Abstract
BACKGROUND: Immune landscape of cancer has been increasingly recognized as a key feature affecting disease progression, prognosis and therapeutic response. Here, we sought to comprehensively characterize the patterns of tumor-infiltrating immune cells (TIIs) in primary oral squamous cell carcinoma (OSCC) and develop immune features-derived models for prognostication and therapeutic prediction. METHODS: A total number of 392 patients with OSCC receiving ablative surgery at three independent centers were retrospectively enrolled and defined as training, testing and validation cohorts. Detailed features of 12 types of TIIs at center of tumor and invasive margin were assessed by immunohistochemistry coupled with digital quantification. TIIs abundance in OSCC was also estimated by bioinformatics approaches using multiple publicly available data sets. Prognostic models based on selected immune features were trained via machine learning approach, validated in independent cohorts and evaluated by time-dependent area under the curves and concordance index (C-index). Immune types of OSCC were further identified by consensus clustering and their associations with genetic, molecular features and patient survival were clarified. RESULTS: Patterns of TIIs infiltration varied among patients and dynamically evolved along with tumor progression. Prognostic models based on selected TIIs were identified as efficient and sensitive biomarkers to stratify patients into subgroups with favorable or inferior survival as well as responders or non-responders to postoperative radiotherapy or immunotherapy. These models outperformed multiple conventional biomarkers and immune-related scores in prognostic prediction. Furthermore, we identified two main immune subtypes of OSCC (immune-hot and immune-cold) which harbored characteristic TIIs infiltrations and genomic and molecular features, and associated with patient survival. CONCLUSIONS: Our results delineated immune landscape and subtypes in OSCC, consolidated their clinical values as robust biomarkers to predict patient survival and therapeutic benefits and reinforced key roles of TIIs and tumor-immune interactions underlying oral tumorigenesis, ultimately facilitating development of tailed immunotherapeutic strategies.