|Home | About | Journals | Submit | Contact Us | Français|
Antigen recognition of peptide-major histocompatibility complexes (pMHCs) by T-cells, a key step in initiating adaptive immune responses, is performed by the T-cell receptor (TCR) bound to CD3 heterodimers. However, the biophysical basis of the transmission of TCR-CD3 extracellular interaction into a productive intracellular signaling sequence remains incomplete. Herein, we used nuclear magnetic resonance (NMR) spectroscopy combined with mutational analysis and computational docking to derive a structural model of the extracellular TCR-CD3 assembly. In the inactivated state, CD3γε interacts with the helix-3 and helix 4-F strand regions of the TCR Cβ subunit while CD3δε interacts with the F and C strand regions of TCR Cα subunit in this model, thereby placing the CD3 subunits on opposing sides of the TCR. Together this work identifies the molecular contacts between the TCR and CD3 subunits thereby identifying a physical basis for transmitting an activating signal through the complex.
The T-cell receptor (TCR) performs the dual role of antigen recognition and signal transduction in T-cell immune responses and plays an important role in viral infections, cancer, inflammation and autoimmunity (Germain and Stefanova, 1999). The TCR is a heterodimeric molecule with each subunit possessing a constant domain (Cα or Cβ) (Davis and Bjorkman, 1988; Krogsgaard and Davis, 2005) and a variable domain (Vα or Vβ) that contains the complementarity-determining regions (CDRs) for peptide-major histocompatibility complex (pMHC) binding. The TCR lacks intracellular signaling domains, but interacts with CD3δε, CD3γε and CD3ζζ subunits, each which possess intracellular immunoreceptor tyrosine-based activation motifs (ITAMs) for phosphorylation (Kane et al., 2000; Kuhns and Davis, 2012). The extracellular interactions of the TCR and CD3 subunits and the orientation of the CD3 subunits on the TCR-CD3 complex have been a subject of much dispute over the last two decades. Previous studies have identified CD3-interacting sites on the TCR, although this was achieved by mutagenesis and other indirect methods (Kuhns and Davis, 2012). The FG loop in TCR Cβ, and the DE loop in TCR Cα have been shown to interact with CD3γε and CD3δε, respectively; whereas the CC’ loop in TCR Cβ was found to interact with both CD3γε and CD3δε (Kim et al., 2010; Kuhns and Davis, 2007). A cavity formed between the Cβ FG loop and the Cα CD and EF loops is hypothesized to accommodate a CD3ε domain (Ghendler et al., 1998). Conformational changes have been implicated in the AB loop of the TCR Cα domain (Beddoe et al., 2009) and on the CD3ε subunit (Martinez-Martin et al., 2009) indicating potential interactions. Moreover, crystal structures of CD3γε (Kjer-Nielsen et al., 2004) and CD3δε (Arnett et al., 2004) suggest other putative TCR binding sites. The Cβ FG loop was shown to act as a lever applying force on the CD3ε after pMHC ligation in the ‘mechanosensor’ model (Kim et al., 2009). Recently, an NMR chemical shift perturbation study identified helix 3 and helix 4 regions as interaction sites for both CD3 subunits (He et al., 2015). However, CD3 subunits interaction with TCR α-subunit was not investigated in the said study. Therefore a complete molecular picture of the extracellular TCR-CD3 complex is still not achieved. Understanding these molecular interactions is necessary to fully understand the signal transduction process during T-cell activation and may identify potential therapeutic targets.
We use nuclear magnetic resonance (NMR) to identify the CD3-TCR contact sites by mapping NMR spectral changes within the TCR upon titration of CD3γε or CD3δε. These data, along with functional measurements in T-cell hybridomas, were used as a filter for all energetically reasonable, computationally docked complexes of CD3γε and CD3δε with TCR to identify the most likely binding sites of the extracellular CD3 subunits on the TCR. A single TCR-CD3 interaction fitting a two-sided model (Sun et al., 2004) was significantly more compatible with experimental measurements than all other models. This model is not compatible with the one-sided model previously proposed (He et al., 2015; Kuhns et al., 2010) but is compatible with TCR-CD3δε SAXS study (Birnbaum et al., 2014). Our data shows that CD3γε interacts with residues of helix 3 and helix 4-F strand on the TCR β-subunit, and CD3δε interacts on the opposite side with residues of F and C strands on the TCR α-subunit.
Previous NMR and crystallographic studies of the TCR have been confounded by its large size (extracellular components ~50 kDa), asymmetry and flexibility (Kass et al., 2014; Rudolph et al., 2006). To overcome these problems, a 2B4 TCR construct has been developed to improve expression (with human LC13 constant domains), solubility and folding (Berg et al., 1989), thereby enabling individual labeling of each subunit to separately incorporate (2H-13C-15N) or (2H-15N)-labeled and 2H-perdeuterated forms. Selective labeling enabled separate backbone NMR assignment of each α and β 2B4 monomer subunit within the heterodimer (Figure S1A, S1D). The backbone resonance assignments were obtained using TROSY-based triple resonance experiments (Salzmann et al., 1998) (Figure S2). Using 1H-15N TROSY, 1H-15N chemical shift changes in the TCR were measured using perdeuterated-15N labeled 2B4-α or 2B4-β subunits separately in the presence of its unlabeled-perdeuterated heterodimer counterpart (2B4-β or 2B4-α respectively). By titrating these two proteins with an increasing amount of deuterated CD3 (Figure S1B-C, S1E-F), which is invisible to the NMR, the chemical shift perturbations (CSPs) that identify changes in the local atomic environment and cross-peak intensities that result from dynamic changes can be localized upon titration. We significantly reduced the spectral complexity by using specific subunit labeling and lessened the signal relaxation effects by using perdeuteration, thus leading us to identify specific residues in TCR Cα and Cβ domains as CD3 interaction sites.
