Responses to Bacterial Superantigens HLA Class II Polymorphisms Determine

The Journal of Immunology

HLA Class II Polymorphisms Determine Responses to Bacterial Superantigens1

Martin Llewelyn,* Shiranee Sriskandan,* Mark Peakman,† David R. Ambrozak,‡ Daniel C. Douek,‡ William W. Kwok,§ Jonathan Cohen,*¶ and Daniel M. Altmann2*

The excessive immunological response triggered by microbial superantigens has been implicated in the etiology of a wide range of human diseases but has been most clearly defined for the staphylococcal and streptococcal toxic shock syndromes. Because MHC class II presentation of superantigens to T cells is not MHC-restricted, the possibility that HLA polymorphisms could influence superantigenicity, and thus clinical susceptibility to the toxicity of individual superantigens, has received little attention. In this study, we demonstrate that binding of streptococcal and staphylococcal superantigens to HLA class II is influenced by allelic differences in class II. For the superantigen streptococcal pyrogenic exotoxin A, class II binding is dependent on DQ ␣-chain polymorphisms such that HLA-DQA1*01 ␣-chains show greater binding than DQA1*03/05 ␣-chains. The functional implications of differential binding on T cell activation were investigated in various experimental systems using human T cells and murine V␤8.2 transgenic cells as responders. These studies showed quantitative and qualitative differences resulting from differential HLA-DQ binding. We observed changes in T cell proliferation and cytokine production, and in the V␤ specific changes in T cell repertoire that have hitherto been regarded as a defining feature of an individual superantigen. Our observations reveal a mechanism for the different outcomes seen following infection by toxigenic bacteria. The Journal of Immunology, 2004, 172: 1719–1726.

uperantigens (SAgs)3 are potent immunostimulatory pro- of all T cells. On the TCR, SAgs bind at the ␤-chain variable (V␤) teins that bind the MHC class II and TCR molecules on the region. Individual SAgs are limited in the V␤ families they stim- S surface of APCs and T lymphocytes (1, 2). The range of ulate (12) and produce a marked skewing of the V␤ repertoire in human diseases with possible SAg etiology includes Kawasaki’s responding T cells known as the V␤ signature. On the MHC class disease (3), psoriasis (4), atopic eczema (5), rheumatic fever (6), II molecule, SAgs adopt four principal binding modes. Some bind Crohn’s disease (7), and sepsis (8). The best-characterized role for the ␣-chain, some the ␤-chain, and some crosslink class II either SAgs in human disease remains that of the SAg exotoxins of by binding both ␣- and ␤-chains or by virtue of two ␤-chain bind- Staphylococcus aureus and Streptococcus pyogenes, which are being sites. SAgs that adopt the same mode of binding do not nec- lieved to trigger the staphylococcal and streptococcal toxic shock essarily compete, suggesting more subtle differences exist within syndromes (9, 10). However, the observation that apparently iden- these binding modes (13). tical toxigenic strains of S. pyogenes or S. aureus cause clinical Mice mount relatively weak in vitro and in vivo responses to syndromes ranging from superficial carriage through pharyngitis to bacterial SAgs compared with humans, a fact that is believed to toxic shock syndrome, suggests that genetic heterogeneity contrib- arise from sequence differences between human and mouse MHC utes to the clinical phenotype following SAg exposure (11). class II, with the mouse class II binding bacterial SAgs poorly Shared structural features allow SAgs to bind, as unprocessed (14). Furthermore, within human HLA class II, differences be- proteins, to the MHC class II molecule and the TCR at sites away tween isotypes in presentation of individual SAgs are also estab- from conventional Ag binding sites, thereby activating up to 20% lished. For example, the staphylococcal enterotoxin A (SEA) and enterotoxin B (SEB) use HLA-DR more efficiently than HLA-DQ in T cell activation, whereas some streptococcal SAgs such as † *Department of Infectious Diseases, Faculty of Medicine, Imperial College, Depart- streptococcal pyrogenic exotoxin A (SPEA) preferentially use ment of Immunology, Guy’s, Kings and St. Thomas’ School of Medicine, London, United Kingdom; ‡Vaccine Research Center, National Institute of Allergy and Infec- HLA-DQ (15). However, during the early characterization of su- tious Diseases, National Institutes of Health, Bethesda, MD 20892; §Virginia Mason perantigenicity, it became clear that a hallmark of this effect was Research Center, Seattle, WA 98101; and ¶Division of Medicine, Brighton and Sus- sex Medical School, Brighton, United Kingdom the lack of classical MHC restriction. This led historically to a Received for publication June 23, 2003. Accepted for publication November focus on the similarities rather than differences in SAg binding by 10, 2003. different HLA alleles, and with the exception of the observation The costs of publication of this article were defrayed in part by the payment of page that SEA and staphylococcal enterotoxin E bind to DRw53 par- charges. This article must therefore be hereby marked advertisement in accordance ticularly poorly (16), the influence of differences within class II with 18 U.S.C. Section 1734 solely to indicate this fact. isotypes has not been studied in detail. 1 This work was supported by the Medical Research Council (U.K.) through a training fellowship (to M.L.). The worldwide resurgence of streptococcal toxic shock syn- 2 drome since the 1990s has been associated with circulation of Address correspondence and reprint requests to Dr. Daniel M. Altmann, Human ϩ Disease Immunogenetics Group, Department of Infectious Diseases, Hammersmith novel spea strains of S. pyogenes (17). An epidemiological link Hospital, London W12 ONN, U.K. E-mail address: [email protected] between HLA haplotype and susceptibility to the SAg-associated 3 Abbreviations used in this paper: SAg, superantigen; SPEA, streptococcal pyrogenic manifestations of S. pyogenes infection was recently demonstrated exotoxin A; SEA, staphylococcal enterotoxin A; SEB, staphylococcal enterotoxin B; B-LCL, B-lymphoblastoid cell line; MFI, mean fluorescence intensity; BLS; bare by Kotb et al. (18). However, because the patients in the Kotb and lymphocyte syndrome. colleagues study were necessarily infected by different S. pyogenes

