Key points

Introduction

Advanced head and neck squamous cell carcinoma (HNSCC) can involve multiple sites of the upper aerodigestive tract, often precluding surgical intervention with curative intent. Additionally, a substantial proportion of patients with HNSCC will progress despite traditional cytotoxic chemotherapy and radiation therapy, and locoregional recurrence following any initial treatment of advanced tumors is relatively common. In concert with efforts to develop new chemotherapy regimens, radiation protocols, and surgical approaches, much work has been focused on understanding the immunobiology of HNSCC and on developing strategies to promote an antitumor immune response.

The idea that the immune system may be able to recognize and control cells undergoing malignant transformation has been around for more than a century. Recent robust and durable clinical responses with immune checkpoint blocking antibodies have led to a resurgence in enthusiasm for this therapeutic strategy for solid malignancies.

Introduction

Advanced head and neck squamous cell carcinoma (HNSCC) can involve multiple sites of the upper aerodigestive tract, often precluding surgical intervention with curative intent. Additionally, a substantial proportion of patients with HNSCC will progress despite traditional cytotoxic chemotherapy and radiation therapy, and locoregional recurrence following any initial treatment of advanced tumors is relatively common. In concert with efforts to develop new chemotherapy regimens, radiation protocols, and surgical approaches, much work has been focused on understanding the immunobiology of HNSCC and on developing strategies to promote an antitumor immune response.

The idea that the immune system may be able to recognize and control cells undergoing malignant transformation has been around for more than a century. Recent robust and durable clinical responses with immune checkpoint blocking antibodies have led to a resurgence in enthusiasm for this therapeutic strategy for solid malignancies.

Immune checkpoint blockade

Immune checkpoints are inhibitory pathways critical for self-tolerance under normal circumstances. It is now clear that tumors often co-opt these pathways by expressing cognate ligands for inhibitory checkpoint receptors and are thus able to induce immune tolerance and suppress responses by tumor-infiltrating lymphocytes (TIL), which would otherwise be activated in the tumor microenvironment. Antibodies that block the interaction between these immune checkpoint receptors and their ligands have demonstrated significant clinical efficacy in recent trials. The first of this class of drugs to be approved by the Food and Drug Administration (FDA) in 2011 was ipilimumab, an antibody to the cytotoxic T-lymphocyte–associated antigen 4 (CTLA-4), designed for use in unresectable or metastatic melanoma. This came following the landmark phase III clinical trial, demonstrating a survival benefit with ipilimumab in this disease. As a single agent, however, a role for anti–CTLA-4 antibodies has not yet been reported in HNSCC.

More recently, another checkpoint inhibitory receptor, the programmed cell death protein 1 (PD-1), has garnered significant interest as a therapeutic target. Expression of PD-1 is induced on activated T cells and, on binding its ligands, PD-L1 (also known as B7-H1 and CD274) or PD-L2 (also known as B7-DC and CD273), inhibits T-cell receptor–induced signaling. Although PD-L2 is primarily expressed on activated macrophages and dendritic cells, PD-L1 expression is induced on both hematopoietic and nonhematopoietic cells, including epithelial cells, both benign and malignant, of head and neck mucosa. Inflammatory cytokines, particularly interferon-γ (IFNγ), upregulate PD-L1 expression on these cells, further promoting an immunosuppressive environment.

Phase I studies of blocking antibodies to either PD-1 or PD-L1 showed surprising responses in solid tumor malignancies, including melanoma, renal cell carcinoma, and traditionally nonimmunogenic cancers, such as non–small cell lung cancer (NSCLC). In 2014, accelerated approval of pembrolizumab, an antibody that blocks PD-1, was granted by the FDA for advanced melanoma. This followed significant responses seen in an open-label, multicenter expansion cohort of a phase I trial (KEYNOTE-001), in which 173 patients with advanced melanoma, who were refractory to ipilimumab, were randomized to either 2 mg/kg or 10 mg/kg every 3 weeks. At both dosing regimens, there was an objective response rate of 26%, and the toxicity profile was similar between the 2 arms, with the most common adverse reactions being fatigue, pruritus, and rash. More recently, a phase III study of patients with previously untreated BRAF wild-type advanced melanoma compared nivolumab (another antibody to PD-1) with dacarbazine. The patients receiving nivolumab had significantly greater overall (72.9%) and median progression-free (5.1 months) survival compared with those receiving dacarbazine (42.1% and 2.2 months).

