Research ArticleImmunology

Deep Sequencing of the Human TCRγ and TCRβ Repertoires Suggests that TCRβ Rearranges After αβ and γδ T Cell Commitment

See allHide authors and affiliations

Science Translational Medicine  06 Jul 2011:
Vol. 3, Issue 90, pp. 90ra61
DOI: 10.1126/scitranslmed.3002536

Abstract

T lymphocytes respond to a broad array of pathogens with the combinatorial diversity of the T cell receptor (TCR). This adaptive response is possible because of the unique structure of the TCR, which is composed of two chains, either αβ or γδ, that undergo genetic rearrangement in the thymus. αβ and γδ T cells are functionally distinct within the host but are derived from a common multipotent precursor. The canonical model for T cell lineage commitment assumes that the γ, δ, and β chains rearrange before αβ or γδ T cell commitment. To test the standard model in humans, we used high-throughput sequencing to catalog millions of TCRγ and TCRβ chains from peripheral blood αβ and γδ T cells from three unrelated individuals. Almost all sampled αβ and γδ T cells had rearranged TCRγ sequences. Although sampled αβ T cells had a diverse repertoire of rearranged TCRβ chains, less than 4% of γδ T cells in peripheral blood had a rearranged TCRβ chain. Our data suggest that TCRγ rearranges in all T lymphocytes, consistent with TCRγ rearranging before T cell lineage commitment. However, rearrangement of the TCRβ locus appears to be restricted after T cell precursors commit to the αβ T cell lineage. Indeed, in T cell leukemias and lymphomas, TCRγ is almost always rearranged and TCRβ is only rearranged in a subset of cancers. Because high-throughput sequencing of TCRs is translated into the clinic for monitoring minimal residual for leukemia/lymphoma, our data suggest the sequencing target should be TCRγ.

Introduction

The ability of T lymphocytes to mount an immune response against a diverse array of pathogens is primarily conveyed by the amino acid sequence of the hypervariable complementarity-determining region 3 (CDR3) regions of the T cell receptor (TCR). The genes that encode the two primary types of TCRs, αβ and γδ, undergo somatic rearrangement during T cell development. TCRβ and TCRδ genes are assembled via recombination of variable (V), diversity (D), and joining (J) gene segments (VDJ recombination), and similarly, the TCRα and TCRγ genes undergo recombination of variable and joining gene segments (VJ recombination) to form productive αβ and γδ surface receptors.

The selection, function, and diversity of αβ T cells have been extensively studied. In the thymus (1), αβ T cells are both positively and negatively selected. Once selected, αβ T cells are activated when they recognize and bind non–self-peptides that are in a protein complex with human leukocyte antigen (HLA) and displayed on antigen-presenting cells (APCs). Because of the combinatorial diversity of αβ TCRs, the adaptive immune system has the potential to recognize a vast number of antigens. Estimates based on direct sequencing of TCRβ chains indicate that at any one time, an individual carries over 3 million unique TCRβ CDR3 chains (2).

γδ TCRs were discovered 4 years after αβ TCRs (3, 4) and were predicted to have a different role in T cell ontogeny based on significant differences in γδ TCR diversity, selection, and distribution (5). Although significant discoveries have advanced the field, important basic questions about γδ T cell activation and function remain. Unlike αβ TCRs, γδ TCRs bind self-antigens, leading many researchers to suggest that γδ T cells are not negatively selected in the thymus (6). Once these cells emigrate from the thymus, γδ TCRs can bind antigens independently of either an HLA scaffold or APCs (7, 8). However, the role of the APC in γδ TCR response is still unclear; mounting evidence indicates that APCs enhance the γδ T cell response (9). The distribution of γδ T cells also differs substantially from αβ T cells: Whereas αβ T cells are the predominant lymphocyte in the blood, γδ T cells are more common in mucosal tissue (1012).

