Glimpse of natural selection of long-lived T-cell clones in healthy life

Significance

A healthy life requires T cells to provide immunity against infections while maintaining immune tolerance to self and commensal antigens. The diversity of T-cell clones evolves in an individual’s life due to competition between preexisting clones and clones continuously generated from the thymus. It is not known whether antigen-experienced T-cell clones are randomly replaced by new clones or selectively retained for long-term keeping in healthy living conditions. Here, we tracked long-lived T-cell clones in mice without infections or immune challenges. Our study revealed that healthy living conditions select regulatory T-cell clones that may be necessary to maintain the immune-tolerant status of long-lived T-cell clones against self or commensal antigens shared across different mice.

Abstract

Homeostatic maintenance of T cells with broad clonal diversity is influenced by both continuing output of young T cells from the thymus and ongoing turnover of preexisting clones in the periphery. In the absence of infection, self and commensal antigens are thought to play important roles in selection and homeostatic maintenance of the T-cell pool. Most naïve T cells are short-lived due to lack of antigen encounter, whereas antigen-experienced T cells may survive and persist as long-lived clones. Thus far, little is known about the homeostasis, antigenic specificity, and clonal diversity of long-lived T-cell clones in peripheral lymphoid organs under healthy living conditions. To identify long-lived T-cell clones in mice, we designed a lineage-tracing method to label a wave of T cells produced in the thymus of young mice. After aging the mice for 1.5 y, we found that lineage-tracked T cells consisted of primarily memory-like T cells and T regulatory cells. T-cell receptor repertoire analysis revealed that the lineage-tracked CD4 memory-like T cells and T regulatory cells exhibited age-dependent enrichment of shared clonotypes. Furthermore, these shared clonotypes were found across different mice maintained in the same housing condition. These findings suggest that nonrandom and shared antigens are involved in controlling selection, retention, and immune tolerance of long-lived T-cell clones under healthy living conditions.

The broad antigenic repertoire of T cells in peripheral lymphoid organs is subject to constant remodeling by many factors, such as continuing production of young naïve T cells from the thymus, gradual turnover of naïve T cells due to lack of antigen encounters, clonal expansion and contraction in response to immunogens, and differential retention of memory T cells. Under healthy living conditions, T-cell homeostasis is achieved to render immune tolerance while maintaining a diverse repertoire for effective adaptive immunity. Most naïve T cells have limited lifespan without encountering antigens in the periphery (1, 2). Approximately 10–20% of peripheral T cells in healthy living mice are effector or memory T cells, which are antigen experienced and have much longer lifespan than naïve T cells (3). Although the decay rate for antigen experienced T cells in response to immune challenge with model antigens has been well defined, the natural turnover rate of effector or memory like T cells that arise in healthy mice without immune challenges is less understood.

Recent studies in both humans and mice have indicated that a significant fraction of circulating T cells retain the ability to recognize peripherally expressed self-antigens (4, 5). These T-cell clones have escaped negative selection in the thymus due to lack or insufficient expression of the corresponding peripheral antigens in the thymic epithelial cells. These self-reactive T cells are immune tolerant in the periphery due to the presence of T regulatory cells (Tregs) that recognize the same peripheral antigens (5). It has been proposed that Treg-mediated peripheral tolerance helps preserve a broader spectrum of immune repertoire, including some self-reactive T cells that would have been otherwise depleted by negative selection, against microbial antigens (6). This model predicts that Treg-mediated suppression of potentially self-reactive T-cell clones must persist throughout life to prevent development of autoimmunity. Indeed, several studies have demonstrated that the T-cell receptor (TCR) is continuously required for immune regulatory function of Tregs in adult mice (7, 8). However, the clonal stability and long-term homeostasis of Tregs involved in life-long immune tolerance of self-reactive T cells has not been examined.

