The T Cell Receptor β-Chain Second Complementarity Determining Region Loop (CDR2β) Governs T Cell Activation and Vβ Specificity by Bacterial Superantigens*

Superantigens (SAgs) are microbial toxins defined by their ability to activate T lymphocytes in a T cell receptor (TCR) β-chain variable domain (Vβ)-specific manner. Although existing structural information indicates that diverse bacterial SAgs all uniformly engage the Vβ second complementarity determining region (CDR2β) loop, the molecular rules that dictate SAg-mediated T cell activation and Vβ specificity are not fully understood. Herein we report the crystal structure of human Vβ2.1 (hVβ2.1) in complex with the toxic shock syndrome toxin-1 (TSST-1) SAg, and mutagenesis of hVβ2.1 indicates that the non-canonical length of CDR2β is a critical determinant for recognition by TSST-1 as well as the distantly related SAg streptococcal pyrogenic exotoxin C. Frame work (FR) region 3 is uniquely critical for TSST-1 function explaining the fine Vβ-specificity exhibited by this SAg. Furthermore, domain swapping experiments with SAgs, which use distinct domains to engage both CDR2β and FR3/4β revealed that the CDR2β contacts dictate T lymphocyte Vβ-specificity. These findings demonstrate that the TCR CDR2β loop is the critical determinant for functional recognition and Vβ-specificity by diverse bacterial SAgs.

a processed peptide antigen presented within the distal groove of self major histocompatibility (pMHC) complexes. A general consensus for typical TCR-pMHC recognition is that germ line encoded complementarity determining region loops (CDR) 1 and CDR2 of both the TCR ␣and ␤-chains primarily recognize the ␣-helices of self pMHC complexes, while peptide specificity is mainly driven by the CDR3 loops (1,2).
Bacterial superantigens (SAgs) are potent immunostimulatory toxins that distort the molecular 'rules' which dictate normal TCR-pMHC class II (pMHC II) interactions. SAgs function by binding to lateral surfaces on both TCR ␤-chain variable domains (V␤) (3) and pMHC II complexes (4), resulting in the activation of T cells in a V␤-restricted manner (5). In the standard TCR-SAg-pMHC II model, the TCR ␤-chain, and pMHC II ␣-chain are wedged apart by the SAg such that the CDR3 loops of the ␣␤ TCR cannot interact directly with the antigenic peptide contained within the pMHC II complexes (6). Because SAgs bind the germ line encoded V␤ region, while essentially ignoring the highly diverse CDR3␤ loop, the frequency of T cells responding to SAg exposure exceeds that of conventional peptide antigens by many orders of magnitude. Through this mechanism, SAgs may ultimately mediate a cytokine storm disease known as the toxic shock syndrome (7).
The bacterial pathogens Staphylococcus aureus and Streptococcus pyogenes are each capable of producing multiple distinct bacterial SAgs that fall into at least five distinct evolutionary groups, designated as I through V (Fig. 1A) (7,8), that share a conserved three-dimensional fold (9) (Fig. 1B). Despite this conserved structure, SAgs can bind to pMHC II through diverse mechanisms (10 -16). In terms of TCR recognition, all characterized SAgs interact with the apical portion of the TCR ␤-chain, but this occurs with distinct geometries (6,(17)(18)(19)(20)(21) (Fig. 1B). For example, the Group I SAg toxic shock syndrome toxin-1 (TSST-1) binds to CDR2␤ and FR3␤ (20); Group II SAgs such as staphylococcal enterotoxin (SE) B, SEC3 and streptococcal pyrogenic exotoxin (Spe) A bind pri-marily to CDR2␤, FR3␤, and HV4␤ (6,19,21); the Group IV SAg SpeC binds to CDR1␤, CDR2␤, CDR3␤, and HV4␤ (21); and the Group V SAg SEK binds to CDR2␤, FR3␤, and FR4␤ (18). Although the V␤ binding site for the Mycoplasma arthritidis mitogen (MAM) nearly overlaps with that of SEB, MAM represents a structurally and phylogenetically distinct SAg that also interacts directly with V␣ (22). Taken together, the one common feature for all of the structurally characterized bacterial SAgs is that each engages the CDR2 loop of V␤ (Fig. 1B), yet the broad functional relevance of CDR2␤ for this large family of toxins is still unknown.
