Staphylococcus aureus iron-regulated surface determinant B (IsdB) protein interacts with von Willebrand factor and promotes adherence to endothelial cells

Staphylococcus aureus is the cause of a spectrum of diseases in humans and animals. The molecular basis of this pathogenicity lies in the expression of a variety of virulence factors, including proteins that mediate adherence to the host plasma and extracellular matrix proteins. In this study, we discovered that the iron-regulated surface determinant B (IsdB) protein, besides being involved in iron transport and vitronectin binding, interacts with von Willebrand Factor (vWF). IsdB-expressing bacteria bound to both soluble and immobilized vWF. The binding of recombinant IsdB to vWF was blocked by heparin and reduced at high ionic strength. Furthermore, treatment with ristocetin, an allosteric agent that promotes the exposure of the A1 domain of vWF, potentiates the binding of IsdB to vWF. Both near-iron transporter motifs NEAT1 and NEAT2 of IsdB individually bound recombinant A1 domain with KD values in the micromolar range. The binding of IsdB and adhesion of S. aureus expressing IsdB to monolayers of activated endothelial cells was significantly inhibited by a monoclonal antibody against the A1 domain and by IsdB reactive IgG from patients with staphylococcal endocarditis. This suggests the importance of IsdB in adherence of S. aureus to the endothelium colonization and as potential therapeutic target.


Results
Identification of a novel S. aureus vWF-binding protein expressed in iron-limited conditions. A property of S. aureus, shared with other Gram-positive bacteria, is the ability to use several different CWA proteins to interact with a specific host component. As an example of this redundancy, S. aureus can adhere to fibrinogen via clumping factors, ClfA 27 and ClfB 28 , and fibronectin-binding proteins, FnBPA 29 and FnBPB 30 . To date, the only CWA protein known to interact with vWF is SpA 7 . In the search for new S. aureus surface component(s) potentially involved in vWF binding, SH1000 spa deletion mutant cells were grown either in iron-rich (BHI) or iron-poor (RPMI) medium. They were digested with lysostaphin, and the released material was subjected to SDS-PAGE under reducing conditions and analyzed for vWF-binding by far Western blotting. A 75 kDa vWF-binding protein was detected in material released from cells grown in RPMI (Fig. 1A, lane 1), whereas no significant signal was detected in proteins originating from cells grown in BHI (Fig. 1A, lane 2) suggesting that binding depends on a protein that is induced by iron starvation. No detection of 75 kDa protein was observed with the anti-vWF IgG alone (data not shown). Additional low molecular weight molecules (25)(26)(27)(28)(29)(30) were observed in the material released from bacteria grown in both media.
In iron starvation conditions, S. aureus expresses several iron-regulated surface proteins. Among these, IsdB with a molecular weight of 75 kDa is a potential candidate receptor for vWF. To validate this hypothesis, lysostaphin-solubilized cell-wall fractions of S. aureus SH1000 spa cells were subjected to SDS-PAGE and immunoblotting and probed with an IsdB antibody (Fig. 1B). A major signal of 75 kDa molecule was detected in the proteins released from cells grown in RPMI, indicating that IsdB is the possible vWF-binding protein (Fig. 1B, lane 1). No signal was observed in the material from cells grown in BHI (Fig. 1B, lane 2).
To confirm this, purified recombinant ligand-binding domains of several S. aureus CWA proteins were screened for the ability to bind vWF by an ELISA-type assay. In particular, vWF showed a binding activity for IsdB NEAT1-NEAT2 comparable to that already reported for SpA, whereas no binding to the N-terminal domains of IsdB-mediated adhesion of bacteria to immobilized vWF. To evaluate the role of IsdB in bacterial adhesion to vWF, we compared the ability of SH1000 WT and its isdB mutant to bind immobilized vWF in an ELISA-type assay. A significantly higher attachment of the WT strain compared to the isdB mutant was observed, indicating the important role of IsdB in bacterial adhesion to vWF (Fig. 3A). The ability of IsdB to promote adhesion to vWF was also examined using the surrogate host strain L. lactis expressing IsdB from a gene cloned into the plasmid vector pNZ8037 (L. lactis pNZ8037::isdB ). When lactococci were tested for adhesion to surface-coated vWF, significantly higher adhesion of L. lactis pNZ8037::isdB was observed compared to that of L. lactis harboring the empty vector (Fig. 3B).

Interference of vWF binding ligands on the interaction of IsdB with vWF.
A competitive inhibition assay was developed to investigate the effect of heparin and other specific vWF-ligands (collagen type I, type III, type IV, and type VI) 31,32 on the binding of IsdB to vWF. Heparin and collagen type IV and VI, each of which interacts with the A1 domain, reduced IsdB binding to vWF by 80% and 50%, respectively. In contrast, Lysostaphin-released material from S. aureus SH1000 spa grown in RPMI (lane 1) and BHI (lane 2) was subjected to Far Western Blotting. The nitrocellulose membrane was probed with human vWF followed by polyclonal rabbit vWF antibody and HRPconjugated goat anti-rabbit IgG. (B) Lysostaphin-released material from S. aureus SH1000 spa grown in RPMI and BHI was subjected to Western Immunoblotting. The nitrocellulose membrane was probed with a rabbit anti-IsdB IgG and HRP-conjugated goat anti-rabbit IgG. Molecular mass standards are indicated on the left of the panels. (C) Binding of vWF to surface proteins from S. aureus. Microtiter wells were coated with purified recombinant A regions of the indicated CWA proteins of S. aureus and incubated with human vWF. Ligand bound to the wells was detected with a polyclonal rabbit vWF antibody followed by HRP-conjugated goat antirabbit IgG. The data points are the means ± SD of two independent experiments. In the inset, binding of the vWF to purified recombinant IsdB NEAT1-NEAT2 protein in a Western blotting assay is shown. The full-length blots are presented in Supplementary Fig. S3

Localization of binding sites in IsdB and vWF and affinity studies.
