Crystal Structure of the C-terminal Region of Streptococcus mutans Antigen I/II and Characterization of Salivary Agglutinin Adherence Domains*♦

The Streptococcus mutans antigen I/II (AgI/II) is a cell surface-localized protein that adheres to salivary components and extracellular matrix molecules. Here we report the 2.5 Å resolution crystal structure of the complete C-terminal region of AgI/II. The C-terminal region is comprised of three major domains: C1, C2, and C3. Each domain adopts a DE-variant IgG fold, with two β-sheets whose A and F strands are linked through an intramolecular isopeptide bond. The adherence of the C-terminal AgI/II fragments to the putative tooth surface receptor salivary agglutinin (SAG), as monitored by surface plasmon resonance, indicated that the minimal region of binding was contained within the first and second DE-variant-IgG domains (C1 and C2) of the C terminus. The minimal C-terminal region that could inhibit S. mutans adherence to SAG was also confirmed to be within the C1 and C2 domains. Competition experiments demonstrated that the C- and N-terminal regions of AgI/II adhere to distinct sites on SAG. A cleft formed at the intersection between these C1 and C2 domains bound glucose molecules from the cryo-protectant solution, revealing a putative binding site for its highly glycosylated receptor SAG. Finally, electron microscopy images confirmed the elongated structure of AgI/II and enabled building a composite tertiary model that encompasses its two distinct binding regions.

The Streptococcus mutans antigen I/II (AgI/II) is a cell surface-localized protein that adheres to salivary components and extracellular matrix molecules. Here we report the 2.5 Å resolution crystal structure of the complete C-terminal region of AgI/ II. The C-terminal region is comprised of three major domains: C 1 , C 2 , and C 3 . Each domain adopts a DE-variant IgG fold, with two ␤-sheets whose A and F strands are linked through an intramolecular isopeptide bond. The adherence of the C-terminal AgI/II fragments to the putative tooth surface receptor salivary agglutinin (SAG), as monitored by surface plasmon resonance, indicated that the minimal region of binding was contained within the first and second DE-variant-IgG domains (C 1 and C 2 ) of the C terminus. The minimal C-terminal region that could inhibit S. mutans adherence to SAG was also confirmed to be within the C 1 and C 2 domains. Competition experiments demonstrated that the C-and N-terminal regions of AgI/II adhere to distinct sites on SAG. A cleft formed at the intersection between these C 1 and C 2 domains bound glucose molecules from the cryo-protectant solution, revealing a putative binding site for its highly glycosylated receptor SAG. Finally, electron microscopy images confirmed the elongated structure of AgI/II and enabled building a composite tertiary model that encompasses its two distinct binding regions.
Dental caries (also called tooth decay or dental cavities) is a ubiquitous worldwide disease that affects humans of all age groups. Streptococcus mutans, a primary etiological agent of human dental caries (1) and an increasingly recognized cause of bacterial endocarditis (2), adheres to proteins contained within the salivary pellicle on the tooth surface, the extracellular matrix, and other microbial species (3). Antigen I/II (AgI/II, 2 also known as P1, B, SpaP, or PAc) of S. mutans has been implicated in bacterial adherence to constituents of the salivary pellicle (4,5) and has been studied for the past three decades as a target for protective immunity against dental caries. Apart from adherence, AgI/II influences biofilm formation (6) and promotes platelet aggregation (7), collagen-dependent bacterial invasion of dentin (8), and cariogenicity (9). Although AgI/II was initially discovered on oral streptococci, it has also been identified in members of the Group A and Group B streptococci (10), suggesting a role for this adhesin in a variety of species.
The AgI/II family proteins range from 140 to 180 kDa in predicted size and have a primary sequence composed of multiple conserved regions (Fig. 1b). Toward the N terminus, repeated sequences of high alanine content constitute the alanine-rich region followed by a segment commonly referred to as the variable (V) region. Further C-terminal in the sequence is a region of high proline content that forms a repetitive prolinerich region. Following the AgI/II proline-rich region is a C-terminal region (60 kDa or 550 amino acids), which is the most conserved region of AgI/II, with 62% identity among strains (11). This region also contains an LPXTG consensus motif recognized by the sortase enzyme responsible for covalently attaching AgI/II to the cell wall peptidoglycan (3).
