Analyses of Vγ-Jγ rearrangements producing the most commonly expressed TCRγ chains in over 200 γδ TCR+ thymocytes showed that assembly of TCRγ V-region genes display properties of allelic exclusion. Moreover, introduction of functionally rearranged TCRγ and δ transgenes results in a profound inhibition of endogenous TCRγ rearrangements in progenitor cells. The extent of TCRγ rearrangements in these cells is best explained by a model in which initiation of TCRγ rearrangements at both alleles is asymmetric, occurs at different frequencies depending on the V or J segments involved, and is terminated upon production of a functional γδ TCR. Approximately 10% of the cells studied contained two functional TCRγ chains involving different V and Jγ gene segments, thus defining a certain degree of isotypic inclusion. However, these cells are isotypically excluded at the level of cell surface expression possibly due to pairing restrictions between different TCRγ and δ chains.
Exons encoding the variable regions of lymphocyte receptors for Ag are assembled during lymphocyte ontogeny from clusters of V, (D), and J segments in a process known as V(D)J recombination (1). T cell precursors can rearrange up to four different loci (α, β, γ, and δ) and express either of two different TCR (αβ and γδ), thus defining two T lymphocyte populations (the αβ and the γδ T cell populations). Most TCRδ, TCRγ, and TCRβ rearrangements take place in a population of thymocytes known as pre-T cells which is characterized by low expression of CD44, high expression of CD25, and lack of expression of the CD4 or CD8 markers (2, 3, 4), whereas most Vα to Jα rearrangements occur later in development in cells coexpressing the CD4 and CD8 molecules (5, 6). Functionally rearranged TCRβ chains associate with the nonrearranging surrogate α-chain (pre-Tα) forming the pre-TCR (7). Only cells expressing the pre-TCR efficiently traverse a developmental checkpoint usually referred to as TCRβ selection (8) and further differentiate into the αβ T cell pathway (9). Such differentiation is accompanied by down-regulation of RAG-1/2 expression, silencing of TCRγ transcription (thus avoiding expression of a γδ TCR in αβ lineage cells), acquisition of the CD4 and CD8 markers and up-regulation of RAG-1/2 expression allowing the initiation of rearrangements at the TCRα locus (10). In contrast, no surrogate chains for γδ lineage cells have been identified and evidence suggest that simultaneous expression of functional TCRγ and TCRδ chains that can efficiently pair are required for differentiation along the γδ T cell lineage (11, 12, 13).
Because lymphocytes are diploid cells, the recombination process could, theoretically, result in the production of two productively rearranged TCR alleles and, therefore, in the expression at the cell surface of more than one TCR specificity. Analyses of a relatively large number of αβ T cells and clones have shown that, as a rule, mature T cells contain only one productive TCRβ rearrangement, whereas the second allele is either nonproductively rearranged or contain their Vβ gene segments in germline configuration (14, 15, 16). In contrast, most mature αβ T cells contain two TCRα rearrangements and 20–30% of them bear two productively rearranged TCRα chains (15, 16). Thus, assembly of TCRβ chain V region is regulated in a way that allows only one of the two alleles to be expressed in each cell (a phenomenon usually referred to as allelic exclusion or genotypic allelic exclusion) whereas that of TCRα chain V-region genes is regulated differently permitting allelic inclusion. Posttranslational mechanisms appear to limit T cells carrying two functional TCRα rearrangements to the expression of a single TCRα chain at the cell surface (17) (a phenomenon usually referred to as phenotypic allelic exclusion). Together, these mechanisms ensure that most αβ T cells express a unique TCR at the cell surface that may be essential for the generation of specific immune responses and for effective tolerance mechanism based on the elimination or inactivation of developing cells reacting with self-Ags.
To explain allelic exclusion at the TCRβ locus it has been postulated that the product of a successful Vβ to DJβ recombination prevents further rearrangements at the same locus via a feedback mechanism (15), similar to what was proposed earlier for the Ig H chain locus in developing B cells (18, 19). Together with the postulate that rearrangement starts initially in one of the two alleles, this hypothesis correctly predicts the extent of TCRβ rearrangement found in mature T cells (15). Also in support of a feedback mechanism are data showing that lymphocytes from TCRβ transgenic (Tg)3 mice that express the transgene early in development are almost completely devoid of endogenous Vβ to DJβ rearrangements (20). More recent experiments analyzing the proportion of progenitor cells containing two completed TCRβ gene rearrangements in normal, pre-Tα-deficient, or SLP-76-deficient mice have indicated an essential role for signals mediated by the pre-TCR in TCRβ allelic exclusion (16, 21).
