Purpose: TP53 is a key gene in cellular homeostasis and is frequently mutated in head and neck squamous cell carcinoma (HNSCC). There is a variety of TP53 mutations, each with its own biological and clinical implication. Aim of the study was to assess the prognostic significance of TP53 mutations in HNSCCs and to identify the most relevant mutation.
Experimental Design:TP53 mutation status was investigated in 141 consecutive HNSCCs treated by surgery with radiotherapy when indicated and with a known human papilloma virus status. The type of mutation was correlated with overall and progression-free survival in a multivariate two-sided Cox regression analysis with wild type as reference.
Results: A TP53 mutation was found in 88 (62.4%) of the carcinomas and was not significantly associated with overall survival (HR = 1.65, P = 0.11). Patients with a mutation resulting in a truncated protein (n = 36, 25.5%) had a significantly worse overall survival (HR = 2.54, P = 0.008) and progression-free survival (HR = 2.65, P = 0.002). Four of these mutations were at a splice site, 13 were nonsense mutations (produces stop codon), and 19 were insertions or deletions resulting in a frameshift. After multivariate analysis, a truncating mutation remained a significant prognosticator. A missense (i.e., nontruncating) mutation did not influence prognosis. Other ways of classification (disruptive vs. nondisruptive, hotspot vs. nonhotspot, and DNA binding vs. non–DNA binding) were less discriminative.
Conclusion: In HNSCCs, a truncating TP53 mutation is associated with a poor prognosis. This patient group seems as a target population for adjuvant therapy with chemoradiation or viral vector–mediated TP53 gene transfer. Clin Cancer Res; 17(11); 3733–41. ©2011 AACR.
Disappointingly, survival of patients with head and neck squamous cell carcinoma (HNCC) has not markedly improved in recent decades. Improved tailoring of existing treatment modalities for the individual patient may improve survival. This study shows that a particular form of a TP53 mutation, namely a truncating mutation, is an independent risk factor for survival of HNSCC patients. We propose that the classification on the TP53 status should be added to the list of classifiers that determine the risk of death: human papilloma virus status, tumor stage, and nodal stage. There is an important message for the oncological community: Some HNSCCs are very aggressive and particularly these should be treated with postoperative chemoradiation.
Head and neck squamous cell carcinomas (HNSCC) contributes to approximately 5% of all cancers in males and 2% in females in the Western world (1). Well-known risk factors for HNSCCs are tobacco smoking (2) and excessive consumption of alcohol (3). An infection with human papilloma virus (HPV) has also been identified as an etiologic factor, in particular in oropharyngeal carcinomas (4–6). HNSCC with biologically active HPV represents a different molecular entity (7–10) and patients generally have a better prognosis (11, 12).
TP53 is a key gene involved in many cellular processes such as apoptosis and cell-cycle control (13). Its inactivation is considered one of the earliest and essential genetic alterations in HNSCC carcinogenesis (14, 15). During carcinogenesis, the gene becomes inactivated either by acquiring mutations or by the viral protein E6 in HPV-infected HNSCCs (4). The majority of TP53 mutations are missense mutations, nucleotide changes that cause a single amino acid alteration. Missense mutations reduce the thermodynamic stability of TP53 resulting in a significant loss of DNA binding and transactivation (16). The impact of the effect is determined by location and type of amino acid alteration. In addition, some missense mutations have been reported to display gain-of-function properties causing an oncogenic effect in carcinogenesis (17–19). Nonsense mutations change a codon into a stop codon and result in a truncated, usually nonfunctional protein. Also, insertions and deletions of one or more nucleotides that cause a frameshift and mutations at splice sites usually lead to a truncated protein (20).
Because of the important role of TP53 in cellular processes and carcinogenesis, a mutation in this gene has been extensively investigated as a marker of prognosis. Despite the large number of studies in HNSCC, no conclusive answer has been obtained about the prognostic value of a TP53 mutation (21–23). Interpretation of these studies is complicated by the small number of included patients due to the fact that HNSCCs from different sites were included, and that different mutation detection techniques have been used. Recent studies have indicated that categorization of mutations in different groups (e.g., hotspot and DNA-binding mutations) has enabled refined prognostication in HNSCCs and other types of cancer (19,24,25). Poeta and colleagues reported on HNSCCs that particularly so-called disruptive mutations seem to be associated with decreased overall survival (24). Another issue is that HPV-infected HNSCCs are generally TP53 wild type, creating a potential confounder as HPV-infected tumors have in general a favorable prognosis. The present study aimed to identify the most relevant classification of TP53 mutations with respect to patient outcome and to analyze the possible confounding effect of HPV.
