High-Risk HPV, Biomarkers, and Outcome in Matched Cohorts of Head and Neck Cancer Patients Positive and Negative for HIV

Cumulative evidence suggests that a proportion of patients with HIV-1 infection develop head and neck squamous cell carcinoma (HNSCC; refs. 1–4). Although it is suspected that immunodeficiency in these patients contributes to lower immune surveillance against a malignant process, little is known about mechanisms of immune suppression that lead to the development and progression of HNSCC in some HIV-positive subjects. Likewise, it is not known whether cancer progression or its aggressiveness is greater in HIV-positive subjects who develop HNSCC relative to HIV-negative subjects with HNSCC. This gap in knowledge limits our ability to successfully treat this cohort of patients with HNSCC. It is postulated that due to compromised immunity, these cancers are aggressive and rapidly progressing. In fact, clinical observations suggest that HNSCCs in this cohort of patients are aggressive and difficult to treat (5, 6). Therefore, an increased understanding of mechanisms that operate in vivo and contribute to the development of HNSCC in HIV-positive subjects is crucial for the selection of the most effective therapies. Studies are necessary to define molecular mechanisms that especially endanger HIV-positive subjects with HNSCC to predict outcome and more effectively treat such patients. In this inter-SPORE collaboration, five Head and Neck SPOREs sought to address these gaps in knowledge. Herein we assess the presence and relative frequency of high-risk HPV in HIV-positive cases and HIV-negative controls with HNSCC matched for sex, age, and tumor site and examine prognostic tumor biomarkers.

This study is a case–control series of HIV-infected and HIV-noninfected head and neck cancer patients seen at five tertiary care referral centers: Emory, Johns Hopkins, MD Anderson, Michigan, and Pittsburgh. IRB approval or exemption to share deidentified data with the study data center was obtained at each study site. This study was funded by a National Cancer Institute Translational Research Program (AARA) supplement to the Head and Neck Cancer Specialized Programs of Research Excellence (HNC-SPORE) collaborative project. HIV-infected patients also diagnosed with HNSCC were retrospectively identified through medical record review at each center and HNSCC tissue located in the pathology archive. To identify controls, pathology records were searched for patients with HNSCC matched to the cases by tumor site, sex, and age at diagnosis (within 18 years to get sufficient matches) and for the presence of archived material. Cases were contributed from five HNC-SPORE sites including MD Anderson Cancer Center (n = 4), Emory University (n = 22), University of Pittsburgh (n = 10), University of Michigan (n = 2), and Johns Hopkins University (n = 3). HIV-negative site, sex, and age-matched head and neck cancer controls were contributed by Emory University (n = 2), University of Michigan (n = 35), and Johns Hopkins University (n = 4). Clinical-pathologic data and outcome were obtained from the medical record and tumor tissue was used for biomarker testing. Of the 82 cases and controls (Table 1), follow-up information was provided by each center for 72 patients (36 case/control pairs) with median follow-up of 65 months. Patients with follow-up information represented a similar distribution of calendar time as the dataset overall, with 54% diagnosed between 2006 and 2011, 31% diagnosed between 2001 and 2005, and 15% diagnosed between 1990 and 2000.

Formalin-fixed, paraffin-embedded (FFPE) tumor tissues from all cases and controls with adequate specimens were collected, barcoded, and tested in a single centralized laboratory (University of Michigan, Ann Arbor, MI). Slides from each FFPE block were evaluated by a board-certified head and neck pathologist (J.B. McHugh) in the University of Michigan Pathology Laboratory and tumor-rich areas were marked on the blocks for construction of a TMA and for DNA extraction. Specimens with too little tissue to be included on the TMA were tested for biomarker expression using individual slides. The TMA was examined using IHC for expression of p16 (CINtec, mtm Laboratories), TP53 (Ab-6 clone DO-1), RB, (Ab-1 clone 1F8), CyclinD1 (clone SP4) (all from Thermo Scientific), and EGFR (Clone 31G7, Life Technologies). IHC staining was quantified using the product of proportion and intensity scores. Scores for proportion of stained tumor cells and for staining intensity were each scored on a four-point scale. Intensity of tumor cell staining: 1, no staining; 2, low; 3, moderate; and 4, high; proportion of tumor cells staining: 1, <5%; 2, 5%–20%; 3, 21%–50%; 4, 51%–100%. IHC scores from each core or tissue section were averaged for each patient and IHC scores of 1–4 were considered negative or low expression; scores of 5–11 were considered moderate expression; and scores of 12–16 were considered positive or high expression. The IHC assays were scored at 400× magnification by the pathologist (J.B. McHugh) who was blinded to the origin of the individual samples.

