Oropharyngeal squamous cell carcinomas (OPSCC) that are associated with human papilloma virus (HPV) infection carry a more favorable prognosis than those that are HPV-negative. However, it remains unclear which biomarker(s) can reliably determine which OPSCC specimens are truly driven by HPV infection. In this study, we analyzed 199 fresh-frozen OPSCC specimens for HPV DNA, viral load, RNA expression patterns typical for cervical carcinomas (CxCaRNA+), and the HPV-targeted tumor suppressor protein p16INK4a as markers for HPV infection. In this set of specimens, there was a 49% prevalence of DNA for the cancer-associated HPV type 16 (HPV+). However, there was only a 16% prevalence of high viral load and only a 20% prevalence of CxCaRNA+, a marker of HPV16 carcinogenic activity. Among the CxCaRNA+ tumors, 78% of the specimens exhibited overexpression of p16INK4a, which also occurred in 14% of the HPV-negative tumors. Using a multivariate survival analysis with HPV negativity as the reference group, CxCaRNA+ as a single marker conferred the lowest risk of death [HR = 0.28, 95% confidence interval (CI), 0.13–0.61] from oropharyngeal cancer, closely followed by high viral load (HR = 0.32, 95% CI, 0.14–0.73). In contrast, a weaker inverse association was found for OPSCC that were HPV+ and p16INK4a high (HR = 0.55, 95% CI, 0.29–1.08). In summary, our findings argued that viral load or RNA pattern analysis is better suited than p16INK4a expression to identify HPV16-driven tumors in OPSCC patient populations. Cancer Res; 72(19); 4993–5003. ©2012 AACR.

See commentary by Méndez, p. 4896

Specific oncogenic types of human papilloma viruses (HPV), most frequently HPV type 16, are causally associated with a subset of oropharyngeal squamous cell carcinomas (OPSCC) (1–4). There are some distinct differences between HPV-negative and HPV-positive OPSCC regarding gender, age, tumor histology, size and lymph node involvement, and also in lifestyle and sexual behavior of the patients (2, 5–11). Of interest, patients with HPV-positive OPSCC have an improved survival compared with patients with HPV-negative OPSCC (1, 2, 5, 10, 12–14).

In several studies, HPV DNA–positive OPSCC were found to be heterogeneous in both biological (e.g., in tumor histology, genetic status of p53, extent of cytogenetic alterations, expression of cell-cycle–related proteins) and clinical behavior (disease progression and patient survival; refs. 1, 3, 9, 15–21). Thus, the markers or marker combinations best suited to identify OPSCC with an active HPV involvement (HPV-driven OPSCC) are still unclear.

HPV DNA detection by in situ hybridization (ISH) theoretically should be best suited to identify HPV-driven tumors. However, ISH was found to lack sufficient sensitivity (22). In contrast, HPV DNA detection by quantitative PCR (qPCR) may lack disease specificity, but it allows precise determination of viral load. Viral load in turn can directly be compared with tumor biology and clinical outcome (8, 9). Overexpression of p16INK4a, a consequence of the inactivation of the cellular retinoblastoma protein (pRb) by the viral oncoprotein E7, was proposed to be a surrogate marker for active HPV involvement in OPSCC carcinogenesis (8, 22, 23). However, high p16INK4a alone also has insufficient sensitivity (13, 24–26) and specificity (22, 25–27). It was therefore suggested to use HPV DNA positivity in conjunction with high p16INK4a to detect HPV-driven OPSCC with increased specificity (14, 26, 27). Because the oncogenic potential of high-risk HPV is mediated through continuous expression of the 2 major viral oncogenes E6 and E7 (28, 29), E6 and/or E7 RNA expression (in the lack of suitable protein reagents) indeed has shown good value as marker for active HPV (13, 16, 20, 21).

In HPV16-positive cervical carcinomas (CxCa) and cervical high-grade intraepithelial lesions, 2 characteristic RNA patterns have been found: upregulated E6*II (226⁁526) and/or E6*I (226⁁409) transcripts with a concurrently low or absent expression of the E1⁁E4 (880⁁3358) transcript (high E6*/E1⁁E4) and/or upregulated E1C (880⁁2582) transcript with a concurrently low or absent L1 (3632⁁5639) transcript (high E1C/L1; ref. 30). Whether HPV-positive OPSCC express viral RNA patterns as in CxCa is unclear. Also, viral load and expression of viral RNA patterns have not been directly compared with regard to clinical behavior and patient survival.

In this large German OPSCC case series (n = 199), the HPV DNA–positive tumors were analyzed in detail for viral load, viral oncogene RNA (E6*II and/or E6*I), and CxCa-like viral RNA patterns (CxCaRNA). We hypothesized that CxCa-like viral RNA patterns should be the best marker to identify truly HPV-driven OPSCC and, therefore, these patterns should be associated with the most favorable patient survival. It was further of interest to examine whether high p16INK4a expression is concordant with these CxCa-like viral RNA patterns.

Patients, tissue samples

Of the patients diagnosed with primary OPSCC and treated at the ENT Department of the University Hospital Heidelberg, Germany, between 1990 and 2008, 199 gave informed consent and provided fresh-frozen tumor biopsies. The study was approved by the Ethics Committee of the Medical Faculty of the University of Heidelberg, study code 176/2002. Biopsies were directly snap-frozen in precooled isopentane/liquid nitrogen and stored at −80°C until further processing. Patient and tumor characteristics were obtained from clinical records. Management decisions for the patients were not guided by knowledge of HPV status or p16INK4a expression. Information on first-line treatment modality was available for 184 patients. In 125 patients, surgery was applied, and in 59 patients with poor performance status or with nonresectable tumors, radiation and/or chemotherapy were applied first.

For nucleic acid extractions, cryosections of 16-μm thickness yielding between 2 and 10 mg of tissue were cut, homogenized in liquid nitrogen, and stored at −80°C.

After each biopsy, the cryostat was cleaned with acetone and ethanol and new gloves, forceps, and blades were used. For each homogenization, a fresh pistil was used. As negative controls, 2 mucosal biopsies from healthy patients were processed in the same way. To verify the presence and to estimate the content of tumor cells in cryosections, adjacent sections were stained with hematoxylin and eosin. The mean tumor cell content was 55% (range, 25%–90%).

Multiplex HPV genotyping and real-time qPCR for viral load determination

DNA was isolated using Qiagen's QIAamp DNA Mini Kit. Twenty-seven mucosal HPV types were analyzed simultaneously by BSGP5+/6+-PCR/MPG as described (31–33). The BSGP5+/6+-PCR/MPG assay comprises the BSGP5+/6+-PCR, which homogenously amplifies all known genital HPV types generating biotinylated amplimers of approximately 150 bp from the L1 region (32, 33) and a multiplex HPV genotyping (MPG) assay with bead-based xMAP Luminex suspension array technology, which is able to simultaneously identify 27 HPV types and the β-globin gene (31, 32). The assay coamplifies cellular β-globin sequences as internal DNA quality control. Quantification of HPV16 signals was accomplished as follows: to yield a relative HPV16 signal (in%), each measured HPV16 signal was first divided by the highest HPV16 signal observed among all tumor samples; the resulting relative HPV16 signal was then divided by the β-globin signal of the same sample to give a normalized viral load value (%HPV signal/β-globin signal). A cutoff between low and high viral load of 0.0007 units was chosen. This high viral load cutoff has been developed for high-risk HPV types and optimized to distinguish cervical smears with normal cytology from those with abnormal cytology (manuscript in preparation).

Tumors positive for HPV16 DNA by BSGP5+/6+-PCR/MPG (HPV+) were further analyzed using real-time qPCR targeting HPV16-specific sequences of the E6 gene to obtain quantitative viral load values and an objective high viral load cutoff. qPCR for β-globin was used to verify the quality of DNA in the sample and to measure the amount of input DNA. Viral load in each specimen was expressed as the number of HPV genome copies per cell. Because on average the specimens contained approximately 50% tumor cells and at least 1 genome copy per cell is expected in HPV-transformed cells, we defined 0.5 copies per cell as cutoff for a high viral load (HPVhigh) and OPSCC below this cutoff had a low viral load (HPVlow). Primer sequences used for real-time qPCR are available upon request (manuscript in preparation).