To identify TCR residues critical to interaction with CD3 subunits we analyzed the binding interaction of CD3γε to 2B4βLα (where βL indicates only β subunit is labeled) by CSP analysis (Figure 1B). Prominent intensity changes occur within the more flexible N-terminal region at edges of β-sheets (A18, F64, I78 and S86) highlighting concerted structural shifts in the TCR upon CD3γε interaction; in CDR regions: Q55, Q56 (both CDR2) and near CDR regions: A92, S93 (near CDR3). However, more localized changes occur at I134, Q138 (Figure 1C, ,2A)2A) and K139 in the helix 3 region and F199, W200, R204 and N205 (Figure 1D, ,2A)2A) in the helix 4-F strand region of the β subunit. CSP changes were seen for residue E133 within helix 3 and T141 contacting helix 3 (Figure 2B). No significant CSP was observed in the CC’ loop, however, residues V160 and V169 flanking the CC’ loop did exhibit CSP. The FG loop residues F213 and S217 showed intensity changes, and T223 showed CSP upon CD3γε addition. The FG loop had three residues that showed spectral changes, indicating that this loop does not interact with the CD3γε in the inactivated state. This titration identified a contiguous site consisting of helix 3 and helix 4-F strand on the β subunit as the CD3γε interaction site (Figure 3A). The interactions that are observed imply weak μM-mM association based on the concentrations needed to achieve detectible chemical and intensity changes. Recently, He et al. identified helix 3 and helix 4 regions as interaction sites for both CD3 subunits (He et al., 2015). Prior studies have not identified these helix 3 and helix 4-F strand as TCR β-CD3γε interaction sites (Beddoe et al., 2009; Kuhns and Davis, 2007, 2012), however NMR has greater sensitivity for detection of weak binding in comparison to other techniques (Zuiderweg, 2002).
Corresponding analysis of the 2B4αLβ (only α is labeled) in the presence of CD3γε displayed few intensity changes or CSP (Figure 2C, 2D). Several variable region residues showed CSP (V6, E7) and intensity changes (E57, Y71, G85) upon CD3γε titration. These residues were present at the edges of β-strands suggesting concerted structural changes upon CD3γε binding.
Peak intensities were reduced in C-strand residues N146, Q149, D152, S134 and CSP was observed in C136 (Figure S3C). The paucity of interactions on the 2B4 α subunit and numerous contacts of CD3γε on the 2B4 β subunit indicate a strong preference for association with the β-subunit.
Analyzing the interactions between 2B4βLα and CD3δε yielded results similar to 2B4αLβ-CD3γε interaction (Figure 3C). CSP was seen for Q224 in the FG loop (Figure S3F) and F120 (Figure S3G), which is present on the opposite face of the CD3γε interaction site of the TCR (Figure 3D). Sporadic residues belonging to Vβ (R42) and Cβ helix 3, Q138 (Figure S3E), show CSP possibly due to non-specific interactions with the ε-subunit shared by both CD3 subunits. No residues in the CC’-loop, previously implicated in CD3δε binding (Kuhns and Davis, 2007), showed significant spectral changes.
Interactions between 2B4αLβ and CD3δε indicated spectral changes in F strand residues N180 and F184 (both CSP) (Figure S3K) and K181 (intensity change). The C strand adjacent to the F strand had residues Q149 and S150 showing CSP. The C-terminal tail residues D198 and T199 showing intensity changes (Figure S3I) and E205 showing CSP (Figure S3J) also showed spectral changes. All these residues are located on the site opposite the CD3γε interaction site (Figure 3D). However, M165 in the DE loop, previously shown to interact with CD3δε (Kuhns and Davis, 2007), shows CSP; but other residues in the DE loop do not. Chemical shift perturbation analysis with higher concentrations of perdeuterated 15N-labeled 2B4α and unlabeled CD3δε showed increased CSP in Cα F and C strands (Figure S4A, S4B) validating our results shown in Figure 2H, ,3D.3D. Our 2B4αLβ-CD3δε interaction experiments determined that CD3δε interaction regions on the α subunit of the TCR include the F strand, C strand and the C-terminal tail.
Further, we performed reciprocal chemical shift perturbation experiments in which the 15N-labeled CD3 subunits were titrated with increasing amounts of unlabeled TCR (Figure 4). Although the human CD3 subunits are unassigned, this indicates the presence of corresponding chemical shifts and intensity changes upon addition of unlabeled CD3 subunit to labeled 2B4, confirming the interaction between CD3s (CD3γε and CD3δε) and 2B4 TCR.
In summary, we show that CD3γε induced spectral changes in helix 3 and helix 4-F strand residues in the TCR β-subunit. The residues of Cβ CC’ loop, previously reported in CD3γε and CD3δε binding (Kuhns and Davis, 2007, 2012), failed to display significant NMR changes upon CD3γε or CD3δε titration, despite its close proximity to the CD3γε interacting helix 3 and helix 4-F regions. The Cβ FG loop region, shown to act as a lever on the CD3ε subunit (Kim et al., 2009) upon antigen ligation, displayed sparse interactions with CD3γε and CD3δε individually. This suggests that the Cβ FG loop may not be involved in CD3 binding in the inactivated state. Similarly, we showed that CD3δε interacts at the site opposite to the CD3γε interaction site through F strand, C strand and C-terminal tail of TCR α-subunit. This contrasts with a previous study wherein Cα F and C strands were implicated in TCR-TCR interactions (Kuhns et al., 2010) leading to TCR clustering, as well as recent NMR results indicating that both CD3γε and CD3δε associate with the Cβ helix 3 and helix 4 on the TCR (He et al., 2015). The only other reported site for CD3δε interaction, Cα DE loop (Kuhns and Davis, 2007), did not produce significant perturbation in our NMR studies. Overall our NMR results are in general agreement with the TCR-CD3δε SAXS structural study (Birnbaum et al., 2014).
