TCR Transgenic Mice in Which Usage of Transgenic α- and β-Chains Is Highly Dependent on the Level of Selecting Ligand¹

Akey event in thymic development of T cells in the αβ lineage is the successful rearrangement of a TCR β-chain gene (1, 2, 3, 4). This first major checkpoint in the thymic developmental process has been called β selection (5, 6). The initial β rearrangement takes place in CD4⁻CD8⁻ thymocytes, and successful rearrangements produce functional TCR β-chains that pair with the pre-Tα-chain (7). The expression of the β/pre-Tα complex on the cell surface leads to three main consequences: 1) proliferation of these thymocytes, 2) up-regulation of the CD4 and CD8 coreceptors, and 3) an end to rearrangement at the β locus (2). The result is an expanded CD4⁺CD8⁺ population of thymocytes that all have at least one β-chain that is competent for cell surface expression. Rearrangement is stopped quickly at the β locus once one functional β-chain has been produced, and it is observed that almost all T cells have only a single functionally rearranged β (8). This mechanism is therefore termed allelic exclusion, but does not exclude the possibility that both β loci can functionally rearrange. Indeed, studies on both human and mouse lymphocytes have shown that 1% of peripheral T cells can have two functional TCR β-chains expressed on the cell surface (9, 10, 11), but it is not known whether this level of allelic exclusion is determined solely by the efficiency of the β selection step or if other mechanisms might be involved.

Once thymocytes become CD4⁺CD8⁺, they can then start rearrangements at the TCR α locus. Double-positive thymocytes will keep trying to rearrange α until they can produce a functional TCR complex on the cell surface that is competent for positive selection (12). Therefore, at this stage many functional α-chains are produced that are not competent for positive selection, but are nevertheless expressed along with the second, selectable TCR. It has been estimated that as many as 30% of peripheral αβ T cells express two α-chains (13).

Others have considered the effect that multiple α- and β-chains might have on an immune response and have suggested that they may be involved in autoimmunity (14). A T cell that has a single TCR against self may not ever be activated and find the site of its self antigen, or it may not have sufficient avidity to react, but if it also has a second TCR that is activated by a pathogen, the T cell could then become competent to attack self via its other TCR. The argument that has been raised against this hypothesis is that T cells with two TCRs are likely to have only one of the TCRs that is specific for self MHC plus peptide (15). This is a natural consequence of the fact that CD4⁺CD8⁺ thymocytes stop trying to rearrange at the α locus as soon as an α-chain is produced that is competent for positive selection (12).

This latter counterargument would certainly be true if a TCR always immediately started the process of positive selection when the proper ligand was also present in the thymus, but expression of a selectable TCR is not sufficient for positive selection, since this process is limited by the number of available selecting niches (16). This concept stems from the fact that in TCR transgenic mice in which all the CD4⁺CD8⁺ thymocytes express a receptor competent for positive selection, only 20% of these CD4⁺CD8⁺ cells are positively selected (17). There are a limited number of niches available to select these TCRs. Although the number of niches can be reduced experimentally by lowering the number of cells that promote selection, a major factor that determines the number of available thymic niches in the mouse is the expression level of selecting molecules. For at least two cytochrome c-specific TCRs, it has been reported that positive selection can depend greatly on the expression level of the selecting ligand (18, 19). Thus, a TCR that is specific for self-MHC may not mediate positive selection due to a lack of a thymic niche, but this TCR may still be able to make it to the periphery if it is expressed on a cell that has a second, more efficiently selected TCR.

We have recently produced TCR transgenic mice that use a TCR specific for the d allele of murine hemoglobin (Hbb^d)⁴ presented by I-E^k. These mice show extensive usage of endogenous TCR β- and α-chains and also show a dramatic dependence on I-E^k levels for positive selection. These mice, therefore, gave us the ability to look at mechanisms other than β selection that might act to keep only one functional TCR β-chain per cell. These mice also gave us an opportunity to examine the effect of selecting ligand density on positive selection and endogenous receptor usage.

