PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J Phys Chem B. Author manuscript; available in PMC 2012 June 30.
Published in final edited form as:
PMCID: PMC3131071
NIHMSID: NIHMS302614

The effect of mutations on the alloreactive T cell receptor/peptide-MHC interface structure: A molecular dynamics study

Abstract

T cells orchestrate adaptive, pathogen-specific immune responses. T cells have a surface receptor (called TCR) whose ligands are complexes (pMHCs) of peptides (derived from pathogens or host proteins) and major histocompatibility complex proteins (MHCs). MHC proteins vary between hosts. During organ transplants, host TCRs interact with peptides present in complex with genetically different MHCs. This usually causes a vigorous immune response—alloreactivity. Studies of alloreactive protein interactions have yielded results that present a puzzle. Some crystallographic studies concluded that the alloreactive TCR/MHC interface is essentially unaffected by changing the TCR peptide-binding region, suggesting that the peptide does not influence the interface. Another biochemical study concluded from mutation data that different peptides can alter the binding interface with the same TCR. To explore the origin of this puzzle, we used molecular dynamics simulations to study the dependence of the TCR/pMHC interface on changes in both the peptide and the TCR. Our simulations show that the footprint of the TCR on the pMHC is insensitive to mutations of the TCR peptide-binding loops, but peptide mutations can make multiple local changes to TCR/pMHC contacts. Therefore, our results demonstrate that the structural and mutation data do not conflict and reveal how subtle, but important, characteristics of the alloreactive TCR/pMHC interface are influenced by the TCR and the peptide.

Keywords: immunology, protein-protein interfaces, peptide contacts, atomistic simulations

Introduction

The adaptive immune system enables higher organisms, like humans, to protect themselves with pathogen-specific responses against a diverse and evolving world of microbes. T lymphocytes (T cells) are key orchestrators of the adaptive immune response. To perform their functions, they must be activated. Activation is predicated on sufficiently strong binding of a T cell’s antigen receptor (T cell receptor, TCR) to a ligand. The ligand consists of a short peptide fragment held in the cleft of a membrane-bound major histocompatibility complex protein (MHC) displayed on the surface of an antigen-presenting cell. T cell activation can lead to a variety of effector immune functions.

Immature T cells undergo development in the thymus, where they interact with pMHC complexes derived from the host proteome. To survive elimination during the development process, the T cells must not interact too strongly with any of these self-pMHC complexes (negative selection), but must bind with sufficient affinity to at least one pMHC (positive selection).18 This selection process largely inhibits autoimmune T cells from joining the immune system and ensures that the surviving T cells can recognize foreign peptides presented on the host’s own MHC molecules with extraordinary specificity.9,10 Because of thymic selection, peptides derived from the hosts’ own proteins do not produce a strong interaction, but foreign-derived peptides do.

Sometimes, such as during organ transplantation, mature T cells encounter pMHC complexes on cells from a genetically different (allogeneic) member of the same species. Since MHC genes are highly polymorphic, allogeneic pMHCs (allo-pMHCs) present previously unseen MHC surfaces to the TCRs. Furthermore, since the most variable regions of the MHC occur along the peptide binding cleft, the peptides presented by allo-MHCs likely differ in sequence and conformation from the self-peptides used to train the TCR in the thymus, even though they originate from the same proteins. Up to 10% of the T cell repertoire can cross-react with any particular pMHC on the allogeneic cells—1000 times as many T cells as the 0.01% of the repertoire activated during the response to a virus.1119 This intense response, known as alloreactivity, makes organ transplantation impossible without immuno-suppression.

Much experimental work has been dedicated to elucidating the roles and relative importance of the peptide and MHC in alloreactivity, and a large number of these studies2229 have examined the interaction footprint—the interface between the TCR and the set of pMHC residues that come into contact with it, which includes the peptide and the α1 and α2 helices of the MHC (Figure 1). A question of particular interest, has been the energetic and structural impact of the peptide on an allo-pMHC/TCR footprint. The question has been actively explored by biochemical mutation experiments and X-ray crystallography.

Figure 1
The 2C TCR binds strongly to the allo-MHC Ld, contacting a “footprint” set of residues on the pMHC. A. A crystal structure (PDB 2OI9) of the variable chains of the 2C TCR in complex with the peptide QL9 and the MHC Ld shows the relative ...

Some biochemical experiments have studied how a TCR interacts with different peptides in the same allo-MHC. But these studies do not provide direct structural data.30,31 One study, by Felix et al., concluded that peptide mutations could have a noticeable impact on the TCR/pMHC footprint by affecting contacts between MHC and TCR.30 In the study, the authors mutated several residues along the α-helices of a particular allo-MHC and observed the effects of those mutations on T cell activation, in order to assess which residues were likely to contact the TCR. The authors found that when different peptides were bound by the same allo-MHC, different subsets of MHC residues impacted recognition, implying that the peptides affected the TCR/MHC contacts.

Since the first complete TCR/pMHC crystal structure in 1996,32 crystal structures have also been used to study several TCR/pMHC footprints. Many of these footprints have been between TCR and peptide/self-MHC complexes,3238 but structural studies have dealt with alloreactive complexes as well.20,3947 Two of these studies42,47 have made direct structural comparisons of the effects of different peptides of varying affinity on the TCR/allo-pMHC interface. A recent crystallographic study by Colf et al. concluded that the peptide has little impact on the footprint. The authors reached this conclusion by showing that mutations on the CDR3α peptide-recognition loop of a TCR bound to an allo-pMHC complex did not impact the structure, despite increasing the binding affinity by over two orders of magnitude.20 Importantly, however, the study did not involve mutation of the peptide itself. A more physical and chemical understanding of the intermolecular interactions involved at the TCR/pMHC interface could shed light on these seemingly conflicting biochemical and crystallographic results. Such insight could also motivate further experiments.

