Structural Conservation and Effects of Alterations in T Cell Receptor Transmembrane Interfaces

Abstract

T cell receptors (TCRs) are octameric assemblies of type-I membrane proteins in which a receptor heterodimer (αβ, δγ, or pre-Tαβ) is associated with three dimeric signaling modules (CD3δε, CD3γε, and ζζ) at the T cell or pre-T cell surface. In the human αβTCR, the α and β transmembrane (TM) domains form a specific structure that acts as a hub for assembly with the signaling modules inside the lipid bilayer. Conservation of key polar contacts across the C-terminal half of this TM interface suggests that the structure is a common feature of all TCR types. In this study, using molecular dynamics simulations in explicit lipid bilayers, we show that human δγ and pre-Tαβ TM domains also adopt stable αβ-like interfaces, yet each displays unique features that modulate the stability of the interaction and are related to sequences that are conserved within TCR types, but are distinct from the αβ sequences. We also performed simulations probing effects of previously reported mutations in the human αβ TM interface, and observed that the most disruptive mutations caused substantial departures from the wild-type TM structure and increased dynamics. These simulations show a strong correlation between structural instability, increased conformational variation, and the severity of structural defects in whole-TCR complexes measured in our previous biochemical assays. These results thus support the view that the stability of the core TM structure is a key determinant of TCR structural integrity and suggest that the interface has been evolutionarily optimized for different forms of TCRs.

Main Text

T cell receptors (TCRs) are complex structures that play a central role in adaptive immunity by linking antigen specificity to the key activating signals that mobilize T lymphocyte responses. Two major subsets of T cells found in most vertebrates express TCRs containing either an αβ or a δγ heterodimer as the variable antigen-recognition module. These are structurally similar, but recognize different types of antigens (1, 2). A third type of TCR, known as the pre-TCR, is transiently expressed early in αβ T cell development (3). Pre-TCR contains a nonvariable α chain (pre-Tα) paired with newly synthesized β chains and does not require any ligand to generate a signal (4, 5). All three forms of the TCR contain conserved basic residues (Lys or Arg) in their transmembrane (TM) domains that direct assembly with three dimeric signaling modules (CD3δε, CD3γε, and ζζ) containing complementary acidic TM residues (Asp or Glu) (6) (Fig. 1). These octameric complexes are assembled in the endoplasmic reticulum (ER) before export to the cell surface (7), where tyrosine-containing sequences in CD3 and ζ cytoplasmic tails couple ligand binding to phosphorylation by the Src-family kinase Lck to initiate the T cell activation program. Although the mechanism governing conversion of ligand binding into signal initiation is not completely understood (8), there is evidence that alterations in TM interactions may be involved in signal transmission (9, 10, 11), making the precise structural arrangement within the membrane a point of significant interest.

Figure 1.

(A and B) Organization of the αβ TCR complex in the membrane. (A) Three conserved basic residues (Arg and Lys; cyan circles) in TCRα and TCRβ TM domains guide assembly with CD3δε, CD3γε, and ζζ signaling dimers containing complementary acidic TM residues (Asp and Glu; red circles) (6). (B) View down the TM axis showing how signaling dimers are arranged around a specific αβ TM structure (12). For clarity, only conserved basic residues (αR27, K32, and βK25) and polar αβ interface residues (αN37, T41, and βY29) are shown. Predicted H-bonds are indicated as dashed lines. To see this figure in color, go online.

In a previous study of the human αβTCR (12), we reported that TCRα and TCRβ TM domains form a coiled-coil structure featuring an interhelical H-bond network formed by evolutionarily conserved polar residues in the C-terminal regions of their TM helices (Asn and Thr in TCRα and Tyr in TCRβ). Importantly, these key residues are conserved across all TCR types (with the exception of an Asn > Asp substitution in pre-Tα), leading us to hypothesize that this interface is a common structural feature in all TCRs. Indeed, disruption of the TM interface through mutagenesis caused defects in the formation of stable TCRαβ-CD3δε-CD3γε-ζζ complexes (12), suggesting that a compact and precisely organized core TM interface is critical for the structural integrity of the fully assembled receptors.

