Introduction

The A6 TCRαβ and B7 TCRαβ (herein we call TCRαβ simply as TCR) are both specific for the same Tax peptide (LLFGYPVYV) of the human T lymphotropic virus 1 (HTLV-1) bound to HLA-A2 (Garboczi et al., 1996a,b; Ding et al., 1998). A6 and B7 derive from from T cell clones isolated from two patients with HTLV-1-associated myelopathic/tropical spastic paraparesis (Utz et al., 1996; Ding et al., 1998). They share the same Vβ germline gene (TRBV6-5) and differ only in the Vα germline gene (A6: TRAV12-2; B7:TRAV29DV5), with sequence similarity of 45% for Vα, 96% for Vβ, and 100% for Cα and Cβ (Ding et al., 1998). The only structural differences between the two TCRs are from the residues of the Vα domain and the highly variable complementarity-determining region 3 (CDR3β) loop of the Vβ domain that is crucial for peptide recognition (Ding et al., 1998; Bourcier et al., 2001; Rudolph et al., 2006). In crystal structures, both TCRs bind in a diagonal orientation to the Tax peptide-bound major histocompatibility complex (pMHC), such that both Vα and Vβ contact the MHC α1 and α2 helices (Garboczi et al., 1996a). The similar diagonal binding modes are achieved by interactions involving different CDR residues of A6 and B7 contacting largely the same sets of pMHC residues (Ding et al., 1998). Only one out of 17 residues contacting pMHC in B7 is also found in the A6 pMHC interaction (Ding et al., 1998). T cell response assays demonstrated that single-residue mutations to the Tax peptide elicit different responses in the two TCRs (Ding et al., 1998; Hausmann et al., 1999). Interfacial interactions may play a role in this TCR-specific response, as residue charges at the surface of the A6 and B7 variable domains show different electrostatic profiles, where the pocket for the Tax peptide Y5 residue is positively charged in A6 but negatively charged in B7 (Ding et al., 1998). Overall, the B7 Vα surface has fewer charged residues exposed than A6 Vα. While A6 and B7 recognize Tax-MHC with similar affinities and kinetics, it has been suggested that they achieve binding via different thermodynamic pathways (Davis-Harrison et al., 2005).

However, since αβTCR is a mechanosensor (Reinherz et al., 2023), the TCR-pMHC bond lifetime under physiological piconewton (pN) level load, rather than equilibrium binding pathway, should be more functionally relevant. In this regard, we have previously used all-atom MD simulations to show that in A6 (Chang-Gonzalez et al., 2024) and JM22 (Hwang et al., 2020) TCRs, contacts with pMHC are stabilized when an adequate 15–20 pN force is applied. The force-induced stabilization occurs as the asymmetric domain motion of the TCR chassis leads to weakening of the interface with pMHC either in the absence of an adequate level of force or if the sequence of the bound peptide is incompatible with maintaining contacts in the loaded state. The goal of this study is to determine whether the load-dependent control of the binding with pMHC is also present in B7, and identify any differences with A6 that may impact the response of the T-cell while responding to the same pMHC.

We find that the mechanism of dynamic allostery is largely conserved in B7, yet the loaded state does not stabilize contacts with pMHC as robustly as in the A6. Thus, while both A6 and B7 possess comparable equilibrium binding affinity for pMHC, A6 appears to exhibit a stronger catch bond behavior under load. Given the nuanced and dynamic nature of TCRs against the same pMHC, including the effect of ligand abundance (Akitsu et al., 2024), difference in mechanical response between A6 and B7 suggests T-cell clones bearing those TCRs may operate differently in vivo. Our study also underscores the importance of dynamics when comparing between TCRs that have similar crystal structures and equilibrium binding affinities.

Results

Our simulation systems include the B7 TCR bound to the Tax-pMHC (Figure 1A,B) with no, low, and high extensions to apply different loads, and an isolated B7 TCR (Table 1). We also used systems without the C-module (Vαβ and Vαβ-pMHC) to study the role of the C-module. We analyzed TCR-pMHC intermolecular and intra-TCR (intramolecular) interactions and domain motion, to test whether the TCR-pMHC interface is stabilized by load, and to find the underlying allostery mechanism that involves the motion of the TCR chassis.

Table 1.

Figure 1.