Residues that display peak intensity changes or CSP may not be directly involved in TCR-CD3 interaction, but could undergo allosteric, structural or conformational change upon CD3 interaction (Chen et al., 2014; Zuiderweg, 2002). We identified regions on the TCR variable domains that show CSP and intensity changes with CD3 addition, but these are clearly not the sites for CD3 interactions since they are involved in pMHC interactions (Rudolph et al., 2006). Based on our NMR results, we devised site-directed mutation and functional studies in T-cell hybridoma 58−/− (Letourneur and Malissen, 1989; Zhong et al., 2010) to determine if NMR-identified residues are important for T-cell activation. The functional assays used a mouse TCR-CD3 system as most of the mutated residues are conserved between human and mouse TCR constant domains (Figure 5A). Residues that undergo intensity changes/CSP upon addition of CD3, or are nearby, were selectively mutated to alanine (Figure 5A). Double alanine mutants were made to decrease the effects of these residues in T-cell activation without blocking TCR expression (Kuhns and Davis, 2007). TCR genes were cloned into a retroviral vector to express mutated mouse 2B4 TCR variants. Specific sites targeted included the Cβ helix 3, Cβ helix 4-F strand, Cβ CC’ loop, Cβ FG loop, Cα F strand and Cα C-terminal tail (Figure 5B). T-cells expressing wild type 2B4 TCR and mutated TCRs were activated by Chinese Hamster Ovary (CHO) cells expressing pMHC (I-Ek/K5) (Krogsgaard et al., 2005), and IL-2 production was quantified by ELISA (Malecek et al., 2013) (Figure S5). The area under the curve for IL-2 production, an overall measure for cumulative response (Hioe et al., 2010; Malecek et al., 2014), was determined and compared to the wild type to identify the change in IL-2 production.
Cβ helix 3 mutants, E133A/I134A, N136A, N136A/K137A and Q138A/K139A, showed decreased IL-2 production compared to wild type (Figure 5C). E133A/I134A produced the least amount of IL-2 (89% decrease) among the helix 3 mutants. These two residues are highly conserved across species, and showed the most CSP (E133) and intensity change (I134) among helix 3 residues upon CD3γε titration. N136A produced more IL-2 in comparison (25% decrease) with the other helix 3 mutants. N136 is not conserved among various species and is replaced by both basic and acidic residues in rabbit and rat, respectively (Figure 5A). N136A/K137A (55% decrease) and Q138A/K139A (57% decrease) produced less than half the IL-2 as the wild type. Q138 and K139 are conserved between various species, yet Q138A/K139A did not have as dramatic effect as E133A/I134A.
To identify if any part of the CC’ loop is involved in CD3 interaction (Kuhns and Davis, 2007), we tested the Cβ CC’ loop mutants, G162A/K163A, K163A/E164A and H166A/S167A, even though we could not assign H166 and S167 in the 2B4β spectra. G162A/K163A (14% decrease) and K163A/E164A (34% decrease) produced reasonable amounts of IL-2 compared with the wild type. H166A/S167A produced significantly less IL-2 (79% decrease) compared with the wild type, suggesting its role in T-cell activation.
The Cβ helix 4-F strand mutants, H201A/N202A, R204A/N205A and N205A/H206A, produced significantly less IL-2 (77.6%, 81.0% and 86.5% decrease, respectively, Figure 5C). However, other Cβ helix 4-F strand mutants N202A/P203A and P203A/R204A produced IL-2 amounts comparable (32.2% and 10.3% increase) to the wild type, suggesting these residues may not have a direct role in T-cell activation, even though R204 showed intensity changes upon CD3γε addition.
The Cβ FG loop mutant, E224A/S226A, produced less IL-2 (34% decrease) compared with the wild type. The FG loop is about 20 residues long (Das et al., 2015; Newell et al., 2011) and the effect of a small number of amino acid changes may not be drastic.
The Cα F strand mutants, N180A/Q181A and F184A/T185A, and Cα C-terminal tail mutant, N198A produced similar amounts of IL-2 compared with the wild type (4.9% increase, 3.7% increase and 4.5% decrease, respectively), suggesting a weaker influence on T-cell activation. The other Cα tail mutant, Y201A/P202A, produced slightly less IL-2 (24.2% decrease) compared with the wild type.
To summarize, we identified E133, I134, K137, Q138, K139 of Cβ helix 3; H201, N205, H206 of Cβ helix 4-F strand; and H166, S167 of Cβ CC’ loop as residues critical for T-cell activation and possibly involved in direct CD3 interactions (residues showing greater than 50% loss in IL-2 production). Due to lack of many spectral changes, only a single Cβ FG loop mutant, E224A/S226A, was analyzed, which showed a 34% decrease in IL-2 production compared with wild type. Previous studies have noted significant loss in T-cell activity upon complete FG loop deletion (Touma et al., 2006). Combined, this suggests that there are CD3 subunit reorientations involved upon antigen presentation that result in FG loop contact with CD3. Similarly, parts of the Cβ CC’ loop (H166 and S167) could also come into contact with both CD3 subunits upon activation, suggesting reorientation. Likewise, NMR spectral changes in N180, K181 and F184 with CD3δε addition and the lack of significant IL-2 reduction upon activation might suggest CD3δε reorientation upon activation.
