Materials and methodsThis study was approved by the Centers for Disease Control and Prevention (CDC) and the Institutional Review Boards of the University of Hawaii and the University of Iowa. This evaluation was part of a larger study by the CDC partnering with seven U.S. population-based cancer registries to examine the distribution of HPV genotypes in anogenital and head and neck cancers diagnosed in the U.S. prior to the implementation of the prophylactic HPV vaccine [6]. The present study was confined to two cancer registries, the Hawaii Tumor Registry and the Iowa Cancer Registry for which additional archival specimens were available for assessment. Both registries are part of the National Cancer Institute’s (NCI) Surveillance, Epidemiology, and End-Results (SEER) program and operate affiliated Residual Tissue Repositories (RTR) [12] comprised of collections of de-identified, formalin-fixed, paraffin-embedded (FFPE) malignant tumor tissue specimens annotated with de-identified, high-quality demographic, clinical, pathologic, and survival data.
Invasive oral cavity cancer was defined based on the International Classification of Diseases for Oncology Version 3 [13]. Oral cavity cancer cases were selected from patients diagnosed in 1993–2004 in Hawaii and Iowa with histologically-confirmed malignancies of the oral tongue (including dorsal, ventral, and anterior 2/3 surfaces and border/ tip) (ICD-O-3 C020-C023); gingiva (upper and lower) (C030-C031); floor of the mouth (including anterior and lateral regions) (C040-C041; C049); hard palate (C050); cheek mucosa (C060); vestibule of the mouth (C061); retromolar area (C062); and overlapping and unspecified parts of mouth (C068). Oropharyngeal subsites were excluded including the base of tongue, tonsils, soft palate, and uvula.
HPV genotyping was conducted at the CDC laboratories. FFPE tumor tissue specimens were prepared following a uniform protocol to avoid cross-contamination. A representative tumor block from each case was selected and sectioned using a new disposable blade for each sample. The first and last tissue sections were stained with hematoxylin and eosin (H&E) and reviewed by a study pathologist (E.R.U.) to confirm the presence of malignancy. DNA was isolated from intervening FFPE tumor sections using high temperature assisted tissue lysis [14] and automated DNA purification with a Chemagic MSM1 (PerkinElmer, Waltham, MA, USA), HPV testing was performed as previously described [15]. Briefly, all samples were tested using the Linear Array HPV Genotyping Test (LA, Roche Diagnostics, Indianapolis, IN), a PCR-based assay targeting a 450-bp consensus region of the L1 HPV genome and detecting 37 HPV genotypes (6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84, 89, IS39). Amplification of the human β-globin gene was included as an internal control. Samples testing negative for β-glohin or HPV in the Roche assay were subsequently tested using the INNO-LiPA HPV Genotyping Assay (LiPA, Innogenetics, Gent, Belgium), a PCR-based assay targeting a 65 base pair region of the HPV L1 gene. INNO-LiPA detects 26 of the same types covered in the Roche assay in addition to 3 additional types (HPV 43, 44, 74). This second assay was included to account for the fixation-related degradation of DNA extracted from FFPE tissue. Specimens testing negative for HPV and the internal control in both assays were considered inadequate for evaluation.
Immunohistochemical evaluation of pl6INK4a was conducted at the University of Hawaii Cancer Center Pathology Core laboratory using tumor tissue obtained from the same tumor specimen used for HPV genotyping. A pl6INK4a mouse monoclonal antibody (clone JC8; Santa Cruz Biotechnology, Santa Cruz, CA, USA; dilution 1:400) was used according to the manufacturer’s specifications. A study pathologist (O.C.), who was blinded to the HPV status of cases, evaluated pl6INK4a based on staining intensity (mild/weak, moderate, strong), intracellular localization (nuclear, cytoplasmic), and the proportion of tumor cells stained. Specimens exhibiting moderate to strong staining of the nucleus and cytoplasm in ≥ 70% of tumor cells were considered definitively positive for pl6INK4a based on established criteria [16,17].
