Role of the Complementarity-Determining Region 3 (CDR3) of the TCR-β Chains Associated with the Vα14 Semi-Invariant TCR α-Chain in the Selection of CD4⁺ NK T Cells¹

Murine T cells, which express both an αβ-TCR and NK cell markers, such as NK1.1 in C57BL/6 mice, are a recently discovered lymphocyte population distinct from T, B, and NK cells (1, 2, 3). When activated through their TCR, they become cytotoxic and quickly release cytokines, such as IL4 and IFN-γ (4, 5). The physiological functions of these cells remain unclear and they have been implicated in very diverse conditions such as immune responses against infection (6, 7, 8) or tumors (9, 10), maintenance of pregnancy (11), granulomatous response (12, 13), and autoimmune responses (14). More precisely, CD4⁺ NK T cells seem to be a population essential to stimulating IL-12 production by APC (15) and to play crucial roles in early stage of Leishmania major infection (16) and to control acceptance of rat islet xenografts in mice (17). The NK T cells are primarily isolated from the thymus, liver, spleen, and bone marrow, in numbers ranging from 0.5 to 1.5 million in each organ, thus accounting for 0.5–1% of total T cells in the thymus and the spleen and 20–30% of all T lymphocytes of the liver and bone marrow (3, 18). NK T cells (i.e., cells defined as TCR-αβ^int NK1.1⁺ CD4⁺, or double negative (DN)³) are heterogeneous. Most of them are TCRαβ^int, CD4⁺, or DN, use an invariant Vα14-Jα281 TCR α-chain (2) preferentially associated to Vβ 2, 7, or 8⁺ TCR β-chains (1, 3, 19, 20), and are restricted by the MHC Ib CD1d molecules: these obey the original definition of NK T cells (3). Recent FACS and molecular analysis have shown that phenotypic NK T cells also contained a variable proportion of cells that were not restricted by CD1d and used all Vα and Vβ segments (5, 18, 21).

The role of the β-chains of the TCR in the selection and in the recognition of CD1d/ligand complexes may be approached by studying the molecular features of the complementarity determining region 3 (CDR3) of the β-chains. The CD1d molecules present GPI proteins, foreign or altered glycolipids, such as α-galactosyl ceramide or cancer glycolipids, as well as self-glycolipids (22, 23, 24, 25, 26, 27). The lipidic moiety of the ligands is buried in the hydrophobic pocket created by the folding of the α1 and α2 domains of CD1d molecule, whereas the carbohydrate moiety of the glycolipids is exposed to the outside and thus made accessible to the TCR. Indeed, all data available so far concerning the stereochemistry of the recognition of α-galactosyl ceramide by the TCR of Vβ8.2⁺ NK T cells point to the specific recognition of the carbohydrate moiety of the ligand, particularly of the OH group at position 2 and of the α-linkage (28). The TCR β-chains of the CD1d-restricted NK T cells are expected to contribute to recognition of highly hydrophilic structures, a role that may be reflected in the physicochemical features of the CDR3 of the β-chains (29) as suggested by studies on T cell recognition of glycopeptides (30) and on recognition of the carbohydrate moiety of glycolipids by T cell hybridomas (31).

