Crystal structure of the superantigen enterotoxin C2 from Staphylococcus aureus reveals a zinc-binding site

Crystal structure of the superantigen enterotoxin C2 from Staphylococcusaureus reveals a zinc-binding site Anastassios C Papageorgiou1 , K Ravi Acharyal*, Robert Shapiro 2, Edward F Passalacqua 1 , Rossalyn D Brehm 3 and Howard STranter 3 1

School of Biology and Biochemistry, University of Bath, Claverton Down, Bath BA2 7AY, UK, 2Center for Biochemical and Biophysical Sciences and Medicine and Department of Pathology, Harvard Medical School, Boston, MA 02115, USA and 3 Developmental Production Department, Production Division, Centre for Applied Microbiology and Research, Porton Down, Salisbury SP4 OJG, UK Background: Staphylococcus aureus enterotoxin C2 (SEC2) belongs to a family' of proteins, termed 'superantigens', that form complexes with class II MHC molecules enabling them to activate a substantial number of T cells. Although superantigens seem to act by a common mechanism, they vary in many of their specific interactions and biological properties. Comparison of the structure of SEC2 with those of two other superantigens - staphylococcal enterotoxin B (SEB) and toxic shock syndrome toxin-1 (TSST-1) - may provide insight into their mode of action. Results: The crystal structure of SEC2 has been determined at 2.0 A resolution. The overall topology of the molecule resembles that of SEB and TSST-1, and the regions corresponding to the MHC class II and T-cell receptor binding sites on SEB are quite similar in SEC2.

A unique feature of SEC2 is the presence of a zinc ion located in a solvent-exposed region at the interface between the two domains of the molecule. The zinc ion is coordinated to Asp83, Hisl 18, Hisl22 and Asp9* (from the neighbouring molecule in the crystal lattice). Atomic absorption spectrometry demonstrates that zinc is also bound to SEC2 in solution. Conclusions: SEC2 appears to be capable of binding to MHC class II molecules in much the same manner as SEB. However, structure-function studies have suggested an alternative binding mode that involves a different site on the toxin. The zinc ion of SEC2 lies within this region and thus may be important for complex formation, for example by acting as a bridge between the two molecules. Other possible roles for the metal cation, including a catalytic one, are also considered.

Structure 15 August 1995, 3:769-779 Key words: enterotoxin, Staphylococcus aureus, superantigen, X-ray crystallography, zinc binding

Introduction Staphylococcus aureus and Streptococcus pyogenes produce a group of related pyrogenic toxins [1] including the staphylococcal enterotoxins (SEs) A, B, C1, C2, C3, D, and E [2,3], toxic shock syndrome toxin-1 (TSST-1 [4]), exotoxins A and B [5], and the streptococcal exotoxins A, B and C [6]. These toxins cause a number of illnesses, including toxic shock syndrome and scarlet fever [7]. The SEs, in particular, are known to be involved in causing the emesis and diarrhoea associated with staphylococcal food poisoning [8]. The pyrogenic toxins have been shown to function as 'superantigens' [9]. Like classical antigens, they bind to MHC class II molecules [10-12] and subsequently form ternary complexes with receptors on T cells, thereby stimulating proliferation and increased cytokine production by these cells [7,9,13]. However, superantigens differ from conventional antigens in that they are not subject to proteolytic processing prior to MHC class II binding and they interact primarily with regions outside the antigenbinding groove. Moreover, superantigen-mediated T-cell activation is achieved predominantly through interaction with the 13-variable region (Vp) of the T-cell receptor (TcR) [14,15], whereas ordinary peptide antigens bind to several variable elements of the TcR. As a consequence,

superantigens stimulate a much larger number of T cells than do other antigens. Despite the common mechanism shared by all superantigens, individual members of this family differ in their interactions with MHC class II molecules and TcRs and in some of their physiological effects. X-ray crystallographic studies on SEB [16], TSST-1 [17,18], and their complexes with an MHC class II molecule (HLA-DR1) [19,20] have recently begun to shed light on the structural basis for these functional differences. Most of the available data on bacterial superantigens have been obtained with SEA, SEB and TSST-1. SED and SEE are only produced in small quantities, thus making structural studies difficult. However, SECs have been purified to homogeneity and have been crystallized [21,22]. The three SEC subtypes differ in amino acid sequence at only 10 residues [23], but nonetheless exhibit differences in their interactions with T cells [9]. In order to further understand the molecular basis of the superantigenic behaviour of SEs, we have determined the structure of SEC2 at 2.0 A resolution using X-ray crystallography. The core topology of the molecule is similar to that reported for SEB [16] and TSST-1 [17,18], with which it shares 63% and -24% sequence identity, respectively. However, SEC2 contains a novel

*Corresponding author.

© Current Biology Ltd ISSN 0969-2126

769

770

Structure 1995, Vol 3 No 8 solvent-accessible, hydrophobic 3-barrel (comprising 13-strands [31-[35) together with three a-helices (a3, oa4 and oa6), and has significant similarity to the known oligosaccharide/oligonucleotide-binding fold ('OB-fold') structure [24]. Domain 2, which resembles the '-grasp motif' [25], contains a long helix (a5), which interacts with helix a6 of domain 1 on one side, and is surrounded by a highly twisted -sheet structure (6-312) on the other. This domain also contains the N-terminal segments (two short helices, oal and ac2) of the molecule.

Fig. 1. The polypeptide fold for SEC2. Helices are coloured in red (the 310-helix being drawn more thinly than the a-helices), -strands in green and loops in yellow. The N and C termini are labelled. The large blue sphere represents the zinc ion. The disulphide bridge is shown in ball-and-stick representation. (Figure drawn with MOLSCRIPT [76] and Raster 3D [77].) zinc-binding site which may be involved in MHC class II molecule interactions, thereby mediating the superantigenic properties of the toxin. Results and discussion Molecular structure The two domains of SEC2 are packed together to form a compact molecule of approximate dimensions 45 A x 53 A x 43 A (Fig. 1). Domain 1 contains a

The overall structure of SEC2 (Figs 1,2a) resembles that of SEB and TSST-1 (Fig. 2b). Sequence alignments for the three toxins, along with the details of their secondary structural elements, are shown in Figure 3. The regions of the SEC2 structure that deviate most significantly from SEB involve residues 51-61 (2-33 loop), 97-110 (the disulphide loop and part of 135), 120-129 (the 31 0-turn and part of 36), 137-140 (6 and 37), 152-155 (P138) and 193-194 (10). The smaller protein, TSST-1 (194 amino acids), which is only distantly related to SEB and SEC2 (both with 239 amino acids), lacks the N-terminal helices o1l, or2 and ot4 of SEC2 and is severely truncated in several loop regions. The zinc-binding site When the protein was crystallized [22] no divalent metal ions were added to the crystallization medium. However, a strong persistent feature of density (above 9u), which could not be accounted for by a water molecule, was always present in the calculated difference Fourier maps during the refinement of the SEC2 structure. This feature was assigned as a zinc ion on the basis of several lines of evidence. Inclusion of the zinc atom in the refinement results in an improvement of both Rfree and Rcryst by 1%

Fig. 2. Stereoviews displaying (a) a Ca trace for SEC2 and (b) a comparison of Ca traces for SEC2 (black), SEB (red) and TSST-1 (green). The rms deviation between the Ca positions of 206 equivalent residues in SEC2 and SEB crystal structures is 0.94 A and the corresponding deviation between the Ca positions of 181 equivalent residues in SEC2 and TSST-1 crystal structures is 2.95 A. Structural alignments were performed using the Structure Homology Program (SHP) [64].

