Discussion
Here we present the targeted mutation spectrum of a large cohort of HPV-positive OPSCCs. We corroborate previously reported aberrations (e.g.
FGFR,
PIK3,
FAT1,
NOTCH2), but also more rare alterations in HPV-positive OPSCCs (e.g.
KALRN,
KMT2A, and
SPTBN1) [
11,
20,
32]. Our finding of 3459 mutations with an average of 10 mutations per megabase is higher than that in the TCGA HNSCC cohort [
33], which may be explained by the fact that the TCGA cohort is predominately HPV-negative tumours.
Among human cancers,
TP53 is the most frequently mutated gene, and TCGA data has shown that p53 ‘gain-of-functions’ mutants bind to, and upregulate, several chromatin regulatory genes including the methyltransferases MLL1 – also known as
KMT2A – which is highly prevalent in our cohort (31%) [
34]. A chromatin mechanism causal for the progression of tumours with gain-of-function p53 prospects possibilities in the design of combinatorial chromatin-based therapies [
34]. Aberrations in
SPTBN1, that encodes a scaffolding cellular protein involved in the formation of the cytoskeleton, is a useful prognostic biomarker in HPV-negative HNSCCs since patients with tumours expressing
SPTBN1 have four times higher mortality, compared with patients not harbouring this mutation [
35]. Although Zhu et al. examines gain-of-function, our data indicates that this gene is also prevalent in HPV+ OPSCCs and could be prognostic in these tumours as well. The known cancer gene
ZNRF3 belonging to the E3 ubiquitin ligase family was also frequent in our cohort (9% of samples). Previously, it has been reported that
ZNRF3 has the ability to inhibit the metastasis and tumorigenesis by suppressing the
Wnt/β-catenin signalling pathway in nasopharyngeal carcinomas (NPC), hence believed to be a potential molecular target for treatment of NPC. Based on our results, it should also be considered in HPV+ OPSCCs [
36]. Due to availability of existing therapeutics and high prevalence of mutations in HPV+ OPSCC, patients with
PIK3CA and CDK4/CDK6 mutations should be recommend for future phase 0 and I trials. Notably, in our cohort, we merely identified an occurrence of 7 and 1% related to
PIK3CA pathways and
CCND1 aberrations. Further, the
FGFR2/3 mutations are of particular interest because they were present in 20% of the tumours, in particular the S249C mutation, which is an oncogenic driven in bladder cancer [
20]. Upon binding of ligand,
FGFRs activate a cascade of downstream signalling pathways, such as the mitogen activated protein kinase (
MAPK), phosphoinositide-3-kinase (PI3K)/Akt pathways signal transducer and activator of transcription (STAT). The mutated
FGFR isoforms result in oncogenic
FGFR signalling, promoting tumorigenesis, and the defect
FGFR signalling pathway is a major contributing factor in the pathogenesis and progression of HNSCC [
37,
38]. Inhibition of FGFRs is a promising therapeutic strategy, and phase I and II trials are progressing [
39,
40] The tumor suppressor
TRAF3 has previously been brought forward as a potential target for therapy development as it was inactivated in 20% of the HPV-positive tumors in the TCGA cohort. [
10] Interestingly, we only identified 5% tumors with this alteration in our cohort.
A previous study [
41] has focused on mutations in HPV-positive and HPV-negative patients in a mixed population of larynx, oral cavity, oropharynx and hypopharynx cancer patients. In this study, a high prevalence of
PIK3CA alterations (mutations and amplifications) was evident, which is not the case in the present study. We speculate that this could be related to 1) a high prevalence of smoking in the present study, 2) ethnical dissimilarities, and 3) the definition of being HPV-positive (RNA-sequencing versus HPV-DNA PCR). These factors could influence the carcinogenic process (smoking vs. no smoking) as well as perhaps not examining the exact same phenotype (being HPV-DNA PCR positive vs. RNA-seq positive for E6/E7), and perhaps tissue specific mutational processes (oropharynx vs. other head and neck cancer subsites).
Other studies evaluating HPV+ OPSCC consistently report mutations in PIK3CA as well as PTEN, TRAF3, NOTCH1. Although, we did identify these aberrations, they were not among the twenty most prevalent mutations in our cohort. This could be explained by the deduction that these aberrations are not as prevalent in Denmark as other countries, or the fact that our specimens or probes were not adequate set for identifying these mutations.
Interestingly,
IL3RA,
HLA-A and
HLA-B have copy number gains in HPV positive oropharyngeal cancers [
3,
42]. Absence of HLA class I (HLA-A and –B) in HPV+ OPSCC has indicated improved outcome which could be used clinically to select patients for trials with de-escalating therapy. The understanding between the genetic background of OPSCC patients and HLA-traits remains incomplete as several HLA-traits have been associated to altered outcome [
43,
44].
