Figure 11.2 illustrates the principal structural features of the CD3 γ, δ, ε, and ζ polypeptides as derived from gene cloning and sequencing,^42,51 and more recently by protein crystal structures of the extracellular domains of γ, δ, and ε.^52,53,54,55 The extracellular domains of the γ, δ, and ε chains show a significant degree of similarity to one another. These domains retain the cysteines that have been shown to form intrachain disulfide bonds and each consists of a single Ig superfamily domain. The spacing of the cysteines in these domains produces a compact Ig-fold, similar to a constant region domain. The γ and δ subunits form distinct heterodimers with ε via highly conserved residues at the dimerizing interface.^52,53,54,55 The connecting peptides of the CD3 γ, δ, and ε chains all contain highly conserved, closely spaced cysteines just before the membrane-spanning regions. These residues are likely candidates for the formation of interchain disulfide bonds and appear to play a role in the assembly of the CD3 and TCR polypeptides.^52,56 The extracellular domain of the ζ chain consists of only nine amino acids and contains the only cysteine, which is responsible for the disulfide linkage of the ζζ homodimer or the ζ F_cεRIγ heterodimer. Each of the γ, δ, ε, and ζ polypeptides contain a conserved, negatively charged amino acids in their transmembrane region complementary to the positive charges seen in the TCR transmembrane regions.^57,58,59,60

The cytoplasmic regions of the γ, δ, ε, and ζ chains are the intracellular signaling “domains” of the TCR heterodimer. Each of these molecules contains one or more amino acid sequence motifs that can mediate cellular activation.⁶¹ The intracellular sequences responsible for this activation are contained within an 18 amino acid conserved ITAM⁶² with the sequence X₂YX₂L/IX₇YX₂L/I. Both of the tyrosines in this motif are absolutely required to mediate signal transduction as mutation of either completely prevents the mobilization of free calcium or cytolytic activity.⁶³ This sequence occurs three times in the ζ chain and once in each of the CD3 γ, δ, ε, and F_cεRI γ chains. There are also pairs of tyrosines present in the cytoplasmic domains of the γ, δ, ε, and ζ chains. This sequence motif is also present in the mβ-1 and B29 chains associated with the Ig β-cell receptor and in the F_cεRI β-chain but there are many more¹⁰ in TCR/CD3 than in any other receptors which use ITAMs. The tyrosines in these cytoplasmic sequences are substrates for the tyrosine phosphorylation that is one of earliest steps in T-cell signaling₆₁ and is thought to occur aberrantly in nonproductive T-cell responses (eg, antagonism; see following). Serine phosphorylation of the CD3γ also occurs upon
antigen or mitogenic stimulation of T cells⁶⁴ and may play a role in T-cell activation as well.

Recently, live-cell imaging studies have shown a close interaction of the CD3epsilon cytoplasmic domain of the TCR with the plasma membrane, using fluorescence resonance energy transfer (FRET) between a C-terminal fluorescent protein and a membrane fluorophore.⁶⁵ Electrostatic interactions between basic CD3epsilon residues and acidic phospholipids enriched in the inner leaflet of the plasma membrane were required for binding. Nuclear magnetic resonance studies of the lipid-bound state of this cytoplasmic domain revealed a deep insertion of the two key tyrosines of the ITAM into the hydrophobic core of the lipid bilayer, likely preventing their phosphorylation.⁶⁵ Similar studies of CD3 zeta have confirmed that this is also the case for that component of the TCR/CD3 complex as well.⁶⁶ This then likely defines the “off” state of the TCR/CD3 complex and is a novel explanation for how it can prevent tyrosine phosphorylation by lck. What is still mysterious is how ligand binding by the TCR disengages these signaling modules from the plasma membrane and triggers the kinase cascade needed to activate T cells.

The assembly of newly formed TCR-α and -β chains with the CD3 γ, δ, ε, and ζ chains and their intracellular fate have been studied in detail.^{39,40,41,61,63} Early studies have focused on mutant hybridoma lines, which fail to express TCR on their cell surface, and on transfection studies using complementary DNA for the different chains in the receptor; but recently, Wucherpfennig and colleagues have developed an elegant in vitro translation and assembly system that has clarified a number of important issues.^54,55

Experiments in a nonlymphoid cell system⁶⁷ have shown that TCR-α can assemble with CD3 δ and ε but not CD3 γ
and ζ. In contrast, the TCR-β chain can assemble with any of the CD3 chains except the ζ chain. When the ζ chain was transfected with either α or β chain genes, or any of the three CD3 chains, no pairwise interaction occurred.

Only when all six complementary DNAs were cotransfected was it shown that the ζ chain could be coprecipitated with the other chains.⁶⁷ Based on these data, a model has been proposed that suggests that TCR-α pairs with CD3 δ and ε chains and that TCR-β pairs with the CD3 γ and ε chains in the completed molecule. The ζ chain is thought to join the TCR and other CD3 polypeptides in that last stage of assembly.