Analyses of TCRδ rearrangements in a panel of T cell hybridomas derived from splenic γδ T cells showed a high percentage of cells carrying two productively rearranged TCRδ chains, indicating that assembly of TCRδ V-region genes, like that of TCRα V-region genes, is not regulated in the context of allelic exclusion (22). In support of this notion, Tg mice carrying a functionally rearranged TCRδ chain showed no evident inhibition of rearrangements at the endogenous TCRδ locus (23). Whether posttranslational mechanisms limit the number of TCRδ chains that may be expressed at the cell surface in γδ T cells is not known, although pairing preferences observed between individual TCRγ and TCRδ chains may contribute to the monoallelic expression of TCRδ chains at the cell surface (24, 25).
Whether assembly of TCRγ V-region genes is regulated in a context of allelic exclusion is a matter of debate. Analyses of human γδ T cell lines have shown that a variable proportion (1–6%) of cells coexpress two functional TCRγ chains at the cell surface (26) and this has been taken to imply that assembly of TCRγ genes is not subjected to allelic exclusion mechanisms. In contrast, analyses of TCRγ rearrangements in mouse γδ thymocyte clones showed that most γδ thymocytes contained only one functionally rearranged allele for most of the TCRγ isotypes (13), which suggests that assembly of TCRγ genes in the mouse may be regulated in the context of allelic exclusion.
To understand the mechanisms regulating assembly of TCRγ V-region genes one needs to take into consideration the genomic structure of the TCRγ locus, which differs greatly between humans and mice. Thus, whereas a human progenitor cell can produce a maximum of two functionally rearranged TCRγ chains, a mouse progenitor cell could, theoretically, produce six to eight functional TCRγ chains, depending on the mouse strain (13, 27). This is due to the fact that the mouse TCRγ locus is organized in four clusters of Vγ, Jγ, and Cγ regions containing seven Vγ gene segments, four Jγ gene segments, and four Cγ regions (hereafter referred to as V1 to V7, J1 to J4, and C1 to C4, respectively, according to the nomenclature of Heilig and Tonegawa; Ref. 28). Each cluster contains a Cγ region, linked to a single Jγ element and one to four Vγ gene segments, which rearrange preferentially to the Jγ segment present in the same cluster. Thus, if expression of more than one TCRγ chain at the cell surface of γδ T cells is to be prevented, TCRγ chains must be regulated not only to avoid simultaneous expression of the two alleles, but also to avoid expression of different TCRγ isotypes. Here we show that assembly of mouse TCRγ V-region genes is regulated in a manner compatible with allelic exclusion mediated by a γδ TCR that can be expressed at the cell surface. Moreover, isotypic exclusion is regulated genotypically, at least in part, as evidenced by the fact that different Vγ and Jγ gene segments display distinct probabilities to participate in a recombination reaction and by the different frequencies at which a rearrangement involving particular Vγ and Jγ gene segments produce a functional chain. Further phenotypic isotypic exclusion results from the different capacity of TCRγ isotypes to pair with a more or less restricted pool of TCRδ chains.
Materials and Methods
C57BL/6JIco (B6) mice were obtained from Iffa-Credo. B6 mice Tg for a functionally rearranged Vγ1Jγ4Cγ4 (Tg-γ; Ref. 29), B6 mice Tg for the same Vγ chain together with a functionally rearranged Vδ6Dδ2Jδ1 chain (Tg-γδ; Ref. 30), and B6 TCRδ-deficient mice (31) were maintained in our animal facilities. B6 mice deficient in the TCRβ enhancer (Eβ−/−; Ref. 32). were obtained from Dr. P. Ferrier (Centre d’Immunologie de Marseille-Luminy, Marseille, France) and also maintained in our animal facilities. All animals were used between 6 and 8 wk of age unless otherwise indicated.
Abs, flow cytometry, and cell sorting
Abs, FACS analyses, and cell sorting were performed as described (13).