To directly show that the A1 domain of vWF carries the IsdB-binding region, the isolated A1 domain of vWF was expressed in E. coli and tested by ELISA-type binding assays (Fig. 5). First, the interaction between immobilized vWF or the A1 domain and IsdB NEAT1-NEAT2 was examined. IsdB bound to the A1 domain in a saturable and dose-dependent manner and the binding profile appeared to be very similar to that obtained for the binding of IsdB to the fulllength vWF (Fig. 5A). This implies that there is a single binding site for IsdB in vWF. To map the vWF-binding region(s) within the IsdB protein, the binding of the A1 domain to immobilized IsdB NEAT1-NEAT2 and the IsdB NEAT1 or IsdB NEAT2 modules was assessed. Our data indicate that the A1 domain bound dose-dependently and saturably to each recombinant module and with a binding pattern resembling that exhibited by IsdB NEAT1-NEAT2 (Fig. 5B).
To determine the affinity of the A1 domain for IsdB proteins by surface plasmon resonance (SPR), IsdB NEAT1-NEAT2 and the single NEAT domains were immobilized on an NTA-Ni +2 sensor chip and incubated with increasing amounts of A1 domains (from 0.08 to 1.25 μM) in a single-cycle operation mode. The sensorgrams shown in Fig. 6 revealed a complex binding system and were best fitted within the framework of a 1:1 stoichiometric heterogeneous ligand binding model 35,36 , whereby a single molecule of the A1 domain binds to a single site on IsdB proteins which, after noncovalent capture on the NTA-Ni +2 sensor chip, can assume different orientations on the chip surface. From this analysis, dissociation constant (K D ) values in the low micromolar range for the A1/IsdB protein complexes were obtained (see the legend to Fig. 6). However, the A1 domain showed a higher affinity for the NEAT2 domain than that observed for the IsdB NEAT1-NEAT2 or NEAT1 domain. Comparative S. aureus SH1000 WT and its spa or isdB mutant were grown in RPMI to stationary phase. The staphylococcal cells were incubated with purified human vWF (5 µg/ ml) and bacteria-bound protein was released by extraction buffer and subjected to SDS-PAGE and Western Immunoblotting. 2 µg of purified vWF was loaded as control. The membrane was probed with a polyclonal rabbit vWF antibody followed by HRP-conjugated goat anti-rabbit IgG. Molecular masses of standard proteins are indicated on the left. The full-length blot is presented in Supplementary Fig. S3. (B) Densitometric analysis of vWF bound to S. aureus WT and its isogenic spa or isdB mutant. Statistically significant differences are indicated (*, P < 0.05; **, P < 0.01). The reported data are the mean values ± SD from three independent experiments. aureus SH1000 WT and its isogenic isdB mutant to immobilized vWF. Microtiter wells coated with vWF were incubated with cells of S. aureus SH1000 WT and its isdB mutant obtained from cultures grown in RPMI. Bacteria bound to vWF were detected by the addition of HRP-conjugated rabbit anti-mouse IgG to the wells. Statistically significant difference is indicated (**, P < 0.01). (B) Adhesion of L. lactis expressing IsdB to immobilized vWF. Microtiter wells coated with vWF were incubated with L. lactis ectopically expressing IsdB (L. lactis pNZ8037::isdB ) or L. lactis carrying the empty vector (L. lactis pNZ8037 ). Bacteria bound to vWF were detected incubating the wells with a polyclonal rabbit anti-L. lactis IgG and HRP-conjugated goat anti-rabbit IgG. Statistically significant difference is indicated (*, P < 0.05). The data points are the means ± SD from three independent experiments, each performed in triplicate. www.nature.com/scientificreports/ analysis of the affinities of IsdB NEAT1-NEAT2 and NEAT1 and NEAT2 revealed a non-additive behavior of the isolated domains in binding to the A1 domain. More specifically, the affinity of IsdB NEAT1-NEAT2 for A1 is much lower than the sum of the affinities of isolated NEAT1 and NEAT2. This finding is suggestive of a different binding mechanism that the two NEAT domains likely exploit for interacting with A1 in the isolated form or when they are embedded in the single-chain IsdB NEAT1-NEAT2 protein.

Electrostatic properties of IsdB and A1 domain.
