Concerted functions of Streptococcus gordonii surface proteins PadA and Hsa mediate activation of human platelets and interactions with extracellular matrix

Summary A range of Streptococcus bacteria are able to interact with blood platelets to form a thrombus (clot). Streptococcus gordonii is ubiquitous within the human oral cavity and amongst the common pathogens isolated from subjects with infective endocarditis. Two cell surface proteins, Hsa and Platelet adherence protein A (PadA), in S. gordonii mediate adherence and activation of platelets. In this study, we demonstrate that PadA binds activated platelets and that an NGR (Asparagine‐Glycine‐Arginine) motif within a 657 amino acid residue N‐terminal fragment of PadA is responsible for this, together with two other integrin‐like recognition motifs RGT and AGD. PadA also acts in concert with Hsa to mediate binding of S. gordonii to cellular fibronectin and vitronectin, and to promote formation of biofilms. Evidence is presented that PadA and Hsa are each reliant on the other's active presentation on the bacterial cell surface, suggesting cooperativity in functions impacting both colonization and pathogenesis.


| INTRODUCTION
Streptococcus, Staphylococcus, and Enterococcus bacteria account for >80% cases of infective endocarditis (Muñoz et al., 2015;Slipczuk et al., 2013) and are able to trigger activation or aggregation of blood platelets into a clot or thrombus (Fitzgerald, Foster, & Cox, 2006;Kerrigan, 2015). Viridans-group streptococci that enter the bloodstream in otherwise healthy subjects almost always originate from the complex microbial communities present within the human oral cavity (Cahill & Prendergast, 2015;McNicol & Israels, 2010;Nilson, Olaison, & Rasmussen, 2015). Accordingly, there is a predictive link between levels of oral hygiene and the risk of cardiovascular disease (Lockhart et al., 2009). Platelet adhesion and activation by oral streptococci occurs by several different mechanisms (Cognasse et al., 2015;McNicol, 2015). For Streptococcus gordonii, S. oralis and S. sanguinis, a common primary interaction occurs with platelet integrin receptor GPIb mediated by a bacterial surface serine-rich repeat protein (Deng et al., 2014). In S. sanguinis, a direct interaction between serine-rich repeat protein SrpA and GPIbα leads to platelet rolling over immobilized bacteria and adhesion at low shear (Kerrigan et al., 2002;Plummer et al., 2005).
The mechanisms by which platelets are activated by streptococci following adhesion are poorly understood. Recent work has identified the importance of specific antibodies in bacterial activation of platelets through receptor FcγRIIa (Arman et al., 2014;Pampolina & McNicol, 2005;Tilley et al., 2013). In S. gordonii, a ubiquitous human oral bacterium, there is direct activation of FcγRIIa integrin (non-antibody mediated), and then inside-out activation of the most highly-expressed platelet receptor GPIIbIIIa (α IIb β 3 ) . Further activation is associated with secretion of granules and secondary mediators, This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. filopodia formation (spreading), generation of ADP and thromboxane A2 (TxA2) to reinforce α IIb β 3 activation (Cox, Kerrigan, & Watson, 2011), and then aggregation with nearby-activated platelets via fibrinogen (Moriarty et al., 2015).
We have identified a protein designated Platelet adherence protein A (PadA), that is expressed on the surface of S. gordonii, and which interacts directly with platelets . PadA precursor (3,646 amino acid [aa] residues) comprises a N-terminal region of 1,328 aa residues containing a von Willebrand Factor (vWF)-like domain (aa residues 72-229), and a C-terminal region comprising 14 blocks of aa residue repeats  with a bacterial cell wall anchor region and sortase (LPxTG) motif (Mazmanian, Ton-That, & Schneewind, 2001) (see Figure 1). PadA, unlike Hsa, does not interact with sialylated GPIb , but was shown to bind directly to the platelet receptor α IIb β 3 over-expressed on the surface of Chinese hamster ovary cells . Binding of PadA to α IIb β 3 on platelets results in platelet activation; however, it is not known if PadA is able to activate platelets on its own, or if it requires a coactivator. Previous results suggest that α IIb β 3 may bind to S. gordonii via common integrin-recognition motifs (RGT and AGD) present within the N-terminal region of PadA (Keane et al., 2013).
Recently, it has emerged that another motif, NGR (Asn-Gly-Arg), found in the D domain of fibrinogen on each of the β and γ chains, plays an important role in the interaction of α IIb β 3 with fibrinogen (Moriarty et al., 2015). A similar motif is found within the N-terminal region of PadA, raising the possibility that this might be a key factor in recognition of α IIb β 3 by PadA.