The landscape of HPV+ and HPV- OPSCCs point to different signatures and structural alterations [
11]. To stratify HPV+ from HPV- patients, the most obvious aberrations from our cohort would include
KALRN,
FGFR2 and
NOTCH2 which are rare in HPV- HNSCC. The
NFE2L2 pathway as well as the promising
PIK3CA mutation and
CCND1 amplication should also be prioritized.
From the TCGA cohort, the majority of driver mutations were found to be clonal (e.g. “early” mutations opposed to subclonal viewed as “late” mutations) where
CDKN2A and
TP53 were nearly completely clonal [
45]. McGranahan et al. identified three “early” signatures in HNSCC: 1) C > T transitions at CpG sites associated with spontaneous deamination of methylated cytosines; 2) an APOBEC-signature (also seen in “late” mutations); and 3) a smoking-associated signature. In our material, we also identify the APOBEC signature in 23% of the tumors, likewise reported in other HPV+ HNSCCs series [
46].
We observed a significant increase in the COSMIC signature 12 associated with certain virus-driven liver cancers. Interestingly, a trial related to these aberrations is progressing. Ribavirin, that target the eIF4E translation initiation factor, is used to treat hepatitis C, and is under investigation for recurrent or metastatic OPSCC (
ClinicalTrials.gov, NCT02308241). Ribavirin is also being evaluated along with induction chemotherapy including afatinib (a tyrosine kinase inhibitor) and weekly carboplatin/paclitaxel for stage IV HPV-associated OPSCC (
ClinicalTrials.gov, NCT01721525). Additionally for HPV+ OPSCC patients with recurrent or metastatic disease, rigosertib (a
PI3K (phosphatidylinositol-3 kinase) and
PLK (Polo-like kinase) signalling pathway inhibitor) is being investigated in a phase II trial (
ClinicalTrials.gov, NCT01807546), and in a phase I trial used as initial treatment before platinum based RT-C (
ClinicalTrials.gov, NCT02107235). For high risk HPV+ OPSCC patients, the
PI3K inhibitor, BYL719, is being tested with induction paclitaxel and cisplatin followed by T-site surgery, neck-disssection, and with post-operative risk adapted IMRT (
ClinicalTrials.gov identifier, NCT02298595).
Although this study is strengthened by a setup, that includes a deep coverage, a high number of genes, and HPV and p16 status of all patients to ensure HPV-active infections, some limitations should be noted. First, there may be a selection bias both in patients but also because we employed a gene panel. Moreover, the particular selection of genes might influence the findings, since we have included quite large genes (e.g. APOB) where an alteration is more likely opposed to smaller genes. The sub analysis including tobacco merely includes 70 patients due to missing data and a higher number of patients might lead to other findings. Finally, as described in “
Methods” section, we strained to reduce artefacts from the use of FFPE tissue (e.g. in the data analysis and tissue preparation) although it should be included as a probable source of bias. Tumor mutational burden (TMB) in the present study was defined as the median number of mutations per megabase examined in the targeted sequencing. As the present study concerns genes previously known to be implicated in carcinogenesis, the TMB would be expected to be higher than studies concerning TMB from e.g. exome-sequencing studies, as there would be fewer mutations per examined base.
When breaking down cosmic signatures from the non-smokers and smokers, we did not observe any difference in enriched cosmic mutational signatures. When looking at mutational signatures in the APOBEC enriched and non-enriched groups, the signature 5 was not enriched in the APOBEC-enriched group, but rather the signature 2. Signature 2 has been attributed to activity of the AID/APOBEC family of cytidine deaminases, and has been related to viral infections.
Both tobacco smoking and defects in DNA repair are known to induce a large number of genetic aberrations, and may be distinct ways to accumulate genetic aberrations required for the emergence of cancer. We found a significant difference between smokers and non-smokers underlining the importance of including tobacco-smoking consumption in prediction models and risk-stratifications. It is likely that the HPV-positive smokers acquire tobacco-related mutations but maintain virus related signatures. Even when stratified on smoking-consumption (none-smokers vs. heavy smokers), a large inter-tumor heterogeneity exists. Thus, if the future aim is to offer a personalized treatment approach it may require a very large battery of anticancer targeted drugs. Although fast-moving technologies have prompted the capability of identifying genetic aberrations promptly and precisely, it remains largely unknown which therapy(−ies) to offer based on the combinations of driver mutations. In order to fuel the development of targeted clinical trials and diagnostic testing, confirmatory studies addressing genetic aberrations in HPV+ OPSCC are needed.