Identification of the Docking Site for CD3 on the T Cell Receptor β Chain by Solution NMR

Background: Understanding T cell signaling requires knowing the structure of the TCR-CD3 complex.

Results: Solution NMR was used to identify the docking site for CD3 ectodomains on the TCR β chain.

Conclusion: The docking site (∼400 Ų) comprises ∼10 Cβ residues at the base of the TCR.

Significance: CD3 is located opposite to the peptide-MHC binding site of the TCR.

Abstract

The T cell receptor (TCR)-CD3 complex is composed of a genetically diverse αβ TCR heterodimer associated noncovalently with the invariant CD3 dimers CD3ϵγ, CD3ϵδ, and CD3ζζ. The TCR mediates peptide-MHC recognition, whereas the CD3 molecules transduce activation signals to the T cell. Although much is known about downstream T cell signaling pathways, the mechanism whereby TCR engagement by peptide-MHC initiates signaling is poorly understood. A key to solving this problem is defining the spatial organization of the TCR-CD3 complex and the interactions between its subunits. We have applied solution NMR methods to identify the docking site for CD3 on the β chain of a human autoimmune TCR. We demonstrate a low affinity but highly specific interaction between the extracellular domains of CD3 and the TCR constant β (Cβ) domain that requires both CD3ϵγ and CD3ϵδ subunits. The mainly hydrophilic docking site, comprising 9–11 solvent-accessible Cβ residues, is relatively small (∼400 Ų), consistent with the weak interaction between TCR and CD3 extracellular domains, and devoid of glycosylation sites. The docking site is centered on the αA and αB helices of Cβ, which are located at the base of the TCR. This positions CD3ϵγ and CD3ϵδ between the TCR and the T cell membrane, permitting us to distinguish among several possible models of TCR-CD3 association. We further correlate structural results from NMR with mutational data on TCR-CD3 interactions from cell-based assays.

Introduction

T cells play a major role in generating adaptive immune responses to microbes and cancers. This process is mediated by the T cell receptor (TCR)⁵-CD3 complex, which is composed of a genetically diverse αβ (or γδ) TCR heterodimer in noncovalent association with invariant CD3 dimers: CD3ϵγ, CD3ϵδ, and CD3ζζ (1, 2). The TCR recognizes antigenic peptides presented by major histocompatibility complex (MHC) molecules, whereas the CD3 subunits transmit activation signals to the T cell. The TCR α and β chains do not possess intracellular signaling motifs, thereby separating peptide-MHC (pMHC) recognition from T cell activation. By contrast, the CD3 ϵ, γ, δ, and ζ chains each contain at least one immunoreceptor tyrosine-based activation motif (ITAM). The TCR-CD3 complex exists in 1:1:1:1 stoichiometry for the TCRαβ/CD3ϵγ/CD3ϵδ/CD3ζζ dimers (3).

X-ray crystallographic studies of TCR-pMHC complexes have revealed the molecular basis for TCR recognition of foreign antigens and self-antigens (4–8). In addition, much is known about the downstream signaling cascade, following phosphorylation of CD3 ITAMs by the Src kinase Lck associated with CD4 or CD8 (9). However, the mechanism(s) by which TCR ligation is communicated to the CD3 signaling apparatus remains a mystery (10, 11). A critical missing element to solving this puzzle is knowledge of the spatial organization of the TCR-CD3 complex and the precise interactions between TCR and CD3 subunits. Although structures of the immunoglobulin (Ig)-like ectodomains of isolated CD3ϵγ and CD3ϵδ heterodimers have been determined (12–15), exactly how these subunits associate with TCR on the T cell surface remains to be established.

A variety of approaches have been used to infer the stoichiometry, pairing, and orientation of subunits within the TCR-CD3 complex. For example, mutational studies have demonstrated the importance of conserved charged residues in the transmembrane helices of the TCR and CD3 chains for complex assembly (1, 16). The following pairings have been established: TCRα-CD3ϵδ, TCRα-CD3ζζ, and TCRβ-CD3ϵγ. Besides these transmembrane interactions, extracellular interactions between the TCR and the Ig-like ectodomains of CD3ϵγ and CD3ϵδ also contribute to the structural integrity and function of the TCR-CD3 complex (2, 11, 17–21). By contrast, the ectodomain of CD3ζ, which is only 9 residues long, has not been implicated in interactions with the TCR. Although possible docking sites on the TCR for CD3 subunits have been proposed (18–20), these key interaction regions remain undefined.

Recently, small angle x-ray scattering (SAXS) was applied to a soluble TCR-CD3ϵδ complex that was engineered for stability by replacing the transmembrane domains with a heterotrimeric coiled coil (22). The SAXS results indicated that the extracellular domains (ECDs) of CD3ϵδ are situated underneath the TCR α chain, although a docking site was not identified. This arrangement was consistent with negative stain electron microscopy (EM) of the entire TCR-CD3 integral membrane complex, which further showed that the CD3ϵγ ECDs sit beneath the TCR β chain (22). Intriguingly, EM images revealed mainly dimeric complexes in which two TCRs projected outward from a central core composed of the CD3 ECDs.