The human receptor for AgI/II is the glycoprotein complex known as salivary agglutinin (SAG), which is secreted by the salivary glands as part of the salivary fluid (12). The main component of SAG is the highly glycosylated 340-kDa glycoprotein scavenger receptor called gp340 (13). The interaction of AgI/II with SAG is complex and multivalent and requires the presence of calcium (14). Segments within both the N termini and the C termini of AgI/II have been implicated in the binding, and evidence from several studies suggests that the C terminus contributes substantially to S. mutans adherence (5,15,16). In addition, monoclonal antibodies recognizing epitopes within the C terminus of AgI/II effectively inhibit the adherence of bacterial cells to SAG immobilized on hydroxyapatite beads (17). Further, the C terminus of AgI/II itself can competitively inhibit the adherence of S. mutans to SAG (5). The peptide QLKTADLPAGRDETTSFVLV within the C terminus inhibits both binding of AgI/II to SAG as well as the recolonization of S. mutans to human teeth in vivo (15). Alanine substitutions within this peptide indicated that residues Gln and Glu and the sequence FVLV were required for inhibition by this peptide (15).
In our previous study, we had shown that the discontinuous alanine-and proline-rich regions associate in a hybrid of ␣-helices and polyproline type II helices to form an extended (A)/(P) stalk, which positions the globular intervening region (V) at the apex and away from the cell surface (16). In addition, we had identified the presence of two distinct AgI/II regions, A 3 VP 1 and the C terminus, that adhere to SAG. Although the functional structure comprising the N-terminal and central regions of S. mutans AgI/II has been well characterized (16), the complete and functional C-terminal region has not yet been described. A partial C-terminal structure of the S. mutans AgI/II (SpaP) (18) and a homologous fragment from the Streptococcus gordonii SspB were recently reported (19), where each structure had two IgG-like domains.
Here we report the structure of the entire C terminus of AgI/II from S. mutans at 2.5 Å resolution. The structure reveals three domains, and each adopts the DE-variant immunoglobulin-like (DEv-IgG) fold. We have biophysically characterized and delineated the minimal binding region within the C terminus through both surface plasmon resonance studies as well as adherence inhibition studies using S. mutans cells in vitro. In addition, several glucose molecules were identified in the C terminus structure, suggestive of a binding site, which could facilitate adherence to the heavily glycosylated SAG. We have further deciphered the AgI/II complex multivalent binding characteristics using competition experiments with AgI/II fragments and provide evidence for the presence of distinct binding sites on SAG. Finally, transmission electron microscopy images of AgI/II enabled the building of a composite structural model for this elongated fibrillar protein.

Expression and Purification of C-terminal Fragments
DNA encoding the C-terminal region (C 123 ) was subcloned from the AgI/II gene (SpaP) of the S. mutans NG8 strain into a pET23d vector as described previously (16). Additional constructs of C 1 , C 12 , C 2 , C 23 , and C 3 were likewise subcloned into the pET23d vector using the primers listed in Table 1, with extents of the fragments shown in Fig. 1b. The resulting C-terminal plasmids were transformed into BL21 (DE3) Escherichia coli cells. 20-ml Terrific broth cultures were grown overnight at 37°C and transferred into 1-liter cultures on the following day with protein expression and cell lysis as described previously (16). Lysates were purified over a HisPrep nickel affinity column (Amersham Biosciences) followed by ion exchange chromatography on a MonoQ HR10/10 column (Amersham Biosciences), and lastly, by size exclusion chromatography over a Superdex 75pg 26/60 column (Amersham Biosciences). Other AgI/II constructs, including the amino-terminal A 3 VP 1 (386 -864) and AVP (199 -992) and the full-length CG14 (without secretion signal peptide, 39 -1566), were expressed and purified as described previously (16). a, the AgI/II C-terminal region is composed of three domains, labeled as C 1 , C 2 , and C 3 . The AgI/II C-terminal domains are shown in a ribbon model overlaid with a surface plot to illustrate the overall contiguous structure of the three domains. Each domain has an isopeptide bond linking two ␤-sheets (shown in red). The C 1 and C 2 domains have five bound glucoses (labeled glc1-5). Additionally, three calcium ions exist in the C 2 and C 3 domains (green), and one magnesium ion exists in the C 1 domain (purple). b, the graphic represents the domain structure of AgI/II, where the alanine-rich and proline-rich repeats, as well as the variable region and C terminus (C 1 , C 2 , C 3 ), are indicated with distinct colors.

TABLE 1 Primers used in this study
Plasmid pDC20 containing the AgI/II gene from S. mutans NG8 (42) was used as the template for PCR.

