Department of Otorhinolaryngology and Head and Neck Surgery, 1st Faculty of Medicine, Charles University in Prague, Motol University Hospital, Prague, Czech Republic
Department of Otolaryngology and Head and Neck Surgery, 1st Faculty of Medicine, Charles University in Prague, Motol University Hospital, Institute for Postgraduate Medical Education, V Úvalu 84, 150 06 Prague 5, Czech Republic
The association of high-risk human papillomaviruses (HR HPVs) with tonsillar cancer (TC) has been documented. Because patients with HPV-associated tumors show better survival rates, modification of their treatment regimen is being considered. It is therefore crucial to find markers for the identification of patients whose tumors are linked to viral infection. A cohort of 109 patients with primary TC was screened for HPV DNA presence in the tumor tissues and HPV-specific antibodies in sera. Data regarding risk factors and clinical parameters were collected. Forty-five specimens were analyzed for the expression of viral E6 and E2-region mRNA, and the p16 and p53 protein expression status was assessed by immunohistochemistry. The overall prevalence of HPV DNA in TC tissues was 65.1%. Ninety-three percent of HR HPV DNA-positive samples expressed E6*I mRNA. E2-region mRNA expression was detected in 36% of positive samples, which implies that the virus is integrated in 64% of HPV DNA/RNA-positive tumors. p16 overexpression and the presence of antibodies specific to HPV16 E6/E7 oncoproteins correlated well with HPV DNA and RNA presence. The disease-specific survival rate of patients with HPV DNA-positive tumors was significantly higher than that of HPV DNA-negative patients. In addition to providing further evidence of the involvement of HPV infection in the etiopathogenesis of a proportion of TC cases, our study demonstrates that p16 immunostaining and anti-E6/E7 antibodies as surrogate markers of HPV involvement represent specific, sensitive and clinically accessible assays for the identification of TC patients who have a considerably better prognosis.
Recent studies have shown that 20–25% of head and neck squamous cell cancers (HNSCCs) appear to be causally linked with high-risk human papillomavirus (HR HPV) infection.1–3 The highest (50–100%) prevalence of HPV DNA is consistently detected in tonsillar cancer (TC).4–6
TC is the most common cancer of the oropharyngeal region in Europe and in the United States, and its incidence, especially in younger age groups, is increasing steadily despite the fact that the well-established risk factors for these tumors—cigarette smoking and alcohol consumption—are relatively stable or decreasing in many countries.4, 7
HR HPV infection has been shown to be the major cause of cervical cancer. Molecular biological studies on cervical cancer have indicated that the essential role in HR HPV-induced malignant transformation is played by two viral oncoproteins, E6 and E7, which interact with cellular proteins and cause disruption of cellular growth control pathways.8, 9 In most cervical cancers, the E6-E7 gene expression is dysregulated because the circular, extrachromosomal HPV genomes became integrated in the cellular genome. The common expression pattern of all integrated HPV genomes analyzed so far consists of E6-E7 gene transcripts spliced into cell sequences and terminated at cellular polyadenylation sites.10 This eliminates an AU-rich mRNA degradation signal at the 3′ end of native HPV early gene mRNAs11 as well as the expression of the viral E2 or E8ˆE2 gene products that have the potential to downregulate E6-E7 transcription (reviewed in Ref.12). Jointly, these changes appear to confer a strong, selective growth advantage upon the cell expressing HR HPV E6-E7 from the integrated form of the virus. However, only limited information is available about the patterns of HR HPV expression in HNSCC.13–16
The HR HPV E6-E7 expression status most likely is critical to HNSCC evaluation. Patients with HPV DNA-positive tumors appear to have a significant prognostic advantage compared to their counterparts with HPV-negative tumors.17–20 However, recent studies have suggested that this prognostic advantage is strictly linked to patients with HNSCC tumors in which the HR HPV genomes are transcriptionally active.21 Therefore, viral involvement in the tumor etiology only may be relevant in those lesions in which the HR HPV E6-E7 oncogenes continue to be expressed.20, 22
To assess the presence of transcriptionally active virus in patients with HNSCC tumors, a variety of direct and indirect markers have been studied.13, 21, 23 Recently, high sensitivity and feasibility have been reported for polymerase chain reaction (PCR) detection of HPV DNA in tumor tissue followed by p16 immunostaining on formalin-fixed, paraffin-embedded (FFPE) tumor specimens.21 Similarly, the absence of detectable p53 protein expression has also been evaluated as a possible indirect marker of the presence of transcriptionally active virus. However, the specificity and sensitivity for identifying patients with HPV-positive tumors by the absence of p53 were low.23, 24
Because of the prognostic advantage of patients with tumors etiologically linked to HPV, the use of less aggressive treatment regimens in such patients is currently being evaluated. This necessitates the development of sensitive, specific and clinically relevant markers for the detection of the eligible patients with HR HPV TC. As no single marker has been shown superior to others, the search for clinically more relevant markers needs to be continued.