In this article, we have investigated the relative functions of Hsa and PadA in S. gordonii platelet interactions, and in bacterial cell binding to various extracellular matrix components that may be exposed at sites of endothelial damage, for example, fibronectin or vitronectin. We have also studied in more detail the interactions of the mature N-terminal F2 region (657 aa residues) of PadA with platelets and matrix components, and more specifically the relative roles of the integrin-recognition motifs and NGR in S. gordonii host interactions.
2 | RESULTS 2.1 | Expression of padA and hsa in deletion mutants and complemented strains A diagrammatic representation of PadA protein is presented in Figure 1 together with a visual expansion of the 690 aa-residue precursor N-terminal domain (F2) of main focus in this paper, and the corresponding aa sequence. Three integrin-like recognition motifs within the F2 domain are highlighted. To investigate further the functional properties of the PadA protein, the entire coding region was cloned into replicative plasmid vector pMSP downstream of a nisin-inducible promoter (see 4). The hsa gene was cloned in the same way as previously described (Jakubovics, Brittan, Dutton, & Jenkinson, 2009). Because Hsa is glycosylated by various products encoded by genes within the hsa-accessory secretion system locus (Zhou & Wu, 2009), authentic Hsa protein has not been expressed in a surrogate host bacterium. The results obtained from PadA Western immunoblot analyses of mutant and complemented strains are shown in Figure 2a. PadA protein was deficient in cell wall extracts of the ΔpadA and ΔpadA Δhsa mutants, and was highly-expressed in the complemented strains following 10 ng nisin ml −1 induction.
Hsa production was detected with succinylated-wheat germ agglutinin (sWGA), which recognizes the glycosylated protein (Takahashi, Yajima, Cisar, & Konishi, 2004). Hsa was well-expressed ectopically in the complemented Δhsa and ΔpadA Δhsa mutants when induced with 50 ng nisin ml -1 (Figure 2b,c). Higher nisin concentrations were growth-inhibitory. Production of Hsa from the chromosomal locus in the ΔpadA mutant was unaffected and was similar to wild type strain DL1 expression levels ( Figure 2c; Petersen et al., 2010). PadA precursor comprises 3,646 aa residues with N-terminal region (1,328 aa), and C-terminal region (2,318 aa) carrying 14 repeat blocks of 148-152 aa residues and a cell wall anchor motif LPKTG. The F2 region of the PadA polypeptide (657 aa) has been expanded to show the extent of the vWF-like domain and the approximate positions of NGR, RGT, and AGD motifs. (b) aa sequence of the N-terminal region 690 aa residues, indicating signal (leader) sequence (in green), the vWF-like domain (purple), and the three integrin-recognition motifs, NGR (214-6 aa), RGT (416-8 aa), and AGD (485-7 aa) in red type 2.2 | Role of PadA and Hsa in platelet adhesion by S. gordonii It is well documented that Hsa interacts with platelets Kerrigan et al., 2007;Takahashi et al., 2004), and we have previously shown that PadA also is involved in S. gordonii binding platelets . This is confirmed (Figure 3) with the ΔpadA mutant being approximately 30% reduced in levels of platelet adherence, and the ΔpadA Δhsa mutant >80% reduced in binding platelets.
Expression of padA in the ΔpadA Δhsa mutant does not restore any level of platelet adhesion, but expression of hsa restored platelet binding levels to~70% of wild type ( Figure 3). These results show that interaction of PadA with resting platelets requires the presence of Hsa.
