Inverse Relationship between Human Papillomavirus-16 Infection and Disruptive p53 Gene Mutations in Squamous Cell Carcinoma of the Head and Neck

The oncogenic virus human papillomavirus 16 (HPV16) is detected in a subset of squamous cell carcinomas of the head and neck (HNSCC). These HPV-positive HNSCCs typically arise in the oropharynx, are less commonly associated with tobacco or alcohol exposure, show enhanced sensitivity to radiation therapy, and are consistently associated with favorable patient outcomes compared with non–HPV-related HNSCCs (1–4). This emerging clinical profile of HPV16-positive HNSCC likely reflects a pattern of molecular genetic alterations that is distinct from its HPV-negative counterpart.

The p53 tumor suppressor gene plays a critical role in regulating key cellular pathways including those involving apoptosis and cell cycle control, and it is a frequent target of inactivation during the development of HNSCC (5). Abrogation of p53 function can be mediated by a variety of mechanisms. Mutations and loss of heterozygosity directly target the p53 gene, whereas expression of the HPV oncoprotein E6 binds and degrades wild-type p53 protein product (6–8). Unlike cervical carcinomas where HPV infection and p53 mutations are mutually exclusive events, HPV infection and p53 mutations sometimes occur together in HNSCC (1, 9–13). The apparent superfluous presence of HPV16 in p53-mutated HNSCCs has raised the suspicion that HPV is incidental, not causal, in the development of these carcinomas.

Not all inactivating events are equivalent in their ability to knock out p53 function. Depending on where they occur, assorted p53 mutations are highly divergent in their effect on p53 protein structure, stability, and DNA binding properties. Those that occur within the core domain affecting p53 protein interaction with sequence-specific DNA completely block DNA binding and entirely abrogate p53 function (14). Those that occur outside of these DNA contact points may affect p53 function in a more limited manner. Furthermore, E6 degradation of p53 protein is not functionally equivalent to a p53 mutation. Even in HPV-infected cells expressing E6 oncoprotein, endogenous wild-type p53 can activate some cellular target genes (15), and the apoptotic response to radiation remains intact (16). These observations would seemingly account for the dual presence of inactivating p53 events, particularly when the effect of any single event on p53 function is incomplete. The purpose of this study was to determine the relationship between HPV16 and the types of p53 mutations in HNSCC.

Patients. The study was nested within a prospective multicenter study (Eastern Cooperative Oncology Group E4393 and Radiation Therapy Oncology Group R9614) of patients with HNSCC treated surgically between 1996 and 2002 with curative intent. Formalin-fixed and paraffin-embedded tumor samples were obtained from a subset of those HNSCCs (i.e., those resected at the Johns Hopkins Hospital), and these tumors were used to construct a tissue microarray. A second highly selected group of tumors consisted of tonsillar carcinomas that were known to carry disruptive p53 mutations. These selected carcinomas were pooled from other study sites.

p53 mutation analysis.p53 analysis had been done on tumor samples that were rapidly frozen at −80°C. Tumor purity was assessed from microscopic analysis of the frozen tumor block. Only samples with at least 70% tumor cells were eligible for p53 analysis. In some cases, samples with a low concentration of tumor cells were microdissected to obtain enriched samples. Mutation status of exons 2 to 11 of the p53 gene was evaluated using the GeneChip p53 assay (Affymetrix) as previously described (17). All mutations detected by GeneChip p53 assay analysis were identified and confirmed by automatic (ABI BigDye cycle sequencing kit) or direct dideoxynucleotide sequencing (17). Based on available information about the functional differences of various p53 mutations, p53 mutations were grouped as “disruptive” and “nondisruptive.” Disruptive mutations were defined as stop mutations, frameshift mutations, or nonconservative mutations occurring within the key DNA binding domain L2/L3. All other mutations were defined as nondisruptive mutations. DNA from blood lymphocytes was also evaluated for p53 status to help discern true allelic differences between tumor DNA and germ line DNA.

HPV16 in situ hybridization. HPV16 detection was done using the in situ hybridization catalyzed signal amplification method for biotinylated probes (DAKO GenPoint). This catalyzed signal amplification system permits visualization of single copies of HPV16 in infected cells (18). Briefly, 5-μm tissue sections underwent deparaffinization, heat-induced target retrieval in citrate buffer, and digestion with Proteinase K (Roche Diagnostics). Slides were subsequently hybridized with a biotinylated HPV16 type–specific probe (DAKO). Signal amplification was done by consecutive application of streptavidin-horseradish peroxidase complex, biotinyl tyramide, and streptavidin-horseradish peroxidase complex. Visualization of positive hybridization signals was done by incubation with the chromogenic substrate diaminobenzidine. Cases were considered positive if hybridization signals visualized as nuclear dots were present within tumor nuclei (Fig. 1). HPV16-positive and HPV16-negative tonsillar carcinomas served as positive and negative controls, respectively. In these control samples, HPV16 status had been rigorously determined using a combination of HPV detection techniques including consensus L1 PCR, E7 type-specific PCR, and Southern blot hybridization of unamplified tumor DNA (1).

Statistical evaluation. The associations between HPV16 status and p53 mutational status were evaluated by use of the Fisher exact test. P values are two sided unless otherwise specified. Statistical analysis was conducted using STATA software, version 7 (STATA).

The tissue microarray contained tumors from 89 patients with squamous cell carcinomas from various anatomic sites of the head and neck including the oral cavity (n = 38), larynx (n = 20), palatine/lingual tonsils (n = 21), hypopharynx (n = 6), and palate (n = 4; Table 1). Of the 89 tumors, 43 (48%) harbored a p53 mutation. Tumors arising from the tonsils were less likely to harbor a p53 mutation than carcinomas arising from nontonsillar sites, but the difference was not statistically significant (38% versus 54%; P = 0.19). For cases with p53 mutations, disruptive mutations were less likely to occur in carcinomas arising from tonsillar than nontonsillar sites, but the difference was not statistically significant (13% versus 38%; P = 0.24).

