Somatic genetics of CDR3 control TCR V- domain rotational probability and germline CDR2 scanning of polymorphic MHC

34


Introduction:
T-cell antigen receptors (TCR) and antibodies (sIg) are individualized within each precursor of a given T-cell or B-cell clone.TCR (α/β type) have genetically variable (V) domains, wherein complementary determining regions (CDR1-CDR3) contain closest amino acid (a.a.) contacts with the peptide (p) plus major histocompatibility complex (MHC) protein (together abbrev., pMHC) composite ligand.CDR1 and CDR2 are encoded in the germline, via the particular Vregion DNA segment involved in the RAG1/RAG2 recombination mechanism responsible for somatic construction, together with the D-and/or J-segment(s), of the third, CDR3 [1][2][3][4][5].This paper seeks to understand two puzzles of the TCR-pMHC interaction, wherein a novel examination of the first can be used to re-examine the second in the context of existing evidence.Firstly, compared to antibodies and indeed other protein:protein binding reactions, TCRs display quite low (µM) binding affinities for pMHC ligands [6][7][8].Nevertheless, like antibodies, TCRs have exquisite specificity, where single peptide or MHC a.a.changes can dictate T-cell reactivity or non-reactivity [9][10][11][12][13][14][15][16][17].Second, the nature of T-cell selection in the thymus prior to seeding secondary lymphoid tissues is thought to operate on the basis of this same relative TCR:pMHC affinity, where overt 'self' peptide plus self-MHC recognition (negative selection) leads to deletion from the repertoire, as does a complete lack of recognition (positive selection) [4].This intermediate TCR:pMHC binding in the fetal thymus forms the basis of 'simultaneous' self plus pathogen recognition in adulthood, i.e., the MHC-restricted adaptive immune response [5,18].Somatic CDR3 loops are roughly positioned atop the peptide component in solved TCR:pMHC complexes [9][10][11][12][13][14][15][16][17][18]; although more recent investigations have challenged the notion that the most diverse component of the ligand (the peptide) is particularly interfaced with the most diverse regions of the receptor (CDR3α/β) [12,19].
Here, the STCR-Dab (www. opig.stats.ox.ac.uk/webapps/stcrdab/) was used to identify all current TCR complexes involving HLA-DR and HLA-DQ (3.0 Å resolution cut-off).Note, also included are PDB 4H1L, at 3.3 Å, and 3T0E, at 4.0 Å (see below).A simple relationship between the observed (i.e., measured) CDR3-CDR2 "pitch" angle for a given V domain and its predicted (i.e., calculated) CDR3-CDR2 pitch angle was observed [20].Here, a linear relationship between calculated pitch and a new measure, "dV"-by multivariable calculus, was found.As is discussed, dV interprets V-domain orientation into a rotational probability involving an apparent CDR2-MHC α-helix scanning function (dθ).While dV were unique for each TCR on each pHLA-DR, one TCR displayed a dramatic restriction in dV for Vβ.This was isolated to a charge-relay H-bonding mechanism for CDR3β; hence, the chemistry of "somatic-TCR" dictates positioning of "germline-TCR".Within the seven pHLA-DQ structures, the two highly-restricted dV TCR (PDB 4OZG & 4OZH) displayed distinct, yet functionally similar mechanisms that shifted the same TCR:MHC H-bond by one MHC a.a.position (relative to the nominal dV structure, 4OZF).This involves an additional Hbond (MHC:MHC) in suitable charge-relay mechanisms of 4OZG and 4OZH.

Results:
PDB files of solved TCR:pHLA-DR and TCR:pHLA-DQ structures were used to investigate Vdomain geometry amongst the available complexes, i.e., involving similar (but different) TCR, and/or similar (but different) pMHC.All of these structures share the canonical (diagonal) orientation of the TCR over pMHC, which was one of the earliest observed similarities between different complexes [13].As indicated in the summary tables (Tables 1A-1D; separation into different tables is simply for clarity) all sequences for each component of all of these structures are available under the appropriate PDB file name at the NCBI (www.ncbi.nlm.nih.gov).