The TSST-1 and SpeC SAgs are two distantly related SAgs, both of which are highly specific for T cells bearing the human V␤2 (hV␤2) TCR, but structural studies indicate that the binding footprints for this common ligand are distinct (20,21). Human V␤2 is somewhat unique as this ␤-chain contains insertions on its amino acid backbone that are lacking in other human ␤-chains. These insertions include Phe 27a in CDR1␤ found uniquely in hV␤2, and Ser 52a in CDR2␤ and Tyr 56 in FR3␤ found only in hV␤2 and hV␤4. The presence of Phe 27a and Ser 52a result in longer than typical CDR1␤ and CDR2␤ loops with non-canonical structures. Therefore, the restrictive V␤ profiles of TSST-1 and SpeC are thought to be due to peculiar V␤ amino acid insertions. More promiscuous SAgs, however, such as SEB, that exhibit a broader range of V␤ targets interact primarily with main-chain atoms of V␤ chains, are thought to recognize V␤s predominantly through a conformational-dependent mechanism (reviewed in Ref. 9).
Within protein-protein interfaces, subsets of the interfacial residues that contribute disproportionately to the overall affinity of the interaction are termed hot spots, and discrete clusters of hot spot residues are termed hot regions (23,24). Biochemical evidence indicates that TSST-1 however, contains two distinct binding hot regions that target residues within CDR2␤ and FR3␤, and these two hot regions exhibit positive binding cooperativity (25). Although SpeC engages all 3 CDRs as well as HV4␤, this SAg contains a functional hot region including SpeC Tyr 15 and SpeC Arg 181 that primarily engages the hV␤2.1 Ser 52a insertion of CDR2␤ (21,26), providing an explanation as to why SpeC is largely specific for V␤s 2 and 4 (27), both of which contain the Ser 52a insertion.
In this work, we sought to further unravel the molecular basis of SAg selectivity for TCR ␤-chains. Exhaustive mutational analysis of the wild-type hV␤2.1 chain expressed from human T cells indicates that the non-canonical length of the CDR2␤ loop is a critical factor for the functional engagement of both TSST-1 and SpeC, and a predicted hotspot within FR3␤ was also critical for TSST-1 function. Since all structurally characterized SAgs engage CDR2␤ (6, 18 -21) (Fig. 1), we hypothesized that the CDR2␤ loop was a key determinant for V␤ discrimination, which is the hallmark activity of SAgs (28). To test this, we created domain swapping mutations between the Group V SAgs staphylococcal enterotoxin K (SEK) and streptococcal pyrogenic exotoxin I (SpeI), which contain distinct SAg domains that interact exclusively with either CDR2␤ or the FR3␤/FR4␤ regions (8,18). Analysis of V␤ expansion by these hybrid SAgs establishes that the CDR2␤ loop is the critical determinant for dictating V␤ specificity.

EXPERIMENTAL PROCEDURES
Crystallization, Data Collection, Structure Determination, and Refinement-TSST-1 and hV␤2.1 proteins were produced and purified as described previously (20,25). A 1:1 TSST-1/hV␤2.1 complex was prepared by mixing the two components together, with hV␤2.1 in slight excess (molar ratio of 1:1.2, respectively), followed by purification on a Superdex S200 10/300 column (GE Healthcare) equilibrated in 50 mM Tris, 25 mM sodium chloride, pH 7.5. The purified TSST-1/hV␤2.1 complex was concentrated to 6.3 mg ml Ϫ1 . Crystals were obtained by hanging drop vapor diffusion, where 1 l of complex was mixed with 1 l of the reservoir solution composed of 22% polyethylene glycol 8000 and 0.46 M lithium sulfate. Crystals were flashed cooled in mother liquor containing 20% glycerol. A data set was collected at the National Synchrotron Light Source, Brookhaven National Laboratory, beam line X6A at 100 K. The data were reduced with iMOSFLM and SCALA in the CCP4 suite of programs (CCP4, 1994). The structure was solved by molecular replacement using the program Phaser with the atomic coordinates of the complex between TSST-1 and the hV␤2.1 affinity-matured variant D10 (Ref. 20; PDB accession code 2IJ0) as a search model. The structure was refined using REFMAC and model building performed using Coot. Coordinates have been deposited in the Protein Data Bank with the accession code 3MFG.
Cloning Procedures-Standard DNA manipulations were performed as described (29) using enzymes supplied from New England Biolabs (Pickering, Ontario) in accordance with the manufacturer's instructions. Oligonucleotides were obtained from Sigma Genosys (Burlington, Ontario). PCRs were performed in a Peltier Thermocycler (MJ Research, Miami, FL) with Vent DNA polymerase (Invitrogen) and PCR products were purified using the QIAquick PCR purification kit (Qiagen Inc., Mississauga, ON). All cloned PCR products were sequenced in their entirety at the Robarts Research Institute Sequencing Facility (London, Ontario, Canada) to ensure correct mutations and PCR fidelity. Escherichia coli XL1blue and BL21(DE3) was cultured aerobically in Luria Bertani broth (Difco Laboratories Inc, Detroit, MI) at 37°C, and solid media was obtained by the addition of 1.5% (w/v) Bacto-agar (Difco). Kanamycin (50 g/ml) was used as selective agent as required. All reagents were made with water purified through a Milli-Q water purification system (Millipore, Mississauga, ON).