The primary aim of our study was to select the best marker or combination of markers accessible in clinical practice for the identification of patients with tonsillar tumors etiologically linked to HR HPV infection and to correlate the markers with prognosis. Another aim was to determine the type-specific prevalence of HR HPV in a homogenous group of primary TCs and to find clinical and epidemiological differences between patients with HPV-linked and HPV-unlinked tumors.
DSS: disease-specific survival; ELISA: enzyme-linked immunosorbent assay; FFPE: formalin-fixed, paraffin-embedded; HNSCC: head and neck squamous cell cancer; (HR) HPV: high-risk human papillomavirus; IHC: immunohistochemistry; NPV: negative predictive value; PCR: polymerase chain reaction; PPV: positive predictive value; TC: tonsillar cancer
Material and Methods
Patients treated in the Department of Otolaryngology and Head and Neck Surgery in 2001–2007 with primary squamous cell TC (ICD-10: C090–C099) who agreed to sign the informed consent form were enrolled. The study received official institutional and ethical approval from the participating institutions. Data on demographics, risk factors for oral cavity and oropharyngeal cancer and risks related to HPV exposure were collected by a questionnaire. The medical and pathology reports were completed for each patient. Altogether 109 patients were included in the study.
All but three patients underwent surgery, and cancer tissue taken during the procedure was sent on dry ice to the pathology department. In three patients who did not receive surgery, tumor biopsies were performed. The pathologist obtained two side-by-side sections of the tumor from the primary site. One of the paired sections from each anatomical location was then labeled, snap frozen in liquid nitrogen and stored for future analysis. The other of the paired sections from each anatomical site was fixed in 10% neutral formalin and paraffin embedded. From each paraffin block, the first and last sections were histologically analyzed to confirm that the sections in between—assigned for the detection of viral nucleic acids and immunohistochemical (IHC) analysis—contained at least 10% of tumor cells in the entire volume of the sample.
Total nucleic acids were extracted from two 20-μm sections. In 62 samples, only DNA extraction was performed by a proteinase K method as described previously.17 Remaining 47 samples had suitable material to perform simultaneous extraction of DNA and RNA by means of the Ambion RecoverAll™ Total Nucleic Acid Isolation Kit for FFPE Tissues (Applied Bioscience, Austin, TX). Care was taken to avoid sample cross contamination.
As an internal control, a 110-bp fragment of the human beta-globin gene was amplified as described previously to check the integrity of DNA and that no PCR inhibitors were present.17
HPV DNA detection was performed by PCR with primers specific for the L1 region (GP5+/GP6+) as described previously,25 and HPV typing was done by reverse line blot hybridization with probes specific for 37 types as specified in detail by van den Brule et al.26
cDNA was prepared by reverse transcription of 1 μg of RNA initially treated with RNase Free DNase (Promega, WI), followed by reverse transcription with M-MLV reverse transcriptase (Promega, WI). The absence of contaminating DNA was confirmed by absence of amplification of the internal control gene DNA, GAPDH,27 on RNA treated with DNase.
