| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Imaging, Diagnosis, Prognosis |
Authors' Affiliations: 1 IARC, Lyon, France; 2 Department of Genetics, The Norwegian Radium Hospital, and Faculty Division, University of Oslo, Oslo, Norway; 3 INSERM U379, ORS PACA, Epidémiologie Sociale Appliquée à l'Innovation, Marseille, France; 4 Department of Oncology (Radiumhemmet), Karolinska Hospital and Institute, Stockholm, Sweden; 5 The Icelandic Cancer Society, Molecular and Cell Biology Research Laboratory, and Faculty of Medicine, University of Iceland, Reykjavik, Iceland; 6 INSERM E 229, CRLC Val d'Aurelle/Paul Lamarque, Montpellier Cedex, France; 7 INSERM U735/Oncogénétique, Centre René Huguenin, St. Cloud, France; 8 Paterson Institute for Cancer Research, Manchester, United Kingdom; 9 Department of Oncogenetics, Centre Jean Perrin, Clermont-Ferrand Cedex, France; Departments of 10 Clinical Genetics and 11 Oncology, Oulu University Hospital, University of Oulu, Oulu, Finland; 12 Molekulargenetisches Labor, Frauenklinik des Universitätsklinikums Düsseldorf, Gebäude, Düsseldorf, Germany; and 13 Department of Clinical Oncology, Institute of Development Aging and Cancer, Aoba-ku, Sendai, Japan
Requests for reprints: Pierre Hainaut, IARC, 150 Cours Albert Thomas, F-69372 Lyon, France. Phone: 33-472-738-532; Fax: 33-472-738-322; E-mail: hainaut{at}iarc.fr.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Several studies have investigated the predictive value of TP53 mutation status for tumor response to treatment and patient outcome in various cancers. Different clinical and methodologic settings have been used and the results are often contradictory. A majority of studies have relied on immunohistochemistry to assess p53 alterations. This approach is, however, a poor surrogate for gene mutation detection because many mutations do not lead to protein accumulation, and because accumulation of wild-type p53 may also occur. Hence, the use of immunohistochemistry leads to an unacceptable number of misclassified cases and to a greater interstudy variability. By contrast, in studies that have screened TP53 mutations by gene sequencing to precisely identify the mutation, the presence of a mutation has been correlated with a shorter survival or a poor response to treatment in several cancers (http://www-p53.iarc.fr/Somatic.html). Moreover, a number of studies have described specific types of mutation that were associated with a worse prognosis compared with other mutations. This is the case for mutations within the DNA-binding domain that have been repeatedly associated with poor prognosis in several types of cancer (4–7). These clinical results are substantiated by in vitro experimental evidence showing that different missense mutations have different functional consequences (see TP53 Function Database, http://www-p53.iarc.fr/). Wild-type TP53 activities rely mainly on the capacity to transactivate specific target genes by binding to specific response elements. In human cancers, >1,800 different TP53 missense mutations have been reported and functional assays have shown that mutant proteins show a great variability in their transactivation activities. Whereas hotspot missense mutations in the DNA-binding domain lead to a general loss of specific transactivation capacity, missense mutations outside the DNA-binding domain more often retain transcriptional activity on a variety of promoters (8, 9).
In breast cancer, more than 20 studies have analyzed the prognostic or predictive value of TP53 mutation (10). In 18 of these studies, TP53 mutation was clearly associated with poor prognosis, mutations at residues involved in DNA contacts being of worse prognosis in several of them. However, it is not clear from these studies whether TP53 is a factor of prognosis that is independent of other clinicopathologic factors. Also, there is no clear consensus on the specific type of mutations carrying a worse prognosis because different classifications of mutations have been used and because the comparison of individual mutations was limited by a lack of statistical power.