Materials and Methods
Patients and tumor samples
We studied HNSCC samples from patients that were collected prospectively for molecular studies (26). Tumor biopsies were taken from surgical specimen after excision, snap-frozen in liquid nitrogen, and stored at −80°C till analysis. The study samples were consecutively collected from patients who were all treated with curative intent at the VU University Medical Center in Amsterdam in the period September 1997 through May 2000. Only patients with primary HNSCCs were enrolled in the study. Written informed consent was obtained from all patients and the study was approved by the Institutional Review Board. Patients tobacco and alcohol use was recorded as described by Hashibe and colleagues (27) and staging was done according to the classification of the International Union Against Cancer (5th edition UICC). In 105 patients, a neck dissection was carried out and pathologic staging of the neck was available. In the remaining 36 patients, a follow-up period of at least 2 years was taken to clinically stage the neck. The margin status was evaluated by histopathologic investigation and was divided into 3 groups: tumor-free margins, when the excised carcinoma was more than 5 mm from the surgical margin; involved margins, when carcinoma was present in the margin; and close margins in the remaining cases (28). Ninety patients were treated by postoperative radiotherapy. The indications were multiple nodal metastases, extranodal spread, and inadequate surgical margins.
Routine hematoxylin and eosin staining was carried out on a 5-μm cryosection to confirm the presence of carcinoma and to guide microdissection. Only carcinomas were included that had sufficient tumor material available for analysis. Neoplastic areas were microdissected on a further set of slides, DNA and RNA were extracted as previously described (29). To identify HNSCCs containing HPV, detection of viral DNA and expression analysis of the oncogenes E6 and E7 was conducted as previously reported (4). Presence of HPV DNA and E6/E7 mRNA expression was considered to reflect biologically relevant viral involvement.
All carcinomas were subsequently analyzed for mutations in the evolutionarily conserved regions of TP53, exons 5 to 9, with direct dideoxynucleotide sequencing as described previously (30, 31). When no mutations were found in the exons 5 to 9, the remaining coding exons 2, 3, 4, 10, and 11 were sequenced in addition.
Four types of mutation classifications were applied, based on either physicochemical or functional consequences as described in previous publications. The first comparison was made between nondisruptive and disruptive mutations. The criteria of Poeta and colleagues were adopted; a disruptive mutation was characterized by either a missense mutation inside the key DNA-binding domain leading to a change in charge or polarity of the altered amino acid or a mutation that result in a stop codon, splice defect, or frameshift in any region (24). In the present study, direct sequencing was used and this enabled the extra detection of insertions and deletions of multiple bases; these were all qualified as disruptive. All other missense mutations were considered nondisruptive.
Second, truncation and missense mutations were discriminated. Truncating mutations are a subgroup within the disruptive mutation group and include nonsense (mutation leading to a stop codon), frameshift, and splice-site mutations (32). Truncating mutations lead to nonfunctional protein and thereby to a loss of function about normally functioning p53. In contrast, missense mutations are characterized by an amino acid alteration and this can lead to a dominant-negative or a gain-of-function phenotype (16, 33).
In addition, the effect of other type of mutation classifications was established as hotspot mutations. On the basis of the IARC TP53 mutation database, 6 hotspot codons (175, 245, 248, 249, 273, and 282) were identified in multiple cancer types and it has been suggested that mutations at these hotspots select for more aggressive tumor growth (32). We chose to set a limit of more than 2% of all missense mutation for defining hotspot mutations. All other mutations were considered nonhotspot mutations.
The DNA-binding domain is the most conserved region of p53 and is known to frequently contain mutations (32). These mutations inactivate p53 by eliminating important DNA-binding contacts or altering structural stability of the core domain. Mutations in codons 102 to 292 were considered as DNA-binding mutations (16). Also, the effect of 2 functional classifications was studied. Within the group of missense mutations, a subdivision was made according to the so-called relative substitution frequency factor. This classification according to Grantham takes the change in composition, polarity, and molecular volume into account when an amino acid is replaced by another and is often used to estimate the possible pathogenic effect of germline mutations (32,34). Missense mutations were also divided according to their effect on the transactivating activity of the protein in model systems based on the results of Kato and colleagues, and nonfunctional, partially functional, and functional p53 proteins were distinguished (35).
As endpoints, overall survival and progression-free survival were calculated, the latter was defined as the time from treatment to local or regional recurrence, a second primary tumor in the respiratory or upper digestive tract, distant metastasis, or death (24). For definition of second primary tumors in HNSCCs, previously reported clinical criteria were used (30). Survival characteristics were assessed using wild type as reference, based on previous findings (24).
