Human Papillomavirus-Related Head and Neck Cancer

Human papillomavirus(HPV)-related carcinogenesis has been attributed to a subset of head and neck cancers [1, 2]. Those cancers comprise squamous cell carcinoma of the head and neck (HNSCC), and primarily arise in the oropharynx (OPSCC) and especially in the tonsils [3, 4]. HPV-related HNSCC are characterized by absent somatic mutations in TP53, deregulation of retinoblastoma protein(Rb)-mediated cell cycle control, and overexpression of the tumor suppressor protein p16^INK4a.

The epidemiology of HPV-related HNSCC is variable and highly dependent on tumor subsite and region. It is assumed that HPV-related HNSCC have different risk factors compared to HPV-negative HNSCC, e.g., sexual behavior instead of tobacco and alcohol consumption. A trend towards higher smoking prevalence from north to south is reported for Europe [5], while the opposite trend in HPV prevalence persists. Although smoking is decreasing, the overall prevalence of OPSCC is remaining constant or even increasing in certain countries over time. This trend of an increase in HPV-related and a decrease in HPV-negative OPSCC was published for the USA in 2011, with a total HPV prevalence in OPSCC of 70% in the last time period evaluated between 2000 and 2004 [6]. In a meta-analysis of data for 12,163 HNSCC patients from 44 countries, the pooled HPV DNA prevalence was 46% for oropharynx, 22% for larynx, and 24% for oral cavity. The overall HPV-attributable fraction was higher in North America (60%) compared to Europe (41%) [7]. A recent review confirmed an increase in OPSCC incidences in many countries and mainly in men [8]. The impact of tumor site and region was also confirmed by an international study with uniform HPV testing of 3,680 specimen from pharyngeal, oral cavity, and laryngeal cancers. Inconsistent with the above study, a higher HPV-attributable fraction was found in women for cancers of the oropharynx in Europe [9].

Numerous papillomaviruses are known, with more than 150 mostly low-risk types in humans. Currently, 15 high-risk HPV types have been identified in precancerous and cancerous lesions, which are considered causative for cancer. HPV infects undifferentiated cells in the basal layer of epithelia through microwounds. HPV DNA replication is coupled with cellular DNA replication and maintained at a low copy number as long as infected cells remain at the basal membrane. Viral gene expression is linked to the differentiation process when infected cells are moved towards the epithelial surface, involving regulatory ‘E' proteins (E1-E8) expressed early in the HPV life cycle. Two proteins expressed late in the life cycle (L1 and L2) build the capsid structure of the viral particles which are released together with cells of the uppermost epithelial layers. In HPV-related cancers, the viral life cycle is interrupted as cancer cells remain in an undifferentiated state. Thus, viral replication cannot be completed, and infectious particles are not produced and released.

Rising incidences of HPV-related OPSCC have been attributed to changes in sexual behavior [6, 10]. Nevertheless, although HPV is highly associated with tonsillar cancer [4, 11], less than 1% of tumor-free tonsillar tissue harbors HPV DNA [12]. Even in women with genital HPV infections, HPV DNA detection in the oropharynx is low (5.7%) and comparable to HPV DNA detection rates in females negative for genital HPV (5.1%), suggesting that sexual transmission of HPV from the cervix uteri to the oropharynx is rare [13]. However, in a recent study in patients undergoing tonsillectomy for benign indications, the HPV prevalence was 3.6% in tonsil brushings and 13.1% in gargles (with an HPV16 prevalence of 2.2 and 4.1%, respectively), indicating significant differences caused by sampling techniques [14] which may account for inconsistency in this field.

Unlike cervical cancer, HPV-induced precancerous lesions are absent in the upper aerodigestive tract. Tonsillar cancers may be surrounded by p16-positive mucosal fields, being an indicator for HPV16 in OPSCC and adjacent dysplasia [15]. However, transcriptionally active HPV could not be detected in the mucosa surrounding HPV-positive OPSCC, suggesting the absence of HPV-induced field cancerization [16].