Minigenes of TCR 3.L2 α- and β-chains were constructed by inserting V-Jα and V-D-J β exon cassettes into TCR α and β shuttle vectors (20). 3.L2 V18 Vα-J and 3.L2 Vβ8.3 V-D-J segments were amplified by PCR using cDNA from the 3.L2 T cell clone (21). The gene fragments were cloned into pBluescript (Stratagene, La Jolla, CA), sequenced using ABI Dye Terminator Cycle Sequencing (Perkin-Elmer, Foster City, CA), and cloned into the TCR shuttle vectors. To verify that the Vα18 and Vβ8.3 chains were involved in the recognition of Hb_64–76, the minigene constructs were transfected into the TCR-negative 58 α-β-hybridoma line, and identical specificity to the original clone was observed (data not shown). The minigene constructs were then coinjected into C57BL/6 pronuclei in the Department of Pathology’s Transgenic Core Facility as previously described (22). Mice were screened by Southern blotting of genomic DNA using probes to the Cα and Cβ regions. The founder mice were bred to the B6 congenic strain B6.AKR-H-2^k/FlaEg or to the RAG-1-deficient strain C57BL/6J-Rag1^{tm1 Mom} (The Jackson Laboratory, Bar Harbor, ME). Transgenic F₂ generation mice were identified that were H-2^k homozygous by FACS analysis of the peripheral blood using anti-H-2K^k and anti-H-2K^b Abs. After the establishment of the line, transgenic mice were identified by PCR amplification of 3.L2 α transgene from tail DNA. The PCR conditions for the 3.L2 α-chain were 35 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 45 s. The forward primer was 5′-AAACTCGAGACCTGTGTGGATAAAAACCTCTCTGATTCTGGTTTGCTTTTCTGTTTCCAAGCAGGGGCAAAGAGCCAATGGGGAGGA-3′, and the reverse primer was 5′-AAAATTGCGGCCGCTTGGGCCCAAGAAACTGTCATCAAAACGTACATGGATGAACCACGAGGCTGGTC-3′. All mice were housed in a specific pathogen-free barrier facility at Washington University.

The 3.L2 CAb was made following a rapid immunization protocol (23). C57BR/cdJ mice (The Jackson Laboratory) were immunized s.c. with 2 × 10⁷ 3.L2 clone T cells emulsified in CFA and boosted 13 days later with 2 × 10⁷ T cells in IFA. Draining lymph node cells were fused 3 days later with the B cell fusion partner P3X63Ag8.653 (24). Supernatants were screened by FACS for positive staining of the 3.L2 hybridoma. The supernatant from one well strongly stained the 3.L2 T cell hybridoma and was shown to be clonotype specific by the absence of staining to a panel of T cell hybridomas expressing Vβ8.3 receptors and those expressing Vα18 family receptors. Specificity for the 3.L2 TCR was further shown by specific staining of the 58α⁻β⁻-transfected cells described above. The CAb line was subcloned twice and was determined to be an IgG2a κ isotype. The CAb Ab was purified from tissue culture supernatants using protein A-Sepharose (Sigma, St. Louis, MO) and biotinylated.

The anti-Vβ8.3 Ab (1B3.3) was made by immunizing Armenian hamsters i.p. with 1 × 10⁷ 3.L2 clones in saline three times, with each injection 14 days apart. Hamsters were rested for 2 mo and then given 5 × 10⁶ T cells i.v. Spleens were fused 3 days later with P3X63Ag8.653 (24). Supernatants were screened by activation of the 3.L2 hybridoma by plate-bound Ab. 1B3.3 clone specificity was shown by the ability to activate or stain by FACS Vβ8.3 cells and not Vβ8.1 or 8.2 hybridomas. The Ab was subcloned four times, and supernatants were purified on protein A-Sepharose and biotinylated.

The peptides used in this study were synthesized by standard F-moc chemistry using either a Synergy Peptide Synthesizer (model 432A, Applied Biosystems, Foster City, CA) or a multiple peptide synthesizer (Symphony/Multiplex, Protein Technologies, Tuscon, AZ). The peptides were purified by HPLC, and their purity and accuracy were confirmed by mass spectrometry (Washington University Mass Spectrometry Resource). The concentration of the peptides was determined on an amino acid analyzer (model 6300, Beckman, Fullerton, CA). Altered peptide ligands have been defined previously and are referred to using the one-letter amino acid code of the substituted amino acid residue and its position (25). The amino acid sequences for the peptides used in this study are: Hb_64–76, GKKVITAFNEGLK; T72, GKKVITAFTEGLK; I72, GKKVITAFIEGLK; A72, GKKVITAFAEGLK; and Q72, GKKVITAFQEGLK.

