The human adaptive immune system employs a vast number (> 10¹¹ [Clark et al., 1999]) of T lymphocytes, to detect and control pathogens. Most T cells express a single T-cell receptor (TCR) variant, which binds antigen in the form of a short peptide presented by the Major Histocompatibility Complex (pMHC) (Davis and Bjorkman, 1988). The TCR has to be specific to distinguish between self- and non-self-pMHC, but due to the large number of possible foreign antigens (> 20⁹) a specific TCR is nevertheless expected to bind many different pMHC (i.e., cross-reactivity) (Mason, 1998; Sewell, 2012). The actual diversity of the TCR repertoire is unknown, but with improved sequencing techniques, estimates have risen by orders of magnitude from 10⁶ (Arstila et al., 1999), 10⁷ (Robins et al., 2009), to over 10⁸ (Qi et al., 2014).

Generation of αβ TCRs occurs in the thymus, where thymocytes randomly rearrange and imprecisely recombine gene segments to create a complete receptor (Nikolich-Zugich et al., 2004). This heterodimer is generated by random recombination of Variable, Diversity, and Joining (V, D and J) segments for TCRβ, and V and J segments for TCRα sequences (Davis and Bjorkman, 1988). Most variability arises due to random nucleotide insertions and deletions where the segments are joined (Murugan et al., 2012). Recent estimates of the potential number of TCRs produced by this V(D)J-recombination process range from > 10²⁰ (Zarnitsyna et al., 2013) to 10⁶¹ (Mora and Walczak, 2019), which vastly outnumbers the number of distinct TCRs present in a human body. After generation of the TCR, T cells undergo positive and negative selection, which selects those T cells that have sufficient, but not too high, affinity for any self-pMHC (McDonald et al., 2015). About 3–5% of thymocytes survive selection (Merkenschlager et al., 1997) and enter the periphery as T cells that have not yet encountered foreign cognate antigen, that is as naive T cells.

The thymic output of new T cells decreases because of thymic involution, making peripheral division of existing cells the main source of naive T cells from early adulthood onwards in humans (den Braber et al., 2012; Kumar et al., 2018). In the periphery, naive T cells compete for cytokines, such as IL-7, and need to interact with self-pMHC to survive (Tanchot et al., 1997; Takada and Jameson, 2009; Jenkins et al., 2010). Competition between T-cell specificities may reduce repertoire diversity when cells with some TCRs outcompete others (De Boer and Perelson, 1994), resulting in differences in TCR frequencies, and heterogeneous naive T-cell clone sizes. Experimental evidence for large heterogeneity in division and survival rates within the naive T-cell pool has been shown in mice (Hogan et al., 2015; Rane et al., 2018; Reynaldi et al., 2019). Such experiments are not feasible in humans, but mathematical modelling has been used to assess how fitness differences between T-cell clones may affect the frequency of clones in the naive repertoire (Stirk et al., 2008; Stirk et al., 2010; Hapuarachchi et al., 2013; Lythe et al., 2016; Desponds et al., 2016; Desponds et al., 2017; Dowling and Hodgkin, 2009; Johnson et al., 2012).

Measuring the distribution of TCRα and TCRβ sequences in samples of naive T cells can inform us about the clone-size distribution of the naive T-cell repertoire. Previous studies have reported large heterogeneity in the frequency of TCRβ sequences in naive repertoires from mice (Quigley et al., 2010) and humans (Robins et al., 2009; Venturi et al., 2011; Qi et al., 2014; Pogorelyy et al., 2017). One important factor shaping the abundance of TCR sequences is their likelihood to be produced during VDJ-recombination. Rearrangements with less N-insertions, for example, tend to be more commonly observed (Robins et al., 2009; Robins et al., 2010; Venturi et al., 2011; Pogorelyy et al., 2017). To study this in more detail, the Mora and Walczak groups developed probabilistic models that predict the generation probability of any specific TCRα or TCRβ sequence (Murugan et al., 2012; Marcou et al., 2018). They showed that these sequences ($\sigma$) differ by several orders of magnitude in their probability $𝒫(\sigma )$ of being produced by V(D)J recombination in the thymus. Differential generation probabilities do not only impact the abundance of TCRα and TCRβ sequences within an individual, but also contribute to sharing among individuals (Robins et al., 2010; Quigley et al., 2010; Venturi et al., 2011; Qi et al., 2014; Pogorelyy et al., 2017; Elhanati et al., 2018). Hence, it is essential to take the likelihood of generating a sequence into account when interpreting sequencing data of immune repertoires.

In this study, we characterize the frequency distribution of TCRα and TCRβ sequences in the naive repertoire. We analyze published and new experimental data on both the TCR α and β chain, and combine a quantitative unique molecular identifier (UMI)-based TCR sequencing pipeline with mathematical modeling to consider carefully the contributions of different mechanisms that may lead to observed abundant TCRα and TCRβ sequences in the naive repertoire. Such mechanisms include experimental confounders, such as the purity of the cell populations and repeated sampling of mRNA from the same cell, and diverse biological processes including distinguishing carefully between repeat generation of identical sequences in different cells, and large naive T-cell clones. We show that all these processes are likely to contribute to the observed abundance profile of TCR sequences in samples of naive repertoires. In particular, even after all other mechanisms are accounted for, we find evidence for naive T-cell clone size heterogeneity. Specifically, the results are compatible with an underlying power-law distribution of naive T-cell clone sizes (Desponds et al., 2016), or more generally by models in which 1–5% of naive T cells represent large clones of 10⁵ - 10⁶ cells. Preferential expansion of some clonotypes, perhaps those occurring early in development of adaptive immunity, therefore plays an important role in shaping the naive T-cell repertoire.

We analysed the frequency distribution of TCR sequences in the naive T-cell compartment, using TCRα and TCRβ sequences published in Oakes et al. (2017). In brief, peripheral blood mononuclear cells (PBMCs) from two adult volunteers were FACS-sorted into naive (CD27⁺CD45RA^high) and various memory CD4⁺ and CD8⁺ populations. TCRα and TCRβ mRNA was reverse transcribed to cDNA molecules to which unique molecular identifiers (UMIs) were attached, followed by PCR-amplification and high-throughput sequencing (HTS) on an Illumina MiSeq platform. We refer to this as experiment 1 below (for further details see Section 'Sequence analysis'). Sequence reads were processed using a customized version of the Decombinator pipeline (Thomas et al., 2013), with an improved error correction on UMIs to more reliably estimate the frequency of nucleotide TCRα and TCRβ sequences in the samples (see Section 'Sequence analysis'). Additionally, we used the RTCR pipeline (Gerritsen et al., 2016) for comparison (Section 'Sequence analysis'). The different memory populations were combined for the purpose of the analysis presented below.

Abundant TCR sequences are frequently shared between naive and memory populations, and are enriched for high V(D)J recombination probabilities

Within the naive T-cell repertoires, the vast majority of TCRα and TCRβ sequences were observed only once, and most frequencies fall within the range from 1 to 5 (Figure 1A). As expected, in the memory repertoires, which contain clonally expanded T cells, much more abundant sequences were present, with a substantial number of α and β chains observed more than 1000 times (Figure 1A). The few sequences observed with a frequency higher than five in the naive samples were shared in most cases (94.6%) with the corresponding memory subset from the same individual. We examined whether this overlap might arise from imperfect sorting of the T-cell populations, despite the tight non-overlapping sort gates applied (see [Oakes et al., 2017]). A prediction of such sorting contamination is that the abundance of the shared TCR sequences in the naive and memory repertoires should be proportional. Such a linear relationship could be observed clearly for CD8⁺ TCRα and TCRβ sequences (Figure 1A), especially for memory abundances greater than 1000. Correlation measurements suggested that the amount of contamination for CD8⁺ T cells was 0.1 - 1.5%. As expected, no correlation was observed between the abundance of TCR sequences shared between naive and memory populations of different donors (Figure 1B).