One direct way to compare the two results is to understand a single system that encompasses both peptide mutation and structural information. To this end, we have performed an in silico analog of the peptide mutation experiments designed by Felix et al. on TCR/pMHC structures obtained by Colf et al.20 Using molecular dynamics simulations,48 we analyzed atomistic models of TCR/allo-pMHC complexes while independently changing both the TCR and the peptide, which allowed us to directly compare the effects of each kind of mutation on the TCR/pMHC interface. In our simulations, as in the crystallographic study, mutation of the CDR3α loop of the TCR did not induce a significant change in the TCR/allo-pMHC footprint, compared to the significant differences between the TCR/allo-pMHC and TCR/self-pMHC footprints. However, our simulations also showed that certain peptide mutations can affect the TCR/pMHC interface. These peptide mutations not only affected the peptide-TCR contacts, but also influenced which MHC residues came into contact with TCR, even though they did not induce a change in the overall orientation of TCR on MHC—a finding that confirms the conclusions of Felix et al.30 Our simulations thus demonstrate that the crystallographic results and biochemical results are not in conflict, because mutations to the CDR3α loop are not necessarily equivalent to peptide mutations. Our results highlight the potential of the peptide to impact the TCR/pMHC interface by making local contact changes and suggest that detailed chemical interactions at the interface between the TCR, peptide, and MHC can all play a part in the ultimate structure of the alloreactive TCR/pMHC interface.

Our findings are consistent with an attractive model for TCR/pMHC interactions in which the TCR docks over the pMHC and scans the ligand for a sufficient number of interactions that confer the TCR/pMHC complex a sufficient lifetime.8,4951 If this necessary condition for recognition is met, structural rearrangements occur to acquire enhanced affinity. The specific character of these relatively modest rearrangements29 depends on the particular TCR/pMHC pair under consideration.

Methods

Structure preparation

The X-ray structures of the C9X and M9X20 complexes (PDB IDs 2OI9 and 2E7L, respectively) were used for the initial coordinates of all calculations. Three mutant-peptide variants were generated from each crystal structure by removing the atoms that corresponded to the mutation (Table 1). This resulted in a total of eight systems to simulate. Hydrogen atoms were introduced into the structures with a stereochemical algorithm.52 The protonation states of titratable residues were assigned based on visual inspection and prior results for similar systems.5355 The structures were minimized with constraints on heavy atom positions and solvated with explicit TIP3P56,57 water molecules in 89 Å rhombic dodecahedra with periodic boundary conditions. 39 K+ ions and 37 –38 Cl ions were added to each system to neutralize the total charge and simulate a 0.15 M KCl concentration. The ions were placed with random initial coordinates, then subjected to 2000 steps of a Metropolis Monte Carlo simulation in order to equilibrate their positions. After solvation, hydrogens, solvent atoms, and protein atoms were all minimized with restraints in stages. First, hydrogens were minimized for 1000 steps without restraints, while heavy atoms were fixed. Then, only the solvent atoms were subjected to 1000 additional steps of constrained minimization, with force constants of 0.5 kcal/mol/Å2, while protein atoms were fixed. After that, all atoms were un-fixed, a 0.1 kcal/mol/Å2 harmonic restraint was placed on the protein atoms, and the system was subjected to 1000 more steps of minimization. Finally, the entire system was minimized again for 1000 steps without restraints. The average number of atoms per system was 47,355.

Table 1
Lower-affinity peptide mutants are good candidates for simulation.

Molecular dynamics simulations

All simulations were performed using the CHARMM58,59 molecular dynamics program (version c34b1) with the CHARMM2256 force field and the CMAP correction for the peptide backbone dihedrals.60 The simulation protocol was influenced by previous TCR/pMHC simulations.5355 Molecular dynamics simulations were performed with holomorphic constraints on the hydrogen atoms,61 which allowed the use of 2 femtosecond integration timesteps. Non-bonded van der Waals interactions were truncated at 9 Å using a force-switching algorithm.62 Electrostatic terms were calculated using the Particle-Mesh Ewald summation method,63 and the real-space terms were evaluated with a 9 Å cutoff. After solvation and the energy minimizations described above, each system was heated and equilibrated in several stages. First, the entire system was heated from 10 K to 300 K over 100 ps without restraints. Then a 3.0 kcal/mol/Å2 harmonic restraint was placed on the heavy protein atoms, and the ions and solvent were equilibrated with three repetitions of a heating and cooling cycle. The cycle ran constant-temperature dynamics for 50 ps at 300 K, then heated the system to 450 K, ran constant-temperature dynamics for 100 ps, cooled the system down to 300 K, then ran another 100 ps of constant-temperature dynamics at 300 K. After the solvent equilibration, a mass-weighted 0.75 kcal/mol/Å2 harmonic restraint was placed on protein heavy atoms and the system was equilibrated for 100 ps with constant-temperature dynamics. The restrained dynamics were followed with 100 ps of unrestrained constant-temperature dynamics, and 200 ps of unrestrained constant-temperature, constant-pressure dynamics. Simulations were then continued for 15 ns. Throughout each simulation, constant temperature was maintained with the Nosé-Hoover thermostat,64,65 and constant pressure was maintained with the Langevin piston method.66

Results

Alloreactive model

The alloreactive model used in our simulations is based on the murine 2C TCR. T cells containing this TCR were found in a mouse—whose own MHC molecules expressed the Kb-haplotype—injected with cells whose MHCs had the Ldhaplotype.69,70 Thus, 2C TCRs recognize foreign and self-peptides on the Kb self-MHC and also bind strongly to peptides presented on the Ld allo-MHC (Figure 1). Crystallographic experiments by Colf et al.20 have obtained the structure of the variable part of a 2C TCR in contact with a peptide (QL9; see Table 1 for sequence) held in the cleft of a stabilized portion of Ld (α12 domains, residues 1 – 180 of the heavy chain). The crystal structure of the same ligand bound to m6, a mutant of 2C with even higher affinity for QL9/Ld, was also obtained. The only difference in sequence between 2C and m6 occurs in residues 94 – 103 of the CDR3α loop, which is changed from the sequence GFASA in 2C to the sequence HQGRY in m6. Despite the 100-fold change in affinity for the ligand caused by this mutation, it had no significant impact on the TCR/pMHC footprint, especially when compared to the structural difference between the 2C/QL9-Ld structure and previously-obtained 2C/Kb self-pMHC structures.20,33 This result led Colf et al. to conclude that the peptide did not play an important structural role in the footprint because, even though interactions between the peptide and TCR had significantly changed, the footprint had not.