HPV16 RNA analysis

HPV16 DNA–positive OPSCC determined by BSGP5+/6+-PCR/MPG were analyzed for viral RNA expression. RNA was isolated using Qiagen's RNeasy Mini Kit. DNase I digestion (Qiagen) was included to ensure an exclusive amplification of RNA. Nucleic acid sequence-based amplification and hybridization to splice-specific probes on Luminex beads were carried out as described (30). Any E6*II and/or E6*I signal above background was considered positive (RNA+). For quantification of E6*II RNA, a calibrator RNA was competitively coamplified in the same reaction and used to normalize the E6*II signals. E6*I copy numbers in E6*II-negative samples were estimated by the same calibrator RNA. Tumors that were positive for E6*I and/or E6*II (RNA+) were further examined quantitatively for viral transcripts E1C (880⁁2582), E1⁁E4 (880⁁3358), and L1 (3632⁁5639) using their respective calibrators. For absolute quantification, each experiment included external standard dilution series of in vitro RNA. Transcript ratios were calculated for E6*II or E6*I over E1⁁E4 and for E1C over L1. Tumors with either an E6*/E1⁁E4 ratio of >1.5 (high E6*/E1⁁E4) and/or an E1C/L1 ratio of >0.003 (high E1C/L1) defined the CxCa-like viral RNA pattern-positive OPSCC (CxCaRNA+). OPSCC below both of these cutoffs defined the tumors without viral RNA patterns (CxCaRNA). RNA of 1 healthy mucosa and of 1 HPV tumor served as negative controls. All RNA samples were positive for the ubiquitin C transcript and the controls were negative for all viral transcripts.

p16INK4a immunohistochemistry on tissue microarray

OPSCCs (188 of 199) of this study were assembled on a tissue microarray (TMA) and were stained for p16INK4a using 2 monoclonal antibodies, MTM-E6H4 (MTM Laboratories AG) and DSC-106 (Progen Biotech). For a detailed description of TMA preparation and immunohistochemistry (IHC), see refs. 34, 35. Tissue cores were evaluated by D. Holzinger and F.X. Bosch without knowledge of patient identities, clinical parameters, and HPV status. For tissue cores with discordant primary reading, a final decision was reached after joint re-evaluation and discussion. Scoring p16-high was in accordance with the literature (14, 22, 27, 36) and required strong nuclear and cytoplasmic staining in the proliferating tumor cells, guided, if necessary, by Ki-67 positivity. Patchy and negative staining was recorded as p16-low.

Statistical analysis

Patient and tumor characteristics were analyzed in relation to their HPV16 DNA and RNA status. Overall survival (OS) was measured as the time from the date of primary tumor diagnosis to the date of OPSCC-related death. One-hundred and eleven patients died during this study, out of which only 9 patients died unrelated to OPSCC. Survival times of patients who were alive at the date of last follow-up or were dead because of causes other than OPSCC were censored. Progression-free survival (PFS) time was calculated from the date of primary tumor diagnosis to the date of the first local recurrence, lymph node or distant metastasis, second primary carcinoma or date of OPSCC-related death (events), or to the date of OPSCC-unrelated death or last follow-up without progression (censored). Follow-up data after diagnosis were obtained from electronic health records of the University Hospital. In case of incomplete records, the resident physicians were contacted to obtain the follow-up data. The cutoff date for follow-up was 31 December, 2009. Thirty-five of 199 (18%) patients were not completely followed. However, the baseline characteristics, for example, age, gender, clinical stage, therapy status, and the distribution of HPV status of these patients were similar to the patients with complete follow-up. The method of Kaplan and Meier was used to estimate survival distributions. Median duration of follow-up was calculated according to Korn (37). Multivariate Cox proportional hazard models were used to analyze the effects of single HPV markers (DNA, viral load, viral E6*II and/or E6*I RNA, and viral RNA patterns) and of HPV marker combinations (HPVtransf/HPVtransf+) as well as of p16INK4a expression on OS and PFS together with covariates age, gender, clinical stage (I–III vs. IV), alcohol and tobacco consumption (never vs. current vs. former user), and first-line therapy (surgery vs. radiation/chemotherapy). By combining the HPVhigh group (n = 33) with the CxCaRNA+ cases among the HPVlow group (n = 9), we formed the HPVtransf+ group (n = 42), whereas HPVlow tumors without viral RNA transcript expression formed the HPVtransf group (n = 54). For all HPV markers or marker combinations, the HPV DNA negative (HPV) group was the reference category. For the IHC marker, tumors with a low p16INK4a expression represented the reference category. Cox models were built using missing value imputation for all covariates by application of bootstrap resampling (38). The proportional hazards assumption, required for valid inference in Cox proportional hazards regression, was tested according to Grambsch and Therneau (39) and was met for all covariates. The ability of the prediction models to classify risk for death and disease progression was assessed by computing the area under the time-dependent receiver operating characteristic (ROC) curves (40). To summarize and compare the prediction accuracy of the different Cox models, we calculated the integrated Brier score as a summary of the prediction error curve over the time interval of 5 years from date of primary tumor diagnosis (41). The prediction error was computed by the 0.632+ bootstrap estimate (42).

In all statistical tests a P value of 0.05 or below was considered as statistically significant. The statistical analyses were carried out using SAS, version 9.2, as well as the statistical software environment R, version 2.12.2.

HPV-type prevalence and viral load

All 199 OPSCC biopsies yielded DNA of good quality with the β-globin gene coamplified in each sample. Of 100 OPSCC (50%) positive for HPV DNA analyzed by BSGP5+/6+-PCR/MPG, 97 contained HPV16 DNA (HPV+, Table 1, Fig. 1). Three tumors contained a single HPV genotype, HPV18, 33, or 35; these were not further analyzed here. No multiple infections were detected. To confirm the quantitative BSGP5+/6+-PCR/MPG results, all HPV+ tumors were again analyzed by qPCR. Eighty-six tumors were HPV16 DNA positive, resulting in 89% concordance (r = 0.93, P < 0.01; Spearman's rank correlation). Eleven tumors with low viral load by BSGP5+/6+-PCR/MPG were below detection level by qPCR but remained in the low viral load group for the further analyses. Thus, of the 97 HPV+ tumors, 64 (66%) had a low (HPVlow) and 33 (34%) had a high viral load (HPVhigh; Fig. 1).

Figure 1.

Overall flow chart showing the grouping of OPSCC in relation to DNA, viral load, and RNA status. HPV, OPSCC negative for HPV DNA; HPV+, OPSCC positive for HPV16 DNA determined by multiplex HPV genotyping (BSGP5+/6+-PCR/MPG); HPVlow or HPVhigh, OPSCC with low or high viral load determined by real-time qPCR; RNA, OPSCC positive for HPV16 DNA, but negative for HPV16 E6*II and E6*I RNA; RNA+, OPSCC positive for HPV16 DNA and positive for HPV16 E6*II and/or E6*I RNA; CxCaRNA, OPSCC positive for HPV16 DNA and for HPV16 E6*II and E6*I RNA, but without CxCa-like viral RNA patterns; CxCaRNA+, OPSCC positive for HPV16 DNA and for HPV16 E6*II and/or E6*I RNA with CxCa-like viral RNA patterns. NASBA, nucleic acid sequence–based amplification. Note that 1 of 64 HPVlow OPSCC was not included in RNA analyses because of insufficient tissue material.

Figure 1.