TCR-pMHC Geometry
Shown in Figures 1A and 1B is an example of the geometry analysis based on fixed a.a.positions in the HLA-DR/HLA-DQ grooves (e.g., PDB 1J8H).We originally used Euler angles (www.mathword.wolfram.com/EulerAngles.html) to establish the basic method (for a TCR-Vα cohort in pHLA-A2 complexes) [20]; here, the method was modified for the available solved TCR-pMHC-II structures (summarized in Tables 1A-1D).In brief, the analysis is based on measuring three angles corresponding to the twist (ω), tilt (λ), and sway (σ) of each V domain over the pMHC.For HLA-DR a vector from the DRA a.a.N62 alpha-carbon (N62:Cα) to DRB1/3/5 a.a.D66 (D66:Cα; Cα used for measurements unless otherwise noted) bisects the MHC-groove from the DRA α-helix to the DRB1/3/5 α-helix, then the angle at D66 to the Vα central cysteine (C22) is computed with the VMD angle-label tool.This ω-angle can be seen in Figure 1A  Specifically, twist is defined by D66:N62:C23, tilt by F40:D66:C23, and sway as D66:F40:C23 (Table 1D).We formulated the equation: that predicts the pitch of a given V domain (pitch = angle ϕ) from the ω, λ, and σ (orientation).
By contrast, measuring the ϕ-angle is based on finding the closest contact between the domain CDR2 and the α-helix that is opposite CDR3 across-the-groove, N62 (for Vα), or D66 (for Vβ) (yellow highlighted in Tables 1A-1D).

Triple Integrals
We observed that the closest CDR2α contacts with the α2-helix 'across-the-groove' from the conserved R65 contact with CDR3α in HLA-A2 structures specified a polymorphic region from a.a.H/R151 to a.a.A158 [20].Similarly, the CDR2α closest contacts range for the DR structures here (Tables 1A-1C) implicate a polymorphic DRB α-helix range from a.a.E69 to a.a.T77; for CDR2β, a DRA α-helix contacts range: a.a.Q57 to a.a.K67.For the DQ complexes: CDR2α contacts range is the polymorphic DQB α-helix a.a.E69 to a.a.D76; and for CDR2β, DQA: Q57 to H68, is also polymorphic (Table 1D).Thus, the hypothesis that if these α-helical a.a.are indeed swept/scanned by CDR2 of the V-domain, that one should be able to model such with integration from spherical coordinates in each structure [22,23].To normalize the equation Vβ measures 22.12 Å)-this is the rho (ρ) distance in the equation.The other measure was the angle φ, which is simply the difference from 180° of the previously determined tilt angle (λ).
The other variables were derived by trigonometry (see Fig. 2B).This is integration of a volume element in spherical coordinates (all dV values in Tables 1A-1D).
For compositional reasons throughout "the triple integral dV" may be referred to as simply "dV."As discussed above it is not surprising that the (π • r / 4) values were close to the actual distances between CDR2 contacts in the aforementioned MHC α-helix ranges using this approach, e.g., 13.27 Å, Figure 2B.Thus, the volume of the cone slice for each V-domain was determined upon integration for the three spherical coordinates, rho (ρ), theta (θ) and phi (φ), where the upper limit of each integral is derived from the measured ρ and φ values of each V domain as shown: cysteine is historically used as the center of any given V domain [10] and each V-domain coneslice is thus the volumetric-density through its CDR2 scanning path (e.g., Fig. 2B).The calculus interprets a comprehensive geometry of the V domain into a probability of scanning using only the ground-state structure, i.e., without crystallographic data on theoretical conformerssuggesting here, a classic protein function, allostery [24].
Accordingly, the mean dV of these TCR V-domains was 955 Å 3 , excluding the unusually large dV values.Overall, it says something about TCR:pMHC that may not be intuitive-that a "flush" Vdomain geometry could limit CDR2 "finding" an optimal binding interface with side-chains involving the aforementioned α-helix regions.This is the clonotypic nature of TCR function.In all cases, the CDR3α:peptide bonds are either non-existent, or isolated from the MHC:peptide centres.In 4E41, 6CQQ and 6CQR, CDR3α may link to CDR3β through both bonding nearby with peptide (blue shading).For 2IAM, 1J8H, and 1FYT there are bonds between intra-V (2IAM), or intra-and inter-V (1J8H, 1FYT), but these are not linked to the peptide (different blue shade).N62 is connected to a peptide:peptide H-bond (4H1L), or to an inter-Vβ bond (3T0E); in both cases these potential networks extend to the Vβ directly binding with N62.1ZGL is the only structure with an inter-V bond potentially linked to a direct Vdomain interaction with MHC, while 4H1L is unique in not showing any DRB H-bonding involving the CDR3 (these two structures involve non-DRB1 isotypes).In 6CQL and 6CQN there is a potential DRB1:peptide link with an MHC:MHC bond, but this does not involve the TCR.The 6CQ series all have CDR3β:peptide bonds also involving a peptide:peptide bond; this is not found in any of the other complexes (mauve shading).Finally, and most significantly here, in 2IAM/2IAN, 1FYT and 3T0E the DRB1:peptide centre is potentially connected through a CDR3β:peptide bond with an MHC:MHC H-bond (Table 2A & 2B, green shades).

