Structural and Functional Characterization of the Streptococcus pneumoniae RrgB Pilus Backbone D1 Domain*

Streptococcus pneumoniae expresses on its surface adhesive pili, involved in bacterial attachment to epithelial cells and virulence. The pneumococcal pilus is composed of three proteins, RrgA, RrgB, and RrgC, each stabilized by intramolecular isopeptide bonds and covalently polymerized by means of intermolecular isopeptide bonds to form an extended fiber. RrgB is the pilus scaffold subunit and is protective in vivo in mouse models of sepsis and pneumonia, thus representing a potential vaccine candidate. The crystal structure of a major RrgB C-terminal portion featured an organization into three independently folded protein domains (D2–D4), whereas the N-terminal D1 domain (D1) remained unsolved. We have tested the four single recombinant RrgB domains in active and passive immunization studies and show that D1 is the most effective, providing a level of protection comparable with that of the full-length protein. To elucidate the structural features of D1, we solved the solution structure of the recombinant domain by NMR spectroscopy. The spectra analysis revealed that D1 has many flexible regions, does not contain any intramolecular isopeptide bond, and shares with the other domains an Ig-like fold. In addition, we demonstrated, by site-directed mutagenesis and complementation in S. pneumoniae, that the D1 domain contains the Lys residue (Lys-183) involved in the formation of the intermolecular isopeptide bonds and pilus polymerization. Finally, we present a model of the RrgB protein architecture along with the mapping of two surface-exposed linear epitopes recognized by protective antisera.

Streptococcus pneumoniae is an important human pathogen responsible for diseases such as otitis media, pneumonia, sepsis, and meningitis (1-6). However, S. pneumoniae is also a common inhabitant of the respiratory tract of children and healthy adults. This carriage state could represent a risk factor for the development of respiratory diseases but also the source for pneumococcal transmission to other individuals (7)(8)(9). Like most streptococci, S. pneumoniae decorates its surface with long filaments known as pili (10 -14). Pneumococcal pili have previously been associated with virulence and the capability of the microorganism to adhere better to epithelial cells and to colonize the nasopharynx (10,15,16). The pneumococcal pilus is a multimeric structure consisting of three proteins (RrgA, RrgB, and RrgC) polymerized by three sortase enzymes (SrtC1, SrtC2, and SrtC3) through the formation of covalent intermolecular isopeptide bonds (17)(18)(19)(20)(21). In particular, multiple copies of RrgB are polymerized to form the scaffold of the pilus, whereas the major adhesin, RrgA, and the putative anchor, RrgC, are localized at the tip and at the base of the pilus, respectively (15,22,23).
Recently, the structure of a major portion of RrgB (residues 184 -627) was solved at a 1.6 Å resolution (24) and revealed an organization into three independently folded IgG-like domains (D2, D3, and D4, residues 184 -326, 326 -446, and 446 -627, respectively). On the contrary, the structure of the RrgB N-terminal region (D1, residues 1-184), likely constituting a fourth independently folded domain, remained unsolved due to the failure to obtain the crystals of the full-length (FL) 3 RrgB (24). Interestingly, each of the D2, D3, and D4 domains is stabilized by one intramolecular isopeptide bond. These covalent linkages, formed between Lys and Asn residues, have been found in other pilus proteins (19,(25)(26)(27)(28) and are thought to play a role similar to that of disulfide bonds; they confer in fact a rigid molecular architecture to the pili and make them less susceptible to proteolytic cleavage (Fig. 1).