Copyright © 2004 by The American Association of Immunologists, Inc. 0022-1767/04/$02.00 1720 HLA CLASS II POLYMORPHISM AND BACTERIAL SAgs strains, each carrying multiple different SAg genes (19), it is im- followed by two rounds of depletion using anti-mouse Ig Dynal beads possible to know from such clinical data whether the HLA asso- (Dynal Biotech, Oslo, Norway) according to manufacturer’s instructions. ciation relates to an individual SAg, and if so, to which SAg. This Depletion of all HLA class II expressing cells was confirmed by failure of purified T cells to proliferate in response to SAg stimulation unless coin- makes it difficult to elucidate molecular mechanisms at the level of cubated with APCs. SAg-HLA class II-TCR interactions. In this work we have set out to study in detail the relationship Flow cytometric binding assays between HLA class II polymorphism and the presentation of bac- A total of 5 ϫ 105 cultured B cells were incubated with biotinylated SPEA terial SAg focusing primarily on SPEA as a protype SAg. Because and washed, then binding visualized using Extravidin-PE (Sigma-Aldrich, SPEA binds HLA-DQ specifically and not HLA-DR or HLA-DP Poole, U.K.) by FACS (FACSCalibur using CellQuest software; BD Bio- (15) we have been able to use HLA homozygous B lymphoblastoid sciences). A total of 20,000 cells falling within a healthy lymphocyte gate cell lines (B-LCLs) to screen a range of HLA-DQ molecules for were analyzed. SPEA binding was measured as mean fluorescence inten- differences in SPEA binding. The data presented in this study dem- sity (MFI) of cells incubated with biotinylated SPEA and Extravidin-PE divided by MFI of cells incubated with unbiotinylated SPEA and Extra- onstrate up to 10-fold higher levels of SPEA binding to cells ex- vidin-PE. Level of HLA-DQ expression was measured using Abs recog- pressing HLA-DQ ␣-chains encoded by the gene HLA-DQA1*01 nizing different conserved regions of the DQ molecule Leu10 (DQ1 and than to cells expressing HLA-DQA1*03 or *05 ␣-chains. In ex- DQ3), L2 (pan DQ␣), and SPV-L3 (pan DQ). Binding of each DQ mAb periments using purified HLA class II, we have extended these was assessed independently for each cell line in each experiment and was measured as the MFI of cells incubated with Ab and FITC-labeled anti- observations to other staphylococcal SAgs. In addition these data mouse Ig second layer divided by MFI of cells incubated with isotype demonstrate that the magnitude of the T cell response to SPEA is control Ab and second layer. DQ expression assessed by each Ab was then determined by HLA-DQ polymorphisms both in terms of prolif- expressed as a percentage of the highest DQ expressing cell line and an eration and cytokine response. Furthermore the V␤ signature of average taken of the three Abs to generate a single value, for each cell line, SPEA is itself determined by the HLA-DQ involved in its presen- of DQ expression compared with the highest expressing cell line in the assay. To correct the amount of SPEA binding by a cell line for its level of tation to T cells. DQ expression, the measured SPEA binding was divided by the percentage Although previous studies have demonstrated that HLA class II of highest DQ expression and multiplied by 1000. isotypes differ in presentation of individual SAgs, these are the first data to demonstrate that HLA class II polymorphisms determine SAg-soluble HLA class II binding assay both the magnitude and the quality of the T cell response to SAgs. ELISA plates (Merck, Poole, U.K.) were coated with L243 or SPV-L3. The data provide a plausible mechanism for observed interindi- Wells were washed (PBS 0.1% Tween 20) and then blocked (PBS 1% vidual differences in disease phenotype following infection by tox- BSA) for 1 h. Reaction tubes containing 30 ␮g/ml purified HLA class II igenic strains of bacteria and the recent identification of HLA class and biotinylated SAg at a range of concentrations were incubated overnight II haplotypes associated with susceptibility to severe S. pyogenes at 4°C. Negative control tubes contained biotinylated SAg alone. Tripli- cates were then set up in the ELISA plate for1hatroom temperature. infections (18). Detection was with avidin-HRP (Jackson ImmunoResearch Laboratories, West Grove, PA) and tetramethylbenzidine (Sigma-Aldrich), stopped using Materials and Methods 1MH2SO4, and read absorbence read at 450 nM. B cell lines Binding of SPEA to DQ␣ peptides A panel of HLA class II homozygous human B-LCLs was used, including three DQ5 cell lines, HOM2, LWAGS (DQA1*0101/DQB1*0501), and Twenty-one amino acid peptides representing the HLA-DQA1*01 ␣-chain BEC11 (DQA1*0101/DQB1*0503); five DQ6 cell lines, TOK 53–73 (KFGGFDPQGALRNMAVAKHNL), a three substitution mutant (DQA1*0103/DQB1*0604), PGF, SCHU (DQA1*0102/DQB1*0602), F/G 61, T/R 64, and I/M 66 (Biosynthesis, Lewisville, TX), and two single HOR, and WT46 (DQA1*0102/DQB1*0604); two DQ7 cell lines, IDF substitution mutants F/G 61 and T/R 64 (Dr. R. Edwards, Imperial College, (DQA1*0501/DQB1*0301) and TISI (DQA1*0505/DQB1*0301); and two London, U.K.) were synthesized. SPEA binding to the peptides was as- DQ8 cell lines WT51 and 600sf (DQA1*0301/DQB1*0302). Bare lym- sessed using peptide, or diluent (DMSO) alone, at concentrations from 1 phocyte syndrome (BLS) cell lines retrovirally transfected to express either mg/ml to 1 ␮g/ml to coat an ELISA plate and capture biotinylated SPEA. DQA1*0301/DQB1*0302, DQA1*0102/DQB1*0602,orDQA1*0102/ DQB1*0604, as previously described, were also used (20). Purified T cell stimulation assays Toxins APCs were incubated with SPEA (1000 ng/ml) or culture medium alone for 40 min at 4°C, washed and fixed using 1% paraformaldehyde. Purified Recombinant SPEA-1 was expressed as previously described (21) and was T cells were added at a range of cell concentrations. At 48 h cells were biotinylated using EZ-link Sulfo-NHS-LC biotinylation kit (Pierce, Rock- pulsed with 1 ␮Ci [3H]thymidine. At 72 h cells were harvested using a ford, IL). SPEA, SEA, SEB, and biotin conjugates of each were purchased Betaplate harvester (Wallac, Milton Keynes, U.K.) from Toxin Technology (Sarasota, FL). HLA class II Abs Mouse splenocyte stimulation assays L243 (DR␣), L2 (pan DQ␣), and SPVL-3 (pan DQ) were purified from APCs were incubated with mitomycin C (100 ␮g/ml) (Sigma-Aldrich) for mouse hybridoma supernatants. TDR31.1 (Ancell, Bayport, MN), WR18 40 min at 37°C and washed. Splenocytes from DO11.10 TCR transgenic (pan class II; Serotec, Oxford, U.K.), and Leu10 (DQ; BD Biosciences, mice were incubated with mitomycin-treated APC lines (5 ϫ 104 cells/ Oxford, U.K.) were purchased from manufacturers. well) and SPEA (0–1000 ng/ml). Culture supernatants were harvested and [3H]thymidine incorporation assayed. Purification of HLA class II HLA-DR11 and HLA-DQ3.1 were affinity purified from IDF cells and TCR repertoire usage assays HLA-DR1501 and HLA-DQ6.2 from PGF cells using L243 and SPV-L3 PBMCs from HLA typed donors were stimulated using SPEA or PHA, columns as previously described (22). Purified class II was Ͼ98% pure as with recombinant human IL-2 (Sigma-Aldrich) 20 IU/ml added at 72 h. assessed by SDS-PAGE. The concentration of purified HLA class II was After 7 days, cells were harvested, stained with anti-CD4-FITC and anti- determined by bicinchoninic acid protein assay (Pierce). V␤-PE, and analyzed by FACS (FACSCalibur using CellQuest software; Preparation of human PBMC and T cells BD Biosciences). For each V␤, percentages of CD4-positive lymphocytes falling within a resting lymphocyte gate for unstimulated cells or a blast- PBMCs were obtained by Ficoll gradient separation. The class II negative ing-lymphocyte gate for stimulated cells were recorded, and pairwise com- T cells were purified by negative selection. A total of 5 ϫ 106 cells were parisons made between unstimulated PHA and SPEA stimulated cells by t incubated with L243, WR18, Leu14 (anti-CD14), and Leu19 (anti-CD19), test. Values for p Ͻ 0.05 were considered significant. The Journal of Immunology 1721

