved throughout the evolution. The 17 amino acid long stretch encoded by codons 237-253 has been demonstrated to be identical in rainbow trout, xenopus, chicken, rat, mouse, monkey, and human (Soussi et al., 1990). Six of the mutations detected by CDGE and DGGE were not confirmed by the sequence analysis. The fact that four of these tumours had nuclear accumulation of p53 protein makes it reasonable to believe that the discrepancy is caused by a lower sensitivity of the sequence analysis than that of the CDGE. It should be noted that the mutant bands were faint in three of the six samples, making the identification of the mutations relyant on the presence of heteroduplexes. The heteroduplexes that easily are recognised in melting gels enable detection of mutations when present in 5% or more of the cell population .
In the present work 76% of the point mutations (22/29) were found at G:C base pairs even though the overall G:C content in the TP53 gene not is higher than 56% . This excess is in accordance with what has been reported by others (Hollstein et al., 1991;Coles et al., 1992;de Fromentel & Soussi, 1992). The nucleoside guanosine has been reported to be a preferential target for most chemical carcinogens (Kriek et al., 1984). CpG nucleotides are preferentially thought to be involved in spontaneous mutations (de Fromentel & Soussi, 1992). In the present work codons 175, 181, 248, 273, and 282 (CGN) that encode arginine accounted for 12 of the 22 mutations in G:C base pairs (55%).
The majority of the point mutations were G:C-*A:T transitions (59%). This frequency is somewhat above the 40% reported in breast carcinomas by Hollstein et al. (1991), who further reported the frequency in colon carcinomas and brain tumours to be 79 and 75%, respectively. G:C-*T:A transversions account for 14% of the point mutations in our breast carcinomas. It should be noted that such transversions never have been found in colon carcinomas, but that they occur at a high frequency in non small cell lung cancer (57%), liver cancer (74%), and oesophageal cancer (24%), which are associated with specific mutagenic factors (Hollstein et al., 1991). The G:C-*T:A transversions observed in sporadic breast carcinomas might indicate that external or internal produced carcinogens take part in the development of these tumours.
The identification of nuclear accumulation of p53 protein in 22% of the primary breast carcinomas in the present study is significantly less than the 57% reported by Thompson et al. (1990) at the mRNA level, the 54% reported by Bartek et al., (1990a) using PAb 1801, 240, and 421, and the 45% reported by Cattoretti et al. (1988) using PAb 1801. On the other end of the scale positive p53 protein immunostainings have been reported in 22 and 24% of primary breast carcinomas (Crawford et al., 1984;Davidoff et al., 1991b;Thor et al., 1992). The varying frequencies can be explained by differences in the proportion of early stage tumours, by varying quality of the antibodies used, by systematical differences in interpretation, and by the observation that frozen samples that have been stored over several years show a lower frequency of positive reactions (Bartek et al., 1990a).
In accordance with Cattoretti et al. (1988), but in contrast to Davidoff et al. (199la,b) we interpreted cases with only few nuclei immunoreactive for p53 protein as immunostaining positive. Several tumours with positive nuclear staining in rare cells were demonstrated to contain mutations in the present study, supporting the interpretation that these tumours did contain accumulated p53 protein.
In agreement with Cattoretti et al. (1988) and Bartek et al. (1990b) we found a discrepancy between the immunoreactivity of the different p53 protein monoclonal antibodies. This could be explained by varying affinity of the antibodies and by differences in the availability of the epitopes when the different mutant p53 proteins complex with other proteins (Yewdell et al., 1986).
The strong association between presence of TP53 mutations and presence of nuclear p53 protein accumulation (P<0.00001) is in agreement with previous reports (Bartek et al., 1990b;Davidoff et al., 1991b), and fits the observation that mutant p53 protein has a half-life exceeding that of wild-type protein (Nigro et al., 1989). It furthermore indicates that nuclear p53 protein accumulation in most cases is due to mutations within the gene.
Mutations were not detected in 7 of the 39 tumours that were p53 protein immunopositive. These tumours might contain mutations outside the screened regions of the gene, or they might contain insufficient amounts of mutant cells. Furthermore, some of the tumours might have alternative mechanisms of TP53 inactivation, such as MDM2 gene amplification (Momand et al., 1992;Oliner et al., 1992) or alterations in TP53 regulatory sequences. Recently, two affected individuals in a cancer prone family were reported to have high levels of p53 protein constitutively expressed in their normal epithelial, endothelial, and stromal cells (Barnes et al., 1992). TP53 was not mutated in these patients, suggesting that mutation in another gene was causing both the high level of p53 protein and the increased risk of neoplasia in this family.
The fact that 7 of 39 samples with mutations were immunostaining negative can be explained by presence of senseand stop codon mutations (shown for tumours no. 78 and 434, Tables I and II), by frameshift mutations that give lead to an abnormal aminoacid sequence or to stop codons down the line, and by the possibility that not all amino acid substitutions lead to stabile complex formation resulting in increased half-life. Furthermore, insufficient quality of the frozen tissue subjected to analysis might lead to false negative results.
The present work demonstrates that TP53 alterations are associated with several clinical and histopathological parameters previously demonstrated to be of prognostic relevance (Table III). The observation that TP53 alterations were detected more frequently in oestrogen and progesterone receptor negative tumours confirms previous findings (Thompson et al., 1990;Iwaya et al., 1991;Ostrowsky et al., 1991;Varley et al., 1991;Isola et al., 1992;Thor et al., 1992), and might reflect lack of differentiation of the breast carcinoma cells.
Recently, wild-type p53 has been postulated to be important in holding up the cell to repair DNA damage (Levine et al., 1991;Kuerbitz et al., 1992;Lane, 1992;. Tumour cells in which TP53 is inactivated do not switch off replication in order to allow extra time for repair. This might explain the observation that primary tumours with TP53 alterations frequently also contain ERBB2 gene amplification (Table III) (Horak et al., 1991;Mazars et al., 1992).
Our data demonstrate that TP53 alterations in primary breast carcinomas may represent an independent prognostic parameter for predicting disease-free and overall survival, but any conclusions based on the multivariate survival analyses should be drawn with caution. The prognostic relevance was most pronounced when the data on mutations and protein accumulation were combined. This probably reflects that neither mutational analysis nor immunohistochemistry alone identifies all TP53 tumour suppressor inactivated tumours.
Node status is the most important established prognostic factor permitting identification of breast cancer patients at risk of relapse. Nevertheless, 20-30% of the node negative patients do relapse (McGuire, 1989). Use of tumour markers to predict unfavourable prognosis independent of node status could permit the identification of a subset of node negative breast carcinoma patients that would benefit from adjuvant treatment. During the last decade several molecular and cellular markers have been proposed as possible prognostic indicators (McGuire et al., 1992). In the present material the node negative patients with TP53 alterations showed trends towards shortened, although not statistically significant reduced survival (Figure 3a and b). Further studies will be required to fully evaluate a possible relevance of TP53 alterations in node negative breast carcinoma. Sigrid Lystad, Merete Hektoen, Ellen Hellesylt, and Mette Myre are gratefully acknowledged for excellent technical assistance. T.I.A. and K.R.H. are research fellows of the Norwegian Cancer Society. This study was supported by grants from the Norwegian Cancer Society and Grete Harbitz legacy.