Overall flow chart showing the grouping of OPSCC in relation to DNA, viral load, and RNA status. HPV, OPSCC negative for HPV DNA; HPV+, OPSCC positive for HPV16 DNA determined by multiplex HPV genotyping (BSGP5+/6+-PCR/MPG); HPVlow or HPVhigh, OPSCC with low or high viral load determined by real-time qPCR; RNA, OPSCC positive for HPV16 DNA, but negative for HPV16 E6*II and E6*I RNA; RNA+, OPSCC positive for HPV16 DNA and positive for HPV16 E6*II and/or E6*I RNA; CxCaRNA, OPSCC positive for HPV16 DNA and for HPV16 E6*II and E6*I RNA, but without CxCa-like viral RNA patterns; CxCaRNA+, OPSCC positive for HPV16 DNA and for HPV16 E6*II and/or E6*I RNA with CxCa-like viral RNA patterns. NASBA, nucleic acid sequence–based amplification. Note that 1 of 64 HPVlow OPSCC was not included in RNA analyses because of insufficient tissue material.

Close modal
Table 1.

Patient and tumor characteristics in relation to HPV16 status studied on OPSCC collected between 1990 and 2008 in Heidelberg, Germany

HPV16 status
TotalHPVHPV+HPVtransfHPVtransf+
N = 196N = 99N = 97N = 54N = 42
Characteristicsn (%)n (%)n (%)n (%)n (%)
Gender 
 Male 146 (74) 79 (80) 67 (69) 39 (72) 27 (64) 
 Female 50 (26) 20 (20) 30 (31) 15 (28) 15 (36) 
Age, y 
 Median 57.0 56.7 57.4 56.1 62.1 
Oropharynx subsite 
 Tonsils 84 (43) 33 (33) 51 (53) 29 (54) 21 (50) 
 Base of tongue 45 (23) 27 (27) 18 (19) 7 (13) 11 (26) 
 Othera 67 (34) 39 (40) 28 (28) 18 (33) 10 (24) 
Tumor size 
 T1–T2 87 (45) 44 (45) 43 (45) 20 (37) 23 (56) 
 T3–T4 107 (55) 54 (55) 53 (55) 34 (63) 18 (44) 
 Missing  
Lymph nodes 
 N0 47 (24) 24 (24) 23 (24) 18 (33) 5 (12) 
 N+ 147 (76) 74 (76) 73 (76) 36 (67) 36 (88) 
 Missing  
Distant metastasis 
 M0 179 (95) 93 (95) 86 (96) 47 (92) 38 (100) 
 M1 9 (5) 5 (5) 4 (4) 4 (8) 0 (0) 
 Missing 
Clinical stage 
 I–III 60 (31) 33 (34) 27 (28) 20 (37) 7 (17) 
 IV 134 (69) 65 (66) 69 (72) 34 (63) 34 (83) 
 Missing  
First-line therapy 
 Surgery 125 (68) 59 (64) 66 (72) 32 (63) 33 (83) 
 Radiation/chemotherapy 59 (32) 33 (36) 26 (28) 19 (37) 7 (17) 
 Missing 12 
Tobacco 
 Never 21 (11) 6 (6) 15 (16) 0 (0) 15 (37) 
 Former 20 (10) 9 (9) 11 (11) 6 (11) 5 (12) 
 Current 153 (79) 83 (85) 70 (73) 48 (89) 21 (51) 
 Missing  
Alcohol 
 Never 15 (8) 9 (9) 6 (6) 2 (4) 4 (10) 
 Former 23 (12) 18 (18) 5 (5) 4 (8) 1 (2) 
 Current 155 (80) 71 (73) 84 (89) 47 (88) 36 (88) 
 Missing 
HPV16 status
TotalHPVHPV+HPVtransfHPVtransf+
N = 196N = 99N = 97N = 54N = 42
Characteristicsn (%)n (%)n (%)n (%)n (%)
Gender 
 Male 146 (74) 79 (80) 67 (69) 39 (72) 27 (64) 
 Female 50 (26) 20 (20) 30 (31) 15 (28) 15 (36) 
Age, y 
 Median 57.0 56.7 57.4 56.1 62.1 
Oropharynx subsite 
 Tonsils 84 (43) 33 (33) 51 (53) 29 (54) 21 (50) 
 Base of tongue 45 (23) 27 (27) 18 (19) 7 (13) 11 (26) 
 Othera 67 (34) 39 (40) 28 (28) 18 (33) 10 (24) 
Tumor size 
 T1–T2 87 (45) 44 (45) 43 (45) 20 (37) 23 (56) 
 T3–T4 107 (55) 54 (55) 53 (55) 34 (63) 18 (44) 
 Missing  
Lymph nodes 
 N0 47 (24) 24 (24) 23 (24) 18 (33) 5 (12) 
 N+ 147 (76) 74 (76) 73 (76) 36 (67) 36 (88) 
 Missing  
Distant metastasis 
 M0 179 (95) 93 (95) 86 (96) 47 (92) 38 (100) 
 M1 9 (5) 5 (5) 4 (4) 4 (8) 0 (0) 
 Missing 
Clinical stage 
 I–III 60 (31) 33 (34) 27 (28) 20 (37) 7 (17) 
 IV 134 (69) 65 (66) 69 (72) 34 (63) 34 (83) 
 Missing  
First-line therapy 
 Surgery 125 (68) 59 (64) 66 (72) 32 (63) 33 (83) 
 Radiation/chemotherapy 59 (32) 33 (36) 26 (28) 19 (37) 7 (17) 
 Missing 12 
Tobacco 
 Never 21 (11) 6 (6) 15 (16) 0 (0) 15 (37) 
 Former 20 (10) 9 (9) 11 (11) 6 (11) 5 (12) 
 Current 153 (79) 83 (85) 70 (73) 48 (89) 21 (51) 
 Missing  
Alcohol 
 Never 15 (8) 9 (9) 6 (6) 2 (4) 4 (10) 
 Former 23 (12) 18 (18) 5 (5) 4 (8) 1 (2) 
 Current 155 (80) 71 (73) 84 (89) 47 (88) 36 (88) 
 Missing 

Abbreviations: HPV, HPV16 DNA negative analyzed by BSGP5+/6+-PCR/MPG; HPV+, HPV16 DNA positive analyzed by BSGP5+/6+-PCR/MPG; HPVtransf, HPV+ with low viral load analyzed by qPCR, but without CxCa-like viral RNA patterns; HPVtransf+, HPV+ with high viral load analyzed by qPCR and/or CxCa-like viral RNA patterns.

aNineteen palate, 18 extending outside oropharynx (hypopharynx, nasopharynx, and/or larynx), 11 uvula, 19 unspecified.

Quantitative assessment of HPV16 E6*I and E6*II RNA expression

Of 96 HPV+ tumors (one HPVlow sample was excluded because of insufficient biopsy material), 48 (50%, RNA+) expressed E6*I and 38 of them additionally expressed E6*II. Of 33 HPVhigh tumors, 32 (97%) were RNA+ and all but 1 positive for both E6*I and E6*II. Of 64 HPVlow tumors, 16 (25%) were RNA+ (Fig. 1) and only 5 positive for both E6*I and E6*II.

HPV16 RNA patterns

Of 48 RNA+ tumors, 40 (83%) displayed CxCa-like viral RNA patterns (CxCaRNA+, Fig. 1). Of these, 12 (25%) displayed high E6*/E1⁁E4, 25 (52%) displayed high E1C/L1, and 3 (6%) displayed both patterns (Fig. 2). The 8 tumors not displaying CxCa-like viral RNA patterns (CxCaRNA) exhibited very low signals for all viral transcripts.

Figure 2.

Viral RNA patterns analogous to CxCa in HPV16 E6*II and/or E6*I positive OPSCC (n = 48). Tumors with high E6*/E1⁁E4 are located in the lower right quadrant (n = 12), tumors with high E1C/L1 in the upper left (n = 25), tumors with both patterns in the upper right (n = 3), and tumors without CxCa-like viral RNA patterns in the lower left quadrant (n = 8). Closed triangles, OPSCC positive for E6*II and E6*I; open triangles, OPSCC positive for E6*I only. Dashed lines represent thresholds for transcript ratios.

Figure 2.

