Immune Cell Activation by Enterotoxin Gene Cluster (egc)-Encoded and Non-egc Superantigens from Staphylococcus aureus¹

Staphylococcus aureus is both a successful colonizer and an important pathogen in humans. These bacteria cause a wide spectrum of infectious diseases including several toxin-mediated diseases. Among the numerous toxins of S. aureus are the 19 known staphylococcal superantigens (SAgs)³: the toxic shock syndrome toxin (TSST-1), the staphylococcal enterotoxins (SEA-SEE, SEG-SEJ), and the staphylococcal enterotoxin-like toxins (SElK-SElR and SElU) (1, 2, 3, 4). SAgs are the causative agents of toxic shock syndrome, and might also contribute to septic shock (3, 5). They directly cross-link conserved regions of the variable domains of the TCR β-chain (TCR Vβ) with MHC class II molecules (outside the peptide-binding cleft) on APCs. This results in a strong stimulation of T cells that express the matching TCR Vβ element on their surface. Activated T cells respond with proliferation and massive cytokine release. In this way, SAgs activate up to 20% of all T cells. In contrast, conventional Ags only stimulate around 0.001% of T cells. They require uptake by APCs, processing into peptides, loading onto MHC-II molecules, and presentation on the cell surface, where their specific recognition is mediated by the hypervariable loops of TCR α- and β-chains.

The recently described enterotoxin gene cluster (egc) harbors 5 to 6 SAg genes (seg, sei, selm, seln, selo, and sometimes selu), which cluster on a staphylococcal pathogenicity island (νSaβ) (6, 7). The egc-genes are the most prevalent SAg genes in commensal and invasive S. aureus isolates; frequencies between 52 and 66% have been reported (8, 9, 10). However, they appear to cause toxic shock only very rarely (11). In fact, egc SAgs are significantly more frequent in commensal strains than in invasive isolates, and their presence is negatively correlated with severity of S. aureus sepsis (12, 13). Because of this, it was suggested that SAgs might differ in their pro-inflammatory potential, and Dauwalder et al. (14) reported that in PBMC from healthy donors the non-egc SAg SEA induces a stronger Th1-response than the egc SAg SEG.

In addition to their superantigenicity, SAgs, like other proteins, also act as conventional Ags and induce a specific Ab response. Abs against non-egc SAgs (e.g., TSST-1, SEA, SEB, SEC) are common in the healthy population (15, 16, 17, 18). In S. aureus carriers, these Abs are highly specific for the SAgs of the colonizing strain and very effectively neutralize their SAg effects (19). In contrast, neutralizing Abs against egc SAgs are very rare, even among carriers of egc-positive S. aureus strains (18) (and unpublished observations). This “egc-gap” in the Ab response of healthy individuals was unexpected, because of the high prevalence of egc SAg genes in clinical S. aureus isolates (8, 9, 10).

In search of an explanation for these counterintuitive observations, we have tested two non-exclusive hypotheses: 1) egc and non-egc SAgs have unique intrinsic properties and drive the immune response into different directions and 2) egc and non-egc SAgs are released by S. aureus under different conditions, which shape the immune response to them. To test these hypotheses, we compared the effects of egc and non-egc SAgs on human blood cells. Their T cell-mitogenic potencies, the elicited cytokine profiles as well as their impact on gene expression were highly similar. Both egc and non-egc SAgs induced a very strong pro-inflammatory response. In contrast, the regulation of SAg release by S. aureus differed markedly between egc and non-egc SAgs.

We used four clinical S. aureus isolates from a strain collection from hospitals in Northeast Germany that were investigated previously (18). Their SAg gene repertoire was determined by multiplex-PCR (pSA17: seb, selp; pSA20: sec, sell, selp; aSA2: seg, sei, selm, seln, selo, selu; aSA4: seg, sei, selm, seln, selo) as described before (10).

