|Home | About | Journals | Submit | Contact Us | Français|
The αβ TCR has recently been suggested to function as an anisotropic mechanosensor during immune surveillance, converting mechanical energy into a biochemical signal upon specific peptide/MHC ligation of the αβ clonotype. The heterodimeric CD3εγ and CD3εδ subunits, each composed of two Ig-like ectodomains, form unique side-to-side hydrophobic interfaces involving their paired G-strands, rigid connectors to their respective transmembrane segments. Those dimers are laterally disposed relative to the αβ heterodimer within the TCR complex. In this paper, using structure-guided mutational analysis, we investigate the functional consequences of a striking asymmetry in CD3γ and CD3δ G-strand geometries impacting ectodomain shape. The uniquely kinked conformation of the CD3γ G-strand is crucial for maximizing Ag-triggered TCR activation and surface TCR assembly/expression, offering a geometry to accommodate juxtaposition of CD3γ and TCR β ectodomains and foster quaternary change that cannot be replaced by the isologous CD3δ subunit’s extracellular region. TCRβ and CD3 subunit protein sequence analyses among Gnathostomata species show that the Cβ FG loop and CD3γ subunit coevolved, consistent with this notion. Furthermore, restoration of T cell activation and development in CD3γ−/− mouse T lineage cells by interspecies replacement can be rationalized from structural insights on the topology of chimeric mouse/human CD3εδ dimers. Most importantly, our findings imply that CD3γ and CD3δ evolved from a common precursor gene to optimize peptide/MHC-triggered αβ TCR activation.
T cell receptors mediate Ag-specific recognition of peptides bound to MHC (pMHC) molecules via their Fab-like heterodimeric clonotypes. The TCR complex consists of αβ, CD3εγ, CD3εδ, and CD3ζζ dimers in a 1:1:1:1 stoichiometry. Whereas the function of the αβ clonotype is to bind pMHC, the associated CD3 subunits mediate signaling through their cytoplasmic (Cyt) tails (1–6). Assembly of the TCR complex is dependent on the transmembrane segment of the individual subunits and their unique disposition of charged residues in the hydrophobic membrane environment, as detailed elsewhere (7). The ectodomain of αβ and the Ig-like CD3εγ and CD3εδ heterodimers make a limited number of contacts with one another such that, in solution, no interactions are definable by nuclear magnetic resonance (NMR) (8, 9). These results imply a loose association of individual dimeric extracellular TCR segments in the resting state.
How pMHC ligation of the αβ heterodimer on the T cell surface evokes intracellular signaling via the adjacent CD3 components has been the subject of intense investigation. Previous models suggest that TCRs transmit activation signals upon specific pMHC ligation via clustering, conformational changes, or combining both (1, 10–14). Recently we proposed that a common TCR quaternary change rather than conformational alterations of αβ clonotypes by pMHCs facilitates signal initiation (15). Given the vast array of TCRs and their pMHC ligands, this appears to represent a practical solution to the T cell immune recognition problem. Clues regarding this mechanism were revealed from structural detailing of TCR complex components and by mAb binding studies and analysis of their effects on T cell function. For example, solution structures of CD3εγ and CD3εδ heterodimers reveal a unique side-to-side hydrophobic interface with conjoined β-sheets involving the G-strands of the two Ig-like ectodomains of each pair (8, 9). The squat and rigid CD3 connecting segments contrast sharply with the long and flexible α and β connecting peptide linking respective constant domains to the transmembrane segments. These opposing features suggested that the αβ heterodimer may move relative to the CD3εγ and CD3εδ dimers upon pMHC ligation. By impeding such movement with H57, an Ab that bridges Cβ and CD3εγ (15, 16), pMHC-triggered activation is blocked, consistent with this view.