Figure 1 with 2 supplements see all

Download asset Open asset

Figure 1—source data 1

Memory and naive counts in Experiment 1.

Number of TCRα and TCRβ sequences exclusively occurring in naive or memory samples within log2-frequency bins, as in Figure 1D. Bin intervals are excluding the left, but including the right number.

https://cdn.elifesciences.org/articles/49900/elife-49900-fig1-data1-v1.txt

Download elife-49900-fig1-data1-v1.txt

We next examined the relationship between V(D)J recombination probabilities and the overlap between naive and memory repertoires. Using the V(D)J-recombination model of Marcou et al. (2018), we predicted the generation probabilities $𝒫(\sigma )$ of all TCRα and TCRβ sequences in our datasets. As expected, we observed a wide range of $𝒫(\sigma )$ values, which were several orders of magnitude higher for TCRα sequences than TCRβ, due to additional recombination of the D segment. The generation probability distributions of sequences derived from naive and memory T cells were indistinguishable (Figure 1C, blue and red, respectively). Thus, our data provide no evidence that the V(D)J-recombination process preferentially produces sequences that are more likely to enter the memory pool during an immune response. However, TCRα sequences shared between memory and the corresponding naive samples, were strikingly enriched for high $𝒫(\sigma )$ (Figure 1C, green). This enrichment is much less evident for TCRβ sequences. The enrichment for sequences with high $𝒫(\sigma )$ in the population of shared memory/naive TCRα is not compatible with overlap derived from contamination during cell sorting, but rather suggests that the sharing may also arise from T cells which use the same TCRα because of identical VJ recombination events in different T cells. It is important to stress that, since such different T cells are highly unlikely to also share TCRβ sequences, the clonotype, and hence specificity of the T cells in the naive and memory compartments may well be different, despite sharing TCRα sequences.

As a control, we also analyzed overlap between the naive sample from one volunteer and the memory sample from the other. In this case, sort contamination of naive repertoires by memory T cells is excluded and a shared sequence can only result from independent identical recombination events, from distinct T-cell clones. For CD4⁺ cells, we find that the number of TCRα sequences shared between naive and memory is similar between and within volunteers, and that the $𝒫(\sigma )$ distribution is nearly identical (Figure 1C, purple). For CD8⁺ cells, the number of sequences shared within an individual is somewhat larger than between individuals, compatible with some degree of sort contamination in this population as discussed above. The small number of TCRβ sequences shared between individuals also had a relatively high $𝒫(\sigma )$, although considerably smaller than for TCRα.

In summary, although contamination with abundant memory T cells may make a small contribution to the TCR sequences which are found in both naive and memory for CD8⁺ cells, multiple identical recombinations arising from high $𝒫(\sigma )$ values is the dominant mechanism leading to overlap in the TCRα repertoires. Nevertheless, in order to stringently exclude any possible contribution of contamination, we included an analysis which excluded all the shared sequences from the further investigations of the relationship between TCR sequence abundance and $𝒫(\sigma )$ (Figure 1D).

The abundances of sequences in all naive repertoires were correlated to $𝒫(\sigma )$ (Figure 1D, blue). The median $𝒫(\sigma )$ of the α chains that were observed at least three times was about 154-fold higher than for those that have only been observed once ($p<{10}^{-15}$, Wilcoxon test). The enrichment for high $𝒫(\sigma )$ in more abundant TCR sequences was weaker for TCRβ (∼2.5-fold, $p<0.01$, Wilcoxon test), but still stronger than for memory subsets (1.65- and 1.03-fold for TCRα and TCRβ, respectively, $p<{10}^{-15}$ and $p=0.27$). In line with this, the number of N-additions tended to be lower for TCRα and TCRβ sequences abundant in the naive samples (Figure 1—figure supplement 1). These correlations suggest that multiple identical recombination events which occur during formation of the naive T-cell repertoire in the thymus due to high generation probabilities, contribute to the observation of abundant TCR sequences. This is especially evident for TCRα, where the probabilities of producing a given sequence are higher because of the absence of a D region. However, abundant TCR sequences with low $𝒫(\sigma )$ are also observed, especially for TCRβ, leaving open the possibility of large naive T-cell clones.

Frequently observed TCR sequences cannot be attributed only to multiple RNA molecules per cell

T cells contain on average in the order of 100 molecules of TCRα and 300 molecules of TCRβ mRNA (Oakes et al., 2017). Because the TCR sequencing pipeline is not 100% efficient, only a small proportion of these molecules are actually sequenced, but the possibility remains that TCR sequences observed multiple times may be due to repeat sampling from the same cell. Because the variance of this number remains undetermined, it is difficult to computationally determine the contribution of this multiple sampling to the data. Instead, we performed an additional experiment (referred to as experiment 2) in which we sorted naive T cells from an additional volunteer, and split the naive T cells into three subsamples before mRNA extraction. We then carried out library preparations and sequenced TCRα and TCRβ sequences from each subsample independently. In this experiment, sequences observed in more than one subsample must have been derived from different cells, and cannot be a result of sequencing multiple mRNA molecules from a single cell. Repeated sequences must therefore derive from different cells, and represent abundant sequences.

In total 16913 (3.4%) TCRα sequences, and 5744 (0.61%) TCRβ sequences, were observed in more than one subsample (Figure 2A), confirming the existence of a substantial number of frequent TCR α and β chains in the full naive repertoire. In order to exclude any contribution from sort contamination, we also plot the data after removing all TCR sequences found in both memory and naive repertoires (Figure 2A, grey bars). A substantial number of α and β chains were still found in multiple subsamples. In order to estimate the impact of multiple sampling on the observed abundances we randomly permuted the TCR sequences between subsamples, and reanalyzed the distributions (see detailed explanation in Section 'Subsampling to exclude inflated abundance through multiple RNA contributions by single cells'). We estimated that ∼25% of α and > 75% of β chains with an abundance of greater than 1 in an individual sample may arise from sampling multiple RNA molecules from single cells. The impact is strongest on TCR sequences observed twice (see Figure 2—figure supplement 1). Thus multiple mRNA sampling is an important confounder of estimating TCR sequence abundances in individual repertoires, especially for TCRβ.

Figure 2 with 3 supplements see all

Download asset Open asset

Figure 2—source data 1

Naive TCRα and TCRβ abundance in the subsamples of Experiment 2.

Per dataset, the number of TCR sequences is shown per incidence and abundance across subsamples. Abundance is log2 binned according to the total UMI count across the three naive subsamples. Pearson correlation coefficient between incidence and total UMI count is shown for each dataset.

https://cdn.elifesciences.org/articles/49900/elife-49900-fig2-data1-v1.txt

Download elife-49900-fig2-data1-v1.txt

Having ruled out the contribution of multiple mRNA sampling experimentally, we examined the relationship between TCR sequence abundance and $𝒫(\sigma )$ in this new data set. The TCRα chains present in more than one naive subsample are dominated by sequences with high $𝒫(\sigma )$. The median generation probability of TCRα sequences observed in two and three subsamples was 56- and 165-fold higher, respectively, than those observed only once (Figure 2B). The relationship for TCRβ sequences was remarkably different, however. While TCRβ sequences observed in two subsamples are mildly enriched for high generation probabilities, those observed in three subsamples have hardly any enrichment for high $𝒫(\sigma )$ (Figure 2B). Instead, their generation probabilities tend to be lower than those of the sequences observed in two subsamples, and more similar to the generation probabilities of TCRβ sequences seen in only one subsample. We obtained similar results when measuring the number of N-additions and VJ-deletions in the rearrangements: abundant α chains (with incidence 2 or 3) tend be closer to germline rearrangements, while this was only the case for β chains with incidence 2, and not for the most abundant β chains with incidence 3 (Figure 2C and D). These trends were observed both with and without removing the sequences that were also observed in memory (Figure 2, grey versus colored bars) and when processing the data with RTCR (Figure 2—figure supplement 2).