SEB, SElQ, TSST-1, SEI, SElM, and SElO were produced by recombinant gene technology. The gene sequences were amplified from the sequenced S. aureus strains N315 (tst, sei, selm, selo) and COL (seb, selq) using primers containing BsaI restriction sites (underlined) (Table I). PCR products were purified with the Qiagen Quick PCR Purification Kit (Qiagen), digested with BsaI and introduced into the Escherichia coli plasmid pPR-IBA1 with Strep-tag II (IBA). The resulting plasmids were amplified in E. coli DH5α and then transfected into E. coli BL21 pLysS for overexpression. Recombinant proteins were purified with Strep-Tactin Superflow columns according to the manufacturer’s instructions (IBA). The purity of the recombinant SAgs was assessed by SDS-PAGE stained with colloidal Coomassie Brilliant Blue. Protein concentrations were determined with the BCA Protein Assay Kit (Pierce). LPS concentrations were determined with a limulus amebocyte lysate assay (QCL1000, Lonza). Contaminating LPS was very efficiently removed by two rounds of LPS depletion with the EndoTrap red columns (Profos). In our proliferation and cytokine secretion assays, the final LPS concentrations ranged between 0.4 (0.004 EU/ml) and 31.6 pg/ml (0.316 EU/ml). The final concentrations of LPS in the stimulation experiments for microarray analysis were always below 31.6 fg/ml (3.14 × 10⁻⁴ EU/ml).

T cell proliferation was assessed after 72 h culture by incorporation of [³H]thymidine. Human PBMC (10⁵/well) were stimulated with recombinant SAgs titrated over a broad concentration range (0.1 pg/ml to 1 μg/ml) as described previously (18, 19). A control protein (S. aureus anti-σ factor RsbW), which was overexpressed and purified in parallel to the recombinant SAgs, did not induce T cell activation. To analyze the growth phase-dependent secretion of SAgs, we cultured two S. aureus strains with egc SAg genes (aSA2, aSA4) and two strains with non-egc SAg genes (pSA17, pSA20) in Luria broth medium. Cultures were sampled at optical densities (OD₆₀₀) of 0.05, 0.15, 0.4, 1.0, 2.0, 3.0, 4.0, 5.0, and 5.5. Culture supernatants were adjusted by dilution in cell culture medium to OD₆₀₀ 0.05 of the S. aureus culture. Their T cell-mitogenic potency was assessed by incubating 10⁵ human PBMC/well with 1000-fold dilutions of the adjusted bacterial culture supernatants in 96-well flat-bottom plates in the presence of RPMI 1640 supplemented with 10% FBS.

Human PBMC (10⁵/well) were stimulated with recombinant SAgs at concentrations 10-fold above the lowest concentration, which triggered the maximal proliferation (100 pg/ml to 100 ng/ml). Culture supernatants were tested for cytokines after 72 h. The concentrations of IL-2, IL-4, IL-5, IL-10, TNF-α, and IFN-γ were measured with the BD Cytometric Bead Assay Human Th1/Th2 Cytokine Kit (BD Biosciences) according to the manufacturer’s instructions.

Differences between egc and non-egc SAgs were assessed using the Kruskal-Wallis test. For T cell proliferation assays, the SAg concentrations inducing maximal proliferation were compared. The p values below 0.05 were considered statistically significant. Median values are depicted in all images.

PBMC from three different blood donors were stimulated with increasing concentrations of the recombinant SAgs SEI and SEB. After 6 h of stimulation, 10⁷ PBMC were lysed with TRIzol (Invitrogen) followed by purification of total RNA using an RNeasy Micro Kit (Qiagen) and subsequent precipitation with sodium acetate. RNA concentration and quality were assessed using a Nanodrop ND-1000 (NanoDrop Technologies) and an RNA 6000 Nano LabChip on the Bioanalyzer 2100 (Agilent Technologies). In parallel, the lowest SAg concentrations which elicited maximal responses in 72 h proliferation assays were determined for each case, and RNA samples obtained by stimulation with 10-fold higher (plateau) SAg concentrations (10 to 100 pg/ml) were selected for microarray analysis.

Starting with 2.5 μg of total RNA, labeling was performed using the One-Cycle Target Labeling protocol according to the manufacturer’s instructions (Affymetrix). Size distribution of biotin-labeled amplified RNA was analyzed using the Bioanalyzer 2100 and concentration was measured with the Nanodrop ND-1000. Gene expression arrays (GeneChip Human Genome U133 Plus 2.0) were used applying standard protocols (Affymetrix) and arrays were scanned on a GeneChip Scanner 3000 (Affymetrix).

