Therapeutic Insights from Genomic Studies of Head and Neck Squamous Cell Carcinomas

Head and neck squamous cell carcinomas (HNSCC) are the fifth most common malignancy worldwide and comprise a diverse set of cancers arising in the upper aerodigestive tract mucosa (1). Unlike many other epithelial cancers, the majority of HNSCCs present at a locally advanced stage with cervical lymph node metastases. More than 90% of patients are treated with curative intent using a combination of surgery, radiotherapy, and chemotherapy (2). To date, treatment approaches have been dictated by the anatomic site of the primary tumor, with oral cavity cancers treated primarily with surgical resection and pharyngeal and laryngeal tumors with chemoradiation (3). Although over one half of patients are cured with initial therapy, these treatments are highly morbid, and therapeutic options for individuals who relapse following initial treatment are limited (4). In the absence of information regarding the biologic underpinnings of individual tumors, predictive biomarkers have been lacking to guide therapy in both the initial treatment setting and in the treatment of relapsed/refractory disease. Establishing robust therapeutic biomarkers in HNSCCs has been challenging for several reasons, including the heterogeneity of these tumors that display diversity in terms of their anatomy, clinical characteristics, and in their association with conventional risk factors, such as tobacco and alcohol exposure, as well as with infection with the oncogenic human papillomavirus (HPV) and Epstein–Barr Virus (EBV; refs. 2, 5). Recent large-scale genomic profiling studies, notably that of The Cancer Genome Atlas (TCGA), have shed light on the molecular underpinnings of the diversity of HNSCCs. We highlight some of the key insights from this and other studies and the implications of these findings on our understanding of HNSCCs and therapeutic approaches.

Despite advances in surgical approaches, radiotherapy, and chemotherapies, treatment outcomes for patients with HNSCCs associated with the traditional risk factors of tobacco use, alcohol exposure, or both, remain disappointing as compared with patients with HPV-driven disease (5, 6). This clinical challenge has stimulated genomic studies focusing on this high-risk group of patients with the goal of identifying molecular aberrations that could be targeted to improve clinical outcomes. However, at this time, there are no agents in clinical use in HNSCCs that show enhanced activity associated with a genetic biomarker. Although EGFR is overexpressed in HNSCC and has been shown to be associated with reduced survival, EGFR-directed therapies have not been especially efficacious (7, 8). Cetuximab, a monoclonal antibody directed against EGFR, is the only FDA-approved molecularly targeted agent for HNSCC, but response rates to this agent given as monotherapy are approximately 10%, and it remains unclear how to predict the subset of patients most likely to respond to cetuximab or other EGFR-directed therapies, despite a large number of studies addressing this topic (7, 9, 10).

Next-generation sequencing (NGS) studies of non–HPV-driven HNSCCs, including the TCGA project that characterized nearly 250 of these individuals (TCGA Network, Nature, in press), have demonstrated a complex landscape of alterations in gene expression, DNA copy number, somatic mutations, gene rearrangements, and gene promoter methylation (11–15). These tumors are characterized by the near-universal loss of TP53 and CKDN2A/RB1 by truncating mutation, deletion, and/or alternative splicing. A summary of somatic alterations in genes regulating a number of key cellular pathways in HPV-negative and HPV-positive HNSCCs is presented in Table 1.

Of note, HPV-negative HNSCCs most closely resemble lung squamous cell carcinomas in terms of their spectra of genomic alterations and contain statistically enriched mutations and copy-number alterations in genes regulating many of the same pathways, in addition to widespread loss of both TP53 and CDKN2A/RB1 (16). These include regulation of squamous differentiation (NOTCH1, RIPK4, IRF6, and TP63), oxidative stress (NFE2L2 and KEAP1), WNT signaling (AJUBA and FAT1), immune evasion (HLA-A, B2M, and TGFBR2), and chromatin remodeling (MLL2 and NSD1; TCGA Network, Nature, in press; ref. 15). Although 20% of HPV-negative HNSCCs in the TCGA cohort displayed amplifications of receptor tyrosine kinase (RTK) genes, such as EGFR, ERBB2, MET, and FGFR1, there were no recurrent mutations or fusions in RTK genes that have been associated with dramatic responses to small-molecule kinase inhibitors in other tumor types, such as lung adenocarcinoma. One possible exception is oncogenic exon 14 skipping in MET, which was reported in two HNSCC cases by TCGA and is found in 4% to 5% of lung adenocarcinoma, and which may be associated with sensitivity to MET small-molecule inhibitors.

