Tumor immune microenvironment states inferred from TLS-associated immune-cell composition stratify prognosis in hepatocellular carcinoma

Background Tertiary lymphoid structures (TLS) are ectopic immune hubs in the tumor immune microenvironment (TIME) associated with prognosis and immunotherapy response, yet commonly used TLS structural descriptors show inconsistent associations with prognosis in hepatocellular carcinoma (HCC). Here, we propose a TLS-associated immune-cell composition framework that quantifies the TIME state and predicts patient outcomes. Methods By integrating an HCC single-cell atlas with literature-curated TLS gene sets, we defined six TLS-associated immune components (TLS6). Using TLS6 as a reference, we applied BayesPrism deconvolution to infer the relative abundance of TLS6 from bulk tumor transcriptomes. Given the divergent prognostic associations across TLS6 fractions, we applied LASSO–Cox regression to derive a two-feature TLS RiskScore retaining regulatory T cells and cDC2 cells. Results The TLS RiskScore stratified overall survival in TCGA-LIHC and ICGC LIRI-JP and was associated with response to PD-1 blockade in an independent anti–PD-1–treated HCC cohort. In multicenter FFPE tissues, a multiplex immunofluorescence (mIF) implementation quantifying TLS-localized CD4 + FOXP3 + regulatory T cells and CD11c + CD1c + cDC2 cells reproduced prognostic stratification without model refitting. Conclusions Collectively, these results support a compact, translatable TLS-associated immune-cell composition framework that provides a computable TIME-state measure associated with prognosis and response to PD-1 blockade in HCC.