Agonist anti-CD3 mAbs like 2C11 and 500A2 footprint to the membrane distal CD3ε lobe that they approach diagonally, adjacent to the lever-like Cβ FG loop that facilitates pMHC-triggered activation (17, 18). This is the most sensitive TCR triggering direction (15). In contrast, a nonagonist mAb, 17A2, binds to the cleft between CD3ε and CD3γ in a perpendicular mode. 17A2 becomes stimulatory only subsequent to application of an external tangential (but not normal) force of ~50 pN introduced via optical tweezers. Specific pMHC, but not irrelevant pMHC, activates a T cell upon application of a similar force. During immune surveillance, when specific TCR-pMHC ligation occurs, an extracellular mechanical torque may be applied to the TCR prior to “stop movement” of the T cell. In this process, the pMHC on the APC functions as a gaff or hook to pull on the ligated TCR on the opposing T cell. Collectively, these findings suggest that the TCR is an anisotropic (i.e., directional) mechanosensor.
We reasoned that detailed analysis of the distinct topology of CD3 heterodimers would provide further insight into the basis of signal transduction involving the ectodomain components of the TCR mechanosensor. In addition, the rationale for differences in CD3 subunit requirements for various T lineage populations and developmental stages might emerge (19–28). In this study we have focused on differential G-strand geometry in CD3γ and CD3δ ectodomain fragments observed in human and other mammalian species. These features make the topology of CD3εγ and of CD3εδ distinct from one another and, in turn, mandate a defined quaternary organization for the TCR. The findings suggest that despite the loose confederacy of dimeric ectodomains constituting the TCR complex, dynamic complementarity of structural elements maximizes sensitivity in the relay of quaternary change essential for signal transduction.
The following fluorochrome-labeled mAbs were used for surface receptor analysis by flow cytometry: FITC–anti-CD3 (2C11), Alexa Fluor 647–anti-CD3 (17A2), Pacific Blue-CD4 (H129.19), Pacific Orange anti-CD8α (53-6.7), PE–anti-Vβ5 (MR9.4), and FITC-conjugated anti-TCR Cβ (H57) (BD Pharmingen, San Diego, CA). For flow cytometry, single-cell suspensions of thymocytes or lymph node (LN) cells were prepared at 5 × 106 cells/ml in PBS containing 2% FCS and 0.05% NaN3. Those cells were triple- or five-color stained with the mAbs at saturating concentrations according to standard procedures. A FACScan or FACSAria (BD Biosciences, San Jose, CA) was used for flow cytometric measurements. Data analysis was performed using FlowJo software (Tree Star, Ashland, OR) after dead cells were excluded from the analysis by forward and side scatter gating.
C57BL/6 mice were purchased from Taconic Farms (Germantown, NY). CD3δ−/− mice and CD3γ−/− mice have been described in detail elsewhere (22, 24). N15TCRtgCD3γ−/− mice were generated by crossing N15TCRtgRAG-2−/− (29) and CD3γ−/− mice. Construction of CD3δ mutant mice will be described in detail elsewhere (V.P. Dave and D.J. Kappes, unpublished observations). Mice were maintained and bred under specific pathogen-free conditions in the animal facility of the Dana-Farber Cancer Institute under a protocol reviewed and approved by the Animal Care and Use Committee. Mutant CD3δ cDNAs A–D (Fig. 2) were generated by PCR, confirmed by sequencing, and inserted into the human CD2 minigene cassette (30). Those transgenic (tg) lines were created by the Fox Chase Cancer Center Transgenic Facility (Philadelphia, PA) according to established protocols.
The cDNA encoding murine CD3γ was subcloned into the pUC18 for mutation and sequencing. To generate the CD3γ6M mutant in which the six amino acids in the G-strand of CD3γ (ETSNPL) were replaced with those of CD3δ (KVVSSV), we used the QuickChange Site-Directed Mutagenesis System (Stratagene, La Jolla, CA). CD3γmδecto or CD3γhδecto mutant was constructed by replacing the whole ectodomain denoted in Fig. 1 by recombinant PCR methods. CD3γ wild-type (wt) or CD3γ mutant constructs all contained a C-terminal FLAG epitope (DYKDDDDK) for quantitation of protein expression. All generated constructs were confirmed by DNA sequencing.