Using a more powerful analysis in order to assess whether the identification of TP53 mutation presents a real benefit over currently available factors of prognosis (such as tumor size, node status, and estrogen and progesterone receptor contents), we collected and pooled clinical and molecular data from 1,794 European patients with breast cancer who were followed-up for at least 10 years and whose tumors were screened for somatic TP53 gene mutation.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Tumor material and TP53 mutation screening. The samples analyzed were either biopsies or surgery specimens, either fresh-frozen or paraffin-embedded. Histopathology, grading, and hormone receptor contents were determined independently for each cohort at their respective institutions. TP53 mutation screening was done on genomic DNA, except for 311 patients from the Swedish cohort, in which RNA was analyzed. Constant denaturing gel electrophoresis/denaturing gradient gel electrophoresis/temporal temperature gradient electrophoresis, single-strand conformational polymorphism, or denaturing high-pressure liquid chromatography prescreening methods were used to detect mutations and sequencing was done to precisely identify the mutation in all cohorts except for the Swedish cohort, in which no prescreening was done (direct sequencing of cDNA was applied). The entire coding sequence of TP53 gene (exons 2-11) was screened in 651 tumors, whereas only exons 5 to 8 were analyzed in 1,124 samples, and exons 5 to 11 in 19 samples. Details of methods and PCR primers have been described previously (13–15, 17).
TP53 mutation classifications. Mutations in exons 5 to 8 (including introns) were classified according to their position, nature, and suspected effect on protein structure and activity (18, 19). The following groups were defined:
, p53AIP1, and Noxa genes. Functional groups were defined as follows: 1, active or partially active on all promoters; 2, inactive on one to two promoters; 3, inactive on three to five promoters; 4, inactive on six to seven promoters. For individual promoters, three groups were considered: 1, inactive; 2, partially active; 3, activity similar to wild-type. Statistical analysis. Statistics were done using SPSS and Minitab software. To avoid a possible selection bias, patients with missing data for clinical prognostic variables were included by creating a category labeled as "missing" for each variable. Patient follow-ups were computed as the time interval between surgery date and the date of last follow-up, or as the time interval between the date of surgery and the date of death. To reduce heterogeneity among hospitals for duration of follow-up (three hospitals, accounting for <30% of the patients, had <10 years of maximum follow-up), follow-up was censored at 120 months (10 years – censoring time). Date of follow-up was set at 10 years after surgery for patients whose survival exceeded 10 years. Death from breast cancer within 10 years after surgery was considered as the primary outcome variable. Due to censoring, all breast cancer deaths after 10 years of follow-up were not taken into consideration. Patients who died from causes other than breast cancer were censored at the time of their death when in the 10-year follow-up period or at 10 years if survival exceeded 10 years.
Mortality rates were computed with a censoring at 10 years and using the cumulated number of person years (PY) in each category as a denominator. Kaplan-Meier survival curves and hazard rates estimated by a Cox proportional hazard model were computed to quantify the effect of TP53 mutation on breast cancer–specific mortality after adjustment for known clinical cofactors (tumor size, node status, ER and PR contents, and age at diagnosis). Histopathologic subtypes and grading were also considered in a descriptive analysis but were not used as predictors of survival because of missing data and possible differences in the classification systems used by participating hospitals.
All Cox models were stratified by hospital using a different baseline hazard function for each hospital to adjust for differences between centers that may be attributable to differences in tumor classification methods, hormone receptor measurements, treatment regimen (surgery and/or chemotherapy), or disease severity (some centers may be more likely than others to admit patients with more severe disease). Patients were grouped according to the mutation present in their tumor as described above. Proportionality assumption for the Cox model was verified by looking at parallelism of survival curves in Kaplan-Meier analysis. To identify possible interactions with TP53, Kaplan Meier curves were also stratified according to the presence of a TP53 mutation.
Because the detection of TP53 mutation outside exons 5 to 8 was done in a limited number of cases, we focused on mutations occurring in exons 5 to 8 and two distinct analyses were done. In the first analysis, all patients were considered and patients with a mutation outside exons 5 to 8 were included in the "no-mutation" group. Multivariate analysis was restricted to patients who had complete data for the variables that remained associated in the final Cox model. Several sensitivity analyses were previewed to validate the results obtained in the final Cox model under the presence of missing data for some clinical cofactors or when using a different definition of the outcome measure:
In the second analysis, only patients with mutations in exons 5 to 8 were included to compare the prognostic value of specific TP53 mutations. Mortality rates and Kaplan-Meier curves and log rank test were computed.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
|
Prognostic value of TP53 gene status. In univariate analysis, known prognostic factors of survival such as tumor size, histopathologic subtype, grading, node status, and hormone receptor status were all associated with patient survival (Table 2). The effect of tumor histopathologic subtype was mainly due to tubular and medullar types that were associated with lower and higher rates of death, respectively. Large tumor size, high histopathologic grade, presence of nodes, and absence of hormone receptors were associated with high mortality rates. Patients with a TP53 mutation had a relative risk of breast cancer–specific death of 
2 over a period of 10 years following surgery compared with patients with no mutation (Table 2).
