Differential Interaction of the Staphylococcal Toxins Panton–Valentine Leukocidin and γ-Hemolysin CB with Human C5a Receptors

Staphylococcus aureus is one of the most common causes of bacterial infections in humans worldwide (1). S. aureus causes various diseases ranging from superficial skin and soft tissue infections to severe invasive diseases. The emergence of hospital-acquired and community-associated methicillin-resistant S. aureus (MRSA) strains has now become a global problem. As no new antibiotic agents are expected to be released in the near future (2), interest in the development of alternative therapeutics and vaccines has increased. Despite promising results in preclinical models, a recent vaccine candidate failed in clinical trials (3).

The pathogen S. aureus is well adapted to the human host. Many of the pathogen’s virulence factors show different specificities across mammalian species frequently used during preclinical in vivo studies (4). As a result, the contribution to pathophysiology of many of these virulence factors cannot be investigated in an integrated model for infection. More importantly, the potential of these virulence factors as vaccine or drug targets cannot be assessed accurately.

Phagocytes play a crucial role in the host defense against infections with S. aureus (4, 5). However, S. aureus has evolved multiple strategies to evade the human immune system. A key mechanism of S. aureus to repel attack by host phagocytes is the production of cytolytic toxins (6). Staphylococcal leukocidins are bicomponent pore-forming toxins that perforate the host cell plasma membrane (7). Based on chromatography elution profiles, the individual leukocidin subunits are designated S (slow) or F (fast) (8). Initial binding of the S-component to the surface of the target cell allows secondary binding of the F-component (9). This subsequently results in the assembly of lytic pore-forming hetero-octamers (10). The genome of human S. aureus isolates can encode up to five leukocidin toxins: Panton–Valentine leukocidin LukSF-PV (PVL) (11), γ-hemolysin AB and γ-hemolysin CB (HlgCB) (12, 13), LukED (14), and LukAB (also known as LukGH) (15, 16). The leukocidin protein components are closely related, and the amino acid sequence of the S-components of PVL (LukS-PV) and HlgC shows highest identity (81%) (17). Of the different staphylococcal leukocidins, the cytotoxic activity of PVL was the first to be described in detail (11). Although rare in methicillin-susceptible S. aureus isolates, the genes encoding PVL are overrepresented in epidemic community-associated MRSA strains (18). Although numerous epidemiological studies suggest a relation between PVL and severe invasive disease (19, 20), the role of PVL during infection is still not fully elucidated. The controversy regarding PVL is mostly caused by species-specific differences in susceptibility of phagocytes toward PVL (21). The HlgCB-encoding genes are present in almost all human S. aureus isolates (22, 23). Near-universal prevalence and consistent toxin expression by various S. aureus strains distinguish HlgCB from the other leukocidins (24). However, the contribution to pathophysiology is not well established because mouse neutrophils are resistant to HlgCB (24). The molecular basis for species specificity of both PVL and HlgCB in different animal models is incompletely understood.

For all leukocidins, recent identification of the myeloid host receptors revealed a specific molecular adaptation to the human host (18, 24–28). The S-components of PVL, HlgAB, HlgCB, and LukED, each target specific receptors belonging to the family of complement and chemokine receptors. This family of receptors shares a seven-transmembrane spanning architecture, with the N terminus and three extracellular loops (ECLs) exposed to the extracellular milieu (29). The involvement of specific host receptors in pore formation is a new concept, and molecular mechanisms of the multistep process of pore formation are only partially understood. PVL and HlgCB both target the complement component C5a receptors C5aR1 (C5aR, CD88) and C5aR2 (C5L2, GPR77) (24, 28). Sharing of receptors by the closely related leukocidins PVL and HlgCB suggests redundancy. However, experimental investigation of this apparent redundancy is seriously hindered by the lack of appropriate animal models for infection (7). The development of improved animal models for infection is urgently needed to address this issue.