A subset of oral cavity cancer cases testing positive for HPV DNA were evaluated for genotype-specific HPV E6/E7 mRNA via real-time quantitative PCR in the laboratory of Dr. Maura Gillison at Ohio State University according to previously described methods [18]. Briefly, purified tumor RNA was reverse transcribed to cDNA and evaluated for HPV E6/E7 mRNA. All samples were tested for HPV 16 mRNA and individually tested for other carcinogenic types based on the corresponding genotypes detected (HPV 18, HPV 39, HPV 51, HPV 52, and HPV 66). The human sapiens ribosomal protein large P0 gene (RPLP0) was used as an internal control.
Statistical analyses were conducted using SAS version 9.4. Overall HPV prevalence was based on the detection of one or more HPV genotypes in tumor tissue. Multiple genotypes detected in a case were not counted more than once in overall prevalence estimates. HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68 were classified as carcinogenic, or high-risk genotypes [19,20]. All other genotypes were considered to be non-carcinogenic, or of low or undetermined risk. Single year age, which was suppressed as part of de-identification for over half of cases, was imputed to the mid-point of 5-year age groups. Chi-square and Kappa statistics were used for group comparisons. Univariate and multivariate unconditional logistic regression were used to evaluate the association of case characteristics with HPV status. Overall survival, which was based on the period from date of cancer diagnosis to date of death or date of last follow-up, was evaluated using the Kaplan-Meier method, the Log-rank test, and Cox proportional hazards regression. All tests were two-sided and a p-value < 0.05 was considered as statistically significant.
ResultsStudy populationTumor tissue specimens from 168 cases of invasive oral cavity cancers cases were initially available for the study. Of these cases, 21 were excluded due to absence of tumor in the tissue specimen (n = 14) or sample insufficiency (i.e., negative PCR result for both HPV and the internal control) (n = 7). Of the remaining 147 cases for which HPV DNA results were available, tissue was available to evaluate pl6INK4a for 122 cases, which constituted the study population for the present evaluation.
Men and women comprised 49% and 51%, respectively, of oral cancer patients (Table 1). Fifty-nine percent were age 60 years and older at diagnosis. Cases included whites (55%), Asians (31%), and other groups (14%). Sixty-two percent were diagnosed in 2000–2004. Tumors of the oral tongue comprised 48% of cases, floor of the mouth 21%, gingiva 10%, hard palate 5%, and unspecified or overlapping sites 16%. Other oral subsites including tumors of the cheek mucosa, vestibule of the mouth, and retromolar area were not represented in the study population. Sixty-three percent of cases were of moderately differentiated grade. Keratinizing squamous cell carcinoma (SCC) was the most common histological subtype comprising 80% of tumors. Early stage (localized) tumors constituted 56% of cases.
HPV DNAHPV DNA was detected in 31% (38/122) of oral cavity tumors. Seven different genotypes were detected including HPV 16 which was detected in 22% (27/122) of cases. Other carcinogenic genotypes included HPV 18, HPV 39, HPV 51, HPV 52, and HPV 66. HPV 6 was the sole non-carcinogenic type detected. Thirty-five of the 38 HPV-positive tumors were positive for a single genotype; 3 cases were positive for multiple genotypes.
HPV DNA positivity did not vary by sex, age, race, tumor grade, histology, or stage. Differences by HPV DNA status were observed by year of diagnosis and subsite (Table 1). HPV DNA was detected in tumors of 19% (9/47) of patients diagnosed in 1993–1999 compared to 39% (29/75) of cases diagnosed in 2000–2004 (p = 0.02). The highest HPV DNA prevalence was in the gingiva (67%, 8/12) followed by the floor of the mouth (38%, 10/26), oral tongue (29%, 17/59), hard palate (17%, 1/6), and overiapping/unspecified sites (11%, 2/19) (p = 0.02).
P16<sup>INK4a</sup>expressionOverall, pl6INK4a was expressed in 30% (36/122) of tumors characterized by moderate or strong staining of the cytoplasm and nucleus in ≥ 70% of tumor cells (Fig. 1). pl6INK4a expression was observed in 24% (9/38) of HPV DNA positive tumors and 32% (27/84) of HPV DNA negative tumors (p = 0.34). pl6INK4a expression was poorly correlated with overall HPV DNA and HPV 16 and/or 18 (Kappa < 0.1). Positivity for both pl6INK4a and HPV DNA was observed in 7% (9/122) of cases. Of the 9 cases positive for both pl6INK4a and HPV DNA, 8 were solely positive for HPV 16 and 1 was positive for HPV 18.
