HPV Serum Antibodies as Predictors of Survival and Disease Progression in Patients with HPV-Positive Squamous Cell Carcinoma of the Oropharynx

The incidence of HPV-positive oropharyngeal squamous cell carcinoma is expected to increase substantially over the next several decades due to the lack of a screening paradigm and despite the availability of a prophylactic vaccine (1). Serum antibodies to HPV proteins, in particular the early (E) proteins E6 and E7, have been detected in several HPV-related cancers and have been found to be reliable indicators of an HPV-positive tumor, including oropharyngeal tumors (2–11). Although patients with HPV-positive oropharyngeal carcinoma are typically diagnosed with late-stage disease, numerous case–control and prospective studies have shown that patients with HPV-positive tumors have better response to treatment and better survival than patients with HPV-negative tumors (12–17). Furthermore, Kreimer and colleagues recently reported a 68% reduced risk of death among patients with oropharyngeal carcinoma who were positive for E6 antibodies versus those who were negative for E6 antibodies after adjustment for smoking (10).

To investigate whether there are differential immune responses to HPV16 proteins with biologic and prognostic relevance, we used a novel protein array to investigate the association between serum antibodies to the entire HPV16 proteome and overall and progression-free survival among patients with oropharyngeal carcinoma treated at our institution.

Patients with newly diagnosed, histopathologically confirmed, and previously untreated oropharyngeal carcinoma who presented to The University of Texas MD Anderson Cancer Center and participated in a large prospective epidemiologic study of head and neck cancer between January 2006 and September 2008 were invited to participate in this separate protocol (presented herein) of serial monitoring of serum markers of disease prognosis. All patients gave written informed consent and provided demographic and exposure data as well as a blood sample for biologic testing. Samples were collected at initial workup, 6 weeks following the end of treatment, and at 6-month intervals up to 36 months for a total possible number of eight samples. Samples were collected using a standardized protocol and stored at −80°C until use. The study was approved by the MD Anderson Institutional Review Board.

Diagnostic in-house paraffin-embedded tumor tissue was obtained following histopathologic confirmation of the diagnosis. A tissue DNA extraction kit (Qiagen Inc.) was used to extract DNA, and a PCR-based type-specific assay was used to test tumor tissue for the presence of HPV16 E6 or E7 regions. β-Actin served as the DNA quality control and positive and negative controls were included in each run. Samples were run in triplicate with 100% concordance between samples.

Bead array ELISAs were performed at Arizona State University essentially as described (11, 18, 19), with the following modifications (20). The in vitro translation products were captured on magnetic microspheres (Bio-Rad), pooled, and blocked with HeteroBlock (Omega Biologicals) diluted to 2 μg/mL into SeaBlock (Thermo Scientific). Sera were diluted 1:80 and incubated with the pooled beads and bound IgG was detected using median fluorescent intensity (MFI). To control for nonspecific and GST-specific antibodies, the ratio of MFI for individual HPV16-specific antibodies to MFI for the control GST antigen was measured. To establish cutoff values for seropositivity, an MFI ratio > (the average +3 standard deviations) of 78 healthy control samples was designated positive. The cutoff values were as follows: E1, 1.4; CE2, 2.4; NE2, 1.5; E4, 2.7; E5, 1.5; E6, 1.7; E7, 2.2; late (L) 1, 1.4; and L2, 1.4. Control sera were obtained from Oregon Health and Science University (Portland, OR) from subjects participating in a head and neck cancer screening event who had no history or evidence on clinical examination of head and neck cancer (20). The control cohort consisted of 47 females and 31 males with a mean age of 51.4 years (SD, 13.5 years). The intraassay coefficient of variation was determined to be 1.04% to 8.34% and the intraassay coefficient of variation was 1.39% to 7.06%. The lower limit of detection of the assay varied based on the serum sample tested, but is shown in representative format in Supplementary Fig. S1. Laboratory researchers were blinded to clinical outcome.

Stata 12.0 (StataCorp) was used for all statistical analyses. A P value of <.05 was considered significant, and all tests were two-sided. Standard descriptive statistical methods were used to describe demographic and clinical characteristics.

