← Back to Home

Why Can’t T Cells Enter Tumors? AI Finds Clues to a Mucin Barrier in Tissue Maps

SPACER analyzes cell locations alongside gene activity, identifying clues to immune cell clustering across 37 datasets. The study links tumor mucins to poorer immunotherapy responses, but further validation is needed before they can become clinical tests or therapeutic targets.

By SURL BioNews

Even when immune cells are present near a tumor, they may not be able to approach cancer cells and launch an attack. Signals that attract them may be missing, or a barrier made of molecules may separate them. SPACER, an artificial intelligence model developed by the University of Texas MD Anderson Cancer Center and collaborators, attempts to identify clues affecting immune cell clustering from gene activity and cell locations in tissues.

The study, published on October 13 in *Nature Biomedical Engineering*, was jointly led by Tao Wang and Linghua Wang. SPACER uses spatial transcriptomics data: these technologies measure gene expression while also preserving the locations of signals within tissue. The model analyzes groups of neighboring cells, linking gene activity in the surrounding environment to immune cell distribution so researchers can investigate which local signals may influence cells to move closer or farther away.

The team applied the model to 37 datasets covering various human tumors and mouse hearts. In benchmark tests, including those assessing the identification of biologically meaningful cell recruitment signals, SPACER achieved 4- to 21-fold improvements in performance compared with existing methods. This figure measures its ability to perform analytical tasks; it does not mean that cancer diagnostic accuracy or treatment effectiveness improved by the same magnitude.

One clue pointed to antigen presentation. Through this process, cells display protein fragments for immune cells to recognize. The study found that expression of genes related to antigen presentation in tumors was closely associated with T cell proximity. This shifts the research focus from “how many T cells are in the tissue” to “which signals near their locations may promote contact.”

Another clue involved mucins. These large proteins, which carry numerous sugar chains, can be produced by tumor cells. The findings support their possible involvement in forming a physical barrier that prevents T cells from approaching. According to MD Anderson’s description of the research, higher mucin expression in melanoma and gastric cancer datasets was also associated with poorer responses to immune checkpoint inhibitors and poorer patient outcomes. However, these associations do not yet prove that reducing mucins can restore treatment effectiveness.

The model was also applied to preclinical data on immunotherapy-related myocarditis, identifying regions and molecular signals in the heart associated with T cell clustering and suggesting that CD4-positive T cells may play a larger role in severe inflammation. This finding provides a research direction for investigating treatment side effects, but it cannot be used directly to assess an individual patient’s risk of myocarditis.

SPACER’s current value lies in turning tissue distribution maps into hypotheses that can be tested experimentally. Before mucins or model scores can be used to select immunotherapies, their reliability across different patient populations must be established, along with whether they can provide information beyond existing tests. Developing new treatments will require intervention experiments to confirm that changing these signals can indeed improve immune cell contact and anticancer effects.

References

  1. UT MD Anderson Cancer Center
  2. UT MD Anderson Cancer Center via Newswise
  3. Medical Xpress, content provided by UT MD Anderson Cancer Center