The integrity of mRNA for cDNA amplification was tested by the inclusion of the beta-globin gene target as a control as specified above. PCR specific for HPV 16E6*I mRNA oncoprotein was performed on cDNA with primers that amplify an 86-nt fragment.21 For the HPV16 E2-region, specific PCR primers that amplify a 184-nt fragment were used.28
IHC examination was performed on 2-μm sections of FFPE tissue with the following antibodies: p16INK4a (Purified Mouse Anti-Human p16, Clone G175-405, BD Pharmingen™, dilution 1:100) and p53 (Monoclonal Mouse Anti-Human p53 Protein, Clone DO-7, Dako, dilution 1:50). As a positive internal control for focal nuclear staining for p53, basal keratinocytes of the normal tonsillar or oropharyngeal squamous epithelium were used. Positive nuclear staining in invasive breast cancer in patients with TP53 gene mutation was used as a positive control for p53. Dysplastic cervical epithelial cells caused by HPV infection were used as the positive control for p16. The intensity of staining (graded + to +++) and the proportion of cells stained (scored in percentages) were evaluated. For p16 immunostaining, the location of the signal (cytoplasmic and/or nuclear) was also specified. A semiquantitative evaluation was performed. The sample was considered positive for p53 protein expression when more than 10% of cells were stained, whereas the samples positive for p16 expression had to show more than 50% of positive cells and reveal nuclear and/or cytoplasmic staining.29
Five milliliters of blood was collected from each subject enrolled. Serum was separated by centrifugation at 1600 rpm for 10 min, aliquoted and stored at −20°C.
The presence of IgG antibodies to antigens derived from HPV-specific proteins was tested as described elsewhere.30 In brief, the enzyme-linked immunosorbent assay (ELISA) was used. The virus-specific antigens were L1 or L1/L2 capsids (virus-like particles, VLP) prepared in a recombinant baculovirus system mimicking HPV16 and HPV16 E6 and E7 oncoproteins overexpressed in bacteria as fusion proteins with glutathione S-transferase at their N-terminus and the C-terminal undecapeptide of the SV40 large T-antigen at their C-terminus. Both systems have been described recently.31
Total cumulative tobacco exposure and alcohol exposure were assessed as specified before,30 and patients were divided in groups according to their smoking and alcohol status as defined elsewhere.17 The cutoff level for smoking was calculated from the average number of cigarettes smoked in the group of HPV-negative cases (33 cigarettes per day) and that for alcohol from the average number of drinks consumed per week in the group of HPV-negative cases (15 drinks per week).
For the comparison of the smoking habits in patients with HPV-positive and HPV-negative tumors, three different models were used. The first compared the numbers of smokers and nonsmokers, the second considered three categories: smokers, ex-smokers and never smokers and the third one divided the subjects into nonsmokers, weak smokers (smokers who smoked less than the cutoff level, e.g., ≤33 cigarettes per day) and heavy smokers (smokers who smoked more than the cutoff level, e.g., >33 cigarettes per day). Continuous variables between groups were compared by t-test. The chi-square test and, in some instances, the Fisher's exact test were used to compare categorical variables between groups. Odds ratios (ORs) with 95% confidence intervals (CI) and two-tailed p values were calculated in 2 × 2 tables using the EPI INFO statistical package (2002) and GraphPad InStat software (GraphPad Software, San Diego, CA). All tests were two sided, and the significance level was α = 0.05. Correlation between variables and HPV positivity was assessed using logistic multivariate regression analysis with a forward stepwise procedure (p value ≥0.1). Survival was measured in days from the date of diagnosis to the date of death or to the date the patient was last known to be alive. Patients who died of causes other than TC were considered censored observations in the disease-specific survival (DSS) analyses. Time-to-event measures were analyzed by the Kaplan–Meier method, log-rank test and Cox multivariate regression. Ninety-five percent CIs for ORs were based on normal approximation. Variables considered in the Cox regression models were presence of HR HPV, age, gender, pack-years, average number of drinks per week, tumor grade, tumor size (pT, according to the TNM Classification of the UICC, 1997), incidence and extent of lymph node metastases (pN, according to the TNM Classification of the UICC 1997). A forward stepwise procedure was performed to find significant covariates (p ≥ 0.1). The analyses were performed using the statistical program SPSS (SPSS, Chicago, IL).
In our study, 83% of the cases were males and 17% females. Out of 109 patients, 65% (71/109) had HR HPV-positive and 35% (38/109) HR HPV-negative tumor tissues. Demographic characteristics are summarized in Table 1. There was no statistically significant difference between HPV-positive and HPV-negative patients in the mean age (56 vs. 58 years, p = 0.309), gender (adjusted p = 0.248) or years of education (≤12 vs. >12 years) (adjusted p = 0.925). Neither was there a significant difference in the factors related to sexual behavior such as the lifetime number of sexual partners (≤6 vs. >6) (adjusted p = 0.489) and engaging in oral-genital sex (yes vs. no) (adjusted p = 0.834) and oral-anal sex (adjusted p = 0.785). However, patients with HPV-positive and HPV-negative tumors differed significantly in smoking habits, regardless of the model of comparison used (for more detail, see Material and Methods). Patients with HPV-positive compared to HPV-negative tumors were more often nonsmokers (28 vs. 0%, adjusted p = 0.001), less often smokers (25 vs. 68%, adjusted p = 0.001) and less often heavy smokers (44 vs. 87%, adjusted p < 0.001). On the other hand, no difference was observed between the two study groups in alcohol consumption.