In an attempt to better understand the process leading to pore formation by PVL and HlgCB, we investigated the interactions of both toxins with their receptors in more detail. By taking advantage of the species-specific interaction of the leukocidins with the receptors, we demonstrate that PVL and HlgCB differentially interact with C5aR1 and C5aR2. In addition, we illustrate that different parts of C5aR1 are involved in the initial binding of the S-components and the subsequent formation of lytic toxin pores. We show that toxicity of PVL, but not of HlgCB, can be neutralized by various C5aR1-antagonists. Together, these findings highlight the implications for our understanding of pore formation, the development and application of improved preclinical models for infection, and the development of drugs antagonizing leukocidin toxicity.

Human neutrophils were isolated after informed consent was obtained from all donors in accordance with the Declaration of Helsinki. For this, approval was obtained from the medical ethics committee of the University Medical Center Utrecht, the Netherlands. Experiments involving animals were reviewed and approved by the animal ethics committees of Lyon, France (CECCAPP, protocol number ENS_2012_033), or Utrecht, the Netherlands (protocol number DEC.2012.II.09.136).

S. aureus strains used to amplify coding sequences of toxins and bacterial proteins include strain ATCC 49775 for LukS-PV and LukF-PV (30), strain Cagnant for HlgC and HlgB (30), strain Newman for chemotaxis inhibiting protein of S. aureus (CHIPS) (31), and strain MW2 for FLIPr-Like (32). Coding sequences were cloned and expressed as described elsewhere (30–32). For the bicomponent pore-forming toxins, single protein-coding sequences omitting the predicted signal peptide were cloned and transformed into Escherichia coli M15 or BL21 pLys. Polyhistidine-tagged proteins were isolated from supernatant of E. coli cell lysates on Ni-nitrilotriacetic acid columns (QIAGEN).

Human neutrophils, obtained from healthy volunteers, macaque (Macaca fascicularis), and rabbit (New Zealand White) neutrophils were isolated by Ficoll/Histopaque centrifugation (31). Neutrophils from cows (Holstein Frisian) were isolated using Percoll (1.09176 g/l) centrifugation. Mouse (C57BL/6) leukocytes were obtained from bone marrow, and immune cells were collected as described (28). Unless specified otherwise, all in vitro experiments with cells were performed using RPMI 1640 (Invitrogen) supplemented with 0.05% human serum albumin (Sanquin), with cell concentrations adjusted to 5 × 10⁶ cell/ml.

HEK293T cells [a human embryonic kidney (HEK) cell line] were used for overexpression of wild-type and chimeric receptors. The genes encoding the C5a receptors of all mammalian species investigated reside on a single exon. Amplifications were performed on QUICK-Clone cDNA of human bone marrow or mouse liver (BD Biosciences Clontech), macaque, rabbit, and cow peripheral blood using PfuTurbo DNA polymerase (Stratagene). N terminus and ECL exchange chimeric receptors were constructed using overlap extension PCR, as described elsewhere (33). The receptor sequences were amplified and fused together in multiple PCR steps. The 5′- and 3′- terminal primers contained, respectively, EcoRI and NotI restriction sites, allowing, respectively, for specific ligation into the multiple cloning sites of the pcDNA3.1 vector (Invitrogen) (28) or the bicistronic expression plasmid pIRESPuro3 (Clontech) (24). To confirm surface expression of the receptors on transiently transfected cells during binding studies of LukS-PV, an N-terminal FLAG-tag (DYKDDDDK) was placed after the first methionine. An additional methionine was placed directly after the FLAG-tag to keep the N-terminal sequence intact. Expression vectors, primer designs, restriction sites, accession numbers, and primer combinations used in this study are listed in Supplemental Table I (wild-type full-length receptors), Supplemental Table II (primer designs used for chimeric receptors), and Supplemental Table III (primer combinations used for C5aR1 or chimeric receptors). Cloning of rabbit C5aR2 resulted in an amino acid sequence slightly divergent from the annotated sequence. All wild-type amino acid sequences used in this study are aligned in Fig. 2. HEK293T cells stably transfected with a plasmid encoding human C5aR2 were donated by Peter N. Monk (Sheffield University Medical School, Sheffield, U.K.) (34).