The Affymetrix GeneChips used in this study analyze the expression of 54,675 probe sets. Affymetrix array image data were processed with the GeneChip Operation Software (GCOS, Version 1.4, Affymetrix) with the MAS5.0 algorithm and default settings. Expression raw data were transferred to the GeneSpring GX software package version 7.3.1 (Agilent Technologies). Very low raw signal intensities were raised to the threshold level of 10 U to avoid artificially high ratios in the subsequent calculations. Expression values were then subjected to a “per chip” and “per gene” median normalization. Statistical analysis was restricted to probe sets flagged present or marginal by GCOS software in at least two of the three samples (different blood donors), in untreated controls and/or in samples stimulated with SEI or SEB.

To identify differentially expressed probe sets, intensity ratios between SAg stimulated and untreated cells were calculated for each donor and each SAg. These ratios (PBMC stimulated with SEI vs untreated controls or PBMC stimulated with SEB vs untreated controls) were used for statistical analysis performed with the rank products method (20), which is especially suited for ratio comparison in experiments with small numbers of replicates (21). Data underwent rank product statistics with 1,000 iterations resulting for each probe set in values for the average ratio and two false discovery rates (FDR) (22), one for up-regulation and one for down-regulation. Probe sets with an FDR value less than 0.05 for either up- or down-regulation were considered to represent significant changes in gene transcription.

The signal intensities of all probe sets displaying significant differences in intensity between untreated controls and samples stimulated with SEI or SEB were used for clustering of samples using Pearson correlation; the intensity ratios of the same probe sets were also subjected to a principal component analysis (PCA). Both analyses were performed with GeneSpring GX (Agilent Technologies). The data discussed in this publication have been deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE11281.

To compare the T cell-mitogenic activity of egc and non-egc SAgs, we stimulated human PBMC from nine healthy blood donors with three egc SAgs (SEI, SElM, and SElO) and three non-egc SAgs (SEB, SElQ, and TSST-1) and measured their proliferation. All six SAgs induced very strong T cell proliferation in a dose-dependent manner with some interindividual variations in the maximal response to single SAgs (Fig. 1). This likely reflects heterogeneity in the TCR Vβ family composition of the T cell pools from our healthy probands as well as MHC gene polymorphisms, which may affect SAg binding. With the exception of SElO, all SAgs induced strong proliferation already in the pg/ml concentration range. We then determined the SAg concentrations inducing half-maximal proliferation to be ∼7 pg/ml for SEB, 1.5 pg/ml for SElQ, 1 pg/ml for TSST-1, 0.5 pg/ml for SEI, 20 pg/ml for SElM, and 2200 pg/ml for SElO (data not shown). Hence, except for SElO (p = 0.00009), we found no significant differences between the T cell-mitogenic potencies of egc and non-egc SAgs.

Non-egc SAgs are notorious for their ability to induce a massive cytokine response in T cells and APCs. Therefore, we next investigated egc SAgs for their ability to trigger the release of pro- and anti-inflammatory cytokines by PBMC and compared this with the effects of non-egc SAgs. Human PBMC from seven different blood donors were stimulated for 72 h with recombinant SAgs and the cytokine concentrations in the supernatants were determined. We used the SAgs at concentrations just above those which elicited maximal proliferation (plateau concentrations) (Fig. 1). All SAgs induced the secretion of large amounts of pro-inflammatory (IFN-γ, TNF-α, IL-2) and anti-inflammatory (IL-4, IL-5, IL-10) cytokines (Fig. 2). There was considerable interindividual variation in strength of the cytokine response: Some individuals released high amounts of all measured cytokines, while in others the concentrations were uniformly low. Notably, individual recombinant SAgs were as potent as the mitogen PHA.

The dominating cytokine was IFN-γ, the lead cytokine of Th1 cells, with median concentrations between 7570 (SElQ) and 21610 pg/ml (SEI). High levels of TNF-α and IL-2 were also found. Besides these proinflammatory cytokines, SAgs also induced secretion of the Th2-cytokines IL-5 (91.5pg/ml) and IL-4 (11.8pg/ml), though at much lower concentrations. In addition, large amounts of the immunosuppressive cytokine IL-10 were detected, which ranged from median values of 170 (TSST-1) to 450 pg/ml (SEB). As shown in Fig. 2, we observed no significant differences between the cytokine profiles triggered by egc and non-egc SAgs.