Although SAXS and negative stain EM have provided important information on the overall organization of the TCR-CD3 complex, these structural techniques lack the resolution necessary to establish the precise orientation of TCR and CD3 subunits in the complex or to define TCR-CD3 interfaces. At present, such information can only come from x-ray crystallography, NMR, or cryo-EM, applied individually or in combination. However, efforts to crystallize TCR-CD3 complexes have been thwarted by the very low affinity of TCR-CD3 interactions in solution. Similarly, previous attempts to map binding epitopes by NMR were unable to detect an interaction between TCR and CD3 ECDs (14). Using optimized protein constructs and isotope labeling conditions, we have revisited the application of NMR to define TCR-CD3 contact sites in solution. Here we demonstrate a specific interaction between TCR and CD3ϵγ and CD3ϵδ ECDs by NMR, identify the docking site for CD3 on the TCR β chain, and correlate our structural results with mutational data on TCRβ-CD3 interactions from cell-based assays.

Experimental Procedures

Production and Purification of CD3ϵγ and CD3ϵδ Heterodimers

Human CD3ϵγ and CD3ϵδ heterodimers were prepared by in vitro folding from inclusion bodies produced in Escherichia coli. CD3ϵδ needed to be folded together with the anti-CD3ϵ antibody UCHT1 in order to form a stable species (15). By contrast, the CD3ϵγ heterodimer was stable on its own. DNA fragments encoding the ECDs of CD3ϵ (residues 1–105), CD3γ (residues 1–90), and CD3δ (residues 1–79) were inserted separately into the vector pET-26b (Novagen) and expressed as inclusion bodies in BL21(DE3) E. coli cells (Novagen). The UCHT1 antibody was likewise produced in inclusion bodies as a single-chain Fv fragment (scFv) in which the light chain variable region was connected to the heavy chain variable region by a (GGGGS)₃ linker. Bacteria were grown at 37 °C in LB medium to A₆₀₀ = 0.6–0.8 and induced with 1 mm isopropyl-β-d-thiogalactoside. After incubation for 3 h, the bacteria were harvested by centrifugation and resuspended in 50 mm Tris-HCl (pH 8.0) containing 0.1 m NaCl and 2 mm EDTA; cells were disrupted by sonication. Inclusion bodies were washed extensively with 50 mm Tris-HCl (pH 8.0) and 2% (v/v) Triton X-100. The CD3ϵ, CD3γ, and CD3δ chains were dissolved in 8 m urea, 50 mm Tris-HCl (pH 8.0), and 10 mm DTT; UCHT1 scFv was dissolved in 6 m guanidinium HCl, 100 mm Tris-HCl (pH 8.0), and 10 mm DTT.

For in vitro folding of CD3ϵγ, solubilized CD3ϵ and CD3γ inclusion bodies were rapidly diluted into ice-cold folding buffer containing 0.8 m arginine HCl, 50 mm Tris-HCl (pH 8.0), 2 mm EDTA, 3.7 mm cystamine, and 6.6 mm cysteamine to final protein concentrations of 59 and 52 mg/liter, respectively. For in vitro folding of scFv-CD3ϵδ, CD3ϵ, CD3δ, and scFv inclusion bodies were diluted into the same buffer at final concentrations of 59, 45, and 130 mg/liter, respectively. After 72 h at 4 °C, the folding mixtures were concentrated 50-fold, dialyzed overnight against 20 mm Tris-HCl (pH 8.0) and 100 mm NaCl, and centrifuged to remove aggregates. Correctly folded CD3ϵγ and scFv-CD3ϵδ proteins were then purified using sequential Superdex S-200 and MonoQ columns (GE Healthcare).

Production of TCR MS2-3C8 with Isotope-labeled β Chain

The β chain of TCR MS2-3C8 (residues 1–244) with U-²H,¹³C,¹⁵N labeling was obtained by inclusion body expression in E. coli BL21(DE3) cells (Agilent) transformed with pET-26b. Transformed cells were grown in 25 ml of LB medium containing 30% D₂O (Isotec) at 37 °C until A₆₀₀ ≥ 1.0 and then transferred in a 1:50 dilution to 25 ml of M9 medium containing 70% D₂O, 1 g/liter [¹⁵N]NH₄Cl (Cambridge Isotope) as the sole source of nitrogen and 4 g/liter [²H,¹³C]glucose (Cambridge Isotope) as the sole source of carbon. The culture was grown until A₆₀₀ = 0.5–1.0 and was transferred in a 1:100 dilution to 100 ml of M9 containing 100% D₂O and grown overnight. The overnight cultures were used to inoculate 1 liter of M9/D₂O to a starting of A₆₀₀ = 0.10. Induction with isopropyl-β-d-thiogalactoside to a final concentration of 1 mm was performed at A₆₀₀ ≥ 0.6, and growth was continued for 3–4 h at 37 °C. The bacteria were disrupted by sonication. Inclusion bodies were washed with and without 5% (v/v) Triton X-100 and then solubilized in 8 m urea, 50 mm Tris-HCl (pH 8.0), and 10 mm DTT.

