Structure of Staphylococcal Enterotoxin E in Complex with TCR Defines the Role of TCR Loop Positioning in Superantigen Recognition

T cells are crucial players in cell-mediated immunity. The specificity of their receptor, the T cell receptor (TCR), is central for the immune system to distinguish foreign from host antigens. Superantigens are bacterial toxins capable of inducing a toxic immune response by cross-linking the TCR and the major histocompatibility complex (MHC) class II and circumventing the antigen specificity. Here, we present the structure of staphylococcal enterotoxin E (SEE) in complex with a human T cell receptor, as well as the unligated T cell receptor structure. There are clear structural changes in the TCR loops upon superantigen binding. In particular, the HV4 loop moves to circumvent steric clashes upon complex formation. In addition, a predicted ternary model of SEE in complex with both TCR and MHC class II displays intermolecular contacts between the TCR α-chain and the MHC, suggesting that the TCR α-chain is of importance for complex formation.


Introduction
T cell activation is a fundamental event in the immune response, which requires T cell receptor (TCR) recognition of a peptide presented by the major histocompatibility complex (MHC) [1].The T cell receptor is a transmembrane protein with an extracellular antigen-binding domain, consisting of an αand a β-chain, each comprising one variable (TRAV and TRBV) and one constant (TRAC and TRBC) domain [2].There are three loops, the CDR1-3, in the variable domains of TCR that predominantly recognize the peptide-MHC complex.A fourth loop, HV4, is also variable, but generally not used for antigen recognition [3].The HV4 loop has, however, been suggested to be important when T cells are activated by certain bacterial toxins, called superantigens (SAgs) [4,5].Superantigens are immune stimulatory toxins that bind directly to TCR and MHC as unprocessed proteins, and hence prevent the TCR from recognizing the peptide presented by MHC [6][7][8].By this cross-linking event, superantigens are capable of evoking an immune response of large proportions, resulting in host disease [9].The superantigens produced by Staphylococcus aureus and Streptococcus pyogenes are divided into five evolutionary groups (I-V), depending on sequence similarity, and each group has structurally diverse ways of engaging TCR and MHC class II [10].Staphylococcal enterotoxin E (SEE) belongs to group III, which generally has one binding site to TCR, to the TRBV domain [11], and two distinct binding sites to MHC class II [12][13][14][15].The first MHC binding site is located in the N-terminal domain of the SAg, which binds to the α-chain of MHC, with relatively low affinity [15], and the other is zinc bridged and located in the C-terminal domain of the SAg, which engages the β-chain of MHC with high affinity [14].SEE has until now evaded crystallographic studies, but the structure of the closely related superantigen SEA has been determined [16].Still, neither SEA, nor SEE has been structurally elucidated in complex with a T cell receptor, even though structures of these complexes are of particular interest since a potential drug for cancer treatment consists of a chimera of these two superantigens [11,17,18].
Here, we present the structures of a T cell receptor, both by itself and in complex with staphylococcal enterotoxin E. The SEE-TCR structure describes the very first interface between a group III superantigen and its TCR.Upon superantigen engagement, the T cell receptor undergoes no global structural changes, but there are several smaller movements of the TCR-loops and a larger conformational change in the HV4 loop upon complex formation.This suggests that SEE-recognition is dependent on flexibility in the T cell receptor antigen binding surface.Moreover, structural alignment of several TRBVs suggests that the conformation of the CDR2 loop is particularly important for SEE recognition.In addition, we generated a TCR-SEE-(MHC) 2 model, which shows that TCR, SEE and one MHC form a triangular complex with an additional interface between TRAV and the β-chain of MHC, as observed for related superantigens [19].