Results (Fig. 1a). When expressed as SPEA bound corrected for level of Binding of SPEA by B cell lines DQ expression (Fig. 1b), a statistically highly significant difference between SPEA binding to DQA1*01 and DQA1*03/05 expressing Binding of SPEA to HLA-DQ expressing cells is almost entirely cell lines was found. Because variations in cell cycle and culture mediated through the ␣-chain of HLA-DQ and not other class II conditions can lead to fluctuations in the level of cell surface HLA determinants such as HLA-DR (15). To confirm this, we demonstrated that the Ab L2, which binds the DQ ␣-chain, almost en- expression, we analyzed the relationship between HLA-DQ ex- tirely blocked SPEA binding to B-LCLs whereas the DR␣-specific pression and SPEA binding in several independent cultures of two Ab L243 produced no detectable blocking (data not shown). If all DQA1*01 lines (PGF and WT46) and two DQA1*05 lines (IDF HLA-DQ molecules were equal in their binding of SPEA, SPEA and TISI). Irrespective of minor fluctuations in HLA expression, binding to B cell lines would be directly related to level of the hierarchy of SPEA binding was maintained (Fig. 1, c and d), HLA-DQ expression. However, in six experiments correlating and again there was a statistically highly significant difference be- SPEA binding with level of DQ expression on a panel of cell lines tween the SPEA binding corrected for level of DQ expression of expressing common HLA-DQ molecules, we found that HLA-DQ DQA1*01 expressing cell lines in comparison with the DQA1*05 genotype markedly influenced SPEA binding. A representative ex- expressing cell lines. periment is shown in Fig. 1, a and b. SPEA binding to cell lines To confirm these observations we assessed SPEA binding to expressing HLA-DQA1*01 ␣-chains was greater than binding to BLS cell lines transfected to express HLA-DQA1*0102/HLA- cell lines expressing DQA1*03 or DQA1*05, and SPEA binding to DQB1*0602, HLA-DQA1*0102/HLA-DQB1*0604,orHLA- HLA-DQA1*01 expressing cell lines related closely to level of DQ DQA1*0301/HLA-DQB1*0302. Although the level of HLA-DQ expression, whereas cell lines expressing DQA1*03 or DQA1*05 expression on these cells was much lower than on the B-LCLs, the bound SPEA poorly, despite comparable levels of DQ expression same heightened level of SPEA binding, corrected for level of DQ