To determine whether ionic forces play a role in the interaction of IsdB with vWF, the effect of increasing NaCl concentrations on IsdB binding was assessed. The addition of salt significantly reduced the binding of IsdB to immobilized vWF. At 0.5 M NaCl IsdB binding was reduced by > 70% (Fig. S2A). These results provide indirect evidence that charge-charge interactions play a role in IsdB-vWF complex formation. Electrostatic potential calculations ( Supplementary Fig. S2B, panels a-d), obtained from the crystallographic structure of IsdB deduced from the IsdB/Hb complex 19 , indicate that the linker region is highly electropositive whereas the NEAT domains display an asymmetric distribution of charges. This is more pronounced in NEAT2, where a predominantly negative face could be identified. On the other hand, NEAT1 is mainly electropositive, with some interspersed negative spots encompassing the β-strand 153-156 and α-helix 163-169.
The A1 domain has a cuboid shape, with a central hydrophobic parallel, eight-stranded parallel β-sheet flanked by three α-helices on each side of the β-sheet 37 . Electrostatic potential calculations carried out on the A1 domain ( Supplementary Fig. S2B, panels e-h) identified two distinct faces of opposite charges, i.e. a strongly electropositive face covering helices α4 and α5 and an electronegative face encompassing helices α1 and α7. Thus, the highly charged nature of both IsdB and A1 suggests that ionic forces may play an important role in macromolecular complex formation. The steric and electrostatic complementarity of the IsdB and A1 structures ( Supplementary Fig. S2B) suggest that the highly electropositive face of the globular A1 domain can preferentially couple with the electronegative surfaces in the concave dumbbell structure of IsdB NEAT1-NEAT2, as observed in the X-ray structure of the IsdB-hemoglobin complex 19 .

IsdB binding to vWF mediates adhesion of S. aureus to endothelial cells. Endothelial cells (ECs)
and megakaryocytes are the only cells that synthesize vWF. Thus, we asked whether S. aureus expressing IsdB can adhere to vWF associated with the ECM of ECs. Increasing amounts of IsdB NEAT1-NEAT2 were incubated with monolayers of human umbilical vein endothelial cells (HUVEC) pre-treated with the calcium ionophore A23187, a compound that induces the fusion of Weibel-Palade bodies with the plasma membrane and secretion of vWF. IsdB bound dose-dependently to activated cells and significantly more than to untreated cells (Fig. 7A). The mechanism of the IsdB binding to resting endothelial cells is unknown.  www.nature.com/scientificreports/ The adhesion of the SH1000 WT and its isogenic isdB mutant to activated HUVEC cells was explored using a bacterial detection method based on the recognition of SpA by a HRP-conjugated rabbit IgG. The WT strain showed a level of adhesion to the monolayers nearly three times higher than the isdB mutant ( Fig. 7B), indicating the active involvement of IsdB in the process. To exclude the possibility that the different levels of adhesion by WT and its isdB mutant to monolayers could be due to the differential expression of IgG-binding proteins, the SH1000 strain and the its isogenic isdB mutant were immobilized onto microtiter wells and then probed with HRP-labelled rabbit IgG. As reported in Fig. 3SA, both bacteria captured similar amounts of antibody, suggesting that they express the same level of protein A and/or other IgG-binding proteins. To rule out a potential interference in the assay due to a presumptive IgG-binding activity of IsdB, we also demonstrated that IsdB does not bind the HRP-conjugated IgG or the anti-L. lactis antibody (Fig. 3SB).
To evaluate the role of the A1 domain of vWF in promoting adherence of S. aureus to HUVEC cells via IsdB, a bacterial adhesion assay was performed in the presence of mAb 6D1. The mAb 6D1, but not an isotype control mAb against ClfA, inhibited binding of IsdB to the A1 domain by 55% at a concentration of 250 ng/well (Fig. 4S), and significantly decreased the adhesion of the WT strain to activated HUVEC cells (Fig. 7B). Conversely, no effect by the mAb 6D1 on adhesion of SH1000 isdB mutant to the monolayers was observed.

Effect of IgG from patients with infective endocarditis on adhesion of S. aureus to HUVEC cells.
Considering the role of adhesion to and invasion of endothelia by S. aureus and the consequent cardiovascular-associated pathologies such as sepsis and endocarditis, a study was designed where the role of IgG against IsdB on bacterial adhesion to HUVEC cells was investigated. A collection of IgG, previously isolated from patients with S. aureus endocarditis 38 , was tested for reactivity to IsdB NEAT1-NEAT2. Although the response varied among the individual IgG preparations, the majority of IgG from patients exhibited a reactivity for IsdB that was higher than that of IgG from sera of healthy donors (Fig. 8A). This observation underlines the in vivo relevance of IsdB as an antigen.