Despite lacking knowledge about the exact ways of oral HPV infection and the very first steps in viral transformation, HPV-driven carcinogenesis is well understood at the molecular level (fig. 1). The same pathways altered in the natural life cycle are involved in HPV-related carcinogenesis and similar to HPV-negative cancers, but deregulation is largely based on protein-protein interactions rather than driven by genetic alteration.

The hallmark of HPV-related carcinogenesis is viral protein-mediated deregulation of cell cycle control and interruption of p53-dependent apoptosis activation. Both depend on the activity of the HPV oncoproteins E6 and E7 (fig. 1). E7 targets Rb for ubiquitin-mediated proteasomal degradation. This releases and activates E2F transcription factors, driving S-phase gene expression and cell cycle progression. Unlike previously thought, p16^INK4A overexpression observed in HPV-related cancers (fig. 1a, b), is not a consequence of this Rb inactivation but triggered by a cellular senescence response via activation of the histone lysine demethylase KDM6B, which has been directly linked to E7 expression in cervical carcinoma cell lines [17].

E6 binds to p53 and recruits ubiquitin ligase E6AP. This leads to p53 ubiquitylation, its proteasomal degradation, and inhibition of p21. As a consequence, growth arrest is released, p53-induced apoptosis is inhibited, and DNA damage repair and cellular senescence are deregulated. Further, induction of the hTERT promoter by E6 via interaction with E6AP (and other proteins) is known, and direct E6/hTERT interactions mediating telomerase activation have been shown recently [18]. Interestingly, p21 expression is described as another surrogate marker of HPV-related tonsillar cancer being strongly associated with favorable patient outcome [19], although reduced p21 expression could be expected with E6-mediated p53 inactivation. Results of other studies are somewhat inconsistent, but a recent study indicates that E7 bypasses the inhibitory effect of p21 on cell cycle progression [20].

Other proteins with altered expression are epidermal growth factor receptor (EGFR), Baculoviral IAP repeat-containing protein 5 (survivin), and metabolic proteins which are related but not essential for HPV-induced carcinogenesis. Survivin is negatively regulated by p53. It suppresses apoptosis and plays a role in cell division. Nuclear survivin expression was correlated with a poor disease-free survival rate and negative HPV status in OPSCC [21].

Thioredoxin (TRX) and epidermal fatty acid binding protein (E-FABP) were identified by proteomics to be upregulated in HPV-related tumors [22]. TRX is a redox mediator promoting cell survival, e.g., under hypoxic conditions and oxidative stress. It has been linked to treatment resistance. E-FABP was discovered as a lipid carrier in human epidermis, but data indicates that E-FABP plays a role in keratinocyte differentiation and other cellular signaling processes; its role in HNSCC is uncertain.

Expression of the EGFR has been in focus for its potential as a molecular marker and therapeutic target in many malignancies. EGFR is normally expressed in low quantities on the surface of most cells. Overexpression is reported for HNSCC and associated with treatment resistance, aggressive clinical behavior, and poor prognosis. A trend toward better prognosis has been shown for EGFR-negative OPSCC, and HPV-related OPSCC tended to have decreased EGFR expression in the same study [23]. Sequencing data support this finding, since amplification of EGFR and fibroblast growth factor receptor 1 (FGFR1) appears to be restricted to HPV-negative HNSCC [24, 25, 26, 27]. Currently, replacement of classical chemotherapy by therapeutic EGFR antibodies is being investigated in clinical trials for HPV-related HNSCC (NCT01855451, NCT01302834, NCT01874171).

Linearization of the viral DNA within the E2 open reading frame abrogates the controlling function of E2 on E6 and E7 protein expression. Thus, viral DNA integration was considered a prerequisite for HPV-driven cancer (fig. 2). Although this classical model is reasonable, in more than 60% of HPV-associated OPSCC only episome-derived DNA was detected by polymerase chain reaction (PCR) [28]. Also, viral oncogene expression has been shown to be independent of viral copy number and DNA integration [29]. This was confirmed by RNA-Seq and whole genome sequencing data showing that 3 possible states of the HPV genome (episomal, integrated, and a hybrid of both forms) were evenly split in HPV16-positive HNSCC [30], indicating overlapping mechanisms in HPV oncoprotein regulation (e.g., by methylation of E2 binding sites (E2BS)).

