Structure of the Chicken CD3ϵδ/γ Heterodimer and Its Assembly with the αβT Cell Receptor*

Background: Chickens possess a CD3δ/γ chain that assembles with T cell receptor to mediate immune signaling. Results: Chicken CD3ϵδ/γ has an atypical heterodimer interface and surface but associates with TCRαβ. Conclusion: Chicken CD3δ/γ represents a hybrid chain possessing features in common with human CD3δ and CD3γ. Significance: Understanding the ancestral TCR signaling complex provides insights into the evolution of this signaling apparatus. In mammals, the αβT cell receptor (TCR) signaling complex is composed of a TCRαβ heterodimer that is noncovalently coupled to three dimeric signaling molecules, CD3ϵδ, CD3ϵγ, and CD3ζζ. The nature of the TCR signaling complex and subunit arrangement in different species remains unclear however. Here we present a structural and biochemical analysis of the more primitive ancestral form of the TCR signaling complex found in chickens. In contrast to mammals, chickens do not express separate CD3δ and CD3γ chains but instead encode a single hybrid chain, termed CD3δ/γ, that is capable of pairing with CD3ϵ. The NMR structure of the chicken CD3ϵδ/γ heterodimer revealed a unique dimer interface that results in a heterodimer with considerable deviation from the distinct side-by-side architecture found in human and murine CD3ϵδ and CD3ϵγ. The chicken CD3ϵδ/γ heterodimer also contains a unique molecular surface, with the vast majority of surface-exposed, nonconserved residues being clustered to a single face of the heterodimer. Using an in vitro biochemical assay, we demonstrate that CD3ϵδ/γ can assemble with both chicken TCRα and TCRβ via conserved polar transmembrane sites. Moreover, analogous to the human TCR signaling complex, the presence of two copies of CD3ϵδ/γ is required for ζζ assembly. These data provide insight into the evolution of this critical receptor signaling apparatus.

The ␣␤T cell receptor (TCR) 7 -CD3 complex represents an extensively studied transmembrane (TM) receptor system. In humans and other mammals, the TCR signaling complex comprises eight type-I membrane-spanning polypeptides that include the TCR ␣␤ heterodimer, the CD3⑀␥ and CD3⑀␦ heterodimers, and the CD3 homodimer (1), although higher order molecular assemblies are considered to exist (2,3). The ␣␤TCR functions to recognize peptide-, lipid-, and vitamin B precursor-based antigens that are presented by major histocompatibility complex (MHC) or MHC-like molecules (4 -7). This ligand-sensing heterodimer has no intrinsic ability to transmit signals to the cell interior due to the short cytoplasmic tails of the ␣ and ␤ chains of the TCR. Instead, signaling is mediated by the CD3⑀␥, CD3⑀␦, and CD3 subunits that associate with the TCR to produce a complex with a 1TCR␣␤-1CD3⑀␦-1CD3⑀␥-1 stoichiometry (1). The signaling subunits contain immunoreceptor tyrosine-based activation motifs (ITAMs) within their intracellular regions that become targets of the Src family kinase Lck upon receptor triggering; each CD3 heterodimer contributes two ITAMS to the signaling capacity of the TCR, whereas the homodimer contributes six (8 -12). The precise mechanism by which signals are transmitted from the extracellular portion of TCR to the intracellular regions of CD3 is a major unresolved issue in T cell biology, with kinetic segregation, conformational change, mechanotransduction, and other models being proposed (13)(14)(15)(16)(17)(18). The uncertainty surrounding TCR triggering mechanisms is in large part due to the lack of structural information on the intact TCR signaling complex.
The structures of mammalian CD3 heterodimer ectodomain fragments have been determined by x-ray crystallography and solution NMR (19 -22). Analysis of their surface features, including the sites of activating monoclonal antibody binding and carbohydrate modifications, have suggested particular regions that are likely to be closely juxtaposed with the TCR ectodomains and may therefore play a role in transmission of ligand sensing from TCR to CD3 modules (17,(23)(24)(25). However, a detailed structural analysis of the extracellular interactions between TCR and CD3 subunits is complicated by the lack of any apparent affinity among these dimeric components in solution. A major energetic contribution to assembly of TCR and CD3 subunits appears to come from a network of polar interactions within the TM domains (26,27). Both the TCR␣ and the TCR␤ TM domains have a central lysine residue that recruits one CD3 heterodimer via a pair of acidic TM residues (28). Within the mature complex, CD3⑀␦ associates exclusively via the TCR␣ TM domain and CD3⑀␥ via the TCR␤ TM domain to create an "asymmetric" hexamer. TCR␣ also contains an arginine residue in the upper half of the TM domain that mediates its interaction with the homodimer (28). The module will only join a preassembled CD3⑀␦-TCR␣␤-CD3⑀␥ hexamer, signaling the completion of assembly by masking an ER retention signal in CD3⑀␦ (29) and allowing export to the cell surface.