To study the importance of TCR residues identified by NMR for CD3 interaction directly, we used three 15N-labeled, partially deuterated TCRβ mutants: E133A/I134A (helix 3), Q138A/K139A (helix 3) and Q201A/N202A (helix 4-F strand). Based on our CSP results, we reasoned that mutations in helix 3 and helix 4- F strand region will impact CD3γε binding as this region had many residues that showed spectral changes upon CD3γε titration (Figure 3A). E133A/I134A did not produce a sufficient 15N-TROSY spectra, indicating a protein bigger than 50 kDa. Light scattering experiments revealed that it oligomerized to a size over 1000 kDa (Figure S6A). Q138A/K139A and Q201A/N202A produced 15N-TROSY spectra comparable in signal quality and peak dispersion to the wild type TCRβ (Figure S6B, S6C). We unambiguously assigned 82% and 65% of Q138A/K139A and Q201A/N202A, respectively, based on the wild type TCRβLα 15N-TROSY spectra. The change in 15N-TROSY spectra upon mutation in critical α-helical regions is not entirely uncommon, as previous instances have shown significantly altered spectra due to higher order oligomerization, transient interactions or unfolding (Park et al., 2002). Light scattering experiments on Q138A/K139A indicated a molecular weight of 60 kDa (Figure S6A). The unassigned regions include helix 3, helix 4-F strand and C-terminal regions for Q201A/N202A and helix 3 region for Q138A/K139A. We could not infer a lack of interaction for Q201A/N202A through CSP analysis by adding unlabeled CD3γε, as helix 3 and helix 4-F strand (CD3γε interacting regions) were unassigned. Q138A/K139A did show reduced spectral changes for R204 and N205 of helix 4-F strand region (Figure 6B), thereby highlighting the importance of Q138 and K139 in helix 3 region for CD3γε interaction. However, some residues on G-strand near helix 4-F strand residues (S235 and W239) along with Q201 (helix 4-F strand) and N161 (CC’ loop) were perturbed with CD3γε addition (Figure 6C), thereby losing the specificity for a contiguous region formed by helix 3 and helix 4-F strand regions in the wild type TCR β-subunit (Figure 3A).
In conclusion, E133A/I134A produced 15N-TROSY spectra evident of higher order oligomer, which was confirmed with light scattering data. 65% of Q201A/N202A was assigned from the wild type 2B4β spectra and we were not able to confirm lack of CD3γε interaction, as helix 3 and helix 4- F strand regions were not assigned. However, Q138A/K139A showed reduced intensity changes for some helix 4- F strand residues upon CD3γε addition, confirming the significance of Q138 and K139 in the Cβ helix 3 region for CD3γε interaction.
To develop a 3D model of TCR-CD3 complex based on NMR data, we used computational docking (Fernandez-Recio et al., 2003) [ICM] to generate all possible energetically favorable conformations of CD3γε and CD3δε structural models docked to a model of the TCR crystal structure. CD3γε and CD3δε subunits were each docked to the TCR, resulting in approximately 45000 TCR-CD3γε and 39000 TCR-CD3δε conformations. We re-ranked these conformations based on the contact surface area between the molecular surfaces of CD3 subunits and the aggregate molecular surface of the set of NMR-identified TCR residues on the TCR structure.
CD3-TCR conformers that buried significantly more of the molecular surface formed by NMR-identified residues were selected and re-ranked using molecular mechanics energy score, which includes van der Waals, solvation electrostatics and entropic terms. Two CD3γε conformers exhibited a significantly better composite contact surface area and energy score, but were minor deviations of each other, likely representing a single strongly predicted conformation. Two other variations of a single conformation had high contact area and significantly worse, but still favorable, energy scores (Table S1). These two likely represent a single alternative conformation to the most strongly predicted conformation. Both the primary and alternative conformations locate the CD3γε at the same specific surface of the TCR. Thus, the NMR data strongly predicts a single CD3γε interaction site on the TCR, with two alternative orientations of CD3γε. Among the CD3δε conformers, there was a single top-scoring conformation and interaction site, in terms of energy score and contact area. The CD3γε and CD3δε orientations do not clash with each other or with the binding sites of pMHC and CD4-pMHC (Yin et al., 2012), even though this information was not used to constrain the docking or the NMR experiment. Thus, integration of the NMR and docking data predicts the arrangement of the TCR-CD3γε-CD3δε assembly with high confidence.
The CD3γε conformers contact the Cβ helix 3 and Cβ helix 4-F strand regions, with some conformers exhibiting contacts with residues of the Cα DE loop and Cβ CC’ loop. The CD3δε conformer exhibited contacts with the Cα AB loop, Cα C strand, Cα F strand and Cβ FG loop. A representation of the TCR-CD3 complex, using the top-scoring CD3γε conformation, is shown in Figure 7A. The CD3δε conformer has a total contact area of 187.2 Å2 with the NMR-identified TCR residues and energy of −24.5 eu (equivalent to kcal/mol). The CD3γε conformer has a total contact area of 225 Å2 with the NMR-identified residues and energy of −23.9 eu. A movie depicting all the high-scoring conformers for CD3γε with TCR-CD3δε and PDB files are provided (Movie S1/ https://nyumc.box.com/s/o5u5jiojvwdbt7myzkmuwz8no5dcoa3z). The top view of the complex is shown in Figure 7B, which indicates that the CD3γε and CD3δε subunits interact on opposite sides of the TCR (Figure 7C). This contrasts the one-sided model (He et al., 2015; Kuhns et al., 2010), but in agreement with TCR-CD3δε SAXS data (Birnbaum et al., 2014).
