Allelic Exclusion of TCR α-Chains upon Severe Restriction of Vα Repertoire

Development of thymocytes through the positive selection checkpoint requires the rearrangement and expression of a suitable T cell receptor (TCR) α-chain that can pair with the already-expressed β-chain to make a TCR that is selectable. That is, it must have sufficient affinity for self MHC-peptide to induce the signals required for differentiation, but not too strong so as to induce cell death. Because both alleles of the α-chain continue to rearrange until a positively-selectable heterodimer is formed, thymocytes and T cells can in principle express dual α-chains. However, cell-surface expression of two TCRs is comparatively rare in mature T cells because of post-transcriptional regulatory mechanisms termed “phenotypic allelic exclusion”. We produced mice transgenic for a rearranged β-chain and for two unrearranged α-chains on a genetic background where endogenous α-chains could not be rearranged. Both Vα3.2 and Vα2 containing α-chains were efficiently positively selected, to the extent that a population of dual α-chain-bearing cells was not distinguishable from single α-chain-expressors. Surprisingly, Vα3.2-expressing cells were much more frequent than the Vα2 transgene-expressing cells, even though this Vα3.2-Vβ5 combination can reconstitute a known selectable TCR. In accord with previous work on the Vα3 repertoire, T cells bearing Vα3.2 expressed from the rearranged minilocus were predominantly selected into the CD8+ T cell subpopulation. Because of the dominance of Vα3.2 expression over Vα2 expressed from the miniloci, the peripheral T cell population was predominantly CD8+ cells.


Introduction
Allelic exclusion of T cell receptor (TCR) genes is regulated differently for the a and b-chains [1][2][3]: for the b-chain, rearrangement stops when the cell detects a productively rearranged membrane-bound b-chain protein associated with pre-Ta, leading to downregulation of Rag1/2 gene expression. Thus only one of the bchain loci is capable of producing full-length, correctly rearranged, b-chain mRNA and therefore protein. In contrast, the TCR a-chain gene does not cease rearranging until the developing T cell undergoes positive selection. During the CD4 + 8 + ''double positive'' (DP) stage of thymocyte development, both of the achain alleles rearrange until a positively-selectable heterodimer is formed with the previously-formed b-chain [4], leading to Rag1/2-turnoff which stops further rearrangement [5][6][7]. Immature thymocytes (DP, TCR lo ) frequently express dual a-chains on the cell surface, but almost all mature (DP, or SP, TCR hi ) thymocytes express a single ab-combination [8,9], in what has been termed phenotypic allelic exclusion [1,10].
Similarly, most peripheral T cells express a single a-chain on the cell surface, despite frequently (20-30%) having two functionally rearranged and expressed achain genes. Estimates of the number of peripheral T cells expressing two cell surface a-chains vary widely from ,5% to 15% [10][11][12][13] in mice, and 30% in humans [14]. The dual receptor cells can cause autoimmunity in some systems [15,16] or be highly alloreactive [17], although other reports did not find them to increase susceptibility to autoimmunity [11,18]. They have also been reported to usefully increase the TCR repertoire [19].
A post-translational mechanism ensures that only one a-chain is generally present on the cell surface of mature T cells [1,9,10,20]. This is caused by selective lack of expression of one of the a-chains on the cell surface [9,10]. This mechanism also operates in transgenic mice expressing two ab TCRs [21,22], where it can also regulate b-chain expression [23]. Several mechanisms were proposed to account for this phenomenon, including competition between the achains for the b-chain, and ''selective retention'' of the selectable a-chain on the cell surface [1,8,9]. Support for the selective retention model and against a-chain competition models has been obtained [24].
The comparative rarity of cells expressing any individual TCR Va-region and the lack of suitable reagents, have made it difficult to study TCR a-chain allelic exclusion in thymocytes. Specifically, there are still only mAbs against four mouse Va-regions: the Va2 family, Va3.2, Va11.1/11.2, and some members of the Va8 family. There is no anti-Ca mAb that can be used to stain live cells for flow cytometry, and no allelic differences between the Ca-regions of different mouse strains. We therefore decided to make a mouse that can express a diverse repertoire limited to two Va-regions, as a tool to allow us to study phenotypic allelic exclusion in the thymus. The ultimate goal of this project was to create a mouse model for allelic inclusion of TCR a-chains, which would permit isolation of sufficient numbers of dual a-chain expressing cells for biochemical and cell biological analysis of posttranslational events resulting in the selective retention of a single a-chain.
Using Va3.2 and Va2 miniloci, each with two different J-region elements, we found that both the Va3.2 and Va2 containing a-chains had a strong ability to be positively selected with either a rearranged Vb5-containing transgene or the natural repertoire of TCR b-chains. As a result, we were unable to find significant numbers of cells that had rearranged and expressed both minigenes, and were therefore unable to detect significant phenotypic allelic exclusion. We did however find that T cells with the Va3.2 minilocus-derived a-chain repertoire dominated the Va2-bearing cells in number. Presumably because of the natural ability of Va3.2 to skew development of CD8 T cells [25][26][27], this also selected strongly for a CD8 + T cell-dominated repertoire.