Molecular dynamics simulations

To study the effect of both TCR and peptide mutation on the TCR/pMHC footprint, we performed molecular dynamics simulations on a total of eight systems. For the peptide mutants, we used four peptides (Table 1), including QL9. The three additional peptides were derived from QL9 by independently introducing two mutations: deletion of the terminal glutamine, and mutation of the carboxy-proximal proline to an alanine. These mutations were chosen for their ability to noticeably affect the binding affinity of the peptide for 2C.67,68 Each of the four pMHCs was simulated with both 2C and m6 TCR, with the m6 simulations intended to test the effect of TCR mutation. The simulated systems are summarized in Table 2, along with their identifying abbreviations.

Table 2
For clarity and conciseness, we use abbreviations to refer to the TCR/pMHC systems discussed in this article.

Since the peptide mutations involved only removal of atoms, the initial coordinates of the mutant peptides were the same as the crystal coordinates of QL9, except for the deleted atoms. The molecular dynamics simulations were carried out for 15 ns, and coordinates were saved at every 2 ps for analysis. Energetic and structural properties of the systems were calculated by averaging over the last 5 ns of the trajectories, after the systems had reached equilibrated states. Equilibration was determined by a consistent plateau in the RMSD (root-mean-square deviations) of the backbone coordinates from the crystal structure (Supplemental Figure 1). The binding free energy of the m6 TCR for pMHC was calculated to be stronger (more negative) than that of 2C (Supplemental Table 1), which corresponds qualitatively to experimental binding affinity measurements.44

To provide more points of reference, a model of the 2C/Kb system was also simulated. The resolution of the crystal structure referenced by Colf et al., originally obtained by Garcia et al.,33 was not high enough for consistent simulations. Instead, a related structure, which replaced the original dEV8 self-peptide with the superagonist SIYR peptide, was used.35 This structure had a higher resolution, yet differed little (RMSD 0.92 for Cα atoms) from the original 2C/Kb/dEV8 structure and could therefore serve as an approximate model.

Average structures from the dynamics highlight the overall effect of peptide mutation

In order to evaluate the qualitative effects of peptide mutation, we calculated average TCR/pMHC structures from each trajectory by taking the mean position of each atom during the last 5 ns of the trajectory. Figure 2 displays some of these average structures superimposed on one another for comparison. The structures in each part of the figure are aligned along their MHC backbone atoms to highlight differences in TCR loop position and TCR/MHC orientation. Figure 2 is designed to show the effect of TCR mutation (Figure 2A), the effect of peptide mutation (Figure 2B), and the effect of switching from self-pMHC to allo-pMHC (Figure 2C). Looking from Figure 2A to Figure 2C, the differences between the structures in each pair increase. In Figure 2A, the qualitative features Colf et al. found are reproduced: the overall orientation of both TCRs is nearly identical,20 despite a few minor loop rearrangements, leading to an average backbone RMSD of 2.7 Å for the loops. By stark contrast, in Figure 2C, almost all of the TCR loops have significantly different average positions, with an average backbone RMSD of 3.5 Å,29 and it is clear that 2C binds to the different pMHCs at different angles. Figure 2A and Figure 2C together demonstrate that our simulations retain qualitative similarity to the crystallographers’ finding that CDR3 loop mutation is insignificant, especially compared to the differences between self- and allo-pMHC.20 Figure 2B shows how the effects of peptide mutation compare with the two “extremes” of Figure 2A (no effect) and Figure 2C (drastic effect). The two peptide mutant systems shown in this subfigure, C8A and C8P, exhibit the greatest difference from one another among all choices of peptide mutants for a given TCR, highlighting the maximum observed extent of peptide mutation. Comparing C8A to C8P shows that, although there is no global rearrangement of binding orientation, as in Figure 2C, there are many differences between the conformations of the 2C CDR loops, and the differences are distributed throughout most of the CDR loops, yielding an average backbone RMSD of 3.4 Å. Every loop except CDR2α exhibits a noticeable difference between the two structures. These unique loop conformations lead to significantly different TCR/pMHC contacts. In particular, the most significant differences are seen in the CDR3 loops, which contact the peptide directly. The peptides, however, are not adopting vastly different conformations (backbone RMSD 1.9 Å). This result suggests that the peptides interact with the TCR in a highly dissimilar manner despite their relatively similar orientation and that mutation can affect the TCR/pMHC contacts through a complex interplay of altered interactions between the TCR and the bound pMHC.

Figure 2
Average structures of the TCR/pMHC footprints compare the effects of TCR, peptide, and pMHC changes. The comparison highlights the ability of the simulations to reproduce qualitative experimental results. The average structures are aligned via their MHC ...

TCR/pMHC contact distributions allow quantitative comparison of the effects of mutation

To measure the impact of peptide mutations on the footprint, we computed the number of contacts that each pMHC residue makes with TCR and averaged these quantities over the structures obtained during the last 5 ns of simulation. We defined any two residues to be in contact when a non-hydrogen atom in one residue was within 4.5 Å of a non-hydrogen atom in the other residue. This average-contact approach qualitatively differs from analyzing any single “representative” structure taken from the simulation trajectory, since a single structure may not accurately represent the overall ensemble of structures well enough, especially when examining the contact footprint. Thus, an analysis based on a single structure would likely suffer from ambiguities in interpreting the results.

