Structural Insights into SraP-Mediated Staphylococcus aureus Adhesion to Host Cells

Staphylococcus aureus, a Gram-positive bacterium causes a number of devastating human diseases, such as infective endocarditis, osteomyelitis, septic arthritis and sepsis. S. aureus SraP, a surface-exposed serine-rich repeat glycoprotein (SRRP), is required for the pathogenesis of human infective endocarditis via its ligand-binding region (BR) adhering to human platelets. It remains unclear how SraP interacts with human host. Here we report the 2.05 Å crystal structure of the BR of SraP, revealing an extended rod-like architecture of four discrete modules. The N-terminal legume lectin-like module specifically binds to N-acetylneuraminic acid. The second module adopts a β-grasp fold similar to Ig-binding proteins, whereas the last two tandem repetitive modules resemble eukaryotic cadherins but differ in calcium coordination pattern. Under the conditions tested, small-angle X-ray scattering and molecular dynamic simulation indicated that the three C-terminal modules function as a relatively rigid stem to extend the N-terminal lectin module outwards. Structure-guided mutagenesis analyses, in addition to a recently identified trisaccharide ligand of SraP, enabled us to elucidate that SraP binding to sialylated receptors promotes S. aureus adhesion to and invasion into host epithelial cells. Our findings have thus provided novel structural and functional insights into the SraP-mediated host-pathogen interaction of S. aureus.


Introduction
The serine-rich repeat glycoproteins (SRRPs) are a family of adhesins encoded by Gram-positive bacteria that mediate attachment to a variety of host cells or bacteria themselves [1]. SRRPs typically consist of a signal peptide at the N-terminus, a short SRR (SRR1, ,50-170 residues), a ligand-binding region (BR, ,250-500 residues) followed by a much longer SRR (SRR2, ,400-4000 residues), and a C-terminal LPXTG motif anchoring to the cell wall [1]. The BRs of SRRPs from different pathogenic bacteria have varying primary sequences and bind to diverse targets from carbohydrates to proteins [1].
In addition to having highly variable sequences, the BRs from different bacteria are composed of distinct modules. The diversity of BR modules and combinations contributes to the multiple functions of SRRPs. The only four known BR structures to date have identified five distinct modules [2][3][4]. The BR of Streptococcus parasanguinis Fap1 contains two modules: an N-terminal helical module and a C-terminal CnaA module [2], whereas Streptococcus gordonii GspB has a BR of three modules: CnaA, Siglec and a unique module of unknown function [3]. In addition, the recently reported BR structures of the two SRRP paralogs (Srr1 and Srr2) from Streptococcus agalactiae defined two immunoglobulin-fold modules, which specifically bind to the host fibrinogen [4]. However, the module composition and corresponding molecular functions of most BRs remain unknown, which largely impedes the understanding of the pathogenesis mechanism of SRRPs.
S. aureus is a human pathogen that causes a wide range of debilitating and life-threatening infections [5]. S. aureus encodes a 2,271-residue SRRP termed serine-rich adhesin for binding to platelets (SraP), that is involved in the pathogenesis of infective endocarditis [6]. Moreover, the BR (residues Phe245-Asn751) of SraP, termed SraP BR , mediates intraspecies interaction and promotes bacterial aggregation [7]. We determine the 2.05 Å crystal structure of SraP BR , revealing a rod-like tandem organization of four discrete modules: a legume lectin-like module, a module with a b-grasp fold, and two tandem cadherin-like modules that create the rigid stem of SraP BR . Further structural and biochemical analyses reveal that the legume lectin-like module specifically binds to N-acetylneuraminic acid (Neu5Ac), which may mediate adhesion to host sialylated receptors. These findings increase our knowledge of the diverse BR modules of SRRPs, and provide structural insights into a novel surface protein that mediates interaction of S. aureus with host epithelial cells.
Each asymmetric unit of the final model at 2.05 Å resolution contains a single SraP BR molecule of residues Thr251-Asn751. The N-terminal residues Phe245-Thr250 are not visible due to their poor electron density. SraP BR folds into a slightly bent, rodlike structure of 160 Å in length that has four discrete modules: a head-like N-terminal module followed a stem of three all-b modules (Fig. 1A). All modules have a dominant b-strand secondary structure. During the model building and refinement process, three peaks of electron density at the 24 s level were observed in the |Fo|2|Fc| Fourier difference map, indicating the presence of three metal ions. Atomic absorption spectroscopy assigned these metal ions to Ca 2+ ; we thus termed them Ca-1, Ca-2 and Ca-3 accordingly. The structure also contains a sucrose and a 2-(N-morpholino)-ethanesulfonic acid (MES) molecule, which were introduced from the cryoprotectant and crystallization buffer, respectively.
The N-terminal module adopts a jelly-roll fold with a bsandwich (b1-b17) core architecture of two antiparallel b-sheets packed against each other (Fig. 1B). Beyond the core structure, two a-helices (a1 and a2) pack on either side of the lateral of the bsandwich and partially seal the hydrophobic lateral openings between the two b-sheets. A molecule of sucrose binds to the protruding loops at the distal end of the N-terminal module (Fig. 1B). A structural similarity search using the DALI server [8] revealed that this module is most similar to legume lectins, despite sharing a sequence identity of #20%. The top hits include lectins from legume plants such as Pisum sativum (PDB 2BQP) [9] and Robinia pseudoacacia (PDB: 1FNY) [10], with a Z-score of 22-23 and root mean square deviation (RMSD) of 2.3-2.5 Å over ,200 Ca atoms. Thus, we termed the N-terminal module L-lectin.
The C-terminal tandem repeat of two modules share a sequence identity of 56% and a similar overall structure with an RMSD of 0.81 Å over 81 Ca atoms. The two modules are linked in a head-to-tail fashion, indicating duplication of the coding region during evolution. Each module consists of a b-sandwich of three b-sheets ( Fig. 1B & 1C). Structural analysis revealed that the two modules resemble eukaryotic cadherins of known-structure (PDB 3UBF, Zscore 8.6, RMSD 2.3 Å over 85 Ca atoms and PDB 4APX, Zscore 9.5, RMSD 2.3 Å over 87 Ca atoms) [12,13]. Thus the tandem cadherin-like (CDHL) modules were termed CDHL-1 and CDHL-2, respectively.