Overall survival was defined as time from first appointment to death from any cause. Progression-free survival was defined as time from first appointment to clinical detection of recurrent cancer (local, regional, or distant) or death from any cause. Participants alive and recurrence-free (for progression-free survival) at last follow-up or lost to follow-up were considered censored. Kaplan–Meier curves were created to compare survival between groups, and the log-rank statistic was used to test the hypothesis of a difference in survival between groups. Cox proportional hazards models were used to calculate hazard ratios (HR) and 95% CIs to assess whether presence of specific HPV16 antibodies was associated with overall and progression-free survival. Akaike information criterion was used to select variables for inclusion in the multivariable models. Overall and progression-free survival within each group of patients were assessed separately and the models with minimum AIC were used as the final models. In addition, smoking is a known predictor for survival and was included in the models. A binary logistic regression model was used to compute the area under the receiver operating characteristic curve (ROC) to compare the performance of all combinations of E antibodies for predicting disease recurrence. The probability cutoff value for the predictions was 0.5.

A total of 209 patients were included in the final analysis. Of these, 114 had tumor HPV16 status available for subgroup analyses; 96 of these patients had HPV16-positive tumors. The median follow-up time for patients who survived was 62.7 months (range, 3.9–96.9 months). The median follow-up time for patients with HPV16-positive tumors who survived was 68.9 months (range, 4.1–92.5 months). There was no difference with respect to survival between patients who had tumor HPV16 status available and those who did not (P = 0.577).

The demographic, exposure, and clinical characteristics of the patients are summarized in Table 1. Ninety percent of patients were positive for at least one E antibody, whereas 16% were positive for at least one L antibody. Of the 114 patients with tumor HPV16 DNA status available, 96 (84%) had HPV16-positive tumors. The characteristics of the patients by tumor HPV16 status are shown in Table 1.

Progression-free survival was better among patients positive for any E antibodies (Fig. 1A), but no survival advantage was noted among patients positive for any L antibodies (Fig. 1B; P < 0.001 and P = 0.657, respectively). Therefore, we excluded L antibody status in subsequent analyses.

Patients positive for any E antibodies had better overall and progression-free survival than patients negative for all E antibodies: 5-year overall survival estimates were 87.4% and 42.2%, respectively, and 5-year progression-free survival estimates were 82.9% and 46.1%, respectively (P < 0.001 for both). In multivariable Cox proportional hazards regression, patients positive for any E antibodies had an 80% lower risk of death (HR, 0.2; 95% CI, 0.1–0.4) and progression (HR, 0.2; 95% CI, 0.1–0.5; Table 2). NE2 positivity, E1 positivity, and E6 positivity reduced the risk of death by up to 80% and the risk of progression by up to 70% (Table 2). Because smoking is a strong predictor of survival and patients with HPV-positive tumors are more likely to be never-smokers or light smokers, we also evaluated a subset of patients with a smoking history of 10 or fewer pack years. Among these patients (n = 130), positivity for any E antibodies was also associated with a significantly reduced risk of death (HR, 0.1; 95% CI, 0–0.7) and progression (HR, 0.1; 95% CI, 0–1.0; Table 2). In never-smokers and light smokers, we did not observe the strong associations with individual antibodies that we observed for the entire cohort; nevertheless, the trend for improved survival among antibody-positive patients was observed in this subset.

Figure 2A shows progression-free survival by E antibody status, and Fig. 2B shows progression-free survival by tumor HPV16 DNA status. Although E antibody positivity and HPV16 tumor positivity were associated with significantly better progression-free survival, E antibody status appeared to be a stronger predictor (P < 0.001 by E antibody status and P = 0.016 by HPV16; Fig. 2A and 2B, respectively). This result was consistent for overall survival as well (P < 0.001 by E antibody status and P = 0.031 by HPV16; data not shown). After multivariable adjustment, both patients positive for E antibodies and those with HPV16-positive tumors had an 80% decreased risk of death (95% CI, 0.1–0.5 and 0.1–0.7, respectively). Likewise, E antibody status and HPV16 status were both strongly associated with progression-free survival (E antibody: HR, 0.2; 95% CI, 0.1–0.5, and HPV16: HR, 0.2; 95% CI, 0.1–0.7). Treating E1, NE2, and E6 as continuous variables in the multivariable analysis resulted in significant associations with both overall and progression-free survival (only E6 was not significant for progression-free survival; data not shown).

