Bivalent binding of staphylococcal superantigens to the TCR and CD28 triggers inflammatory signals independently of antigen presenting cells

Staphylococcus aureus superantigens (SAgs) such as staphylococcal enterotoxin A (SEA) and B (SEB) are potent toxins stimulating T cells to produce high levels of inflammatory cytokines, thus causing toxic shock and sepsis. Here we used a recently released artificial intelligence-based algorithm to better elucidate the interaction between staphylococcal SAgs and their ligands on T cells, the TCR and CD28. The obtained computational models together with functional data show that SEB and SEA are able to bind to the TCR and CD28 stimulating T cells to activate inflammatory signals independently of MHC class II- and B7-expressing antigen presenting cells. These data reveal a novel mode of action of staphylococcal SAgs. By binding to the TCR and CD28 in a bivalent way, staphylococcal SAgs trigger both the early and late signalling events, which lead to massive inflammatory cytokine secretion.


Introduction
Staphylococcus aureus is a gram-positive bacterium, which persists as commensal in the human population. It can cause serious infections leading to toxic shock syndrome (TSS) and sepsis, with an overall mortality rate of 25% (1)(2)(3). Type I toxins secreted by Staphylococcus aureus are the most important virulence factors. They, indeed, act as superantigens (SAgs) and stimulate T cells to produce massive inflammatory cytokines, thus leading to vascular damage, multiorgan system breakdown and fatal shock (4)(5)(6). By binding MHC class II and B7 molecules on antigen presenting cells (APCs) as well as specific variable regions within the TCR b chain (TCRVb) and the costimulatory receptor CD28, staphylococcal SAgs activate polyclonally a large proportion of T cells, which in turn produce high levels of inflammatory cytokines (4,7).
Staphylococcal enterotoxin A (SEA) and B (SEB) were the first SAgs to be identified (8,9) and their three-dimensional structures were solved by crystallography (10,11), followed by other staphylococcal SAg structures (12). They share a compact twodomain fold with a b-barrel (oligonucleotide OB-fold) at the Nterminal domain and a b-grasp fold at the C-terminal domain. Both SEB and SEA use hydrophobic residues at the N-terminal OBdomain to bind the a-chain of MHC class II molecules on APCs, outside the MHC class II peptide-binding groove (13,14). Furthermore, SEA binds the b-chain of MHC class II on the opposite side through a tetravalent zinc bridge at the C-terminal domain (15). The interaction of both SEB and SEA with particular elements within the TCRVb, mainly the complementarydetermining region 2 (CDR2) and, to a lesser extent, the hypervariable region 4 (HV4), is mediated by distinct residues located in a shallow groove between the N-terminal b-barrel and the a (2)-helix (16)(17)(18). Thus, staphylococcal SAgs engage TCRVb and MHC class II molecules through distinct regions enabling the formation of a ternary complex, which transmits TCR-dependent signals (19).
Here we explored the capability of staphylococcal SAgs to interact with the TCR and CD28 molecule and to trigger inflammatory signals in the absence of APCs. Indeed, our computational modelling suggested that both SEB and SEA may bind to the TCR and CD28 by using distinct regions on the same side facing the membrane bilayer, even in the absence of MHC class II and B7 molecules. Consistently with our computational simulations, the bivalent binding of both SEB and SEA to the TCR and CD28 triggered both early and late signalling events leading to inflammatory cytokine production in human primary CD4 + T cells. Finally, our data on both the colocalization of the TCR and CD28 in SEB/SEA-stimulated cells demonstrate that, by bridging simultaneously the TCR and CD28, staphylococcal SAgs favour the cooperative activation of inflammatory signals even in the absence of MHC class II-and/or B7-expressing APCs.

Materials and methods
Structural modelling of the interaction between SEA, SEB, TCR and CD28 Structure predictions of full-length proteins were performed in a standalone version of AlphaFold2 and Alpha Fold-Multimer (28), as implemented in ColabFold, which was run on a local computer with Ubuntu 22 operating system and accelerated with two NVIDIA GeForce RTX 2080 Ti GPU. "Template mode" using the following experimentally determined structures was used to this purpose: SEA (PDB: 1ESF) (11); SEB (PDB: 1SEB) ( (21). Standalone MultiLZerD was used for multiprotein docking (35). Other parameters were kept at their default values. Protein sequence manipulations, superpositions and modelling were carried out using PyMod 3.0 (36). Root-mean-square deviation (RMSD) values between structures were calculated using PyMod 3.0.