Cells were either transiently or stably transfected, as described elsewhere (24, 28). Briefly, cells were transfected with 4 μg DNA and 5 μl Lipofectamine 2000 (Life Technologies), according to the manufacturer’s protocol. For transient transfections, cells were harvested after 24 h with 0.05% trypsin/0.53 mM EDTA. For stable transfections, selection pressure was applied using 1 μg/ml puromycin in DMEM with 10% FCS, and cells were subcloned after several weeks.

The expression of receptors was confirmed using 10 μg/ml mAb (PE-conjugated mouse anti-human C5aR1 clone S5/1, 1:50, BioLegend; FITC-conjugated rat anti-mouse C5aR1 clone 20/70, 1:50, BioLegend; PE-conjugated mouse anti-human C5aR2 clone 1D9-M12, 1:50, BioLegend) or 240 nM FITC-conjugated C5a (Bachem) where possible. Specific susceptibility toward PVL and/or HlgCB toxicity of cells stably transfected with plasmids encoding wild-type C5a receptors from different mammalian species was interpreted as a confirmation of receptor expression. In the case of resistance of cells toward both PVL and HlgCB toxicity, stable expression of wild-type C5a receptors was confirmed (Supplemental Fig. 1A, 1B). Owing to inability to bind mAbs or C5a, expression of chimera Mouse C5aR1 Human ECL1+2+3 and chimera Mouse C5aR2 Human ECL1+3 on stably transfected cells could not be verified. The expression of the remaining chimeric receptors was verified similarly (Supplemental Fig. 1C–P).

Cells were exposed to recombinant toxins in a volume of 50 μl for 30 min at 37°C. Human, macaque, rabbit, and cow neutrophils, and cell lines, were subsequently analyzed by flow cytometry and gated for forward and side scatters. At least 10,000 cells per sample were analyzed. For graphical presentation, pore formation was defined as the percentage of cells positive for intracellular staining by DAPI.

After treatment with FcγR block, mouse leukocytes were stained with mAbs directed toward CD11b (clone M1/70, 1:400; BD), Ly6G (clone 1A8, 1:100; BD), and Ly6C (clone AL-21, 1:200; BD). After exposure to toxins, mouse cells were incubated on ice with the fixable viability dye eFluor 780 (1:1000, eBiosciences). Cell viability of mouse neutrophils was analyzed by flow cytometry in the presence of Flow-Count Fluorospheres (BD) and defined as eFluor 780–negative cells relative to buffer-treated cells.

For C5aR1-competition experiments, human neutrophils were preincubated for 10 min at room temperature with CHIPS (710 nM), mouse anti-human C5aR1 mAb (clone S5/1, AbD SeroTec; 67 nM), NDT 9513727 (Tocris, R&D Systems; 10 μM), W-54011 (Calbiochem, Merck Millipore; 10 μM), PMX 205 (Tocris, R&D Systems; 10 μM), or AS-65122 (AnaSpec; 10 μM). Concentrations of C5aR1 antagonists used are based on efficacy titration for PVL and HlgCB.

Analysis of pore formation in transiently transfected HEK293T cells was corrected for the fraction of receptor-positive cells. Transiently transfected or stably transfected cells were reciprocally compared in functional experiments.

Equimolar concentrations of toxin protein components were used.

For binding experiments with LukS-PV, cells transiently transfected with plasmids encoding receptors with an N-terminal FLAG tag (28) were prelabeled with a mouse anti-FLAG mAb (clone M2, 1:500; Sigma-Aldrich), followed by PE-labeled goat–anti-mouse Ab (1:80, Dako). For experiments with HlgC, stably transfected cells were directly used for subsequent binding. After incubation with the protein on ice for 30 min in a volume of 50 μl, cells were washed and incubated with an FITC-conjugated mouse anti-his mAb (1:80; LifeSpan Biosciences). Binding was detected using flow cytometry, and ≥10,000 cells were analyzed. Analysis of binding to transiently transfected cells was limited to receptor-positive cells. Binding of the S-components was expressed as the percentage of mean fluorescence in relation to saturated binding of the S-component to human C5aR1 at 31 nM.