Construction of Wild-type and Mutant hV␤2.1 vectors-The full-length wild-type hV␤2.1 cDNA (26) was directionally subcloned into the KpnI and BamHI sites of the expression vector pMAX (Amaxa). The various hV␤2.1 mutant constructs were generated using an overlapping megaprimer PCR method using the oligonucleotides listed in supplemental Table S1. All hV␤2.1 mutations were verified by sequencing both strands.
Construction of hV␤2.1-expressing HuT78 T Cells and Activation Measurements-To generate functional epitopes of SpeC and TSST-1 on the surface of hV␤2.1, the HuT78 T cell line (30) was used to express hV␤2.1-chain conjugated to en-dogenous hV␣. 20 g of pMAX vector harboring either wildtype or mutated hV␤2.1 constructs was electroporated into 8 million HuT78 T cells using 250 V and 250F (Bio-Rad). HuT78 T cells transfected with pMAX::eGFP alone was used as negative control. Surface expression of hV␤2.1 paired with endogenous hV␣ was confirmed using FACS analysis with FITC-conjugated anti-V␤2 antibody (eBioscience) (data not shown). Anti-V␤2 antibody and anti-CD28 antibody were coupled onto beads (Dynabeads M-450 Tosylactivated, Invitrogen) as per the manufacturer's instruction. The final preparation of anti-V␤2/anti-CD28 beads were resuspended in complete R-10 media. 24 h after electroporation, transfected HuT78 T cells were counted, and 200,000 cells were incubated with SpeC or TSST-1 at 1 g/ml, or anti-V␤2/anti-CD28 beads at one bead per transfected HuT78 T cell for 18 h in 5% CO 2 at 37°C. Activation was monitored using IL-2 ELISA (eBioscience).
Superantigens-Cloning of the wild-type genes for SpeC, TSST-1, SpeI and SEK has been described (8,18,25,26). Mutagenesis was conducted using megaprimer-PCR designed to introduce domain-swapping mutants between SpeI and SEK using oligonucleotides listed in supplemental Table S2. Recombinant SAgs were produced by overexpression from E. coli BL21(DE3) with N-terminal His 6 tags, purified by nickel affinity chelation chromatography, and His 6 tags were removed by cleavage of an engineered tobacco etch virus (TEV) protease cleavage site using autoinactivation resistant His 6 ::TEV protease, as described (8,18). All recombinant SAgs ran as discrete homogenous bands by SDS-PAGE.
Primary Cells and Quantitative Real-time PCR-Experiments using human lymphocytes were reviewed and approved by The University of Western Ontario Research Ethics Board for Health Sciences Research Involving Human Subjects. Ficoll gradient purified human PBMCs were stimulated ex vivo in 24-well microtiter plates (1 ϫ 10 6 cells/well) in R10 media (26) with the various SEK or SpeI wild-type or hybrid proteins at 1 g/ml, in triplicate. Cells were incubated in 5% CO 2 at 37°C for 4 days. Afterward cells were harvested and total RNA prepared using RNeasy Spin columns (Qiagen). RNA (500 ng) was reverse transcribed using Superscript II (Invitrogen), and qPCR was performed on cDNA using a specific forward primer for hV␤5, hV␤6.7, hV␤9, an hV␤21.3, coupled with a reverse C␤ transcript specific primer. qPCR reactions were performed using SybrGreen chemistry (Bio-Rad) on a Rotogene 6000 instrument, and the V␤ specific transcript levels were normalized to a C␤1/C␤2 transcript primer pair as a measure of total TCR ␤-chain expression using oligonucleotides listed in supplemental Table S3.
Statistics-Results are represented as the mean Ϯ S.E. where statistical significance between two conditions was determined by the Student's t test using Prism 4.0 (GraphPad Software).

RESULTS AND DISCUSSION
Crystal Structure of Wild-type hV␤2.1-TSST-1 Complex-TSST-1 is unique among the SAg toxin family and is classified as the only Group I SAg (Fig. 1A). TSST-1 is believed to be responsible for essentially all cases of the menstrual-associ-ated TSS cases, and many non-menstrual TSS cases (7). It seems counterintuitive that TSST-1, while likely the most clinically significant bacterial SAg, is also one of the most V␤ specific, being exclusive for hV␤2-expressing T cells (31).