As part of the phase I KEYNOTE-001 trial previously mentioned, an expansion cohort of patients with NSCLC receiving pembrolizumab also was analyzed. An objective response rate of 19.4% and a median duration of response of 12.5 months were observed. Of note, in this study, the proportion of tumor cells expressing PD-L1 correlated with improved efficacy of pembrolizumab. In a randomized phase III trial of advanced NSCLC, in which 272 patients were randomized to nivolumab or docetaxel, the response rate, overall survival, and progression-free survival were significantly greater in the nivolumab cohort. The patients included in this study had recurrent disease after 1 previous platinum-containing regimen. Unlike the phase I trial, PD-L1 expression did not correlate with response.

The histologic similarities between NSCLC and HNSCC and the recent success observed in NSCLC and other solid malignancies have generated much hope for similar success of PD-1:PD-L1 blockade in the treatment of the HNSCC. Of an interesting historical note, the initial preclinical studies demonstrating the blockade of this immune checkpoint axis in the treatment of cancer was actually shown in preclinical models of HNSCC.

Clinical studies of HNSCC have reported that PD-L1 is expressed in 68% to 77% of HNSCC tumors, although different thresholds and antibodies have been used. At the American Society of Clinical Oncology (ASCO) annual meeting in 2015, Seiwert and colleagues presented on an HNSCC expansion cohort of KEYNOTE-012 ( NCT01848834 ), in which pembrolizumab was given to patients with recurrent/metastatic HNSCC. In the study, 132 patients were enrolled; 99 were available for preliminary analysis. The objective response rate per RECIST 1.1 criteria was 18.2% with 18 partial responses and 31.3% with stable disease. Drug-related adverse events of any grade occurred in 47% of patients; the most common were fatigue, decreased appetite, pyrexia, and rash. Adverse events of grade ≥3 occurred in only 7.6%. Another study reported by Segal and colleagues at the ASCO annual meeting in 2015 was of an ongoing phase I/II, multicenter, open-label study that is evaluating the safety and efficacy of MEDI4736, a human immunoglobulin (Ig)G1 monoclonal antibody against PD-L1 ( NCT01693562 ). In this study, 62 patients were enrolled, and 51 with 24 months or more of follow-up were analyzed. The overall objective response rate was 12% and was 25% for patients with tumors that stained positive for PD-L1. The overall disease control rate was 16% at 24 months and 25% for PD-L1+ tumors. The median duration of response had not been reached at the time of the report, indicating meaningful durability of the responses. These promising results have led to an ongoing international open-label phase III trial (KEYNOTE-040, NCT02252042 ) in which 466 patients with recurrent/metastatic HNSCC, whose tumors had progressed following a platinum-based regimen, will be randomized to receive either pembrolizumab or the investigator’s choice of standard-of-care single systemic agent (methotrexate, docetaxel, or cetuximab), and a similar trial of nivolumab versus standard chemotherapy is under way as well ( NCT02105636 ).

While the effect of PD-1:PD-L1 blockade and the durability of response in the approximately 20% of patients who benefit from this treatment is remarkable, a clearer understanding of the precise mechanism is needed to better select patients to treat and improve response rates. There is a growing sense that greater PD-1 and PD-L1 expression by the tumor-infiltrating immune cells correlates with a greater response to PD-1:PD-L1 blockade. This may reflect a state of activation of the tumor-infiltrating immune effector cells that can be released from inhibition by blocking the PD-1 checkpoint. A novel window-of-opportunity study ( NCT02296684 ), in which neoadjuvant pembrolizumab is given approximately 2 to 3 weeks before standard-of-care surgery for locally advanced, surgically resectable HNSCC, is currently being conducted. Preinfusion and postinfusion biopsies of the tumors will hopefully provide important insight into biologic parameters that can predict responses to PD-1 blockade.