A prerequisite for this analytical study is the availability of homogeneous populations of NK T cells. Because CD4⁺ NK T cells seem essential in immune responses to infection autoimmunity and transplantation, we have focused our study on this population. Thus, TCR-αβ^intNK1.1⁺ CD4⁺ T cells were sorted out of the thymus, spleen, bone marrow, and liver of 6-wk-old C57BL/6 mice and studied for their Vα and Vβ usage. Nearly homogeneous populations of CD4⁺ T cells characterized by the usage of the invariant Vα14-Jα281 α-chain and a skewed Vβ usage (thus, CD1d restricted-NK T cells) were identified in the thymus and the liver. The repertoire of their TCR β-chain could thus be determined and compared with that of the same Vβ-Jβ rearrangements present in conventional, class II-restricted, NK1.1⁻CD4⁺ T cells. No distinctive CDR3 length or amino acid sequence and composition that would be characteristic of CD1d-restricted NK T cells could be found. Moreover, the amino acid distributions within the CDR3s of the β-chains of CD4⁺ NK T cells could not be distinguished from those of conventional CD4⁺ T cells. The invariant α-chain can pair with all β-chains in CD8⁺NK1.1⁻ T cells isolated from Vα14-Jα281 transgenic (Tg) mice on a Cα^−/− background, whereas the vast majority CD4⁺NK1.1⁺ T cells of the Tg mouse used Vβ 2, 7, and 8 segments. Thus, these data suggest that the invariant α-chain is dominant for the interaction of the TCR with the CD1d molecules and may constitute the imprinting of the selection by CD1d molecules.

The C57BL/6 mice used in this study were 6–8 wk old and were obtained from IFFA-Credo (L’Arbresle, France). Vα14-Jα281 Tg mice on Cα^−/− background have been previously described (14).

Single-cell suspensions were prepared from liver, spleen, thymus, and bone marrow. Total liver cells were resuspended in a 80% isotonic Percoll solution (Pharmacia, Uppsala, Sweden) and overlaid with a 40% isotonic Percoll solution. Centrifugation for 30 min at 3000 rpm resulted in the concentration of the mononuclear cells at the 40–80% interface. The collected cells were washed once with PBS supplemented with 2% FCS.

Thymus cells were dissociated and freed of connective tissue by filtration. The thymocytes were depleted of CD8⁺ cells using biotinylated anti-αCD8 Ab (CT-CD8α; Caltag, South San Francisco, CA) and streptavidin beads (Dynal, Oslo, Norway).

Total bone marrow cells were collected by flushing bones (tibia, femur) with 2% FCS in PBS and spleen cells were isolated as for thymus. In both cases, the cell suspension was first incubated with anti-Fcγ III/II (Fcblock, 2.4G2) and then depleted using anti-B220 (RA3-6B2), anti-Mac1 (M1/70), anti-Gr1 (RB6-8C5) biotinylated Abs, and magnetic streptavidin beads. All Abs were purchased from PharMingen (San Diego, CA). After depletion, the cell suspension were kept at 4°C in RPMI 1640 supplemented with 2% FCS.

Cells were first incubated 10 min with Fcblock followed by a 30-min exposure to anti-TCRβ-FITC (H57-597), NK1.1-PE (PK136), CD4-biotin (RM4-5), or CD8-Cy-Chrome (53-6.7) Abs (PharMingen). After two washes, tricolor-streptavidin (Caltag) was added. After further washes, the cells were resuspended in PBS containing 2% FCS and analyzed using a FACSCalibur BD Becton Dickinson (San Jose, CA).

The cells were sorted on FACStar (BD Becton Dickinson) as TCRαβ⁺ NK1.1⁺CD4⁺, TCRαβ⁺ NK1.1⁻CD4⁺, and TCRαβ⁺NK1.1⁻CD4⁺ at a flow rate of 3500 events/s and collected in RPMI 1640 supplemented with 20% FCS.

Total RNA was isolated by ultracentrifugation through a discontinuous CsCl gradient as described (32). cDNA was synthesized from 10 μg of total RNA using a (dT)₁₇ primer, 25 U of Rnasin (Promega, Madison, WI), and 10 U of avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim, Indianapolis, IN) in the provided buffer.