Previous work using yeast display of hV␤2.1 coupled with affinity maturation experiments identified and characterized a number of high-affinity hV␤2.1 variants that bound TSST-1 with increasing affinities up to 12,800-fold for the D10 variant (25,32). The D10 variant was subsequently used to solve the co-crystal structure with wild-type TSST-1, and modeling of the wild-type hV␤2.1 onto this complex revealed the structural basis for how TSST-1 engages the human hV␤2 TCR (20). Here we report the crystal structure of TSST-1 in complex with the wild-type version of hV␤2.1 called Ep8, previously used for stabilization on the surface of yeast. The amino acid sequence of Ep8 is identical to naturally occurring hV␤2.1 except for two mutations (A13V and S88G) that would fall into the V␤:C␤ interface and allowed for stabilized expression on yeast, but the mutations do not occur within the TSST-1 interface (32). The TSST-1/hV␤2.1 crystal structure was refined to 2.3 Å resolution. Data collection and refinement statistics are shown in Table 1.
Functional Epitope Mapping on the Surface of the hV␤2.1 TCR for TSST-1 and SpeC-From prior co-crystallization studies (12,16,20,21), and using the wild-type hV␤2-TSST complex solved here, coupled with corroborating mutagenesis data (11,26,(33)(34)(35), supramolecular T cell activation complexes can be modeled for both TSST-1 and SpeC (Fig. 2). For comparison, a typical TCR-pMHC complex of conventional antigen recognition, where the CDR3 loops from both ␣and ␤-chains center on the antigenic peptide is shown in Fig. 2A (36). For TSST-1 (Fig. 2B) and SpeC (Fig. 2C), both SAgs engage the TCR V␤ and pMHC II molecules using lateral surfaces, forming a bridge (e.g. TSST-1) or a wedge (e.g. SpeC) between the two immune receptors. In both cases, the SAg displaces the CDR3 loops away from the antigenic peptide preventing peptide recognition by the TCR, while SpeC contains an additional high-affinity, zinc-dependent MHC II interaction (16,34).
Crystallographic analysis for TSST-1 and SpeC provides us with the molecular contacts required for SAg engagement of hV␤2, as well as the plasticity of the SAg binding interface on the surface of this common ligand (Fig. 2, D and E). Although the wild-type Ep8-TSST-1 complex is essentially identical in orientation to the high-affinity D10-TSST complex (20), it is radically different from the other characterized SAg-V␤ structures (Fig. 1B). TSST-1 binds more laterally to the apical region of V␤, interacting with a contiguous segment of residues stretching from Gly 51 to Lys 64 comprising the CDR2␤ loop and FR3␤ region of hV␤2.1 (Fig. 2D). The engagement of hV␤2.1 by SpeC is more similar to other characterized SAgs, except SpeC binds in a slightly altered orientation with a larger binding interface engaging residues from all three CDR␤ loops and HV4␤, which coalesce into a distinct binding region on the surface of hV␤2.1 (Fig. 2E) (21).
To understand the molecular requirements for hV␤2 selectivity and activation by TSST-1 and SpeC, 24 surface residues in the wild-type hV␤2.1 involved in binding these SAgs were TCR CDR2␤ Recognition by Superantigens FEBRUARY 11, 2011 • VOLUME 286 • NUMBER 6 targeted for alanine-scanning mutagenesis (Fig. 2, D and E). Three naturally occurring alanines present in hV␤2.1 (Ala 29 , Ala 54 , and Ala 96 ) were instead replaced with both glycine and valine. In addition, three deletion mutants were generated including the non-canonical insertions Phe 27a and Ser 52a in CDR1␤ and CDR2␤ loops, respectively (21), as well as the unique insertion Tyr 56 in FR3␤ (25). The non-canonical CDR2␤ insertion Ser 52a was also mutated to Cys (S52aC), as some naturally occurring hV␤2 alleles carry Cys at this position, and to Gly (S52aG), to remove the interacting side-chain and to introduce more flexibility, yet maintain the length of the CDR2␤ loop. In total, 31 hV␤2.1 mutants were generated to elucidate the molecular basis of selectivity and activation of hV␤2-expressing T cell by both TSST-1 and SpeC.
Wild-type hV␤2.1 was transfected into HuT78 T cells (37) and stimulated with titrations of TSST-1 or SpeC (Fig. 3A). HuT78 T cells naturally express endogenous hV␤23, which is an irrelevant V␤ for the SAgs used in this work, and thus the hV␤2.1-chain paired with the endogenous V␣-chain. HuT78 T cells also express endogenous HLA-DR4 molecules (38), which can cross-present TSST-1 or SpeC to neighboring HuT78 T cells. Using this system, both TSST-1 and SpeC were able to activate these cells in a dose-dependent manner (Fig. 3A).