Enhancement of response to PD-1:PD-L1 blockade may potentially occur with tumor antigen vaccines (discussed later), in combination with radiation therapy, and/or in combination with other checkpoint inhibitors. The idea that combination of immune checkpoint inhibitors with radiation may be synergistic comes from observations of an abscopal effect against lesions outside of the radiation field when a patient with melanoma was treated with ipilimumab. By modeling of this abscopal response in mice, it was observed that the upregulation of PD-L1 following treatment is a dominant component of immune evasion and that dual blockade of PD-L1 and CTLA-4 significantly augments the therapeutic effect of radiation. This effect was related to an increase in the CD8 ⁺ /regulatory T-cell (Treg) ratio promoted by CTLA-4 blockade and a reinvigoration of exhausted CD8 ⁺ T cells by PD-L1 blockade. Thus, combination checkpoint blockade, possibly with radiation (which presumably promotes release of tumor antigens) or vaccine strategies, may prove to be promising therapeutic strategies in the future.

Finally, activation of the innate immune system may also enhance the response to the PD-1:PD-L1 blockade. As an example, activation of the stimulator of interferon genes (STING) pathway appears to augment the response to anti-PD-1 antibodies. STING is a cytosolic receptor that senses exogenous and endogenous cyclic dinucleotides (CDN). In preclinical models of poorly immunogenic solid tumor malignancies, an engineered CDN designed to have greater in vivo stability and formulated with a GM-CSF secreting whole-cell vaccine (STINGVAX) is able to induce regression of solid tumors alone and synergize with anti-PD-1 antibodies in tumors resistant to PD-1 blockade. It will be interesting to see if these observations are translated to the clinic and if other combinations, which include activation of the innate immune system (such as antibody-induced natural killer [NK] cell activation), can synergize with checkpoint blockade.

Augmenting cetuximab-based immunotherapy

Cetuximab is a human-mouse chimeric IgG1 antibody directed against an epidermal growth factor receptor (EGFR); it has been shown to be a useful adjunct to standard therapeutic regimens in patients with recurrent or metastatic HNSCC. Although direct inhibition of EGFR signaling may contribute to the antitumor effect of cetuximab, appreciable evidence indicates that a prominent component of any therapeutic effect can be attributed to the extracellular consequences of antibody binding to EGFR. For instance, the constant region (Fc) of cetuximab can bind to Fc receptors on a variety of immune cells, including the low-affinity, activating Fc receptor expressed on NK cells (CD16/FcγRIII). This interaction induces antibody-dependent cell-mediated cytotoxicity (ADCC), and the degree to which ADCC occurs has been correlated with clinical responses. Thus, cetuximab is actually the first “immunotherapeutic” agent approved for use in HNSCC.

Augmentation of ADCC induced by cetuximab is possible and is a focus of recent clinical investigation. One way in which cetuximab-induced ADCC can be augmented is by the activation of CD137 (also known as 4-1BB), which is an inducible, costimulatory molecule that is upregulated on activated NK cells and T cells following exposure to certain cytokines and Fc fragments. CD137 expression is induced on NK cells following exposure to cetuximab and EGFR-expressing targets. The addition of an agonist CD137 antibody can augment NK cell–mediated, cetuximab-dependent cytotoxicity, resulting in apoptosis of cultured cells, inhibition of tumor growth in vivo, and prolonged survival of tumor-bearing mice. This effect was largely dependent on NK cells, although depletion of CD8 ⁺ T cells partially abrogated the observed response. Several additional findings further indicated that treatment with cetuximab followed by CD137 stimulation was capable of concurrently stimulating an adaptive immune response. Specifically, mice previously cured with cetuximab/anti-CD137 agonist therapy rejected further tumor engraftment and additionally rejected tumors with coincident epitopes, indicating both immunologic memory and epitope spreading. Previous work had demonstrated that CD8 ⁺ T cells significantly contribute to the antitumor effect of cetuximab, and this reaction can be elicited through NK cell–mediated dendritic cell maturation and their subsequent interface with CD8 ⁺ T cells. As a result of these findings, a phase Ib open-label, multicenter trial ( NCT02110082 ) of combination cetuximab and anti-CD137 antibody (urelumab) in patients with advanced/metastatic colorectal cancer or advanced/metastatic head and neck cancer is currently under way.

