Modeling colorectal liver metastasis to leverage type I interferon-induced innate immune cell anti-tumor response

Colorectal cancer liver metastasis (CRLM) is the leading cause of death in patients with colorectal cancer, and its immune landscape is profoundly shaped by tumor-intrinsic pathways. Among the key regulators of liver immunity, type 1 innate lymphoid cells (ILC1), natural killer (NK) cells, and metastasis-associated macrophages (MAMs) orchestrate tissue residency, effector activation, and immunoregulation. However, the features of the tumor microenvironment, particularly those associated with mismatch-repair-deficient (dMMR) and proficient (pMMR) tumors, shape their recruitment, while functional polarization of these populations remain poorly understood. This study investigates how tumors with distinct genetic features and type I interferon programs regulate the immune architecture of CRLM. By comparing two orthotopic metastasis models, MC38 (dMMR, immunogenic) and SL4 (pMMR, non-immunogenic), we dissect mechanisms controlling NK/ILC1 recruitment, differentiation, and interaction with MAMs, and assess whether STING pathway activation can convert poorly immunogenic lesions into immune-responsive metastases. We uncovered pronounced differences between the two CRC models; MC38 metastases displayed abundant, functionally competent type 1 ILC/NK populations, including strongly infiltrating CD49a⁺ tissue-resident NK cells with mature, cytotoxic profiles (CD69, CD11b, high effector activity). SL4 lesions instead showed limited NK recruitment, reduced maturation, and weaker effector responses. Myeloid profiling revealed that MC38 metastases were enriched in mature, CXCL9-producing MAMs with a type I IFN-driven program, whereas SL4 MAMs expressed ECM-remodeling. Co-culture of MAMs and NK cells showed a stronger NK activation and IFN- production when cultured with MC38-derived MAMs compared to SL4 MAMs. Both MC38 tumor cells and MAMs exhibited a more active baseline I IFN pathway. In vivo, diABZI treatment activated type I IFN programs in both models, promoting lymphocyte activation, phenotypic remodeling of lymphoid and myeloid cells, and a marked increase in PD-L1 expression, with especially strong induction on CD45- tumor cells in SL4. This observation prompted testing of whether STING activation could sensitize poorly immunogenic SL4 metastases to immune checkpoint blockade. Whereas MC38 metastatic model was more responsive anti-PD-L1 alone was effective only in MC38, combining diABZI with anti-PD-L1 in SL4 profoundly reduced metastatic burden. This response was accompanied by expansion of proliferative, tissue-resident CD8⁺ T cells, activation of NK/ILC1 subsets, and a shift toward less immunosuppressive macrophage states. These findings support STING-based combination strategies as a promising therapeutic avenue for poorly immunogenic CRLM.