Freshly isolated macaque and human neutrophils were incubated in a total volume of 50 μl at 5 × 10⁶ cells/ml on ice with 10 μg/ml mouse anti-human chemokine receptor mAbs for C5aR1 (clone S5/1; AbD SeroTec) or C5aR2 (clone 1D9-M12; BioLegend), followed by PE-conjugated goat–anti-mouse Ab (1:80; Dako). Samples were analyzed using flow cytometry. At least 10,000 cells were analyzed, and cells were gated for forward and side scatters. Ab binding was quantified by calibration to defined Ab binding capacity units using QIFIKIT (Dako) and corrected for isotype controls.

Flow cytometric data were analyzed with FlowJo (TreeStar Software). Statistical analyses were performed using Prism (GraphPad Software).

Some leukocidins exert cytotoxic activity in a species-specific manner (21, 24, 26, 28). For PVL, species-specific toxicity is well described (21). Although PVL and HlgCB have shared receptors, species-specific susceptibility of primary cells toward HlgCB is not completely understood. To compare species specificity of HlgCB with that of PVL, we tested toxin-mediated cell permeability in freshly isolated neutrophils from different mammalian species, using a membrane-impermeant fluorescent dye. Cells from mouse and rabbit were investigated, as both mammals are frequently used in preclinical experimental models of S. aureus infections (21). In addition, macaque neutrophils were investigated because this primate is closely related to humans. Finally, cow neutrophils were included, as this species is a natural host to S. aureus (35).

As expected, PVL was active only toward human and rabbit neutrophils, but not toward neutrophils from macaque, cow, and mouse (Fig. 1A). However, HlgCB was active toward neutrophils from all species tested, except for mice (Fig. 1A).

These data identify that the rabbit is the only compatible species for establishing the contribution toward staphylococcal pathophysiology of both PVL and HlgCB in vivo. Furthermore, the divergent susceptibility of neutrophils from different mammalian species toward PVL and HlgCB suggests a differential interaction of both toxins with their receptors.

To further investigate the molecular interaction of PVL and HlgCB with neutrophils from different mammalian species, we screened HEK cells overexpressing the respective orthologs of human C5aR1 and C5aR2 for their susceptibility to these two toxins.

In line with susceptibility of neutrophils, only cells expressing human or rabbit C5aR1 were susceptible toward PVL. As illustrated in Fig. 1B, cells expressing macaque, cow, or mouse C5aR1 were fully resistant toward pore formation induced by PVL. In accordance with the results obtained with primary cells, cells expressing different mammalian C5a receptors, with the exception of mouse C5aR1, were susceptible toward HlgCB-induced toxicity (Fig. 1B). PVL was cytotoxic toward cells expressing human or rabbit C5aR2, whereas mouse and cow C5aR2 were incompatible with PVL (Fig. 1C). In contrast to the resistance of neutrophils, HEK cells expressing macaque C5aR2 were susceptible to pore formation induced by PVL (Fig. 1C). Just as in humans (28), expression of C5aR2 on macaque neutrophils is very low compared with expression of C5aR1 (Supplemental Fig. 1Q). Similarly to C5aR1, screening of C5aR2 with HlgCB showed broad species compatibility, again with the exception of mouse C5aR2 (Fig. 1C). Cells transfected with plasmids encoding PVL- and HlgCB-incompatible receptors expressed equal levels of receptors compared with cells transfected with plasmids encoding PVL- or HlgCB-compatible receptors (Supplemental Fig. 1A, 1B).

Jointly, these analyses demonstrate that the mouse is a poor model for investigating both PVL and HlgCB as a virulence factor owing to a molecular incompatibility of the toxins with its respective receptors. Furthermore, these data show that although PVL and HlgCB target the complement receptors C5aR1 and C5aR2, both toxins do so in a species-dependent differential manner.