Cloning of individually sorted γδ thymocytes was performed as described (13). Sorted V1+ and V4+ cells were cloned in plates coated with anti-V1 mAb or anti-V4 mAb, respectively, thus providing a second control for the TCRγ chain expressed at the cell surface.
Single cell and single clone PCRs, cloning, and sequencing
The single clone PCR to detect the rearrangement status of the TCRγ locus have been described in detail (13). The protocol was identical except that only primers specific for the V1 and V4 V regions were used, and the germline status was only analyzed for the V1 and the V4 regions. Reverse primers for the germline V4 region were: VG4GLext CTGAACAGCAGGTGGTTGCC and VG4GLint CCAAGCTAAGAAGGATGTGG. Single cell PCR was performed as described (33) using the VG1: CCGGCAAAAAGCAAAAAAGTT and JG4: GCAAATATCTTGACCCATGA primers
Most γδ T cells express one Ag receptor at the cell surface
To quantitate the proportion of γδ thymocytes expressing two different TCRγ or TCRδ chains at the cell surface we used the available anti-Vγ and anti-Vδ mAbs. Cells staining simultaneously with an anti-Vγ1 and a mixture of anti-Vγ4 and anti-Vγ7 mAbs are rare, representing <1% of all γδ thymocytes in B6 mice (Fig. 1 A). Because together, these Abs recognize >90% of the γδ thymocytes in this strain, these data demonstrate that the vast majority of γδ T cells express only one TCRγ isotype at the cell surface.
In contrast, double staining of the same γδ T cell population with Abs specific for the Vδ4 chain on one hand and a mixture of Abs recognizing Vδ5 and three of the four Vδ6 gene segment present in B6 mice on the other shows a small but reproducible population of double-stained cells, representing 2–4% of the total γδ thymocytes (Fig. 1,B). Of those, ∼20% (i.e., 0.8% of the total γδ thymocyte population) do express two different TCRδ chains at the cell surface as evidenced by sorting and re-staining of the putative dual-expressor cells (Fig. 1 C). Because about one-third of the γδ thymocytes bear TCRδ chains that are not recognized by the mAbs used and because cells expressing two different TCRδ chains containing the same V-gene segment are not identified by these analyses, these data allow us to estimate to ∼3% the maximum frequency of γδ thymocytes expressing two different TCRδ chain at the cell surface.
Extent of allelic and isotypic inclusion of the functionally most relevant V4 and V1 gene segments
To quantitate the extent of genotypic allelic and isotypic inclusion at the TCRγ locus and to gain a better understanding on the mechanisms controlling assembly of TCRγ V-region genes, we analyzed the rearrangement status of the V1 and V4 genes in progenies of a total of 234 γδ thymocytes (141 V1+ and 93 V4+) using a previously described two-step PCR (Fig. 2 and Ref. 13). Cells bearing at the cell surface either V1 or V4 chains represent >80% of the γδ thymocytes in B6 mice (25). For simplicity we only analyzed V1 to J4 and V4 to J1 rearrangements, as well as the germline status of the V1 and V4 gene segments. This was based on our previous observation that in adult γδ thymocytes >98% of the rearrangements involving the J4 gene segment also involved the V1 gene segment and >90% of the rearrangements involving the J1 gene segment also involved the V4 gene segment (13). These analyses also showed that about half of the V1+ and V4+ thymocytes bear functional V2-J2 rearrangements and the mechanisms precluding detectable expression of V2 chains at the cell surface of these cells have been previously discussed (13). The PCR amplification products of the rearrangements were sequenced to determine their functionality and a summary of the results is shown in Fig. 2.
Approximately 1% of the cells contained two functionally rearranged J4 or J1 segments (1 of 141 V1+ clones (clone 66) and 1 of 93 V4+ clones (clone 60) contained the two J4 or the two J1 segments productively rearranged, respectively), indicating that the vast majority of γδ thymocytes contain only one functional rearrangement at the corresponding alleles. Contrasting with the paucity of allelically included cells, 18 clones (i.e., 8%; clones 18–24 and 70–77) were found to harbor two productive rearrangements involving different J segments. Interestingly, all these clones were found in the V1+ population showing an apparent dominance of V1 chains over V4 chains for their expression at the cell surface as part of a γδ TCR. As discussed previously (13), such apparent dominance may result, at least in part, from the fact that V4 chains pair with a more restricted pool of TCRδ chains than V1 chains.