Most antigen experienced T cells present in healthy living conditions without infections and immune challenges are thought to be against self or commensal antigens (3). It remains to be determined whether antigen experienced T cells and Tregs generated under homeostatic conditions are either selectively maintained or frequently replaced by new T-cell clones from the thymus. In this study, we used a newly established tool to examine T-cell clones that have been naturally aged in mice without any immune challenge for 1.5 y. Our study revealed the presence of long-lived T-cell clones in both the non-Treg and Treg compartments. Furthermore, these long-lived T-cell clones in aged mice displayed an enrichment of shared clonotypes between the non-Treg and Treg clones and across different animals, suggesting that common antigens are involved in selecting their long-term retention.

Results

A Method for Tracking a Single Wave of αβ T Cells from Thymus.

To allow long-term tracking of T-cell clones generated at a given time in the thymus without relying on cell transfers, we designed a lineage-marking system that couples activation of a genetic marker with the birth of αβ T cells. The design is based on the recent established TCRδ^CreER strain (Fig. 1A) (9). The Tcrd gene is actively transcribed in all developing T cells (10) and subsequently deleted on completion of TCRα gene rearrangement during the transition from the double positive (DP) stage to the single positive (SP) stage during thymopoiesis (11). Consequently, the Cre-ER cassette in Tcrd^CreER mice is present in DP and absent in SP and later stages of αβ T-cell development. Furthermore, the Cre-ER fusion protein is inactive in developing T cells until mice are treated with tamoxifen, which allows Cre activity to be regulated in a temporal manner in addition to lineage and stage restrictions. We crossed Tcrd^CreER mice with the R26^ZsGreen reporter strain, in which ZsGreen expression provides a permanent genetic label on tamoxifen-induced Cre/lox recombination (12). We tested the efficiency of this marking system by injecting a single dose of tamoxifen to young adult mice at 1 mo of age, followed by examination of developing T cells in the thymus and T cells in the periphery at different time points (Fig. 1 B–D). One week after tamoxifen treatment, ZsGreen labeling was primarily detected in the DP and SP cells in the thymus (Fig. 1C). DP labels disappeared after 2 wk, whereas SP labels returned to the background level after 3 wk. Examination of CD4 T cells in the periphery also revealed a wave of labeled T cells, which peaked at 2 wk after tamoxifen injection, followed by a sharp decline to less than half of the peak value in the following week and gradual decline thereafter (Fig. 1C). The observed speed of thymic egress was consistent with previous observations based on BrdU pulse chasing (13). Analysis of other peripheral lymphoid organs showed similar kinetics of labeled CD4 T cells as that observed in the spleen. However, the magnitude of peak labeling varies at different sites (Fig. 1D). The difference in peak labeling indicates that Peyer’s patch and bone marrow have fewer recent thymic emigrants (RTEs) in comparison with the spleen and lymph nodes.

Fig. 1.

Generation and decay of age-tracked T cells in young adult mice. (A) Schematic illustration of the Tcrd^CreER knockin allele and its application in activation of the R26 reporter. (B) Scheme of tamoxifen treatment and age-tracking of T cells. (C) Representative flow cytometry analysis of ZsGreen-labeled T-cell fractions from day 7 to day 60 after single-dose tamoxifen. (D) Time course tracking of labeled CD4 T cells in thymus, spleen, peripheral lymph nodes (LNs), mesentery lymph nodes (mLNs), Peyer’s patch (PP), and bone marrow (BM) between 1 wk and 60 d after tamoxifen injection. (E) Time course tracking of Tregs as in D. (F) Bar graph summary of Treg frequency among labeled CD4 T cells at different time points. Each data point is an average with SD of three to six mice for graphs presented in D–F.

Tregs and Tconvs Show Different Kinetics in Development.