FIGURE 1. Binding of biotinylated SPEA to B cell lines expressing HLA-DQ. a and b, SPEA binding to a panel of B-LCLs expressing different HLA-DQ molecules. HLA-DQA1*01/DQB1*06 cell lines (F), HLA-DQA1*01/DQB1*05 cell lines (), HLA-DQA1*03/DQB1*03, or HLA-DQA1* 05/DQB1*03 cell lines (E), BLS cell line (Ⅺ). Cell lines used were BLS, 600sf, WT51, IDF, TISI, BEC11, HOM2, LWAGS, HOR, TOK, WT46, PGF, and SCHU, each numbered 1–13, respectively. c, Relationship between level of HLA-DQ expression and SPEA binding for individual cell lines: PGF (F), WT46 (f), IDF (E), TISI (Ⅺ). b and d, SPEA binding has been corrected for level of DQ expression, cell lines expressing HLA-DQA1*01 ␣-chains (), cell lines expressing HLA-DQA1*03/05 ␣-chains (ƒ), p Ͻ 0.05 and p Ͻ 0.01, respectively, by t test. e and f, SPEA binding corrected for level of DQ expression on BLS cell lines transfected to express different HLA-DQ molecules, cell lines expressing HLA-DQA1*01 ␣-chains (), cell lines expressing HLA-DQA1*03 ␣-chains (ƒ), p Ͻ 0.05 and p Ͻ 0.01, respectively, by t test. 1722 HLA CLASS II POLYMORPHISM AND BACTERIAL SAgs expression, was observed in cell lines expressing HLA-DQA1*01 DQA1*01 ␣-chain) with HLA-DQ3.1 (HLA-DQA1*05 ␣-chain). (Fig. 1, e and f). In keeping with the results from whole cell binding experiments, binding of SPEA by purified HLA-DQ6.2 was markedly superior Binding of SAgs to purified HLA class II by ELISA to binding by purified HLA-DQ3.1 (Fig. 2c). SEA-HLA-DR bind- To establish whether an effect of HLA class II polymorphism on ing is dependent on the presence of zinc. Performing the binding SAg binding could be observed in a cell-free system and extended assay in the presence of 1 mM EDTA, as would be predicted, more broadly to other bacterial SAgs, we designed an ELISA to completely blocked binding of this SAg to purified HLA-DR15 compare SAg–HLA class II binding using purified HLA class II. (Fig. 2d). SPEA binding to the HLA-DQ ␣-chain has been pre- The binding of SEA and SEB to HLA-DR4, HLA-DR11, and dicted by analogy with SEB to be independent of the presence of HLA-DR15 were assessed. SEA binding to HLA-DR is predom- zinc, although a zinc-binding pocket has been identified in the inantly through the polymorphic DR ␤-chain. Binding to HLA- crystal structure of SPEA (23, 24). SPEA-DQ binding was how- DR4 and -DR15 was markedly greater than to HLA-DR11 (Fig. ever not altered in the presence of 1 mM EDTA (Fig. 2e) con- 2a), indicating that just as HLA-DQ polymorphisms influence firming that like the SEB-DR interaction, SPEA-DQ binding is SPEA binding to HLA-DQ, HLA-DR polymorphisms influence zinc independent. SEA binding to HLA-DR. In contrast, SEB binding to HLA-DR is Comparison of the amino acid sequences of the DQ ␣-chains through the nonpolymorphic DR ␣-chain. We found that accord- coded by DQA1*01, DQA1*03, and DQA1*05 revealed that at ingly, differences in binding of SEB by different HLA-DR mole- positions of predicted SPEA binding, the DQA1*03 and DQA1*05 cules were small (Fig. 2b). Using HLA-DQ purified from the same ␣-chains share three amino acid substitutions that are distinguish- cell lines, we compared SPEA binding to HLA-DQ6.2 (HLA- able from the DQA1*01 ␣-chain (Table I) (25). To confirm the

FIGURE 2. Binding of bacterial SAgs to purified HLA class II by ELISA. SEA (a) and SEB (b) binding by HLA-DR; HLA-DR4 (F), HLA-DR15 (), HLA-DR11 (E), no DR control (f). c, SPEA binding by HLA-DQ; HLA-DQ6.2 (F), HLA-DQ3.1 (E), no DQ control (f). d,Influence of zinc chelation on binding of SEA to HLA-DR15 (F) and no DR control (f). e,Influence of zinc chelation on binding of SPEA to HLA-DQ6.2 (F) and no DQ control (f). Points joined by a dashed line show zinc chelation using 1 mM EDTA. Points show mean OD Ϯ 1 SD. The Journal of Immunology 1723

Table I. Sequence alignment of the bacterial SAg binding domains of HLA-DQ ␣-chainsa