The TMA was also tested for high-risk HPV by in situ hybridization (ISH; Ventana INFORM HPV III, Ventana Medical Systems) according to the manufacturer's protocol. The INFORM HPV III assay (Ventana Medical Systems) is designed to detect the presence of 12 high-risk (oncogenic) HPV types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 66), but does not distinguish the specific HPV type(s) present. ISH was scored for the presence of blue signals indicating the presence of HPV DNA in tumor cell nuclei as single punctate (integrated) or diffuse (episomal) signals.

DNA isolated from tumor cores was tested for oncogenic HPV DNA using the HPV MultiPlex PCR-MassArray assay designed to detect and identify 15 high-risk HPV types (HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, and 73), using multiplex, type-specific, competitive PCR and single base extension followed by MALDI-TOF mass spectrometry analysis, as described previously (7). For any sample with adequate DNA that was positive by p16^ink4a or ISH but negative by HPV PCR-MA, the DNA was re-examined by consensus PCR targeting the L1 region of the viral genome using PGMY primers (8) and sequencing of the PCR product to identify the unknown type. All specimens that were found to contain an identifiable high-risk HPV type were scored as HPV-positive.

Correlations among biomarkers were explored using Spearman correlation coefficients and differences in patient characteristics between HIV-positive/negative or HPV-positive/negative groups were analyzed using logistic regression. Kaplan–Meier estimates for overall survival and Cox models taking into account the matched nature of the data were employed to analyze outcome. Overall survival was defined as time (months) from cancer diagnosis to date of death or last follow-up. Patients lost to follow-up were censored at date they were last known to be alive. All analysis was performed using SAS v9.3.

Forty-four HIV-positive cases with head and neck cancer were diagnosed between June 1990 and July 2011 at five institutions. HIV-negative patients with comparable tumors were selected from the general HNSCC patient population. Control patients were matched on the basis of gender, disease anatomic subsite, and age at cancer diagnosis. One reported HIV-positive case with esophageal cancer was later found to be HIV-negative and was excluded with its control from further consideration. There was insufficient tissue for any testing from two cases, leaving 82 HNSCC patients, 41 evaluable HIV-positive cases, and 41 HIV-negative controls (Table 1). For the entire group, the median age at cancer diagnosis was 52 years (range 29–76 years). Ninety-six percent of patients were male. After the analysis was completed, it was discovered that one pair had incorrectly matched a female HIV case, EU1852, to a male noninfected control, UM2499. The data from these subjects was retained in the study and data analysis. Median follow-up for survival was 65 months. At the time of survival analysis, 38 patients (19 cases and 19 controls) were deceased. There was no evidence of a significant difference in survival between HIV-positive and HIV-negative cases (P = 0.58). The majority of tumors were from the oral cavity (22 HIV-positive cases/22 HIV-negative controls), roughly a quarter were oropharynx tumors (11 cases/11 controls), 15% were larynx tumors (6 cases/6 controls), and a minority of tumors were hypopharynx (1 case/1 control) or classified as other (1 case/1 control; Table 1 summary).