For a comprehensive view of the SAg response, we next selected the egc SAg SEI and the non-egc SAg SEB and used them to stimulate PBMC from three different donors. After 6 h of stimulation, the cells were harvested and their transcription profiles analyzed with Affymetrix expression arrays. The gene expression data showed that 391 genes (511 probe sets, FDR < 0.05) were influenced by stimulation of PMBC with SEI and/or SEB. As shown in Fig. 3^,A, two-thirds of these genes (262 genes) were up-regulated and one-third was down-regulated (129 genes). The vast majority of genes was influenced to a similar extent by SEI and SEB (supplemental Tables I and II).⁴ In general, the stimulation with SEI was slightly stronger than with SEB, so that more of the observed changes in gene expression reached significance (350 genes vs 249 genes). This fits to the observation that 91% (171 of 188 genes) of all SEB-induced genes were also significantly induced by SEI, while only 17 genes were significantly up-regulated only by SEB (Fig. 3 A, upper panel). Repression of gene expression was generally less impressive, but still more than half of the genes influenced by SEB (61%, 37/61 genes) were also significantly affected by SEI. Notably, no gene was regulated in opposite directions by SEI and SEB (data not shown). The complete list of regulated genes is documented in supplemental Tables I and II.

We then conducted a hierarchical clustering analysis of all probe sets significantly affected by SEI and/or SEB to group samples based on the similarity of their expression data. Fig. 3 B illustrates the results of this analysis as a condition tree. As expected, unstimulated and SAg-stimulated samples clustered separately. Strikingly, the SAg-exposed samples of the three healthy blood donors clustered separately, but not the two SAgs used for stimulation as one would have expected, if SEI and SEB had different effects on PBMC. Thus, the condition tree clearly shows that interindividual differences between the blood donors outweighed the differences between the stimulating SAgs.

In an independent approach, the data were submitted to a PCA. This PCA was based on the ratios between the probe set signals of stimulated and non-stimulated PBMC, which had been calculated separately for each donor and each SAg. PCA converges the majority of statistical variables of complex data sets to three informative dimensions (principal components), which are visualized in Fig. 3 C. Similar to the hierarchical clustering, the PCA grouped our samples into three pairs, which comprised the SEI- and SEB-stimulated samples of each donor. Therefore, both analysis tools show that the effects of SEI and SEB on PBMC gene expression were very similar, so that differences between them had less impact than the interindividual variation between the blood donors.

Table II displays those genes, whose transcription was induced >10-fold by SAg treatment. No gene was repressed by a factor of 10 or more. Among the most strongly induced genes were those encoding for IL-2 (SEI: 318.4fold; SEB: 166.1-fold), IFN-γ (SEI: 165.8-fold; SEB: 127.2-fold) and IL-17A (SEI: 98.5-fold; SEB: 57.0-fold), indicative of a very strong Th1- and Th17-response. But also the cytokines IL-22, IL-3, and IL-27, the chemokines CXCL9, CXCL10, CXCL11, XCL2, and CCL8 as well as the T cell activation marker CD64 were strongly induced. These findings are in good agreement with the results of our T cell proliferation and cytokine measurements (Figs. 1 and 2). The T cell activation marker genes CD69, CD40L, and CD25; the cell cycle genes cyclin D2 and cyclin-dependent kinase 6 as well as the transcripts for the cytokines IL-4, IL-5, and TNF-α were also significantly induced by both SEI and SEB but <10-fold (see supplementary Table I). The only discrepancy concerns IL-10, where protein secretion was detected by cytokine measurements but no induction of gene transcription by SEI or SEB was observed. However, IL-10 is typically produced in the later phase of the immune response, while in this study the gene expression patterns were assessed already after 6 h of stimulation.

In search of an explanation for the lack of neutralizing Abs against egc SAgs even in carriers of egc-positive S. aureus strains, we finally focused on the regulation of the SAg release by S. aureus. It has been reported that the egc-operon is transcribed in the exponential but not in the stationary growth phase, as it is typical for non-egc SAgs (6, 23). We cultured two clinical S. aureus strains exclusively with egc (aSA2: seg, sei, selm, seln, selo, selu; aSA4: seg, sei, selm, seln, selo) and two strains with only non-egc SAgs genes (pSA17: seb, selp; pSA20: sec, sell, selp) and sampled supernatants at optical densities (OD₆₀₀) between 0.05 and 5.5. To correct for the higher cell densities and resulting higher concentrations of secreted products at the late growth phase, all samples were adjusted to OD₆₀₀ 0.05 by dilution in cell culture medium (Fig. 4). These normalized bacterial supernatants from different growth phases of the S. aureus isolates were then used at 1/1000 dilutions to assess their T cell-mitogenic potency in a proliferation assay. Supernatants from S. aureus strains pSA17 and pSA20 showed mitogenic activity only at high OD₆₀₀ (>OD₆₀₀ 3.0), indicating that these non-egc SAg proteins are secreted primarily in the late stationary growth phase of S. aureus. In contrast, egc SAgs were released during exponential growth but secretion stopped at higher bacterial densities. The SAgs were not degraded (data not shown), but on a per cell basis their concentration decreased to submitogenic levels with increasing bacterial densities.