Uni- and multivariate Cox regression was used to investigate confounding and effect modifying influence of clinical parameters on the HRs of the mutation groups of interest. Effect modification was concluded when the HR of the interaction term of clinical parameter and mutation group indicator was found to be significant. In the absence of effect modification, the parameter ΔB was computed to be the relative difference of the ln(HR) of the determinant in the model without and with the clinical parameter, where ln(HR) is the logarithm of the HR. Confounding was concluded, when the absolute value of the relative difference ΔB exceeded the generally accepted threshold of 10%. Therefore, multivariate confounding was investigated with a variant of stepwise Cox regression, in which the stepping criterion was abs(ΔB), compared with its threshold. Thus, in each step, the covariate with the highest abs(ΔB) was entered in the regression, as long as the threshold was exceeded. This implicated that for each mutation category, the range of confounders and the order of entry was different. The analysis was carried out for overall and progression-free survival.
Survival curves were constructed with the Kaplan–Meier method. Frequency analysis was done with the χ2 test (Pearson) with Bonferroni correction for multiple comparisons.
Statistical tests were taken 2 sided and differences were considered significant at α = 0.05. For all statistical analyses, SPSS 15.0 for Windows (SPSS Inc.) was used.
Clinical factors, unclassified TP53 mutations, and prognosis
Patient and tumor characteristics of the 141 patients are shown in Table 1. The median overall follow-up time was 54.2 months and ranged from 2.8 to 120.1 months. Fifty patients had died having a median survival time of 18.9 months. The median survival time of patients alive at the end of follow-up was 63.4 months. Cancer recurred in 52 patients, divided into local recurrent disease in 15, regional recurrent disease in 7, distant metastases in 15, and second primary tumors in 15 patients. The relation between patient and tumor characteristics and overall and progression-free survival are listed in Table 1. Transcriptionally active HPV type 16 was found in 12 patients (8.5% of total) and its presence was associated with a seemingly better overall survival as indicated by an HR of 0.372 (Table 1; Supplementary Fig. S1A). However, this advantage in survival was not statistically significant (P = 0.172). No TP53 mutations were observed in HPV-positive HNSCCs.
Mutation details for each patient are shown in Supplementary Table S1 and the relationship between type of mutation and patient variables in Supplementary Table S2. A TP53 mutation was observed in the carcinomas of 88 patients (62%). The presence of a TP53 mutation tended to be associated with a worse overall survival but this was not significant (P = 0.113; Table 2 and Fig. 1A; the multivariate analysis is shown in Table 3). On the other hand, the association with progression-free survival was significant in the univariate (HR = 1.802, P = 0.028; Table 2) but not in the multivariate analysis (Table 3).
Disruptive and nondisruptive mutations
Patients with a disruptive mutation (n = 47) seemed to have a worse overall survival as compared with the patients with wild-type TP53 (HR = 1.995, P = 0.044; Table 2 and Fig. 1B) and this effect was more pronounced for progression-free survival (Table 2). Multivariate analysis for overall and progression-free survival showed somewhat lower HR values without statistical significance (Table 3). No effect modification could be traced.
Truncating and missense TP53 mutations
Fifty-two HNSCCs had a missense mutation, whereas 36 (25.5% of total) had a truncating mutation, namely 4 at a splice site, 13 nonsense mutations (produces stop codon), and 19 mutations resulting in a frameshift (Supplementary Table S1 for details). Patients with a truncating TP53 mutation had a significantly worse survival when compared with the wild-type group (HR = 2.545, P = 0.008; Table 2 and Fig. 1C). Stepwise multivariate analysis indicated that tumor stage (T-stage), HPV status, and age had a confounding effect, and after adjustment, a lower HR was obtained with a borderline significant P value (P = 0.044; Table 3). This indicates that a truncating mutation may be an independent prognostic variable. Similar effects were seen about progression-free survival (Tables2 and 3). The group of disruptive mutations consists of specific missense (n = 11) and truncating mutations (n = 36) and it can be appreciated that the decrease in survival in the disruptive group can be attributed to the group with a truncating mutation (Supplementary Fig. S1B). Patients with a missense mutation did not differ from the wild-type TP53 group about survival. The associations between the truncating mutation status and a patient or tumor variable are listed in the Supplementary Table S2. None of the features that have been studied were found to be significantly associated with a truncating mutation. No effect modification could be established.
Other types of mutation categories
A mutation at a hotspot (n = 20) had no prognostic significance (Tables2 and 3; Supplementary Fig. S1C). There was also no difference in overall and progression-free survival about DNA-binding mutations (n = 72; Tables2 and 3; Supplementary Fig. S1D). In addition, of the 52 HNSCCs with missense mutations, the type of mutation was scored according to the Grantham criteria (34). In our material, this score ranged from 10 to 194 with a median of 87.5. When distributing the tumors above and below the median score, there was no difference of overall or progression-free survival between either group and the wild-type group (data not shown).
The group with a missense mutation contained 1 carcinoma with the protein label functional and 1 with supertrans according to the publication of Kato and colleagues (35). Excluding these cases from the TP53-mutated group did not influence the survival characteristics (data not shown).