The Complete C-terminal Structure of AgI/II Crystallization and Data Collection
C 123 in a buffer containing 50 mM Tris, 150 mM sodium chloride at pH 8.0 was concentrated to 12.1 mg/ml using a 10-kDa molecular mass cutoff membrane in an Amicon concentrator under nitrogen gas pressure and screened by vapor diffusion at 22°C using commercial crystallization screens with hanging drops of 1 l of protein to 1 l of well solution. C 123 crystals were optimized from condition 1 of Crystal Screen II (Hampton Research) to a well solution containing 11% polyethylene glycol (PEG) 6000, 2 M sodium chloride, and 55 mM sodium succinate at pH 4.5. Crystals were flash-frozen in well solution augmented with cryoprotectants of either 20% PEG 400 for data collection at a home source (Raxis-IV mounted on an RU-H3R x-ray generator operated at 50 mA and 100 KV) or 30% D-glucose at the Northeastern Collaborative Access Team (NE-CAT) beamline at 93 K under a gaseous N 2 stream. High resolution data were collected at the NE-CAT beamline on a Quantum4 CCD detector, using a wavelength () of 1.0 Å, an exposure time of 2.0 s, an oscillation angle () of 1.0°, and a distance (D) of 150 mm from crystal to detector. The diffraction data were integrated and scaled with HKL2000 (20).

Multiple Isomorphous Replacement
Three heavy atom derivatives of the C 123 region were obtained by soaking freshly grown crystals. Crystals placed in well solution augmented with 1 M sodium iodide darkened to a brown/yellow tinge after 1 week, indicating incorporation of iodide. Derivatives containing thallium (III) acetate were stable under a brief 30-s quick soak containing 1 mM thallium (III) acetate followed by flash freezing, whereas derivatives of thimerosal were soaked in well solution augmented with 10 mM thimerosal for 48 h. For phasing, low resolution derivative and native datasets were collected at the home source. These data were processed with HKL2000 (20), and multiple isomorphous replacement solutions were identified by SHARP (version 2.6). 12 sites were identified in the sodium iodide derivative, and an additional four thallium and six thimerosal sites were identified. The figure of merit of the initial solution was 0.35 and 0.40, respectively, for the acentric and centric reflections. After solvent flattening, the total figure of merit improved to 0.914, with a resulting highly interpretable electron density map having clear boundaries between the protein and solvent within the map. The models were manually built using Coot (21), with refinements under CNS (22) over several cycles. At later stages of refinement, a higher resolution (2.5 Å) native dataset collected at NE-CAT was used to complete the model building. The first 8 residues (992-1000) could not be identified on the electron density maps. A complete summary of crystallographic parameters and refinement statistics is included in Table 2.

Surface Plasmon Resonance
Adherence of AgI/II Fragments-SAG was prepared from pooled unstimulated saliva from healthy human volunteers as described previously (17). The adherence of AgI/II polypeptides to human SAG was assessed by surface plasmon resonance (SPR) on the Biacore 2000 (Biacore Life Sciences, Uppsala, Sweden). SAG was immobilized on a CM5 chip, with buffer and flow conditions as described previously (16). 80 l of each AgI/II polypeptide, in concentrations ranging from 0.25 to 4 M, was injected over the chip surfaces, and dissociations were measured over 8-min time periods. Between experiments, 10 l of 10 mM HCl was used to regenerate the surface. Signals from the uncoated control cell FC1 were subtracted from those of the SAG-coated cell FC2 to produce sensorgrams (resonance units, or RU). Binding data were fitted with a 1:1 Langmuir kinetic model using the BIAevaluation software (version 4.2, Biacore Life Sciences). Competition Binding-To determine whether AgI/II domains bound to the same site on SAG, competition SPR experiments were conducted as follows. One fragment of AgI/II (CG14, A 3 VP 1 , AVP, or C 123 ) was flowed over the chip surface for 60 s to saturate available SAG binding sites, and immediately afterward, the second fragment was flowed over the chip for 60 s to test for adherence, with each experiment performed in triplicate. The resonance unit values (of the second fragment) were determined as the difference in the maximal resonance unit subtracted by the baseline prior to second injection.

Inhibition of Adherence of S. mutans to SAG-coated Hydroxyapatite
The ability of fluid-phase recombinant AgI/II polypeptides to compete with S. mutans cell-associated AgI/II for binding to SAG-coated hydroxyapatite beads was determined as described previously (17). Assays were performed using S. mutans strain NG8, and all assays were performed in triplicate.

Purification of AgI/II from S. mutans and Electron Microscopy
S. mutans (Guy's strain) was grown in supplemented basal medium, and full-length AgI/II was prepared from the culture supernatant by ion exchange and gel filtration chromatography as described previously (23). Purified AgI/II was dissolved at a range of protein concentrations (0.2-1 M) in 50% (w/w) glycerol and sprayed onto freshly cleaved mica, and after drying for 15 min at 10 Ϫ5 mbars, sample was unidirectionally shadowed with platinum at low angle (7.5 ). Replicas were floated off onto distilled water. All specimens were examined and photo-graphed in a Siemens 102 transmission electron microscope, with the magnification calibrated using a diffraction grating replica. Electron micrographs were measured on an Apple digitizer. At least 100 measurements of length were recorded for each specimen.

