Molecular Classification Identifies a Subset of Human Papillomavirus–Associated Oropharyngeal Cancers With Favorable Prognosis
the Departments of Medical Oncology, Otolaryngology, Therapeutic Radiology, Pathology, and Comparative Anatomy, Yale University School of Medicine, New Haven, CT
ABSTRACT
PURPOSE: We sought to determine the prevalence of biologically relevant human papillomavirus (HPV) in oropharyngeal squamous cell carcinoma (OSCC). Retinoblastoma (Rb) downregulation by HPV E7 results in p16 upregulation. We hypothesized that p16 overexpression in OSCC defines HPV-induced tumors with favorable prognosis.
METHODS: Using real-time polymerase chain reaction for HPV16, we determined HPV16 viral load in a cohort of 79 OSCCs annotated with long-term patient follow-up. A tissue microarray including these cases was also analyzed for p53, p16, and Rb utilizing in situ quantitative protein expression analysis. Seventy-seven tumors were classified into a three-class model on the basis of p16 expression and HPV-DNA presence: class I, HPV–, p16 low; class II, HPV+, p16 low; and class III, HPV+, p16 high.
RESULTS: Sixty-one percent of OSCCs were HPV16+; HPV status alone was of no prognostic value for local recurrence and was barely significant for survival times. Overall survival was improved in class III (79%) compared with the other two classes (20% and 18%; P = .0095). Disease-free survival for the same class was 75% versus 15% and 13% (P = .0025). The 5-year local recurrence was 14% in class III versus 45% and 74% (P = .03). Only patients in class III had significantly lower p53 and Rb expression (P = .017 and .001, respectively). Multivariable survival analysis confirmed the prognostic value of the three-class model.
CONCLUSION: Using this system for classification, we define the molecular profile of HPV+ OSCC with favorable prognosis, namely HPV+/p16 high (class III). This study defines a novel classification scheme that may have value for patient stratification for clinical trials testing HPV-targeted therapies.
INTRODUCTION
Numerous lines of molecular and epidemiologic evidence demonstrate that some of the oncogenic human papillomaviruses (HPVs) are etiologically related to a subset of oropharyngeal tumors.1,2 Individuals with HPV-associated anogenital cancers,3 post-transplant patients, HIV-infected men and patients with Fanconi anemia are prone to develop head and neck cancers.4
Since 1986, the first time that HPV16 DNA was detected in an invasive squamous cell carcinoma (SCC) of the head and neck (HNSCC) by Southern blot hybridization,5 HPV sequences have been detected repeatedly in a variable proportion of HNSCC, from less than 10% to up to 100%,6-8 apparently biased by the anatomic location of tumors and HPV detection techniques. Despite numerous reports confirming high frequency of HPV DNA presence in oropharyngeal cancers, the prevalence of biologically relevant HPVs in these tumors is unclear. HPV DNA detection in tumors per se does not indicate causal association.
Molecular studies in cervical cancer, the most widely accepted HPV-associated malignancy, indicate that HPV-driven malignant conversion is associated with certain molecular events. First, the viral DNA usually becomes integrated into the host genome. After integration, the expression of the main viral transcription/replication factor E2, which represses the transcription of E6 and E7 oncogenes, is disrupted. The disruption of E2 expression leads to dysregulated expression of the E6 and E7 oncogenes. The E6 and E7 genes of the oncogenic HPVs encode oncoproteins that bind and degrade p53 and retinoblastoma (Rb) tumor suppressors, respectively.9,10 Most cervical carcinomas harbor wild-type p53 and Rb tumor-suppressor genes. Thus, the tumor-suppressor pathways are intact but dormant in these cells because of the continuous expression of E6 and E7 genes.
Overexpression of p16 protein has been reported repeatedly in HPV-associated cancers. The p16 protein functions as a tumor suppressor by binding to the cyclin D1 CDK4/CDK6 complex, preventing phosphorylation of the Rb protein.11 In one study in cervical and genital lesions, levels of p16 protein expression were associated with HPV oncogenic potential.12 A recent study of lymph node metastases in HNSCC reported that p16 overexpression is a surrogate marker for oropharyngeal origin and HPV-association.13 To the contrary, in tobacco-related HNSCC loss of p16 protein expression is a common and early event.14-16 Thus, immunohistochemical analysis of p16 protein is a reliable method of detection of p16 alterations in HNSCC.
We previously studied p16 protein expression levels by immunohistochemistry (IHC) on a tissue array of oropharyngeal cancer cases annotated with long-term patient follow-up data.17 We found the prevalence of p16 overexpression in our cohort to be 20%; in addition, p16 overexpression was associated with favorable patient outcome. If indeed p16 overexpression is a surrogate marker of HPV association, then our prevalence of HPV-associated cancers would be inconsistent with most studies reporting a prevalence of 50% in oropharyngeal location. On the basis of these findings, we hypothesized that p16 overexpression defines a subgroup of HPV-positive oropharyngeal tumors with a more favorable prognosis. Thus, there may be a subgroup of HPV-positive p16 nonoverexpressing oropharyngeal tumors of multifactorial pathogenesis. In this model, tobacco may induce loss of p16 function and negate the favorable impact that the virus would have on prognosis. This hypothesis was also supported by the conflicting data in the literature regarding the impact of HPV on prognosis of patients with oropharyngeal cancer1,7,18-23; it is possible that the results of these studies are affected by the prevalence of the p16 overexpressors within the group of HPV-positive tumors.
In this study, we sought to determine the role of HPV in pathogenesis and clinical behavior of oropharyngeal squamous cell cancers. We studied a cohort of 107 oropharyngeal squamous cell cancers for HPV16 DNA viral load by real-time polymerase chain reaction (PCR). In addition, we constructed a tissue array composed of these tumors and studied expression of p53, Rb, and p16 proteins using a quantitative, in situ method of protein analysis. We hypothesized that for HPV DNA–positive cases, p16 expression status would differentiate the biologically relevant ones. Our results indeed delineate three biologically and clinically distinct types of oropharyngeal squamous cell cancers based on HPV-DNA determination and p16 expression status: one class of HPV-negative/p16-nonexpressing, 1 class of HPV-positive/p16-nonexpressing, and one class HPV-positive/p16-expressing oropharyngeal tumors. We were able to show that only the HPV-positive/p16-expressing tumors fit the cervical carcinogenesis model and that they are the ones associated with favorable prognosis.
METHODS
Tumor Specimens and Tissue Microarray Construction
After institutional review board approval, paraffin-embedded specimens from 107 patients were collected from Yale–New Haven Hospital (New Haven, CT) archives. Inclusion criteria were primary or recurrent histologically confirmed SCC of the oropharynx treated between 1980 and 1999 with primary external beam radiotherapy (RT) or gross total surgical resection and postoperative RT. Exclusion criteria included presentation with metastatic disease, recurrent disease when the primary tumor was available, or failure to receive full course of radiation therapy. Hematoxylin and eosin–stained (H&E) slides were reviewed by a pathologist (D.K.), evaluated for the presence of squamous cell carcinoma, and marked for 0.6-mm core extraction. Cores were placed on the recipient microarray block using a Tissue Microarrayer (Beecher Instrument, Silver Spring, MD). All tumors were represented with two-fold redundancy. Cores from HPV18-positive HeLa and HPV16-positive Caski cell lines fixed in formalin and embedded in paraffin were selected for controls and included in the array. The tissue microarray was then cut to yield 5-μm sections and placed on glass slides using an adhesive-tape transfer system (Instrumedics, Hackensack, NJ).
DNA Extraction
Subsequently, DNA was extracted from tumor samples. H&E slides of specimen blocks were reviewed by a pathologist (D.K.) and tumor areas were identified. Four tissue cores were extracted from these areas of SCC using sterile, blunt 18-gauge needles (VWR Scientific, West Chester, PA). Tissue cores were subsequently extracted via chelex-100/proteinase K protocol as previously described.24 Extracted DNA was aliquoted for storage at –20°C until used. DNA was extracted from cell lines (HeLa and Caski) using the same protocol, substituting a cell pellet from one 75-cm2 culture flask grown to confluence for the tissue cores.
Quantitative PCR
Quantitative PCR (qPCR) was carried out in a BioRad iCycler (Hercules, CA) 96-well thermal cycler equipped with fluorescence detection. Primer concentrations used are indicated in Table 1. Complete details of qPCR methods can be found in the Appendix (online only). As a confirmation step, a majority of samples were additionally subjected to real-time PCR analysis using a TaqMan double-dye based assay (Eurogentec Corp, Seraing, Belgium) as previously described. For this analysis, a SmartCycler (Cepheid, Sunnyvale, CA) qPCR instrument was used, and samples were run in duplicate. TaqMan primer/probe sequences and concentrations used are also indicated in Table 1.
Conventional IHC for p16 Expression Status
We previously determined p16 expression status of tumor samples in this cohort17 using a commercially available monoclonal p16 antibody kit (DAKO Cytomation, Carpinteria, CA). HeLa cell lines embedded in paraffin were used for positive control, and nonspecific mouse immunoglobulin G was substituted for p16 antibody as a negative control. In our previous investigation, we determined p16 expression to be essentially dichotomous (present or absent), therefore tumors were classified as p16 expressors (strong, diffuse staining) or p16 nonexpressors (weak or absent staining). These results were also confirmed separately by quantitative AQUA (HistoRx, New Haven, CT) immunofluorescence analysis for p16 using the same antibody (see Fluorescence IHC).
Tumor Classification
On the basis of the results of our previous study of p16 expression in oropharyngeal SCC, the existence of a distinct subgroup of p16 nonexpressing tumors with HPV DNA presence was hypothesized. Therefore, tumors were classified according to a 2 x 2 table by p16 expression and HPV16 DNA presence. Possible groups included HPV16 negative/p16 nonexpressing, HPV16 negative/p16 expressing, HPV16 positive/p16 nonexpressing, and HPV16 positive/p16 expressing (Fig 1).
Fluorescence IHC
The use of AQUA quantitative fluorescence IHC has been described previously25 and is also available in the Appendix (online only). Primary antibody to p16, p53 and Rb were used at 4°C overnight. The primary antibodies and concentrations used are summarized in Table 2.
Statistical Analysis
Comparison of HPV16 viral load and age by HPV/p16 classification (three-class model) was made by Kruskal-Wallis analysis followed by Wilcoxon rank sum test with Bonferroni corrections. Comparison of protein expression (AQUA score) by HPV/p16 classification was made using one-way analysis of variance (ANOVA) with post hoc comparisons by Dunnett's T3 test. Comparison of HPV/p16 (three class model) classification status with the clinical and pathologic variables sex, TNM stage, T stage, N stage, histologic grade, management (primary external beam RT [EBRT] v primary surgical excision plus RT), alcohol use, tobacco use and clinical response was made by Fisher's exact test. Self-reported alcohol use was classified as a dichotomous variable: social or none versus current drinker. Patients who reported no alcohol use for more than 2 years prior were included in the social/none category. Tobacco use was similarly classified as reported tobacco history of less than one pack per day or quit more than 10 years prior, versus current smoker of one or more packs per day. Clinical response was recorded as complete response versus no response/partial response after completion of treatment.
Disease-free survival (DFS), overall survival (OS), and local recurrence were assessed by Kaplan-Meier analysis with log-rank for determining statistical significance. Relative risk and prognostic independence was assessed by the multivariate Cox proportional hazards model including clinical and pathologic variables HPV/p16 classification, T stage, N stage (±), histologic grade, tumor type (primary v recurrent), and management (primary EBRT v surgery + RT). All survival analyses were performed at 5-year cutoffs.
Fisher's exact test comparisons were made using SAS version 8.2 (SAS Institute, Cary, NC). All other calculations and analyses were performed with the Statistical Package for the Social Sciences version 11.5 for Windows (SPSS Inc, Chicago, IL) and where appropriate were two tailed. Box/whisker plots were generated with Prism version 4.0 (GraphPad Software, San Diego CA).
RESULTS
Clinical and Pathologic Variables
Our study included 107 patients with histologically confirmed oralpharyngeal SCC (94 primary, 13 recurrent). Median follow-up time for survivors was 32.1 months, and 22.0 months for the entire cohort. Demographic and clinicopathologic variables for the cohort are summarized in Table 3.