Cells, antibodies and reagents
Human primary CD4 + T cells were isolated from peripheral blood mononuclear cells (PBMC) by negative selection using a EasySepTM isolation kit (#17952, STEMCELL Technology, CAN) and cultured in RPMI 1640 supplemented with 5% human serum (Euroclone, UK), L-glutamine, penicillin and streptomycin. The purity of the sorted population was > 95%, as evidenced by staining with anti-CD3 plus anti-CD4 Abs. PBMCs were derived from buffy coats of anonymous healthy blood (HD) donors provided by the Policlinico Umberto I (Sapienza University of Rome, Italy). Written informed consent was obtained from blood donors and both the informed consent form and procedure were approved by the Ethics Committee of Policlinico Umberto I (ethical code N. 1061bis/2019, 13/09/2019).
RelA/NF-kB nuclear translocation in primary CD4 + T cells unstimulated or stimulated for the indicated times with 1 mg ml -1 SEB or 0.1 mg ml -1 SEA was performed by using anti-human RelA (1:300 dilution) followed by goat anti-rabbit Alexa-flour 488. Nuclei were labelled with DAPI. Images of RelA localization were obtained with a computer-controlled Nikon Eclipse 50i epifluorescence microscope and a plan achromat microscope objective 100XA/ 1.25Oil OFN22 WD 0.2 and QImaging QICAM Fast 1394 Digital Camera, 12-bit,Mono (Minato, Tokyo, Japan). The fluorescence intensities of cytoplasmic and nuclear RelA, overlapping with DAPI, were quantified by using Fiji ImageJ software and intensities ratio were calculated for each cell. At least seventy cells were examined quantitatively for each condition in three independent experiments.
Confocal microscopy analysis of the colocalization of CD3 and CD28 with SEB or SEA Primary CD4 + T cells incubated for 5 minutes with 1 mg ml -1 SEB-Alexa-fluor-594 or 0.1 mg ml -1 SEA-Alexa-fluor-594 were fixed by 3% paraformaldehyde and seeded on poly-L-lysin (#8920, Sigma) coated-cover glasses (12 mm). After washing and permeabilization with 0.1% saponin in PBS containing 1% BSA, CD28 and CD3 staining were performed by using rabbit anti-human CD28 (1:100 dilution) followed by goat anti-rabbit Alexa-flour 488, and mouse anti-human CD3 (OKT3) followed by goat anti-mouse Alexa-flour 488 (1:150 dilution). Nuclei were stained with DAPI. Confocal observations were performed with a 63X oil objective, Nikon Eclipse Ti2 confocal microscope, and Z stack images were processed by NIS Elements AR 5.30 software (Nikon Europe B.V.) using the same acquisition settings. The colocalizations of CD28 or CD3 with SEB or SEA were quantified by using Fiji ImageJ software. The Pearson correlation coefficient (PCC) in each cell was calculated with a range +1 (perfect correlation) to -1 (perfect exclusion) (42).