The result of the average contact calculations, shown in Figure 3 and Figure 4, is a picture of the interaction footprints as functions of pMHC sequence. In each plot, the ordinate lists the positional indices and amino acid abbreviations for the residues that constitute a particular segment of the pMHC—either the α1 helix or the α2 helix. Around each point, which represents an individual pMHC residue, a group of bars is clustered. The set of all bars of a single color represents a particular TCR/pMHC system, as described by the legend and Table 2, and the length of a bar represents the average number of contacts that a pMHC residue made with TCR. Figure 4 shows the effect of peptide mutation on the 2C footprint by comparing all simulated systems that involve the binding of 2C TCR to different peptides: C9P, C9A, C8P, and C8A, respectively. Figure 3 serves as a control by comparing the footprints of the 2C and m6 crystal structures (C9X, M9X) with their corresponding simulations (C9P, M9P). The control set provides two methods of testing the simulations against experimental results, showing (1) how well the simulations reproduce the experimental conclusions that mutation from 2C to m6 does not have a significant effect on the footprint, and (2) how similar the footprints generated by the simulations are to their experimental counterparts. To be numerically significant, any changes observed from peptide mutation would have to be larger than the differences between crystal structures and their simulation counterparts, and also larger than the differences between the 2C and m6 TCR simulations.

Figure 3
MHC/TCR contact distributions show a quantitative description of the TCR/pMHC footprint. The figure compares the crystal structure footprints to their corresponding simulations in order to establish the accuracy of the simulations. In general, the simulations ...
Figure 4
Peptide mutation can induce local TCR/MHC contact changes. MHC/TCR contact distributions of peptide-mutant systems show the effects of peptide mutation on the interface. Some of these effects are significant: the C8A distribution is noticeably different ...

Figure 4 directly shows how peptide mutation changes local contacts between MHC residues and TCR in several locations throughout the footprint, in direct support of the findings of Felix et al.,30 that peptide mutations can rearrange the finely-paired contacts between TCR and MHC. For a more quantitative comparison of the footprints, we present a single system’s footprint (the set of all bars of the same color) as a discrete “distribution” of contacts over the sequence of the pMHC. The distributions of different systems can be compared segment-by-segment by their “means,” [r with macron]c, shown in Figure 5. The [r with macron]cs are calculated as weighted averages

Figure 5
The changes in TCR/MHC contacts upon peptide mutation from C8P to C8A are numerically significant. The values of [r with macron]c for each system are plotted above on the same axes for comparison. [r with macron]c represents the “center ...

equation M1
(1)

where An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg [set membership] {α1, α2, pep} is a label for a pMHC segment (either the α1 helix, the α2 helix, or the peptide) of a TCR/pMHC system A, r is the sequence number of a residue in An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg (denoted along the ordinate), and c(r) is the average number of contacts that residue r makes with TCR (the length of the bar at position r that corresponds to system A in Figure 3 and Figure 4). Thus, [r with macron]c(A; An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg) represents the average position in the sequence where the TCR contact distribution for pMHC segment An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg is centered in system A, i.e. the contact center of mass. Differences between the means of two distributions representing different systems reflect underlying differences between the TCR/pMHC contact distributions of those systems (Figure 5), and we will use the notation Δ[r with macron]c(A1A2; An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg) Ȭ |[r with macron]c(A1; An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg) – [r with macron]c(A2; An external file that holds a picture, illustration, etc.
Object name is nihms302614ig1.jpg)| to refer to these differences throughout the paper.

Mutations to the CDR3α loop of the TCR do not produce significant changes to the footprint

To verify the predictive quality of the simulations, we first examined how well they reproduce the experimental result that CDR3 mutations do not significantly alter the footprint.20 Figure 3 and Figure 5 compare the contact distributions and [r with macron]cs of C9P to M9P, and C9X to M9X—an intra-method, inter-TCR comparison—to assess the effect of the TCR mutation (2C → m6) on TCR/MHC contacts. The contact distributions of the crystallographic C9X and M9X systems are very similar to one another, except for a few stray contacts in M9X near the ends of the helices and two contacts with Arg75 of the MHC, which are only present in the C9X structure (Figure 3). Like their crystallographic counterparts, M9P and C9P are also similar to one another, as reflected by the small difference in their [r with macron]cs (Figure 5). However, M9P has approximately one more average TCR contact than C9P in several places where MHC/TCR contacts already exist in C9P. The additional average contacts are distributed almost evenly throughout the entire contact interface. Due to the normalization applied in Eq. (1), this difference is not reflected in the [r with macron]c(M9P; {α1, α2}) values, and therefore, no significant shift is observed in the location of the footprint on the MHC. The even distribution of additional contacts suggests that m6 TCR interacts strongly with the pMHC. This is consistent with the experimental finding that the m6 TCR has a higher affinity for the Ld-QL9 ligand20 and with the computed binding free energies reported in Supplemental Table 1.

As a further test of quality, we compare the QL9 simulations (C9P and M9P) to their corresponding crystal structures (C9X and M9X)—the only direct experimental comparison available—in the bottom of Figure 5. This inter-method, intra-TCR comparison yields larger differences in [r with macron]c between simulation and experiment than the previous comparison, especially for the α1 helix. The differences may reflect the large numbers of crystal contacts (476 crystal contacts for C9X and 874 for M9X), that are present only in the crystal structures and are absent in our simulations. Further analysis shows that, in the α1 helix, 74% of Δ[r with macron]c(C9P–C9X; α1) is due to the two TCR contacts with residue Arg62 on the α1 helix of the MHC which develop in the C9P simulation, but are absent in the C9X structure. Other contact patterns are mostly conserved between simulation and experiment. Since such a large part of the difference can be accounted for by only one new contact location, the difference in [r with macron]cs does not reflect the widespread shifts in pMHC contact locations that would accompany a significant change in the footprint. Although the inter-method Δ[r with macron]c(M9X–M9P; α1) is larger than Δ[r with macron]c(C9X–C9P; α1), and both are relatively large, they are still smaller than the largest Δ[r with macron]c obtained from peptide mutation and therefore do not affect qualitative conclusions about its potential effects.