RESULTS
Overall Structure-The C terminus of AgI/II (SpaP) from S. mutans NG8 strain contains three domains, C 1 (992-1142), C 2 (1143-1332), and C 3 (1333-1486) (Fig. 1a), that possess 11-12 ␤-strands, which are interspersed with short helical stretches. The high percentage of ␤-strands within the C-terminal region is consistent with the circular dichroism studies on this protein fragment (16). Each of these domains adopts the DEv-IgG fold (24). The DEv-IgG like domains possess a set of antiparallel strands A-G similar to classical IgG folds, whose essential variations (additional strands and helices) occur between the D and E strands (Fig. 2). The A, B, E, and D strands form one major ␤-sheet, whereas the C, F, and G strands form the second major ␤-sheet. The variations between the D and E strands are most prominent in the C 2 and C 3 domains, whereas the C 1 domain has the least variation. The C 1 domain connects to the C 2 domain through a short proline-rich segment, whereas the C 2 domain connects to the C 3 through one very long ␤-strand. The surface of the C-terminal domains are comprised of an even mix of polar (44%) and non-polar (56%) residues (as calculated by Naccess 2.1.1). Each of these domains also carries an isopeptide bond, a characteristic signature present in Gram-positive pili proteins (25,26) that are typically coordinated through 3 principal residues, lysine, asparagine, and aspartate. Within the C 1 , C 2 , and C 3 domains, the isopeptide bonds link the residues Lys-1006 -Asn-1121, Lys-1161-Asn-1311, and Lys-1338 -Asn-1473, respectively. In each of the three C-terminal domains of AgI/II, this isopeptide bond covalently links the A and F strands, locking these strands together into a stable conformation (Fig. 2a).
Metal Ions-Three ions were identically present on each monomer: two within the C 2 domain and one within the C 3 domain. Each ion is well coordinated (coordination number of 6 -7) through backbone and side-chain oxygens (Fig. 3a) and have average bond distances on the order of 2.5-2.6 Å. These ions, in addition to existing within a region of AgI/II implicated as the minimal calcium binding region for members of the AgI/II family (1168 -1250 of SspB or 1248 -1330 of AgI/II) (27), also have coordination properties consistent with that of calcium. Further, in the high resolution crystal structures of SspB (1.7 Å) (19) and AgI/II (2.2 Å) (18), two of the sites were earlier identified as calcium ions. Therefore these ions were assigned as Ca 2ϩ in our current model. In the C 2 domain, the first calcium ion interacts with Asn-1155, Asp-1189, Asp-1191, and Gln-1192 side chains, as well as the main-chain oxygen of Tyr-1156 and an apex water molecule, resulting in the pentagonal bipyramidal coordination geometry. A second calcium ion is also observed within the C 2 domain and is well coordinated through the side-chain oxygens of Asp-1212 and Glu-1215 and the backbone oxygens of Tyr-1213, Lys-1265, and Ala-1267 that are located at loops between the C and D strands. The C 3 domain has the third calcium ion that is coordinated by the side-chain oxygens of residues Asp-1388 and Gln-1391 as well as by backbone oxygens of Tyr-1389, Lys-1434, and Gly-1435 and a water molecule once again located at loops between the C and D strands similar to the second Ca 2ϩ of the C 2 domain. Although most of the coordinating residues are also conserved in the loops between the C and D strands of the C 1 domain, the presence of a metal ion site is precluded by the substitution of a FIGURE 3. Interaction of carbohydrates within the C 12 cleft. a, the stereo image shows the three glucose molecules, which are bound within the cleft formed between the C 1 and C 2 domains of AgI/II. Sites Glc1 and Glc3 are present in both molecule A and molecule B, whereas Glc2 is only present in molecule A of the asymmetric unit. The DH2-helix of C 2 domain is in close proximity to the cleft, and residues from DH2 interact with Glc1. Each glucose forms bonds with surrounding residues and waters. Glc1 and Glc2 are bridged through a shared water molecule (W). b-d, the interactions between Glc1 (b), Glc2 (c), and Glc3 (d) and AgI/II residues are shown below in a Ligplot diagram (39). e, the electron density map shown in stereo for the Glc2 sugar, which is stabilized in the cleft by hydrogen bonding with nearby residues including the backbone oxygen of isoleucine 1157. The backbone oxygen of the neighboring tyrosine 1156 forms part of a coordination site for the nearby calcium ion.