Joint HPV16/18 and pl6INK4a status significantly varied by oral cavity subsite. Positivity for both markers was observed in 33% (4/12) of tumors of the gingiva compared to 0% (0/6)-8% (2/26) for other sites (p = 0.01) (Fig. 2). HPV 16/18 and pl6INK4a status also varied by year of diagnoses: None of the 47 cases diagnosed in 1993–1999 were positive for both markers compared to 12% (9/75) of cases diagnosed in 2000–2004 (p = 0.01).
HPV E6/E7 mRNAHPV 16 E6/E7 mRNA was evaluated in 106 of the 122 samples. All were negative for HPV 16 E6/E7 mRNA. The 11 cases positive for HPV 18, HPV 39, HPV 51, HPV 52, and/or HPV 66 were individually tested for E6/E7 mRNA of the corresponding genotypes. All samples were negative. The housekeeping gene, RPLP0, was low in all 106 samples which may be indicative of low levels of RNA. As the RNA quality from these archival FFPE tumor samples may have reduced the ability to detect viral E6/E7 mRNA, these results were not considered in the determination of HPV status.
Subgroup comparisonsTumor subsite differences were further evaluated (data not shown). Patients with gingiva tumors were older (ages ≥ 60: 92%, 11/12) compared to those with tumors of other oral sites (ages ≥ 60: 55%, 61/ 110) (p = 0.02). Cancers of the gingiva were comprised of a greater proportion of advanced tumors (92%, 11/12) compared to other oral cavity subsites (39%, 40/104, respectively) (p < 0.001). The distribution of subsites did not significantly differ by sex, race, calendar time, tumor grade, or histology. When HPV 16/18 and pl6INK4a were examined in multivariate logistic regression models that included tumor subsite, age, and stage, the associations of gingival tumors and age with HPV markers remained significant (p < 0.05) (data not shown). There was no significant difference in the proportion of oral tumor specimens requiring testing with the LiPA assay for cases diagnosed before 2000 (30%, 14/67) and those diagnosed in 2000–2004 (38%, 29/75) (p = 0.32), suggesting that the observed differences in HPV detection by time period were not attributable to possible degradation of nucleic acid in older specimens.
Overall survivalOverall survival of oral cavity cancer cases did not vary by HPV DNA positivity (any HPV DNA, HPV 16 DNA, HPV 16/18), p16INK4a expression, or joint positivity for both markers (Log-rank p-value > 0.05) (data not shown). None of these markers were associated with overall survival in multivariate models adjusting for age, stage, tumor subsite, and diagnosis year (data not shown).
DiscussionOur study of oral cavity cancers from two U.S. registry-based populations demonstrates wide variability in the estimates of the etiologic role of HPV depending on the measures utilized. The strongest evidence for the etiologic role of HPV was restricted to only 7% of cases based on positivity for HPV DNA 16 and/or 18 accompanied by elevated expression of pl6INK4a. In contrast, based on detection of either HPV DNA 16 and/or 18 or p16 INK4a alone, our findings suggests the potential for HPV involvement in nearly a third of cases. Our findings are in general agreement with prevalence estimates from international pooled studies which range from 3.5% for joint HPV DNA - pl6INK4a [21] to 24% and 28% for HPV DNA and pl6INK4a, respectively [22,18].
HPV DNA and pl6INK4a expression were poorly correlated in oral cavity tumors. PCR-based detection of viral DNA and pl6INK4a expression in tumor tissue each has limitations as measures of viral exposure. HPV DNA measured through PCR has high sensitivity for the presence of viral DNA but cannot distinguish between HPV acting oncogenically within tumor and productive HPV in adjacent benign tissue [23]. Conversely, pl6INK4a is a non-specific measure of viral exposure. E7 inactivation of pRb results in overexpression of pl6INK4a in HPV carcinogenesis [10]. High expression of pl6TNK4a can also result from cellular alterations unrelated to infection [23]. In HPV carcinogenesis, pl6INK4a expression may be lost with increasing host genetic and epigenetic chromosomal instability [21], which may, in part, explain the poor correlation with HPV DNA in oral cavity tumors.
