|
As evaluated by multiplexed immunostaining on tumor microarrays (TMA; including matched tumor subregions, tumor margin, and nontumor liver; Supplementary Fig. S1B), the three immune groups displayed comparable densities of CD3+ T cells, CD8+ T cells, and CD68+ macrophages in tumor margins and nontumor livers but significantly differed in tumor subregions (Fig. 2D). T-cell exclusion was sequentially aggravated from highly infiltrated to sparsely infiltrated tumors (Fig. 2E). In contrast, macrophages were more abundant in tumor regions than in surrounding tissues (Supplementary Fig. S5C), suggesting that T cells were the targets of exclusion. Immunogenomic analysis showed that highly infiltrated tumors displayed comparable cancer testis antigen (CTA) load, chromosomal instability, and ploidy with heterogeneously and sparsely infiltrated tumors but had significantly higher TNB and the lowest tumor purity (Fig. 2F). Sparsely infiltrated tumors had comparable TMB and TNB with heterogeneously infiltrated tumors but exhibited universally fewer immune and stromal cells. Meanwhile, multiple chemokines, including CXCL9, CXCL10, and CXCL11, were inadequate in sparsely infiltrated tumors compared with highly infiltrated tumors, and heterogeneously infiltrated tumors displayed the distinct distribution of chemokines between tumor subregions with low and high immune infiltration (Fig. 2F; Supplementary Fig. S5D). We then compared transcriptomic features among the three immune groups using hallmarks from the Molecular Signatures Database (MSigDB; ref. 34). Immune-related pathways, angiogenesis, and classical cancer pathways were upregulated in highly infiltrated tumors, while sparsely infiltrated tumors upregulated cell-cycle signaling and several metabolic pathways such as oxidative phosphorylation and cholesterol homeostasis (Supplementary Fig. S5E). These results revealed distinct immune landscape, genomic, and transcriptomic characteristics among the three immune groups. We further calculated an immune ITH index for each patient and observed significant positive correlations between ITH of genomic features and immune contexts, indicating their potential interaction (Supplementary Fig. S5F).
|