Mutational Landscape and Predictors of Survival in Head and Neck Mucosal Melanoma

Introduction

Head and neck mucosal melanoma (HNMM) is an aggressive malignancy most commonly arising from the nasal cavity, paranasal sinuses, and oral cavity with a 5-year overall survival (OS) rate <25%.¹ HNMM is clinically distinct from cutaneous melanomas in terms of etiology, tumor genetics, clinical outcomes, and treatment paradigms.² For cutaneous melanoma, specific driver mutations (e.g., BRAF V600E mutations) predict survival and influence response to targeted and immunotherapies.³ However, HNMM infrequently harbor BRAF mutations (0–8%), yet NRAS is frequently mutated in up to 20% of cases, for which no targeted therapy exists.⁴ Additionally, HNMM have lower tumor mutational burden (TMB) compared to cutaneous melanomas, likely influencing their heterogenous response to immunotherapy.⁵ In this study, we utilized a publicly accessible clinic-genomic dataset to investigate the mutational landscape of HNMM and the influence of top driver mutations on OS.

Methods

This study was exempt from institutional review board approval due to the publicly accessible and deidentified nature of this database. The AACR Project Genomics Evidence Neoplasia Information Exchange (GENIE)^® database was accessed through cBioPortal (v13.1-public), which aggregates data from 19 international cancer centers.⁶ The database is distinct from The Cancer Genome Atlas (TCGA), with few samples having overlap between databases. Therapeutic response data and disease burden is limited to specific cancer subtypes. Institutions may have different pipelines and targeted panels of up to 555 genes (or unbiased whole genomic sequencing) for sequencing analysis. For this study, the database was queried for all patients with HNMM. The dataset included clinical demographics (e.g., age and race), genomic information (e.g., somatic mutations, copy number alterations, structural variants), and clinical outcomes (i.e., OS). Tumor mutational burden was estimated based on number of somatic mutations detected. Statistical analyses were performed using R/R Studio (R Foundation for Statistical Computing, Boston, MA). P<0.05 was considered statistically significant. Kaplan-Meier log-rank test evaluated differences in OS based on mutational status and TMB. Student’s t-tests were used to compare differences in means between groups for continuous variables.

Results

6,794 patients had melanoma, of which 108 (1.6%) had HNMM. The average age of the cohort was 65.8 ± 14.8 years and 86 (79.6%) were White race. The average TMB was 6.1 ± 5.2. Of the 108 patients, only 106 had OS data available. The 1-, 2,-, and 5-year OS rates for the cohort were 84.2% (95% CI 77.4–91.6%), 69.6% (95% CI 60.9–79.6%), and 47.5% (95% CI 37.6–60.0%) (Figure 1A). Patients with high TMB had improved OS compared to patients with low TMB (p=0.026) (Figure 1B). The top-10 mutations in the cohort were NRAS (n=19, 17.6%), NF1 (n=15, 13.9%), ROS1 (n=14, 13.0%), TERT (n=11, 10.2%), BRAF (n=11, 10.2%), TP53 (n=9, 8.3%), KIT (n=8, 7.4%), KMT2C (n=7, 6.5%), KMT2A (n=7, 6.5%), and EP300 (n=4, 3.7%). Of the 19 patients with NRAS mutations, 7 (47.4%) had Q61R, 5 (26.3%) had Q61K, 3 (15.8%) had G12A, and 2 (10.5%) had G12C mutations. Of the 11 patients with BRAF mutations, 4 (36.3%) had V600E mutation. Additionally, we evaluated OS differences in patients with NRAS (Figure 2A), NF1 (Figure 2B), ROS1 (Figure 2C), KIT (Figure 2D), TERT (Figure 2E), or BRAF (Figure 2F) mutations. We only observed a significant difference in OS for patients with ROS1 mutation, where those with this driver mutation had improved OS (p=0.023) (Figure 2C). ROS1-mutated HNMM patients had a significantly higher TMB of 10.8 ± 7.0 compared to 5.3 ± 4.0 for ROS1-wild-type HNMM patients (p=0.012).

Figure 1:

A) Overall survival across the entire cohort. B) Overall survival comparing high vs. low tumor mutational burden.

Figure 2:

Overall survival differences for patients harboring A) NRAS, B) NF1, C) ROS1, D) KIT, E) TERT, or F) BRAF mutations.

Discussion

In this study, we demonstrated that HNMM patients are older, have a low TMB with poor OS, and harbor one or more of NRAS, NF1, ROS1, TERT, BRAF, or KIT mutations. Additionally, no driver mutations occurred OS benefit, as previously demonstrated in other studies, except we identified that patients with mutations in ROS1 had improved OS, likely due to their higher TMB. Prior whole-genomic sequencing studies for mucosal melanoma have identified genomic alterations which have informed targeted therapy selection, such as use of palbociclib for tumors with CDK4 amplification.³ Thus, the AACR Project GENIE^® database may provide yet another platform for biomarker discovery leading to precision medicine treatment selection in HNMM.

ROS1 is a proto-oncogene receptor tyrosine kinase (RTK) whose activation signals the downstream RAS/RAF/MAPK pathway. It is frequently implicated in lung, gastric, ovarian, colorectal, and biliary cancers. Small molecule RTK inhibitors are common systemic agents for solid tumors. Recently, entrectinib and crizotinib, both RTK inhibitors, have been approved for treating ROS1-positive solid tumors.⁷ In cutaneous melanoma, de novo mutations in ROS1 has been shown to result in improved OS for patients treated with immunotherapy, along with having a higher TMB.⁸ We also observed in our cohort ROS1-mutated cases to have a higher TMB, likely contributing to their improved OS. Furthermore, one case report describes complete response to crizotinib therapy in a HNMM patient harboring a ROS1 mutation.⁹

The treatment paradigm for HNMM is wide surgical resection for local control with adjuvant radiotherapy and/or immunotherapy.⁵ The role of immunotherapy remains unclear, with poor understanding of factors conferring additional OS benefit, but studies have elucidated that there may be benefit in select patients post-surgery or those with recurrent or metastatic disease.¹⁰ Beyond clinicopathologic information, such as TNM staging, profiling the whole-genomic landscape in HNMM patients prior to immunotherapy use may provide insights into biomarkers to stratify patients which may benefit, as performed for cutaneous melanoma patients.

These results should be interpreted considering the limitations. First, we lacked information from the database on clinical management and outcomes (e.g., systemic/immunotherapy and targeted therapy use), limiting the ability to interpret therapy response based on mutational status. Second, this database does not provide associated transcriptomic signatures to correlate mutation with signaling pathway activation. Third, there may be an over- or underestimation of mutation frequency across these patients as different centers use different sequencing platforms and either targeted or unbiased sequencing approaches. Lastly, the associated histology was not available to correlate immunohistochemical expression with mutational status. However, despite these limitations, this study provides rationale for screening ROS1 as a potential HNMM biomarker in the era of precision medicine.

Key Points.

• Head and neck mucosal melanomas have a diverse mutational landscape with low mutational burden

• A molecular subset (~13%) have ROS1 mutations, which is an actionable driver mutation

• ROS1 mutated patients have improved overall survival likely due to high mutational burden