Superantigen enterotoxin C2 Papageorgiou et al.

Fig. 3. Structure-based alignment of the amino acid sequences of SEC2, SEB and TSST-1 as determined using the program SHP 164]. Every tenth residue of each sequence is numbered. In SEC2, the solvent-inaccessible residues (<20 A2 of exposed surface as calculated using the program DSSP [78]) are indicated by stars (*) and residues involved in crystal-packing contacts (as given by the program X-PLOR [67]) are indicated by the hash (#) symbol. Secondary-structure elements as determined by DSSP are also shown. The residues in SEB [16,19] and TSST-1 [18,20] involved in MHC class II and TcR binding are highlighted in red and green, respectively. (SEB residues Tyr89 and Gln92 have been implicated in both interactions.) SEC2 residues corresponding to those in SEB which are involved in MHC and TcR interactions are indicated by the same colours. In addition, two SEC2 peptides proposed to bind MHC molecules [46] are highlighted in yellow. The zinc ligands in SEC2 are shown highlighted in blue. (Figure drawn with ALSCRIPT [79].) (Rfree from 27.5% to 26.4% and Rcryst from 22.8% to 21.8%). The putative zinc ion (Fig. 4a) is in close proximity to two imidazole groups (from His18 and His122) and two carboxylate groups (from Asp83 and Asp9* [on a symmetry-related molecule]). Such groups are amongst those that commonly serve as ligands in binding sites for this metal [26]. The spacing between His118 and His122 in the primary structure is also typical of Zn2 + sites, which in general include two ligands that are separated by one to three amino acids. The metal-ligand distances (1.93-2.23 A) and coordination angles (93.6-123.8 °) are within the range seen for tetrahedrally coordinated zinc ions in proteins whose three-dimensional structures have been determined previously (for examples, see [27-29]).

The low temperature factors for the zinc ion and its ligands (9.5-13.9 A2 ) in SEC2 are also consistent with those expected for a tightly bound metal ion. Direct physical evidence that zinc is a constituent of SEC2 was obtained by atomic absorption spectrometry, which revealed the presence of 0.53 mole of zinc bound per mole of protein in the preparation used to grow crystals for X-ray diffraction (see the Materials and methods section). Because the occupancy of the zinc atom observed in the crystal structure was nearly 1.0, it may be that the zinc-SEC2 complex crystallizes more readily than the apoprotein. Indeed, this is reasonable in light of the position of the zinc ion as a bridge between

771

772

Structure 1995, Vol 3 No 8

Fig. 4. The zinc-binding site of SEC2. (a)Amino acid residues that interact with zinc are shown. The zinc ion is represented by a large blue sphere. Two protein molecules contribute atoms to this site. The ligands Asp83, Hisll18 and His122 are from one molecule and Asp9* is from the other. (Figure drawn using O [70].) (b) Stereo diagram showing the superposition of all non-hydrogen atoms for residues His-Glu-X-X-His in SEC2 (black), thermolysin (red) and the catalytic domain of fibroblast collagenase (green), performed using the 'Analyze Structure Homology' (ASH) program [64]. The rms deviations for Cacs are: SEC2 versus thermolysin, 1.05 A; SEC2 versus collagenase,

1.12 A, and thermolysin versus collagenase, 0.20 A. (Figure drawn using MOLSCRIPT [76].) neighbouring molecules in the crystal. Consistent with this view, a preparation of SEC2 which had failed to crystallize was found to contain only a trace quantity of zinc (less than 0.06 mole per mole of protein). A possible explanation for the variable, substoichiometric quantities of zinc measured in the SEC2 preparations is that part of the purification procedure [22] was performed at a pH (pH 5) at which appreciable dissociation of the metal ion may occur. The apoSEC2 preparation was used to further examine the zinc-binding properties of the toxin. Three molar equivalents of zinc nitrate were added to the protein and the amount bound was determined by analyzing the final retentate and filtrate obtained after ultrafiltrationdialysis. The retentate contained 57.5 ptM zinc (representing bound plus free metal) together with 53.2 plM

SEC2; the zinc concentration in the filtrate (representing free metal) was 1.8 p.M. Thus, 1.05 mole of zinc was incorporated per mole of toxin. These data indicate that the dissociation constant for the zinc-SEC2 complex is well below 1 uM and, therefore, that a zinc atom is an intrinsic component of SEC2 in solution under normal physiological conditions. The zinc-binding site observed in the SEC2 crystal structure (Figs 1,4a) is located on the surface of the protein at the interface between two SEC2 molecules. It is formed by residues from three elements of secondary structure: the 3-sheet of domain 1 (Asp83 [on strand 34] and Hisll18 [on strand 35]); the 13-turn that connects strand 35 with the 31 0-helix of domain 2 (Hisl22), and the or1 helix of domain 2 of the symmetry-related molecule (Asp9*). An analogous recruitment of zinc-binding

Superantigen enterotoxin C2 Papageorgiou et al. ligands from neighbouring molecules in a protein crystal was reported for the serine protease tonin [30], although in that case the protein, unlike SEC2, was crystallized from a buffer that contained a high concentration of zinc ions and the metal is not thought to be part of the native structure in solution. The intermolecular contacts of SEC2 in the crystalline state are presumably absent when the protein is in solution. Thus the zinc ion, although tightly bound, may be coordinated only to the three ligands, Asp83, Hisl 18, and His122, that derive from a single molecule. Coordination of a zinc ion by three protein ligands (plus a water molecule) is, in general, indicative of a catalytic role for the metal [31]. SEC2, however, has no known enzymatic activity and a catalytic role for the zinc ion seems unlikely. The zinc-binding site of SEC2 is not conserved in SEA, SEB, SED or SEE. Although Asp83 and Hisll8 are conserved in all SEs, His122 is replaced by arginine (in SEA and SEE), glutamine (in SEB) and lysine (in SED) and these residues are relatively poor ligands for zinc ions. Mutagenesis results on other zinc-containing enzymes (e.g. leukotriene A4 hydrolase [32]) suggest that the residues corresponding to Asp83 and Hisll18 in the other SEs would not, by themselves, be sufficient to coordinate the metal effectively. Thus, a catalytic role for the zinc ion in SEC2 might imply that the mechanism of action of this toxin differs in some fundamental respect from that of the others. This seems improbable in view of the apparent similarity of the superantigenic, pyrogenic and emetic properties of these toxins. Despite this consideration, the possibility that SEC2 is a zinc-dependent enzyme cannot be excluded. Indeed, the presence of the zinc ion in SEC2 may point to a hitherto unsuspected aspect of the protein's function. In this regard, we have investigated the potential signiificance of the His-Glu-Gly-Asn-His sequence that forms part of the metal-binding site in SEC2. His-Glu-X-X-His is a common zinc-bindirig motif in zinc proteases, and several proteins - including leukotriene A4 hydrolase [33], tetanus toxin [34], botulinum neurotoxin [35,36], and the enterotoxin of Bacteroidesfiragilis [37] - were discovered to possess peptidase activity only after they were recognized to contain this sequence. However, the conformation of these residues in SEC2 differs markedly from those observed thus far in crystal structures of these proteases. The zinc-coordinating pentapeptides in thermolysin (Protein Data Bank [PDB] code 7TLN [29]) and collagenase (PDB code 1CGL [38]) are compared with the corresponding residues in SEC2 in Fig. 4b. A potentially significant observation is that the glutamate side chain in SEC2 has a different orientation from that of the glutamate in the proteases, where it is thought to participate in catalysis by removing a proton from the water molecule that provides the fourth zinc ligand. In fact, examination of the SEC2 crystal structure reveals that no residue is situated appropriately to act on the zincbound water molecule that would replace the Asp9* carboxylate when the toxin is in solution. Although these