Following this, IsdB was pre-incubated with either IgG from the most reactive patients (P14 and P19) or healthy donors (C2 and C7) and then tested for binding to activated HUVEC cells. Patients' IgG showed a strong blocking effect, while the control IgG affected the binding marginally (Fig. 8B). When SH1000 WT was tested for adhesion to activated HUVEC cells in the presence of IsdB IgG and control IgG, a low but significant blocking effect on adhesion by patients' IgG was noticed, while no action by the control IgG was recorded (Fig. 8C). The limited inhibitory effect of IsdB IgG can be explained by considering both the action of SpA as a vWF receptor The monolayers were incubated with the indicated amounts of IsdB NEAT1-NEAT2 and binding of the ligand to the cells was detected by addition to the wells of a rabbit polyclonal IsdB antibody followed by an HRP-conjugated goat anti-rabbit IgG. The binding of IsdB NEAT1-NEAT2 to the ionophore-untreated cells is also reported. Statistically significant difference is indicated (***, P < 0.001). (B) Endothelial cell monolayers treated as reported in A were incubated with S. aureus SH1000 WT and isdB mutant bacteria. Adherence to the monolayers was detected by the addition of HRP-conjugated rabbit anti-mouse IgG to the wells. The effect of antibodies on the adhesion of bacteria to endothelial cells was tested by incubating S. aureus cells with monolayers in the presence of an anti-A1 6D1 antibody or an isotype control mAb. Binding of HRPconjugated rabbit anti-mouse IgG to monolayers in the absence of bacteria is also reported. Statistically significant difference is indicated (***, P < 0.001). Error bars show S.D. of the means from three independent determinations, each performed in triplicate. www.nature.com/scientificreports/ and its activity as an agent capturing the antibodies. Along this line, we investigated the effect of patients' IgG on IsdB-mediated bacterial adhesion to activated HUVEC by L. lactis pNZ8037::isdB. The L. lactis heterologously expressing IsdB exhibited a level of adhesion that was increased by 40% compared to L. lactis harboring the empty vector (Fig. 8D). Moreover, pre-incubation of L. lactis pNZ8037::isdB with patients' IgG almost completely reduced bacterial adhesion to the monolayers, whereas the controls' IgG did not affect the adhesion.

Discussion
The ability of S. aureus to adhere to and colonize the endothelia is strongly associated with the severity of cardiovascular diseases. Several S. aureus surface proteins are involved in attachment to endothelial cells or the surface-associated ECM. The fibronectin-binding proteins FnBPA and FnBPB recognize fibronectin in the ECM, and FnBPA/B-mediated adhesion is the prerequisite for endothelial cell invasion by S. aureus 39 . Furthermore, ClfA binds to surface-anchored integrin α V β 3 via fibrinogen 40 , and IsdB binds to Vn 26 and SpA interacts with vWF 7 in the ECM or with gC1qR on the surface of activated endothelial cells 41 . However, the bacterial and host patients' sera on IsdB NEAT1-NEAT2 binding to HUVEC monolayers. Confluent HUVEC monolayers were treated with calcium ionophore A23187 and incubated with recombinant IsdB NEAT1-NEAT2 in the presence of the indicated IgG isolated from patients' sera. Bound IsdB NEAT1-NEAT2 was detected by the addition of a rabbit polyclonal IsdB antibody followed by HRP-conjugated goat anti-rabbit IgG. The binding of IsdB NEAT1-NEAT2 to the monolayers in the presence of control IgG is also reported. Binding observed in the absence of antibodies was set as 100% binding. (C) Adhesion of S. aureus SH1000 to HUVEC monolayers in the presence of IgG isolated from patients' sera. HUVEC cell monolayers treated with ionophore A23187 were incubated with cells of S. aureus SH1000 WT in the presence of the indicated IgG isolated from patients' sera or healthy human sera. Adhesion was determined by the addition of an HRP-conjugated rabbit anti-mouse IgG. Bacterial attachment observed in the absence of antibodies was set as 100% adhesion. Statistically significant differences are indicated (*, P < 0.05; **, P < 0.01). (D) Adhesion of L. lactis ectopically expressing IsdB to HUVEC monolayers in the presence of IgG isolated from patients' sera. HUVEC cell monolayers treated with ionophore A23187 were incubated with cells of L. lactis pNZ8037 or L. lactis pNZ8037::isdB in the presence of the indicated IgG isolated from patients' sera or healthy human donors. Adhesion of bacteria to HUVEC monolayers was detected through rabbit anti-L. lactis IgG followed by an HRP-conjugated goat anti-rabbit IgG. Statistically significant difference is indicated (***, P < 0.001). Bars reported in (B-D) represent means ± SD of triplicate tests. www.nature.com/scientificreports/ determinants of endothelial binding have not been fully elucidated. Here we identified vWF as a ligand for S. aureus IsdB and provide a comprehensive analysis of the interaction between the bacterial protein and its host binding partner. As reported for Vn, vWF binding to IsdB was strictly related to the conditions required for the optimal expression of the protein (stationary phase of growth and iron-starvation). Thus, IsdB-expressing bacteria captured soluble vWF and adhered to the immobilized molecule. L. lactis ectopically expressing IsdB also attached to surface-coated vWF. The observation that heparin blocked IsdB binding to vWF and the finding that IsdB directly interacted with the A1 domain clearly indicated that the A1 contains the IsdB-binding site. Thus, IsdB and SpA share the same binding site on vWF. It is noteworthy that SpA-mediated adhesion of S. aureus to vWF has been demonstrated in conditions of low shear stress 12 , whereas our findings regarding IsdB-binding to vWF were performed in static conditions. However, it cannot be ruled out that IsdB may also play a role in mediating staphylococcal adhesion under shear stress conditions. IsdB binding to vWF significantly increased when the protein was treated with ristocetin, a compound that reproduces in vivo high shear stress-induced conformational change of vWF. Thus, it can be concluded that the IsdB binding site is partially hidden in the compact conformation of vWF.