A wave of ZsGreen-labeled CD25⁺FoxP3⁺ Tregs was also detected in the thymus and periphery (Fig. 1E and Fig. S1A). The kinetics of labeled Tregs was different in several ways from that of total CD4 T cells, most of which are conventional CD4 T cells (Tconvs). First, the peak of the labeled Tregs in the thymus was approximately 1 wk behind the wave of labeled CD4 SP cells, which is consistent with the notion that Treg development takes place after positive selection. Labeled Tregs reached the peak frequency at the same time in the thymus and spleen, indicating that newly generated Tregs egress without any delay. Second, the peak frequency of labeled Tregs in the periphery was less than one-half of that observed for CD4 T cells (Fig. 1E). This observation suggests that the peripheral Treg compartment is replaced by RTE at a lower rate than that of Tconvs. Third, a small fraction of labeled Tregs was found in the thymus along with age-tracked CD44^hiCD62L⁻ effector memory-like CD4 T cells (Fig. S1B) after 2 mo. This phenomenon is consistent with the idea that some of the thymic resident Tregs have gone through a life outside the thymus before taking residence in the thymus (14). Finally, the relative frequency of Tregs among labeled CD4 T cells gradually increased with time, starting from 1% to 3% at 1 wk in all peripheral tissues to ∼10% in the spleen and lymph nodes and 30% in the thymus and bone marrow 2 mo after marking (Fig. 1F). This result indicates that newly generated Tconvs and Tregs follow different survival or proliferative kinetics during the 2-mo period before reaching to homeostasis.

Fig. S1.

Kinetics of age-tracked Tregs within the first 2 mo after labeling. Four-week-old Tcrd^CreERR26^ZsGreen mice were treated with single-dose tamoxifen followed by FACS analysis at indicated time points. (A) Analysis of ZsGreen-labeled Treg cells in the thymus (Upper) and spleen (Lower). Representative dot plots display Foxp3 and ZsGreen expression among gated CD4⁺CD8⁻TCRβ⁺ cells. Each FACS plot represents at least three independent mice. (B) Frequency of lineage-tracked Tregs and effector memory CD4⁺ T cells in the thymus. (Top) Frequency of ZsGreen-positive cells among CD4SP cells. Plots were pregated for TCRβ⁺CD4⁺ T cells in the thymus. (Middle) Quantification of Treg cells with CD25 and Foxp3 (upper right quadrant) staining of the lineage-tracked cells (gated in the top panel). (Bottom) Memory-like phenotype reviewed by CD44 and CD62L staining of lineage-tracked TCRβ⁺CD4⁺ T cells (gated in the top panel). Relative percentages of naïve (CD44^loCD62L^hi) and effector memory (CD44^hiCD62L⁻) are indicated in the plot. Each FACS plot represents analysis of two to three independent mice.

Eight-Week Tracking Reveals Clonal Selection of Lineage-Labeled T Cells.

To gain a better understanding of the clonal diversity and stability of T-cell clones generated during a single administration of tamoxifen, we performed TCRβ repertoire analysis of labeled T cells in the spleen at either 2 or 8 wk after the treatment (Datasets S1 and S2). The complementarity determining region 3 of the TCRβ chain (CDR3β) has been shown to dictate the specificity and affinity of antigen recognition by making major contacts with the peptide antigen presented by the MHCII (15, 16). Therefore, we used CDR3β clonotype as an approximation of antigenic specificity of the TCR. Tconvs and Tregs were separated into either unlabeled or ZsGreen-labeled fractions. Pairwise distance calculation of TCRβ nucleotide sequences (SI Materials and Methods) (17, 18) among all samples collected at 2 wk did not reveal any strong patterns except an overall weak similarity between samples from the same animals (Fig. 2A, Left). The same unbiased analysis of 8-wk samples revealed a prominent cluster containing age-tracked Tregs (TRGP) from three mice, although the same cluster also contained age-tracked Tconvs (TCGP) from one of the three mice (Fig. 2A, Right). We then examined the similarity between labeled and nonlabeled cells within the same animals (Fig. 2B). The labeled populations became less similar to the corresponding nonlabeled populations with increasing tracking time, suggesting an age-dependent repertoire shifting among lineage-tracked cells. Indeed both labeled Tconvs and Tregs were enriched with high-frequency clones in comparison with unlabeled population at 8 wk after lineage tracking in all three mice analyzed (Fig. 2C and Fig. S2). Therefore, Tconvs and Tregs generated in young adult mice had already preferentially expanded after surviving 8 wk. These age-tracked Tregs were phenotypically indistinguishable from nontracked total Tregs in terms of CD25 and FoxP3 expression (Fig. S3). We further examined the similarity across different animals for age-tracked T cells at 2 vs. 8 wk after labeling (Fig. 2D). Intermouse similarity was increased for both Tconvs and Tregs with increasing age, although the overall intermouse similarity for Tregs is higher than that of Tconvs. This result indicated that common antigens shared across different mice may be involved in selective retention of certain Tconv and Treg clones.