Viral RNA patterns analogous to CxCa in HPV16 E6*II and/or E6*I positive OPSCC (n = 48). Tumors with high E6*/E1⁁E4 are located in the lower right quadrant (n = 12), tumors with high E1C/L1 in the upper left (n = 25), tumors with both patterns in the upper right (n = 3), and tumors without CxCa-like viral RNA patterns in the lower left quadrant (n = 8). Closed triangles, OPSCC positive for E6*II and E6*I; open triangles, OPSCC positive for E6*I only. Dashed lines represent thresholds for transcript ratios.

Close modal

HPV16 marker combinations

The groups HPVhigh and CxCaRNA+ were in good agreement as 31 of 33 (94%) HPVhigh tumors were CxCaRNA+. However, 9 CxCaRNA+ tumors were not detected by the marker HPVhigh (Fig. 3). Assuming that CxCaRNA+ was a critical parameter for tissues transformed by HPV16 (HPVtransf), we combined the HPVhigh group (n = 33) with the CxCaRNA+ cases among the HPVlow group (n = 9) to form the HPVtransf+ group (n = 42), whereas HPVlow tumors without CxCa-like viral RNA formed the HPVtransf group (n = 54).

Figure 3.

Correlation between HPV16 DNA copy numbers determined by qPCR and estimated HPV16 viral load using BSGP5+/6+-PCR/MPG in OPSCC (n = 97). Each symbol depicts 1 tumor; HPVtransf and HPVtransf+ tumors are displayed as open and closed symbols, respectively. Tumors were either positive for CxCa-like viral RNA patterns (CxCa+transf+, closed triangles) or positive for viral E6*II and/or E6*I only (RNA+, squares). Circles denote E6*II- and E6*I-negative tumors (RNA). Dashed lines represent cutoffs between low viral load (HPVlow) and high viral load (HPVhigh) groups for each analysis (COBSGP5+/6+-PCR/MPG 7 × 10−4; COcopy#/cell 0.5). Note that 9 of 64 HPVlow tumors had CxCa-like viral RNA patterns; 1 HPVhigh tumor was RNA negative. Viral load values (y-axis) analyzed by BSGP5+/6+-PCR/MPG correlated significantly with viral DNA copy numbers (x-axis), resulting from qPCR analysis (r = 0.93, P < 0.001; Spearman's rank correlation).

Figure 3.

Correlation between HPV16 DNA copy numbers determined by qPCR and estimated HPV16 viral load using BSGP5+/6+-PCR/MPG in OPSCC (n = 97). Each symbol depicts 1 tumor; HPVtransf and HPVtransf+ tumors are displayed as open and closed symbols, respectively. Tumors were either positive for CxCa-like viral RNA patterns (CxCa+transf+, closed triangles) or positive for viral E6*II and/or E6*I only (RNA+, squares). Circles denote E6*II- and E6*I-negative tumors (RNA). Dashed lines represent cutoffs between low viral load (HPVlow) and high viral load (HPVhigh) groups for each analysis (COBSGP5+/6+-PCR/MPG 7 × 10−4; COcopy#/cell 0.5). Note that 9 of 64 HPVlow tumors had CxCa-like viral RNA patterns; 1 HPVhigh tumor was RNA negative. Viral load values (y-axis) analyzed by BSGP5+/6+-PCR/MPG correlated significantly with viral DNA copy numbers (x-axis), resulting from qPCR analysis (r = 0.93, P < 0.001; Spearman's rank correlation).

Close modal

Patient and tumor characteristics in relation to HPV16 DNA and RNA status

HPV and HPV+ tumors showed similar clinical characteristics except that the tonsillar subsite and alcohol consumption were associated with the HPV+ group. The HPVtransf+ group in comparison to the HPVtransf group was associated with older age, more lymph node involvement, advanced clinical stages, less receipt of radiation/chemotherapy for first-line treatment, and it contained all never-smokers of the HPV+ group (Table 1).

Prevalence of HPV and high p16INK4a expression in the tumors grouped according to HPV16 DNA and RNA status

HPV16 DNA prevalence increased from 37% in tumors collected between 1990 and 1999 to 63% in the tumors collected from 1999 to 2008. In contrast, the prevalence of HPVtransf+ or CxCaRNA+ tumors did not change over time (Table 2). One-hundred and seventy eight of the 188 OPSCC included on the TMA could be evaluated. Fifty-four (30%) showed high p16INK4a expression (Table 2, Supplementary Fig. S1). As expected, high p16INK4a expression was significantly associated with the presence of HPV16. However, even in the HPVtransf+ group, there was a sizeable fraction of discordant cases with 9 of 41 (22%) not showing high p16INK4a levels. In the HPV+ group (HPV DNA positivity alone), only 42 out of 92 (46%) of the cases showed high p16INK4a levels, whereas the HPVtransf group showed a prevalence of high p16INK4a expression (10 of 50; 20%) similar to the HPV reference group (12 of 85; 14%, Table 2).

Table 2.

Prevalence of HPV16 DNA and RNA and p16INK4a expression according to HPV DNA and RNA status in a German OPSCC cohort, collected between 1990 and 2008 in Heidelberg

HPV16 status
Total (N = 196)HPV (N = 99)HPV+ (N = 97)RNA (N = 56)CxCaRNA+ (N = 40)HPVtransf (N = 54)HPVtransf+ (N = 42)
n (%)n (%)n (%)Pn (%)n (%)Pn (%)n (%)P
Diagnosis date 
 5/1990–7/1999 98 (50) 62 (63) 36 (37)  20 (36) 15 (38)  19 (35) 16 (38)  
 8/1999–7/2008 98 (50) 37 (37) 61 (63) <0.001 36 (64) 25 (62) 0.9 35 (65) 26 (62) 0.8 
p16INK4a, a 
 Low 124 (70) 73 (86) 50 (54)  41 (79) 8 (21)  40 (80) 9 (22)  
 High 54 (30) 12 (14) 42 (46) <0.001 11 (21) 31 (79) <0.001 10 (20) 32 (78) <0.001 
 Missing 19 14    
HPV16 status
Total (N = 196)HPV (N = 99)HPV+ (N = 97)RNA (N = 56)CxCaRNA+ (N = 40)HPVtransf (N = 54)HPVtransf+ (N = 42)
n (%)n (%)n (%)Pn (%)n (%)Pn (%)n (%)P
Diagnosis date 
 5/1990–7/1999 98 (50) 62 (63) 36 (37)  20 (36) 15 (38)  19 (35) 16 (38)  
 8/1999–7/2008 98 (50) 37 (37) 61 (63) <0.001 36 (64) 25 (62) 0.9 35 (65) 26 (62) 0.8 
p16INK4a, a 
 Low 124 (70) 73 (86) 50 (54)  41 (79) 8 (21)  40 (80) 9 (22)  
 High 54 (30) 12 (14) 42 (46) <0.001 11 (21) 31 (79) <0.001 10 (20) 32 (78) <0.001 
 Missing 19 14    

Abbreviations: HPV, HPV16 DNA-negative analyzed by BSGP5+/6+-PCR/MPG; HPV+, HPV16 DNA-positive analyzed by BSGP5+/6+-PCR/MPG; RNA, HPV+ OPSCC without viral RNA patterns (CxCaRNA) and without viral oncogene expression (E6*II, E6*I transcripts); CxCaRNA+, OPSCC with viral RNA patterns; HPVtransf, HPV+ with low viral load analyzed by qPCR, but without CxCa-like viral RNA patterns; HPVtransf+, HPV+ with high viral load analyzed by qPCR and/or CxCa-like viral RNA patterns

aBased on immunohistochemical analysis.

OS and PFS in relation to HPV DNA, RNA, and p16INK4a

The median follow-up time for this study was 103 months. In the HPV+ group, both OS and PFS were better than in the HPV group (median survival times 61 and 32 months in HPV+ patients vs. 26 and 12 months in HPV patients, respectively; Fig. 4A and B). However, stratifying the patients by viral load or RNA status revealed larger differences, and the strongest differences between survival curves was seen in the CxCaRNA+ or the HPVtransf+ group (Fig. 4C–F). High p16INK4a expression was also associated with better OS (median survival time 122 months in patients with p16-high vs. 34 months in patients with p16-low) and PFS times (median survival time 67 months in patients with p16-high vs. 20 months in patients with p16-low; Fig. 4G and H), but performed much weaker compared with high viral load or RNA patterns. In addition, the combination p16-high and HPV+ did not provide improved OS or PFS over p16-high alone (Fig. 4I and J).

