Human GM-CSF/IL-3 enhance tumor immune infiltration in humanized HCC patient-derived xenografts

Response to immunotherapy in hepatocellular carcinoma (HCC) is suboptimal with no biomarkers to guide patient selection. "Humanized" mice represent promising models to address this deficiency but are limited by variable chimerism and underdeveloped myeloid compartments. We hypothesized that expression of human GM-CSF and IL-3 increases tumor immune cell infiltration, especially myeloid-derived cells, in humanized HCC patient-derived xenografts.

NOG (NOD/Shi-scid/IL-2Rγ^null) and NOG-EXL (huGM-CSF/huIL-3 NOG) mice conditioned with busulfan underwent i.v. injection of human CD34+ cells. HCC patient-derived xenograft tumors were then implanted subcutaneously or orthotopically. Following serial blood sampling, mice were euthanized at defined tumor sizes. Tumor, blood, liver, and spleen were analyzed by flow cytometry and immunohistochemistry.

Humanized NOG-EXL mice demonstrated earlier and higher levels of human chimerism compared to humanized NOG mice (82.1% vs. 43.8%, p <0.0001) with a greater proportion of human monocytes (3.2% vs. 1.1%, p = 0.001) and neutrophils (0.8% vs. 0.3%, p = 0.02) in circulation. HCC tumors in humanized NOG-EXL mice exhibited greater human immune cell infiltration (57.6% vs. 30.2%, p = 0.04) with higher proportions of regulatory T cells (14.6% vs. 6.8%, p = 0.04), CD4+ PD-1 expression (84.7% vs. 32.0%, p <0.01), macrophages (1.2% vs. 0.6%, p = 0.02), and neutrophils (0.5% vs. 0.1%, p <0.0001). No differences were observed in tumor engraftment or growth latency in subcutaneous tumors, but orthotopic tumors required implantation at 2 rather than 4 weeks post-humanization for successful engraftment. Finally, utilizing adult bone marrow instead of fetal livers enabled partial HLA-matching to HCC tumors but required more CD34+ cells.

Human GM-CSF and IL-3 expression in humanized mice resulted in features more closely approximating the immune microenvironment of human disease, providing a promising model for investigating critical questions in immunotherapy for HCC.

This study introduces a unique mouse model at a critical point in the evolution of treatment paradigms for patients with hepatocellular carcinoma (HCC). Immunotherapies have become the first-line treatment for advanced HCC; however, response rates remain low with no clear predictors of response to guide patient selection. In this context, animal models that recapitulate human disease are greatly needed. Leveraging xenograft tumors derived from patients with unresectable HCCs and a commercially available immunodeficient mouse strain that expresses human GM-CSF and IL-3, we demonstrate a novel but accessible approach for modeling the HCC tumor microenvironment.