The importance of IsdB-mediated adhesion of S. aureus to vWF could be even more relevant in vivo considering that the role of SpA as a vWF receptor could be of minor impact in a milieu such as blood where the immunoglobulins could successfully compete with vWF for binding to SpA expressed on the surface of bacteria 7 .
To localize the vWF-binding site(s) on IsdB, the recombinant NEAT1 and NEAT2 modules were tested for their ability to interact with the A1 domain by ELISA. The A1 domain bound dose-dependently to the individual modules and with a binding profile resembling that exhibited using IsdB NEAT1-NEAT2. As determined by SPR, each module interacted with the A1 domain with an affinity comparable to that of IsdB NEAT1-NEAT2. However, the NEAT2 module showed an affinity for A1 that was higher than that of NEAT1 that is consistent with the low sequence identity and the different functionality of the modules.
The ability of A1 to bind both the modules is reminiscent of Vn binding to each NEAT domain. However, Vn did not compete with the binding of A1 to IsdB (data not shown), indicating that A1 and Vn recognize and bind to different subsites within the IsdB domains and, possibly, with different binding mechanisms. The evidence that high ionic strength significantly reduced the IsdB interaction with the A1 domain, as well as molecular docking analysis, provide clues that electrostatic bonds could be involved in the vWF-IsdB interaction.
We also demonstrated that IsdB mediates S. aureus adhesion to vWF on activated HUVEC cells. Furthermore, we show that adhesion directly involves the A1 domain as indicated by the specific inhibitory effect of the A1 mAb named 6D1 on the process. Thus, a scenario can be envisaged where activated endothelial cells secrete ultra large vWF fibers that run parallel to the direction of flow in the blood 42 . vWF multimers can also rebind to the endothelial surface via integrin α v β 3 42 or, in case of endothelial damage or inflammation, to subendothelial matrix molecules such as fibrin monomers 43 , collagens [44][45][46] or fibronectin 47 . Due to the effect of shear stress, the globular, concealed A1 domain in vWF fibers becomes exposed and allowed to interact with S. aureus via specifically up-regulated IsdB in the iron-limited conditions of the blood. These events allow the pathogen to withstand the strong current of the blood flow and create the prerequisite of vascular infections such as infective endocarditis. In support of the crucial role of A1 in bacterial adherence, in the presence of heparin, a binder of A1, both fiber formation and bacterial adhesion are simultaneously reduced 5 .
The importance of bacterial receptor/vWF interaction in vascular infections is underlined by the consideration that, among the staphylococci, only S. aureus 7,8,34 and the coagulase-negative Staphylococcus lugdunensis are able to bind vWF and are more effective in causing endocarditis compared to other staphylococcal species 48 . The nature of the vWF receptor in S. lugdunensis remains elusive, despite its potential importance in the pathogenesis associated with this bacterium 49,50 . Perhaps the IsdB orthologue of S. lugdunensis binds vWF and promotes endothelial colonization and invasion.
The functional aspects of S. aureus binding to vWF can be correlated with the pathophysiological consequences of this interaction such as infective endocarditis, sepsis, and cardiovascular complications. Infection of the heart valves is triggered by the attachment of circulating bacteria to the endocardium and the formation of bacterial vegetations, which are embedded in fibrin and platelets. Bacterial growth occurs within cells and the matrix inside vegetations, making it difficult for the host immune system to control or eradicate the ongoing infection. Therefore, in the perspective of future therapeutic interventions, the acquisition of information on the immune response by the host remains essential.
Along this line, we examined the reactivity of IgG isolated from patients with S. aureus endocarditis and their possible interference with the vWF/IsdB interaction. A considerable proportion of isolated IgG from a collection of human sera showed a significant reactivity to IsdB. Moreover, anti-IsdB antibodies blocked the binding of recombinant IsdB to activated endothelial cells and interfered with the adhesion of L. lactis ectopically expressing IsdB to HUVEC cells. On the other hand, the antibodies affected less, although significantly, S. aureus adhesion to endothelial cells. The reduced efficacy of immune IgG on adhesion of SH1000 to HUVEC cells may be attributable to the action of SpA as a vWF receptor per se and/or to its IgG neutralizing activity. Moreover, the involvement of other staphylococcal vWF-binding partners in the process cannot be excluded. Finally, although our data suggest a role for IsdB as a host colonization/invasion factor and as a potential player in S. aureus pathogenesis, it remains to be determined whether IsdB can be used as potential vaccine component. Indeed, it has been found that a vaccine including IsdB induced a higher mortality rate than the placebo in human patients 51,52 . Thus, additional studies are needed to better estimate the pathophysiological role of IsdB and its therapeutic value.  55 or harboring the isdB gene (pNZ8037::isdB) 25 were grown overnight in BHI medium supplemented with chloramphenicol (10 μg/ml) at 30 °C without shaking. Cultures of L. lactis were diluted at 1:100 in the same medium and allowed to reach the exponential phase of growth. Nisin (6.4 ng/ml) was added, and cultures were allowed to grow overnight as above. In those experiments where a defined number of cells were used, bacteria were harvested from the cultures by centrifugation, washed, and suspended in phosphatebuffered saline (PBS) at OD 600nm = 1.0. Escherichia coli BL21 (DE3) (Invitrogen, Carlsbad, CA, USA) transformed with vector pQE30 or pET28a (Integrated DNA Technologies, Leuven, Belgium) was grown in Luria agar and Luria broth (VWR International Srl) containing 100 μg/ml ampicillin or kanamycin, respectively, at 37 °C with shaking.