Specific binding of the L-lectin module to N-acetylneuraminic acid
The L-lectin module is structurally similar to legume lectins ( Fig. 2A), a large family of carbohydrate-binding proteins with diverse activities [14]. Legume lectins commonly coordinate a Ca 2+ in addition to a transition metal ion, usually Mn 2+ [15]. In contrast, the L-lectin module has only a Ca 2+ -binding site. At the apex of the L-lectin module, the Ca 2+ (named Ca-1) was embedded in a 7-coordinate geometry (Fig. 2B). The seven coordinates are from the side-chain oxygen atoms of Asp365 (bidentate coordination from Od1 and Od2), Asp382 (Od2), Asn369 (Od1), the main-chain oxygen atom of Tyr367 and two water molecules (Wat1 and Wat2). The two water molecules were further stabilized by the main-chain oxygen atoms of Asp330,

Author Summary
Staphylococcus aureus is an important pathogen that causes a range of human diseases, such as infective endocarditis, osteomyelitis, septic arthritis and sepsis. The increasing resistance of S. aureus to most of the current antibiotics emphasizes the need to develop new approaches to control staphylococcal infections. As a surfaceexposed serine-rich repeat glycoprotein (SRRP), S. aureus SraP is involved in the pathogenesis of infective endocarditis via its ligand-binding region (BR) adhering to human platelets. However, little is known about how SraP interacts with its host receptor(s). Through structural and functional analyses of the BR domain, we have discovered a specific binding of SraP to N-acetylneuraminic acid (Neu5Ac), in agreement with a recent report of the trisaccharide ligand of SraP. Further mutagenesis analysis showed that SraP binding to Neu5Ac and the trisaccharide promotes S. aureus adhesion to and invasion into host epithelial cells. These findings increase our knowledge of surface protein mediated interaction of S. aureus with host epithelial cells.
A molecule of sucrose is fixed by a cluster of loops protruding from the apex of the L-lectin module. The sucrose molecule has a ''bent-back'' conformation with the glucose and fructose moieties perpendicular to each other (Fig. 2B). The glucose moiety is inserted in the pocket, and adopts a conformation nearly parallel to the two b-sheet layers. The sugar ring makes hydrophobic interactions with Tyr367 and the morpholine ring of MES, whereas the hydroxyl groups are stabilized by hydrogen bonds with Ala478-Na, Asp330-Od1, and three water molecules (Wat3-5), which are further fixed by residues Asp330, Asn347 and Ala349 (Fig. 2B). In contrast, the solvent-exposed fructose moiety is bent through interactions with MES and two water molecules (Wat6 and Wat7) (Fig. 2B). The sucrose binding residues, especially the stacking residue Tyr367, are structurally conserved in the legume lectins (Fig. 2C).
To identify the favored saccharide of SraP, we first detected the binding affinity of the L-lectin module towards eight common monosaccharides using the surface plasmon resonance assays. Among these monosaccharides, only Neu5Ac bound to the Llectin module (Fig. 2D). In consequence, we determined that the L-lectin module has an equilibrium dissociation constant (K d ) of 0.54 mM towards Neu5Ac (Fig. 2E), comparable to previously reported values of legume lectins [16].
Afterwards, we attempted to obtain the Neu5Ac-complexed structure without success. Therefore, we docked Neu5Ac to the structure of the L-lectin module with the sucrose-binding pocket as the search grid. Neu5Ac was docked at a position overlapping the glucose moiety of sucrose, with a shift of ,1.5 Å towards Ca-1. In the model, Neu5Ac is stacked against Tyr367 via hydrophobic interactions, and makes direct polar interactions with the side chains of Ser293, Asn347, Tyr367 and Asn369 and the Na atoms of Gly477 and Ala478 (Fig. 2F).