Fig. 2.

TCRβ repertoire analysis of age-tracked T cells in young adult mice. (A) Heat map clustering of T-cell subsets from either 2- or 8-wk age-tracking experiments. Pairwise distances were calculated based on Bhattacharyya similarity index for CDR3β nucleotide (NT) sequences. TCGN, Tconv ZsGreen negative; TCGP, Tconv ZsGreen positive; TRGN, Treg ZsGreen negative; TRGP, Treg ZsGreen positive. (B) Similarity comparison between age-tracked T-cells and their nontracked counterparts for both Tconvs and Tregs. (C) Cumulated proportion of the 100 most abundant CDR3β clones in each compartment for three 8-wk age-tracked mice. Open, remaining rare clones; solid, sum of 100 dominant clones. (D) Age-dependent increase of intermouse similarity. Left, Tconvs; Right, Tregs. Bars, mean value. Statistics were based on two-tailed Student t test.

Fig. S2.

Repertoire diversity was compared using the Simpson index for each T-cell subset of three 8-wk lineage-tracked mice. AA, amino acid sequence; NT, nucleotide sequence; TCGN, Tconv ZsGreen negative; TCGP, Tconv ZsGreen positive; TRGN, Treg ZsGreen negative; TRGP, Treg ZsGreen positive. Horizontal bar, mean diversity; error bar, SD. Statistics are based on two-tailed Student t test.

Fig. S3.

Foxp3 and CD25 expression in age-tracked Tregs. CD4 T lymphocytes from spleen, LN, mLN, and PP were analyzed 60 d after tamoxifen treatment of 4-wk-old mice. (Top) Frequency of Foxp3 positive Treg cells in lineage-positive (ZsGreen⁺) and -negative (ZsGreen⁻) CD4 T cells. (Bottom) CD25 expression in lineage-positive (green line) and -negative (gray line) Treg cells defined by Foxp3 staining in the top panel. Each FACS plot or histogram represents three independent mice.

Tracking for 1.5 y Identifies Long-Lived T-Cell Clones.

We next used the lineage-tracking tool to evaluate the age composition of T-cell clones in aged mice. To distinguish between truly aged T-cell clones and newly generated T cells in aged animals, we treated 1-mo-old mice or 1.5-y-old mice with tamoxifen before analyzing the labeled T cells at different time points (Fig. 3A). Short-term tracking showed that the thymus of 1.5-y-old mice produced fewer T cells than the thymus of young adult mice, although the relative frequencies of naïve T cells and Tregs were similar between the two age groups (Fig. 3 B–G). Long-term tracking also detected long-lived T-cell clones in all peripheral lymphoid organs examined including spleen, lymph nodes, Peyer’s patch, and bone marrow (Fig. 3 B and C). The long-lived CD4 T-cell clones were mostly memory-like T cells based on up-regulation of CD44 and down-regulation of CD62L (Fig. 3 D and E). Importantly, the Treg ratio among labeled CD4⁺ T cells rose significantly after long-term tracking in comparison with the short-term tracking in either young or aged mice (Fig. 3 F and G). The long-term labeled Tregs and CD4 Tconvs can be further distinguished from each other based on differential expression of cytokines (Fig. S4).

Fig. 3.