Previously, we have suggested that the RGT and AGD motifs have little or no role in supporting platelet adhesion, but are involved in the transformation of the platelet biconcave disc through formation of filopodia and lamellipodia to a fully spread cell (Keane et al., 2013). Accordingly, we tested the effect of NGR AAA mutation on platelet spreading and found that for those platelets that adhered, spreading was unaffected ( Figure 5). Taken collectively, these results suggest that NGR plays a major role in directing adhesion of platelets, while AGD and RGT promote spreading. Adhesion of non-activated or TRAP-activated platelets to immobilized recombinant PadA-F2 region fragments under static conditions. Recombinant PadA protein fragments were immobilized onto microtitre plate wells (10 μg per well), and non-specific binding sites were blocked with BSA. Gel-filtered platelets were either non-activated (a) or activated by addition of TRAP (Thrombin Receptor Activating Peptide; b) and were incubated with the immobilized proteins at 37°C for 45 min (2 × 10 7 platelets per well). Platelet adherence was determined by phosphatase assay. RGT AAA , AGD AAA , NGR AAA , and so forth indicate the motifs within the various F2 fragments that were alanine-substituted to AAA. Error bars represent ±SEM from three independent experiments (n = 3). * P < 0.05. In (a), NS = not statistically significant (F2 v F2 RGT AAA AGD AAA , P = 0.429; F2 v F2 NGR AAA RGT AAA AGD AAA , P = 0.962) FIGURE 5 Platelet spreading on immobilized recombinant PadA-F2 region fragments under static conditions. (a) BSA (negative control), (b) fibrinogen (positive control), (c) recombinant PadA protein fragment F2, or (d) fragment F2 NGR AAA were immobilized onto microtitre plate wells (10 μg per well), and non-specific binding sites were blocked with BSA. Gel-filtered platelets were incubated with the immobilized substrates at 37°C for 45 min (2 × 10 7 platelets per well), and platelet spreading was visualized by confocal microscopy. Scale bar = 15 μm. (b) Percentage of platelets spread on BSA (negative control), fibrinogen (positive control), and recombinant PadA protein F2 fragments immobilized onto glass slides. Error bars represent ±SEM from three independent experiments (n = 3). NS = not statistically significant

| Functions of PadA and Hsa in binding fibronectin
We then utilized an affinity chromatography proteomics approach to determine if PadA interacted with specific host proteins present in human plasma. Purified PadA protein with a ×6 His C-terminal tag (PadA 6His ) was linked to Ni-NTA magnetic beads and incubated with plasma. The beads were collected, washed, and the interacting proteins were eluted, subjected to SDS-PAGE, in-gel digested with trypsin, and analysed by tandem mass spectrometry. Data analysis (see 4) identified Fn1 protein (fibronectin) (B7ZLE5_HUMAN) as the highest scoring protein to be pulled down (99% identity confidence/ SEQUEST, 31.16% sequence coverage, 48 unique peptides) relative to plasma controls. Also, uncharacterized protein Q6GMX0_HUMAN was identified (47% sequence coverage, 1 unique peptide), recently annotated as anti-polyhydroxybutyrate antibody Fv light chain, the significance of which is unclear. Fibronectin and Q6GMX0 peptides were low scoring or non-detectable, respectively, in parallel plasma controls.
Fibronectin is an extracellular matrix (ECM) protein often found at sites of endothelial cell damage to which platelets and bacteria are attracted. Plasma fibronectin (pFn) is a major component of the fibrin clot (early wound repair), while cellular fibronectin is involved in later repair events (To & Midwood, 2011). It is known that Hsa is involved in S. gordonii binding to pFn by recognition of sialylated regions on the molecule (Jakubovics et al., 2009). Accordingly, we tested the ΔpadA or Δhsa mutants and complemented strains in adherence to pFn and to cFn. Levels of binding were slightly higher to cFn, but the overall adherence patterns of the strains were identical; therefore, we only present the data for binding to cFn. The ΔpadA mutant was approximately 25% reduced in binding cFn, while the complemented strain was above wild type levels of binding ( Figure 6). These adherence events were reduced by 50-60% when the cFn was desialylated ( Figure 6). Deletion of hsa led to ablation of cFn binding, while complementation of the Δhsa mutant led to part-restoration of cFn-binding activity ( Figure 6). Complementation of the ΔpadA Δhsa mutant with padA led to a small increase in cFn binding, while complementation with hsa restored binding levels to just below wild type ( Figure 6).
We therefore conclude that Hsa is a major mediator of Fn binding under these conditions and that PadA plays a minor, but significant, role in the process. The fact that adhesion levels are only 50-60% reduced for desialylated cFn is consistent with the presence of PadA and other proteins that interact with non-sialylated Fn (Jakubovics et al., 2009). Interestingly, complementation of ΔpadA Δhsa with either padA or hsa alone did not complement the Vn-binding phenotype. These results strongly indicate that Hsa requires the presence of functional PadA in order to efficiently bind Vn. Lastly, adherence of streptococcal cells to Vn was ablated by sialidase (neuraminidase) treatment of Vn, implying that Vn adherence is sialylation dependent (Figure 7).