For retroviral transduction, we used the pLZRS-IRES-eGFP vector encoding the enhanced GFP downstream of an internal ribosome entry site (31). The virus supernatant was prepared as described previously (17, 32, 33). For retroviral transduction with CD3γwt and mutant constructs, total LN cells from N15TCRtgCD3γ−/− mice were stimulated with 4 μg/ml concanavalin A for 2 d, and T cells were purified by removing I-Ab–positive cells using anti–I-Ab and magnetic beads. N15TCRtgCD3γ−/− T cells were placed in a 24-well plate at 106 cells/well (volume of 300 μl) in complete RPMI 1640 medium, and 300 μl viral supernatant containing 20 μg/ml Lipofectamin (Life Technologies-BRL, Carlsbad, CA) was added to each well and the plate centrifuged at 2000 rpm for 1 h at room temperature. Transduced T cells were cultured for 3 d with 50 μ/ml human rIL-2 until used for the assay. To assess the TCR signaling of transduced N15TCRtgCD3γ−/− T cells, cells were washed and stimulated with VSV8 peptide-loaded irradiated splenocytes from C57BL/6 mice or with plate-coated 2C11 (5 μg/ml) for 4 h in media containing GolgiPlug reagent (BD Pharmingen). Then, cells were stained for surface CD8 and for intracellular IFN-γ, using Fix/Perm solution (BD Pharmingen).
For retroviral transduction with CD3γwt and mutant constructs, fetal thymi were removed from day 14.5 CD3γ−/− fetuses (observation of vaginal plug is day 0.5), and thymocytes at 100,000 cells/well (volume of 100 μl) in fetal thymic organ culture (FTOC) medium supplemented with 50 ng/ml IL-7 and stem cell factor were placed in a 96-well plate. Next, 100 μl viral supernatant containing 20 μg/ml Lipofectamin (Life Technologies-BRL) was added to each well and the plate centrifuged at 2000 rpm for 1 h at room temperature, then incubated at 37°C overnight. The next day, cells were collected and 30 μl/well placed in a Terasaki plate. One fetal thymic lobe was placed in each well. The plate was inverted and incubated for 2 d. Thymic lobes from C57BL/6 were treated with 1.35 mM 2′-deoxyguanosine (Sigma-Aldrich, St. Louis, MO) in transwell inserts (Costar, Cambridge, MA) for 5 d before use to remove hematopoietic cells, but not epithelial tissue capable of allowing the differentiation of T cell precursors. After 2 d of hanging drop culture, lobes were transferred to ATTP 0.8 mM filters (Millipore, Billerica, MA) on Gelfoam (Pharmacia & Upjohn Company, Kalamazoo, MI). After 7 d, thymocytes were counted and analyzed by FACS. To determine transduction efficiency, 50,000 of the transduced cells were used for the FACS analysis of GFP expression.
Solution structures of the murine CD3 (mCD3)εγ heterodimer and a chimeric CD3εδ heterodimer (mCD3ε and sheep CD3δ) reveal that both CD3ε domains adopt a virtually identical conformation with a root-mean-square-deviation <1.45 Å (8, 9). Each heterodimer is composed of two Ig-like ectodomains with a hydrophobic interface brought together by hydrogen bond-paired terminal G-strands forming conjoined β-sheets. Unlike CD3ε, the CD3γ and CD3δ ectodomains within these heterodimers are more divergent from one another. In particular, structural comparison between CD3γ and CD3δ ectodomains shows that there is differential G-strand geometry resulting in a pronounced cleft between the two CD3 ectodomains in CD3εγ that is partially occluded in CD3εδ (Fig. 1A). The difference in G-strand disposition likely results from two factors. First, fewer hydrogen bonds are formed between the two G-strands in CD3εγ relative to those in CD3εδ, at least in part owing to an amino acid sequence of CD3γ that does not support the optimal packing between CD3ε and CD3γ at the N terminus (or top) of the G–G interaction surface. Second, relative to CD3γ, the CD3δ BC loop is five residues shorter, containing only four amino acids (Fig. 1B). The side view of the heterodimers in Fig. 1A (right panel) highlights the difference in CD3γ versus CD3δ FG and BC loops. The longer BC loop and the presence of the C′-strand force the entire GFCC’ face of CD3γ, along with the FG loop, to bow away from the heterodimeric interface, whereas the shorter BC loop in CD3εδ must cut across the two β-sheets (ABED and FGCC′ faces), thereby preventing the FG loop from bending. The crystal structures of the human CD3 (hCD3)εγ and hCD3εδ heterodimers (Supplemental Fig. 1) identified a similar difference in G-strand geometry, as observed in the mouse orthologs (34, 35). Note that the top of the hCD3εδ FG loop (Gly52–Lys57) is missing in the existing crystal structure coordinates, however, most likely owing to the inherent flexibility in this region of the human ortholog.