The evolutionary basis for the inclusion of two globally similar CD3 modules in this receptor complex is unclear. There is evidence to suggest that the mammalian CD3␦ and CD3␥ subunits have evolved novel, nonredundant functions. Both CD3␦ and CD3␥ knock-out mice exhibit profound defects in the production of mature ␣␤T cells, confirming that both CD3⑀␦ and CD3⑀␥ modules are absolutely required (30,31). Furthermore, substitution of the extracellular immunoglobulin (Ig) domain of CD3␥ with that of CD3␦ failed to rescue surface TCR expression in CD3␥-deficient human Jurkat T cells (32), although exchange of TM and cytoplasmic domains was tolerated; thus the folded ectodomains, at least, are non-interchangeable. Nonetheless, the precise nature of their unique functional roles and the significance of their strictly asymmetric assembly with TCR remain unknown.
Chickens (ch) and the amphibian Xenopus laevis also express a CD3⑀-like protein but do not express separate CD3␦ and CD3␥ chains. Instead, they encode a protein that shares equal homology with both mammalian CD3␦ and CD3␥ and has thus been designated CD3␦/␥ (33). At the amino acid sequence level, chicken and human CD3⑀, -␦, and -␥ chains have low extracellular (32-34%) and high TM (44 -52%) and intracellular (49 -55%) sequence identity. Analysis of the CD3 locus suggests that mammalian CD3␦ and CD3␥ arose from a gene duplication event that occurred 230 million years ago (34). Accordingly, it is likely that the ch-CD3 represents a primordial form that has not diversified in a manner analogous to its mammalian counterpart. To provide further insight into the relationship between the mammalian TCR signaling complex and its evolutionary precursors, we have undertaken a structural and biochemical analysis of the chicken CD3 proteins and their assembly into the chicken TCR signaling complex. The solution NMR structure of the ch-CD3⑀␦/␥ ectodomain dimer reveals significant differences from the mouse and human CD3 heterodimers in both domain orientation and surface chemistry. Furthermore, the ch-TCR signaling complex assembly demonstrates that despite the lack of CD3 asymmetry in the chicken receptor system, two CD3⑀␦/␥ dimers are required to form a fully assembled complex that is capped by association.

EXPERIMENTAL PROCEDURES
Cloning, Expression, Refolding, and Purification of Chicken CD3-Gene fragments encoding the extracellular domains of mature ch-CD3⑀ (residues 24 -91) and CD3␦/␥ (residues 18 -97) excluding the cysteine-rich stalks were synthesized de novo (GenScript). To generate a single chain construct, the C terminus of ch-CD3⑀ was covalently linked to the N terminus of ch-CD3␦/␥ via a 26-amino acid flexible peptide using splice-byoverlap PCR. ch-CD3⑀␦/␥ was cloned into a pET28b expression vector downstream of the thrombin-cleavable histidine tag and expressed as inclusion bodies in Escherichia coli BL21(DE3) cells. Inclusion bodies were solubilized in 0.2 M Tris-HCl (pH 9.5), 6 M guanidine HCl, 0.1 M DTT, 10 mM EDTA and refolded essentially as described (35). Refolded protein was buffer-exchanged into 10 mM Tris (pH 8) containing 0.5 M NaCl using tangential flow filtration prior to loading on a HisTrap HP nickel column (GE Healthcare) and eluted with 0.5 M imidazole. Histidine tag cleavage was performed using agarose-linked thrombin beads (Sigma) according to the manufacturer's instructions. The final purification step involved gel filtration chromatography using a Superdex75 16/60 column (GE Healthcare) pre-equilibrated in 25 mM HEPES (pH 7.6) containing 50 mM NaCl and 0.5 mM EDTA.
NMR-Suitable NMR buffer conditions were identified as 0.5 mM CD3, 50 mM HEPES, pH 7.6, 125 mM arginine, 125 mM glutamate, 0.01% azide, 0.01% Roche Applied Science protease inhibitor, 0.5 mM EDTA using crystallography dialysis buttons. All NMR samples contained 10% 2 H 2 O, and the spectra were recorded at 293 K. The following NMR spectra were recorded on a Bruker AVANCE TM 600-MHz spectrometer with cryoprobe using a 13 C, 15 13 C, 15 N-CD3 sample was used to acquire transverse relaxation optimized spectroscopy versions of an HNCACB, HN(CO)CACB, HN(CA)CO, and HN(CO)CA on an 800-MHz Bruker AVANCE fitted with a cryoprobe, and 13 C NOESY-HSQC (aliphatic) and 13 C NOESY-HSQC (aromatic) ( m 110 ms) spectra were acquired on the same spectrometer using the 13 C, 15 N-labeled sample. Spectra were processed using Topspin version 3.0. Backbone amide, and CA, CB, HA, and HB resonances were assigned manually using XEasy (36). Automated side-chain assignments were made using the ASCAN algorithms of UNIO and verified and supplemented by manual assignments using the HCCH-TOCSY spectra.