observations suggest that SEC2 does not have proteolytic or other hydrolytic activity, this question must remain open in the absence of direct functional information. The major non-catalytic role for zinc ions in proteins is 'structural', that is, helping to stabilize a local conformation required for interactions with DNA and other molecules [26]. As all structural zinc-binding sites defined by crystallography or NMR comprise four protein ligands and no bound water molecules, a role of this type for the metal in SEC2 again appears to be somewhat improbable. Zinc ions have also been proposed to be capable of acting as bridges that participate directly in protein-protein interactions, although this has not yet been confirmed by structural determinations. Binding of human growth hormone (hGH) to the human prolactin receptor is strongly dependent on zinc and the results of mutagenesis studies indicate that the zinc-binding site comprises three residues from hGH and one from the receptor [39]. In contrast with SEC2, hGH by itself does not bind zinc to any measurable extent - the additional presence of the receptor is required. Zinc has also been shown to be important for the binding of SEA and SEE to MHC class II molecules and, on the basis of indirect evidence, three of the ligands are again thought to be provided by one of the proteins, namely SEA or SEE, and the fourth by the MHC molecule [40,41]. It is possible that the zinc ion in SEC2 plays an analogous role (although its binding site is distinct from that in SEA or SEE) and contributes to the interaction of the toxin with MHC (see below) or some other molecule that supplies an additional ligand. Finally, we note that Asp83, Hisll18 and His122 of SEC2 are conserved in the related S. pyogenes exotoxin A, as are the residues between the two histidines (Glu-Gly-Asn) and a large proportion of the neighbouring residues [42]. Thus, it is likely that this toxin, which is involved in scarlet fever, also contains zinc and that it shares with SEC2 whatever functional property is dependent on the metal ion. The MHC class IIbinding site Detailed pictures of the interactions of SEB and TSST-1 with HLA-DR1 molecules have recently been obtained by X-ray crystallography [19,20]. These studies reveal both similarities and differences in the binding modes of the two toxins. Thirteen of the contact residues on SEB and TSST-1 (out of 19 and 24, respectively) are structurally equivalent, and 11 of the DR1 residues involved are the same (Fig. 3). In the SEB-DR1 complex, the major interactions link the 3-barrel of domain 1 of the toxin to the a-chain oal helix of DR1, close to the edge of the peptide-binding groove. Although the P-barrel of TSST-1 also makes important contacts with DR1 al, this toxin is positioned differently. Consequently, it covers almost half of the peptide-binding site and interacts with helices from both the axl and 31 domains of DRI and with the bound peptide. Despite the apparent overlap of their binding sites on DR1, SEB and TSST-1 do not

773

774

Structure 1995, Vol 3 No 8 compete with each other for binding to cells that express DR1 molecules, perhaps reflecting the participation of cellular factors such as peptides [43]. The high degree of structural similarity and amino acid sequence identity between SEC2 and SEB suggests that these toxins might bind in analogous fashion to MHC class II molecules. In order to examine this question, we have modelled the SEC2-MHC molecule complex by docking the SEC2 structure onto DR1 in the SEB-DR1 complex [19] in the same orientation as SEB. As the SEB-DR1 structure has not yet been refined at high resolution, we shall confine our discussion to the most salient features. The modelled SEC2-DR1 complex shows a reasonable fit of the two molecules without any obvious stereochemical impediments. The interface consists of residues from the 31-32 loop, 33 strand, ot4 helix, [34-35 loop, 35 strand and a6 helix of SEC2 (Figs 3,5) and the oal domain of DR1. Twelve of the SEC2 residues are identical to those in SEB that make contact with DR1, and these superimpose well with their SEB counterparts, participating in a similar set of molecular interactions at the interface. In particular, the major hydrogen-bonding interactions of SEB with Tyrl3, Gln18, Lys39 and Lys67 of DR1 oil are all maintained in the modelled SEC2-DR1 complex. There are seven SEB residues at the interface that are not conserved in SEC2: Glu43 is replaced by lysine, Tyr46 by alanine, Phe47 by histidine, Arg65 by lysine, Lys69 by leucine, Lys71 by glutamate and Gln92 by asparagine. Of these, Lys43,

Ala46, His47 and Leu69 in SEC2 are involved in similar main-chain hydrogen bonds and/or side-chain van der Waals contacts with DR1 residues as are found in the SEB-DR1 complex. However, the van der Waals contact involving the side chain of residue 65 is absent in SEC2. The side chain of Glu71 does not appear to interact with DR1 and our modelling exercise therefore suggests that it may not be involved in DR1 binding. On the other hand, Asn92 seems to make a better hydrogen bond with Gln57 from DR1 than does Gln92 of SEB. As in the SEB complex,. the DRI peptide is located some distance (10 A) away from the interface. The SECs bind to MHC class II molecules of the HLADQ isotype as well as to HLA-DR, but apparently do not bind HLA-DP molecules to a significant extent [44,45]. DQ and DP differ from DR at numerous positions involved in the SEB-DR and modelled SEC2-DR interfaces (6-9 for DQ and 7 for DP). Jardetzky et al. [19] have noted that many of the side chains that make key contacts with SEB are maintained or conservatively replaced in DQ, and have suggested that the substitution of Gln18 by proline in DQ is an important reason for the weaker binding of SEB to DQ relative to its binding to DR. If SEC2 does indeed bind MHC class II molecules as in this model, it would seem that any conformational disruption caused by the proline replacement has no appreciable influence on the overall interaction with SEC. DP has several substitutions in the segment containing residue 18 that may diminish binding and, moreover, has a replacement of Lys67 by asparagine that might eliminate this major stabilizing interaction [19]. Although an SEB-like mode for binding of SEC2 to MHC class II molecules seems structurally plausible, the limited experimental data available in fact suggest that the region of the SECs involved in the interaction may differ from that utilized in SEB. These results, which were obtained with SEC1 mutants and peptides and interpreted with reference to the unpublished SEC3 crystal structure [46], point to the 'a5 groove' between the two domains as the binding site (Figs 1,5). Peptides representing residues on either side of the lower portion of this groove - that is, 74-86 (a4 helix to 4 strand in domain 1) and 148-162 (end of 37 strand to the middle of a5 helix in domain 2) - stimulated T-cell proliferation. Deletion mutants lacking residues in the a3 helix in the upper portion of the groove retained strong MHCbinding capacity, suggesting that the lower part may be of greater importance.