By analyzing the individual residue contributions to the overall contact area in a particular conformation, we found that Q138 (Cβ helix 3), T198 and R204 (both Cβ helix 4-F strand) had significant contacts (> 10% contribution) with CD3γε in at least three of the four different predicted conformations. Notably, Q138A/K139A, E133A/I134A (nearer to Q138) and H201A/N202A (nearer to T198) produced significantly less IL-2, suggesting that these residues are important to T-cell signaling. In our model, Q138 and T198 contact Y80 (in CD3γ) and Y124 (in CD3ε), respectively (Figure 7D). Similarly, in the CD3δε conformer, D132 in Cα AB loop, and N180 in Cα F strand, make significant contributions to the overall contact area. D132 contacts S133, and N180 contacts T112 of CD3ε in the conformer (Figure 7E).
In summary, top-scoring conformations from docking CD3 to TCR were ranked by (i) high total contact area between the NMR-perturbed amino acids and the docked CD3 conformations and (ii) molecular energetics, revealing a strongly preferred extracellular TCR-CD3 assembly. In the absence of a large amount of induced fit upon CD3-TCR association, which cannot be assessed by the method we used, these data clearly suggest that CD3γε associates with an interaction surface that contains the Cβ helix 3 and helix 4-F strand regions, and CD3δε associates with the Cα F strand and C strand regions. This yields a two-sided model for the TCR-CD3 interaction wherein CD3γε and CD3δε interact at opposing sides of the TCR (Figure 7A, 7B). The Cβ FG loop did not provide significant contacts with the CD3γε in our docking studies, although the CD3δε conformation showed interactions with the FG loop. The Cα AB loop, which undergoes conformational change upon agonist ligation (Beddoe et al., 2009), interacts with CD3δε in our model. The Cβ CC’ loop did not produce sufficient spectral changes in our NMR experiments and, therefore, was not used as a guide in our docking analysis.
In the present study, NMR spectroscopy, T-cell hybridoma functional assays, and computational docking were used to visualize the TCR-CD3 complex. The size and spectral complexity problems of a full-length TCR were overcome by using perdeuteration and partial subunit labeling. While single-chain TCR has been used previously to look at CDR mobility (Varani et al., 2007), this study is the first in which NMR backbone assignments have been obtained on both subunits of a full-length TCR. While one limitation is that the perturbations identified by NMR may represent an aggregate imprint of a dynamic ensemble of CD3-TCR interactions, the consistency between docking and experimental observations converging to one interaction site each for CD3γε and CD3δε with TCR argues against this scenario. This is further supported by the complementarity, rather than clash, of the independently docked CD3γε and CD3δε sites on opposing sides of the TCR.
The important question of how the TCR is triggered has been obscured by the lack of atomic-detail structural information. Understanding how TCR-ligand engagement results in phosphorylation of the CD3 cytoplasmic tail requires knowledge of how the TCR, which lacks signaling capacity, transmits the information received via its variable region through interaction of its constant region with the CD3 ectodomain.
We have determined that CD3γε interacts with residues of the helix 3 and helix 4-F strand regions on the TCR β-subunit with some conformers in contact with the Cα DE loop and Cβ CC’ loop as well (Kuhns and Davis, 2007), and that CD3δε interacts with residues of the F and C strands on the TCR α-subunit (Figure 7A). Previous work has identified several Cβ helix 4-F strand mutants (F199, N202 and R204) showed reduced TCR surface expression, however, T-cell activation was not assessed with those mutants (Fernandes et al., 2012). Recently, NMR chemical shift perturbation analysis (He et al., 2015) identified Cβ helix 3 and helix 4 residues as the binding site for both CD3 subunits, which is incompatible with the two-sided model. We have shown through reciprocal chemical shift analysis that CD3γε and CD3δε can bind the TCR individually without the requirement of the other CD3 subunit (Figure 4), arguing against the one-sided model (He et al., 2015). That each CD3 subunit can bind independently was recently quantified with 2D binding analysis (Huang et al., 2010) using immobilized TCR and CD3 extracellular subunits (Ge, Natarajan, Krogsgaard, Zhu; Direct measurement of extracellular interactions within the TCR complex) (in submission).
Identification of specific interacting residues may provide insight into the nature of the TCR and CD3 interaction. Notably, contacts between TCR and CD3 were observed that are consistent with hydrogen bonds. These CD3 interacting regions are on opposite sides of the TCR, as predicted previously (Sun et al., 2004) in the resting condition, contradicting the one-sided model (He et al., 2015; Kuhns et al., 2010). However, this two-sided model cannot rule out CD3 subunit reorientations upon antigen ligation to expose intracellular ITAMs for phosphorylation. The TCR-CD3δε SAXS and pMHC-TCR-CD3 EM structure also points to the two-sided model. (Birnbaum et al., 2014). Our model bears similarity to the proposed CD8- CD3δ-α-connecting peptide motif interactions needed for proper pMHC binding (Werlen et al., 2003). The current NMR-derived structure is physiologically relevant to the resting state interactions between TCR-CD3γε and TCR-CD3δε. This may be crucial for understanding the conformational changes or subunit movement that may occur upon antigen ligation.