Pulse-chase experiments have shown that all six chains are assembled in the endoplasmic reticulum (ER), transported to the Golgi apparatus, and then transferred to the plasma membrane. It also appears that the amount of ζ chain is rate limiting, as it is synthesized at only 10% the level of the other chains. This results in the vast majority of newly synthesized α, β, or CD3 components being degraded within 4 hours of their synthesis. The remaining nondegraded chains are long lived due to the formation of complete TCR/CD3 complexes with the limiting ζ chain.⁶⁸ TCR/CD3 lacking CD3 ζ chains migrate through the ER and Golgi intact but then are transported to and degraded in the lysosomes. The immunologic significance of this pre-Golgi degradation pathway is most evident in CD4+CD8+thymocytes where, despite high levels of synthesis of both messenger ribonucleic acid and protein for all the TCR, CD3, and ζ chains, surface expression is relatively low. The TCR chains in immature thymocytes seem to be selectively degraded.⁶⁸ Thus posttranslation regulation appears to be an important means of controlling the cell surface expression of TCR heterodimers.

The TCR and CD3 γ, δ, and ε chains contain ER retention signals.^68,69 If the γ and δ signals are removed, then the chains are transported through the Golgi and rapidly degraded in the lysosomes. In contrast, removal of the CD3 ε ER retention signal allows this chain, and any associated chains, to be transported to the cell surface. Thus, association of the TCR and other CD3 chains with ε renders their ER retention signals inoperative. However, the ε ER retention signal remains functional. This prevents the surface expression of partial complex intermediaries until CD3 ζ is incorporated into the complex, which then masks the ε ER retention signal and allows the transport of mature complexes to the cell surface.

The overall stoichiometry of the αβ TCR/CD3 complex is controversial. The work of Call and colleagues⁵⁷ has shown the relationships between different CD3 dimers (γε, δε, and ζ ζ) and a single TCR αβ heterodimer (as shown in Fig. 11.3). Using mutagenesis, they found very specific interactions based on the positive charges in each TCR transmembrane domain with complementary negative charges in the transmembrane domains of the different CD3 components. The two positively charged residues of TCR-α mediate interactions with the negatively charged residues of the of CD3δε and CD3ζζ dimers, while the single positively charged residue of TCR-β mediates interactions with the negatively charged residues of CD3γε. In addition, their data suggest a highly ordered assembly process as they found that the TCR/CD3δε association facilitates the assembly of CD3γε into the complex and that the association of the TCR and CD3 heterodimers was a prerequisite for incorporation of CD3ζζ into the complex. Importantly, the data show that there is only one TCR heterodimer per nascent TCR/CD3 complex in their in vitro expression system.

In contrast, a number of groups have found evidence that there can be two TCR heterodimers in a given TCR/CD3 cluster on T-cell surfaces. In particular, Terhorst and colleagues⁴² showed that in a T-T hybridoma, a monoclonal antibody against one TCR-αβ pair could comodulate a second αβ heterodimer. In addition, sucrose gradient centrifugation of TCR/CD3 showed a predicted molecular weight of 300 kDa, more than 100 kDa larger than expected from a minimal δ subunit complex (α, β, γ, δ, ε₂, ζ₂).⁷⁰ Another study suggesting that there are least two TCRs in a given CD3 complex is the Scatchard analysis indicating that the number of CD3ε molecules on a T-cell surface equals the number of αβ TCRs.^71,72,73 Finally, there is the work of Fernandez-Miguel et al.,⁷⁴ who showed that in T cells which have two transgenic TCR-β chains, antibodies to one Vβ can immunoprecipitate the other. It was also found that they are often close enough to allow fluorescence energy transfer, meaning that the two TCR-βs in a cluster are within 50 Angstroms of each other.⁷⁴ Interestingly, it appears that the TCR complexes with CD3 either have CD3 γ or CD3 δ, but not both, and these two
receptor types are expressed in different ratios in different cells. These data may not be irreconcilable, because while the initial TCR/CD3 assembly may involve only one TCR, these may dimerize or multimerize later on the cell surface.^50,75

The composition of TCR/CD3 complexes on γδ T cells is distinct from that of αβ T cells and changes with the activation state of the cell. Biochemical analysis showed that most murine γδ TCRs contain only CD3 γε dimers. Interestingly, a differentially glycosylated form of CD3 γ was found to associated with γδ TCRs dependent on the activation state of the cells.⁷⁶ In addition, while C3ζζ is incorporated into the complexes of naïve cells, activation results in the expression and incorporation of F_cεRI γ into the γδTCR complex.⁷⁶ Using quantitative immunofluorescence, Hayes and Love have derived data and proposed a model of murine γδTCR stoichiometry in which there are two CD3γε dimers, as well as one CD3ζ dimer in each TCR complex.⁷⁷ Taken together, these findings strongly suggest that signal transduction through the TCR will occur differently in γδ versus αβ T cells.

In humans and in mice, there is a single α-chain C-region gene that is composed of four exons encoding: 1) the constant region domain, 2) 16 amino acids including the cysteine that forms the interchain disulfide bond, 3) the transmembrane and intracytoplasmic domains, and 4) the 3′ untranslated region (see Fig. 11.4). The entire α/δ locus spans about 1.1 MB in both mice and humans. There are 50 different J-region gene segments upstream of the C-region in the murine locus. At least eight of these J-regions are nonfunctional because of in-frame stop codons or rearrangement and splicing signals that are likely to be defective. A similar number of α-chain J-regions are present in the human locus. This very large number of J-regions compared to the Ig loci may indicate that
the functional diversity contributed by the J segment of the TCR (which constitutes a major portion of the complementarity-determining region [CDR]3 loop) makes an important contribution to antigen recognition (see the following).