In the pilus backbone assembly RrgB molecules are linked together by sortases through intermolecular isopeptide bond formation between a Thr within the C-terminal LPXTG motif of a molecule and a Lys located at the N terminus of the following molecule (18,26,29) (Fig. 1) established, the functional Lys is located within a conserved YPKN "pilin" motif (18,27,30). Nevertheless, this sequence is not absolutely required for polymerization as demonstrated by studies on the Spy0128 pilin of Streptococcus pyogenes, where the lysine forming the intermolecular isopeptide bond and responsible for pilus polymerization is located into 159 GSKVPI 164 motif even though the YPKN pilin motif is also present (26,31). RrgB, along with the other two pilus proteins RrgA and RrgC, was previously shown to confer protection in mouse models of infection and therefore is regarded as a potential candidate for a new generation of protein-based vaccines (32,33). We have investigated the protective ability of the single recombinant D1, D2, D3, and D4 domains of RrgB in a mouse model of sepsis, and here we provide evidence that D1 is the most protective, followed by D4. Furthermore, we present the solution structure of the recombinant D1 obtained by NMR spectroscopy and show that Lys-183 of D1 is engaged in the intermolecular isopeptide bond formation during pilus polymerization. Finally, we propose a possible model of the entire RrgB molecule and show the positions of two linear epitopes possibly involved in the protection mechanism.

MATERIALS AND METHODS
Bacterial Strains and Cultures-For the animal experiments, the S. pneumoniae TIGR4 (serotype 4) strain was used. Bacteria were grown, frozen in aliquots at Ϫ80°C, and titrated as already reported (32). Immediately prior to challenge, frozen aliquots were thawed and diluted in PBS to reach the working concentration.
Cloning and Protein Expression and Purification-Standard recombinant DNA techniques were used to construct the expression plasmids (pET21b ϩ ; Novagen) and to express and purify the recombinant C-terminal His-tagged proteins (for details, see supplemental Materials and Methods, and the primers used are listed in supplemental Table S1). The affinity-purified proteins were subsequently used to immunize CD1 mice or rabbits for antibody generation (Charles River Laboratory) and BALB/c mice to evaluate the protective efficacy.
Complementation Plasmids-Wild-type or mutant rrgB genes were amplified from chromosomal DNA of TIGR4 strain by PCR by using the primers listed in supplemental Table S2; point mutations were introduced by overlap extension PCR by using specific primers (supplemental Table S2). PCR products were then cloned into the complementation plasmid pMU1328 between the BamHI and SalI restriction sites (34). Expression of RrgB or RrgB mutated forms was under control of the erythromycin constitutive promoter (Pc) which was amplified with the primers listed in supplemental Table S2 and cloned immediately upstream rrgB (EcoRI, BamHI). All plasmids were confirmed by sequencing.
Generation of rrgB Deletion Mutants and rrgB Complementants-A TIGR4 ⌬rrgB isogenic mutant was generated by allelic exchange. Fragments of ϳ500 bp upstream and downstream the target gene were amplified by PCR (oligonucleotides are listed in supplemental Table S2) and spliced into a kanamycin resistance cassette by using overlap extension PCR; the PCR fragments were then cloned into pGEMt (Promega) and transformed in S. pneumoniae with conventional methods (35). To select the bacteria in which the target gene was replaced with the resistance cassette, bacteria were plated on blood-agar plates with kanamycin (500 g/ml). The presence of the isogenic mutation was confirmed by PCR and Western blot (WB) analysis. To obtain RrgB-complemented mutants, pMU1328 plasmids containing WT rrgB or rrgB mutated forms were transformed into TIGR4 ⌬rrgB with conventional methods. Transformants selection was performed by supplementing media with kanamycin (500 g/ml) and erythromycin (1 g/ml). The complemented mutants were then analyzed by PCR; expression of FL WT RrgB or RrgB mutants was detected by WB analysis of whole cell lysates.
SDS-PAGE and Western Blot Analysis-SDS-PAGE analysis was performed using NuPAGE TM 4 -12% BisTris gradient gels (Invitrogen) according to the manufacturer's instructions. Hi-Mark TM prestained HMW (Invitrogen) served as protein standard. Gels were processed for WB analysis by using standard protocols. Mouse polyclonal antibodies raised against recombinant His-tagged RrgB were used at 1/3000 dilution. Secondary goat anti-mouse IgG alkaline phosphatase-conjugated antibodies (Promega) were used at 1/5000, and signals were developed by using Western Blue Stabilized Substrate for Alkaline Phosphatase (Promega).