Quantification of PCR products was conducted using a locally developed competitive PCR strategy based on size-altered CD3ε cDNA. Briefly, known number of copies of a plasmid (10²–10⁶ copies) harboring the CD3ε cDNA in which a 4-bp deletion has been introduced were mixed with the cDNA solution and PCR-amplified in the same tube in the presence of CD3ε-specific primers (5′ primer: GCCTCAGAAGCATGATAAGC; 3′ primer: CCTTGGCCTTCCTATTCTTG). PCR was run at saturation (40 cycles; 94°C, 30 s; 60°C, 30 s; 72°C, 30 s). The resulting PCR products were submitted to 5 cycles (94°C, 30 s; 60°C, 30 s; 72°C, 30 s) of extension using a nested fluorescent primer specific of CD3ε (FAM-CCCAGAGTGATACAGATGTC) and analyzed for size and peak area on an automated 373A sequencer (Perkin-Elmer, Norwalk, CT). The number of CD3ε copies contained in the samples was determined from a calibration curve.

A volume of cDNA solution containing 10⁴ copies of cDNA CD3ε was PCR-amplified using each of the 24 Vβ-specific primers (33) and a fluorescently labeled Cβ-specific primer, (94°C, 30 s; 60°C, 30 s;72°C, 30 s) during 31 cycles, thus remaining within the exponential phase of amplification. Products resulting of this PCR were analyzed on an automatic sequencer. The size and the intensity of each band were recorded and then analyzed using Immunoscope software (Applied Biosystems, Foster City, CA) (34). A similar quantitative approach was conducted for analysis of the Vα repertoire. Vα-specific primers, with the exception of Vα 9 and 15, have been described previously (12, 35) (Vα9 primer: ACACCGTTGTTAAAGGCACC; Vα 15 primer: GAGCCAAAGACTTATAGTTTT).

One microliter of the PCR product solution resulting from a first amplification step using the Vβ 8.2-specific primer were further amplified for 20 more cycles (94°C, 45 s; 60°C, 45 s; 72°C, 1 min) using a nested primer Jβ2.5-chain and the PFU DNA polymerase (Stratagene, La Jolla, CA) (32).

The PCR products could be further studied by run-off analysis using fluorescently labeled Jβ-specific primers or were cloned for sequencing (32).

Blunt PCR products were ligated into the commercially available vector Blunt II-Topo (Invitrogen, Groningen, The Netherlands) and cloned in Escherichia coli (Invitrogen). Sequence reactions were conducted on positive clones using standard protocols (Perkin-Elmer), and nucleotide sequences were determined using the ABI-Prism 373 sequencing equipment (Applied Biosystems).

Single-cell suspensions prepared from thymus, spleen, liver, and bone marrow were first depleted using a mixture of anti-CD19, Mac-1, Gr-1, and CD8 mAbs. The CD4⁺TCRαβ^intNK1.1⁺ T cells (which accounted for 15, 30, 1.2, and 5% of the depleted cell population in the thymus, liver, bone marrow, and spleen, respectively) were further sorted out as described in Materials and Methods. The purity of the sorted cells exceeded 98% for all four organs (the isolation of spleen cells is given as example, Fig. 1, B1 and B2). The average final recoveries for individual mice of sorted NK T CD4⁺ cells were ∼10⁵, 2 × 10⁵, 3500, and 9000 cells in the thymus, liver, bone marrow, and spleen, respectively. Conventional NK1.1⁻CD4⁺TCRαβ^high T cells were sorted out from spleen using settings that excluded TCRαβ^intNK1.1⁻ T and were also found 98% pure (Fig. 1, C1 and C2).

The usage of all Vα segments by sorted NK1.1⁺TCRαβ^intCD4⁺ T cells was determined using a semiquantitative PCR-based analysis and compared with that of conventional CD4⁺ T cells isolated from age- and sex-matched, naive C57BL/6 mice. The Vα14 segment accounted for 75, 95, 40, and 40% of total Vα segments used by thymus, liver, spleen, and bone marrow NK1.1⁺TCRαβ^intCD4⁺ T cells, respectively, compared with 3–5% detected in conventional CD4⁺ T cells and was thus overrepresented in all four organs (Fig. 2). The size distribution of the CDR3 of the rearranged Vα14 chains was determined using the Immunoscope technique. The analysis of total T cells of the spleen showed that the Vα14⁺ α-chain could use any size of the CDR3 (Fig. 3). By contrast, the Vα14-Cα Immunoscope analysis of the NK1.1⁺TCRαβ^intCD4⁺ T cell populations sorted out of all four organs showed the presence of a unique 10 aa long CDR3 peak, further identified by sequencing and use of a clonotypic primer, as being the Vα14-Jα281 NK T cell-specific rearrangement (Fig. 3).