Structural similarity of the b-GF module to Ig-binding proteins
The b-GF module adopts a ubiquitin-like b-grasp fold in the Igbinding superfamily [11]. DALI search [8] suggested the module resembles the B1 domain of mucus-binding protein type 2 repeat Mub-R5 from Lactobacillus reuteri [17] (PDB code 3I57, Z-score 9.1, RMSD 1.6 Å , over 68 Ca atoms), and protein L (PpL) from Peptostreptococcus magnus [18] (PDB code 1HEZ, Z-score 4.5, RMSD 2.5, over 55 Ca atoms). The B1 domain belongs to a family of Igbinding proteins [19] that have a core structure of an a-helix packed against a four-stranded b-sheet [17,20]. The major differences are from the helix a3 and the two lateral b-strands (b19 and b21). In addition, the b-GF module of SraP BR contains two extra b-strands, b20 and b22, which are substituted by loops or a-helix extensions in the B1 domain (Fig. 3). Mub-R5 interacts in vitro with a large repertoire of mammalian Ig proteins including secretory IgA, whereas PpL binds to the V L domain of Ig k chain [19,21]. Complex structures indicated that formation of a b-zipper is necessary for the binding of PpL to the V L domain of Ig k chain [18] and IgG or IgM [22]. However, the corresponding b-strands b19 and b21 are much shorter in the b-GF module, which might not be capable of forming a b-zipper.
The small-angle scattering of X-rays (SAXS) is usually applied to address the flexibility and conformational states of biological macromolecules in solution [24]. To explore the role of Ca 2+ , we used SAXS to compare the overall structure of SraP BR in the presence or absence of Ca 2+ . A difference in wide-angle scattering curves of SAXS indicated that SraP BR adopts different conformations with or without Ca 2+ . Compared to the Ca 2+ -free form, the envelope of Ca 2+ -bound SraP BR correlates much better with the SraP BR crystal structure (Fig. 5A). The larger discrepancies of the Ca 2+ -free SraP BR are mainly resulted from the tandem CDHL modules, indicating that binding of Ca 2+ makes the tandem CDHL modules more rigid, and thereby facilitates the extended conformation of SraP BR in solution.
We also performed molecular dynamics simulations of SraP BR in the presence or absence of Ca 2+ . Ca 2+ -bound SraP BR remains extended and shows slight conformational changes over the simulation time, whereas removal of Ca 2+ caused the curling of SraP BR (Fig. 5B). Geometric analysis of the Ca 2+ -coordinating residues revealed larger fluctuations at the two Ca 2+ -binding junctions of Ca 2+ -free SraP BR , suggesting the importance of Ca 2+ for the structural integrity of SraP BR . Moreover, the increased RMSD values indicated the two junctions in Ca 2+ -free SraP BR undergo dramatic conformational changes. Given the CDHL-2 modules superimposed, the other three modules adopt a more curved conformation and are projected to the opposite side upon the loss of Ca 2+ (Fig. 5B).  The rod-like, four-module structure of SraP BR has three junctions with a buried interface of 700, 400 and 220 Å 2 , respectively (Fig. 1B). The interface between L-lectin and b-GF is maintained by several hydrogen bonds including Asn257-Glu493, Arg303-Asn519, Arg303-Ser494, Leu341-Asn519, and Asn466-Asn562 (Fig. 5C). The second interface is formed by residues from b-GF and CDHL-1 via hydrogen bonds of Lys536-Asn602, Gly537-Asn602 and Asp503-Thr604 (Fig. 5D). At the third junction, Glu587 and Val588 of CDHL-1 form two hydrogen bonds with Asn691 of CDHL-2 (Fig. 5E). The second and third interfaces are relatively small; however, they are more stable due to the contribution from coordinate bonds of Ca-2 to Asn602 and Ca-3 to Asn691.
To further investigate the plasticity of these junctions, we determined three crystal structures for each of two consecutive modules: L-lectin&b-GF (Phe245-Lys575), b-GF&CDHL-1 (Ser494-Thr663), and CDHL-1&2 (Ala576-Asn751). We superimposed each of these three structures over the corresponding two modules of the full-length SraP BR structure, always with the Cterminal modules aligned. The results revealed slight twisting and/or translation with intermodule angle changes of 9.4u, 11.7u and 14.6u for the three junctions, respectively (Fig. 5C,E). In detail, residues Val496-Gln498 in L-lectin&b-GF, Phe571&Thr572 in b-GF&CDHL-1, and Asp661-Thr663 in CDHL-1&2 undergo slight conformational changes. Except for a short disordered segment at the N-terminus of CDHL-1&2 (Fig. 5E) due to the deletion of two Ca-2 coordinate residues Asp573 and Lys575, we did not find secondarystructure change in any module pairs. Moreover, merging these three structures via sequential alignment of the same module resulted in a total twist of 5.4u along the axis of SraP BR structure (Fig. S1), suggesting that the slight interdomain twists of the three junctions are randomly occurred and could be canceled out. Together, the results indicated that in the presence of calcium, SraP BR adopts a relatively rigid rod-like structure under all conditions tested.