Although the role of HPV DNA integration may be challenged, chromosomal instability seems to be associated with HPV DNA integration and cancer progression [31]. In an in-vitro study of HPV-transfected keratinocytes, integration sites were detected in various chromosomal locations but also in or near genes encoding growth control proteins [32]. This supports the idea of clonal selection by HPV integration towards more aggressive tumors. Interestingly, HPV DNA integration in the AKR1C3 gene (a hydroxysteroid dehydrogenase involved in the regulation of androgens and estrogens) has been attributed to malignant transformation in a patient with HPV type 6-related juvenile-onset recurrent respiratory papillomatosis [33]. However, in tonsillar cancer, HPV DNA integration was found to be associated with a favorable prognosis [31].

Certain genetic alterations are shared in HNSCC independent of HPV, while others are restricted to either HPV-related or HPV-negative HNSCC [34]. Chromosomal amplifications of 3q, 8q, and 20p have been frequently detected in HNSCC, irrespective of HPV status, by comparative genomic hybridization in earlier studies [35] and by sequencing later on [24, 26, 36]. Important genes like PIK3CA, TP63, and SOX2 are located in the 3q26-28 region. HPV-negative HNSCC show frequent loss of chromosome arms 3p and 9p, and amplification of 11q13 [35, 37]. CDKN2A and CCND1 are located on 9p and 11q13, encoding p16^INK4A and cyclin D1. Both are involved in Rb signaling and therefore, unsurprisingly, less frequently altered in HPV-related HNSCC. The same applies for TP53 mutations which are found in 60-80% of HPV-negative but not in HPV-related HNSCC. Also, amplifications of EGFR and FGFR1 appear to be restricted to HPV-related HNSCC [24, 25, 26, 27]. In contrast, deletion (14%) or mutation (8%) of tumor necrosis factor receptor-associated factor 3 (TRAF3, encoded on 14q32.32), involved in the innate and acquired antiviral immune response, is overrepresented in HPV-related HNSCC [24]. Innate immunity appears to be significant for HPV-related HNSCC, since genes coding for HLA I components are frequently mutated and higher numbers of CD56-positive natural killer cells have been reported for HPV-related OPSCC recently [38].

Currently available data indicates that higher TpC mutation frequencies are found in HPV-positive HNSCC, and CpG transversions in HPV-negative HNSCC, while the overall mutation rate does not differ by HPV status [24]. However, regarding the still small number of genome-wide analyzed samples, general statements should be handled with care. For example, KRAS mutations were found in 5.8% (3/51) of HPV-positive tumors in a comparative genomic analysis of HPV-positive and -negative HNSCC [36], which could not be confirmed in another study [39].

So far, microRNA (miRNA) expression analyses according to HPV status have been performed in 5 studies for HNSCC [40, 41, 42, 43, 44]. In the most recent study, 15 HPV-negative and 11 HPV-related OPSCC were analyzed by microarrays targeting 1,719 sequences. 25 differentially expressed miRNAs were detected. Biological importance was in silico predicted for PI3K and Wnt signaling, focal adhesion, and cytoskeleton regulation [40]. However, only 11 differentially expressed miRNAs were reported in more than 1 study, with 2 of them (miR-381 and miR-101) regulated in opposite directions (table 1). It is important to note that different methods have been used and that the number of analyzed miRNA sequences ranges between 96 and 1,719 among the studies.

Even HPV-encoded miRNAs have been identified and experimentally validated. Their putative target genes are within the viral genome itself [45], but reciprocal interactions between viral and host genes and miRNAs can be assumed. Recently, a target site for human miRNAs (miR-875 and miR-3144) was found in the HPV E6 gene. Both miRNAs inhibit E6 oncogene expression, and in HPV16-positive cell lines growth was inhibited and apoptosis promoted by high-level expression of both miRNAs [46].

Modification, e.g., by acetylation, methylation, ubiquitylation, or phosphorylation, alters histone structure and binding to DNA. Therefore, chromosomal regions become more or less accessible for transcription. Interactions of HPV E7 with histone-modifying proteins are known. One consequence is p16^INK4A overexpression due to the activity of histone lysine demethylase KDM6B as described above [17]. Further, E7 recruits histone deacetylase HDAC1 and histone demethylase KDM5B to the Toll-like receptor 9 (TLR9) regulatory region and to promoters containing interferon regulatory factor 1 (IRF1) response elements, both involved in immune response, apoptosis regulation, DNA damage, and tumor suppression [47, 48].