Of the 82 cases and controls, sufficient valid DNA for HPV testing by PCR-MA was available from 74 samples, including 34 HIV-positive cases and 40 HIV-negative controls. In the PCR-MA assay, 16 of 74 tumors were positive for high-risk HPV. Of the 58 remaining samples with adequate DNA that were HPV negative by PCR-MA, 13 were strongly positive for p16 expression and were retested by L1 consensus HPV PCR. Of these, 5 were found to contain hrHPV (Table 1), for a total of 21 out of 74 hrHPV-positive tumors (Table 2). Three of the five detected by the L1 consensus primer assay were HPV types not included in the PCR-MA assay: HPV26, HPV69, and HPV82. These HPV types all belong to alpha papillomaviruses group 5, and have been implicated in cancer or rare aggressive dysplastic lesions (9–11). Two of the five contained HPV16 but were not detected by PCR-MA, suggesting a possible rearrangement in the E6 region where the PCR-MA primers bind. Seventy-three tumors were tested by ISH; of these 7 (2 cases and 5 controls) were HPV-positive by ISH (Table 1; Fig. 1). As with the PCR-MA assay, HPV types 26, 69, and 82 are not represented in the INFORM HPV III ISH assay (Fig. 2). All 7 HPV ISH-positive tumors contained HPV16 or HPV18 and all were strongly positive for p16 expression (Table 1). Two examples of HPV-positive case tumors with positive in situ HPV detection are illustrated in Fig. 1. Examples of HPV33- and HPV69-positive tumors from HIV-positive cases not positive for HPV ISH are illustrated in Fig. 2 (note that both tumors strongly express p16). HPV DNA was detected in tumor tissue of 11 out of 34 (32%) evaluable HIV-positive cases and 10 of 40 (25%) site-matched HIV-negative controls (Tables 1 and 2). Overall, HPV16 was present in 16 of 21 (76%) of the HPV-positive tumors.

There was greater diversity of hrHPV types in the tumors of the HIV-positive cases compared with the HIV-negative controls. While the majority (7/11) of HPV-positive cases contained HPV16, 4 of 11 cases had less common hrHPV types: HPV26, HPV33, HPV69, and HPV82. In the HPV-positive controls, 9 of 10 tumors contained HPV16 and one contained HPV18 (Table 2). Among the HPV-positive cases, 10 of 11 tumors expressed high p16 expression, indicating HPV E7 oncogene activity, whereas of the HPV-positive controls, 3 of 10 tumors failed to express strong p16 (Table 2). Of HPV-positive tumors, 10 of 21 (47.6%) were oropharynx and 11 of 21 (52.4%) were from other sites. Of the 21 HPV-positive tumors, 5 of 11 in the HIV-positive case group and 5 of 10 in HIV-negative control group were from tonsil, base of tongue, or vallecula; other sites of HPV-positive tumors in the cases were tongue, pyriform sinus, mandible, hard palate, alveolar ridge, and medial canthus of the eye. In the HPV-positive controls, tumor sites included tongue (1), floor of mouth (2), conjunctiva (1), and larynx (1; Table 2).

HPV-positive tumors had higher p16 (P < 0.001) and lower RB (P < 0.0001) expression (Figs. 1 and 2). Among HPV-positive tumors, RB was moderately expressed in only 3 of 21 (14%) tumors (Table 2). In contrast, among the HPV-negative tumors (Table 1; Figs. 3 and 4), RB was strongly expressed (IHC score ≥12) in 15 of 59 (25%) samples. In 14 of 15 (93%) RB-positive, HPV-negative tumors, p16 expression was moderate or absent. Expression of TP53 was generally low in the HPV-positive tumors and only 4 of 21 (19%) of these exhibited IHC scores ≥12, and three of these had low to moderate (1, 8, 8.5) p16 scores (Table 2). Of the HPV-negative tumors, 22 of 60 (37%) exhibited strong TP53 staining (Table 1; Figs. 3 and 4). Cyclin D1 expression was absent in all of the HPV-positive cases and was strongly expressed in only 3 of the HPV-negative tumors. Two Cyclin D1–positive tumors and one Cyclin D1 moderate tumor are shown in Fig. 4. EGFR overexpression was common in the entire set with 32 of 82 tumors exhibiting IHC scores ≥12; among the HPV-positive tumors, 9 of 21 had EGFR scores ≥12 (Tables 1 and 2; Fig. 1). HIV-positive patients had higher p16 (7.6 vs. 5.1, P = 0.05) and lower TP53 (5.6 vs. 8.3, P = 0.03) scores than HIV-negative controls. Strong TP53 staining was significantly more common (P = 0.03) in the HIV-negative group (18/41, 44%) than in the HIV-positive group (8/40, 20%; Table 1; Fig. 4). Among HPV-positive tumors, the biomarker expression levels were not significantly different (P = 0.34–0.78) between HIV-positive patients and HIV-negative controls (Table 2 summary).