HPV16 DNA Viral Load
Standard curves for -globin DNA demonstrated linear sensitivity (range, 4,000 to one human genome/μL). Standard curves for HPV16 E6 demonstrated linear sensitivity (range, 230,000 to 10 copies/μL). Melt curve (Tm) analysis was performed for all SYBR reactions (-globin and HPV16 E6) and demonstrated consistent Tms. Example melt analysis and standard curves for -globin and HPV16 E6 are shown in Figure 2.
Seventy-nine tumors had sufficient amplifiable -globin (>1 human genome/μL) by qPCR and were analyzed for HPV16 E6 DNA presence. Of the 79 tumors analyzed, 61% had evidence of significant HPV16 DNA presence using the criteria of at least one copy per 10 cell genomes' worth of DNA. Thirty-nine percent of samples were judged as negative for HPV16. The median copy number of HPV16 E6 DNA in HPV16-positive tumors was 14.1copies per human genome (range, 0.11 to 3,948 copies/cell genome).
p16 Protein Expression (conventional IHC and AQUA)
In our previous study,17 p16 expression status was determined in these tumors by conventional IHC. Tumors were classified as p16 expressors (strong and diffuse) or p16 nonexpressors (weak, absent, or focal staining). Analysis was restricted to the 79 tumors with amplifiable DNA. There were 78 (99%) of 79 tumors with sufficient tissue for determining p16 expression. Of these, 19 (24%) of 79 tumors were designated as p16 expressors, as were seven of seven cell line positive controls. These results were confirmed for this study using quantitative AQUA analysis of p16 expression. Of the patients included in this study, 53 (67%) of 79 had sufficient tissue for AQUA analysis of p16 expression. As visualized by conventional IHC (Fig 3), p16 displayed a diffuse cytoplasmic/nuclear pattern of expression that was essentially dichotomous (highly expressed v low/no expression). Results seen in the traditional IHC experiment were confirmed by AQUA analysis, with p16 nonexpressors having a median AQUA score of 3.4 versus 112.3 for p16 expressors (Fig 4).
Tumor Classification
Based on the results of our previous study of p16 expression in oropharyngeal SCC, the existence of a distinct subgroup of p16 nonexpressing tumors with HPV DNA presence was hypothesized. Therefore, tumors were classified according to a 2 x 2 table by p16 expression and HPV16 DNA presence (Fig 1). There were 78 tumors for which both HPV16 status and p16 expression status were available. Thirty tumors (39%) were class I (HPV16 negative/p16 nonexpressors). Twenty-nine tumors (37%) were class II (HPV16 positive/p16 nonexpressors), and 18 (23%) were class III (HPV16 positive/p16 expressors). There was one tumor with no amplifiable HPV16 E6 but p16 expression. Because this represented a classification group with only one member, it was excluded from further analysis, leaving a cohort of 77 tumors in three classes for subsequent analyses.
Comparison of HPV16 Viral Load by HPV/p16 Classification
Using the three-class model, there was a significant difference in HPV16 viral load between groups (P < .001). Post hoc comparison demonstrated that class II and class III tumors had significantly higher viral load than class I tumors (P = .003 each). The median viral load of class III tumors was higher than class II tumors; however, the difference was not significant at the .05 level (P = .078 after Bonferroni correction). The median viral loads (copies HPV16/human genome) were class I (0.0 copies), class II (3.6 copies), and class III (46.0 copies). These results are summarized graphically in Figure 5.
Quantitative AQUA Analysis of p53 and Rb Protein Expression
Of the 79 patients with amplifiable DNA, 57 (72%) had sufficient tissue for analysis of p53 protein expression by AQUA. Using the three-class model, there was a significant difference in p53 expression between groups (P = .026). Post hoc comparisons demonstrated that class III tumors had significantly lower p53 AQUA scores than class I and class II tumors (P = .017 and .015 respectively). For Rb, 65 (82%) of 79 had sufficient tissue for analysis of protein expression by AQUA. Comparison of Rb expression by HPV/p16 classification demonstrated a significant difference between groups (P = .001). Post hoc comparisons demonstrated that class III tumors had significantly lower AQUA scores than class I and class II tumors (P < .001 and .001 respectively). Results of fluorescence IHC and AQUA analysis are summarized in Figure 4 and Table 4.
Clinicopathologic Variables by HPV/p16 Classification (three-class model)
Tumors assigned to HPV/p16 class III (p16 expressors/HPV16 positive) were less likely to be recurrent versus primary tumor type, with 100% of class III tumors being primary tumor type (P = .028). Class III tumors were also associated with more advanced TNM stage (P = .022). This was a result of N stage contribution, as class III tumors were more likely to have lower T stage (P = .004) but higher N stage (P = .021). Class III tumors were also closely associated with a poorly differentiated histology, with 80% of class III tumors being poorly differentiated (P = .030). One hundred percent of patients with class III tumors experienced a clinical complete response to initial therapy, but this was not significant at the .05 level (P = .18). Patients with class I tumors (p16 nonexpressors/HPV16 negative) were statistically less likely to report low/no tobacco history, representing only 4% of patients with class I tumors (P = .006). Patients with class III tumors were less likely to report frequent alcohol consumption, although this was not significant at the .05 level (P = .06). Clinicopathologic variable correlations are summarized in Table 3.
Survival Analysis by Three-Class Model Only
When all HPV16-positive cases were compared with the HPV16-negative cases, HPV positivity was found to have no prognostic value for local recurrence (P = .66). For DFS and OS, HPV positivity did show some prognostic value. Patients with HPV-positive tumors had improved DFS (P = .05) and OS (P = .03) compared with the HPV-negative cohort. When the HPV16-positive cases were split by the three-class model, the predictive power was increased dramatically for local recurrence and DFS and OS.
For OS, patients with class III tumors had markedly improved OS compared with class I and class II tumors. Patients with class III tumors had a 5-year OS of 79% versus 20% for class I and 18% for class II (P = .0095). For DFS the same relationship was noted, with class III tumors demonstrating a 5-year DFS of 75% versus 15% for patients with class I and 13% for patients with class II tumors (P = .0025).
For 5-year local recurrence, patients with class III tumors had very low recurrence rates compared with both class I and class II tumors (P = .027). Class III patients had 5-year local recurrence rates of 14% versus 45% for class I and 74% for class II. Univariate survival analyses are summarized in Figure 6 and Table 5.
Multivariate Survival Analysis by Three-Class Model and Clinicopathologic Variables
Using the Cox proportional hazards model, we performed multivariable analysis to assess the independent predictive value of the three-class HPV/p16 classification model for local recurrence and overall and DFS.
For local recurrence, class III tumors were 4.8 times less likely to recur than class I tumors, although this was not significant at the .05 level (P = .054). For DFS, HPV/p16 classification was the most significant independent predictor. Class III patients were five times less likely to die or suffer recurrence as compared with class I patients (P = .005). For OS, HPV/p16 classification was the only significant independent predictor. Patients with class III tumors were 5.3 times less likely to die as a result of any cause than patients with class I tumors (P = .013). Complete results of multivariable survival analysis are presented in Table 6.
DISCUSSION
In this study, we identified a subset of patients with oropharyngeal SCC whose tumors display a unique molecular phenotype (ie, class III tumors). We believe this molecular phenotype defines HPV-positive tumors with favorable prognosis. These class III tumors represented 23% of the cohort, and were characterized by dramatically reduced local recurrence rates, as well as markedly improved 5-year DFS and OS.
Although the causal link between HPVs, especially type 16, and a subset of oropharyngeal cancers is becoming more firmly established,2,22 the prevalence of biologically relevant HPVs in these tumors has yet to be determined. Although HPVs can initiate carcinogenesis in epithelial cells,26 HPV infection per se is not sufficient to induce malignant transformation. The number of HNSCCs that test positive for HPV DNA presence far exceeds the number that actually express HPV oncoproteins. Only this latter group of tumors are likely to be HPV induced. In this study, by measuring protein levels of proteins of p53 and Rb pathways, we provide indirect evidence for E6 and E7 oncoprotein production, respectively, by a subset of HPV DNA–positive tumors. Prior studies have also shown that not all HPV DNA–positive HNSCC contain transcriptionally active HPV. In a recent study of 143 oral and oropharyngeal SCCs by Braakhuis et al,27 17% of tumors contained HPV16 DNA, but only 8% expressed E6/E7 mRNA. Another recently published case-control study investigated the relationship between HPV-DNA and tp53 mutations in oral cancer.28 The authors found that only tumors that had both HPV DNA and presence of antibodies to HPV E6 in patient serum displayed a reduction in tp53 mutations. Tumors with HPV DNA but no antibodies to HPV E6 had a similar mutation frequency to that of HPV-negative tumors. These studies included oral cavity tumors in addition to oropharyngeal tumors, thus the lower overall prevalence of HPV16 positive tumors in Brakhuis et al's article. This difference notwithstanding, the concept of only a subset of HPV-positive tumors being HPV induced is strongly supported by this evidence. In this study, we demonstrate an indirect method to detect transcriptionally active HPV in paraffin-embedded material. In addition, this study provides strong evidence that among HPV DNA–positive oropharyngeal tumors, only the transcriptionally active ones are biologically as well as clinically relevant.
In our study, we found that only a subset (class III) of HPV-DNA positive tumors match the expected molecular phenotype of HPV-induced malignancy, low p53 and Rb expression, and high p16 expression.
As previously mentioned, we classified oropharyngeal tumors into three distinct categories by determining HPV DNA presence and p16 status. HPV DNA positive/p16 high (class III) tumors displayed high HPV16 viral load, low p53 and low Rb protein levels. Tumors in this group were poorly differentiated and associated with lower alcohol and tobacco usage history. Despite the adverse histologic features, patients had lower local recurrence rate and improved DFS and OS. Several previous studies have found that HPV-associated head and neck cancers tend to share these features.2,18,29,30 To the contrary, HPV DNA positive/p16 nonexpressing (class II) tumors displayed a similar molecular profile to the HPV-negative tumors. These tumors had a median viral load of 3.6 viral copies per human genome, making them unlikely to represent false-positive results. Interestingly, patients in class II had a higher local recurrence rate than patients with HPV-negative (class I) tumors. A smaller previous study of p16 expression in tonsillar carcinomas found 2 (11%) of 18 HPV DNA–positive tumors lacked strong p16 expression.31 Both of these patients also suffered recurrences. It is possible that in class II tumors, HPV super-infection of a pre-existing neoplasm may be acting in an additive or synergistic fashion to the traditional carcinogenic insults of alcohol and tobacco, but this remains conjecture.
Our prevalence of HPV16 positivity of 61% is similar to other reports in oropharyngeal cancers.2,18,32,33 However, only 23% displayed the molecular profile that, on the basis of the cervical carcinogenesis model, defines HPV-causal association. Thus, the conflicting reports in the literature regarding the prognosis of HPV-associated head and neck cancers1,7,18-23 may be explained by the molecular heterogeneity of the HPV-positive group. Knowledge of the p16 protein status clarifies delineation of the proportion of HPV-induced oropharyngeal cancers.
We found that HPV DNA–positive/p16-expressing tumors (class III) had a low recurrence rate and markedly improved DFS and OS. Strikingly only 14% of class III tumors developed local recurrence whereas 45% of class I and 74% of class II recurred locally in 5 years. In addition the overwhelming majority of class III patients (approximately four of five) were alive in 5 years whereas only one of five of class I or II patients were alive in 5 years. The favorable outcome of HPV-induced oropharyngeal cancers may be due to the absence of field cancerization or enhanced radiation sensitivity2,34; all of the patients in this study received radiation therapy either as primary treatment modality or postoperatively. From the molecular biology standpoint, it is possible that the high p16 levels of HPV-induced oropharyngeal SCC account for the improved prognosis. Specifically, studies exploring the events of replicative and accelerated senescence in normal fibroblasts have shown that p16 is required for maintenance of growth arrest in senescent cells whereas Rb may not play an active role in the maintenance of senescent phenotype; although Rb levels decrease after the onset of senescence in normal fibroblasts,35 p16 becomes constitutively upregulated.35,36 Therefore, maintenance of senescence in tumor cells after treatment such as radiation may be dependent on p16 levels and that would explain the difference in outcome between p16 expressors and nonexpressors. However, a prospective randomized trial adjusting for other important prognostic variables is needed to definitively determine the impact of HPV association on patient prognosis.