Figure 4.

OS (A, C, E, G, I) and PFS (B, D, F, H, J) in relation to HPV16 status and p16INK4a expression. Tumors were grouped by HPV16 DNA status (A, B), viral load (C, D), viral marker combinations (HPVtransf+, HPVhigh tumors combined with HPVlow tumors positive for CxCa-like viral RNA patterns; HPVtransf, HPVlow tumors without CxCa-like viral RNA patterns; E, F), p16INK4a expression (p16-high, p16-low; G, H), and combination of p16INK4a with HPV16 DNA (p16-low+HPV+, p16-high+HPV+; I, J). Note that survival was not significantly different if only HPV16 DNA status was assessed. Grouping by viral load status or by marker combinations revealed significant differences in OS and PFS. HPV or HPV+, OPSCC negative or positive for HPV16 DNA; HPVlow or HPVhigh, OPSCC with low or high viral load.

Figure 4.

OS (A, C, E, G, I) and PFS (B, D, F, H, J) in relation to HPV16 status and p16INK4a expression. Tumors were grouped by HPV16 DNA status (A, B), viral load (C, D), viral marker combinations (HPVtransf+, HPVhigh tumors combined with HPVlow tumors positive for CxCa-like viral RNA patterns; HPVtransf, HPVlow tumors without CxCa-like viral RNA patterns; E, F), p16INK4a expression (p16-high, p16-low; G, H), and combination of p16INK4a with HPV16 DNA (p16-low+HPV+, p16-high+HPV+; I, J). Note that survival was not significantly different if only HPV16 DNA status was assessed. Grouping by viral load status or by marker combinations revealed significant differences in OS and PFS. HPV or HPV+, OPSCC negative or positive for HPV16 DNA; HPVlow or HPVhigh, OPSCC with low or high viral load.

Close modal

Multivariate Cox regression using only demographic, clinical, and behavioral factors showed a statistically significant association of clinical stage IV (OS only), current tobacco consumption (PFS only), and the first-line therapy with a shorter survival (Table 3). Most notable, Cox proportional hazard models of HPV markers adjusting for demographic, clinical, and behavioral factors revealed a statistically significant association of HPV viral load and viral RNA patterns with both OS and PFS. Although DNA positivity alone was not related with a strong inverse association with outcomes, HPVhigh patients had a 68% and CxCaRNA+ patients a 72% lower risk of death from OPSCC compared with HPV patients. The HRs with respect to PFS were HR = 0.45 for HPVhigh patients and HR = 0.50 for CxCaRNA+ patients compared with HPV patients. The HPVtransf+ group (HPVhigh and/or CxCaRNA+) had a comparable survival advantage as the single marker groups HPVhigh and CxCaRNA+ (Table 3). Importantly, tumors without transcriptionally active HPV16 represented by the HPVtransf group conferred similar risks for OPSCC cancer death and tumor relapse as patients with HPV tumors (Table 3). Of interest, the combination p16-high and HPV+ conferred a decreased risk of OS but did not perform as well as CxCaRNA+ or high viral load in the multivariate analysis. In contrast, p16INK4a alone did not confer a decreased risk of OS. Both, p16-high and the combination of p16-high and HPV+ had no effect on PFS, in contrast to CxCaRNA+ and high viral load (Table 3). Compared with the Cox model of demographic, clinical, and behavioral factors alone, prediction accuracy of OS was always improved by including the single HPV markers or marker combinations into the model, but not by including p16INK4a alone. The strongest improvement resulted from using the HPV16 marker combination and CxCa-like RNA patterns. For OS, the cumulative prediction error decreased from 0.192 for the clinical factors model to 0.183 using additional CxCa-like RNA patterns and to 0.182 using additional HPV marker combination. On the contrary, no improvement in prediction accuracy was observed for PFS by adding HPV markers, p16INK4a, or marker combinations to clinical factors (Supplementary Table S1). Similar results were obtained with respect to the area under the time-dependent ROC curves (Supplementary Fig. S2).

Table 3.

Multivariate analysis of OS and PFS for OPSCC patients collected between 1990 and 2008 in Heidelberg, Germany

OSPFS
ParameterHR (95% CI)P valueHR (95% CI)P value
Demographic, clinical, and behavioral factors 
 Age (10 y)  1.01 (0.82–1.26) 0.9 1.02 (0.84–1.24) 0.8 
 Gender (female vs. male)  1.09 (0.67–1.79) 0.7 0.94 (0.60–1.46) 0.8 
 Clinical stage (IV vs. I–III)  2.10 (1.25–3.54) 0.005 1.45 (0.94–2.24) 0.09 
 Therapy status (Radiation/chemotherapy vs. surgery)  1.95 (1.26–3.03) 0.003 1.71 (1.13–2.59) 0.01 
 Tobacco 
  Current vs. never  2.13 (0.84–5.38) 0.1 2.56 (1.09–6.00) 0.03 
  Former vs. never  1.24 (0.39–3.92) 0.7 1.39 (0.50–3.88) 0.5 
 Alcohol 
  Current vs. never  1.27 (0.50–3.22) 0.6 1.12 (0.48–2.63) 0.8 
  Former vs. never  1.41 (0.49–4.06) 0.5 1.30 (0.50–3.38) 0.6 
Single HPV markers     
 HPV (reference) 99 1.00  1.00  
 HPV+ 97 0.67 (0.44–1.03) 0.07 0.77 (0.53–1.12) 0.2 
 HPV (reference) 99 1.00  1.00  
 HPVlow 64 0.82 (0.53–1.26) 0.4 0.89 (0.60–1.31) 0.5 
 HPVhigh 33 0.32 (0.14–0.73) 0.007 0.45 (0.23–0.90) 0.02 
 HPV (reference) 99 1.00  1.00  
 RNA 48 0.91 (0.56–1.46) 0.7 0.92 (0.60–1.43) 0.7 
 RNA+ 48 0.44 (0.24–0.80) 0.008 0.61 (0.37–1.02) 0.06 
 HPV (reference) 99 1.00  1.00  
 RNA plus CxCaRNA 56 0.95 (0.61–1.49) 0.8 0.94 (0.62–1.42) 0.8 
 CxCaRNA+ 40 0.28 (0.13–0.61) 0.001 0.50 (0.27–0.92) 0.02 
HPV marker combinations 
 HPV (reference) 99 1.00  1.00  
 HPVtransf 54 0.96 (0.62–1.51) 0.9 0.96 (0.63–1.45) 0.8 
 HPVtransf+ 42 0.30 (0.14–0.62) 0.001 0.50 (0.28–0.90) 0.02 
p16INK4a Immunohistochemistry 
 p16-low (reference) 123 1.00  1.00  
 p16-high 54 0.90 (0.55–1.49) 0.7 1.07 (0.69–1.66) 0.8 
p16INK4a HPV DNA combination 
 HPV (reference) 99 1.00  1.00  
 p16-low+HPV+ 50 0.70 (0.43–1.13) 0.1 0.74 (0.48–1.14) 0.2 
 p16-high+HPV+ 42 0.55 (0.29–1.08) 0.08 0.83 (0.47–1.44) 0.5 
OSPFS
ParameterHR (95% CI)P valueHR (95% CI)P value
Demographic, clinical, and behavioral factors 
 Age (10 y)  1.01 (0.82–1.26) 0.9 1.02 (0.84–1.24) 0.8 
 Gender (female vs. male)  1.09 (0.67–1.79) 0.7 0.94 (0.60–1.46) 0.8 
 Clinical stage (IV vs. I–III)  2.10 (1.25–3.54) 0.005 1.45 (0.94–2.24) 0.09 
 Therapy status (Radiation/chemotherapy vs. surgery)  1.95 (1.26–3.03) 0.003 1.71 (1.13–2.59) 0.01 
 Tobacco 
  Current vs. never  2.13 (0.84–5.38) 0.1 2.56 (1.09–6.00) 0.03 
  Former vs. never  1.24 (0.39–3.92) 0.7 1.39 (0.50–3.88) 0.5 
 Alcohol 
  Current vs. never  1.27 (0.50–3.22) 0.6 1.12 (0.48–2.63) 0.8 
  Former vs. never  1.41 (0.49–4.06) 0.5 1.30 (0.50–3.38) 0.6 
Single HPV markers     
 HPV (reference) 99 1.00  1.00  
 HPV+ 97 0.67 (0.44–1.03) 0.07 0.77 (0.53–1.12) 0.2 
 HPV (reference) 99 1.00  1.00  
 HPVlow 64 0.82 (0.53–1.26) 0.4 0.89 (0.60–1.31) 0.5 
 HPVhigh 33 0.32 (0.14–0.73) 0.007 0.45 (0.23–0.90) 0.02 
 HPV (reference) 99 1.00  1.00  
 RNA 48 0.91 (0.56–1.46) 0.7 0.92 (0.60–1.43) 0.7 
 RNA+ 48 0.44 (0.24–0.80) 0.008 0.61 (0.37–1.02) 0.06 
 HPV (reference) 99 1.00  1.00  
 RNA plus CxCaRNA 56 0.95 (0.61–1.49) 0.8 0.94 (0.62–1.42) 0.8 
 CxCaRNA+ 40 0.28 (0.13–0.61) 0.001 0.50 (0.27–0.92) 0.02 
HPV marker combinations 
 HPV (reference) 99 1.00  1.00  
 HPVtransf 54 0.96 (0.62–1.51) 0.9 0.96 (0.63–1.45) 0.8 
 HPVtransf+ 42 0.30 (0.14–0.62) 0.001 0.50 (0.28–0.90) 0.02 
p16INK4a Immunohistochemistry 
 p16-low (reference) 123 1.00  1.00  
 p16-high 54 0.90 (0.55–1.49) 0.7 1.07 (0.69–1.66) 0.8 
p16INK4a HPV DNA combination 
 HPV (reference) 99 1.00  1.00  
 p16-low+HPV+ 50 0.70 (0.43–1.13) 0.1 0.74 (0.48–1.14) 0.2 
 p16-high+HPV+ 42 0.55 (0.29–1.08) 0.08 0.83 (0.47–1.44) 0.5 