In this study, we computationally tested these hypotheses by evaluating the ability of human δγ and pre-Tαβ TM domains to adopt stable αβ-like interfaces, and probing the effects of mutations in the human αβ TM interface using all-atom molecular dynamics (MD) simulations in explicit bilayers. Starting with the TM sequences of human TCR α, β, δ, γ, and pre-Tα proteins (Fig. 2 A), initial structures of δγ and pre-Tαβ were prepared by substituting the corresponding residues in our previously reported αβ TM model (12). Mutations in TCRαβ (Fig. S1) were prepared by the same procedure. For all systems, the conserved charged Arg and Lys residues, which are normally neutralized by assembly with acidic residues in signaling modules, were mutated to Leu to allow stable membrane integration of isolated TCR TM domains. The initial TM models were then inserted into explicit lipid bilayers using Membrane Builder (13, 14) in CHARMM-GUI (15) and 200-ns MD simulations were carried out using NAMD (16, 17) to track interface stability. Five independent simulations were performed for αβ, δγ, and pre-Tαβ TM models, and three simulations were performed for each mutant αβ TM model. All computational methods and analyses are described in detail in the Supporting Material.

Figure 2.

TCR heterodimers adopt compatible interfaces among αβ, δγ, and pre-Tαβ TM assemblies. (A) Sequences of TCR TMs used in the study. Charged residues (Arg and Lys, shown in bold italic) for the interface to signaling molecules are mutated to Leu in simulations and the conserved residues in the C-terminal side (Asn/Asp, Thr, and Tyr) are shown in bold. (B) Representative TCR TM structure models from MD simulations, and interhelical H-bonds are shown as red dotted lines. The structure of pre-Tαβ is that with protonated Asp in pre-Tα. The positions of Arg and Lys residues (mutated to Leu) are marked by cyan spheres. (C) RMSD between TCR TM models and the previously reported αβ TM model (12) over the 200-ns simulation time, where five independent simulations for each TM model are shown in different colors. (D) Distribution of distances between conserved polar residues. For αβ TM, distances between Cβ atom of residues α37/α41 and Cζ atom of β29 (d_α37–β29/d_α41–β29) were measured. For TCRδγ, the distances were measured between δ37/δ41 and γ29 residues. Similarly, the distances between pre-Tα37/pre-Tα41 and β29 were measured for pre-Tαβ. To guide visual inspection, the corresponding distances for the previously reported αβ TM model are shown in blue lines. The density of the distribution is shown in a rainbow color scheme. In (C) and (D), ionization states of Asp in pre-Tα are indicated in the corresponding panels. To see this figure in color, go online.

From the simulation trajectories, representative structures of each receptor TM heterodimer were obtained by clustering sampled conformations (see Supporting Material for details). Both δγ and pre-Tαβ (with a protonated Asp residue) maintained stable αβ-like Asn/Asp-Thr-Tyr H-bond networks (Fig. 2 B and Table 1). In addition to the conserved H-bond networks, the receptor TMs show comparable (pre-Tαβ) or lower (δγ) overall root mean square deviations (RMSD) from their initial models (Fig. 2 C), and similar distributions of distances between conserved polar residues compared to αβ TM (Fig. 2 D). These data indicate that both δγ and pre-Tαβ TM sequences are fully compatible with the formation of stable, αβ-like interfaces.

Table 1.

Interhelical H-Bonds Involving Conserved Residues

TM Model	Residue Paira	Occupancy (%)b
WT-TCRαβ	αN37-βA26	96.4 ± 1.0
	αT41-βY29	94.5 ± 4.2
TCRδγ	δN37-γS26	97.8 ± 0.8
	δT41-γY29	83.1 ± 9.2
	δT33-γS26	98.1 ± 1.2
pre-Tαβ (Asp0)	pTαD37-βA26	97.1 ± 0.2
	pTαT41-βY29	99.3 ± 0.4
pre-Tαβ (Asp−1)	pTαD37-βY29	26.7 ± 21.4
αN37A	αT41-βY29	97.1 ± 1.9
αN37L	αT41-βY29	28.4 ± 23.2
αN37F	αT41-βY29	4.9 ± 4.3
βV33F	αN37-βA26	21.7 ± 11.1
	αT41-βY29	27.3 ± 22.1
αT41A	αN37-βA26	40.3 ± 29.9
βY29F	αN37-βA26	52.2 ± 19.9
αT41A-βY29F	αN37-βA26	58.4 ± 29.2
Triple mutation	−	−

^aOne or more residues in the pair are evolutionarily conserved.