Previous NMR binding experiments involving chemical shift and cross-saturation analyses (15) showed that 17A2 contacts both CD3ε and CD3γ domains, providing a structural explanation for preferential interaction of 17A2 with CD3εγ on the T cell surface. To investigate regions of CD3γ critical for native 17A2 binding, we used tg mice generated on the CD3δ−/− background in which one of four segments of mCD3δ was replaced with a corresponding segment from mCD3γ (mutant [mut] A–D), as shown in Fig. 1B, and introduced into the mouse germline via transgenesis. The flow cytometric analysis in Fig. 2A reveals that both CD4 and CD8 T cells from CD3δ−/− mice express low levels of residual 2C11-FITC (2C11-F) reactivity (mean fluorescence intensity [MFI] = 127 and 119, respectively; red curves), consistent with low-level CD3εγ surface expression in the absence of CD3εδ (24). This 2C11-F staining is completely blocked by unlabeled 17A2 (MFI = 10–14; shaded blue curves). In contrast, the introduction of wtCD3δ on the CD3δ−/− background restores 2C11-F reactivity (MFI = 1362–2060), and this binding is inhibited only by ~50% upon preincubation with unlabeled 17A2 (see also schematic in Fig. 2A, right panel). Similar patterns of reactivity and partial blockade of 2C11-F staining by unlabeled 17A2 preincubation are seen with introduction of mutA, mutB, and mutC CD3δ transgenes. However, mutD, involving substitution of the G-strand and a portion of the FG loop from CD3γ into CD3δ, results in T cells whose 2C11 binding is virtually completely blocked by 17A2 (Fig. 2A, mutD; shaded curve). The completeness of this 17A2-mediated inhibition of 2C11-F binding is comparable to that observed in the CD3δ−/− T cells, indicating that 17A2 is able to bind to this mutant CD3εδ heterodimer (Fig. 2A, right panel). Further competitive binding experiments shown in Fig. 2B indicate that 17A2 inhibits 2C11 binding completely on αβ T cells from mutD mice, whereas in contrast, 17A2 inhibits 2C11 binding by only ~50% on αβ T cells from wt mice. Because murine γδ T cells lack CD3εδ, instead expressing two CD3εγ heterodimers per TCR complex (20), equivalent 2C11-F blocking activity by 17A2 on γδ T cells in both wt and mutD mice was expected and is observed (Fig. 2B).
Collectively, these findings show that the CD3γ G-strand segment is critical for 17A2 binding and that this segment can be introduced into CD3δ with no loss of TCR expression or function (Fig. 2). Such is not the case when residues in the CD3δ G-strand region are introduced into CD3γ (vide infra).