Table 1. Demographic/Epidemiological characteristics of TC patients
Clinical pathological characteristics
In our cohort, HPV-positive tumors were larger, even though not statistically significantly. Patients with HPV-positive tumors presented more often with regional metastases at the time of diagnosis (84 vs. 69%, adjusted p = 0.053). The HPV-positive tumors were of higher stage (III and IV vs. I and II) (89 vs. 72%, adjusted p = 0.038). HPV-positive tumors tended to be undifferentiated, but the difference was not significant (G3 vs. G1,2) (43 vs. 26%, adjusted p = 0.429) (Table 1).
HPV DNA prevalence and type specificity
The overall prevalence of HR HPV DNA in tumor tissue was 65% (71/109 patients). HPV 16 was the most prevalent type detected in 94% of HPV-positive samples. Additionally, HPV types 26, 33, 52 and 58 were found in one sample each. Multiple infection (HPV 16 and 18) was detected in a single sample.
Prevalence of HPV-specific antibodies
The presence of antibodies specific for HPV16 VLP correlated fairly with the presence of HR HPV DNA in the tumor tissue (kappa value = 0.396). HPV16 VLP-specific antibodies were found in 65% of the HPV-positive cases, but only in 21% of the HPV-negative cases (adjusted p = 0.003) (Table 2). Antibodies to HPV16 E6 and E7 oncoproteins were present in 80 and 70% of HR HPV-positive patients, respectively. Of the HPV negatives, only one patient was positive for HPV 16 E6-specific antibodies and two patients showed positivity for HPV 16 E7-specific antibodies (adjusted p < 0.001 for both E6 and E7). Of 71 patients with HR HPV DNA-positive tumors, 84% had antibodies specific to at least one of the HPV 16 E6 and E7 antigens and 66% had antibodies specific to both antigens (adjusted p = 0.006 and p < 0.001, respectively). The detection of HPV 16 E6- and E7-specific antibodies correlated well with the presence of HR HPV DNA in the tissue (kappa value = 0.719 and 0.584).
Table 2. Tissue and serological markers in the subgroup of patients from which material for RNA extraction was available
Detection of markers of active viral infection
Forty-seven samples were available for RNA extraction and IHC analyses. For surrogate marker analysis, cases negative for HPV DNA and positive for HPV16 and HPV16 and 18 DNA were selected (samples positive for HPV 33, 52 and 58 were excluded).
When the demographic and clinical characteristics of this subgroup of cases were stratified according to HPV 16 E6 mRNA presence, no difference was found in gender, age and education. RNA-positive cases had more sexual partners (>6) in the lifetime (64 vs. 41%, adjusted p = 0.04). However, no difference in the parameters reflecting the route of transmission (oral-genital/oral-anal contact) was found. Similarly to the entire cohort, tumors positive for HPV 16 E6 mRNA tended to be diagnosed at a more advanced stage (III and IV vs. I and II, 30 vs. 11%, p = 0.017) than those without viral mRNA expression. The HPV E6 mRNA-positive tumors also were more often associated with regional metastases than RNA-negative tumors (N+ vs. N0, 81 vs. 56%, p = 0.022). However, the significance of both of these parameters was lost after adjustment. The only additional characteristic differing significantly between HPV E6 mRNA-positive and RNA-negative tumors was tumor grade. E6 mRNA-positive tumors tended to be undifferentiated compared to RNA-negative ones.
Table 2 shows the sensitivity and specificity of direct and indirect markers for the detection of active viral infection as assessed by the detection of HPV E6 mRNA in those patients whose samples were available for RNA analysis. For detailed list of results of indirect marker analyses, see Supporting Information Table 1.
All but two HPV DNA-positive cases (93%) were positive for the expression of viral HPV16 E6-specific mRNA. All but one HPV DNA-positive cases (97%) were negative for p53 expression, whereas 8 (50%) of 16 HPV DNA-negative cases expressed p53. The p16 protein was expressed in all but two HPV DNA-positive cases (93%). Similarly to the entire cohort, this subgroup showed a very good agreement between HPV DNA positivity and the presence of HPV16 E6- and E7-specific antibodies. Antibodies were present in all but two HPV DNA-positive cases (93%). Additionally, one HPV DNA-negative patient had antibodies specific to HPV16 E6 and E7.