SraP promotes S. aureus adhesion and invasion to A549 cells through sialylated receptors
Bacterial attachment and colonization at the surface of host cells have been thought to be mediated by specific binding of BRs to glycoconjugates [1]. It has recently been reported that SraP binds to the salivary agglutinin gp340 via the Neu5Ac moiety of the trisaccharide Neu5Aca(2-3)Galb(1-4)GlcNAc [25]. In fact, gp340 and homologs are also expressed in lung epithelial cells [26]. To determine if the L-lectin module mediates SraP BR adhesion, we incubated a monolayer of human lung epithelial A549 cells with green fluorescent protein (GFP)-fused SraP BR and individual modules. The results indicated that full-length SraP BR and the Llectin module, but not the CDHL-1&2 modules or the b-GF module, specifically adhered to the A549 cells (Fig. 6). Tyr367 is a key residue to make hydrophobic interactions with sucrose and the docked Neu5Ac, we thus constructed the Y367G mutant proteins to test the adhesion to A549 cells. As shown in the CD spectra, the mutation of Y367G did not introduce significant structural changes to SraP BR or the L-lectin module (Fig. S2). In contrast, a Y367G mutation in both the full-length SraP BR and the L-lectin module almost completely abolished the adhesion capacity. Moreover, the addition of 5 mM Neu5Ac completely inhibited the adhesion of SraP BR to the A549 cells (Fig. 6). Quantification of the fluorescent signal of three representative frames for each image further confirmed that only the full-length SraP BR and the L-lectin module are capable of specific binding to A549 cells (Fig. S3). These results demonstrate that the adhesion of SraP BR to A549 cells is mediated by the specific recognition of the L-lectin module towards Neu5Ac.
Furthermore we performed comparative assays of bacterial adhesion and invasion to A549 cells using S. aureus strain NCTC 8325 and an isogenic DsraP mutant. Deletion of sraP resulted in an approximately 40% decrease in adhesion, as compared to the wild-type (Fig. 7A). As a result the level of invasion was decreased by ,50% (Fig. 7B). These results indicate that SraP contributes to S. aureus adhesion to and invasion into host cells.
Together with the recently reported SraP ligand, the trisaccharide Neu5Aca(2-3)Galb(1-4)GlcNAc [25], our results strongly suggested that the specific binding of SraP BR to the ligand promotes the S. aureus adhesion to host cells. We initially tried to determine the complex structure of the trisaccharide with SraP BR or the L-lectin module without success. As an alternative, we docked the trisaccharide to the structure of the L-lectin module (Fig. 7C). In the docking model, the Neu5Ac moiety of the trisaccharide adopts an almost same position to that of Neu5Ac (Fig. 2F). In addition, the moieties of Gal and GlcNAc are stabilized via hydrogen bonds by residues Asn374 and Ser371, respectively (Fig. 7C).
To determine whether the interactions between the L-lectin module and the trisaccharide contribute to the SraP BR -mediated adhesion, we used neuraminidase, b-galactosidase or N-acetylglucosaminidase to differentially remove Neu5Ac, Gal, and GlcNAc from the surface of A549 cells (Fig. 7D). Treating A549 cells with neuraminidase alone resulted in an approximately 36% decrease in the adhesion of S. aureus NCTC 8325. In contrast, the digestion with either b-galactosidase or N-acetylglucosaminidase did not significantly affect the adhesion. Further addition of b-galactosidase and N-acetylglucosaminidase to the neuraminidase-treated A549 cells did not significantly lower the adhesion level (Fig. 7D, the 6 th and 7 th columns). These results indicated that Neu5Ac is the moiety at the non-reducing end of the trisaccharide, and a major receptor to the SraP-mediated bacterial adhesion of A549 cells. Consistently, the wild-type S. aureus and DsraP mutant showed comparable levels of adhesion to the neuraminidase-treated epithelial cells ( Fig 7D, the 4 th and 8 th columns). The adhesion level of the wild-type S. aureus to the neuraminidase-treated A549 cells (Fig. 7D, the 4 th column) is similar to that of the DsraP mutant to the untreated A549 cells (Fig. 7A). We thus concluded that SraP is the major S. aureus adhesin that recognizes the sialylated host receptors.