Similarly, Toll-like receptor (TLR)-8 stimulation of peripheral blood mononuclear cells has been shown to enhance cetuximab-induced NK cell–mediated ADCC of HNSCC cells and lead to dendritic cell maturation and priming of EGFR-specific CD8 ⁺ T cells. These findings have led to a phase I dose-escalation study ( NCT01334177 ) of a TLR8 agonist (VTX-2337) given in conjunction with cetuximab in patients with locally advanced, recurrent, or metastatic HNSCC.

Tregs (CD4 ⁺ FOXP3 ⁺ ) can constitute a significant proportion of the TIL population following cetuximab therapy. The presence of Tregs correlates with a functional impairment of NK cells. Depletion of these Tregs with anti-CTLA-4 antibody can partially correct this immunosuppressive environment and restore NK-mediated cytotoxicity. This provides rationale for the combination of cetuximab and ipilimumab in the clinic. A phase I trial of cetuximab, ipilimumab, and intensity-modulated radiation therapy is currently under way.

Vaccines

There is abundant evidence that the adaptive immune system actively participates in the elimination of susceptible cells (a process termed “editing”) during tumor formation, and that the more resistant cells may achieve a state of equilibrium with the host immune system. One principal goal of immunotherapy is to bias an immune response toward elimination of these more resistant cells. This can be achieved, in part, through augmenting the recognition of cancer cells through therapeutic vaccination, with the intent to induce an antigen-specific, cytotoxic CD8 ⁺ T-cell response. Practically, this involves promoting the uptake and presentation of antigen by antigen-presenting cells, the most efficient of which are dendritic cells. Following antigen capture and processing, maturation, and recruitment of CD4 ⁺ T cells, dendritic cells prime a CD8 ⁺ T-cell response, which is then integrally involved in the antitumor immune response. There are several different approaches to using dendritic cells in tumor vaccination, ranging from direct delivery of antigen to loading of autologous dendritic cells with tumor antigen ex vivo, followed by maturation and reintroduction. These strategies have been used in various ways to target HNSCC and, in most studies, are predicated on the availability of a specific tumor-associated antigen.

A potential source of antigens in a substantial proportion of HNSCC is the human papilloma virus (HPV) virus. HPV is a double-stranded DNA virus that can infect the epithelium of the upper aerodigestive tract and significantly contribute to malignant transformation. Typical infection can elicit a T-cell–mediated immune response and subsequent clearance, and much work has centered on understanding and adapting this response to the treatment of HPV ⁺ HNSCC. Previous studies have confirmed that patients with HPV ⁺ HNSCC have a significantly higher number of E7-specific cytotoxic CD8 ⁺ T cells than patients with HPV ⁺ HNSCC or healthy controls, demonstrating a preexisting level of endogenous immunity. Interestingly, this study also demonstrated that components of the antigen-processing machinery were relatively lower in HPV ⁺ HNSCC compared with surrounding squamous epithelium, suggesting a degree of immune escape in the development of these malignancies and underscoring the complexity of the immune response to developing tumors. There have subsequently been varied efforts to augment the immune response to HPV-associated HNSCC. One study of peptide/protein vaccines used a “Trojan peptide” approach, in which HPV-16 and MAGE-A3 tumor-associated antigens were engineered to contain sequences intended to promote modified processing and subsequent generation of HLA-I and HLA-II responses. Patients with advanced recurrent or metastatic HNSCC were injected with either the HPV-16 or MAGE-A3 Trojan peptide vaccine at 4-week intervals for up to 4 total doses. Although no complete or partial responses were documented, apart from 1 serious adverse event associated toxicities were generally well tolerated. Interestingly, although there was an appreciable response to the HLA-II restricted epitopes in most patients, there was no detectable response to the HLA-I restricted epitopes. In 1 of the 2 patients treated with the HPV-16 vaccine who underwent a neck dissection 3 months after vaccination, there was marked a CD4 ⁺ and CD8 ⁺ infiltrate that was notably enriched for HPV-specific T cells. This study illustrates that peptide/protein vaccines targeting HPV-16 can induce an immune reaction to HNSCCs, although the response can be relatively complex and may bias toward HLA-II–restricted epitopes and is not yet of clinical utility in its present form.