We further explored whether the more abundant sequences were also more ‘public’ (found in the repertoires of multiple individuals), which would be predicted if they are more likely products of V(D)J-recombination. We measured the degree of sharing between those TCR sequences observed in 1, 2, or 3 naive subsamples, and the TCRα and TCRβ repertoires of unfractionated blood samples collected from 28 healthy donors (details in Section 'Sharing of TCRα and TCRβ sequences'). Both TCRα and TCRβ sequences observed in two or three subsamples were found to be significantly more often shared with this independent cohort than those observed once (Figure 2—figure supplement 3A). The most frequent TCRα sequences, which were seen in three subsamples, showed the highest sharing degree, consistent with their strongest enrichment for high generation probabilities. The relatively small number of most frequent TCRβ sequences (i.e., those observed in three subsamples), did not show increased inter-individual sharing compared to the TCRβ sequences observed in two subsamples. Additional comparison with publicly available TCRβ data from a large cohort (Emerson et al., 2017) (see Section 'Sharing of TCRα and TCRβ sequences') showed that the most frequently observed β chains, which were observed in all three subsets in experiment 2, were less public than sequences observed in two subsamples (Figure 2—figure supplement 3B). The seemingly paradoxical finding that the most abundant TCRβ sequences (observed in all three subsamples) have lower $𝒫(\sigma )$, and are less public than those found twice, is explored in more detail below.

Computational models of TCR repertoire generation suggest the presence of a small proportion of large T-cell clones in the naive repertoire

In order to more rigorously test our ideas about the frequency distribution of clonotypes in the naive T-cell repertoire, we explored a number of possible computational models of repertoire generation and sampling, and compared model predictions with the experimental data discussed above. The first simplest scenario we considered was a neutral model of repertoire formation, similar to Hubbell’s Neutral Community Model (Hubbell, 2001; Figure 3A, details in Section 'Neutral model for dynamics of naive T cells'). The model assumes that there is no selective advantage of one TCR over another, and therefore the TCR of a naive T cell does not affect its lifespan or division rate. Consider a pool of $N$ naive T cells, from which cells are removed by cell death or by priming with antigen, leading to differentiation into a memory population. A fraction θ of these cells is replaced by thymic production of new clones and the remaining fraction 1 - θ gets replaced by division of cells present in the pool. When simulating the naive T-cell pool with this model, the clone-size distribution approaches a ‘steady state’ (not shown). We use this steady-state distribution, for which we have an analytical expression (Section 'Neutral model for dynamics of naive T cells') to predict the size of clones in the naive T-cell pool. As the contribution of thymic output decreases during aging (Steinmann et al., 1985), we evaluated the model for a wide range of values for θ. The clone-size distribution which emerges from the neutral model is approximately geometric for clone sizes larger than the introduction size $c$ (Figure 3B, Section 'Neutral model for dynamics of naive T cells'). We compared this basic model to models in which we impose other distributions on the underlying clonotype abundances (model details in Section 'Clone-size distributions of the naive T-cell pools'). We specifically focused on heavy-tailed distributions such as log-normal and power-law distributions, which have previously been associated with T-cell repertoires (Desponds et al., 2016). The shape of each of these distributions is controlled by a single parameter (as shown in Figure 3B), allowing us to compare distributions with different degrees of heterogeneity. In all cases, we normalized the clone-size distribution such that the total number of cells $N$ is constant. Since we had separate experimental data for CD4⁺ and CD8⁺ cells, we considered CD4⁺ and CD8⁺ cells separately, setting $N(CD4)=7.5\times {10}^{10}$, and $N(CD8)=2.5\times {10}^{10}$.

Figure 3

Download asset Open asset

From all model clone-size distributions we simulate three subsamples, so as to compare with the data from the second experiment described above. Each sampled TCR is assigned a TCRα and a TCRβ sequence that were generated with IGoR (Marcou et al., 2018). Previous studies showed that α and β chains with higher generation probabilities tend to have a higher probability to survive selection (Elhanati et al., 2014). Therefore, we train a simple $𝒫(\sigma )$-dependent selection model on the data from the single naive T-cell samples shown in Figure 1. First, we assume that productively rearranged chains have an overall 1/3 probability to survive thymic selection. Then we bias the probability for bins of sequences based on their $𝒫(\sigma )$, such that the resulting set of α and β chains has the same generation probability distribution as in the experimental repertoire data (Section 'In silico samples from modelled clone-size distributions'). The models also incorporate the expected number of cells that contribute at least one mRNA molecule. This parameter is also learnt from the data, by setting the number of cells that contributed mRNA such that the predicted diversity of a subsample matches the observed diversity (Section 'In silico samples from modelled clone-size distributions'). Taken together, the subsamples we take from the various model clone-size distributions are such that they match the generation probabilities and diversity of the experimental subsamples as closely as possible. We compare the number and median $𝒫(\sigma )$ of the TCR sequences that are predicted to occur in only one, in two or in three subsamples, with the equivalent experimental data from experiment two above (Figure 4).

Figure 4 with 3 supplements see all

Download asset Open asset

We consider first the neutral model (Figure 4A and C). For the $\alpha$ chain, a wide range of thymic output rates predict the number of chains occurring in 1, 2 and 3 subsamples reasonably well (Figure 4A). The model does not predict the median $𝒫(\sigma )$ of TCRα found in 2 and 3 subsamples well, although qualitatively the model does predict the increasing $𝒫(\sigma )$ with increasing abundance (Figure 4C). For the β chains, there is no range of thymic output rates for which the model correctly predicts the number of sequences observed in 2 and 3 subsamples. Moreover, the observation that incidence two chains have higher $𝒫(\sigma )$ than incidence three chains was not predicted for any value of θ (Figure 4C). Thus, although the neutral model captures some features of the observed TCRα sequence abundances, it cannot account for observed TCRβ distributions. A similarly poor match between observed and predicted data is observed for log-normal clonotype model (Figure 3B) distributions (not shown).

In contrast, there is a much better fit between observed and predicted data is obtained when the model clonotype frequencies are modelled by a power-law distribution (Figure 4B and D). Like the distributions discussed above, in the parameter range where clone-size heterogeneity is limited (i.e. a steep slope), a power-law distribution predicts both the number of TCRα sequences found in 2 and 3 samples, and their larger median $𝒫(\sigma )$. The number of TCRβ sequences is also predicted well if the slope is close to 2.3 (Figure 4B). Remarkably, for this slope the median $𝒫(\sigma )$ of TCRβ sequences found in two samples is higher than the median $𝒫(\sigma )$ of TCRβ sequences found in three samples (Figure 4D). Intuitively, we can understand this observation as reflecting the properties of power-law distributions, combined with the lower generation probabilities of TCRβ. Identical TCRβ recombinations occur frequently enough to make a detectable contribution to the TCR sequences observed in two samples, but not to those detected in three samples. Therefore, a significant proportion of TCRβ sequences observed twice are in fact derived from two or more different naive T-cell clones. In contrast, TCR sequences observed three times (or more) must be derived from large naive T-cell clones. Abundant TCRα sequences arise both from large clones and summation of identical TCRα from multiple smaller clones, but due to their higher generation probabilities, the latter dominates the $𝒫(\sigma )$ for TCRα sequences found twice and three times. Finally, we note that although TCR sequence abundance in the single samples from experiment one is likely to incorporate multiple mRNA from single cells, the power-law distribution also predicts abundances in the single samples of experiment one reasonably well (Figure 4—figure supplement 1).