Although C5aR1 and C5aR2 are relatively conserved among different mammalian species, amino acid sequence identities vary from 60 to 93% when compared with the human receptors (Fig. 2A, 2B). In an attempt to define the regions of the receptors responsible for the observed species-specific toxin–receptor interaction, we tested cells expressing chimeric C5a receptors for susceptibility toward PVL and HlgCB. Because most amino acid sequence divergence between the C5a receptors of the different mammalian species is clustered in the N termini and ECLs (Fig. 2C, 2D), we focused on the extracellular domains of the respective receptors.

Although macaque C5aR1 is most closely related to human C5aR1, this receptor is incompatible with PVL (Figs. 1B, 2C). Therefore, we used macaque and human C5aR1 as chimeric templates to investigate the importance of the different extracellular receptor domains for interaction with PVL on receptor-expressing HEK cells. Receptor expression was similar for all constructs tested (Supplemental Fig. 1C–L). For PVL, exchange of the N terminus of human C5aR1 with macaque or mouse N-terminal C5aR1 did not result in loss of susceptibility (Fig. 3A). Exchange of the individual human ECL2 or ECL3 with those of macaque resulted in loss of susceptibility to PVL; however, replacement of ECL1 showed no effect (Fig. 3B). Because the amino acid sequence of human and macaque ECL1 is nearly identical (Fig. 2C), we additionally tested a chimeric human C5aR1 with mouse ECL1. Cells expressing this receptor were resistant to pore formation induced by PVL, just as were cells expressing a human C5aR1 chimera with mouse ECL2 or ECL3 (Fig. 3C). Of interest, although replacement of human ECL2 with macaque ECL2 is sufficient to prevent PVL toxicity, the sequences of ECL2 differ in only a few amino acids (Fig. 2C). Thus, we introduced single amino acid mutations in the ECL2 of human C5aR1 following the macaque sequence. Cells expressing a P183S mutant human C5aR1 were resistant to pore formation induced by PVL (Fig. 3D), highlighting the critical role of C5aR1-ECL2 in human-specific cytotoxicity of PVL. In addition, mutation of the human amino acid cassette sequence SHDK to the respective macaque sequence NNDT resulted in loss of susceptibility toward PVL. Introduction of the individual human ECL2 or ECL3 in macaque C5aR1 did not restore susceptibility of cells toward PVL (Fig. 3E). Cells expressing a macaque C5aR1 chimera with both human ECL2 and ECL3 showed a partially restored susceptibility toward PVL, just as did cells expressing a mouse C5aR1 chimera with human ECL1, 2, and 3 (Fig. 3F). This observation suggests that, in addition to the extracellular domains, other parts of the receptor are involved in the interaction with PVL as well.

Mouse C5aR1 is the most divergent receptor from the panel of mammalian C5a receptors investigated, and it is the only HlgCB-incompatible C5aR1 (Figs. 1B, 2C). Thus, we used human and mouse C5aR1 as chimeric templates to study the interaction between HlgCB and C5aR1. Similar to PVL, exchange of the N terminus of human C5aR1 with mouse N-terminal C5aR1 did not result in loss of susceptibility of cells toward HlgCB (Fig. 4A). In contrast to PVL, introduction of mouse ECL2 in human C5aR1 did not affect susceptibility of cells toward HlgCB (Fig. 4B). Introduction of mouse ECL1 or ECL3 in human C5aR1 resulted in partial and complete loss of cell susceptibility toward HlgCB-induced pore formation, respectively (Fig. 4B). ECL3 of mouse C5aR1 is highly divergent from the human amino acid sequence (Fig. 2C). Introduction of single amino acid mutations in the ECL3 of human C5aR1 following the mouse sequence did not completely abolish susceptibility of cells toward HlgCB (Fig. 4C), indicating that the interaction between HlgCB and human C5aR1 involves a combination of multiple amino acid residues in ECL3. As expected based on the studies with loss of function chimeric receptors, introduction of human ECL1 in mouse C5aR1 was insufficient to result in susceptibility of cells toward HlgCB (Fig. 4D). However, introduction of human ECL3 in mouse C5aR1 resulted in a substantial restoration of susceptibility toward HlgCB (Fig. 4D). Introduction of both human ECL1 and ECL3 did not further enhance susceptibility of cells toward HlgCB. These data indicate that specificity of HlgCB for human C5aR1 is mostly defined by ECL3 and ECL1.