The finding that high HPV16 viral load is associated with better prognosis has been reported previously.37 Other investigators have also shown that transcription of HPV16 E6/E7 mRNA in tonsillar carcinomas is not necessarily dependent on viral DNA integration and that the virus is predominately in episomal form. How the virus can remain in cancer tissues as episomes with high copy numbers is not fully understood. A recent report by Van Tine et al38 revealed that, in addition to its role as a transcriptional regulator, the HPV E2 protein may serve as an "anchor" to bind episomal HPV to cellular mitotic spindles, thus ensuring faithful maintenance of the episomal state.
On the basis of our results, we are proposing a model for HPV-associated oropharyngeal cancer (Fig 7). The classic oncogenic insult in HNSCC has been attributed to excess alcohol or tobacco exposure producing mutations or epigenetic inactivations of p53, p16, and Rb in a multistep progression from healthy cell to dysplasia to frank carcinoma. Alternatively, high-risk HPV may also lead to tumor formation, in a similar fashion to its role in cervical cancer. In these HPV-induced tumors, oncogenic HPV E6 and E7 proteins act to inactivate p53 and Rb pathways, with subsequent upregulation of p16 expression through loss of feedback inhibition. This obviates the need for mutational inactivation of these genes. In addition to these two classes of oropharyngeal tumors, we describe a novel class in which HPV16 is present but p53, p16, and Rb expression are similar to non–HPV-positive (class I) tumors. These class II tumors could be multifactorial in origin, possibly formed when tobacco-/alcohol-related tumors are infected by high-risk HPVs. These tumors may or may not represent a separate group biologically from HPV-negative tumors and need to be investigated further.
The single patient with a p16 expressing tumor that was HPV16 negative may represent presence of another, less common oncogenic subtype of HPV such as 18 or 33. Previous studies have demonstrated that although the vast majority of oncogenic subtypes present in HNSCC are type 16, other high-risk types such as 18 or 33 are present in 1% to 3% of oropharyngeal tumors.31
In this study, using AQUA technology, we were able to elucidate protein expression relationships using routine paraffin-embedded tissue that previously would have required fresh- or snap-frozen tissue resources on a large scale. As seen in Figure 4, the quantitative protein expression data generated from AQUA analysis allows much more powerful analysis than that provided by simple IHC, which relies on arbitrary cut points and qualitative scoring. However, it is important to note that for a protein with dichotomous expression patterns, such as p16, traditional IHC provides essentially the same information as AQUA analysis.
Our results have important diagnostic, prognostic, and clinical implications. IHC for p16 can be used as a simple test to determine whether biologically relevant HPV is present in tumors. HPV-targeted strategies such as therapeutic vaccines are currently undergoing clinical trials in head and neck cancer patients. The addition of p16 IHC to existing protocols for determination of HPV DNA presence would seem to be of vital importance for patient selection in these trials. We hypothesize that HPV DNA presence combined with low p16 expression indicates that these tumors do not harbor transcriptionally active HPV. In fact, p16 promoter methylation leading to abrogation of p16 function has been reported in cervical cancer cells that do not express viral E7 transcripts.39 The rationale for HPV-targeted therapies in these patients, in our view, is therefore lacking. In addition to potentially affecting clinical trial patient selection, our results have additional therapeutic implications. Patients with p16 expressing oropharyngeal tumors seem to have a favorable prognosis and may be spared the devastating sequelae of aggressive therapies. We also feel that current clinical trials in patients with oropharyngeal cancer should stratify by HPV DNA status/p16 expression, or at least include them as a prognostic variable. Although our evidence is compelling, these findings need to be confirmed in a prospective fashion.
Appendix
Several candidate primer pairs for each target gene (human -globin and human papillomavirus [HPV] E6) were identified using OligoPerfect software (Invitrogen Corp, Carlsbad, CA, http://www.invitrogen.com/content.cfmpageid=9716) and a review of the literature (Tucker RA, Unger ER, Holloway BP, et al: Real-time PCR-based fluorescent assay for quantitation of human papillomavirus types 6, 11, 16, and 18. Mol Diagn 6:39–47, 2001; and Peitsaro P, Johansson B, Syrjanen S: Integrated human papillomavirus type 16 is frequently found in cervical cancer precursors as demonstrated by a novel quantitative real-time PCR technique. J Clin Microbiol 40:886–91, 2002). Criteria specified for the OligoPerfect software included amplimer length of 80 to 120 base pairs (bp), primer length of 18 to 25 bp, primer melting threshold (Tm) between 58°C and 62°C, primer percentage guanine-cytosine between 30 and 60, optimized for 50 nM salt concentration. Candidate primers were synthesized by the Yale University Critical Technologies in Molecular Medicine core facility (New Haven, CT) and screened using an SYBR Green–based quantitative real-time polymerase chain reaction (qPCR) assay with known positive and negative controls for each gene target. Primers were selected based on sensitivity as evidenced by threshold cycle (Ct) qPCR analysis, and specificity as evidenced by Tm analysis and agarose gel electrophoresis visualization of reaction end products. Selected primers are indicated in Table 1 of this article. All primers utilized were optimized using the primer-chessboard method described by Gunson, Gillespie, and Carman (Optimisation of PCR reactions using primer chessboarding. J Clin Virol 26:369–373, 2003). Briefly, forward and reverse primers are combined across concentration gradients, with all other reaction conditions equal. This method allows selection of optimum forward and reverse primer concentrations, which are not always equimolar. qPCR was carried out in a BioRad iCycler (Hercules, CA) 96-well thermal cycler equipped with fluorescence detection. Final reaction conditions were 1x iQ SYBR Green Supermix (BioRad; 50 mM KCl; 20 mM Tris-HCl; 0.2 mM each deoxynucleotide triphosphate, 25 U/mL iTaq hot start DNA polymerase, 3 mM MgCl2, SYBR Green I dye, 10 nM floresein, and stabilizers). Primer concentrations used are indicated in Figure 1. Each reaction vial consisted of 20 μL master mix to which 5μL target DNA was added for a final reaction volume of 25 μL. All reactions (-globin and HPV16 E6) for a patient were each run in triplicate on the same 96-well plate. Each plate included a three-log serial dilution positive control for -globin and HPV16 E6, and three negative controls for each. The PCR reaction was preceded by 3 minutes at 95°C to activate the iTaq DNA polymerase. Reactions were run at 40 cycles of 95°C for 15 seconds followed by 60°C for 30 seconds. Fluorescence was detected during the combined annealing/extension phase of each cycle. The cycle threshold (Ct) was determined as the point at which each sample’s fluorescence crossed 30 units. Following 40 cycles, a melt curve analysis was performed by ramping the temperature from 60°C to 95°C at a rate of 0.1° per second, while monitoring resulting fluorescence. Following melt curve analysis, a final extension step at 60°C for 1 minute to reform double-stranded DNA was performed. As a confirmation step, a majority of samples were additionally subjected to PCR analysis using a TaqMan double-dye based assay as previously described. For this analysis, a SmartCycler (Cepheid, Sunnyvale, CA) qPCR instrument was used, and samples were run in duplicate. Primers and probes for the TaqMan assay were assembled by Eurogentec Corp (Seraing, Belgium). TaqMan primer/probe sequences and concentrations used are also indicated in Table 1. Reaction conditions for the TaqMan assay were 1x PCR supermix (UDG Platinum Supermix, Invitrogen Corporation) with 2 mM added MgCl2. Reactions were carried out with 23 μL master mix plus 2 μL target, for 25 μL final volume. Thermal parameters for -globin included a 2-minute incubation at 50°C to allow uracil-deglycosylase (UDG) activity and a 95°C Taq-polymerase activation step for 2 minutes followed by 45 cycles consisting of 95°C for 15 seconds, 61°C annealing step for 15 seconds plus optical analysis, and a 70°C extension step for 30 seconds. HPV16 E6 conditions were similar, with substitution of a combined extension/annealing step of 65°C for 30 seconds plus optics. Some patients had insufficient extracted DNA to run both assays (minimum required volume of 45 μL for the SYBR assay and 8μL for the TaqMan assay), in which case the TaqMan assay was used in triplicate with no SYBR assay.
PCR Data Analysis
Standard curves for both assays (TaqMan and SYBR) were developed for -globin using serial dilutions of purified high-molecular weight human DNA (Promega Corp, Madison, WI) from 4,000 genomes/μL to 1 genome/μL. Standard curves for HPV16 E6 were developed using a purified whole-length HPV16 plasmid at concentrations from 230,000 copies/μL to 10 copies/μL. This allowed determination of HPV16 E6 concentration and expression of the final quantity as the number of viral copies present per human genome (Tucker et al). Samples with -globin values less than 1 human genome/μL indicated lack of amplifiable DNA and were discarded from qPCR analysis. HPV viral load quantities derived that were greater than 0.1 (> 1 HPV genome copy/10 cells) were scored as positive for HPV16. This threshold was based on the assumption that the presence of less than 1 HPV genome copy/10 cellular genomes indicates that HPV is present in only a minority of the dominant clonal population. A viral load between 0.1 and 1 may represent stromal/nontumor DNA (not harboring HPV) being included in a tissue core and yielding an apparent viral load less than 1 HPV/cell. Previous investigators have utilized this cutoff as a viral load of 1 HPV/cell is considered sufficient to cause transformation (Hernandez-Avila M, Lazcano-Ponce EC, Berumen-Campos J, et al: Human papilloma virus 16–18 infection and cervical cancer in Mexico: a case-control study. Arch Med Res 28:265–71, 1997; Berumen J, Casas L, Segura E, et al: Genome amplification of human papillomavirus types 16 and 18 in cervical carcinomas is related to the retention of E1/E2 genes. Int J Cancer 56:640–5, 1994; Kutler DI, Wreesmann VB, Goberdhan A, et al: Human papillomavirus DNA and p53 polymorphisms in squamous cell carcinomas from Fanconi anemia patients. J Natl Cancer Inst 95:1718–21, 2003; and Ha PK, Pai SI, Westra WH, et al: Real-time quantitative PCR demonstrates low prevalence of human papillomavirus type 16 in premalignant and malignant lesions of the oral cavity. Clin Cancer Res 8:1203–9, 2002). Similarly, SiHa is an immortalized human cervical cancer cell line that contains one to two HPV 16 copies per cell.
Fluorescent Immunohistochemistry
In brief, slides were deparaffinized with xylene followed by ethanol. After rehydration in dH20, antigen retrieval was accomplished by pressure cooking in 0.1 M citrate buffer (pH, 6.0). Endogenous peroxidase activity was blocked by incubating in 0.3% hydrogen peroxide in methanol for 30 minutes. Nonspecific antibody binding was then blocked with 0.3% bovine serum albumin (BSA) for 30 minutes at room temperature. Following these steps, slides were incubated with primary antibody to p16, p53 or Rb at 4°C overnight. The primary antibodies and concentrations used are summarized in Table 2. Of note, the retinoblastoma antibody utilized detects both phosphorylated and unphosphorylated forms of retinoblastoma. Similarly, the p53 antibody utilized detects both mutant and wild-type p53. Subsequently, slides were incubated with goat antimouse secondary antibody conjugated to a horseradish peroxidase-decorated dextran polymer backbone (Envision; DAKO Corp, Carpinteria, CA) for 1 hour at room temperature. Tumor cells were distinguished from stroma by use of anticytokeratin antibody cocktail (rabbit antipancytokeratin antibody z0622; DAKO Corp) with subsequent goat antirabbit antibody conjugated to Alexa546 fluourophore (A11035; Molecular Probes, Eugene, Oregon). We added 4', 6-diamidino-2-phenylindole (DAPI) to visualize nuclei. Target (p53, Rb, p16) molecules were subsequently visualized with a fluorescent chromogen (Cy-5-tyramide; PerkinElmer Corp, Wellesley, MA). Slides were mounted with a polyvinyl alcohol-containing aqueous mounting media with antifade reagent (n-propyl gallate; Acros Organics, Geel, Belgium). Appropriate positive controls were included in the tissue microarray, including HeLa cell lines embedded in paraffin (for p16), colon cancer previously determined positive for p53 (for p53), and squamous cell carcinoma (for Rb). For negative control, the primary antibody step was eliminated.
