BACKGROUND & AIMS: The tumour microenvironment (TME) is a crucial mediator of cancer progression and therapeutic outcome. The TME subtype correlates with patient response to immunotherapy in multiple cancers. Most previous studies have focused on the role of different cellular components in the TME associated with immunotherapy efficacy. However, the specific structure of the TME and its role in immunotherapy efficacy remain largely unknown. METHODS: We combined spatial transcriptomics with single-cell RNA-sequencing and multiplexed immunofluorescence to identify the specific spatial structures in the TME that determine the efficacy of immunotherapy in patients with hepatocellular carcinoma (HCC) receiving anti-PD-1 treatment. RESULTS: We identified a tumour immune barrier (TIB) structure, a spatial niche composed of SPP1 macrophages and cancer-associated fibroblasts (CAFs) located near the tumour boundary, which is associated with the efficacy of immune checkpoint blockade. Furthermore, we dissected ligand‒receptor networks among malignant cells, SPP1 macrophages, and CAFs; that is, the hypoxic microenvironment promotes SPP1 expression, and SPP1 macrophages interact with CAFs to stimulate extracellular matrix remodelling and promote TIB structure formation, thereby limiting immune infiltration in the tumour core. Preclinically, the blockade of SPP1 or macrophage-specific deletion of Spp1 in mice led to enhanced efficacy of anti-PD-1 treatment in mouse liver cancer, accompanied by reduced CAF infiltration and increased cytotoxic T-cell infiltration. CONCLUSIONS: We identified that the TIB structure formed by the interaction of SPP1 macrophages and CAFs is related to immunotherapy efficacy. Therefore, disruption of the TIB structure by blocking SPP1 may be considered a relevant therapeutic approach to enhance the therapeutic effect of immune checkpoint blockade in HCC. IMPACT AND IMPLICATIONS: Only a limited number of patients with hepatocellular carcinoma (HCC) benefit from tumour immunotherapy, which significantly hinders its application. Herein, we used multiomics to identify the spatial structure of the tumour immune barrier (TIB), which is formed by the interaction of SPP1+ macrophages and cancer-associated fibroblasts in the HCC microenvironment. This structure constrains immunotherapy efficacy by limiting immune cell infiltration into malignant regions. Preclinically, we revealed that blocking SPP1 or macrophage-specific deletion of Spp1 in mice could destroy the TIB structure and sensitize HCC cells to immunotherapy. These results provide the first key steps towards finding more effective therapies for HCC and have implications for physicians, scientists, and drug developers in the field of HCC.