Plasmid and DNA manipulation. Plasmid DNA (   www.nature.com/scientificreports/ re-suspended in lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, pH 8.0) containing 1 mM phenyl-methanesulfonyl-fluoride (PMSF) (Sigma-Aldrich) and 20 μg/mL DNase (Sigma-Aldrich) and lysed by sonication (70% amplitude, 12 × 30″ on/off, 1′30″ interval between sonication steps). The cell debris was removed by centrifugation and proteins purified from the supernatants by Ni +2 -affinity chromatography on a HiTrap chelating column (GE Healthcare, Buckinghamshire, UK Reagents, proteins, and antibodies. BSA (bovine serum albumin), protease-free DNase I, skim milk, von Willebrand factor, heparin, chondroitin sulphate, heparan sulphate, lysostaphin, nisin, protein A (SpA 37-485 ), biotin, avidin-HRP, trypsin, calcium ionophore A23187 were purchased from Sigma-Aldrich. Collagens type I, type III, type IV, and type VI were purchased by Merck (Darmstadt, Germany). Ristocetin was from Hyphen BioMed (Neuville-sur-Oise, France). IgG were isolated from patients' with infective endocarditis as previously reported 38 , all methods were carried out in accordance with relevant guidelines and regulations, all experimental protocols were approved by the ethical board of the University of Pavia and informed consent was obtained from all human participants. Anti-A1 6D1 monoclonal antibody was raised as previously described 34 . αThrombin and PPACK were bought from Haematologic Technologies (Essex Junction, VT, USA). Sensor Chip NTA and Ni 2+ Sepharose 6 Fast Flow resin were purchased by Cytiva Lifesciences (Washington, USA). IsdB antibody was raised in a rabbit by routine immunization procedure using purified IsdB NEAT1-NEAT2 as the antigen. Anti-L. lactis antibody was raised in a rabbit by routine immunization procedure using heat-inactivated L. lactis NZ9800 cells as the antigen. Rabbit anti-human von Willebrand factor IgG and rabbit anti-mouse or goat anti-rabbit horseradish peroxidase (HRP)-conjugated secondary antibodies were purchased from Dako Cytomation (Glostrup, Denmark). Monoclonal mouse HRP-conjugated α-polyHistidine antibody was purchased from Abcam (Cambridge, UK).

Release of CWA proteins from S. aureus and detection of vWF-binding activity. Lysostaphin
digestion. CWA proteins from S. aureus were released by following the protocol described by Pietrocola et al 26 .
Briefly, bacteria were grown to the stationary phase, either in RPMI or BHI medium, harvested by centrifugation at 7000 × g for 15 min at 4 °C, washed three times with PBS, and resuspended to an A 600 = 2.0 in lysis buffer (50 mM Tris-HCl, 20 mM MgCl 2 , pH 7.5) supplemented with 30% (w/v) raffinose. Cell wall proteins were solubilized from S. aureus by incubation with lysostaphin (200 μg/ml) for 20 min at 37 °C in the presence of protease inhibitors (Complete Mini; Sigma-Aldrich). Protoplasts were recovered by centrifugation at 6000 × g for 20 min, while the supernatant taken as the wall fraction was concentrated by treatment with 20% (v/v) trichloroacetic acid (TCA) for 30 min at 4 °C. The precipitated proteins were washed twice with ice-cold acetone and dried overnight.

ELISA-type solid-phase binding assays. Binding of vWF to
The binding of 1 μg of IsdB to surface coated vWF (1 μg/well) in presence of increasing concentrations (from 0 to 10 μg/well) of heparin, chondroitin sulfate, or heparan sulfate was detected with IsdB antibody as reported above. www.nature.com/scientificreports/ To assess the effect of ionic strength on the IsdB-vWF interaction, microtiter wells coated with 1 μg of vWF were incubated with 1 μg of IsdB in presence of increasing concentrations of NaCl (from 0 to 1 M). Complex formation was detected by incubation of the wells with the IsdB antibody.
The dose-dependent binding of IsdB to surface-coated human vWF or A1 domain (1 μg/well) was evaluated by incubation of the plates with increasing concentrations of IsdB (from 0 to 5 μg/well). Bound protein was revealed as reported above.
Binding of IsdB to the vWF A1 domain was performed incubating immobilized A1 domain (1 μγ/well) with IsdB (1 μg/well) in the presence of 250 ng of 6D1 or anti-ClfA monoclonal antibodies and detected as reported above.
Binding of increasing concentrations of the A1 domain (from 0 to 5 μg/well) to immobilized IsdB NEAT1-NEAT2 or its subregions NEAT1 and NEAT2 (1 μg/well) was also assessed and detected by the addition of anti-human vWF IgG.
To assess the allosteric conformation change of vWF and the A1 domain exposure by ristocetin treatment, vWF (10 μg /ml) was treated with 0.5 mg/ml of ristocetin and the complex used to coat microtiter wells (100 μl). The A1 domain exposure was evaluated by the addition of the anti-A1 domain monoclonal antibody 6D1 (250 ng/ well) followed by HRP-conjugated rabbit anti-mouse IgG (1:1000) to the wells.