The vast majority of TCR sequences in samples of naive T cells are observed only once, and hence we cannot infer anything about their frequency in the whole repertoire, except that it is likely to be below a given abundance threshold. Therefore we explored whether a more generalised model, which does not make any assumptions about the distribution of the low abundance T cells, would predict our experimental data as well as the power-law model. In this simple mixture model we generate a population in which the majority of cells are present only once, and a minority are present many times. We scanned the parameter space of this model, varying both the proportion of cells in each population, and the size of clones in the larger population. The prediction of the model for each parameter pair was compared to the experimental data from experiment 2, both for the number of TCRs (combining α and β sequences) observed in one, two or three subsamples, and for the median $𝒫(\sigma )$ of these TCR sequences. The best agreement between model and data was observed when 1–5% of the cells were derived from abundant T-cell clones (between 10⁵ and 10⁶ cells in the whole repertoire) (Figure 4E).

Abundant T-cell sequences are enriched for zero insertions and for antigen-association

In human prenatal thymocytes, the enzyme terminal deoxynucleotidyl transferase (TdT) is not expressed, leading to the production of TCR sequences with zero insertions of N-nucleotides. Pogorelyy and colleagues showed that enrichment of zero insertion TCR sequences can be used to detect fetal clones even in adults, and that their contribution to the overall repertoire decays slowly with age (Pogorelyy et al., 2017). Interestingly, the proportion of zero-insertion sequences was strongly enhanced in those sequences observed more than once in the three subsamples examined in experiment 2 (Figure 5A). The interpretation of this finding is not straightforward, since zero-insertion TCR sequences have higher median generation probabilities, and this is also a property of abundant sequences as discussed above. Nevertheless, the data are compatible with a model in which the large clones observed in the repertoire are generated preferentially during early prenatal development of the naive T-cell repertoire.

Figure 5 with 1 supplement see all

Download asset Open asset

We next examined if the abundant sequences in our data showed characteristics of semi-invariant NKT and MAIT cell populations. Classical NKT cells are characterized by an invariant TRAV24-TRAJ18 α chain and β chains with TRBV11 (Dellabona et al., 1994). MAIT cells are enriched for TCRα rearrangements of TRAV1-2 with TRAJ33, TRAJ12 and TRAJ20 (Reantragoon et al., 2013), and TCRβ sequences predominantly using TRBV20 and TRBV6 (Lepore et al., 2014). Since our HTS data does not contain information on αβ pairing, we studied both chains separately. A substantial fraction of the observed TCRβ sequences matches the characteristics of MAIT cells, and to a lesser extent NKT cells (Figure 5B and C). For both cell types, however, this fraction does not show a clear relation to incidence, and does not suggest enrichment for MAIT or NKT cells among abundant sequences. The most abundant TCRα sequences are enriched for NKT sequences, but these still account for only a small fraction of the total (0.3% and 1.7% for CD4⁺ and CD8⁺, respectively, Figure 5B). Hence, we conclude that only a small fraction of the abundant sequences are derived from clones with a MAIT or NKT cell phenotype.

Finally we analysed whether the abundant TCR sequences in the naive population could be detected in a database of TCR sequences with known antigen specificity (Shugay et al., 2018). Interestingly, there was a striking enrichment of TCR sequences with known antigen-specific annotation within the high abundance TCRα sequences observed in more than one subsample from experiment 2, and to a lesser extent for TCRβ sequences (Figure 5C). Interpretation is again not straightforward, because the high generation probabilities of the abundantly observed chains could lead to these sequences being over-represented in the database (ascertainment bias). Additionally, the observation may also reflect the fact that the naive T-cell populations we sequenced contained some antigen-experienced T cells with a naive phenotype (Pulko et al., 2016). Finally, the observation is also compatible with the hypothesis that TCR recombination has evolved to preferentially generate TCRs specific to common pathogens like CMV or EBV as discussed in Thomas and Crawford (2019).

The diversity and clone size distribution of the naive T-cell repertoire has been the subject of considerable debate, fueled by the difficulty of obtaining more than a very small sample of the total repertoire, and by a variety of other technical considerations which we address in this study. We use a quantitative UMI-based sequencing protocol, and careful error correction to analyse the naive and memory repertoires from three healthy human volunteers. We convincingly demonstrate that a small proportion of the TCR sequences are present more than once in a sample of naive T cells from blood, corresponding to expected frequencies greater than 1 in 10⁵. This number of abundant TCRα sequences is higher than the number of abundant TCRβ sequences.

We carefully considered different mechanisms that could give rise to these abundant TCR sequences. We examined the contribution of potential contamination of the naive population with abundant T cells from the memory compartment during the sorting process, but the extent of such contamination was small (for CD8⁺ cells) or not detectable (for CD4⁺ cells). Furthermore, exclusion of all TCR sequences which occurred in both memory and naive populations did not alter the subsequent conclusions of the analysis. We also considered the possibility that abundant TCR sequences were observed due to sampling multiple mRNA molecules from the same cell. In order to exclude this possibility, we carried out an experiment where we divided up a sample of sorted naive T cells into three subsamples prior to lysis, and sequencing. In this experimental paradigm, TCR sequences found in more than one subsample must arise from different T cells. We observed that repeat sampling of mRNA from the same T cell did indeed occur, and might account for as much as 75% of the high abundance TCRβ sequences (for which there are more mRNA molecules per cell, [Oakes et al., 2017]), and as much as 25% of the TCRα sequences. However, this effect was mostly restricted to TCR sequences observed twice, and made little contribution to TCR sequences observed three or more times.

Having excluded methodological causes of high abundance TCR sequences, we examined two biological mechanisms which could explain the data. The first mechanism we consider is that abundant sequences derive from identical TCRα and TCRβ rearrangements occurring in multiple clones. In this model, abundance arises not from multiple sampling from the same large clone of T cells, but from summation over many different clones of T cells, each of which share an α or β chain. The second mechanism is that the naive repertoire clone size distribution is not uniform, but contains many small and some large clones. We combine computational models with experimental data to provide evidence that both mechanisms are required to explain the observed data. The first mechanism dominates the repertoire of TCRα, and is likely to contribute to the majority of observed abundant sequences. Interestingly, the model suggests that those TCRα which have the highest probability of generation are produced hundreds of thousands, or even millions of times within an individual, and must therefore be produced extremely frequently in the thymus. In contrast, the first mechanism has a smaller impact on the TCRβ repertoire, and abundant TCRβ sequences are more likely to arise from large clones in the naive repertoire.

The experimental limitations of sampling small volume of blood which contains only a tiny proportion of the total repertoire has dramatic effects on the observed TCR frequency distribution. One can use the analytical solution of the neutral model (Section 'Neutral model for dynamics of naive T cells') with thymic introduction size $c=1$ to illustrate this extreme sampling effect: ${\widehat{F}}_i\approx F_i{(\frac{s}{\theta })}^i$, where $\widehat{F_i}$ and $F_i$ are the number of clones present with $i$ cells in the sample, and in the pool, respectively, and $s$ is the fraction of the repertoire that was sampled (here s ∼ 10^-6). Since $s/\theta$ is of order 10^-5 and this is raised to the $i$^th power, even very large TCR clones become rare in such a sample. Because of this, it is difficult to be definitive about the exact underlying T-cell clone frequency distribution which gives rise to the abundant TCR sequences we observe. The data are certainly compatible with a power-law distribution, as has been suggested previously (Desponds et al., 2016). But many distributions made up of a mixture of rare clones and a small proportion (1–5%) of large clones (10⁵ - 10⁶) are compatible with the data we observe.