Although only 5 to 10% of circulating T cells express γδ TCR, in primates, most of the circulating γδ T cells use the same Vγ and Vδ gene segments, Vγ9/Vδ2 (13, 14). After exposure to certain pathogens, including tuberculosis, leprosy, and malaria (13, 14), or tumor cells, including Daudi cells (15), T cells with Vγ9/Vδ2 chains expand rapidly. In some patients, this T cell population (Vγ9/Vδ2) expands to more than 50% of all circulating T cells (including αβ T cells) during bacterial infection (16). This immune response appears to be essential; the circulating γδ T cell population in AIDS patients is depleted of Vγ9/Vδ2 type T cells, and the reduction of these T cell types is associated with increased susceptibility to bacterial pathogens and lymphomas (17). Given that this γδ TCR chain type is the most common and responds to a variety of pathogens and tumors, circulating γδ T cells are often considered a bridge between the innate and adaptive immune system (16).

Both αβ and γδ T cells are derived from the same multipotent precursor cells. These two types of T cells are defined by the expression of TCRs on their surface, which are either γδ or αβ heterodimers. The TCR variable regions do not rearrange simultaneously during T cell development; TCRδ rearranges first, followed by TCRγ and TCRβ. The TCRα locus rearranges last, after the surface expression of both pre-TCRα and TCRβ chains (18). The effect of TCR rearrangement and expression on γδ and αβ T cell lineage commitment remains controversial (1, 1923). The canonical model proposes that relative expression level of αβ and γδ surface receptors drives precursor T cells to differentiate into either αβ or γδ T cells.

To study T cells in depth, we developed a method to sequence millions of TCRβ chains from genomic DNA in parallel from a single sample (24). This method was adapted to sequence TCRγ chains from genomic DNA and uses a multiplex set of polymerase chain reaction (PCR) primers to simultaneously amplify each possible combination of V and J. Here, we sequence the TCRγ chains from three unrelated healthy people and elucidate the properties of the TCRγ repertoire. To explore the timing of T cell differentiation into αβ and γδ T lineages, we sorted peripheral blood samples into αβ and γδ T cells and then sequenced TCRβ and TCRγ from both lineages. Although the canonical model (22, 24) suggests that TCRβ rearrangement precedes lineage fate decisions, we find that γδ T cells have few (or possibly no) rearranged TCRβ genes, suggesting that cell fate is decided before TCRβ rearrangement. TCRγ genes are rearranged in most or all αβ T cells, consistent with the evidence that TCRγ rearranges before cell fate is determined.

Results

Using high-throughput sequencing, we sequenced both TCRβ and TCRγ repertoires from sorted αβ and γδ T cell subsets collected from the blood of three healthy individuals. The TCRγ and TCRβ CDR3 chains were sequenced from genomic DNA extracted from about 150,000 T cells for each sample. We oversampled from the PCR pool and used redundancy to correct errors inherent to the PCR amplification and sequencing processes (25) (Table 1).

Table 1.

Sequenced TCRγ CDR3 chains.

View this table:

TCRγ sequences in γδ and αβ T cells

To assess the distribution of TCRγ CDR3 sequences in both γδ and αβ T cells, we amplified TCRγ chains from all samples using a multiplex PCR and sequenced them using the Illumina platform. A mean of 1.6 million TCRγ sequences were amplified from αβ T cell samples, and a mean of 3 million TCRγ sequences were amplified from γδ T cell samples (Table 1). The overall diversity was markedly different between γδ and αβ T cells: Threefold more unique TCRγ CDR3 sequences were amplified from αβ T cells than γδ T cells, with an average of 44,799 and 15,021, respectively (Table 1). The TCRγ repertoire of γδ T cells was dominated by one clone; this most frequent TCRγ CDR3 sequence represented >45% of all amplified TCRγ sequences (Fig. 1A). In all three samples, this most abundant TCRγ CDR3 sequence (5′-TGTGCCTTGTGGGAGGTGCAAGAGTTGGGCAAAA-3′) used Vγ9 and JγP gene segments and was nucleotide-identical across the CDR3 region between individuals (Fig. 2). Most of the remaining TCRγ sequences from γδ T cells also used the gene segments Vγ9 and JγP (Fig. 3A). The 10 most common unique TCRγ sequences account for 80% of the total γδ T cells. The remaining sequences, which are too rare to have been observed with lower-throughput technologies, encode a diverse set of sequences and differed substantially between individuals.