On the basis of the structural comparison and amino acid sequence alignment in Fig. 1, we generated a number of CD3γ mutants, including CD3γ6M, a variant in which the six amino acids in the G-strand of CD3γ (ETSNPL) at positions 70–75 were replaced with those of CD3δ (KVVSSV) (Fig. 1A). This swap is predicted to force the preceding segment and loop of the CD3γ6M mutant to adopt a more vertical trajectory than that of the CD3γwt counterpart. To test the functional significance of the CD3γ ectodomain modification, we exploited N15TCR tg mice specific for the vesicular stomatitis virus nuclear protein octapeptide (VSV8) bound to Kb (29, 36) and bred onto a CD3γ−/− genetic background (22). Retroviral transduction of peripheral T cells from these animals with pLZRS-IRES-eGFP–containing CD3γwt or CD3γ mutant cDNAs encoding proteins schematically depicted in Fig. 3A gives rise to GFP-expressing transductants (GFP+). As shown in Fig. 3B, CD3γwt retrovirally transduced N15tg CD3γ−/− T cells restores TCR surface expression, as monitored by flow cytometric analysis with each of four anti-TCR mAbs (MR 9.4, anti-Vβ5; 2C11, anti-CD3ε; H57, anti–TCR-Cβ; and 17A2, anti-CD3εγ). Note that this increase occurs only in the GFP+ population (Fig. 3B). Although the CD3γ6M mutant restores much of the surface TCR expression (~60% of wtCD3γ), as assessed by MR9.4, 2C11, and H57 reactivity, 17A2 mAb binding to TCRs incorporating the CD3γ6M mutant is lost (Fig. 3B). These results support the notion that CD3γ6M adopts an extended G-strand more like mCD3δ, thereby preventing 17A2 mAb from binding in the cleft between CD3ε and CD3γ. Because the direct interaction residues mapped by proton-induced relaxation are exclusively in the CD3ε side (15), the trivial explanation that mutants abrogate 17A2 binding per se is excluded. Subsequently, we shortened the BC loop (LTDKT) in CD3γ by deleting the LTDKT sequence to create a second mutant, CD3γ6M-ΔBC. This foreshortened BC loop likely reinforces the upward trajectory of the FG loop in the CD3γ6M-ΔBC variant. Strikingly, the surface TCR expression rescue observed with CD3γ6M was not detected by any of the four Abs tested after CD3γ6M-ΔBC transduction. Collectively, these data suggest that this kinked CD3γ G-strand geometry is important for TCR surface expression, permitting CD3εγ to occupy a position beneath the Cβ FG loop of the β subunit and consistent with our previous TCR quaternary model (15) (Supplemental Fig. 2). In further support of this hypothesis, a third mutant in which the entire CD3γ ectodomain is replaced with that of CD3δ (CD3γmδecto) also fails to reconstitute TCR surface expression. Cellular protein expression for each of the mutant CD3γ transductants was comparable to that of wtCD3γ, as assessed by intracellular staining using the FLAG-tag appended to the Cyt tail of each construct and anti-FLAG Ab (Supplemental Fig. 3). However, we cannot exclude that in the case of CD3γ6M-ΔBC, incorrect protein folding may contribute to absent surface expression.
Next, following retroviral transduction, N15 T cells were stimulated with VSV8-pulsed Kb-expressing APCs. As shown by the representative experiment in Fig. 3C, no significant IFN-γ production was obtained from GFP+ empty vector control transductants, whereas N15 T cells transduced with wtCD3γ cDNA showed specific responsiveness to VSV8 peptide in a concentration-dependent manner. N15 T cells expressing the CD3γ6M cDNA induced less IFN-γ production compared with that of CD3γwt, emphasizing the importance of the unique CD3γ G-strand geometry for Ag-triggered activation. As expected, VSV8 stimulation failed to induce IFN-γ production in N15 T cells transduced with CD3γ6M-ΔBC or CD3γmδecto.