NOTE: HR for age, gender, clinical stage, therapy status, and alcohol and tobacco consumption were calculated using a Cox model including only these factors. HR for single HPV markers were computed using 4 different models: (i) DNA model (HPV+, HPV16 DNA–positive OPSCC), (ii) viral load model (HPVlow, OPSCC with low viral load; HPVhigh, OPSCC with high viral load), (iii) RNA models (RNA, DNA-positive and RNA-negative OPSCC; RNA+, DNA-positive and RNA-positive OPSCC, or (iv) RNA pattern model (RNA plus CxCaRNA, DNA-positive and RNA-negative OPSCC plus DNA-positive and RNA-positive OPSCC without CxCa-like viral RNA patterns; CxCaRNA+, DNA-positive and RNA-positive with CxCa-like viral RNA patterns). HRs were also calculated for HPV marker combinations: HPVtransf, OPSCC with low viral load, but without CxCa-like viral RNA patterns; HPVtransf+, OPSCC with high viral load and/or CxCa-like viral RNA patterns. HPV group (HPV16 DNA-negative OPSCC) resulting from BSGP5+/6+-PCR/MPG analysis was reference category for all models. p16-low or p16-high, OPSCC with low or high p16INK4a expression level analyzed by IHC. Models were adjusted by age, gender, clinical stage, therapy status, and alcohol/tobacco consumption. Missing values in covariates were imputed. Statistically significant values are in bold. P values were calculated for each marker model separately.

We analyzed 199 OPSCC for the active involvement of HPV16 using viral load, viral oncogene RNA, CxCa-like viral RNA patterns, and p16INK4a as markers. For the first time, we assessed whether the HPV16 DNA–positive OPSCC displayed similar viral RNA patterns as seen in HPV16-positive CxCa. By HPV DNA status only, HPV+ and HPV tumors were marginally distinct entities, since they showed similar survival curves, and in multivariate Cox proportional hazard models, patients with HPV+ tumors had no statistically significantly better outcome in OS and PFS compared with patients with HPV tumors (Table 3).

We are aware that some of the HPVlow tumors in our cohort likely in respect to HPV-driven OPSCC were false positive because of the presence of low amounts of HPV DNA in the tissues resulting from past infection that had not progressed to malignancy or from recent oral exposure to the virus. However, the use of high sensitivity detection technology with high viral load quantitation was justified since we clearly detected HPVlow tumors in which HPV16 was transcriptionally active (see below). Compared with HPV DNA status, expression of the viral E6*I and/or E6*II transcripts (RNA+ group) was superior in defining patients with better survival, a finding in full agreement with a recent French study (17). However, both high viral load and the expression of CxCa-like viral RNA patterns exhibited similar, and stronger, inverse associations with survival. Although high viral load identified only 33 patients, the CxCaRNA+ group consisted of 40 patients, suggesting that the HPV RNA pattern analysis in comparison to viral load analysis has higher sensitivity for identifying patients with improved survival.

The analysis of CxCa-like viral RNA patterns is complex and not feasible for every clinical laboratory. A feasible and precise algorithm to identify truly HPV-driven OPSCC was provided by taking the HPVhigh tumors, which showed a very high overlap with CxCaRNA+ tumors (Fig. 1), and combining them with HPVlow tumors positive for CxCa-like RNA patterns. With this algorithm, analyses of RNA patterns can be limited to HPVlow tumors, without losing sensitivity. The HR of the HPVtransf+ group was comparable to the best single marker CxCaRNA+ (Table 3).

Importantly, the CxCa-like viral RNA patterns as well as high viral load were more strongly associated with improved survival [CxCaRNA+, HR = 0.28, 95% confidence interval (CI), 0.13–0.61 and HPVhigh, HR = 0.32, 95% CI, 0.14–0.73] than overexpression of p16INK4a (HR = 0.90, 95% CI, 0.55–1.49) because in our study cohort by far, not all CxCaRNA+ and HPVhigh tumors showed increased p16INK4a expression, but a few HPV-negative tumors did (Table 2, Supplementary Fig. S1). We believe that the large number of discordant cases in our study population was not because of some technical problems, but reflects true differences in the biology of HPV16-associated OPSCC and HPV16-positive CxCa, because in a parallel study on CxCa, only 1 of 58 cervical tumors lacked p16INK4a overexpression (43). Thus, our results are in conflict with some reports in the literature in which only very few HPV-positive OPSCC or HNSCC were not showing high p16INK4a expression and hence p16INK4a was proposed to provide a cellular surrogate marker for HPV involvement and predictor of improved survival (8, 22, 23, 44). Our results are in agreement with other studies in which the combination of high p16INK4a with positive HPV DNA typing was proposed to better correlate with active HPV involvement and improved patient survival than high p16INK4a alone (14, 26, 27). However, the combination p16-high and HPV+ did not match high viral load or viral RNA pattern analysis (p16-high and HPV+ vs. HPV, HR = 0.55; 95% CI, 0.29–1.08; Table 3). Surprisingly, both p16-high and the combination p16-high and HPV+ had some beneficial effect on OS but not on PFS (Table 3). The reason for this is unclear. However, our data did not reveal any differential impact of p16INK4a overexpression in HPV DNA–positive versus HPV DNA–negative OPSCC as was observed by Smith and colleagues (26).

A limitation of this study is that we could not confirm improved prediction accuracy of the CxCa-like RNA patterns over RNA+ by statistical analysis, probably because of the small number of RNA+ tumors without CxCa-like RNA patterns (n = 8). However, because the HR for the 8 CxCaRNA tumors was similar to the HR of the tumors without transcriptionally active HPV16 (RNA and HPV group), and since they also showed similar histology, tumor subsite distribution, and patient outcome (data not shown), we conclude that CxCaRNA tumors behave like HPV tumors. It may be helpful for future studies to develop a quantitative cutoff for viral RNA copy numbers per OPSCC tumor cell.