Fig. 5. Worm representation of the SEC2 molecule illustrating potential sites of interaction with MHC class II and TcR molecules. The orientation of the molecule is similar to that in Figure 1. Regions implicated in MHC class II binding on the basis of homology modelling from the SEB-DRI complex structure [19] are in red and those implicated by peptide data [461 are in yellow. A minor overlap occurs between these two regions in helix a4 (residues 78-79) and is indicated in red. Segments implicated in TcR binding are in green. Residues lying within both the SEB-like MHC-binding site and the TcR-binding site are indicated in purple. The zinc atom is shown in blue.

It is striking that the zinc ion identified in our SEC2 crystal structure (and presumably also present in SEC1 and SEC3) lies within this region of the toxin, well away from the SEB-like site. Thus, if the proposed model for SEC1 is correct, it would appear that the zinc ion is ideally situated to contribute to MHC binding. Moreover, it is tempting to speculate that it does this by acting as a bridge: in other words, that the MHC class II molecule provides a fourth protein ligand to complement the

Superantigen enterotoxin C2 Papageorgiou et al. three from the toxin. Crystallographic studies on complexes between SEC2 and MHC molecules and investigations of the functional involvement of zinc in MHC class II binding will ultimately determine the validity of this hypothesis and reveal whether both of the potential MHC-binding modes suggested by the present structural data are utilized in SEC2. A similar bridging role for the zinc ion in the interactions of SEA and SEE with HLA-DR1 was proposed previously [40] in order to account for the marked zincdependence of DR1 binding, although kinetic evidence [41] had suggested that the zinc-binding site was confined to the toxin. The residues coordinating zinc in SEA and SEE have been designated as His187, His225 and Asp227 near the C-terminal end of the molecule; mutations of these amino acids reduce zinc binding, MHC class II binding, and T-cell stimulating activity (unpublished data of Fraser and Hudson cited in [40]). The analogous three residues in SEC2 (Ile189, Lys229 and Glu231) cannot coordinate zinc and are located on 3strands 19 and 112 of domain 2, quite distant from both of the putative MHC-binding sites as well as the region corresponding to the observed site on TSST-1. These observations, together with the results of additional mutational and cross-competition studies [47,48], raise the possibility that bacterial superantigens have developed four or more distinct MHC-binding modes. The TcR-binding site The V, region of the TcR is the major site of interaction with staphylococcal superantigens. Putative TcR-binding sites in the three-dimensional structures of SEB [16,19] and TSST-1 [18,20] have been assigned on the basis of mutational evidence and topological considerations. In SEB, this site lies between the two domains of the protein in a shallow cavity formed by residues 22-33 (mostly ca2 helix), 55-61 (2-P3 loop), 87-92 (4 strand and ,4-15 loop), 112 (5 strand) and 210-214 (oa5 helix). Replacement of residues 23, 60 and 61 in this region affect T-cell activation without impairing MHC class II binding [49]. In contrast, the TSST-1 residues implicated in TcR binding - Tyr115, Glu132, Hisl35, Ile140, Hisl41 and Tyr144 [50-53] - are located in domain 2, on the long helix a2 (4 in SEB; a5 in SEC2) and the 17-18 and ot2-P9 loops. The SECs have VB specificities similar to that of SEB [9,54], suggesting that the site of interaction of these toxins with TcRs is analogous to that on SEB (Fig. 5). Indeed, deletion mutants lacking various residues between positions 19 and 33 (3 helix, corresponding to a2 in SEB) are mitogenically inactive but can inhibit both T-cell activation and MHC class II binding by the native toxin [46]. Moreover, 15 of the 21 residues lining the cavity that comprises this site are conserved in the SEC2 (and SEC1) sequence (Fig. 3). Comparison of the SEC2 and SEB [19] crystal structures reveals that these residues are positioned similarly in the two toxins. The six amino acid replacements (Glu22-+Gly, Val26-*Tyr,

Asn31l-His, Ala87-Ser, Tyr91-Val and Gln92-+Asn) may then account for some or all of the observed differences in TcR interactions. The substantially greater divergence in V specificity between SEC2 (and SEB) relative to TSST-1 [9,54] is not surprising: TSST-1 has no site that is structurally equivalent to the putative TcR-binding cavity of SEC2 and SEB [18] and the TcR-binding region of TSST-1 is poorly conserved in the SEC2 and SEB structures. SEA and SEE also vary considerably from SEC2 and SEB in their TcR preferences [9,55]. Although mutagenesis results suggest that the same cavity may be involved in TcR recognition [56], 60% of the corresponding residues in both SEA and SEE are different. The disulphide loop All SEs possess a disulphide loop. The amino acid sequence and length of this loop vary amongst the SEs; however, all such residues are conserved among SEC subtypes. In the crystal structure of SEC2, this loop (between residues 93-110) is partially disordered and has diffuse density (this loop is known to be extremely susceptible to proteolysis [57]) indicative of considerable flexibility. Residues on the C-terminal side of this loop have been suggested to be involved in conferring the emetic property of the SEs [58,59], as they are well conserved in these toxins, but not in TSST-1, which lacks this biological activity. Some support for this view is provided by the lack of emetic activity of mutants of both SEC1 [59] and SEA [60] in which alanine replaces one or both of the cysteines. Swaminathan et al. [16] have proposed that one particular residue in this region (Hisl21 in SEB and His118 in SECs) may be important because chemical modification of histidines in SEA [61] also eliminates emetic activity. However, in the SECs this histidine participates in an interaction that is absent in the other SEs: namely, it is a ligand for the zinc ion. This may imply that the histidine in question is not actually involved in the emetic activity of these toxins. Alternatively, the histidine may contribute to the emetic activity of other SEs but not of SEC (i.e. the emetic site may vary in the different SEs) or the zinc may not be bound to SEC under the physiological conditions where the critical interaction responsible for emesis occurs (e.g. if this is in the stomach, the acidic pH would release the zinc ion). It appears that emesis involves a complex set of events and further investigations are required in order to address these questions.