CDR3:pMHC Chemistry
H-bonds computed in Swiss are potential H-bonds, and to assess a given network the chemistry of each bond must be examined [24,25].The analyses of Tables 2A-2C indicated that a potential H-bonding network where CDR3β is connected through peptide-bonding to DRB1:peptide and DRB1:DRB1 H-bonds might correlate with the highly-restricted dV of Vβ in 2IAM (297 Å 3 ) and 2IAN (270 Å 3 ).However, these networks would need to be distinct from this same type of potential network observed in 1FYT and 3T0E (Table 2B), which have dV for Vβ of 1747 Å 3 and 1024 Å 3 , respectively (Table 1B).Shown is the analysis of the 2IAM-Vβ versus 1FYT-Vβ for contacts made by each CDR3β (Fig. 3).Illustrating the importance of the Swiss computation, in Figure 3C it was anticipated that the R71K polymorphism would effect three H-bonds-involving K71 in a manner similar to R71.
However, as is shown in Figure 3F, R71K actually shifted peptide contacts to Q70; in-fact, K71 did not show any H-bonds.Importantly, this confirms the change in peptide conformation between the two structures as originally reported [13], and suggests that alloreactivity of 1FYT/1J8H involves the single G98:0 to K315:N bond in maintaining the ≈ 1740 Å 3 dV of the two Vβ (Table 1B).Still, the chemistry of each possible bond must be examined to arrive at feasible mechanisms; arguably, this is necessary to understand binding.

