Geometrical characterization of T cell receptor binding modes reveals class-specific binding to maximize access to antigen

Abstract

Recognition of antigenic peptides bound to major histocompatibility complex (MHC) proteins by αβ T cell receptors (TCRs) is a hallmark of T cell mediated immunity. Recent data suggests that variations in TCR binding geometry may influence T cell signaling, which could help explain outliers in relationships between physical parameters such as TCR-pMHC binding affinity and T cell function. Traditionally, TCR binding geometry has been described with simple descriptors such as the crossing angle, which quantifies what has become known as the TCR’s diagonal binding mode. However, these descriptors often fail to reveal distinctions in binding geometry that are apparent through visual inspection. To provide a framework for better relating TCR structure to T cell function, we developed a comprehensive system for quantifying the geometries of how TCRs bind peptide/MHC complexes. We show that our system can discern differences not clearly revealed by the more common methods. As an example of its potential to impact biology, we used it to reveal differences in how TCRs bind class I and class II peptide/MHC complexes, which we show allow the TCR to maximize access to and “read out” the peptide antigen. We anticipate our system will be of use in not only exploring these and other details of TCR-peptide/MHC binding interactions, but also addressing long-held questions about how TCR binding geometry relates to T cell function, as well as modeling structural properties of class I and class II TCR-peptide/MHC complexes from sequence information.

1. Introduction

Specificity and signaling in cellular immunity is driven by the T cell receptor (TCR), a heterodimeric cell surface protein of the immunoglobulin superfamily. Architecturally, TCRs resemble antigen binding fragments of antibodies and use a similar arrangement of complementarity determining region (CDR) loops to bind their ligand. One key difference between TCRs and antibodies though is that while antibodies recognize targets of remarkable structural and chemical diversity, TCRs recognize small peptide antigens bound and presented by major histocompatibility complex (MHC) proteins. This phenomenon, termed MHC restriction, is a fundamental facet of cellular immunity that ensures T cell responses are appropriately directed towards immunologically-relevant targets.

Early structural studies of TCR-peptide/MHC complexes established that TCRs bind peptide/MHC complexes “head on” so they can sample the antigenic peptide ^1,2. Although the vast majority of TCRs investigated to-date bind in this general manner, TCRs still adopt a large range of positions over peptide/MHC complexes, varying twist, tilt, and planar positioning ³. These variations allow the TCR to best optimize its binding orientation in order to minimize binding free energy, as would be expected for the interaction between a diverse receptor and a diverse ligand.

Despite the variances that are seen, TCR binding geometries are believed to be somewhat biologically constrained. Indeed, recent observations show that when TCRs bind with unusual outlier geometries, deficient T cell signaling can occur ^4,5. Biological constraints on binding geometry could be imparted through the need to assemble functional signaling complexes involving co-receptors and other molecules on T cell surfaces ⁶. TCR binding geometries are also believed to be influenced by co-evolution of TCRs and MHC proteins. Co-evolution leads to a built-in compatibility between TCRs and the structural and chemical features of MHC proteins, biasing them towards binding in particular fashions ^7,8. Accordingly, there has been significant discussion regarding TCR binding geometries and how they might influence signaling, MHC restriction, and target selection.

The first structures of TCR-peptide/MHC complexes established what is commonly termed the diagonal binding mode, determined by the orientation of the TCR Vα/Vβ domains relative to the peptide. This geometry places the loops of the Vα chain over the N-terminal half of the peptide and the loops of the Vβ chain over the C-terminal half of the peptide. Diagonal binding is traditionally quantified by the crossing angle, measured by a line through the centers of mass of the two variable domains relative to the line through the peptide backbone ⁹. A large variety of crossing angles have been observed ³, and lately TCRs that bind with reverse crossing angles (Vβ near the peptide N-terminus) have even been reported ^5,10. A complementary means to describe TCR geometries is the incident angle, which describes the inclination of the TCR relative to the plane of the MHC peptide-binding groove ¹¹. However, as we recently demonstrated ⁷, crossing and incident angles alone do not fully describe the variation in TCR binding modes, which limits our ability to derive relationships between TCR binding and function or study their various biological and physical determinants.

