Antigen Presentation

T cells do not see proteins; they see peptides bound in the groove of an MHC molecule. Class I molecules, on almost all nucleated cells, carry fragments of about eight to ten residues generated by the proteasome, and are read by CD8 T cells. Class II molecules, restricted to dendritic cells, macrophages and B cells, carry longer fragments of roughly thirteen to twenty-five residues produced in endolysosomes, and are read by CD4 T cells.

Because this is a peptide interaction, it obeys peptide rules. Each MHC allele has anchor positions with defined side-chain preferences, so HLA-A2 favours leucine or methionine near the amino end and a hydrophobic residue at the carboxy end, which is what makes computational epitope prediction possible. The pathway is also a drug safety problem: HLA-B*57:01 screening became standard before abacavir because the drug alters which self peptides that allele presents, turning a normal repertoire into an apparently foreign one.

This is why immune responses are genetically restricted. An epitope works only in people carrying an allele that presents it, so an immunotherapy built around one peptide is inherently a subgroup therapy. It also explains why tumours with high mutational burden respond better to checkpoint blockade, since more mutations mean more novel peptides available for display.

The common overreach is treating predicted binding as demonstrated immunogenicity. Prediction scores affinity for the groove, which is necessary but far from sufficient: the peptide must still be produced by proteasomal cleavage, survive transport and compete with the rest of the repertoire. Results in HLA-A2 transgenic mice are likewise generalised to whole populations, though that allele covers a minority of people worldwide and is much less common outside European ancestry.