Egc SAgs are by far the most prevalent staphylococcal SAgs (18, 24). Despite this, they are not a prominent cause of toxic shock syndrome (4, 11, 25) but their presence appears to be associated mainly with symptom-free carriage and inversely correlated with the severity of S. aureus infection (12, 26). Astonishingly, neutralizing anti-egc serum Abs are very uncommon in healthy individuals (18). In search of an explanation for these counterintuitive observations, we proposed two hypotheses: 1) egc and non-egc SAgs have unique intrinsic properties and drive the immune system into different directions and 2) egc and non-egc SAgs are released by S. aureus under different conditions, which shape the immune response to them. Our results lend support only to the latter.

Our purified and rigorously LPS-depleted recombinant egc (SEI and SElM) and non-egc SAgs (TSST-1, SEB, SElQ) were of similarly high mitogenic potency, inducing half-maximal proliferation at concentrations between ∼0.5 and 20 pg/ml. Only the egc SAg SEIO was significantly less potent (∼2200 pg/ml). These findings corroborate and extend earlier studies where mitogenic concentrations for non-egc staphylococcal SAgs were determined to be in the pg/ml or even fg/ml range (27, 28, 29). To date, the mitogenic potency of the other egc SAgs, SEG, SElN, and SElU, has not been determined in a human T cell proliferation assay. The group of Munson et al. (23) reported that concentrations of 1.84 nM SEG (∼40 ng/ml) were necessary to induce maximal proliferation in murine splenocytes, but murine cells are less susceptible to SAg stimulation than human PBMC by several orders of magnitude (25, 30).

If intrinsic properties of egc and non-egc SAgs were responsible for the different response of the immune system, one might expect them to elicit contrasting cytokine profiles. However, we observed no systematic differences between egc and non-egc SAgs. Both groups induced the release of high amounts of pro-inflammatory (IFN-γ, TNF-α, IL-2) and lower concentrations of anti-inflammatory (IL-4, IL-5, IL-10) cytokines. IFN-γ was the lead cytokine in our panel, reaching average concentrations of ∼17 ng/ml after 72 h. Others have also reported induction of Th1- and Th2-cytokines after stimulation with a number of non-egc staphylococcal and streptococcal SAgs (31, 32, 33). In contrast, Dauwalder and coworkers (14) recently reported that stimulation of human PBMC with very high concentrations (100 ng/ml) of SEA and the egc SAg SEG did not elicit anti-inflammatory cytokine (IL-10) and chemokine (TARC) secretion. Moreover, only SEA but not SEG induced the release of TNF-α and the Th1-chemokine MIP-1α in their system. This suggests that SEG might have unique properties among the egc SAgs, but differences in the SAg concentrations and/or SAg production and purification procedures should also be considered as an explanation (34).

These differences prompted us to aim for a comprehensive picture of the response of blood cells to SAg stimulation. We used transcriptional profiling to compare one egc SAg (SEI) with one non-egc SAg (SEB). Stimulation with SEI and/or SEB changed the transcription of 391 genes, two thirds of which were up-regulated. This is in agreement with Mendis and coworkers (35), who also treated PBMC with SEB. Stimulation with SEI was slightly stronger than with SEB, probably because this SAg activates a larger fraction of T cells. But the hierarchy of induced genes was very similar, and not a single gene was regulated in opposite directions by the two SAgs. In fact, there were interindividual variations of the blood donors in their SAg response, as has also been reported by others (32, 33), and these outweighed the differences between the transcription profiles induced by SEI and SEB. Among the most strongly induced genes were proinflammatory mediators of Th1- as well as cytotoxic T cell-responses: IL-2 and IFN-γ were up-regulated more than 100-fold, supporting our observations on the protein level. Fitting the same response profile, CD40L, lymphotoxin-α, granzymes A and B, TNF-α, IL-12 (p35) and IL-27 were also highly induced. Additionally, we noticed highly increased transcription levels of IL-17A and IL-17F that indicate activation of Th17 cells. We interpret these findings to represent an extremely strong proinflammatory in vitro reaction to both SAgs.