Single cell suspensions were prepared from thymus and spleen. One million cells per sample were stained at 4°C for 30 min with the Abs diluted in PBS with 0.5% BSA and 0.02% NaN₃, and then washed. When necessary, cells were also stained in an identical manner with a second step reagent. The Abs used in this study were 53.6.7-FITC (rat anti-mouse CD8α; PharMingen, San Diego, CA), H129.19-phycoerythrin (rat anti-mouse CD4; PharMingen), H57-597-biotin (hamster anti-mouse TCRβ; PharMingen), 17-3-3-biotin (mouse anti-mouse I-E^k; PharMingen), AF6-88.5-biotin (mouse anti-mouse H-2K^b; PharMingen), 36-7-5-biotin (mouse anti-mouse H-2K^k; PharMingen), G155-178-biotin (mouse anti-TNP (control IgG2a; PharMingen), KT4-biotin (rat anti-mouse Vβ4; PharMingen), 14-2-biotin (rat anti-mouse Vβ14; PharMingen), TR310-FITC (rat anti-mouse Vβ7; PharMingen), streptavidin-tricolor (Caltag, South San Francisco, CA), streptavidin-FITC (Caltag), CT-CD8a-tricolor (rat anti-mouse CD8α; Caltag), CAb-biotin (rat anti-mouse 3.L2 clonotype), and 1B3.3-biotin (hamster anti-mouse Vβ8.3). Cells were analyzed on a FACScan flow cytometer (Becton Dickinson, Mountain View, CA) using CellQuest analysis software.

Proliferation of primary 3.L2tg T cells was performed as described previously (21). Briefly, 5 × 10⁵ splenocytes were cultured in RPMI 1640 supplemented with 10% FCS, 5 × 10⁻⁵ M 2-ME, 1 mM Glutamax (Life Technologies, Gaithersburg, MD), and 50 μg/ml gentamicin in a 200-μl total volume in a 96-well tissue culture plate with the indicated concentration of peptide. Culture wells were pulsed at 48 h with 0.4 μCi [³H]thymidine and harvested 18 to 24 h later.

For these studies, we have generated a TCR transgenic founder line that expresses the rearranged Vα18 and Vβ8.3 genes of the Th1 clone 3.L2 (21, 26). This founder was generated by coinjection of the constructs into C57BL/6J (H-2^b)-fertilized eggs, and the resulting line was named 3.L2tg. Because the 3.L2 clone is specific for residues 64 to 76 of murine hemoglobin (Hbb^d) complexed to I-E^k, the founder was bred to the congenic strain B6.AKR (H-2^k) to introduce the I-E^k selecting ligand and also to maintain homozygosity of the nonstimulatory Hbb^s allele.

To follow expression of the 3.L2 TCR in these mice, we have generated two mAbs: one is against the 3.L2 clonotype, which we call CAb, and the other is specific for Vβ8.3. Initial analysis of these mice was performed on thymocytes of 3.L2tg mice on the H-2^k/k background. As shown in Figure 1,A, the percentages of CD4 and CD8 single-positive cells are similar to those observed in nontransgenic B6.AKR mice. The thymocytes express the transgenic β-chain (Fig. 1,B) and the combination of the transgenic α and β as shown by CAb staining (Fig. 1,C). The CAb⁺ cells can be divided into three groups: low, intermediate, and high. The intermediate cells are mostly the immature double-positive cells, and the high CAb cells are those we consider to be undergoing selection. When we gated on these high CAb cells, we observed that they were preferentially selected into the CD4 lineage (Fig. 1,D). The histograms in Figure 1, E and F, demonstrate the specificity of these Abs. These panels show staining performed on the CD4⁺ thymocytes from nontransgenic B6.AKR mice. There are essentially no CAb⁺ cells, whereas there are about 7% Vβ8.3⁺ cells among CD4⁺ thymocytes in these mice.

To further demonstrate that we have produced a transgenic line that uses the same receptor as the 3.L2 Th1 clone, we assayed the proliferation of 3.L2tg splenocytes in response to Hb peptides. As shown in Figure 2, the Hb_64–76 peptide induced strong proliferation in 3.L2tg splenocytes at a concentration as low as 10 nM. Furthermore, peptide analogues of Hb_64–76 that are known to stimulate the 3.L2 clone also stimulated the 3.L2tg splenocytes with a relative activity similar to that found with the original T cell clone (26). Therefore, the 3.L2tg mice on an H-2^k/k background have cells that express the same TCR as the 3.L2 clone. These cells are selected into the CD4 lineage and recapitulate the reactivity observed in the 3.L2 clone.