In both the murine and human loci, the Cδ, Jδ, and two Dδ gene segments are located between the Vα and Jα gene segments. In the murine system, there are two Jδ and two Dδ gene segments on the 5′ side of Cδ and the Cδ exons are approximately 75 kb upstream of the Cα gene and approximately 8 kb upstream of the most 5′ known Jα gene segments. The human organization is similar, with three Dα gene segments and three Jδs. Surprisingly, in both species all of the D elements can be used in one rearranged gene rather than alternating as is the case with TCR-β or IgH. That is, in mice one frequently finds Vδ D₁, D₂, and Jδ rearrangements,⁸⁶ and in humans Vδ, D₁, D₂, D₃, and Jδ.⁸⁷ This greatly increases the junctional or CDR3 diversity that is available, especially because of the potential for N-region addition in between each gene segment. This property makes TCR-δ the most diverse of any of the antigens receptors known, with approximately 10¹² to 10¹³ different amino acid sequences in a relatively small (10 to 15 amino acid) region.⁸⁶ The implications for this and comparisons with other antigen receptor genes are discussed subsequently.

The location of Dδ, Jδ, and Cδ genes between Vα and Jα gene segments suggests that TCR-δ and -α could share the same pool of V gene segments. While there is some overlap in V gene usage, in the murine system, four of the commonly used Vδ genes (Vδ1, Vδ2, Vδ4, Vδ5) are very different than known Vα sequences and they have not been found to associate with Cα.⁸⁸ The other four Vδ gene families overlap with or are identical to Vα subfamilies (Vδ3, Vδ6, Vδ7, and Vδ8, with Vα6, Vα7, Vα4, and Vα11, respectively).

The mechanisms that account for the preferential usage of certain gene segments to produce δ versus α chain are not known. While some Vδ genes are located closer to the Dδ and Jδ fragments than Vα genes (such as Vδ1), other Vδs (such as Vδ6) are rarely deleted by Vα Jα rearrangements and thus seem likely to be located 5′ of many Vα gene segments.

One of the Vδ gene segments, Vδ5, is located approximately 2.5 kb to the 3′ of Cδ in the opposite transcriptional orientation and rearranges by inversion. Despite its close proximity to Dδ Jδ gene segments, Vδ5 is not often found in fetal γδ T cells. Instead, the Vδ5 → DJδ rearrangement predominates in adult γδ T cells.

An implicit characteristic of the α/δ gene locus is that a rearrangement of Vα to Jα deletes the entire D-J-C core of the δ-chain locus. In many αβ T cells, the α-chain locus is rearranged on both chromosomes and thus no TCR δ could be made. In most cases, this is due to Vα → Jα rearrangement, but evidence suggesting an intermediate step in the deletion of TCR-δ has been reported.⁸⁹ This involves rearrangements of an element termed T early alpha (TEA) to a pseudo-Jα 3′ of Cδ. The rearrangement of TEA to this psuedo-Jα would eliminate the δ-chain locus in αβ T cells. Gene targeting of the TEA element resulted in normal levels of αβ and γδ T cells, but usage of the most Jαs was severely restricted,⁹⁰ suggesting that its function is to govern the accessibility of the most proximal 5′ Jαs for recombination.

The entire human 685 kb β chain gene locus was originally sequenced by Hood and coworkers⁹¹ (Fig. 11.5). One interesting feature is the tandem nature of J_β–C_β in the TCR-β locus. This arrangement is preserved in all higher vertebrate species that have been characterized thus far (mouse, human, chicken, frog). The two C_β coding sequences are identical in the mouse and nearly so in humans and other species. Thus it is unlikely that they represent two functionally distinct forms of C_β. However, the J_β clusters have relatively unique sequences, and thus this may be a mechanism for increasing the number of J_β gene segments. Together with the large number of Jα gene segments, there is far more combinatorial diversity (Jα × Jβ = 50× 12=600) provided by J regions in αβ TCRs than in Igs.

Most of the V-regions are located upstream of the joining and constant regions and in the same transcriptional orientation as the D and J gene element, and rearrange to DβJβ gene via deletion. Similar to the case of Vδ5, a single Vβ gene, Vβ14 is located 3′ to C-regions and in the opposite transcriptional orientation, thus rearrangements involving Vβ14 occur via inversion.