HPV16 was detected in 12 of 89 (13%) HNSCCs. When stratified by site of origin, HPV16 was detected in 12 of 21 (57%) HNSCCs of the lingual/palatine tonsils, but in none of 68 carcinomas from nontonsillar sites (55% versus 0%; P < 0.00001; Table 1). HPV16-positive tumors were less likely to harbor p53 mutations than HPV16-negative tumors (25% versus 52%; P = 0.12). Disruptive p53 mutations were not identified in any of the HPV16-positive cancers (Table 2). To determine whether the inverse correlation between HPV16 and disruptive p53 mutations persisted in a larger group of tonsillar carcinomas, tonsillar carcinomas known to harbor disruptive p53 mutations were pooled from multiple study sites. Although HPV16 was present in 12 of 20 (60%) tonsillar carcinomas that did not harbor disruptive p53 mutations, HPV16 was not detected in the 7 tonsillar carcinomas with disruptive p53 mutations (P = 0.008).

The presence of transcriptionally active HPV genome in the nuclei of clonally expanded transformed cells has helped establish an etiologic link between HPV16 and a subset of HNSCC, but the occurrence of p53 mutations in a subset of HPV-positive HNSCCs has prompted a resurgent skepticism over the strength of this link. Mutation of the p53 gene is accepted as an important mechanism of p53 pathway inactivation, and its presence in a HPV-positive HNSCC would seemingly diminish the role of HPV in tumor development. Most studies of HNSCC have noted a reduced but persistent prevalence of p53 mutations in HPV-positive tumors with the dual presence of HPV DNA and p53 mutations ranging from 0% to 42% (1, 9–13). We noted the paradoxical overlap of these p53 pathway–inactivating events in 25% of the HPV16-positive HNSCCs.

A detailed understanding of p53 protein structure indicates that variant p53 mutations are not equivalent in their capacity to disrupt p53 function. Nonconservative mutations located within the L2/L3 region—the region of the p53 protein directly involved in DNA contact—have limited consequences on protein structure but profoundly disrupt binding to DNA (14). Mutations in the β sandwich of the core domain, on the other hand, generally lead to an extended denaturation of the p53 protein. The effect of these variant mutations goes well beyond differences in protein structure. In contrast to mutations that do not disrupt DNA binding, disruptive mutations both (a) selectively eliminate the wild-type p53 allele (19) and (b) transactivate cellular proliferation genes by mechanisms independent of direct DNA binding (20, 21), and thereby confer “gain-of-function” characteristics as characterized by enhanced tumor growth and increased resistance to antitumor therapies (20–22). For HNSCCs with disruptive p53 mutations, acquisition of a malignant phenotype seems to be complete. In these tumors, the coexistence of other events targeting the p53 pathway is neither required nor anticipated. Indeed, we did not identify HPV16 in any of the HNSCCs with disruptive p53 mutations, including those carcinomas most likely to be associated with HPV16 infection (i.e., tonsillar carcinomas). For HNSCCs with nondisruptive p53 mutations, the additive effect of other inactivating events such as transcriptionally active HPV could heighten the overall effect of a p53 functional abrogation. Of the HNSCCs that showed the overlapping presence of HPV16 and p53 mutations, the p53 mutations were all of the nondisruptive type.

Retrospective case-series have consistently shown that patients with HPV-positive HNSCCs have an improved prognosis when compared with patients with HPV-negative tumors (1, 23–25). Several hypotheses have been offered to account for this difference. The immune system likely plays a pivotal role in modulating the behavior of HPV-related cancer of the head and neck, and strategies designed to up-regulate the immune response hold promise for further improving patient outcomes. The effects of “field cancerization” are diminished in patients with HPV-positive HNSCCs. Whereas chronic exposure to chemical carcinogens induce genetic alterations that tend to be distributed across large tracts of epithelium (26–28), HPV16 specifically targets the tonsillar epithelium resulting in a more restricted distribution of HPV-positive cancers. In effect, patients with HPV16-positive HNSCCs are less likely to succumb to synchronous and metachronous tumors (24, 29). Like other studies, we did not detect the presence of HPV16 in any cancers arising outside of the oropharynx (1, 30, 31).

HPV16 infection does not track with certain genetic alterations that characterize most HNSCCs, and the segregation of genetic profiles as a function of HPV16 status may also contribute to differences in clinical outcomes. For example, Braakhuis et al. (32) observed that HPV16-positive tumors are not associated with widespread allelic loss that characterizes most HNSCCs. Our results indicate that the presence of HPV16 occurs at the exclusion of disruptive p53 mutations. Importantly, disruptive p53 mutations greatly affect survival outcomes. In breast cancer, p53 mutations that reside within DNA binding sites are associated with treatment resistance and early cancer relapse compared with those p53 mutations that reside outside of these DNA binding sites (33). In HNSCC, disruptive p53 mutations likewise have been correlated with accelerated tumor progression and reduced therapeutic responsiveness (19). We do not report clinical follow-up in our small nested study population, but in the complete cohort of study patients with HNSCC, disruptive p53 mutations were found to be a significant predictor of survival independent of stage and treatment (34).

HPV16 infection and p53 mutation may coexist in a subset of HNSCC, but there is an inverse correlation between HPV16 and the types of p53 mutations that most severely disrupt DNA binding. A less calamitous genetic profile such as the absence of disruptive p53 gene mutations may underlie the emerging clinical profile of HPV16-positive HNSCC including improved response to therapy and patient outcome.