H-Bonding Mechanisms
Shown in Figure 4A is a suitable reaction mechanism for the 2IAM CDR3β binding to the peptide (triosephosphate isomerse, 15mer)-HLA-DR complex (PDB ref.,15).When considering the probability of hydrogen bonds in a mechanism, all possibilities were drawn using ChemSketch (ACD Labs; www.acdlabs.com/resources)and electron pathways traced using standard evaluations of electron configuration [24,25].Thus DRB1 (B1) Q70:OE1 would be in an H-bond with B1R71:NH1 in the absence of the peptide, with the downstream effect of relieving the charge on the B1R71:NH2.Bound peptide has the same effect on R71:NH2, where the N30:O attacks the R71:NH2 (orange arrows).Here, the peptide would indirectly break the H-bond between Q70 and R71 (orange blocks).When the TCR binds, the βH96:NE2 charge is preferentially attacked by the N30:O (purple arrows); this reverses the previous bond between the peptide and DRB1 (purple block) and has the downstream effect of re-forming the intra-MHC bond between Q70 and R71 (purple blocks on the orange blocks).Note, this is the only obvious mechanism that relieves both the H96 and R71 charges, and would be favoured over an A31:O attack on H96:NE2 by the neighbouring βY95 H-bond with A31:O (Fig. 4A).Therefore, just three of the possible H-bonds are predicted; and the overall effect is a charge-relay [28,32,33] between the MHC, the peptide, and the TCR.
Physiologically, the charge on R71:NH2 could be initially neutralized via the neighbour Q70:OE1-subsequently replaced during processing by CLIP, and/or at HLA-DM exchange, with antigenic peptide (ref.9).During T-cell conjugation with the antigen presenting cell, the TCR would replace the A30:O to R71:NH2 attack with an A30:O to H96:NE2 attack, which relays the charge back to R71:NH2-subsequently, neutralized again by Q70:OE1.While the hallmark charge-relay network of the 'catalytic triad' within serine proteases stabilizes formation of a covalent acyl-enzyme intermediate [24]; here, a relay is utilized in a similar (albeit, noncovalent) role.Undoubtedly, this would still be a particularly stabilised transition-state, i.e., given that these H-bonds are de-localised across 3/5 components of the structure [24][25][26][27].such that an attack directly on R71:NH2 is not possible.Thus, because of a remaining charge on R71:NH2, the T97:O attack on N312:ND2 (with its downstream effects) would not be favoured (orange blocks).While similarly to 2IAM there is a neighbouring H-bond (in this case to Q311), the net loss of a CDR3β H-bond clearly distinguishes the two mechanisms, and shows that the 1FYT CDR3β does not favour a charge-relay mechanism; i.e., would not have the stabilized CDR3β tether at B1R71-ostensibly, essential chemistry driving restricted dV in 2IAM.Note that the 2IAN-Vβ mechanism appears to be identical to that of 2IAM (structures differ by a single, sufficiently distant, a.a. of the peptide).Next, 3T0E also showed the TCR:peptide:MHC:MHC type of potential network by contacts analysis (Table 2B). Figure 5A shows the a.a.involved in this potential network using Swiss, and the proposed mechanism is shown in Figure 5B.Note that the DRB1-4 allotype of 3T0E has the R71K polymorphism, and that the involved peptide a.a. is R9.In terms of distinguishing the mechanism from 2IAM, the question is whether the CDR3β N100 H-bond to R9 could exist with a B1:peptide H-bond.In fact, if the Vβ closest contact N100:O is to attack the positive charge of R9:NH2 (purple arrows), it would have to compete (indirectly) with the H-bond of B1Q93:OE1 to R9:NH1 (orange arrows).Together with the B1Q99:NE2 possible bond to the same N100:O, the B1Q93 bond would be favoured; thus, the 3T0E mechanism is not similar to 2IAM, most notably because CDR3β is not likely to form an H-bond with the peptide.Note that in the related structure, PDB 3O6F, this conclusion holds as well, where B1D95:OD2 has a possible H-bond to R9:NH2 at 1.6 Å; again, not favouring a CDR3β N100:O bond to the peptide [28].
Accordingly, H-bonding was analysed with Swiss-Deepview as before (Fig. 6).While very similar, we noted the absence (Fig. 6A & B) of the DQA1, T61:OG1 role in H-bonding to Vβ, R109:NH1 (Fig. 6C).This is accomplished in 4OZH by preferential H-bonding to Vβ a.a.S110 (Fig. 6A), or sterically in 4OZG by Vβ a.a.F110 (Fig. 6B).Thus, only in 4OZF is T61:OG1 free to H- and V-domain "dV" calculated by: This suggests that reduced pitch limits dθ, i.e., the CDR2-scanning function.In retrospect, this might confirm intuition (excluding the broader TCR-CD3 complex, of course) on possibilities for a mechanism involving just the five components.Indeed, H-bonding chemistry which effects relatively simple physics is a hallmark of quite diverse protein machinery [24,30].Also, TCR had k values that varied in either direction-indicating that conformation adjustments might moderate flush and open pitch without much increasing or decreasing of the dV, viz., R 2 ≈ 0.900 (Suppl.1.II).Thus, pitch calculated from a given V-domain's twist-tilt-sway (orientation) might be a kind of "hidden" correlate of TCR-selection; although there is not available crystallography on any thymic (selecting) ligands [31].For 2IAM, there are data for the uncomplexed TCR, and the CDR3β backbone is displaced ≈ 3.4 Å upon binding pMHC, while βY95 moves ≈ 9.0 Å to form the H-bond shown (Fig. 3A); indeed, the key to understanding the TCR may be in this relationship between the "two" induced-fit TCR conformations [32][33][34][35].Suggested here is a conservation/approximation of dV for V-domains in TCR binding the selecting (thymus) and activating (peripheral) pMHC.As shown (Fig. 3E vs. 3F), alloreactivity displays shared dV accomplished by quite different binding chemistry.Alternatively/historically, intrinsic TCR affinity (i.e., equilibrium and rate binding constants specified by the α/β TCR protein) has been used to explain both thymic TCR selection and peripheral TCR recognition [4,36,37].Clearly TCR distinguish rare target (agonist) pMHC from thousands of nonagonist pMHC on the APC surface; and this is not mutually exclusive of whether or not a physical force (TCR loading) is an obligate component of TCR function [8,37].In these and other regards, there is a stereochemical alternative to an affinity-limited binding reaction [25][26][27].
Briefly, we assume TCR:pMHC reactions involve a high(er) energy "scanning" conformerbecause, scanning (leading to a suitable CDR2:MHC interface) has the effect of lowering the transition-state free energy, ∆∆G ǂ (i.e., Curtin-Hammett control) [25,27,38].For example, in TCR like E8 (of 2IAM/2IAN) very little CDR2β scanning is apparently needed, due to the effect of the proposed charge-relay stabilised CDR3β:pHLA-DR transition-state.Thus, dV is a universal consequence of CDR3 chemistry; but usually the reaction requires more CDR2-scanning (i.e., usually V-domains have larger dV values; Tables 1A-1D).Recently, ensemble refinement of crystal data with MD simulations has suggested conformational diversity in the microsecond range, but the conformational changes implicated here-in would be on the millisecond to minutes scale; in this regard, NMR of membrane-bound receptors offers promise [39], and may elucidate the potential of internal water(s) in these H-bonding networks [32].Finally, spherical to clonotype [41].While there are myriad downstream implications, taken together, these data support that V-domain rotation and germline to germline contacts between TCR and pMHC both depend upon CDR3 H-bonding with highly-conserved MHC α-helix motifs.Thus, unlike TCR-affinity, V-domain dynamics are clonotypic.In this regard, similar to the class-I, R65-motif [20,42], the class-II motifs presented here are found in Galago sp.(NCBI Acc.No. AAA96291) as well as in both tarsiers and lemurs; suggesting conservation for at least 63 million years [43,44].