In the NZW strain of mouse, there is a deletion in the β chain locus that spans from Cβ1, up to and including the Jβ 2 cluster.⁹² In SJL, C57BR and C57L mice, there is a large deletion⁹³ in the V-region locus from Vδ5-Vβ9. These mice
also express a V gene, Vβ17, that is not expressed in other strains of mice. Deletion of about half of the V genes (in SJL, C57BR and C57L mice) does not seem to have any particular effect on the ability of these mice to mount immune responses whereas mice that have deleted the J_β2 cluster show impaired responses.⁹⁴

The organization of the mouse and human γ-chain loci are shown in Figure 11.4. The human γ genes span about 150 kb⁸⁵ and are organized in a fashion similar to that of the β chain locus with two JγCγ regions. There is more than one nomenclature commonly used to describe the γ chain genes.^85,96,97,98 Here, we use that of Lefranc and Rabbitts⁹⁹ and Tonegawa and colleagues⁸⁵ for the human and mouse γ chains, respectively. The organization of the human γ chain genes consists an array of Vγs in which at least six of the V-regions are pseudogenes is located 5′ to these JγCγ clusters, and each of the V genes are potentially capable of rearranging to any of the five J-regions. The sequences of the two human Cγ regions are very similar overall and only differ significantly in the second exon. In Cγ2, this exon is duplicated two or three times and the cysteine that forms in the interchain disulfide bond is absent. Thus, Cγ2-bearing human T cells have an extra large γ-chain (55,000 MW) that is not disulfide bonded to its δ-chain partner.

The organization of the murine γ chain genes is very different than that of the human genes in that there are three separate rearranging loci that span about 205 kb.^100,101 Of four murine Cγ genes, Cγ3 is apparently a pseudogene in BALB/c mice, and the Jγ3 Cγ3 region is deleted in several mouse strains including C57 Bl/10. Cγl and Cγ2 are very similar in coding sequences. The major differences between these two genes is in the five amino acid deletion in the Cγ2 gene that is located in the C II exon at the amino acid terminal of the cysteine residue used for the disulfide formation with the δ chain. The Cγ4 gene differs significantly in sequences from the other Cγ genes (in 66% overall amino acid identity). In addition, the Cγ4 sequences contains a 17 amino acid insertion (compared to Cγ1) in the C II exon located at similar position to the five amino acid deletions in the Cγ2 gene.^101a Each Cγ gene is associated with a single Jγ gene segment. The sequences of Jγ1 and Jγ2 are identical at the amino acid level, whereas Jγ4 differs from Jγ1 and Jγ2 at 9 out of 19 amino acid residues.

The murine Vγ genes usually rearrange to the Jγ Cγ gene that is most proximal and in the same transcriptional orientation. Thus Vγ1 rearranges to Jγ4; Vγ2 to Jγ2; and Vγ4, Vγ5, Vγ6, and Vγ7 to Jγ1. Interestingly, some Vγ genes are rearranged and expressed preferentially during γδ T-cell ontogeny and in different adult tissues as well.¹⁰¹ In particular, Vγ5+ and Vγ6+ T cells are generated in the fetal thymus with very limited/no junctional diversity. Instead, the adult thymus produces γδ T cells expressing Vγ1, Vγ2, Vγ4, and Vγ7 gene segments with highly diverse junctional sequences. Moreover, γδ T cells that localize to the secondary lymphoid organs tend to express Vγ4, Vγ1, and Vγ2, whereas those that localize to the intestinal epithelium express Vγ7.^{88,102,103,104,105,106} There are also reports suggesting that some intestinal epithelial γδ T cells develop extrathymically.¹⁰⁷ Regardless, it has been suggested that the Vγ gene rearrangement is a programmed process.^108,109

It has become increasingly apparent that transcriptional accessibility and rearrangement of TCR and Igloci are closely linked, following the early work of Alt and Yancopoulous.¹¹⁰ Factors governing accessibility and rearrangement include histone methylation,^{111,112,113,114} DNA methlyation, and the presence of enhancer and specific promoter elements.¹¹⁵ Even specific variations in the recombination signal sequences have been shown to elicit specific biases in V(D)J joining.⁸⁶ With respect to enhancer elements in the TCR loci, these were first identified in the TCR-β locus, 3′ of Cβ2,^116,117 and subsequently for the other TCR loci as well,¹¹⁵ as indicated in Figure 11.4. These TCR enhancers all share sequence similarities with each other. Some of the transcriptional factors that bind to the TCR genes are also found to regulate Ig gene expressions. It has been shown that TCR-α enhancer (Eα) is not only important for normal rearrangement and expression for the α chain locus but also is required for the normal expression level of mature TCR-δ transcripts.¹¹⁸ Also interesting is the work of Lauzurica and Krangel,^119,120 who have shown that a human TCR-δ enhancer-containing minilocus in transgenic mice is able to rearrange equally well in αβ T cells as in αδ T cells but that an Eα-containing construct was only active in αβ lineage T cells. Similar to Ig genes, promoter sequences are located 5′ to the V gene segments. Although D → J_β rearrangement and transcription occur fairly often in B cells and in B-cell tumors,¹²¹ V_β rearrangement and/or transcription appears highly specific to T cells. In addition to enhancers, there also appear to be “silencer” sequences 3′ of Cα^122,123 and in the Cγ1 locus.¹²⁴ It has been suggested that these “repressor sites” could turn off the expression of either of these genes, influencing T-cell differentiation toward either the αβ or the γδ T-cell lineage.