Ethics statement
Animal work was performed at TSRI and was approved by the Institutional animal care and use committee of TSRI (protocol #06-0340).

Bone marrow reconstitution
Bone marrow cells were isolated from donor mice, and 2610 6 bone marrow cells were injected i.v. into lethally irradiated (11 Gy in two equally split doses) CD45.1 and MHC o/o recipient mice. Thymocytes were analyzed 8 weeks postreconstitution.

Transgenic TCR Va3.2 minilocus
In order to produce a mouse with the ability to make two independent TCR achains to allow us to study allelic exclusion, we made use of a TCR a-chain minilocus transgene that allows a Va2 gene to recombine with Ja26 or Ja2 [32]. Specifically, this is the Va2.3 (TRAV14-1) variable region exon, which is the same Va gene used in the MHC class I-restricted OT-I TCR [33]. When bred to the achain knockout (Tcra 2/2 ) background, this mouse (called ''Va2Var'') makes a very limited TCRa repertoire, but with a good degree of diversity in CDR3a [32]. The minilocus construct is based on an earlier vector for making TCR transgenes that uses a natural Va-promoter [34], such that the transgene is rearranged and expressed at the correct developmental stage. The Va2 minilocus-derived TCR is expressed in both CD4 and CD8 subsets, with a slight bias to CD4s, although when bred to the Vb5-containing OT-I TCR b-chain transgene, its expression is skewed to the CD8 cells [32].
We replaced the Va2 gene with a Va3 gene (specifically Va3.2 (TRAV9D-4)) flanked by appropriate recombination signal sequences. With the resulting construct, we made transgenic mice on the B6 background. Va3.2 was chosen because a good antibody against this V-region exists, and because earlier studies showed that normal Va3.2 is preferentially expressed in CD8 T cells [25,27,29].
We obtained several transgenic founder lines. These were characterized for expression of Va3.2 and the number of copies of the transgene was estimated by Southern blotting (Table 1). Those with several copies of the transgene (lines #16, 17, and 20) expressed significantly more Va3.2 in peripheral blood T cells than B6 mice. Line #8 had a single copy of the transgene and expressed about twice the normal amount of Va3.2 ( Table 1). The Line #8 Va3.2 minilocus mice were bred to the Va2 minilocus mice on the Tcra 2/2 background, with or without the Vb5 OT-I b-chain transgene.
The data in Table 1 also confirmed our previous work showing that Va3.2 is preferentially selected into the CD8 subset of T cells [25][26][27], as there were about 3.4-fold more Va3.2 + cells in the CD8 subset than the CD4 subset, in each of the Va3.2 minilocus strains.

Preferential selection of Va3.2 + T cells over Va2 + T cells
We tested expression of Va2, Va3.2 and Va5 in mice transgenic for the two Va miniloci and deficient in endogenous a-chain expression (Tcra 2/2 ), and in the presence or absence of the OT-I Vb5 transgene. In WT B6 mice, Va2 is expressed in about 12% of T cells, and Va3.2 in about 2%. However, in the dual minilocus mice, Va3.2 was much more prevalent than Va2, whether or not the b-chain transgene was expressed (Fig. 1A,B). Given that the Va2 minilocus can preferentially reconstitute the OT-I TCR with this b-chain transgene [32], it was a surprise to find such a low representation of the Va2 + cells in the mice expressing the OT-I b-chain. In the WT B6 mice, Va2 was the more frequently expressed protein, particularly in CD4 + cells (Fig. 1B), but the presence of the Va3.2 minilocus strongly favored Va3.2 expression, especially among the CD8 + cells (Fig. 1B). Analysis of peripheral T cells failed to reveal any significant enrichment of dual Va expressing cells (Figs. 1 and 2). It was recently noted that the Va2-Vb5 combination is disfavored when Va2 can ''choose'' a different Vb element than Vb5 [35].
When lymph node T cells from mice expressing both miniloci as their only achains were gated for expression of each of the Va-regions, the normal CD4skewing of Va2 + cells was reduced to close to 1:1 (Fig. 3A, left panels). When the Vb5 transgene was also present (bottom left panel), skewing was reversed so that the majority of Va2 + cells were CD8 + . This is as expected from earlier work where Va2 expression occurred predominantly in the CD4 + compartment in B6 mice [28,36], was roughly equally expressed in CD4 + and CD8 + T cells in the Va2 minilocus (Va2Var) mouse, and was skewed to CD8 + cells when the OT-I Vb5 transgene was coexpressed with the Va2 minilocus (''Limited mouse'') [32]. Va3.2 expression in the minilocus mice was even more strongly skewed to the CD8 subset than in WT (Fig. 3A, right panels), and this bias was almost absolute in the double-minilocus, Vb5 transgene-expressing mice (bottom right panel). These effects were reflected in the CD8:CD4 ratio (Fig. 3B). This ratio is less than 1:1 in B6 lymph nodes [36]. The Va3.2 minilocus caused the ratio to increase to ,4:1, but expression of the Vb5 transgene massively increased the ratio to over 30:1. Similarly the Va2 minilocus plus the Vb5 transgene skewed the population to more than a 10-fold excess of CD8 T cells. This is not surprising given that the OT-I TCR transgene causes development of predominantly CD8 + mature T cells, but it is perhaps surprising that the Va3.2 minilocus had an even stronger impact than the Va2. This indicates that the potential Va3-Ja26 and Va3-Ja2 rearrangements were able to produce very positively selectable TCRs in combination with the OT-I b-chain, just as they were with a complete b-chain repertoire. It also shows the strength of the preferential selection of Va3.2 into MHC class I-restricted T cells [25,27].