Amino acid residues at sites of SPEA contact with the DQ ␣-chain

DQA1* 16 20 21 22 23 39 40 41 42 58 60 61 63 64 66 67 70

0101/2 YGP SGLERKDQGLRMAK 0103 K 0301/2 FTI 0501/5 GFTI

a Numbering corresponds to the DQA1*0101 product (25). importance of these substitutions in altering the SPEA binding there was an absolute difference in the TNF-␣ response, with properties of this region of the DQ ␣-chain, two 21-amino acid DQA1*05 APCs eliciting no TNF-␣ release. peptides were compared. The first peptide represented the In the second approach the aim was to use a humanized system DQA1*01 ␣-chain 53–73, and the second, a three amino acid sub- and explore the contribution made by density of Ag presentation stitution mutant in which the three substitutions distinguishing the on differences attributable to HLA-DQ polymorphisms. Purified DQA1*03/*05 ␣-chains at sites of proposed SPEA binding had human T cells, selected for HLA class II negativity from healthy been made. Using an ELISA of SPEA binding to immobilized donor PBMCs, were used as responder T cells, and using SPEA peptide, a log2 higher concentration of the mutant peptide than of pulsed B-LCLs to present SPEA, the level of stimulation was ad- the DQA1*01 ␣-chain peptide was required to produce comparable justed by altering the APC to T cell ratio. Comparing low and high SPEA binding (Fig. 3). To determine the impact of these substi- binding HLA-DQ types, at ratios of 2:1 and 10:1, DQA1*01 express- tutions in isolation, single substitution mutant peptides were used. ing APCs supported significantly higher levels of T cell activation In each case, binding of SPEA was markedly reduced (data not than DQA1*05 expressing cells ( p Ͻ 0.05 by t test) (Fig. 5). shown). In the third approach, to confirm that the effect of HLA-DQ polymorphism remains apparent, notwithstanding other host fac- Influence of HLA-DQ polymorphism on presentation of SPEA tors involved in the response to SAgs, PBMCs purified from HLA The ability of high and low SPEA-binding HLA-DQ molecules to typed donors were stimulated with SPEA. Proliferation response support SPEA activation of T cells was assessed using three ap- was assayed. Comparing three HLA-DQA1*01 homozygous do- proaches. The aim of the first approach was to control maximally nors with three HLA-DQA1*03/05 homozygous donors, in which for variability attributable to donor T cell differences and the con- no differences were observed between the groups in response to founding effect of T cell HLA class II expression. This was medium alone or PHA, the level of both proliferation in response achieved by using HLA homozygous B-LCLs to present SPEA to to SPEA was greater for HLA-DQA1*01 donors than HLA- murine TCR V␤8.2 transgenic responder T cells, mV␤8.2, with DQA1*03/05 donors at concentrations of SPEA ranging from 0.1 this being one of the murine V␤ families targeted in the response to 100 ng/ml (Fig. 6). Corresponding differences in IFN-␥ produc- to SPEA. APCs expressing DQA1*01 were found to support tion were also detected (data not shown). higher levels of response to SPEA than were APCs expressing ␤ DQA1*05, assessed by proliferation or by TNF-␣, IFN-␥, or IL-4 Influence of HLA-DQ polymorphism on V repertoire of SPEA- (Fig. 4, aÐe). Furthermore, the lowest concentration of SPEA as- stimulated T cells sociated with detectable T cell activation was two orders of mag- SPEA has been previously reported to produce changes in the pro- nitude lower in the presence of DQA1*01 expressing APCs than of portions of V␤12- and V␤14-positive lymphocytes (27, 28). To DQA1*05 APCs, despite comparable levels of HLA-DQ expres- assess impact of HLA-DQ polymorphism on the V␤ repertoire of sion. Elevated TNF-␣ levels are of central importance in the le- the T cell response to SPEA, we stimulated PBMCs from donors thality of SAg-mediated shock (26). It is noteworthy therefore that, homozygous for either HLA-DQA1*01 or for HLA-DQA1*03/05 within the likely in vivo SPEA concentration range (Ͻ100 ng/ml), using either PHA or SPEA. The proportion of CD4-positive lymphocytes binding Abs to 21 different V␤ types was assessed before and after stimulation. In three experiments, expansion of V␤12 and V␤14 was noted following SPEA stimulation, relative to V␤ percentages of unstimulated or PHA-stimulated PBMCs irrespective of donor HLA type. However in HLA-DQA1*01 homozygous donors, an additional 2-fold expansion of V␤13.1 CD4 cells was found ( p ϭ 0.026 by t test). Changes for these and four representative other V␤s are shown in Fig. 7. PHA stimulation produced no changes in the V␤ repertoire compared with unstimulated cells (data not shown) and no differences between donor groups were observed following SPEA stimulation for the other sixteen V␤s assessed.

Discussion SAg binding to the MHC class II molecule is a prerequisite for the excessive T cell activation that underlies SAg toxicity (1, 29). FIGURE 3. Binding of SPEA to DQA1*01 (53–73) peptide (F) and Differential SAg presentation by HLA-DQ and HLA-DR is well DQA1*01 (53–73)3S mutant (E) and control (f). Points show mean OD Ϯ established (30). But because these class II molecules are coex- 1 SD. pressed on APC, such data cast no light on the question of whether 1724 HLA CLASS II POLYMORPHISM AND BACTERIAL SAgs

FIGURE 4. Inﬂuence of HLA-DQ polymorphism on murine TCR V␤8.2 transgenic T cell proliferation and cytokine response to SPEA. aÐd, Pre- sentation by B-LCLs PGF (DQA1*0102/ DQB1*0602), IDF (DQA1*0501/DQB1*0301), and BLS (no DQ control). eÐh, Presentation by HLA-DQ transfected BLS lines DQA1*0102/ DQB1*0602, DQA1*0301/DQB1*0302,noDQ control. a and e, Proliferation in response to presentation of SPEA by DQ6.2 (F), DQ3.1 (E), and no DQ control (f). bÐd and fÐh, Cytokine response to presentation by DQ6.2 (f), DQ3.1 (u), and no DQ control (s). Mean Ϯ 1 SD is shown in each case.