Structures were calculated using the AtnosCandid automated NOE peak picking and assignment algorithms with CNS torsion angle dynamics starting from an extended chain. The resulting structures were refined in CNS using simulated annealing with Cartesian dynamics. During refinement, dihedral angle restraints predicted from TALOS were incorporated along with hydrogen bond restraints in regions of canonical secondary structure where unique donor-acceptor pairs could be identified by convergence. The 10 lowest energy conformers with no NOE violations Ͼ0.3 Å, no bond violations Ͼ0.05 Å, and no improper or dihedral angle violations Ͼ5°were chosen to represent the solution structure of CD3. The ch-CD3 structures have been deposited with the Protein Data Bank (PDB code: 2MIM), and the chemical shifts have been deposited with the BMRB (code 19687).
Epitope Tags and Affinity Reagents-Epitope tags were used to facilitate immunoprecipitation of chicken TCR components. In all cases, epitope tags were installed at the C terminus of protein with a Gly-Ser linker, which was encoded by the BamHI site used for cloning. Anti-HA-agarose beads (Sigma, A2095) were used to immunoprecipitate chains carrying the hemagglutinin (HA) epitope tag (GYPYDVPDYA). Streptavidin-conjugated beads (Sigma, S1638) were used to precipitate chains carrying the streptavidin-binding peptide (SBP) tag (DEKTT-GWRGGHVVEGLAGELEQLRARLEHHPQGQREPSSSGG-SKLG). Anti-protein C (PC)-agarose beads (Roche Applied Science, 11815024001) were used to immunoprecipitate chains carrying the PC tag (EDQVDPRLIDGK).
cDNA Constructs and in Vitro Transcription-cDNA encoding CD3⑀, CD3␦/␥, and were synthesized as G-blocks (Integrated DNA Technologies) and cloned into pSP64 (modified by M. Kozak). The chicken cell line MSB1 (37) was the source of cDNA to amplify by standard PCR conditions the coding sequence for TCR␣ chain (primers: forward CMGTGSGAS-RAAATGRATTTTS and reverse GCACCCAATGCTCCAG-TAAT), which was cloned into the pJET3.1 vector (CloneJet, Fermentas), and for TCR␤ chain (primers: forward TTCCST-GCTGGTTTCTTACG and reverse TCCTCTTCAAGGCAA-AGCAT), which was cloned into the filled BamHI site of pGE-M3zf (Promega). In all cases the natural signal sequence was replaced with that of mouse H2-K b (MVPCTLLLLLAAALAP-TQTRA), ensuring robust and equal targeting of all polypeptide chains to ER microsomes (38). Sequences were cloned into pSP64 constructs modified to append an HA, PC, SBP, or no tag (Gly-Ser) to the C terminus as appropriate and a poly(A) tail to the mRNA. mRNA was transcribed in vitro using the RiboMAX large scale RNA T7 production system (Promega, P1300) according to the manufacturer's instructions. m7G Cap analog (Promega, P1712) was supplied in 5-fold excess with respect to ribo-GTP to ensure that mRNA was correctly capped. After mRNA transcription, plasmid DNA was cleaved with RQ1 DNase (Promega, M6101), and mRNA was purified with an RNeasy mini kit (Qiagen, 74104). 25-l transcription reactions typically yielded 60 l of mRNA at a concentration of 250 g/ml.
In Vitro Transcription and Assembly-Each mRNA was translated alone to ensure that targeting to ER microsomes, signal peptide cleavage, and glycosylation occurred as expected (data not shown). mRNA concentration was adjusted to ensure equivalent translation of each polypeptide. Adjustment factors were determined by densitometry (see below) of deglycosylated (endoglycosidase H-treated) samples. 25-l in vitro translation reactions contained 17.5 l of rabbit reticulocyte lysate (Promega, L4960), 0.5 l of amino acids without Cys or Met (Promega, L5511), 0.5 l of SUPERase⅐In (Promega, N2511), 2.0 l of EXPRE 35 (28)). Total mRNA concentrations were generally kept below 250 ng/reaction but varied depending on efficiency of translation as determined by test translation experiments. mRNA was denatured at 65°C for 3 min prior to addition to lysate mixtures. mRNA was translated for 15-30 min at 30°C, and then oxidized glutathione was added to a final concentration of 4 mM. Assembly proceeded in oxidizing conditions for 2 h (three chain experiments) or 4 h (five chain experiments). Reactions were stopped by the addition of 900 l of ice-cold Tris-buffered saline containing 10 mM iodoacetamide. ER microsomes were pelleted at 20,000 ϫ g for 10 min. Pellets were washed with an additional 500 l to Tris-buffered saline containing 10 mM iodoacetamide. After discarding the wash, ER microsome pellets were extracted in 0.5% digitonin, 200 g/ml BSA, 10 mM iodoacetamide for 30 min at 4°C. For experiments involving anti-PC pulldown, 2 mM CaCl 2 was included in the extraction buffer. Mixing controls were performed by translating and oxidizing certain mRNAs (detailed in the figure legends) in separate, halfvolume in vitro translation reactions and mixing after reactions were stopped with ice-cold buffer.