Among patients with HPV16-negative tumors, E antibody positivity had no demonstrable predictive value for either overall or progression-free survival, although due to the small number of patients this result should be interpreted with caution (data not shown). However, when the analysis was restricted to patients with HPV16-positive tumors (by PCR), patients serologically positive for E antibodies remained at a survival advantage compared with patients who were serologically negative for E antibodies (HR, 0.3; 95% CI, 0.1–0.9 for overall survival and HR, 0.3; 95% CI, 0.1–0.8 for progression-free survival; Table 2 and Fig. 3). In particular, antibodies to NE2, E1, and E6 remained strong predictors of overall and progression-free survival. Treating E1, NE2, and E6 as continuous variables in the multivariable analysis resulted in significant or borderline significant (P < 0.10) associations (data not shown).

We used ROC curves to determine the best combination of E antibodies for predicting disease recurrence for all 209 patients. The combination of NE2, E4, E6, and E7 showed the highest accuracy; the optimal operating point was 71.4% at a specificity of 70.2% (AUC, 0.71). Adding E1, CE2, and E5 did not improve this. In addition, the accuracy for NE2 and E6 alone was good relative to the combined antibodies (AUC, 0.69 and AUC, 0.61; data not shown).

Figure 4 shows the median antibody levels of E1, NE2, and E6 over time for patients with HPV-positive oropharyngeal carcinoma according to recurrence status. A subset of 23 patients chosen at random who did not recur had higher median antibody levels than the 8 patients who did recur, although due to the small sample size we were not able to detect statistically significant differences between the groups.

We used a novel protein array serologic assay to determine the prognostic significance of HPV16 antibodies for predicting survival among patients with oropharyngeal carcinoma and HPV-positive oropharyngeal carcinoma. We found that positivity for HPV16 antibodies was strongly associated with both overall and progression-free survival. This association was limited to E antibodies, in particular antibodies to the early proteins NE2, E1, and E6; positivity for antibodies to coat protein L were not predictive for survival. This preliminary evidence suggests that serologic response to antigens involved in HPV-mediated carcinogenesis, but not serologic response to antigens important to the infectious process (L proteins), may differentially predict cancer outcomes. Our findings support the hypothesis that cancer outcomes are in part dictated by the immune response and that E antibodies may be biomarkers of carcinogenic changes and prognosis.

We previously showed that a small subset of patients with HPV16-positive tumors who were negative for E6 and E7 antibodies were positive for E1 and NE2 antibodies, which illustrates the need for including multiple markers in HPV16 serologic studies and for diagnosis (19). More recently, we showed a dramatically increased risk for oropharyngeal carcinoma among individuals who were positive for E antibodies in a study that included 256 cases and 250 controls (K.S. Anderson; submitted for publication). We found that those positive for any E antibody had 244 times the oropharyngeal carcinoma risk of those negative for all E antibodies. Furthermore, patients with tumors positive for HPV16 DNA were 4 times as likely to be positive for any E antibody as were patients with tumors negative for HPV16 DNA, and this association was particularly strong among patients with HPV16-positive tumors who were never-smokers or light smokers.

Although HPV16 capsid proteins, L1 and L2, have been used to establish the causal association between HPV16 and oropharyngeal carcinoma and are present several years before the development of an HPV16-positive cancer (3, 8, 21, 22), we did not find that antibodies to capsid proteins were associated with survival among patients with oropharyngeal carcinoma. This is not surprising biologically given that capsid antibodies are not expressed following tumor development and are lost on HPV DNA integration into host DNA. Capsid antibodies thus are not expected to be useful as a diagnostic tool or as predictors of cancer outcome.

The immunobiology that might explain differences or absences in antibody responses to one or another specific antigen is not understood. Serologic immune responses to HPV16-positive tumors may correlate with high viral load. For example, in the Costa Rica HPV16 vaccine trial, high viral load was predictive of HPV16 seropositivity among 646 young, HPV16-infected women (23). It has also been reported that viral load is a reliable indicator of an HPV16-driven tumor and subsequent risk of death (24, 25). Mellin and colleagues found in a small number of subjects that viral load had a wide distribution among patients with tonsillar cancer and that it was positively correlated with survival (25). This may reflect changes during carcinogenesis with decreased dependence on HPV early antigens as the tumors progress genetically. HPV16-positive tumors that fail to elicit or lose a serologic response may no longer be completely driven by HPV-based mechanisms.