The smaller size of both of the intra-method, inter-TCR comparison Δ[r with macron]cs, relative to the inter-method, intra-TCR Δ[r with macron]cs, suggests that the simulations and crystal structures might have systematic, localized shifts from one another. Such systematic shifts may be caused, for example, by the absence of crystal contacts in the simulations. The two comparisons, taken together, suggest that despite some localized differences with the crystal structures, our simulations are consistent with the experimental result that mutation of the 2C CDR3α loop does not induce significant perturbations in TCR/MHC contacts with QL9-Ld.

Mutations to the shortened antigenic peptide produce noticeable, local changes to TCR/MHC contacts

Among all the peptide mutant systems tested, two (C8P and C8A) show the largest shifts in [r with macron]c values, larger than any other pair of systems with the same TCR by at least a factor of two (Figure 5). The exact [r with macron]c values that underlie these differences are shown in Supplemental Table 2.

Unlike the case of TCR mutation, the simple creation or annihilation of a single MHC/TCR contact cannot account for the large shift in [r with macron]c. Rather, the shift is caused by the rearrangement of several contacts and an overall shift in the contact distributions of C8A and C8P to opposite ends of the sequence (relative to the wild-type C9P). This can be seen clearly in Figure 2B, where the CDR1 and CDR3 loops for both Vα and Vβ have noticeably different average conformations in C8A than they do in C8P.

We also determined how widely distributed along the MHC the differences between C8P and C8A are (Figure 5). When comparing two footprints, we define a pMHC site as “changed” if it has 1.0 or more average contacts in one footprint, and 0.5 average contacts or less in the other. This choice of cutoff values is reasonable, as a pMHC residue with less than 0.5 average TCR contacts makes no contact with TCR more than half the time, but a pMHC residue with 1.0 average TCR contacts is in contact with the TCR on average once every time step in the trajectory. Using the above definition of change to compare C8A to C8P yields 8 changed sites out of 22, while comparing C9P to C9X or C9P to M9P yields only 4 out of 21 sites. The results are not very sensitive to changes of the cutoff values: if either parameter were varied by up to 10% in either direction, only two changed sites would lose their label, and in both cases, it is because the high value is very close to 1.0, and would cease to qualify as a contact if the upper threshold were raised to 1.1. The comparatively large number of changed sites demonstrates that differences between the C8A and C8P footprint are widely distributed and represent changes throughout the footprint.

Figure 5 also shows that the wild-type C9P contact distributions are centered between those of the C8P and C8A systems for both MHC helices. Thus, relative to the original, wild-type footprint, the lone deletion (C8P) and deletion-with-mutation (C8A) have opposite, non-additive effects, possibly due to how greatly deletion can affect the space available to the peptide. Taken as a whole, these effects are relatively localized to the TCR-MHC interface, and they do not involve the kind of structural rearrangement of TCR/MHC orientation observed when comparing the allo 2C/QL9-Ld(C9P) to 2C/SIYR-Kb (CSK) (Figure 2C), nor do they involve large rearrangement of the CDR loops themselves. This perspective is consistent with assessments of TCR/pMHC interfaces that highlight the relative inflexibility of the CDR loops.28,29 Nevertheless, it is clear that mutation of the antigenic peptide, unlike mutation of the CDR3α, can induce noticeable rearrangement of local contacts in the footprint.

Most importantly, these peptide mutation results are completely consistent with the findings of Felix et. al,30 which implied that changing the peptide can make changes to contacts between MHC and TCR throughout the footprint. They are also consistent with a previous crystallographic study of dissimilar peptides in the same TCR/allo-MHC complex.42 Performing [r with macron]c calculations on the crystal structures compared in that study showed that the contact distributions of the different peptide mutants differed by as much as two residue positions—much more than C9X differed from M9X (Supplemental Table 2).

Recent crystal structures obtained by MacDonald et. al47 show the LC13 TCR in complex with different peptides in the same allo-MHC. In both cases, the TCR produces a remarkably similar footprint on the substrate. In this situation, however, no residues were removed from the peptide, so the physical impact of mutation was not as severe, despite several sequence differences between the two peptides. Such results indicate that not all kinds of peptide mutation are equivalent: while allopeptide mutations can have an impact on the TCR/pMHC footprint, this is not necessarily the case for a particular TCR/pMHC pair and a particular mutation—in the end, the specific interplay of TCR, peptide, and MHC interactions determines the ultimate interface structure.

Peptide mutations can impact the topology of the interface by changing the ratio of TCR Vα and Vβ chain contacts

A particular feature of the 2C contact distributions is that the ranking of [r with macron]c values for the α1 helix is the inverse of the ranking for the α2 helix. That is, [r with macron]c(C8A; α1) > [r with macron]c(C8P; α1), but [r with macron]c(C8A; α2) < [r with macron]c(C8P; α2) (Figure 5). Since the sequence of the MHC α helices is organized in such a way that the N-terminus of one helix is close in space to the C-terminus of the other helix and vice versa (Figure 1), what appears as shifts in opposite directions along the sequence actually corresponds to a shift of the contact footprints in the same direction in space. Specifically, the shifts indicate that for C8P, more TCR contacts occur on the part of the MHC near the N-terminus of the peptide (the left end in Figure 1B), while for C8A, more TCR contacts occur on the part of the MHC near the C-terminus of the peptide (the right end in Figure 1B). The more space available to 8-mer peptides may allow this “rocking” flexibility, as is discussed in the following section.