However, in agreement with our cytokine measurements, the Th2-cytokine genes IL-4 and IL-5 were also induced. As expected, the T cell activation marker genes (CD69, CD40L, and CD25) and cell cycle genes (cyclin D2 and cyclin-dependent kinase 6) were up-regulated, and the transcription of chemokines guiding the migration of activated T cells, CXCL9, CXCL11, CXCL10, and CCL8 was also increased (36). In addition, we found enhanced transcription levels of the apoptosis-related genes trail, fas, casp10, casp7, which is not surprising, because very strong T cell stimulation may drive activation-induced cell death, and the elimination of SAg-reactive T cell subpopulations is considered to be a hallmark of SAg action (37). In summary, the egc SAg SEI and the non-egc SAg SEB induced very similar gene expression patterns corresponding to an extremely strong activation of T cells and APCs such as one might expect from SAgs.

Taken together, the immune cell-activating properties of egc and non-egc SAgs, their superantigenicity, proved very similar in every aspect studied and cannot explain the striking differences in the immune response to egc and non-egc SAgs, with a high prevalence of neutralizing Abs specific for non-egc but not for egc SAgs (15, 16, 17, 18).

Because the amino acid sequences of the egc SAgs are more closely related to those of individual non-egc SAgs than to each other (6, 38), it also appears unlikely, that the two groups of SAgs differ systematically in their immunogenicity.

We found only one important aspect, in which egc and non-egc SAgs were fundamentally different: S. aureus strains harboring egc SAg genes secreted these toxins during early exponential growth, while strains with non-egc SAgs released them mainly in the late stationary phase. Similar observations have been reported on the RNA-level: The polycistronic egc-mRNA accumulated maximally during exponential growth (Refs. 6 ,23 , and unpublished data), while most non-egc SAg genes were transcribed in the postexponential growth phase.

Considering the extraordinary stability of SAg proteins, why did the egc-related T cell proliferation drop away so rapidly with increasing bacterial density? Presumably transcription of the egc-operon was shut down in the late exponential growth phase, so that the concentration of the egc SAgs did not further increase with increasing bacterial cell numbers in culture. This would lead to a rapid decline in mitogenic potency of bacterial supernatants on a per cell basis. We have no evidence for proteolytic degradation of the egc SAgs, not very surprisingly, because SAg proteins have been shown to be exquisitely stable.

Why the egc-operon is selectively transcribed during early exponential growth, is not understood. The regulation of staphylococcal SAgs is complex, and to date, besides the accessory gene regulator, the staphylococcal accessory regulator, the alternative σ factor, and the regulator of toxins also have been shown to play a role (39, 40, 41, 42, 43). However, this information has been obtained by investigation of S. aureus gene regulation in bacterial cell culture. The challenge will now be to elucidate, which of these processes are effective during the interactions of the microorganism with its host. There is first evidence that the regulatory circuits S. aureus employs during infection (and probably also colonization) differ from those characterized in vitro (44, 45). It remains to be seen, how this affects the release of egc and non-egc SAgs in vivo and the immune response against them.

At this stage, the high prevalence of neutralizing serum Abs against non-egc SAgs (18) allows the conclusion that most healthy adults have been exposed to these toxins during their encounters with S. aureus. For egc SAgs, it remains an open question whether they are 1) expressed in vivo, 2) in which quantities, and 3) under which conditions. In vitro data suggest that egc SAgs are generally produced in very small amounts (46). Because S. aureus secrets egc and non-egc SAgs in distinct functional states, they are likely released during different phases of its interaction with the human host. Colonization may not be the same as invasion. Consequently it is possible that the cell populations, which are exposed to egc- vs non-egc SAgs, differ, and that only in the case of non-egc SAgs they are able to orchestrate an efficient adaptive immune response.

In summary, egc and non-egc SAgs were very similar in all studied aspects of immune cell activation: gene regulation, cytokine secretion and induction of T cell proliferation, but their release from the bacteria was regulated in a fundamentally different manner.

We are grateful to Anne-Kathrin Ziebandt and Annette Dreisbach for their help with some of the experiments, and to Robert S. Jack for critical reading of the manuscript.

The authors have no financial conflict of interest.