Automated Quantitative Protein Expression Analysis (AQUA)
Automated image acquisition and analysis using automated quantitative protein expression analysis AQUA (HistoRx, New Haven, CT) has been described previously (Camp RL, Chung GG, Rimm DL: Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 8:1323–7, 2002). In brief, monochromatic, high-resolution images were obtained of each histospot after immunofluorescent staining as described herein. We distinguished areas of tumor from stromal elements by creating a mask from the cytokeratin signal. A tumor nuclei–specific compartment was created by using DAPI signal to identify nuclei, and the cytokeratin signal to define cytoplasm/membrane. The target signal (AQUA score) was scored on a normalized scale of 1 to 255 expressed as pixel intensity divided by the target area (tumor nuclei compartment). AQUA scores for duplicate tissue cores were averaged to obtain a mean AQUA score for each tumor. Histospots containing less than 10% tumor, as assessed by tumor mask area (automated), were deemed too small for accurate representation of tumor protein expression and were excluded from further analysis.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
AuthorsEmploymentLeadershipConsultantStockHonorariaResearch FundsTestimonyOther
Robert L. CampHistoRx (A)HistoRx (A)
David L. RimmHistoRx (A)HistoRx (A)
Dollar Amount Codes (A) < $10,000 (B) $10,000-99,999 (C) $100,000 (N/R) Not Required
Author Contributions
Conception and design: Paul M. Weinberger, Robert L. Camp, David L. Rimm, Amanda Psyrri
Financial support: Amanda Psyrri
Provision of study materials or patients: Bruce G. Haffty, Clarence Sasaki
Collection and assembly of data: Bruce G. Haffty
Data analysis and interpretation: Paul M. Weinberger, Ziwei Yu, Diane Kowalski, Malini Harigopal, Janet Brandsma, John Joe, Robert L. Camp, David L. Rimm, Amanda Psyrri
Manuscript writing: Paul M. Weinberger, David L. Rimm, Amanda Psyrri
Final approval of manuscript: Paul M. Weinberger, Ziwei Yu, Bruce G. Haffty, Diane Kowalski, Malini Harigopal, Janet Brandsma, Clarence Sasaki, John Joe, Robert L. Camp, David L. Rimm, Amanda Psyrri
GLOSSARY
AQUA (automated quantitative protein expression analysis): This technology overcomes limitations associated with traditional "brown stain" immunohistochemistry (IHC). In traditional IHC, protein expression is reported on a quantized scale such as 0, 1, 2, 3. Biologic material rarely is expressed in such neatly quantized packets, but rather is expressed on a continuous scale. AQUA is a method of computerized interpretation of fluorescence IHC images that allows protein expression to be automatically assigned to subcellular compartment and expressed on a continuous scale.
HPV (human papillomavirus) tumor classification: A proposed system of classifying oropharyngeal tumors based on both HPV16 presence (as indicated by detection of HPV16 DNA) and cellular p16 expression status (as determined by immunohistochemistry or automated quantitative protein expression analysis). Proposed groups include class I (HPV16 negative/p16 nonexpressing), class II (HPV16 positive/p16 nonexpressing), and class III (HPV16 positive/p16 expressing).
qPCR (quantitative polymerase chain reaction): Quantitative polymerase chain reaction (qPCR), also known as real-time PCR, consists of detecting PCR products as they accu-mulate. It can be applied to gene expression quantification by reverse transcription of RNA into cDNA, thus receiving the name of quantitative reverse transcriptase PCR (qRT-PCR). In spite of its name, quantitative, results are usually normalized to an endogenous reference. Current devices allow the simultaneous assessment of many RNA sequences.
Tissue microarray: Used to analyze the expression of genes of interest simultaneously in multiple tissue samples, tissue microarrays consist of hundreds of individual tissue samples placed on slides ranging from 2 to 3 mm in diameter. Using conventional histochemical and molecular detection techniques, tissue microarrays are powerful tools to evaluate the expression of genes of interest in tissue samples. In cancer research, tissue microarrays are used to analyze the frequency of a molecular alteration in different tumor type, to evaluate prognostic markers, and to test potential diagnostic markers.
Viral load: Using qPCR, a starting concentration of template (eg, human papillomavirus) DNA can be determined. If human DNA concentration in the same tumor sample is also determined from a cellular housekeeping gene (eg, -globin), then simple arithmetic can be used to determine viral load per human cell within the tumor sample.
Acknowledgment
We thank Daniel DiMaio, MD, PhD, and Edward Goodwin, PhD, for helpful discussions, and Jim Dziura, PhD, for excellent statistical advice.
NOTES
Supported by the Yale University Institutional Startup Fund (A.P.), the Virginia Alden Wright Fund (C.S.), and the Doris Duke Charitable Foundation (P.M.W.).
Terms in blue are defined in the glossary, found at the end of this article and online at www.jco.org.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
REFERENCES
Andl T, Kahn T, Pfuhl A, et al: Etiological involvement of oncogenic human papillomavirus in tonsillar squamous cell carcinomas lacking retinoblastoma cell cycle control. Cancer Res 58:5-13, 1998
Gillison ML, Koch WM, Capone RB, et al: Evidence for a causal association between human papillomavirus and a subset of head and neck cancers. J Natl Cancer Inst 92:709-720, 2000
Frisch M, Biggar RJ: Aetiological parallel between tonsillar and anogenital squamous-cell carcinomas. Lancet 354:1442-1443, 1999
Swoboda A, Fabrizii V: Tonsillar carcinoma in a renal graft recipient treated with cyclosporine A. Clin Nephrol 39:272-274, 1993
de Villiers EM, Weidauer H, Otto H, et al: Papillomavirus DNA in human tongue carcinomas. Int J Cancer 36:575-578, 1985
Franceschi S, Munoz N, Bosch XF, et al: Human papillomavirus and cancers of the upper aerodigestive tract: A review of epidemiological and experimental evidence. Cancer Epidemiol Biomarkers Prev 5:567-575, 1996
Snijders PJ, Scholes AG, Hart CA, et al: Prevalence of mucosotropic human papillomaviruses in squamous-cell carcinoma of the head and neck. Int J Cancer 66:464-469, 1996
McKaig RG, Baric RS, Olshan AF: Human papillomavirus and head and neck cancer: Epidemiology and molecular biology. Head Neck 20:250-265, 1998
Psyrri A, DeFilippis RA, Edwards AP, et al: Role of the retinoblastoma pathway in senescence triggered by repression of the human papillomavirus E7 protein in cervical carcinoma cells. Cancer Res 64:3079-3086, 2004
Alani RM, Munger K: Human papillomaviruses and associated malignancies. J Clin Oncol 16:330-337, 1998
Serrano M, Hannon GJ, Beach D: A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4. Nature 366:704-707, 1993
Agoff SN, Lin P, Morihara J, et al: P16(INK4a) expression correlates with degree of cervical neoplasia: A comparison with Ki-67 expression and detection of high-risk HPV types. Mod Pathol 16:665-673, 2003
Begum S, Gillison ML, Ansari-Lari MA, et al: Detection of human papillomavirus in cervical lymph nodes: A highly effective strategy for localizing site of tumor origin. Clin Cancer Res 9:6469-6475, 2003
Cairns P, Mao L, Merlo A, et al: Rates of p16 (MTS1) mutations in primary tumors with 9p loss. Science 265:415-417, 1994
Merlo A, Herman JG, Mao L, et al: 5' CpG island methylation is associated with transcriptional silencing of the tumour suppressor p16/CDKN2/MTS1 in human cancers. Nat Med 1:686-692, 1995
Reed AL, Califano J, Cairns P, et al: High frequency of p16 (CDKN2/MTS-1/INK4A) inactivation in head and neck squamous cell carcinoma. Cancer Res 56:3630-3633, 1996
Weinberger PM, Yu Z, Haffty BG, et al: Prognostic significance of p16 protein levels in oropharyngeal squamous cell cancer. Clin Cancer Res 10:5684-5691, 2004
Haraf DJ, Nodzenski E, Brachman D, et al: Human papilloma virus and p53 in head and neck cancer: Clinical correlates and survival. Clin Cancer Res 2:755-762, 1996
Brandwein M, Zeitlin J, Nuovo GJ, et al: HPV detection using "hot start" polymerase chain reaction in patients with oral cancer: A clinicopathological study of 64 patients. Mod Pathol 7:720-727, 1994
Klussmann JP, Weissenborn SJ, Wieland U, et al: Human papillomavirus-positive tonsillar carcinomas: A different tumor entity Med Microbiol Immunol (Berl) 192:129-132, 2003
Strome SE, Savva A, Brissett AE, et al: Squamous cell carcinoma of the tonsils: A molecular analysis of HPV associations. Clin Cancer Res 8:1093-1100, 2002
Li W, Thompson CH, O'Brien CJ, et al: Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil. Int J Cancer 106:553-558, 2003
Koskinen WJ, Chen RW, Leivo I, et al: Prevalence and physical status of human papillomavirus in squamous cell carcinomas of the head and neck. Int J Cancer 107:401-406, 2003
Coombs NJ, Gough AC, Primrose JN: Optimisation of DNA and RNA extraction from archival formalin-fixed tissue. Nucleic Acids Res 27:15, 1999
Camp RL, Chung GG, Rimm DL: Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 8:1323-1327, 2002
McDougall JK: Immortalization and transformation of human cells by human papillomavirus. Curr Top Microbiol Immunol 186:101-119, 1994
Braakhuis BJ, Snijders PJ, Keune WJ, et al: Genetic patterns in head and neck cancers that contain or lack transcriptionally active human papillomavirus. J Natl Cancer Inst 96:998-1006, 2004
Dai M, Clifford GM, le Calvez F, et al: Human papillomavirus type 16 and TP53 mutation in oral cancer: Matched analysis of the IARC multicenter study. Cancer Res 64:468-471, 2004
Koch WM, Lango M, Sewell D, et al: Head and neck cancer in nonsmokers: A distinct clinical and molecular entity. Laryngoscope 109:1544-1551, 1999
Ritchie JM, Smith EM, Summersgill KF, et al: Human papillomavirus infection as a prognostic factor in carcinomas of the oral cavity and oropharynx. Int J Cancer 104:336-344, 2003
Klussmann JP, Gultekin E, Weissenborn SJ, et al: Expression of p16 protein identifies a distinct entity of tonsillar carcinomas associated with human papillomavirus. Am J Pathol 162:747-753, 2003
Paz IB, Cook N, Odom-Maryon T, et al: Human papillomavirus (HPV) in head and neck cancer: An association of HPV 16 with squamous cell carcinoma of Waldeyer's tonsillar ring. Cancer 79:595-604, 1997
Ringstrom E, Peters E, Hasegawa M, et al: Human papillomavirus type 16 and squamous cell carcinoma of the head and neck. Clin Cancer Res 8:3187-3192, 2002
Schwartz SR, Yueh B, McDougall JK, et al: Human papillomavirus infection and survival in oral squamous cell cancer: A population-based study. Otolaryngol Head Neck Surg 125:1-9, 2001
Stein GH, Drullinger LF, Soulard A, et al: Differential roles for cyclin-dependent kinase inhibitors p21 and p16 in the mechanisms of senescence and differentiation in human fibroblasts. Mol Cell Biol 19:2109-2117, 1999
Alcorta DA, Xiong Y, Phelps D, et al: Involvement of the cyclin-dependent kinase inhibitor p16 (INK4a) in replicative senescence of normal human fibroblasts. Proc Natl Acad Sci U S A 93:13742-13747, 1996
Mellin H, Dahlgren L, Munck-Wikland E, et al: Human papillomavirus type 16 is episomal and a high viral load may be correlated to better prognosis in tonsillar cancer. Int J Cancer 102:152-158, 2002
Van Tine BA, Dao LD, Wu SY, et al: Human papillomavirus (HPV) origin-binding protein associates with mitotic spindles to enable viral DNA partitioning. Proc Natl Acad Sci U S A 101:4030-4035, 2004
Nuovo GJ, Plaia TW, Belinsky SA, et al: In situ detection of the hypermethylation-induced inactivation of the p16 gene as an early event in oncogenesis. Proc Natl Acad Sci U S A 96:12754-12759, 1999, 百拇医药
ABSTRACT
PURPOSE: We sought to determine the prevalence of biologically relevant human papillomavirus (HPV) in oropharyngeal squamous cell carcinoma (OSCC). Retinoblastoma (Rb) downregulation by HPV E7 results in p16 upregulation. We hypothesized that p16 overexpression in OSCC defines HPV-induced tumors with favorable prognosis.