Stabilization of the TCRαβ-pMHC interface with load

Compared to the no– (B7⁰) and low-load (B7^low) cases, the number of high-occupancy contacts were more numerous for the high-load case (B7^high), indicative of a catch bond behavior. Absence of the C-module (Vαβ-pMHC) also promoted more contacts with pMHC, suggesting the allosteric role of the C-module for binding with pMHC (Figure 1C). These results agree well with the behaviors seen in A6 and JM22 TCRs in our previous studies (Chang-Gonzalez et al., 2024; Hwang et al., 2020). However, the number of pMHC contacts with B7 was reduced compared to A6 (Figure 1C). This is despite their comparable number of contacts with the pMHC in crystal structures and comparable equilibrium binding affinity in solution (Ding et al., 1998; Davis-Harrison et al., 2005). In fact, we had to modify our simulation protocol to avoid premature breakage of contacts between B7 and pMHC when preparing the system for production run under load (see Methods).

Time-dependent behavior of the TCR-pMHC interface further supports the load-mediated enhancement of binding. For this, we calculated instantaneous force and contact occupancy in 40-ns overlapping intervals starting from 200 ns (Figure 1D). B7^high had overall more inter– and intramolecular contacts (more orange-yellow circles) than B7^low (more purple), suggesting that the increased load stiffens the TCR-pMHC complex. B7⁰ had fewer TCR-pMHC contacts while intra-TCR contacts increased (horizontal bars in Figure 1D), indicating decoupling of the TCR from pMHC in the absence of load.

Occupancy heat maps for individual contact residue pairs show reduced or fragmented contacts for B7⁰ and B7^low (Figure 2A,B, more red compared to blue) while B7^high and Vαβ-pMHC exhibit more persistent contact profiles (Figure 2C,D, more blue compared to red). The heat maps and contact counts suggest that interfacial contacts were dominated by MHC-Vα (Figure 3A). Comparing average counts of high-occupancy pMHC contacts for both A6 and B7 TCRs indicates that Vα formed more contacts with MHC than Vβ, while Vβ formed more contacts with the peptide than it does with MHC (Figure 3A). Temporal progression of the number of contacts was measured via the Hamming distance ℋ, the number of the initial high-occupancy contacts lost over time (Figure 3B; see Methods). For B7⁰, ℋ rapidly increased and by 200 ns, most of the initial high-occupancy contacts were lost. While the increase in ℋ for Vαβ-pMHC was comparable to that of B7^low (Figure 3B), the contact occupancy heat maps reveal that Vαβ-pMHC maintained contacts after a brief initial adjustment (gray arrow in Figure 2D) while contacts were lost in B7^low (red Figure 2B).

Figure 2.

Figure 3.

Location of V-module residues forming contacts with pMHC with greater than 50% average occupancy were concentrated along the peptide for B7^high and Vαβ-pMHC, but scattered in B7⁰ and B7^low (Figure 2E). This trend was also observed in A6 and JM22. Such concentration of highoccupancy contacts may protect them from breakage by water. In A6 and JM22 TCRs, the greater number of contacts with pMHC in the high-load cases and Vαβ-pMHC correlated with larger buried surface area (BSA) of the residues forming contacts (Chang-Gonzalez et al., 2024; Hwang et al., 2020). B7 did not follow this trend, as B7^high and Vαβ-pMHC had reduced total and per-residue BSA than B7⁰ and B7^low (Figure 3C). This is likely because the fewer high-occupancy contacts in B7 (Figure 1C) tend to be more exposed, making the relationship between the BSA and load less direct. Consistent with this, the total BSA of B7 was 67.4 % (B7⁰) to 44.2% (B7^high) of the corresponding values of A6.

The residues of pMHC forming greater than 50% average occupancy under high load differed between A6 and B7 (Figure 3D, top row), as did the location of V-module residues forming respective contacts with the pMHC residues (Figure 3D, bottom row), further highlighting their divergence in the interfacial footprint under load. Despite this, the root-mean square fluctuation (RMSF) of C_α atoms of the Tax peptide measured after 500 ns was reduced in B7^high and Vαβ-pMHC compared to B7⁰ and B7^low (Figure 2F), as observed for A6 (Chang-Gonzalez et al., 2024), which supports loadinduced stabilization of the complex.