HPV-positive patients had significantly better survival than HPV-negative patients [P = 0.03, Cox; HR (95% CI) = 0.49 (0.25–0.94)]. There was no significant difference in overall survival in HIV-positive patients compared with HIV-negative patients [P = 0.58, Cox; HR (95% CI) = 0.85 (0.46–1.54]. Among all patients, high TP53 expression was associated with a trend for poorer survival (P = 0.20). In subset analysis, high TP53 expression was significantly associated with decreased survival (P = 0.04) in the HIV-positive cases but not the HIV-negative controls (P = 0.92, Fig. 5). A test for interaction in the combined model was not significant (P = 0.34) failing to provide evidence that the effect of TP53 is significantly different in HIV-positive cases versus HIV-negative controls. Higher TP53 expression in the HPV-positive patients was associated with poor survival (P = 0.04) but the numbers in our sample were very small and the P values should be interpreted with caution.

The goals of this multi-institutional collaborative study were to investigate the role of hrHPV in head and neck cancer in individuals with HIV infection, and compare the rate of HPV involvement in control HIV-negative individuals with similar head and neck cancers. We assessed frequency of HPV involvement and the site of HPV-positive HNSCC in the cases, and then selected controls matched for sex, tumor site, and age at cancer diagnosis from pathology archives. Going into the study, we expected that the HIV-positive cohort might be much more susceptible to HPV-induced head and neck cancers and the proportion of HPV-positive tumors would be much higher in the HIV-infected group. This was based on studies demonstrating the higher prevalence and persistence of oral HPV infection in HIV-positive groups compared with a healthy population (12, 13). Furthermore, it was expected that the HIV-infected individuals with head and neck cancer would have poorer outcome due to immunosuppression from their underlying HIV infection and associated health risks, including higher rates of smoking, intravenous drug use, complications of AIDS, and susceptibility to other infections and inflammation (2, 14–16).

The population for this study was drawn from a collaboration between five head and neck cancer SPORES (Emory, Johns Hopkins, MD Anderson, Pittsburgh, and Michigan). No prior knowledge of HPV status was available. We previously reported (2) the median HIV-HNSCC survival (a cohort that included the same cases in this study) was not appreciably lower than US-HNSCC survival (63% vs. 61%). In that study, we also reported that the median CD4 count in the cohort was 300 cells/μL, which is indicative of good control of the HIV infection in most of the cases. Nevertheless, poorer survival was associated with CD4 > 100 cells/μL. In addition, tumor sites of larynx and hypopharynx, as well as current tobacco use, were adverse predictors of survival (2). Of the 82 subjects included in the current study, 41 cases and 41 controls, we did not see a significant difference in survival between the two groups.

The rate of HPV involvement in the head and neck cancers did not differ significantly between cases and controls, with 11 HPV-positive tumors among the cases and 10 HPV-positive tumors among the controls. However, the small size of some of the tumor biopsies from the HIV-positive cases did not allow for isolation of sufficient DNA for HPV testing. The tumor sites with HPV involvement included the usual preponderance of oropharynx (10/21); however, HPV16 involvement was also found in squamous cancers of the eye in one case and one control, raising a question of how the virus was acquired at this site. Most HPV infections are thought to be acquired by direct sexual contact, although the transfer of infectious virus from one site to another by an autoinoculation by a contaminated finger (17) or by contact with a partner's HPV-infected anogenital tissue or fluids. In cattle, transmission of BPV by fly bites (18) is thought to be a vector in ocular squamous cell carcinomas, so HPV transmission by other means cannot be excluded.