Discussion
The specific recognition of the L-lectin module to sialylated receptors may be a universal mechanism for staphylococcal adhesion to host cells Structural analyses combined with epithelial adhesion experiments demonstrate that SraP plays an important role in mediating bacterial adhesion to host cells by recognizing the L-lectin module. Legume lectins are a large family of proteins primarily found in the seeds of legume plants that have a similar fold but distinct carbohydrate-binding specificities [14]. They typically adopt a quaternary structure of dimers or tetramers that enhances sugar binding specificity or affinity [27]. The architecture of the legume lectin fold has also been found in the animal calcium-dependent lectin ERGIC-53/MR60, a mannose-binding protein involved in the export of soluble glycoproteins from the endoplasmic reticulum [28]. We report here the first legume lectin foldcontaining protein in bacteria. This prokaryotic monomeric Llectin module mediates adhesion to host cells by recognizing Neu5Ac, usually the non-reducing terminal residue of glycoconjugates of extracellular receptors. Bioinformatics analysis suggested that the L-lectin module might also exist in other proteins in Staphylococci and Streptococci (Fig. S4A). Moreover, the residues involved in Neu5Ac binding are relatively conserved among these putative L-lectins (Fig. S4B). Therefore, we hypothesize that this adhesion mechanism mediated by the L-lectin module operates in other staphylococcal and streptococcal species.
The L-lectin module is projected outwards mainly by two cadherin-like modules To scan the host receptors, the L-lectin module should be projected outwards from the bacterial surface. In addition to a long region of SRR2 that crosses the bacterial cell wall, SraP has a relatively rigid stem of three modules: a b-GF and two CDHL modules. The classic cadherins in vertebrates bridge the intermembrane space between neighboring cells by forming trans-adhesive homodimers through membrane-distal extracellular domains [29]. The extracellular domain of most cadherins often contains three conserved Ca 2+ -coordination sites at the interdomain junction [30]. In contrast, each CDHL module in SraP BR binds to only a single Ca 2+ that does not superimpose on any of the three Ca 2+ ions in eukaryotic cadherins. Nevertheless, multiplesequence alignments suggested strong conservation among Grampositive bacteria of SraP BR Ca 2+ -binding residues (Fig. 8A).
Unlike the monomeric form of SraP BR , the recombinant CDHL1&2 exists as a dimer in solution as confirmed by sizeexclusion chromatography and chemical cross-linking assays (Fig.  S5). The structure revealed that the homodimer of CDHL1&2 buries an interface area of ,600 Å 2 /per subunit, as calculated by PISA (http://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver) server [31]. The interface is mainly stabilized by polar interactions, in contrast to the hydrophobic dimeric interfaces of eukaryotic cadherins [29]. CDHL-2 from one subunit packs against the junction between CDHL-1 and CDHL-2 from the symmetric subunit, and vice versa (Fig. 8B). This dimerization pattern may explain the result that SraP mediates intraspecies interaction and promotes aggregation of S. aureus ISP479C [7]. However, we did not observe significant decrease of aggregation or biofilm formation upon the deletion of sraP in S. aureus NCTC 8325. The different results might be due to the variation of S. aureus strains. Sequence homology search indicated that some surface proteins in Gram-positive bacteria contain two or more CDHL modules, suggesting that those proteins with multiple CDHL modules have a high potential to mediate intraspecies aggregation.

SraP BR possesses a unique functional modularization pattern
SRRPs have been identified in a variety of Gram-positive bacteria and function as virulence factors in a wide spectrum of infections [1]. The diversity of these infections (e.g. endocarditis, meningitis and pneumonia) correlates with the variability of BRs. Furthermore, the few BR ligands that have been identified to date range from carbohydrates (such as sialyl-T antigen) [32] to proteins (such as keratins) [33,34]. In addition to the five distinct modules defined in the four previously reported BR structures [2][3][4], our SraP BR structure identified three types of unique modules. Notably, although the b-GF module resembles the Ig-binding proteins, such as B1 domain of Mub-R5 and PpL [19,21], the binding of the b-GF module towards human IgG, IgA or IgM could not be detected, in agreement with our structural analysis. We thus propose that the b-GF module, in addition to the two CDHL modules, functions as a relatively rigid stem to project the L-lectin module outwards.
The two structure-known SRRPs, Fap1 and GspB, undergo significant inter-module angle changes, which are regulated by pH and ligand binding, respectively [2,3]. In contrast, SraP BR appears to adopt a relatively rigid, rod-like conformation when colonizing a host, for the concentration of free Ca 2+ in the extracellular space or blood is stringently maintained at 1.1-1.3 mM [35]. The rigid architecture of SraP BR enables the globular L-lectin module to extend outwards from the bacterial surface for scanning host receptors. This strategy to expose the functional modules has been observed for other bacterial adhesins such as the fibrillar antigen I/II from S. mutans [36], and the rod-like surface protein SasG from S. aureus [37].

Cloning, expression, and purification of SraP BR
The genomic DNA from Staphylococcus aureus NCTC 8325 was prepared for gene cloning. The DNA sequences (GeneBank, the accession number of YP_501439.1) encoding SraP BR (Phe245-Asn751) and other SraP truncates were cloned into pET28a with an N-terminal His 6 -tag or pET28a with a C-terminal GFP-tag, respectively. The constructs were overexpressed in E. coli strain BL21 (Novagen) using LB culture medium (10 g NaCl, 10 g Bacto-Tryptone, and 5 g yeast extract per liter). The cells were Figure 6. Contribution of the L-lectin module to bacterial adhesion to human lung epithelial cells. A549 cells were incubated with GFPfused SraP BR (termed BR for short) and truncations (L-lectin, b-GF and CDHL-1&2) or mutants (BR Y367G and L-lectin Y367G ), respectively. GFP is used as a negative control. The nuclei were stained with DAPI (blue) and the adhered proteins were detected with anti-GFP mouse IgG, followed by FITC conjugated goat anti-mouse IgG (green). doi:10.1371/journal.ppat.1004169.g006 grown at 37uC to an OD 600nm of 0.6. Expression of the recombinant protein was induced with 0.2 mM isopropyl b-D-1thiogalactopyranoside (IPTG) at 16uC for another 20 hr before harvesting. Bacteria were collected by centrifugation at 8,0006g for 10 min and resuspended in 30 ml lysis buffer (20 mM Tris-Cl, pH 8.8, 100 mM NaCl). After sonication for 2.5 min followed by centrifugation at 12,0006g for 25 min, the supernatant containing the His-tagged protein was collected and loaded onto a Ni-NTA column (GE Healthcare) equilibrated with the binding buffer (20 mM Tris-Cl, pH 8.0, 100 mM NaCl). The target protein was eluted with 300 mM imidazole, and loaded onto a Superdex 200 column or Superdex 75 column (GE Healthcare; 20 mM Tris-Cl, pH 8.0, 100 mM NaCl). The purity of protein was assessed by electrophoresis and the protein sample was stored at 280uC.
The selenium-Met (SeMet) labeled L-lectin&b-GF protein was expressed in E. coli strain B834 (DE3) (Novagen). Transformed cells were grown at 37uC in SeMet medium (M9 medium with 25 mg/ml SeMet and the other essential amino acids at 50 mg/ml) containing 30 mg/ml kanamycin until the OD 600nm reached 0.6, and were then induced with 0.2 mM IPTG at 16uC for 20 hr. SeMet substituted protein was purified with the same procedure as the native protein.