Through a structural informatics analysis of TCR-peptide/MHC complexes, here we developed a novel coordinate system to better describe the geometries by which TCRs bind. Our approach identified complexes with different binding geometries which cannot be distinguished by earlier approaches. Demonstrating its utility, we found previously unrecognized differences in the ways TCRs bind class I and class II peptide/MHC complexes. For both classes, TCRs bind in order to maximize their ability to interact with and “read out” exposed peptide side chains. However, because the two classes of MHC proteins present peptides differently, the average TCR binding geometry differs for class I and class II proteins. The extent to which these differences are physically or biologically imparted is unknown: while they likely reflect optimal TCR binding modes given the architectural differences between class I and class II proteins and how they present peptides, they may also be attributable to differences between the CD4 and CD8 coreceptors or co-evolution of TCR and MHC genes. We anticipate that as more data emerges, the coordinate system we developed will be valuable not only in determining this, but also in quantifying other relationships between TCR binding and immunological outcomes. Lastly, we expect that our findings and new coordinate system will be of use in modeling TCR complexes from sequence data, an area of growing interest.

2. Methods

2. 1. Center of mass analysis using a spherical coordinate system

The binding mode of the T cell receptor (TCR) to peptide-major histocompatibility complex (pMHC) was described in a spherical coordinate system by three parameters (r, θ, and ϕ). The center of mass (COM) of the α carbon atoms in the MHC peptide-binding groove (MHC COM) was determined from residues 1–180 for class I proteins and from α chain residues 1–83 and β chain residues 1–93 for class II proteins. The origin of the coordinate system was assigned to the MHC COM. A y axis was defined as a vector along the peptide binding groove using singular value decomposition of the α carbon atoms of the groove α helices ⁹. The x-axis was defined normal to the y-axis in the plane of the MHC helices, and the z-axis normal to the xy plane. The magnitude of the vector connecting the COM of the Vα/Vβ domain of the TCR and the MHC COM was defined as r, and is given as:

$$r=\sqrt{x^2+y^2+z^2}$$

The ϕ angle is the angle between the z-axis and TCR COM – MHC COM vector and is given as:

$$\mathrm{\varphi }={\cos }^{-1}\frac{Z}{r}$$

The θ angle is the angle between the x-axis and the projection of the TCR COM – MHC COM vector onto the xy plane and is given as:

$$\mathrm{\theta }=\mathrm{atan}\frac{y}{x}$$

Spherical plots were generated with OriginPro 2018. The analysis was performed on TCR-pMHC complexes available in the Protein Data Bank (PDB) as of July 2018. This included 121 overall structures with class I MHC and 50 overall structures with class II MHC. PDB codes and data for these structures are listed in supplemental Table S1. A Python script for computing geometrical parameters using PyMOL is available at https://github.com/EsamTolba/TCR_CoM, and the system has been integrated into the TCR structural repertoire database server ¹² at https://tcr3d.ibbr.umd.edu/tcr_com. Due to differences between the Python script and the web code, the two yield slightly different values, but these differences are systematic and do not impact the conclusions; all values reported here herein generated using the Python script.

2.2. Global z-depth analysis

The global z-depth analysis was performed on X-ray crystal structures of pMHC complexes available in the PDB as of November 2018. This included 297 structures for nonamers and 130 structures for decamers bound to class I MHC proteins and 127 structures of peptides bound to class II MHC proteins (all peptide lengths). In each structure, the distance along the z axis between each α carbon and the xy plane of the binding groove was determined. The means of all the z-distance values for each α carbon were calculated for class I and class II proteins and depicted with a blue to red color spectrum respectively on structures of HLA-A2 and HLA-DR1.