METHODS: Using real-time polymerase chain reaction for HPV16, we determined HPV16 viral load in a cohort of 79 OSCCs annotated with long-term patient follow-up. A tissue microarray including these cases was also analyzed for p53, p16, and Rb utilizing in situ quantitative protein expression analysis. Seventy-seven tumors were classified into a three-class model on the basis of p16 expression and HPV-DNA presence: class I, HPV–, p16 low; class II, HPV+, p16 low; and class III, HPV+, p16 high.
RESULTS: Sixty-one percent of OSCCs were HPV16+; HPV status alone was of no prognostic value for local recurrence and was barely significant for survival times. Overall survival was improved in class III (79%) compared with the other two classes (20% and 18%; P = .0095). Disease-free survival for the same class was 75% versus 15% and 13% (P = .0025). The 5-year local recurrence was 14% in class III versus 45% and 74% (P = .03). Only patients in class III had significantly lower p53 and Rb expression (P = .017 and .001, respectively). Multivariable survival analysis confirmed the prognostic value of the three-class model.
CONCLUSION: Using this system for classification, we define the molecular profile of HPV+ OSCC with favorable prognosis, namely HPV+/p16 high (class III). This study defines a novel classification scheme that may have value for patient stratification for clinical trials testing HPV-targeted therapies.
INTRODUCTION
Numerous lines of molecular and epidemiologic evidence demonstrate that some of the oncogenic human papillomaviruses (HPVs) are etiologically related to a subset of oropharyngeal tumors.1,2 Individuals with HPV-associated anogenital cancers,3 post-transplant patients, HIV-infected men and patients with Fanconi anemia are prone to develop head and neck cancers.4
Since 1986, the first time that HPV16 DNA was detected in an invasive squamous cell carcinoma (SCC) of the head and neck (HNSCC) by Southern blot hybridization,5 HPV sequences have been detected repeatedly in a variable proportion of HNSCC, from less than 10% to up to 100%,6-8 apparently biased by the anatomic location of tumors and HPV detection techniques. Despite numerous reports confirming high frequency of HPV DNA presence in oropharyngeal cancers, the prevalence of biologically relevant HPVs in these tumors is unclear. HPV DNA detection in tumors per se does not indicate causal association.
Molecular studies in cervical cancer, the most widely accepted HPV-associated malignancy, indicate that HPV-driven malignant conversion is associated with certain molecular events. First, the viral DNA usually becomes integrated into the host genome. After integration, the expression of the main viral transcription/replication factor E2, which represses the transcription of E6 and E7 oncogenes, is disrupted. The disruption of E2 expression leads to dysregulated expression of the E6 and E7 oncogenes. The E6 and E7 genes of the oncogenic HPVs encode oncoproteins that bind and degrade p53 and retinoblastoma (Rb) tumor suppressors, respectively.9,10 Most cervical carcinomas harbor wild-type p53 and Rb tumor-suppressor genes. Thus, the tumor-suppressor pathways are intact but dormant in these cells because of the continuous expression of E6 and E7 genes.
Overexpression of p16 protein has been reported repeatedly in HPV-associated cancers. The p16 protein functions as a tumor suppressor by binding to the cyclin D1 CDK4/CDK6 complex, preventing phosphorylation of the Rb protein.11 In one study in cervical and genital lesions, levels of p16 protein expression were associated with HPV oncogenic potential.12 A recent study of lymph node metastases in HNSCC reported that p16 overexpression is a surrogate marker for oropharyngeal origin and HPV-association.13 To the contrary, in tobacco-related HNSCC loss of p16 protein expression is a common and early event.14-16 Thus, immunohistochemical analysis of p16 protein is a reliable method of detection of p16 alterations in HNSCC.
We previously studied p16 protein expression levels by immunohistochemistry (IHC) on a tissue array of oropharyngeal cancer cases annotated with long-term patient follow-up data.17 We found the prevalence of p16 overexpression in our cohort to be 20%; in addition, p16 overexpression was associated with favorable patient outcome. If indeed p16 overexpression is a surrogate marker of HPV association, then our prevalence of HPV-associated cancers would be inconsistent with most studies reporting a prevalence of 50% in oropharyngeal location. On the basis of these findings, we hypothesized that p16 overexpression defines a subgroup of HPV-positive oropharyngeal tumors with a more favorable prognosis. Thus, there may be a subgroup of HPV-positive p16 nonoverexpressing oropharyngeal tumors of multifactorial pathogenesis. In this model, tobacco may induce loss of p16 function and negate the favorable impact that the virus would have on prognosis. This hypothesis was also supported by the conflicting data in the literature regarding the impact of HPV on prognosis of patients with oropharyngeal cancer1,7,18-23; it is possible that the results of these studies are affected by the prevalence of the p16 overexpressors within the group of HPV-positive tumors.
In this study, we sought to determine the role of HPV in pathogenesis and clinical behavior of oropharyngeal squamous cell cancers. We studied a cohort of 107 oropharyngeal squamous cell cancers for HPV16 DNA viral load by real-time polymerase chain reaction (PCR). In addition, we constructed a tissue array composed of these tumors and studied expression of p53, Rb, and p16 proteins using a quantitative, in situ method of protein analysis. We hypothesized that for HPV DNA–positive cases, p16 expression status would differentiate the biologically relevant ones. Our results indeed delineate three biologically and clinically distinct types of oropharyngeal squamous cell cancers based on HPV-DNA determination and p16 expression status: one class of HPV-negative/p16-nonexpressing, 1 class of HPV-positive/p16-nonexpressing, and one class HPV-positive/p16-expressing oropharyngeal tumors. We were able to show that only the HPV-positive/p16-expressing tumors fit the cervical carcinogenesis model and that they are the ones associated with favorable prognosis.
METHODS
Tumor Specimens and Tissue Microarray Construction
After institutional review board approval, paraffin-embedded specimens from 107 patients were collected from Yale–New Haven Hospital (New Haven, CT) archives. Inclusion criteria were primary or recurrent histologically confirmed SCC of the oropharynx treated between 1980 and 1999 with primary external beam radiotherapy (RT) or gross total surgical resection and postoperative RT. Exclusion criteria included presentation with metastatic disease, recurrent disease when the primary tumor was available, or failure to receive full course of radiation therapy. Hematoxylin and eosin–stained (H&E) slides were reviewed by a pathologist (D.K.), evaluated for the presence of squamous cell carcinoma, and marked for 0.6-mm core extraction. Cores were placed on the recipient microarray block using a Tissue Microarrayer (Beecher Instrument, Silver Spring, MD). All tumors were represented with two-fold redundancy. Cores from HPV18-positive HeLa and HPV16-positive Caski cell lines fixed in formalin and embedded in paraffin were selected for controls and included in the array. The tissue microarray was then cut to yield 5-μm sections and placed on glass slides using an adhesive-tape transfer system (Instrumedics, Hackensack, NJ).
DNA Extraction
Subsequently, DNA was extracted from tumor samples. H&E slides of specimen blocks were reviewed by a pathologist (D.K.) and tumor areas were identified. Four tissue cores were extracted from these areas of SCC using sterile, blunt 18-gauge needles (VWR Scientific, West Chester, PA). Tissue cores were subsequently extracted via chelex-100/proteinase K protocol as previously described.24 Extracted DNA was aliquoted for storage at –20°C until used. DNA was extracted from cell lines (HeLa and Caski) using the same protocol, substituting a cell pellet from one 75-cm2 culture flask grown to confluence for the tissue cores.
Quantitative PCR
Quantitative PCR (qPCR) was carried out in a BioRad iCycler (Hercules, CA) 96-well thermal cycler equipped with fluorescence detection. Primer concentrations used are indicated in Table 1. Complete details of qPCR methods can be found in the Appendix (online only). As a confirmation step, a majority of samples were additionally subjected to real-time PCR analysis using a TaqMan double-dye based assay (Eurogentec Corp, Seraing, Belgium) as previously described. For this analysis, a SmartCycler (Cepheid, Sunnyvale, CA) qPCR instrument was used, and samples were run in duplicate. TaqMan primer/probe sequences and concentrations used are also indicated in Table 1.
Conventional IHC for p16 Expression Status
We previously determined p16 expression status of tumor samples in this cohort17 using a commercially available monoclonal p16 antibody kit (DAKO Cytomation, Carpinteria, CA). HeLa cell lines embedded in paraffin were used for positive control, and nonspecific mouse immunoglobulin G was substituted for p16 antibody as a negative control. In our previous investigation, we determined p16 expression to be essentially dichotomous (present or absent), therefore tumors were classified as p16 expressors (strong, diffuse staining) or p16 nonexpressors (weak or absent staining). These results were also confirmed separately by quantitative AQUA (HistoRx, New Haven, CT) immunofluorescence analysis for p16 using the same antibody (see Fluorescence IHC).
Tumor Classification
On the basis of the results of our previous study of p16 expression in oropharyngeal SCC, the existence of a distinct subgroup of p16 nonexpressing tumors with HPV DNA presence was hypothesized. Therefore, tumors were classified according to a 2 x 2 table by p16 expression and HPV16 DNA presence. Possible groups included HPV16 negative/p16 nonexpressing, HPV16 negative/p16 expressing, HPV16 positive/p16 nonexpressing, and HPV16 positive/p16 expressing (Fig 1).
Fluorescence IHC
The use of AQUA quantitative fluorescence IHC has been described previously25 and is also available in the Appendix (online only). Primary antibody to p16, p53 and Rb were used at 4°C overnight. The primary antibodies and concentrations used are summarized in Table 2.
Statistical Analysis
Comparison of HPV16 viral load and age by HPV/p16 classification (three-class model) was made by Kruskal-Wallis analysis followed by Wilcoxon rank sum test with Bonferroni corrections. Comparison of protein expression (AQUA score) by HPV/p16 classification was made using one-way analysis of variance (ANOVA) with post hoc comparisons by Dunnett's T3 test. Comparison of HPV/p16 (three class model) classification status with the clinical and pathologic variables sex, TNM stage, T stage, N stage, histologic grade, management (primary external beam RT [EBRT] v primary surgical excision plus RT), alcohol use, tobacco use and clinical response was made by Fisher's exact test. Self-reported alcohol use was classified as a dichotomous variable: social or none versus current drinker. Patients who reported no alcohol use for more than 2 years prior were included in the social/none category. Tobacco use was similarly classified as reported tobacco history of less than one pack per day or quit more than 10 years prior, versus current smoker of one or more packs per day. Clinical response was recorded as complete response versus no response/partial response after completion of treatment.
Disease-free survival (DFS), overall survival (OS), and local recurrence were assessed by Kaplan-Meier analysis with log-rank for determining statistical significance. Relative risk and prognostic independence was assessed by the multivariate Cox proportional hazards model including clinical and pathologic variables HPV/p16 classification, T stage, N stage (±), histologic grade, tumor type (primary v recurrent), and management (primary EBRT v surgery + RT). All survival analyses were performed at 5-year cutoffs.
Fisher's exact test comparisons were made using SAS version 8.2 (SAS Institute, Cary, NC). All other calculations and analyses were performed with the Statistical Package for the Social Sciences version 11.5 for Windows (SPSS Inc, Chicago, IL) and where appropriate were two tailed. Box/whisker plots were generated with Prism version 4.0 (GraphPad Software, San Diego CA).
RESULTS
Clinical and Pathologic Variables
Our study included 107 patients with histologically confirmed oralpharyngeal SCC (94 primary, 13 recurrent). Median follow-up time for survivors was 32.1 months, and 22.0 months for the entire cohort. Demographic and clinicopathologic variables for the cohort are summarized in Table 3.