The limited size, fixation conditions, and minimal remaining tissue of many of the biopsy specimens precluded isolation of high quality RNA. Thus, the evidence of viral oncogene activity as a driver of the tumors was assessed by expression of p16 by IHC. By this indicator, only one of the HPV-positive tumors (UP1880, p16 score = 1) among the HIV-infected cases was not driven by the HPV16 viral DNA found in the tumor. Among the HPV-positive tumors from HIV-negative controls, one (UM2495) failed to express p16, and two others (UM1349 and UM 1803) expressed only low levels of p16, raising a question about whether the virus is a driver in these cases or if p16 has been completely or partially inactivated by other mechanisms such as methylation or copy loss (19–22).

Three HPV types that were detected and identified by L1 consensus PCR but not included in the PCR-MA assay or HPV IHC test were HPV26, HPV69, and HPV82. The tumors from the HIV-positive cases UP1864, EU1851, and UM2359 containing HPV26, HPV69, and HPV82 viral DNA, respectively, all expressed high levels of p16, suggesting that all three tumors might be driven by these HPV types, resulting in the high p16 expression. These HPV types are considered possibly carcinogenic (carcinogenic group 2B according to the International Agency for Research on Cancer; ref. 23), and they are grouped together in alpha papillomaviruses group 5 with HPV type 51, which is among the definitive carcinogenic types. HPV26 and HPV82 have been found to be associated with invasive cervical cancer, albeit a rare occurrence: a meta-analysis found 25 positive of 9,265 tested (0.27%) for HPV82, and 8 positive of 6,111 tested (0.13%) for HPV26 (24). Another study examined single HPV type–infected cervical cancer tissue biopsies with probable/possible carcinogenic HPV types and found expression of five markers suggestive of HPV transformation (type-specific E6*I HPV spliced viral transcripts, high p16, low RB, low cyclin D1, and low p53 expression) for those harboring HPV26, and four of these markers (type-specific E6*I HPV spliced viral transcripts, high p16, low RB, and low cyclin D1 expression) for HPV82. Only one of the four HPV82 cervical cancer cases had low TP53 expression (25). The HPV26-positive case in our study demonstrated similar expression to the cervical cases with high p16, low RB, low cyclin D1, and moderate p53 expression, supporting HPV26 as a carcinogenic driver in this tumor. The HPV82-positive case in our study also exhibited the same expression pattern as seen in the previous study: high p16, low RB, low cyclin D1, and high TP53.

TP53 is typically wild type in tumors with HPV as a driver of carcinogenesis (26–28). Generally when TP53 is wild type, little or no TP53 protein expression is seen in the tumor; however, TP53 protein expression is often high in tumors with mutant TP53 (22). Among the HPV-positive tumors from HIV-positive cases, TP53 expression was low or moderate in all but one tumor (UM2359, p53 score = 12). In the HIV-negative control HPV-positive tumors, three tumors expressed high TP53 levels; these were the same tumors that had no or low p16 expression. Taken together, these findings suggest the possibility that mutant TP53, not HPV, might be the primary driver of these cancers. While we did not expect to see indications of mutant TP53 in the HPV-positive cases or controls, a limited number of studies have demonstrated a higher frequency of HPV and mutant TP53 coinvolvement in HNSCC cases than previously reported, which may be dependent on specific tumor site or cocarcinogenic factors such as tobacco, alcohol, betel quid, or areca nut use (29, 30) Furthermore, the commonly used HNSCC HPV16-positive cell line VU-SCC-147, established from a floor of mouth tumor removed from a patient with a history of smoking, harbors a TP53 (L257R) mutation (29).