Expression of a functionally rearranged γδ TCR inhibits rearrangements at the endogenous TCRγ locus
The low frequency of allelically included cells may suggest that the products of functionally rearranged V1-J4 or V4-J1 gene segments are involved in the termination of TCRγ rearrangements in progenitor cells. They could do so by participating in a signaling complex either alone, together with a putative surrogate TCRδ chain, or together with a TCRδ chain with which they can form a γδ TCR. In either case, it will be expected that introduction of a functionally rearranged TCRγ chain in progenitor cells will result in a certain degree of inhibition of TCRγ rearrangements at the endogenous loci. Moreover, comparison of the extent of inhibition resulting from ectopic expression of either a functional TCRγ chain alone or together with a functional TCRδ chain will be indicative of whether a surrogate receptor containing a TCRγ chain or a complete γδ TCR is involved in this process.
We have previously produced mice Tg for a functionally rearranged V1J4C4 chain (Tg-γ) or for the same TCR-γ chain together with a functionally rearranged Vδ6DδJδ chain (Tg-γδ) (29, 30). Because among all TCRγ chains V1 chains are known to pair with the largest number of TCRδ chains (24, 25), these animals are most suitable to analyze this issue. Because a large majority of V1+ thymocytes contain at least one V4-J1 rearrangement (Fig. 2) we analyzed and compared the occurrence of this type of rearrangement in V1+ thymocytes isolated from wild-type, Tg-γ and Tg-γδ mice. Introduction of a functionally rearranged V1 chain reduced the frequency of endogenous V4-J1 rearrangements by ∼3-fold (Fig. 3). A further 3-fold reduction was observed when a functional TCRδ chain was coexpressed with the same TCRγ chain, resulting in a 84% inhibition of endogenous V4-J1 rearrangements in γδ T cells from Tg-γδ mice (Fig. 3). This extent of inhibition of endogenous TCRγ rearrangements is of a similar magnitude to that previously observed in endogenous Vβ to DJβ rearrangements as the result of early expression of a functionally rearranged TCRβ transgene (34). These data strongly suggest that both TCRγ and TCRδ chains participate in the regulation of TCRγ V-region gene assembling. The inhibition observed following expression of a Tg TCRγ chain alone could reflect the increased probability of progenitor cells in Tg-γ mice to express a γδ TCR.
γδ T cell progenitors do not attempt all possible TCRγ rearrangements
The low frequency (12.8%) of V1+ cells harboring rearrangements at the two J4 alleles (Fig. 2) indicates that γδ T cell progenitors do not attempt all possible TCRγ rearrangements. This could result from the low probability at which the V1 and J4 gene segments rearrange in progenitor cells in a given period of time (13). Also, a receptor different from the γδ TCR may signal back to stop further rearrangements at the TCRγ locus. An obvious candidate will be the pre-TCR, which may be expressed in γδ T cell precursors that produce a functional TCRβ chain and that is known to be essential in the process of allelic exclusion at the TCRβ locus (16). To investigate whether signals through the pre-TCR inhibit rearrangements at the TCRγ locus and, if so, to examine the relevance of such mechanism in αβ and γδ T cell fate decisions, we quantitated, by single cell PCR, the frequency of cells containing two V1-J4 rearrangements in sorted V1+ cells isolated from wild-type mice and from mice deficient in the TCRβ enhancer (Eβ−/−). This mutation completely inhibits rearrangement at the TCR-β locus and, therefore, γδ T cells in these mice develop in the absence of putative pre-TCR mediated signals. Moreover, comparison between these two strains of mice also eliminates competition between the TCRγ and the TCRβ loci for components of the V(D)J recombinase as an explanation for possible differences.
