The following investigators contributed to the study: Nubia Muñoz, IARC, Lyon, France; Javier Pintos, McGill University, Montreal, Canada; Frank Kee, Queen's University Belfast, Belfast, Northern Ireland, United Kingdom; Leticia Fernández, Instituto Nacional de Oncología y Radiobiología, Havana, Cuba; Ali Idris, Toombak and Smoking Research Center, Khartoum, the Sudan; María José Sánchez, Escuela Andaluza de Salud Pública, Granada, Spain; Adoración Nieto, Facultad de Medicina, Seville, Spain; F. Xavier Bosch, Institut Català d'Oncologia, Barcelona, Spain; Renato Talamini, Centro di Riferimento Oncologico di Aviano, Aviano, Italy; Alessandra Tavani, Instituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy; Ulrich Reidel, Deutsches Krebsforschungszentrum, Heidelberg, Germany; Jolanta Lissowska, Cancer Center, Warsaw, Poland; Barbara Rose, Sydney Head and Neck Cancer Institute, Royal Prince Alfred Hospital, Sydney, Australia; Hema Sridhar, Kidwai Memorial Institute of Oncology, Bangalore, India; Prabha Balaram, Regional Cancer Centre, Trivandrum, India; Thangarajan Rajkumar, Cancer Institute (WIA), Chennai, India.
Numerous studies have provided consistent evidence that human papillomavirus (HPV), the necessary cause of cervical cancer, is present in tumor biopsies from approximately 20-50% of oropharyngeal squamous cell carcinomas (SCCs) and a smaller subset of oral SCCs.1, 2, 3, 4 Among HPV DNA-positive oropharyngeal SCCs, 90% are positive for HPV16.1, 2, 3, 4 Nonetheless, HPV DNA detection in tumor biopsies may not be sufficient evidence of causation. HPV16 DNA from tumor specimens analyzed jointly with markers of expression of the viral oncogene E6, mutational patterns of the cancer suppressor gene TP53 and levels of allelic loss, have helped identify a subset of these cancers that may be the consequence of HPV infection.1, 2, 5, 6, 7, 8
HPV viral load, a measure of the amount of viral DNA in biopsy specimens, alone or in conjunction with well-characterized HPV serologic assays, may clarify the role of HPV among oral and oropharyngeal cases. Antibodies against HPV E6 and E7 are markers of an invasive HPV-associated malignancy9, 10 and are rarely present among individuals with HPV DNA-negative oral and oropharyngeal tumors.1 Antibodies against HPV virus-like particles (VLPs) are considered a marker of cumulative, lifetime HPV infection,11, 12, 13, 14, 15 and are associated with HPV-related disease, but not as strongly as E6 and E7 antibodies. While these markers do not allow for inferences on causality, evaluation of the associations between high and low viral load with HPV16 serologic markers among HPV16 DNA-positive and -negative oral and oropharyngeal SCCs may delineate the subset more likely the result of HPV infection. Viral load assessment may also compensate for the less than optimal sensitivity in each of the HPV serologic markers currently available.
Material and methods
Oral and oropharyngeal SCC cases from the International Agency for Research on Cancer (IARC) multinational case-control study of HPV and oral cancer were selected for this analysis. The methods and main results of the parent study have been previously reported.1 Briefly, incident cases with oral and oropharyngeal SCC were recruited from referral centers and hospitals in 9 countries (Australia, Canada, Cuba, Italy, India, Northern Ireland, Poland, Spain and Sudan) from 1996 to 1999. Following informed consent, biologic samples were collected prior to cancer treatment and included cancer biopsies and serologic samples. Specimens were stored at −70 or −40°C (depending on study site) and shipped to designated laboratories for analysis.
Only cases with HPV DNA results from beta-globin positive tumor biopsy specimens were included (n = 908); cancer-free controls were not used in the present analysis. Cases positive for high-risk HPV types other than HPV16 were excluded (n = 3) because only antibodies against HPV16-specific VLP, E6 and E7 were available. Our final analytic population consisted of 53 HPV16 DNA-positive oral (n = 28) and oropharyngeal (n = 25) SCCs and 852 HPV DNA-negative cases.