Together, these findings demonstrate that PVL and HlgCB differentially interact with C5aR1 in a complex manner. Results with chimeric receptors suggest that both PVL and HlgCB interact with multiple extracellular domains of C5aR1. Whereas human C5aR1 specificity of PVL is critically driven by ECL2, human C5aR1 specificity of HlgCB is mostly determined by ECL3 and ECL1. Incomplete functional restoration of macaque and mouse C5aR1 chimeric receptors toward PVL and HlgCB indicate that, in addition to the extracellular domains, other domains of the receptor are likely involved in the interaction with both toxins as well.

In addition to C5aR1, we tested cells expressing chimeric C5aR2 receptors for susceptibility toward PVL and HlgCB. Mouse C5aR2 is the most divergent receptor from the mammalian C5aR2 receptors tested, and the only PVL- and HlgCB-incompatible C5aR2 (Fig. 1C). The N terminus of mouse C5aR2 is most divergent compared with the amino acid sequence of other mammalian species (Fig. 2D). Thus, we used human and mouse C5aR2 as chimeric templates to study the interaction of PVL and HlgCB with receptor-expressing HEK cells. Receptor expression was similar for all constructs tested (Supplemental Fig. 1L–P).

Exchange of the N terminus of human C5aR2 with the N terminus of mouse C5aR2 resulted in a complete and partial loss of susceptibility for PVL and HlgCB, respectively (Fig. 5A, 5B). Although introduction of mouse ECL1 and ECL3 into human C5aR2 caused a complete loss of susceptibility of cells toward PVL (Fig. 5C), none of the mouse ECLs substantially affected susceptibility toward HlgCB toxicity (Fig. 5D). Introduction of mouse ECL2 into human C5aR2 only moderately affected susceptibility of cells toward both PVL (Fig. 5C). Introduction of individual human ECLs, or all three human ECLs together, into mouse C5aR2 was insufficient to restore susceptibility toward PVL and HlgCB (Fig. 5E, 5F). Additional introduction of the human N terminus did not result in PVL and HlgCB susceptibility either (Fig. 5E, 5F).

These findings indicate that PVL and HlgCB interact with human C5aR2 in a different manner. As the introduction of all human extracellular domains in mouse C5aR2 did not result in susceptibility of cells toward both toxins, other parts of the receptor are clearly also involved in the interaction with the toxins.

Recently, it was identified that staphylococcal leukocidins act via cell surface receptors (24–26, 28). Although expression of the specific receptor is a prerequisite for the bicomponent toxins to form pores in the cell membrane, the role of the receptors during the process of binding, oligomerization, and pore formation itself has not been fully elucidated. This fact led us to investigate, in addition to pore formation, the binding of LukS-PV and HlgCB (HlgC) to an exemplary selection of HEK cells overexpressing chimeric C5a receptors.

Fig. 6A confirms that LukS-PV cannot bind to HEK cells expressing macaque C5aR1, in contrast to human C5aR1 (28). In line with the loss of pore-forming capacity of PVL (Fig. 6B), introduction of macaque ECL2 in human C5aR1 resulted in a complete loss of detectable binding of LukS-PV (Fig. 6A). Although a replacement in macaque C5aR1 with human ECL2 did not restore the pore-forming ability of PVL (Fig. 6B), binding of LukS-PV to cells expressing this chimeric receptor was similar when compared with wild-type human C5aR1 (Fig. 6A). Similarly, HlgC could not bind to HEK cells expressing mouse C5aR1 (Fig. 6C). Partial restoration of susceptibility toward HlgCB cytotoxicity in cells expressing a mouse C5aR1 with human ECL3 (Fig. 6D) was associated with suboptimal binding of HlgC when compared with cells expressing wild-type human C5aR1 (Fig. 6C). Although introduction of mouse ECL3 in human C5aR1 resulted in a nearly complete loss of susceptibility toward HlgCB toxicity (Fig. 6D), HlgC binding was still observed (Fig. 6C).