Fig. 1

Frequency of the 25 most common TCR sequences. For each sample, we plot the proportion of productive sequences accounted for by the 25 most numerous productive TCR sequences. (A) TCRγ chains amplified from γδ T cells and αβ T cells. (B) TCRβ chains amplified from αβ T cells.

Fig. 2

Shared nucleotide-identical TCRγ CDR3 sequences. Nine nucleotide-identical TCRγ CDR3 sequences amplified from γδ T cells are shared by all three individuals. For each shared sequence, the copy count detected for each individual is indicated on the y axis.

Fig. 3

Average V-J gene utilization of sequenced TCRγ and TCRβ sequences across three samples. (A to C) Average V-J utilization of gene segments in TCRγ CDR3 sequences amplified from γδ T cells (A), TCRγ CDR3 sequences amplified from αβ T cells (B), and TCRβ sequences amplified from αβ T cells (C).

The TCRγ chains amplified from αβ T cells were more diverse than those from γδ cells. In the αβ T cells, no single TCRγ sequence represented more than 7% of the overall TCR population (Fig. 1B) and no single Vγ or Jγ gene segment dominated the repertoire (Fig. 3B). Most TCRγ sequences used only three of the nine Vγ segments; however, the assay does capture all possible VJ combinations as detected by the VJ usage of unique out-of-frame TCRγ CDR3 sequences in αβ T cells (Fig. 4). Many TCRγ CDR3 sequences amplified from αβ T cells used Vγ10 gene segments, a gene segment predicted to have a nonconsensus donor splice site in the first exon, which results in the absence of splicing of the leader intron and termination of translation of the leader peptide (26, 27).

Fig. 4

Average copy number, by V-J pairing, of unique out-of-frame TCRγ sequences. Average copy number, by VJ gene segment usage, of unique out-of-frame TCRγ sequences amplified from αβ T cells.

Previously observed TCRγ sequences

We identified that seven of the TCRγ CDR3 sequences amplified in this study have been previously sequenced and reported in at least seven publications. More than 80% of in-frame peripheral blood mononuclear cell (PBMC) TCRγ amplified by this study use Vγ9JγP gene segments (Fig. 3A), consistent with previous observations (28, 29). Most of these are nucleotide-identical and have been reported in five scientific publications to be part of γδ heterodimers that bind biologically important epitopes (15, 2933). Six additional sequences that use Vγ9JγP gene segments found are also reported in one publication to bind biologically important epitopes (33), and one Vγ8 chain shared by two individuals was identified as a public TCR that responds to cytomegalovirus (CMV) (34).

Productive versus nonproductive TCRγ rearrangements in γδ T cells and αβ T cells

Because TCRs rearrange before thymic selection, random insertions and deletions in the CDR3 region would result in about one-third in-frame and two-thirds out-of-frame rearrangements. Thymic selection then skews this ratio, because a T cell must contain at least one productive TCR (which has to have both components of the heterodimer in-frame) to survive and leave the thymus. In our data, the observed fraction of unique in-frame TCRγ sequences is 53.5 ± 2.6% in γδ T cells and 31% in αβ T cells (Table 1). In αβ T cells, the observed fraction of unique in-frame TCRβ sequences is 82%, with 18% of TCRβ CDR3 sequences being out-of-frame (Table 2).

Table 2.

Sequenced TCRβ CDR3 chains.

View this table:

TCRβ sequences in γδ T cells and αβ T cells

To assess the presence and diversity of TCRβ CDR3 chains in both γδ and αβ T cells, we amplified the TCRβ chains at high throughput for all samples. More than 1.4 million TCRβ CDR3 chains were sequenced from about 250,000 haploid genomes. Within this population, there were 95,648 unique TCRβ CDR3 sequences (Table 2). The TCRβ chains amplified from αβ T cells represented a relatively diverse population, with varied V(D)J usage (Fig. 3C), and the most common TCRβ CDR3 sequence representing just 0.3% of the sequenced sample (Fig. 1B). Just 2582 unique TCRβ CDR3 chains were sequenced from a corresponding population of 250,000 haploid genomes extracted from γδ T cells. This number of TCRβ chains is within the expected range for contamination of the γδ cell population with αβ T cells during cell sorting (<4%).