Statistical analysis
The sample size was chosen based on previous studies to ensure adequate power. Parametrical statistical analysis (mean and standard deviation) was performed to evaluate differences between continuous variables through Prism 8.0 (GraphPad Software, San Diego, CA), using Student's t test or one-way ANOVA with the Fisher's LSD test for multiple comparisons. For all tests, P values < 0.05 were considered significant.
The obtained models are in agreement with previous data showing the interaction of SEB with the TCR, CD28 and MHC class II molecules (21). CD28 is predicted to bind to a distant region of SEB ( Figure 1A) or SEA ( Figure 1B) compared to the TCR binding sites, but on the same side facing the membrane bilayer. In agreement with Arad et al. (20) and Levy et al. (23), SEB adopts a wedge-like conformation in binding both the TCR and CD28, with the residues 150-161 in the b-strand 8 /hinge/a-helix 4 domain of SEB in close contact with the CD28 homodimer interface ( Figure 1A; Supplementary Movie S1). A very similar (RMSD ≈ 2.0 Å) conformation and orientation is predicted to be adopted by SEA in TCR and CD28 binding ( Figure 1B; Supplementary Movie S2), as also evidenced by the co-crystallized structure of SEA with the TCR (Supplementary Movie S2) (18). Indeed, as shown by contacts mapping on the surface of SEB (Figure 2A), most of the accessible surface area of SEB is engaged in contacts of roughly the same extent with CD28 (≈ 1320 Å 2 ) and the TCR (≈ 1380 Å 2 ) and, to a lesser extent, with MHC class II (≈ 950 Å 2 ) and B7 (≈ 500 Å 2 ). Given the very similar interaction binding mode of SEA, the values of SEA contact area ( Figure 2B) are in good agreement with those measured for SEB (TCR ≈1266 Å 2 ; CD28 ≈1580 Å 2 ). Therefore, the structural modelling of SAgs and their full-length interactors and the analysis of the interacting regions suggest that SEB and SEA might bind simultaneously to the TCR and CD28 even in the absence of MHC class II and B7 molecules, thus triggering inflammatory signals. To verify the functional relevance of our computational simulations, we performed a kinetic analysis of inflammatory cytokine secretion in culture supernatants of highly purified (> 95%) primary CD4 + T cells (Supplementary Figures S1A, B) stimulated with SEB alone. Consistently with our previous data (27,43), CD4 + T cells isolated from the peripheral blood of healthy donors (HD) expressed CD28 (mean ± SEM = 96.37 ± 2.41) but neither HLA-DR (mean ± SEM = 3.1 ± 0.9) nor B7.1/CD80 (mean ± SEM = 0.98 ± 0.59) nor B7.2/CD86 (mean ± SEM = 1.81 ± 0.88). Stimulation of CD4 + T cells with SEB alone induced a strong production of IFN-g ( Figure 3A), IL-2 ( Figure 3B), IL-6 ( Figure 3C), TNF-a ( Figure 3D), IL-17A ( Figure 3E), IL-22 ( Figure 3F) and GM-CSF ( Figure 3G) after 48-72 hours. To ascertain that inflammatory cytokine production by CD4 + T cells was independent of contaminating accessory cells as found in some samples, we compared SEB-induced inflammatory cytokine production in highly purified CD4 + T cells (> 99%, mean = 99.4), expressing very low levels of HLA-DR (mean = 1.  Figures S2F, G). A significant increase in TNF-a was observed in T cells stimulated with SEB in the presence of 5% (mean ± SEM = 1390 ± 534) and 10% 5-3.1/ B7 cells (mean ± SEM = 1644 ± 282) compared to SEB alone (mean ± SEM = 830 ± 162) (Supplementary Figure S2H). Likewise, stimulation of T cells with SEB in the presence of 2.5% (mean ± SEM = 1545 ± 81), 5% (mean ± SEM = 1700 ± 216) and 10% 5-3.1/B7 cells (mean ± SEM = 2056 ± 375) significantly enhanced GM-CSF production compared to SEB alone (mean ± SEM = 1143 ± 305) (Supplementary Figure S2I). Similarly to SEB, stimulation of CD4 + T cells with SEA elicited comparable amounts of inflammatory cytokines after 48-72 hours of stimulation ( Figures 4A-G) Altogether these data are consistent with our computational modelling and demonstrate that SEB and SEA are able to directly induce inflammatory cytokine release from T cells independently of MHC class II-and B7-expressing APCs.

Binding of SEB and SEA to TCR and CD28 elicit inflammatory signals in T cells
To investigate whether SEB binding to T cells required TCR and/or CD28, we labelled SEB with AlexaFluor-594 and we performed fluorescence microscopy analyses in a CD28-negative CH7C17 Jurkat T cell line expressing TCR Vb3.1 (37) that specifically interacts with SEB (44), CH7C17 cells reconstituted with human CD28WT (38), and 31.13 Jurkat cells (39) that express CD28 but not TCR/CD3 ( Supplementary  Figures S1F, G). The kinetic analysis performed in CD28WT cells evidenced that SEB efficiently bound to the cell surface with a peak of mean fluorescence intensity (MFI = 636) at 5 minutes ( Figure 5A). SEB was also able to bind CD28-negative CH7 cells ( Figure 5B), although with less efficacy as demonstrated by the lower fluorescence intensity (MFI = 427). The requirement of both the TCR and CD28 for optimal SEB binding was also confirmed by the reduced fluorescence intensity (MFI = 170) observed in the TCR-negative 31.13 Jurkat T cell line ( Figure 5B).
We next analysed the capability of SEB and SEA to directly bind the surface of primary CD4 + T cells and to trigger TCR/CD28 inflammatory signals. Consistently with CD28WT cells, the binding of SEB to T cells was detected after 1 minute of incubation and its MFI increased at 5-15 minutes ( Figure 5C). Similarly, after 5 minutes of incubation, SEA efficiently bound to the surface of T cells with a fluorescence intensity comparable to SEB ( Figure 5D).
Taken together these findings show that SEB and SEA might directly bind to both TCR and CD28 and activate the early and late signalling events associated with inflammatory cytokines.