The demonstration of large clones in the naive repertoire raises the question of what determines the different sizes of different clonotypes. The neutral model already excludes repeated thymic production as explanation for large clones, because the combined probability of repeated αβ-clone production is very low (Dupic et al., 2019). We confirmed that the abundant TCR sequences were not strongly enriched for sequences characteristic of iNKT and MAIT cells (Figure 5B&C). An alternative explanation is that the large clones may actually be antigen-experienced, but with a naive phenotype such as memory stem T cells (Gattinoni et al., 2011; Lugli et al., 2013a; Lugli et al., 2013b; Fuertes Marraco et al., 2015; Pulko et al., 2016). However, a number of alternative explanations for this enrichment exist, as discussed above. Furthermore, antigen-experienced T cells should be present in the memory populations, and large clones could still be observed even after removal of all cells which occur in both memory and naive populations. So, although we cannot exclude that some of the frequent TCR sequences may be derived from T cells that are not truly naive, we believe the data argue for the existence of truly naive large clones.

We speculate that the most likely mechanism for large clones is preferential growth/survival of some clones, presumably due to preferential selection on self-peptide/MHC (Rudd et al., 2011; Lythe et al., 2016). Intriguingly, the abundant TCR sequences we observed were enriched for sequences without N-insertions, a characteristic of TCRs produced prenatally (Pogorelyy et al., 2017). The large clones may therefore be established very early in the development of T-cell adaptive immunity, before homeostasis of the immune system is achieved and when more rapid division and clonal expansion may be favoured.

In conclusion, our study highlights the huge impact of subsampling on correct interpretation of TCR repertoire data. It provides evidence for two different mechanisms which give rise to abundant TCR sequences in the naive human repertoire. The first mechanism, driven by multiple identical recombination events, is frequently overlooked in the analysis of T-cell repertoires, but has important implications in interpretation of observed sharing between different T-cell subpopulations of an individual, and between individuals (public TCR sequences). The second mechanism suggests that the TCR sequence plays a critical role in naive T-cell homeostasis. Further experiments will be required to fully elucidate the cellular and molecular mechanisms which underlie the heterogeneity of the naive T-cell repertoire.

Sequencing data have been deposited in the Short Read Archive as experiments SRP109035, SRP045430 and SRP151125.

1. 1. Oakes T
   2. Heather JM
   3. Ismail M

   (2018) SRP151125

   ID NCBI Sequence Read Archive. T cell receptor repertoire sequencing of healthy individuals.

   https://trace.ncbi.nlm.nih.gov/Traces/sra/?study=SRP151125

1. 1. Oakes T
   2. Heather JM
   3. Best K
   4. Byng-Maddick R
   5. Husovsky C
   6. Ismail M
   7. Joshi K
   8. Maxwell G
   9. Noursadeghi M
   10. Riddell N
   11. Ruehl T
   12. Turner CT
   13. Uddin I
   14. Chain B

   (2017) NCBI Sequence Read Archive

   ID SRP109035. TCR repertoire sequencing of T cell subsets from healthy individuals.

   https://trace.ncbi.nlm.nih.gov/Traces/sra/?study=SRP109035
2. 1. Heather JM
   2. Best K
   3. Oakes T
   4. Gray ER
   5. Roe JK
   6. Thomas N
   7. Friedman N
   8. Noursadeghi M
   9. Chain B

   (2014) NCBI Sequence Read Archive

   ID SRP045430. T-cell receptor repertoires in HIV-infected patients and healthy controls.

   https://trace.ncbi.nlm.nih.gov/Traces/sra/?study=SRP045430
3. 1. Emerson R
   2. DeWitt W
   3. Vignali M
   4. Gravley J
   5. Hu J
   6. Osborne E
   7. Desmarais C
   8. Klinger M
   9. Carlson C
   10. Hansen J
   11. Rieder M
   12. Robins H

   (2017) immuneACCESS

   Immunosequencing identifies signatures of cytomegalovirus exposure history and HLA-mediated effects on the T-cell repertoire.

   https://doi.org/10.21417/B7001Z

1. 1. Arstila TP
   2. Casrouge A
   3. Baron V
   4. Even J
   5. Kanellopoulos J
   6. Kourilsky P

   (1999) A direct estimate of the human alphabeta T cell receptor diversity

   Science 286:958–961.

   https://doi.org/10.1126/science.286.5441.958

   • PubMed
   • Google Scholar
2. 1. Clark DR
   2. de Boer RJ
   3. Wolthers KC
   4. Miedema F

   (1999) T cell dynamics in HIV-1 infection

   Advances in Immunology 73:256–304.

   https://doi.org/10.1016/s0065-2776(08)60789-0

   • Google Scholar
3. 1. Davis MM
   2. Bjorkman PJ

   (1988) T-cell antigen receptor genes and T-cell recognition

   Nature 335:744.

   https://doi.org/10.1038/335744b0

   • Google Scholar
4. 1. De Boer RJ
   2. Perelson AS

   (1994) T cell repertoires and competitive exclusion

   Journal of Theoretical Biology 169:375–390.

   https://doi.org/10.1006/jtbi.1994.1160

   • PubMed
   • Google Scholar
5. 1. Dellabona P
   2. Padovan E
   3. Casorati G
   4. Brockhaus M
   5. Lanzavecchia A

   (1994) An invariant V alpha 24-J alpha Q/V beta 11 T cell receptor is expressed in all individuals by clonally expanded CD4-8- T cells

   The Journal of Experimental Medicine 180:1171–1176.

   https://doi.org/10.1084/jem.180.3.1171

   • PubMed
   • Google Scholar
6. 1. den Braber I
   2. Mugwagwa T
   3. Vrisekoop N
   4. Westera L
   5. Mögling R
   6. de Boer AB
   7. Willems N
   8. Schrijver EH
   9. Spierenburg G
   10. Gaiser K
   11. Mul E
   12. Otto SA
   13. Ruiter AF
   14. Ackermans MT
   15. Miedema F
   16. Borghans JA
   17. de Boer RJ
   18. Tesselaar K

   (2012) Maintenance of peripheral naive T cells is sustained by Thymus output in mice but not humans

   Immunity 36:288–297.

   https://doi.org/10.1016/j.immuni.2012.02.006

   • PubMed
   • Google Scholar
7. 1. Desponds J
   2. Mora T
   3. Walczak AM

   (2016) Fluctuating fitness shapes the clone-size distribution of immune repertoires

   PNAS 113:274–279.

   https://doi.org/10.1073/pnas.1512977112

   • PubMed
   • Google Scholar
8. Preprint

   1. Desponds J
   2. Mayer A
   3. Mora T
   4. Walczak AM

   (2017) Population dynamics of immune repertoires.

   arXiv.

   https://arxiv.org/abs/1703.00226

   • Google Scholar
9. 1. Dowling MR
   2. Hodgkin PD

   (2009) Modelling naive T-cell homeostasis: consequences of heritable cellular lifespan during ageing

   Immunology and Cell Biology 87:445–456.

   https://doi.org/10.1038/icb.2009.11

   • PubMed
   • Google Scholar
10. 1. Dupic T
    2. Marcou Q
    3. Walczak AM
    4. Mora T

    (2019) Genesis of the αβ T-cell receptor

    PLOS Computational Biology 15:e1006874.