RB expression was also low in most of the HPV-positive tumors, which is consistent with HPV viral oncogene expression (31, 32). Only one control HPV-positive tumor expressed moderate RB (UM1349, RB score = 11). Low RB expression is consistent with HPV oncogene expression; however, many HPV-negative tumors also expressed low RB.

Although survival did not differ between HIV-positive cases and HIV-negative controls, HPV-positive status of the tumor was associated with significantly better survival than that of the HPV-negative cases and controls. As the survival among the HIV-positive cases was slightly better than the HIV-negative controls, HIV status did not adversely affect outcome for HPV-positive patients. Curiously, TP53 expression segregated the HIV-positive subjects for survival. Those with the lowest TP53 expression enjoyed the best survival and those with the highest TP53 expression had significantly poorer survival. In contrast, TP53 expression was not associated with survival for HIV-negative controls.

R.L. Ferris reports receiving commercial research grants from Astra-Zeneca MedImmune, Bristol Myers Squibb, Merck, VentiRX Pharmaceuticals, and is a consultant/advisory board member for Astra-Zeneca/MedImmune, Bristol Myers Squib, Lilly, Merck, and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T.E. Carey, L.A. Peterson, D.M. Shin, R.L. Ferris

Development of methodology: H.M. Walline, T.E. Carey, L.A. Peterson, R.L. Ferris

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.M. Walline, C.M. Goudsmit, L.A. Peterson, J.B. McHugh, S.I. Pai, R.L. Ferris

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.M. Walline, T.E. Carey, C.M. Goudsmit, E.L. Bellile, G. D'souza, J.J. Lee, D.M. Shin

Writing, review, and/or revision of the manuscript: H.M. Walline, T.E. Carey, C.M. Goudsmit, E.L. Bellile, J.B. McHugh, S.I. Pai, J.J. Lee, D.M. Shin, R.L. Ferris

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.A. Peterson, D.M. Shin, R.L. Ferris

Study supervision: T.E. Carey, L.A. Peterson, D.M. Shin, R.L. Ferris

The SPORE HNC network contributed collectively to this study. Biospecimens were provided by the sites and processed by the centralized testing laboratory. In addition to the leading contributions of the authors listed above, other important contributions were made by the following: Pathology contributors: Jonathan B. McHugh, Martin Graham (University of Michigan, Ann Arbor, MI); Raja Seethala, Simion Chiosea (University of Pittsburgh, Pittsburgh, PA); Marina Mosunjac (Emory University, Atlanta, GA); Adel K. El-Naggar (MD Anderson Cancer Center, Houston, TX); William H. Westra (Johns Hopkins University, Baltimore, MD). Data coordination: Jeff Lewis (M.D. Anderson Cancer Center, Houston, TX); Nicole Kluz, Alicia Wentz (Johns Hopkins School of Public Health, Johns Hopkins University, Baltimore, MD); Rachel Moreno (Emory University, Atlanta, GA); James Riddell IV, MD (Medicine-Infectious Disease, University of Michigan, Ann Arbor, MI). The SPORE Directors are listed as follows:

Dong Moon Shin, Director, Emory University Head and Neck Cancer SPORE.

David Sidransky, Director, Johns Hopkins Head and Neck Cancer SPORE.

Jeffrey Myers, Director, MD Anderson Head and Neck Cancer SPORE.

Gregory T. Wolf, Director, University of Michigan Head and Neck Cancer SPORE.

Jennifer G. Grandis, Director, University of Pittsburgh Head and Neck Cancer SPORE.

This work was supported by the NCI HIV Supplement to the Head and Neck SPORE Consortium. This study was also supported by grants to the following institutions: ARRA: University of Michigan: P50 CA097248 R01 CA194536 (to T. Carey); University of Michigan Cancer Center Core Grant P30 CA46592; MD Anderson: 5P50 CA097007; University of Pittsburgh: P50 CA097190; Johns Hopkins University, P50 DE019032 and 3P50 DE019032-14S2; Emory University: P50 CA128613, R01 DE021395 (to G. D'Souza), P50 CA128613, and P50 CA128613-02S1.

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.