This study has several limitations. First, as a large proportion of patients whose tumors were negative for HPV16 DNA were positive for at least one E antibody (13 of 18, 72.2%) and, conversely, a subset of patients whose tumors were positive for HPV16 DNA were negative for all E antibodies (10 of 96, 10.4%), these patients may have been misclassified. If we had evaluated HPV16 status using other methods in addition to PCR, such as p16 expression or in situ hybridization, we might have found more HPV16-positive cases among patients serologically positive for E antibodies but with tumors negative for HPV16 DNA by PCR. In addition, although HPV16 is the predominant type found in oropharyngeal carcinoma tumor tissue, the HPV16-negative cases could have been positive for another HPV type. We were unable to further analyze this subset of HPV16-negative, antibody-positive patients due to the small sample size and lack of availability of additional tumor tissue for analysis. Second, although the study included 209 patients with follow-up, we had HPV16 status for only 114, which limited our sample size for analysis to 96 patients with HPV16-positive tumors. Furthermore, most patients were positive for at least one HPV16 antibody, limiting our ability to isolate effects of individual antibodies. These limitations may have reduced the power to detect differences in survival between individual antibody-positive and antibody-negative subgroups. We are currently expanding the study to increase the sample size as well as using p16 expression and in situ hybridization to determine tumor HPV status. Given the high accuracy of these assays, along with increased study power, we will be able to explore more fully the association between tumor HPV DNA and antibody status.

Our results show that HPV16 antibody status, in particular E antibody status, is a strong predictor of survival among patients with HPV-positive oropharyngeal carcinoma. Specific antibody status has the potential to be a useful prognostic indicator that may identify subsets of patients diagnosed with HPV16-positive tumors who may benefit from altered monitoring and/or treatments. In addition, the suggestion that immune response to HPV16 antigens is important to cancer outcomes suggests the potential of augmenting immune responses to improve treatment of patients with HPV-driven oropharyngeal carcinoma. Using serology to identify HPV16-positive tumors may be useful as tumor tissue is not always available and serology is less invasive and therefore more acceptable to patients. Furthermore, combining multiple biomarkers may improve the validity of a prognostic test. In light of this, larger prospective studies are needed to confirm our findings.

K.S. Anderson is a consultant/advisory board member for and has ownership interest in Provista Dx. M. Posner is listed as a co-inventor on a patent for the serologic test, which is owned by ASU. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.S. Anderson, G. Li, M. Posner, E.M. Sturgis

Development of methodology: K.S. Anderson, J.N. Cheng, E.M. Sturgis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.S. Anderson, J.N. Cheng, G. Li, E.M. Sturgis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.R. Dahlstrom, K.S. Anderson, J.N. Cheng, D. Chowell, M. Posner, E.M. Sturgis

Writing, review, and/or revision of the manuscript: K.R. Dahlstrom, K.S. Anderson, J.N. Cheng, G. Li, M. Posner, E.M. Sturgis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.N. Cheng, M. Posner, E.M. Sturgis

Study supervision: K.S. Anderson, M. Posner, E.M. Sturgis

The authors thank Alison Goulder for technical assistance and Stephanie Deming for article editing.

This research was supported in part by the NIH through MD Anderson's Cancer Center Support Grant, CA016672, NIH grant R03 Grant CA128110-01A1 (to E.M. Sturgis), an institutional research grant from The University of Texas MD Anderson Cancer Center (to E.M. Sturgis), Arizona State University institutional funds (to K.S. Anderson), and National Cancer Institute Early Detection Research Network grant U01 CA117374 (to K.S. Anderson). K.S. Anderson received travel support from Luminex Corporation in 2011 and serves on the scientific advisory board and receives consultant fees from ProvistaDx. This research was accomplished within the Oropharynx Program at The University of Texas MD Anderson Cancer Center and funded in part through the Stiefel Oropharyngeal Research Fund.

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.