The unlabeled TCR MS2-3C8 α chain (residues 1–205) was expressed as inclusion bodies in E. coli BL21(DE3) cells transformed with pET-26b. Bacteria were grown at 37 °C in LB medium to A₆₀₀ = 0.5–1.0 and induced with 1 mm isopropyl-β-d-thiogalactoside. The inclusion bodies were washed with 5% (v/v) Triton X-100 and dissolved in 8 m urea, 50 mm Tris-HCl (pH 8.0), and 10 mm DTT.

For in vitro folding, the TCR MS2-3C8 α and β chains were mixed in a 3.6:1 molar ratio and diluted into a folding mixture containing 5 m urea, 0.4 m l-arginine HCl, 5 mm EDTA, 3.7 cystamine dihydrochloride, and 6.6 mm cysteamine to a final concentration of 50 mg/liter. The folding mixture was dialyzed against H₂O for 72 h at 4 °C and then dialyzed against 10 mm Tris-HCl (pH 8.0) for 48 h at 4 °C. After removal from dialysis, the folding mixture was concentrated and dialyzed against 50 mm MES (pH 6.0) at 4 °C overnight. Disulfide-linked TCR MS2-3C8 heterodimer was purified using sequential MonoQ and Superdex S-200 columns.

A U-²H,¹⁵N-labeled MS2-3C8 β chain with U-¹³C, ¹⁵N-labeled Ile, Leu, and Val was produced as described above. [¹H,¹²C]glucose (Sigma) was used as the sole carbon source, and 50 mg/liter of each labeled amino acid (Sigma/Isotec) was added to M9/D2O 1 h prior to induction. A U-²H,¹⁵N-labeled MS2-3C8 β chain for titration experiments was produced similarly except for the use of [¹H,¹²C]glucose as the sole carbon source.

NMR Spectroscopy

TCR MS2-3C8α[β-²H,¹³C,¹⁵N] sample concentrations in the range of 200–240 μm were used in 50 mm NaP_i, 100 mm NaCl (pH 7.0) with the addition of 0.1 mg/ml Pefabloc (Sigma). NMR spectra were acquired at 25 °C on Bruker AVANCE III 600 and 900 Hz spectrometers fitted with Z-gradient ¹H/¹³C/¹⁵N cryoprobes. Backbone resonance assignment was carried out with TROSY versions of the following standard three-dimensional triple resonance experiments: HNCACB, HN(CO)CACB, HNCA, HN(CO)CA, HN(CA)CO, and HNCO. The HNCACB and HN(CO)CACB experiments were also recorded with optimization on the C^β signals. Three-dimensional spectra were acquired using non-uniform sampling (25%) and reconstructed employing the method of Hyberts et al. (23). Three-dimensional HNCO, HN(CA)CO, HN(CO)CA, and HNCA experiments were also recorded on an MS2-3C8 sample where the β chain was deuterated and selectively [¹³C,¹⁵N]Ile/Leu/Val(ILV)-labeled. NMR spectra were processed using NMRPipe (24) and analyzed with SPARKY (25).

NMR titration experiments were carried out at 25 °C using 100 μm MS2-3C8α[β-²H,¹⁵N] and either unlabeled CD3ϵγ, scFv-CD3ϵδ, or CD3ϵγ plus scFv-CD3ϵδ. Samples were exhaustively exchanged into identical buffer conditions (50 mm NaP_i, 100 mm NaCl, 0.1 mg/ml Pefabloc, pH 7.0) prior to titration. Unlabeled CD3 samples (typically 50 μm in 0.8–2.4 ml of buffer) were added to 200 μl of 100 μm labeled TCR and concentrated back to 200 μl using Centricon membrane filtration. In each titration, a two-dimensional ¹H-¹⁵N HSQC spectrum was acquired for the MS2-3C8 TCR alone followed by spectra of successive CD3 additions. The following molar ratios were used for titrations: MS2-3C8/CD3ϵγ, 1:2, 1:4, and 1:5.8; MS2-3C8/scFv-CD3ϵδ, 1:2 and 1:3.3; MS2-3C8/CD3ϵγ/scFv-CD3ϵδ, 1:2:2, 1:4:4, and 1:6:6. Chemical shift perturbations were determined from Δδ_total = [(W_HΔδ_H)² + (W_NΔδ_N)²]^½, where Δδ_H and Δδ_N represent ¹H and ¹⁵N chemical shift differences, respectively, between free and CD3-bound states of MS2-3C8, with weighting factors of W_H = 1 and W_N = 0.2. Values of Δδ_total above 0.01 ppm were estimated to be significant based on errors in the two-dimensional ¹H-¹⁵N HSQC measurement of chemical shifts in multiple identical MS2-3C8 samples. Experimentally significant decreases in peak intensity were estimated as being above the average peak intensity decrease plus one S.D. value.

Generation of TCR-CD3 Mutants

A CD3 multicistronic construct containing CD3ϵ^WT, CD3δ^WT, CD3ζ^WT, and CD3γ^WT and an AND TCR multicistronic vector containing TCRα^WT and TCRβ^WT were generated as described previously (26–28). Briefly, the 2A peptide-linked CD3 and AND TCR constructs were generated by recombinant PCR and cloned into pMIG and pMIA, respectively (MSCV-based retroviral vectors containing an IRES-GFP or IRES-Ametrine1.1 (Addgene) cassette). Specific mutations in TCRβ (V166R, T199R, N203R, and R205E) were made by recombinant PCR, and subsequent mutant constructs were generated by appropriate subcloning.