Biological implications The bacterial toxins secreted by Staphylococcus aureus and Streptococcus pyogenes produce a wide range of clinical conditions including food poiscning, toxic shock syndrome and scarlet fever but their precise mode of action remains unclear. These molecules are capable of stimulating a large proportion of T lymphocytes bearing certain VP elements and are therefore called 'superantigensq Like normal peptide antigens, superantigens

775

776

Structure 1995, Vol 3 No 8

require antigen-presenting cells expressing MHC class II molecules to induce T-cell proliferation. The T-cell response is known to involve production of several cytokines including interleukins-2, 4 and 6, tumour necrosis factors-a and 3, and interferon-y - which may result in acute systemic illness and clinical shock. Crystallographic techniques have allowed considerable progress to be made in understanding the superantigenic properties of these toxins. To date, three-dimensional structures for staphylococcal enterotoxin B and toxic shock syndrome toxin-1 and their complexes with MHC class II molecules have been determined. These structures have revealed important differences in the interactions of the toxins with MHC class II molecules. Results of mutagenesis and competition studies suggest that other superantigens may utilize yet additional binding modes to induce their effects. Interactions of these toxins with T-cell receptors have yet to be observed crystallographically, but various models have been proposed to account for available functional data. We have now determined the three-dimensional structure of staphylococcal enterotoxin C2. This structure is similar to that of enterotoxin B, most notably in the regions corresponding to the binding sites for MHC molecules and T-cell receptors. A striking and unexpected feature, found only in the enterotoxin C2 structure to date, is the presence of a zinc ion which is coordinated to residues located in a solvent-exposed groove. The metal ion is bound tightly in solution and is therefore intrinsic to the protein, suggesting that it may participate in the biological action of this toxin. The zinc-binding site contains the characteristic His-Glu-X-X-His sequence found in many zinccontaining proteases, but in enterotoxin C2 this pentapeptide adopts a significantly different conformation. The role of the zinc ion in enterotoxin C2 has yet to be established. Although a catalytic role cannot be ruled out, the position of the zinc ion suggests a possible bridging role in binding to MHC class II molecules.

Materials and methods Protein crystallization Details of the purification and crystallization of SEC2 have been described elsewhere [22,62]. Crystals of purified SEC2 were grown at 16°C using the hanging drop method with 25-28% PEG 8000, 0.15 M ammonium sulphate, 0.02% (w/v) sodium azide, 0.1 M cacodylate buffer (pH 6.5) in the reservoir and 2 mg ml-1 protein in the drop. The crystals belong to space group P4 3 212 with unit cell dimensions a=b=43.05 A, c=290.0 A. The crystals diffract to minimum Bragg spacings of 2.0 A at a synchrotron radiation source. There is one molecule of SEC2 per crystallographic asymmetric unit and approximately 50% of the crystal volume is occupied by solvent.

Diffraction measurements and data processing Data for the native protein crystals at 2.6 A (AD-lab data set) were collected at room temperature from two crystals using a Siemens area detector mounted on a Siemens rotating anode (CuKoa X-rays, 50 kV, 80 mA). Overall, 1674 frames of data were collected (0.250 per frame, crystal to detector distance 29 cm, 20 angle 250, 300-360 s per frame exposure). The data were processed using the XDS package [63]. The two data sets (Table 1) were scaled using the program 3DSCALE [64]. Table 1. X-ray data collection statistics.

Resolution Wavelength (X) N* Nu? Overall completeness I/u(l) Completeness of outer shell Rmerge t (%)

AD-lab set

SRS set I

SRS set II

2.6 A 1.5418 A 36679 9011

3.0 A 1.009 A 39546 5672

2.0 A 0.88 A 119513 19015 (1+11 19302)

86.9% 23.2

92% 13.6

95.3% (1+11 97.6) 13.8 (1+11 12.7)

87.0% 74.9% 96.7% 7.5 (10.4 from 5.6 (9.3 from 4.4 (8.8 from 2.67-2.58 A) 3.1-3.0 A) 2.07-2.0 A) (1+11 5.0)

fNumber of measurements. tNumber of unique reflections. *Rm,,,ge=( I- I1 )/<> where Ij is the observed intensity of reflection j and is the average intensity of multiple observations.

Two additional data sets, one at medium (3.0 A) resolution (SRS data set I) and the second one at higher (2.0 A) resolution (SRS data set II), were collected on station 9.5 of the Synchrotron Radiation Source, Daresbury, UK, using a 30 cm diameter MAR-research image plate with X-ray beam (0.2 mm collimator) of wavelengths 1.009 A and 0.88 A respectively. One crystal per data set was used with an oscillation range of 0.750 per image for the 3.0 A data set and 0.400 per image for the 2.0 A data set. The data were indexed, integrated and corrected for Lorentz and polarization effects using the program DENZO [65]. The details of data processing statistics for the three data sets used in the structure determination are presented in Table 1.

Structure determination The structure of SEC2 was determined by the method of molecular replacement using the program AMoRe [66] with a polyalanine search model based on the 2.7 A resolution SEB-DR1 complex structure [19]. SRS data set I was used during molecular replacement. Data in the range 10-4 A were used in the rotation function searches and in the range 10-4.5 A for the translation function searches. Using a Patterson cut-offradius of 25 A, a list of 100 rotation function peaks was obtained, with the top peak having a correlation coefficient of 14.5 (5.8cr above the mean). The translation function gave two symmetry-related solutions with correlation coefficients of 39.0 and 39.7 respectively (4.0a higher than any other peak). During the data collection there was some ambiguity regarding the space group assignment for SEC2 crystals. In order to resolve this problem, translation function search and rigid-body refinement were performed separately for space groups P4 1212 and P4 3 212. The correlation coefficient for the latter space group was significantly higher, and remained so, after rigid-body minimization using the program AMoRe (37.7 and 55.9 respectively), indicating that P43 21 2 was the correct space group.

Superantigen enterotoxin C2 Papageorgiou et al.

Refinement The output model from AMoRe was subjected to rigid-body refinement (polyalanine model) with X-PLOR [67] using SRS data set I from 10-8 A. The stereochemical parameters of Engh and Huber [68] were used throughout refinement. The crystal-

lographic R-factor, Rct (defined as

IIFo I-IF II/ IFo I),

was 46.1% at this stage. AD-Lab data were then used in the refinement from 8.0-2.6 A. Using the amino acid sequence of SEC2, a model was built with the program FRODO [69] on an Evans and Sutherland PS390 graphics system and at later stages of refinement on a Silicon Graphics Indigo (SGI)2 using O (version 5.10) [70]. Electron-density maps were calculated using either X-PLOR [67] or the CCP4 package of programs [71]. There was no density for the residues 56-60, 96-107, 119-128 and 173-181 in the initial maps. The starting R value was 48.5% and the Rfree [67] was 47.3% with 5% of the data omitted. The refinement proceeded smoothly. Three rounds of positional refinement dropped the Rcryst value to 35.9% and the Rfee to 44.8% with all the temperature factors