Discussion

High-throughput sequencing technology allows analysis of the TCRγ and TCRβ repertoire at an unprecedented depth. Using this technique, we found that both circulating αβ and γδ T cells carried rearranged TCRγ chains. Indeed, the ratio of productive to nonproductive rearrangements observed suggests that, at least in γδ T cells, both TCRγ alleles rearrange before thymic selection. However, most γδ T cells did not carry a rearranged TCRβ chain, consistent with TCRβ rearranging after T cell lineage bifurcation. We also observed that although the TCRγ repertoire of αβ T cells contains no predominant clone, most γδ T cells sampled from peripheral blood used an identical germline TCRγ rearrangement: Vγ9JγP gene segments that lack insertions at the V-J junction, which is identical to a TCRγ receptor with well-studied binding affinities (30). However, 20% of the circulating γδ T cells had a diverse TCRγ repertoire that was nearly undetectable by older methods (29). Although a few of the sequences from the remainder of the γδ T cell repertoire have known functions, most have not been previously studied or observed (33, 34). This distribution of TCRγ sequences in blood suggests that γδ T cells share qualities with both the innate and adaptive immune system, with universal sequences shared across individuals likely to perform innate functions, whereas the diverse background repertoire may play a more adaptive role.

Most models of T cell lineage decision assume that TCRδ, TCRγ, and TCRβ rearrange before T cell fate commitment (19, 22, 35). However, most (>90%) γδ T cells included in this study lacked a rearranged TCRβ chain. These results are consistent with a previous study that provided evidence that TCRβ chains rearrange after both TCRδ and TCRγ (18) and that TCRβ chains are not rearranged in γδ T cells (36). These results indicate that TCRβ rearranges after T cell lineage bifurcation and that successful TCR rearrangement is an unlikely lineage commitment signal.

However, our data indicate that both TCRγ alleles rearrange in most T cells, in agreement with previously published data (37). Other evidence also suggests that TCRγ rearranges before cell fate determination and, therefore, before thymic selection (18, 19). Our data are consistent with both TCRγ alleles rearranging before selection. If TCRs rearrange before thymic selection, random insertions and deletions in the CDR3 region should result in about one-third in-frame and two-thirds out-of-frame rearrangements. In γδ T cells, thymic selection then would skew this ratio, because a T cell must contain at least one productive TCR to survive and leave the thymus. Therefore, under this model, we expect to observe T cells with one in-frame TCRγ rearrangement with the other allele either in-frame or out-of-frame. With an expectation of two-thirds of TCRγ rearrangements being out-of-frame, 2/3 × 2/3 = 4/9 of all cells should fail to generate a productive TCRγ chain. Of the remaining five-ninths, four-ninths are estimated to have one in-frame and one out-of-frame allele. The final one-ninth should have two in-frame alleles. Therefore, the ratio of in-frame to out-of-frame TCRγ chains in a pool of γδ T cells is predicted to be 3:2. In our data, the observed fraction of in-frame TCRγ is 53.5 ± 2.6. These data suggest low-level negative selection, in contrast with evidence that γδ T cells do not undergo antigen selection within the thymus [for a review, see (6, 38)]. If, on the other hand, the second allele only rearranges if the first allele is nonproductive, we would expect an even higher ratio of in-frame to out-of-frame TCRγ genes. An estimated one-third of the cells should have the first allele in-frame. The other two-thirds should then rearrange their second allele. Thus, one-third of the cells would have one in-frame allele and 1/3 × 2/3 = 2/9 would have one in-frame and one out-of-frame allele. The remaining cells would have two out-of-frame alleles and would be expected to die. The result is a ratio of 2:1 of in-frame to out-of-frame. This is a far worse match to the observed data, and thus, our data are consistent with both TCRγ alleles rearranging before thymic selection.