| Functions of PadA and Hsa in adherence to salivary glycoproteins and in biofilm formation
S. gordonii is normally found in the oral cavity and produces a spectrum of adhesins that interact with salivary components (Nobbs, Lamont, & Jenkinson, 2009). To test the effects of padA or hsa deletions on initial streptococcal adherence to salivary pellicle, bacteria were incubated with saliva-coated glass cover slips, and levels of adhesion determined by crystal violet staining. We detected significant differences in adherence of the ΔpadA and ΔpadA Δhsa mutants compared with wild type DL1 (Figure 8). Complementation with padA restored adherence levels while complementation with hsa did not ( Figure 8).

FIGURE 6
Adhesion of S. gordonii strains to human cellular fibronectin (cFn). Microtitre plate wells were coated with cFn (1 μg per well), blocked with BSA, and then incubated with streptococcal cells (5 × 10 7 per well) for 2 hr at 37°C. Bacterial cells adhered (black columns) were quantified by staining with crystal violet as described in 4. Wells coated with cFn were also incubated with 0.001 U neuraminidase (sialidase) for 2 hr at 37°C, washed, blocked with BSA and then incubated with streptococcal cells (grey shaded columns). Expression of PadA or Hsa proteins by complemented strains was induced with 10 ng or 50 ng nisin ml −1 . Error bars represent ±SEM from three independent experiments (n = 3). * P < 0.05 for comparisons indicated; § P < 0.05 for neuraminidase-treated cFn versus untreated In subsequent biofilm formation, the ΔpadA mutant was 50% decreased in biomass compared to wild type, and complementation with padA restored biomass to slightly above wild type levels ( Figure 8). Deletion of hsa resulted in similar effects, and complementation of the Δhsa mutant was highly effective in enhancing biofilm formation. The ΔpadA Δhsa mutant was ablated in biofilm formation ( Figure 8). Complementation with padA restored biofilm formation, but complementation with hsa did not ( Figure 8). These results suggest that PadA must be fully functional for Hsa to promote biofilm formation. However, it also appears that PadA alone can provide the necessary function for biofilm formation in the absence of Hsa ( Figure 8).
2.7 | PadA integrin-recognition motifs are not involved in binding cFn, Vn, or salivary pellicle Because the previous data strongly suggest that PadA is involved in binding ECM substrata and salivary pellicle, we tested the ability of the PadA-F2 region to mediate these interactions, and the role of the three integrin-recognition motifs. In binding assays of purified F2 fragments to immobilized substrata, it was found that levels of binding to cFn were higher than to Vn and pellicle ( Figure 9). Substitution of all three integrin-recognition motifs (RGT, AGD, and NGR) with AAA failed to significantly affect levels of adhesion of the PadA-F2 fragment to Fn, but there was slightly elevated interaction with Vn and with salivary pellicle (Figure 9). Sialidase treatment of the substrata did not affect levels of PadA-F2 binding compared to untreated substrata ( Figure 9). Microtitre plate wells were coated with Vn (0.05 μg per well), blocked with BSA, and then incubated with streptococcal cells (5 × 10 7 per well) for 2 hr at 37°C. Bacterial cells adhered (black columns) were quantified by staining with crystal violet as described in 4. Wells coated with Vn were also incubated with 0.001 U neuraminidase (sialidase) for 2 hr at 37°C, washed, blocked with BSA, and then incubated with streptococcal cells (grey columns). Expression of PadA or Hsa proteins by complemented strains was induced with 10 ng or 50 ng nisin ml −1 , respectively. Error bars represent ±SEM from three independent experiments (n = 3). * P < 0.05 for comparisons indicated; § P < 0.05 for neuraminidase-treated Vn versus untreated FIGURE 8 S. gordonii strains adherence to salivary pellicle and biofilm formation. Cover slips were coated with salivary pellicle and incubated with streptococcal cells (5 × 10 7 per well) for 2 hr at 37°C for adherence (grey columns) or in YPT-Glc medium for 16 hr at 37°C for biofilm formation (black columns). Bacterial cells adhered, and biofilm biomass values were quantified by staining with crystal violet as described in 4. Expression of PadA or Hsa proteins by complemented strains was induced with 10 ng or 50 ng nisin ml −1 , respectively. Error bars represent ±SEM from three independent experiments (n = 3). * P < 0.05 for biofilm comparisons indicated; § P < 0.05 for significantly different adherence versus DL1 Adhesion of PadA-F2 region fragments to cellular fibronectin (cFn), vitronectin (Vn), or salivary pellicle (SP). cFn, Vn, or saliva were immobilized onto microtitre plate wells and non-specific binding sites were blocked with BSA (black columns). Wells coated with cFn, Vn, or salivary glycoproteins were also incubated with 0.001 U neuraminidase (sialidase) for 2 hr at 37°C, washed, and blocked with BSA (grey columns). Recombinant proteins (100 μg ml −1 ) were incubated with substrata for 2 hr at 37°C and amounts bound measured with anti-tetra-His mouse antibodies and HRP-conjugated anti-mouse IgG antibodies as described in 4. F2, unmodified fragment; F2*, fragment containing NGR, RGT, and AGD motifs all alanine-substituted to AAA. Error bars represent ±SEM from three independent experiments (n = 3). * P < 0.05 compared with corresponding F2 values 3 | DISCUSSION S. gordonii is a component of the normal microbiota of the human oral cavity and plays a pivotal role in the development of microbial communities on the tooth surfaces and gingival crevices (Jenkinson, 2011;Wright et al., 2013). These communities provide a reservoir for bacteria to enter the circulation where they may interact with blood platelets and cause unwanted thrombus generation. This can lead to infective endocarditis, which is characterized by the formation of vegetations on the heart valves (Moreillon & Que, 2004). Significant advances in our understanding of how S. gordonii, and other mitisgroup oral bacteria, interact with platelets have been made in recent years.