Fig. 3D represents a composite histogram of IFN-γ production from multiple experiments using VSV8 stimulation of retrovirally transduced N15tgCD3γ−/− T cells. Both CD3γwt and CD3γ6M transductants (GFP+) produce IFN-γ upon VSV stimulation, but with cytokine production from CD3γwt greater than that from CD3γ6M transductants. Note the lack of detectable activation by pMHC in GFP− cells in either set of cultures. In contrast, with anti-CD3ε mAb triggering, IFN-γ production is observed in transduced (GFP+) and nontransduced (GFP−) populations, implying that 2C11 stimulation via CD3εδ induces detectable IFN-γ production. Nonetheless, transduction of CD3γwt significantly augments that 2C11-stimulated response, compared with the vector control. Similar 2C11-stimulated IFN-γ production in CD3γwt and CD3γ6M transductants suggests that differential responsiveness to pMHC is not a consequence of TCR expression level. Together, the findings imply that TCRβ-CD3εγ juxtaposition involving the kinked CD3γ G-strand affords an optimized geometry for effective Ag-triggered T cell activation.
Whereas the striking elongation of the Cβ FG loop among mammalian species is well conserved, sequence comparison with nonmammalian vertebrate species (chicken, fish, and frog) reveals that the lengthy Cβ FG loop is not observed in the latter (Fig. 4A). This rigid Cβ FG loop in mouse has been shown to facilitate both selection of thymocytes and activation of T cells (17, 18). The absence of distinct CD3γ and CD3δ subunits in the above-mentioned nonmammalian species and expression of a single precursor CD3γ/δ gene have been shown based on genomic and biochemical analyses, as well as theoretical predictions dating the required CD3 duplication event (37–39). Although recently it has been demonstrated that the jawless vertebrates (agnathans) have alternative adaptive immune systems with variable lymphocyte receptors, all jawed vertebrates (gnathostomates) possess fully developed adaptive immune systems with TCR and Ig genes (40). Both the elongated Cβ FG loop and the distinct CD3γ and CD3δ genes are unique in the mammalian species among Gnathostamata (Fig. 4B). These analyses support the notion that TCRCβ and CD3γ have been evolutionarily coupled for TCR assembly and signaling in the mammalian species. Furthermore, these findings imply that the distinct topology of CD3 heterodimers coevolved with TCRC domains to optimize the quaternary TCR structure for pMHC-triggered αβ TCR activation. TCRαβ heterodimer assembly studies with various CD3 complexes, using a phylogenetic approach with TCR complex subunits from mammalian and nonmammalian vertebrates, support this conclusion (41).
Despite the above results and conclusions, it has been reported that hCD3δ can partially restore thymic development in the absence of mCD3γ in CD3γ−/− mice (42, 43), suggesting that the hCD3δ ectodomain may be accommodated in juxtaposition to the murine β subunit. We therefore assessed whether a chimeric CD3γ protein containing the hCD3δ ectodomain fused with transmembrane and Cyt tail segments of mCD3γ (CD3γhδecto) could restore TCR surface expression and T cell activation in N15tgCD3γ−/− T cells. In contrast to the inability of CD3γmδecto chimera to restore TCR expression and function, CD3γhδecto was competent to do both. As shown in Fig. 5A, CD3γhδecto transduction induced surface TCR expression comparable to that of CD3γwt, although, not surprisingly, with loss of 17A2 mAb reactivity. In vitro stimulation experiments likewise showed that CD3γhδecto possesses a signaling capacity equivalent to that of CD3γwt to induce IFN-γ production upon pMHC- (Fig. 5B) or 2C11- mediated stimulation (Fig. 5C). Thus, CD3γhδecto serves as a suitable structural surrogate for both VSV8- and 2C11-triggered activation.