Expression of the a-chain miniloci in thymocytes
We analyzed expression of the two a-chain minilocus transgenes in Tcra 2/2 thymocytes, finding that, as with the peripheral T cells, the vast majority of thymocytes expressed Va3.2 rather than Va2, whether or not Vb5 was also expressed (Fig. 4A). This was true of the TCR int (before or during positive selection) cells, as well as the TCR hi post-positive selection cells. There was no evidence for increased numbers of dual TCR a-chain-expressing cells compared to WT thymi in any of these developmental subsets. Because we had previously noted that some cells express two a-chains intracellularly, with only one being expressed on the cell surface [10], we also tested for expression of the two transgenes in permeabilized cells. Intracellular expression of either Va3.2 or Va2 from the miniloci was found, but there were essentially no cells expressing both Va3.2 and Va2 (Fig 4B). Separate analysis of the DN subset and CD4 and CD8 SP subsets did not reveal significantly enriched dual Va-expressing cells (Fig. 5). Rare dual Va positive cells were indeed observed in all specimens, however regardless of the phenotype this population represented no more than several hundred cells per mouse. Because a-chain rearrangement and expression from one chromosome does not inhibit rearrangements on the other chromosome [1][2][3], we were initially surprised that there was not a population of cells expressing both miniloci. However, it was previously found that that only a relatively small proportion (,0.5% of normal) of the thymocytes in the Va2 minilocus mice rearrange and express the locus [32]. This suggested that the lack of cells expressing both Va3.2 and Va2 was due to the low numbers that actually rearrange either of the miniloci. As noted earlier, both of the miniloci produce highly selectable a-chains, so that the lack of dual Va-expressing thymocytes could be due to positive selection occurring before there is a chance for rearrangement of the other minilocus.  Va expression in thymocytes in the absence of a selecting signal We reasoned that lack of dual Va-expressing cells might be due to efficient positive selection of both the single-Va3.2 or single-Va2-expressing cells, before they had a chance for a second rearrangement. We therefore tested whether the dual minilocus thymocytes could be induced to co-express both Va2 and Va3.2 if they were allowed to develop in a non-selecting background. Chimeric mice were made by reconstituting lethally irradiated CD45.  (Fig. 6B, lower panel).
In another experiment to test the effect of eliminating positive selection, we used the OP9-DL1 co-culture system [30,31] to analyze thymocyte development.
In this system, thymocytes develop from DN to DP TCR int cells, but few undergo positive selection to produce TCR hi cells. Here, the triple-transgenic (Va3.2 and Va2 miniloci, rearranged OT-I Vb5, Tcra 2/2 ) cells developed from DN to DP, Effective Allelic Exclusion in Va-Restricted Mice expressing TCR lo or TCR int (TCR hi cells were not distinguishable, and were counted with TCR int ). Again, negligible numbers of dual Va-expressing cells were found (Fig. 7).

Conclusions
Our initial goal was to produce a mouse expressing two defined Va's that would allow us to study the induction of phenotypic allelic exclusion in detail, and particularly biochemically. This goal was not realized, likely because of the low frequency of cells in these transgenic mice that actually rearranged the a-chain miniloci. Moreover, once expressed, both the Va3.2 and Va2 containing a-chains were efficiently positively selected. This was the case with endogenously rearranged TCR b-chains as well as when the b-chain repertoire was limited to the OT-I (Vb5) b-chain. This likely contributed to the failure to find a distinct population of cells that expressed dual TCRs, even intracellularly. Va3.2 transgene-expressing cells were much more frequent than the Va2 transgeneexpressing cells, especially in the presence of the OT-I b-chain. This was surprising given that the OT-I TCR uses the identical Va2 element to the minilocus and that the OT-I CDR3a can be recreated by rearrangement of the minilocus [32], although other work also suggests that the combination of Va2 with Vb5 is not particularly strongly selected when other Va-Vb combinations are possible [35]. The predominant selection of Va3.2-bearing cells to the CD8 + population [25][26][27] was recapitulated in the minilocus mice, and the dominant expression of Va3.2 compared to Va2 resulted in the majority of peripheral T cells being CD8 + cells. The Va3.2 minilocus mouse strain will potentially be useful for studies of repertoires using the Va3.2 element, similar to the utility of the previouslydescribed Va2 minilocus strain.