Generation and retention of T cells in aged mice. (A) Schematic illustration of three experiments to assess thymus output from young mice (experiment 1), thymus output from old mice (experiment 2), and long-term retention of aged T cells (experiment 3). (B) Representative flow cytometry analysis of age-tracked (ZsGreen⁺) cells among splenic CD4 T cells. (C) Bar graph summary of CD4⁺ZsGreen⁺ cell frequency as shown in B for various lymphoid organs. (D) Representative flow cytometry analysis of age-tracked CD4 T cells with activation markers to distinguish naïve (CD44^loCD62L⁺), effector memory (CD44^hiCD62L^lo), and central memory (CD44^hiCD62L⁺) cells. Samples were pregated on CD4⁺ZsGreen⁺ cells as depicted in B. (E) Bar graph summary of total memory-like cell frequency (CD44^hi) in various lymphoid organs (abbreviations as in Fig. 1D). (F) Representative FACS analysis of Tregs with Foxp3 and CD4 staining of TCRβ⁺ZsGreen⁺ lymphocytes from lymph nodes. (G) Bar graph summary of Treg frequency among age-tracked CD4 T cells. All bar graphs were based on the averages of three mice in each experiment.

Fig. S4.

Foxp3 and cytokine expression in age-tracked Treg (CD4⁺CD25⁺) and Tconv (CD⁺CD25⁻) cells. (A) CD25 and Foxp3 expression in Treg and Tconv cells was examined at 1.5 y after tamoxifen treatment of 4-wk-old mice. (B) IFNγ and IL-17 expression in Tconv and Treg cells of the mice as in A. Each FACS plot represents three independent mice.

Shared Antigens Drive Selection of Long-Lived T-Cell Clones.

To further elucidate the clonal diversity of these long-lived T-cell clones in aged mice, we analyzed the TCRβ repertoire of six independent mice aged between 16 and 18 mo after tamoxifen treatment (Dataset S1). Both age-tracked Tconvs and Tregs showed increased intermouse similarities in comparison with nonlabeled cells (Fig. 4A). Intermouse sharing of CDR3 peptide sequence was always higher than that of the nucleotide sequence, indicating a selection pressure on common peptide sequences (Fig. S5). Age-tracked Tconvs and Tregs within each individual mouse also showed increased similarities in comparison with the nonlabeled cells (Fig. 4B). TCRβ repertoires in Peyer’s patch showed higher inter- and intramouse similarities compared with those in the spleen. To further examine the phenomenon of shared clonotypes between different mice, we separated clonotype from clonal frequency by calculating a sharing index based on overlapping clonotypes between paired samples (Fig. 4C). This analysis confirmed that clonotype sharing is a unique phenomenon in age-tracked Tregs or Tconvs in aged mice. Among age-tracked T cells in younger mice, only 8-wk-old Tregs showed a significant enrichment of shared clonotypes across three mice analyzed (Fig. 4C). This result indicated that clonotype sharing in Tregs and (or) Tconvs in aged mice is a result of selection of dominant clonotypes carrying the fittest TCRs that recognize common antigens.

Fig. 4.

CDR3β repertoire analysis of long-lived T cells in aged mice. (A and B) Similarity analysis of CDR3β repertoires between mice (A) or within each mouse (B) of specified T-cell subsets isolated from the spleen and Peyer’s patch of six aged mice. Horizontal bar, mean similarity; error bar, SEM. (C) Clonotype analysis. Clonotype sharing is defined as the proportion of shared nucleotide sequences between two mice. The percent of intersection was calculated for each mouse (Left). Clonotype sharing was evaluated for age-tracked T cells obtained at three different time points: 2 wk (three mice), 8 wk (three mice), and 1.5 y (six mice) after tamoxifen treatment. Bar, mean proportion, error bar, SD. (D) List of shared 25 CDR3β sequences and clonal frequencies of age-tracked Treg from 1.5-y tracking and 2-wk tracking experiments. NT coding is average per mouse. (E) The average frequencies of individual CDR3β clonotypes (listed in D) in different T-cell subsets in young and aged mice. Each dot represents the average frequency of one particular CDR3β clonotype in the specified cell type among different mice of the specified age group. Bars, mean frequencies. Statistics were based on two-tailed Student t test. (F) Age-dependent enrichment of common CDR3β clonotypes among lineage-tracked Tregs. (Left) Cumulative clonal frequency of the shared TCRβ sequences. Among the 25 shared CDR3β clonotypes, 15 were identified as truly shared TCRβ clones based on the analysis of nucleotide sequences (NT) among all six age-tracked mice. Each dot represents the cumulative clonal frequencies of these 15 TCRβ clones among age-tracked Tregs in a single mouse. Three age groups were included in this analysis. (Right) Cumulative frequency of the 25 common CDR3β clonotypes (AA as amino acid sequences). Each dot represents the cumulative frequency of 25 clonotypes among age-tracked Tregs in a single mouse. Statistical analysis was performed between different age groups. error bar, SEM.