DQ and DR polymorphisms influence susceptibility to a particular Our observation that SPEA binding by cell lines expressing SAg. Although the genes of the HLA region are the most poly- DQA1*01 ␣-chains is a factor of the level of DQ expression morphic in the human genome, the influence that HLA class II whereas SPEA binding to cell lines expressing DQA1*03 or polymorphisms might exert over SAg activation of T cells has not DQA1*05 ␣-chains is virtually undetectable under the conditions been investigated in detail. The streptococcal SAg SPEA has been used, indicates that DQ ␣-chain polymorphisms influence the bind- implicated epidemiologically with cases of streptococcal toxic ing of this SAg to HLA-DQ. The sites at which SPEA interacts shock in North America and Europe, and one reason for the with the HLA-DQ ␣-chain have been predicted from the DR1-SEB marked increase in incidence of streptococcal toxic shock since the crystal structure and the structure of the DQ ␣-chain binding strep- 1980s may have been the emergence of new variants of SPEA tococcal SAg (25). Comparison of the amino acid sequences of the (31). In addition to this clinical relevance, SPEA is a useful pro- DQ ␣-chains encoded by the polymorphic DQA1 gene demon- totypic SAg to study using the techniques used in this study, be- strates that those ␣-chains coded by DQA1*01 and its variants are cause it is exclusively presented by HLA-DQ, whereas many of the virtually nonpolymorphic at sites of SPEA binding, whereas the staphylococcal SAgs are more promiscuous in their binding to ␣-chains encoded by DQA1*03 or DQA1*05 share three amino HLA class II isotypes. acid substitutions at sites of SPEA binding: F/G 61, T/R 64 and The Journal of Immunology 1725