Animal Experiments-Animal studies were done in compliance with the current law approved by the local Animal Ethics Committee and authorized by the Italian Ministry of Health. Female, 6-week-old, specific pathogen-free BALB/c mice (Charles River) received three intraperitoneal immunizations, 2 weeks apart. Each dose was composed of 20 g of either the single RrgB domains or the FL RrgB, or of a combination of the four RrgB domains (D1ϩD2ϩD3ϩD4), 10 g each, along with 400 g of aluminum hydroxide as an adjuvant, in a final volume of 200 l of PBS. Control animals received the same course of saline plus adjuvant. Ten days after the third immunization, samples of sera were obtained from each animal and pooled according to the immunization group to be used in passive serum transfer experiment. Two weeks after the third immunization, each mouse was challenged intraperitoneally with a mean dose of 1.6 ϫ 10 2 cfu of TIGR4. Bacteremia was evaluated 24 h after challenge, and mortality course monitored for 10 days after challenge as already reported (32). The animals were euthanized when they exhibited defined humane end points that had been preestablished for the study in agreement with Novartis Animal Welfare Policies. Pilus scaffold is composed by multiples copies of RrgB protein in a head-to-tail arrangement. Pilus polymerization occurs through intermolecular isopeptide bonds (red), whereas each RrgB protein is stabilized by intramolecular isopeptide bonds (black). Lys-183 as a residue involved in the intermolecular bond has been identified in the present work.
For passive protection experiments, 8-week-old mice were used. Fifteen minutes before TIGR4 intraperitoneal challenge (10 2 cfu/mouse), each mouse received intraperitoneally 50 l of pooled mouse sera against recombinant D1 or D4, or of control sera obtained immunizing with adjuvant plus saline.
Statistical Analysis-Bacteremia and mortality course were analyzed by the Mann-Whitney U test. Survival rates were analyzed by Fisher's exact test. One-tailed or two-tailed tests were used to compare immunized groups with the control group or each other, respectively. Values of p Յ 0.05 were considered and referred to as significant. Values of p Յ 0.1 were referred to as a trend.
Flow Cytometry on Entire Bacteria-TIGR4 were grown in Todd-Hewitt yeast extract broth to an exponential phase (A 600 ϭ 0.25), fixed with 2% formaldehyde, and then stained with pooled mouse antisera raised against FL RrgB or RrgB domains at 1:400 dilution. Mouse IgG were detected with FITC-conjugated goat anti-mouse IgG (Jackson Laboratories) at 1:100 dilution, and bacterial staining was analyzed by using a FACS-Calibur cytometer (Becton Dickinson). Sera from mice immunized with PBS plus adjuvant were used as negative control.
ELISA-96-well MaxiSorp TM flat-bottom plates (Nunc) were coated with 0.2 g/well FL RrgB overnight at 4°C. Plates were then washed three times with PBS/0.05% Tween 20 and saturated for 1 h at 37°C with PBS/1% BSA. Following three washing steps with PBS/0,05% Tween 20, the plates were incubated for 2 h at 37°C with serial dilutions of the pooled mouse sera. After another three washing steps, bound antigen-specific mouse IgGs were revealed with alkaline phosphatase-conjugated goat anti-mouse IgG (Sigma), followed by the phosphatase alkaline substrate p-nitrophenyl phosphate (Sigma). The intensity of color was quantified with an ELISA plate reader at A 405 . The antibody titer was expressed as the log10 of the reciprocal of the serum dilution that gave A 405 ϭ 1.