HPV DNA and viral load analysis
The general primer-mediated GP5+/6+ polymerase chain reaction enzyme immunoassay (PCR-EIA) was used to detect and type 20 HPV types6 from biopsy material.1 Of the 53 HPV16-positive SCCs included in this analysis, the PCR-EIA optical density (OD) values obtained after a 1-hr substrate incubation were used to assess the relative amount of HPV16 DNA.16 A previous study demonstrated that, within the range of 1 × 10 to 1 × 106 genome equivalents, the EIA OD value obtained after 1 hr of substrate incubation shows a linear relation with the amount of input DNA.17 Consequently, EIA ODs can semiquantitatively assess the relative viral load.
Detection of antibodies against HPV16 VLPs
Plasma samples were tested for antibodies against HPV16 VLPs by ELISA18 and results were dichotomized as seropositive or seronegative as described previously.1
Detection of antibodies against HPV16 E6 and E7
Antibodies against HPV16 E6 and E7 proteins were detected in an ELISA that utilizes the glutathione S-transferase (GST) capture method with bacterially-expressed full-length E6 and E7 fused to GST as the antigens.19 The assay cutoff point for seropositivity was set and samples were dichotomized as seropositive or seronegative, as described previously.1
The minimum, maximum and median values of HPV16 viral load were described for the 53 HPV16 DNA-positive cases. The median values of viral load for oral and oropharyngeal cases were compared by use of the nonparametric Mann-Whitney test.
HPV16 DNA-positive cases were dichotomized as “high” and “low” viral load based on the median viral load value (OD = 1.92) from the 53 HPV16 DNA-positive cases. Three categories were therefore established for analysis: (i) HPV DNA-negative cases (n = 852), (ii) HPV16 DNA-positive cases with low viral load (n = 27) and (iii) HPV16 DNA-positive cases with high viral load (n = 26).
Tumor site and stage,20 seropositivity to HPV16 VLPs, E6 and E7, and a combined measure of seropositivity to E6 and/or E7, were evaluated by use of maximum-likelihood multinomial (polytomous) logistic regression to estimate odds ratios (ORs) and corresponding 95% confidence intervals (CIs). HPV DNA-negative cases served as the reference category and the high and low viral load groups were each compared to the reference category. Age (in 10-year categories), tumor site, tobacco use (cigarettes per day), alcohol use (drinks per day) and seropositivity to HPV16 serologic markers beyond the 1 in question, were each investigated as possible confounders in adjusted multinomial logistic regression models.
Among the HPV16 DNA-positive cases, viral load OD values ranged from 0.27 to 2.53; the median value was 1.92 (Fig. 1). The median viral load was significantly higher among cases with oropharyngeal SCC (OD = 2.40) compared to oral SCC (OD = 1.59; p = 0.006). Compared to oral cancers, cancers arising in the oropharynx, the head and neck cancer site most commonly associated with HPV infection, were more than twice (95% CI, 1.1–6.2) as likely to have low viral load and 12 times (95% CI, 5.2–27.5) as likely to have high viral load (Table I). Large tumor size,20 a possible confounding factor for viral load in tumor specimens, was not associated with HPV DNA-positivity or viral load (Table I).
Table I. Tumor-Related Classifications and HPV16 Serologic Markers Among HPV16 DNA-Positive Cases with Low and High Viral Load Compared to HPV DNA-Negative Cases*
Serologic biomarkers of HPV16 were assessed among HPV DNA-negative, HPV16 DNA-positive cases with low viral load, and HPV16 DNA-positive cases with high viral load. HPV16 VLP seroprevalence significantly increased from 9.7% of HPV DNA-negative cases to 22.2% of low viral load HPV16 DNA-positive cases (OR adjusted for tumor site [AdjOR], 2.7; 95% CI, 1.1–6.9) to 60.0% of high viral load HPV16 DNA-positive cases (AdjOR, 14.6; 95% CI, 6.0–35.6; p for trend <0.001). Compared to HPV DNA-negative cases, low viral load was not significantly associated with E6 seropositivity. However, high HPV viral load significantly increased the odds of HPV16 E6 seropositivity 57-fold (AdjOR, 57.6; 95% CI, 21.4–155.3) (Table I). High viral load was also significantly associated with HPV16 E7 seropositivity (AdjOR, 25.6; 95% CI, 9.3–70.8) compared to HPV DNA-negative cases (Table I). Adjustment for age (in 10-year categories) did not affect the magnitude or significance of the associations of viral load with tumor site, grade, or HPV16 serologic markers (data not shown).