2.3. Peptide solvent accessible surface area analysis

The peptide solvent accessible surface area (SASA) analysis was performed with the Python module of the FreeSASA software ¹³ on the same set of pMHC structures used for the global z-depth analysis. The SASA value of each peptide residue was calculated with a 1.4 Å probe radius. The mean of all the SASA values for each residue was calculated and indicated as the radius of the ribbon cartoon representation of the peptide. Solvent accessible surface areas were normalized to glycine-X-glycine tripeptides where X was the amino acid under consideration.

2.4. Statistics

For angular data, the mean direction $\overline{a}$ was calculated as:

$$\overline{a}=atan2\left(\frac{1}{n}\sum _{i=1}^n\sin a_i,\frac{1}{n}\sum _{i=1}^n\cos a_i\right)$$

where a_i represents the value of the measured angle and n is the number of crystal structures examined. The circular standard deviation (v) and circular variance (V) were calculated as:

$$v=\sqrt{-2\ln \overline{R}}$$

$$V=1-\overline{R}$$

where $\overline{R}$ is the length of the calculated the mean direction $\overline{a}$. All calculations were performed using the circular library in the R package. Statistical significance of angular data was calculated using Watson’s two sample test for angular data ^14–17. Statistical analyses of non-angular data were performed with t-tests. Variances were compared with f-tests.

2.5. Principal Component Analysis

For principal component analysis of TCR binding geometries, a secondary structure superimposition of the class I and class II peptide binding grooves onto a reference structure (PDB 1DUZ ¹⁸) was first performed with the MatchMaker function in Chimera ¹⁹ in order to focus on variance in TCR binding geometry. In the case of class II complexed TCRs, following the structural superimposition the coordinates for the MHC molecule were replaced with those of the reference 1DUZ molecule in order to permit proper sequence alignment. Following this, a sequence alignment for all structures was generated with MUSCLE ²⁰. Principal component analysis was then performed with the pc.xray function in the Bio3D R package. For structural insight, projections of the first and second principal component were generated through use of the mktrj.pca function in the Bio3D R package ²¹.

3. Results

3.1. Quantifying the differences seen by visualizing TCRs over MHC proteins

We recently showed that TCR binding to the human class I MHC protein HLA-A*0201 (HLA-A2) occurs with a more restricted positioning than TCR binding to other human class I MHC proteins ⁷. This was apparent from examining the position of the centers of mass (COM) of the TCR Vα/Vβ domains over HLA-A2 (Fig. 1A). In examining this data, we observed that some TCRs were positioned differently over the MHC but nonetheless had similar crossing and incident angles. In examining a larger set of TCR-pMHC complexes, we readily found similar instances of this for TCR complexes with both class I (Fig. 1B) and class II MHC proteins (Fig. 1C). The differences in binding are apparent not only from examining the COMs, but also from visual inspections of the structures. Thus, the commonly used crossing and incident angles are incomplete descriptors of TCR binding geometries.

Figure 1. Traditional analyses of crossing and incident angles do not describe variations in TCR binding modes.

A) Positions of the COMs of TCR variable domains over the class I MHC protein HLA-A2 (left, orange spheres) or other human class I MHC proteins (right, yellow spheres) shows the restricted positioning over HLA-A2 (illustrated using coordinates from PDB ID 1BD2 ³⁴; see also ref. 7). B) The BM3.3-VSV8/K^{b 35} and TK3-HPVG/B35 ³⁶ TCR-pMHC complexes bind with nearly identical crossing and incident angles but are clearly positioned differently over the MHC protein as revealed by superimposing the MHC peptide binding domains (left) or viewing the position of the TCR Vα/Vβ COMs over the pMHC (right). C) As with panel B, the differences in the MS23C8-MBP/DR4 ³⁷ and D2-gliadin/DQ2 ³⁸ TCR-pMHC complexes are not revealed by crossing and incident angles. For panels B and C, PDB codes and angles are indicated in the insets.