Crystallization, data collection and processing
All crystals were grown using the hanging drop vapor diffusion method, with a drop of 1 ml protein solution mixed with 1 ml of reservoir solution equilibrated against 500 ml of the reservoir solution. The proteins for crystallization were concentrated by ultrafiltration (Millipore Amicon) to 30, 38, 20 and 20 mg/ml for the full-length SraP BR , L-lectin&b-GF, b-GF&CDHL-1 and CDHL-1&2, respectively. The SeMet substituted L-lectin&b-GF protein for crystallization was concentrated to 38 mg/ml. Crystals of SraP BR were grown at 28uC, whereas others were grown at 16uC. Crystals were obtained from 0.  All crystals in the cryoprotectant were flash-cooled with liquid nitrogen prior to X-ray diffraction. Data for a single crystal were collected at 100 K in a liquid nitrogen stream using beamline 17U with a Q315r CCD (ADSC, MARresearch, Germany) at the Shanghai Synchrotron Radiation Facility (SSRF). All diffraction data were integrated and scaled with the program HKL2000 [38].

Structure determination and refinement
The crystal structure of L-lectin&b-GF was determined using single-wavelength anomalous dispersion (SAD) phasing [39] method from a single crystal of SeMet-substituted protein to a maximum resolution of 2.10 Å . The AutoSol program implemented in PHENIX [40] was used to locate the selenium atoms and calculate the phase, which was further improved with the program Buccaneer [41]. Automatic model building was carried out using Autobuild in PHENIX. The initial model was refined in REFMAC5 [42] and Phenix.refine and rebuilt interactively using the program COOT [43]. The model was used as the search model against 2.05 Å SraP BR data by molecular replacement using Molrep program as part of CCP4i [44] program suite. Electron density maps showed clear features of secondary structural elements for automatically building the C-terminal tandem cadherin-like modules using IPCAS [45]. The structures of b-GF&CDHL-1 and CDHL-1&2 were determined by molecular replacement using the corresponding modules in the fulllength SraP BR structure as the search model. The initial models were refined by simulated annealing using Phenix.refine to reduce the phase bias. Then the models were refined interactively using COOT and REFMAC5 until the R-factor and R-free values converged. All final models were evaluated with the programs MOLPROBITY [46] and PROCHECK [47]. Crystallographic parameters were listed in Table 1. The |Fo|-|Fc| omit map of the sucrose molecule contoured at 3 s was calculated by FFT implemented in CCP4i. All structure figures were prepared with PyMOL (http://www.pymol.org/).

Determination of metals binding to the protein
The purified SraP BR , CDHL-1&2, BR Y367G and L-lectin Y367G in 20 mM Tris-Cl, pH 8.0, 100 mM NaCl were concentrated to 30, 38, 30 and 21 mg/ml, respectively, and applied to the analyses. Briefly, 500 ml of protein sample was subjected to digestion by the aqueous method using the HNO 3 and HClO 4 (4:1, v/v) method. Afterwards, the digested samples were diluted with deionized water and analyzed by atomic absorption spectroscopy (Atomscan Advantage, Thermo Ash Jarell Corporation, USA).

Surface plasmon resonance (SPR) assays
The binding affinities of the L-lectin module towards varying monosaccharides were determined by SPR. SPR experiments were performed at 25uC using a Biacore 3000 instrument using HBS (10 mM HEPES, pH 7.5, 150 mM NaCl) containing 0.005% (v/v) Tween 20 and a flow rate of 5 ml/min. The L-lectin module was covalently immobilized on the carboxymethyldextran surface of the CM5 chip. The chip was activated with EDC (N-ethyl-N- [3dimethylaminopropyl] carbodi-imide)/NHS (N-hydroxysuccinimide) solution, and the L-lectin module in 10 mM acetate buffer (pH 5.5) was injected into the flow channel. At the end, the sensor surface was blocked with 1 M ethanolamine. The blank channel was treated in the same way without protein injected. Each monosaccharide in the running buffer was incubated for 1 min in the flow-cells using the kinject mode. Both injection and dissociation steps last for 5 min. The sensor surface was regenerated with 50 mM NaOH. All analyses were performed with the BIAeval software. The equilibrium responses were plotted versus monosac-charide concentrations and fitted to a 1:1 Langmuir binding model using the Origin 8.0 software (OriginLab Corp.).