Protein production and purification
The T cell receptor with variable domains TRAV22/TRBV7-9 was prepared as described previously [20,21], apart from minor changes stated here.Expression was carried out in Escherichia coli BL21 (DE3) Star (Invitrogen).Inclusion bodies were solubilized in 50 mM Tris-HCl pH 8.0, 6 M guanidinium chloride, 100 mM NaCl, 10 mM EDTA, 5 mM DTT and then refolded in 100 mM Tris-HCl pH 8.0, 5 M urea, 400 mM L-arginine, 0.83 mg/l cysteamine.Purification was done by anion exchange chromatography on an ÄKTA explorer (GE Healthcare), with a Resource Q 6 ml column (GE Healthcare), followed by size exclusion chromatography on a Superdex 200 column (GE Healthcare) in TBS buffer.Purified staphylococcal enterotoxin E, a gift from Active Biotech Research AB, was prepared according to a previously published protocol [22].

Crystallisation and structure determination
The TCR was crystallized by vapour diffusion at 5.0 mg/ml in 22% PEG 2,000 MME, 0.1 M ammonium chloride pH 8.5 and 0.1 M NaCl.For cryo protection, 20% glycerol was used and data were collected at 100 K and 1.000 Å at ESRF beamline ID 23-1 (Table 1).A high resolution pass with an oscillation angle of 0.1°, using helical oscillation, and a low resolution pass with an oscillation angle of 0.5°on a single location was collected from a single crystal.The TCR data were indexed, integrated and merged with XDS [23] and aimless [24,25], within autoPROC [26].Subsequently, the TCR structure was solved using molecular replacement in Phaser [27], with TRAV, TRAC, and TRBC domains from 2IAL [28] and TRBV domain from 2DX9 [29] as search models.Differing amino acids were initially omitted and then built using Buccaneer [30,31].Refinement was carried out in autoBUSTER [32] and refmac5 [33] with manual modeling in Coot [34], with anisotropic B-factors for protein atoms and isotropic Bfactors for solvent molecules.Finally, a composite omit map was generated using CNS [35] with 5% of the structure omitted.The final TCR model comprised residues 2-203 in TCRα and 3-243 in TCRβ, along with one glycerol molecule and 309 waters.Ramachandran statistics, calculated with SFCHECK [36], for the TCR structure were 91.9% in preferred, 7.1% in allowed, and 1.0% in generously allowed regions.
The co-crystallization of SEE and TCR was carried out by vapour diffusion at an equimolar ratio of the two proteins at 6.5 mg/ml total protein concentration, in 15% PEG 20,000, 0.1 M glycine pH 9.0 and 0.1 M NaCl.Crystals were soaked in cryo protectant containing 20% glycerol and flash-frozen in liquid nitrogen.Data of the SEE-TCR complex were collected at ESRF beamline ID 23-1 at a wavelength of 0.9763 Å and 100 K, with 0.5°oscillation, from a single crystal (Table 1).The data were processed using XDS [23] and aimless [24,25] within the CCP4 suite [37] with 5% of the data chosen as a subset for cross-validation.The SEE-TCR structure was solved using the TCR structure presented here and a polyalanine model of SEE, generated from the SEA structure 1ESF [16] in Phaser [27], and amino acids were built manually in Coot [34].Refinement was carried out using refmac5 [33] and autoBUSTER [32], with automatically generated TLS groups, and manual modeling was done in Coot [34].A composite omit map was generated using CNS [35] with 5% of the structure omitted.The model In addition, the model includes 50 water molecules, two sodium ions and one zinc ion.Ramachandran statistics, calculated with SFCHECK [36], were 89% in preferred, 10.3% in allowed, and 0.7% in generously allowed regions.

