T cells are part of the adaptive immune system that recognize pieces of harmful pathogens or cancer cells to target them for elimination. This recognition is mediated by the T cell receptor (TCR) a heterodimer that binds to antigen-presenting complexes on other cells and subsequently leads to T cell activation. TCRs are composed of an alpha and a beta chain, and the structure of both chains dictates which TCR can bind to a specific antigen.
Structural diversity in the alpha and beta chains and combinatorial pairing of these chains is crucial to producing diverse T cells that can recognize all the different epitopes they could encounter, but efforts to associate TCR structural features with the epitope it recognizes have been slow-moving. TCRs are generated by a random gene-shuffling process and the potential diversity of the receptor far exceeds the number of T cells in an individual. This means that different individuals use distinct TCRs to respond to the same epitope, complicating efforts to formulate universal rules for specificity. “If we look at TCRs that recognize the same epitopes, they kind of look the same, but the similarity is unequally distributed. Sometimes, we see that all the beta chains look the same, but the alphas are different. Sometimes, all the alpha chains look the same, but the betas are different,” explains Dr. Anastasia Minervina, a staff scientist in the Thomas lab in the Vaccine and Infectious Disease Division.
This problem led Minervina and colleagues to develop a quantitative way to measure how much the structure of each individual chain and the combination of the chains influence epitope recognition. She did this by replacing an endogenous TCR alpha or beta chain in donor immune cells with an exogenous chain known to recognize a specific epitope in certain contexts. From there, they allowed the exogenous chain to form “hybrid” TCRs with many endogenous chains in the donor cells. They next analyzed which chains could pair and which pairs could still recognize the known epitope. “Somewhat surprisingly, we found that most [TCR chain pairs] could pair, meaning they could form a stable receptor and go to the surface of the cell,” says Minervina, “But, when we measured recognition, [most of the pairs] lost binding.” At most, they saw that 10% of hybrid TCRs could recognize the known epitope, highlighting the pitfalls of inferring TCR specificity using structural information from a single chain alone.
Next, the team wanted to know if the epitope-recognizing hybrid TCR pairs they observed mirror those observed in nature. They compared their TCR pairs to those from a public database of TCR pairs generated from natural epitope exposure. They found several identical matches from the natural TCRs and even more highly similar matches, demonstrating that their in vitro system can recapitulate TCRs generated from natural epitope exposure.
Even though the team found many TCR pairs that could bind their tested epitopes, the chain replacement strategy left them with even more pairs that were able to pair but unable to bind epitopes. Given that the nonbinding and binding TCR pairs are almost identical in structure, Minervina and Thomas wanted to know if they could use their nonbinding TCR data to improve machine learning algorithms that predict TCR-epitope binding. They integrated their positive and negative binding data with public databases of labeled TCRs and used a machine learning algorithm based on TCR sequence similarity to predict TCR-epitope binding. With the negative binding data generated by Thomas’s lab, their nearest neighbor classification successfully predicted which TCRs would bind a specific epitope, but using the same classifier trained on just the public data led to a high rate of false positives. This observation indicates that training models on highly similar positive and negative binding data are crucial to computationally predicting specificity boundaries.
The team next wanted to see if they could identify new binding partners for rarer TCRs. To do this, they introduced the alpha or beta chains of a TCR that targets a neoepitope from a rare type of cancer. The neoepitope differs from the healthy variant by only one amino acid, and the only known natural TCR was ineffective at discriminating between the epitopes in healthy and malignant cells. Using the chain replacement strategy, Minervina identified several previously unknown TCRs that were able to discriminate between epitopes from healthy and malignant cells.
Together, these data show that their chain replacement strategy can accelerate TCR identification at a scale that previously would have required sampling thousands of individuals. In the future, Thomas is hopeful that this sort of strategy could have implications for patients. “If we could look at your T cell receptor and tell what their epitope targets were going to be, whether that was in a tumor or infection, that would have therapeutic or diagnostic potential.” Thomas also hopes their data contributes to new predictive models of TCR binding. “Making much, much better, more subtle models that could really advance the field…There’s lots of people here and elsewhere that [these data] will be useful for.”
This work was supported by funding from St. Jude Children’s Research Hospital, the American Lebanese Syrian Associated Charities at St. Jude, the National Institutes of Health, the Agence Nationale de la Recherche, the European Research Council, the Foundation pour la Recherche Medical, and a Paul Barrett endowed Fellowship.
Fred Hutch/University of Washington/Seattle Children’s Cancer Consortium Members Drs. Paul Thomas and Phil Bradley contributed to this work.
Minervina AA, Pogorelyy MV, Mayer-Blackwell K, Tirtakusuma R, Schattgen SA, Fiore-Gartland A, Walczak AM, Mora T, Bradley P, Thomas PG. 2026. Uneven TCR chain pairing constraints govern epitope recognition. Science. 2026 Sep 3;393(6815):eadx3863. doi: 10.1126/science.adx3863.
Kelsey Woodruff is a PhD candidate in the Termini Lab at Fred Hutch Cancer Center. She studies how acute myeloid leukemia cells remodel the sugars on their membranes to reprogram cancer cell signaling. Originally from Indiana, she holds a bachelor's degree in Biochemistry from Ball State University. Outside of lab, you can find her crocheting and enjoying the Seattle summers.