Nine of 71 (12.7%) V1+ cells from wild-type mice contained two V1-J4 rearrangements (Table I). This frequency is nearly identical to that found in the V1+ clones shown in Fig. 2, indicating that the culture step did not bias the experimental sample. In contrast, 18 of 78 (23.1%) of the V1+ cells isolated from Eβ−/− mice harbored two V1-J4 rearrangements. The differences between the two V1+ populations are statistically significant (p = 0.04 according to a Fischer’s exact test), demonstrating a quantitatively minor but evident role of the pre-TCR in the termination of rearrangements at the TCRγ locus in T cell progenitors.
|Strain .||Cell Analyzed .||Cells with Two V1-J4 Rearrangementsa .|
|WT||V1+ clones||17 of 141 (12.06%)|
|WT||V1+ single cells||9 of 71 (12.68%)|
|Eβ−/−||V1+ single cells||18 of 78 (23.08%)|
|Strain .||Cell Analyzed .||Cells with Two V1-J4 Rearrangementsa .|
|WT||V1+ clones||17 of 141 (12.06%)|
|WT||V1+ single cells||9 of 71 (12.68%)|
|Eβ−/−||V1+ single cells||18 of 78 (23.08%)|
Differences between WT and Eβ−/− cells are statistically significant according to the Fisher’s exact test (p = 0.043).
The extent of TCRγ rearrangements found in γδ T cells is best explained by a model in which initiation of TCRγ rearrangements at both alleles is asymetric and is terminated upon production of a functional γδ TCR
Whereas the feedback mechanism certainly contributes to the inhibition of rearrangements at the second allele, it alone cannot explain allelic exclusion unless the rearrangement step is generally very inefficient (35). This is due to the fact that the product of a rearrangement must be tested for its ability to encode a functional chain. Attempting to calculate the rates of different TCRγ rearrangements and to investigate whether other mechanisms regulate TCRγ V-gene assembly, we constructed probabilistic models and compared the extent of V1 to J4 and V4 to J1 rearrangements found in V1+ and V4+ thymocytes with those predicted by the models (36). Models that only take into consideration the time given to a progenitor cell to rearrange and a feedback mechanism to stop further rearrangements, although compatible with the experimental data, only gave marginal statistical significance, strongly suggesting that additional mechanisms regulate TCRγ V-gene assembly. Interestingly, a very accurate match between the experimental data and the expected values was observed when a differential accessibility of the two TCRγ alleles was imposed in the model, assuming that after a period during which only one allele is accessible both alleles become accessible (36). Thus, on statistical basis, the experimental data is compatible with the hypothesis that assembly of TCRγ V-region genes is regulated by the products of functional TCRγ rearrangements and the experimental data is better explain if the two alleles do not become accessible to the V(D)J recombinase simultaneously. Importantly, the models make a number of predictions that can be experimentally analyzed and used to test their robustness. Thus, the best fitting model predicts that a progenitor cell will rearrange the V4 gene 12 times more often than the V1 gene, what compares well with the value of 11 found experimentally (13). Moreover, the model also predicts a ratio betweenV4+ and V1+ cells of 1.4, which also compares well with the ratio of 1.67 obtained by staining thymocytes with Vγ-specific mAbs ex vivo. Finally, the model predicts that ∼0.5% of the V1+ cells will contain two functional V1-J4 rearrangements, compatible with the 0.7% found in the clones analyzed here (Fig. 2). Interestingly, the model also predicts that ∼44% of the progenitor cells will produce functional V1 or V4 chains. Given that a few more progenitor cells may produce functional V2 or V7 chains, the expected value of γδ T cell precursors that will produce a functional TCRγ chain may not differ significantly from the maximum of 56% of the αβ T cell precursors expected to produce a functional TCRβ chain (15).
The number of γδ thymocytes correlates with the probability of producing a selectable γδ TCR
A proper comparison of the models with the data requires that newly differentiated γδ T cells do not expand in the thymus or, if they do, that this expansion is independent of the TCRγ isotype they express. Although it is generally believed that formation of a selectable γδ TCR in γδ-lineage progenitor cells induces maturation of the cells in a process that involves little or no proliferation (37), it was important to ascertain that this was indeed the case. If maturation of γδ thymocytes occurs without significant expansion, it would be expected that the number of γδ T cells present in the thymus is a direct function of the probability of assembling a selectable γδ TCR. Therefore, we quantified and compared γδ thymocyte numbers in mice carrying only one functional allele of the Cδ chain and in wild-type littermates. As shown in Fig. 4, TCRδ+/− mice contained about half the number of γδ thymocytes than their wild-type littermates, demonstrating that the number of γδ thymocytes is a linear function of the probability of a progenitor cell to produce a functional γδ TCR, and suggesting that no homeostatic mechanism regulates γδ T cell numbers in the thymus.