Garcia et al. suggested that CD3γε and CD3δε interact “below” TCRβ and TCRα, respectively, near the plasma membrane, but residue specific detail is unattainable in EM studies due to low-resolution (Birnbaum et al., 2014). Our work predicts that the CD3δε subunit interacts adjacent to (on the “side” of) the constant domain of TCR α-subunit rather than “below” it. Conformational changes in the TCR have been implicated in playing a role in differential T-cell activation (varying activation potential for different antigenic peptides) (Krogsgaard et al., 2003). This involves residues in different regions of the TCR interacting with the CD3 subunits based on the type/strength of antigen. Thus, it is important to know the CD3 interacting residues on the TCR under resting conditions.
The Cα regions, namely F strand, C strand and AB loop, implicated in TCR dimerization (Kuhns et al., 2010) are involved in CD3δε binding based on our data. One possible explanation could be that antigen ligation induces a torque or conformational change on the TCR (Krogsgaard et al., 2003) that frees these regions and exposes them for dimerization, leading to TCR clustering. More recently, mechanical forces (Das et al., 2015; Kim et al., 2009; Liu et al., 2014; Ma and Finkel, 2010; Pryshchep et al., 2014) have been shown to induce such changes and trigger T-cell activation. Conformational changes upon antigen ligation could possibly lead to CD3γε interacting with Cβ FG loop, thereby stabilizing the FG loop and influencing the TCR-pMHC bond lifetime (Das et al., 2015). The permissive-geometry (Adams et al., 2011) and lipid-based (Shi et al., 2013) models wherein the CD3 ITAMs are resistant to phosphorylation in the resting condition still hold in this two-sided model. The two-sided binding model does not necessarily rule out pre-clustering of TCR on the T-cell membrane (Alarcon et al., 2006; Lillemeier et al., 2010; Roh et al., 2015) and has implications for TCR clustering, seen as a mechanism for signal amplification upon antigen binding. According to the model, clustering might occur through the interactions between CD3 subunits of neighboring TCR-CD3 complexes or might require CD3 subunit rearrangement upon antigen binding to promote TCR-TCR interaction. Hypothetically, CD3 subunit rearrangement could bring the intracellular CD3 ITAMs into close vicinity with each other, thereby promoting trans-autophosphorylation (Kuhns et al., 2010). A similar conformational change at the CD3ζζ juxtamembrane region has been recently shown (Lee et al., 2015). Moreover, the pMHC-TCR-CD3 EM structure has suggested TCR-CD3 subunit rearrangement upon antigen ligation (Birnbaum et al., 2014) for signal propagation.
Our results answer a long-standing question regarding the orientation of CD3 subunits in the TCR-CD3 complex during resting conditions. The study suggests an answer to one of the fundamental questions in immunology and receptor recognition: how ligand engagement of TCR mediates critical information transfer from the antigen recognition site in the external environment to the intracellular compartment via the CD3 signaling complex. This may also allow us to understand other receptor-mediated cell activation systems. An understanding of the molecular organization of the TCR-CD3 signaling complex has wide significance for manipulating immune functions to cancer, viral infections, inflammation and autoimmunity, including antibodies (Martin et al., 2013), modified T-cells (Rosenberg and Restifo, 2015), and small molecules that either dampen or enhance T-cell signaling. This work also illustrates the capabilities of NMR spectroscopy in the study of other immune and protein complexes. Our subunit-specific labeling scheme could be used to study a variety of large multimeric proteins such as γδTCR, NKT TCR, B-cell receptors, various Ig molecules, reverse transcriptases and kinesins.
Soluble full-length 2B4 TCR (containing constant domains from human LC13 TCR) was generated by E. coli inclusion body expression, protein refolding and purification as previously described (Newell et al., 2011). Only one of the subunits (either α or β) is isotopically labeled (2H/13C/15N) and linked through disulfide linkage to the other deuterated (2H/12C/14N) subunit (either β or α). The labeled α linked to unlabeled β is hereby referred as 2B4αLβ sample and the labeled β linked to unlabeled α is hereby referred as 2B4βLα sample. Perdeuterated (> 95% deuteration) subunits were made by culturing BL21(DE3) E. coli cells expressing 2B4 subunits step-wise in media containing increasing proportions of D2O (25%, 50%, 75%) and finally into 100% D2O media containing 15NH4Cl + 13C-2H-glucose or 14NH4Cl + 12C-2H-glucose (Gardner and Kay, 1998). The refolded protein was concentrated and purified by anion-exchange chromatography (DEAE column) and gel filtration (S200) (Figure S1D). The protein was exchanged to the buffer containing 200 mM MES, 100 mM arginine, pH 7.0 for NMR experiments.
TROSY based triple resonance experiments (Salzmann et al., 1998) were performed on the perdeuterated 300 μM sample of 2B4βLα at 37 °C on 600 MHz and 800 MHz spectrometers to obtain the backbone N, HN, Cα and C’ assignments. The data were processed by NMRPipe (Delaglio et al., 1995) and analyzed by NMRView (Johnson, 2004). About 75% of the assignable backbone N and HN were assigned. 65% of these assignments were transferred to the 22 °C 15N-TROSY spectra without ambiguity (80% of the constant domain was assigned) (Figure S2B). This transfer was needed as the CD3 subunits were not stable at 37 °C. Similar NMR experiments on perdeuterated 200 μM sample of 2B4αLβ on 600 MHz spectrometer yielded about 60% of the assignable backbone N and HN assignments (60% of the constant domain were assigned) (Figure S2A). 15N-TROSY spectra of 2B4αLβ at concentrations higher than 200 μM indicated aggregation that involved the α-subunit. This restricted us from using higher concentrations of 2B4αLβ to get better quality triple resonance experimental data.
