The interaction between T cell receptors (TCRs) and the peptide-presenting major histocompatibility complex (pMHC) molecule is a nanometer-scale recognition event with macroscopic consequences for health and control of disease. In αβ T cells, the six complementarity-determining region (CDR) loops of the TCR are responsible for scanning the pMHC surface and determining the appropriate immunological response. Since the first crystal structures of a TCR-pMHC complex were solved in 1996 (Garcia et al., 1996; Garboczi et al., 1996), general ‘rules of engagement’ between TCR and pMHC have been identified and reliably reproduced in subsequent structures (Garcia et al., 1999; Yin et al., 2012). First, the peptide is presented in the ‘groove’ formed by the two α-helices of either the class I or class II MHC molecule (Bjorkman et al., 1987; Brown et al., 1993). The TCR then scans this peptide primarily using the highly diverse CDR3 loops of the TCRα and TCRβ chains, whose sequences are determined by V(D)J recombination (Wu et al., 2002; Jung and Alt, 2004). The sequences of the remaining two CDR1 and CDR2 loops of the α and β chains are entirely determined by germline TRAV and TRBV genes, respectively. In a majority of structures, these CDR1 and CDR2 loops make a number of key contacts with well-defined regions of the likewise germline-encoded MHC $\alpha$-helices (Gras et al., 2012).

Importantly, previous work has found a large list of deviations from these classic ‘rules of engagement’. The hypervariable CDR3 residues can dominate the interactions with the germline-encoded MHC α-helices (Piepenbrink et al., 2013; Singh et al., 2022), germline-encoded CDRs can interact strongly with peptide (Piepenbrink et al., 2013), an exceptionally long peptide can ‘bulge’ causing separation between germline-encoded regions (Tynan et al., 2005), or the docking orientation can even be entirely reversed (Beringer et al., 2015; Gras et al., 2016; Zareie et al., 2021). When considering the large number of possible TCR-pMHC interactions and the substantial crossreactivity of TCRs (Riley et al., 2018; Sewell, 2012), these historically non-canonical interactions may in fact be rather common. Despite this great variability inherent in TCR-pMHC complex formation, the docking orientation of the TCR over MHC molecules (Figure 1A and B) falls within a tight distribution of angles for the majority of structures solved thus far (Figure 1C and D) nearly independent of the strong variation in the CDR3 and peptide sequences involved. This suggests that interactions between germline-encoded regions play fundamentally important roles.

Figure 1

Download asset Open asset

However, it remains unclear whether germline-encoded residues enforce the orientational preference highlighted in Figure 1, or if this preference is a consequence of thymic selection (Van Laethem et al., 2007; Tikhonova et al., 2012; Van Laethem et al., 2012), whereby T cells are first positively selected on TCR recognition of MHC molecules and later negatively selected against autoreactivity towards MHC molecules loaded with organismal self-peptides (Blackman et al., 1990; Juang et al., 2010; Germain, 2002). This selection occurs in the presence of the potentially orientation-determining co-receptors CD4 or CD8. However, evidence for this hypothesis of docking determined by selection comes largely from a study utilizing a knockout system that deviates strongly from the physiological norm (Van Laethem et al., 2007), or from the identification of rare ‘reverse-docking’ TCRs (Beringer et al., 2015; Gras et al., 2016; Zareie et al., 2021). Alternatively, the extent of this orientational preference may reflect a co-evolution of interacting, germline-encoded regions of TCR and MHC molecules. In support of this hypothesis, a number of mouse studies have identified key residue pairs critical for TCR-pMHC binding which were later structurally validated and found in other, distantly related organisms (Scott-Browne et al., 2009; Krovi et al., 2019; Feng et al., 2007; Blevins et al., 2016; Scott-Browne et al., 2011; Dai et al., 2008). While these studies find evidence of conserved interactions for a small subset of alleles, no work thus far has found consistent conservation across the full range of possible TCR-MHC molecule interactions.

These observations underscore the need for a more thorough understanding of the fundamental role of CDR1 and CDR2 in the recognition of pMHC complexes, either as sculptors of the TCR-pMHC interface, or as opportunistic reinforcements to a protein-protein interaction largely determined by the CDR3 loops and peptide involved. Due in part to the concentration of diversity at the CDR3-peptide interface, a majority of studies characterizing TCR-pMHC interactions have focused on describing immune recognition events in the context of this interface (Birnbaum et al., 2014; Sharon et al., 2016; Sibener et al., 2018; Krovi et al., 2019; Burrows et al., 2010; Ishigaki et al., 2022). This runs counter to the relative contributions of the germline-encoded regions that account for a majority (75–80%) of the binding interface (Garcia et al., 2009). The progress made thus far can largely be attributed to X-ray crystallography, the gold standard for the study of TCR-pMHC interactions. However, this approach is severely limited by its low-throughput nature, calling its utility for this particular combinatorial problem into question. While a majority of the 45 productive TCRα-encoding TRAV and 48 productive TCRβ-encoding TRBV genes in humans have been crystallized at least once in complex with pMHC (Gowthaman and Pierce, 2019), the total of 149 crystallized human TCR-pMHC complexes pales in comparison to the 2160 possible productive TRAV-TRBV pairings. When we further consider the nearly 900 known unique amino acid sequences of class I MHC in humans (also referred to as class I human leukocyte antigen - HLA), we find that structural studies to date have covered less than 0.01% of the total possible TCR-MHC germline interaction space.

While structural modeling is an intriguing alternative (Milighetti et al., 2021), predictions of the conformations adopted by flexible TCR or antibody CDR loops and the placement of the antigen-contacting amino acid side chains on these loops have been found to be somewhat unreliable (Evans et al., 2022). In the work reported here, we explore how much insight about TCR-MHC interactions can be gained using a simplified, pseudo-structural representation of the biophysics of protein interactions that scores amino acid interactions as either productive, neutral, or counter-productive with regard to complex formation. Using the Automated Immune Molecule Separator (AIMS) software (Boughter et al., 2020; Boughter and Meier-Schellersheim, 2023) we generated the first systematic characterization of every germline-encoded TRAV-MHC and TRBV-MHC interaction through such a pseudo-structural approach. The AIMS architecture allows us to analyze the CDR1 and CDR2 contributions to interactions with MHC molecules independent of CDR3, providing a not previously explored antigen-independent perspective on the TCR-MHC complex. We find that the great diversity present in the CDR1 and CDR2 loops encoded by TRAV and TRBV genes in humans, as well as the dynamic nature of amino acid side chains, makes strongly conserved interactions between TCR germline-encoded residues and MHC $\alpha$-helices highly unlikely. This suggests that the few strongly evolutionarily conserved interactions that have been identified previously in the literature should be considered exceptions and not the rule. Importantly, however, analyzing the biophysical compatibility for each germline-encoded CDR residue and those on the MHC α-helices, we find evidence of an evolutionary conserved region of permissible binding on the MHC molecular surface responsible for the observed canonical orientation of the TCR-MHC molecular complex. These measures of biophysical compatibility between germline CDRs and MHC α-helices can be used as a basis for understanding TCR-pMHC interactions, and as a tool for understanding how perturbations to these regions alter these interactions.

Starting from the TRAV and TRBV protein entries from the ImMunoGeneTics (IMGT) database (Brochet et al., 2008; Lefranc, 2014) and the IMGT-HLA database containing all identified HLA alleles (Mack et al., 2013), we include only productive genes and unique amino acid sequences in our final analysis dataset. This final dataset included a total of 882 unique class I MHC sequences (HLA-A, B, and C), 44 class II α chain sequences (HLA-DQα, DRα, and DPα), and 431 class II β chain sequences (HLA-DQβ, DRβ, and DPβ). Due to significant differences in the alignments and more nuanced differences in the structures of these MHC molecules, we analyzed these three groups separately and compare the results as appropriate.

Polymorphic HLA molecules are far less diverse than human TRAV and TRBV genes

We first searched for potentially conserved interactions between TCR and MHC molecules that would suggest a strong coevolution between the two. Such biases in the amino acid usage of TCR and MHC molecules at specific sites can occur either in the form of full amino acid conservation, or as pairs of meaningfully co-varying amino acids found across multiple diverse sites. Without the structural data necessary to precisely identify these conserved interactions, we can characterize the diversity present in TRAV, TRBV, and HLA amino acid sequences using bioinformatic approaches and determine whether a conserved germline-encoded interaction is possible for each unique molecular combination. We quantified the diversity in these molecular subsets using concepts from information theory. Specifically, the amino acid diversity of each TRAV, TRBV, and HLA sequence encoded into an alignment matrix (Figure 2A) was quantified by calculating the Shannon entropy (Shannon, 1948) for each individual dataset (Figure 2B). The alignment matrix of Figure 2A provides the basis for all of our pseudo-structural analyses throughout this manuscript. The incorporated structural information goes as far as picking out the CDR loops of each TRAV/TRBV sequence and $\alpha$-helices of each MHC sequence, but does not make explicit assumptions about relative structural orientations.