We calculated the distance between the V-module and pMHC as another measure of the interfacial stability (Figure 3E; Methods). The distance was stably maintained in Vαβ-pMHC and B7^high whereas it fluctuated more in B7⁰ and B7^low. Of note, the former two systems maintained the distance greater than that in the crystal structure by 0.3–0.9 Å. Thus, a slight separation engendered by force or in the absence of constraint imposed by the C-module provides room for adjusting residues to form more stable contacts.

CDR3 positions are controlled by load-dependent Vα-Vβ motion

The greater number of Vα-Vβ contacts in B7⁰ (Figure 4A) is consistent with the increase in total intra-TCR contact occupancy (horizontal bar in Figure 1D bottom panel). Without load this does not translate to a stronger TCR-pMHC interface explained above. B7 in general had fewer Vα-Vβ contacts (11.0–16.3) than A6 (15.9–23.1) (Chang-Gonzalez et al., 2024). The ∼70% reduction in Vα-Vβ contacts for B7 is comparable to the ∼50% reduction in contacts with pMHC between the two TCRs (Figure 1C).

Figure 4.

Vα-Vβ motion was measured via triads (orientational markers) assigned to respective domains and by performing principal component analysis (PCA) (Figure 4B; Methods). PC amplitude was the lowest for B7^high and Vαβ-pMHC (Figure 4C), which is consistent with the greater number of Vα-Vβ contacts. Regarding the direction of motion, the mutually orthogonal PC directions can be difficult to interpret (arrows in Figure 4B). We instead measured angles between the matching arms of the two triads named ∠e_i (i = 1, 2, 3), to examine the Vα-Vβ motion in structurally interpretable directions (Figure 4D) (Chang-Gonzalez et al., 2024). For example, ∠e₁ is the angle between each e₁arm from Vα and Vβ, which describes a ‘flapping’ or ‘twisting’ motion of the two domains. Since e₂ and e₃ lie approximately parallel to the Vα-Vβ interface, they vary reciprocally, corresponding to a ‘scissoring’ motion (Hwang et al., 2020).

Measuring the distance between CDR3α and CDR3β (“CDR3 distance”) revealed that this distance is the shortest for B7^high followed by Vαβ-pMHC. Comparing CDR3 distance versus triad angles (Figure 4E) shows that CDR3 distance varied in opposite directions with ∠e₂ and ∠e₃, which reflects their reciprocal relation (opposite slopes in Figure 4E). In comparison, the CDR3 distance of the B7 crystal structure is 12.0 Å, which is larger than those of B7^high and Vαβ-pMHC (Figure 4E). The slight separation between the V-module and pMHC (Figure 3E) in Vαβ-pMHC and B7^high allows CDR3 loops to come closer together compared to the crystal structure, akin to pinching the central protrusion of the peptide.

Asymmetric V-C bending in the B7 TCR is suppressed with applied load

Next we considered the motion between the V– and C-modules (“V-C motion”). The number of high-occupancy contacts for the Cα-Cβ interface (26.2–27.5) was considerably greater than those for Vα-Vβ (11.2–16.3), indicating that the C-module acts as a single base for the V-C motion, as noted for A6 and JM22 TCRs (Hwang et al., 2020; Chang-Gonzalez et al., 2024). Continuing this general feature, there were fewer high-occupancy contacts for the Vα-Cα interface compared to the Vβ-Cβ interface (Figure 4F). With two exceptions, residues involved in Vβ-Cβ contacts were the same regardless of force.

The V-C motion was analyzed by using the bead-on-chain (BOC) model that tracks individual domains and the hinge between them (Figure 4G). PC motion directions were compared by calculating the dot products between the corresponding PC vectors (Figure 4H). PC1 corresponding to the V-C bending in B7^high (Figure 4G) was similar in other systems, whereas B7^low differed the most (darker color for B7^low in Figure 4H, PC1). Amplitudes of PC1 shows a clear distinction where the unliganded Tαβ and B7^low were more mobile than B7⁰ and B7^high (Figure 4I). Comparing PC amplitudes of the elements of the BOC between α and β chains revealed that Vα moves more relative to the C-module than Vβ (Figure 4J), similar to A6 (Chang-Gonzalez et al., 2024). Amplitude of the hinge motion in the two chains varied, where Hα had greater amplitude in PC1 for B7^low compared to B7^high (Figure 4J, PC1 in bottom row). This suggests a more pronounced asymmetric motion in B7^low. The no-load B7⁰ Hβ amplitude was larger compared to Hα (Figure 4J, bottom, negative value for PC1 of B7⁰). For B7⁰, the small PC1 amplitude of the overall V-C motion without load (Figure 4I) does not suppress the motional asymmetry between α and β chains, while in B7^high, the chassis becomes less mobile under load.