To develop an alternative system to describe TCR docking geometries, we considered how to quantify the position of TCR variable domains over MHC proteins. The COM of a class I or class II MHC peptide binding domain provided a convenient point to which the Vα/Vβ COM could be related. We therefore derived a coordinate system with an origin defined by the MHC peptide binding domain COM. A y-axis was defined with a best fit line through the Cα atoms of the two α helices of the binding groove ⁹. An x-axis was defined normal to the y-axis, yielding a plane through the peptide-binding domain. A z-axis was then determined normal to the xy-plane (Fig. 2).

Figure 2. Construction of the spherical coordinate system and definition of terms.

The COM of the MHC peptide binding domain defines the origin, illustrated as a black sphere. A y axis is defined by a vector through the α carbons of the MHC helices, with an x axis defined normal to y, generating an xy plane at the base of the binding groove. A z axis is defined normal to xy. The position of the TCR Vα/Vβ COM, illustrated here as a cyan sphere, defines three parameters: r, the distance from the origin to the TCR COM; θ, the angle formed between the x axis and the projection of the TCR COM onto the xy plane; and ϕ, the angle formed between the TCR COM, the MHC COM, and the z axis. In general, r describes the distance of the TCR from the MHC, θ the rotation of the TCR over the MHC, and ϕ the tilt of the TCR relative to the MHC.

With this coordinate system, we defined three new parameters. The first, r, is the distance between the MHC peptide binding domain COM and the TCR Vα/Vβ COM. The second parameter, ϕ, is the angle formed by the z-axis, the MHC COM, and the Vα/Vβ COM, or the tilt of the TCR relative to the MHC. The third, θ, is the angle formed between the x axis and the projection of the Vα/Vβ COM onto the xy plane, or the rotation of the TCR around the MHC. These three parameters and their determinations from three-dimensional coordinates are shown in Fig. 2.

Examining all three geometrical parameters together illustrated the more clustered nature of the TCR complexes with HLA-A2 vs. TCR complexes with other human class I MHC proteins that was seen earlier through visual inspection (Fig. 3A). The variances for both distance (r) and tilt (ϕ) were statistically different between the HLA-A2 and non-HLA-A2 data (f-test p value < 0.05). The variances of the crossing and incident angles, on the other hand, were indistinguishable between the HLA-A2 and non-HLA-A2 data.

Figure 3. Quantitative comparison of human TCRs bound to HLA-A2 vs. other human class I MHC proteins.

A) HLA-A2 and non-HLA-A2 data compared by the COM analysis system. B) HLA-A2 and non-HLA-A2 data compared by more traditional crossing and incident angles. C) Spherical coordinates better illustrate the differences in the HLA-A2 and non-HLA-A2 data. Rotation (θ) is shown in the circular graph, with tilt (ϕ) indicated by the scale on the left. Distance (r) is indicated by shading as shown in the bottom scale. The graphs are superimposed over the structure of a class I MHC protein, scaled to correctly illustrate the relative position of the TCR COMs.

To enhance visualization of the data, we used spherical coordinates to represent TCR rotation (θ) and tilt (ϕ), with distance (r) represented by shading (Fig. 3C). This analysis recapitulated what was seen by direct visualization of the TCR COMs over the MHC, demonstrating the restricted nature in how TCRs bind HLA-A2 complexes, while adding quantitative information (compare Fig. 1A with Fig. 3C).

3.2. TCRs show MHC class-dependent binding geometries

The geometrical analysis quantified differences in how TCRs bind HLA-A2 and non-HLA-A2 human class I MHC proteins. This difference had been observed previously and attributed to charges on the α1 helix that are unique to the HLA-A2 protein ⁷. To explore if other differences are apparent in the TCR-pMHC structural database, we compared different non-redundant sets of TCR-pMHC complexes, e.g., mouse vs. human class I or class II, self vs. non-self, etc. (supplemental Tables S2, S3). In some cases, the comparisons did not show significant differences, or the datasets were too small to yield conclusions. For example, comparison of human vs. mouse structures of complexes with class I MHC proteins showed no indication of a species-dependence in TCR binding geometries. Comparison of human vs. mouse complexes with class II suggested a more restricted tilt (ϕ) range for mouse class II complexes, but the number of mouse complexes is too small to be statistically relevant (Fig. S1).