    https://doi.org/10.1371/journal.pcbi.1006874

    • PubMed
    • Google Scholar
11. 1. Elhanati Y
    2. Murugan A
    3. Callan CG
    4. Mora T
    5. Walczak AM

    (2014) Quantifying selection in immune receptor repertoires

    PNAS 111:9875–9880.

    https://doi.org/10.1073/pnas.1409572111

    • PubMed
    • Google Scholar
12. 1. Elhanati Y
    2. Sethna Z
    3. Callan CG
    4. Mora T
    5. Walczak AM

    (2018) Predicting the spectrum of TCR repertoire sharing with a data-driven model of recombination

    Immunological Reviews 284:167–179.

    https://doi.org/10.1111/imr.12665

    • PubMed
    • Google Scholar
13. 1. Emerson RO
    2. DeWitt WS
    3. Vignali M
    4. Gravley J
    5. Hu JK
    6. Osborne EJ
    7. Desmarais C
    8. Klinger M
    9. Carlson CS
    10. Hansen JA
    11. Rieder M
    12. Robins HS

    (2017) Immunosequencing identifies signatures of Cytomegalovirus exposure history and HLA-mediated effects on the T cell repertoire

    Nature Genetics 49:659–665.

    https://doi.org/10.1038/ng.3822

    • PubMed
    • Google Scholar
14. 1. Fuertes Marraco SA
    2. Soneson C
    3. Cagnon L
    4. Gannon PO
    5. Allard M
    6. Abed Maillard S
    7. Montandon N
    8. Rufer N
    9. Waldvogel S
    10. Delorenzi M
    11. Speiser DE

    (2015) Long-lasting stem cell-like memory CD8+ T cells with a naïve-like profile upon yellow fever vaccination

    Science Translational Medicine 7:282ra48.

    https://doi.org/10.1126/scitranslmed.aaa3700

    • PubMed
    • Google Scholar
15. 1. Gattinoni L
    2. Lugli E
    3. Ji Y
    4. Pos Z
    5. Paulos CM
    6. Quigley MF
    7. Almeida JR
    8. Gostick E
    9. Yu Z
    10. Carpenito C
    11. Wang E
    12. Douek DC
    13. Price DA
    14. June CH
    15. Marincola FM
    16. Roederer M
    17. Restifo NP

    (2011) A human memory T cell subset with stem cell-like properties

    Nature Medicine 17:1290–1297.

    https://doi.org/10.1038/nm.2446

    • PubMed
    • Google Scholar
16. 1. Gerritsen B
    2. Pandit A
    3. Andeweg AC
    4. de Boer RJ

    (2016) RTCR: a pipeline for complete and accurate recovery of T cell repertoires from high throughput sequencing data

    Bioinformatics 32:3098–3106.

    https://doi.org/10.1093/bioinformatics/btw339

    • PubMed
    • Google Scholar
17. 1. Gonçalves P
    2. Ferrarini M
    3. Molina-Paris C
    4. Lythe G
    5. Vasseur F
    6. Lim A
    7. Rocha B
    8. Azogui O

    (2017) A new mechanism shapes the naïve CD8⁺ T cell repertoire: the selection for full diversity

    Molecular Immunology 85:66–80.

    https://doi.org/10.1016/j.molimm.2017.01.026

    • PubMed
    • Google Scholar
18. 1. Hapuarachchi T
    2. Lewis J
    3. Callard RE

    (2013) A mechanistic model for naive CD4 T cell homeostasis in healthy adults and children

    Frontiers in Immunology 4:366.

    https://doi.org/10.3389/fimmu.2013.00366

    • PubMed
    • Google Scholar
19. 1. Hogan T
    2. Gossel G
    3. Yates AJ
    4. Seddon B

    (2015) Temporal fate mapping reveals age-linked heterogeneity in naive T lymphocytes in mice

    PNAS 112:E6917–E6926.

    https://doi.org/10.1073/pnas.1517246112

    • PubMed
    • Google Scholar
20. Book

    1. Hubbell SP

    (2001) The Unified Neutral Theory of Biodiversity and Biogeography

    Princeton University Press.

    https://doi.org/10.4249/scholarpedia.8822

    • Google Scholar
21. 1. Jenkins MK
    2. Chu HH
    3. McLachlan JB
    4. Moon JJ

    (2010) On the composition of the preimmune repertoire of T cells specific for Peptide-major histocompatibility complex ligands

    Annual Review of Immunology 28:275–294.

    https://doi.org/10.1146/annurev-immunol-030409-101253

    • PubMed
    • Google Scholar
22. 1. Johnson PL
    2. Yates AJ
    3. Goronzy JJ
    4. Antia R

    (2012) Peripheral selection rather than thymic involution explains sudden contraction in naive CD4 T-cell diversity with age

    PNAS 109:21432–21437.

    https://doi.org/10.1073/pnas.1209283110

    • PubMed
    • Google Scholar
23. 1. Kumar BV
    2. Connors TJ
    3. Farber DL

    (2018) Human T cell development, localization, and function throughout life

    Immunity 48:202–213.

    https://doi.org/10.1016/j.immuni.2018.01.007

    • PubMed
    • Google Scholar
24. 1. Lepore M
    2. Kalinichenko A
    3. Kalinicenko A
    4. Colone A
    5. Paleja B
    6. Singhal A
    7. Tschumi A
    8. Lee B
    9. Poidinger M
    10. Zolezzi F
    11. Quagliata L
    12. Sander P
    13. Newell E
    14. Bertoletti A
    15. Terracciano L
    16. De Libero G
    17. Mori L

    (2014) Parallel T-cell cloning and deep sequencing of human MAIT cells reveal stable oligoclonal tcrβ repertoire

    Nature Communications 5:3866.

    https://doi.org/10.1038/ncomms4866

    • PubMed
    • Google Scholar
25. 1. Lugli E
    2. Dominguez MH
    3. Gattinoni L
    4. Chattopadhyay PK
    5. Bolton DL
    6. Song K
    7. Klatt NR
    8. Brenchley JM
    9. Vaccari M
    10. Gostick E
    11. Price DA
    12. Waldmann TA
    13. Restifo NP
    14. Franchini G
    15. Roederer M

    (2013a) Superior T memory stem cell persistence supports long-lived T cell memory

    Journal of Clinical Investigation 8:33.

    https://doi.org/10.1172/JCI66327

    • Google Scholar
26. 1. Lugli E
    2. Gattinoni L
    3. Roberto A
    4. Mavilio D
    5. Price DA
    6. Restifo NP
    7. Roederer M

    (2013b) Identification, isolation and in vitro expansion of human and nonhuman primate T stem cell memory cells

    Nature Protocols 8:33–42.

    https://doi.org/10.1038/nprot.2012.143

    • PubMed
    • Google Scholar
27. 1. Lythe G
    2. Callard RE
    3. Hoare RL
    4. Molina-París C

    (2016) How many TCR clonotypes does a body maintain?

    Journal of Theoretical Biology 389:214–224.

    https://doi.org/10.1016/j.jtbi.2015.10.016

    • PubMed
    • Google Scholar
28. 1. Marcou Q
    2. Mora T
    3. Walczak AM

    (2018) High-throughput immune repertoire analysis with IGoR

    Nature Communications 9:561.

    https://doi.org/10.1038/s41467-018-02832-w

    • PubMed
    • Google Scholar
29. 1. Mason D

    (1998) A very high level of crossreactivity is an essential feature of the T-cell receptor

    Immunology Today 19:395–404.

    https://doi.org/10.1016/S0167-5699(98)01299-7

    • PubMed
    • Google Scholar
30. 1. McDonald BD
    2. Bunker JJ
    3. Erickson SA
    4. Oh-Hora M
    5. Bendelac A

    (2015) Crossreactive αβ T cell receptors are the predominant targets of thymocyte negative selection

    Immunity 43:859–869.

    https://doi.org/10.1016/j.immuni.2015.09.009

    • PubMed
    • Google Scholar
31. 1. Merkenschlager M
    2. Graf D
    3. Lovatt M
    4. Bommhardt U
    5. Zamoyska R
    6. Fisher AG

    (1997) How many thymocytes audition for selection?

