Background: Syngeneic cell line based cancer model systems have advanced our understanding of host tumor interactions in developing cancers and in immunotherapeutic approaches. Our group initiated development of murine head and neck squamous cell carcinomas (HNSCC) model systems in 2007-2008 as no organ specific models had been described. The best described animal model systems for HNSCC utilized the hamster cheek pouch model where the carcinogen 7,12-Dimethylbenz(a)anthracene was the initiating and tumor promoting agent. We adapted this model to murine systems and generated primary oral squamous cell carcinomas that we used to make transplantable cell line models. These mouse oral carcinoma cell lines (MOC model) have a conserved carcinogen origin and retain pathologic, genomic and biologic features of the human disease and thus represent a robust preclinical surrogate. Here we describe considerations in development of immunocompetent models of HNSCC, validation approaches used for the MOC model and highlight recent work on checkpoint responses in the MOC model.

Methods: To define the genomic landscape and assess the fidelity of the MOC model to human disease, whole exome sequencing (WES) and expression arrays were performed. In order to evaluate checkpoint-based immunotherapeutic approaches in murine OSCC, independent cohorts of mice were treated with blocking monoclonal antibodies to PD-1. Tumor growth was monitored by twice-weekly measurement. Whole exome sequencing data were processed via MHC Class I binding algorithms including NetMHC to predict high-affinity neoantigen candidates. Mutant peptides were synthesized and used in Enzyme-Linked ImmunoSpot (ELISPOT) assays using tumor-infiltrating lymphocytes generated from mice bearing MOC1 and MOC22 tumors and ELISPOT data was validated using H-2Kb dual-color tetramer assays.

Results: Next-generation sequencing of the MOC cell lines demonstrated significant homology to human head and neck squamous cell carcinoma (HNSCC), including identification of driver pathway mutations in Trp53, mitogen-activated protein kinase, phosphoinositide 3-kinase, NOTCH, JAK/STAT, and Fat1-4. Furthermore, the most aggressive MOC cell line, MOC2, demonstrated an expression signature consistent with more aggressive human HNSCC tumors, including upregulation of pERK and CD44. This finding led to a biomarker window-of-opportunity clinical trial of a MEK inhibitor, trametinib, for patients with aggressive OSCC.

Next, looking at checkpoint inhibitors, we identified three distinct phenotypes in response to anti-PD1 in the three murine OSCC tumors. Anti-PD1 had no impact on the growth kinetics or frequency of metastases in MOC2. MOC1, a moderately immunogenic tumor, exhibited significant tumor regression followed by outgrowth in a subset of treated mice. When this “escape” tumor was harvested, expanded in vitro and retransplanted into mice, it was more aggressive than the parental and no longer exhibited any growth delay with anti-PD1. MOC22 demonstrated complete response, consistently rejecting 20-30 days following treatment.

We then applied a neoantigen discovery pipeline to MOC22 as these tumors demonstrated the most robust immune response with checkpoint blockade. Predicted high-affinity neoepitopes were screened using tumor-infiltrating lymphocytes assessing for interferon-gamma (IFNg) response with ELISPOT. TIL from MOC22 demonstrated reactivity to a mutated ICAM-1 (mICAM1) protein with a proline to leucine (P315L) substitution which induced a 145-fold increase in H-2Kb affinity (2.45 nM for mutant versus 357.21 nM for wild type). In addition to screening for mutation-derived neoepitopes, we also identified that MOC22 expressed an endogenous retroviral derived H2-Kb restricted antigen from the p15e protein, previously described in several C57BL/6 derived tumor models. Likewise, in ELISPOT assays, we found strong reactivity to the p15e antigen as well. To validate the ELISPOT data, we utilized dual-color H-2Kb restricted tetramers with our two candidate antigens, miCAM1 (TVYNFSAL) and p15e (KSPWFTTL) and identified a subset of T cells with specificity for each antigen.

Ongoing work is focused on defining the time course of TIL infiltration in these tumors and how to optimize use of these tumor antigens in a vaccine setting.

Conclusions: We have developed and characterized three syngeneic mouse models of OSCC that demonstrate conservation of mutations and transcriptional pathway activation with human HNSCC. Findings in this model have demonstrated translational relevancy leading to demonstration of MEK targeting as an option in aggressive OSCC. Furthermore, we have identified a genomically-derived tumor neoantigen and an endogenous retroviral antigen that are candidates for checkpoint inhibitor-induced tumor rejection in a murine OSCC model. These data represent the translational relevance of the MOC model and provide a preclinical foundation for development of personalized and targeted cancer immunotherapeutics in OSCC.

Citation Format: Ravindra Uppaluri. Immunocompetent models of head and neck cancer [abstract]. In: Proceedings of the AACR-AHNS Head and Neck Cancer Conference: Optimizing Survival and Quality of Life through Basic, Clinical, and Translational Research; April 23-25, 2017; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2017;23(23_Suppl):Abstract nr IA21.