Figure 2 with 4 supplements see all

Download asset Open asset

Figure 2B indicates that the TRAV and TRBV genes are very diverse, even more so than the polymorphic HLA alleles. Despite the low number of TRAV and TRBV genes significant diversity persists in CDR1 and CDR2, suggesting limited conservation in those residues most likely to contact the MHC surface. Further, most of the entropy in MHC class I alleles is concentrated in the peptide-binding region. In this more diverse region the entropy tops out at around 2 bits, which corresponds to at most 4 equiprobable amino acids, or a distribution of a slightly larger number of amino acids strongly skewed towards one particular amino acid. In the α-helical regions that are in contact with the germline-encoded CDR loops, the class I entropy reaches a maximum of 1.5 bits, or at most 3 equiprobable amino acids. These same trends persist for the analysis of class II HLA alleles, with sequence diversity more limited in the class II α-helices when compared to the germline-encoded CDR loops (Figure 2C and D).

HLA alleles are polymorphic, which can be misconstrued as having high sequence diversity in MHC molecules across the human population. However, the polymorphic sites are limited to a minority of the total variable positions within the MHC nucleotide sequences (Robinson et al., 2017), which suggests even more limited variability at the amino acid level. While there are many unique alleles, these alleles are largely similar, especially when compared to other, more diverse protein families that are strongly structurally conserved but sequentially divergent, as in the neuronal Dpr-DIP system in Drosophila melanogaster (Nandigrami et al., 2022). Conversely, despite the low number of TCR germline genes present in humans, the diversity is substantial, with multiple sites across the CDR loops harboring over 10 unique amino acids. This mismatched diversity between the TCR germline-encoded CDR loops and the MHC α-helices suggests that conserved germline-encoded interactions are unlikely to exist for every possible molecular combination, either within a single individual or across the human population more broadly. Mutual information calculations (see Methods) corroborate these observations by quantifying this lack of co-variation of germline-encoded TCR and MHC residues in both humans (Figure 2—figure supplement 1) and in a wider range of mammals (Figure 2—figure supplements 2 and 3). If we limit these mutual information calculations solely to those TCR-MHC pairs that have been crystallized in complex, we see this lack of co-variation persists (Figure 2—figure supplement 4).

Biophysical diversity in TRAV and TRBV genes makes strongly conserved contacts across HLA molecules unlikely

The diversity in TRAV and TRBV germline-encoded regions that we had found excluded the possibility of conserved contacts as key mediators across all TCR-MHC germline-encoded interactions. However, this did not rule out the possibility that within this diversity, there are conserved biophysical interactions. For example, a site on CDR2β could be highly diverse, but only allow expression of hydrophilic residues of neutral charge. Pairing this with an α-helical site on HLA molecules which likewise only permits hydrophilic residues would represent a likely site of conserved interaction. We can test for such relations by quantifying the position-sensitive amino acid biophysical properties for both the germline-encoded TCR (Figure 3A) and HLA (Figure 3B) residues. Looking at both the position-sensitive charge (Figure 3) and the position-sensitive hydropathy (Figure 3—figure supplement 1), we see that the substantial diversity in the TRAV and TRBV sequences corresponds to a similar biophysical diversity in the amino acids found across the germline-encoded CDR loops.

Figure 3 with 2 supplements see all

Download asset Open asset

To put these results into perspective, we compared them to corresponding variabilities found for another class of molecules that interact with the MHC. The killer cell immunoglobulin-like receptors (KIRs) largely recognize the MHC class I subset HLA-C, subsequently relaying either inhibitory or stimulatory signals to the natural killer cell expressing these KIRs (Thielens et al., 2012). While each individual possesses a fixed subset of KIRs, with no analogous diversity introduced by a V(D)J-like process, polymorphisms in these KIRs exist across the human population (Robinson et al., 2010). The interactions between these KIRs and HLA-C molecules are more well-defined than the germline-encoded interactions between TCR and MHC molecules, with strong salt bridge and hydrogen bond networks in place between two immunoglobulin-like domains of the KIR and each α-helix of HLA-C (Figure 3—figure supplement 2A, B; Moradi et al., 2015; Moradi et al., 2021).

We found that from an initial dataset of 369 KIR polymorphisms compiled from the IPD-KIR database (Robinson et al., 2010), only 31 productive alleles have unique amino acid sequences in the MHC molecule-binding region with very limited diversity (Figure 3—figure supplement 2C, D). Further, within the few moderately diverse sites, we found almost no variation in the biophysical properties of the amino acids contacting the MHC molecule, or in those of the amino acids on the HLA-C α-helices (Figure 3C and D). We further contextualized these findings with specific examples from KIR-MHC and TCR-MHC molecular interactions (Figure 3E and F). We selected a key hydrogen bonding network in the KIR2DL2-HLA-C*07:02 interface (Moradi et al., 2021) and compared this to the evolutionarily conserved YXY motif of CDR2β (Feng et al., 2007; Scott-Browne et al., 2011) found in a small subset of human TRBV genes. From this, we can see how diversity in each region reinforces or disrupts potentially conserved interactions at each site. Across all KIRs tested, the interacting residues are fully conserved, suggesting diversity is not tolerated at these sites. Similarly, diversity is exceptionally limited at the interacting sites on the HLA-C α2 helix.

Contrasting this with the regions involved in the YXY motif interactions, we see much higher diversity across both the TCR and MHC molecules. Considering specifically the conserved lysine on the α1 helix, we see that some of the residues utilized by TRBV-encoded CDR2 would create a strong conflict with the proposed conserved interaction site. While the YXY motif present on a few TRBV genes is clearly evolutionarily conserved (Scott-Browne et al., 2011) it is evident that substitutions either on CDR2β or on the MHC α1 helix would disrupt this interaction. These results strongly suggest that conserved interactions between germline-encoded CDR loops and MHC α-helices are deviations from the norm, and that CDR1 and CDR2 loops of TCRα and TCRβ likely adopt a more opportunistic binding strategy.

Calculation of TCR-MHC interaction potentials finds strong differences in germline-encoded recognition strategies

While our information-theoretic and biophysical analyses suggested limited conserved contacts between TCR α/β CDR1 and CDR2 loops and MHC helices, they did not rule out coevolution at the level of biophysical compatibility over a broader interaction region. To explore the possibility of such coevolution that does not manifest itself at the level of genetic correlations, we analyzed TCR-MHC compatibility using the previously-validated (Nandigrami et al., 2022) AIMS interaction potential. Using a simplified potential to estimate protein interaction propensities, analyses performed with AIMS contain few implicit assumptions and compare favorably with more detailed and computationally expensive models. In a binary classification of a large database of structurally similar protein complexes, a combination of the AIMS interaction potential and a linear discriminant analysis-based classifier was capable of distinguishing binders and non-binders to an accuracy of 80%, whereas calculations run on over 45 µs of simulated all-atom trajectories could only distinguish to an accuracy of 50%. While not as biophysically descriptive as methods such as alchemical free energy perturbation (Gumbart et al., 2013b) or potential of mean force-based calculations (Gumbart et al., 2013a), the AIMS interaction potential generates accurate predictions for problems that are intractable with current limitations to computation due to the number of protein complexes of interest.

These interaction scores are calculated between all possible interacting amino acid sequences of HLA alleles and TRAV or TRBV genes (i.e. without preference for a canonical binding orientation), providing a useful quantification of the relative contribution of germline-encoded CDR1 and CDR2 loops to a given TCR-MHC complex (See Methods). Figure 4 shows the results of this scoring metric, focusing on the resultant patterns evident in both the x-axis, corresponding to the TRBV (Figure 4A) or TRAV (Figure 4B) genes, and the y-axis, corresponding to each HLA allele. We note that the variance in the strength of interaction with MHC class I is greater for the CDR1 and CDR2 loops of TCRα than those of TCRβ (Figure 4—figure supplement 1). Our analysis showed that TRAV38-1, 38–2, and 19 have the strongest potential interactions with class I HLA molecules of any TCR genes. Conversely, the TRBV genes 7–2 and 7–3 are predicted to have a severely limited interaction potential with HLA molecules (Figure 4C and D). Interestingly, these alleles, TRBV7-2 and 7–3, are of significant interest in autoimmunity as they have been associated with celiac disease (Gunnarsen et al., 2017; Qiao et al., 2014) and multiple sclerosis (Sethi et al., 2013), respectively. Similarly, those TRBV alleles most strongly enriched in celiac patients (7–2, 20–1, and 29–1) (Qiao et al., 2014) likewise all have a corresponding low AIMS interaction potential with class I and class II HLA molecules (Figure 4—figure supplement 2).

Figure 4 with 2 supplements see all

Download asset Open asset

These interaction potentials suggest that not only do germline-encoded CDR loops not contact the same conserved regions of the HLA molecule, but further imply that different TRAV/TRBV gene usage encourages differential utilization of the CDR loops. For instance, TRBV6-5 has a high CDR1 interaction potential and a negligible CDR2 interaction potential. Conversely, the suggested CDR bias is reversed for TRBV11-2. These trends are similarly found for analysis of germline-encoded interactions with HLA class II molecules (Figure 4—figure supplement 2). These results further highlight the variability across the full repertoire of germline-encoded CDR1 and CDR2 loops and the role of this variability in interactions with MHC molecules.