TCR-CD3 Expression Analysis in HEK-293T Cells

Transient transfection of HEK-293T cells was performed as described previously with some modifications (29). HEK-293T cells were incubated in 6-well plates at 2 × 10⁵/well overnight at 37 °C. The AND TCRαβ pMIA plasmid (0.3 μg) and CD3 pMIG plasmid (0.3 μg) were transfected into HEK-293T cells using 6 μl of TransIT LT1 transfection reagent (Mirus). Cells were harvested 48 h after transfection and stained with anti-CD3ϵ and anti-TCRβ monoclonal antibodies for flow cytometric analysis.

Results

NMR Assignment of TCR MS2-3C8α[β-²H,¹³C,¹⁵N]

For detailed NMR studies, we used a human autoimmune TCR, MS2-3C8, which recognizes a self-peptide from myelin basic protein (MBP) and the multiple sclerosis-associated MHC class II molecule HLA-DR4 (30). The pathogenic potential of MS2-3C8 was demonstrated using mice transgenic for this TCR and HLA-DR4 (31). An x-ray crystal structure of TCR MS2-3C8 bound to its pMHC ligand, MBP-HLA-DR4, has been determined (30), as well as that of the ternary complex formed by MS2-3C8, MBP-HLA-DR4, and CD4 (32). The interactions of MS2-3C8 with several of the key players in T cell signaling are therefore well characterized structurally, making this TCR a good candidate for studying critical associations with CD3ϵγ and CD3ϵδ. The MS2-3C8 TCR employed here was modified slightly from the wild-type used for previous x-ray studies, with three mutations in the α chain and four mutations in the β chain. The mutations were made to improve either expression (VβT55A), folding yield, and stability (CαT159C, CαC206(deleted), CβS171C, CβS189A, and CβC245(deleted)) or binding to HLA-DR4 (VαA31T). None of these mutations are at positions that contact CD3 (see below). Thus, the TCR MS2-3C8 ectodomain studied here is a heterodimer consisting of an unlabeled 205-residue α chain and a ²H,¹³C,¹⁵N-labeled 244-residue β chain for a total molecular mass of ∼50 kDa. Due to this relatively high molecular mass, transverse relaxation-optimized (TROSY)-based triple resonance experiments were utilized for the NMR assignment process (33). Interresidue connectivities were obtained through sequential matching of C^α and C^β chemical shift patterns and confirmed with CO correlations in most cases. A selectively labeled ILV sample provided further validation of the assignments using methods described previously (34). Overall, 200 (87%) of the 230 possible main chain amide assignments were obtained, with extensive coverage of both TCR variable β (Vβ) and constant β (Cβ) domains (Fig. 1, A–D). Assignments of H^N, ¹⁵N, C^α, C^β, and CO resonances of the MS2-3C8α[β-²H,¹³C,¹⁵N] sample have been deposited in the Biological Magnetic Resonance Bank (BMRB accession code 26569).

FIGURE 1.

NMR assignment of the TCR ectodomain MS2-3C8α[β-²H,¹³C,¹⁵N]. A, extent of backbone amide assignments for the TCR β chain mapped onto the MS2-3C8 x-ray structure. Assigned regions of the β chain are shown in red, whereas unassigned regions are shown in gray. The α chain is in green. The sites of the CαT159C and CβS171C mutations to form an interchain disulfide (see “Results”) are highlighted in magenta. B, two-dimensional ¹H-¹⁵N TROSY-HSQC spectrum of MS2-3C8α[β-²H,¹³C,¹⁵N] at 25 °C with backbone amide assignments. C and D, expanded regions from B with assignments. A two-dimensional HNCA spectrum of MS2-3C8α[β-²H;ILV-¹³C,¹⁵N] (blue) is superimposed on the two-dimensional HSQC spectrum (black). Peaks labeled in blue are due to ILV residues or residues with an (i − 1) ILV.