Contacts Analysis
All measures were performed with Swiss or VMD-1.9.

CDR3 Joint Analysis
Nucleotide sequences for all CDR3 of TCR were specified from PDB files (Suppl.1.III).TCR a.a.
sequences were reverse translated using the SMS tool at www.bioinformatics.org.These were then imported into IMGT algorithms for joint analysis (www.imgt.org).Amino acid sequences of resulting CDR3 joints were determined by the IMGT algorithm; consensus IMGT numbering.

H-Bonding Mechanisms
Hydrogen (H) bonds were estimated either with VMD, or (more accurately) with Swiss.In Swiss, H-bonds are computed after computing hydrogens to the structures.H-bond distances were used as a factor in determining suitable organic reaction mechanisms, where relevant side-chains were reproduced with Chemsketch (www.acdlabs.com/resources).Standard evaluations of electron configuration per relevant atoms were used to predict electron flows [25].

Figures
Figure 1 The

Figure 1
Figure1The twist/tilt/sway of TCR-Vα and -Vβ relative to pMHC-II (A & B, respectively).Illustrated by the example of 1J8H.This same analysis was performed on all 19 structures (Tables1A-1D).All angle measurements were from Cα in VMD-1.9.1 (www.ks.uiuc.edu)used to examine the diversity through the 38 V-domains: (i) in-plane to the MHC-groove {twist = ω}, with (ii) displacements perpendicular to the groove {tilt = λ}, including (iii) side-to-side variation {sway = σ}.The a.a.positions used as coordinates for angular measures (dotted orange vectors) across structures were fixed and are labelled (see text).The measured incline of a V domain, or "pitch" {pitch = ϕ m }, is shown by white dotted vectors and could be approximated by the equation:ϕ c = [σ ÷ (λ + σ)]ω (see text, Tables1A-1D).Side chains of a.a. in measurements are shown by CPK licorice; Cα-backbones are in new cartoon and labelled; peptides are in black.Note from A to B the view of the structure rotates 180°.Figures are the original output of the PDB file as analysed in VMD-1.9.1.Best viewed at 100%.
across structures and use defined geometry, the central cysteine (C22/23) was chosen as a point rotating from the fixed a.a.position that defines twist, tilt, and sway (Figs.1 & 2).As shown in Figure2A& 2B, the coordinates model a 'slice' of a cone, where for each of the 38 Vdomains a measured distance was taken between the C22/23 and said α-helix a.a.(e.g., 2IAM

