An engineered signaling protein and receptor pair with little cross-talk with natural signals directed T-cells to tumors producing the protein in mice and helped them form a “swarm” in cell culture
An engineered signaling protein and receptor pair with little cross-talk with natural signals directed T-cells to tumors producing the protein in mice and helped them form a “swarm” in cell culture
On October 14, 2026, Wendell Lim’s group at the University of California, San Francisco posted a preprint, a manuscript that has yet to undergo peer review, on bioRxiv. The authors modified a natural pair: the signaling protein CCL20 and its receptor CCR6.
A chemokine is a protein that cells release as a signal: immune cells with a matching receptor move toward higher concentrations of it. In humans, 46 chemokines and 23 receptors form a network with more than a hundred confirmed interactions, so a chemokine that recruits tumor-attacking T-cells can also attract cells that suppress the immune response.
In 2014, Lim’s laboratory had already used an artificial signal to direct T-cells: cells equipped with a receptor for a drug that natural cells do not respond to moved toward implanted drug-containing beads in a mouse. The experimenter supplied that signal. In the new pair, both the signal and the receptor are encoded in the cells’ genes, and the pair is designed to operate independently of the natural network.
The researchers chose CCL20 and CCR6 because, according to the authors, they signal exclusively through each other, and their small, shallow binding interface is easier to modify. The authors changed the charges of three amino acids at this interface, one in the chemokine and two in the receptor, to preserve its shape while making the modified partners compatible with each other. The chemokine substitution reduced its activity at the natural receptor by a factor of 20–30; the receptor substitutions restored the response to the modified chemokine.
ProteinMPNN, a neural network that selects amino acids to fit a protein’s structure, identified two additional substitutions for the receptor and independently proposed the same chemokine substitution that the authors had chosen. The receptor with four substitutions was more selective for the modified chemokine, so the researchers used it in subsequent experiments. None of the 44 chemokines tested appreciably activated the modified receptors, and the modified chemokine strongly activated only those receptors. Human blood cells that migrate toward CCL20, including regulatory T-cells that suppress the immune response, did not migrate toward the modified chemokine.
Immunodeficient mice each received two tumors derived from human lung cancer cells: one produced the modified chemokine, and the other produced CCL20 or a control protein. Intravenously injected T-cells carrying the modified receptor accumulated preferentially in the tumor producing the modified chemokine. Injecting a virus carrying the chemokine gene into the tumor produced the same result.
Because the signal and receptor are genetically encoded, cells can form a “swarm”: a cluster of cells that both release the signal and move toward it. The researchers built the swarm using synNotch, a receptor developed in Lim’s laboratory in 2016 that switches on a chosen gene when it encounters a specified protein on another cell. In cell culture, the target was a spheroid, a small ball of tumor cells expressing CD19. The T-cells were equipped with the modified receptor and a CD19-recognizing synNotch receptor that switched on the gene for the modified chemokine.
The first T-cells, the “pioneers,” moved toward the spheroid in response to the natural chemokine CXCL12. Upon encountering CD19, a pioneer began releasing the modified chemokine. This attracted additional cells carrying the modified receptor, which encountered CD19 and released more chemokine themselves: a small first wave was followed by a much larger second wave. Only the first wave reached a spheroid lacking CD19, and T-cells that lacked the modified receptor and responded only to CXCL12 did not join the second wave. The authors write:
These tools allow us to program macroscopic behavior, determining which anatomical site a cell population moves toward within the complex environment of the body, rather than altering direct molecular recognition of target cells […].