Analogous to the triad angle, the V-C angles (Figure 4G, ∠TCRα and ∠TCRβ) reveal the motional asymmetry in addition to PCA. As in the case of low-load A6 (Chang-Gonzalez et al., 2024), ∠TCRα of B7^low shows a wide bimodal distribution (Figure 4K). ∠TCRα decreased at around 700 ns as B7^low bent more (Figure 4M). This would put the V-module in an unfavorable orientation to bind pMHC (Hwang et al., 2020). Also, about the dependence of the CDR3 distance on V-C angles, it was the most steady in B7^high (Figure 4L). This supports the allosteric mechanism by which the asymmetric V-C motion of the whole TCR controls the Vα-Vβ motion and the V-C orientation that in turn affects the stability of the TCR-pMHC interface.

Concluding Discussion

As a protein-protein complex, TCR-pMHC forms a weak interface. A typical heterodimeric protein-protein interface with BSA comparable to that of TCR-pMHC (∼1700 Ų) has sub-βM binding affinity (Chen et al., 2013) while the affinity of the TCR-pMHC complex ranges between βM to hundreds of βM (Rudolph and Wilson, 2002; Rudolph et al., 2006). Given the low equilibrium affinity, recent findings highlight the importance of mechanosensing, where force generated during immune surveillance of αβ T-cells is utilized to discriminate cognate versus non-cognate pMHCs (reviewed in Zhu et al. (2019); Liu et al. (2021); Reinherz et al. (2023)). Our earlier simulation studies of the JM22 and A6 TCRs showed that the dynamical motion of the TCR chassis is responsible for the force-driven stabilization of the contacts with pMHC, i.e., catch bond formation (Hwang et al., 2020; Chang-Gonzalez et al., 2024). Examining additional TCRs via all-atom MD simulation can inform how the mechanism is adopted. By finding similarities and differences between A6 and B7 TCRs, which recognize the same Tax pMHC, we can elucidate how mechanical force is utilized by each to impact the differential function of T cells.

The present study shows that the dynamic allostery of the TCR chassis observed in A6 (Chang-Gonzalez et al., 2024) and JM22 (Hwang et al., 2020) is largely conserved in B7. Cα and Cβ domains form extensive contacts to construct a base, and the Vβ-Cβ contacts are more extensive than Vα-Cα contacts. This leads to an asymmetric V-C motion where Vα is more mobile relative to the C-module than Vβ. In turn, the motional asymmetry affects the relative positioning of CDR loops, as measured by the CDR3 distance as well as the orientation of the TCR-pMHC interface relative to the loading direction. Unless an adequate load is applied to the complex to suppress the motion, destabilization of the interface occurs. As noted in A6 and JM22, the C domain renders a disadvantage to binding except under force. This is indicated by the more interfacial contacts and decreased Vα-Vβ motion of Vαβ-pMHC (no C-module) simulations. Therefore, while the Variable domain dictates TCR fit with pMHC, a logical evolutionary pressure would be for the C domains to maximize discriminatory power by adding instability to the TCR chassis.