Mutually exclusive mutations in RAS family genes, notably RHOA, KRAS, and HRAS, are present but infrequently found in HNSCC (6% of TCGA cases), though recurrent mutations in RHOA at amino acid position 40 are worth noting; however, the biologic significance of these RHOA mutations is unclear. Amplification of chromosome 3q, a region containing the TP63, SOX2, and PIK3CA genes, is seen in the majority of both HPV-negative and HPV-positive HNSCCs, and PIK3CA mutations are commonly found in both HPV-negative and HPV-positive disease, in agreement with prior studies (11, 14, 15, 17).

HPV-negative HNSCCs arise from a number of anatomic sites, including the larynx, oral cavity, and oropharynx, and generally occur in the setting of heavy alcohol and/or tobacco exposure, or less commonly in patients without these well-established risk factors. The TCGA cohort did not identify any mutated genes specific to an anatomic site, though the numbers of cases in each of these categories was insufficient to comprehensively address this question. It should be noted that a prior report suggested that TERT promoter mutations are enriched in tongue cancers (18). In contrast with lung cancers in which many targetable genomic alterations have been identified specifically in patients who lack exogeneous carcinogen exposure in the form of tobacco, two small studies of HNSCCs arising in HPV-negative individuals with minimal tobacco or alcohol histories did not identify any recurrent kinase alterations (19, 20).

HPV-negative HNSCCs demonstrate clear evidence of molecular diversity, as suggested by expression profiling studies that clearly demonstrate diverse biologic subclasses within HPV-negative disease, including a class of tumors without EGFR amplification and/or overexpression, previously termed “atypical” HNSCCs, which consist of approximately 20% of HPV-negative cases and the vast majority of HPV-positive HNSCCs (21, 22). An intriguing mutational pattern identified by TCGA was a subset of HPV-negative HNSCCs originating in the oral cavity with few to no copy-number alterations statistically enriched for HRAS, CASP8, and PIK3CA mutations and lack of TP53 mutation (TCGA Network, Nature, in press).

It is well accepted that the clinical features of patients with HPV-driven HNSCC are distinct from those of patients with HPV-negative disease. These include distinct sites of origin of the disease (e.g., tonsil and base of tongue), younger age, and improved relapse-free survival following initial definitive treatment. Clinically, an emphasis on protocol development for patients with HPV-driven HNSCCs has consisted largely of de-escalation of standard therapy with a lower dose of radiation, more limited resection, or deintensification of chemotherapy.

Initial NGS studies of HPV-positive patients confirmed that these individuals harbor few genomic alterations in TP53 and/or CDKN2A, presumably due to the activity of the HPV E6 and E7 viral oncoproteins (11). These studies also suggested that HPV-driven HNSCCs display less genomic complexity as compared with HPV-negative disease, though TCGA and other more recent cohorts did not confirm this finding, perhaps due to tobacco use in the HPV-positive individuals in these studies. This possibility is supported by the prevalence of both the virally associated Tp*Cp(A/C/T) substitution mutation in the HPV-positive individuals as well as CpG transversions, a mutation class typically associated with smoking. A major limitation in most studies reported to date has been the relatively small numbers of characterized HPV-positive tumors.

HPV-driven HNSCCs are distinct from HPV-negative disease in that they lack focal RTK amplifications but do display a higher rate of focal PIK3CA amplification and mutation. PIK3CA alterations have been reported as therapeutic biomarkers in this patient population based on cell line and patient-derived xenograft studies (14). HPV-associated HNSCCs also demonstrate enrichment for copy-number gains in TRAF3 and E2F1 and a lack of CCND1 amplification when compared with HPV-negative disease.