HPV16 DNA Viral Load
Standard curves for -globin DNA demonstrated linear sensitivity (range, 4,000 to one human genome/μL). Standard curves for HPV16 E6 demonstrated linear sensitivity (range, 230,000 to 10 copies/μL). Melt curve (Tm) analysis was performed for all SYBR reactions (-globin and HPV16 E6) and demonstrated consistent Tms. Example melt analysis and standard curves for -globin and HPV16 E6 are shown in Figure 2.
Seventy-nine tumors had sufficient amplifiable -globin (>1 human genome/μL) by qPCR and were analyzed for HPV16 E6 DNA presence. Of the 79 tumors analyzed, 61% had evidence of significant HPV16 DNA presence using the criteria of at least one copy per 10 cell genomes' worth of DNA. Thirty-nine percent of samples were judged as negative for HPV16. The median copy number of HPV16 E6 DNA in HPV16-positive tumors was 14.1copies per human genome (range, 0.11 to 3,948 copies/cell genome).
p16 Protein Expression (conventional IHC and AQUA)
In our previous study,17 p16 expression status was determined in these tumors by conventional IHC. Tumors were classified as p16 expressors (strong and diffuse) or p16 nonexpressors (weak, absent, or focal staining). Analysis was restricted to the 79 tumors with amplifiable DNA. There were 78 (99%) of 79 tumors with sufficient tissue for determining p16 expression. Of these, 19 (24%) of 79 tumors were designated as p16 expressors, as were seven of seven cell line positive controls. These results were confirmed for this study using quantitative AQUA analysis of p16 expression. Of the patients included in this study, 53 (67%) of 79 had sufficient tissue for AQUA analysis of p16 expression. As visualized by conventional IHC (Fig 3), p16 displayed a diffuse cytoplasmic/nuclear pattern of expression that was essentially dichotomous (highly expressed v low/no expression). Results seen in the traditional IHC experiment were confirmed by AQUA analysis, with p16 nonexpressors having a median AQUA score of 3.4 versus 112.3 for p16 expressors (Fig 4).
Tumor Classification
Based on the results of our previous study of p16 expression in oropharyngeal SCC, the existence of a distinct subgroup of p16 nonexpressing tumors with HPV DNA presence was hypothesized. Therefore, tumors were classified according to a 2 x 2 table by p16 expression and HPV16 DNA presence (Fig 1). There were 78 tumors for which both HPV16 status and p16 expression status were available. Thirty tumors (39%) were class I (HPV16 negative/p16 nonexpressors). Twenty-nine tumors (37%) were class II (HPV16 positive/p16 nonexpressors), and 18 (23%) were class III (HPV16 positive/p16 expressors). There was one tumor with no amplifiable HPV16 E6 but p16 expression. Because this represented a classification group with only one member, it was excluded from further analysis, leaving a cohort of 77 tumors in three classes for subsequent analyses.
Comparison of HPV16 Viral Load by HPV/p16 Classification
Using the three-class model, there was a significant difference in HPV16 viral load between groups (P < .001). Post hoc comparison demonstrated that class II and class III tumors had significantly higher viral load than class I tumors (P = .003 each). The median viral load of class III tumors was higher than class II tumors; however, the difference was not significant at the .05 level (P = .078 after Bonferroni correction). The median viral loads (copies HPV16/human genome) were class I (0.0 copies), class II (3.6 copies), and class III (46.0 copies). These results are summarized graphically in Figure 5.
Quantitative AQUA Analysis of p53 and Rb Protein Expression
Of the 79 patients with amplifiable DNA, 57 (72%) had sufficient tissue for analysis of p53 protein expression by AQUA. Using the three-class model, there was a significant difference in p53 expression between groups (P = .026). Post hoc comparisons demonstrated that class III tumors had significantly lower p53 AQUA scores than class I and class II tumors (P = .017 and .015 respectively). For Rb, 65 (82%) of 79 had sufficient tissue for analysis of protein expression by AQUA. Comparison of Rb expression by HPV/p16 classification demonstrated a significant difference between groups (P = .001). Post hoc comparisons demonstrated that class III tumors had significantly lower AQUA scores than class I and class II tumors (P < .001 and .001 respectively). Results of fluorescence IHC and AQUA analysis are summarized in Figure 4 and Table 4.
Clinicopathologic Variables by HPV/p16 Classification (three-class model)
Tumors assigned to HPV/p16 class III (p16 expressors/HPV16 positive) were less likely to be recurrent versus primary tumor type, with 100% of class III tumors being primary tumor type (P = .028). Class III tumors were also associated with more advanced TNM stage (P = .022). This was a result of N stage contribution, as class III tumors were more likely to have lower T stage (P = .004) but higher N stage (P = .021). Class III tumors were also closely associated with a poorly differentiated histology, with 80% of class III tumors being poorly differentiated (P = .030). One hundred percent of patients with class III tumors experienced a clinical complete response to initial therapy, but this was not significant at the .05 level (P = .18). Patients with class I tumors (p16 nonexpressors/HPV16 negative) were statistically less likely to report low/no tobacco history, representing only 4% of patients with class I tumors (P = .006). Patients with class III tumors were less likely to report frequent alcohol consumption, although this was not significant at the .05 level (P = .06). Clinicopathologic variable correlations are summarized in Table 3.
Survival Analysis by Three-Class Model Only
When all HPV16-positive cases were compared with the HPV16-negative cases, HPV positivity was found to have no prognostic value for local recurrence (P = .66). For DFS and OS, HPV positivity did show some prognostic value. Patients with HPV-positive tumors had improved DFS (P = .05) and OS (P = .03) compared with the HPV-negative cohort. When the HPV16-positive cases were split by the three-class model, the predictive power was increased dramatically for local recurrence and DFS and OS.
For OS, patients with class III tumors had markedly improved OS compared with class I and class II tumors. Patients with class III tumors had a 5-year OS of 79% versus 20% for class I and 18% for class II (P = .0095). For DFS the same relationship was noted, with class III tumors demonstrating a 5-year DFS of 75% versus 15% for patients with class I and 13% for patients with class II tumors (P = .0025).
For 5-year local recurrence, patients with class III tumors had very low recurrence rates compared with both class I and class II tumors (P = .027). Class III patients had 5-year local recurrence rates of 14% versus 45% for class I and 74% for class II. Univariate survival analyses are summarized in Figure 6 and Table 5.
Multivariate Survival Analysis by Three-Class Model and Clinicopathologic Variables
Using the Cox proportional hazards model, we performed multivariable analysis to assess the independent predictive value of the three-class HPV/p16 classification model for local recurrence and overall and DFS.
For local recurrence, class III tumors were 4.8 times less likely to recur than class I tumors, although this was not significant at the .05 level (P = .054). For DFS, HPV/p16 classification was the most significant independent predictor. Class III patients were five times less likely to die or suffer recurrence as compared with class I patients (P = .005). For OS, HPV/p16 classification was the only significant independent predictor. Patients with class III tumors were 5.3 times less likely to die as a result of any cause than patients with class I tumors (P = .013). Complete results of multivariable survival analysis are presented in Table 6.
DISCUSSION
In this study, we identified a subset of patients with oropharyngeal SCC whose tumors display a unique molecular phenotype (ie, class III tumors). We believe this molecular phenotype defines HPV-positive tumors with favorable prognosis. These class III tumors represented 23% of the cohort, and were characterized by dramatically reduced local recurrence rates, as well as markedly improved 5-year DFS and OS.
Although the causal link between HPVs, especially type 16, and a subset of oropharyngeal cancers is becoming more firmly established,2,22 the prevalence of biologically relevant HPVs in these tumors has yet to be determined. Although HPVs can initiate carcinogenesis in epithelial cells,26 HPV infection per se is not sufficient to induce malignant transformation. The number of HNSCCs that test positive for HPV DNA presence far exceeds the number that actually express HPV oncoproteins. Only this latter group of tumors are likely to be HPV induced. In this study, by measuring protein levels of proteins of p53 and Rb pathways, we provide indirect evidence for E6 and E7 oncoprotein production, respectively, by a subset of HPV DNA–positive tumors. Prior studies have also shown that not all HPV DNA–positive HNSCC contain transcriptionally active HPV. In a recent study of 143 oral and oropharyngeal SCCs by Braakhuis et al,27 17% of tumors contained HPV16 DNA, but only 8% expressed E6/E7 mRNA. Another recently published case-control study investigated the relationship between HPV-DNA and tp53 mutations in oral cancer.28 The authors found that only tumors that had both HPV DNA and presence of antibodies to HPV E6 in patient serum displayed a reduction in tp53 mutations. Tumors with HPV DNA but no antibodies to HPV E6 had a similar mutation frequency to that of HPV-negative tumors. These studies included oral cavity tumors in addition to oropharyngeal tumors, thus the lower overall prevalence of HPV16 positive tumors in Brakhuis et al's article. This difference notwithstanding, the concept of only a subset of HPV-positive tumors being HPV induced is strongly supported by this evidence. In this study, we demonstrate an indirect method to detect transcriptionally active HPV in paraffin-embedded material. In addition, this study provides strong evidence that among HPV DNA–positive oropharyngeal tumors, only the transcriptionally active ones are biologically as well as clinically relevant.
In our study, we found that only a subset (class III) of HPV-DNA positive tumors match the expected molecular phenotype of HPV-induced malignancy, low p53 and Rb expression, and high p16 expression.
As previously mentioned, we classified oropharyngeal tumors into three distinct categories by determining HPV DNA presence and p16 status. HPV DNA positive/p16 high (class III) tumors displayed high HPV16 viral load, low p53 and low Rb protein levels. Tumors in this group were poorly differentiated and associated with lower alcohol and tobacco usage history. Despite the adverse histologic features, patients had lower local recurrence rate and improved DFS and OS. Several previous studies have found that HPV-associated head and neck cancers tend to share these features.2,18,29,30 To the contrary, HPV DNA positive/p16 nonexpressing (class II) tumors displayed a similar molecular profile to the HPV-negative tumors. These tumors had a median viral load of 3.6 viral copies per human genome, making them unlikely to represent false-positive results. Interestingly, patients in class II had a higher local recurrence rate than patients with HPV-negative (class I) tumors. A smaller previous study of p16 expression in tonsillar carcinomas found 2 (11%) of 18 HPV DNA–positive tumors lacked strong p16 expression.31 Both of these patients also suffered recurrences. It is possible that in class II tumors, HPV super-infection of a pre-existing neoplasm may be acting in an additive or synergistic fashion to the traditional carcinogenic insults of alcohol and tobacco, but this remains conjecture.
Our prevalence of HPV16 positivity of 61% is similar to other reports in oropharyngeal cancers.2,18,32,33 However, only 23% displayed the molecular profile that, on the basis of the cervical carcinogenesis model, defines HPV-causal association. Thus, the conflicting reports in the literature regarding the prognosis of HPV-associated head and neck cancers1,7,18-23 may be explained by the molecular heterogeneity of the HPV-positive group. Knowledge of the p16 protein status clarifies delineation of the proportion of HPV-induced oropharyngeal cancers.
We found that HPV DNA–positive/p16-expressing tumors (class III) had a low recurrence rate and markedly improved DFS and OS. Strikingly only 14% of class III tumors developed local recurrence whereas 45% of class I and 74% of class II recurred locally in 5 years. In addition the overwhelming majority of class III patients (approximately four of five) were alive in 5 years whereas only one of five of class I or II patients were alive in 5 years. The favorable outcome of HPV-induced oropharyngeal cancers may be due to the absence of field cancerization or enhanced radiation sensitivity2,34; all of the patients in this study received radiation therapy either as primary treatment modality or postoperatively. From the molecular biology standpoint, it is possible that the high p16 levels of HPV-induced oropharyngeal SCC account for the improved prognosis. Specifically, studies exploring the events of replicative and accelerated senescence in normal fibroblasts have shown that p16 is required for maintenance of growth arrest in senescent cells whereas Rb may not play an active role in the maintenance of senescent phenotype; although Rb levels decrease after the onset of senescence in normal fibroblasts,35 p16 becomes constitutively upregulated.35,36 Therefore, maintenance of senescence in tumor cells after treatment such as radiation may be dependent on p16 levels and that would explain the difference in outcome between p16 expressors and nonexpressors. However, a prospective randomized trial adjusting for other important prognostic variables is needed to definitively determine the impact of HPV association on patient prognosis.