TCR MS2-3C8 β Chain Conformation and Dynamics in Solution

Chemical shift index analysis (35) using the assigned C^α, C^β, and CO resonances identified the secondary structure elements for the MS2-3C8 β chain in solution (Fig. 2A). These results were compared with the x-ray structure of MS2-3C8 in complex with MBP-HLA-DR4 (Protein Data Bank accession code 3O6F) (30). The only region that appears to be slightly different is strand D in the Cβ domain, which is a short 3-residue β-strand in the x-ray structure but has a coil conformation in solution. Interestingly, this is the region where we engineered a CαCys¹⁵⁹-CβCys¹⁷¹ interchain disulfide (36) to increase yields of paired TCR αβ heterodimers for the NMR sample. The wild-type protein used for x-ray analysis did not contain this disulfide (30). Generally, however, the secondary structure elements derived from chemical shifts are very similar to those observed in the x-ray structure, indicating comparable fold topologies for MS2-3C8 in solution and crystal forms. The chemical shift data were also used to obtain insights into the flexibility of the β chain in the context of the TCR αβ heterodimer (37). Thus, the high order parameters (S²) for residues in the β chain of MS2-3C8 correspond with more rigid secondary structure elements, whereas lower values reflect less structured and more mobile loop regions (Fig. 2B). Moreover, most high S² regions in MS2-3C8 correlate with the lower temperature (B) factors of more ordered regions in the x-ray structure of the MS2-3C8-MBP-HLA-DR4 complex (Fig. 2, B and C). There are two notable exceptions, however. First, the complementarity-determining region loops have low S² values, consistent with being disordered in the free MS2-3C8 in solution, but have low B factors, consistent with order in the x-ray structure. This difference between the NMR and x-ray observations is to be expected because complexation with MBP-HLA-DR4 in the x-ray study rigidifies the complementarity-determining region loops, whereas the MS2-3C8 is uncomplexed in the solution NMR study. Second, there is a distinct difference between the main chain mobility inferred from order parameters versus B factors for the two α-helical regions in the Cβ domain, αA and αB (Fig. 2, B and C). For MS2-3C8 alone in solution, the relatively high S² values (∼0.9) indicate that αA and αB are ordered α-helices. In contrast, these structured helical regions have high B factors (∼120–140) in the x-ray structure of the MS2-3C8-MBP-HLA-DR4 complex, indicating increased mobility relative to other secondary structured regions in the complex. The differences in main chain flexibility of the αA and αB helices in free and pMHC-bound states are discussed further below.

FIGURE 2.

Secondary structure and main chain flexibility of the β chain of αβ TCR MS2-3C8 in solution. A, comparison of secondary structure elements from x-ray (Protein Data Bank accession code 3O6F) and NMR data. The Vβ and Cβ domains are colored blue and red, respectively. Residues for which backbone amide assignments were not obtained, including prolines and the N-terminal amino acid, are underlined. B, estimated order parameters (S²) from chemical shift data as a function of residue number. Secondary structured and other regions of interest are highlighted as follows: β-strands in the Vβ domain (blue), β-strands in the Cβ domain (red), α-helices in the Cβ domain (green), and complementarity-determining region (CDR) loops in the Vβ domain (yellow). C, temperature B factors for C^α atoms in the β chain of the MS2-3C8-MBP-HLA-DR4 complex (30). Regions are highlighted as in B.

TCR MS2-3C8 Interactions with CD3 ECDs

Unlabeled samples of recombinant CD3ϵγ and scFv-CD3ϵδ were checked for proper folding using one-dimensional ¹H NMR and circular dichroism spectra. ¹⁵N-Labeled samples were also prepared, and two-dimensional ¹H-¹⁵N HSQC spectra indicated well folded proteins (Fig. 3). The addition of unlabeled CD3ϵγ to MS2-3C8α[β-²H,¹⁵N] gave minimal chemical shift perturbations, even at the highest TCR/CD3ϵγ molar ratio of 1:5.8 (Fig. 4, A and B). Most shifts in the NMR spectrum had Δδ_total values less than 0.01 ppm, which is within the estimated experimental error of measurement. The few small changes that did occur were mainly due to residues in the Cβ domain, in the αA and αB helical regions and adjoining loops. Similarly, very few differences in the MS2-3C8α[β-²H,¹⁵N] NMR spectrum resulted when only unlabeled scFv-CD3ϵδ was added. The largest change (Δδ_total ∼0.013 ppm) corresponded to Arg¹⁹⁵ in the Cβ domain, a residue that is at the interface with the Cα domain (Fig. 4, A and B). In contrast to the titrations with separate CD3 ECDs, the binding experiments where CD3ϵγ and scFv-CD3ϵδ were added together to MS2-3C8α[β-²H,¹⁵N] produced more extensive chemical shift perturbations. In this case, 32 residues gave Δδ_total values ranging from 0.01 to 0.045 ppm (Fig. 4, A and B) such that 4 were in the Vβ domain and 28 were in the Cβ domain. These changes are consistent with a low affinity but highly specific interaction between CD3 ECDs and the TCR β chain, where only a subset of the NMR peaks are affected upon ligand binding. Moreover, the finding that both CD3ϵγ and CDϵδ must be added together to produce significant chemical shift perturbations suggests that the CD3 ECDs bind cooperatively to the TCR.

FIGURE 3.

NMR spectra showing that CD3ϵγ and scFv-CD3ϵδ are folded. A, two-dimensional ¹H-¹⁵N HSQC spectrum of CD3ϵγ with the γ-chain ¹⁵N-labeled. B, two-dimensional ¹H-¹⁵N TROSY HSQC spectrum of scFv-CD3ϵδ with the δ-chain ¹⁵N-labeled. Conditions were as in Fig. 1.

FIGURE 4.

