18 SEP 2026 — Researchers at Dana-Farber Cancer Institute predicted 175 peptide fragments that mutated p53 ought to display on the surface of a cancer cell. They could robustly detect five.
The missing 170 are the finding. These tumours are not simply carrying a mutation the immune system overlooks. They are destroying the evidence before it reaches the surface.
What the study reports
The work appeared in Immunity on 17 September, from the laboratories of Ellis L. Reinherz and David A. Barbie at Dana-Farber, with Koji Haratani as first author and collaborators at Harvard Medical School, Vanderbilt University, Argonne National Laboratory and Boston Children's Hospital.
TP53 is mutated in roughly half of all human cancers, which is why it has been an obvious immunotherapy target and an unrewarding one. The reasoning is straightforward. A mutated protein yields fragments unlike the healthy version, those fragments appear on the cell surface bound to HLA molecules, and T cells can be directed at them.
But that only works if the fragments actually appear. Of 175 predicted candidates, five were robustly detected.
"Immunotherapy cannot attack what the immune system cannot see," Reinherz said.
Two ways a target disappears
The team found different failures behind two different mutations, so the work does not rest on a single mechanism.
For the p53 I195F mutation, the problem is an enzyme called ERAP1. ERAP1 trims peptides during normal antigen processing, and in these cancer cells its activity was high enough to destroy a strongly immunogenic p53 fragment before it could be presented. Deleting ERAP1, or blocking it with an inhibitor, restored p53-specific T-cell recognition.
For R175H — a common mutation already being pursued with engineered T-cell therapies in the clinic — the failure is physical, not enzymatic. T-cell receptors could respond to very sparse amounts of the target, but the mutant fragment bound its HLA molecule so weakly that the complex was unstable and short-lived. The target forms, then falls apart.
That second result is the more uncomfortable one, because it concerns a mutation that clinical programmes are already built around.
What this does not show
The release accompanying the paper says so directly, which is rarer than it should be. The study does not show that a drug has already converted cold human tumours into hot ones in patients. The results come from ultrasensitive immunopeptidomics, engineered human T cells and tumour-cell killing assays.
No patient was treated. ERAP1 also has a day job in healthy tissue across the body, and what systemic inhibition would do is not established here. That question is what a clinical programme would spend its early years on.
The 175 figure is a prediction, not a census. It is the output of a computational model, so the gap to five is partly tumour biology and partly the tendency of such models to over-call.
Why a negative result helps
Programmes targeting p53 have underperformed for years relative to how common the mutation is. If the fragments they aim at are largely not being presented, or are presented too briefly to matter, that explains a long run of disappointment better than assuming the immune response was simply too weak.
Barbie put the implication plainly: "A tumor may carry an ideal mutation in every cancer cell, but if that mutation is not processed into a stable surface target, making a stronger T cell may not be enough."
It reframes the engineering problem. Reinherz said the field's working question has been how to make the immune system attack harder. This work argues that, for these targets, the binding constraint is upstream, in what the tumour displays. We covered a related question last week in a model built to predict which lung-cancer patients respond to immunotherapy — that work asks who responds, this one asks why the target is absent.
What to watch
The first practical requirement is an ERAP1 inhibitor selective enough to spare normal antigen processing. Whether one exists is not answered here.
The larger question is whether the mechanism generalises to other proteins and tumour types. This is one enzyme, one set of cancers, one mutated protein. The possibility that tumours routinely edit their surface to conceal mutations needs replication across tumour types before anyone builds on it.
The R175H finding deserves separate attention. Engineered T-cell therapies against that mutation are already in the clinic. If the limiting factor is complex stability rather than T-cell potency, those programmes can act on it without waiting for an ERAP1 drug.
View author profile → · Editorial policy