While dynamic allostery is overall conserved, A6 and B7 differ in the behavior of the TCR-pMHC interface under load. The crystal structures of the two complexes have a comparable number of TCR-pMHC contacts, with a total of 33 for A6 and 29 for B7 in terms of residue pairs, while the number of distinct atom pairs forming contacts are 46 for A6 (Garboczi et al., 1996a) and 63 for B7 (Ding et al., 1998). During MD simulations, many of these contacts become transient, resulting in fewer high-occupancy contacts. In the thermally fluctuating state, there is a 4-fold increase in the difference in number of contacts between A6 and B7 over that difference for the static crystal structures (Figure 1C). Previously, contacts found in crystal structures have been used to explain differences in responses to point mutations to the Tax peptide. For example, the sensitivity of B7 to the mutation of Y5 of Tax was explained based on the Y5-D30α hydrogen bond and stacking of Y5 with Y101β of B7 (Y101β in the present study was numbered Y104β in Ding et al. (1998) and Hausmann et al. (1999)). In comparison, the more amenable pocket for Y5 on A6 was suggested to be responsible for the greater tolerance of A6 under mutations of Y5. In MD simulations of B7, the Y5-D30α hydrogen bond is formed only in B7^high, after about 300 ns (Figure 2C), and the Y5-Y101β nonpolar contact breaks during simulations of B7⁰ and B7^low (Figure 2A,B) while it persists with 69% occupancy in B7^high and 52% occupancy in Vαβ-pMHC (Figure 2C,D). Changes to T cell specificity by point mutations on Y5 are thus unlikely to arise solely from the size or the polarity of its binding pocket as a static structure. Instead, point mutations affect the organization and dynamics of the surrounding contacts in addition to fit (Chang-Gonzalez et al., 2024). In a related vein, intermolecular motional network has been suggested to be a controlling factor for allostery in a peptide-SH3 domain complex (Gomez et al., 2024), which aligns with the nonlocal and dynamic role of contacts at the TCR-pMHC interface. The inter-domain motion of TCR enables long-range allosteric discrimination of pMHCs, integrating the motion of the two entities such that they influence each other.

The 14.5-pN force on B7^high was applied under 190.0-Å extension (Table 1), whereas the corresponding high-load simulation for A6 was with 18.2 pN at a slightly lower 187.7-Å extension (Chang-Gonzalez et al., 2024). The lower force of the B7 complex despite the larger extension and the overall smaller number of contacts with pMHC in comparison to A6 (Figure 1C) leads us to predict that B7 exhibits a weaker catch bond where the peak of the TCR-pMHC bond lifetime is at a lower force. As a related example, PA59 and PA25 TCRs share the same TRBV and TRBJ genes, and they share 11-aa CDR3β loops that differ only at a single position (Trp vs. Leu) (Akitsu et al., 2024). Recognizing the same pMHC (PA_224−233/D^b), their maximum bond lifetimes and the peak catch bond forces differ considerably, 75 s at 21 pN for PA59, and 13 s at 15 pN for PA25 (Akitsu et al., 2024). These differences may facilitate actions of the corresponding T-cells in tissues that present diverse mechanical environments provided by cell movement, adhesion molecules, tight junction, and non-compliant inflammation.

Likewise, T-cells bearing A6 and B7 may also perform differently in vivo depending on tissue localization (Akitsu et al., 2024) even though they behave similarly in vitro in terms of cytotoxicity and secretion of select cytokines (γ-IFN, MIP-1α, and TNFα) (Ding et al., 1998; Hausmann et al., 1999). Disparate biological outcomes between structurally similar TCRs recognizing the same pMHC are mechanically possible since slight changes in interfacial contacts can result in altered distribution of loads across the TCR chassis so that changes in its dynamics affect interaction with CD3 signaling subunits (Reinherz et al., 2023). Structural details of the dynamic amplification and propagation of the recognition signal warrant further investigation.

Computational Methods

Structure preparation

B7 TCRαβ-pMHC was built from PDB 1BD2 (Ding et al., 1998) using CHARMM (Brooks et al., 2009; Hwang et al., 2024). Non-numeral residue IDs in the PDB were renumbered to follow sequential numbering used in the present study. We used MODELLER (Šali and Blundell, 1993) to generate coordinates for missing loops in the Cα domain (S133-K136 and S170-D172 in the PDB numbering scheme) followed by a brief energy minimization. We visually verified MODELLER results, comparing generated loops to those of the related A6 TCR (Garboczi et al., 1996a; Ding et al., 1998). The constant domain of TCRα (Cα) was also missing coordinates for F204-S210 (F198-S204 after renumbering), which were added with the TCRα linker as detailed below. Disulfide bonds were assigned between cysteine residues as defined in the PDB file. Crystal waters within 2.8 Å from the protein were kept for the truncated structures, and all waters were kept for the full structure.

Histidine protonation state was determined to promote hydrogen bond formation with neigh-boring residues. The histidine N^δ atom was protonated as follows: MHC residues 3, 93, 114, 145, 151, 188, 260; β2m residues 13, 51; TCRα all histidine residues; and TCRβ residues 29, 47, 154. For the remaining histidine residues, the N^є atom was protonated.