Interaction scores show good agreement with structural data

To test our scoring of CDR-MHC interactions, we analyzed all TCR-pMHC complexes crystallized thus far. Using the TCR3D Database (Gowthaman and Pierce, 2019), we were able to process 182 total human TCR-pMHC complexes and quantify germline-encoded contacts. Final interactions included in the analysis were defined using somewhat generous cutoffs of 6 Å for charged interactions and 4.5 Å for van der Waals interactions (Methods). This comparison shows good agreement between our bioinformatic results and structural analyses (Figure 5). The results show that a normalized version of our AIMS interaction potential (Figure 5A) is in qualitative agreement with a symmetrized version of the count matrix generated from crystal structure contacts between TCR and class I MHC molecule side chains (Figure 5B). Normalization of these matrices is necessary, as residues with a negative or zero interaction potential are indistinguishable in the empirical contact map; that is, a negative number of contacts are not possible. Non-normalized versions of these matrices can be found in Figure 5—figure supplement 1A, B. The largest discrepancies between the interaction potential and the empirical matrix occurs in regions of hydrophobic interactions, as strongly hydrophobic residues are relatively rare on the MHC α-helices and TCR CDR loops (Figure 5—figure supplement 1C). Further, biases in the PDB from repeated crystallization of identical molecular species are evident, particularly in Ala-Tyr and Gln-Tyr MHC-TCR interactions (Figure 5—figure supplement 1D).

Figure 5 with 3 supplements see all

Download asset Open asset

While the AIMS interaction potential only scores possible contacts between amino acid sidechains, there exists evidence in the literature that interactions with backbone atoms are often critical for overall complex stability (Huseby et al., 2006). However, incorporation of such information for a given interaction requires extensive knowledge of the structural details of the complex of interest, which as discussed previously, is not available for all possible TRAV/TRBV pairings. Quantification of sidechain-backbone (Figure 5—figure supplement 2A, B), backone-sidechain (Figure 5—figure supplement 2C, D), and backbone-backbone (Figure 5—figure supplement 2E, F) contacts across TCR-pMHC complexes show some intuitive results, such as increased involvement of alanine and glycine, but other results are clearly counter-intuitive, or biased by the structures crystallized thus far. For instance, we shouldn’t expect a priori that a Tyr-Gln interaction would be a common backbone-sidechain or sidechain-sidechain interaction, as both of these sidechains are relatively bulky. While the ‘biases’ introduced by the structures solved thus far may not be biases at all, and could in fact be signatures of TCR recognition, it is difficult to decouple the signal from the noise without more structural information.

Despite the challenges associated with aligning model predictions with structural data, AIMS interaction scores are useful to identify those germline TRAV- and TRBV-encoded CDR loops that may make more frequent contacts with the MHC α-helices (Figure 5C and D). We find significant increases in the number of CDR1/2 sidechain contacts with both MHC helix sidechain and backbone atoms in TRAV sequences predicted by the AIMS interaction potential to be stronger interaction partners. While a similar agreement between interaction potential and sidechain contacts is seen for TRBV, these differences are not statistically significant. This effect is less pronounced for TCR backbone atom interactions with MHC sidechain and backbone atoms, as one would expect given that these interactions are not accounted for in the AIMS interaction potential. Interestingly, these interaction potentials show better experimental agreement and predictive power for TRAV-encoded sequences compared to TRBV-encoded sequences. This could be due either to a fundamental difference in how TCRα and TCRβ contact the MHC α-helices, or in part due to the aforementioned higher interaction potential variance for TRAV-encoded CDR loops (Figure 4—figure supplement 1).

However, there are still clear deviations from the predictions we can achieve at this point. In these structures, it appears that the CDR3-peptide interactions dominate the interface, overruling all other trends in germline-encoded interactions. We can consider an extreme example, looking at the germline-encoded and peptide contacts of a so-called ‘super-bulged’ peptide bound to MHC class I (Figure 5—figure supplement 3; Tynan et al., 2005). In this structure, the TRAV19 and TRBV6-1 encoded CDR loops, which are both identified as having a high interaction potential with HLA class I molecules, make limited germline-encoded contacts with MHC side-chains. There is only a single germline-encoded sidechain interaction between TRAV19 and the MHC α2-helix. As is evident from this structure, significant CDR3-peptide interactions are capable of strongly altering the TCR-MHC interface, regardless of which TRAV and TRBV genes are used. Absent these dominating CDR3-peptide contacts, the interaction potentials serve as a strong, structurally validated quantification of the likelihood of each germline-encoded CDR loop to interact with MHC α-helices.

Biophysical constraints on the MHC surface create germline-encoded biases in docking angles

Having experimentally validated the interaction matrices in Figure 4, we explored whether there exist structural biases in the interactions between germline-encoded TCR CDR loops and MHC α-helices. First, we divided these matrices into separate datasets for each HLA group A, B, and C. We then calculated the interaction potentials across every HLA allele and TRAV or TRBV gene and identified the relative contribution of each CDR loop interacting with each HLA α-helix. From these distributions of contributions, our interaction potentials suggest that the constructive TCR β-chain interactions with the HLA α1-helix (Figure 6A), and perhaps to a lesser extent destructive interactions between the TCR α-chain and the MHC α1-helix (Figure 6—figure supplement 1), may moderately bias the binding orientation between TCR and class I HLA molecules. This reliance on TCR β-chain interactions to guide the overall TCR-MHC interaction has been suggested previously in the literature based upon multiple early observations in mice (Marrack and Kappler, 1987). It is also in line with previous findings that CDR1 and CDR2 of the TCR β-chain are key drivers of the TCR-MHC docking orientation (Feng et al., 2007; Garcia et al., 2009). Conversely, our calculations suggest the α2-helix is only weakly involved in the determination of binding orientation (Figure 6B). The exposed residues on the α2-helix of HLA class I molecules are enriched in alanine and glycine relative to the β1-helix, which are highly unlikely to be involved in a specific, orientation-altering productive interaction.

Figure 6 with 1 supplement see all

Download asset Open asset

Interestingly, the TCR-HLA interaction potential appears flipped for class II molecules: interactions between TCR germline-encoded regions and HLA class II molecules are broadly driven instead by the β-chain helix (Figure 6C and D). We note that the interaction potential incorrectly predicts that the TCR β-chain CDR loops should be the strongest interaction partners with the HLA class II β-chain helix. Nearly every crystallized TCR-HLA class II complex solved thus far adopts the canonical docking orientation whereby the TCR β-chain binds to the HLA α-chain helix, while the TCR α-chain binds to the HLA β-chain helix, with few notable exceptions (Beringer et al., 2015). It is important to note that these interaction potentials take an unbiased approach, calculating every possible interaction between TCR and MHC residues to produce this final score. Therefore, this inconsistency between our interaction potential distributions and crystallized structures may arise from variations across a given MHC helix obscuring the interaction potentials within the canonical binding region, highlighting the need for a more precise approach. To more carefully characterize the CDR loop interactions with MHC α-helices and attempt to alleviate this inconsistency, we isolated the contributions of each individual residue in the HLA class I and class II sequence alignments to the interaction potentials (Figure 7).

Figure 7

Download asset Open asset

Figure 7A and B show a structural visualization of the amino acids of interest on the class I (Figure 7A) and class II (Figure 7B) HLA molecules. These TCR-exposed residues are colored based approximately upon the register a TCR would adopt upon binding each subset, whereby TCRs adopting the canonical binding orientation would contact the residues colored in cyan and pink, and non-canonical binders would contact the blue and red residues. For both HLA class I (Figure 7C, top) and class II (Figure 7D, top), we found a conserved pattern of specific regions of peak interaction potentials surrounded by immediate drop-offs in these interaction potentials. In all but the HLA class I α1-helix, the interaction potentials adopt a distinctly bimodal distribution over the surface of the HLA molecules. In the case of class I molecules, our interaction analysis suggests that the TCR β-chain binds to the HLA class I α1-helix nearly indiscriminately, with potential to form many interactions. The TCR α-chain then contributes to the orientation of the interactions by binding to the HLA class I α2-helix in one of two discrete regions. Conversely, interactions between the αβ TCR and HLA class II molecules appear to have a more defined register for binding, with both the HLA class II α-chain and β-chain interactions with TCR forming distinct bimodal distributions.

We then compared these by-residue interaction breakdowns to crystallized TCR-pMHC structures, counting the number of side-chain interactions between CDR loops and the solvent-exposed residues of the MHC helices. For HLA class I (Figure 7C, bottom), we found good agreement between our predictions and the crystal contacts. Particularly for contacts with the α2-helix, we see that the majority of CDR1/2α contacts fall directly within the predicted regions of non-zero interaction potential. Comparisons to HLA class II (Figure 7D, bottom) are more difficult, again due to the relative dearth of crystallized structures. Overall, these results further contextualize the seemingly inconsistent findings of Figure 6D. While the interactions with the class II β-chain are stronger, we see that the average over all interactions with the α-chain are subdued by the very weak interaction potential of the non-canonical binding region. Within the canonical binding regions of the class II molecule, the TCR α-chain and β-chain CDR loops have similar binding potentials with either α-helix.