Fig. S5.

Evaluation of intermouse repertoire similarity. Dot plot depicts Bhattachayya pairwise analysis among six mice that have been lineage-tracked for 1.5 y. Each cell type is grouped together with each dot representing one sample pair. Similarity index for TCRβ NT clones and CDR3 AA clones for each paired sample is compared with a connecting line. Nomenclature as in Fig. S2. (Left) Peyer's patch. (Right) Spleen. Statistics are based on two-tailed paired Student t test.

Age-Dependent Enrichment of Shared Clonotypes Between Tconvs and Tregs.

Clonotype analysis identified 25 unique CDR3β sequences shared by age-tracked Tregs from all six aged mice (Fig. 4D). Each CDR3 sequence was encoded by an average of 5, ranging from 3 to 12, unique nucleotide sequences (NT). In addition, these CDR3 loops were presented by a broad diversity of Vβ and Jβ regions, indicating its diversified clonal origin (Dataset S3). Because the CDR3 loop is the major structural features recognizing antigen peptide in the MHC groove (15, 16), we reasoned that the best explanation for accumulation of the high-frequency and coding-degenerate CDR3β clonotypes is common antigen-driven selection and retention. Consistent with this interpretation, we found that most of this CDR3β sequences can be found in 2-wk tracked Tregs, but at a lower frequency and reduced coding degeneracy (Fig. 4 D and E). To evaluate selection of these 25 clonotypes at different age, we compared the aggregate frequency among labeled Tregs at 2 wk, 8 wk, or 1.5 y (Fig. 4F). This analysis revealed a trend of enrichment of these common CDR3β by age, suggesting that selection of these clonotypes occurred gradually with age. Finally, the age-dependent enrichment of these 25 clonal types were also observed in the nonlabeled Tconvs and Tregs (Fig. 4E), confirming that the selection pressure applies not only to the single wave of labeled T cells but also to the general T-cell population.

SI Materials and Methods

Sequencing Read Adjustment.

Adjustment is triggered when numbers of unique NT clones ≥ numbers of input cells. Adjustment is based on two assumptions: (i) real clones exhibit higher abundance than sequencing errors after deep sequencing; and (ii) each cell contributes to equal mRNA transcripts to the sequencing pool. Raw sequencing reads were adjusted based on the following workflow: step 1, record the number of input cells as Ni and extract the most abundant clones down to the Ni-th (based on assumption 1); step 2, calculate the unity abundance as 1/Ni*100% (based on assumption 2); step 3, for each extracted clone, divide the original frequency by unity to obtain the adjusted frequency; step 4, start from loop = first, pick the loop abundant clone (the adjusted frequency) into real list; step 5, calculate the sum abundance in real list and stop the loop when the total proportion ≥100%; step 6, finalize the real list as the adjusted NT clones with adjusted frequencies.

Computational Methods and Statistical Analysis.