The murine TCR Cγ1 gene cluster comprises four closely linked Vγ gene segments, in the order Vγ7, 4, 6, and 5, which rearrange to a single common downstream J gene segment, Jγ1 (see Fig. 11.4). In early fetal thymocytes, rearrangements of Vγ5 and Vγ6 genes predominate, and the resulting Vγ3+ and Vg4+ cells migrate to the skin or reproductive tissue, respectively. Later in ontogeny, Vγ4 and Vγ7 rearrangements predominate, and cells expressing these V regions migrate from the adult thymus to the secondary lymphoid organs and the intestinal epithelium.^100,101 At least two cis-acting, enhancer/locus control region (LCR) elements are present in the Cγ1 cluster. One is a T cell-specific transcriptional enhancer, 3γEγ, located 3 kb downstream of the Cγ1 gene segment.¹²⁵ A second element, “has,” was found between the Vγ7 and Vγ4 genes, based on DNAse I hypersensitivity.¹²⁶ Similar enhancers have also been found to be associated with the Cγ2 and Cγ3 genes.⁹⁵ Experiments suggest that simultaneous deletion of both enhancer elements in Cγ1 cluster severely diminishes TCR-γ transcription, but only modestly reduces TCR-γ gene rearrangement, while deletion of each element
separately has little effect.¹²⁷ In contrast to these results in thymocytes, deletion of “has” alone reduces transcription of one Vγ gene specifically in peripheral γδ T cells. Thus, the two elements not only exhibit functional redundancy in thymocytes but also have unique functions in other settings.

In Igs, normally only one allele of the heavy chain locus and one of the light chain alleles is productively rearranged and expressed, a phenomenon termed “allelic exclusion.” With respect to αβ TCR expression, while TCR-β exhibits allelic exclusion,¹²⁸ TCR-α seems much less constrained,^129,130 and many mature T cells express two functional TCR-α chains. As the chances of forming an in-frame joint with any antigen receptor is only one in three, the probability that a T cell would have two productively rearranged TCRαs is only 1/3 × 1/3 = 1/9, or 11%. However, even when this happens, the two TCR-α chains may not form heterodimers equally well with the single TCR-β that is expressed; thus, only one heterodimer may be expressed. But this simple calculation is complicated by the likelihood that only thymocytes that make at least one productive TCR rearrangement of each type will have a chance at maturation, which would eliminate almost half of the T cells (four-ninths, which is the product of a two-thirds chance of failure on one chromosome, followed by the same failure rate on the second). Secondly, it has been found that there is a mechanism called receptor editing, which means that a given rearrangement that is not productive, either because of an out-of-frame joint or for reasons of self reactivity, can induce a V region further 5′ or “upstream” of the initial VJ joint to rearrange to one of the remaining Jas.¹³¹ In any event, it has been reported that one-third of human T cells express two TCRs,¹³² which is slightly higher than what is expected from the probabilities discussed previously (approximately 20%), perhaps reflecting the effect of receptor editing.

There also appears to be an important role for the pre-TCR heterodimer (eg, pre-Tα:TCR-β) in blocking further TCR-β rearrangement and thus ensuring allelic exclusion at that locus.^133,134 In particular, pre-Tα-deficient mice had a significant increase in the number of cells with two productive TCR=β rearrangements, compared with wildtype mice.¹³³

Although the basic organization and V(D)J recombination machinery are shared between TCR loci and Igs, there are a number of striking differences. One of these is somatic hypermutation. In antibodies, this form of mutation typically raises the affinities of antigen specific Igs several order of magnitude, typically from the micromolar (10^-6 M) to the nanomolar (10^-9 M) range.^135,136 We now know that most cellsurface receptors that bind ligands on other cell surfaces, including TCRs, typically have affinities in the micromolar range (see later section) but that they compensate for this relatively low affinity by engaging multiple receptors simultaneously (eg, increasing the valency) and by functioning in a confined, largely two-dimensional volume (eg, between two cells). Cells employing such receptors require weak, but highly specific, interactions so that they can disengage quickly.^137,138 The rapid off-rates seen with TCRs (see later section) may even amplify the effects of small numbers of ligands.^139,140

There has also been no enduring evidence for a naturally secreted form of either an αβ or γδ TCR. Here again, it can be argued that such a molecule would have no obvious use as the affinities are too low to be very useful in solution. Thus, for most TCRs, the concentration of protein would have to be extremely high in order to achieve an effect similar to soluble antibodies (in the milligram/milliliter range).

A third mechanism seen in antibodies but not TCRs is C_H switching, which allows different Ig isotypes to maintain a given V region specificity and associate it with different constant regions that have different properties in solution (such as complement fixation, basophil binding, etc.). As there is no secreted form of the TCR, this feature would also lack any obvious utility.