Computational modeling using Rosetta Dock
The initial TCR-SEE-(MHC) 2 model was generated by modelling the missing residues 44-49 in SEE according to the SEA D227A -MHC structure (1LO5) [15].The MHC structure and its position with respect to the SAg in the SEA D227A -MHC structure was used to generate the TCR-SEE-MHC model and side chains were energy minimized with the Rosetta Docking Prepack Protocol.Constraints based on the SEA D227A -MHC interface (S1 Table ) were placed and 5000 models were generated using the high resolution protocol in Rosetta Dock [38][39][40][41].Models were scored based on the interface between SEE-TCR and MHC and their RMSD from the starting model (S1 Fig) , and the model with lowest interface score was chosen and refined by generating 1000 additional models, again with the high resolution protocol only.This was repeated, and the final model chosen was tested by running an unconstrained docking simulation according to the full Rosetta Dock protocol, which revealed a funnel-like shape of the plot of interface score versus RMSD for the models (S1 Fig) , suggesting that the final model is energetically favourable, given that the MHC is located approximately in that region.For the high affinity zinc-bridged MHC site, the MHC molecule from the SEH-MHC class II structure (1HXY) [14] was placed according to the position of the MHC with respect to the SAg in the SEI-MHC structure (PDB: 2G9H) [42], this due to the identical conservation of the zinc site between SEE and SEI.The zinc-coordinating residues, His187, His225 and Asp227 in SEE, as well as His81 in MHCβ were constrained with respect to the Zn 2+ and due to the conservation of Gln135 between SEE and SEH, this residue was constrained to contact the peptide as seen in the SEH-MHC structure [14] (S1 Table ).Modelling of this site was carried out in the same manner as for the low-affinity site (S2 Fig) .The models of the low and high affinity sites were then aligned and combined to a single TCR-SEE-(MHC) 2 model.

Overall structure of the T cell receptor
The crystal structure of extracellular domains of a human chimeric TCR was determined to 1.35 Å resolution.The TCR was crystallized in space group P2 1 with one molecule in the asymmetric unit.Data collection and refinement statistics are summarized in Table 1.The variable domains TRAV22 and TRBV7-9 were isolated from two TCRs specific against HLA-A2 in complex either with a telomerase peptide (sequence ILAKFLHWL) or a survivin peptide (sequence ELTLGEFLKL), respectively.The TCR exhibited the conventional TCR fold described previously [2].Briefly, the TCR is a heterodimeric protein consisting of an αand a β-chain, each with one membrane-distal variable domain (TRAV and TRBV) and one membrane-proximal constant domain (TRAC and TRBC).All four domains have Ig-like folds, almost exclusively consisting of β-sheets.The variable domains have four loops each, CDR1-3 and HV4, and together this surface is responsible for pMHC recognition.
The structure of TCR in complex with staphylococcal enterotoxin E In order to study superantigen recognition by TCR, the structure of the TRAV22/TRBV7-9 TCR was determined in complex with staphylococcal enterotoxin E, to a resolution of 2.5 Å (Fig 1A).The complex crystallized in space group P2 1 2 1 2 1 with one protein complex in the asymmetric unit.Data collection and refinement statistics are summarized in Table 1.The T cell receptor exhibited the same fold as described in the previous section, and the superantigen shares a similar fold with other bacterial superantigens, as first described for SEB [43].SEE consists of two domains, an N-terminal domain resembling an oligosaccharide binding fold and a C-terminal β-grasp motif.The N-terminal domain consists of β-sheets (β 1 -β 5 ) and a short α-helix (α 3 ), and the C-terminal domain consists of an antiparallel β-sheet (β 6 , β 7 , β 9 , β 10 , and β 12 ) packed against three α-helices (α 2 , α 4 and α 5 ), as well as a small two-stranded β-sheet (β 8 , β 11 ).SEE is structurally and sequentially similar to SEA, with RMSD values between main chain atoms of 0.79 and 0.77 for the respective copies of SEA in the published three-dimensional structure (PDB ID: 1ESF) [16], and a sequence identity of 82%.In general, SEE engages the TRBV domain of TCR with the TCR binding site described for most other bacterial  There are six previously determined superantigen structures from S. aureus with TCR available in the Protein Data Bank (SEB, SEC3, SEG, SEH, SEK and TSST-1), and two from S. pyogenes (SPE-A and SPE-C) [6,44,[48][49][50][51][52].A structural alignment using PROMALS3D was preformed to investigate whether the residues that are important for TRBV recognition in SEE are conserved among the other superantigens [53] However, in seven out of the nine structures investigated, the corresponding residue to Asn25s is present and forms hydrogen bonds to TCR in their respective structures.Moreover, Tyr64s is rather conserved (five out of nine) but is not involved in binding TCR in the other structures.Taken together, SEE utilizes, to a large extent, different amino acids to bind TCR compared to the previously determined SAgs, except for Asn25s.This correlates well with little overlap in TRBV specificity between SEE and the other SAgs [54].However, superantigens in group III, as SEA, will possibly have a more SEE-like TCR recognition interface.