As done in Hwang et al. (2020) and Chang-Gonzalez et al. (2024), we extended the MHC and TCRαβ termini as handles for applying positional restraints. For MHC, we used UniProt P01892 to add ²⁷⁶LSSQPTIPI²⁸⁴. For TCRα we used GenBank AAA60627.1 to add ²⁰⁵CDVKLVEKSFETDT²¹⁸. For TCRβ we used GenBank AAC08953.1 to add ²⁴⁵CGFTSESYQQGVLSA²⁵⁹. We placed an interchain disulfide bond between αC205-βC245. Added strands in the initially straight conformations were relaxed to a state similar to that in Figure 1A by performing a series of brief energy minimization and MD simulation with the FACTS implicit solvent model (Haberthür and Caflisch, 2008).

Truncated structures were built based on the prepared B7 TCR-pMHC complex as:

• Vαβ: The last residues were αP110 and βV113.

• Tαβ: The last residues were αD206 and βG246 (no C-terminal strands).

• Vαβ-pMHC: includes Vαβ, pMHC, and β2m. The last residue of MHC was L276.

• B7⁰: includes Tαβ, pMHC, and β2m. The last residue of MHC was L276.

MD simulation protocol

Solvation, energy minimization, heating, and equilibration of the B7 complexes followed the protocol in Chang-Gonzalez et al. (2024), except for systems which include the pMHC, where we modified the preparation protocol prior to production runs as detailed below.

Laddered extensions

Applying the same protocol as done for A6 to achieve the laddered extensions in B7 resulted in substantial breakage of the TCR-pMHC contacts within the first 50 ns in several production runs. To mitigate this, we introduced distance restraints to selected atom pairs forming contacts between the TCR and pMHC to prevent them from breaking during preparatory simulations. This ensured that the complex could structurally adapt as we modified the extension distance, yet all laddered extensions maintained a core set of initial TCR-pMHC contacts. Atom pair distance restraints were removed in production runs.

Twelve atom pairs between TCR and pMHC were selected that were within 5 Å of each other in the equilibration restart file of the TCR-pMHC complex with added linkers. A 2-kcal/[mol⋅Ų] flat-bottom harmonic restraint potential was applied to keep the atom pair distance within the value at the end of the equilibration run. Then a 2-ns CPT simulation was carried out while also applying a 1-kcal/[mol⋅Ų] harmonic potential to the C_α atoms of the C-terminal MHC (I284), TCRα (T218), and TCRβ (A259) residues. Production run followed upon removing the atom pair distance restraints. During the production run, the 1-kcal/[mol⋅Ų] harmonic potential to the TCR and MHC end-residues and a 10 Å distance restraint between αT218 and βA259 were applied. Throughout the production run, we intermittently measured average and rolling force in 40 ns intervals and found these to be around 10 pN, an ideal target for the low-load simulation. This simulation is the 173.7 Å B7^low system reported in Table 1.

Using the structure at the end of the 2-ns simulation with the TCR-pMHC atom-pair distance restraints, we increased the extension by 8 Å by moving the center of the 1-kcal/[mol⋅Ų] harmonic potential on the end-residue C_α atoms by 4 Å at each end. While keeping the TCR-pMHC atom-pair distance restraints applied, we launched another 2-ns simulation under the increased extension. Atom pair distance restraints were then removed and the production run was launched. The extension averaged after 500 ns of this simulation was 181.7 Å. Following the same way, we increased the extension by another 8 Å, which led to the 190.0 Å B7^high system reported in Table 1. We also decreased the extension by 8 Å from the initial 173.7-Å extension, which led to a 165.7-Å extension.

Among the 4 extensions tested, the 181.7-Å extension was not selected primarily because the average force of the simulation from 500 ns to 850 ns (the total length of the simulation), was 9.26 pN, only barely higher than the reported load for B7^low. For the 165.7-Å simulation, the average force from 500 ns to 900 ns (total length of the simulation) was 15.7 pN. We had observed this high force at low extension for A6 TCR (Chang-Gonzalez et al., 2024) and attribute this to folding of the flexible added strands leading to contacts between the stands and the TCR constant domains.