^bThe standard errors were calculated over the replicas.

However, within this conserved structural framework, the δγ and pre-Tαβ TM heterodimers each exhibited a unique feature that modulated the stability of their interfaces compared to the αβ TM reference. In the δγ TM, there is an additional H-bond across the interface, resulting in lower overall backbone RMSD from the αβ TM model (12) and tighter packing of the TM interface during the simulations (Fig. 2, B and C and Table 1). The amino acids forming this additional H-bond (Ser/Thr) are conserved in vertebrate δγ sequences, but are absent in most vertebrate (and all mammalian) αβ sequences (12). With respect to the pre-Tαβ interface, the most notable feature is that the position of the conserved Asn in TCRα and TCRδ is Asp in all known pre-Tα sequences (12). As shown in Fig. 2, C and D, pre-Tαβ TM dynamics are similar to αβ only when the Asp is protonated, and the ionized form is highly destabilizing in the TM interface (see also Fig. S2). These deviations from the TCRαβ TM interface are linked to sequences that are highly conserved within (but not across) TCR types, suggesting that the interaction may have been shaped by unique evolutionary pressures in different functional contexts (see further discussion below).

In our previous study, we identified mutations in the αβ TM interface that generated poorly assembled or unstable TCR-CD3 complexes (12). To further explore the structural consequences of these alterations on αβ TM packing and stability, we performed simulations probing the effects of interface mutations to address two scenarios: 1) introduction of steric blocks through substitutions at positions that are buried in the TM interface and 2) disruption of interhelical H-bonds by the removal of polar groups using the most structurally conservative substitutions possible. For the first scenario, we mutated the conserved polar residue Asn in TCRα to Ala, Leu, and Phe (αN37A, αN37L, and αN37F), and mutated an interface Val in TCRβ to Phe (βV33F). For the second scenario, we introduced single and double mutations at the conserved TCRα Thr and TCRβ Tyr positions to remove polar hydroxyl groups (αT41A, βY29F, and αT41A-βY29F), as well as a triple mutation of all key polar residues to eliminate both H-bonds (αN37A-αT41A-βY29F). The representative structures from the mutant αβ TM simulations are shown in Fig. 3 A (see also Figs. S3 and S4), with the structural similarity to wild-type αβ TM in terms of RMSD time series and distributions of distances between conserved polar residues (Fig. 3, B and C).

Figure 3.

Effects of mutations on TCRαβ TM structure and stability. In the first four mutations (αN37A, αN37L, αN37F, and βV33F), alterations were made to the residues buried in the intimately packed regions. In the latter four mutations (αT41A, βY29F, αT41A-βY29F, and triple mutant), interhelical H-bonds were disrupted, and all three polar residues were eliminated in the triple mutation αN37A-αT41A-βY29F. (A) C-terminal interfaces of mutant TCRαβ TM models in which mutated residues are shown in orange and interhelical H-bonds are shown as red dotted lines. The Arg and Lys residues (mutated to Leu) interfacing to signaling modules are marked by cyan spheres. (B) RMSD between mutant TCRαβ TM models and the previously reported TCRαβ TM model (12) over the 200-ns simulation time, where three independent simulations for each TM model are shown in different colors. (C) Distribution of distances between residues α37 and β29 (d_α37–β29), and α41 and β29 (d_α41–β29), where the distances are measured between the Cβ atom of α37/α41 and the Cζ atom of β29. To guide visual inspection, the corresponding distances for the previously reported αβ TM model are shown in blue lines. To see this figure in color, go online.