Pro-1051 for the Asp-1212 or Asp-1388, which are present in the C 2 and C 3 sites. Additionally, an extra ion (Fig. 1a, purple) is observed coordinated between the two molecules within the asymmetric unit through residues Asp-1150 and Arg-1018 of molecule A and Thr-1397 of molecule B. This ion has a coordinating number of 5, with shorter bond distances (average Ͻ 2.4 Å) than those previously described for calcium sites and was therefore modeled as magnesium.
Carbohydrates-In the final stages of refinements, eight positions near the surface of the AgI/II C-terminal domains resolved as large ring-like structures in the difference electron density map. The native data had been obtained from a crystal soaked in a cryoprotectant solution that contained 30% w/v glucose, and these densities were well fitted with glucose molecules (Glc1-5 shown in Fig. 1a). Glc1, Glc3, and Glc4 are conserved in both molecules of the asymmetric unit, whereas Glc2 (mol1 only) and Glc5 (interface of mol1 and mol2) each exist only at one position within the asymmetric unit (Fig. 1a). Glc1, Glc2, and Glc3 were in close proximity to each other (12 Å) within the cleft formed between the C 1 and C 2 . This cleft may represent a binding site for a complex carbohydrate structure such as a glycosylation or sugar polymer (Fig.  3a). Glc4 and Glc5, which lie outside the cleft (Fig. 1a), are observed at intermolecular contact sites between different C 123 molecules within the crystal. The oxygens of Glc1, Glc2, and Glc3 make direct hydrogen bonds with the residues of AgI/II (Fig. 3, b-d), as well as water-mediated interactions with nearby amino acids that appear to stabilize the position of these glucoses. Glc1 interacts with side-chain nitrogens of Asn-1320 and Gln-1024, backbone nitrogen of Gly-1321, and nitrogen and oxygens of the backbone of the C 2 DH2helical residues Lys-1259 and Ala-1260 (Fig. 3b). Neighboring Glc1 and Glc2 are linked through a single water molecule (Fig. 3a), mimicking a linked sugar chain. Glc2 is additionally hydrogen-bonded with Ile-1157 and Ala-1323 (Fig. 3, c and  e). Glc3 is stabilized through interactions with Asn-1076 and Gly-1055 (Fig. 3d).
Similarity between C 23 Domains of AgI/II of S. mutans and SspB of S. gordonii-The recently resolved structures of C 23 from SpaP (18) and C 23 domains within the present structure have nearly identical structure and superpose with an r.m.s.d. of 0.586 Å. The C 23 domains of the present structure and SspB (19) also display high similarity and superpose well with an r.m.s.d. (on C␣) of 0.906 Å (Fig. 4a). The isopeptide bonds observed in the C 23 domains of the C-terminal region of S. mutans AgI/II occur at equivalent positions in the structure as those identified in the homologous C 23 domains of S. gordonii SspB (19). The DH2-helix of the C 2 domain, involved in the interaction between S. gordonii SspB and Porphyromonas gingivalis (28), is also very similar between SspB and AgI/II structures and superposes with an r.m.s.d. of 0.452 Å. (Fig. 4b).
Studies "Surface Plasmon Resonance"-AgI/II has two separate segments capable of adherence to immobilized salivary agglutinin (16): first, the N-terminal and central regions comprising the third A-repeat through the first P-repeat (A 3 VP 1 ), and second, the C terminus. With the identification of three domains within the C terminus, we attempted to further delin-eate the binding within the C terminus. The recombinant fulllength AgI/II (CG14) and C-terminal polypeptides comprising C 123 , C 12 , C 23 , and the individual C 1 , C 2 , and C 3 domains were investigated using SPR for their ability to interact with immobilized SAG. Among the C-terminal fragments, C 123 and C 12 exhibited measurable binding (Fig. 5a). Kinetics experiments conducted at multiple concentrations (Fig. 5b) exhibited a relatively higher affinity for C 12 with an estimated dissociation constant (K D ) of 57 nM as compared with C 123 and recombinant full-length AgI/II (CG14), which had K D values of 410 nM and 69 nM, respectively.
Next we tested whether the N-terminal (AVP or subset A 3 VP 1 ) and C-terminal (C 123 ) polypeptides of AgI/II could compete with each other for binding to immobilized SAG (Fig.  5c) to determine whether these regions adhere to the same or different sites on SAG. Binding of full-length CG14 construct substantially reduced the subsequent binding of C 123 , A 3 VP 1 , and AVP by 48, 73, and 77%, respectively. However, C 123 did not greatly inhibit the binding of the amino-terminal constructs, nor did the amino-terminal fragments inhibit the binding of C 123 . Control SPR experiments with immobilized N-and C-terminal AgI/II fragments did not show measurable self-or cross-interactions with other AgI/II fragments. Lack of competition between the AgI/II N-terminal and C-terminal regions for SAG suggests that these regions interact with independent or distinct sites on the immobilized SAG.