In addition to peptide-based vaccines, there have been several preclinical studies focused on the development of DNA-based vaccine strategies to induce tumor-specific immunity in HPV ⁺ HNSCC. For instance, treatment of tumor-bearing mice with a construct containing the HPV-16 E7 gene modified to disrupt Rb-binding and transforming capability was shown to result in a significant inhibition of tumor growth and an induction of antigen-specific CD8 ⁺ T cells. This effect was further augmented by cyclophosphamide administration, which correlated with a reduction in regulatory T cells. This finding suggests an inherent constraint on vaccine effect and again highlights the relatively complex nature of the immune response to cancer vaccines.

Another potential source of antigen in HNSCC is p53, a central component of the cellular response to a diverse array of stressors and an essential tumor suppressor. Mutations in p53 can be detected in just more than half of patients with HNSCC undergoing definitive surgical resection, and the presence of disruptive p53 mutations is associated with significantly decreased overall survival. Subsequent studies have confirmed the frequent occurrence of p53 mutations in HNSCC, as well as an inverse relationship between p53 mutations and HPV infection. There has thus been much interest in developing strategies to elicit an immune reaction toward p53. The frequency of p53 antigen-specific CD8 ⁺ T cells is higher in the peripheral blood of patients with HNSCC than healthy controls, though this relative enrichment interestingly does not correlate with overexpression of p53 in the tumor itself. Subsequent studies have demonstrated that p53-specific T cells are not only found in the circulation, but are appreciably enriched in primary tumors or tumor-involved lymph nodes. Again, this accumulation of TILs was not necessarily related to p53 overexpression, as it was seen in several tumors with no notable p53 staining. This population of TILs is, however, functionally deficient, as demonstrated by a relative inability to secrete interferon-gamma after stimulation.

Building on this previous work, a dendritic cell–based vaccine strategy was developed and used in a recent phase 1 trial. The study accrued 16 patients with advanced HNSCC treated with either surgery alone (10 patients) or surgery and adjuvant chemotherapy (6 patients). These patients were then randomized to receive 1 of 3 different vaccines by ultrasound-guided inguinal lymph node injection: autologous dendritic cells loaded with 2 optimized p53 peptides, dendritic cells loaded with optimized p53 peptides and tetanus helper peptide, or dendritic cells loaded with optimized p53 peptides and HLA class II p53 helper peptide. Apart from a local injection site reaction in 2 patients, the vaccines were well tolerated and disease-free survival compared favorably with a historical cohort. After vaccination, 69% of patients showed some peptide-specific response, with 38% developing cytotoxic T lymphocytes (CTLs) reactive to either optimized peptide and another 31% developing CTLs reactive to either peptide. One additional, notable finding was the effect of vaccination on Treg populations. The average frequency and absolute number of Tregs was significantly reduced after vaccination, and this outcome was postulated to be a component of any potential benefit.

In addition to vaccination centered on a clearly defined, specific antigen, there has been interest in a broader approach using dendritic cells loaded with antigens derived from autologous apoptotic tumor cells. One study enrolled patients with operable stage III or IV HNSCC. Tumor cells harvested during surgical resection were purified and stored. Approximately 2 months following conventional therapy, monocytes were harvested from these patients and coincubated with the previously harvested autologous tumor cells induced to undergo apoptosis by UV treatment. Dendritic cells exposed to apoptotic tumor cells were then induced to undergo maturation and administered to disease-free patients by ultrasound-guided intranodal injection in 2 separate doses 6 weeks apart. Because of strict inclusion criteria and appreciable issues with the number and sterility of autologous tumor cells, 4 patients could be given the planned 2 doses of the dendritic cell–based vaccine. Apoptotic tumor cell–reactive T cells were detected after vaccination in 3 patients, and no adverse events occurred, illustrating that, in principle, this strategy was immunogenic.