The binding register of MHC molecules is well conserved by areas of low interaction potential

The identification of these localized regions of low interaction potential contrasts with what has been postulated previously to be the root of the conserved TCR-pMHC docking orientation. Instead of confirming that conserved interactions determine the docking orientation, our results suggested that regions that are less likely to contact the TCR generate this preference. To further characterize these regions of low interaction potential flanked by regions of increased interaction potential on the MHC surface, we examined position-sensitive biophysical properties to characterize the exposed residues of the MHC α-helices. Specifically, we directly compared the position-sensitive Shannon entropy to the position-sensitive hydropathy of these residues across all MHC class I (Figure 8A) and class II (Figure 8B) molecules.

Figure 8 with 1 supplement see all

Download asset Open asset

We found that in many of the regions where the hydropathy is at or near zero, meaning the region is neither hydrophilic nor hydrophobic, there is drop in the entropy to a value near zero. These residues with a hydropathy near zero likewise have an interaction potential near zero. This suggests that these residues with a lower interaction potential are well-conserved across HLA molecules. Indeed, looking again at alignments of HLA class I (Figure 8C) and class II (Figure 8D) sequences, we found that these regions of alanine and glycine usage are well conserved in a sampling of the so-called ‘parental’ HLA alleles (Robinson et al., 2017). Further inspection of a matrix encoding of a broader set of mammalian MHC molecules (Figure 8—figure supplement 1) showed that alanine and glycine are well conserved in specific regions across species in the α2-helix of MHC class I molecules, as well as in both helices of MHC class II molecules.

These observations completed our final model for the source of bias in the conserved TCR-MHC docking orientation (Figure 8E and F). The structures of HLA class I (Figure 8E) and HLA class II (Figure 8F) are colored based on their average interaction potentials as shown in Figure 7. We readily see the well-conserved areas of negligible interaction potential (black) occur near the centers of the α-helices for each MHC. By overlaying typical CDR contact regions, we can visualize how these regions of low interaction potential may be capable of dictating the docking angle. The α2 and β-chain helices, due in part to the centrally located region of reduced interaction potential and in part due to the kink in helix, appear to play a key role in determining the conserved docking angle.

The concept of evolutionarily conserved interactions guiding immune recognition dates back half a century to work by Jerne, 1971. Despite predating the first TCR-MHC structures the prediction has withstood the test of time. Canonical TCR-MHC docking orientations have repeatedly been observed and specific instances of evolutionarily conserved contacts have been identified in a subset of structures (Feng et al., 2007; Blevins et al., 2016; Scott-Browne et al., 2011). However, the identification of TCRs that are capable of binding to non-MHC ligands (Van Laethem et al., 2007) or in a reversed docking orientation (Beringer et al., 2015; Gras et al., 2016; Zareie et al., 2021) call into question the general validity of these findings of evolutionary conservation and highlight the need for systematic analyses covering the entire space of TRAV, TRBV, and MHC alleles. Even without these eccentric exceptions, evidence for evolutionarily conserved interactions have only been convincingly shown for a small subset of TRAV and TRBV genes. Here, we report the results of the first systematic study across these variable alleles.

Biophysically compatible regions of MHC play a key role in determining the canonical docking orientation

These by-allele interaction potentials can further be broken down on a per-residue basis for each MHC α-helix. Surprisingly, this analysis may suggest why, in spite of the observed sequence variability, we find rather well-conserved TCR-pMHC docking orientations. Across nearly all tested human polymorphisms, we found that class I and class II MHC helices are composed of moderately diverse regions of increased interaction potential interrupted by centrally located, well-conserved regions of low interaction potential. These regions of low interaction potential define a conserved docking orientation, with room for slight variability in orientation about the flanking regions of higher interaction potential. Our findings do not contradict the previously proposed concept of ‘interaction codons’ whereby specific TCR-MHC pairs are predisposed to interact in multiple registers dependent upon the peptide bound to the MHC (Adams et al., 2016; Feng et al., 2007; Garcia et al., 2009). However, these codons as currently posited suggest the existence of multiple rigid docking modes, consistent with their ideation deriving from solved crystal structures. The results of our analyses suggest a broader, dynamic interpretation of the TCR-pMHC interface (Baker et al., 2012; Devlin et al., 2020; Smith et al., 2021; Borbulevych et al., 2009; Scott et al., 2011), with the TCR CDR loops sampling local regions of increased interaction potential on the MHC surface, utilizing a more opportunistic approach to binding. Further, while the interaction codon hypothesis suggests co-evolved interaction interfaces between the TCR and MHC, our analysis suggests that each TRAV-TRBV pairing finds a unique approach to binding within the constraints permitted by the MHC molecular surface. In other words, the MHC molecule largely defines the interface.

In further extending these dynamic interpretations of the TCR-pMHC interface, all-atom molecular dynamics simulations highlight the inherently fleeting nature of sidechain interactions. As a case study, we can consider a well-studied tyrosine-alanine-glutamine interaction which appears to be an example of exceptional shape complementarity between the TCR and MHC surfaces (Figure 9A). Triplicate, all-atom molecular dynamics simulations of a TCR-pMHC complex (PDB: 1FYT Hennecke et al., 2000) highlight the short lifetimes of such intricate molecular interaction networks (Figure 9B). Over the course of these simulations (Figure 9—figure supplement 1), the intrinsic motility of the glutamine sidechain in the interface significantly alters the interpretation of the tyrosine-alanine interaction from a well-evolved notch reserved for TCR binding to merely a consequence of other factors determined strongly by the potentially more dominant CDR3 interactions with the interface.

Figure 9 with 1 supplement see all

Download asset Open asset

These simulations are hardly the first investigations into TCR and MHC dynamics (Baker et al., 2012), but highlight that care should be taken when discussing sidechain interactions within these protein-protein complexes. While we focus briefly on only a trio of interacting residues, one should generally expect that sidechain bonds in crystal structures are unlikely to persist over the entire microsecond-scale binding process. Further, crystal structures should be considered snapshots in a local energy minimum, which at near-physiological temperatures likely sample a wide range of contacts (Mei et al., 2020; Bradford et al., 2021). These dynamic interpretations are consistent with the broad biophysical compatibility suggested by the interaction potential results.

The AIMS software package used to generate the analysis in this manuscript is available here, including the original Jupyter Notebook used to generate the Figures in this manuscript as well as a generalized Notebook and a python-based GUI application for analysis of novel datasets. The AIMS software is constantly evolving, so to ensure exact recapitulation of results presented here AIMS v0.8 should be used. Detailed descriptions of the analysis and the instructions for use can be found at https://aims-doc.readthedocs.io.

All data and code used for the analysis in this manuscript are freely available online with no restrictions. All input FASTA sequences and code needed to recreate the analysis can be found via the AIMS GitHub page. Specific analysis for structural comparisons between interaction potentials and TCR-pMHC complexes are found via a separate repository, called PRESTO, also hosted on GitHub. Due to the significant time required to calculate the interaction scores calculated via AIMS, the calculated scores can be found on Zenodo. In case of future updates to either AIMS or PRESTO, the specific versions used for this manuscript are also hosted on Zenodo, as AIMS v0.8 and PRESTO v0.1.

1. 1. Adams JJ
   2. Narayanan S
   3. Liu BY
   4. Birnbaum ME
   5. Kruse AC
   6. Bowerman NA
   7. Chen W
   8. Levin AM
   9. Connolly JM
   10. Zhu C
   11. Kranz DM
   12. Garcia KC

   (2011) T cell receptor signaling is limited by docking geometry to peptide-major histocompatibility complex

   Immunity 35:681–693.

   https://doi.org/10.1016/j.immuni.2011.09.013

   • PubMed
   • Google Scholar
2. 1. Adams JJ
   2. Narayanan S
   3. Birnbaum ME
   4. Sidhu SS
   5. Blevins SJ
   6. Gee MH
   7. Sibener LV
   8. Baker BM
   9. Kranz DM
   10. Garcia KC

   (2016) Structural interplay between germline interactions and adaptive recognition determines the bandwidth of TCR-peptide-MHC cross-reactivity

   Nature Immunology 17:87–94.

   https://doi.org/10.1038/ni.3310

   • PubMed
   • Google Scholar
3. 1. Al-Lazikani B
   2. Lesk AM
   3. Chothia C

   (2000) Canonical structures for the hypervariable regions of T cell alphabeta receptors

   Journal of Molecular Biology 295:979–995.

   https://doi.org/10.1006/jmbi.1999.3358

   • PubMed
   • Google Scholar
4. 1. Baker BM
   2. Scott DR
   3. Blevins SJ
   4. Hawse WF

   (2012) Structural and dynamic control of T-cell receptor specificity, cross-reactivity, and binding mechanism

   Immunological Reviews 250:10–31.

   https://doi.org/10.1111/j.1600-065X.2012.01165.x

   • PubMed
   • Google Scholar
5. 1. Beringer DX
   2. Kleijwegt FS
   3. Wiede F
   4. van der Slik AR
   5. Loh KL
   6. Petersen J
   7. Dudek NL
   8. Duinkerken G
   9. Laban S
   10. Joosten A
   11. Vivian JP
   12. Chen Z
   13. Uldrich AP
   14. Godfrey DI
   15. McCluskey J
   16. Price DA
   17. Radford KJ
   18. Purcell AW
   19. Nikolic T
   20. Reid HH
   21. Tiganis T
   22. Roep BO
   23. Rossjohn J