|
|
|
Because data on tumor grade was lacking for too many cases to be included in the multivariate analysis, a separate analysis on a subset of patients with information on grading was carried out. Patients predicted to have a favorable outcome were selected (tumor grade <3, tumor size <5 cm, no node invasion, positive ER or PR receptor). Ten-year survival analysis of these patients stratified by TP53 mutation status showed that patients with a TP53 mutation had a large and significant reduction in survival compared with patients without mutation (close to 60% at 10 years; Fig. 1B).
Prognostic values of specific TP53 mutations. Mutations within exons 5 to 8 were classified in different groups according to the effect or position of the mutation in the primary or tertiary sequence of the protein (see Materials and Methods). Kaplan-Meier survival analysis of patients grouped according to the type of TP53 mutation found in their tumor showed that non–missense mutations (any mutation other than missense) and missense mutations in the DNA-bindings motifs (DBM) were associated with a strong reduction in survival compared with patients without mutations, whereas missense mutations outside the DNA-bindings motifs (non-DBM) were associated with an intermediate reduction of survival (Fig. 2A). The 10-year mortality rates for non-DBM and DBM mutations were, respectively, 43.92 and 73.42 (per 1,000 persons, P = 0.0897; Table 4). If the non-DBL missense mutations were grouped with silent mutations (associated with similar mortality rate) and used as a reference group, the relative risk associated with DBM mutations was 1.8 (1.03-3.18) and the one of non–missense mutations was 2.9 (1.44-5.38). These results remained valid after adjustment for tumor size, node status, and hormone receptor status. Analysis of mortality rates for the most frequent (hotspots) missense mutations in this series identified mutations with higher or lower mortality rates compared with other missense mutations (Table 4). Figure 2B shows that missense mutations at codon 179 and the R248W mutation were associated with reduced survival, whereas the G245S and Y220C mutations were associated with better survival compared with any other missense mutation. Of note, other mutation hotspots, which are general hotspots for all breast cancers (R175H, R248Q, R273H/C, codons 163, 249, and 282), were associated with mortality rates similar to those of non–hotspot missense mutations (these mutations were included in the "other" category in Fig. 2).
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
There was a linear relationship between the size of the tumor and the frequency of TP53 mutations, and a strong association between the presence of a TP53 mutation and high grade, positive node status, and loss of hormone receptors (ER and PR), which is in agreement with previous reports (http://www-p53.iarc.fr/Somatic.html). TP53 mutations have also been found to be associated with increased global genomic instability (21–23) and with markers of increased cell proliferation such as high mitotic frequency, high expression of Ki-67, and high cyclin E expression (24, 25). These results show that TP53 mutations are generally associated with an advanced and aggressive tumor phenotype.
TP53 mutations were significantly more frequent in young women and in medullar carcinoma as reported by others (25–27). Both early age at onset and medullar subtype may be indicative of inherited cancer due to BRCA1 germ line mutations (28, 29). BRCA1 breast cancers may represent up to 5% of the cases in a breast cancer series and these cancers usually present a high frequency of TP53 somatic mutations (19). Because family history is not well documented in our series and <5% of the patients have been tested for BRCA1, we cannot exclude that some BRCA1 cases may contribute to the observed mutation frequency. To rule out any confounder effect of possible BRCA1 cases in subsequent analyses, results were systematically verified on subsets excluding patients <40 years and medullar cases.
The prognostic value of TP53 mutation has been shown to be independent of tumor size, node status, or ER in a number of reports (13, 15, 25, 30, 31). The present study confirms these observations. In addition, we found an interaction between TP53 mutation and PR content that has not been previously reported. TP53 mutation combined with low PR had a very bad prognosis independently of tumor size, node status, and ER status. PR status, which reflects estrogen pathway integrity, has been shown to be more relevant than ER status for tumor response to tamoxifen and prediction of patient survival (32–34). In two independent retrospective series of patients (14, 35), one of them being included in our series (35), TP53 mutation has been shown to affect tumor response to tamoxifen. These results suggest that TP53 pathway may play a role in the response to antihormone therapy. However, the lack of information on treatment for a significant number of patients prevented us from exploring this hypothesis. Further studies are thus required to explore the possible interplay between TP53 and ER pathways and the consequences on tumor development and behavior.
