Science
Keystone Epitope Theory
A framework for asking which microbial, self, drug-altered and donor peptide-HLA complexes share T-cell recognition, and whether that recognition contributes to disease.
Keystone Epitope Theory proposes that persistent, human-adapted organisms repeatedly stimulate selected immune responses within the tissues they occupy. We test whether T cells recognizing these microbial peptide-HLA complexes can also recognize particular self, drug-altered, or donor peptide-HLA complexes, and whether that recognition contributes to disease.
The framework generates candidate mechanisms and experiments. Each disease-specific connection requires direct validation.
Where it applies
-
Drug hypersensitivity
Abacavir provides a defined example of how a drug can change immune recognition. The parent drug binds noncovalently within the HLA-B*57:01 peptide-binding groove and changes the repertoire of presented self peptides. KET asks whether prior immune experience contributes to particular drug-reactive T-cell responses. That proposed connection requires direct testing.
-
Autoimmunity
The framework asks whether some pathogen-reactive T cells also recognize self peptide-HLA complexes. Sequence similarity alone does not establish cross-reactivity or tissue injury; natural presentation and functional effects must be tested.
-
Transplantation
The framework examines whether prior microbial exposure helps shape T-cell recognition of donor peptide-HLA complexes. Individual receptor interactions and their contribution to graft injury require direct evidence.
-
Vaccine design
KET proposes evaluating conserved epitopes together with presentation context, the fitness cost of escape, and other immune interactions. These are research propositions to test in specific vaccine settings.
How we look for triplets
- 01
Seed
Start from an experimentally characterized epitope of a persistent, human-adapted organism, such as a herpesvirus, with a defined HLA restriction and a relevant T-cell response.
- 02
Search
Prioritize candidate peptides using receptor-facing features and structural context. Computational similarity generates candidates; it does not establish T-cell recognition.
- 03
Restrict
Use HLA-binding predictions to prioritize candidates, then test binding and natural presentation. Distinguish predicted binding, measured binding, and presentation in cells.
- 04
Localize and test
Assess source-protein expression, cell state, and natural peptide presentation in the affected tissue. Test recognition by the same TCR and relevant functional effects before linking a candidate to disease.
How we assess evidence
These are distinct kinds of evidence. A positive result in one category does not establish the others.
- Prediction
- Computational prioritization of a candidate peptide, HLA restriction, or receptor interaction.
- Binding
- Measured peptide-HLA binding or receptor binding, with the assay and molecular partners specified.
- Recognition
- Demonstrated recognition of the relevant complexes by the same TCR, with appropriate controls.
- Natural presentation
- Evidence that relevant cells process and display the candidate peptide-HLA complex.
- Functional effect
- A relevant T-cell response, such as activation or cytotoxicity, in a defined experimental system.
- Evidence from affected tissue
- Evidence connecting the receptor, presented target, and response in disease-relevant tissue. Causal conclusions require further controls and, where feasible, perturbation.
A reported negative result
A reported negative result helps define the limits of current evidence. Sooda and colleagues tested two HLA-B*57:01-restricted EBV-specific TCRs and found that abacavir inhibited their viral recognition but did not cause self-reactivity in the tested system. This is not a demonstration of the proposed viral-to-self cross-reactivity, and it does not exclude every possible receptor-target pair.
The framework is set out in full in two open-access reviews in Pathogens and Immunity, August 2026.
See the publications