In TCR transgenic mice that have been reported previously, expression of a transgenic β-chain leads to severe inhibition of rearrangement at the endogenous β locus (12, 27). We were therefore astonished when we made the observation that the 3.L2 transgenic mice express many endogenous TCR β-chains. Figure 3,B shows Vβ expression on CD4⁺ splenocytes in 3.L2tg mice. Although 85% of these CD4⁺ splenocytes express the transgenic Vβ8.3 chain, they also express significant levels of the three endogenous β-chains that we examined (Vβ4, Vβ14, and Vβ7). There are many other Vβ-chains that we did not examine, but we would presume that all the endogenous chains are being used. The extent of endogenous β usage in these mice suggests that most of these T cells are expressing more than one β-chain on the cell surface. Similar results were obtained with CD4⁺ thymocytes. As a control, we also examined Vβ usage on CD4⁺ splenocytes in a class II-restricted TCR transgenic mouse that was generated using the same expression constructs as the 3.L2tg mouse, but with the 3A9 TCR (20). This TCR is comprised of the Vβ8.2 and Vα3 chains, and in these mice (as expected) there is very little usage of endogenous, nontransgenic TCR β-chains (Fig. 3 A). We also found that the transgenic Vβ8.2 chain is expressed on >97% of CD4⁺ splenocytes in 3A9 transgenic mice (data not shown).

The results shown in Figure 3 indicate that most T cells in the 3.L2tg mouse are expressing more than one β-chain on the cell surface. We wanted to show this directly by dual staining of CD4⁺ splenocytes for expression of the transgenic Vβ8.3 and an endogenous TCR β-chain. Figure 4 shows staining of CD4⁺ splenocytes for both Vβ8.3 and Vβ14. As expected, the normal B6.AKR mouse expresses these two TCR chains on distinct populations of T cells (Fig. 4,A). In contrast, in the 3.L2tg mouse, Vβ14 is expressed almost exclusively on cells that also have the transgenic Vβ8.3 chain (Fig. 4 B). In addition, dual staining of CD4⁺ splenocytes in 3.L2tg mice for the two endogenous chains Vβ14 and Vβ4 found that these two receptors are expressed on distinct cell populations (data not shown). The data indicate that almost all the CD4⁺ T cells in this mouse have both the transgenic Vβ8.3 and one additional, endogenous Vβ on the surface.

We next wanted to analyze the development of T cells that express both transgenic and endogenous TCR β-chains. As shown in Figure 1, both the transgenic α and β are expressed on most thymocytes in H-2^k/k mice. In these mice, although most cells are expressing more than one TCR, the majority of these cells express the transgenic TCR at high levels, and it is this TCR that is likely to be used for positive selection. This fits with models that suggest that in thymocytes that have two TCRs, only one will be able to mediate positive selection (15). We next wanted to test this idea by asking whether the transgenic TCR would continue to be used on a nonselecting MHC background and also how the transgenic TCR would be used on a background that limited the number of available selecting niches by reducing the levels of I-E^k.

We were able to investigate these questions by breeding the 3.L2tg, H-2^k/k mice to the H-2^b strain C57BL/6 and the H-2^h4 strain B10.A(4R). By such breeding, we were able to obtain 3.L2tg mice that expressed varying levels of I-E^k. H-2^k/k mice have the highest levels of I-E^k, whereas H-2^h4 mice do not express I-E^k due to a defect in the Eα promoter (28). H-2^k/h4 mice have levels of I-E^k that are reduced compared with those of H-2^k/k because only one chromosome has a functional Eα gene. H-2^k/b mice have levels of I-E^k that are even lower than those of H-2^k/h4. This is because the H-2^b chromosome is also unable to contribute a functional Eα gene, and the functional Eα-chain from the H-2^k chromosome must be shared between Eβ^k and Eβ^b (see Table I). The result is that I-E^k levels vary in these MHC backgrounds: k/k > k/h4 > k/b, and h4/h4 is I-E^k negative. We have confirmed that these mice really do have these relative amounts of I-E^k, both by staining with an Ab against Eβ^k and by measuring the ability of APCs from these strains to stimulate a hybridoma derived from the 3.L2 clone (Table I).