SEB and SEA colocalization with the TCR and CD28 in primary CD4 + T cells
To demonstrate the capability of both SEB and SEA to bind to the TCR and CD28 in a bivalent manner, we evaluated the colocalization of either SEB or SEA with the TCR and CD28 in primary CD4 + T cells. Confocal microscopy analysis evidenced that soon after 5 minutes of stimulation, both SEB and SEA ( Figure 8A) efficiently colocalize with either CD3 or CD28, as demonstrated by the values of the Pearson correlation coefficient (PCC) (Figures 8B,  C). Moreover, in stimulated CD4 + T cells ( Figure 8D), we also found a substantial colocalization of both CD28 and CD3 with either SEB or SEA ( Figure 8E Altogether these data support our computational and functional data on the capability of staphylococcal SAgs to promote the activation of pro-inflammatory signals by simultaneously engaging the TCR and CD28 even in the absence of APCs expressing MHC class II and B7 molecules.

Discussion
For several decades, T-cell activation by bacterial SAgs was believed to require only the TCR and MHC class II molecules (12). However, the discovery of CD28 and B7 molecules as novel co-ligands of staphylococcal SAgs (20,21,23,26) together with our recent data showing that the binding of SEB to B7 molecules on APCs and to CD28 on T cells leads to a productive engagement of the TCR even in the absence of MHC class II, changed this classical view (27). Here, we show that SEB-and SEA-mediated activation of inflammatory signals in T cells also can occur in the absence of MHC class II-and B7-expressing APCs.
Starting from the available experimental data on the complex between SEB, the TCR and MHC class II (19), we initially rebuilt a tentative model of the whole SEB-MHCII-TCR-CD3gϵ-CD3dϵ-CD3zz signalling hexamer by exploiting the recently released cryo-EM structure (32). This first model permitted us to pinpoint the orientation of SEB relative to the lipid bilayer and the surface regions left for interaction with CD28. Then, using the crystal structures of CD28 (29, 30) and a spatial-restrained protein-protein docking approach based on previous experimental data (18,19,21), we obtained a final complex that permitted us to analyse, from a structural point of view, the interaction network of SEB with its ligands ( Figure 1A; Supplementary Movie S1). The same procedure was adopted for the highly similar SEA ( Figure 1B; Supplementary Movie S2), and the two obtained models suggested that most of the interaction surfaces of SEA and SEB were occupied by contacts with the TCR and CD28 (Figure 2). These data suggest that the coengagement of the TCR and CD28 by staphylococcal SAgs might suffice for the activation of inflammatory signals. Indeed, we demonstrate here that both SEA and SEB induce a strong production of inflammatory cytokines in highly purified CD4 + T cells in the absence of APCs (Figures 3, 4; Supplementary Figure S2). These data support the capability of SEA and SEB to cooperatively interact with both the TCR and CD28 eliciting inflammatory signals.
Efficient T-cell activation and inflammatory cytokine production require both TCR and CD28 signals (48, 51). Generally, both signals are initiated at the immunological synapse (IS) following the engagement of the TCR and CD28 by peptide-MHC complexes and B7 molecules on APCs, respectively (52). Similarly to the IS, in the superantigen synapse, the engagement of CD28 and B7 by staphylococcal SAgs together with the TCR triggers both TCR-and CD28-dependent signals, thus leading to optimal T cell activation (24,51,53) and inflammatory responses (20,21,23,26), even in the absence of MHC class II co-engagement (27). However, primary CD4 + T cells may generate mechanical forces through the TCR and CD28 when stimulated by high affinity ligands also in the absence of cell-cell interaction, thus leading to productive TCR-and CD28activating signals (54,55). Here we show that staphylococcal SAgs efficiently bind to the surface of TCR + CD28 + Jurkat cells and primary CD4 + T cells ( Figure 5). Notably, the binding of SAgs was maximal when both TCR and CD28 were expressed, as evidenced by the reduction of SEB binding observed in TCR -CD28+ 31.13 or TCR + CD28 -Jurkat cells compared to TCR + CD28WT cells ( Figure 5B). Although the