set to 20.0 A2. At this stage, sigma-weighted 21 F I-IF c I

maps were calculated using the program SIGMAA [72] and careful examination of the maps allowed corrections to be incorporated into the model. Alternating cycles of manual rebuilding, conventional positional refinement and simulated annealing, using the slow-cool protocol as implemented in X-PLOR [67], allowed residues 56-60, 119-128 and 173-181 to be placed in the electron-density map. The behaviour of Rfree was monitored throughout the refinement procedure. When the cry't was about 26% and the Rfree 35%, a new high-resolution data set (SRS data set II) was collected and the merged data set (SRS data sets I and II) was used as the master set during the remainder of the refinement. Extension of refinement from 2.6-2.0 A was performed in 0.1 A resolution steps. Further refinement, model rebuilding and individual temperature factor (B-factor) refinement (using all data, 18932 reflections, 8.0-2.0 A) gave a final model with R c of 20.5%, a last-recorded Rfree of 25.5% (917 reflections) an rms deviation of bond lengths and bond angles from ideality of 0.009 A and 1.460, respectively. During the final stages of refinement, water molecules were inserted into the model only if there were peaks in the IF I- I F c electron-density maps with heights greater than 3.0a and they were at hydrogen bond forming distances from appropriate atoms. 21 Fo I - I Fc I maps were also used to check the consistency in peaks. Water molecules with B-factors higher than 50.0 A2 were excluded from subsequent refinement steps. The average B-factors for protein atoms, main-chain atoms, water molecules and the zinc ion are 19.9 A2 , 19.5 A2 , 21.1 A2 and 11.7 A2 , respectively. All

Fig. 6. Representative portion of the 21 Fo-I Fc I electron-density map of SEC2 contoured at 1.0a using the refined structure at 2.0 A resolution.

computations were carried out on SGI2 and Vax workstations. A portion of the SIGMAA-weighted 2 Fo I- I Fc I final electron-density map is shown in Figure 6. Validation of the model The final refined model at 2.0 A comprised 238 residues, 70 water molecules and one zinc ion. The N-terminal glutamate and the two C-terminal residues (Lys237 and Asn238) have been modelled as alanines due to poor density in the electron-density map. Residues 99-103 (part of the disulphide loop which is exposed to the solvent) have been modelled using the O database [69] but the side chains have been omitted from the final model due to poor density. The stereochemistry of the model was assessed using X-PLOR [67], O [70] and PROCHECK [73]. Analysis of the Ramachandran (4,) plot showed that 89% of the residues lie in the most favoured regions, and two others (Leu58 and Ala237) are in generously allowed regions. Determination of the zinc content of SEC2 Two preparations of lyophilized SEC2 (0.3-0.5 mg) [62] were each reconstituted in 200 Il of 2 mM HEPES, pH 7.5, and 50 I1 aliquots were analyzed for zinc content by atomic absorption spectrometry (AAS). One preparation was found to contain a significant amount of zinc, whereas the other was virtually devoid of the metal (see the Results and discussion section). The concentration of free zinc (i.e. not bound to SEC2) in the first preparation was determined by performing AAS on an ultrafiltrate obtained by centrifugation of this material at 4000 g in a Centricon 10 microconcentrator (Amicon, Beverly, MA). The ratio of bound versus free zinc was found to be about 9:1. The zinc-binding capacity of the second SEC2 preparation was investigated by adding 25 nmol of zinc nitrate to 8.3 nmol of the protein in 500 il of 20 mM HEPES, pH 7.5. This mixture was then washed by performing two cycles of centrifugation and dilution (each time with 10 volumes of 2 mM HEPES, pH 7.5) in a Centricon 10 device and the zinc concentrations in the final protein solution and filtrate were measured by AAS. Centricon membranes were prepared for use in these experiments by centrifugation with 2 ml of 1 mM HCI followed by 2 x 1 ml of 2 mM HEPES, pH 7.5. The zinc concentration in the last filtrate from this washing procedure was <20 nM in all cases. Zinc concentrations in protein samples were determined with a Perkin Elmer 2280 flame atomic absorption instrument (Norwalk, CT), whereas the substantially lower concentrations in the filtrates were measured by the more sensitive graphite

777

778

Structure 1995, Vol 3 No 8 furnace AAS technique using a Perkin Elmer Model 4100 ZL instrument equipped with an AS-70 autosampler. Absorbance peak heights (flame) or areas (graphite furnace) were measured at 213.9 nm. Zinc concentrations were quantitated by comparing samples (diluted in MilliQ water [flame] or 0.2% nitric acid [graphite furnace]) with standards containing 1.5 jM, 3.1 AM and 6.1 pM (flame) or 38 nM, 76 nM, 153 nM, 229 nM and 306 nM (graphite furnace) zinc reference solution (Fisher, Pittsburgh, PA). Protein concentrations were determined by amino acid analysis using 6-aminoquinolinolyl-N-hydroxysuccinimidyl carbamate precolumn derivatization [74]. The atomic coordinates of SEC2 will be deposited in the Brookhaven Protein Data Bank [75]. Acknowledgements: We thank Don Wiley and his colleagues for providing the atomic coordinates of SEB-HLA-DR1 and TSST-HLA-DR1 complexes, Bert Vallee and Peter Hambleton for support and encouragement. KRA would specially like to acknowledge Anthony Rees for encouragement to initiate the superantigen work at Bath University. We are grateful to the staff at SRS, Daresbury, Demetres Leonidas and Ashley Pike for assistance with data collection, Daniel Strydom for amino acid analysis, James Riordan, David Auld, Vasanta Subramanian and Laurie Irons for helpful discussions. This work was supported by the Medical Research Council and the Nuffield Foundation, UK through grants to KRA. We thank the Science and Engineering Research Council, UK for a CASE award to EFP.

References 1. Alouf, J.E., Knoll, H. &Kohler, W. (1991). The family of mitogenic, shock-inducing and superantigenic toxins from staphylococci and streptococci. In Sourcebook of Bacterial Protein Toxins. (Alouf, J.E. &Freer, J.H., eds), pp. 367-414, Academic Press, London. 2. Bergdoll, M.S., Robbins, R.N., Weiss, K., Borja, C.R., Huang, I.Y. & Chu, F.S. (1973). The staphylococcal enterotoxins: similarities. Contrib. Microbiol. Immunol. 1, 390-396. 3. Bergdoll, M.S. (1983). Enterotoxins. In Staphylococci and Staphylococcal Infections. (Easmon, C.S.F. &Adlam, C., eds), pp. 559-598, Academic Press, London. 4. Bergdoll, M.S. & Schlievert, P.M. (1984). Toxic-shock syndrome toxin. Lancet 2, 691. 5. Schlievert, P.M. (1980). Purification and characterization of staphylococcal pyrogenic exotoxin type B. Biochemistry 19, 6204-6208. 6. Cremer, N. &Watson, D.W. (1960). Host-parasite factors group A streptococcal infections. A comparative study of streptococcal pyrogenic toxins and Gram-negative bacterial endotoxin. J. Exp. Med. 112, 1037-1053. 7. Bohach, G.A., Fast, D.J., Nelson, R.D. &Schlievert, P.M. (1990). Staphylococcal and streptococcal pyrogenic toxins involved in toxic shock syndrome and related illness. Crit. Rev. Microbiol. 17, 251-272. 8. Tranter, H.S. (1990). Food borne staphylococcal illness. Lancet 336, 1044-1046. 9. Marrack, P. &Kappler, J. (1990). The staphylococcal enterotoxins and their relatives. Science 248, 705-711. 10. Fraser, J.D. (1989). High affinity binding of staphylococcal enterotoxins A and B to HLA-DR. Nature 339, 221-223. 11. Mollick, .A., Cook, R.G. & Rich, R.R. (1989). Class II MHC molecules are specific receptors for staphylococcal enterotoxin A. Science 244, 817-820. 12. Scholl, P., Diez, A., Mourad, W., Personnet, J., Geha, R.S. &Chatila, T. (1989). Toxic shock syndrome toxin-1 binds to major histocompatibility complex class II molecules. Proc. Nat. Acad. Sci. USA 86, 4210-4214. 13. Herman, A., Kappler, J.W., Marrack, P. & Pullen, A.M. (1991). Superantigens - mechanism of T-cell stimulation and role in immune responses. Annu. Rev. Immuno. 9, 745-772. 14. White, J., Herman, A., Pullen, A.M., Kubo, R., Kappler, J.W. &Marrack, P. (1989). The V3-specific superantigen staphylococcal enterotoxin B: stimulation of mature T cells and clonal deletion in neonatal mice. Cell 56, 27-35. 15. Kappler, J., et al., & Marrack, P. (1989). V -stimulation of human T cells by staphylococcal toxins. Science 244, 811-813.