Titration of TCR MS2-3C8α[β-²H,¹⁵N] with CD3. A, (top) regions from two-dimensional ¹H-¹⁵N TROSY HSQC spectra of MS2-3C8 with increasing amounts of CD3ϵγ plus scFv-CD3ϵδ. Molar ratios of MS2–3C8/CD3ϵγ/scFv-CD3ϵδ were as follows: MS2–3C8 alone (black), 1:2:2 (red), 1:4:4 (green), and 1:6:6 (magenta). Center, two-dimensional HSQC spectra of MS2–3C8 titrated with CD3ϵγ. Molar ratios of MS2–3C8/CD3ϵγ are as follows: control (black), 1:2 (red), 1:4 (green), and 1:5.8 (magenta). Bottom, spectra of MS2–3C8 titrated with only scFv-CD3ϵδ. Molar ratios of MS2–3C8/scFv-CD3ϵδ are as follows: control (black), 1:2 (red), and 1:3.3 (green). B, chemical shift perturbations (Δδ_total) as a function of residue number for MS2-3C8 with CD3ϵγ plus scFv-CD3ϵδ (black), CD3ϵγ alone (red), and scFv-CD3ϵδ alone (yellow). Values above the horizontal dashed line are experimentally significant. C, the percentage decrease in intensity of MS2-3C8 amide peaks in two-dimensional HSQC spectra as CD3ϵγ and scFv-CD3ϵδ are added together. The MS2-3C8/CD3ϵγ/scFv-CD3ϵδ molar ratios for peak intensity data are as follows: 1:2:2 (black), 1:4:4 (orange), and 1:6:6 (cyan). The corresponding horizontal dashed lines indicate the average value plus one S.D. value.

Control experiments were also carried out where two-dimensional ¹H-¹⁵N HSQC spectra were acquired at different TCR concentrations ranging from 75 to 250 μm. Almost all of the TCR β chain residues have backbone amide chemical shifts that are invariant (Δδ_total ≤ 0.005 ppm) over this range, with a few exceptions in the Vβ domain. The largest Δδ_total value in these control experiments was for His⁸³ (∼0.03 ppm) with smaller values (0.005–0.01 ppm) for Met⁴⁵, Asp⁸⁶, Gly⁹⁶, and Gly⁹⁷. The chemical shift perturbations observed for these residues in the CD3 titrations may therefore have some contribution from small changes in TCR concentration-dependent effects.

Mapping of the most perturbed residues onto the x-ray structure of MS2-3C8 showed significant clustering in the Cβ domain, particularly in the αA and αB helices and neighboring residues, strongly suggesting that this region contains the binding epitope for CD3 (Fig. 5, A and B). Analysis of peak intensities indicated that decreases were uneven as CD3ϵγ and scFv-CD3ϵδ were added, with larger losses corresponding to the two α-helices in the Cβ domain, αA and αB (Fig. 4C). Such disproportionate and localized decreases in peak intensity are indicative of binding interactions and provide further support for the chemical shift perturbation data. Based on the intensity changes, a K_D was estimated for the interaction between MS2-3C8 and CD3ϵγ/ϵδ of ∼320 ± 100 μm (Fig. 6).

FIGURE 5.

The CD3 docking site on TCR MS2-3C8. A, NMR chemical shift perturbations mapped onto the MS2-3C8 x-ray structure (Protein Data Bank code 3O6F). Values of Δδ_total ≥0.01 ppm from the titration of MS2-3C8α[β-²H,¹⁵N] with CD3ϵγ plus scFv-CD3ϵδ are indicated (red). B, surface representation of A. C, results from mutagenesis data obtained here (magenta spheres) and from Fernandes et al. (20) (orange spheres) mapped onto the MS2-3C8 structure with the same orientation as in A. Side-chain heavy atoms are highlighted for residues where mutation led to less than 60% TCR surface expression relative to wild type. D, surface representation of C.

FIGURE 6.

Estimate of K_D for the TCR-CD3 interaction. Plot of the percentage peak intensity decrease versus concentration of each CD3 in the titration of MS2-3C8 with CD3ϵγ and scFv-CD3ϵδ. Amide peak intensity changes are shown for αA and αB residues as follows: Glu¹³² (filled circles), Glu¹³⁴ (open circles), Ile¹³⁵ (filled squares), Ser¹³⁶ (open squares), His¹³⁷ (filled triangles), Thr¹⁹⁹ (open triangles), Phe²⁰⁰ (filled diamonds), Asn²⁰³ (open diamonds). A binding curve is shown for each individual residue (dashed lines) and for the mean values (bold line).

The TCR β chain binding epitope thus identified by NMR is centered on the αA and αB helices, which are located at the base of the TCR (Fig. 5, A and B). This would position CD3ϵγ and CD3ϵδ between the TCR and the T cell membrane, in agreement with the shorter length of the CD3 stalks compared with the TCR α and β chain stalks. The epitope is mainly hydrophilic, including the following solvent-accessible residues: Glu¹³², Ser¹³⁶, His¹³⁷, Ser¹⁹⁷, Thr¹⁹⁹, Phe²⁰⁰, Asn²⁰³, Arg²⁰⁵, Asn²⁰⁶, and possibly also Trp²⁴⁰ and Arg²⁴². The surface area is relatively small (∼400 Ų), consistent with the weak interaction between TCR and CD3, although interactions from the α chain are as yet undetermined and may also contribute to binding with CD3. Importantly, the TCR β chain binding epitope is devoid of glycosylation sites, which are located well away from the αA/αB docking site with CD3.