S. mutans Adherence Inhibition-To further establish the physiological relevance of C 12 binding, competition experiments with S. mutans were performed. The binding of 3 H-labeled S. mutans to SAG-coated hydroxyapatite beads was evaluated in the presence and absence of fluid-phase competitors. In the first set of experiments (Table 3 (upper)), the abilities of recombinant full-length AgI/II (CG14) and the N-terminal (A 3 VP 1 ) and C-terminal (C 123 ) fragments to inhibit the interaction of S. mutans with immobilized SAG were measured. Both A 3 VP 1 and the C-terminal fragments demonstrated partial inhibition of bacterial adherence at 5 M concentration (28 and 43%, respectively), whereas the full-length molecule demonstrated substantial inhibition (70%) at the lower 1 M concentration. Adding the A 3 VP 1 and the C-terminal polypeptides together did not improve the ability to inhibit adherence, suggesting that the orientation or spacing of two separate binding sites within the full-length adhesin may contribute to the strength of the binding interaction. Next we tested the abilities of the C 12 and C 23 fragments to competitively inhibit bacterial adherence as compared with the recombinant full-length AgI/II (CG14) and the entire C-terminal fragment (C 123 ) at 1 and 5 M concentrations (Table 3 (lower)). The C 12 , but not the C 23 fragment, was able to serve as a competitive inhibitor. Consistent with the results of the SPR binding studies, the C 12 fragment again demonstrated a greater ability to interact with SAG than the entire C-terminal fragment (C 123 ).
Transmission Electron Microscopy Studies-Native AgI/II purified from S. mutans Guy's strain was visualized by transmission electron microscopy as an elongated molecule that contains larger bulges at its termini (Fig. 6a). The average length of AgI/II molecules in these electron microscopic images was estimated to be 65.8 nm and conforms well with the 68-nm length of the full-length molecule estimated from our previous structural study (16). The length of the C terminus was measured to be 13 nm based on the crystal structure described here. This is twice the measured diameter (6 nm) of the variable region domain in crystal structures containing the V-region (16,29). These longer carboxyl termini can be clearly seen on one end of the molecule, with the smaller V-region at the other end of the molecule, and these carboxyl termini are connected through the alanine-and proline-rich stalk. A composite comprehensive tertiary structure that contains two distinct and independent binding sites for SAG was built (Fig. 6b) based upon the current and previous x-ray crystallographic studies of AgI/II regions, the measured dimensions from the current elec-tron microscopy, and prior analytical ultracentrifugation (16). This model now well describes the majority of the AgI/II structure.

DISCUSSION
The C-terminal region consists of three (C 1 , C 2 , and C 3 ) DEv-IgG domains (24), each of which displays distinct variations in the structural elements between the D and E strands on the common underlying architecture of the IgG fold. This fold is mainly observed in proteins connected with protein-protein interactions, such as Staphylococcus aureus ClfA (24) and CNA (30). The domains of the C terminus of AgI/II are aligned to give an extended appearance similar to the fibronectin-like domains  (31). Each of the C 1 , C 2 , and C 3 domains of S. mutans AgI/II contains an isopeptide bond, located at an identical position of the DEv-IgG fold. The isopeptide bonds present in the AgI/II C terminus follow an architecture similar to the domains of the Corynebacterium diphtheriae pilin SpaA, with the isopeptide bond joining the A and F strands and linking the two opposing ␤-sheets (26). These isopeptide bonds likely provide stability to the fold for protection from proteolytic enzymes and shear forces brought by salivary flow (1 liter/day) that pose particular challenges to the design and function of bacterial surface proteins within the oral cavity.

The Complete C-terminal Structure of AgI/II
SPR experiments on the C-terminal fragments C 1 , C 2 , C 3 , C 12 , and C 23 indicated that a cooperative surface formed by C 12 is required for adherence to SAG. This result is in concurrence with two other important studies on AgI/II. First, it is in immediate proximity to the previously identified inhibitory peptide (mapping to residues 1030 -1049 on NG8 AgI/II) (15) on C 1 , which is capable of blocking S. mutans adhesion to SAG and recolonization in human subjects. Secondly, the DH2-helix present within the C 2 AgI/II family member of S. gordonii SspB was shown to specifically recognize the minor fimbrial protein Mfa1 of P. gingivalis (28,32). In addition, this kind of specific adherence utilizing multiple DEv-IgG domains to latch on to receptors has also been observed for other Gram-positive sur-face adhesins, such as the S. aureus CNA and Staphylococcus epidermis SdrG (33). Therefore it appears likely from these results that the C-terminal SAG adherence site is contained within the C 12 domains.