From the diagram of the footprint in Figure 1B, it is clear that the Vα chain of the TCR is oriented closer to the C-terminal end of the α2 MHC helix (on the right in the figure), and the Vβ chain of the TCR is oriented closer to the C-terminal end of the α1 helix (on the left in the figure). Combining this structural knowledge with the contact data discussed above, we expect that C8P would contain more MHC contacts with Vα, and C8A would contain more MHC-Vβ contacts. In other words, the change from C8P to C8A would increase the ratio of Vβ to Vα contacts. Table 3 shows that this is the case. In C8P, Ld makes 27 contacts with Vα and 21 contacts with Vβ, while in C8A, Ld makes only 15 contacts with Vα and 23 contacts with Vβ. The changes result in a shift of the Vβ/Vα ratio from 0.79 to 1.46 as C8P changes to C8A. Figure 2 provides a structural explanation of the drop in Vα contacts, showing that in C8A, the CDR1α loop tends away from the MHC, unlike in C8P.

Table 3
Peptide mutation can affect the ratio of Vβ/Vα contacts with pMHC

These results demonstrate how the mutation of the fifth position of the shortened peptide from a proline to an alanine (C8P → C8A) can cause rearrangement of the contact topology, affecting not only which MHC residues come into contact with TCR, but also which TCR residues contact the MHC.

Mutations to the 9-mer peptide produce less contact rearrangement than mutations to the 8-mer peptide

By looking at all four peptide variants together, we can compare the relative impact of the mutation of the Ala6/5 residue on the footprint in both the 9-mer and 8-mer peptides. When we compare Δ[r with macron]c(C8A–C8P; {α1, α2}) to Δ[r with macron]c(C9A–C9P; {α1, α2}), it is clear that the footprint is significantly more sensitive to mutation at the Ala5 position of the 8-mer peptide than it is to mutation at the corresponding Ala6 position of the 9-mer peptide. While mutation of the 9-mer peptide does change contacts throughout the footprint, these changes do not affect the values of [r with macron]c(C9P; {α1, α2})/[r with macron]c(C9A; {α1, α2}) more than inherent differences between crystal structure and simulation. Thus, although they may be due to underlying differences in the contact interface, these differences are not large enough to be seen as significant.

The shorter peptide evidently allows more freedom for the α helices—and thereby the entire footprint—to rearrange in response to a mutation in the center of the peptide. Despite such rearrangement, the mutation seems to have a similar effect on TCR binding in both the 8-mer and 9-mer peptide systems, lowering the TCR/pMHC affinity by two orders of magnitude (Table 1). This lack of correlation between binding affinity changes and structural rearrangement may seem counterintuitive, but it is in fact consistent with other results that describe sets of nearly identical TCR/pMHC crystal structures with highly varying TCR binding affinities.34

The structural distinction that the footprint makes between mutations to the shortened and full-length peptide suggests that TCR and allo-pMHC have a complicated interaction landscape at their interface and therefore can respond to subtle changes in unique ways.

Discussion

The intent of our study has been to provide insight into the structural impact of peptide mutations on the alloreactive TCR/pMHC interface by comparing atomistic models of systems that have not yet been crystallized. In our simulations, mutation of the TCR’s CDR3α loop did not induce significant rearrangement of the TCR/MHC contacts, but certain mutations of the peptide made noticeable changes to local contacts throughout the TCR/pMHC interface.

Our results contribute to a large body of work aimed at elucidating the nature of the 2C TCR alloreactive response in particular and alloreactive responses in general. That work has yielded two differing pictures of the peptide’s structural influence on the TCR/pMHC footprint: a peptide-centric model,22,30,31,37 in which the peptide plays an important role in determining the footprint, and an MHC-centric model,20,34,44,45 in which the footprint is determined by the MHC, with the peptide having virtually no effect. These views are sometimes simplified as a “tail wagging the dog/dog wagging the tail” debate, with the dog being the MHC and the tail being the peptide. Some experimental results, such as the MHC mutation experiments of Felix et al.30 and the alanine-substitution work of Conolly et al.,22,31 have been cited as support for the peptide-centric idea. Other experiments, specifically the work of Colf et al.,20 have been explained as support for the MHC-centric model.

Our findings serve as a synthesis and explanation of these apparently conflicting views. We show that, while CDR3 mutations do not necessarily influence the alloreactive contact footprint, peptide mutation, by contrast, can have an influence. The difference between the observed effects of these two kinds of mutations highlights the difference between influencing the TCR/peptide interactions by changing only the TCR’s CDR3 residues and mutating the peptide itself, which is a more direct way to test the peptide-centric model. Our simulations reveal that the two experimental approaches are not equivalent, as mutations to the peptide can directly impact the MHC helices and thus affect the chemical surface to which the TCR binds in ways that mutations of the CDR3 loops cannot. However, even in the cases where we found that peptide mutations broadly affected the footprint, these effects were relatively localized compared to the difference between self- and allo-MHC footprints made by the 2C TCR.20

Conclusions

We thus predict that both MHC and peptide have a direct impact in determining the TCR/allo-pMHC footprint, with the peptide being able to “edit” the specific contacts after initial contact between the TCR and pMHC. One mechanism for this multilevel impact may involve the TCR “reading” the pMHC ligand in search of enough sufficiently favorable interactions. If enough interactions exist and the TCR/pMHC complex survives long enough, modest, specific rearrangements of the interface residues would occur to maximize the affinity.8,29 This mechanism is consistent with our findings and is also consistent with diverse other findings, including those of MacDonald et al.,47 because rearrangements after the initial scanning induced by the peptide could produce a similar final structure in initially different pMHC landscapes.

Looking forward, we believe that free energy simulations of TCR/allo-pMHC similar to those previously performed for TCR/self-pMHC5355 would be informative, providing more detailed free energetic data, such as a systematic comparison of the relative contributions of peptide mutation and MHC changes on the TCR binding affinity. Most importantly, testing the predictions made by our simulations by obtaining crystal structures of the C8P and C8A systems (or some analogues) and performing analysis similar to our own with these structures will shed more light on the interplay between peptide and MHC on alloreactivity. We hope that our study will motivate such experiments.