PepScan Analysis-Arrayed peptides were synthesized in situ on glass fiber membranes. Membranes were conditioned by wetting with ethanol and washing three times for 5 min in TTBS (50 mM Tris-HCl, pH 7.0, 137 mM NaCl, 2.7 mM KCl, 0.05% Tween 20). After overnight blocking at 4°C in MBS (2% dry milk in TTBS), membranes were incubated for 1.5 h at 37°C with polyclonal antisera (1:3000 in MBS) followed by secondary goat anti-mouse IgG alkaline phosphatase-conjugated antibodies (1:5000 in MBS; Promega), and signals were developed by using Western Blue Stabilized Substrate for Alkaline Phosphatase. For image processing, membranes were scanned using an Epson V750 Pro scanner at 800 dpi, 48-bit color depth and with gamma 1.0 full linear response.
NMR Characterization of RrgB D1 Domain-Expression and purification of labeled D1 were carried out as described in the supplemental Materials and Methods. NMR spectra were acquired at 298 K on Avance 900, 800, 700, and 500 MHz Bruker spectrometers, all equipped with a triple resonance cryoprobe. The NMR experiments, used for the backbone and the aliphatic side chain resonances assignment recorded on 13 C/ 15 N and 15 N enriched samples or on unlabeled D1 samples, are summarized in supplemental Table S3. Backbone dihedral angle constraints were derived from 15 N, 13 C', 13 C␣, 13 C␤, and Ha chemical shifts, using TALOSϩ (36). Distance constraints for structure determination were obtained from 15 N-edited and 13 C-edited three-dimensional NOESY-HSQC. 3131 meaningful proton-proton distance restraints (supplemental Table S4), with 114 and 120 backbone dihedral angles restraints were included in structure calculations. The exchangeability of the backbone amide hydrogen nuclei with solvent protons was investigated through a 1 H-15 N HSQC experiment performed on a protein sample dialyzed against deuterated buffer for 3 days. Hydrogen bond constraints for the slowly deuterium-exchanging amide protons of the ␤-strands were introduced at later steps of the structure calculations.
Structure calculations were performed through iterative cycles of CYANA-2.1 (37) followed by restrained energy minimization with the AMBER 10.0 Package in explicit water solvent (38). The quality of the structures was evaluated by using the iCING validation program (for details, see supplemental Table S4). The program MOLMOL was subsequently used for structure analysis (39).
The final bundle of 20 conformers of D1 has an average target function of 1.36 Ϯ 0.13 (CYANA units S2). They are essentially homogeneous along the entire polypeptide sequence, with the exception of residues located at the C and N termini and three loop regions (56 -69, 148 -162, and 173-177). The correlation time for the molecule tumbling ( c ), as estimated from the R 2 /R 1 ratio, is 11.49 Ϯ 1.9 ns, consistent with the molecular mass of the monomeric protein. The relaxation data were analyzed according to the model-free approach of Lipari and Szabo (42,43) using TENSOR2 (44) (supplemental Fig. S2).
Rigid Body Fitting-The procedure used to accommodate the NMR D1 structure into the EM map of the whole pilus previously generated (24) followed the same procedures used for the fitting of the RrgBD2-D4 x-ray crystal structure and the D1 computer model into the EM pilus map (22). A preliminary rigid body fitting of the D2-D4 crystal fragment was performed by using CHIMERA (45)(46)(47) followed by a rigid body fitting of the D1 NMR coordinates into the leftover N-terminal apical volume of the pilus. The D1 NMR coordinates were first fitted manually using CHIMERA by placing as much of the atomic structure as possible into the EM density map, approximately in the position thought to be correct. This step was then followed by a rigid-body fitting using the "fit model in map" tool from CHIMERA. This tool calculates, for the selected atoms, a position that maximizes locally the sum of the densities. The evaluation of the correlation coefficient values, the resulting average map value at the fit atoms, the number of fits atoms outside the lowest value contour surface displayed were the parameters used to assess the fit of the D1 molecule.

Distinct Domains of RrgB Provide Protection in Active
Immunization Experiments-The protective efficacy of each of the four RrgB domains or of a mixture of them was assessed in a mouse model of intraperitoneal challenge (TIGR4) in two distinct experiments, performed under the same conditions, which were combined to reach n ϭ 13-16 mice/group. The results are summarized in Fig. 2, A and B, and detailed statistical analysis is provided in supplemental Table S5A.