The one case with HPV16 and 18 DNA-positive tumor tissue was one of two with no HPV 16 E6 mRNA expression. The results of the detection of other surrogate markers of active viral infection in this case suggested a possible involvement of HPV18 or nonviral etiology. In contrast, in the second case positive for HPV16 DNA with no expression of viral HPV16 E6 mRNA, other analyzed surrogate markers suggested active viral gene expression (see Supporting Information Table 1).
To assess the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for the tested markers, we divided the cases into HR HPV expression-positive (HPV DNA-positive with HPV 16 E6 mRNA expression) and -negative (DNA and RNA-negative or DNA-positive and RNA-negative) tumors. The optimal combination of sensitivity and specificity was found for p16 (96 and 94%, respectively), followed by the detection of HPV 16 E6- and/or E7-specific antibodies (96 and 89%, respectively). Comparable sensitivity but low specificity was observed for p53 detection (96 and 56%, respectively).
Only 9 (36%) of 25 samples expressing E6 mRNA were also positive for E2-region mRNA expression. Therefore, 16 of 25 (64%) of the HPV mRNA-expressing cases exhibited the viral mRNA expression pattern characteristic of viral integration.
The survival analyses have been performed on the entire cohort of patients (N = 109). Of the 109 patients, 77 (71%) were alive with no evidence of the disease 0.02–4.9 years after the treatment. The median follow-up period was 2.4 years. Of the 32 patients who died, 26 had a recurrence of TC, one died of a second primary cancer and five died of unrelated disease.
Using the Kaplan–Meier method and log-rank test, we observed a significantly better DSS in patients with HR HPV-positive tumors in comparison to HPV-negative tumors (mean survival time 4.4 vs. 3.3 years, p = 0.004) (Fig. 1a) (Table 3). Moreover, the survival benefit in patients with HPV-positive tumors retained its statistical significance regardless of which marker had been used to identify possible HPV involvement. Patients positive for antibodies against HPV16 E6 and/or HPV16 E7 oncoproteins also had a longer mean survival time (4.4 vs. 3.4 and 4.3 vs. 3.5, p = 0.004 and p = 0.033, respectively). Not surprisingly, patients with smaller tumors (T1 and T2) also showed a significantly better DSS than those with T3 and T4 (mean survival time 4.3 vs. 3.0, p = 0.003).
Table 3. Factors with impact on patients' survival
Additionally, current smokers in comparison to ex-smokers as well as heavy smokers in comparison to weak smokers had significantly worse survival (p = 0.026, p = 0.045, respectively). No effect of alcohol consumption on survival was observed. Further analysis of the influence of tumor HR HPV positivity and smoking on the survival revealed that both smokers and nonsmokers with HPV-positive tumors had a significantly better survival than HPV-negative cases (all of whom had been smokers in our study group). The differences remained significant when adjusted in the multivariate Cox regression analysis for additional factors—age, alcohol consumption, tumor size and incidence and extent of lymph node metastasis (adjusted p = 0.048, adjusted p = 0.013, respectively) (Fig. 1b).
In the multivariate Cox regression analysis, the improved survival of patients with HR HPV-positive tumors was also confirmed in the presence of other cofactors (adjusted p = 0.023)—age, alcohol and tobacco use, tumor size, tumor stage, tumor grade and incidence and extent of lymph node metastasis. We have also confirmed the importance of HR HPV positivity of the tumor for the survival of patients by the omnibus test, showing the worst prognosis for the model without tissue positivity for HR HPV as a variable. The only other prognostic factor found to be significant by the multivariate analysis was tumor size (adjusted p = 0.049) (Table 3).
In a survival analysis performed separately for the smaller RNA-positive cohort, a similar trend was observed as in the large cohort of 109 TC patients, but the differences were not statistically significant. The survival of patients according to surrogate marker status is visualized in Supporting Information Figure 1.
In this prospective study, we have analyzed a well-defined cohort of patients with primary TCs for the presence, expression and integration of HR HPVs as well as for several additional potential prognostic markers. Our results provide additional strong support for the hypothesis that patients with HR HPV-dependent tumors have a better prognosis and may merit a differentiated therapeutic approach. In addition, we have identified a potential set of biomarkers that, in combination, appear to be highly sensitive and specific for the identification of patients with tonsillar tumors causally linked to HPV infection and can be utilized in clinical practice.