To study the 2B4 TCR interaction with CD3 subunits by NMR, the constructs for ectodomain human CD3γε and CD3δε subunits (Kjer-Nielsen et al., 2003; Kjer-Nielsen et al., 2004), wherein the two subunits are linked by a 26-residue linker, were used. E. coli BL21(DE3) expressing CD3 proteins were grown at 37 °C in LB media and transferred into 250 ml minimal media containing 12C-glucose, 14NH4Cl and 75% D2O (Marley et al., 2001) before induction and cultured. CD3 inclusion bodies were refolded and purified in a previously described manner (Kjer-Nielsen et al., 2004) (Figure S1B-F). After purification, the CD3 protein samples were exchanged using Amicon concentrator to the same NMR buffer used for TCR sample.
Unlabeled deuterated CD3γε titrations with perdeuterated 15N-labeled 2B4βLα were done at 22 °C on a 800 MHz spectrometer equipped with 1.7 mm cryoprobe. 15N-TROSY was collected on samples with the following 2B4βLα: CD3γε concentration ratios (μM): 100:0, 100:200, 100:500 and 100:900 and adjusted to a final volume of 30 μL (Figure 1B-D). All other CSP experiments were performed at 22 °C on spectrometers equipped with 5 mm cryoprobe. Unlabeled deuterated CD3γε titrations with perdeuterated 15N-labeled 2B4αLβ were done on a 600 MHz spectrometer. 15N-TROSY was collected on samples with the following 2B4αLβ: CD3γε concentrations ratios (μM): 38:0 (1:0), 35.4:35.4 (1:1), 33:66 (1:2) and 29:116.5 (1:4) (Figure S3A). Unlabeled deuterated CD3δε titrations with perdeuterated 15N-labeled 2B4βLα were done on a 500 MHz spectrometer.15N-TROSY was collected on samples with the following 2B4βLα: CD3δε concentration ratios (μM): 58.5:0 (1:0), 46.7:93.4 (1:2), 42.4:127.3 (1:3) and 38.9:155.5 (1:4) (Figure S3D). Unlabeled deuterated CD3δε titrations with perdeuterated 15N-labeled 2B4αLβ were done on a 600 MHz spectrometer. 15N-TROSY was collected on samples with the following 2B4αLβ: CD3δε concentration ratios (μM): 37:0 (1:0), 33.5:33.5 (1:1), 30.6:61.2 (1:2) and 25:112.6 (1:4.5) (Figure S3H). Unlabeled CD3δε was added to perdeuterated 15N-labeled 2B4αLβ to achieve concentration (μM) of 2B4αLβ: CD3δε: 80:290 (1:3.6) for the higher concentration CSP experiment. In each titration case, assignments were carefully transferred from the CD3 free TROSY spectra to saturated TROSY spectra. The scaled chemical shift, δppm = ((Δ1H)2+0.11(Δ15N)2)0.5 (Natarajan et al., 2012). The residues that showed spectral changes were identified if (a) they showed scaled chemical shift changes greater than 1σ above the 25% trimmed mean deviation when comparing in the presence and absence of saturated amounts of CD3 state, and (b) lesser than 1σ below the 25% trimmed mean deviation in the ratio of NMR signal intensity between CD3 saturated-state and CD3 free state (Figure 2).
Deuterated 15N-labeled CD3γε was produced using LB media prior induction and 15NH4Cl in M9 minimal D2O media after induction. Protonated 15N-labeled CD3δε was produced using 15NH4Cl in M9 minimal media. Unlabeled protonated 2B4 TCR were produced using LB media. Unlabeled 2B4 titrations with deuterated 15N-labeled CD3γε in NMR buffer were done on a 600 MHz spectrometer equipped with 5 mm cryoprobe at 22 °C.15N-TROSY experiments were collected at each titration point with 256 incremental points in the indirect dimension with 128 scans. 15N-TROSY was collected on samples with the following CD3γε:2B4 concentrations (μM): 40:0 (1:0), 35:71 (1:2) and 34:93 (1:2.7) (Figure 4A). Unlabeled 2B4 titrations with 15N-labeled CD3δε in PBS were done on a 600 MHz spectrometer equipped with 5 mm cryoprobe at 22 °C.15N-TROSY was collected on samples with the following CD3δε:2B4 concentration ratios (μM): 120:0 (1:0), 110:165 (1:1.5) and 105.5: 263.8 (1:2.5) (Figure 4B).
Mouse 58−/− T-cell hybridoma cells (Letourneur and Malissen, 1989), which expresses mouse CD3 but not TCRαβ, (from David Kranz, University of Illinois) and CHO cells expressing I-Ek (Krogsgaard et al., 2005) (from Mark M. Davis, Stanford University) were cultured in RPMI 1640 and DMEM media respectively, supplemented with 10% FBS, sodium pyruvate, non-essential amino acids, glutaMAX-1 and penicillin-streptomycin. The mutant mouse 2B4 TCR construct were generated by PCR using overlapping primers containing the mutant sequences and cloned into retroviral pMX vector (Kitamura et al., 2003). Retroviral transductions of the hybridoma cells were done as previously described (Malecek et al., 2014; Malecek et al., 2013; Zhong et al., 2013; Zhong et al., 2010). The transduced cells were stained with PE anti-CD3ε and APC anti-TCRβ (clone H57-597) antibodies. The cells were then sorted using a SONY SY3200 sorter. The sorted cells were expanded for 6 days, quantified for TCR expression (Figure S5G-J) and prepared for cytokine assay. 105 CHO-IEk cells were loaded with different concentrations of K5 peptide and incubated with 105 T-cell hybridoma clones for 16 h at 37 °C, 5% CO2. A standard ELISA sandwich was used to quantify the cytokine IL-2 production (Malecek et al., 2013) (Figure S5A-F). The area under the curve for the wild type and mutant IL-2 production was calculated after non-linear fitting using Prism (Graphpad software).
