Cancer and Immunity · global
AI-designed protein directs T cells to attack gastric cancer, while dendritic cells bolster the antitumor response
Bringing immune cells close to cancer cells may not be enough to sustain an attack. A new study in gastric cancer cells, organoids, and mice found that AI-assisted protein molecules had cancer-killing activity, but adding dendritic cells further improved tumor control.
One challenge in cancer immunotherapy is ensuring that immune cells continue to function after finding a tumor. A team at the First Affiliated Hospital of Nanjing Medical University used AI-assisted protein design to direct T cells toward gastric cancer cells; experiments showed, however, that the initial tumor-suppressing effect gradually weakened. Adding dendritic cells improved the antitumor response, suggesting that treatment success also depends on whether sufficient immune support is available around the tumor.
The study was published in the Journal of Translational Medicine on August 28, and Bioengineer.org and Scienmag covered its findings on October 9. The team used the protein design tool RFdiffusion to generate and screen candidate proteins capable of binding the gastric cancer target CLDN18.2, then assembled them into “T-cell engagers” (TCEs): one end binds the target on the cancer cell surface, while the other binds CD3 on T cells, bringing the two cells together to enable an immune attack. AI assisted with molecular design, while subsequent experiments tested biological function.
Validation included engineered gastric cancer cells, patient-derived tumor cells, and three-dimensional organoids grown from patient tissue. In these co-culture systems, the molecules triggered CLDN18.2-dependent cancer cell killing and activated CD8-positive T cells. Tests detected limited binding to cells expressing the related protein CLDN18.1; this provides preliminary evidence of selectivity, but does not establish safety for use in humans.
Mouse experiments revealed another limitation. In syngeneic tumor models and models implanted with patient tumors and supplemented with human immune cells, the researchers observed early tumor suppression, but tumor control weakened over time. Accompanying findings included insufficient T-cell accumulation within tumors, increased PD-1 expression, scarce dendritic cells, and weak signals from immune-cell-attracting molecules such as CXCL9 and CXCL10. These results point to insufficient local immune support, although the causal role of each factor requires further clarification.
The team therefore added dendritic cells and found improvements in both cancer-killing activity in co-culture and tumor control in mice. In addition to presenting antigens to T cells, dendritic cells can provide cytokines that regulate immune responses; in this set of experiments, their contribution appeared to extend beyond antigen presentation.
To investigate the mechanism, the researchers separately blocked MHC-II, which participates in antigen presentation, and the cytokine IL-12. Blocking the former had a smaller effect on the enhancement, whereas blocking the latter more consistently weakened the benefits provided by dendritic cells. This supports an important role for IL-12-related cytokine support under the tested conditions and provides a testable direction for the subsequent design of combination therapies.
The evidence currently comes from culture systems, organoids, and mice, with no human clinical trial results yet. Scienmag noted that the source of dendritic cells, manufacturing methods, timing of administration, and delivery routes remain unresolved questions for clinical translation; the journal also noted that the current version is a peer-reviewed accepted manuscript that may still undergo editorial revisions. The next step proposed by the study is to improve the targeting molecules while testing how to sustain immune support within tumors and whether such a combination can achieve both efficacy and safety.