Immunoprecipitation and Analysis-Digitonin extracts were passed through Spin-X filters (Corning, 8160) to remove insoluble material, and 10 -15 l of the appropriate affinity bead slurry was added to immunoprecipitate polypeptides of interest. Immunoprecipitations were incubated between 2 and 24 h before being washed. For nondenaturing elution, SBP pulldowns were eluted with 100 M biotin in 0.5% digitonin containing 200 g/ml BSA in Tris-buffered saline before being transferred to anti-HA-agarose for sequential immunoprecipitation. When nondenaturing elution was not required, samples were eluted by heating at 95°C for 3 min in 10 l of 0.5% SDS in 50 mM citrate, pH 5.6. Eluted complexes were incubated at 37°C after the addition of 0.5 l of endoglycosidase H (New England Biolabs, P0702) to remove sugar moieties. Deglycosylated chains were separated by 12% NuPAGE (Invitrogen, NP0341 or NP0343) in MES buffer (Invitrogen, NP0002) and blotted onto PVDF (Millipore, ISQ00010). To measure incorporated [ 35 S]methionine and [ 35 S]cysteine, PVDF membranes were placed on storage phosphor screens (BAS IP TR2040) and scanned by a Typhoon (GE Healthcare, FLA-9410). The gel files were converted to tiff format, and densitometry was performed using Image Gauge V4 software (Fujifilm). Data were plotted using Prism V5 (GraphPad Software Inc.).

RESULTS
For NMR structural studies we expressed, refolded, and purified a single-chain construct comprising the extracellular Ig domains of ch-CD3⑀ and ch-CD3␦/␥ joined via a 26-amino acid flexible linker. Of the 178 residues present, backbone amide assignments could be made for all CD3⑀ residues with the exception of Gly 13 , and all CD3␦/␥ residues except Ser 27 -Asn 33 and His 75 -His 77 (excluding the linker sequence). Approximately 85% of other nonlabile side-chain resonances were assigned ( Chicken CD3 Domain Structure-The ch-CD3⑀ monomer forms an Ig domain comprising seven ␤-strands making up two antiparallel ␤-sheets (sheet 1, ABED, and sheet 2, CFG) that pack via a hydrophobic core and are bridged by a disulfide bond originating from two conserved cysteine residues (Cys 20 and Cys 55 ) (Fig. 1, A and B). Accordingly, ch-CD3⑀ belongs to the C1-set Ig fold but is atypical in that, like the C H1 antibody domains, it lacks the short CЈ strand in sheet 2 (39). This is in contrast to both mouse and human CD3⑀, which contain seven and eight ␤-strands and fall into the C2 and I-set Ig folds, respectively (20,21).
The ch-CD3⑀ Ig domain has low sequence identity with human (22%) and mouse (24%) CD3⑀, but overall is structurally very similar with an r.m.s.d. of 1.65-1.75Å over the entire domain (61 C␣ atoms). The primary deviations from the structure of human CD3⑀ are related to the configuration of the Ig domain and the length and conformation of loops connecting certain ␤-strands (Fig. 1C I ). Two loop regions in ch-CD3⑀ are significantly different from the human CD3⑀ structure: the C-D loop, which in ch-CD3⑀ adopts a compact 3 10 helical turn, and the FG loop present at the membrane distal face, which forms a tight hairpin loop. Due to these differences, ch-CD3⑀ has a considerably diminished accessible surface area (4430 Å 2 ) relative to the human CD3⑀ (5640 Å 2 ).
ch-CD3␦/␥ comprises two antiparallel ␤-sheets (sheet 1, ABE, and sheet 2, CCЈFG) linked by a single disulfide bond between Cys 24 and Cys 64 and supported by a hydrophobic core comprising residues Leu 22 , Trp 36 , Leu 51 , Pro 58 , Tyr 62 , and Leu 74 (Fig. 1, A and B). Accordingly, ch-CD3␦/␥ adopts a C2-set Ig fold. The single N-linked glycosylation site at Asn 71 is present within the FG loop (Fig. 1B) in a location that would likely protrude away from the cell membrane.
The ch-CD3␦/␥ fold is similar to that observed for human CD3␥ (20) but differs from the C1-set Ig fold found in human CD3␦ due to a translocation of the D strand from the ABED face to the CFG face, where it becomes a CЈ strand (Fig. 1, C II/III and D). Despite this difference, the remainder of the ch-CD3␦/␥ domain is reasonably structurally conserved when compared with human CD3␦, with 47 C␣ atoms having an r.m.s.d of 2.15 Å. However, ch-CD3␦/␥ aligns relatively poorly to human CD3␥ (r.m.s.d 2.5 Å over 50 C␣ atoms) primarily due to a number of structural alterations distributed throughout the molecule. Apart from the DE loop (CЈ-E loop in human CD3␥), the major structural difference lies in the FG loop, which adopts a flattened conformation in ch-CD3␦/␥. In both human CD3␥ and sheep CD3␦ (this loop is not visible in the human CD3␦ structure), the F-G loop extends away from the core of the CD3 module (20,22).