Where TCRs are equal—and in fact generally superior— to Igs is in the sheer number of possible receptors that can be generated through recombination alone. Table 11.1 summarizes the potential V region diversity that TCRs are capable of when the number of V region gene segments is multiplied by D, J, and N region diversity. It can be seen from this table that while the V region number is generally lower in murine TCRs, particularly TCR-δ and TCR-γ, this is more than compensated for by the degree of junctional diversity (where V and J or V, D, and J come together) and chain combinations, such that overall TCRs have orders of magnitude greater potential diversity than Igs. This junctional region corresponds to CDR3 as originally defined by Kabat and Wu for Igs.¹⁴¹ With respect to αβ TCRs, the concentration of diversity in this region (in both chains) can be explained by the key role that these sequences play in recognizing diverse peptides in MHC molecules (see later section), as supported by mutagenesis and structural studies. For γδ TCRs and Igs, however, the diversity is almost all in just one chain (TCRδ and IgH, respectively), and the implications of this are discussed subsequently. Recent work using high throughput sequencing techniques are in remarkable agreement with these crude early estimates of TCR-β diversity.¹⁴²

Because CDR regions are loops between different β strands of an Ig or TCR V region (see later section), the configurations they adopt are generally very sensitive to their length, such that a difference of even one amino acid may produce a significant change in the overall structure.^3,143 A comparison of CDR3 length distributions between the αβ TCRs, γδ TCRs, and Igs (see Fig. 11.5)^144,145 showed that those of TCR-α and -β have a very constrained distribution of lengths and that these are nearly identical in size. These length constraints may reflect a requirement for both the α
and β chains of TCRs to contact both the MHC molecules and bound peptides on the same plane, as borne out by structural studies (see later section). In contrast, the CDR3s of Ig heavy chains are long and variable, whereas those of Ig light chains are short and constrained. This may reflect the fact that Igs recognize both very small molecules (eg, haptens) as well as very large ones (eg, proteins). Surprisingly, γδ TCR CDR3 length distributions are similar to those of Igs in that the CDR3 lengths of TCR-δ chains are long and variable, whereas those of the TCR γ chains are short and constrained. Thus, on the basis of this measure of ligand recognition, one might expect γδ TCRs to be more similar to Igs than to αβ TCRs. This has been validated in subsequent biochemical and structural studies (see later sections).

The chromosomal locations of the different TCR loci have been delineated in both mouse and humans, and the results are summarized in Table 11.2. One significant factor in cancers of hematopoietic cells are chromosomal translocations that result in the activation of genes normally turned off or the inactivation of genes that are normally turned on. Thus, B- or T-lymphocyte neoplasia is frequently associated with inter- or intrachromosomal rearrangements of Ig or TCR loci or in some cases both.^146,147

These translocations seem to mediated by the V(D)J recombinase machinery, indicating the inherent danger and need for tight regulation of this pathway. Such rearrangements are particularly common in the α/δ locus, perhaps because this locus spans the longest developmental window in terms of gene expression, with TCR-δ being the first and TCR-α the last gene to rearrange during T-cell ontogeny (as discussed in more detail in the following). In addition, the α/δ locus is in excess of 1 mb in size, and this provides a larger target for rearrangement than either TCR-β or TCR-γ. Interestingly, in humans, TCR-α/δ is on the same chromosome as the IgH locus and V_H → J_α rearrangements (by inversion) have been observed in some human tumor material.^148,149 The functional significance of this is not known.

Particularly frequent is the chromosome 8-14 translocation [t(8;14) (q24;q11)] that joins the α/δ locus to the c-myc gene, analogous to the c-myc → IgH translocation in many mouse myeloma tumors and in Burkitt lymphomas in humans. In one cell line, a rearrangement occurred between the Jα-region coding sequences, and a region 3′ of c-myc.¹⁵⁰ In both B- and T-cell malignancies, the translocation of c-myc into IgH or TCR-α/β appears to increase the expression of c-myc and may be a major factor in the unregulated cell growth that characterizes cancerous cells. Other
putative proto-oncogenes that have been found translocated into the TCR-α/β locus are the LIM domain-containing transcription factors Ttg-1¹⁵¹ and Ttg-2,^152,153 which are involved in neural development; the helix-loop-helix proteins Lyl-1¹⁵⁴ and Scl,¹⁵⁵ which are involved in early hematopoietic development; and the homeobox gene Hox 11,¹⁵⁶ which is normally active in the liver. How these particular translocations contribute to malignancy is unknown, but they presumably causes aberrations in gene expression that contribute to cell growth or escape from normal regulation. In patients with T-cell leukemia infected with the human T-cell lymphotrophic-I virus, there are large numbers of similar translocations; it is thought that human T-cell lymphotrophic-I itself is not directly leukemogenic but acts by causing aberrant rearrangements in the T cell that it infects, some of which become malignant.