    The Journal of Experimental Medicine 186:1149–1158.

    https://doi.org/10.1084/jem.186.7.1149

    • PubMed
    • Google Scholar
32. Book

    1. Mora T
    2. Walczak A

    (2019)

    Quantifying lymphocyte receptor diversity

    In: Das J, Jayaprakash C, editors. Systems Immunology. CRC Press. pp. 183–198.

    • Google Scholar
33. 1. Murugan A
    2. Mora T
    3. Walczak AM
    4. Callan CG

    (2012) Statistical inference of the generation probability of T-cell receptors from sequence repertoires

    PNAS 109:16161–16166.

    https://doi.org/10.1073/pnas.1212755109

    • PubMed
    • Google Scholar
34. 1. Nikolich-Zugich J
    2. Slifka MK
    3. Messaoudi I

    (2004) The many important facets of T-cell repertoire diversity

    Nature Reviews Immunology 4:123–132.

    https://doi.org/10.1038/nri1292

    • PubMed
    • Google Scholar
35. 1. Oakes T
    2. Heather JM
    3. Best K
    4. Byng-Maddick R
    5. Husovsky C
    6. Ismail M
    7. Joshi K
    8. Maxwell G
    9. Noursadeghi M
    10. Riddell N
    11. Ruehl T
    12. Turner CT
    13. Uddin I
    14. Chain B

    (2017) Quantitative characterization of the T cell receptor repertoire of naïve and memory subsets using an integrated experimental and computational pipeline which is robust, economical, and versatile

    Frontiers in Immunology 8:1267.

    https://doi.org/10.3389/fimmu.2017.01267

    • PubMed
    • Google Scholar
36. 1. Pogorelyy MV
    2. Elhanati Y
    3. Marcou Q
    4. Sycheva AL
    5. Komech EA
    6. Nazarov VI
    7. Britanova OV
    8. Chudakov DM
    9. Mamedov IZ
    10. Lebedev YB
    11. Mora T
    12. Walczak AM

    (2017) Persisting fetal clonotypes influence the structure and overlap of adult human T cell receptor repertoires

    PLOS Computational Biology 13:e1005572.

    https://doi.org/10.1371/journal.pcbi.1005572

    • PubMed
    • Google Scholar
37. 1. Pulko V
    2. Davies JS
    3. Martinez C
    4. Lanteri MC
    5. Busch MP
    6. Diamond MS
    7. Knox K
    8. Bush EC
    9. Sims PA
    10. Sinari S
    11. Billheimer D
    12. Haddad EK
    13. Murray KO
    14. Wertheimer AM
    15. Nikolich-Žugich J

    (2016) Human memory T cells with a naive phenotype accumulate with aging and respond to persistent viruses

    Nature Immunology 17:966–975.

    https://doi.org/10.1038/ni.3483

    • PubMed
    • Google Scholar
38. 1. Qi Q
    2. Liu Y
    3. Cheng Y
    4. Glanville J
    5. Zhang D
    6. Lee JY
    7. Olshen RA
    8. Weyand CM
    9. Boyd SD
    10. Goronzy JJ

    (2014) Diversity and clonal selection in the human T-cell repertoire

    PNAS 111:13139–13144.

    https://doi.org/10.1073/pnas.1409155111

    • PubMed
    • Google Scholar
39. 1. Quigley MF
    2. Greenaway HY
    3. Venturi V
    4. Lindsay R
    5. Quinn KM
    6. Seder RA
    7. Douek DC
    8. Davenport MP
    9. Price DA

    (2010) Convergent recombination shapes the clonotypic landscape of the naive T-cell repertoire

    PNAS 107:19414–19419.

    https://doi.org/10.1073/pnas.1010586107

    • PubMed
    • Google Scholar
40. 1. Rane S
    2. Hogan T
    3. Seddon B
    4. Yates AJ

    (2018) Age is not just a number: naive T cells increase their ability to persist in the circulation over time

    PLOS Biology 16:e2003949.

    https://doi.org/10.1371/journal.pbio.2003949

    • PubMed
    • Google Scholar
41. 1. Reantragoon R
    2. Corbett AJ
    3. Sakala IG
    4. Gherardin NA
    5. Furness JB
    6. Chen Z
    7. Eckle SB
    8. Uldrich AP
    9. Birkinshaw RW
    10. Patel O
    11. Kostenko L
    12. Meehan B
    13. Kedzierska K
    14. Liu L
    15. Fairlie DP
    16. Hansen TH
    17. Godfrey DI
    18. Rossjohn J
    19. McCluskey J
    20. Kjer-Nielsen L

    (2013) Antigen-loaded MR1 tetramers define T cell receptor heterogeneity in mucosal-associated invariant T cells

    The Journal of Experimental Medicine 210:2305–2320.

    https://doi.org/10.1084/jem.20130958

    • PubMed
    • Google Scholar
42. 1. Reynaldi A
    2. Smith NL
    3. Schlub TE
    4. Tabilas C
    5. Venturi V
    6. Rudd BD
    7. Davenport MP

    (2019) Fate mapping reveals the age structure of the peripheral T cell compartment

    PNAS 116:3974–3981.

    https://doi.org/10.1073/pnas.1811634116

    • PubMed
    • Google Scholar
43. 1. Robins HS
    2. Campregher PV
    3. Srivastava SK
    4. Wacher A
    5. Turtle CJ
    6. Kahsai O
    7. Riddell SR
    8. Warren EH
    9. Carlson CS

    (2009) Comprehensive assessment of T-cell receptor beta-chain diversity in alphabeta T cells

    Blood 114:4099–4107.

    https://doi.org/10.1182/blood-2009-04-217604

    • PubMed
    • Google Scholar
44. 1. Robins HS
    2. Srivastava SK
    3. Campregher PV
    4. Turtle CJ
    5. Andriesen J
    6. Riddell SR
    7. Carlson CS
    8. Warren EH

    (2010) Overlap and effective size of the human CD8+ T cell receptor repertoire

    Science Translational Medicine 2:47ra64.

    https://doi.org/10.1126/scitranslmed.3001442

    • PubMed
    • Google Scholar
45. 1. Rudd BD
    2. Venturi V
    3. Li G
    4. Samadder P
    5. Ertelt JM
    6. Way SS
    7. Davenport MP
    8. Nikolich-Žugich J

    (2011) Nonrandom attrition of the naive CD8+ T-cell pool with aging governed by T-cell receptor:pmhc interactions

    PNAS 108:13694–13699.

    https://doi.org/10.1073/pnas.1107594108

    • PubMed
    • Google Scholar
46. 1. Sewell AK

    (2012) Why must T cells be cross-reactive?