The Chicken CD3 Heterodimer-The ch-CD3 subunits associate in a side-by-side manner, resulting in a buried surface area of 1,470 Å 2 . The heterodimer interface is dominated by interactions between the parallel G-strands, each of which contains a continuous string of residues (Asp 59 -Tyr 65 in CD3⑀ and Asn 71 -His 75 in CD3␦/␥) that come together to form an interlocking ladder (Fig. 2A). The CD3⑀␦/␥ interface is further supported by contacts between the F-strand of CD3⑀ and a triplet of residues (Met 11 -Val 13 ) at the base of the CD3␦/␥ A-strand that make extensive van der Waals interactions with Thr 52 and Tyr 65 of CD3⑀.
Despite playing a central role at the heart of the ⑀-␦/␥ interface, the conformation of the ch-CD3⑀ G-strand is considerably different from that found in human and murine CD3s (Fig. 2B). Due to a tight turn in the FG loop, the ch-CD3⑀ G-strand, which is extended at its base, packs tightly against the C-terminal portion of the F-strand, presenting an unusual surface to ch-CD3␦/␥ (Fig. 2B). Accordingly, the nature of the CD3⑀-CD3␦/␥ interactions are distinct from those observed in human CD3s. For example, of the 8 CD3⑀ residues whose side chains contribute to the dimer interface, only Gln 4 and Tyr 65 are present in human CD3⑀; and of the 10 interacting residues in CD3␦/␥, 3 are conserved with human CD3⑀␥ and 4 are conserved with CD3⑀␦, most of which (Leu 74 , His 77 , Tyr 78 , and Arg 79 ) lie at the base of the G-strand.
The unusual conformation adopted by the base of the CD3⑀ G-strand has a considerable impact on the overall domain organization, giving the ch-CD3 molecule an unusual appearance relative to all mammalian CD3 structures studied to date. Although human and mouse CD3 heterodimers have a distinctive upright arrangement, the ch-CD3 heterodimer has a more skewed/asymmetric configuration, as evidenced by the center of mass of the ␦/␥ subunit that is rotated by up to 45°when viewed from side-on and 35°viewed from the top relative to human CD3⑀␦ (Fig. 2C). As a consequence, the overall structural similarity between ch-CD3⑀␦/␥ and human CD3⑀␦ is low, with an r.m.s.d of 3.6 Å over 106 C␣ atoms.  3A). In ch-CD3⑀ the bulky CЈ-D loop is notably absent, and in its place is a more compact and neutral 3 10 helix derived from Pro 33 -Leu 35 (Fig. 1, C I and D). This, coupled with a change in the nature of the D-E loop, which contains a single basic residue (Lys 40 ), contributes to the relatively flat, featureless appearance of ch-CD3⑀ (Fig. 3A). Although many of the amino acid residues conserved between chicken and human CD3 lie within the core and maintain the Ig fold, several conserved residues are surface-exposed. When mapped onto the structure of ch-CD3, these residues appear to cluster to a single face (the ⑀-ABE, ␦/␥-CЈCFG) and are particularly prominent in the membrane-proximal region adjacent to where the cysteine-rich stalks (not present in our construct) would protrude (Fig. 3B). In contrast, the opposing face (⑀-CFG, ␦/␥-ABE) is almost devoid of any conserved, surface-exposed residues.
Incorporation of Chicken CD3⑀␦/␥ into a Complete Chicken TCR Signaling Complex-To relate our structural insights on the isolated CD3 heterodimer to the broader picture of subunit organization within the complex, we performed a series of biochemical experiments interrogating the basis of chicken TCR complex assembly. Analysis of the TM domains of chicken TCR␣ and TCR␤ chains reveals that basic residues are present at all the same positions as those in mammalian sequences (Fig.  4), and each CD3 chain contains the conserved acidic and hydroxyl-bearing TM residues that are crucial for human TCR signaling complex assembly. Interestingly, the first basic residue in the TCR␣ chain is a lysine rather than the typical arginine found in other vertebrates (including fish). We mutated each TCR TM lysine to alanine and assessed each mutant's ability to assemble with ch-CD3⑀␦/␥ heterodimers in isolated The structures shown span from Glu-6⑀ to Val 70 ⑀ and from Leu 9 ␦/␥ to Met 79 ␦/␥ and exclude the 26-amino acid linker. C, secondary structure-matched superimposition of human and chicken CD3⑀ chains (I). In II and III, the structure of ch-CD3␦/␥ was aligned to that of human CD3␦ and CD3␥, respectively. Human CD3 structures are colored gray, and ch-CD3 structures are in pink (CD3⑀) or cyan (CD3␦/␥). Only residues that were assigned in the ch-CD3 structure and the corresponding residues in human CD3 are shown. Dashed lines highlight regions of significant structural deviation. D, amino acid sequence alignment of the immunoglobulin domains of human, mouse, and chicken CD3⑀, ␦, and ␥ chains. The secondary structure for human and chicken variants is displayed above and below the sequences, respectively. Potential N-linked glycosylation sites are highlighted in bold. * indicates absolutely conserved residues.