The prevalence in OPSCC of both HPV16 DNA (49%) and, in particular, the CxCa-like RNA patterns (20%; the latter reflecting biological activity of HPV16) in this study are in line with several previous Western European studies (16, 17, 20, 21, 23, 44), but are distinctly lower than reported in the major Scandinavian and North American studies (5, 7, 13, 15, 45). In these, the reported HPV prevalence ranged from 65% to 80% and absent HPV RNA expression was exceptional, indicating that nearly all of these tumors harbored active virus. The reasons for these relative prevalence differences in active HPV16 involvement between the North American/Scandinavian and Western European study populations are unclear, at present.

In conclusion, by using highly sensitive HPV DNA and RNA detection technologies, we not only detected OPSCC with high viral load and with CxCa-like viral RNA patterns but also a fraction of tumors with low viral load in which the transcriptional activity of the virus indicated a carcinogenic role. These tumors represent a clinically distinct subgroup with significantly better OS and PFS. The viral markers were superior over p16INK4a expression as diagnostic and prognostic biomarkers. Patients with HPV-driven OPSCC have been shown to have better survival even with radiochemotherapy as primary treatment modality. In situations like in Germany, with heterogeneity among the HPV DNA–positive tumor patients, precise definition of HPV activity in the tumor as suggested here could help in selection of the primary treatment modality.

M. Schmitt and M. Pawlita received research funding from Qiagen and Roche, respectively. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Holzinger, M. Pawlita, F.X. Bosch

Development of methodology: D. Holzinger, M. Schmitt, M. Pawlita, F.X. Bosch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Dyckhoff, M. Pawlita

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Holzinger, M. Schmitt, A. Benner, M. Pawlita, F.X. Bosch

Writing, review, and/or revision of the manuscript: D. Holzinger, M. Schmitt, G. Dyckhoff, A. Benner, M. Pawlita, F.X. Bosch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Schmitt

Study supervision: M. Pawlita, F.X. Bosch

The authors thank Prof. Plinkert, Executive Director, and Jochen Heß, Head of Experimental Oncology, Department of Otolaryngology, Head and Neck Surgery, Heidelberg University, for constant support and discussions. We appreciate the excellent technical assistance of Antje Schuhmann, Ines Kaden, and Nataly Henfling.