In addition to the knowledge about S. gordonii Hsa and GspB proteins (Takahashi et al., 2002;Xiong, Bensing, Bayer, Chambers, & Sullam, 2008), and how they are able to interact with platelet integrin GPIb (Takamatsu et al., 2005), we have demonstrated that PadA surface protein can bind to platelets in a α IIb β 3 -dependent manner  causing dense granule secretion and full platelet spreading . The N-terminal region of 1328 aa residues interacts with platelets (Keane et al., 2013;Petersen et al., 2010), and here, we show that the same region binds Vn, Fn, and salivary pellicle. The PadA C-terminal region comprising 2,318 aa residues has no defined function at this stage. It is thought that this region containing aa residue repeat blocks ( Figure 1) may act as a flexible stalk holding the N-terminal binding-region of the protein out into the environment. C-terminal region repeat-block regions of other S. gordonii surface proteins, such as CshA and Hsa, are reported to provide extended conformations (McNab et al., 1999;Takahashi et al., 2002) that may assist in capture of their ligands even under conditions of shear or flow (Kerrigan et al., 2007).
In the first part of the present study, we focused on the roles of three integrin-like recognition motifs within the N-terminal PadA-F2 fragment (657 aa residues) in the interactions of PadA with platelets.
We showed previously that neither RGT (416)(417)(418) nor AGD (485)(486)(487) were necessary for supporting static or shear-induced platelet adhesion, but both motifs contributed to platelet spreading (Keane et al., 2013). This is in keeping with the data suggesting that Hsa is essential for platelet capture under shear and thus responsible for platelet rolling (Bensing et al., 2004;Kerrigan et al., 2007). Firm adhesion is then complete when platelet integrin α IIb β 3 interacts with PadA . Recent evidence indicates that another motif, NGR, within fibrinogen is responsible for interaction with α IIb β 3 and triggering platelet activation (Moriarty et al., 2015). In light of this discovery, we investigated the role of an identical motif NGR (214)(215)(216) within the N-terminal region of PadA to direct interactions with platelets. In our experiments with platelets that were specifically in resting state, we found that the PadA-F2 fragment bound these only weakly ( Figure 4a). However, the immobilized PadA-F2 region bound TRAP-activated platelets, so this would be in keeping with the notion that PadA preferentially interacts with α IIb β 3 that is in an activated complex with GPIb following Hsa-binding. Confirming our previous studies, we showed that RGT and AGD were not essential for PadA-F2 binding to TRAP-activated platelets. However, binding of the PadA-F2 fragment to activated platelets required the combined activities of NGR, RGT, and AGD, because alanine-substitutions of all of these motifs ablated adhesion (Figure 4b). Moreover, we could not demonstrate that the NGR motif alone was necessary for full platelet spreading, while RGT and AGD motifs clearly contribute to the full spreading process (Keane et al., 2013). In summary, we are nearing the situation in which we could potentially target these three motifs within PadA as a means to controlling unwanted platelet activation by circulating S. gordonii, and by other oral streptococci that express PadA-like proteins.