3.L2tg mice bred to these different MHC backgrounds all developed CD4⁺ T cells, although they were produced most efficiently in H-2^k/k mice (Table II). What is striking about the CD4⁺ T cells that do develop is that their usage of the 3.L2 transgenic α- and β-chains is highly dependent on the level of the I-E^k-selecting ligand. Figure 5,A shows CAb staining performed in a single experiment on CD4⁺ thymocytes in 3.L2tg mice bred onto the four indicated MHC backgrounds, and Table II shows cumulative data from numerous mice and experiments. The data show that as the level of I-E^k is reduced, the percentage of thymocytes maturing with high levels of the 3.L2 TCR is also reduced. Correspondingly, Figure 5,B shows that the percentage of CD4⁺ thymocytes using the transgenic β-chain is also reduced as I-E^k levels are lowered. Surprisingly, in the absence of I-E^k, usage of Vβ8.3 in CD4⁺ cells is reduced to levels seen in nontransgenic mice, even though 85% of CD4⁺ T cells use this chain on a 3.L2tg;H-2^k/k background (Fig. 3 B). The down-regulation of transgenic β must mean that these cells are using an endogenous β-chain, since levels of total TCR β remain uniformly high (data not shown). These data demonstrate that this TCR is highly dependent on the density of the selecting ligand for positive selection. Furthermore, CD4⁺CD8⁺ thymocytes that express more than one TCR specificity will favor expression of the TCR used for positive selection (29). However, in conditions where the number of available selecting niches is limiting (e.g., 3.L2tg;H-2^k/b) these mice use mostly nontransgenic TCRs for positive selection even though the transgenic TCR has the requisite specificity for positive selection.

Since the usage of the transgenic β-chain is reduced in mice that have lower levels of I-E^k, one might expect the percentages of cells using endogenous β-chains to increase. Figure 6 displays dual staining for Vβ8.3 and Vβ14 on CD4⁺ splenocytes from 3.L2tg mice with four different levels of I-E^k. Indeed, as I-E^k levels were reduced we did observe an increase in the percentages of cells using endogenous Vβ14. Similar increases were seen with Vβ4 and Vβ7 (data not shown). Overall, on the H-2^k/h4 and H-2^k/b backgrounds, similar levels of endogenous TCR β-chains were used, and this was at a level similar to that found in nontransgenic mice. Therefore, our conclusion is that under conditions where the number of thymic niches for the transgenic TCR is limiting, the endogenous TCR repertoire is expressed at near normal levels. In 3.L2tg;H-2^h4 mice, the percentage of cells using high levels of Vβ8.3 is reduced to the levels found in normal mice, although a fairly high percentage of CD4⁺ cells maintains Vβ8.3 expression at a low level.

Why do we see endogenous β-chains expressed in the 3.L2tg mice? We cannot say for sure, but a clue comes from the phenotype that we observed when we bred the 3.L2tg mice to a background homozygous for a deficiency in the RAG-1 gene (RAG^−/−) (30). On this background, the 3.L2tg mice will not rearrange endogenous α or β loci and will therefore only be able to express the transgenic TCR. 3.L2tg;RAG^−/− mice with an H-2^k/k MHC have thymi about 10 times smaller than those in 3.L2tg;RAG^+/+ mice. These small thymi are made up of about 85% CD4⁻CD8⁻ cells that are also CD44 low and CD25⁺ (Fig. 7 and data not shown). CAb staining of the different subsets in these mice shows that the CD4⁻CD8⁻ cells are negative for surface TCR, but cells at the CD4 single-positive stage express uniformly high levels of the 3.L2 TCR (Fig. 7 B). Surprisingly, we also see cells that are CD8 single positive and expressing high levels of 3.L2 TCR. We are currently investigating the selection and function of these cells. Overall, these data suggest that the process of positive selection works well in these mice, but the 3.L2tg;RAG^−/− mice are very inefficient at the transition from CD4⁻CD8⁻ to CD4⁺CD8⁺. Because we see this phenotype when endogenous TCR chains are unable to rearrange, the endogenous β must be required in 3.L2tg;RAG^+/+ mice to mediate efficient β selection.

At this time, we do not know why the transgenic 3.L2 TCR is inefficient at mediating β selection. One possibility is that the transgene is expressed only after the time of normal endogenous β rearrangement, and the transgene cannot mediate β selection because it arrives too late. Another, more intriguing possibility is that the transgenic β is present early enough but is unable to function. This could be due to an inability of this Vβ8.3 to signal in immature thymocytes or to poor pairing between pTα and 3.L2 Vβ8.3. A precedent for the latter explanation has been seen in the inability of certain Ig heavy chains to mediate maturation of B cell precursors in adult bone marrow due to poor pairing with the surrogate light chain (31).