However, a clear difference in binding geometries emerged when comparing TCR binding to class I and class II MHC proteins. Although the r distances for both classes were indistinguishable, the ϕ and θ angles diverged (Watson’s U² test for circular means, p values < 0.001). As shown in Fig. 4A, the COMs of TCRs bound to class I MHC tended to cluster over the C-terminal half of the peptide, whereas the COMs of TCRs bound to class II MHC tended to be more centrally located. As with the previous analyses, traditional crossing and incident angles could not distinguish TCR from binding class I or class II MHC proteins (Fig. 4B) (although we note that early structural studies suggested that TCRs might bind class I and class II MHC differently ²²).

Figure 4. TCRs bind class I and class II peptide/MHC complexes with different geometries.

A) Spherical coordinate analysis of TCRs positioned over class I MHC complexes (left) and class II complexes (right). Plot parameters are as in Fig. 3; the structures of the MHC proteins are scaled to correctly illustrate the relative positions of the TCR COMs. In general, TCRs that bind class I are positioned closer to the peptide C-terminus, whereas TCRs that bind class II are positioned more centrally over the MHC. The two outlier points in the class I data are for two TCRs that bind the murine class I MHC protein H-2D^d with “reverse polarity” ⁵. B) Crossing and incident angles do not reveal the distinctions shown in panel A.

3.3. Class I and class II restricted TCRs focus on distinct regions of peptide/MHC complexes to maximize accessibility to peptide

The differences in binding geometries identified for TCRs binding class I or class II MHC proteins suggest TCRs might focus on structural features distinct to the two classes of proteins. To examine this, we first depicted the mean vertical distance between each Cα atom and the xy plane of the binding groove as defined in Fig. 2 (i.e., the distance along the z-axis). Both MHC proteins showed an overall similar topology of the top surface, with the highest spots at 1) the kink region in the α2 helix of class I and β1 helix of class II; and 2) the helical segments preceding the 3₁₀ helices in the α1 helices of class I and class II (Fig. 5A). The average crossing angles for TCRs binding both class I and class II cleanly bisected these high points, indicating that the commonly-observed TCR diagonal binding mode allows the receptor to avoid MHC high points and maximize TCR-peptide interactions, an observation consistent with a hypothesis that first emerged from mutagenesis data prior to the availability of structural information on TCR-peptide/MHC complexes ^23,24.

Figure 5. TCRs bind class I and class II MHC proteins maximize accessibility to the peptide.

A) Top view of the peptide binding grooves of class I (left) and class II (right) MHC proteins. The blue to red spectrum represents the average height of the protein Cα atoms from the xy plane at the base of the binding groove according to the scale indicated. In each panel, the cyan and orange spheres represent the highest two atoms in the binding groove. The red dotted line represents the average of the crossing angles, with the shading showing one standard deviation. The green spheres show the average position of the TCR COMs over the MHC proteins as determined from the mean θ, ϕ, and r parameters, with the green shading representing one standard deviation in this position. B) Average relative solvent accessible surface area (SASA) of nonameric and decameric peptides bound to MHC class I (left) and the core residues (p1 to p9) bound to MHC class II (right), indicating regions of the peptides most exposed. C) Side views of the class I and class II proteins, colored according to depth as in panel A, with the peptide shown as a ribbon whose width is proportional to the SASA values in panel B. The green lines and shading show the convergence of the mean θ, ϕ, and r parameters as in panel A. For both systems, and for both nonamers and decamers presented by class I, the TCR COMs indicate the TCRs bind such that the CDR loops converge on peptide regions that are maximally solvent exposed and between the high points of the flanking α helices.