The impact of the conformational change of the ristocetin-treated vWF was evaluated by measuring the binding of soluble vWF (1 μg/well) to immobilized IsdB (1 μg/well). The complex formation was detected using anti-human vWF IgG as previously reported.
Bacterial adhesion to surface-coated vWF. The ability of S. aureus or L. lactis ectopically expressing IsdB to adhere to surface-coated vWF was evaluated by ELISA-based assay. Microtiter wells coated with vWF were incubated with cells (A 600 = 1.0) of S. aureus SH1000 and its isdB mutant obtained from cultures grown to stationary phase in RPMI and suspended in 0.5% (v/v) BSA. The wells were extensively washed with PBST, blocked with 2% (v/v) BSA, and incubated with 100 µl bacterial suspensions for 1 h at 22 °C. The expression of SpA on cell surface was exploited to detect bacteria adhesion by incubating the plates for 45 min with HRP-conjugated rabbit anti-mouse IgG (1:1000).
Binding of non-immune IgG such as HRP-conjugated goat anti-rabbit IgG or anti-L. lactis IgG to IsdB was determined by incubating surface-coated recombinant IsdB (1 μγ/well) with the antibodies and antibody binding was detected as reported above.
Adhesion of L. lactis to vWF was performed by incubating plates coated with vWF with cells of L. lactis expressing IsdB (L. lactis pNZ8037::isdB ) and the strain carrying the empty vector (L.lactis pNZ8037 ) obtained from cultures grown in BHI. L. lactis binding to surface-coated vWF was detected by incubating for 1 h at 22 °C with rabbit polyclonal anti-L. lactis IgG (1 µg/well) followed by an HRP-conjugated goat anti-rabbit IgG (1:1000).

Reactivity of IgG from patients with infective endocarditis against IsdB.
To test the reactivity of IgG from the collection of infective endocarditis sera IsdB NEAT1-NEAT2 was immobilized onto microtiter wells (1 µg/well). After blocking with BSA, the wells were incubated with antibodies (1 μg/well) from patients and healthy donors. The binding of antibodies was revealed by the addition of a polyclonal rabbit anti-human IgG (1:1000).

Capture of vWF by S. aureus cells.
Cells of S. aureus strain SH1000 WT or the isdB mutant, grown to stationary phase in RPMI, were harvested by centrifugation at 7000 × g at 4 °C for 15 min, washed three times with PBS, and resuspended to an A 600 = 1.0 in PBS. Cells were then incubated with human vWF (5 µg/mL) for 1 h at 22 °C under constant shaking. The extraction of vWF captured by bacteria was conducted as previously described 22 . Briefly, bacteria were treated with the extraction buffer (125 mM Tris-HCl, pH 7.0, 2% (w/v) SDS) for 3 min at 95 °C and finally centrifuged at 10.000 × g for 3 min. The supernatants were subjected to 5% (w/v) SDS-PAGE under reducing conditions, and the proteins were electrotransferred to a nitrocellulose membrane. The membrane was incubated with a rabbit polyclonal vWF antibody followed by HRP-conjugated goat antirabbit IgG. The band intensities were quantified with Quantity One software (Bio-Rad).
Surface plasmon resonance. For performing SPR measurements on a NTA-Ni 2+ sensor chip (see below), the recombinant 6xHis-tag-A1 was treated with α-thrombin to remove the 6xHis tag to yield vWF A1 with an additional 17-amino acid segment at the N-terminus. The fused protein (1 mg/ml, 0.5 ml) was treated at an enzyme:substrate molar ratio of 1:200 in PBS, for 1 h at 25 °C. The reaction was quenched by adding 1 μM (D)-Phe-Pip-Arg-chloromethylketone, as an irreversible thrombin inhibitor, while A1 was purified by the batch mode procedure. Briefly, the reaction mixture was incubated with 25 μl of Ni 2+ -Sepharose-6 Fast Flow resin at 22 °C for 1 h under gentle stirring on an orbital shaker. The supernatant, containing the purified A1, was then collected and protein purity and chemical identity assessed by non-reducing 4-12% (w/v) SDS-PAGE, high-resolution mass spectrometry and by Dot Blot analysis (not shown), using anti-His tag IgG as a probe, all confirming the removal of the 6xHis-tag.
