The mutations acquired by a tumor can generate neoepitopes that mediate immune recognition and rejection of the tumor. But there are conflicting reports as to whether a tumor's potential neoepitope load is useful as a reliable indicator of immunogenicity across cancer types. Two recent papers in Nature shed light on this issue, showing that neoepitope quality and not quantity is what matters. The work makes strides not only because of its potential implications for patient selection and therapy development but also because the insights uncovered tell an important story on the immunological basis of tumor regression.

Many factors have been implicated in a responder signature for patients receiving checkpoint blockade, an antibody-based approach to unleash tumor-specific T cells. But none has been sufficient to reliably select patients who will respond, and the search has continued for predictors of tumor immunogenicity. One puzzling factor has been the presence of potential neoepitopes—peptides derived from tumor-specific mutations that are presented on tumor cells by the major histocompatibility complex (MHC) and recognized by T cells. High tumor mutation burden is not always associated with immunogenicity, and tumors with low mutational load do sometimes respond to checkpoint blockade or other forms of immunotherapy.

The work of Łuksza et al.1 and Balachandran et al.2 focuses on the 'quality' of neoepitopes instead of their numbers and demonstrates that computed neoepitope characteristics are important determinants of tumor immunogenicity in several cancer types, including low mutational burden and aggressive tumor types, such as pancreatic cancer.

Łuksza et al.1 develop an algorithm that considers two major aspects of neoepitope recognition: the difference in inferred binding affinity to MHC of the mutant peptide versus its wild-type version, and the likelihood that the mutant peptide will be recognized by T-cell receptors (TCRs) in a patient's repertoire. For the latter estimation, the model considers how well the mutant peptide aligns to thousands of T-cell epitopes known to be recognized by human TCRs. The idea is that if neoepitopes share characteristics of epitopes known to be recognized by the immune system, the chances that they will be seen as 'non-self' and targeted by T cells are increased.

Credit: Figure reproduced with permission from ref. 1.

Robert Vonderheide, director of the Abramson Cancer Center at the University of Pennsylvania, says of the approach, “This model will be a first-generation attempt; there's bound to be refinements, but it's a very clever way of assessing the likelihood that a human peptide has a match in the human TCR repertoire.”

Applying their algorithm to cancer types with a relatively high mutational burden (melanoma and non-small cell lung carcinoma), Łuksza et al.1 find that the model can stratify patients who respond or don't respond to checkpoint blockade therapy. This finding suggests that what makes a tumor different from its host in terms of immune recognition can be quantified, and that these scores have predictive value in terms of patient survival.

Balachandran et al.2 examine a cohort of pancreatic cancer patients with long-term (median 6 years) and short-term (median 0.8 years) survival. Long-term survivors, who are rare, had greater density of cytotoxic CD8+ T cells, a majority of which were tumor-specific. Together, but not separately, CD8+ T-cell infiltrate and predicted neoepitope quantity were associated with patient survival. The researchers then looked at the homology of neoepitopes to known infectious disease-related epitopes, as in the work of Łuksza et al.1, aiming to estimate the probability that a neoepitope will be recognized as 'non-self' by the TCR repertoire. They also find that neoepitope quality, but not quantity, can stratify long-term survivors.

These results “reaffirm with modern immunological understanding that even in pancreatic cancer, the immune system plays a role,” says Vonderheide. The idea of estimating differences between the mutant and wild-type epitope in the thermodynamics of MHC binding and TCR recognition is not new. But demonstrating that a model incorporating these factors can help predict patient survival underscores the importance of this biology to immune recognition. It could also help design therapeutic cancer vaccines that incorporate neoepitopes.

“Now we understand that there can be immune recognition. We have to be smarter and innovative in finding strategies that target this immunology,” says Vonderheide. “And it will undoubtedly go further than checkpoint blockade.”