To next assess whether this CD3γhδecto chimera can provide differentiation signals during development, an FTOC system was employed. Thymocyte progenitors from CD3γ−/− fetal mice were transduced with CD3γwt or CD3γhδecto cDNA containing retroviruses of comparable viral titer. Subsequently, thymocytes generated within the reconstituted FTOC after 7 d of culture were prepared and analyzed for GFP expression and surface phenotype. As shown by CD4 and CD8 surface expression patterns in Fig. 5D (top), thymic development in the CD3γhδecto transduced thymocytes was equivalent or better than that of CD3γwt transduced thymocytes. In contrast to nontransduced cells (GFP−) within the same FTOC cultures, the number of double positive thymocytes was increased 2- to 5-fold. Fig. 5D (middle) shows that double negative (DN) cell development from DN1 (CD44+CD25−) plus DN2 (CD44+CD25+) stages to DN3 (CD44−CD25+) and DN4 (CD44−CD25−) stages was also accelerated by the transduction of CD3γhδecto. Thus, for example, in CD3γhδecto transductants, DN3 = 9.8% and DN4 = 42.2%, compared with 1.1 and 3.9%, respectively, for the GFP− control. Furthermore, surface TCRβ expression on DN3/4 stage cells was only slightly higher than on nontransduced cells (Fig. 5D, bottom), suggesting that pre-TCR expression is supported by CD3γhδecto, as it is with wtCD3γ without nonphysiological overexpression of the pre-TCR. These results imply that the CD3γhδecto protein can replace that of the wtCD3γ component in the pre-TCR during T cell development.
Our findings with the chimeric CD3γhδecto protein are in agreement with results generated through transgenesis in CD3γ−/− mice using the hCD3δ (human only) construct (42, 43). Why, then, might these mouse-human interspecies CD3εδ heterodimers function to foster signaling (Fig. 6)? In this regard, it is noteworthy that the aromatic ring of Phe89 in human CD3ε makes hydrophobic interactions with both the γ-methyl of Thr62 and the hydrophobic β-methylene position of the Gln64 side chain of human CD3δ. This interaction creates a tightly packed interface between hCD3ε and hCD3δ molecules at the top of the G–G-strand interface (Supplemental Fig. 4). This preferential interaction would not be formed in the mCD3εγhδecto dimer, as the Phe89 in hCD3ε is a Thr in the corresponding position mCD3ε. As a result, the top of the G-strand in CD3γhδecto and hCD3δ proteins, when dimerized with mCD3ε, can bend and slot into the area normally accommodating mCD3εγ in the TCRβ–CD3εγ junction, without steric clash. Furthermore, hCD3δ contains four alternating charged residues at the end of the FG loop (K57/D58/K59/E60). This charge cluster would destabilize a straight-up β-strand conformation with two positively charged lysines on one side and two negatively charged residues on the back side. Thus, the FG loop of hCD3δ is most likely unstructured, consistent with its crystal structure (Supplemental Fig. 1) (35), and could be readily bent toward the back ABED face when heterodimerized with mCD3ε in a fashion analogous to that of mCD3γ. Differential glycosylation of mCD3δ versus hCD3δ with one fewer N-linked adduct in the human ortholog may also contribute to the functionality of this replacement.
Analysis of CD3 sequence divergence indicates that CD3γ, CD3δ, and CD3ε genes arose from a common ancestor in a two-step process of gene duplication (38). Mammals have three CD3 genes (γ, δ, and ε), whereas nonmammalians (birds, fish, and amphibians) have only two: a CD3γδ precursor and a CD3ε gene. Protein sequence comparison indicates that each CD3γ and CD3δ subunit evolved with highly homologous heterodimeric interfaces and membrane proximal segments for efficient and specific signaling transfer when paired with CD3ε. The compact orientation of the CD3γ FG loop and single horizontally attached glycan in the mouse is a feature of the CD3εγ heterodimer that affords lateral support of the C region domains for αβ as well as γδ TCRs. In addition, CD3εγ appears to have adapted to optimally interact with the Cβ FG loop. Our findings that the elongation of the structured Cβ FG loop coevolved with appearance of the CD3γ gene from a CD3γ/δ precursor are strong support for this notion. By contrast, the more vertical CD3δ FG loop trajectory and greater number of N-linked glycan adducts in CD3εδ heterodimers assume a more extended geometry that cannot fit into the homologous TCRβ–CD3εγ interaction site (15). The CD3εδ disposition on the TCRα “side” of the complex occupies intervening space between the coreceptor (CD4 or CD8) and the αβ heterodimer (44, 45). This bulky CD3εδ component may also be entropically advantageous to help preconfigure the coreceptor as a TCR, pMHC, and coreceptor ternary complex forms.