The Vβ segments used by cells of the four NK1.1⁺TCRαβ^intCD4⁺ T cell populations and of CD4⁺NK1.1⁻TCRαβ^high conventional T cells sorted from the spleen and liver were determined (Fig. 4). The Vβ repertoire of the liver and thymus NK1.1⁺TCRαβ^intCD4⁺ T cell populations was nearly completely skewed toward the usage of the Vβ 2, 7, and 8 segments. The skewing in the Vβ usage was less marked in the spleen and bone marrow NK1.1⁺TCRαβ^intCD4⁺ T cells but was still apparent. No such skewing was observed in conventional T cells. Altogether, these results confirm on a molecular basis that the most of NK1.1⁺TCRαβ^intCD4⁺ liver T cells use the invariant Vα14-Jα281 α-chain associated with a limited set of β-chains, namely Vβ 2, 7, and 8, a set of markers that identifies the CD1d-restricted NK T cells. These cells primarily also predominate among thymus NK1.1⁺TCRαβ^intCD4⁺ T cells and still contribute to 40–50% of the spleen and bone marrow NK1.1⁺TCRαβ^intCD4⁺ T cells. The second NK1.1⁺TCRαβ^intCD4⁺ T cell population found in spleen and bone marrow displays Vα and Vβ usages more similar to that of conventional T cells: their α and β usages can be estimated by computing the expected contributions of the Vα14⁺-associated TCR β-chains and removing the resulting values from the total Vβ values determined for total spleen and bone marrow NK1.1⁺TCRαβ^intCD4⁺ T cells; then, a Vβ usage close to that of conventional CD4⁺ T cells could be calculated (data not shown). Organ-specific heterogeneity, recently described among NK1.1⁺, TCRαβ^int, DN T cells both by Immunoscope and FACS analyses (18, 21), seems thus to be a general feature of the NK1.1⁺TCRαβ^int T cells. Liver and, to a lesser extent, thymus NK1.1⁺TCRαβ^intCD4⁺ T cell populations are markedly enriched in molecularly defined CD4⁺ NK T cells, which in addition are CD1d restricted because they are not detectable in CD1d^−/− mice (18, 36) (data not shown).

The bias in the Vβ usage observed in CD4⁺ NK T cells could be due to structural constraints associated with the particular folding of the semi-invariant α-chain. Due to the absence of the proper reagents, this problem had only be partly answered by FACS analysis of Vα14-Jα281 Tg mice on a Cα^−/− background (36). Thus, CD4⁺NK1.1⁺TCRαβ cells were sorted out of the liver, spleen, and bone marrow of these Tg mice. The Vβ usage was determined using the same strategy as above (Fig. 5). Eighty-five percent of liver CD4⁺ NK T cells used the Vβ 2, 7, and 8 segments. The remaining 15% were almost equally distributed among the other Vβ segments with a bias in favor of Vβ 1, 9, 10, 12, and 14. By contrast, an identical analysis performed on the CD8⁺NK1.1⁻ T cell population of the same mice showed a use of all Vβ segments, as in wild-type mice. Thus, invariant Vα14-Jα281 α-chain can associate to any Vβ-chain in conventional T cells, but not in CD4⁺ NK T cells.