The mutant 2B4β TCR (E133A/I134A, Q138A/K139A and Q201A/N202A) construct were generated using the QuikChange mutagenesis kit (Agilent Technologies) using overlapping primers containing the mutant sequences. E. coli BL21(DE3) expressing 2B4 proteins (α wild type, β mutants) were grown at 37 °C in LB media and transferred into 250 ml minimal media containing 15NH4Cl (14NH4Cl for α-subunit) and 75% D2O (Marley et al., 2001) before induction. Unlabeled CD3γε were expressed in LB media. The proteins were refolded and purified as described above. Unlabeled protonated CD3γε additions with partially deuterated 15N-labeled 2B4βLα mutants in NMR buffer were done on a 600 MHz spectrometer equipped with 5 mm cryoprobe at 22 °C. experiments were collected before and after CD3γε with 256 incremental points in the indirect dimension. 15N-TROSY was collected on samples with the following 2B4: CD3γε concentration ratios (μM): a) Q201A/N202A: 60:0 (1:0) and 50:200 (1:4); and b) Q138A/K139A: 40:0 (1:0) and 30:120 (1:4) (Figure S6B, S6C). MALS experiments on the 2B4 wild type and mutants were performed on a Wyatt miniDAWN TREOS detector.
All docking, energy calculations, and contact area calculations were done with ICM (Molsoft, Inc.). A molecular model of the 2B4 TCR αβ dimer was made based on the crystal structure 3QJF (PDB ID); CD3εγ was made with 1SY6; CD3δε was made with 1XIW. The ICM protein-protein docking algorithm was used as previously described (Fernandez-Recio et al., 2003) to dock the TCR to CD3εγ and CD3δε.
The contact surface area between each docked conformation of CD3εγ or CD3δε, respectively, and the TCR was calculated with the calcResContactArea macro in ICM. For TCR-CD3εγ, the contact area contributed by the following NMR-derived residues of the TCR β chain was summed: V11, A18, E41, Q55, Q56, F64, S65, I78, S86, C91, A92, S93, E133, I134, Q138, K139, T141, D152, S157, V169, Q179, S190, S196, F199, W200, R204, N205, F213, S217, T223, Q232. Similarly, for each TCR-CDδε conformation, the contact area contributed by the following NMR-derived residues of the TCR α chain was summed: Q5, S11, T19, S21, C25, Q38, R42, L48, L51, K68, A91, N99, G104, T107, S134, Q149, S150, M165, N180, K181, F184, D198, T199, E205. The residue numbers are all based on the NMR constructs used in this study.
The distribution of contact surface areas of the docked conformations was observed to be a Boltzmann distribution, and was modeled as such. Outliers with significantly higher contact surface areas between CD3εγ or CD3δε, respectively, and the TCR were kept and re-ranked by energy score. Conformations with similar energy scores and RMSD<0.001 Å2 were quantitatively considered a single conformation. Conformations targeting the same specific interaction site on the TCR may be qualitatively considered to be equivalent, even if they exhibit different rotational orientations, given the margin of error of the procedure (e.g. static structure in a crystal lattice used for docking, dynamic NMR measurements in solution).
We thank the staff at New York Structural Biology Center (NYSBC) for the assistance provided while accessing the NMR spectrometers. We also thank Evan Newell (Singapore Immunology Network) and Mark M. Davis (Stanford University) for protocols and advice on soluble TCR expression and purification. We thank James McCluskey and Lars Kjer-Nielson (University of Melbourne) for providing us with human CD3γε and CD3δε gene constructs. We also thank David Kranz (University of Illinois) for providing us 58 −/− T-cell hybridoma. We thank Duane Moogk, Karolina Malecek, Stevan Hubbard, Cheng Zhu and Baoyu Liu (Georgia Institute of Technology) and Ranajeet Ghose (City College of New York) for helpful discussions and critical reading of the manuscript. M.K. was a Pew Scholar in the Biomedical Sciences supported by the Pew Trust. This work was supported by the NIH grant NIGMS 5R01GM085586 (to M.K.), 4DP2OD004631 (to T.C.), S10-RR023694-01EWOF (to C.B.), 1S10-OD016343 (to M.K.) and Perlmutter Cancer Center at NYU Langone Medical Center. Use of flow cytometry was sponsored from NYUCI Center Support Grant, NIH/NCI 5 P30CA16087-31. Facilities at the NYSBC are supported by National Institutes of Health Grants CO6RR015495 and P41GM066354. NMR spectrometers at NYU chemistry shared instrumentation facility is supported by NYU and NIH 1S10-OD016343.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conceptualization, A.N., T.C., C.B., and M.K; Methodology, A.N., T.C., C.B., and M.K.; Investigation, A.N., V.N., W.W., V.R.J., R.W., and M.K; Software, K.V.; Writing – Original Draft, A.N.; Writing – Review & Editing, A.N., K.V., T.C., C.B., and M.K.; Funding Acquisition, T.C., C.B., and M.K.; Resources, T.C., C.B. and M.K.; Supervision, T.C., C.B., and M.K.
There is no conflict of interest among the authors.