The prevalence of HPV DNA in our cohort (65%), with HPV16 being the most prevalent type (94%), was in agreement with other studies.4, 6 Even though our method of HPV testing allows for the detection of a large number of different HPV types as well as of multiple HPV types in a single sample, we only identified one sample with two HPV types in our study (1%). This is less than up to 4% of oropharyngeal tumors with more than one HPV type listed in a meta-analysis report.32 HPV type 33 was detected in 1% of our samples, which is in accordance with the meta-analysis results.32 Other HPV types (26, 52 and 58), not detected in other studies, were found in a single sample each. In contrast to other papers,33 no LR HPV types were detected in our study.
The two groups of HPV-negative and HPV-associated TC patients in our study were similar in terms of demographics (age, gender, education, sexual behavior and alcohol intake). They differed significantly only in their tobacco use: much higher proportions of smokers and heavy smokers were found in the HPV-negative patient group. This observation is in agreement with our previous results17 as well as with other recent data18, 34, 35 and shows that smoking as the most important risk factor for HNSCC does not play a significant role in the etiology of HPV-positive tumors.
In contrast to other studies, the patients with HPV-positive and HPV-negative tumors did not differ in their reported sexual behavior (see Table 1).
Previous studies have suggested that the presence of HR HPV DNA in the tumor tissue may not indicate a continued involvement of the viral genes in TC. Viral E6-E7 mRNA levels have been shown to better reflect active HPV involvement in cervical lesions.36 We therefore analyzed both viral and surrogate markers of active viral expression in our study. Although only a subset of the cases had tumor tissues available for HPV mRNA testing, this set of cases was comparable with the entire cohort in demographic and clinical characteristics.
HPV16 E6*I mRNA, the major E6-E7 gene transcript detected in HPV-16-dependent cancers, was detected in 93% of samples positive for HPV16 DNA and in none of the HPV-negative tumors. Only a limited number of studies have analyzed the HPV oncogene expression in HNSCC to date. Similarly to our data, Lindquist et al. have reported E6 mRNA in 79% of HPV16 DNA-positive TC. They have also detected E7 mRNA in 94% of samples.22 Two other studies have revealed lower oncogene expression rates. Smeets et al. detected E6 mRNA expression in six (75%) of eight HPV16 DNA-positive oropharyngeal cancer samples.21 Jung et al. recently reported HPV oncogene expression in only 12 of 30 (40%) HR HPV DNA-positive oropharyngeal cancer samples.37 These discrepant results may be attributed to differences in the prevalence of HPV-containing tumors, geographical and socioeconomical differences as well as to possible technical differences in the adequacy of tumor RNA samples and the choices of the viral RNA regions, primers and methods used in viral mRNA detection.
As the first prospective study on TC, we further tested our E6 mRNA-positive tumor samples for the absence of HPV 16 E2-region-encompassing mRNA sequences as the gold-standard marker of HR HPV integration.11, 16 HR HPV integration in cervical cancer invariably results in the expression of the viral E6-E7 genes from mRNAs correctly initiated at the major viral early gene promoter just upstream of E6, but spliced across a truncated 3′ end of the HPV early region directly into cellular sequences. This eliminates the expression of the HR HPV E8ˆE2 repressor as well as the full-length E2 gene product that downregulate E6-E7 transcription12 and eliminates a 3′ AU-rich element in the downstream noncoding region of the viral early transcripts that destabilizes viral mRNAs10, 11 No E2-region mRNA was observed in 16 (64%) of 25 E6 mRNA expressing TC tumor samples, implying E6-E7 expression from integrated HPV genomes. This is a proportion comparable to those reported by Hafkamp et al. and Koskinen et al.14, 15 The E6-E7 mRNAs in the nine remaining TC samples that also express the 3′ E2-region are transcribed from extrachromosomal, unintegrated plasmid HPV genomes; although more detailed DNA analysis would be required to exclude the possibility that these lesions also harbor additional, integrated HPV forms.