To map the functional epitopes of TSST-1 and SpeC on the surface of hV␤2.1, each hV␤2.1 mutant was transfected into HuT78 T cells and stimulated with either wild-type TSST-1 or SpeC at 1 g/ml (Fig. 3, B-F). To control for transfection efficiencies and TCR expression, each transfection was also stimulated with anti-V␤2/anti-CD28 (␣-V␤2/␣-CD28) beads, at one bead per T cell. The IL-2 levels from controls lacking SAg, and the TSST-1 and SpeC stimulations, were normalized to the ␣-V␤2/␣-CD28 beads response which was set at 100%. For 5 of the hV␤2.1 mutations (K53A, A54V, K62A, D63A, and K64A), IL-2 produced from these transfections stimulated with the ␣-V␤2/␣-CD28 beads was severely compromised, possibly due to disruption of the ␣-V␤2 binding epitope. The IL-2 responses of mutant K53A was therefore normalized to that of the wild-type hV␤2.1 transfections, and since A54G could be stimulated with ␣-V␤2/␣-CD28 (Fig.  3C), the A54V data were discarded. As SpeC does not engage FR3␤ residues, the IL-2 responses from FR3␤ mutants K62A, D63A, and K64A were normalized to that of SpeC-stimulated HuT78 T cells transfected with wild-type hV␤2.1 (Fig. 3F). Lastly, functional forms of the Tyr 56 deletion, or the Y56A mutation, could not be detected on the surface of HuT78 T cells by flow cytometry and thus these mutants could not be evaluated.
The Molecular Basis of hV␤2.1 Recognition by TSST-1-TSST-1 interacts with fifteen contiguous residues including Glu 51 through Thr 55 in CDR2␤, and continuing into the FR3␤ region from residues Tyr 56 to Lys 64 (Fig. 2D). Within CDR2␤, E51A, G52A, and K53A each showed moderately impaired IL-2 responses to TSST-1 and these phenotypes were consistent with residues identified from the generation of the highaffinity TSST-1 binding hV␤2.1 variants previously described (32). Two well studied high-affinity hV␤2.1 variants named C10 and D10 contain 10 and 14 mutations relative to the yeast stabilized wild-type Ep8, with ϳ6,800-fold (K D ϭ 340 pM) and ϳ13,000-fold (K D ϭ 180 pM) stronger affinities for TSST-1, respectively, relative to Ep8 (K D ϭ 0.6 M) (25,32). In C10, mutagenesis of only three of the affinity-matured residues (Phe 52a , Asn 53 , Val 61 ) to alanine demonstrated faster off-rates than C10 (32) and only 4 affinity-matured residues within D10 (Glu 51 , Phe 52a , Lys 53 , and Val 61 ) when mutated back to the wild-type residues showed significant changes in affinity (25,32). Consistent with our findings, each of these residues (other than Ser 52a discussed below) showed reduced responses in the wild-type hV␤2.1 backbone providing biological significance to the prior biochemical analysis. In the wild-type Ep8-TSST-1 structure there are no direct contacts of Ser 52a with TSST-1 and the Ser52a 3 Ala mutation retained wild-type function (Fig. 3C), but from the affinity maturation experiments with D10, Ser 52a 3 Phe was demonstrated to dramatically increase the affinity of this interaction (25,32). To further evaluate if Ser 52a itself was important, or if the non-canonical conformation of CDR2␤ was necessary for TSST-1 activation, three other mutations at Ser 52a were assessed. S52aC and S52aG each demonstrated neutral profiles upon TSST-1 and ␣-V␤2/␣-CD28 beads stimulation. The most dramatic mutation however, was the deletion of Ser 52a , which abrogated IL-2 secretion by TSST-1 stimulation (Fig.  3D). The fact that TSST-1 could activate S52aG indicates that the presence of a specific side-chain at position 52a is not a requirement for efficient T cell activation by TSST-1.
Within the FR3␤ residues involved in the interaction with TSST-1, E61A had a moderate phenotype, and K62A was a critical mutation (Fig. 3F). The E61V mutation was also identified in the D10 affinity maturation pathway following a likely random PCR error, and the subsequent Val 3 Ala mutation within the C10 variant demonstrated faster off-rate kinetics compared with C10 (32). In the wild-type hV␤2-TSST structure, Lys 62 hydrogen bonds with both TSST-1 residues Glu 132 and His 135 , each of which are individually critical for TSST-1 function (33). Although Lys 62 was not identified through the affinity maturation experiments, mutation to Ala at this position in C10 also demonstrated faster off-rate kinetics (32). Based only on CDR2␤ sequences, TSST-1 would be expected to activate T cells expressing hV␤4; however, hV␤4 alleles contain an Ile at position 62 in FR3, explaining how TSST-1 shows such fine specificity for hV␤2-expressing T cells.
