Related Experiment Video
Updated: Jun 20, 2026
Predictive Immune Modeling of Solid Tumors
Published on: February 25, 2020
Spatial transcriptomics of immune ecotypes for predicting immunotherapy outcomes in head and neck squamous cell
Yunchao Xin1, Lihang Yang1, Chuan Liu1
1The First Affiliated Hospital of Hebei North University, Zhangjiakou, Hebei 075000, China.
Background:
Head and neck squamous cell carcinoma (HNSCC) exhibits heterogeneous tumour-immune microenvironments that limit the utility of single biomarkers such as programmed death-ligand 1 (PD-L1) and tumour mutational burden (TMB) for guiding immune checkpoint inhibitor (ICI) therapy. This study developed and validated the Ecotype-Integrated Response Model for HNSCC (EIRM-HN), integrating single-cell states, spatial transcriptomic niches, and bulk transcriptomes to derive immune ecotypes that stratify ICI outcomes.
Methods:
This retrospective multi-cohort study analysed 370 HNSCC cases (80 molecular, 210 immunotherapies, 80 control) profiled by single-cell RNA sequencing, spatial transcriptomics, and bulk RNA sequencing. Immune ecotypes were derived from integrated single-cell and spatial features, converted into weighted gene signatures, and projected into bulk ICI cohorts to train penalised Cox and logistic models and compare performance against PD-L1, tumour mutational burden, and published signatures, with external prognostic validation in the GSE65858 bulk cohort.
Results:
Among 210 ICI-treated patients, four ecotypes occurred at similar frequencies. The most suppressive ecotype showed low CD8+ T-cell abundance, high regulatory T-cell abundance, and increased stromal fraction, with median progression-free survival of 3.8 months and overall survival of 9.1 months, versus 9.8 and 20.3 months in the lymphoid-enriched ecotype. EIRM-HN achieved progression-free and overall survival concordance indices of 0.71 and 0.70, improving to 0.75 and 0.74 after adding clinical covariates, and exceeding PD-L1 and TMB. In GSE65858, overall survival concordance index was 0.67 with a hazard ratio of 1.89 for high- versus low-risk strata.
Conclusion:
An ecotype-based, spatially anchored risk model integrating single-cell, spatial, and bulk transcriptomic data provides improved prognostic stratification of HNSCC relative to established biomarkers and generalises to an external bulk cohort.