FIGURE 5. Influence of HLA-DQ on magnitude of proliferation in response to SPEA. Proliferation of purified T cells to SPEA presented by ␤ B-LCLs. PGF (DQ6.2), IDF (DQ3.1), BLS (DQ negative). Two levels of FIGURE 7. Influence of donor HLA type on V repertoire of T cells p Ͻ 0.01 by t test. proliferating in response to 100 ng/ml SPEA. HLA-DQA1*01 homozygous ,ء .stimulation are shown (s), HLA-DQA1*03/05 homozygous (f), prestimulation (hatched bars), and following stimulation (filled bars) are shown. A total n ϭ 3 were in Ϯ ␤ ␤ ␣␤ each group, and means 1 SD are shown. Expansion of V 12 and V 14 I/M 66 (Table I). Constraints on HLA-DQ heterodimer forma- occurs following SPEA stimulation irrespective of donor HLA-DQA1 ge- tion mean that the great majority of HLA-DQ5 and HLA-DQ6 notype whereas only T cells from HLA-DQA1*01 donors show expansion molecules comprise a DQA1*01 ␣-chain paired with a DQB1*05 of V␤13.1. Values for p Ͻ 0.05 are by t test. or DQB1*06 ␣-chain (32), as do all the DQ5 and DQ6 cell lines used in this study. Conversely the DQ3 expressing cell lines used in this study all comprise an ␣-chain encoded by DQA1*03 or the presence or absence of a TNF-␣ response, particularly because DQA1*05 paired with a DQB1*03 ␤-chain. Our observation of high TNF-␣ levels are thought to underlie the lethality of toxic poor SPEA binding by HLA-DQA1*03/*05 cell lines compared shock (26). with HLA-DQA1*01 cell lines, irrespective of level of DQ expres- Although V␤-specific T cell expansion is considered to be a sion, is likely therefore to be due to DQ ␣-chain polymorphism at hallmark of superantigenicity, definition of the V␤ signature of sites of SPEA binding. individual SAgs is complicated by the fact that SAg activation may Using purified HLA class II and peptide fragments of the DQ be variably associated with T cell proliferation or deletion and ␣-chain in ELISAs of SAg–class II binding, we have confirmed apparent loss of SAg responsive V␤ types. Previous studies have that HLA-DQ polymorphisms influence SPEA binding in isolation reported principally V␤12 and V␤14 changes associated with from other cell surface factors and antigenic peptide. Furthermore SPEA stimulation (27, 28). It is noteworthy that V␤13 is the clos- we have demonstrated that HLA class II polymorphisms influence est relative of V␤12 in humans (33). Our observation of an addi- the binding of at least one other important bacterial SAg, namely tional expansion of V␤13.1, seen only in HLA-DQA1*01 homozy- SEA. It is likely, therefore, that HLA class II polymorphisms in- gous individuals, is in keeping with previous observations that the fluence the properties of many, if not all, SAgs. nature of the class II SAg interaction determines V␤ specificity. In The differences observed in magnitude of proliferation and cy- particular, an effect of mouse MHC class II isotype on the response of tokine response to SPEA presented by different HLA-DQ alleles V␤-specific T cell hybridomas has been reported (34) and SAgs with were to some extent quantitative, following the trend observed for mutations at class II binding sites show both altered T cell mitoge- differences in binding. However, it is noteworthy that qualitative nicity and V␤ specificity (35). It is not surprising therefore that V␤- differences were also found. At lower concentrations of SAg such specific changes following SAg stimulation are determined, at least in as might be encountered pathologically during sepsis, we observed part, by intra-isotype HLA class II polymorphisms. an absolute difference between strong and weak binding alleles in Our findings demonstrate that allelic differences can account for both quantitative and qualitative differences in the SAg response. They therefore offer a molecular explanation for why individuals vary in their susceptibility to toxic shock syndrome during the course of infection by toxigenic organisms. They reinforce and greatly extend the observation by Kotb et al. (18) of an association between HLA haplotype and susceptibility to SAg-mediated manifestations of S. pyogenes infection. Because Kotb and coworkers were using epidemiological data and clinical isolates, they were not able to focus on specific SAgs and elucidate the mechanisms at work. The HLA class II haplotypes they identified contrast with those identified in this study. Kotb found HLA DQA1*0102/ DQB1*0602 for example to be protective against the severest manifestations of infection. This is likely to be because the patients in Kotb’s study will necessarily have been infected with different S. pyogenes strains each carrying multiple different SAg genes (19). FIGURE 6. Proliferation of PBMCs from HLA typed donors. HLA- The association identified may therefore relate to other SAgs such DQA1*01 homozygous donors (F) HLA-DQA1*03/05 donors (E), n ϭ 3 as SMEZ. Kotb et al. (19) demonstrated that HLA class II haplo- in each group. Means Ϯ 1 SD are shown. Background counts were 150– types associated with severe disease were also associated with 300 cpm and not statistically different between groups. greater proliferation of PBMCs and purified T cells stimulated in 1726 HLA CLASS II POLYMORPHISM AND BACTERIAL SAgs the presence of HLA homozygous B-LCLs. However the stimuli 14. Sriskandan, S., M. Unnikrishnan, T. Krausz, H. Dewchand, S. Van Noorden, used in these experiments were partially purified culture superna- J. Cohen, and D. M. Altmann. 2001. Enhanced susceptibility to superantigen- associated streptococcal sepsis in human leukocyte antigen-DQ transgenic mice. tants containing an undetermined range of SAgs. Again therefore, J. Infect. Dis. 184:166. no specific relationship between the SAg and the class II involved 15. Imanishi, K., H. Igarashi, and T. Uchiyama. 1992. Relative abilities of distinct isotypes of human major histocompatibility complex class II molecules to bind in its presentation could be studied. Kotb et al. (19) speculate that streptococcal pyrogenic exotoxin types A and B. Infect. Immun. 60:5025. the association between class II haplotype and susceptibility may 16. Herman, A., G. Croteau, R. Sekaly, J. Kappler, and P. Marrack. 1990. HLA-DR be through modification of the inflammatory response to the SAg. alleles differ in their ability to present staphylococcal enterotoxins to T cells. J. Exp. Med. 172:706. By focusing on individual SAgs and performing detailed binding 17. Musser, J. M., V. Kapur, S. Kanjilal, U. Shah, D. M. Musher, N. L. Barg, analyses, the work presented in this study demonstrates that dif- K. H. Johnston, P. M. Schlievert, J. Henrichsen, D. Gerlach, et al. 1993. Geo- ferences in the response to a SAg relate rather to differences in graphic and temporal distribution and molecular characterization of two highly pathogenic clones of Streptococcus pyogenes expressing allelic variants of py- class II binding determined by HLA class II polymorphisms. rogenic exotoxin A (Scarlet fever toxin). J. Infect. Dis. 167:337. In the autoimmunity field, disease associations with HLA-DQ 18. Kotb, M., A. Norrby-Teglund, A. McGeer, and D. Low. 2002. An immunogenetic and molecular basis for differences in outcomes of invasive group A streptococcal have been both striking and perplexing in light of the relatively infections. Nat. Med. 8:1398. poor expression of HLA-DQ by APCs and the fact that HLA-DQ- 19. Ferretti, J. J., W. M. McShan, D. Ajdic, D. J. Savic, G. Savic, K. Lyon, restricted T cells are not commonly isolated in most systems (36). C. Primeaux, S. Sezate, A. N. Suvorov, S. Kenton, et al. 2001. Complete genome sequence of an M1 strain of Streptococcus pyogenes. Proc. Natl. Acad. Sci. USA This pattern of HLA-DQ association with responses to a patho- 98:4658. genic bacterial SAg argues that the list of HLA-DQ associations 20. Ettinger, R. A., A. W. Liu, G. T. Nepom, and W. W. Kwok. 2000. ␤57-Asp plays can now be extended from autoimmunity to bacterial immunity an essential role in the unique SDS stability of HLA-DQA1*0102/DQB1*0602 ␣␤ protein dimer, the class II MHC allele associated with protection from insulin- and septic shock. HLA genotype is likely to be a useful predictor dependent diabetes mellitus. J. Immunol. 