These data show that in the 3.L2tg mice, β selection is not working efficiently to prevent multiple β-chain expression on individual thymocytes. This has given us an opportunity to determine whether any other mechanisms might be at work to limit multiple β expression. We have seen one possibility in the examination of transgenic TCR levels on CD4⁺CD8⁺ thymocytes. Figure 8 shows the levels of CAb staining on double-positive thymocytes in 3.L2tg mice bred onto the four MHC backgrounds that differ in their levels of I-E^k. The 3.L2tg;H-2^k/k mouse that is shown is representative of most of the 3.L2tg;H-2^k/k mice examined; the levels of CAb staining on CD4⁺CD8⁺ thymocytes is mostly in a peak that we call TCR intermediate. On the MHC backgrounds that have reduced levels of I-E^k, we have observed that the double-positive thymocytes use lower levels of the transgenic TCR, even though the level of total TCR β is the same in all these mice (data not shown). This is surprising given the fact that interactions between the TCR and MHC ligands are only thought to influence the few CD4⁺CD8⁺ cells that are starting the process of positive selection (32). Therefore, these data suggest that in normal mice, the few CD4⁺CD8⁺ thymocytes that have two functional β-chains due to the occasional imperfection of the β selection process may have a tendency to down-regulate any β-chain that does not result in efficient positive selection.

Here we have described TCR transgenic mice that show a heavy dependence on the expression level of the selecting ligand. The level of I-E^k in these mice determines how many cells get positively selected and how many of these cells use the transgenic α- and β-chains. Surprisingly, these mice use many endogenous β-chains despite the fact that the transgenic α- and β-chains are expressed well on both CD4⁺CD8⁺ and CD4 single-positive thymocytes. Although the transgenic chains are only expressed at low levels on some backgrounds, this does not indicate a down-regulation of TCR in general, because T cells in these mice always have normal, uniform levels of TCR αβ, as measured by staining with an anti-TCR β Ab. Therefore, in these mice, the level of ligand expression is critical in determining TCR usage.

This observation has important implications for the mechanism of determining the naive T cell repertoire. The data suggest that many cells at the immature CD4⁺CD8⁺ stage have multiple α-chain rearrangements and, in the case of our 3.L2tg mice, α- and β-chain rearrangements. However, during the process of positive selection, a specific TCR is up-regulated as cells progress from the double-positive to the single-positive stage. It is striking that in our 3.L2tg mice under conditions where the 3.L2 TCR is poorly selected, most positively selected cells fail to up-regulate the 3.L2 TCR. Instead, this TCR seems to remain on many cells, but only at a low level. Thus, up-regulation of TCR during positive selection seems to be specific for the TCR that can efficiently mediate this process. This type of phenomenon has also been observed in a careful analysis of multiple α-chain usage on immature vs mature thymocytes (29).

At this time, we do not know the mechanism by which the developing thymocyte determines that a particular TCR αβ pair is useful for selection and therefore up-regulates only that particular TCR. T cells seem to have a type of feedback mechanism that instructs a cell that a particular TCR has bound ligand. In mature cells this results in specific down-regulation of the ligated TCR (33). We would argue that a similar mechanism acts in immature thymocytes such that cells whose TCR binds a ligand will up-regulate that TCR specifically. In addition, we cannot rule out that competition for pairing among α- and β-chains may play some role, particularly in the 3.L2tg mice that can potentially have two β- and three α-chains functionally rearranged in double-positive thymocytes.

Previous studies have reported that mice transgenic for the 2B4 and 5C.C7 TCR are also both dependent on ligand density, whereas the AND TCR transgenic mice show little dependence on ligand levels (18, 19). This phenomenon may also explain observations regarding selection of 3A9 TCR transgenic mice. Thymocytes bearing this TCR are selected well on an H-2^k background, but poorly on an H-2^k/b background (34). Although it is unknown exactly what makes some TCRs highly dependent on ligand density, it seems that those TCRs that are not selected efficiently will be more dependent on the level of selecting ligand. Thymocytes with these TCRs need more time to find their selecting niche, and if the number of niches is effectively reduced by lowering the level of ligand expression, they may not find the niche before death. One way that these cells might be saved is for them to express a second TCR specificity that might be more efficiently selected. We would conclude that the efficiency of selection of the 3.L2tg TCR is low, and when the number of niches is lowered, thymocytes have a better chance for selection using an endogenous TCR specificity.