   (2015) T cell receptor reversed polarity recognition of a self-antigen major histocompatibility complex

   Nature Immunology 16:1153–1161.

   https://doi.org/10.1038/ni.3271

   • PubMed
   • Google Scholar
6. 1. Birnbaum ME
   2. Mendoza JL
   3. Sethi DK
   4. Dong S
   5. Glanville J
   6. Dobbins J
   7. Ozkan E
   8. Davis MM
   9. Wucherpfennig KW
   10. Garcia KC

   (2014) Deconstructing the peptide-MHC specificity of T cell recognition

   Cell 157:1073–1087.

   https://doi.org/10.1016/j.cell.2014.03.047

   • PubMed
   • Google Scholar
7. 1. Bjorkman PJ
   2. Saper MA
   3. Samraoui B
   4. Bennett WS
   5. Strominger JL
   6. Wiley DC

   (1987) Structure of the human class I histocompatibility antigen, HLA-A2

   Nature 329:506–512.

   https://doi.org/10.1038/329506a0

   • PubMed
   • Google Scholar
8. 1. Bjorkman PJ
   2. Parham P

   (1990) Structure, function, and diversity of class I major histocompatibility complex molecules

   Annual Review of Biochemistry 59:253–288.

   https://doi.org/10.1146/annurev.bi.59.070190.001345

   • PubMed
   • Google Scholar
9. 1. Blackman M
   2. Kappler J
   3. Marrack P

   (1990) The role of the T cell receptor in positive and negative selection of developing T cells

   Science 248:1335–1341.

   https://doi.org/10.1126/science.1972592

   • PubMed
   • Google Scholar
10. 1. Blevins SJ
    2. Pierce BG
    3. Singh NK
    4. Riley TP
    5. Wang Y
    6. Spear TT
    7. Nishimura MI
    8. Weng ZP
    9. Baker BM

    (2016) How structural adaptability exists alongside HLA-A2 bias in the human αβ TCR repertoire

    PNAS 113:E1276–E1285.

    https://doi.org/10.1073/pnas.1522069113

    • PubMed
    • Google Scholar
11. 1. Borbulevych OY
    2. Piepenbrink KH
    3. Gloor BE
    4. Scott DR
    5. Sommese RF
    6. Cole DK
    7. Sewell AK
    8. Baker BM

    (2009) T cell receptor cross-reactivity directed by antigen-dependent tuning of peptide-MHC molecular flexibility

    Immunity 31:885–896.

    https://doi.org/10.1016/j.immuni.2009.11.003

    • PubMed
    • Google Scholar
12. 1. Boughter CT
    2. Borowska MT
    3. Guthmiller JJ
    4. Bendelac A
    5. Wilson PC
    6. Roux B
    7. Adams EJ

    (2020) Biochemical patterns of antibody polyreactivity revealed through a bioinformatics-based analysis of CDR loops

    eLife 9:e61393.

    https://doi.org/10.7554/eLife.61393

    • PubMed
    • Google Scholar
13. 1. Boughter CT
    2. Meier-Schellersheim M

    (2023) An integrated approach to the characterization of immune repertoires using AIMS: An Automated Immune Molecule Separator

    PLOS Computational Biology 19:e1011577.

    https://doi.org/10.1371/journal.pcbi.1011577

    • PubMed
    • Google Scholar
14. 1. Bradford SYC
    2. El Khoury L
    3. Ge Y
    4. Osato M
    5. Mobley DL
    6. Fischer M

    (2021) Temperature artifacts in protein structures bias ligand-binding predictions

    Chemical Science 12:11275–11293.

    https://doi.org/10.1039/d1sc02751d

    • PubMed
    • Google Scholar
15. 1. Brochet X
    2. Lefranc MP
    3. Giudicelli V

    (2008) IMGT/V-QUEST: the highly customized and integrated system for IG and TR standardized V-J and V-D-J sequence analysis

    Nucleic Acids Research 36:W503–W508.

    https://doi.org/10.1093/nar/gkn316

    • PubMed
    • Google Scholar
16. 1. Brooks BR
    2. Brooks CL
    3. Mackerell AD
    4. Nilsson L
    5. Petrella RJ
    6. Roux B
    7. Won Y
    8. Archontis G
    9. Bartels C
    10. Boresch S
    11. Caflisch A
    12. Caves L
    13. Cui Q
    14. Dinner AR
    15. Feig M
    16. Fischer S
    17. Gao J
    18. Hodoscek M
    19. Im W
    20. Kuczera K
    21. Lazaridis T
    22. Ma J
    23. Ovchinnikov V
    24. Paci E
    25. Pastor RW
    26. Post CB
    27. Pu JZ
    28. Schaefer M
    29. Tidor B
    30. Venable RM
    31. Woodcock HL
    32. Wu X
    33. Yang W
    34. York DM
    35. Karplus M

    (2009) CHARMM: the biomolecular simulation program

    Journal of Computational Chemistry 30:1545–1614.

    https://doi.org/10.1002/jcc.21287

    • PubMed
    • Google Scholar
17. 1. Brown JH
    2. Jardetzky TS
    3. Gorga JC
    4. Stern LJ
    5. Urban RG
    6. Strominger JL
    7. Wiley DC

    (1993) Three-dimensional structure of the human class II histocompatibility antigen HLA-DR1

    Nature 364:33–39.

    https://doi.org/10.1038/364033a0

    • PubMed
    • Google Scholar
18. 1. Burrows SR
    2. Chen ZJ
    3. Archbold JK
    4. Tynan FE
    5. Beddoe T
    6. Kjer-Nielsen L
    7. Miles JJ
    8. Khanna R
    9. Moss DJ
    10. Liu YC
    11. Gras S
    12. Kostenko L
    13. Brennan RM
    14. Clements CS
    15. Brooks AG
    16. Purcell AW
    17. McCluskey J
    18. Rossjohn J

    (2010) Hard wiring of T cell receptor specificity for the major histocompatibility complex is underpinned by TCR adaptability

    PNAS 107:10608–10613.

    https://doi.org/10.1073/pnas.1004926107

    • PubMed
    • Google Scholar
19. 1. Cheong R
    2. Rhee A
    3. Wang CJ
    4. Nemenman I
    5. Levchenko A

    (2011) Information transduction capacity of noisy biochemical signaling networks

    Science 334:354–358.

    https://doi.org/10.1126/science.1204553

    • PubMed
    • Google Scholar
20. 1. Ciacchi L
    2. Farenc C
    3. Dahal-Koirala S
    4. Petersen J
    5. Sollid LM
    6. Reid HH
    7. Rossjohn J

    (2022) Structural basis of T cell receptor specificity and cross-reactivity of two HLA-DQ2.5-restricted gluten epitopes in celiac disease

    The Journal of Biological Chemistry 298:101619.

    https://doi.org/10.1016/j.jbc.2022.101619

    • PubMed
    • Google Scholar
21. 1. Dai S
    2. Huseby ES
    3. Rubtsova K
    4. Scott-Browne J
    5. Crawford F
    6. Macdonald WA
    7. Marrack P
    8. Kappler JW

    (2008) Crossreactive T Cells spotlight the germline rules for alphabeta T cell-receptor interactions with MHC molecules

    Immunity 28:324–334.

    https://doi.org/10.1016/j.immuni.2008.01.008

    • PubMed
    • Google Scholar
22. 1. Devlin JR
    2. Alonso JA
    3. Ayres CM
    4. Keller GLJ
    5. Bobisse S
    6. Vander Kooi CW
    7. Coukos G
    8. Gfeller D
    9. Harari A
    10. Baker BM

    (2020) Structural dissimilarity from self drives neoepitope escape from immune tolerance

    Nature Chemical Biology 16:1269–1276.

    https://doi.org/10.1038/s41589-020-0610-1

    • PubMed
    • Google Scholar
23. Preprint

    1. Evans R
    2. O’Neill M
    3. Pritzel A
    4. Antropova N
    5. Senior A
    6. Green T
    7. Žídek A
    8. Bates R
    9. Blackwell S
    10. Yim J
    11. Ronneberger O
    12. Bodenstein S
    13. Zielinski M
    14. Bridgland A
    15. Potapenko A
    16. Cowie A
    17. Tunyasuvunakool K
    18. Jain R
    19. Clancy E
    20. Kohli P
    21. Jumper J
    22. Hassabis D

    (2022) Protein complex prediction with alphafold-multimer

    bioRxiv.

    https://doi.org/10.1101/2021.10.04.463034

    • Google Scholar
24. 1. Feng D
    2. Bond CJ
    3. Ely LK
    4. Maynard J
    5. Garcia KC

    (2007) Structural evidence for a germline-encoded T cell receptor-major histocompatibility complex interaction “codon.”