Figure 2
Figure2Derivation of the V-domain rotational volumetric-density equation.The same a.a.positions used for the tilt (λ) angles were used to define the phi (φ) angles (e.g., 49.85° for this structure, 2IAM Vβ, gold) by subtraction of λ from a theoretical 180° vector.This is illustrated with actual positions in 2IAM (A), showing the 180° approximating, 176.32° vector; the ~ 90°, 88.91° angle; the ~ 14.27 Å, 14.14 Å line segment; and the ~ 16.90 Å, 17.29 Å line segment, i.e., to show the essential geometry within a structure; these are more accurately calculated by trigonometry in the derivation (B).Additionally, the rho (ρ) segment from the appropriate α-helix position (see λ definitions) to a given V-domain's C22/23 was measured (here, 22.12 Å).The cone (full rotation) includes a probability (slice volume) through a scanning path (dθ) for CDR2. Figure (A) is the original output of the PDB file as analysed in VMD-1.9.1.Figure (B) is a diagram constructed in MS-3D-paint.Best viewed at 100%.
Since ρ and φ are measured for each V domain, derivations of upper limits for the first and third integrals simply provide trigonometric relationships.Two example solutions are given below (parameters as defined, Figs. 1 & 2; text)-for all 38 triple integral solutions, see Suppl.1.I*.The (π • r / 4) circumference segment is used as the upper limit of the dθ integral because it is accurate to a path, i.e., a distance; formally (for θangle in degrees): π • r (θ) / 180 = arc length in Å [22].Hence, dφ is the only integrand for conversion (multiplying by π ÷ 180) to yield the cubic angstrom (Å 3 ) unit of volume.The C22/23 of the 1ZGL Vα.Importantly, with the exclusion of 1ZGL (an apparent outlier) there is a linear relationship between calculated pitch (ϕ c ) and the dV triple integral (see plotted values from Tables 1A-1D in the Suppl.1.II)-corresponding to the equation: y = 83.60x-719.40;where R 2 ≈ 0.900, by linear regression analysis.The lowest dV structures have the lowest calculated pitch All these structures, even those with the highest resolution, are subject to limitations of crystallography; and indeed, computed H-bonds are based on these coordinates.Nevertheless, a correlation between relative dV and contact distances for hydrogen bonds of the five components could be investigated; see Tables 2A-2C.For example, asking if particularly strong (close) H-bonds are linked to a restricted dV? 2IAM-Vα 2IAM-Vβ *for all (n = 38) triple-integral solutions, see Suppl.1.I.The structures were examined in Swiss-PDB Viewer/Deepview-v4.1 (www.wpdbv.vital-it.ch); in Swiss H-bonds can be computed after hydrogens are added based on coordinates; complexity of certain structures (2IAN, 1ZGL) precluded computing hydrogens, and the shown H-bonding distances are thus ≈ 1 Å larger (Tables2A, 2B).Here, an obvious feature of all the complexes is that there are two principal foci of H-bonding; one involving N62 of DRA (A1) (shaded orange in the tables), which may include a separate secondary grouping at N69 (light-orange shade), and the DRB1 (B1) centre involving R71 (green shades).Note, R71 is polymorphic, so structures involving allotypes/isotypes that have a different a.a. at pos. 71 (1J8H, 3T0E, 1ZGL, 6CQQ & 4H1L) involve a different a.a.(1J8H, 3T0E, 6CQQ), or simply do not have a corresponding betachain H-bonding centre (4H1L & 1ZGL).Superimposed upon the H-bonding centres are the H-bonds between the CDR3α and peptide, the CDR3β and peptide, any peptide to peptide Hbonding, any intra-or inter-bonds between the TCR V-domains, any MHC to MHC H-bonding, and direct H-bonds between TCR V-domains and the MHC (Tables 2A-2C, columns).

Figure 4 H
Figure 4 H-bonding mechanism for CDR3β binding to peptide and MHC.ChemSketch drawn mechanisms are shown for 2IAM (A) and 1FYT (B) and were based on analyses in VMD and Swiss (Fig. 3).Not necessarily to scale.Electrons shown with arrows (purple, involving the TCR; orange, related to the peptide).Possible H-bonds are shown in blue (darker shade for key TCR bond) and are crossed with the color-coded block line if not probable (see text).Best viewed at 125%.