We next computed the average solvent accessible surface area of each residue of peptides presented by MHC proteins (Fig. 5B). For class I MHC proteins, we focused on the most commonly presented nonameric and decameric peptides, structures of which are highly represented in the Protein Data Bank. For these peptides, which are typically constrained at their termini and bulge away from the MHC binding groove, maximum exposure occurred between positions 4 and 8 for nonamers and positions 4 and 9 for decamers. For class II MHC proteins, which do not bulge from the groove, our analysis revealed alternating patterns of high and low exposure, with highest exposure occurring at positions 2, 5, and 8.

We next determined the average position of TCR Vα/Vβ COMs for class I and class II complexes relative to the MHC COMs from the mean θ, ϕ, and r parameters (Table 1). To control for the possibility that the mean values were influenced by redundant TCR-pMHC structures with the same values (e.g., multiple structures of the same TCR recognizing nearly-identical peptides), we also computed the median values for the class I and class II data. The median values were nearly identical to the means, indicating that the data were not overly influenced by repeated values describing essentially the same structure.

Table 1.

Geometrical parameters describing TCRs binding to class I and class II pMHC complexes

	r (Å)	θ	Φ	crossing	incident
Class I complexes
mean	28.9	77.7°	11.4°	47.0°	10.9°
standard deviation	1.1	65.0°	8.4°	12.3°	6.0°
median	29.1	71.3°	9.0°	44.7°	10.9°
Class II complexes
mean	29.1	201.2°	7.4°	44.7°	11.3°
standard deviation	1.1	88.0°	5.0°	20.0°	8.6°
median	28.9	190.1°	6.3°	41.3°	9.9°

Strikingly, for both class I and class II complexes, the mean geometrical parameters indicated that, on average, TCRs bind such that their COMs and thus their CDR loops are directed towards regions of the peptides that are most solvent exposed. For class I complexes with both nonamers and decamers, the mean TCR COM converged on the bulge of the peptide following position 4, bisecting the high points of the α1 and α2 helices (Fig. 5C, left panel). For class II complexes, the convergence was on the center of the peptide, again bisecting the high points of the α and β helices (Fig. 5C, right panel). In both cases, this convergence would provide the various CDR loops of TCRs (and the hypervariable CDR3 loops in particular) ready access to the most exposed peptide side chains. The differences in TCR binding geometries for class I and class II MHC proteins thus maximize the opportunity for the TCRs to “read-out” features of the peptide antigen.

To independently evaluate differences between class I and class II complexes, we performed principal component analyses (PCA) on the crystallographic coordinates of the variable domains of TCRs bound to class I and class II pMHC complexes. In agreement with the binding geometry analysis, differences were revealed between class I and class II complexes (Fig. 6). The first principal component, which accounts for 48% of the variance, scans along the length of the peptide, with class I focusing primarily on the central and C-terminal regions of the peptide and class II focusing on the central region of the peptide. The second principal component, which accounts for 25% of the total variance, involves scanning to either side of the peptide. This motion would direct the variable domain towards one of the two α helices.

Figure 6. Principal Component Analysis of Class I and Class II Complexed TCR Variable Domains.

A) Principal component analysis of TCR Vα/Vβ Cα atoms after structural superimposition to a reference MHC protein. Data for class I complexes are indicated in green and class II complexes are in pink. The first principal component accounts for 48.3% of the total variance and the second principal component accounts for 24.6% of the variance. B) As in panel A but mapped onto reference class I and class II MHC molecules. The orientation of the axes was determined by mapping the COMs of the TCR variable domains in the projected principal components, and the magnitude determined by the position of the variable domain COM in the crystallographic structure with the highest magnitude in the principal component 1 or 2 axis.