    Nature Immunology 8:975–983.

    https://doi.org/10.1038/ni1502

    • PubMed
    • Google Scholar
25. 1. Garboczi DN
    2. Ghosh P
    3. Utz U
    4. Fan QR
    5. Biddison WE
    6. Wiley DC

    (1996) Structure of the complex between human T-cell receptor, viral peptide and HLA-A2

    Nature 384:134–141.

    https://doi.org/10.1038/384134a0

    • PubMed
    • Google Scholar
26. 1. Garcia KC
    2. Degano M
    3. Stanfield RL
    4. Brunmark A
    5. Jackson MR
    6. Peterson PA
    7. Teyton L
    8. Wilson IA

    (1996) An alphabeta T cell receptor structure at 2.5 A and its orientation in the TCR-MHC complex

    Science 274:209–219.

    https://doi.org/10.1126/science.274.5285.209

    • PubMed
    • Google Scholar
27. 1. Garcia KC
    2. Teyton L
    3. Wilson IA

    (1999) Structural basis of T cell recognition

    Annual Review of Immunology 17:369–397.

    https://doi.org/10.1146/annurev.immunol.17.1.369

    • PubMed
    • Google Scholar
28. 1. Garcia KC
    2. Adams JJ
    3. Feng D
    4. Ely LK

    (2009) The molecular basis of TCR germline bias for MHC is surprisingly simple

    Nature Immunology 10:143–147.

    https://doi.org/10.1038/ni.f.219

    • PubMed
    • Google Scholar
29. 1. Garcia KC

    (2012) Reconciling views on T cell receptor germline bias for MHC

    Trends in Immunology 33:429–436.

    https://doi.org/10.1016/j.it.2012.05.005

    • PubMed
    • Google Scholar
30. 1. Germain RN

    (2002) T-cell development and the CD4-CD8 lineage decision

    Nature Reviews. Immunology 2:309–322.

    https://doi.org/10.1038/nri798

    • PubMed
    • Google Scholar
31. 1. Gowthaman R
    2. Pierce BG

    (2019) TCR3d: The T cell receptor structural repertoire database

    Bioinformatics 35:5323–5325.

    https://doi.org/10.1093/bioinformatics/btz517

    • PubMed
    • Google Scholar
32. 1. Gras S
    2. Burrows SR
    3. Turner SJ
    4. Sewell AK
    5. McCluskey J
    6. Rossjohn J

    (2012) A structural voyage toward an understanding of the MHC-I-restricted immune response: lessons learned and much to be learned

    Immunological Reviews 250:61–81.

    https://doi.org/10.1111/j.1600-065X.2012.01159.x

    • PubMed
    • Google Scholar
33. 1. Gras S
    2. Chadderton J
    3. Del Campo CM
    4. Farenc C
    5. Wiede F
    6. Josephs TM
    7. Sng XYX
    8. Mirams M
    9. Watson KA
    10. Tiganis T
    11. Quinn KM
    12. Rossjohn J
    13. La Gruta NL

    (2016) Reversed T cell receptor docking on a major histocompatibility class i complex limits involvement in the immune response

    Immunity 45:749–760.

    https://doi.org/10.1016/j.immuni.2016.09.007

    • PubMed
    • Google Scholar
34. 1. Gumbart JC
    2. Roux B
    3. Chipot C

    (2013a) Efficient determination of protein-protein standard binding free energies from first principles

    Journal of Chemical Theory and Computation 9:ct400273t.

    https://doi.org/10.1021/ct400273t

    • PubMed
    • Google Scholar
35. 1. Gumbart JC
    2. Roux B
    3. Chipot C

    (2013b) Standard binding free energies from computer simulations: What is the best strategy?

    Journal of Chemical Theory and Computation 9:794–802.

    https://doi.org/10.1021/ct3008099

    • PubMed
    • Google Scholar
36. 1. Gunnarsen KS
    2. Høydahl LS
    3. Risnes LF
    4. Dahal-Koirala S
    5. Neumann RS
    6. Bergseng E
    7. Frigstad T
    8. Frick R
    9. du Pré MF
    10. Dalhus B
    11. Lundin KE
    12. Qiao S-W
    13. Sollid LM
    14. Sandlie I
    15. Løset GÅ

    (2017) A TCRα framework-centered codon shapes A biased T cell repertoire through direct MHC and CDR3β interactions

    JCI Insight 2:17.

    https://doi.org/10.1172/jci.insight.95193

    • PubMed
    • Google Scholar
37. 1. Hahn M
    2. Nicholson MJ
    3. Pyrdol J
    4. Wucherpfennig KW

    (2005) Unconventional topology of self peptide-major histocompatibility complex binding by a human autoimmune T cell receptor

    Nature Immunology 6:490–496.

    https://doi.org/10.1038/ni1187

    • PubMed
    • Google Scholar
38. 1. Hennecke J
    2. Carfi A
    3. Wiley DC

    (2000) Structure of a covalently stabilized complex of a human alphabeta T-cell receptor, influenza HA peptide and MHC class II molecule, HLA-DR1

    The EMBO Journal 19:5611–5624.

    https://doi.org/10.1093/emboj/19.21.5611

    • PubMed
    • Google Scholar
39. 1. Hopkins CW
    2. Le Grand S
    3. Walker RC
    4. Roitberg AE

    (2015) Long-time-step molecular dynamics through hydrogen mass repartitioning

    Journal of Chemical Theory and Computation 11:1864–1874.

    https://doi.org/10.1021/ct5010406

    • PubMed
    • Google Scholar
40. 1. Humphrey W
    2. Dalke A
    3. Schulten K

    (1996) VMD: visual molecular dynamics

    Journal of Molecular Graphics 14:33–38.

    https://doi.org/10.1016/0263-7855(96)00018-5

    • PubMed
    • Google Scholar
41. 1. Huseby ES
    2. Crawford F
    3. White J
    4. Marrack P
    5. Kappler JW

    (2006) Interface-disrupting amino acids establish specificity between T cell receptors and complexes of major histocompatibility complex and peptide

    Nature Immunology 7:1191–1199.

    https://doi.org/10.1038/ni1401

    • PubMed
    • Google Scholar
42. 1. Ishigaki K
    2. Lagattuta KA
    3. Luo Y
    4. James EA
    5. Buckner JH
    6. Raychaudhuri S

    (2022) HLA autoimmune risk alleles restrict the hypervariable region of T cell receptors

    Nature Genetics 54:393–402.

    https://doi.org/10.1038/s41588-022-01032-z

    • PubMed
    • Google Scholar
43. 1. Jerne NK

    (1971) The somatic generation of immune recognition

    European Journal of Immunology 1:1–9.

    https://doi.org/10.1002/eji.1830010102

    • PubMed
    • Google Scholar
44. 1. Jo S
    2. Kim T
    3. Im W

    (2007) Automated builder and database of protein/membrane complexes for molecular dynamics simulations

    PLOS ONE 2:e880.

    https://doi.org/10.1371/journal.pone.0000880

    • PubMed
    • Google Scholar
45. 1. Jo S
    2. Kim T
    3. Iyer VG
    4. Im W

    (2008) CHARMM-GUI: A web-based graphical user interface for CHARMM

    Journal of Computational Chemistry 29:1859–1865.

    https://doi.org/10.1002/jcc.20945

    • PubMed
    • Google Scholar
46. 1. Juang J
    2. Ebert PJR
    3. Feng D
    4. Garcia KC
    5. Krogsgaard M
    6. Davis MM

    (2010) Peptide-MHC heterodimers show that thymic positive selection requires a more restricted set of self-peptides than negative selection

    The Journal of Experimental Medicine 207:1223–1234.

    https://doi.org/10.1084/jem.20092170

    • PubMed
    • Google Scholar
47. 1. Jung D
    2. Alt FW

    (2004) Unraveling V(D)J recombination; insights into gene regulation

    Cell 116:299–311.

    https://doi.org/10.1016/s0092-8674(04)00039-x

    • PubMed
    • Google Scholar
48. 1. Krovi SH
    2. Kappler JW
    3. Marrack P
    4. Gapin L

    (2019) Inherent reactivity of unselected TCR repertoires to peptide-MHC molecules

    PNAS 116:22252–22261.

    https://doi.org/10.1073/pnas.1909504116

    • PubMed
    • Google Scholar
49. 1. Lee J
    2. Cheng X
    3. Swails JM
    4. Yeom MS
    5. Eastman PK
    6. Lemkul JA
    7. Wei S
    8. Buckner J
    9. Jeong JC
    10. Qi Y
    11. Jo S
    12. Pande VS
    13. Case DA
    14. Brooks CL
    15. MacKerell AD
    16. Klauda JB
    17. Im W

    (2016) CHARMM-GUI Input Generator for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simulations Using the CHARMM36 Additive Force Field

    Journal of Chemical Theory and Computation 12:405–413.

    https://doi.org/10.1021/acs.jctc.5b00935

    • PubMed
    • Google Scholar
50. 1. Lefranc MP

    (2014) Immunoglobulins: 25 years of immunoinformatics and IMGT-ONTOLOGY

    Biomolecules 4:1102–1139.