Another disorder, which frequently associates with TCR and Ig locus translocation, is ataxia telangiectasia, an autosomal recessive disorder characterized by ataxia, vascular telangiectasis, immunodeficiency, increased incidence of neoplasia, and an increased sensitivity to ionizing radiation. Peripheral blood lymphocytes from patients with ataxia telangiectasia have an especially high frequency of translocations involving chromosomes 7 and 14.¹⁵⁷ These sites correspond to the TCR-γ, -β, and -α loci, and the Ig heavy chain locus. Thus, it appears as though one of the characteristics of patients with ataxia telangiectasia is a relatively error-prone rearrangement process that indiscriminately recombines genes that have the TCR and Ig rearrangement signals.¹⁵⁸

As discussed previously, the sequences of TCR polypeptides show many similarities to Igs, and thus it has long been suggested that both αβ and γδ heterodimers would be antibodylike in structure.^24,25,159 The similarities between TCRs and Igs include the number and spacing of specific cysteine residues within domains, which in antibodies form intrachain disulfide bonds. Also conserved are many of the inter- and intradomain contact residues and, in addition, secondary structure predictions are largely consistent with an Ig-like “β barrel” structure. This consists of three to four antiparallel β strands on one side of the “barrel” facing a similar number on the other side, with a disulfide bridge (usually) connecting the two β “sheets” (sets of β strands in the same plane). All Ig variable and constant region domains have this structure, with slight variations in the number of β strands in variable region domains (by convention including V, D, and J sequences) compared with constant domains.

Efforts to derive x-ray crystal structures of TCR heterodimers and fragments of heterodimers presented many technical hurdles.¹⁶⁰ One difficulty is that structure determination required engineering the molecules into a soluble form. A second problem is that many of the TCRs are heavily glycosylated, and it was necessary to eliminate most or all of the carbohydrates on each chain to achieve highquality crystals. An alternative is to express soluble TCRs in insect cells, where they have compact N-linked sugars, or in Escherichia coli, where they are unglycosylated. The first successes in TCR crystallization come from the laboratory of Mariuzza and collaborators who solved the structure of first a Vβ Cβ polypeptide¹⁶¹ and then a Vα fragment.¹⁶² In the following year, the first complete αβ TCR structures were solved.^163,164 The structure of the 2C TCR, by Garcia and colleagues, is shown in Figure 11.6.¹⁶³ In general, as predicted from sequence homologies, these domains are all Iglike, with the classical β-barrel structure in evidence in all three domains. At each end of the barrel in each V-region domain there are four loops between the β sheets, three of which form the CDRs of Igs, which are numbered in Figure 11.6. The fourth loop, between the D and E strands, has been implicated in superantigen binding. The six CDR loops from the two variable domains form the antigenbinding surface in both Igs and TCRs. The major anomaly in terms of similarity of TCRs to Igs is the structure of the Cα.¹⁶⁵ Cα consists of one-half of the classical β-barrel, that is, one set (or “sheet”) of β strands while the rest of the partially truncated domain exhibits random coils. This type of structure is unprecedented in the Ig gene family. The functional significance of such a variant structure in unknown, but it has been suggested that this incompletely formed Iglike domain may be responsible for the observed lability of TCR-α, and this may allow greater flexibility in the regulation of its expression. Another possible explanation is that this configuration is designed to accommodate one or more of the CD3 molecules.

The now large number of solved αβ TCR structures can be compared to the three γδ heterodimers (discussed in more detail in the following), and while these also resemble the Fab fragment of an antibody, there are several features that are unique to the αβ molecules, which may be significant. These include the following:

There are now three γδ heterodimer structures: a γδ TCR from a human T-cell clone G115,¹⁶⁷ which can be activated by natural or synthetic pyrophosphomonoesters, a γδ TCR from the murine T cell clone G8 together with its ligand, the nonclassical MHC class I molecule T22,¹⁶⁸ and a human MHC class I chain-related-reactive γδ TCR (δ1A/B-3),^169a which was determined as a single-chain Fv construct. The G8 structure is shown in Figure 11.6, alongside the 2C αβ TCR. The structure of a single human Vδ domain also has been determined.¹⁶⁹ The Vδ domain of the G115 structure is similar to the isolated Vδ domain and the quaternary structure of G8 is similar to that of G115.¹⁶⁸

The most distinctive feature of both the G115 and the G8 TCR, when compared with αβ TCRs and Igs, is that the C domains “swing out” from under the V domains. This unusual shape is highlighted by both a small elbow angle of 110 degrees, defined as the angle between the pseudo twofold symmetry axes that relate V to V and C to C, and a small V-C interdomain angle. This contrasts with an average of 149 degrees for αβ TCR structures. The small angle between the Vγ and Cγ domains shifts both Cδ and Cγ to one side. Moreover, the molecular surfaces of the constant domains are different than those of αβ TCRs with no clear similarities either in the shape or the nature of the CαCβ and CγCδ surfaces; there are only a few solvent-exposed residues that are conserved in both Cβ and Cγ domains as well. Thus, it is unclear where or how the extracellular domains of the CD3 subunits interact with the extracellular portions of γδ TCRs compared with αβ TCRs. This may explain why the CD3 components of αβ TCRs are so different from those of γδ TCRs.