Repertoire diversity based on an RNA template was evaluated as previously described. Simpson diversity index, which considers both the number of species and the relative abundance of each species, was calculated as

$$\text{Simpson}\hspace{0.17em}\text{diversity}\hspace{0.17em}\text{index}={\log }_{10}\left[1-{\displaystyle \sum }_{i=1}^c\frac{n_i\left(n_i-1\right)}{n\left(n-1\right)}\right],$$

where n is the total number of clones in its repertoire, c is the number of unique sampling clones, and n_i is the abundance of sampling clones in its repertoire (17). Formula was scaled by log10 to facilitate the influence of sampling size. Diversity index is a statistics variable derived from ecologic studies to estimate the biodiversity of a given environment based on limited sampling size (counting the absolute number of organism inside is somehow impossible). Similarly, because the measuring for every T-cell clone in whole lymphoid organs is impracticable, we introduced this formulation to evaluate the diversity of interest repertoire. More uneven clonal frequencies result in a smaller corresponding index. If all clonotypes in the repertoire of interest have equal distribution, the index takes the unity value. To quantify the similarity between each TCR repertoire, we adopted the Bhattacharyya similarity index based on the percentage and homogeneity of abundance of shared sequences within two populations (18) as follows:

$$\text{Bhattacharyya}={\displaystyle \sum }_{i=1}^n\sqrt{f\left(i,1\right)\times f\left(i,2\right)},$$

where n is the number overlap clones, and f(i,1/2) is the abundance of overlapped clones in two repertoires. A value ranging from 0 (no overlap) to 1 (identical repertoire) was calculated for each pair of clonotypes in the repertoire. Differences between two groups were determined by the Mann-Whitney u test. MATLAB software was used for all statistical analysis. All data were analyzed using two-tailed tests, and P < 0.05 was considered statistically significant unless otherwise specified.

Discussion

In this study, we showed that long-lived T-cell clones arising from long-term homeostasis exhibited restricted clonotypes. These clonotypes are defined by the enrichment of a set of CDR3β sequences shared among aged-tracked T cells across different animals housed in the same SPF facility. Each of these restricted CDR3β clonotypes is associated with multiple unique Vβ and Jβ sequences (Dataset S3), indicating a convergent selection on CDR3β. However, each of these CDR3β clonotypes may still recognize more than one antigen because the TCRα chain also contributes to antigen binding specificity. Therefore, more than one antigen may be involved in the selection of each CDR3β clonotype that has been enriched among age-tracked T cells.

Although the source of the antigens cannot be determined from the current study, the enrichment of these antigen-specific Tconvs and Tregs in the Peyer’s patch indicated the possibility of gut-associated self or commensal antigens. We cannot exclude the possibility that the long-lived Tconvs and Tregs found in Peyer’s patches were originally derived from other anatomic locations because many of the shared clonotypes were also found among age-tracked T cells isolated from the spleen. In considering the source of the antigens, two important facts are worth pointing out: first, individual clones of these shared and lineage-tracked clonotypes invariably existed in low frequencies at the population level, indicating that antigens involved in their selection did not drive vigorous clonal expansion. Second, these clonal types were not unique to the lineage-tracked populations because they were also detected among nonlabeled Tconvs and Tregs. The gradual accumulation of these clonal types in nonlabeled Tconvs and Tregs further indicates that these clonal types are continuously been selected out from newly generated T cells. Thus, persistent antigens, either self or commensal origin, are most likely involved in driving the homeostatic retention of these long-lived T-cell clones.

The lineage relationship between long-lived Tconv and Treg clones remains to be further defined. The long-lived Treg clones may be descendents of thymic-derived Tregs or peripherally converted from Tconvs in response to gut antigens. Although our lineage-tracing method cannot distinguish between these two possibilities, results from 2-wk tracking provided some clues. Most of the shared clonotypes identified from truly aged T-cell populations can be found in the 2-wk tracking samples, albeit at a lower frequency. In fact, the clonal frequency of these unique clonotypes is higher in 2-wk labeled Tregs than 2-wk labeled Tconvs. Our kinetic analysis indicated that 2-wk tracking only captures newly generated Tconvs and Tregs in the spleen. Therefore, clonotype sharing between Tconvs and Tregs may occur as early as they are first generated in the thymus. This is consistent with a recent study indicating that thymus-derived Tregs are the major source of clonal diversity of the Treg population present in the gut (19). However, we cannot exclude the possibility that lineage conversion may also contribute to clonotype sharing among truly aged Tconvs and Tregs. It has been shown that commensal antigens and commensal metabolites can induce lineage conversion from Tconvs to Tregs and thus shaping the Treg compartment in the gut (20–22). A recent study also reported that inflammation plays an important role in shaping the Treg population in the periphery (23). We cannot rule out the possibility that undetectable and subclinical inflammation may also contribute to the accumulation of long-lived Tregs. It remains to be tested whether the clonotypes identified in our experiment are subject to perturbation in germfree or different housing conditions.