To determine if viral load assessment may contribute to defining the subset of HPV-positive oral and oropharyngeal cancers likely due to HPV infection, we investigated associations among HPV16 viral load and established viral markers, namely, HPV16 VLP, E6 and E7 seropositivity. High viral load among HPV16 DNA-positive oral and oropharyngeal cancers was both highly and significantly associated with these serologic biomarkers compared to HPV DNA-negative cases. In contrast, low viral load was only modestly associated with serologic markers, and these associations often did not attain statistical significance.
HPV16 VLP seropositivity was elevated 14-fold among cases with high viral load, but only 3-fold among cases with low viral load, compared to HPV DNA-negative cases. These associations suggest that VLP seropositivity, a marker of past HPV infection,11, 12, 13, 14, 15 rather than HPV-associated tumorigenesis, may actually be the result of HPV exposure that preceded, and possibly contributed to, HPV-related carcinogenesis. More importantly, seropositivities for HPV16 E6 or E7, recognized markers of an HPV-associated malignancy,9, 10 were remarkably higher among cases with high viral load compared to HPV DNA-negative cases. Again, the corresponding associations for low viral load cases with E6 or E7 seropositivities were modest. While the serologic assays are not site-specific and may be the result of HPV infection and cancer at other mucosal sites, the magnitude of the associations are unlikely to be a result of unidentified tumors rather than tumors of the head and neck. The marked difference between the associations of serologic markers with low compared to high viral load provides evidence that high viral load in biopsy specimens, in conjunction with seropositivity to markers of an HPV-associated malignancy, may identify a subset of these cancers in which HPV is biologically active.
Large studies on cervical smears suggested that persistent HPV infection,21 cytological progression,21, 22, 23, 24 and the risk of a high-grade cervical lesion21, 22, 23, 24, 25 are each associated with high HPV viral load in the cervix. As a consequence, viral load assessment was proposed to distinguish clinically relevant HPV infections in the cervix,26 yet has only been assessed in a few studies of sites beyond the cervix. Indeed, 1 study showed a markedly higher viral load in HPV DNA containing oral and oropharyngeal SCCs with E6 mRNA expression than in those without.6 Another study revealed that viral load was significantly higher among HPV DNA-positive cancers occurring in the tonsil, a subsite of the oropharynx, compared to HPV DNA-positive nonoropharyngeal cancers.27
Most of these studies employed the gold-standard for HPV viral load assessment: real-time PCR.27, 28 However, based on this study, it seems that the less costly, more feasible semiquantitative method may also have utility in epidemiologic studies in which laborious real-time PCR assays may not be possible. Important limitations may exist for both semiquantitative and quantitative viral load assessment. The amount of cancer cells in a biopsy specimen may impact viral load quantitation, and substantial variability may exist in viral load measurements within a single tumor.29 Additionally, it remains uncertain whether the measurement of viral load is the result of a few cells with greater number of HPV DNA copies, or many cells with few copies. Nonetheless, it seems that estimated viral load might contribute to defining the subset of HPV-positive oral and oropharyngeal cancers that are more likely the result of HPV infection. Varying the cutoff point for high vs. low viral load did not appreciably affect the results (for receiver operating characteristic [ROC] curve analysis of viral load OD predicting E6/E7 seropositivity, the area under the curve was 0.81).
While we would not advise the use of a semiquantitative viral load assay for predictive value on an individual basis, in the absence of a quantitative assay, this estimate of the actual viral load may importantly contribute to population-based studies on the etiology of HPV-associated oral and oropharyngeal cancer.