SPR analyses were carried out on a dual flow-cell Biacore X-100 instrument (Cytiva, Uppsala, Sweden) as described 65 . 6xHis-tag-IsdB proteins (i.e. the ligands) were non-covalently immobilized onto a Ni 2+ -chelated nitrilotriacetate (NTA) carboxymethyldestrane sensor chip and incremental concentrations of A1 domain (i.e. the analyte) lacking the 6xHis-tag were loaded at incremental concentrations in the mobile phase, following the single-cycle operation mode. The Ni 2+ -NTA/6xHis-IsdB chip assembly was prepared as follows: the NTA chip (Cytiva) was first washed (flow-rate: 30 μl/min) with 0. 35  . After each set of measurements, the NTA chip was regenerated by a pulse of regeneration buffer (350 mM EDTA). Each sensogram was subtracted for the corresponding baseline, obtained on the reference flow cell and accounting for nonspecific binding, i.e. typically less than 2% of RU max . The binding data were analyzed using the BIAevaluation software vs 2.0. The sensorgrams (Fig. 6, black curves) were fitted with theoretical curves obtained by simulations based on several different 1:1 stoichiometric binding models: (i) simple analyte-ligand interaction, (ii) bivalent analyte, and iii) heterogeneous ligand binding model 35,36 . The best fit, as evaluated from the χ 2 values of experimental and simulated sensorgrams, was obtained using the heterogeneous ligand binding model, which assumes the existence of two differently populated orientations (i.e., L1 and L2) of the immobilized ligand, allowing the exposure of different ligand surfaces which are variably accessible for interaction with the analyte and bind the analyte (i.e. A1 domain) with different affinities (i.e. K D1 and K D2 ). This is even more likely to occur with highly charged IsdB proteins bound on the NTA sensor chip, formed by the negatively charged chelating agent NTA and the positive Ni 2+ ions. The relative abundance of L 1 and L 2 were estimated from their RU max values, obtained as a fitting parameter, e.g. L 1 = [RU max1 /(RU max1 + RU max2 )] × 100 35,36 . Computational methods. Electrostatic potential calculations were carried out using the APBS program 66,67 , run on the crystallographic structure of IsdB_E chain (5vmm.pdb) 18 , after removal of Hb coordinates, and on the crystallographic structure of the A1 domain bound to the platelet thrombin receptor GpIbα (1u0n.pdb) 68 , after removal of the receptor coordinates. Calculations were performed using a solvent dielectric of 78.14 and a protein dielectric of 2.0 at 310 K in 150 mM NaCl.

Binding and adhesion assays to endothelial cells. IsdB binding to endothelial cells. Human umbili-
cal vein endothelial cells (HUVEC) from a single donor (Lonza, Spain) were kindly provided by the Researchers of the Neonatal Unit and Neonatal Intensive Care Unit, Fondazione IRCCS Policlinico S. Matteo, Pavia, Italy and cultured as previously reported 40 . To examine the binding of recombinant IsdB NEAT1-NEAT2 to endothelial cells, HUVECs were cultured onto 96-microtiter wells. Monolayers (8 × 10 4 cells/well) were treated with 0.1 mM calcium ionophore A23187 (Sigma-Aldrich) for 10 min at 22 °C, washed three times with PBS, and then fixed with 3% (w/v) paraformaldehyde in PBS for 10 min. The wells were thoroughly rinsed with PBS, blocked with BSA (v/v) 2% in PBS for 1 h, and then incubated with increasing concentrations of recombinant IsdB (0.63-2.5 µg/well) in PBS for 1 h. After extensive washing, IsdB binding to the wells was detected by addition to the wells of rabbit polyclonal IsdB IgG followed by HRP-conjugated goat anti-rabbit IgG.
Bacterial adherence to endothelial cells. The ability of S. aureus cells to adhere to HUVEC cells was assessed by an ELISA-based assay. 100 μl of bacterial suspensions (A 600 = 1.0) of S. aureus strain SH1000 WT and its isogenic isdB mutant obtained from cultures grown in RPMI were added to ionophore-treated HUVEC monolayers and the wells incubated for 1 h. The attached bacteria were detected by incubating the wells with HRP-conjugated rabbit anti-mouse IgG (1:1000) for 45 min at 22 °C. To test the effect of the anti-A1 monoclonal antibody (mAb) 6D1 on the adhesion of S. aureus SH1000 to HUVEC monolayers, the assay was performed in the presence of 250 ng/well of the 6D1 or isotype matched anti-ClfA mAb and bacterial attachment determined as above.
Inhibitory activity of patients' IgG on the interaction of IsdB with vWF expressed on endothelial cells. The ability of the patients' IgG to interfere with the binding of recombinant IsdB NEAT1-NEAT2 to ionophore-treated and fixed endothelial cells was determined by incubating the wells with 2.5 μg IsdB in the presence of the indicated IgG (10 μg/well) for 1 h at 22 °C. The binding of IsdB to the cells was detected as reported above. To analyse the effect of patients' IgG on adherence of staphylococci to ionophore-treated HUVEC cells, 100 μl of a S. aureus SH1000 WT suspension (A 600 = 1.0) and 10 μg/well of the indicated IgG were simultaneously added to the monolayers and the wells incubated for 1 h at 22 °C. Bacterial adherence was detected by incubating the wells with an HRP-conjugated rabbit anti-mouse IgG (1:1000). A similar adhesion protocol was used to test the effect of patients' IgG on adhesion of L. lactis pNZ8037::isdB and L. lactis pNZ8037 to the monolayers. L. lactis adherence to the cells was determined by adding to the wells rabbit anti-L. lactis IgG (1 µg/well) followed by an HRP-conjugated goat anti-rabbit IgG (1:1000).
Statistical methods. Analyses were performed using Prism 4.0 (GraphPad). Comparison of more than two groups were performed with the one-way ANOVA followed by Bonferroni's post hoc tests. The two-tailed Student's t-test was employed to compare two groups. P values < 0.05 were considered statistically significant (*, P < 0.05, **, P < 0.01, ***, P < 0.001).