Analyses of the regulatory T (TregEur. J. Immunol. 2015Immunol. . 45: 1524Immunol. -1534 Molecular immunology 1525 been described so far: natural Treg cells, which arise in the thymus [7], renamed tTreg cells [8], and induced Treg cells, which differentiate from naïve CD4 + precursors in the periphery [9,10].Recently, the wide range of Treg-cell functional properties has been associated with phenotypical heterogeneity, such as their homing [6] and their activation status, that defines their anatomical location and homeostasis [11][12][13][14]. How Treg-cell antigen specificity relates to these functions remains poorly understood. Indeed, while Treg-cell activation is largely antigen-specific, the nature of their recognized antigens is still unclear. Theoretically, protecting normal tissues with Treg cells could be handled by a very restricted set of Treg cells specifically recognizing ubiquitously expressed self-antigens. However, many studies concluded that the Treg TCR repertoire is as complex as the effector T (Teff) cell repertoire [15][16][17][18], suggesting that the set of self-antigens recognized by Treg cells is quite complex and may be different at distinct anatomical location [11,19]. Deciphering the nature of the Treg-cell repertoire is of utmost importance for the understanding of their biology.Indeed, it is believed that during thymocyte differentiation, the affinity of the interaction between the TCRs and antigens presented onto thymic antigen-presenting cells (APCs) determines the fate of each cell [20]. The current paradigm is that of a mostly instructive process in which signaling from high-affinity TCRs, likely self-antigen specific, leads to negative selection except for a fraction of cells that will be selected to become Treg cells [21]. Hierarchical clustering of perturbation scores (Fig. 2B, top) discriminates Teff samples from nTreg-cell and amTreg-cell samples. Treg-cell subsets are dispersed into two clusters (I and II), while all but two Teff-cell samples are found in cluster III. Principal component analysis (PCA) projection (Fig. 2B, bottom) depicts this overlap between amTreg-cell and nTreg-cell repertoires.We then focused on amTreg-cell and nTreg-cell samples only and used their perturbation scores against Teff-cell samples to perform a hierarchical clustering according to their LNs localization. Hierarchical clustering leads to two clusters (I and II) (Fig. 2C, top) discriminating deep LNs samples in cluster I, while cluster II mainly gathers superficial LNs samples. Noteworthy, the first component of the PCA projection (PC1), which captures 43% of variability, mainly correlates with sample localization (Fig. 2C, bottom), indicating an increased gradation of perturbation from superficial LN nTreg cells to deep LN amTreg cells.Altogether, hierarchical clustering and PCA separated the different population repertoires, with amTreg cells repertoire being the more perturbed and more distant from that of Teff cells. Furthermore, the PCA projection also showed a rather high interindivi...