In this study, we have also utilized gene knockouts and respective remains platelet bound (Parker, Stone, White, & Seinshaw, 1989). Vn forms complexes with plasminogen activator inhibitor-1 (PAI-1) and thus behaves as a physiological inhibitor of active thrombin. Conversely, Vn also appears to stabilize the thrombus and, so, plays a dual role in mediating platelet adhesion (Thiagarajan & Kelly, 1988) and aggregation (Reheman et al., 2005). Vn carries binding sites for heparin, PAI-1, integrins, collagen and plasminogen (Ekmekçi & Ekmekçi, 2006) and also binds α IIb β 3 . Thus, S. gordonii might be able to interact with platelets via Vn, which may be bound to β 3 integrins or to surfaceassociated collagen. Furthermore, bacteria coated with Vn are able to evade the membrane attack complex (MAC) by blocking complement components (C5b-7 complex and C9) and conferring serum resistance (Singh et al., 2010). This of course would be a crucially important property for augmenting systemic survival of bacteria. Our data suggest that binding of streptococcal cells to Vn is dependent on glycan moieties, in particular sialic acid, because sialidase treatment of Vn led to ablation of streptococcal cell adhesion. The major N-linked oligosaccharides of Vn consist of the N-acetyllactosamine type, with a major proportion of them sialylated (Ogawa et al., 1995 Experiments investigating adherence to salivary pellicle and biofilm formation set out to determine if there were roles for the PadA and Hsa proteins in oral cavity colonization outside of pathogenesis within the circulatory system. It is already known that Hsa and GspB bind salivary proteins (Takamatsu, Bensing, Prakobphol, Fisher, & Sullam, 2006), and that Hsa is required for intergeneric coaggregation with Veillonella species (Zhou, Liu, Li, Takahaski, & Qi, 2015). However, this is not sialic acid-dependent, providing further evidence for the presence of additional binding sites within the BR of Hsa. The ability of S. gordonii to adhere to pellicle is antecedent to biofilm formation.
Both PadA and Hsa appear to contribute to biofilm formation and complementation of ΔpadA or Δhsa single knock-out mutants restores the ability to form biofilms in each case. However, only complementation of the ΔpadA Δhsa mutant with PadA restored biofilm formation, not complementation with Hsa. We conclude with an important interpretation that the role of Hsa in biofilm formation is only effective in the presence of PadA. Clearly, therefore, PadA has the ability to mediate S. gordonii binding to salivary pellicle in the absence of sialic acid receptors (Figure 9), and to promote cell-cell interactions that also do not involve sialic acid receptors and result in biofilm formation. This could mean that biofilm formation is a two-step process, and that PadA provides the first step (as opposed to the secondary step in binding platelets). Alternatively, Hsa and PadA may independently contribute to biofilm formation but Hsa is not functionally expressed without the presence of PadA. This might suggest that a complex is formed between PadA and Hsa on the S. gordonii cell surface, and this possibility is currently under investigation.
In conclusion, PadA is a multi-domain adhesin that interacts with activated platelets via α IIb β 3 facilitating firm adhesion, granule release, and full platelet spreading (see Figure 10). These properties depend upon the presence of NGR, RGT, and AGD integrin-like recognition motifs within the N-terminal F2 fragment of 657 aa residues. These motifs are not involved in PadA-F2 fragment binding to cFn, Vn, or salivary pellicle, suggesting that they are more critical to platelet interactions, and that the alanine substitutions did not significantly affect the general adhesion properties of the F2 region. PadA functions in concert with Hsa, mediating firm adhesion of platelets following FIGURE 10 Diagrammatic representation of some of the processes involved in platelet activation by S. gordonii. Cell wall-anchored proteins Hsa and PadA interact with platelet membrane integrins GPIb and α IIb β 3 (GPIIbIIIa). Hsa captures platelets under flow (rolling) by binding GPIb, and possibly also α IIb β 3 , and activates signalling cascades including FcγRIIa phosphorylation, leading to dense granule release (see Arman et al., 2014). PadA binds activated α IIb β 3 , thus amplifying signals leading to shape change, thrombin production, coagulation, and thrombus formation. Platelet activation by S. gordonii can occur in the absence of specific IgG. However, with IgG present, there is evidence for activation (phosphorylation) of spleen tyrosine kinase (Syk-P) through FcγRIIa. Conserved streptococcal surface protein antigens such as antigen I/II proteins (e.g., SspA/B) may also be involved in the overall process (Kerrigan et al., 2007). Physiologically, collagen activates Syk through GPVI, which is closely associated with FcγRIIa (not shown). Fibrinogen engages GPIIbIIIa (α IIb β 3 ), which also associates with FcγRIIa. CD40L (otherwise known as CD154) is up-regulated in the platelet cell membrane and binds CD40 + cells such as endothelial cells and neutrophils, while soluble (released) CD40L further activates platelets capture of platelets by Hsa via GPIb, and firm adhesion to cFn. PadA and Hsa also act cooperatively in mediating binding of bacteria to Vn, salivary pellicle, and biofilm formation. In the latter phenotype, Hsa is clearly reliant on PadA expression to mediate its function. Taken collectively, these results strongly suggest that in S. gordonii at least two cell wall-anchored proteins work in concert to mediate hostinteractive processes relevant to both bacterial colonization and pathogenesis.

