Dynamics of TCRβ repertoires from serial sampling of healthy individuals

Abstract

T-cell receptor (TCR) repertoires provide a historical record of antigen exposure. However, the dynamics of TCR repertoires in healthy individuals remain largely uncharacterised. How much of the repertoire is under immune selection in healthy individuals? Do groups of sequences under immune selection share similar dynamics due to convergent specificity? What is the relationship between dynamic similarity and sequence similarity of TCRs? Here we develop a statistical framework for identifying clonotypes under immune selection in time series repertoire data. Applying this framework to serially sampled repertoires collected over the course of a year from 3 healthy volunteers, we are able to detect hundreds of TCRs undergoing strong immune selection whereby clonotype frequencies can change by orders of magnitude over timescales as short as a month. Clonotypes under immune selection belong to a handful of distinct dynamic clusters each of which show highly coordinated temporal behaviour suggesting a common immunogenic stimulus. Whilst a subset of clonotypes within dynamic clusters show shared amino acid motif usage, most do not, suggesting the same immunogenic stimulus elicits a diverse TCR response. Conversely, shared amino acid motif usage alone identifies far fewer clonotypes under immune selection and these clonotypes do not routinely exhibit correlated temporal behaviour. These results highlight the potential of using information contained in the dynamics of TCR repertoires for identifying clonotypes responding to the same immunogenic stimulus in a sequence agnostic way.