Vαβ-pMHC and B7⁰

We also applied a 2-kcal/[mol⋅Ų] flat-bottom harmonic distance restraint during preparatory simulations of Vαβ-pMHC and B7⁰. We attempted to use the same set of atom pairs as in the laddered extension simulations, but considerable interface breakage occurred, likely due to changes in interfacial contacts after equilibration in these systems. We thus selected different atom pairs for Vαβ-pMHC and B7⁰. For consistency, we selected 12 atom pairs, the same number as in the laddered extensions, and distributed in the same way between the Tax peptide or MHC residues to Vα or Vβ residues. The distance restraint was applied to the atom pairs for a 2-ns CPT simulation, then released for production runs.

Systems without load

The following additional restraints were used for systems without load.

• Vαβ, Tαβ: no positional restraints were applied.

• Vαβ-pMHC: we applied a weak 0.01-kcal/[mol⋅Ų] harmonic positional restraint to the backbone C_α atoms of MHC α3 (P185-L276) to prevent large transverse rotation of the whole molecule in the orthorhombic box.

• B7⁰: we applied a 0.2-kcal/[mol⋅Ų] harmonic positional restraint to the backbone C_α atoms of MHC α3 with RMSF less than 0.5 Å calculated from the simulations of B7^low. These residues were: L201-Y209, T240-Q242, T259-H263.

Production runs

Production runs were performed similar to Chang-Gonzalez et al. (2024). We used OpenMM (Eastman et al., 2017) with the CHARMM param36 all-atom force field (MacKerell Jr et al., 2004) and the particle-mesh Ewald method to calculate long-range electrostatic interactions. We used an Ewald error tolerance of 10⁻⁴ which is 1/5 of the default value in OpenMM and a 12 Å cutoff distance for nonbonded interactions. The complexes were simulated at 300 K with a 2-fs time step using the Nose-Hoover integrator in OpenMM. Production run lengths are in Table 1.

Trajectory analyses

Analysis methods are detailed in Chang-Gonzalez et al. (2024). Below we mainly explain B7-specific residue selections. Out of ∼1-βs production runs, we excluded the initial 500 ns when calculating the average and standard deviation of the number of contacts, BSA, CDR3 distance, PCA, and triad and V-C angles. With a coordinate saving rate of 20 ps, this leaves at least 25,000 frames for analysis.

V-module to pMHC distance

The distance from TCR V-module to pMHC (Figure 3E) was measured between the center of mass of the C_α atoms of the same residues used to build the V-module triads (described below) to the center of mass of five C_α atoms from each of the central 4 strands forming the β-sheet floor located above the α1 and α2 helices of MHC (20 MHC atoms in total; Figure 1A). These were R6–T10, I23–Y27, Q96–G100, and Y113–A117. RMSF of these residues after 500 ns was below 1.4 Åin all B7 systems, so the measured distance is minimally affected by the intra-domain conformational motion.

CDR3 distance

Distance between CDR3α and CDR3β (Figure 4E,L) was measured using the midpoint between backbone C_α atoms of two residues at the base of each CDR3, which are E93 and K97 for CDR3α and S94 and E102 for CDR3β.

V-module triads

We assigned triads (Figure 4B) based on the backbone C_α atoms of the stably folded β-sheet core of each variable domain (Hwang et al., 2020; Chang-Gonzalez et al., 2024). Selected residues for triad assignment of the B7 systems were as follows. For Vα, I19-Y24, F32-K37, H71-I76, and Y87-M92. For Vβ, T20-Q25, M32-Q37, D73-L78, and Y89-S94. Prior to triad assignment we aligned all complexes to the first frame of B7^low using the selected residues to monitor the relative motion between the two triads without global translation nor rotation.

V-C BOC

We assigned BOCs (Figure 4G) as detailed in Hwang et al. (2020); Chang-Gonzalez et al. (2024). To place beads for the C-module, we used the following residues. For Cα, A118-R123, V132-D137, Y153-T158, and S171-S176. For Cβ, L143-T148, L157-N162, L190-R195, and F208-Q213. For the hinges, we used: αN114 for Hα, and βD116 and βL117 for Hβ. We aligned all complexes to the backbone C_α atoms of the selected residues of B7^low then built BOCs to monitor the motion of the V-module relative to the C-module.

Acknowledgements

This work was funded by US National Institutes of Health Grants P01AI143565 and R01AI136301. Simulations were performed by using computers at the Texas A&M High Performance Research Computing facility.