The robust enrichment of a unique group of TCR clonotypes among age-tracked Tregs suggests that these Tregs are actively involved in maintaining lifelong immune tolerance. This hypothesis is consistent with the recent finding that TCR is continuously required for Treg function after lineage commitment (7, 8). The high frequency of clonotype match between long-lived Tconv and Treg clones further suggests that antigen-specific immune suppression plays a major role during lifelong T-cell homeostasis. Identification of the antigens involved in the generation and maintenance of long-lived Tregs may help understand how lifelong immune tolerance is achieved in healthy conditions and perturbed in cancer or autoimmune disease situations. The unique TCR clonotypes identified from the age-tracked T cells provide crucial leads for further investigating the behavior and function of T-cell clones involved in lifelong immune tolerance.

Materials and Methods

Mice.

The Tcrd^CreER knock-in allele was generated and maintained on 129/sv and Bl/6 background (9). The R26^ZsGreen strain was purchased from the Jackson Laboratory. All mouse strains were bred and maintained in the SPF facility managed by Duke University Division of Laboratory Animal Research. Animal procedures were reviewed and approved by Duke Institutional Animal Care and Use Committee.

Tamoxifen Treatment.

For long-term tracing experiments, Tcrd^CreERR26^ZsGreen mice (4 wk old) were treated with three doses of 1 mg tamoxifen i.p. every other day. For kinetic analysis and short-term tracing experiment, Tcrd^CreERR26^ZsGreen mice (4 wk old) were treated with one dose of tamoxifen (1 mg/mouse).

Flow Cytometry Analysis and Cell Sorting.

Lymphocytes from indicated organs of tamoxifen treated Tcrd^CreERR26^ZsGreen mice were isolated and stained with anti-CD4, CD25, CD8, CD44, and CD62L antibodies. Foxp3 staining was performed according to intracellular staining procedure from eBioscience. The analysis was performed on BD FACSCanto II analyzer. Cells from spleen and PP were stained by anti-TCRβ, CD4, and CD25 antibodies, and ZsGreen⁺ conventional (TCRβ⁺CD4⁺CD25⁻) and regulatory (TCRβ⁺CD4⁺CD25⁺) T cells were sorted by MoFlo XDP sorter. FACS data were analyzed with the Flowjo software.

Ex Vivo Cytokine Analysis of Treg and Tconv Cells.

Lymphocytes from spleen and PP were stained by anti-CD4, CD25 and TCRβ antibodies. ZsGreen⁺CD25⁺ and ZsGreen⁺CD25⁻ cells were sorted by FACS and stimulated by phorbol 12-myristate 13-acetate and ionomycin with brefeldin A and monensin for 4 h in vitro. Cell suspensions were fixed and permeablized according to Foxp3 kit and stained with Foxp3, IFNγ, and IL-17 antibodies. The IFNγ and IL-17 expression in Foxp3⁺ and Foxp3⁻ cells was analyzed on BD FACSCanto II analyzer.

High-Throughput TCRβ CDR3 Sequencing.

Total RNA was extracted from trizol-lysed samples using the RNeasy Mini Kit (QIAGEN) and converted to cDNA (RevertAid First Strand cDNA Synthesis Kit; Fermentas) with a constant region-specific primer (RT primer: 5′-ATCTCTGCTTCTGATGGCTCA-3′). A multiplex PCR system was introduced to amplify the CDR3 region of rearranged TCRB loci. A set of forward primers, each specific to one or a set of functional TCR Vβ segments, and a reverse primer specific to the constant region of TCRB (Dataset S2) were used to generate amplicons that cover the entire CDR3-region. Purified products were sequenced using the Ion Torrent PGM platform (Life Technologies).