Structural rearrangement in the TCR upon enterotoxin binding
Differences in the overall structure of the unbound and SEE-bound TCR are small, with RMSD for main chain atoms of 1.3 and 0.91 for TCRα and TCRβ, respectively.This might be due to the introduced disulfide bond between the TRAC and TRBC domains, which potentially could lock the constant domains in certain positions and thus inhibit potential conformational changes upon binding [75].In the TRAV domain, there are no large changes in the loop conformations.The majority of the loop rearrangements occur in the TRBV domain, with the HV4 loop and CDR1 loops in different positions, whereas CDR2 is only slightly shifted (Fig 4A ).
The largest change upon SEE binding is seen in the HV4 loop, which previously has been suggested to be of importance for SAg engagement [4,5].This loop undergoes a conformational change (Fig 4B and 4C), as indicated by with an RMSD of 2.4 Å for residues 69-74 between bound and unbound TCR, aligned with respect to the Cα atoms of the TRBV domains.The loop moves away from the superantigen upon complex formation to avoid steric clashes (Fig 4).The largest movement in the HV4 loop occurs at Gly73b with a magnitude of 4.3 Å.Subsequently, this results in two hydrogen bonds both between Arg70b and Gln28s, as well as van der Waals contacts from Arg70b to Tyr32s and Trp63s (S2 and S3 Tables).As mentioned above, the corresponding glycine in TRBV7-2 Ã 01 (Gly84, IMGT numbering) [63] has previously been shown to be involved in TRBV specificity for SEE [5].When substituting TRBV7-2 Ã 01 to TRBV7-2 Ã 02, with a Gly84Glu substitution, SEE could no longer activate the T cells.This is likely due to a combination of electrostatic repulsions caused by Glu34s, which is located close to the HV4 loop and steric hindrance because of a restrained HV4 loop upon substitution [5].Moreover, Glu69b adopts a single conformation in the bound TCR structure instead of a dual, as in the unligated TCR, since one of the conformations clashes with Tyr32s in SEE.The single conformation allows Glu69b to interact with Arg27s, Gln28s and Tyr32s (S2 and S3 Tables).Pro71b is significantly shifted in order to accommodate for SEE, resulting in van der Waals contacts with Tyr32s.In contrast to this, no conformational changes have previously been seen in the HV4β loop upon SAg engagement in other studied TCR-SAg complexes (SEB, SEC3, SPE-A and SEG) [44,48,50,52,56,76].There are two sets of structures available with TCRs bearing the same TCR as studied here (TRBV7-9), alone and in complex with pMHC.An analysis of differences within each pair, aligned with respect to the Cα atoms of the TRBV domains, reveals no movements in the HV4 loop [29] [55].Taken together, this suggests that the flexibility and hence the possibility to move the HV4 loop is of importance for SEE recognition, but is necessarily not directly coupled to general T cell activation by superantigens nor by conventional antigens.