A number of the residues undergoing chemical shift perturbations are not solvent-accessible and so cannot directly contact CD3. Notably, many of these are at the interface between the β and α chains of the MS2-3C8 TCR. Given that the chemical shift is an exquisitely sensitive measure of structure, these changes may reflect small adjustments in orientation between the α and β subunits due to cooperative binding of CD3ϵγ and CD3ϵδ. Even binding of CD3ϵγ and scFv-CD3ϵδ individually to MS2-3C8 both appear to contribute to weak perturbations at the α/β-interface (e.g. Arg¹⁹⁵ for scFv-CD3ϵδ and Val¹²⁵, Phe¹²⁸, and Ser¹³¹ for CD3ϵγ). If small domain rearrangements between the α and β chains of TCR are important, then it is possible that the non-native interchain disulfide between CαCys¹⁵⁹ and CβCys¹⁷¹ may influence CD3 binding. Small long range effects like those seen here are commonly observed in chemical shift mapping of protein binding interactions (38).

Comparison of Binding Epitope with Mutational Data on TCR-CD3 Interactions

The docking site for CD3 on the TCR β chain identified by NMR (Fig. 5, A and B) may be compared with information on this site obtained from cell-based assays. In an extensive analysis of TCRβ-CD3 interactions, Fernandes et al. (20) mutated all Cβ residues whose side chains are >50% solvent-exposed and tested the effect of these mutations on surface expression of the TCR-CD3 complex in transduced Jurkat T cells lacking a functional TCR β chain. In Fig. 5, C and D, mutations in Cβ that reduced TCR expression by >60% compared with wild type are shown mapped onto the x-ray structure of MS2-3C8. Strikingly, both the NMR and cell-based approaches identified residues in the αA and αB helices as interacting with CD3. However, mutations that affected TCR surface expression were broadly distributed over the Cβ domain, in sharp contrast to the clustering of residues showing the largest chemical shift perturbations at the base of Cβ (Fig. 5, A and B). We conclude that the higher number of Cβ residues implicated by mutagenesis (20) than by NMR indicates that reduced TCR surface expression most likely reflects not only disruption of TCR-CD3 contacts but also deleterious effects on TCR folding, stability, and/or intracellular trafficking.

Impact of Mutations in the TCRβ-CD3 Interface on Cell Surface Expression

Finally, we assessed the importance of the TCRβ-CD3 interaction interface that we defined by NMR on TCR-CD3 cell surface expression. Because the key TCRβ residues at this interface are conserved between mice and humans (Val¹⁶⁶/Thr¹⁹⁹/Asn²⁰³/Arg²⁰⁵), we made non-conservative mutations in the H-2E^k-restricted, moth cytochrome c 88–102-specific AND TCR at these residues to determine whether this led to the disruption of the TCRβ-CD3 interaction. Multicistronic 2A-peptide-linked vectors were generated that contained the AND TCRαβ with either a wild type or mutant TCRβ (TCRβ^WT or TCRβ^Mut (V166R, T199R, N203R, or R205E)), and an IRES.mAmetrine cassette to facilitate detection of transfectants by flow cytometry (26–29). Wild type or mutant AND TCR-encoding vectors were co-transfected with a similarly constructed multicistronic vector containing all four wild type CD3 chains (CD3ϵ^WT/CD3δ^WT/CD3ζ^WT/CD3γ^WT) that included an IRES-GFP cassette into HEK 293T cells. Cells were gated on GFP/mAmetrine, and the mean fluorescence of CD3ϵ and TCRβ was determined. A substantial reduction in TCR-CD3 expression (∼66%) was reproducibly observed with TCRβ^Mut- versus TCRβ^WT-expressing cells (Fig. 7). These data suggest that this interface is required for optimal TCR-CD3 cell surface expression.

FIGURE 7.

Impact of structure-based mutations in TCRβ (V166R, T199R, N203R, and R205E) on surface expression of TCR-CD3 in transfected 293T cells. HEK 293T cells were co-transfected with a multicistronic vector containing the AND-specific TCRα^WT with either TCRβ^WT or TCRβ^Mut (V166R, T199R, N203R, and R205E) and IRES-mAmetrine and a multicistronic vector containing CD3ϵ^WT/CD3δ^WT/CD3ζ^WT/CD3γ^WT and IRES-GFP. GFP⁺/mAmetrine⁺ cells were gated, and the mean fluorescence of CD3ϵ and TCRβ was determined. The data were calculated as the percentage expression of WT. The data are the mean ± S.E. (error bars) of six independent experiments (****, p < 0.0001).

Discussion

We have demonstrated the feasibility of using NMR chemical shift mapping to define contacting surfaces between TCR and CD3 in solution, despite the low affinity of the interaction (K_D ∼320 μm). An earlier NMR study did not detect binding between an unlabeled mouse TCR and ¹⁵N-labeled CD3ϵδ (14). In fact, this result is consistent with our findings that the addition of unlabeled CD3ϵδ or CD3ϵγ alone to MS2-3C8α[β-²H,¹⁵N] produced minimal chemical shift changes and that significant perturbations necessitated the addition of both CD3ϵδ and CD3ϵγ. This suggests that CD3 ECDs bind cooperatively to the TCR, in agreement with evidence from cell-based mutagenesis experiments that extracellular TCR-CD3ϵγ interactions require TCR-CD3ϵδ interactions for stabilization (18). Cooperativity could also arise from CD3ϵγ–CD3ϵδ interactions, although direct contacts between CD3 ECDs remain to be shown.