We also assessed our TC tumor tissues for potential indirect markers of active viral oncogene expression, p16 and p53. p16 has been reported previously as a surrogate marker of active HPV presence in cervical tumors.36 In HNSCC, Smith et al.29 in a recent study reported 20% disagreement in the detection of HPV DNA and p16 in their samples (one marker positive, the other negative and vice versa) and suggested that p16 cannot serve as a reliable surrogate marker. In contrast, our results presented here, in agreement with others,38–40 indicate that the detection of p16 expression together with HPV DNA detection is a sensitive and specific approach to the identification of patients with HPV-associated TC.
We found that the sensitivity and specificity of HR HPV DNA detection in combination with p16 IHC detection were 100 and 88%, respectively. These rates concur with the results of others21 and indicate that the joint assessment of HPV DNA and p16 IHC could potentially replace HR HPV RNA E6 mRNA detection, which is labor intensive and costly.
We further assessed p53 protein expression by IHC in 45 study samples. All except 1 of the 27 HPV-positive tumors in our study lacked detectable levels of p53, whereas 50% of HPV-negative tumors showed p53 overexpression. However, the specificity and sensitivity of p53 as an “inverse” marker of HPV positivity were low. This can be partly explained by the fact reported by others that IHC does not precisely reflect TP53 gene mutational status.41, 42 Taken together, our data clearly imply that absence of p53 overexpression is not a reliable marker of active HPV expression in TC.
Serological markers used in our study showed high correlation with HPV DNA and RNA status in tumor tissues. Serological studies in cervical cancer patients have shown that although antibodies against HR HPV virions, detected by VLP antigens, reflect lifetime exposure to HPV infection, they do not provide information about the presence of the disease.43 In contrast, antibodies against E6 and E7 oncoproteins are mostly present in sera of patients with invasive HPV-associated tumors and are rarely present in healthy and asymptomatic individuals.44
In agreement with this concept, the entire 109 TC patient cohort showed high and statistically significant correlation between HPV DNA and/or RNA status, and seropositivity to E6/E7 oncoproteins. The presence of anti-E6/E7 antibodies also correlated with significantly better patient survival, whereas no correlation was observed for HPV-specific antibodies to VLPs. Similar data have been reported by others.45 Taken together, our results indicate that antibodies against HPV 16 E6 and E7 oncoproteins represent highly sensitive (96%) and specific (89%) markers of HPV-associated TC.
The multivariate analysis of the entire cohort showed that HPV positivity of the tumor is the strongest predictor of patient survival in TC. The survival benefit did not depend on the marker used for the determination of HPV positivity. All of the markers—both direct and indirect—correlated with better DSS, even though this correlation was not statistically significant in the smaller cohort, most likely because of the small sample size.
The only other factor influencing survival was tumor size as reported by others.14, 46 In our study, nodal status, traditionally used for treatment decision making, did not show any significant effect on survival. We25 and others14 have found similar results previously, whereas Hoffmann et al.47 have reported nodal status to influence survival irrespective of HPV status.
A survival benefit in patients with HPV-positive tumors has been suggested by a number of studies,13, 46, 48 further summarized in a meta-analysis.19 A variety of hypotheses have been proposed but the issue has not been clarified yet. It is important to note here that all but two39, 49 available studies on this subject have been retrospective, whereas ours represents a prospective study design.
With the aim to better define the role of HPV in relation to smoking, we compared the survival rates of patients stratified according to joint HPV status and tobacco use. The Kaplan–Meier survival curves showed a significant difference in survival between smokers by HPV status but not between HPV-dependent TC patients depending by tobacco use (Fig. 1b). This observation, in agreement with those of others,50 supports the role of HR HPV etiology as the strongest prognostic factor in head and neck cancer.
We observed two etiologically different groups of TCs. Those linked with HR HPV have a significant prognostic advantage. This fact should be considered in clinical decision making. Patients with HPV-positive TC, thus, may benefit from a different, possibly less aggressive, treatment regimen. This conclusion will need to be validated by further studies in order not to compromise existing, excellent treatment outcomes in patients with HPV-dependent tumors. We have shown that the combination of PCR detection of HPV DNA in the tumor tissue with p16 IHC and/or serological detection of anti-E6 and/or anti-E7 antibodies is highly sensitive and specific for the identification of patients with tonsillar tumors causally linked to HPV infection. These tests can be easily performed in clinical practice.
The authors thank Prof. Roman Kodet for expert help in analysis of histological specimens, Prof. Lubomír Turek for expert consult and language correction and Marie Neradová and Blanka Langová for excellent technical assistance.