Abstract:
Objective: To develop and validate an immunophenotyping system based on tumor budding (TB) and tertiary lymphoid structures (TLS) for head and neck squamous cell carcinoma (HNSCC). Methods: A total of 1405 patients undergoing radical resection from January 2016 to December 2022 were enrolled as the training cohort, and 502 consecutive patients enrolled from March 2023 to December 2024 were enrolled as the prospective time validation cohort. HE staining was used to evaluate TB, and CD20/CD3, CD21 and Ki-67 were used to evaluate TLS and its maturity and spatial location. According to the level of TB and TLS activity, the patients were divided into immune activation type, immune balance type, immune rejection type and immune depletion type. Kaplan-Meier method, Cox regression, interaction analysis, and time-dependent ROC curve were used to evaluate the prognostic value. Exploratory spatial immunoassay was performed in a subset of 100 mIHC patients in the training cohort. Results: In the training cohort, the 5-year overall survival rates were 84.6%, 67.8%, 55.1% and 29.7%, respectively (P<0.001). Compared with the immune activation type, the immune balance type, immune rejection type, and immune depletion type gradually increased the adjusted risk of death, with HR of 1.82, 2.46, and 4.28, respectively. In the validation cohort, the 2-year OS rates of the four types were 94.0%, 84.9%, 76.2% and 57.8%, respectively (P<0.001), and the adjusted HR was 3.91 for the immunodepleted type. After adding TB-TLS, the C-index increased from 0.724 to 0.769 in the training cohort and from 0.711 to 0.751 in the validation cohort. mIHC analysis showed that there were gradient differences in the density of effector cells, the effector-depletion ratio and the distribution of immune cells around TB among the four types. Conclusions: The combined classification of TB and TLS may provide additional prognostic stratification information on the basis of traditional clinicopathological factors. However, its clinical application and treatment prediction value still need to be further verified by multi-center independent cohorts.