16. Swaminathan, S., Furey, W., Pletcher, J. & Sax. M. (1992). Crystal structure of staphylococcal enterotoxin B, a superantigen. Nature 359, 801-806. 17. Prasad, G.S., Earhart, C.A., Murray, D.L., Novick, R.P., Schlievert, P.M. & Ohlendorf, D.H. (1993). Structure of toxic shock syndrome toxin-1. Biochemistry 32, 13761-13766. 18. Acharya, K.R., et al., & Tranter, H.S. (1994). Structural basis of superantigen action inferred from crystal structure of toxic-shock syndrome toxin-1. Nature 367, 94-97. 19. Jardetzky, T.S., et al., & Wiley, D.C. (1994). Three-dimensional structure of a human class II histocompatibility molecule complexed with superantigen. Nature 368, 711-718. 20. Kim, J., Urban, R.G., Strominger, J.L. & Wiley, D.C. (1994). Toxic shock syndrome toxin-1 complexed with a class II major histocompatibility molecule HLA-DR1. Science 266, 1870-1874. 21. Bohach, G.A., Chi, Y. & Stauffacher, C.V. (1992). Crystallization and preliminary X-ray diffraction analysis of staphylococcal enterotoxin type C. Proteins 13, 152-157. 22. Passalacqua, E.F., Brehm, R.D., Acharya, K.R. & Tranter, H.S. (1993). Crystallization and preliminary X-ray analysis of a microbial superantigen staphylococcal enterotoxin C2. J. Mol. Biol. 233, 170-172. 23. Hovde, C.J., Hackett, S.P. & Bohach, G.A. (1990). Nucleotide sequence of the staphylococcal enterotoxin C3 gene: sequence comparison of all three type C staphylococcal enterotoxins. Mol. Gen. Genet. 220, 329-333. 24. Murzin, A.G. (1993). OB (oligonucleotide/oligosaccharide binding)fold: common structural and functional solution for non-homologous sequences. EMBO J. 12, 861-867. 25. Overington, J.P. (1992). Comparison of three-dimensional structures of homologous proteins. Curr. Opin. Struct. Biol. 2, 394-401. 26. Vallee, B.L. & Auld, D.S. (1990). Zinc coordination, function, and structure of zinc enzymes and other proteins. Biochemistry 29, 5647-5659. 27. Holmes, M.A. & Matthews, B.W. (1982). Structure of thermolysin refined at 1.6 A resolution. J. Mol. Biol. 160, 623-639. 28. Rees, D.C., Lewis, M. & Lipscomb, W.N. (1983). Refined crystal structure of carboxypeptidase at 1.54 A resolution. J. Mol. Biol. 168, 367-387. 29. Kannan, K.K., Ramanadham, M. & Jones, T.A. (1984). Structure, refinement and function of carbonic anhydrase isozymes - refinement of carbonic anhydrase I. Ann. N. Y.Acad. Sci. 429, 49-60. 30. Fujinaga, M. & James, M.N.G. (1987). Rat submaxillary gland serine protease, tonin - structure solution and refinement at 1.8 A resolution. J. Mol. Biol. 195, 373-396. 31. Vallee, B.L. & Auld, D.S. (1990). Active site zinc ligands and activated H2 0 of zinc enzymes. Proc. Natl. Acad. Sci. USA 87, 220-224. 32. Medina, J.F., et al., & Samuelsson, B. (1991). Leukotriene A4 hydrolase - determination of the 3 zinc binding ligands by site directed mutagenesis and zinc analysis. Proc. Natl. Acad. Sci. USA 88, 7620-7624. 33. Haeggstrom, J.Z., Wetterholm, A., Shapiro, R., Vallee, B.L. & Samuelsson, B. (1990). Leukotriene A4 hydrolase - a zinc metalloenzyme. Biochem. Biophys. Res. Commun. 172, 965-970. 34. Schiavo, G., Poulain, B., Rossetto, O., Benfenati, F., Tauc, L. & Montecucco, C. (1992). Tetanus toxin is a zinc protein and its inhibition of neurotransmitter release and protease activity depend on zinc. EMBOJ. 11, 3577-3583. 35. Schiavo, G., Rossetto, O., Santucci, A., Dasgupta, B.R. & Montecucco, C. (1992). Botulinum neurotoxins are zinc proteins. J. Biol. Chem. 267, 23479-23483. 36. Montecucco, C. & Schiavo, G. (1993). Tetanus and botulinum neurotoxins - a new group of zinc proteases. Trends Biochem. Sci. 18, 324-327. 37. Moncrief, J.S., et al., & Wilkins, T.D. (1995). The enterotoxin of Bacteroides fragilis is a metalloprotease. Infect. Immun. 63, 175-181. 38. Lovejoy, B., et al., & Jordan S.R. (1994). Structure of the catalytic domain of fibroblast collagenase complexed with an inhibitor. Science 263, 375-377. 39. Cunningham, B.C., Bass, S., Fuh, G. & Wells, J.A. (1990). Zinc mediation of the binding of human growth hormone to the human prolactin receptor. Science 250, 1709-1712. 40. Fraser, .D. & Hudson, K.R. (1993). Superantigens - remnants of a past process? Res. Immunol. 144, 88-93. 41. Fraser, J.D., Urban, R.G., Strominger, J.L. & Robinson, H. (1992). Zinc regulates the function of two superantigens. Proc. Natl. Acad. Sci. USA 89, 5507-5511. 42. Weeks, C.R. & Ferretti, J.J. (1986). Nucleotide sequence of the type A streptococcal exotoxin (erythrogenic toxin) gene from Streptococcus pyogenes bacteriophage T12. Infect. Immun. 52, 144-150.