165:3232. of outcome during clinical toxic shock and the molecular interac- 21. Sriskandan, S., D. Moyes, L. K. Buttery, T. Krausz, T. J. Evans, J. Polak, and J. Cohen. 1996. Streptococcal pyrogenic exotoxin A release, distribution, and role tions analyzed in this study offer clear potential for development of in a murine model of fasciitis and multiorgan failure due to Streptococcus pyo- immunotherapeutic interventions. genes. J. Infect. Dis. 173:1399. It is intriguing that S. aureus and S. pyogenes, being distinctly 22. Peakman, M., E. J. Stevens, T. Lohmann, P. Narendran, J. Dromey, A. Alexander, A. J. Tomlinson, M. Trucco, J. C. Gorga, and R. M. Chicz. 1999. the most pathogenic of the Gram-positive bacteria that infect man, Naturally processed and presented epitopes of the islet cell autoantigen IA-2 are also distinct in their use of SAgs and have each evolved mul- eluted from HLA-DR4. J. Clin. Invest. 104:1449. tiple SAg toxins. The findings we report suggest that host genetic 23. Papageorgiou, A. C., C. M. Collins, D. M. Gutman, J. B. Kline, S. M. O’Brien, H. S. Tranter, and K. R. Acharya. 1999. Structural basis for the recognition of heterogeneity may have played a role in driving the evolution of superantigen streptococcal pyrogenic exotoxin A (SpeA1) by MHC class II mol- bacterial SAg diversity. ecules and T-cell receptors. EMBO J. 18:9. 24. Baker, M., D. M. Gutman, A. C. Papageorgiou, C. M. Collins, and K. R. Acharya. 2001. Structural features of a zinc binding site in the superantigen streptococcal References pyrogenic exotoxin A (SpeA1): implications for MHC class II recognition. Pro- 1. Fraser, J. D. 1989. High-affinity binding of staphylococcal enterotoxins A and B tein Sci. 10:1268. to HLA-DR. Nature 339:221. 25. Sundberg, E., and T. S. Jardetzky. 1999. Structural basis for HLA-DQ binding by 2. Marrack, P., and J. Kappler. 1990. The staphylococcal enterotoxins and their the streptococcal superantigen SSA. Nat. Struct. Biol. 6:123. relatives. Science 248:705. 26. Miethke, T., C. Wahl, K. Heeg, B. Echtenacher, P. H. Krammer, and H. Wagner. 3. Meissner, H. C., and D. Y. Leung. 2000. Superantigens, conventional antigens 1992. T cell-mediated lethal shock triggered in mice by the superantigen staph- and the etiology of Kawasaki syndrome. Pediatr. Infect. Dis. J. 19:91. ylococcal enterotoxin B: critical role of tumor necrosis factor. J. Exp. Med. 4. Davison, S. C., M. H. Allen, E. Mallon, and J. N. Barker. 2001. Contrasting 175:91. patterns of streptococcal superantigen-induced T-cell proliferation in guttate vs. 27. Abe, J., J. Forrester, T. Nakahara, J. A. Lafferty, B. L. Kotzin, and D. Y. Leung. chronic plaque psoriasis. Br. J. Dermatol. 145:245. 1991. Selective stimulation of human T cells with streptococcal erythrogenic 5. Bunikowski, R., M. E. Mielke, H. Skarabis, M. Worm, I. Anagnostopoulos, toxins A and B. J. Immunol. 146:3747. G. Kolde, U. Wahn, and H. Renz. 2000. Evidence for a disease-promoting effect 28. Tomai, M. A., P. M. Schlievert, and M. Kotb. 1992. Distinct T-cell receptor V␤ of Staphylococcus aureus-derived exotoxins in atopic dermatitis. J. Allergy Clin. gene usage by human T lymphocytes stimulated with the streptococcal pyrogenic Immunol. 105:814. exotoxins and pep M5 protein. Infect. Immun. 60:701. 6. Smoot, L. M., J. K. McCormick, J. C. Smoot, N. P. Hoe, I. Strickland, R. L. Cole, 29. Marrack, P., M. Blackman, E. Kushnir, and J. Kappler. 1990. The toxicity of K. D. Barbian, C. A. Earhart, D. H. Ohlendorf, L. G. Veasy, et al. 2002. Char- staphylococcal enterotoxin B in mice is mediated by T cells. J. Exp. Med. acterization of two novel pyrogenic toxin superantigens made by an acute rheu- 171:455. matic fever clone of Streptococcus pyogenes associated with multiple disease 30. Norrby-Teglund, A., G. T. Nepom, and M. Kotb. 2002. Differential presentation outbreaks. Infect. Immun. 70:7095. of group A streptococcal superantigens by HLA class II DQ and DR alleles. Eur. 7. Dalwadi, H., B. Wei, M. Kronenberg, C. L. Sutton, and J. Braun. 2001. The J. Immunol. 32:2570. Crohn’s disease-associated bacterial protein I2 is a novel enteric T cell superan- 31. Nelson, K., P. M. Schlievert, R. K. Selander, and J. M. Musser. 1991. Charac- tigen. Immunity 15:149. terization and clonal distribution of four alleles of the speA gene encoding py- 8. Bannan, J., K. Visvanathan, and J. B. Zabriskie. 1999. Structure and function of rogenic exotoxin A (scarlet fever toxin) in Streptococcus pyogenes. J. Exp. Med. streptococcal and staphylococcal superantigens in septic shock. Infect. Dis. Clin. 174:1271. North Am. 13:387. 32. Begovich, A. B., W. Klitz, L. L. Steiner, S. Grams, V. Suraj-Baker, 9. Kappler, J., B. Kotzin, L. Herron, E. W. Gelfand, R. D. Bigler, A. Boylston, J. Hollenbach, E. Trachtenberg, L. Louie, P. A. Zimmerman, V. R. S. Hill, et al. S. Carrel, D. N. Posnett, Y. Choi, and P. Marrack. 1989. V␤-specific stimulation 2000. HLA-DQ haplotypes in 15 different populations. In Major Histocompati- of human T cells by staphylococcal toxins. Science 244:811. bility Complex: Evolution, structure and function. M. Kasahara, ed. Springer- 10. McCormick, J. K., J. M. Yarwood, and P. M. Schlievert. 2001. Toxic shock Verlag, Tokyo, p. 412. syndrome and bacterial superantigens: an update. Annu. Rev. Microbiol. 55:77. 33. Arden, B., S. P. Clark, D. Kabelitz, and T. W. Mak. 1995. Human T-cell receptor 11. Haukness, H. A., R. R. Tanz, R. B. Thomson, Jr., D. K. Pierry, E. L. Kaplan, variable gene segment families. Immunogenetics 42:455. B. Beall, D. Johnson, N. P. Hoe, J. M. Musser, and S. T. Shulman. 2002. The 34. Wen, R., M. A. Blackman, and D. L. Woodland. 1995. Variable influence of heterogeneity of endemic community pediatric group a streptococcal pharyngeal MHC polymorphism on the recognition of bacterial superantigens by T cells. isolates and their relationship to invasive isolates. J. Infect. Dis. 185:915. J. Immunol. 155:1884. 12. Hudson, K. R., H. Robinson, and J. D. Fraser. 1993. Two adjacent residues in 35. Newton, D. W., M. Dohlsten, C. Olsson, S. Segren, K. E. Lundin, P. A. Lando, staphylococcal enterotoxins A and E determine T cell receptor V␤ specificity. T. Kalland, and M. Kotb. 1996. Mutations in the MHC class II binding domains J. Exp. Med. 177:175. of staphylococcal enterotoxin A differentially affect T cell receptor V␤ specific- 13. Sundberg, E. J., Y. Li, and R. A. Mariuzza. 2002. So many ways of getting in the ity. J. Immunol. 157:3988. way: diversity in the molecular architecture of superantigen-dependent T-cell 36. Douek, D. C., and D. M. Altmann. 2000. T cell apoptosis and differential HLA signaling complexes. Curr. Opin. Immunol. 14:36. class II expression in human thymus. Immunology 99:249.