Model of the quaternary TCR-SEE-(MHC) 2 complex
As mentioned, the group III SAgs is distinguished by one TCR binding site and two MHC class II binding sites [12][13][14][15].One site is between the N-terminal domain of the SAg and the αchain of MHC utilizing Lys39 on MHC class II [15], and the other is in the C-terminal domain of the SAg to the β-chain of MHC, bridged by a zinc ion utilizing His81 on MHC class II [14].In contrast, the group II SAgs, such as SEB, only has one TCR and one MHC class II binding site, to the α-chain of MHC [37].Due to the similarity between SEE and SEA in their dependence of His81 and Lys39 for MHC binding, and the sequential conservation of both the N-terminal and zinc dependent C-terminal MHC class II binding sites (Fig 5A ), it is likely that SEE, as SEA, is able to cross-link two MHC class II molecules, as also previously been suggested (Fig 5B) [15,77,78].A zinc ion is visible in the SEE-TCR structure presented here, coordinated by residues His187s, His225s, and Asp227s, as well as Asp225b from the β-chain of a symmetry-related TCR, due to crystal packing.The three equivalent residues in SEA, SEI and two in SEH are known to coordinate the zinc ion in these SAgs [42,79,80] and are crucial for biological activity [14,81].Since the zinc binding site to MHC is conserved (Fig 5A ), it is likely that SEE will engage the β-chain of MHC, using its C-terminal domain, in a manner similar to what has been observed for SEI [42].In addition, many of the known MHC-coordinating residues in the N-terminal MHC binding site (to the α-chain) are conserved between SEE, SEA and SEB (Fig 5A ).
Recently, the three-dimensional structure describing the ternary complex of TCR-SEB-MHC was published [19].The structure clearly showed that in addition to contacts between the superantigen and the TRBV domain, the TRAV domain of TCR contributed to the complex formation by contacting the MHC class II molecule directly.This supports previous findings by Andersen and co-workers, who proposed that there is an interface between the TRAV domain of TCR and MHC, upon SEC3 binding [82].Since SEE has a similar MHC binding site as SEB, a putative contribution from the TRAV domain upon complex formation is not unlikely.This is supported by previously published data where the particular sequence of the TRAV domain influenced the level of T cell activation by SEE in cells expressing in MHCβH81Y [83].By combining the available structure of the superantigens SEA D227A -MHC (PDB ID: 1LO5) [15] and the SEI-MHC (PDB ID: 2G9H) [42] as well as the SEH-MHC structures (PDB ID: 1HXY) [14], we were able to build an initial structural model of a TCR-SEE-(MHC) 2 complex.This was then used to generate models of the complex using Rosetta Dock (  [38][39][40][41].The initial model showed distances larger than 7.5 Å between the TRAV domain and the MHC β-chain, suggesting that the TRAV may not be able to have a stabilizing effect on ternary complex formation (Fig 5B).However, after running docking simulations, the final model displayed a pMHC shifted towards the TRAV domain resulting in a contact area between these proteins (Fig 5C).This is in line with what has been observed for SEB, where these domains can interact (Fig 5D) [19].Thus, it is plausible that this TRAV-MHC contact is able to form in solution for SEE as well.Residues that are involved in the TRAV-MHCβ model interface are Asp66, Glu69, Gln70, Arg72, Ala73, Asp76 and Thr77 in the MHC β-chain, to mainly the CDR2 loop of the TRAV domain, but contacts to HV4 and CDR1 are also present.Notably, all of these residues except Gln70 and Arg72 are in contact with TRAV in the TCR-SEB-MHC structure [19].Compared to the SEA D227A -MHC structure, our modeled SEE-MHCα interface buries the same surface area (approximately 1120 Å 2 ) but has considerably fewer hydrogen bonds and the MHC has moved significantly.The TRAV-MHCβ interface buries approximately 1020 Å 2 in total, which is more than double the size observed in the TCR-SEB-MHC structure.This is likely the cause for the quite large differences seen in the SEE-MHCα interface compared to the SEA D227A -MHC and SEB-MHC interfaces.It is also worth considering that no flexibility between SEE and TCR is allowed in the Rosetta model.In solution, flexibility in this region could allow for a more SEB-like position of MHC but still allowing for the larger TRAV-MHCβ interface.In line with these results, SEE has previously been suggested to activate T cells partly dependent on the TRAV domain, in addition to the clear specificity for the TRBV domain [83].From our structural and modeling data, we can conclude that it is likely that an interface can be formed between the TRAV domain and MHCβ, and hence that the main reason for the observed TRAV specificity for SEE is a direct contact between the TRAV and the MHCβ, instead of the speculated indirect conformational changes of the TRBV domain [83].