As TSST-1 makes only main chain contacts with hV␤2.1 residues Glu 57 and Val 60 , these two residues were not subjected to mutagenesis, and the contribution of Tyr 56 could not be assessed in this work because neither the Tyr 56 deletion nor T56A mutation was detectable on the surface of T cells (data not shown). Nevertheless, from mutational analysis in the C10 hV␤2.1 variant, C10 Y56A demonstrated an almost 200-fold reduction in affinity. Tyr 56 is located at the center of the cЉ ␤-strand and forms many intramolecular contacts within hV␤2.1. This residue has been proposed to be important for linking the long-range cooperative binding regions for TSST-1 between CDR2␤ and FR3␤ (25) and although we could not assess the function in the current work, this particular residue is also deemed critical for TSST-1 recognition of hV␤2.1.
A few of the hV␤2.1 mutations demonstrated slight, but consistent, increases in IL-2 production and included T30A, I67A, and S98A. None of these residues make direct contacts with TSST-1, or are located in close proximity to the putative ␣-V␤2 epitope, but in each case the corresponding IL-2 responses for SpeC were also marginally increased. As such we do not believe these increased responses reflect direct SAgbinding responses.
In the TSST-1-hV␤2.1 complex, there are six TSST-1 interacting residues in the interface that are known to be individually critical for the activation of hV␤2 ϩ T cells (Gly 16 , Arg 68 , Ser 72 , Glu 132 , His 135 , and Gln 139 ) (25,33,35). These residues form two distinct hot regions on the surface of TSST-1 surrounding CDR2␤ and FR3␤ of hV␤2.1 and are perfectly juxtaposed with the two hot-regions on hV␤2.1 identified here and from prior biochemical studies (Glu 51 , Gly 52 , Lys 53 , Tyr 56 , Glu 61 , and Lys 62 ) (25,32) (Fig. 4A). Thus the molecular basis of the extreme V␤ specificity for TSST-1 is a combination of both the non-canonical CDR2␤ conformation coupled with the hotspot residues on FR3␤ region, in particular the unique residues Tyr 56 and Lys 62 .
The Molecular Basis of SpeC Recognition by hV␤2.1-The crystal structure of SpeC in complex with hV␤2.1 (21) demonstrates that SpeC interacts with three residues (Gln 28 , Ala 29 , and Thr 30 ) from CDR1␤, seven residues (Ser 49 , Glu 51 , Gly 52 , Ser 52a , Lys 53 , Ala 54 , and Thr 55 ) from CDR2␤, four residues (Ile 67 , Asn 68 , His 69 , and Leu 72 ) from HV4␤ and three residues (Ala 96 , Gly 97 , and Ser 98 ) from CDR3␤. Based on IL-2 FIGURE 2. Exposition of the differences between conventional TCR-pMHC and TCR-SAg-pMHC T cell signaling complexes. A, cartoon representation of conventional TCR-pMHC ternary complex (36), where the TCR adopts a semi-conserved diagonal docking mode on pMHC II. By and large, the CDR1 and CDR2 loops engage the ␣-helices of MHC II and the CDR3 loops engage the antigenic peptide. B, cartoon representation of the TSST-1-mediated T cell activation complex (20), where TSST-1 acts a bridge between TCR and pMHC II complex and direct TCR contact with pMHC II is abrogated, thus removing antigenic peptide recognition. C, cartoon representation of the SpeC-mediated T cell activation complex (11,16,21) where SpeC cross-links MHC II molecules through zinc-mediated, high-affinity binding to the pMHC II ␤-chain (left) (16), and as a wedge, by binding to the low-affinity binding site on the pMHC II ␣-chain (11) and preventing direct contacts between TCR V␤ and the antigenic peptide. Colors are as follows: MHC ␣-chain, green; MHC ␤-chain, red; antigenic peptide, black; TCR ␣-chain, orange; TCR ␤-chain, gray; Zn 2ϩ ion, magenta; TSST-1, yellow and SpeC, blue. D, close up view of hV␤2.1 (PDB code 3MFG) showing all the residues involved in binding to TSST-1. E, close up view of hV␤2.1 (PDB code 1KTK) showing all residues involved in binding to SpeC. stimulation profiles for SpeC, residues within the CDR1␤, HV4␤, and CDR3␤ loops play no major individual role for the activation of hV␤2 ϩ T cells (Fig. 3, B and E). The data for CDR1␤ was surprising as this loop is unique in that it contains an additional residue, Phe 27a , not present in the CDR1␤ loops from any of the other V␤ families. Although  For panels (B-F), electroporated HuT78 T cells expressing eGFP and wild-type hV␤2.1 were used as negative and positive controls, respectively, with SpeC and TSST-1 added at a final concentration of 1 g/ml, and the internal control anti-V␤2/anti-CD28 beads added at one bead per electroporated HuT78 T cells. Data shown are the average Ϯ S.E. from at least three independent experiments each conducted in triplicate (*, p Ͻ 0.05 compared with HuT78-V␤2.1, and **, p Ͻ 0.05 compared with HuT78 transfected with wild-type hV␤2.1 but not significantly different compared with HuT78 transfected with the eGFP negative control).