The finding that high HPV16 viral load is associated with better prognosis has been reported previously.37 Other investigators have also shown that transcription of HPV16 E6/E7 mRNA in tonsillar carcinomas is not necessarily dependent on viral DNA integration and that the virus is predominately in episomal form. How the virus can remain in cancer tissues as episomes with high copy numbers is not fully understood. A recent report by Van Tine et al38 revealed that, in addition to its role as a transcriptional regulator, the HPV E2 protein may serve as an "anchor" to bind episomal HPV to cellular mitotic spindles, thus ensuring faithful maintenance of the episomal state.
On the basis of our results, we are proposing a model for HPV-associated oropharyngeal cancer (Fig 7). The classic oncogenic insult in HNSCC has been attributed to excess alcohol or tobacco exposure producing mutations or epigenetic inactivations of p53, p16, and Rb in a multistep progression from healthy cell to dysplasia to frank carcinoma. Alternatively, high-risk HPV may also lead to tumor formation, in a similar fashion to its role in cervical cancer. In these HPV-induced tumors, oncogenic HPV E6 and E7 proteins act to inactivate p53 and Rb pathways, with subsequent upregulation of p16 expression through loss of feedback inhibition. This obviates the need for mutational inactivation of these genes. In addition to these two classes of oropharyngeal tumors, we describe a novel class in which HPV16 is present but p53, p16, and Rb expression are similar to non–HPV-positive (class I) tumors. These class II tumors could be multifactorial in origin, possibly formed when tobacco-/alcohol-related tumors are infected by high-risk HPVs. These tumors may or may not represent a separate group biologically from HPV-negative tumors and need to be investigated further.
The single patient with a p16 expressing tumor that was HPV16 negative may represent presence of another, less common oncogenic subtype of HPV such as 18 or 33. Previous studies have demonstrated that although the vast majority of oncogenic subtypes present in HNSCC are type 16, other high-risk types such as 18 or 33 are present in 1% to 3% of oropharyngeal tumors.31
In this study, using AQUA technology, we were able to elucidate protein expression relationships using routine paraffin-embedded tissue that previously would have required fresh- or snap-frozen tissue resources on a large scale. As seen in Figure 4, the quantitative protein expression data generated from AQUA analysis allows much more powerful analysis than that provided by simple IHC, which relies on arbitrary cut points and qualitative scoring. However, it is important to note that for a protein with dichotomous expression patterns, such as p16, traditional IHC provides essentially the same information as AQUA analysis.
Our results have important diagnostic, prognostic, and clinical implications. IHC for p16 can be used as a simple test to determine whether biologically relevant HPV is present in tumors. HPV-targeted strategies such as therapeutic vaccines are currently undergoing clinical trials in head and neck cancer patients. The addition of p16 IHC to existing protocols for determination of HPV DNA presence would seem to be of vital importance for patient selection in these trials. We hypothesize that HPV DNA presence combined with low p16 expression indicates that these tumors do not harbor transcriptionally active HPV. In fact, p16 promoter methylation leading to abrogation of p16 function has been reported in cervical cancer cells that do not express viral E7 transcripts.39 The rationale for HPV-targeted therapies in these patients, in our view, is therefore lacking. In addition to potentially affecting clinical trial patient selection, our results have additional therapeutic implications. Patients with p16 expressing oropharyngeal tumors seem to have a favorable prognosis and may be spared the devastating sequelae of aggressive therapies. We also feel that current clinical trials in patients with oropharyngeal cancer should stratify by HPV DNA status/p16 expression, or at least include them as a prognostic variable. Although our evidence is compelling, these findings need to be confirmed in a prospective fashion.
Appendix
Several candidate primer pairs for each target gene (human -globin and human papillomavirus [HPV] E6) were identified using OligoPerfect software (Invitrogen Corp, Carlsbad, CA, http://www.invitrogen.com/content.cfmpageid=9716) and a review of the literature (Tucker RA, Unger ER, Holloway BP, et al: Real-time PCR-based fluorescent assay for quantitation of human papillomavirus types 6, 11, 16, and 18. Mol Diagn 6:39–47, 2001; and Peitsaro P, Johansson B, Syrjanen S: Integrated human papillomavirus type 16 is frequently found in cervical cancer precursors as demonstrated by a novel quantitative real-time PCR technique. J Clin Microbiol 40:886–91, 2002). Criteria specified for the OligoPerfect software included amplimer length of 80 to 120 base pairs (bp), primer length of 18 to 25 bp, primer melting threshold (Tm) between 58°C and 62°C, primer percentage guanine-cytosine between 30 and 60, optimized for 50 nM salt concentration. Candidate primers were synthesized by the Yale University Critical Technologies in Molecular Medicine core facility (New Haven, CT) and screened using an SYBR Green–based quantitative real-time polymerase chain reaction (qPCR) assay with known positive and negative controls for each gene target. Primers were selected based on sensitivity as evidenced by threshold cycle (Ct) qPCR analysis, and specificity as evidenced by Tm analysis and agarose gel electrophoresis visualization of reaction end products. Selected primers are indicated in Table 1 of this article. All primers utilized were optimized using the primer-chessboard method described by Gunson, Gillespie, and Carman (Optimisation of PCR reactions using primer chessboarding. J Clin Virol 26:369–373, 2003). Briefly, forward and reverse primers are combined across concentration gradients, with all other reaction conditions equal. This method allows selection of optimum forward and reverse primer concentrations, which are not always equimolar. qPCR was carried out in a BioRad iCycler (Hercules, CA) 96-well thermal cycler equipped with fluorescence detection. Final reaction conditions were 1x iQ SYBR Green Supermix (BioRad; 50 mM KCl; 20 mM Tris-HCl; 0.2 mM each deoxynucleotide triphosphate, 25 U/mL iTaq hot start DNA polymerase, 3 mM MgCl2, SYBR Green I dye, 10 nM floresein, and stabilizers). Primer concentrations used are indicated in Figure 1. Each reaction vial consisted of 20 μL master mix to which 5μL target DNA was added for a final reaction volume of 25 μL. All reactions (-globin and HPV16 E6) for a patient were each run in triplicate on the same 96-well plate. Each plate included a three-log serial dilution positive control for -globin and HPV16 E6, and three negative controls for each. The PCR reaction was preceded by 3 minutes at 95°C to activate the iTaq DNA polymerase. Reactions were run at 40 cycles of 95°C for 15 seconds followed by 60°C for 30 seconds. Fluorescence was detected during the combined annealing/extension phase of each cycle. The cycle threshold (Ct) was determined as the point at which each sample’s fluorescence crossed 30 units. Following 40 cycles, a melt curve analysis was performed by ramping the temperature from 60°C to 95°C at a rate of 0.1° per second, while monitoring resulting fluorescence. Following melt curve analysis, a final extension step at 60°C for 1 minute to reform double-stranded DNA was performed. As a confirmation step, a majority of samples were additionally subjected to PCR analysis using a TaqMan double-dye based assay as previously described. For this analysis, a SmartCycler (Cepheid, Sunnyvale, CA) qPCR instrument was used, and samples were run in duplicate. Primers and probes for the TaqMan assay were assembled by Eurogentec Corp (Seraing, Belgium). TaqMan primer/probe sequences and concentrations used are also indicated in Table 1. Reaction conditions for the TaqMan assay were 1x PCR supermix (UDG Platinum Supermix, Invitrogen Corporation) with 2 mM added MgCl2. Reactions were carried out with 23 μL master mix plus 2 μL target, for 25 μL final volume. Thermal parameters for -globin included a 2-minute incubation at 50°C to allow uracil-deglycosylase (UDG) activity and a 95°C Taq-polymerase activation step for 2 minutes followed by 45 cycles consisting of 95°C for 15 seconds, 61°C annealing step for 15 seconds plus optical analysis, and a 70°C extension step for 30 seconds. HPV16 E6 conditions were similar, with substitution of a combined extension/annealing step of 65°C for 30 seconds plus optics. Some patients had insufficient extracted DNA to run both assays (minimum required volume of 45 μL for the SYBR assay and 8μL for the TaqMan assay), in which case the TaqMan assay was used in triplicate with no SYBR assay.
PCR Data Analysis
Standard curves for both assays (TaqMan and SYBR) were developed for -globin using serial dilutions of purified high-molecular weight human DNA (Promega Corp, Madison, WI) from 4,000 genomes/μL to 1 genome/μL. Standard curves for HPV16 E6 were developed using a purified whole-length HPV16 plasmid at concentrations from 230,000 copies/μL to 10 copies/μL. This allowed determination of HPV16 E6 concentration and expression of the final quantity as the number of viral copies present per human genome (Tucker et al). Samples with -globin values less than 1 human genome/μL indicated lack of amplifiable DNA and were discarded from qPCR analysis. HPV viral load quantities derived that were greater than 0.1 (> 1 HPV genome copy/10 cells) were scored as positive for HPV16. This threshold was based on the assumption that the presence of less than 1 HPV genome copy/10 cellular genomes indicates that HPV is present in only a minority of the dominant clonal population. A viral load between 0.1 and 1 may represent stromal/nontumor DNA (not harboring HPV) being included in a tissue core and yielding an apparent viral load less than 1 HPV/cell. Previous investigators have utilized this cutoff as a viral load of 1 HPV/cell is considered sufficient to cause transformation (Hernandez-Avila M, Lazcano-Ponce EC, Berumen-Campos J, et al: Human papilloma virus 16–18 infection and cervical cancer in Mexico: a case-control study. Arch Med Res 28:265–71, 1997; Berumen J, Casas L, Segura E, et al: Genome amplification of human papillomavirus types 16 and 18 in cervical carcinomas is related to the retention of E1/E2 genes. Int J Cancer 56:640–5, 1994; Kutler DI, Wreesmann VB, Goberdhan A, et al: Human papillomavirus DNA and p53 polymorphisms in squamous cell carcinomas from Fanconi anemia patients. J Natl Cancer Inst 95:1718–21, 2003; and Ha PK, Pai SI, Westra WH, et al: Real-time quantitative PCR demonstrates low prevalence of human papillomavirus type 16 in premalignant and malignant lesions of the oral cavity. Clin Cancer Res 8:1203–9, 2002). Similarly, SiHa is an immortalized human cervical cancer cell line that contains one to two HPV 16 copies per cell.
Fluorescent Immunohistochemistry
In brief, slides were deparaffinized with xylene followed by ethanol. After rehydration in dH20, antigen retrieval was accomplished by pressure cooking in 0.1 M citrate buffer (pH, 6.0). Endogenous peroxidase activity was blocked by incubating in 0.3% hydrogen peroxide in methanol for 30 minutes. Nonspecific antibody binding was then blocked with 0.3% bovine serum albumin (BSA) for 30 minutes at room temperature. Following these steps, slides were incubated with primary antibody to p16, p53 or Rb at 4°C overnight. The primary antibodies and concentrations used are summarized in Table 2. Of note, the retinoblastoma antibody utilized detects both phosphorylated and unphosphorylated forms of retinoblastoma. Similarly, the p53 antibody utilized detects both mutant and wild-type p53. Subsequently, slides were incubated with goat antimouse secondary antibody conjugated to a horseradish peroxidase-decorated dextran polymer backbone (Envision; DAKO Corp, Carpinteria, CA) for 1 hour at room temperature. Tumor cells were distinguished from stroma by use of anticytokeratin antibody cocktail (rabbit antipancytokeratin antibody z0622; DAKO Corp) with subsequent goat antirabbit antibody conjugated to Alexa546 fluourophore (A11035; Molecular Probes, Eugene, Oregon). We added 4', 6-diamidino-2-phenylindole (DAPI) to visualize nuclei. Target (p53, Rb, p16) molecules were subsequently visualized with a fluorescent chromogen (Cy-5-tyramide; PerkinElmer Corp, Wellesley, MA). Slides were mounted with a polyvinyl alcohol-containing aqueous mounting media with antifade reagent (n-propyl gallate; Acros Organics, Geel, Belgium). Appropriate positive controls were included in the tissue microarray, including HeLa cell lines embedded in paraffin (for p16), colon cancer previously determined positive for p53 (for p53), and squamous cell carcinoma (for Rb). For negative control, the primary antibody step was eliminated.