Figure 5 H
Figure 5 H-bonds and H-bonding mechanism for 3T0E CDR3β.Swiss was used as before (non-involved a.a.deselected for simplicity) to compute possible H-bonds for 3T0E (A).Mechanism for 3T0E TCR binding via CDR3β to peptide and DRB1 (B) shows possible H-bonds in blue, with improbable bonds blocked with grey lines (ChemSketch).Electron flow shown with purple arrows (TCR), or orange (MHC).CDR3β in yellow (top) or a.a.labelled yellow (bottom); DRB1 in mauve (top) or labelled a.a. in pink (bottom); peptide by atom (top) or a.a.labelled grey (bottom).Figure (A) is the original output of the PDB file as analysed in Swiss-Deepview-v4.1.Figure (B) is a diagram constructed in ChemSketch.Best viewed at 100%.
Figure 5 H-bonds and H-bonding mechanism for 3T0E CDR3β.Swiss was used as before (non-involved a.a.deselected for simplicity) to compute possible H-bonds for 3T0E (A).Mechanism for 3T0E TCR binding via CDR3β to peptide and DRB1 (B) shows possible H-bonds in blue, with improbable bonds blocked with grey lines (ChemSketch).Electron flow shown with purple arrows (TCR), or orange (MHC).CDR3β in yellow (top) or a.a.labelled yellow (bottom); DRB1 in mauve (top) or labelled a.a. in pink (bottom); peptide by atom (top) or a.a.labelled grey (bottom).Figure (A) is the original output of the PDB file as analysed in Swiss-Deepview-v4.1.Figure (B) is a diagram constructed in ChemSketch.Best viewed at 100%.
bond to the pivotal arginine (CDR3β, R109); note, this shift from R109 contacting MHC at a.a.N62 to T61 indirectly disrupts the MHC:MHC component of the charge-relay mechanism (Fig. 7C vs. 7A & B).Crucially, the chemistry of S110 versus F110 versus A110 in these TCR is ultimately the result of somatic genetic construction of the CDR3β loop [Suppl.1.III; ref. 4].

Figure 6 H
Figure 6 H-bonds for three different TCR CDR3β, (from germline-identical TCR) on the same pHLA-DQ.(A) 4OZH, showing S110:T61 H-bonds (orange arrow), (B) 4OZG, showing bulky F110 (orange arrow); both of which could preclude (C), the conserved R109 H-bonds to T61 (orange star).Swiss-Deepview analysis as previously; DQ and CDR3β backbones in white ribbon; peptides in yellow wire; involved side chains in CPK; computed possible H-bonds and distances in green.Figures are the original output of the indicated PDB files as analysed in Swiss-Deepview-v4.1.Best viewed at 150%.
coordinates have been used by others to calculate TCR center-of-mass variation in TCR:pMHC complexes [40].These investigators did not consider Vα and Vβ independently, although Hoffmann et al., had shown the angle between Vα and Vβ characterized different TCR in pMHC complexes; allowing these researchers to group a panel of TCR into six different clusters linked twist/tilt/sway of TCR-Vα and -Vβ relative to pMHC-II (A & B, respectively).Please see manuscript .pdffor full caption.

Figure 4 H
Figure 4

Figure 5 H
Figure 5

Figure 6 H
Figure 6

Figure 7 H
Figure 7 For the DQ Vβ's, angle a.a.are the same as in DR structures.
Availability of data and materials: PDB files are public and available at www.ncbi.nlm.nih.gov;allanalyticsdata are either in the paper, or supplementary materials; Competing interests: Dr. Murray and Xenolaüs Genetics LLC declare no competing interests regarding the research, or its publication; Author Contributions: J.S.M. did the research and wrote the manuscript.-1.9.1 software (www.ks.uiuc.edu)wasusedforPDBfiles downloaded via NCBI (www.ncbi.nlm.nih.gov)from the RSCB-PDB (www.rscb.org);viewsnormalizedwith the VMD xyz-axis tool; alpha carbon (Cα) main-chains in new cartoon; all alleles named per NCBI annotation.Euler's methods (www.mathword.wolfram.com/EulerAngles.html)were the basis for the specific angle analyses, as previously reported.Briefly, three angles corresponding to the twist, tilt, and sway of each domain over the pMHC were measured from fixed Cα through the 19 structures: (i) in-plane to the MHC-groove {twist = ω}, with (ii) displacements perpendicular to the groove {tilt = λ}, including (iii) side-to-side variation {sway = σ}.The a.a.positions used as coordinates for angular measures across structures were fixed; see previous.The incline of a V-domain, {pitch = ϕ m }, was approximated (calculated) by the equation:ϕ c = [σ ÷ (λ + σ)] ω (see also Results, Tables1A-1D).Pitch was also measured by using the closest determined CDR2 contact Cα for an angle across-the-groove to the closest CDR3 contact with an α-helix side-chain (vertex), then back to said CDR2 closest contact within the opposite αhelix (≈ 2-fold symmetry); angular value in degrees via the VMD angle-label tool.Triple integrals for all 38 V-domains based on the fixed geometry were solved as described in Results (Suppl.1.I).Linear regression analysis by MS-Excel (Suppl.1.II); and statistics by paired twotailed Student's t-test (www.insilico-net/tools/statistics/ttest).