HPV-driven cancers display both mutations and fusions in the FGFR3 gene, with mutations at position 249 reported at 14% in one study of 50 cases of locoregionally advanced disease, and FGFR3–TACC3 fusions have been reported in multiple cases by the TCGA and other groups (15, 23). These two FGFR3 alterations have been associated with therapeutic response to FGFR small-molecule inhibitors in preclinical (24) and clinical studies (25, 26). TCGA did not detect any genes displaying statistical enrichment for mutation in HPV-positive individuals as compared with HPV-negative individuals, though B2M truncating mutations most closely approached significance. In addition to FGFR3 mutation, other studies have identified the RNA helicase DDX3X, which is mutated in medulloblastoma and is a regulator of β-catenin, as a gene more commonly altered in HPV-positive individuals (15). HPV-positive disease shares many common altered genes and pathways with HPV-negative HNSCC (e.g., NOTCH, MLLs, RAS, and WNT; ref. 15).

In addition to the genomic context in which HPV resides, it has become increasingly clear that the virus itself plays an important role in oncogenesis beyond expression of E6 and E7 (27, 28). Studies of the interaction of HPV with the human genome in HNSCCs have shown that HPV may be present in integrated or nonintegrated forms, may be associated with the presence or absence of ongoing E6/E7 expression, and also may be present in extrachromosomal elements. Sites of HPV integration are nonrandom, occurring in gene- and microRNA-rich areas of the genome as well as in sites that are commonly associated with somatic copy number alterations. HPV integration can have a profound impact on local gene structure and function and result in high-level amplifications, gene disruptions, alternative splicing, novel gene fusions, and changes in global promoter methylation and transcription. An intriguing finding in the field of HPV integration is recurrent disruptive integration in the RAD51 gene, perhaps facilitating further HPV integration by hindering DNA repair (29). Although the study of host genome–HPV interactions is still maturing, it is clear that the role of HPV extends beyond the production of E6 and E7 in HNSCCs.

A very important feature of the TCGA data and other cohorts is the demonstration that HPV may be detected in HNSCCs using a number of methods, including mass spectrometry and massively parallel RNA and DNA sequencing, and that these methods are far more sensitive than those currently applied in the clinic. These methods identify patients without conventional clinical features (e.g., larynx cancers and p16-negative) who appear to have HPV-driven disease. As NGS methods are increasingly applied in the clinic, it will be a challenge moving forward to further define the sensitivity and specificity of these newer methods for HPV detection and if there is any significance related to quantitative differences in the amount of HPV detected in a given tumor. Although p16 immunohistochemistry is a simple assay, these recent studies suggest that it should not be regarded as an appropriate surrogate for direct assessment of HPV status. p16 assessment alone will misclassify individuals in whom HPV is present in the absence of E6/E7 expression, and more importantly, will overlook HPV in tumors in which both HPV is present and p16 is lost by an independent mechanism, an especially relevant issue in patients with both HPV and a tobacco history. For studies moving forward, it will be critical to accurately determine HPV status by direct measurement of HPV to avoid systematic errors in patient classification and stratification.

The application of targeted therapeutics in HNSCCs has been disappointing to date as compared with other cancer types. This has been due, in part, to the slower development of therapeutic biomarkers and a lack of understanding of the genomic landscape of these diseases. As noted above, cetuximab is the only targeted agent approved for HNSCC; its use in the metastatic setting is associated with a low response rate of 10% to 15%, and most studies have failed to find an association between EGFR expression and/or gene amplification with response to EGFR inhibitors, including cetuximab, in HNSCC cohorts. There are no prospectively validated biomarkers to enable the selection of patients for EGFR-directed therapy.

Before the TCGA and other recent studies, earlier reports noted the prevalence of a number of genomic alterations in HNSCCs, which are associated with therapeutic response to targeted agents in other cancer types. These include EGFR mutations and ALK and ROS1 fusions. It is now clear that these events are extremely rare in HNSCC, if present at all, and that the therapeutic opportunities in HNSCCs more closely resemble squamous cell cancers from other tissue types. A few specific examples are discussed below.