    Nature Reviews Immunology 12:669–677.

    https://doi.org/10.1038/nri3279

    • PubMed
    • Google Scholar
47. 1. Shugay M
    2. Bagaev DV
    3. Zvyagin IV
    4. Vroomans RM
    5. Crawford JC
    6. Dolton G
    7. Komech EA
    8. Sycheva AL
    9. Koneva AE
    10. Egorov ES
    11. Eliseev AV
    12. Van Dyk E
    13. Dash P
    14. Attaf M
    15. Rius C
    16. Ladell K
    17. McLaren JE
    18. Matthews KK
    19. Clemens EB
    20. Douek DC
    21. Luciani F
    22. van Baarle D
    23. Kedzierska K
    24. Kesmir C
    25. Thomas PG
    26. Price DA
    27. Sewell AK
    28. Chudakov DM

    (2018) VDJdb: a curated database of T-cell receptor sequences with known antigen specificity

    Nucleic Acids Research 46:D419–D427.

    https://doi.org/10.1093/nar/gkx760

    • PubMed
    • Google Scholar
48. 1. Steinmann GG
    2. Klaus B
    3. Müller-Hermelink HK

    (1985) The involution of the ageing human thymic epithelium is independent of puberty. A morphometric study

    Scandinavian Journal of Immunology 22:563–575.

    https://doi.org/10.1111/j.1365-3083.1985.tb01916.x

    • PubMed
    • Google Scholar
49. 1. Stirk ER
    2. Molina-París C
    3. van den Berg HA

    (2008) Stochastic niche structure and diversity maintenance in the T cell repertoire

    Journal of Theoretical Biology 255:237–249.

    https://doi.org/10.1016/j.jtbi.2008.07.017

    • PubMed
    • Google Scholar
50. 1. Stirk ER
    2. Lythe G
    3. van den Berg HA
    4. Molina-París C

    (2010) Stochastic competitive exclusion in the maintenance of the naïve T cell repertoire

    Journal of Theoretical Biology 265:396–410.

    https://doi.org/10.1016/j.jtbi.2010.05.004

    • PubMed
    • Google Scholar
51. 1. Takada K
    2. Jameson SC

    (2009) Naive T cell homeostasis: from awareness of space to a sense of place

    Nature Reviews Immunology 9:823–832.

    https://doi.org/10.1038/nri2657

    • PubMed
    • Google Scholar
52. 1. Tanchot C
    2. Lemonnier FA
    3. Pérarnau B
    4. Freitas AA
    5. Rocha B

    (1997) Differential requirements for survival and proliferation of CD8 naïve or memory T cells

    Science 276:2057–2062.

    https://doi.org/10.1126/science.276.5321.2057

    • PubMed
    • Google Scholar
53. 1. Tanno H
    2. Gould TM
    3. McDaniel JR
    4. Cao W
    5. Tanno Y
    6. Durrett RE
    7. Park D
    8. Cate SJ
    9. Hildebrand WH
    10. Dekker CL
    11. Tian L
    12. Weyand CM
    13. Georgiou G
    14. Goronzy JJ

    (2020) Determinants governing T cell receptor α/β-chain pairing in repertoire formation of identical twins

    PNAS 117:532–540.

    https://doi.org/10.1073/pnas.1915008117

    • PubMed
    • Google Scholar
54. 1. Thomas N
    2. Heather J
    3. Ndifon W
    4. Shawe-Taylor J
    5. Chain B

    (2013) Decombinator: a tool for fast, efficient gene assignment in T-cell receptor sequences using a finite state machine

    Bioinformatics 29:542–550.

    https://doi.org/10.1093/bioinformatics/btt004

    • PubMed
    • Google Scholar
55. 1. Thomas PG
    2. Crawford JC

    (2019) Selected before selection: a case for inherent antigen Bias in the T-cell receptor repertoire

    Current Opinion in Systems Biology 18:36–43.

    https://doi.org/10.1016/j.coisb.2019.10.007

    • Google Scholar
56. 1. Uddin I
    2. Joshi K
    3. Oakes T
    4. Heather JM
    5. Swanton C
    6. Chain B

    (2019) An economical, quantitative, and robust protocol for high-throughput T cell receptor sequencing from tumor or blood

    Cancer Immunosurveillance 1184:15–42.

    https://doi.org/10.1007/978-1-4939-8885-3_2

    • Google Scholar
57. 1. Venturi V
    2. Quigley MF
    3. Greenaway HY
    4. Ng PC
    5. Ende ZS
    6. McIntosh T
    7. Asher TE
    8. Almeida JR
    9. Levy S
    10. Price DA
    11. Davenport MP
    12. Douek DC

    (2011) A mechanism for TCR sharing between T cell subsets and individuals revealed by pyrosequencing

    The Journal of Immunology 186:4285–4294.

    https://doi.org/10.4049/jimmunol.1003898

    • PubMed
    • Google Scholar
58. 1. Wertheimer AM
    2. Bennett MS
    3. Park B
    4. Uhrlaub JL
    5. Martinez C
    6. Pulko V
    7. Currier NL
    8. Nikolich-Žugich D
    9. Kaye J
    10. Nikolich-Žugich J

    (2014) Aging and Cytomegalovirus infection differentially and jointly affect distinct circulating T cell subsets in humans

    The Journal of Immunology 192:2143–2155.

    https://doi.org/10.4049/jimmunol.1301721

    • PubMed
    • Google Scholar
59. 1. Westera L
    2. van Hoeven V
    3. Drylewicz J
    4. Spierenburg G
    5. van Velzen JF
    6. de Boer RJ
    7. Tesselaar K
    8. Borghans JA

    (2015) Lymphocyte maintenance during healthy aging requires no substantial alterations in cellular turnover

    Aging Cell 14:219–227.

    https://doi.org/10.1111/acel.12311

    • PubMed
    • Google Scholar
60. 1. Zarnitsyna VI
    2. Evavold BD
    3. Schoettle LN
    4. Blattman JN
    5. Antia R

    (2013) Estimating the diversity, completeness, and cross-reactivity of the T cell repertoire

    Frontiers in Immunology 4:485.

    https://doi.org/10.3389/fimmu.2013.00485

    • PubMed
    • Google Scholar

Author details

1. Peter C de Greef

   Theoretical Biology and Bioinformatics, Utrecht University, Utrecht, Netherlands

   Contribution

   Conceptualization, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing

   Contributed equally with

   Theres Oakes and Bram Gerritsen

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9802-629X

2. Theres Oakes

   Division of Infection and Immunity, University College London, London, United Kingdom

   Contribution

   Validation, Investigation, Writing - review and editing, Experimentation - T cell sorting and receptor sequencing

   Contributed equally with

   Peter C de Greef and Bram Gerritsen

   Competing interests

   No competing interests declared

3. Bram Gerritsen

   1. Theoretical Biology and Bioinformatics, Utrecht University, Utrecht, Netherlands
   2. Department of Pathology, Yale School of Medicine, New Haven, United States

   Contribution

   Conceptualization, Software, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing

   Contributed equally with

   Peter C de Greef and Theres Oakes

   Competing interests

   No competing interests declared

4. Mazlina Ismail

   Division of Infection and Immunity, University College London, London, United Kingdom

   Contribution

   Software, Formal analysis

   Competing interests

   No competing interests declared

5. James M Heather

   Division of Infection and Immunity, University College London, London, United Kingdom

   Contribution

   Software, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

6. Rutger Hermsen

   Theoretical Biology and Bioinformatics, Utrecht University, Utrecht, Netherlands

   Contribution

   Formal analysis, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4633-4877

7. Benjamin Chain

   Division of Infection and Immunity, University College London, London, United Kingdom

   Contribution

   Conceptualization, Supervision, Writing - review and editing

   For correspondence

   b.chain@ucl.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7417-3970

8. Rob J de Boer

   Theoretical Biology and Bioinformatics, Utrecht University, Utrecht, Netherlands

   Contribution

   Conceptualization, Supervision, Writing - review and editing

   For correspondence

   R.J.DeBoer@uu.nl

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2130-691X