The data presented in this manuscript indicate that the vast majority of mature γδ thymocytes exhibit properties of allelic exclusion at the TCRγ locus at the level of V-region gene assembly, at least in what concerns the most commonly used J1 and J4 regions. Such allelic exclusion is achieved, in part, through a feedback mechanism mediated by the products of the functionally rearranged TCRγ and TCRδ chains and requires a proper interaction between both chains to form a functional γδ TCR. The existence of a signaling mechanism is strongly suggested by the inhibition of endogenous TCRγ rearrangements observed in Tg-γδ mice. However, the possibility that a rapid development of T cells may not provide adequate time for RAG mediated recombination in the Tg model cannot be formally excluded at the moment. Independent evidence consistent with the existence of a feedback mechanism came from the comparison of the extent of TCRγ rearrangements found in γδ T cells with those predicted by statistical models (36).
For signaling through a γδ TCR to be effective in ensuring allelic exclusion at the TCRγ locus it is required that, in general, TCRδ rearrangements precede TCRγ rearrangements in progenitor cells. Although complete rearrangements at both loci are found at the same developmental stage (3, 4), indirect evidence suggests that this may indeed be the case. Thus, cells with rearrangements at the TCRδ locus and with the TCRγ locus in germline configuration were easily found in fetal thymocyte hybridomas (38). Moreover, analyses of thymocytes from SCID mice (39) showed that rearrangements at the TCRδ locus exceed by far those occurring at the TCRγ or TCRβ loci (40). SCID mice produce dsDNA breaks mediated by the RAG proteins at the initiation of the recombination process at normal levels (41) but their defect in the DNA-dependent protein kinase catalytic subunit protein, which is involved in the resolution of coding ends, precludes the formation of V(D)J coding joints at appreciable frequency. Therefore, the SCID mice data provide evidence for a different accessibility of the TCRδ and TCRγ loci to participate in a recombination reaction in early progenitor cells. It should be pointed out that we are not implying a strict temporal order in the rearrangements at both loci, but just an increased probability of a TCRδ chain to be rearranged (or expressed) before a TCRγ chain in progenitor cells. Absence of a strict temporal order in the rearrangement at both loci is suggested by the isolation of fetal hybridomas containing rearrangements at the TCRγ locus and the TCRδ locus in germline configuration (38) and by our data showing that ectopic expression of a functionally rearranged TCRγ chain only partially inhibits endogenous rearrangements at the TCRγ locus.
Although the feedback mechanism certainly contributes to the inhibition of rearrangements at the second allele, it alone cannot explain allelic exclusion unless the rearrangement step is generally very inefficient (35). This is due to the fact that the product of a rearrangement must be tested for its ability to encode a functional chain. The actual frequencies of B and T lymphocytes containing, respectively, V(D)J rearrangements at the two IgH or TCRβ alleles were shown to be very close to those expected if rearrangements at the two alleles were attempted sequentially and every progenitor cell had enough time to rearrange both alleles (15, 19). These results prompted the generalized notion that accessibility to the recombination machinery differs between two identical alleles. However, the same results would be predicted if the efficiency of recombination is generally low but the time given to a progenitor cell to recombine is long, because these experiments analyze a single recombination event (V to DJ) on each chromosome. The genomic organization of the TCRγ locus in the mouse, permitting multiple V to J rearrangements in the same chromosome, allows us to differentiate between these two possibilities. This is due to the fact that, although the initiation of recombination at different V and J segments is likely to be independent from each other (see below), the fact that the products of these recombination events terminates recombination in the whole locus results in their functional dependence. In other words, the extent of V1-J4 rearrangements will depend not only on the probability of producing a functional V1 chain, but also on that of producing a functional V4 chain in the same progenitor cell and vice versa. In this scenario, a feedback mechanism together with a limited time given to the progenitor cell to rearrange their TCRγ genes fail to fit the experimental data because they always overestimate the number of cells containing the two alleles rearranged. Fitting the extent of expected rearrangements to the experimental values at one J region invariably results in predicted values for rearrangements at the other J region higher than those observed experimentally, indicating that other mechanisms participate in the process of allelic exclusion at the TCRγ locus. Imposing a different accessibility of the two alleles to the action of the V(D)J recombinase result in an almost perfect match between the observed and expected values, suggesting that this is one of the mechanisms by which allelic exclusion is achieved. In this line, recent experiments have shown the existence of differential epigenetic modifications at two Igκ alleles, established early during embryonic development, that lead to a preferential accessibility of one of the alleles to the V(D)J recombinase (42). These epigenetic modifications are clonally transmitted and also correlate with the asynchronous replication of different TCR and Ig alleles in mature lymphocytes, with Igκ rearrangements in mature B cells found preferentially in the early replicating allele (42, 43). Interestingly, DNA demethylation, which is one of the epigenetic modifications increasing chromatin accessibility, was shown to occur with different kinetics in both Igκ alleles in a manner compatible with a different probability of each allele to be demethylated, more than with a specific mechanism imposing demethylation at a single allele (44). Independent evidence for the stochastic nature of the preferential accessibility of one of the alleles has also been obtained in studies analyzing the chromatin structure surrounding the transcriptional enhancers associated with the Igκ locus (45).
