Seeking the selectors

The nature of the peptides that mediate thymic positive selection has been much debated. Reporting in the September issue of Immunity, Santori and colleagues have characterized peptides that induce positive selection using bioassay- and bioinformatics-based approaches. Reassuringly, both lines of investigation picked out the same two peptide ligands.

Positive selection is the process by which CD4⁺CD8⁺ double-positive thymocytes that interact with self-peptide–MHC complexes are selected to survive and mature. This developmental checkpoint is thought to ensure that all mature T cells can interact with self-MHC molecules. But, what makes a peptide a positive selector and how common are they?

The authors addressed these issues by characterizing naturally occurring peptides that mediate positive selection in a T-cell receptor (TCR)-transgenic model. OT-1 mice express a TCR that is specific for an ovalbumin (OVA) peptide presented by the MHC class I molecule H2-K^b.

In the first approach, bound peptides were extracted from H2-K^b molecules that had been purified from various cell lines and tissues. Peptide fractions were then tested for activity in an assay for positive selection — the 'co-receptor-dulling assay'. This measures the ability of a peptide ligand to induce the downregulation of expression of the CD4 or CD8 co-receptors on thymocytes, which is an in vitro correlate of positive selection. Candidate peptides in positive fractions were sequenced, then synthesized and tested for their ability to induce positive selection in fetal thymic organ culture. This process identified two positively selecting peptides: one from β-catenin (Catnb_329–336) and the other from F-actin capping protein (Cappa1_92–99).

The Catnb_329–336 and Cappa1_92–99 peptides have some structural homology to the OVA peptide that is the OT-1 TCR agonist. So, can positively selecting ligands be predicted on the basis of sequence homology? The second bioinformatics-based approach tested this theory. A database of predicted mouse MHC ligands was assembled, then algorithms were devised to select ligands that were similar to the OVA peptide at key TCR contact sites. Only ligand peptides for which the parent protein is expressed in the thymus were considered. Seven candidate peptides were identified which were then tested for their ability to bind to H2-K^b and for activity in the co-receptor-dulling assay. Ruling out peptides that are not actually presented by H2-K^b in the thymus gave an end result of two peptides — Catnb_329–336 and Cappa1_92–99 — which, remarkably, are the same peptides that were identified by the bioassay-based approach.

So, it seems that the peptides that induce positive selection are rare and have key TCR contact residues in common. Moreover, this study sets an important precedent for the general use of bioinformatics to predict the ligands that mediate positive selection.