When comparing the prognostic value of different types of mutations, we found that the more severe mutations were non–missense mutations followed by missense mutations in the DBMs (L2/L3 and LSH), among which mutations at codon 179 and the R248W mutant were associated with the highest mortality rates. Grouping of missense mutations according to their loss of transcriptional activities measured by systematic yeast-based assays, or to their predicted effect on protein structure, did not correlate with patient outcome. Although these structural and functional analyses of p53 mutations are the most extensive that are currently available, they may not give an accurate assessment of the changes induced by mutation that really have an effect on clinical outcome. Indeed, other variables such as protein-protein interactions, transcriptional repression, and transactivation of other genes, not taken into account here, play an important role in the antiproliferative activity of p53 and in the activities of mutant proteins (36, 37). Mutations affecting the L2/L3 motif involved in specific DNA-binding and zinc coordination have been repeatedly described as "bad" mutations in breast cancer based on their association with poor tumor response to treatment and short patient survival (4, 26, 30, 38–41). Functional assessments of some of these mutations in human cells have shown not only loss of transcriptional activity and defects in the capacity to induce cell cycle arrest or apoptosis, but also gain of function properties and/or dominant-negative effects, resulting in growth-promoting activities and resistance to drug-induced apoptosis (http://www-p53.iarc.fr/p53MUTfunction.html). How these properties specifically affect tumor response to treatment and patient outcome is still under debate. Because deletions/insertions mutations are expected to result in a null phenotype (truncated and unfolded proteins), our results suggest that loss of transcriptional activity is the main determinant of the poor prognostic value of TP53 mutations in breast cancer. Moreover, loss of transcriptional activity may be sufficient to promote breast tumor development, as suggested by studies on germ line mutations. First, non–missense mutations and missense mutations in the DBMs were associated with an earlier age at onset of breast tumors compared with missense mutations outside these motifs (32 versus 42 years; ref. 42). Second, TP53 null mammary epithelium isolated from TP53 null mice and transplanted into cleared mammary fat pads of TP53 wild-type mice showed that the absence of TP53 is sufficient to cause the development of primary tumors (43). If loss of transcriptional activity seems to be the main determinant of breast tumor development and behavior, it cannot be excluded that some specific mutants, such as codon 179 and R248W might exert dominant-negative effects and gain of function activities responsible for their very bad prognostic value.
Some limitations, mainly due to the multicenter structure of the study, have to be acknowledged. First, intercenter heterogeneity could be mostly controlled through stratification (44), as mentioned in Materials and Methods. In addition, differences of follow-up between hospitals were homogenized by censoring of follow-up. Second, TP53 mutation detection was done in five different laboratories with four different prescreening methods that may differ in their sensitivities. Thus, hazard risk estimates for TP53 mutation may have been underestimated. Third, the presence of missing values in known predictors of survival prevented the inclusion of some variables in multivariate analysis. It was particularly true for histopathologic grade and subtype with, respectively, 48% and 17% of missing values. However, histopathologic subtype and grade may reflect the presence of molecular alterations including TP53 mutation and hormone receptor expression and thus may be collinear with TP53 when entered in the multivariate model. The sensitivity analyses confirmed results when missing values for variables such as tumor size and nodal status were integrated in the analysis, and when censoring was made at the 5-year follow-up, or when using a more extended definition survival (overall mortality). These sensitivity analyses show the stability of the results. Nonetheless, the lack of information on family history, adjuvant treatments, and other markers such as ERBB2 amplification, prevented us from investigating the influence of these variables on our results. In a recent study, Bull et al. have found that TP53 gene mutations were more frequent than ERBB2 amplification in women with node-negative breast cancer and that TP53 mutation may be beneficial in identifying women at higher risk of disease recurrence and death when their tumor has ERBB2 amplification (45). In two other studies, the use of expression microarrays has shown that TP53 gene mutation is highly associated with groups of patients with similar gene expression profiles (46), and that tumor classification based on these profiles was a stronger predictor of outcome than any of the classical clinicopathologic markers, TP53 mutation status being equally significant.14 These results strongly support the fact that TP53 mutations have a prognostic value in various groups of patients. However, further studies will be required to precisely identify which groups of patients would benefit or not from TP53 mutation screening.
