Science
The Keystone Epitope Theory
A small number of epitopes carry a disproportionate share of the immune system's attention. Finding them explains a great deal that is currently called idiopathic.
The immune system does not distribute its memory evenly. Persistent, human-adapted pathogens — herpesviruses above all — focus postnatal immune memory on a small set of conserved epitopes, in the specific tissue niches where those pathogens are controlled. Conservation is not an accident: these are positions where the virus cannot easily mutate without paying a fitness cost, so the response that targets them stays useful for decades.
We call those epitopes keystone epitopes, after the ecological term. They hold a disproportionate share of the structure around them. Remove one, or imitate one, and a great deal moves.
The consequence is that a large, long-lived, tissue-resident T cell population sits waiting for a very specific shape. Anything that presents a close enough copy of that shape inherits the response — a drug metabolite bound in the HLA groove, a self-protein in an inflamed tissue, a mismatched allele on a transplanted organ. Diseases that look unrelated turn out to share a mechanism.
Where it applies
- Drug hypersensitivity. The HLA-B*57:01 association with abacavir hypersensitivity is the worked example: a defined allele, a defined drug, a defined T cell response, and a screening test that changed prescribing worldwide. The theory asks which other reactions have the same shape.
- Autoimmunity. Where a viral epitope and a self-protein are similar enough at the receptor-facing positions, an anti-viral response becomes an anti-self response. Tissue restriction of the self-protein then predicts which tissue is damaged.
- Transplantation. Pre-existing memory against keystone epitopes can cross-react with mismatched donor HLA, which reframes part of alloreactivity as heterologous immunity rather than a response to something genuinely new.
- Vaccine design. Three principles follow: prioritize epitopes whose substitutions carry a measurable fitness cost, avoid immunogens dominated by mutable targets, and account for presentation context and inhibitory NK signaling when choosing what to include.
How we look for triplets
- 01
Seed
Start from an epitope of a persistent, human-adapted pathogen — a herpesvirus, HIV, a virus the immune system has been holding in check for a lifetime.
- 02
Search
Look across human proteomes for peptides that a T cell receptor raised against that seed could also bind, scoring similarity at the receptor-facing positions rather than across the whole peptide.
- 03
Restrict
Keep only candidates that the relevant HLA allele actually presents, using binding prediction and the allele frequencies that matter for the disease in question.
- 04
Localize
Ask where the candidate protein is expressed, and in which cell state. A cross-reactive target only explains a disease if it is present in the tissue the disease damages.
The theory is set out in full in two open-access reviews in Pathogens and Immunity, August 2026.
See the publications