    https://doi.org/10.3390/biom4041102

    • PubMed
    • Google Scholar
51. 1. Lu J
    2. Van Laethem F
    3. Bhattacharya A
    4. Craveiro M
    5. Saba I
    6. Chu J
    7. Love NC
    8. Tikhonova A
    9. Radaev S
    10. Sun X
    11. Ko A
    12. Arnon T
    13. Shifrut E
    14. Friedman N
    15. Weng N-P
    16. Singer A
    17. Sun PD

    (2019) Molecular constraints on CDR3 for thymic selection of MHC-restricted TCRs from a random pre-selection repertoire

    Nature Communications 10:1019.

    https://doi.org/10.1038/s41467-019-08906-7

    • PubMed
    • Google Scholar
52. 1. Mack SJ
    2. Cano P
    3. Hollenbach JA
    4. He J
    5. Hurley CK
    6. Middleton D
    7. Moraes ME
    8. Pereira SE
    9. Kempenich JH
    10. Reed EF
    11. Setterholm M
    12. Smith AG
    13. Tilanus MG
    14. Torres M
    15. Varney MD
    16. Voorter CEM
    17. Fischer GF
    18. Fleischhauer K
    19. Goodridge D
    20. Klitz W
    21. Little AM
    22. Maiers M
    23. Marsh SGE
    24. Müller CR
    25. Noreen H
    26. Rozemuller EH
    27. Sanchez-Mazas A
    28. Senitzer D
    29. Trachtenberg E
    30. Fernandez-Vina M

    (2013) Common and well-documented HLA alleles: 2012 update to the CWD catalogue

    Tissue Antigens 81:194–203.

    https://doi.org/10.1111/tan.12093

    • PubMed
    • Google Scholar
53. 1. Marrack P
    2. Kappler J

    (1987) The T cell receptor

    Science 238:1073–1079.

    https://doi.org/10.1126/science.3317824

    • PubMed
    • Google Scholar
54. 1. Mason D

    (1998) A very high level of crossreactivity is an essential feature of the T-cell receptor

    Immunology Today 19:395–404.

    https://doi.org/10.1016/s0167-5699(98)01299-7

    • PubMed
    • Google Scholar
55. 1. McGibbon RT
    2. Beauchamp KA
    3. Harrigan MP
    4. Klein C
    5. Swails JM
    6. Hernández CX
    7. Schwantes CR
    8. Wang LP
    9. Lane TJ
    10. Pande VS

    (2015) MDTraj: a modern open library for the analysis of molecular dynamics trajectories

    Biophysical Journal 109:1528–1532.

    https://doi.org/10.1016/j.bpj.2015.08.015

    • PubMed
    • Google Scholar
56. 1. Mei Z
    2. Treado JD
    3. Grigas AT
    4. Levine ZA
    5. Regan L
    6. O’Hern CS

    (2020) Analyses of protein cores reveal fundamental differences between solution and crystal structures

    Proteins 88:1154–1161.

    https://doi.org/10.1002/prot.25884

    • Google Scholar
57. 1. Milighetti M
    2. Shawe-Taylor J
    3. Chain B

    (2021) Corrigendum: predicting t cell receptor antigen specificity from structural features derived from homology models of receptor-peptide-major histocompatibility complexes

    Frontiers in Physiology 12:790998.

    https://doi.org/10.3389/fphys.2021.790998

    • PubMed
    • Google Scholar
58. 1. Mora T
    2. Walczak AM
    3. Bialek W
    4. Callan CG

    (2010) Maximum entropy models for antibody diversity

    PNAS 107:5405–5410.

    https://doi.org/10.1073/pnas.1001705107

    • PubMed
    • Google Scholar
59. 1. Moradi S
    2. Berry R
    3. Pymm P
    4. Hitchen C
    5. Beckham SA
    6. Wilce MCJ
    7. Walpole NG
    8. Clements CS
    9. Reid HH
    10. Perugini MA
    11. Brooks AG
    12. Rossjohn J
    13. Vivian JP

    (2015) The structure of the atypical killer cell immunoglobulin-like receptor, KIR2DL4

    The Journal of Biological Chemistry 290:10460–10471.

    https://doi.org/10.1074/jbc.M114.612291

    • PubMed
    • Google Scholar
60. 1. Moradi S
    2. Stankovic S
    3. O’Connor GM
    4. Pymm P
    5. MacLachlan BJ
    6. Faoro C
    7. Retière C
    8. Sullivan LC
    9. Saunders PM
    10. Widjaja J
    11. Cox-Livingstone S
    12. Rossjohn J
    13. Brooks AG
    14. Vivian JP

    (2021) Structural plasticity of KIR2DL2 and KIR2DL3 enables altered docking geometries atop HLA-C

    Nature Communications 12:2173.

    https://doi.org/10.1038/s41467-021-22359-x

    • PubMed
    • Google Scholar
61. 1. Murugan A
    2. Mora T
    3. Walczak AM
    4. Callan CG

    (2012) Statistical inference of the generation probability of T-cell receptors from sequence repertoires

    PNAS 109:16161–16166.

    https://doi.org/10.1073/pnas.1212755109

    • PubMed
    • Google Scholar
62. 1. Nandigrami P
    2. Szczepaniak F
    3. Boughter CT
    4. Dehez F
    5. Chipot C
    6. Roux B

    (2022) Computational assessment of protein-protein binding specificity within a family of synaptic surface receptors

    The Journal of Physical Chemistry. B 126:7510–7527.

    https://doi.org/10.1021/acs.jpcb.2c02173

    • PubMed
    • Google Scholar
63. 1. Nguyen AT
    2. Szeto C
    3. Gras S

    (2021) The pockets guide to HLA class I molecules

    Biochemical Society Transactions 49:2319–2331.

    https://doi.org/10.1042/BST20210410

    • PubMed
    • Google Scholar
64. 1. Petersen J
    2. Montserrat V
    3. Mujico JR
    4. Loh KL
    5. Beringer DX
    6. van Lummel M
    7. Thompson A
    8. Mearin ML
    9. Schweizer J
    10. Kooy-Winkelaar Y
    11. van Bergen J
    12. Drijfhout JW
    13. Kan W-T
    14. La Gruta NL
    15. Anderson RP
    16. Reid HH
    17. Koning F
    18. Rossjohn J

    (2014) T-cell receptor recognition of HLA-DQ2-gliadin complexes associated with celiac disease

    Nature Structural & Molecular Biology 21:480–488.

    https://doi.org/10.1038/nsmb.2817

    • PubMed
    • Google Scholar
65. 1. Petersen J
    2. Ciacchi L
    3. Tran MT
    4. Loh KL
    5. Kooy-Winkelaar Y
    6. Croft NP
    7. Hardy MY
    8. Chen Z
    9. McCluskey J
    10. Anderson RP
    11. Purcell AW
    12. Tye-Din JA
    13. Koning F
    14. Reid HH
    15. Rossjohn J

    (2020) T cell receptor cross-reactivity between gliadin and bacterial peptides in celiac disease

    Nature Structural & Molecular Biology 27:49–61.

    https://doi.org/10.1038/s41594-019-0353-4

    • PubMed
    • Google Scholar
66. 1. Piepenbrink KH
    2. Blevins SJ
    3. Scott DR
    4. Baker BM

    (2013) The basis for limited specificity and MHC restriction in a T cell receptor interface

    Nature Communications 4:1948.

    https://doi.org/10.1038/ncomms2948

    • PubMed
    • Google Scholar
67. 1. Qiao S-W
    2. Christophersen A
    3. Lundin KEA
    4. Sollid LM

    (2014) Biased usage and preferred pairing of α- and β-chains of TCRs specific for an immunodominant gluten epitope in coeliac disease

    International Immunology 26:13–19.

    https://doi.org/10.1093/intimm/dxt037

    • PubMed
    • Google Scholar
68. 1. Riley TP
    2. Hellman LM
    3. Gee MH
    4. Mendoza JL
    5. Alonso JA
    6. Foley KC
    7. Nishimura MI
    8. Vander Kooi CW
    9. Garcia KC
    10. Baker BM

    (2018) T cell receptor cross-reactivity expanded by dramatic peptide-MHC adaptability

    Nature Chemical Biology 14:934–942.

    https://doi.org/10.1038/s41589-018-0130-4

    • PubMed
    • Google Scholar
69. 1. Robinson J
    2. Mistry K
    3. McWilliam H
    4. Lopez R
    5. Marsh SGE

    (2010) IPD--the immuno polymorphism database

    Nucleic Acids Research 38:D863–D869.

    https://doi.org/10.1093/nar/gkp879

    • PubMed
    • Google Scholar
70. 1. Robinson J
    2. Guethlein LA
    3. Cereb N
    4. Yang SY
    5. Norman PJ
    6. Marsh SGE
    7. Parham P

    (2017) Distinguishing functional polymorphism from random variation in the sequences of >10,000 HLA-A, -B and -C alleles

    PLOS Genetics 13:e1006862.

    https://doi.org/10.1371/journal.pgen.1006862

    • PubMed
    • Google Scholar
71. 1. Román-Roldán R
    2. Bernaola-Galván P
    3. Oliver JL

    (1996) Application of information theory to DNA sequence analysis: A review

    Pattern Recognition 29:1187–1194.

    https://doi.org/10.1016/0031-3203(95)00145-X

    • Google Scholar
72. 1. Scott DR
    2. Borbulevych OY
    3. Piepenbrink KH
    4. Corcelli SA
    5. Baker BM