In contrast to the classical HPV DNA integration-based idea of HPV E2 deregulation, and in particular in tumors with non-integrated HPV genomes, methylations at E2BS in the upstream regulatory region of E6 and E7 appear to regulate their expression in the presence of E2 [49, 50]. DNA methylation patterns have been published for viral-driven cancers including HPV-associated neoplasms [51, 52]. Based on viral DNA integration and the E2BS methylation status, distinct subgroups of HPV-associated SCC causally linked with viral oncogene expression have been identified [28, 49, 50]. Interestingly, the clinical outcome of HNSCC patients, including HPV-related OPSCC, could be predicted depending on differentially methylated promoters of 5 host genes (ALDH1A2, OSR2, IRX4, GRIA4, and GATA4) [51, 53].

HPV diagnosis relies on HPV DNA detection and proof of its oncogenic activity. PCR-based methods are advantageous compared to in-situ hybridization with regard to time, cost, and sensitivity. HPV-typing can be performed using amplified DNA for hybridization to type specific probes or for type determination by sequencing. Additional p16^INK4A immunostaining or HPV mRNA detection can compensate the lower specificity of HPV DNA PCR. Although HPV mRNA detection is considered the ‘gold standard' of HPV diagnosis, mRNA expression is not necessarily equal to protein expression or activity. In contrast, cellular p16^INK4A overexpression has been directly linked to viral oncogenic activity, and p16^INK4A overexpression is rare in other cancers. Although p16^INK4A expression can also be detected in tumor-free tonsils [12], p16 immunostaining has become a reliable surrogate for HPV-driven cancers, especially for high-risk HPV-induced HNSCC and OPSCC [11, 15, 54, 55] and has recently been introduced into TNM staging [56]. HPV serology as an alternate biomarker also seems to be relevant to the etiology and natural history of HPV16-related OPSCC. HPV16 E6 antibodies were found to be strongly associated with HPV16-related OPSCC in a large case control study [57], but further studies are needed in this field.

It is accepted that HPV is a risk factor and indicator of improved survival in HNSCC patients. Nevertheless, in patients with advanced non-resectable tumors, HPV status had a positive impact on progression-free survival but not on overall survival [58]. This indicates that in subsets of HPV-related HNSCC, additional factors may overcome the influence of HPV, and stratification strategies have to be refined.

Treatment is either with a focus on radiotherapy or upfront surgery, depending largely on regional/national preference rather than on evidence-based criteria, with no clear advantage of one over the other. Currently, de-escalation is being evaluated in several clinical trials [59]. For example, in ECOG3311, an intermediate risk group of OPSCC patients is randomized to receive a reduced or standard radiation dose (50 vs. 60 Gy) after transoral tumor resection (NCT01898494). In the ADEPT trial, patients with complete tumor resection but with extracapsular spread are randomized to receive platinum-based radiochemotherapy versus radiation only (NCT01687413). Both studies are ongoing.

Recent molecular findings suggest novel treatment options, especially in HPV-related HNSCC, using demethylating agents or PI3K and immune checkpoint pathway modulation. However, the identification of suitable patient subgroups for targeted treatment is still under investigation.

In contrast to cancers associated with other risk factors, HPV-related HNSCC are driven by viral oncoprotein activity and have individual profiles regarding protein expression, and genetic and epigenetic alterations. Patients with HPV-related HNSCC have improved survival compared to those with HPV-negative tumors, but no specific or toxicity-reduced treatment modalities have been established for this tumor entity so far. Although the still incomplete and partially inconsistent data in this field needs further study, certain features of HPV-related HNSCC have the strong potential to be implemented in future diagnostic and therapeutic concepts.

The authors declare no conflict of interest.

Steffen Wagner and Shachi Jenny Sharma equally contributed as first authors. Claus Wittekindt and Jens Peter Klussmann equally contributed as senior authors.