The CD3 docking site on Cβ, as defined here by NMR, comprises 9–11 residues that form a contiguous, mainly hydrophilic patch measuring ∼400 Ų. Assuming a total buried surface of ∼800 Ų (2 × 400 Ų), the TCRβ-CD3 interface would be below the lower limit of the average value of 1600 ± 400 Ų for stable protein-protein complexes, as defined by x-ray crystallography (39). However, it would be comparable in size with the interfaces of transient protein-protein complexes that mediate weak interactions in intracellular signaling or cell-cell adhesion (40). Although a variety of mechanisms have been proposed to explain TCR triggering, they all converge on the idea that TCR ligation somehow results in perturbation of the extracellular region of the TCR-CD3 complex through rigid body movements or intermolecular conformational changes (10, 11, 17). In this regard, the weak and transient TCRβ-CD3 interaction revealed by our NMR analysis would facilitate shifts in the orientation of CD3 subunits with respect to TCR. By contrast, a more stable interaction mediated by a larger TCRβ-CD3 interface might rigidify the TCR-CD3 complex and thereby prevent effective T cell signaling. This hypothesis could be tested by engineering a high affinity TCRβ-CD3 interface through structure-guided mutagenesis.

We have shown that CD3 docks on the αA and αB helices of Cβ. This contact site positions CD3 at the base of the TCR, rather than alongside it, in agreement with results from SAXS and negative stain EM (22). As a consequence, CD3 would be situated between the TCR and T cell membrane, which is consistent with the greater length of the TCR α and β chain stalk regions (21 and 16 residues, respectively) than those of CD3ϵγ and CD3ϵδ (each ∼9 residues). Mutational studies have demonstrated the importance of these highly conserved stalks for the assembly and function of the TCR-CD3 complex (41–45). Extension of our NMR analysis to the MS2-3C8 α chain will define the docking site for CD3 ECDs on the TCR more completely.

X-ray studies have shown that the ternary complex formed by MS2-3C8, MBP-HLA-DR4, and the CD4 co-receptor resembles a pointed arch in which TCR and CD4 are each tilted relative to the T cell membrane (Fig. 8) (32, 46). This structure, in conjunction with the location of the CD3 docking site established by NMR, places CD3ϵγ and CD3ϵδ inside the TCR-pMHC-CD4 arch, wedged between the TCR and T cell membrane and facing CD4 (Fig. 8). Importantly, this particular arrangement would achieve intracellular juxtaposition of CD3 ITAMs with the tyrosine kinase Lck bound to CD4, thereby promoting phosphorylation of CD3 ITAMs and T cell triggering.

FIGURE 8.

Model of the TCR-CD3 complex bound to MHC and CD4. Shown in a ribbon representation is the crystal structure of a TCR-pMHC-CD4 complex (MS2-3C8-MBP-DR4-CD4) oriented as if the TCR and CD4 molecules are attached to the T cell at the bottom and the HLA-DR4 MHC class II molecule is attached to an opposing antigen-presenting cell at the top (Protein Data Bank accession code 3TOE) (32). Green, TCR α chain; wheat, TCR β chain; magenta, MHC α chain; blue, MHC β chain; brown, CD4. Highlighted in red is the docking site for CD3 on the TCR β chain identified by NMR. The ectodomains of CD3 (orange oval) are wedged between the TCR and T cell membrane, opposite CD4. The tyrosine kinase Lck (yellow oval) is bound to the cytoplasmic tail of CD4. ITAMs on the cytoplasmic tails of CD3 subunits are drawn as purple spheres. The intracellular proximity of Lck to CD3 ITAMs promotes ITAM phosphorylation and T cell activation.

Although virtually all regions of the MS2-3C8 β chain that were ordered in the NMR structure (high S² values) were also ordered in the x-ray structure (low B factors) (30), intriguing exceptions are the αA and αB helical regions that constitute the CD3 docking site. These α-helices are ordered in solution but display high B factors in the crystal. Whereas MS2-3C8 was bound to its pMHC ligand (MBP-HLA-DR4) in the x-ray structure, it was unbound in the NMR study. It is therefore tempting to speculate that binding of pMHC to the TCR could increase the flexibility of the CD3 docking site, thereby transmitting an allosteric signal from Vβ to Cβ and its associated CD3 subunits. Indeed, recent studies in several systems have demonstrated that ligand binding can alter protein flexibility at distant sites, resulting in long range transmission of biological signals, even in the absence of crystallographically observed structural changes (47–50). This process, known as dynamic allostery, is of special interest in cases such as the TCR, where x-ray crystallographic studies of multiple TCRs in free form and bound to pMHC have so far failed to identify clear and consistent conformational changes in the TCR Cα or Cβ domains that could be unambiguously attributed to antigen binding (4, 8, 10). Future studies will examine the possibility that MBP-HLA-DR4 ligation by MS2-3C8 induces allosteric changes in TCR dynamics that are transmitted to the CD3 signaling apparatus.

Author Contributions

Y. H., S. R., M. K., M. L., Y. C., Q. W., and Y. Y. expressed recombinant proteins and performed NMR experiments. C. J. W. and K. M. V. carried out mutational analysis. D. A. A. V., R. A. M., and J. O. wrote the manuscript.