| Generation of S. gordonii complemented mutant strains
To complement the S. gordonii UB2723 ΔpadA mutant, the entire padA coding sequence was cloned into pMSP, a derivative of pMSP7517 (Hirt et al., 2000) in which the prgB gene had been replaced by DNA encoding three alanine residues, together with the transcriptional ter- To purify PadA from S. gordonii, plasmid pMSP-padA was used as template in inverse PCR with 5′ phosphorylated primers His.padAR 5′ATGATGATGTGCTCCGTCTTTAATAGATG and His.
padAF CACCACCACTAACTCGAGGAATTAGGTTG to substitute the sequence encoding the PadA wall anchorage motif (LPKTG) with a sequence encoding ×6 His (underlined in primers), followed by a stop codon. Amplicons were ligated following removal of original template by DpnI digestion, and resulting plasmids were transformed into E. coli JM109. Several plasmids were confirmed by sequencing, and then transformed into S. gordonii UB2723 ΔpadA. A representative strain was selected in which padA gene expression was induced with nisin (50 ng ml −1 ), and PadA protein, in the absence of cell wall anchorage motif, was secreted into the growth medium ( Figure S1).

| Site-directed mutagenesis and protein expression
Site-directed mutagenesis of two potential integrin-recognition sites in the F2 N-terminal region of PadA have been previously described (Keane et al., 2013). Briefly, alanine-substitution mutagenesis was performed by the use of mutagenic primers containing base mis- peptide data were filtered to satisfy false discovery rate (FDR) of 5%.
The Proteome Discoverer software generates a reverse "decoy" database from the same protein database, and any peptides passing the initial filtering parameters that were derived from this decoy database are defined as false positive identifications. The minimum crosscorrelation factor (Xcorr) filter was readjusted for each individual charge state separately to optimally meet the predetermined target FDR of 5% based on the number of random false positive matches from the reverse decoy database. Thus, each data set has its own passing parameters.

| Platelet adherence assay
Whole blood was collected from six donors and added to 1.5 ml acid citrate dextrose per 10 ml blood collected. The donors were healthy subjects who had abstained from taking any non-steroidal antiinflammatory drugs (NSAIDs) in the previous 10 days. Informed consent was obtained from all subjects, and the study was approved by the Royal College of Surgeons in Ireland Ethics Committee (REC679b).

| Platelet spreading assay
Poly-L-lysine-coated glass slides were coated with PadA protein fragment (100 μg ml −1 ) for 16 hr at 4°C. Slides were then blocked with 1% BSA; washed and gel-filtered platelets (5 × 10 6 platelets ml −1 ) were allowed to spread on the PadA fragments for 45 min at 37°C. The slides were washed, the platelets were fixed and permeabilized, and then the samples were stained with Alexa 546 phalloidin and examined by confocal microscopy as described in detail elsewhere (Keane et al., 2013). The percentage of platelets spread was calculated from five areas, selected randomly, containing 200 or more total cells.

| Bacterial cell adhesion assays
Extracellular matrix proteins in coating buffer (20 mM Na 2 CO 3 , 2 mM NaHCO 3 , pH 9.3) were added to the wells of a high-binding 96-well plate (Immulon 2HB) and incubated for 16 hr at 4°C. In some experiments, immobilized substrata were then desialylated with 0.001 U neuraminidase from Clostridium perfringens (Sigma) in 0.1 M sodium acetate buffer (pH 5.0) containing 2 mM CaCl 2 , for 2 hr at 37°C. Wells were washed once in TBSC (10 mM Tris-HCl pH 7.6, 0.15 M NaCl, 5 mM CaCl 2 ), and non-specific binding sites blocked with 3% BSA in TBSC containing 0.05% Tween-20 (TBST) for 1 hr at 37°C. Bacterial cultures were grown for 16 hr at 37°C, and cells were harvested by centrifugation (5000 × g, 7 min). Cells were washed once in TBSC, adjusted to OD 600 0.5, and portions (0.1 ml) incubated with the immobilized proteins at 37°C for 2 hr. The suspensions were then removed, and the wells were washed twice in TBS. Adherent cells were fixed with 25% formaldehyde for 30 min at room temperature.