In the early 1990s, the rapid accumulation of data on TP53 mutations in human cancer raised high expectations for clinical exploitation. However, most studies relied on immunohistochemistry to assess TP53 status, a method prone to misclassification, as many TP53 mutations do not correlate with protein accumulation. In the present study, the strongest association with poor survival was found for non–missense mutations, predicted to generate a negative immunostaining. It is thus highly recommended to perform gene sequencing to precisely identify the mutation. Several common polymorphisms in TP53 (in exons 3, 4, and 6) may also deserve further investigation. Although a recent study argues against a role for these polymorphisms in breast cancer susceptibility (47), there is experimental evidence that codon 72 polymorphism (arginine to proline) may influence wild-type p53 activity in response to cytotoxic drugs (48). Mutation detection in our series was done by PCR-based prescreening methods followed by DNA sequencing, which are the gold standards for gene mutation identification, but they are labor-intensive, and thus, are not suitable for clinical practice. New techniques have been developed recently that may be more easily implemented for routine use, such as microarray-based methods (Affymetrix, arrayed primer extension assay; ref. 49). As these techniques will soon be available at an effective cost, TP53 gene mutation may become an important marker for patient management.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
14 A. Langerød and A-L. Børresen-Dale, unpublished data. 
Received 5/10/05; revised 11/ 4/05; accepted 12/ 8/05.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. X. Real p53: It Has It All, But Will It Make It to the Clinic As a Marker in Bladder Cancer? J. Clin. Oncol., December 1, 2007; 25(34): 5341 - 5344. [Full Text] [PDF]
|
|
|
|
|
|
B. George, R. H. Datar, L. Wu, J. Cai, N. Patten, S. J. Beil, S. Groshen, J. Stein, D. Skinner, P. A. Jones, et al. p53 Gene and Protein Status: The Role of p53 Alterations in Predicting Outcome in Patients With Bladder Cancer J. Clin. Oncol., December 1, 2007; 25(34): 5352 - 5358. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Harris, H. Fritsche, R. Mennel, L. Norton, P. Ravdin, S. Taube, M. R. Somerfield, D. F. Hayes, and R. C. Bast Jr American Society of Clinical Oncology 2007 Update of Recommendations for the Use of Tumor Markers in Breast Cancer J. Clin. Oncol., November 20, 2007; 25(33): 5287 - 5312. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 U. Cardellino, Y. Ciribilli, V. Andreotti, P. Modesto, P. Menichini, G. Fronza, C. Pellegrino, and A. Inga Transcriptional properties of feline p53 and its tumour-associated mutants: a yeast-based approach Mutagenesis, November 1, 2007; 22(6): 417 - 423. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. R. Grossman, S. Lyle, M. B. Resnick, E. Sabo, R. T. Lis, E. Rosinha, Q. Liu, C.-C. Hsieh, G. Bhat, A. R. Frackelton Jr., et al. p66 Shc Tumor Levels Show a Strong Prognostic Correlation with Disease Outcome in Stage IIA Colon Cancer Clin. Cancer Res., October 1, 2007; 13(19): 5798 - 5804. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Lonning, S Knappskog, V Staalesen, R Chrisanthar, and J. Lillehaug Breast cancer prognostication and prediction in the postgenomic era Ann. Onc., August 1, 2007; 18(8): 1293 - 1306. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. R. Dearth, H. Qian, T. Wang, T. E. Baroni, J. Zeng, S. W. Chen, S. Y. Yi, and R. K. Brachmann Inactive full-length p53 mutants lacking dominant wild-type p53 inhibition highlight loss of heterozygosity as an important aspect of p53 status in human cancers Carcinogenesis, February 1, 2007; 28(2): 289 - 298. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. A. Klein Random mutations, selected mutations: A PIN opens the door to new genetic landscapes PNAS, November 28, 2006; 103(48): 18033 - 18034. [Full Text] [PDF]
|
|
|
|
|
|
A. C. Joerger, H. C. Ang, and A. R. Fersht From the Cover: Structural basis for understanding oncogenic p53 mutations and designing rescue drugs PNAS, October 10, 2006; 103(41): 15056 - 15061. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