Even if the two chromosomes become accessible to the action of the V(D)J recombinase with different probabilities, that will only have a limited effect on the isotypic exclusion of mouse TCRγ genes due to the genomic structure of the mouse TCRγ locus (Fig. 2). Possible mechanisms of isotypic exclusion of IgL chains have been extensively discussed and two general models (a sequential model and a probabilistic model) have been proposed (46). The most relevant difference between the two models is that, whereas the sequential model proposes that λ rearrangements are regulated by κ rearrangements, the probabilistic model presumes complete independence of both loci. By analogy with the IgL chain loci, the same two general models could be proposed to explain isotypic exclusion at the TCRγ locus in the mouse. However, the fact that a fraction of V1+ cells contain the two J1 alleles in germline configuration (Fig. 2), argues against a strict temporal order in the rearrangements at different Jγ regions and strongly suggests that rearrangements at different Jγ regions initiate independently. In these conditions, concomitant rearrangement of two or more isotypes and, therefore, expression of two functionally rearranged TCRγ chains becomes possible and their frequency will depend on the rearrangement rates at different J regions. Obviously, the chances that a progenitor cell will produce two functional chains decrease as the rates of rearrangement of the different TCRγ chains diminish but with a concomitant decrease in the fraction of progenitor cells that will succeed in producing a functional TCRγ chain. Interestingly, increasing the rate of recombination of one isotype relative to the other will have as a consequence a relatively high rate of success keeping the probability of producing isotypically included cells at ∼1% (36). However, in these conditions, the majority of the rescued cells will express the most abundantly rearranged isotype (V4 in adult mouse γδ thymocyte). An elevated production of V1 cells can be obtained by reducing the probability of obtaining functional V4 chains without altering the rate of rearrangement of the V4 gene segment. This is precisely the consequence of the presence of an in-frame stop codon at the end of the V4 gene, which decreases by about 2-fold the chance that a V4-J1 rearrangement will result in a functional V4 chain (4, 47). Thus, the 11-fold difference in the rates at which V1 and V4 rearrange (13), together with the presence of a stop codon at the end of the V4 gene segment are sufficient to ensure the development of relatively high numbers of V4+ and V1+ γδ thymocytes from a common precursor, keeping the probability that a cell may contain two functionally rearranged isotypes below 1% (36). The selective advantage for the coexistence of both cell populations may relate to their different functions, as has been suggested in several infectious models (reviewed in Ref. 48).
Although assembly of Vγ genes displays properties of allelic exclusion, it is expected that a small fraction of γδ cells will contain two functionally rearranged alleles. This is predicted by stochastic models of TCRγ gene rearrangement and our data indicates that γδ cells containing two functional V1 or V4 chains are in the order of 1%. More importantly, the models also predict a very small fraction of isotypically included cells, which is in apparent contradiction with the experimental data. Thus, ∼12% of the V1+ thymocytes also contain a functional V4 chain (Fig. 2) and about half of the γδ thymocytes contain a functional V2 chain in addition to the Vγ chain that is expressed at their cell surface (13). These cells are isotypically included at the genetic level but excluded at the level of TCRγ expression at the cell surface as demonstrated by their staining with Vγ-specific Abs (Fig. 1 and Ref. 13). At least part of this phenotypic exclusion is due to the restriction in Vδ chain pairing displayed by V2 and possibly V4 chains (13), reminiscent of the allelically included B cells that were shown to be allelically excluded at the level of pre-BCR surface expression (49). Altogether, these and previous data analyzing the extent of TCRγ rearrangements in a smaller number of γδ thymocytes (13), are consistent with the hypothesis that cessation of rearrangements at the TCRγ locus (and possibly at the TCRδ and TCRβ loci) and final development along the γδ T cell pathway are mediated by signals trough a γδ TCR that can be expressed at the cell surface.