These data show, as an example, that binding of the S-component to the specific receptor itself is not necessarily sufficient to allow subsequent pore formation.

C5aR1 is considered the major host receptor for PVL, and C5aR1 competition interferes with PVL toxicity on human neutrophils (28). Because PVL and HlgCB target the human C5a receptors in a specific but divergent manner, we investigated whether toxicity of PVL and HlgCB toward human neutrophils can be equally neutralized by different C5aR1 antagonists in vitro.

As observed before, pretreatment of neutrophils with CHIPS (31), which binds human C5aR1, significantly shifted the EC₅₀ of PVL (Fig. 7A, 7D). However, CHIPS could not neutralize toxicity of HlgCB (Fig 7A, 7D). The staphylococcal protein FLIPr-Like, targeting two unrelated receptors, did not confer protection to neutrophils for either of the toxins. Pretreatment of neutrophils with a mAb, directed toward the N terminus of human C5aR1 (clone S5/1), resulted in a minimal shift of the EC₅₀ for PVL, but, again, not for HlgCB (Fig 7B, 7D). In the next step, we tested the potency of different commercially available C5aR1 antagonists to neutralize PVL and HlgCB toxicity. NDT 9513727 and W-54011 are small molecules with potent C5aR1 antagonistic properties (36, 37). PMX 205 and AS-65122 are C5a-mimetic peptides blocking the binding of C5a (38, 39). Although pretreatment of human neutrophils with all compounds resulted in 5- to 10-fold increases in EC₅₀ values for PVL, none of the compounds reduced susceptibility of cells toward HlgCB (Fig 7C, 7D). Remarkably and in contrast to the other C5aR1-antagonists tested, both small molecules NDT 9513727 and W-54011 significantly enhanced susceptibility to HlgCB, and pretreatment of neutrophils with these molecules resulted in 5-fold decreased EC₅₀ values for HlgCB (Fig. 7C, 7D).

Together, these data show that although PVL and HlgCB both target the human C5a receptors, only toxicity of PVL can be antagonized in vitro by C5aR1 competition on human neutrophils. This observation provides further evidence that both toxins differentially interact with human C5aR1.

Susceptibility of host cells to staphylococcal leukocidins is determined by surface expression of specific receptors (24–28). Both PVL and HlgCB interact with their shared receptors, C5aR1 and C5aR2, in a host-adapted manner (24, 28). C5aR1 is highly expressed on phagocytic cells. Compared with C5aR1, expression of C5aR2 on the surface of neutrophils is very low (28, 40, 41). Although macaque C5aR2 is compatible with PVL, macaque neutrophils are resistant to the toxin. Incompatibility of macaque C5aR1 with PVL apparently drives resistance of macaque neutrophils toward the toxin. For this reason, the abundantly expressed C5aR1 can be considered the major host receptor for PVL, and presumably for HlgCB as well.

Although PVL and HlgCB both target C5aR1 and C5aR2, the toxins interact with these receptors in a divergent manner. Although multiple receptor domains are involved in the interaction with the toxins, human C5aR1 specificity is mostly defined by ECL2 for PVL, and ECL3 and ECL1 for HlgCB. Although HlgCB tolerates significant variations in the C5a receptors, receptor compatibility of PVL is more restricted. Tolerance of receptor variations by HlgCB is illustrated by a substantial restoration of cytotoxic susceptibility of mouse C5aR1 chimera with human ECL3, which still harbors significant amino acid sequence divergence in the other extracellular domains when compared with human C5aR1. In contrast to this observation, restricted receptor compatibility of PVL is reflected by the incompatibility of PVL with macaque C5aR1, which has high amino acid sequence identity with human C5aR1. S. aureus is well adapted to the human host, and high-affinity protein–protein interactions are human specific for many of its virulence factors (4). Whereas HlgCB is core-genome encoded (22, 23), PVL is encoded on a mobile genetic element (42). It is possible that PVL is a result of the adaptation of S. aureus to the human host.