ER microsomes using an established in vitro translation-based assay (28). To facilitate immunoprecipitation (IP) and SDS-PAGE analyses, HA and SBP tags were installed at the C termini of TCR and CD3⑀ polypeptides, respectively. Incorporation of 35 S isotope-labeled methionine and cysteine during translation provided a quantitative method to measure the relative levels of each subunit recovered from digitonin extracts of the ER microsomes. Mutation of each central lysine residue in TCR␣ and TCR␤ resulted in loss of CD3⑀␦/␥ association with the individual TCR chains, whereas the lysine in the upper half of the TCR␣ TM had no effect (Fig. 5). Interestingly, the ability of TCR␤ to support CD3⑀␦/␥ assembly was ϳ50% when compared with TCR␣. This property is similar when human TCR␤ is compared with TCR␣ for its ability to associate with CD3 heterodimers; TCR␣-CD3⑀␦ assembly is severalfold more efficient that TCR␤-CD3⑀␥ assembly.
We next tested the ability of ch-CD3⑀␦/␥ to associate with a chicken TCR␣␤ heterodimer and homodimer in in vitro assembly reactions. To facilitate isolation of TCR␣␤ heterodimers for analysis, we installed an SBP tag on the TCR␣ chain and an HA tag on the TCR␤ chain. This allowed the products of a first-step capture with streptavidin-coupled agarose beads to be gently eluted with biotin and reprobed with anti-HA-agarose beads to specifically isolate folded TCR␣␤ heterodimers. Densitometry measurements were used to determine the ratios of different CD3 subunits that co-precipitated with TCR. For these experiments, a PC epitope tag was installed at the C terminus of the CD3⑀ chain to improve electrophoretic separation from CD3␦/␥. Again, mutation of each central lysine residue of TCR␣ and TCR␤ resulted in reduced CD3 association, but only when both chains lacked the central lysine residue was CD3⑀␦/␥ recruitment to the complex completely abolished (Fig. 6). Surprisingly, mutation of a single lysine residue did not result in a 50% drop in CD3 association, implying that in the context of the whole receptor system, additional contacts may be made that partially compensate for loss of each lysine.
In contrast to previous results in the human system (28), a clear product at the expected molecular weight of disulfidelinked chicken homodimer was not evident in the full assembly reactions. Because chicken migrates very close to free TCR chains when separated by SDS-PAGE, we used a -chain IP strategy (via C-terminal PC tag) to determine what percentage of interacted with TCR␣␤. We co-translated quantities The black line connects the center of mass of the CD3 subunits. Center of mass calculations included only those residues for which NMR spectra were assigned (for ch-CD3) and the corresponding residues in the human CD3 variants. Subunits are colored as in Fig. 1. Although human CD3⑀ possesses a highly electronegative surface, ch-CD3⑀ is relatively neutral (white) in this region. B, surface-exposed conserved residues in human and chicken CD3. The ch-CD3 heterodimer is displayed as a surface representation in two orientations (left, ⑀-ABE, ␦/␥-CЈCFG; right, ⑀-CFG, ␦/␥-ABE) with the CD3 chains colored as in Fig. 1. Residues conserved in human CD3 are colored as follows: CD3⑀ (orange), CD3␦ and CD3␥ (orange), CD3␦ (red), and CD3␥ (yellow). Amino acid residues shown are as in Fig. 1. of mRNA that produced TCR␣, TCR␤, CD3⑀, CD3␦/␥, and proteins at a molar ratio of 5:5:10:10:1 and probed the digitonin extracts with anti-PC-coupled agarose beads. CD3⑀ was again tagged with SBP to facilitate separation from CD3␦/␥ by SDS-PAGE. Anti-PC IP yielded substantial amounts of homodimer and monomeric chain (both signal peptide-cleaved and uncleaved) in addition to TCR␣␤, CD3⑀, and CD3␦/␥ chains, but only when all lysines in the TCR TM domains were present (Fig.  7). Densitometric ratios in the all-wild-type assembly reaction (lane 1) indicated that only ϳ25% of was associated with TCR␣␤, explaining why a prominent homodimer was not seen when TCR␣␤ was the IP target. Despite the lack of effect on CD3 heterodimer recruitment, mutation of the lysine in the upper half of the TCR␣ TM reduced TCR␣␤ recovery in the IP to near background levels. Mutation of the central lysines of TCR␣ and TCR␤ also interrupted assembly. These results are consistent with the observation that in human TCR, is the final species to join the complex and prevention of CD3 association at either the TCR␣ or the TCR␤ TM domain is sufficient to prevent recruitment (28). Residues that are identical between human (hu) and chicken sequences are marked with asterisks. In the comparison of chicken CD3␦/␥ with human CD3␦ and CD3␥, residues identical to human CD3␦ are marked above the alignment, and those identical to human CD3␥ are marked below the alignment. The predicted membranespanning regions based on analysis of TMHMM 2.0 scores (55) are shaded with gray rectangles. The polar TM residues that are known to be essential for assembly of the human TCR and CD3 subunits and the corresponding positions in the chicken sequences are marked in blue (basic), red (acidic), and orange (hydroxyl-bearing).