4. Discussion

The extent to which TCR binding geometries impact T cell function has been a source of discussion for many years. The commonly observed diagonal binding mode, which generally places the Vα domain over the N-terminal half of the peptide and the Vβ domain over the C-terminal half of the peptide, has led to suggestions that binding orientations are biologically constrained, implying that variations in binding geometry could influence biological outcomes. Such influences could stem from how the architecture of the TCR-peptide/MHC complex affects the build-up of larger signaling complexes in a cellular interface, to include interactions with co-receptors, adjacent TCR-pMHC complexes, or other signaling modules. Additionally, binding geometry could impact the occurrence of through-protein conformational or dynamic changes that influence signaling ^25–27. Indeed, some receptors that bind with unusual, “outlier” binding modes have been shown to yield limited or altered T cell responses ^4,5 and select TCRs associated with autoimmunity bind with unusual geometries ²⁸.

An issue that complicates our understanding of how TCR structure influences geometry is that the traditional topological descriptors of TCR binding geometry – most commonly the crossing (or diagonal) angle – are incapable of characterizing different binding geometries. We first observed this in our study of how features on the common class I MHC protein HLA-A2 contribute to biased receptor binding geometries ⁷. To better describe TCR binding, we developed a coordinate system that describes the position of the center of mass of the TCR variable domain over a peptide/MHC complex. Designed to supplement rather than replace other descriptors (such as crossing and incident angles), the system allows us to quantify differences in binding modes that other descriptors do not always reveal. For example, our system identifies the distinct, restricted nature in how TCRs bind HLA-A2 compared to other class I MHC proteins. As the TCR-pMHC structural database grows, our system thus provides a new framework with which questions that link TCR binding geometry to function can be addressed, such as whether other MHC haplotypes “instruct” TCRs to bind in certain fashions, as well as the physiological consequences of restricted or geometrically altered binding.

One intriguing observation we did make is a divergence in the geometries with which TCRs bind class I and class II pMHC complexes. TCRs that bind class I MHC proteins tend to bind in a way that focuses the center of the variable domains on the C-terminal half of the peptide, whereas TCRs that bind class II MHC proteins tend to bind more centrally. These differences are correlated with differences in how class I and class II MHC proteins present peptides: owing to how their termini are fixed in the binding groove, peptides presented by class I proteins bulge away from the binding groove base, with the bulge usually beginning at peptide position 4 and continuing through the penultimate residue. In contrast, peptides bound to class II MHC proteins are extended and lie flat in the groove, with maximum side chain exposure in the middle and near both termini. In recognizing class I and class II MHC proteins then, TCRs bind in a fashion that maximizes the opportunities for the CDR loops to access the most exposed peptide side chains while simultaneously avoiding the high points of the flanking α helices.

Our observations do not speak to the origin of the differences between class I and class II binding. It is likely that these differences reflect, in part, the most optimal “fit” between TCRs and class I vs. class II ligands given the need to engage solvent exposed regions of the peptide. The differences may also be imparted by differences between the CD4 and CD8 co-receptors and/or co-evolution between TCRs and MHC proteins, which could bias TCRs towards binding class I and class II ligands differently ^6,8. Regardless of the underlying mechanisms, the distinctions and their connection to peptide exposure highlight the crucial importance of the peptide in TCR binding and subsequent peptide-directed immune responses.

In addition to addressing biological questions, we expect our findings and the coordinate system will be of use in modeling TCR-pMHC complexes, particularly in capturing distinctions between complexes with class I and class II MHC proteins. With the growing ease of obtaining TCR sequence data, there is increasing interest in obtaining structural information about TCRs and their complexes in order to assess properties such as specificity and cross-reactivity ^29,30. To this end, strategies for structurally modeling TCR-pMHC complexes have been described ^31–33. The distinction we observe between class I and class II complexes may be of use in constraining such models to increase the accuracy of structural predictions.

Abbreviations

MHC

major histocompatibility complex

pMHC

peptide/MHC complex

TCR

T cell receptor

TCR-pMHC

complex between a TCR and a pMHC

COM

center of mass

PCA

principle component analysis

PDB

Protein Data Bank