All RrgB domains except D3 afforded significant protection against bacteremia ( Fig. 2A), giving a reduction of the cfu geometric mean by 1-2 logs with respect to the controls. These values were similar to those afforded by the FL RrgB and by the combination of the four domains D1ϩD2ϩD3ϩD4. The results of mortality are reported in Fig. 2B and supplemental Table S5A. D1 and D4 conferred significant increase of survival time, similar to FL RrgB and the combination D1ϩD2ϩD3ϩD4. In particular, the median survival for the D1 group was 2.5 days higher respect to that of the control group (4 versus 1.5 days). At the end of the mortality observation, a significant survival rate was found for D1 (44% survival), D4 (27%), FL RrgB (44%), and D1ϩD2ϩD3ϩD4 (31%) groups.
N-terminal D1 Domain Is the Most Protective in Vivo-Among the single RrgB domains, D1 and D4 showed the most significant protective efficacy and were therefore analyzed further in a larger group of mice. Four different experiments, carried out under the same conditions, were combined to reach n ϭ 31 animals/group. The results are shown in Fig. 2, C and D, and detailed statistical analysis is provided in supplemental Table S5B. In terms of bacteremia (Fig. 2C), both D1 and D4 afforded highly significant protection, with a cfu geometric mean by 2.6 and 1.5 logs lower, respectively, than that of the control group, and 8 animals from the D1 group in which cfu were undetectable. The reduction of bacteremia was significantly superior in D1 than in the D4 group.
In terms of mortality course (Fig. 2D), both D1 and D4 conferred significant protection. The increase of survival time afforded by D1 showed a better trend than that of D4. In particular, only for the D1 group was the median survival time higher than that of the control group (7.5 versus 1.5 days). At the end of the mortality observation, the D1 group showed the highest survival rate, i.e. 45% versus 21% observed for D4. An evident difference of survival rates between D1 and D4 groups was observed.
The possible relevance of antibodies in the protection elicited by D1 and D4 was investigated by a passive serum transfer experiment, with groups of 8 mice. The results are shown in Fig.  3, and detailed statistical analysis is provided in supplemental Table S5C. Both anti-D1 and anti-D4 sera elicited significant protection against bacteremia (Fig. 3A), with cfu geometric means lower by 2.6 and 1.6 logs, respectively, than that of the control group. Only anti-D1 serum afforded significant protection against mortality, giving 100% survival rate, whereas santi-D4 gave a protective trend, with 60% survival rate. These results indicate that immunization with D1 domain and at a lower extent D4 domain, elicits functional antibodies that may play a role in the protection.
D1 and D4 Antisera Recognize the Native Pilus and Linear Epitopes within RrgB-To investigate the differences in protective efficacy exerted by the isolated RrgB domains with respect to the FL RrgB in the in vivo assays, mouse sera were tested for their capability of recognizing the native pilus and the recombinant FL RrgB. Sera were probed against entire TIGR4 bacteria by FACS analysis. Sera raised against D1, D4, and D1ϩD2ϩD3ϩD4 gave a fluorescence intensity comparable with that obtained with anti-FL RrgB, whereas D2-and D3-specific antisera recognized the native pilus less effectively. To explain the lower recognition efficacy observed with D2 and D3, the same sera were titrated in ELISA against the recombinant FL RrgB. As shown in supplemental Fig. S3, antibody titers elicited by D2 and D3 immunization toward FL RrgB were about 10 times lower than those obtained by immunization with D1, D4, D1ϩD2ϩD3ϩD4, and FL RrgB, consistent with the results obtained by FACS analysis on the polymerized native RrgB.