This study was funded in part by the European Commission, grant HPV-AHEAD (FP7-HEALTH-2011-282562) and a grant from BMBG/HGF-Cancéropôle Grand-Est. D. Holzinger was supported by a PhD grant of the German Research Foundation (DFG), Graduiertenkolleg 793: “Epidemiology of communicable and chronic, non-communicable diseases and their interrelationships”.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Andl
T
,
Kahn
T
,
Pfuhl
A
,
Nicola
T
,
Erber
R
,
Conradt
C
, et al
Etiological involvement of oncogenic human papillomavirus in tonsillar squamous cell carcinomas lacking retinoblastoma cell cycle control
.
Cancer Res
1998
;
58
:
5
13
.
2.
Gillison
ML
,
Koch
WM
,
Capone
RB
,
Spafford
M
,
Westra
WH
,
Wu
L
, et al
Evidence for a causal association between human papillomavirus and a subset of head and neck cancers
.
J Natl Cancer Inst
2000
;
92
:
709
20
.
3.
Paz
IB
,
Cook
N
,
Odom-Maryon
T
,
Xie
Y
,
Wilczynski
SP
. 
Human papillomavirus (HPV) in head and neck cancer. An association of HPV 16 with squamous cell carcinoma of Waldeyer's tonsillar ring
.
Cancer
1997
;
79
:
595
604
.
4.
Snijders
PJ
,
Cromme
FV
,
van den Brule
AJ
,
Schrijnemakers
HF
,
Snow
GB
,
Meijer
CJ
, et al
Prevalence and expression of human papillomavirus in tonsillar carcinomas, indicating a possible viral etiology
.
Int J Cancer
1992
;
51
:
845
50
.
5.
Ang
KK
,
Harris
J
,
Wheeler
R
,
Weber
R
,
Rosenthal
DI
,
Nguyen-Tan
PF
, et al
Human papillomavirus and survival of patients with oropharyngeal cancer
.
N Engl J Med
2010
;
363
:
24
35
.
6.
D'Souza
G
,
Agrawal
Y
,
Halpern
J
,
Bodison
S
,
Gillison
ML
. 
Oral sexual behaviors associated with prevalent oral human papillomavirus infection
.
J Infect Dis
2009
;
199
:
1263
9
.
7.
Gillison
ML
,
D'Souza
G
,
Westra
W
,
Sugar
E
,
Xiao
W
,
Begum
S
, et al
Distinct risk factor profiles for human papillomavirus type 16-positive and human papillomavirus type 16-negative head and neck cancers
.
J Natl Cancer Inst
2008
;
100
:
407
20
.
8.
Hafkamp
HC
,
Manni
JJ
,
Haesevoets
A
,
Voogd
AC
,
Schepers
M
,
Bot
FJ
, et al
Marked differences in survival rate between smokers and nonsmokers with HPV 16-associated tonsillar carcinomas
.
Int J Cancer
2008
;
122
:
2656
64
.
9.
Klussmann
JP
,
Weissenborn
SJ
,
Wieland
U
,
Dries
V
,
Kolligs
J
,
Jungehuelsing
M
, et al
Prevalence, distribution, and viral load of human papillomavirus 16 DNA in tonsillar carcinomas
.
Cancer
2001
;
92
:
2875
84
.
10.
Lindel
K
,
Beer
KT
,
Laissue
J
,
Greiner
RH
,
Aebersold
DM
. 
Human papillomavirus positive squamous cell carcinoma of the oropharynx: a radiosensitive subgroup of head and neck carcinoma
.
Cancer
2001
;
92
:
805
13
.
11.
Smith
EM
,
Ritchie
JM
,
Summersgill
KF
,
Klussmann
JP
,
Lee
JH
,
Wang
D
, et al
Age, sexual behavior and human papillomavirus infection in oral cavity and oropharyngeal cancers
.
Int J Cancer
2004
;
108
:
766
72
.
12.
Fakhry
C
,
Westra
WH
,
Li
S
,
Cmelak
A
,
Ridge
JA
,
Pinto
H
, et al
Improved survival of patients with human papillomavirus-positive head and neck squamous cell carcinoma in a prospective clinical trial
.
J Natl Cancer Inst
2008
;
100
:
261
9
.
13.
Shi
W
,
Kato
H
,
Perez-Ordonez
B
,
Pintilie
M
,
Huang
S
,
Hui
A
, et al
Comparative prognostic value of HPV16 E6 mRNA compared with in situ hybridization for human oropharyngeal squamous carcinoma
.
J Clin Oncol
2009
;
27
:
6213
21
.
14.
Weinberger
PM
,
Yu
Z
,
Haffty
BG
,
Kowalski
D
,
Harigopal
M
,
Brandsma
J
, et al
Molecular classification identifies a subset of human papillomavirus–associated oropharyngeal cancers with favorable prognosis
.
J Clin Oncol
2006
;
24
:
736
47
.
15.
Attner
P
,
Du
J
,
Nasman
A
,
Hammarstedt
L
,
Ramqvist
T
,
Lindholm
J
, et al
The role of human papillomavirus in the increased incidence of base of tongue cancer
.
Int J Cancer
2010
;
126
:
2879
84
.
16.
Braakhuis
BJ
,
Snijders
PJ
,
Keune
WJ
,
Meijer
CJ
,
Ruijter-Schippers
HJ
,
Leemans
CR
, et al
Genetic patterns in head and neck cancers that contain or lack transcriptionally active human papillomavirus
.
J Natl Cancer Inst
2004
;
96
:
998
1006
.
17.
Jung
AC
,
Briolat
J
,
Millon
R
,
de Reynies
A
,
Rickman
D
,
Thomas
E
, et al
Biological and clinical relevance of transcriptionally active human papillomavirus (HPV) infection in oropharynx squamous cell carcinoma
.
Int J Cancer
2010
;
126
:
1882
94
.
18.
Lindquist
D
,
Romanitan
M
,
Hammarstedt
L
,
Nasman
A
,
Dahlstrand
H
,
Lindholm
J
, et al
Human papillomavirus is a favourable prognostic factor in tonsillar cancer and its oncogenic role is supported by the expression of E6 and E7
.
Mol Oncol
2007
;
1
:
350
5
.
19.
Mellin
H
,
Dahlgren
L
,
Munck-Wikland
E
,
Lindholm
J
,
Rabbani
H
,
Kalantari
M
, 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
2002
;
102
:
152
8
.
20.
van Houten
VM
,
Snijders
PJ
,
van den Brekel
MW
,
Kummer
JA
,
Meijer
CJ
,
van Leeuwen
B
, et al
Biological evidence that human papillomaviruses are etiologically involved in a subgroup of head and neck squamous cell carcinomas
.
Int J Cancer
2001
;
93
:
232
5
.
21.
Wiest
T
,
Schwarz
E
,
Enders
C
,
Flechtenmacher
C
,
Bosch
FX
. 
Involvement of intact HPV16 E6/E7 gene expression in head and neck cancers with unaltered p53 status and perturbed pRb cell cycle control
.
Oncogene
2002
;
21
:
1510
7
.
22.
Rischin
D
,
Young
RJ
,
Fisher
R
,
Fox
SB
,
Le
QT
,
Peters
LJ
, et al
Prognostic significance of p16INK4A and human papillomavirus in patients with oropharyngeal cancer treated on TROG 02.02 phase III trial
.
J Clin Oncol
2010
;
28
:
4142
8
.
23.
Reimers
N
,
Kasper
HU
,
Weissenborn
SJ
,
Stutzer
H
,
Preuss
SF
,
Hoffmann
TK
, et al
Combined analysis of HPV-DNA, p16 and EGFR expression to predict prognosis in oropharyngeal cancer
.
Int J Cancer
2007
;
120
:
1731
8
.
24.
Begum
S
,
Cao
D
,
Gillison
M
,
Zahurak
M
,
Westra
WH
. 
Tissue distribution of human papillomavirus 16 DNA integration in patients with tonsillar carcinoma
.
Clin Cancer Res
2005
;
11
:
5694
9
.
25.
Hoffmann
M
,
Tribius
S
,
Quabius
ES
,
Henry
H
,
Pfannenschmidt
S
,
Burkhardt
C
, et al
HPV DNA, E6(*)I-mRNA expression and p16(INK4A) immunohistochemistry in head and neck cancer—how valid is p16(INK4A) as surrogate marker?
Cancer Lett
2012
;
323
:
88
96
.
26.
Smith
EM
,
Wang
D
,
Kim
Y
,
Rubenstein
LM
,
Lee
JH
,
Haugen
TH
, et al
P16INK4a expression, human papillomavirus, and survival in head and neck cancer
.
Oral Oncol
2008
;
44
:
133
42
.
27.
Smeets
SJ
,
Hesselink
AT
,
Speel
EJ
,
Haesevoets
A
,
Snijders
PJ
,
Pawlita
M
, et al
A novel algorithm for reliable detection of human papillomavirus in paraffin embedded head and neck cancer specimen
.
Int J Cancer
2007
;
121
:
2465
72
.
28.
Dyson
N
,
Howley
PM
,
Munger
K
,
Harlow
E
. 
The human papillomavirus-16 E7 oncoprotein is able to bind to the retinoblastoma gene product
.
Science
1989
;
243
:
934
7
.
29.
Werness
BA
,
Levine
AJ
,
Howley
PM
. 
Association of human papillomavirus types 16 and 18 E6 proteins with p53
.
Science
1990
;
248
:
76
9
.
30.
Schmitt
M
,
Dalstein
V
,
Waterboer
T
,
Clavel
C
,
Gissmann
L
,
Pawlita
M
. 
Diagnosing cervical cancer and high-grade precursors by HPV16 transcription patterns
.
Cancer Res
2010
;
70
:
249
56
.
31.
Schmitt
M
,
Bravo
IG
,
Snijders
PJ
,
Gissmann
L
,
Pawlita
M
,
Waterboer
T
. 
Bead-based multiplex genotyping of human papillomaviruses
.
J Clin Microbiol
2006
;
44
:
504
12
.
32.
Schmitt
M
,
Dondog
B
,
Waterboer
T
,
Pawlita
M
. 
Homogeneous amplification of genital human alpha papillomaviruses by PCR using novel broad-spectrum GP5+ and GP6+ primers
.
J Clin Microbiol
2008
;
46
:
1050
9
.
33.
Schmitt
M
,
Dondog
B
,
Waterboer
T
,
Pawlita
M
,
Tommasino
M
,
Gheit
T
. 
Abundance of multiple high-risk human papillomavirus (HPV) infections found in cervical cells analyzed by use of an ultrasensitive HPV genotyping assay
.
J Clin Microbiol
2010
;
48
:
143
9
.
34.
Karsai
S
,
Abel
U
,
Roesch-Ely
M
,
Affolter
A
,
Hofele
C
,
Joos
S
, et al
Comparison of p16(INK4a) expression with p53 alterations in head and neck cancer by tissue microarray analysis
.
J Pathol
2007
;
211
:
314
22
.
35.
Nasser
W
,
Flechtenmacher
C
,
Holzinger
D
,
Hofele
C
,
Bosch
FX
. 
Aberrant expression of p53, p16INK4a and Ki-67 as basic biomarker for malignant progression of oral leukoplakias
.
J Oral Pathol Med
2011
;
40
:
629
35
.
36.
Klaes
R
,
Friedrich
T
,
Spitkovsky
D
,
Ridder
R
,
Rudy
W
,
Petry
U
, et al
Overexpression of p16(INK4A) as a specific marker for dysplastic and neoplastic epithelial cells of the cervix uteri
.
Int J Cancer
2001
;
92
:
276
84
.
37.
Korn
EL
. 
Censoring distributions as a measure of follow-up in survival analysis
.
Stat Med
1986
;
5
:
255
60
.
38.
Harrell
F
. 
Regression modeling strategies
. 1st ed.
New York
:
Springer
; 
2001
.
39.
Grambsch
P
,
Therneau
T
. 
Proportional hazards tests and diagnostics based on weighted residuals
.
Biometrika
1994
;
81
:
515
26
.
40.
Heagerty
PJ
,
Zheng
Y
. 
Survival model predictive accuracy and ROC curves
.
Biometrics
2005
;
61
:
92
105
.
41.
Gerds
TA
,
Schumacher
M
. 
Efron-type measures of prediction error for survival analysis
.
Biometrics
2007
;
63
:
1283
7
.
42.
Efron
B
,
Tibshirani
R
. 
Improvements on cross-validation: the 0.632+ bootstrap method
.
J Am Stat Assoc
1997
;
92
:
548
60
.
43.
Halec
G
,
Schmitt
M
,
Dondog
B
,
Sharkhuu
E
,
Wentzensen
N
,
Gheit
T
, et al
Biological activity of probable/possible high-risk human papillomavirus types in cervical cancer
.
Int J Cancer
. 
2012
Apr 19.
[Epub ahead of print]
.
44.
Hoffmann
M
,
Ihloff
AS
,
Gorogh
T
,
Weise
JB
,
Fazel
A
,
Krams
M
, et al
p16(INK4a) overexpression predicts translational active human papillomavirus infection in tonsillar cancer
.
Int J Cancer
2010
;
127
:
1595
602
.
45.
D'Souza
G
,
Kreimer
AR
,
Viscidi
R
,
Pawlita
M
,
Fakhry
C
,
Koch
WM
, et al
Case-control study of human papillomavirus and oropharyngeal cancer
.
N Engl J Med
2007
;
356
:
1944
56
.

Supplementary data