Integrated multi-omic analysis characterizes an IFNγ-responsive-to-myeloid-suppressive immune continuum in immune-active hepatocellular carcinoma

Immune-active hepatocellular carcinoma (HCC) is commonly treated as a single inflamed phenotype, although its internal immune heterogeneity remains incompletely defined. We integrated public bulk transcriptomic, single-cell, spatial transcriptomic, proteomic, phosphoproteomic, and treatment-cohort data to test whether this heterogeneity is better represented by discrete states or a continuous immune axis. In 112 checkpoint-high TCGA-LIHC tumors, an inflamed-minus-myeloid axis showed structured heterogeneity but no robust empirical bimodality: the observed kernel-density curve had one prominent mode and none of 1,000 bootstrap resamples contained two or more prominent modes. IFNγ-responsive-associated and myeloid-suppressive-associated labels were therefore retained only as operational summaries of opposing axis tendencies, and the biological polarity persisted across alternative definitions of immune-active entry. A fixed 10-gene within-sample rank template-contrast score transferred the axis to independent HCC cohorts. Transcriptomic analyses based on gene sets non-overlapping with the compact panel and original state-defining genes supported the IFNγ-oriented end, whereas spatial analyses showed local cross-end coexistence with reduced immediate-neighbor mixing relative to a count-preserving random-label null, and paired proteomic/phosphoproteomic analyses showed concordant myeloid-oriented associations involving macrophage, fibroblast/extracellular-matrix, hypoxia, SPP1-integrin-FAK, and myeloid-receptor programs. Direct application of the frozen transfer score to four public HCC immunotherapy cohorts yielded heterogeneous, imprecise response associations (pooled Hedges’ g = 0.10, 95% CI −0.42 to 0.61) and did not establish clinical response prediction. These results support a structured IFNγ-responsive-to-myeloid-suppressive continuum within immune-active HCC. The framework is transferable across public datasets but requires prospective HCC-specific clinical validation, while functional perturbation studies are needed before causal or therapeutic claims.