Comprehensive analysis of αβT-cell receptor repertoires reveals signatures of thymic selection.

Thymic selection is crucial for forming a pool of T-cells that can efficiently discriminate self from non-self using their T-cell receptors (TCRs) to develop adaptive immunity. In the present study we analyzed how a diverse set of physicochemical and sequence features of a TCR can affect the chances of successfully passing the selection. On a global scale we identified differences in selection probabilities based on CDR3 loop length, hydrophobicity, and residue sizes depending on variable genes and TCR chain context. We also observed a substantial decrease in N-glycosylation sites and other short sequence motifs for both alpha and beta chains. At the local scale we used dedicated statistical and machine learning methods coupled with a probabilistic model of the V(D)J rearrangement process to infer patterns in the CDR3 region that are either enriched or depleted during the course of selection. While the abundance of patterns containing poly-Glycines can improve CDR3 flexibility in selected TCRs, the "holes" in the TCR repertoire induced by negative selection can be related to Arginines in the (N)-Diversity (D)-N-region (NDN) region. Corresponding patterns were stored by us in a database available online. We demonstrated how TCR sequence composition affects lineage commitment during thymic selection. Structural modeling reveals that TCRs with "flat" and "bulged" CDR3 loops are more likely to commit T-cells to the CD4+ and CD8+ lineage respectively. Finally, we highlighted the effect of an individual MHC haplotype on the selection process, suggesting that those "holes" can be donor-specific. Our results can be further applied to identify potentially self-reactive TCRs in donor repertoires and aid in TCR selection for immunotherapies.