Computational docking
The docking of Neu5Ac or the trisaccharide Neu5Aca(2-3)Galb(1-4)GlcNAc to the L-lectin module of SraP BR was performed with AutoDock Vina software (version 1.0) [48], which uses a unique algorithm that implements a machine learning approach to its scoring function. This docking allowed a population of possible conformations and orientations for the ligand at the binding site to be obtained. Using AutoDock Tools (ADT) 1.5.4 [49], polar hydrogen atoms were added to the Llectin structure, and its non-polar hydrogen atoms were merged. The protein and ligands were converted from a PDB format to a PDBQT format. All single-bonds within Neu5Ac were set to allow rotation. A grid box covering the entire sucrose-binding site was used to place Neu5Ac freely. The results were sorted by binding affinity and visually analyzed using PyMOL.

Chemical cross-linking
Chemical cross-linking of purified CDHL-1&2 was performed using formaldehyde and bis(sulfosuccinimidyl) suberate (BS 3 ), which is a homobifunctional sulfo-N-hydroxysuccinimide ester analog with a spacer arm length of 1.14 nm (Pierce). Briefly, for formaldehyde cross-linking assays of CDHL-1&2, 20 ml of the recombinant protein (2 mg/ml) was mixed with 20 ml PBS containing 2% formaldehyde, and the samples were incubated at 25uC for 30 min and 1 hr, respectively. For BS 3 cross-linking assay of CDHL-1&2, 100 ml recombinant protein (2 mg/ml) was incubated with 5 mM BS 3 at 25uC for 30 min and 1 hr, respectively. The reaction was quenched by the addition of 20 mM Tris-HCl, pH 8

Bacterial strains, media and growth conditions
Staphylococcus aureus NCTC 8325 and its derivative strains were grown in LB medium, and when necessary, erythomycin (2.5 mg/ ml) and chlorampenicol (15 mg/ml) were added. To generate the insertion-deletion mutagenesis, approximately 500 bp of upstream and downstream fragments of the BR region of sraP gene was amplified by polymerase chain reaction (PCR), digested with PstI/ SalI and BamHI/XbaI, respectively, and then ligated to either end of a double-digested (BamHI/SalI) erythromycin-resistance gene (Erm R ), which was amplified from plasmid pEC1. The three fragments were ligated with the erythromycin-resistance gene in the middle, and then cloned in the temperature-sensitive shuttle vector pBT2. The resulting plasmid was transformed by electroporation into S. aureus strain RN4220 for propagation, and then transformed into S. aureus NCTC 8325 for allelic exchange. Mutants were screened and further checked by PCR and sequencing.

Cell adhesion and invasion assays
Adhesion and invasion experiments were performed as described previously [50]. A549 human respiratory epithelial cells were grown in Dulbecco's modified Eagle's medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS), 5 mM glutamine, penicillin (5 mg/ml) and streptomycin (100 mg/ml). Approximately 5610 5 cells were seeded into 24-well tissue culture plates, and allowed to grow in 5% CO 2 at 37uC. Before use, the monolayers were washed twice with PBS. For adhesion assays, S. aureus grown to OD 600nm of 0.5 in LB medium were resuspended in cell culture medium without serum. Bacteria were diluted to a concentration of 2610 7 CFU/ml and were used to infect the confluent cell monolayers at 37uC for 1 hr. After incubation, the infected monolayers were washed five times with PBS to remove non-adhered bacteria, and treated with 200 ml trypsin (2.5 mg/ml) at 37uC for 3 min to release the adhered bacteria. A549 cells were lysed with 0.05% Triton X-100. The number of adhered bacteria was determined by plating serial dilutions of the recovered bacterial suspensions onto LB agar. For the deglycosylation of A549 cells, The cells were incubated with DMEM media containing 0.008 units ml 21 purified Clostridium perfringens neuraminidase (Sigma), 40 nM purified b-galactosidase (Spr0565) from S. pneumoniae or 0.0025 units ml 21 of S. pneumoniae b-N-acetylglucosaminidase (Sigma) at 37uC for 4 hr in 5% CO 2 [34].
For invasion assays, the bacteria in each well, after incubated in 0.5 ml DMEM for 1 hr, were incubated in 1 ml fresh DMEM containing 14 mg/ml of gentamicin (Sigma) for another 1 hr. Cell monolayers were washed three times with sterile PBS and lysed with 0.05% Triton X-100. The internalized bacteria were counted by plating serial dilutions of the recovered bacterial suspensions onto LB agar. Experiments were performed in triplicate. Data corresponding to adhesion and invasion were compared using the Mann-Whitney tests. Statistical differences were determined with the t-test.

Immunofluorescence assays
Immunofluorescence assays were performed as described previously [51]. Briefly, GFP and GFP-fused proteins at 10 mM were suspended in the media in the absence of serum and antibiotics, and incubated on ice for 2 hr with the monolayer of A549 cells grown on the 12-mm glass coverslips. For inhibition assays, Neu5Ac at 5 and 10 mM was pre-incubated with GFPfused SraP BR protein, respectively. After incubation, cells were sequentially washed three times with cold PBS, fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100 for 2 min, and incubated with GFP-tag mouse antibody and a fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG antibody. The nuclei were stained with 49,6-diamidino-2-phenylindole (DAPI) reagent. Slides were examined with a Zeiss LSM710 confocal scanning fluorescence microscope (Carl Zeiss, Jena, Germany) with a Plan-Apocromat 206/0.8NA objective. Confocal parameters set for immunofluorescence detection were taken as standard settings. The excitation wavelength is 405 nm and the emission wavelength is 410-492 nm. The confocal images were collected and processed with the software ZEN 2009. Experiments were performed in triplicate, with three or more replicate wells tested for each experimental condition.