FGFRs have been shown to be activated by amplification, mutation, and translocation in a wide range of cancer types. In HNSCC, FGFR1 amplifications are found in HPV-negative patients at a rate of approximately 10% and appear to be enriched in non-oropharynx tumors. FGFR1 amplification has been associated with therapeutic response to FGFR tyrosine kinase inhibitors (TKI) in lung squamous cell cancers, though response rates represent only a modest improvement as compared with chemotherapy (30–32). Multiple explanations for this disappointing result have been reported, including the presence of co-mutations activating the RAS–MAPK pathway, a lack of correlation between FGFR1 amplification and expression or activation of the protein, and difficulty in standardizing assays for detection of amplification by FISH or NGS methods. Several early-phase clinical trials are ongoing or planned in patients with HNSCC with FGFR1 amplifications who have relapsed/refractory HNSCC (e.g., NCT01962532, NCT01004224, and NCT01948297), though no public data have been reported on efficacy. Given that FGFR inhibitors appear to be well tolerated and may also be radiosensitizers, the combination of these agents in the curative treatment setting with the current standard of care may be reasonable in high-risk HPV-negative patients.

FGFR2 and FGFR3 mutations and FGFR3–TACC3 fusions are of particular interest, as these genetic lesions have been associated with dramatic responses to FGFR TKIs in preclinical models and in early-phase clinical settings, including a case report of a dramatic response to pazopanib in a patient with an FGFR2-mutated tongue cancer (24). In contrast to FGFR1 amplification, FGFR2/3 mutations and FGFR3 fusions appear to occur largely in HPV-positive individuals at a prevalence of 10% to 20%, and clinical trials are currently targeting this patient population. However, given that these trials are focusing on patients with relapsed/refractory disease, they may encounter difficulty with accrual, as the number of HPV-positive patients who are candidates for such studies is small.

PIK3CA is commonly amplified and/or mutated in patients with HNSCCs (37% of cases in the TCGA), and PIK3CA alterations are enriched in HPV-positive patients. If one examines the PI3K–AKT–mTOR pathway in detail, more than one half of patients with HNSCC have a somatic alteration that can activate this pathway. As such, there is tremendous interest in developing small-molecule inhibitors of components of this pathway for individuals with HNSCC, and ample preclinical data suggest that this may be an effective therapeutic strategy, though it should be noted that the activity of PI3K inhibitors as monotherapy in lung squamous cell cancers in patients with PIK3CA or PTEN mutations has been disappointing. However, initial data have been more encouraging when combining these agents with chemotherapy or other targeted agents.

Clinical concepts moving forward include both recruiting patients with relapsed/refractory disease with PI3K pathway lesions as well as using these agents in “window of opportunity” trials in the upfront setting or in combination with chemoradiotherapy (e.g., NCT01816984, NCT01195922, NCT01852292, and NCT01133678). Preclinical data have shown that inhibition of the PI3K pathway may sensitize cancer cells to radiation and that PI3K inhibitors may be most efficacious as radiosensitizers in patients with NFE2L2 or KEAP1 mutations, genomic events commonly seen in high-risk HPV-negative individuals (33). A clinical trial in this high-risk population is now ongoing based on these preclinical data (NCT02113878).

In the TCGA study, 32% of HNSCCs displayed an amplification or mutation of CCND1, CDK4, or CDK6, with the majority of these alterations found in HPV-negative patients. It should be noted that the genomic region on chromosome 11 containing the CCND1 locus also contains other cancer-related genes such as FADD, and it is not clear that CCND1 is the focus of amplification in all cases with chromosome 11q13 amplification. However, impressive early clinical data in breast cancer have suggested that CDK4/6 inhibitors may be effective in patient cohorts with high rates of CCND1 amplification and that these agents are well tolerated as both single agents and in combination with other therapies. Clinical trials are moving forward at this time in other cancer types with frequent alterations of CCND1/CDK4/CDK6, and a subset of these trials will include patients with relapsed/refractory HNSCC (NCT02101034). Many HPV-negative patients with CCND1/CDK4/CDK6 amplifications also harbor RTK amplifications, suggesting that combination strategies may be needed in this setting.