Phe 27a does not interact directly with SpeC, this non-canonical insertion was deleted because this extra residue gives the hV␤2.1 CDR1␤ loop a non-canonical structure that projects toward SpeC (21). Furthermore, the presence of Phe 27a in CDR1␤ has also been predicted to make hV␤2 more reactive to SpeC than hV␤4, which lacks Phe 27a , but has Ser 52a in CDR2␤ loop. However, deletion of Phe 27a had no impact on function (Fig. 3B).
Within CDR2␤, mutating Ser 49 to alanine resulted in a significant phenotype, although this mutant was fully competent for activation by TSST-1 and ␣-V␤2/␣-CD28 beads (Fig. 3C). hV␤2.1 Ser 49 appears to function to anchor SpeC Tyr 15 , a SpeC residue required for hV␤2.1 binding (26), allowing for the non-canonical CDR2␤ to engage the SpeC binding pocket. Despite the critical requirement for hV␤2.1 Ser 49 , this residue is found in roughly one third of all human V␤s, and thus it does not likely contribute much in terms of fine V␤-selectivity. CDR2␤ mutants E51A and G52A had moderate phenotypes (Fig. 3C). hV␤2.1 Glu 51 makes two van der Waals contacts with SpeC Ile 77 , and hV␤2.1 Gly 52 makes two van der Waals interactions with SpeC Leu 78 , and both these SpeC residues are known hot-spots on SpeC, although the G52A mutation may also affect the flexibility of CDR2␤. The remainder of the CDR2␤ loop residues Lys 53 , Ala 54 , and Thr 55 had neutral phenotypes (Fig. 3C).
Previous work identified the O ␥ atom of Ser 52a to be the focal point of several hydrogen bonds with SpeC residues (21,26). Deletion of Ser 52a abrogated IL-2 production from SpeC, yet SpeC was able to activate all of the various Ser 52a point mutations, with only a slight reduction for S52aC (Fig. 3D). It was proposed that the S ␥ atom from the thiol group of Cys 52a , which occurs naturally in some hV␤2 alleles, could replace the O ␥ atom of Ser 52a while maintaining hydrogen bonds with the interacting residues from SpeC (26). Although this may be the case, it was again surprising that S52aG was as efficient as either S52aA or S52aC for activation by SpeC.
Five SpeC residues within the SpeC-hV␤2.1 interface are each individually critical for the activation of hV␤2 ϩ T cells (11,26). These residues (Tyr 15 , Phe 75 , Ile 77 , Leu 78 , and Arg 181 ) form a contiguous concave surface surrounding Ser 52a within CDR2␤ (Fig. 4B). Like TSST-1, the hot regions from either SpeC or hV␤2.1 are perfectly juxtaposed within the SpeC-V␤2.1 interface (Fig. 4B) and taken together demonstrates that the non-canonical conformation of the CDR2␤ loop is the critical determinant for the molecular recognition of hV␤2 by SpeC.
CDR2␤ Contacts Determine V␤ Specificity for Group V SAgs-Existing structural information indicates that despite divergent binding architectures, all bacterial SAgs engage the CDR2␤ of their respective V␤ ligand (Fig. 1) (9, 39). We have shown that CDR2␤ contributes disproportionately in terms of functionality to form complexes with the Group I SAg TSST-1 and the Group IV SAg SpeC and prior biochemical studies have also confirmed that CDR2␤ loop contributes disproportionately in terms of energetics to form complexes with the Group II SAg SEC3 (40). In the absence of co-crystal structures for Group III SAgs in complex with V␤, no such conclusions can be drawn for this SAg subset.