To gain more insights on the epitopes recognized by the protective D1 and D4 polyclonal antibodies, a PepScan approach, suitable for identifying linear epitopes, was applied. Overlapping 15-mer peptides with an offset of 5 residues were synthesized in situ on a glass fiber membrane. The library of peptides tested covered residues 25-190 of D1 and 444 -628 of D4. Incubation with the anti-D1 serum (previously used in passive protection experiments) revealed a single linear epitope covering residues 40 -59 (D1-1) (Fig. 4B), whereas anti-D4 serum detected a unique linear epitope (D4-1) spanning residues 494 -508 (Fig. 4C).
The structure was well defined with the exception of three long loops corresponding to the stretches 56 -69, 148 -162, and 173-177 (supplemental Fig. S1). Heteronuclear relaxation measurements revealed that residues in the first two loops had heteronuclear NOE and longitudinal R 1 values lower and higher than average, respectively (supplemental Fig. S2). This behavior was a consequence of local internal motions occurring on a faster time scale with respect to the overall reorientational correlation time ( r ) of the molecule and accordingly, a correlation time ( e ) for these fast motions can be fitted for the previous mentioned loops (supplemental Fig. S2). In addition, conformational exchange processes, occurring on the millisecond-microsecond scale, affected some residues located in the region 155-164, as monitored by transverse R 2 relaxation rates higher than the average (supplemental Fig. S2). These data indicated that these loops experience higher flexibility, showing accordingly a low number of long ranges 1 H-1 H NOEs (supplemental Fig. S5). For a few residues, the assignment of the backbone resonances was also not achieved (Glu-143, His-145, Ser-146, Ser-148, Thr-149, Tyr-150, Val-152, and Gly-160), likely as a consequence of an increased local mobility. The loop between strands ␤ 9 and ␤ 10 (residues 148 -162) was the most disordered, with a r.m.s.d. value of about 2 Å (supplemental Fig. S1). Conformational exchange processes on the millisecond-microsecond time scale are observed also for the loop between strands ␤ 10 and ␤ 11 comprising residues 174 -178.
The protein core is characterized by hydrophobic interactions between residues located on the first (strands ␤ 1 , ␤ 8 , ␤ 11 ) and second (strands ␤ 4 , ␤ 7 , ␤ 9 , ␤ 10 ) sheets. A salt bridge between two complementarily charged side chains of residues Lys-41 and Glu-143 was also present. The aliphatic side chains of residues Met-48, Ile-52, Ala-53, and Leu-56, all located on one side of the ␣-helix, formed hydrophobic interactions with  3.4 Å). Like D1, none of these domains contained intramolecular isopeptide bonds. However, only the SpaA domain is located at the N terminus of pilus backbone protein as D1; their overlay is presented in Fig. 6A.
In an attempt to determine the orientation of D1 with respect to the D2-D4 RrgB portion, a rigid body fitting of all the domains into the shape of the native pilus obtained from cryoelectron microscopy (cryo-EM) was attempted (24). Initial rigid body fitting of the D2-D4 crystal structure into the cryo-EM density map of the native pilus left an apical unoccupied volume, likely due to the absence of D1. However, when the simultaneous fitting of the D1 and the D2-D4 domains was carried out, some portions of D1 could not be accommodated into the apical empty volume (supplemental Fig. S6 and supplemental Table S6). To model the FL RrgB, we merged the D1 and D2-D4 structures into a single molecule by overlapping residues 188 -191, shared by the C terminus of D1 and the N terminus of D2-D4. The relative orientation of D1 with respect to the others domains was then varied to best fit into the cryo-EM map (Fig. 6B). The final model, which represents only one of the possible orientations of D1 with respect to the rest of the protein, does not present steric clashes between D1 and the D2-D4 domains. The two linear epitopes, previously identified to be located within the D1 and D4 domains by peptide hybridization with specific protective polyclonal antibodies (Fig. 4), when mapped onto the FL model of RrgB, confirmed their superficial localization (Fig. 6B).