For scenario 1 (introducing steric blocks), αN37F was the most severe mutation, introducing a bulky hydrophobic residue in the most intimately packed region. In simulation, this mutation caused substantial variation from the starting TM structure (Fig. 3 B), prevented the formation of the αT41-βY29 interhelical H-bond (Table 1), and generated a wide distribution of interhelical distances in this region (Fig. 3 C). Mutations αN37L and βV33F caused similar but lesser defects (most visible in distance plots (Fig. 3 D) and H-bond occupancy (Table 1)), whereas αN37A had only small effects by all measures. For scenario 2 (removal of polar groups without introducing steric blocks), αT41A caused larger RMSD variation than βY29F or the αT41A-βY29F double mutant (Fig. 3 B). This suggests that leaving an unpaired TCRβ Tyr hydroxyl was particularly destabilizing in simulations. All three mutants showed significantly reduced occupancy of the side-chain-to-backbone H-bond between αN37 and βA26 (Table 1), and the additional elimination of this H-bond in the triple mutant had no compounding effects. Together, these data indicate that loss of H-bonds is not in itself sufficient to cause a breakdown in the TCRαβ TM interface on the timescale of these simulations (200 ns), but that interrupting the close packing or leaving TCRβ Tyr 29 without a H-bonding partner causes significant instability and structural alterations.

These results parallel our previous experimental analysis of the same mutants in the context of full-length proteins, where we interrogated their effects on the stable assembly of αβTCR with signaling subunits by measuring recovery of intact TCR-CD3 complexes from detergent lysates of ER microsomes producing assembled receptors in vitro (12). In these experiments, substitutions at αN37 caused defects that increased with side-chain size, whereas other mutations had intermediate or no effects (Fig. 4 A). To generate a single quantitative measure of TCRαβ TM structural alterations for comparison with the biochemical data, we examined the correlation between contact maps for the coiled-coil TM interface (18) of each mutant against wild-type TCRαβ (Fig. 4 B; see Supporting Material for analysis). As shown in Fig. 4, A and B, the pattern observed in simulations closely mirrors the biochemical results for most mutants, showing that the degree of departure from the WT interface is related to the severity of structural defects in whole-TCR complexes. However, mutants αT41A and βY29F do not follow this pattern, showing similar intermediate effects in the biochemical assay (Fig. 4 A) but very different degrees of correlation to the WT contact map in simulations (Fig. 4 B). The biochemical effects of these mutations in the context of assembled receptors may therefore be more complex than what is reflected in isolated TCRαβ TM simulations.

Figure 4.

The precise interhelical interfaces and their stability are key determinants of structural integrity in the assembled TCR-CD3 complex. (A) Recovery of intact TCR-CD3 complexes with indicated TM mutations relative to the wild-type human TCRαβ TM sequences, reprinted from (12). The statistical significance (ns: not significant; ^∗P < 0.05; ^∗∗∗P < 0.001; and ^∗∗∗∗P < 0.0001) was determined using an ordinary one-way ANOVA uncorrected Fisher’s least significant difference test with single pooled variance. (B) Correlation of interhelical contact maps between mutant αβ TM models and the previously reported wild-type αβ TM model. Statistical significance could not be evaluated in this analysis due to the large variations inherent in the measurements from destabilized mutants. To see this figure in color, go online.

Given the relationship between TCR dimer TM structure and the stability of intact receptor complexes, what may be the functional significance of the unique features we observed in δγ and pre-Tαβ interfaces? Compared to αβTCRs, which bind primarily to “classical” peptide-MHC complexes, δγTCRs recognize a more structurally diverse set of antigens (1), some of which exhibit radically different TCR binding geometry (19) and/or unusually high affinity. δγTCRs may therefore require a core TM structure with enhanced stability, enforced by the additional H-bond, to set a higher energetic threshold against structural alterations upon interaction with ligands. The pre-TCR, on the other hand, signals without requiring ligands at all (4, 5), is expressed at very low surface levels (20), and forms weak (detergent-sensitive) complexes with the ζζ signaling subunits (21). The strong destabilizing influence of Asp deprotonation in the pre-Tαβ TM simulations raises the interesting possibility that one or more of these characteristics are related to the ionizability of this residue. Although Asp is likely to be protonated in a lipid bilayer composed of mostly neutral lipids (22), as found in the ER during assembly (23), the high concentration of anionic lipids in the inner leaflet of the plasma membrane could favor deprotonation at the cell surface and perturb the interface, as in our simulations. The results presented here thus provide a strong rationale for further investigation of TCR TM interface features in a functional context.

Author Contributions

S.P., L.K., M.J.C., M.E.C., and W.I. conceived the study. S.P. performed research. S.P., M.E.C., and W.I. wrote the manuscript. All authors analyzed data and edited the manuscript.

Editor: Markus Deserno.