Superantigen enterotoxin C2 Papageorgiou et al. 43. Thibodeau, J., et al., & Sekaly, R.-P. (1994). Subsets of HLA-DR1 molecules defined by SEB and TSST-1 binding. Science 266, 1874-1878. 44. Herrmann, T., Accolla, R.S. & MacDonald, H.R. (1989). Different staphylococcal enterotoxins bind preferentially to distinct major histocompatibility complex class II isotypes. Eur. . Immunol. 19, 2171-2174. 45. Mollick, J.A., Chintagumpala, M., Cook, R.G. & Rich, R.R. (1991). Staphylococcal exotoxin activation of T-cells - role of exotoxin MHC class II binding affinity and class II isotype. J. Immunol. 146, 463-468. 46. Hoffmann, M.L., et al., & Bohach, G.A. (1994). Predictions of T-cell receptor- and major histocompatibility complex- binding sites on staphylococcal enterotoxin C1. Infect. Immun. 62, 3396-3407. 47. Chintagumpala, M., Mollick, .A. & Rich, R.R. (1991). Staphylococcal toxins bind to different sites on HLA-DR. J. Immunol. 147, 3876-3881. 48. Scholl, P.R., Diez, A. & Geha, R.S. (1989). Staphylococcal enterotoxin B and toxic shock syndrome toxin-1 bind to distinct sites on HLA-DR and HLA-DQ molecules. J. Immunol. 143, 2583-2588. 49. Kappler, J.W., Herman, A., Clements, J. & Marrack, P. (1992). Mutations defining functional regions of the superantigen staphylococcal enterotoxin B. J. Exp. Med. 175, 387-396. 50. Blanco, L., Choi, E.M., Connolly, K., Thompson, M.R. & Bonventre, P.F. (1990). Mutants of staphylococcal toxic shock syndrome toxin 1: mitogenicity and recognition by a neutralizing monoclonal antibody. Infect. Immun. 58, 3020-3028. 51. Bonventre, P.F., Heeg, H., Cullen, C. & Lian, C.-J. (1993). Toxicity of recombinant toxic shock syndrome toxin 1 and mutant toxins produced by Staphylococcus aureus in a rabbit infection model of toxic shock syndrome. Infect. Immun. 61, 793-799. 52. Deresiewicz, R.L., Woo, J.-H., Chan, M., Finberg, R.W. & Kasper, D.L. (1994). Mutations affecting the activity of toxic shock syndrome toxin-1. Biochemistry 33, 12844-12851. 53. Bonventre, P.F., Heeg, H., Edwards, C.K., III & Cullen, C.M. (1995). A mutation of histidine residue 135 of toxic shock syndrome toxin yields an immunogenic protein with minimal toxicity. Infect. Immun. 63, 509-515. 54. Janeway, C.A., et al., & Buxser, S. (1989). T cell responses to MIs and to bacterial proteins that mimic its behaviour. Immunol. Rev. 107, 61-68. 55. Irwin, M.J., Hudson, K.R., Ames, K.T., Fraser, J.D. & Gascoigne, N.R.J. (1993). T-cell receptor -chain binding to enterotoxin superantigens. Immunol. Rev. 131, 60-78. 56. Irwin, M.J., Hudson, K.R., Fraser, J.D. & Gascoigne, N.R.J. (1992). Enterotoxin residues determining T-cell receptor V. binding specificity. Nature 359, 841-843. 57. Spero, L., Griffin, B.Y., Middlebrook, J.L. & Metzger, J.F. (1976). Effect of single and double peptide bond scission by trypsin on the structure and activity of enterotoxin C. J. Biol. Chem. 251, 5580-5588. 58. Huang, I.-Y., Hughes, J.L., Bergdoll, M.J. & Schantz, E.J.(1987). Complete amino acid sequence of staphylococcal eterotoxin A. J. Biol. Chem. 262, 7006-7013. 59. Hovde, C.J., et al., & Bohach, G.A. (1994). Investigation of the role of the disulphide bond in the activity and structure of staphylococcal enterotoxin C1. Mol. Microbiol. 13, 897-909. 60. Harris, T.O., Grossman, D., Kappler, J.W., Marrack, P., Rich, R.R. & Betley, M.J. (1993). Lack of complete correlation between emetic and T-cell stimulatory activities of staphylococcal enterotoxins. Infect. Immun. 61, 3175-3183.

61. Stelma, G.N. & Bergdoll, M.J. (1982). Inactivation of staphylococcal enterotoxin A by chemical modification. Biochem. Biophys. Res. Commun. 105, 121-124. 62. Brehm, R.D., Tranter, H.S., Hambleton, P. & Melling, J. (1990). Large scale purification of SEA, B and C2 by dye-ligand affinity chromatography. Appl. Environ. Microbiol. 56, 1067-1072. 63. Kabsch, W. (1988). Evaluation of single crystal X-ray diffraction data from a position sensitive detector. J. Appl Crystallogr. 21, 916-924. 64. Stuart, D.I., Levine, M., Muirhead, H. & Stammers, D.K. (1979). The crystal structure of cat pyruvate kinase at a resolution of 2.6 A. J. Mol. Biol. 134, 109-142. 65. Otwinowski, Z. (1993). Oscillation data reduction program. In Data Collection and Processing. SERC Daresbury Laboratory, Warrington, UK. 66. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallogr. A 50, 157-163. 67. Branger, A.T. (1992). X-PLOR Version 3.1: System for X-ray Crystallography and NMR. Yale University, New Haven, CT. 68. Engh, R.A. & Huber, R. (1991). Accurate bond and angle parameters for X-ray protein structure refinement. Acta Crystallogr. A 47, 392-400. 69. Jones, T.A. (1985). Interactive computer graphics: FRODO. Methods Enzymol. 115, 157-1 71. 70. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110-119. 71. Collaborative Computational Project, Number 4 (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D 50, 760-763. 72. Read, R.J. (1986). Improved Fourier coefficient for maps using phases from partial structures with errors. Acta Crystallogr. A 42, 140-149. 73. Laskowski, R.J., MacArthur, M.W., Moss, D.S. & Thornton, J.M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J.Appl. Crystallogr. 26, 283-291. 74. Strydom, D.J. & Cohen, S.A. (1994). Comparison of amino acid analyses by phenylisothiocyanate and 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate precolumn derivatization. Anal. Biochem. 222, 19-28. 75. Bernstein, F.C., et al., & Tasumi, M. (1977). The protein data bank: a computer based archival system for macromolecular structures. J. Mol. Biol. 112, 535-542. 76. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of proteins. . Appl. Crystallogr. 24, 946-950. 77. Merritt, E.A. & Murphy, M.E.P. (1994). RASTER 3D version 2.0: a program for photorealistic molecular graphics. Acta Crystallogr. D 50, 869-873. 78. Kabsch, W. & Sander, C. (1983). Dictionary of protein secondary structure: pattern recognition of hydrogen bonded and geometrical features. Biopolymers 22, 2577-2637. 79. Barton, G.J. (1993). ALSCRIPT - a tool to format multiple sequence alignments. Protein Eng. 6, 37-40. Received: 29 Mar 1995; revisions requested: 21 Apr 1995; revisions received: 18 May 1995. Accepted: 5 Jun 1995.

779