Earlier we identified the presence of two distinct SAG adherence regions, A 3 VP 1 and the C terminus (16). The present SPR studies now further elucidate that the cell surface proximal (C-terminal) and distal (A 3 VP 1 ) ends of AgI/II bind non-competitively to SAG (Fig. 5c). The ability of the two different AgI/II fragments to simultaneously adhere to immobilized SAG indicates that the AVP and C terminus regions bind to distinct sites on SAG. Given that the AVP region (antigen I) is prone to proteolysis, whereas the antigen II region (C-terminal region) is more resistant to proteolysis (23,34), the presence of two independent adherence sites for SAG suggests that S. mutans has evolved heterogeneous adherence mechanisms for bacterial colonization of the tooth surface.
The significant glycosylations of SAG have long been studied as a possible mode for the binding of AgI/II with SAG (14,35,36). Interactions of S. mutans and SAG can be inhibited by various amino sugars as well as other compounds with primary amine groups and less efficiently by lactose and melibiose (14). In addition, binding of AgI/II homologue SspB from S. gordonii to SAG is sialic acid-dependent (35). The structural report on the V-region implicated a hydrated pocket as a potential sugar binding site (29), and this site was occupied by a non-carbohydrate ligand in the structure of A 3 VP 1 (16). However, the identification and measurement of an interaction between a simple or complex sugar molecule with any region of AgI/II are yet to be substantiated. In the current structure, the serendipitous identification of multiple glucose molecules within a large trench formed between the C 12 domains suggests a possible mechanism for recognition of the highly glycosylated SAG.
It is unlikely, however, that a simple glucose would represent an ideal carbohydrate motif to be recognized by the AgI/II C terminus. The presence of multiple bound carbohydrates (Glc1, Glc2, and Glc3) in near proximity (12 Å) also suggests that a more complex glyco-structure would fit the cleft, with increased affinity and better recognition than a simple monosaccharide. Branched glycosylations or polymers like dextran might be candidates for such recognition. Glucose and other simple monosaccharides have little inhibitory effect on the interaction of S. mutans with SAG (14), whereas the interactions of oral streptococci with SAG are impaired upon enzy- FIGURE 5. Evaluation of AgI/II polypeptide-SAG interactions using surface plasmon resonance. a, SAG was immobilized, and AgI/II polypeptides (2 M each) were injected over the CM5 chip surface. Recombinant full-length AgI/II (CG14, black), C 123 (orange), and C 12 (red) each displayed a strong binding response (RU, resonance units; 1 resonance unit ϭ ϳ1 pg/mm 2 ) accompanied by slower dissociations that are indicative of adherence to the SAG-coated surface. b, the interaction between C 12 with immobilized SAG at multiple concentrations is shown. Kinetics fitting with multiple C 12 concentrations estimated the K D to be 57 nM. c, competition experiments were performed for the binding of AVP, A 3 VP 1 , or C 123 to immobilized SAG. The adherence of AgI/II fragments to SAG without competition (labeled as None) is shown first followed by measured adherence when competed against other AgI/II fragments (labeled CG14, C 123 , AVP, or A 3 VP 1 ). All experiments were carried out in triplicate. Error bars indicate mean Ϯ S.D.

The Complete C-terminal Structure of AgI/II
matic removal of the complex sugar structures on SAG (36). The multiple globular scavenger-receptor (SRCR) domains of SAG are interspersed with segments predicted to have O-glycosylation sites that separate the SRCR domains, and these glycosylations are predicted to produce the elongated structure for SAG (12). The binding sites of Glc1, Glc2, and Glc3 could interact with the above mentioned SRCR interspersed segments, acting alone or in addition to other direct protein-protein interactions between AgI/II and SAG. Structural changes in the presence or absence of calcium have previously been suggested to be important for the binding of AgI/II and SAG (27). Within the cleft region, structural elements link the two calcium ion sites within the C 2 domain to the Glc1 and Glc2 sites. The main-chain oxygen of Tyr-1156 interacts with a calcium ion and its adjacent residue (Ile-1157) interacts with Glc2 (Fig. 4). Separately, the leading loop to the C 2 DH2-helix hosts a second calcium ion, and Glc1 interacts with Lys-1259 and Ala-1260 of the C 2 DH2-helix. These ions could stabilize the orientations of residues to preferably interact with carbohydrate chains. Such an arrangement may explain the observed requirement for calcium in SAG binding (14), and it is tempting to speculate that either of these calcium ions could provide a mechanism for conformational changes that modulate an interaction of the AgI/II C terminus to carbohydrate chains.