The most interesting finding of the present study is the relatively strong association between truncating mutations in TP53 and a poor overall and progression-free survival when compared with wild-type or missense mutations. Truncating mutations lead to nonfunctional protein and are caused by nonsense mutations (resulting in stop codons), splice-site mutations, and frameshift deletions and insertions. As we were interested to establish how independent this parameter might be, the analysis was adjusted for important clinical and pathologic factors including the HPV status. Nevertheless, after correction for these potential confounders, it seemed that a truncating mutation still had a significant association in relation to survival. Apparently, loss of function of p53 has greater biological consequences in HNSCCs than the effect of missense point mutations, which often result in dominant-negative or even to gain-of-function properties as postulated by some (20).
The question arises why a truncating mutation leads to a more aggressive carcinoma. The analysis of the association of the mutation type with clinical features provided no clear explanation in this respect. Although no significant association was observed, there was overrepresentation of more aggressive HNSCCs in the truncating mutation group, based on the notion of a higher T-stage, the presence of extranodal spread, and an involved margin status. With regard to a possible influence of therapy, we could establish that the poor survival of patients with this type of mutation seems not to be related to a radiation-resistant phenotype as well. When the HNSCCs were divided in 2 populations with and without radiation, no difference about outcome became apparent between the groups within these classifications (Supplementary Fig. S2). A biological explanation may be found in the effect that p53 and mutated p53 has on microRNA biogenesis. It has recently been revealed that particular missense mutations cause an overall decrease in microRNA biogenesis (36). At present, it is unknown whether this may affect specific microRNAs in squamous cell carcinomas and this requires further investigation
Based on our data, HNSCC patients with truncating TP53 mutations may benefit from more intense treatment. In this study, patients with such mutation were treated by surgery, followed in the majority of cases (25 of 36) by adjuvant radiotherapy based on staging and histologic examination of the specimen. It may be worthwhile to investigate whether the group of patients with a truncating TP53 mutation in their tumor would benefit from postoperative chemoradiation. A recent meta-analysis provided evidence that the addition of concomitant chemotherapy to locoregional treatment improved survival (37).
In addition, one might envision experimental approaches to correct p53 protein levels in the tumor cells. Because the majority of recurrences (9 of 16) in patients with a truncating p53 were locoregional, a promising option might be locoregional adjuvant treatment with adenoviral vectors that contain p53 expression cassettes (Advexin). Recent trials have shown success in patients with recurrent HNSCCs (38). Interestingly, tumors showing a TP53 mutation and lack of p53 expression by immunohistochemical staining had a higher chance of response when compared with tumors showing a mutation with expression of p53 (39). In the report of that study, no details on the type of TP53 mutation were described, but it is likely that the tumors without p53 expression are the ones that have a truncated TP53 mutation. It has been shown before that a truncating mutation leads to an undetectable protein (23).
When the mutations were classified in disruptive and nondisruptive, the group of tumors with a disruptive mutation had a worse overall and progression-free survival compared with patients with wild type TP53, confirming a previous report (24). A novel finding is that it is particularly the subgroup of truncating mutations within the disruptive group that seems to be responsible for the poor prognosis in HNSCCs. In our study, tumors with a disruptive missense mutation showed a similarly favorable prognosis when compared with the ones with nondisruptive mutations or with wild type TP53. In our series, the significance of a disruptive mutation was just lost in the multivariate model when we corrected for, among others, T-stage and HPV infection status. This effect was more pronounced for the overall than for the progression-free survival.
When analyzing the prognostic effect of a type of TP53 mutation, HPV status may play an important confounding role; this could also be interpreted in the present study from the multivariate analyses (Table 3). The explanation is that HPV-positive tumors are typically wild type TP53, as the p53 protein is inactivated by the viral oncoprotein E6 (5, 40) and that HPV-infected tumors have been shown to lead a better survival (5, 41). The univariate analysis indicates that infection with HPV may lead to a better overall survival, though there was only a not statistically significant trend, likely related to the small number of HPV16-infected tumors and outcome events in our cohort.
In conclusion, this study reports an important impact of functionally important mutations in TP53 on the survival in oral and oropharyngeal cancer. In particular, a truncating mutation was shown to have predictive value for a shorter survival. A study that investigates the effect of the addition of postoperative chemotherapy to radiotherapy in HNSCC patients with a truncating mutation in the tumor is warranted. In addition to the HPV status, a truncating TP53 mutation seems an important stratification factor in HNSCC prospective clinical trials.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
This work was supported by the Cancer Center Amsterdam Foundation.
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/).
- Received January 21, 2011.
- Revision received March 8, 2011.
- Accepted March 26, 2011.
- ©2011 American Association for Cancer Research.