FIGURE 5. Both TCR␣ and TCR␤ can bind CD3⑀␦/␥ and the interaction is dependent on the central TM lysines.
HA-tagged TCR chains were translated with CD3⑀ and CD3␦/␥ in in vitro translation and immunoprecipitated with anti-HA-agarose. Shown is a representative gel of the species bound to the beads in each immunoprecipitation and a gel showing 5% of the input before immunoprecipitation. The data presented on the graph represent three repeats of the same experiment. Two-tailed t tests indicated that TCR␣ could pull down more CD3⑀␦/␥ than TCR␤, but that the amount pulled down by TCR␤ was still more than the mixing control where TCR␤ and the CD3 mRNAs were translated in separate in vitro translation reactions and mixed prior to IP. ** indicates 0.001 Ͻ p Յ 0.01, * indicates 0.01 Ͻ p Յ 0.05, ns indicates p Ͼ 0.05. SBP-tagged TCR␣ and HA-tagged TCR␤ were translated with CD3⑀, CD3␦/␥, and , and a sequential immunoprecipitation for TCR␣ followed by TCR␤ was used to select TCR␣␤ heterodimers. In the mixing control, TCR␣ and TCR␤ were translated in a separate reaction from the CD3 and components and mixed after stopping the assembly reaction. A 5% sample before immunoprecipitation was taken to ensure that input was similar. Densitometry of TCR␣␤ heterodimers and CD3 chains was performed, and the results were calculated as a ratio of CD3 heterodimer/TCR heterodimer. The contribution of individual TCR␣ and TCR␤ chains to the CD3 signal was included in the ratio calculations. ** indicates 0.001 Ͻ p Ͻ 0.01, * indicates 0.01 Ͻ p Ͻ 0.05, and ns indicates p Ͼ 0.05. FIGURE 7. incorporation requires the upper TM lysine of TCR␣ and all CD3 chains. TCR␣ and TCR␤ were translated with CD3⑀-SBP, CD3␦/␥ and -PC with at 10-fold lower concentration than CD3 chains and 5-fold lower than TCR chains. was targeted for immunoprecipitation and the amount of co-precipitated TCR heterodimer was measured by densitometry. In mixing controls, TCR␣ and TCR␤ mRNA were translated separately from CD3 and chains and processed as described in the legend for Fig. 6. A 5% sample of the digitonin extraction was loaded on a gel as an input control. The ratio of TCR heterodimer to homodimer is graphed. * indicates 0.01 Ͻ p Ͻ 0.05.

DISCUSSION
All jawed vertebrates have an adaptive immune system that includes the genes to make antibodies, MHC and TCR (40). Furthermore, all mature TCR genes cloned to date (excluding pre-T␣ (41,42)) encode proteins with very short cytoplasmic domains that are unlikely to contain signaling motifs and are therefore thought to depend on CD3 or CD3-like signaling modules to transmit information to the intracellular signaling machinery (43). Analysis of CD3 loci across jawed vertebrate species suggests that the three genes present in mammals arose from two separate gene duplication events, the latter of which facilitated the divergence of CD3␥ and CD3␦ genes to produce two distinct CD3⑀␥ and CD3⑀␦ signaling modules. However, in non-mammalian species, the CD3 locus encodes only CD3⑀ and a second protein that has features similar to both CD3␦ and CD3␥ (34) and is likely to be their evolutionary precursor. Avian and mammalian CD3 proteins have diverged significantly enough that chicken and human TCR and CD3 are unable to form hybrid complexes (44), although the more highly conserved chicken chain can replace its mouse counterpart in -deficient T cells (45). In an effort to gain further insights into the evolutionary relationship among mammalian and non-mammalian CD3 components, we have determined the structure of the chicken CD3⑀␦/␥ heterodimer extracellular domains and examined the biochemical requirements for incorporation of the CD3⑀␦/␥ module into a complete chicken TCR signaling complex.
Herein we provide evidence that ch-CD3␦/␥ is a genuine hybrid chain. Based on amino acid sequence, ch-CD3␦/␥ is CD3␦-like in its extracellular domain and CD3␥-like in its intracellular region (34). However, our structure reveals that the extracellular region of ch-CD3␦/␥ has a fold more reminiscent of that found in human CD3␥ while retaining significant structural homology to CD3␦. The most striking feature of ch-CD3 is the unexpected domain juxtapositioning, which deviates significantly from the rigid, side-by-side, almost pseudosymmetric arrangement observed in murine and human CD3 dimers (20,21). Surprisingly, this difference appears to stem from the more conserved CD3⑀ chain rather than from the hybrid CD3␦/␥ chain.
Precisely how the unusual subunit arrangement in ch-CD3 impacts on the overall TCR signaling complex structure is unclear, but based on our structure, the ch-CD3 heterodimer is likely to present a unique molecular surface to the associated FIGURE 8. Schematic of the chicken TCR signaling complex. The gray disc represents the plasma membrane. Residues that are important in guiding TCR assembly are color-coded: light blue, lysine; red, aspartic acid. The extracellular domains are color-coded to identify each polypeptide: light blue, TCR␣; green, TCR␤; orange, CD3⑀; red, CD3␦/␥. ITAMs in the cytoplasmic tails of each species are represented as gray rectangles. Although this schematic implies that both CD3⑀ TM domains would fall closest to in the assembled complex, this information is not known, and this uncertainty is indicated with question marks.