Although the main chains of the C 2 DH2-helices of both AgI/II and SspB have very similar conformation (superposition r.m.s.d. of 0.452 Å 2 ), there certainly exist differences, particularly at the side chains where the amino acids are different (Fig.  4b). When Asn-1182 and Val-1185 on SspB were mutated to match the equivalent residues on AgI/II, Gly-1261 and Pro-1264, respectively, each conferred a loss of adherence to P. gingivalis (37). Although neither of these mutations appear to greatly affect the backbone structure within this region, the Asn-1182/Gly-1261 mutational site neighbors the Glc1 sugar in the AgI/II C-terminal structure (Fig. 4b). It is quite possible that these mutational sites on the C 2 DH2-helix of SspB are perhaps involved in targeted adherence to specific glycosylated moieties that are present on Mfa1 (38).
With the resolution of the C-terminal structure of AgI/II, almost the entire AgI/II protein structure is now known with only a minimal N-terminal portion of the AgI/II preceding the alanine-rich region remaining uncharacterized (Fig. 6b). The C 123 structure, which is 13 nm in length, in addition to the 50-nm extended AVP region (16), produces an estimated overall AgI/II length of 63 nm (Fig. 6b), which concurs well with our earlier results from analytical ultracentrifugation (16). The electron microscopy studies (Fig. 6a) confirm the extended nature and dumbbell shape of AgI/II, particularly the visualization of the larger extended C-terminal domain on one end of the molecule and the smaller V-region that is positioned at the opposite apex of the molecule. The C 123 appears as a continuous van der Waals surface in the electron micrographs. Unlike . Electron micrographs and composite model of AgI/II of S. mutans. a, electron microscopy of native AgI/II obtained from streptococcal cells (S. mutans Guy's strain) is shown with a composite of electron micrographs of unidirectionally metal-shadowed streptococcal AgI/II molecules at magnification ϫ235,000 (bar ϭ 50 nm). The micrographs of AgI/II illustrate an elongated model with globular domains present at the termini. The longer globular C terminus (ϳ13 nm) present at an opposing end from the smaller V-region (ϳ6 nm) can be clearly visualized in these micrographs. The end-to-end length of the molecules is 65.8 (S.D. Ϯ 3.6) nm, similar to predicted dimensions from analytical ultracentrifugation of AgI/II (16). Southern blotting (40) and partial DNA sequencing of AgI/II (41) have displayed a high degree of homology (Ͼ95%) between S. mutans NG8 and Guy's strains. b, illustrated is a composite model of tertiary structure for AgI/II, including the extended hybrid of the ␣-helical A-region (red) and the polyproline type II helical P-region (blue), that separates the globular V-region(green) and C-terminal region (light blue, C 1 ; yellow, C 2 ; blue, C 3 ). The two distinct and non-competing binding sites for SAG are contained within the A 3 VP 1 and C 12 regions, which are located at opposite ends of the A/P stalk. Following the C 3 domain is the LPXTG motif that anchors the protein to the cell wall peptidoglycan layer.

The Complete C-terminal Structure of AgI/II
the A/P repeat motifs, the C terminus has utilized structural motifs consisting of repetitions of the DEv-IgG fold. The use of repeated sequences and structural units to extend Gram-positive surface proteins from the cell surface is now being seen as a common and widespread architecture for bacterial adhesion proteins.
In summary, the structure of the C terminus of AgI/II has revealed three domains of the DEv-IgG fold (C 1 , C 2 , and C 3 ) and has enabled the construction of a comprehensive structural model of the adhesin as visualized from electron microscopy images (Fig. 6). SAG binding function was localized within the first two domains (C 12 ) of the C terminus, which also showed multiple interactions with carbohydrates within the crystal structure. The presence of two distinct N-and C-terminal SAG binding sites within AgI/II, as well as the localization within the C 12 fragment, were confirmed by the ability of these regions to inhibit S. mutans adherence to SAG, further indicative of its physiological relevance in the context of the cell surface-localized adhesin as it is displayed on S. mutans cells. The distinct types of interaction between AgI/II and SAG warrants further studies that would elaborate on the mechanisms of the two independent forms of adherence. In this context, structural and functional studies on the interaction between the gp340 SRCR domains and AgI/II fragments would provide intimate details on these bacterial-host interactions, and investigations aimed at disrupting these interactions could produce novel therapeutics to impede the adherence of oral streptococci to tooth surface.