Getting CAR-T Cells In—and Keeping Them Going: Dual Gene Editing Clears Lung Tumors in Mice

CAR-T cells have repeatedly achieved strong results in blood cancers, but in solid tumors they often resemble troops that cannot find a way in and, even if they arrive, rapidly lose their ability to fight. A team from Gladstone Institutes and the University of California, San Francisco used an in vivo genome-wide CRISPR screen to identify P2RY8 and GNAS, which respectively constrain T-cell migration and function. Knocking out both produced a markedly stronger therapeutic response in a mouse lung tumor model.

Most previous CRISPR screens have been conducted in culture dishes, making it difficult to reproduce the hypoxia, nutrient deprivation, acidification, and multiple immunosuppressive signals found within tumors. The researchers therefore infused primary human T cells carrying a genome-wide knockout library into tumor-bearing mice and engineered A375 melanoma cells into a model capable of attracting T cells through an anti-CD3 single-chain antibody fragment, allowing enough cells to be recovered from each tumor for large-scale analysis.

A screen based on the number of T cells within tumors first pointed to the P2RY8–Gα13 signaling pathway. P2RY8 senses metabolites released by tumors and restricts cell migration; after P2RY8 was knocked out, human T cells entered tumors more readily. This receptor has no direct mouse ortholog, also highlighting how screening only mouse-derived immune cells could miss regulatory mechanisms with therapeutic relevance in humans.

Another screen, which distinguished T-cell function by interferon-γ production, identified GNAS. The Gαs protein it encodes lies downstream of multiple inhibitory G protein-coupled receptor signals. T cells lacking GNAS were better able to maintain metabolism, survival, and effector function in acidic conditions and in the presence of adenosine, prostaglandin E2, and β-adrenergic receptor signaling. In other words, P2RY8 knockout improved tumor “entry,” while GNAS knockout helped cells “keep fighting” inside the tumor.

The research team then incorporated both edits into CAR-T cells. In a xenograft model using engineered A549 human lung cancer cells expressing CD19, what the study described as low-dose, dual-knockout CAR-T cells produced complete responses in 4 of 6 mice. The group with only GNAS knocked out had 1 complete response, while neither the unedited control group nor the group with only P2RY8 knocked out had any complete responses. The results support the possibility that combining complementary cellular traits may be more effective than enhancing either one alone.

However, this does not mean CAR-T cells can already clear lung cancer in patients. The screening system artificially increased the number of T cells entering tumors by using an anti-CD3 fragment, and the A549 cells in the efficacy experiment were also given the CD19 target. The study used immunodeficient mice, which cannot fully represent the human immune system, tumor heterogeneity, or antigen loss. The related research remains a preprint that has not yet undergone journal peer review, and the treatment data included only 6 animals per group.

Before moving toward clinical development, the team must confirm efficacy in models that more closely resemble naturally occurring human tumors and assess off-target editing, long-term cellular behavior, and toxicity associated with the dual gene knockout. UCSF has listed the screening platform and related findings as patent pending. At this stage, its most concrete value may not be to declare that a new therapy is mature, but to provide a method for finding combinations of T-cell modifications under the pressures of an actual tumor environment.