The most common TCRγ sequence does not have a corresponding in-frame or out-of-frame TCRγ sequence with similar copy count. This, combined with the observation that the dominant clone lacks nontemplated insertions, suggests that the dominant clone developed under different circumstances than the underlying diverse array of γδ T cells. The lack of a corresponding sequence of similar copy count suggests that either recombination was restricted to one allele, or both alleles were restricted to the same rearrangement, and the lack of insertions suggests that the terminal deoxynucleotidyl transferase (TdT) enzyme is not present during the rearrangement. These observations suggest indirectly that this dominant clone is formed very early in human development, likely a prenatal rearrangement before gestational week 20 when TdT is first expressed in the fetal thymus. This hypothesis has precedent: During mouse fetal development, γδ T cells with determinate TCRs develop and home to specific tissues. Although some work indicates that this TCRγ chain is not common in neonates (39), work by McVay and Carding found that this TCRγ chain was present in fetal livers at 11 weeks (40). Later work indicates that this population of γδ T cells develops extrathymically in the fetal liver and primitive gut (41). These observations support the hypothesis that the dominant clone may play an innate role.

Several biologically important epitopes have been reported to bind γδ heterodimers formed from TCRγ chains amplified by this study. The most abundant TCRγ sequence chain in all three individuals is identical to a TCRγ CDR3 chain reported by Wang et al. to be part of a γδ heterodimer that reacts to a variety of nonpeptide prenyl phosphates, including (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP) and isopentenyl phosphate (IPP) (30). Nonpeptide prenyl phosphates are metabolites common to all organisms and are associated with Vγ9/Vδ2 T cell expansion (31). HMBPP, a metabolite of the 2-C-methyl-d-erythritol-4 phosphate pathway for isoprenoid biosynthesis, is common in Eubacteria and some protozoa. This same TCRγ sequence is found in the TCRs of T cells that respond to Daudi Burkitt lymphoma cells (15, 32) and Molt-4 tumor cells (29). In addition to the dominant sequence, multiple sequences amplified in this study were previously asserted by Chen et al. to bind epitopes displayed on the surface of SKOV3 tumor cells and hepatitis B virus (HBV)–infected cells. These epitopes include human mutS homolog 2 (hMSH2) and heat shock protein (HSP) 60 (33). Finally, a chain that uses Vγ8, CATWDTTG, which was present in two sampled individuals (samples B and C), is a public TCRγ chain that reacts with CMV and is expanded in infants infected with CMV (34). Because the frequency of Vγ9/Vδ2 T cells is positively correlated with natural repression of the HIV virus (42) and rapid expansion of this population is associated with rapid clearance of tuberculosis in rhesus monkeys (13), the frequency of the major V9JP clone may be a useful biomarker.

For the analysis presented, we limited each sample to 250,000 haploid genomes because of the relatively small fraction of γδ T cells in human blood. However, the assay presented has the potential to sequence many millions of TCRγ chains. This depth of coverage readily allows detection of a TCR clone at one part in 100,000 or even lower. All T cells appear to rearrange TCRγ, including αβ T cells, so this assay allows us to track clones with unprecedented sensitivity and significant potential for clinical applications. Given that the most common TCRγ clone has likely important biological functions, the clone’s frequency may be an important indicator of immune health and/or immune response. In addition, given that TCRγ rearranges in both γδ and αβ T cells, TCRγ clones can be useful for tracking minimal residual disease in blood cancers. With the technology we used, we can follow the clonal expansion and contraction for hundreds of thousands of T cell clones over time.

Materials and Methods

Samples

We collected 40 ml of blood from three, one male and two female, healthy unrelated adults. PBMC was isolated from whole blood with a Ficoll Histoplaque gradient. After separation, the total number of cells was estimated based on the number of cells counted in 1 mm3 of sample.

Sorting αβ and γδ T cells

For each individual, isolated PBMCs were separated into two fractions of about 20 million cells (18 to 25 million). Cells were labeled with either anti-γδ TCR antibodies conjugated with hapten or anti-αβ TCR antibodies conjugated with phycoerythrin (PE). These labeled cells were treated with either anti-hapten antibodies or anti-PE antibodies coupled to microbeads (Miltenyi Biotech). Microbead-labeled cells were positively selected with Miltenyi cell separator columns. To achieve maximum purity (>96%) for rare cells, we passed anti-γδ TCR-labeled cells over two purification columns. The number of recovered cells was estimated from the number of cells counted in a 1-mm3 subset. Between 5 million and 10 million αβ T cells and 200,000 and 1 million γδ T cells were recovered for each sample. DNA was extracted from circulating γδ T cells and circulating αβ T cells with Qiagen Maxi DNA isolation kits (Qiagen Inc.).