TCR. For example, an acidic patch on the surface of human CD3⑀ that has previously been suggested to participate in interactions with TCR␤ (20) is notably absent in the chicken structure. It is noteworthy that the appearance of separate CD3␦ and CD3␥ chains in mammals correlates with the appearance of a surface-exposed loop (the FG loop) in the mammalian TCR␤ constant domain (46), a region that has long been implicated in CD3⑀␥ association (23,25,(47)(48)(49). Given that one face of ch-CD3 (⑀-CFG, ␦/␥-ABE) contains several unique surface-exposed residues and the opposing face (⑀-ABE, ␦/␥-CЈCFG) is decorated with residues conserved in human CD3, it is tempting to speculate that the latter may be involved in CD3⑀␦-TCR␣ interactions that have been conserved during evolution, whereas the former has diverged to accommodate the relatively newly evolved TCR␤ surface. This hypothesis is supported by the observations that many of the conserved residues in the putative CD3⑀␦-TCR␣ interface are located in the membraneproximal region, adjacent to the cysteine-rich stalks that have been previously implicated in association of the TCR and CD3 subunits (48,50,51).
What unique functional features may have accompanied the CD3 diversification in mammals remains an open question. One possibility is that development of separate CD3␦ and CD3␥ proteins was concomitant with incorporation of two CD3 dimers in the TCR signaling complex where the evolutionary precursor utilized only one. However, our analysis of chicken TCR and CD3 subunit assembly demonstrates that, as suggested by the conservation of basic residues in the TM domains of chicken TCR␣␤, the chicken TCR associates with two copies of the CD3⑀␦/␥ heterodimer in a manner that is analogous to the incorporation of both CD3⑀␦ and CD3⑀␥ modules in mammals (Fig. 8). Thus the octameric arrangement of one TCR with three dimeric signaling modules is an intrinsic feature of the TCR signaling complex even in organisms without separate CD3␦ and CD3␥ chains. This theme is echoed in the structure of the mouse ␥␦TCR complex, which does not incorporate the CD3⑀␦ module but instead contains two copies of CD3⑀␥ (30,52). Whether this symmetric CD3 arrangement confers any unique signaling properties on the mouse ␥␦TCR is unclear, but it does not appear to be a general feature common to mammals because primary human ␥␦ T cells contain both CD3⑀␥ and CD3⑀␦ (53). Determination of a chicken ␣␤TCR structure would allow comparison of putative CD3-interacting surfaces among all of these receptor types to reveal whether particular features have co-evolved with asymmetric versus symmetric CD3 assemblies.
A notable feature is that, like human TCR␤, the chicken TCR␤ is less efficient than its ␣ chain counterpart in associating with CD3 in three-chain assembly experiments (28). This property in human TCR is expected to ensure that the requisite asymmetric arrangement of CD3 heterodimers is achieved during biosynthesis. Observation of a similar phenomenon in the chicken TCR, where no such asymmetry exists, suggests that this feature is intrinsic to the TCR␣ and TCR␤ proteins and may reflect a fundamental requirement for cooperative assembly with the two CD3 modules. Accordingly, we consistently recovered more ch-TCR␣␤ heterodimer in samples where only one of the central lysine residues was mutated to alanine when compared with samples where both were mutated (Figs. 5 and 6), and poorest TCR␣␤ heterodimer formation was seen in control reactions where ␣ and ␤ chains are folded in the absence of CD3. TCR␣␤ formation and assembly with two CD3 modules are thus cooperatively linked. As in human TCR, joining of the homodimer appears to be the final step in chicken TCR assembly. It is unusual that the interaction between and TCR is mediated by a lysine instead of an arginine in the TM domain of TCR␣ because in the human TCR, mutation of this arginine to lysine results in a significant drop in affinity for (28). Indeed, the chicken TCR is comparatively poor at recruiting the homodimer based on our biochemical analysis. Whether this has functional consequences for chicken TCR signaling or is simply an artifact introduced by detergent extraction is unclear.
The unique structural features of ch-CD3 described here may reflect the distinctive demands of a symmetric CD3 assembly with the inherently asymmetric TCR; although human CD3⑀␦ and CD3⑀␥ each appear to have evolved unique surface features that optimize extracellular complementarity with TCR␣ and TCR␤, respectively, the ch-CD3⑀␦/␥ surface must complement both TCR constant domains simultaneously. Due to its reduced subunit complexity when compared with the mammalian receptor, the chicken TCR signaling complex may represent a more tractable experimental system for studying ectodomain interactions between the TCR and CD3 subunits. Further structural studies on the chicken proteins may therefore provide access to some of the more conserved features of the TCR signaling complex molecular organization that will impact on our models of receptor function.