Conclusions
The structure presented here features staphylococcal enterotoxin E in complex with a T cell receptor.This structure, in combination with the unligated TCR structure, shows that flexibility of the HV4 loop is of importance for SEE binding to TCR.In addition, the structure suggests that SEE discriminates between TRBV domains primarily by a mechanism dependent on CDR2 loop conformation and HV4 loop flexibility, and secondarily by CDR2 and HV4 loop sequence.Lastly, a computer model of the TCR-SEE-(MHC) 2 complex concludes that an interface between the TRAV domain of TCR and the MHC molecule, upon binding to the low affinity site of SEE, is possible and that it could stabilize ternary complex formation.

Accession Numbers
Coordinates and structure factors have been deposited in the Protein Data Bank with accession numbers 4UDT and 4UDU for TCR and SEE-TCR, respectively.

Fig 1 .
Fig 1. X-ray structure of the SEE-TCR complex.(A) Overall structure of the complex, with SEE in beige, the TCR α-chain in purple and the β-chain in blue.(B) Close-up of the SEE α 2 -helix and contacting residues in TCR, (C) the hydrophobic patch, (D) the α 4 -β 9 loop, and (E) the upper part of the α 5 -helix.Hydrogen bonds are marked as dotted lines.doi:10.1371/journal.pone.0131988.g001

Fig 2 .
Fig 2. Presentation of the buried surface areas in the SEE-TCR interface.(A) The areas in the TRBV domain which are buried upon binding are marked in colors corresponding to the CDR1 loop (red), CDR2 loop (green), FR3 region (blue), HV4 loop (purple) and FR4 region (orange).(B) The buried surface area in SEE is marked in colors corresponding to the α 2 -helix and following loop (purple), the hydrophobic patch consisting of the β 2 -β 3 (green) and β 4 -β 5a (blue) loops, the α 4 -β 9 loop (red), and the upper part of the α 5 -helix (orange).doi:10.1371/journal.pone.0131988.g002 . As shown in S3 Fig, most of the residues in the SEE-TCR interface are not conserved among the other SAgs (S3 Fig).

Fig 4 .
Fig 4. Comparison between the TCR and SEE-TCR structures, aligned with respect to the TRBV domain, in cross-eyed stereo view.(A) Overall differences between the TRBV domains, with the SEE-TCR structure shown in beige and blue and the TCR structure in grey.(B) Close-up of the HV4 loop with residues shown as sticks.(C) Close-up of the HV4 loop shown as Cα trace with 2F o -F c electron density maps shown for the two structures, with the SEE-TCR map in blue and the TCR map in grey.doi:10.1371/journal.pone.0131988.g004 Fig 5B, S1 Fig, S2 Fig and S1 Table)

Fig 5 .
Fig 5. Modelling of the TCR-SEE-(MHC) 2 quaternary complex.(A) Sequence alignment of SEE with SEA, SEB, SEH and SEI, displaying the conservation of both MHC binding sites in SEE, made using ClustalW2 [84, 85].The N-terminal binding site to the MHC α-chain is marked in green and the C-terminal binding site to the MHC β-chain is marked in purple.(B) The initial model of TCR-SEE-(MHC) 2 .The TCR is shown in purple and blue (TCRα and TCRβ respectively), the SEE in beige, and MHC molecules in green.(C) The final model of TCR-SEE-(MHC) 2 .(D) The TCR-SEB-MHC structure, with SEB shown in orange.doi:10.1371/journal.pone.0131988.g005

Table 1 .
Data collection and refinement statistics.