binding affinities of SEA and SEB for the TCR and CD28 are quite similar (20,56,57), the binding of both SAgs to the cell surface increased when both the TCR and CD28 were co-expressed ( Figure 5B). These data are consistent with the findings B C D A FIGURE 5 SEB and SEA efficiently bind T cell surface in the absence of APCs. (A) Fluorescence microscopy imaging of CH7C17 Jurkat cells expressing human CD28WT incubated for the indicated times with SEB-Alexa Fluor 594 (SEB-594, 1 mg ml -1 ). Nucleus was stained with DAPI (blue). Scale bar = 10 mm. that TCR stimulation activates an inside-out signalling that induces the formation of a more stable CD28 homodimer interface (58)(59)(60), likely favouring an optimal binding of staphylococcal SAgs as well as signalling (20). For instance, the mitogenicity of CD28 ligands depends on the epitope location (29) as evidenced by ability of superagonistic Abs to induce CD28 clustering and signalling by binding in a bivalent way a loop close to the homodimer interface (61). Likewise, by simultaneously binding the TCR and CD28, SEB and SEA might act in a superagonistic way triggering optimal activating signals. It is well known that the TCR and CD28 cooperate to activate the early and late signalling events regulating cytokine production (48, 51,   70 recruitment and activation (71) elicits a PLC-g1-dependent Ca 2+ response that in turn induce a conformational change in the CD28 cytoplasmic domain facilitating CD28 signalling (70). On the other hand, CD28 enhances and sustains early TCR signalling (24,(72)(73)(74) by favouring the actin cytoskeleton events required for the recruitment of critical signalling mediators to both the TCR and CD28 (38,41,(75)(76)(77). Consistently, we found that SEB and SEA binding to the TCR and CD28 triggers a sustained activation of both ZAP-70 and PLC-g1 ( Figures 6A-D; Figures 7A, B). Moreover, both SEB and SEA also induced the nuclear translocation of RelA/NF-kB ( Figures 6E, F;  Figures 7C, D), an important transcription factor that is involved in inflammatory cytokine gene expression and is mainly regulated by the CD28 signalling axis (41,48,(78)(79)(80)(81). Finally, our data on the colocalization of both SEB and SEA with either CD3 or CD28 together with the colocalization of both CD3 and CD28 with SEB or SEA on the surface of primary CD4 + T-cells (Figure 8), suggest that staphylococcal SAgs might function as bivalent stimulators of the TCR and CD28. For instance, by binding simultaneously to the TCR and CD28, staphylococcal SAgs may induce the dimerization and/or conformational changes required for triggering the early tyrosine phosphorylation events, which in turn will promote the recruitment of important signalling mediators and the activation of downstream inflammatory cascades, as occur for receptor tyrosine kinases (82,83).
Altogether, our data reveal a novel model of T-cell activation by bacterial SAgs, which share the ability to bind both TCR and CD28/ B7 molecules, exemplified by staphylococcal and streptococcal toxins (23, 26), and provides new insight for the design of novel SAg antagonists able to interfere with the formation of SAg-TCR-CD28 complex, thus inhibiting massive inflammatory signals.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by Ethical Committee of the Policlinico Umberto I (Sapienza University, Rome, Italy). The patients/participants provided their written informed consent to participate in this study.

Author contributions
MK performed most of the experiments, analysed the data, interpreted the results and helped in writing the manuscript. CA performed part of the experiments and data analyses. FS performed part the experiments and contributed to discussions and editing the manuscript. SC contributed with technical support. MF contributed to discussions and editing the manuscript. AP performed the computational analyses and contributed to discussions, writing and editing the manuscript. RK contributed to discussions and editing the manuscript. LT designed the study, coordinated the work and wrote the manuscript. All authors contributed to the article and approved the submitted version.

In memoriam
This paper is dedicated to the memory of our friend and colleague Silvana Caristi who prematurely died in 2022.