This interpretation has implications for the involvement of second TCRs in autoimmune recognition. There is evidence to suggest that a T cell that has a receptor against self may be ignorant of self until that cell is activated through a second TCR against some foreign Ag (14). Once activated, this cell may then find the self Ag and become competent to mediate some effector function in response to the self Ag. This scenario depends on the cell having two receptors that are both restricted to self MHC. It has been argued that this is an unlikely event, because once a CD4⁺CD8⁺ thymocyte rearranges a TCR that is competent for positive selection, it should begin that process and stop rearrangements (12, 15). That would be true if TCRs were all uniformly efficient, like the AND TCR. The data presented here, however, would argue that under conditions where selecting niches are limiting, a thymocyte may express one self-restricted TCR, but get selected on a second self-restricted TCR because rearrangements continue until an efficient TCR is produced. The second TCR may not be up-regulated during positive selection, but could be maintained at a low level until activation of the T cell makes its specificity relevant. Due to the large number of T cells produced, it seems inevitable that T cells with two self-restricted TCR specificities will exist in the mature T cell repertoire.

These data also have implications for the mechanism of allelic exclusion at the TCR β locus. A variety of experiments have pointed to a direct link between the β selection step and allelic exclusion at the β locus (3, 12). It is clear that when a functional β-chain makes it to the cell surface of a CD4⁻CD8⁻ thymocyte, a very efficient shut-off of rearrangement at the β locus ensues. What we have observed in the 3.L2tg mouse also points to such a linkage. Breeding of the 3.L2tg mouse to the RAG^−/− background shows that the 3.L2 transgenes are not capable of inducing efficient β selection on their own. These mice are not able to produce a greatly expanded CD4⁺CD8⁺ population, although some cells using the 3.L2 TCR do make it through development. These transgenes are not efficient at β selection, and they are also not efficient at excluding endogenous TCR β rearrangements. We would conclude that the same signal that goes through β/pre-Tα to induce β selection also induces allelic exclusion at the β locus. This has been suggested by studies involving lck transgenes and studies on the role of pre-Tα in allelic exclusion of TCR β (35, 36).

We know that this mechanism of allelic exclusion is not 100% accurate. Studies in both humans and mice have found that about 1% of peripheral T cells have two TCR β-chains (9, 10, 11). It was therefore of interest to determine whether there might be any other mechanisms to help cells guard against expressing two functional β-chains. We did observe one possible mechanism in the 3.L2tg mice. Only on the H-2^k/k background did we observe good expression of the 3.L2 TCR in the double-positive population. In cases where the ligand density was limiting or the selecting ligand was absent, we saw a much lower level of transgenic TCR on the CD4⁺CD8⁺ cells. This phenomenon seems to act on the whole double-positive population and may not be related to the process of TCR up-regulation on cells undergoing positive selection. Further evidence for this interpretation comes from our observation that 3.L2tg;H-2^k/k mice bred to express the stimulatory Hbb^d allele have a level of 3.L2 TCR on the CD4⁺CD8⁺ cells even higher than that in 3.L2tg;H-2^k/k;Hbb^s mice (data not shown). This occurs despite the fact that this ligand is sufficient to mediate negative selection. We can therefore offer two possible interpretations of this data: 1) CD4⁺CD8⁺ thymocytes may up-regulate a useful TCR even before the process of positive selection starts; or 2) positive selection starts at an early double-positive stage and is noticeably working before negative selection is evident.

From our studies on 3.L2tg mice we conclude that a TCR used during positive selection is the predominant TCR expressed on the surface of the mature T cell. The efficiency with which a TCR mediates positive selection will determine whether it is used for that step. Thus, a self-restricted TCR may get to the periphery on the surface of a T cell even if it was not the receptor used for positive selection. This finding can have important implications for T cell cross-reactivity and autoimmunity.

We thank Ed Palmer for providing a Vβ8.3 T cell hybridoma, Darren Kreamalmeyer and Donna Thompson for help with animal care and breeding, Larry Kane and Daniel Peterson for critical review of the manuscript, and Jerri Smith for assistance with preparation of this manuscript.