Because the CDR2␤ loop is critical for activation by Group I, II, and IV SAgs, we evaluated the role of CDR2␤ for V␤specificity in Group V SAgs. SEK and SpeI are both homologous Group V SAgs that show divergent V␤-specificity. SpeI expands T cells with TCRs bearing hV␤6.7, hV␤9, and hV␤21 (8), while SEK almost exclusively targets T cells expressing hV␤5.1 (41). Recent crystallographic data for SEK with human hV␤5.1 showed that this subclass of SAg are unique be- cause the intermolecular contacts with V␤ utilize distinct SAg domains to engage CDR2␤ and FR3/4␤ regions (18). The Nterminal ␣-helix of SEK (herein referred to as N␣) makes a number of contacts with the CDR2␤ loop, while the SEK ␣3-␤8 loop makes contacts with both FR3␤ and FR4␤ regions (Fig. 5, A and B). To test which combination of these distinct binding domains where responsible for specific V␤ recognition, we constructed domain-swapped mutants of the N-terminal ␣-helix and the ␣3/␤8 loop for SpeI and SEK with all possible combinations (Fig. 5, C and D). We used qRT-PCR to compare mature V␤ transcripts for hV␤5, hV␤6.7, hV␤9, and hV␤21 expressed from primary human lymphocytes activated with the various SAg hybrids. As predicted, wild-type SEK resulted in increased expression of hV␤5 transcripts relative to SpeI, while wild-type SpeI resulted in increased expression of hV␤6.7, hV␤9, and hV␤21 transcripts relative to SEK (Fig.  6). Remarkably, swapping of the CDR2␤ targeting N-terminal ␣-helix completely reversed this pattern for all tested V␤s such that the recombinant SAgs containing the SpeI N␣ all resembled the V␤-skewing pattern for SpeI, while recombinant SAgs containing the SEK N␣ all resembled the V␤-skewing pattern for SEK (Fig. 6). Although the ␣3/␤8 loop did not affect hybrid reactivity to V␤5 expressing T cells, it was apparent that the SEK ␣3/␤8 loop was more efficient for the activation of human T cells expressing V␤6.7 or 9. In these cases, the inclusion of both the N␣ and ␣3-␤8 loop domains from SpeI grafted onto SEK resulted in fewer transcripts compared with N␣ alone. These differences can be explained by sequence comparisons within the ␣3-␤8 loops and the V␤ FR4 sequences. SEK His 142 and Tyr 158 are known to be individually necessary for T cell activation of hV␤5.1 expressing T cells (18) and the corresponding residues in SpeI are Phe 155 and Gln 168 , respectively (8). SEK His 142 hydrogen bonds with Ser 63 in FR3␤, a residue conserved in all four analyzed V␤s. SEK His 142 also hydrogen bonds with hV␤5.1 Asn 75 , and SEK Tyr 158 hydrogen bonds to hV␤5.1 Thr 78 . These FR4␤ residues are polymorphic between the different V␤s, thus either individually or in combination the SpeI ␣3-␤8 loop residues are less efficient for engaging FR4␤. Thus, although the FR3/4␤ contacts influence the magnitude of the response, this data establishes that CDR2␤ is the critical determinant for V␤ selectivity by Group V SAgs.
CDR2␤ as the Universal SAg Target-The current understanding of SAgs with promiscuous V␤ profiles is that these toxins interact primarily with main-chain atoms of V␤s, and thus V␤ discrimination is more dependent on overall CDR conformation, rather than specific amino acid side chains. SEB and SEC3 are two such promiscuous SAgs and the crystal structure of their common ligand, mouse V␤8.2 (mV␤8.2)chain, has been solved unbound and in complex with two different pMHC II complexes (42), as well as in complex with SEC3 (6) and SEB (19). SEC3 and SEB interact primarily with the CDR2␤ and HV4␤ loops of mV␤8.2 and superposition of the CDR1␤, CDR2␤, and HV4␤ loops from the various mV␤8.2 structures reveals that there are minimal movements of the CDR1␤, CDR2␤, and HV4␤ loops when in complex with SEB, SEC3, and pMHC II, in comparison to the unbound structure. Hence SEC3 and SEB bind to V␤s with similar architectures in a lock and key fashion. However, superposition of the crystal structures of V␤2.1 in complex with TSST-1, SpeC (21) and pMHC II (43) shows the CDR2␤ loop undergoes significant conformational changes upon SAg binding. Assuming that the conformation of CDR2␤ in complex with pMHC is similar to the unbound state, as in other TCR-pMHC complexes (42,44,45), the CDR2␤ loop is pushed away from its native state by both TSST-1 and SpeC. Hence the molecular basis for the V␤ specificity of both TSST-1 and SpeC, similar to the more promiscuous SAgs, is also primarily driven by conformational recognition of CDR2␤, but is also coupled to plasticity of CDR2␤.
Collectively, this information reveals that although there are V␤ residues that are individually essential for the generation of functional SAg-mediated T cell activation complexes, molecular recognition by both TSST-1 and SpeC is driven primarily not by sequence dependence of CDR2␤, but similar to the more promiscuous SAgs, by conformational recognition of CDR2␤. This provides new insight into the molecular basis of V␤ targeting by highly specific bacterial SAgs. In addition, the domain swapping experiments establish that, at least for the Group V SAgs, that the CDR2␤ contacts allow SAgs to discriminate between different V␤s.
Analysis of multiple TCR-pMHC complexes with the same V␤-chain has shown that differences in CDR3␤ sequence did not influence MHC footprints on the TCRs. Rather, germ line encoded regions of TCR and MHC, so called V␤ structural codons, influence rotational orientations implying the V␤ domain plays a critical role in TCRbased recognition (42,46). It seems that bacterial SAgs have evolved to target an intrinsic structural feature of the TCR, imposed by the MHC molecule, critical for the specific recognition and activation of T cells.