Automated Quantitative Protein Expression Analysis (AQUA)
Automated image acquisition and analysis using automated quantitative protein expression analysis AQUA (HistoRx, New Haven, CT) has been described previously (Camp RL, Chung GG, Rimm DL: Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 8:1323–7, 2002). In brief, monochromatic, high-resolution images were obtained of each histospot after immunofluorescent staining as described herein. We distinguished areas of tumor from stromal elements by creating a mask from the cytokeratin signal. A tumor nuclei–specific compartment was created by using DAPI signal to identify nuclei, and the cytokeratin signal to define cytoplasm/membrane. The target signal (AQUA score) was scored on a normalized scale of 1 to 255 expressed as pixel intensity divided by the target area (tumor nuclei compartment). AQUA scores for duplicate tissue cores were averaged to obtain a mean AQUA score for each tumor. Histospots containing less than 10% tumor, as assessed by tumor mask area (automated), were deemed too small for accurate representation of tumor protein expression and were excluded from further analysis.
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following authors or their immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
AuthorsEmploymentLeadershipConsultantStockHonorariaResearch FundsTestimonyOther
Robert L. CampHistoRx (A)HistoRx (A)
David L. RimmHistoRx (A)HistoRx (A)
Dollar Amount Codes (A) < $10,000 (B) $10,000-99,999 (C) $100,000 (N/R) Not Required
Author Contributions
Conception and design: Paul M. Weinberger, Robert L. Camp, David L. Rimm, Amanda Psyrri
Financial support: Amanda Psyrri
Provision of study materials or patients: Bruce G. Haffty, Clarence Sasaki
Collection and assembly of data: Bruce G. Haffty
Data analysis and interpretation: Paul M. Weinberger, Ziwei Yu, Diane Kowalski, Malini Harigopal, Janet Brandsma, John Joe, Robert L. Camp, David L. Rimm, Amanda Psyrri
Manuscript writing: Paul M. Weinberger, David L. Rimm, Amanda Psyrri
Final approval of manuscript: Paul M. Weinberger, Ziwei Yu, Bruce G. Haffty, Diane Kowalski, Malini Harigopal, Janet Brandsma, Clarence Sasaki, John Joe, Robert L. Camp, David L. Rimm, Amanda Psyrri
GLOSSARY
AQUA (automated quantitative protein expression analysis): This technology overcomes limitations associated with traditional "brown stain" immunohistochemistry (IHC). In traditional IHC, protein expression is reported on a quantized scale such as 0, 1, 2, 3. Biologic material rarely is expressed in such neatly quantized packets, but rather is expressed on a continuous scale. AQUA is a method of computerized interpretation of fluorescence IHC images that allows protein expression to be automatically assigned to subcellular compartment and expressed on a continuous scale.
HPV (human papillomavirus) tumor classification: A proposed system of classifying oropharyngeal tumors based on both HPV16 presence (as indicated by detection of HPV16 DNA) and cellular p16 expression status (as determined by immunohistochemistry or automated quantitative protein expression analysis). Proposed groups include class I (HPV16 negative/p16 nonexpressing), class II (HPV16 positive/p16 nonexpressing), and class III (HPV16 positive/p16 expressing).
qPCR (quantitative polymerase chain reaction): Quantitative polymerase chain reaction (qPCR), also known as real-time PCR, consists of detecting PCR products as they accu-mulate. It can be applied to gene expression quantification by reverse transcription of RNA into cDNA, thus receiving the name of quantitative reverse transcriptase PCR (qRT-PCR). In spite of its name, quantitative, results are usually normalized to an endogenous reference. Current devices allow the simultaneous assessment of many RNA sequences.
Tissue microarray: Used to analyze the expression of genes of interest simultaneously in multiple tissue samples, tissue microarrays consist of hundreds of individual tissue samples placed on slides ranging from 2 to 3 mm in diameter. Using conventional histochemical and molecular detection techniques, tissue microarrays are powerful tools to evaluate the expression of genes of interest in tissue samples. In cancer research, tissue microarrays are used to analyze the frequency of a molecular alteration in different tumor type, to evaluate prognostic markers, and to test potential diagnostic markers.
Viral load: Using qPCR, a starting concentration of template (eg, human papillomavirus) DNA can be determined. If human DNA concentration in the same tumor sample is also determined from a cellular housekeeping gene (eg, -globin), then simple arithmetic can be used to determine viral load per human cell within the tumor sample.
Acknowledgment
We thank Daniel DiMaio, MD, PhD, and Edward Goodwin, PhD, for helpful discussions, and Jim Dziura, PhD, for excellent statistical advice.
NOTES
Supported by the Yale University Institutional Startup Fund (A.P.), the Virginia Alden Wright Fund (C.S.), and the Doris Duke Charitable Foundation (P.M.W.).
Terms in blue are defined in the glossary, found at the end of this article and online at www.jco.org.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
REFERENCES
Andl T, Kahn T, Pfuhl A, et al: Etiological involvement of oncogenic human papillomavirus in tonsillar squamous cell carcinomas lacking retinoblastoma cell cycle control. Cancer Res 58:5-13, 1998
Gillison ML, Koch WM, Capone RB, et al: Evidence for a causal association between human papillomavirus and a subset of head and neck cancers. J Natl Cancer Inst 92:709-720, 2000
Frisch M, Biggar RJ: Aetiological parallel between tonsillar and anogenital squamous-cell carcinomas. Lancet 354:1442-1443, 1999
Swoboda A, Fabrizii V: Tonsillar carcinoma in a renal graft recipient treated with cyclosporine A. Clin Nephrol 39:272-274, 1993
de Villiers EM, Weidauer H, Otto H, et al: Papillomavirus DNA in human tongue carcinomas. Int J Cancer 36:575-578, 1985
Franceschi S, Munoz N, Bosch XF, et al: Human papillomavirus and cancers of the upper aerodigestive tract: A review of epidemiological and experimental evidence. Cancer Epidemiol Biomarkers Prev 5:567-575, 1996
Snijders PJ, Scholes AG, Hart CA, et al: Prevalence of mucosotropic human papillomaviruses in squamous-cell carcinoma of the head and neck. Int J Cancer 66:464-469, 1996
McKaig RG, Baric RS, Olshan AF: Human papillomavirus and head and neck cancer: Epidemiology and molecular biology. Head Neck 20:250-265, 1998
Psyrri A, DeFilippis RA, Edwards AP, et al: Role of the retinoblastoma pathway in senescence triggered by repression of the human papillomavirus E7 protein in cervical carcinoma cells. Cancer Res 64:3079-3086, 2004
Alani RM, Munger K: Human papillomaviruses and associated malignancies. J Clin Oncol 16:330-337, 1998
Serrano M, Hannon GJ, Beach D: A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4. Nature 366:704-707, 1993
Agoff SN, Lin P, Morihara J, et al: P16(INK4a) expression correlates with degree of cervical neoplasia: A comparison with Ki-67 expression and detection of high-risk HPV types. Mod Pathol 16:665-673, 2003
Begum S, Gillison ML, Ansari-Lari MA, et al: Detection of human papillomavirus in cervical lymph nodes: A highly effective strategy for localizing site of tumor origin. Clin Cancer Res 9:6469-6475, 2003
Cairns P, Mao L, Merlo A, et al: Rates of p16 (MTS1) mutations in primary tumors with 9p loss. Science 265:415-417, 1994
Merlo A, Herman JG, Mao L, et al: 5' CpG island methylation is associated with transcriptional silencing of the tumour suppressor p16/CDKN2/MTS1 in human cancers. Nat Med 1:686-692, 1995
Reed AL, Califano J, Cairns P, et al: High frequency of p16 (CDKN2/MTS-1/INK4A) inactivation in head and neck squamous cell carcinoma. Cancer Res 56:3630-3633, 1996
Weinberger PM, Yu Z, Haffty BG, et al: Prognostic significance of p16 protein levels in oropharyngeal squamous cell cancer. Clin Cancer Res 10:5684-5691, 2004
Haraf DJ, Nodzenski E, Brachman D, et al: Human papilloma virus and p53 in head and neck cancer: Clinical correlates and survival. Clin Cancer Res 2:755-762, 1996
Brandwein M, Zeitlin J, Nuovo GJ, et al: HPV detection using "hot start" polymerase chain reaction in patients with oral cancer: A clinicopathological study of 64 patients. Mod Pathol 7:720-727, 1994
Klussmann JP, Weissenborn SJ, Wieland U, et al: Human papillomavirus-positive tonsillar carcinomas: A different tumor entity Med Microbiol Immunol (Berl) 192:129-132, 2003
Strome SE, Savva A, Brissett AE, et al: Squamous cell carcinoma of the tonsils: A molecular analysis of HPV associations. Clin Cancer Res 8:1093-1100, 2002
Li W, Thompson CH, O'Brien CJ, et al: Human papillomavirus positivity predicts favourable outcome for squamous carcinoma of the tonsil. Int J Cancer 106:553-558, 2003
Koskinen WJ, Chen RW, Leivo I, et al: Prevalence and physical status of human papillomavirus in squamous cell carcinomas of the head and neck. Int J Cancer 107:401-406, 2003
Coombs NJ, Gough AC, Primrose JN: Optimisation of DNA and RNA extraction from archival formalin-fixed tissue. Nucleic Acids Res 27:15, 1999
Camp RL, Chung GG, Rimm DL: Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 8:1323-1327, 2002
McDougall JK: Immortalization and transformation of human cells by human papillomavirus. Curr Top Microbiol Immunol 186:101-119, 1994
Braakhuis BJ, Snijders PJ, Keune WJ, et al: Genetic patterns in head and neck cancers that contain or lack transcriptionally active human papillomavirus. J Natl Cancer Inst 96:998-1006, 2004
Dai M, Clifford GM, le Calvez F, et al: Human papillomavirus type 16 and TP53 mutation in oral cancer: Matched analysis of the IARC multicenter study. Cancer Res 64:468-471, 2004
Koch WM, Lango M, Sewell D, et al: Head and neck cancer in nonsmokers: A distinct clinical and molecular entity. Laryngoscope 109:1544-1551, 1999
Ritchie JM, Smith EM, Summersgill KF, et al: Human papillomavirus infection as a prognostic factor in carcinomas of the oral cavity and oropharynx. Int J Cancer 104:336-344, 2003
Klussmann JP, Gultekin E, Weissenborn SJ, et al: Expression of p16 protein identifies a distinct entity of tonsillar carcinomas associated with human papillomavirus. Am J Pathol 162:747-753, 2003
Paz IB, Cook N, Odom-Maryon T, et al: Human papillomavirus (HPV) in head and neck cancer: An association of HPV 16 with squamous cell carcinoma of Waldeyer's tonsillar ring. Cancer 79:595-604, 1997
Ringstrom E, Peters E, Hasegawa M, et al: Human papillomavirus type 16 and squamous cell carcinoma of the head and neck. Clin Cancer Res 8:3187-3192, 2002
Schwartz SR, Yueh B, McDougall JK, et al: Human papillomavirus infection and survival in oral squamous cell cancer: A population-based study. Otolaryngol Head Neck Surg 125:1-9, 2001
Stein GH, Drullinger LF, Soulard A, et al: Differential roles for cyclin-dependent kinase inhibitors p21 and p16 in the mechanisms of senescence and differentiation in human fibroblasts. Mol Cell Biol 19:2109-2117, 1999
Alcorta DA, Xiong Y, Phelps D, et al: Involvement of the cyclin-dependent kinase inhibitor p16 (INK4a) in replicative senescence of normal human fibroblasts. Proc Natl Acad Sci U S A 93:13742-13747, 1996
Mellin H, Dahlgren L, Munck-Wikland E, et al: Human papillomavirus type 16 is episomal and a high viral load may be correlated to better prognosis in tonsillar cancer. Int J Cancer 102:152-158, 2002
Van Tine BA, Dao LD, Wu SY, et al: Human papillomavirus (HPV) origin-binding protein associates with mitotic spindles to enable viral DNA partitioning. Proc Natl Acad Sci U S A 101:4030-4035, 2004
Nuovo GJ, Plaia TW, Belinsky SA, et al: In situ detection of the hypermethylation-induced inactivation of the p16 gene as an early event in oncogenesis. Proc Natl Acad Sci U S A 96:12754-12759, 1999, 百拇医药