In contrast to oral cavity tumors, pl6INK4a and HPV DNA in oropharyngeal cancer are generally correlated [11]. However, as with oral cavity tumors, HPV DNA detection alone may also overestimate viral involvement in oropharyngeal cancers. In a prior evaluation of oropharyngeal cancers identified through seven U.S. cancer registries, including the two included in the present analysis, overall HPV DNA prevalence was 70% [24]. When combined with pl6INK4a or other markers such as HPV E6/E7 mRNA, the attributable fraction of HPV in oropharyngeal cancer is lower but still substantially higher than for oral cavity cancer [21,22]. HPV E6/E7 mRNA, which was negative for the subset of cases tested, was not considered in our final evaluation given the indications of low quality of RNA yielded from these archival FFPE samples. We observed no association of HPV status, including viral DNA and/or pl6INK4a, with overall 5-year survival in oral cavity cancer. This contrasts with oropharyngeal cancer for which tumor HPV is established as a predictor of favorable survival [8,9].
Our findings also provide evidence for a limited, albeit increasing, role of HPV in cancers of the oral cavity. The joint prevalence of HPV 16/18 and pl6INK4a in oral tumors increased from null to twelve percent over the 11-year period of cancer diagnoses. The observed increase in HPV-associated oral cavity cancers over calendar time parallels trends for oropharyngeal cancers for which the increasing incidence in the U.S. appears to be largely driven by HPV. In a study which included tumor samples from two SEER registries included in the present analysis, Chaturvedi et al. [7] reported an increase in HPV-positive or opharyngeal cancers in the U.S. over time but no increase in HPV-negative tumors. The increase in HPV-associated oropharyngeal cancers in the U.S. over time is more pronounced in men than in women and may be attributed to increasing oral HPV exposure through sexual behaviors [25]. A limitation of the present analysis is the lack of information on tobacco and alcohol use, as well as sexual history. HPV-negative head and neck cancers are more closely associated with tobacco and alcohol use, while HPV-positive tumors are related to sexual exposures [26,27].
Although overall prevalence of joint positivity for HPV DNA 16/18 and pl6INK4a was low, there was significant variation across subsites. Positivity for both markers was highest in the gingiva, in which 33% of tumors were jointly positive. Our findings are consistent with a metaanalysis reporting pooled HPV DNA prevalence estimates to be higher in the gingival tumors compared to the tumors of the tongue, floor of the mouth, and hard palate [22]. We observed that joint HPV 16/18 - pl6INK4a positivity for other oral cavity tumors - including the oral tongue, hard palate, and floor of the mouth - was substantially lower than the gingiva and did not exceed 8%. Tumors of the oral tongue, the most common cancer of the oral cavity, have experienced increasing incidence in the U.S. for several decades, particularly among young whites and are not attributed to changes in exposure to HPV, alcohol, or tobacco [4] suggesting other yet unrecognized risk factors. The bacterial microbiome varies across subsites of the oral cavity. In particular, the gingiva, including the sub-gingiva and supra-gingiva, have been shown to have the most distinct bacterial composition within oral cavity subsites, including the buccal mucosa, hard palate, and tongue [28]. The particular susceptibility of the gingiva to bacterial infection is well-established, clinically manifesting as gingivitis and periodontal disease [29], Whether the gingiva is more susceptible to HPV infection is not known. Given the limited number of tumor samples for distinct subsites, including the gingiva, the confirmation of these findings in larger study populations is needed.
The strengths of our study, including a relatively large study population derived from two population-based cancer registries, are countered by a number of limitations including the use of retrospectively-collected, archival FFPE tissue. Variation in tissue preservation with sample age and fixation conditions may have influenced the detection of HPV DNA and pl6INK4a via PCR and IHC, respectively.
We conclude that HPV is etiologically linked to a limited subset of oral cavity cancers overall, but its relative contribution may be substantial for certain subsites. Future research focusing on specific subsites, including the gingiva, are needed, and prospectively collection of population-based samples would be ideal. Confirmation of the increasing role of HPV in oral carcinogenesis in the U.S. is critical for informing prevention strategies.