Sequencing CDR3 regions

TCRG and TCRB CDR3 regions were amplified and sequenced from 800 ng of extracted DNA by Adaptive TCR Corp. with ImmunoSeq. Amplification and sequencing of TCRβ CDR3 regions were carried out as previously described by Robins et al. (25). To amplify the TCRγ template, we designed a muliplex PCR method to amplify all possible rearranged genomic TCRγ CDR3 sequences. The method uses nine forward primers specific to all Vγ gene segments and five reverse primers specific to all Jγ gene segments, including Vγ and Jγ gene segments that are predicted open reading frames (IMGT). Amplification and sequencing protocols were modified for TCRγ from methods designed for amplifying TCRβ (25).

Identifying CDR3 sequences

Both the TCRβ and TCRγ CDR3 region was delineated according to the definition established by the International ImMunoGeneTics collaboration (43). Sequences that did not match CDR3 sequences were removed from the analysis. A standard algorithm was used to identify which V, D, and J segments contributed to each TCRβ CDR3 sequence and which V and J segments contributed to each TCRγ CDR3 sequence (43). Rearranged CDR3 sequences were classified as nonproductive if they included insertions or deletions that resulted in sequences with frameshifts or premature stop codons.

Identifying productive rearranged TCRγ and TCRβ in T cells

The number of productive and nonproductive rearranged TCRβ and TCRγ in sampled circulating αβ and γδ T cells was calculated for each individual. CDR3 sequences, in which translating a functional TCR is impossible, due to insertions or deletions that cause frameshifts or stop codons, are considered nonproductive rearrangements, whereas CDR3 sequences, from which a functional TCR is possible, are considered productive rearrangements.

Literature search for functional data on TCRγ

We searched PubMed for articles that matched the query “(T cell receptor[Title/Abstract] OR TCR[Title/Abstract]) AND peptide[Title/Abstract].” This resulted in 4899 records. We then used a Python script to download the full-text PDF of these articles, which lead to at least one PDF file for 3520 articles. These files were converted to ASCII text with the tool pdftotext from the xpdf package (http://www.foolabs.com/xpdf). We ran Peptide::Peptide (44) at default settings over all text files to find the peptide sequences in them. Peptide sequences were searched with BLAT (45) against the CDR3 sequences, and the results were inspected manually.

Footnotes

    References and Notes

    1. Acknowledgments: H.R. thanks Y.-H. Chien for useful comments and suggestions. Funding: This study was supported by a grant from the NIH (AI081860). Author contributions: A.M.S., C.S.C., and H.R. designed the study. J.A. recruited subjects, acquired written informed consent, and arranged collection of blood samples. R.J.L. and A.M.S. sorted T cells, extracted DNA, and sequenced TCR and TCRβ chains. M.H. applied a computational tool of his own design to search the literature for TCR sequences matching those observed in this study. A.M.S., C.D., and H.R. analyzed the results, created the figures, and wrote the manuscript. All authors discussed the results and commented on the paper. Competing interests: A.M.S., C.D., R.J.L., and J.A. are employed by Adaptive TCR Corp., which commercialized, for research purposes, the TCRβ and TCRγ sequencing assays used in this manuscript under the brand name ImmunoSeq. C.S.C. and H.R. are co-founders of Adaptive TCR Corp. and own stock. H.R. and C.S.C. are inventors on the filed patent no. 003.001P2 titled “A method to measure immune diversity.” The other authors declare they have no competing interests.
    • Citation: A. M. Sherwood, C. Desmarais, R. J. Livingston, J. Andriesen, M. Haussler, C. S. Carlson, H. Robins, Deep Sequencing of the Human TCRγ and TCRβ Repertoires Suggests that TCRβ Rearranges After αβ and γδ T Cell Commitment. Sci. Transl. Med. 3, 90ra61 (2011).

    View Abstract

    Stay Connected to Science Translational Medicine

    Navigate This Article