Immunogenicity of non-canonical HLA-I tumor ligands identified through proteogenomics
Lozano-Rabella, M. et al. (BioRxiv) DOI: 10.1101/2022.11.07.514886
The ability of the immune system to recognize cancerous cells is dependent on their expression of antigens that flag them as foreign and warranted for destruction. Neoantigens, derived from nonsynonymous mutations in the tumor genome and presented on MHC class I molecules, are thought to be the most immunogenic antigens since they are not expressed on healthy cells. Identification of these mutations from next generation sequencing of the tumor allows for the development of personalized therapeutics, but these mutations are often unique to an individual and prevents development of off-the-shelf therapeutics that can benefit multiple patients. In contrast, tumor associated antigens (TAA) are shared across patients and can be leveraged for off-the-shelf vaccines. Cancer germline antigens or cancer testis antigens (CGA/CTA) consist of 276 genes that are typically expressed exclusively on germ cells but cannot be presented on HLA, are expressed in some cancers due to demethylation or histone modification and are highly immunogenic. Notable examples are MAGE-A1 and NY-ESO-1, which have been used for T cell based therapies and vaccines. Melanoma-associated antigens, such as PMEL, are proteins essential for melanocytes but are over-expressed in cancer, and are also potent therapeutic targets. Yet these antigens are derived from the small percentage of the genome that is thought to be protein-coding. In fact, up to 75% of the genome can be transcribed and translated and many mutations in cancer are in these regions of the genome that are not annotated as protein-coding. Non-canonical HLA-I tumor ligands have recently been described in cancer, and they consist of proteins from the +1 or +2 reading frame of canonical ORFs, non-coding transcripts from 5’ or 3’ UTR, introns, intergenic regions, and noncoding RNA. These can be shared across patients and are hard to detect via WES, but since they are HLA-bound, they can be detected by mass spectrometry methods. In this preprint, the authors sought to identify the presentation and immunogenicity of these noncanonical tumor antigens.
Using 9 tumor cell line with matched ex vivo expanded TIL and PBMCs, and a powerful mass spectrometry and computational pipeline, they identified 839 nonC peptides on HLA-I, mainly from 5’UTR and bound primarily to A*03:01 and A*11:01 of which 517 were expressed on tumors and termed them nonC-TL (non-canonical tumor ligands). Importantly, 10% of these nonC-TL were shared in at least 2 patients, with one in found in 5 of the 9 patients and three found in 4 patients, whereas the neoantigens were private to an individual. Within previous datasets they found 76 of these nonC-TL found in immunopeptidomics datasets from melanoma, showing that they can be presented naturally. The abundance of nonC-TL was much higher than neoantigens or tumor associated antigens, which totaled less than 40 each.
To assess the role of these nonC-TL in cancer immune surveillance, they used an ex vivo coculture system of patient TILs and antigen presenting cells that had been pulsed with single peptide or peptide pools of the different classes of antigens. Recall responses to nonC-TL were absent in TILs quantified by IFN-g ELISA and 4-1BB expression via flow cytometry in contrast to responses to 39% of neoantigens and 5-7% of TAA, suggesting that they do not play a role in immune surveillance. Despite this, the authors reasoned that perhaps these nonC-TL could still be immunogenic. They performed in vitro sensitization of peripheral blood lymphocytes, which allows naïve T cells to be expanded due to activation by peptide pools of 170 nonC-TLs that they had previously identified. They were successfully able to expand and isolate three T cell clones, each specific to one of the candidate nonC-TLs, and mapping to 5’U-HOXC13, 5’U-ZKSCAN1, and non-coding C5orf22C. so shows sharing of nonC TL across tumor types and can be presented adn recognized by T cell s
In order to be therapeutically applicable, these nonC-TLs should be shared across patients but not be expressed on healthy tissue. Using RNA-seq data, the authors detected expression of the genes for the 3 immunogenic nonC-TL in tumors and healthy samples. To test this experimentally, they synthesized the TCRs from their expanded TIL pools and cloned them into PBLs, and then cocultured them with healthy or tumor cells lines artifically expressing the relevant HLA. They confirmed the tumor-specificity of the 5’U-HOXC13 peptide and 5’U-ZKSCAN1; T cells reacted to melanoma cell lines expressing the relevant HLA, but barely or not at all to healthy
The limited success of TAA vaccines and engineered T cells has been due to toxicities, but the tumor-specificity of nonC-TLs points to perhaps greater specificity and perhaps reduced toxicity, but this can only be directly assessed in clinical trials. Moreover, to more definitively conclude the tumor-specificity of these ligands, the authors should also test the reactivity of healthy cells from the same patient from which they detected the antigen.
However, the specificity reported contrasts with other studies which have suggested that they are indeed expressed on healthy tissue. The method used are ultimately limited in their detection capacity and using fresh tissue would enhance detection of more ligands. Biologically, it would be interesting to understand the benefit of nonC-TL expression for the tumor, since it increases its visibility to the immune system but does not produce functional protein that contributes to tumor fitness. If the authors finding that preexisting nonC-TL specific TILs were absent from the tumor and thus not contributing to immune surveillance is correct, perhaps they are expressed later in tumor development.
Overall, this preprint characterizes for the first time the immunogenicity and therapeutic applicability of an emerging class of tumor antigens that has the potential to become a potent shared, off-the-shelf therapeutic that can benefit many patients.
This study is significant for aiding the development of off the shelf vaccines or T cell therapies that could potentially be applied to many patients. Their novel proteogenomic and computational pipeline to detect a variety of HLA ligands has the potential to identify even more candidate ligands, especially when a larger sample size is used that incorporates more HLA haplotypes.
Reviewed by Miriam Saffern as part of the cross-institutional journal club of the Immunology Institute of the Icahn School of Medicine, Mount Sinai (U.S.A.), the Kennedy Institute of Rheumatology, University of Oxford (U.K.), and the Center for Molecular Medicine, Karolinska Institutet (Sweden).