Liver and thymic NK1.1⁺TCRαβ^intCD4⁺ T cells isolated from C57BL/6 wild-type mice were thus used in the study of the putative Ag recognition region (CDR3) of the β-chains of their TCR, in an analysis aimed at defining the clonality of the populations. A Vβ-Cβ Immunoscope analysis of thymic and liver CD4⁺ NK T cells yielded a gaussian distribution of the CDR3 length of the Vβ2, Vβ7, Vβ 8.1, 8.2, and 8.3 TCR β-chains (Fig. 6), with no evidence of a distortion that would be indicative of dominant expansions. Similarly, the CDR3 size distribution observed for Vβ2, Vβ7, Vβ 8.1, 8.2, and 8.3 TCR β-chains of NK T cells of Vα14-Jα281 Tg mice were also polyclonal (data not shown). In addition to excluding major clonal expansions, these data also exclude the existence of structural constraints due to the length of the CDR3 of the β-chain that would limit its association with the invariant α-chain.

The repertoire of the Vβ 8.2⁺ β-chain of liver and thymus CD4⁺ NK T cells was studied in more detail because of the high frequency of the segment in NK T cells. Nested fluorescent Jβ-specific primers were used to analyze the Vβ8.2-Cβ PCR products (Fig. 6). The Jβ 2.1 to Jβ 2.7 regions were found preferentially used but a similar bias in the Jβ usage has already been reported in conventional T cells (37) (data not shown). All Vβ 8.2-Jβ Immunoscope patterns were found gaussian, with no evidence of expansions due to an Ag-specific proliferation, which would have appeared as a peak distorting the gaussian profile. These gaussian V-C and V-J profiles demonstrate the high degree of polyclonality of the TCR β-chains of the CD4⁺ NK T cells. Assuming that NK T cells respond to Ag stimulation by clonal expansions as conventional T cells do, the gaussian patterns that we observed are indicative of the absence of a specific, acute, or chronic, antigenic stimulation of NK T cells.

The Vβ8.2-Jβ2.5 rearrangements predominate among NK T cells because they represent ∼7 and 4% of all Vβ-Jβ rearrangements used by liver and thymus NK T cells, respectively, on the basis of quantitative PCR analysis on thymic and liver CD4⁺ NK T cells. The Vβ8.2-Jβ2.5 PCR products derived from thymus and liver CD4⁺ NK T cells were thus cloned in E. coli and sequenced. A total of 278 and 372 sequences, respectively, were determined, of which 171 and 213 were different. The differences between “total” and “different” numbers were accounted for by the determination of the same nucleotide sequence in several clones. The size of the clones cannot be determined accurately due to the small number of cells of the original samples; however, it is most probably smaller than that of conventional T cells (38) due to the low frequency of redundant sequences.

The Vβ8.2-Jβ2.5 PCR products of conventional CD4⁺ T cells isolated from spleen were also cloned and sequenced, resulting in a total of 200 independent sequences that were taken as representative of the overall diversity of the CDR3 of Vβ8.2-Jβ2.5 rearranged β-chains, irrespective of the α-chain with which they are associated. The absence of systematic biases in the cloning/sequencing procedures was assessed by comparing the distribution of the CDR3 length, determined by sequencing and determined using the Immunoscope analysis; no obvious bias due to PCR amplification and cloning could be seen (data not shown). The contribution of the CDR3 sequences derived from CD1d-restricted Vα14-Jα281⁺ CD4⁺NK1.1⁻ T cells to the total number of determined sequences (39, 40) was calculated by comparison of Vα14-Cα Immunoscope profiles in CD1^+/+ and CD1^−/− mice and found to be in the 1–2% range and was thus considered as negligible (Ronet et al., manuscript in preparation).