Wells were washed twice in TBS, 0.5% crystal violet was added, and the plates were incubated for 2 min at room temperature. Wells were washed three times in TBS, 10% acetic acid (0.1 ml) was added and after 5 min, levels of adhesion were quantified by measuring absorbance at 595 nm (A 595 ). Previous work has shown that bacterial cell numbers are proportional to A 595 over the range employed in the assays (Jakubovics et al., 2005).

| Salivary pellicle adhesion assay
Saliva was collected from at least six healthy adult human subjects, who provided written informed consent (approved by the National Research Ethics Committee South Central Oxford C. 165 # 08/ H0606/87 + 5). Exclusion criteria were: antimicrobial medication within 7 days previously, continuous medication, gross dental caries, or unstable periodontal disease. Samples were pooled, mixed with 0.25 M dithiothreitol on ice for 10 min and clarified by centrifugation (8000 × g, 10 min). The supernatant was diluted to 10% with sterile water, filter sterilized (0.22 μm pore membrane) and aliquots were stored at −20°C.
Saliva-coated cover slips were prepared by placing sterile 19mm-diameter cover slips (Menzel-Glaser, Braunschweig, Germany) into 12-well plates (Greiner) and adding 10% saliva (1 ml) to each well. The cover slips were incubated at 4°C for 16 hr, washed with PBS, and transferred to a fresh 12-well plate. Bacterial cell suspension (OD 600 0.5) was added to each well as described above, incubated at 37°C for 2 hr, and aspirated from the wells, and the cover slips were washed twice in TBSC. Adherent cells were fixed and stained with crystal violet, and levels of adhesion were quantified by A 595 measurement.

| Biofilm formation
Bacterial cultures grown for 16 hr (10 ml) were harvested by centrifugation (5000 × g, 7 min), washed in modified C (mC) medium (0.25% Difco Proteose Peptone #2, 0.75% yeast extract, 10 mM K 2 HPO 4 , 0.4 mM MgSO 4 .7H 2 O, 17 mM NaCl, pH 7.5, containing 0.2% glucose) and adjusted to OD 600 1.0 in mC medium. Saliva-coated cover slips were prepared as above in 12-well plates (Greiner) before bacterial cell suspension (OD 600 0.5) was added to each well. The plates were incubated at 37°C for 1 hr with shaking at 50 r.p.m. The cell suspensions were removed and the cover slips washed twice in mC medium. Fresh mC medium was then added to each well and the plates were incubated at 37°C for a further 15 hr. Cover slips were then removed, washed twice in PBS, and stained with crystal violet as described above.

| Recombinant protein binding assays
ECM proteins in coating buffer (50 μl) were added to wells of a highbinding 96-well plate (Immulon 2HB) and incubated for 16 hr at 4°C.
Wells were washed once in TBSC and non-specific binding sites were blocked with 3% BSA in TBSC for 1 hr at 37°C. Wells were washed once in TBSC and recombinant protein (0-5 μg) diluted in TBSC was applied to the wells and incubated for 1 hr at 37°C. Unbound protein was removed and wells washed once in TBS. Primary antibody diluted in TBST was added to the wells and incubated for 1 hr at 37°C. Wells were washed twice in TBST before adding HRP-linked secondary antibody diluted in TBST containing 3% BSA and incubating for 1 hr at 37°C.
Wells were washed once in TBST, twice in TBS, and detection reagent (0.102 M Na 2 HPO 4 , 0.049 M citric acid, 0.012% H 2 O 2 , 3.7 mM o-phenylenediamine) was added to wells. Plates were incubated in the dark for 10 min at room temperature, 0.56 M H 2 SO 4 was added to stop the reactions and A 490 measured. x6His-tagged proteins were detected using anti-tetraHis antibody (Qiagen) at 1:1000 dilution and HRP-conjugated anti-mouse antibody (Dako) at 1:2000 dilution. Fibrinogen was detected using rabbit anti-human fibrinogen antibody (Dako) at 1:1000 dilution and HRP-conjugated swine anti-rabbit antibody (Dako) at 1:2000 dilution.