The crystal structure of a γδ TCR heterodimer reveals a Cγ–Cδ domain symmetry, in contradistinction to the Cα–Cβ domain asymmetry observed in αβ TCRs (16, 46, 47). The γδ TCR heterodimer also differs by lacking an elongated Cβ FG loop equivalent. γδ TCR lineage commitment is associated with more robust signaling relative to that of the αβ TCR, with greater TCR copy number and/or ligand density likely affecting γδ T lineage signaling strength (22, 48). In contrast, αβ TCR pMHC ligands are present at low levels, mandating additional TCR modifications to compensate for weak signals promoting αβ T cell fate and function. During immune surveillance, continued cell movement following ligation of the TCR αβ clonotype by specific pMHC fosters quaternary change; the Cβ FG loop interacts with CD3εγ on one side of the TCR and the Cα domain with the bulky CD3εδ heterodimer on the other. It is the tangential rather than normal (i.e., perpendicular to the membrane) directional force that triggers TCR activation post-pMHC ligation, as shown by optical tweezer experiments (15).
Our current results emphasize how the αβ TCR quaternary structure is optimal for surface expression and signaling. Fig. 6A gives a side view of the surface-exposed TCR complex based upon existing structural information of individual components and molecular modeling (8, 15, 49). The substantial N-linked glycosylation of TCR subunits is indicated by the brown space-filling molecular representations. On the TCRβ “side,” the Cβ FG loop (17, 18) and the compact orientation of the CD3γ FG loop (Fig. 1, Supplemental Fig. 2) are key features contributing to the asymmetry optimizing TCR signaling. Lateral movement of the TCRαβ heterodimer can apply a torque on CD3εγ via the Cβ FG loop appendage. Fig. 6B schematically demonstrates that the extended CD3δ subunit ectodomain would sterically clash with the Cβ FG loop above, whereas that of CD3γ or the chimeric heterodimer does not. On the TCRα “side,” the bulky glycans and vertical CD3δ FG loop disposition may also likewise relay quaternary change to CD3εδ after tangential force-induced torque. Alternatively, the torque on CD3εδ could be applied through the highly conserved connecting peptide at the base of the TCRCα domain (30). Further studies aimed at rigidifying or derigidifying segments of the TCR complex, without altering pMHC binding, should show an impact on TCR signaling, consistent with a mechanosensor mechanism of action.
The ability of the hCD3δ ectodomain to pair with mCD3ε and foster TCR complex expression signaling, as well as murine thymocyte development, might appear, at first glance, contradictory to our view that CD3γ and CD3δ ectodomains evolved to occupy a different side of the TCR complex. However, that is not the case. Isologous subunit ectodomain substitution is not permitted, whereas the heterologous hCD3δ ectodomain can replace that of mCD3γ. That functional substitution, as noted in Supplemental Fig. 4, is possible because mCD3ε has a Thr residue in lieu of Phe89 in hCD3ε, allowing the top of the G-strand of hCD3δ, when paired with mCD3ε, to avoid steric clash with the TCRβ subunit. We emphasize that the geometry of mCD3γ and mCD3δ G-strand N-terminal residues (residues 70–75 and residues 58–63) are distinct from each other, as are the corresponding segments in the respective human orthologs.
As structural and functional analyses of these and other Ig-like domains of receptors become more sophisticated, additional subtleties and their biological implications will be revealed. The details as described in this study for CD3 heterodimers demonstrate the important functional consequences of structural evolution. Understanding these differences will help with elucidating the function of multisubunit receptors, such as the TCR.
We thank Haesook Kim for statistical analysis, Maris Handley for flow cytometry, and Jiahuai Wang for careful review of this manuscript.
This work was supported by National Institutes of Health Grants AI19807 (to E.L.R.), CA74620 (to D.J.K.), AI37581, GM47467, and EB002026 (to G.W.), and Canadian Institutes of Health Research Grant 81145 (to V.P.D.).
The online version of this article contains supplemental material.
The authors have no financial conflicts of interest.