Supplementary Material

1_si_001

Acknowledgments

We thank Prof. Martin Karplus, Prof. Herman Eisen, and Dr. Abhishek Jha for their fruitful discussions and comments.

Financial support was provided by NIAID grant 5P01/AI071195/02 and an NIH Director’s Pioneer Award to A.K.C.

Footnotes

Supporting Information Available

Quantitative information about the stability of the molecular dynamics situations and the effects of peptide mutation. This material is available free of charge via the Internet at http://pubs.acs.org/.

References

1. Hogquist KA, Jameson SC, Bevan MJ. Curr Opin Immunol. 1994;6:273. [PubMed]
2. Jameson SC, Hogquist KA, Bevan MJ. Annu Rev Immunol. 1995;13:93. [PubMed]
3. Werlen G, Hausmann B, Naeher D, Palmer E. Science. 2003;299:1859. [PubMed]
4. Starr TK, Jameson SC, Hogquist KA. Annu Rev Immunol. 2003;21:139. [PubMed]
5. von Boehmer H, Aifantis I, Gounari F, Azogui O, Haughn L, Apostolou I, Jaeckel E, Grassi F, Klein L. Immunol Rev. 2003;191:62. [PubMed]
6. Hogquist KA, Baldwin TA, Jameson SC. Nat Rev Immunol. 2005;5:772. [PubMed]
7. Siggs OM, Makaroff LE, Liston A. Curr Opin Immunol. 2006;18:175. [PubMed]
8. Košmrlj A, Jha AK, Huseby ES, Kardar M, Chakraborty AK. Proc Natl Acad Sci USA. 2008;105:16671. [PubMed]
9. Huseby E, White J, Crawford F, Vass T, Becker D, Pinilla C, Marrack P, Kappler J. Cell. 2005;122:247. [PubMed]
10. Huseby E, Crawford F, White J, Marrack P, Kappler J. Nat Immunol. 2006;7:1191. [PubMed]
11. Lindahl KF, Wilson DB. J Exp Med. 1977;145:508. [PMC free article] [PubMed]
12. Lynes MA, Flaherty L, Michaelson J, Collins JJ, Rinchik EM. J Immunogenet. 1984;11:189. [PubMed]
13. Lechler RI, Lombardi G, Batchelor JR, Reinsmoen N, Bach FH. Immunol Tod. 1990;11:83. [PubMed]
14. Kaufman CL, Gaines BA, Ildstad ST. Annu Rev Immunol. 1995;13:339. [PubMed]
15. Joyce S, Nathenson SG. Immunol Rev. 1996;154:59. [PubMed]
16. Le NT, Chen BJ, Chao NJ. Cytotherapy. 2005;7:126. [PubMed]
17. Hauben E, Bacchetta R, Roncarolo MG. Cytotherapy. 2005;7:158. [PubMed]
18. Felix NJ, Allen PM. Nat Rev Immunol. 2007;7:942. [PubMed]
19. Nikolich-Zugich J. Nat Immunol. 2007;8:388. [PubMed]
20. Colf LA, Bankovich AJ, Hanick NA, Bowerman NA, Jones LL, Kranz DM, Garcia KC. Cell. 2007;129:135. [PubMed]
21. DeLano W. The PyMOL User’s Manual. 2002.
22. Alexander-Miller MA, Burke K, Koszinowski UH, Hansen TH, Connolly JM. J Immunol. 1993;151:1. [PubMed]
23. Basu D, Horvath S, Matsumoto I, Fremont DH, Allen PM. J Immunol. 2000;164:5788. [PubMed]
24. Eisen HN. Annu Rev Immunol. 2001;19:1. [PubMed]
25. Huseby ES, Crawford F, White J, Kappler J, Marrack P. Proc Natl Acad Sci USA. 2003;100:11565. [PubMed]
26. Panina-Bordignon P, Corradin G, Roosnek E, Sette A, Lanzavecchia A. Science. 1991;252:1548. [PubMed]
27. Udaka K, Tsomides TJ, Eisen HN. Cell. 1992;69:989. [PubMed]
28. Garcia K, Adams E. Cell. 2005;122:333. [PubMed]
29. Armstrong KM, Piepenbrink KH, Baker BM. Biochem J. 2008;415:183. [PMC free article] [PubMed]
30. Felix NJ, Donermeyer DL, Horvath S, Walters JJ, Gross ML, Suri A, Allen PM. Nat Immunol. 2007;8:388. [PubMed]
31. Hornell TM, Martin SM, Myers NB, Connolly JM. J Immunol. 2001;167:4207. [PubMed]
32. Garboczi D, Ghosh P, Utz U, Fan Q, Biddison W, Wiley D. Nature. 1996;384:134. [PubMed]
33. Garcia KC, Degano M, Pease LR, Huang M, Peterson PA, Teyton L, Wilson IA. Science. 1998;279:1166. [PubMed]
34. Ding Y, Baker B, Garboczi D, Biddison W, Wiley D. Immunity. 1999;11:45. [PubMed]
35. Degano M, Garcia K, Apostolopoulos V, Rudolph M, Teyton L, Wilson I. Immunity. 2000;12:251. [PubMed]
36. Hahn M, Nicholson MJ, Pyrdol J, Wucherpfennig KW. Nat Immunol. 2005;6:490. [PMC free article] [PubMed]
37. Borbulevych OY, Piepenbrink KH, Gloor BE, Scott DR, Sommese RF, Cole DK, Sewell AK, Baker BM. Immunity. 2009;31:885. [PMC free article] [PubMed]
38. Burrows S, Chen Z, Archbold J, Tynan F, Beddoe T, Kjer-Nielsen L, Miles J, Khanna R, Moss D, Liu Y, Gras S, Kostenko L, Brennan RM, Clements CS, Brooks AG, Purcell AW, McCluskey J, Rossjohn J. Proc Natl Acad Sci USA. 2010;107:10608. [PubMed]