    (2011) Disparate degrees of hypervariable loop flexibility control T-cell receptor cross-reactivity, specificity, and binding mechanism

    Journal of Molecular Biology 414:385–400.

    https://doi.org/10.1016/j.jmb.2011.10.006

    • PubMed
    • Google Scholar
73. 1. Scott-Browne JP
    2. White J
    3. Kappler JW
    4. Gapin L
    5. Marrack P

    (2009) Germline-encoded amino acids in the alphabeta T-cell receptor control thymic selection

    Nature 458:1043–1046.

    https://doi.org/10.1038/nature07812

    • PubMed
    • Google Scholar
74. 1. Scott-Browne JP
    2. Crawford F
    3. Young MH
    4. Kappler JW
    5. Marrack P
    6. Gapin L

    (2011) Evolutionarily conserved features contribute to αβ T cell receptor specificity

    Immunity 35:526–535.

    https://doi.org/10.1016/j.immuni.2011.09.005

    • PubMed
    • Google Scholar
75. 1. Sethi DK
    2. Gordo S
    3. Schubert DA
    4. Wucherpfennig KW

    (2013) Crossreactivity of a human autoimmune TCR is dominated by a single TCR loop

    Nature Communications 4:2623.

    https://doi.org/10.1038/ncomms3623

    • PubMed
    • Google Scholar
76. 1. Sewell AK

    (2012) Why must T cells be cross-reactive?

    Nature Reviews. Immunology 12:669–677.

    https://doi.org/10.1038/nri3279

    • PubMed
    • Google Scholar
77. 1. Shannon CE

    (1948) A mathematical theory of communication

    Bell System Technical Journal 27:623–656.

    https://doi.org/10.1002/j.1538-7305.1948.tb00917.x

    • Google Scholar
78. 1. Sharon E
    2. Sibener LV
    3. Battle A
    4. Fraser HB
    5. Garcia KC
    6. Pritchard JK

    (2016) Genetic variation in MHC proteins is associated with T cell receptor expression biases

    Nature Genetics 48:995–1002.

    https://doi.org/10.1038/ng.3625

    • PubMed
    • Google Scholar
79. 1. Sibener LV
    2. Fernandes RA
    3. Kolawole EM
    4. Carbone CB
    5. Liu F
    6. McAffee D
    7. Birnbaum ME
    8. Yang X
    9. Su LF
    10. Yu W
    11. Dong S
    12. Gee MH
    13. Jude KM
    14. Davis MM
    15. Groves JT
    16. Goddard WA
    17. Heath JR
    18. Evavold BD
    19. Vale RD
    20. Garcia KC

    (2018) Isolation of a Structural Mechanism for Uncoupling T Cell Receptor Signaling from Peptide-MHC Binding

    Cell 174:672–687.

    https://doi.org/10.1016/j.cell.2018.06.017

    • PubMed
    • Google Scholar
80. 1. Singh NK
    2. Alonso JA
    3. Devlin JR
    4. Keller GLJ
    5. Gray GI
    6. Chiranjivi AK
    7. Foote SG
    8. Landau LM
    9. Arbuiso AG
    10. Weiss LI
    11. Rosenberg AM
    12. Hellman LM
    13. Nishimura MI
    14. Baker BM

    (2022) A class-mismatched TCR bypasses MHC restriction via an unorthodox but fully functional binding geometry

    Nature Communications 13:7189.

    https://doi.org/10.1038/s41467-022-34896-0

    • PubMed
    • Google Scholar
81. 1. Smith AR
    2. Alonso JA
    3. Ayres CM
    4. Singh NK
    5. Hellman LM
    6. Baker BM

    (2021) Structurally silent peptide anchor modifications allosterically modulate T cell recognition in a receptor-dependent manner

    PNAS 118:e2018125118.

    https://doi.org/10.1073/pnas.2018125118

    • PubMed
    • Google Scholar
82. 1. Stadinski BD
    2. Trenh P
    3. Duke B
    4. Huseby PG
    5. Li G
    6. Stern LJ
    7. Huseby ES

    (2014) Effect of CDR3 sequences and distal V gene residues in regulating TCR-MHC contacts and ligand specificity

    Journal of Immunology 192:6071–6082.

    https://doi.org/10.4049/jimmunol.1303209

    • PubMed
    • Google Scholar
83. 1. Thielens A
    2. Vivier E
    3. Romagné F

    (2012) NK cell MHC class I specific receptors (KIR): from biology to clinical intervention

    Current Opinion in Immunology 24:239–245.

    https://doi.org/10.1016/j.coi.2012.01.001

    • PubMed
    • Google Scholar
84. 1. Tikhonova AN
    2. Van Laethem F
    3. Hanada K
    4. Lu J
    5. Pobezinsky LA
    6. Hong C
    7. Guinter TI
    8. Jeurling SK
    9. Bernhardt G
    10. Park J-H
    11. Yang JC
    12. Sun PD
    13. Singer A

    (2012) αβ T cell receptors that do not undergo major histocompatibility complex-specific thymic selection possess antibody-like recognition specificities

    Immunity 36:79–91.

    https://doi.org/10.1016/j.immuni.2011.11.013

    • PubMed
    • Google Scholar
85. 1. Tynan FE
    2. Burrows SR
    3. Buckle AM
    4. Clements CS
    5. Borg NA
    6. Miles JJ
    7. Beddoe T
    8. Whisstock JC
    9. Wilce MC
    10. Silins SL
    11. Burrows JM
    12. Kjer-Nielsen L
    13. Kostenko L
    14. Purcell AW
    15. McCluskey J
    16. Rossjohn J

    (2005) T cell receptor recognition of a “super-bulged” major histocompatibility complex class I-bound peptide

    Nature Immunology 6:1114–1122.

    https://doi.org/10.1038/ni1257

    • PubMed
    • Google Scholar
86. 1. Van Laethem F
    2. Sarafova SD
    3. Park J-H
    4. Tai X
    5. Pobezinsky L
    6. Guinter TI
    7. Adoro S
    8. Adams A
    9. Sharrow SO
    10. Feigenbaum L
    11. Singer A

    (2007) Deletion of CD4 and CD8 coreceptors permits generation of alphabetaT cells that recognize antigens independently of the MHC

    Immunity 27:735–750.

    https://doi.org/10.1016/j.immuni.2007.10.007

    • PubMed
    • Google Scholar
87. 1. Van Laethem F
    2. Tikhonova AN
    3. Singer A

    (2012) MHC restriction is imposed on a diverse T cell receptor repertoire by CD4 and CD8 co-receptors during thymic selection

    Trends in Immunology 33:437–441.

    https://doi.org/10.1016/j.it.2012.05.006

    • PubMed
    • Google Scholar
88. 1. Vinga S

    (2014) Information theory applications for biological sequence analysis

    Briefings in Bioinformatics 15:376–389.

    https://doi.org/10.1093/bib/bbt068

    • PubMed
    • Google Scholar
89. 1. Wu LC
    2. Tuot DS
    3. Lyons DS
    4. Garcia KC
    5. Davis MM

    (2002) Two-step binding mechanism for T-cell receptor recognition of peptide MHC

    Nature 418:552–556.

    https://doi.org/10.1038/nature00920

    • PubMed
    • Google Scholar
90. 1. Yin L
    2. Scott-Browne J
    3. Kappler JW
    4. Gapin L
    5. Marrack P

    (2012) T cells and their eons-old obsession with MHC

    Immunological Reviews 250:49–60.

    https://doi.org/10.1111/imr.12004

    • PubMed
    • Google Scholar
91. 1. Ysern X
    2. Li H
    3. Mariuzza RA

    (1998) Imperfect interfaces

    Nature Structural Biology 5:412–414.

    https://doi.org/10.1038/nsb0698-412

    • PubMed
    • Google Scholar
92. 1. Zareie P
    2. Szeto C
    3. Farenc C
    4. Gunasinghe SD
    5. Kolawole EM
    6. Nguyen A
    7. Blyth C
    8. Sng XYX
    9. Li J
    10. Jones CM
    11. Fulcher AJ
    12. Jacobs JR
    13. Wei Q
    14. Wojciech L
    15. Petersen J
    16. Gascoigne NRJ
    17. Evavold BD
    18. Gaus K
    19. Gras S
    20. Rossjohn J
    21. La Gruta NL

    (2021) Canonical T cell receptor docking on peptide-MHC is essential for T cell signaling

    Science 372:eabe9124.

    https://doi.org/10.1126/science.abe9124

    • PubMed
    • Google Scholar

Author details

1. Christopher T Boughter

   Computational Biology Section, Laboratory of Immune System Biology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, United States

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing – review and editing

   For correspondence

   christopher.boughter@nih.gov

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7106-4699

2. Martin Meier-Schellersheim

   Computational Biology Section, Laboratory of Immune System Biology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, United States

   Contribution

   Supervision, Funding acquisition, Project administration, Writing – review and editing, Conceptualization, Resources

   For correspondence

   mms@niaid.nih.gov

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8754-6377

• 922

  views

• 117

  downloads

• 1

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.