In terms of ligand binding surfaces, we note that the Vδ CDR3 of G8 protrudes significantly away from the other CDRs, as shown in Figure 11.6. This has significance in that this is the major region of contact with the T22 ligand (see later section). In the case of G115, both Vδ and Vγ CDR3 loops protrude from the rest of the putative binding surface and create a cleft between them. Portions of the CDR1γ and δ and CDR2γ combine with the clefts between the CDR3 loops to form a pocket, which is surrounded by positively charged amino acid residues contributed by CDR2γ and δ, and CDR3γ. The jagged surface of this TCR resembles the surface of an antibody that binds a small-molecule antigen. Although this would be consistent with the supposition that this TCR binds the negatively charged phosphate compounds,¹⁷⁰ direct binding between the TCR and phosphoantigen including crystal-soaking and cocrystallization experiments have not been successful. Instead, a soluble G115 was found to bind a soluble form of adenotriphosphate (ATP) synthase F1 and apolipoprotein A-1.¹⁷¹

While the δ1A/B-3 TCR maintains an overall fold similar to the other γδ TCR structures, it was noted that unlike the G115 and G8 CDR3 regions, which are protruding out, the δ1A/B-3 CDR loops together generate a nearly flat surface on the combining site. This difference is anticipated, as the CDR3 length distribution of the TCRδ chains is quite variable as discussed previously, and like antibodies should have a broad range of binding site shapes.

Although it has long been established that this type of T cell generally recognizes a peptide bound to an MHC molecule, a formal biochemical demonstration that this was due to TCR binding to a peptide/MHC complex took many years to establish. Part of the difficulty in obtaining measurements of this type has been the intrinsically membrane-bound nature of MHC and TCR molecules. Another major problem is that the affinities are relatively low, in the micromolar range, which is too unstable to measure by conventional means.

To some extent, the problem of measuring the interactions of membrane-bound molecules can be circumvented by expressing soluble forms of TCR and MHC, which is also essential for structural studies (see previous discussion). For TCRs, many successful strategies have been described, including replacing the transmembrane regions with signal sequences for glycolipid linkage,¹⁷² expressing chains without transmembrane regions in either insect or mammalian cells,¹⁷³ or a combination of cysteine mutagenesis and E. coli expression.¹⁶⁴ Unfortunately, no one method seems to work for all TCR heterodimers, although the combination of insect cell expression and leucine zippers at the c-terminus to stabilize heterodimer expression has been successful in many cases.¹⁷⁴ The production of soluble forms of MHC molecule has a much longer history, starting with the enzymatic cleavage of detergent solubilized native molecules¹⁷⁵ as well as some of the same methods employed for TCR such as glycophosphatidylinositol (GPI) linkage,¹⁷⁶ E. coli expression and refolding,^177,178 and insect cell expression of truncated
(or leucine zippered) molecules.¹⁷⁹ One interesting variant that seems necessary for the stable expression of some class II MHC molecules in insect cells has been the addition of a covalent peptide to the N-terminus of the β chain.¹⁸⁰

The first measurements of TCR affinities binding to peptide/MHC complexes were performed by Matsui et al.¹⁸¹ and Weber et al.¹⁸² Matsui and colleagues used a high concentration of soluble peptide/MHC complexes to block the binding of a labeled anti-TCR Fab fragment to T cells specific for those complexes, obtaining an equilibrium binding affinity (K_d) value of approximately 50µM for several different T cells and two different cytochrome peptide/I-E^k complexes (as shown in Table 11.3). Weber and colleagues used a soluble TCR to inhibit the recognition of a flu peptide/I-E^d complex by a T cell and obtained a K_D value of approximately 10 µM. While these measurements were an important start in TCR biochemistry, they gave no direct information about the kinetics of TCR-ligand interactions. Fortunately, the development of surface plasmon resonance instruments, particularly the BIAcore^TM (Pharmacia Biosensor, Uppsala, Sweden) with its remarkable sensitivity to weak macromolecular interactions,¹⁸³ has allowed rapid progress in this area. In this technique, one component is covalently crosslinked to a surface and then buffer containing the ligand is passed in solution over it. The binding of even approximately 5% of the surface-bound material is sufficient to cause a detectable change in the resonance state of gold electrons on the surface. This method allows the direct measurement of association and dissociation rates, that is, kinetic parameters, and also has the advantage of requiring neither cells nor radioactive labels. Recently, microcalorimetry has also been used to measure some TCR ligand affinities; and these analyses have confirmed the surface plasmon resonance (SPR) values,¹⁸⁴ but do not allow kinetic measurements. These and other data^138,160 showed definitively that TCR and peptideloaded MHC molecules alone are able to interact and also that expression in a soluble form has not altered their ability to bind to each other. As shown in Table 11.3, SPR measurements show that while the on-rates of TCRs binding to peptide/MHC molecules vary from very slow (1,000 M sec) to moderately fast (200,000 M sec), their off-rates fall in a relatively narrow range (0.5 to 0.01 sec^-1) or a t_1/2 of 12 to 30 seconds at 25°C. This is in the general range of other membrane bound receptors that recognize membrane molecules on other cells,¹³⁷ but it has been noted that most TCRs have very slow on-rates,¹³⁸ which reflects a flexibility in the binding site that might help to foster cross-reactivity (see the following). In the case of a class I MHC-restricted TCR, 2C, this relatively fast off-rate may be stabilized (10-fold) if soluble CD8 is introduced,¹⁶⁵ but this result is controversial.¹⁸⁵ CD8 stabilization of TCR binding has been seen by Luescher et al. in cell-based TCR labeling assay¹⁸⁶; however, no enhancement of TCR binding has been seen using soluble CD4¹⁸⁷ (see the following for more discussion of CD4 and CD8).