Molecular dynamics simulations
Molecular dynamics simulations were performed for SraP BR in the Ca 2+ -bound and the Ca 2+ -free states, respectively. For each simulation, the system was placed in a TIP3P [52] water box with a distance of at least 12 Å to the box boundaries. Ions were added to neutralize the system and to result in a concentration of 0.15 M NaCl. The solvated protein was subjected to energy minimization employed the steepest descent algorithm and conjugate gradient, respectively. Simulations were performed with a parallel implementation of the GROMACS (version 4.5.5) package [53] using the AMBER03 force field [54]. MD productions were run for 20 ns using a time step of 2 fs and the NPT ensemble [55]. Covalent bonds were constrained using the LINCS algorithm [56], while the cutoff distances for the Coulomb and van der Waals interactions were set to 0.9 and 1.4 nm, respectively. The long-range electrostatic interactions were treated by the PME algorithm [57] with a tolerance of 10 25 and an interpolation order of 4. Structure visualization was performed with VMD [58].

Small-angle X-ray scattering (SAXS) experiments
SAXS was used to investigate the overall conformations of SraP BR in the presence or absence of calcium. The full-length SraP BR at 1.0 and 5.0 mg/ml was analyzed, either in 10 mM calcium chloride or in 50 mM EDTA. SAXS data were collected on the 12ID beamline of Advanced Photon Sources (APS) at the Argonne National Laboratory using the Pilatus 2M detector (DECTRIS, Switzerland). The scattering patterns were measured with a 1-2 second exposure time for each collected frame, and twenty frames were taken for each sample to optimize the signalto-noise ratio. To reduce the radiation damage, a flow cell made of a cylindrical quartz capillary with a diameter of 1.5 mm and a wall of 10 mm was used during the data collection process. No concentration effect was observed. All SAXS curves were measured at the room temperature over the range of momentum transfer 0.006,s,0.82 Å 21 (where s = 4p sin(h)/l, 2h is the scattering angle, and the X-ray wavelength l is 1.033 Å ). The data were processed using the PRIMUS [59] program package and standard procedures. The forward scattering (I(0)) and the radius of gyration (R g ) were evaluated using the Guinier approximation assuming that at very small angles (s,1.3/R g ) the intensity is represented as I(s) = I(0)exp(2(sR g ) 2/3 ).
The program GNOM [60] was used to calculate D max and the interatomic distance distribution function p(r). The particle shape of each measured sample was reconstructed ab initio using the programs DAMMIN [61] and GASBOR [62]. The scattering patterns of the atomic crystal structures for SraP were calculated using the program CRYSOL. For the ab initio analyses and modeling, multiple runs were performed to verify the stability of the solution. Figure S1 The intermodule twist along the axis of SraP BR . The merged structure of SraP BR (in orange) was generated by sequentially superimposing the same module against each other from the three structures (L-lectin&b-GF, b-GF&CDHL-1 and CDHL-1&2). The L-lectin, b-GF, CDHL-1 and CDHL-2 modules of SraP BR are shown in cyan, red, yellow and green, respectively. The merged SraP BR was superimposed against the crystal structure of SraP BR with the two CDHL-2 modules aligned. (TIF) Figure S2 The CD spectra of A) SraP BR and B) the L-lectin module. The results demonstrated that mutation of Y367G did not introduce significant changes to the protein structures. (TIF) Figure S3 Quantitation of the GFP fluorescent signals of SraP BR and mutants. The fluorescent signals for each protein were quantified by calculating the mean gray values of three representative frames using the ImageJ software (http://imagej.nih.gov/ij/). The average mean gray value and standard error of the mean (SEM) derived from triplicate treatments are indicated as bar graph. (TIF) Figure S4 The L-lectin module is conserved in Staphylococci and some species of Streptococci. A) Schematic of Staphylococcal and Streptococcal proteins containing an L-lectin module. The first line represents the SraP protein from S. aureus. Similar modularization is observed in S. epidermidis and S. warneri (2 nd and 3 rd line). The 4 th line represents uncharacterized protein from S. salivarius. The 5 th is a serine threonine rich antigen from S. canis, and the last is a putative uncharacterized protein from S. australis. SP: signal peptide, CWA: cell wall anchor motif, Rib: a repeat motif in Rib protein of group B Streptococcus, Ig/ albumin-bd: immunoglobulin and albumin-binding domain. B) Multiple-sequence alignment of the L-lectin module and homologs. The Neu5Ac binding residues (red triangle) are conserved. (TIF) Figure S5 CDHL1&2 exists as a dimer in solution. A) The size-exclusion chromatograph profile of CDHL-1&2. The standard curve was inserted as an inlet. Chemical cross-linking of the purified CDHL-1&2 proteins using B) formaldehyde (FA) and C) bis(sulfosuccinimidyl) suberate (BS 3 ). The protein samples were separated by 10% SDS-PAGE. The bands corresponding to the monomer and the dimer of CDHL1&2 are labeled. (TIF)