Because its CDR3 length occurs at the highest frequency among rearranged Vβ8.2-Jβ2.5 using β-chains, the nucleotide sequences that encode the 10 aa long CDR3 of Vβ8.2-Jβ2.5 β-chains of CD4⁺ NK T cells and of conventional CD4⁺ T cells were analyzed in detail. The study of 48 liver CD4⁺ NK T sequences, 54 thymus CD4⁺ NK T sequences, and 79 spleen conventional CD4⁺ T cell sequences showed that both N and P additions were about as frequent and extensive in the CD4⁺ T and CD4⁺ NK T cell populations. No germline clone was found among the sequences studied. The Dβ1 and Dβ2 segments were used at random. The genomic Vβ8.2 segment was shortened by 12 or 13 nucleotides in all NK T cell-derived sequences as it was in conventional CD4⁺ T cells (Table I). Thus, recombinational and diversity processes seem to be primarily the same in CD4⁺ NK T and in conventional CD4⁺ T cells.

The 10 aa long CDR3 sequences encoded by the same Vβ8.2-Jβ2.5 β nucleotide sequences described above were aligned (Fig. 7). Positions 1 and 2 (Ser-Gly) were encoded by the genomic BV8S2A2 segment. All sequences retained for analysis differed at least at one position along the sequence, redundant sequences having been excluded. Position 3 (except D encoded by the genomic segment BV8S2A1) was always created by N diversity in both CD4⁺ conventional and CD4⁺ NK T cells. Except at position 3 (and to a much lesser extent at position 4) the diversity of which correlates with the extent of N diversity in the two cell populations, the amino acids present at each of the 10 positions of the CDR3 were primarily the same in CD4⁺ T and CD4⁺ NK T cells. Also, the nature of the amino acids detected at position 3 was primarily the same in both cell types, and these amino acids were used at similar frequencies; they were A, D, E, G, K, N, T, V and A, D, E, G, K, P, T, and V, respectively. No statistical bias in the usage of the protonated or hydrophilic amino acid was associated specifically to CD4⁺ NK T cells, although the CDR3 of CD4⁺ NK T cells was expected to be enriched in amino acids able to interact with the OH side chains of the carbohydrate moiety of glycolipids. Finally, the 3-dimensional display of the frequency of the amino acid used by CD4⁺ NK T and CD4⁺ T cells at defined positions along the CDR3 sequences illustrates the absence of NK T cell- or T cell-specific distribution of amino acids (Fig. 7).

These data confirm and extend the emerging conclusion that NK1.1⁺TCRαβ^int T cells are heterogeneous and, in addition to being either DN or CD4⁺, consist in at least two populations. The usage of an invariant Vα14-Jα281 α-chain and of a set of β-chains skewed toward the nearly exclusive usage of the Vβ 2, 7, and 8 segments, characterizes one of the two populations. The second population is defined by the same apparent cell surface phenotype but differs from the previous one by a usage of the α- and β-segments similar to that of conventional T cells. The existence of two cell populations within the DN, NK1.1⁺TCRαβ^int T cells, has been recently found by FACS and Immunoscope analyses (18, 21). Thus, the phenotypic description of DN or CD4⁺, NK1.1⁺ TCRαβ^int T cells does not define unequivocally a cell population. We propose to restrict the use of the expression NK T cells to the DN or CD4⁺, NK1.1⁺ TCRαβ^int T cells that use the Vα14-Jα281 invariant α-chain in combination with the Vβ 2, 7, and 8⁺ segments. In that respect, the liver and thymus CD4⁺ NK1.1⁺TCRαβ^int T cells obey this definition of NK T cells, and the cells sorted out from these two organs on the basis of the CD4⁺ NK1.1⁺TCRαβ^int phenotype are rather homogeneous populations of NK T cells.