Immunotherapy approaches have garnered a great deal of excitement in the oncology community based on the early clinical success of immune checkpoint inhibitors in melanoma, renal cell carcinoma, and lung cancer. Trials of immunotherapeutic agents are ongoing in HNSCC in both the initial treatment setting with ipilimumab and for recurrent disease with agents targeting the PD1:PDL1 checkpoint and other immune effectors (e.g., NCT01860430 and NCT01935921), with promising data presented recently with pembrolizumab in relapsed/refractory disease (34). PDL1 (CD274) expression is a biomarker reported by some groups to enrich for response to blockade of the PD1:PDL1 checkpoint, and it has been reported in other cancer types, including lymphoma and gastric cancer, that PDL1 expression is associated with virally induced cancers (35). Although there are several reports on this topic in HNSCCs, with variable results, RNA sequencing data from the TCGA are consistent with higher levels of PDL1 expression in HPV-negative individuals. However, it should be noted that these samples contain both tumor and stroma and it is still unclear what the optimal methods are for PDL1 measurement in the context of patient stratification. In addition to variable expression of PDL1, HNSCCs display a wide range of somatic alterations in genes involved in antigen presentation, inflammation, and immune evasion, including HLA-A, B2M, TGFBR2, and TRAF3. Although the mechanisms governing immune evasion in HNSCCs remain poorly understood, it is likely that somatic alterations in these and other genes are likely to play a key role in immune surveillance of HNSCCs and may affect the responsiveness of cancers to specific immunotherapeutic approaches. One particular pathway of interest in this regard is PI3K, given reports in other cancer types that it may be associated with response to PD1:PDL1 checkpoint inhibitor therapy (36).

A number of additional therapeutic targets have been proposed for HNSCC based on genomic discovery studies and preclinical models. HNSCCs frequently display hyperactivation of STAT3 via a variety of mechanisms, and STAT3 pathway inhibitors are currently being explored both inpreclinical models and in early-phase clinical trials. HNSCCs display frequent degregulation of pro- and antiapoptotic genes such as CASP8, and cell line studies have suggested that inhibitors of BCL2 family proteins may demonstrate activity against HNSCCs. HNSCCs also commonly display concurrent amplification of two or more putative “drivers,” and preclinical studies have demonstrated synergy in combining inhibitors in cell lines that demonstrate multiple activated kinase pathways, such as concurrent EGFR and FGFR pathway activation. Novel EGFR-targeting strategies with more potent anti-EGFR TKIs and antibodies are moving forward in HNSCCs, as are efforts to better define the subset of patients most likely to benefit from anti-EGFR therapy.

HNSCCs are less common than other cancer types in which substantial strides in biomarker-based clinical trials have been made, and the number of clinical trials currently available for individuals with HNSCCs is approximately one half the number of trials for patients with lung or breast cancers. Furthermore, there are no genetic tests routinely incorporated into the management of HNSCC, and patient stratification is largely done based on clinical features and HPV status. With the widespread incorporation of NGS diagnostics into the routine care of patients with a wide array of cancer types, it will be important to consider how this information can be used to improve the efficacy of therapies for HNSCCs, and here we propose a few possible approaches.

First, the molecular alterations identified in the HNSCC TCGA project and other cohorts are not unique and are shared with a number of other epithelial cancer types, most notably lung squamous cell carcinoma. We suggest that clinical trials of novel agents in lung SCCs also include patients with HNSCCs given this overlap, and that “basket trial” approaches be considered across these disease types. Natural candidates for this approach include agents targeting FGFRs, PI3K/AKT, cyclin-dependent kinases (CDK), and immunotherapies. Second, HNSCC would seem to be an ideal cancer type for “window-of-opportunity” trials, given the relative accessibility of tumors to sample and the widespread use of surgical resection as a standard curative approach. In this study approach, a novel agent could be deployed before definitive therapy to assess its efficacy with biopsy samples taken before and during treatment to identify biomarkers of response and resistance. Third, given that we have now defined a number of potential targets in HNSCC, the use of these agents in the definitive setting with analysis of tissue before and after therapy should be conducted so that we might better define cohorts of patients for specific therapeutic approaches in which novel agents may be used in combination with chemoradiotherapy in the initial or adjuvant setting. Finally, it will be critical for the HNSCC community of surgeons, radiation oncologists, and medical oncologists to appreciate the value of a molecular understanding of HNSCCs to facilitate moving the field forward in the era of genomic medicine.

No potential conflicts of interest were disclosed.

P.S. Hammerman is supported by NCI grant K08 CA163677.