Previous analysis of human γδ T cell clones with Vγ-specific Abs showed that a significant fraction of cells (1–7%) contained two functional TCRγ chains, suggesting that assembly of human Vγ genes is not regulated in the context of allelic exclusion (26). This is in contrast with our data showing that the vast majority of mouse γδ thymocytes express a unique TCRγ chain at the cell surface. However, as also shown here, up to 3% of the mouse γδ thymocytes express two different TCRδ chains at the cell surface. If signals to terminate rearrangements at the TCRγ (and possibly TCRδ) loci are mediated by a selectable γδ TCR and not by a surrogate receptor, only the chain that rearranges later can be allelically excluded efficiently. Furthermore, a certain degree of isotypic exclusion of TCRγ chains in the mouse is due to their restricted pairing with TCRδ chains. Thus, differences in the time at which TCRγ and TCRδ rearrangements take place in human progenitors and/or a less evident restriction for TCRδ chains displayed by human TCRγ chains could explain, at least in part, the different phenotype observed in mouse and human γδ T cells with regard to allelic exclusion of TCRγ chains.
Because assembly of TCRδ chains does not exhibit properties of allelic exclusion it is expected that ∼20% of the γδ T cells containing complete V(D)J rearrangements at both alleles will contain two functional TCRδ chains (22). The actual number of γδ T cells bearing two functional TCRδ chains is higher than that (∼28% in the combined data from Ref. 22 and 13). This is also a possible consequence of the restricted pairing of TCRγ and TCRδ chains and of the fact that γδ T cell progenitors do not attempt all possible rearrangements at the TCRγ locus. In these circumstances, progenitor cells containing two functional TCRδ chains have a greater chance to be rescued by a TCRγ chain to enter the γδ T cell pool than progenitor cells containing a single functional TCRδ chain. Because∼37% of the γδ T cells contain only one complete TCRδ rearrangement (13, 22) it is expected that ∼18% (0.28 × 0.63 × 100) of the γδ T cells contain two functional TCRδ chains. Of those, a relatively large fraction will harbor a TCRδ chain unable to pair with the TCRγ chain and, therefore, will be allelically excluded at the cell surface. Consistent with this interpretation is the fact that, of 11 V4+ clones containing two functional TCRδ chains, 7 (63%) contained a Vδ6 chain (L. Boucontet and P. Pereira, unpublished observations), which is known to be rarely expressed with V4 (25). Thus, in the absence of any other constraint it can be estimated that ∼6% of the γδ thymocytes may express two different TCRδ chains at the cell surface. This value is somewhat higher than the maximum of 3% found by staining with Vδ-specific mAbs (Fig. 1) suggesting that additional mechanisms exist to further restrict the expression of two functional TCRδ chains.
Finally, our results suggest a mechanism to explain, at least in part, how the absence of a functional pre-TCR results in an increased number of γδ thymocytes (7, 50). Thus, signals through the pre-TCR may inhibit TCRγ rearrangements in cells that, otherwise, could still become γδ T cells. However, this mechanism appears to operate in a small number of progenitors, likely because Vβ to DJβ rearrangements mostly take place after the majority of TCRγ and TCRδ rearrangements have been completed (3, 4).
The authors have no financial conflict of interest.
We thank Paulo Vieira and Ana Cumano for critical reading of the manuscript and Pierre Ferrier for the kind gift of TCRβ−/− mice.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by institutional grants, grants from the “Association pour la Recherche sur le Cancer” (to P.P.) and Fundaçao para a Ciência e Tecnologia (to J.C.). N.S. received a fellowship from the Gulbenkian Institute for Science.
Abbreviation used in this paper: Tg, transgenic.