Efficacy of immunotherapy in hepatocellular carcinoma (HCC) is blocked by its high degree of interâ and intraâtumor heterogeneity and immunosuppressive tumor microenvironment. However, the correlation between tumor heterogeneity and immunosuppressive microenvironment in HCC has not been well addressed. Here, we endeavored to dissect interâ and intraâtumor heterogeneity in HCC and uncover how they contribute to the immunosuppressive microenvironment. We performed consensus molecular subtyping with nonânegative matrix factorization (NMF) clustering to stratify the interâheterogeneity profile of HCC tumors. We grouped HCC tumors from the Cancer Genome Atlas (TCGA) patients into three subtypes (S1, S2 and S3), where S1 was characterized as a âhot tumorâ profile with high expression level of T cell genes and rate of immune scores. S2 was characterized as a âcold tumorâ profile with the highest tumor purity score, and S3 as an âimmunosuppressed tumorâ profile with the poorest prognosis and a high expression level of immunosuppressive genes such as cytotoxic Tâlymphocyteâassociated proteinâ4,
TIGIT,
and
PDCD1
. Moreover, we combined weighted gene coâexpression network analysis and singleâcell regulatory network inference and clustering (SCENIC) in the singleâcell dataset of the S3âlike subtype (CS3) and identified a transcription factor,
BATF
, which could upregulate immunosuppressive genes. Finally, we identified a cell interaction network in which a myeloidâderived suppressor cellâlike macrophage subtype could promote the formation of immunosuppressive Tâcells.