The exclusive usage of an invariant α-chain in thymus and liver NK T cells permits the direct approach of the diversity of the associated β-chains. The Immunoscope profiles of the Vβ-Cβ TCR β-chains associated to liver and thymus NK T cells showed a random usage of the CDR3 length of the associated Vβ 2, 7, and 8 TCR β-chains. This was confirmed by further analysis of NK T cells in Vα14Jα281 Tg mice on Cα^−/− background. No significant skewing of the Jβ usage was detected on a sequence and Immunoscope basis. The Vβ-Jβ Immunoscope analysis confirmed that the thymus and liver Vα14⁺ NK T cells were polyclonal with no evidence for clonal expansions, a profile reminiscent of that of naive CD4⁺ T cells. These findings are difficult to reconcile with the observation using FACS analysis that NK T cells display cell surface markers of activated and memory T cells (data not shown) (18, 21), whereas clonal expansions are a hallmark of conventional activated and memory T cells. This becomes understandable if the replication rate of NK T cells is more limited than that of T cells, as already suggested (3). The absence of clonal expansion may also be explained by the absence of activation of NK T cells at the periphery by nonself-ligands, the activated phenotype being due to repeated exposure to a large array of diverse self-glycolipids (25).

The comparison of the CDR3 sequences of CD4⁺ NK T and T cells did not suggest the existence of any distinctive feature that could unequivocally be attributed to the β-chain of the TCR of NK T cells. On the contrary, the 10-aa-long sequences of the CDR3 Vβ8.2-Jβ2.5 rearrangements of NK T and of conventional CD4⁺ T cells shared similar features; the details of the machinery generating the β-chain diversity were also the same in T and NK T cells. Thus, most of the properties of the TCR of NK T cells seem to be dictated by the presence of the invariant α-chain. The differentiation pathway of NK T cells, as well as the moment at which they segregate from the mainstream of T cell differentiation pathway, are not well understood. The CDR3s of the β-chains of NK T cells are not different from those found associated to randomly taken TCR α-chains of conventional CD4⁺ T cells. Studies on the Vα14 Tg mice on a Cα^−/− background revealed that peripheral CD4⁺ NK T cells used mainly Vβ 2, 7, or 8, whereas the same invariant α-chain can associate with any Vβ⁺ chain in conventional CD8⁺NK1.1⁻ T cells. Thus there is no structural constraint caused by a particular folding of the invariant α-chain. The predominant role of the α-chain in the selection of NK T cells on CD1d molecules is strongly suggested by our data. Thus the clonotypic α-chain behaves as some kind of a print of the recognition of the CD1d molecules.

In this respect, the similarity of the CDR3 sequences of naive CD4⁺ T and of CD4⁺ NK T cells, although the two populations are clearly endowed by distinct functional properties, questions the role of the β-chain in the selection of NK T cells. The role of the β-chain in carbohydrate recognition is supported by the finding that some NK T cell clones, and also ex vivo NK T cell populations, can discriminate between sugar-derivatized ceramides and thus identify differences in the structure of carbohydrates (41, 42). Recently acquired data are more in favor of the conclusion that the CDR3 of the TCR β-chain of NK T cells is endowed with a highly degenerated recognition ability. Indeed, the analysis of the human Vα24⁺ Vβ11⁺ NK T cells in vitro expanded in the presence of α-galactosyl-ceramide was indicative of a polyclonal activation of NK T cells with no evidence for predominantly expanding clones (43). Also, the isolation of CD1d-restricted NK T cells with CD1d tetramers loaded with αGalCer results in the sorting of the entire population that uses the Vα14-Jα281 α-chain, whatever NK1.1⁺ or NK1.1⁻ (40) rather than the sorting of discrete, oligoclonal, Ag-specific populations, as for class I tetramers loaded with peptides (44). Only the 3-dimensional structure of CD1d-glycolipid-TCR complexes will definitely settle the problem of the relative contribution of TCR α- and β-chains to the recognition of glycolipids.

We thank Drs. P. Bousso and D. Guy-Grand for helpful discussions and critical reading of the manuscript. We thank A. Lim and S. Darche for technical advice, S. Dalle for construction of the CD3ε plasmid, and A. Louise and H. Kiefer-Biasizzo for the cell sorting.