39. Reiser JB, Darnault C, Guimezanes A, Gregoire C, Mosser T, Schmitt-Verhulst AM, Fontecilla-Camps JC, Malissen B, Housset D, Mazza G. Nat Immunol. 2000;1:291. [PubMed]
40. Luz J, Huang M, Garcia K, Rudolph M, Apostolopoulos V, Teyton L, Wilson I. J Exp Med. 2002;195:1175. [PMC free article] [PubMed]
41. Reiser JB, Grégoire C, Darnault C, Mosser T, Guimezanes A, Schmitt-Verhulst AM, Fontecilla-Camps JC, Mazza G, Malissen B, Housset D. Immunity. 2002;16:345. [PubMed]
42. Reiser J, Darnault C, Grégoire C, Mosser T, Mazza G, Kearney A, van der Merwe P, Fontecilla-Camps J, Housset D, Malissen B. Nat Immunol. 2003;4:241. [PubMed]
43. Archbold J, Macdonald W, Miles J, Brennan R, Kjer-Nielsen L, McCluskey J, Burrows S, Rossjohn J. J Biol Chem. 2006;281:34324. [PubMed]
44. Jones LL, Colf LA, Bankovich AJ, Stone JD, Gao YG, Chan CM, Huang RH, Garcia KC, Kranz DM. Biochemistry. 2008;47:12398. [PMC free article] [PubMed]
45. Jones L, Colf L, Stone J, Garcia K, Kranz D. J Immunol. 2008;181:6255. [PMC free article] [PubMed]
46. Gras S, Burrows SR, Kjer-Nielsen L, Clements CS, Liu YC, Sullivan LC, Bell MJ, Brooks AG, Purcell AW, McCluskey J, Rossjohn J. Immunity. 2009;30:193. [PubMed]
47. MacDonald WA, Chen Z, Gras S, Archbold JK, Tynan FE, Clements CS, Bharadwaj M, Kjer-Nielsen L, Saunders PM, Wilce MCJ, Crawford F, Stadinsky B, Jackson D, Brooks AG, Purcell AW, Kappler JW, Burrows SR, Rossjohn J, McCluskey J. Immunity. 2009;31:897. [PubMed]
48. Karplus M, McCammon J. Nat Struct Biol. 2002;9:646. [PubMed]
49. Wu L, Tuot D, Lyons D, Garcia K, Davis M. Nature. 2002;418:552. [PubMed]
50. Kosmrlj A, Chakraborty AK, Kardar M, Shakhnovich E. Phys Rev Lett. 2009;103:068103. [PubMed]
51. Eisen H, Chakraborty A. Proc Natl Acad Sci USA. 2010;107:22373. [PubMed]
52. Brünger A. Prot Struct Func Genet. 1988;4:148. [PubMed]
53. Michielin O, Karplus M. J Mol Biol. 2002;324:547. [PubMed]
54. Wan SZ, Coveney PV, Flower DR. J Immunol. 2005;175:1715. [PubMed]
55. Zoete V, Michielin O. Prot Struct Func Bioinf. 2007;67:1026. [PubMed]
56. MacKerell AD, Jr, Bashford D, Bellott Dunbrack LR, Evanseck JD, Field MJ, Fischer S, Gao J, Guo H, Ha S, Joseph-McCarthy D, Kuchnir L, Kuczera K, Lau FTK, Mattos C, Michnick S, Ngo T, Nguyen DT, Prodhom B, Reiher WE, Roux B, Schlenkrich M, Smith JC, Stote R, Straub J, Watanabe M, Wiorkiewicz-Kuczera J, Yin D, Karplus M. J Phys Chem B. 1998;102:3586.
57. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. J Chem Phys. 1983;79:926.
58. Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan S, Karplus M. J Comput Chem. 1983;4:187.
59. Brooks BR, Brooks CL, III, Mackerell AD, Jr, Nilsson L, Petrella RJ, Roux B, Won Y, Archontis G, Bartels C, Boresch S, Caflisch A, Caves L, Cui Q, Dinner AR, Feig M, Fischer S, Gao J, Hodoscek M, Im W, Kuczera K, Lazaridis T, Ma J, Ovchinnikov V, Paci E, Pastor RW, Post CB, Pu JZ, Schaefer M, Tidor B, Venable RM, Woodcock HL, Wu X, Yang W, York DM, Karplus M. J Comput Chem. 2009;30:1545. [PMC free article] [PubMed]
60. Mackerell A, Jr, Feig M, Brooks C., III J Comput Chem. 2004;25:1400. [PubMed]
61. van Gunsteren WF, Berendsen HJC. Mol Phys. 1977;34:1311.
62. Steinbach P, Brooks B. J Comput Chem. 1994;15:667.
63. Essmann U, Perera L, Berkowitz M, Darden T, Lee H, Pedersen L. J Chem Phys. 1995;103:8577.
64. Nosé S. J Chem Phys. 1984;81:511.
65. Hoover WG. Phys Rev A. 1985;31:1695. [PubMed]
66. Feller S, Zhang Y, Pastor R, Brooks B. J Chem Phys. 1995;103:4613.
67. Eisen HN, Sykulev Y, Tsomides TJ. Adv Prot Chem. 1996;49:1. [PubMed]
68. Kageyama S, Tsomides TJ, Fukusen N, Papayannopoulos I, Eisen HN, Sykulev Y. J Immunol. 2001;166:3028. [PubMed]
69. Kranz DM, Sherman DH, Sitkovsky MV, Pasternack MS, Eisen HN. Proc Natl Acad Sci USA. 1984;81:573. [PubMed]
70. Sha WC, Nelson CA, Newberry RD, Kranz DM, Russell JH, Loh DY. Nature. 1988;336:73. [PubMed]
71. Humphrey W, Dalke A, Schulten K. J Mol Graph. 1996;14:33. [PubMed]