For PVL and HlgCB, multiple extracellular domains of both C5aR1 and C5aR2 are involved during the formation of lytic pores. However, different parts of the receptor mediate initial S-component binding and subsequent pore formation of the leukocidin, as illustrated by LukS-PV and HlgC for C5aR1. This finding suggests the occurrence of conformational changes of receptors and toxins during the process of binding, hetero-oligomerization, and pore formation. Introduction of human extracellular domains into macaque and mouse C5aR1 only partially restored susceptibility of cells toward both PVL- and HlgCB toxicity. For C5aR2, no gain of function could be achieved in a similar approach. These findings imply that interactions of PVL and HlgCB with C5aR1 and C5aR2 not only are mediated by the extracellular domains of the receptors but involve other parts of the receptors as well. The ligand of the C5a receptors, the complement component C5a, is known to interact with the receptor N termini and ECLs, but also with binding pockets formed by hydrophobic residues of the receptor transmembrane regions (29). The multifaceted sequence of pore formation by PVL and HlgCB potentially involves conformational changes of toxin and receptor, which are mediated by the transmembrane regions of the receptors. Whether the interaction between receptor and toxin is limited to the S-component only or directly involves the F-component as well remains to be determined.

The contribution of PVL and HlgCB to staphylococcal pathophysiology is not completely understood. Our data underscore previous observations that the mouse is a suboptimal mammalian species to investigate PVL and HlgCB in vivo (21, 24). Use of rabbits or transgenic mice offers opportunities for future studies to explore the relative contribution to pathophysiology of both toxins. Despite many attempts over decades, a safe and effective vaccine against S. aureus is still not available. Again, use of appropriate animal models for staphylococcal infections is essential for future vaccine development. Owing to the increasing resistance of S. aureus to traditional antimicrobial agents, innovative therapeutics are urgently needed. One approach is the development of toxin-neutralizing mAbs (43). The complement and chemokine receptors, which are highly expressed on phagocytic cells and targeted by the staphylococcal leukocidins, are, potentially, host candidate drug targets (44, 45). The divergent interaction of PVL and HlgCB with the C5a receptors is illustrated by the differential competitive potency of various C5aR1 antagonists on human neutrophils. Although several C5aR1-binding molecules reduced the toxicity of PVL, none of the molecules tested had a protective effect on HlgCB cytotoxicity. Although CHIPS antagonizes the proinflammatory responses of neutrophils induced by sublytic concentrations of both PVL and HlgCB (24, 28), at lytic concentrations, CHIPS could only antagonize PVL but not HlgCB. The molar ratios between HlgCB and CHIPS likely determine the competitive potency of the antagonist. Intriguingly, two small-molecule C5aR1 antagonists substantially potentiated susceptibility of neutrophils toward HlgCB. In line with this finding, others have recently observed an amplification of calcium mobilization induced by sublytic concentrations of HlgCB in the presence of one of these antagonists (46). Likely, distinctive interference of the various C5aR1 antagonists with the conformational state of the receptor affects the susceptibility of cells toward HlgCB (47). Our data provide a proof of concept for the protective potential of small C5aR1-antagonizing molecules against PVL toxicity. Future additional screening of other compounds is needed to identify molecules antagonizing cytotoxicity of both PVL and HlgCB. From there, the efficacy of molecules blocking the toxin–receptor interaction could be investigated in improved animal models for infection.

We thank Marie Daoud El-Baba (IUT Lyon 1, Lyon, France), Fabrice Taborik (Cynbiose, Marcy l’Etoile, France), and Manouk Vrieling (UMC Utrecht, Utrecht, the Netherlands) for providing rabbit, macaque, and bovine blood, and Kaila Bennett and Ron Gorham (University Medical Center Utrecht) for critically reviewing the manuscript.

The authors have no financial conflicts of interest.