The task of identifying the ligand that stimulates a specific T cell receptor (TCR) is one that has proven incredibly complex due to the vast diversity in the TCR, human leukocyte antigens (HLA), and peptide components. While powerful methods such as mass spectrometry or epitope prediction algorithms can narrow the list of possibilities, these strategies require a priori information about the target, and the identification process is lengthy and cumbersome. In an effort to better determine the specificity of unidentified TCRs, also known as “orphan” TCRs, Gee et al. developed an affinity-based TCR ligand identification strategy utilizing a diverse yeast display library of surface-displayed peptide:HLA (pHLA) molecules to identify multiple selected antigen targets of TCRs isolated from human tumor-infiltrating lymphocytes, and used bioinformatics approaches to infer the actual human peptidome epitopes.

To limit the complexity of variables in their study, the researchers screened only the HLA-A*02:01 allele (highly prevalent in a number of populations) and therefore could only identify HLA-A*02:01-restricted TCRs. They generated multiple peptide libraries, accounting for peptide lengths from eight to eleven amino acids. Their combined libraries represented approximately 400 million unique peptides.

To confirm that the HLA-A complex used in their library was able to properly fold and present peptides, they used the library, along with deep sequencing and cluster analysis, to identify an antigen for a TCR with a known specificity (the DMF5 TCR for MART-1). While they did identify a wide range of peptides that were able to bind the TCR, many in one cluster were closely related to the known antigen, and ultimately MART-1 was identified as the most likely target.

Upping the difficulty level, the researchers next performed a blinded validation study in which they attempted to correctly identify the target antigens of three TCRs (derived from TILs from a human melanoma patient) with known specificities to neoantigens. Using their system, the researchers were able to correctly match one of the three TCRs to a known predicted neoantigen target (mutated CDK4). They further showed that the top predicted peptides were actually able to potently stimulate TCR-transduced T cells, despite sequence differences from the actual epitope; some peptides were as potent as the epitope itself.

To test whether their system could then work to identify truly unknown antigen targets, the researchers identified target TCRs from resected tumors from two patients with colorectal adenocarcinoma who were homozygous for the HLA-A*02 allele. Both tumor samples contained some T cell clones that were expanded in the tumor and not found in a more limited sampling of healthy colon tissue, possibly indicating antitumor reactivity. Based on a profile of local expansion, cytotoxic profile (inferred from single-cell RNA sequencing), and other characteristics, the researchers selected twenty candidate TCRs to test. Of the twenty, four exhibited strong selection for peptides presented in the library. After initial analysis using currently available algorithms failed, the researchers developed two new methods (an improved statistical approach and a neural network approach) that allowed them to identify probable antigen targets for three of the four TCRs, all of which were self-antigens, not neoantigens.

Interestingly, two of the TCRs, isolated from different patients, expressed the same TCRα chain, a highly similar TCRβ chain, and selected for a subset of related peptides. After complete analysis, both TCRs were found to most likely bind a peptide derived from U2AF2, a protein involved in an RNA splicing complex that is known to be overexpressed in many cancers, including colorectal cancer. As a point of difference, however, one TCR was found to have high cross-reactivity to multiple pHLA, while the other was fairly rigid in its specificity. The ability to determine the degree of cross-reactivity of a TCR using this system could one day become relevant in identifying TCRs for adoptive cell therapy that are more selective in order to prevent on-target, off-tumor effects.

While this affinity-based strategy for identifying the antigen targets of orphan TCRs is still in its very early stages and is subject to a number of limitations, this proof-of-concept paper offers a major step forward in tackling an incredibly complex challenge. The ability to quickly and easily identify putative antigen targets of orphan TCRs without extensive expansion or guesswork that current strategies require would allow for major advances in immunotherapy research, both in the lab and in the clinic.

by Lauren Hitchings