Lysine 183 of D1 Is Required for Intermolecular Isopeptide Bond Formation and Pilus Polymerization-Pili of Gram-positive bacteria are polymerized by means of intermolecular isopeptide bonds occurring between the Thr of the C-terminal LPXTG motif of a RrgB molecule and a Lys located at the N terminus of the following molecule (18,29,30). To identify the Lys implicated in the intermolecular isopeptide bond formation between two consecutive RrgB monomers, a TIGR4 RrgB deletion mutant (no longer able to assemble a pilus on its surface) was created. Subsequently, RrgB expression and pilus polymerization were restored in TIGR4⌬RrgB by transforming the mutant with plasmids expressing either wild-type RrgB or RrgB mutated in single Lys residues (Lys 3 Ala substitutions). Sequence analysis revealed that the D1 contains the canonical 181 YPKN 184 pilin motif, with Lys-183 nicely superimposing onto the functional SpaA Lys-190, as shown in Fig. 6A. Noteworthy, D1 also presents the sequence fragment 160 GSKAVP 165 , similar to the motif containing the Lys residue functional for the pilus assembly in S. pyogenes Spy0128 (31). Lys-183 and Lys-162, along with two additional Lys residues (Lys-138 and Lys-309), located on D1 and D3 domains, respectively, were selected and mutated.
WB analysis performed on whole cell bacterial lysates using rabbit polyclonal antisera raised against RrgA, RrgB, or RrgC revealed that all of the pilus proteins were expressed in all of the complemented mutants (Fig. 5). However, the typical pilus-associated high molecular weight ladder was revealed only in the case of TIGR4⌬RrgB complemented with RrgB wild type or RrgB mutated in Lys-162, Lys-138, or Lys-309. In contrast, in the mutant complemented with K183A, RrgB was detectable only as a monomer (Fig. 7), clearly indicating that Lys-183 is the residue implicated in intermolecular isopeptide bond formation.
Moreover, in the presence of the RrgB K183A substitution only hetero-oligomers composed of RrgA-RrgC (about 160 kDa) or RrgB-RrgC (about 110 kDa) could be detected by WB analysis (see arrows in Fig. 7, A and C). The absence of RrgA-RrgB heterodimers suggests that RrgA and RrgB are linked exclusively through intermolecular isopeptide bonds involving the D1 residue Lys-183 of RrgB.

DISCUSSION
Ever since their initial discovery, pili of Gram-positive bacteria have received considerable attention because they are associated with a number of different virulence mechanisms (15,16,33) and elicit protection in animal models (32, 49 -52). In particular, the S. pneumoniae pilus was found to be implicated in the initial attachment to epithelial cells (15,16,33), and the pilus components, being protective in mouse models of infection, are regarded as potential vaccine candidates (32,33). The major pilin RrgB is organized into four Ig-like domains (D1-D4), as shown by combined structural approaches (Ref. 24 and this work). We have tested the four RrgB domains in animal model experiments and have shown that the protective efficacy exerted by the combination of the four RrgB domains D1ϩD2ϩD3ϩD4 is comparable with that afforded by the FL RrgB. Among the single domains, D1 is the most protective, and D4 retained an important part of the protective efficacy of the FL protein. The lower protection achieved by D2 and D3 compared with the FL protein, as well as with the D1 and D4, is probably because the antibodies elicited by the former two domains recognize the FL protein less efficiently, in both its native and recombinant forms.
This may be the result of smaller exposed surface areas experienced by D2 and D3 in the FL RrgB compared with the D2 and D3 isolated domains and with D1 and D4. It is possible, in fact, that the antibodies generated against the isolated domains are recognizing areas of D2 and D3 that are buried in the FL protein. Alternatively, D2 and D3 could assume a slightly different conformation when expressed as single domains, thus generating nonfunctional antibodies. On the other hand, the two linear epitopes (Fig. 6B) identified within D1 and D4 by PepScan analysis performed with protective polyclonal antibodies raised against the two domains (conformational epitopes are not detectable with this method) are well exposed on the surface of the RrgB molecule and could contribute to the protective activity exerted by D1 and D4. Further experiments are needed to understand to what extent these linear epitopes contribute to the protective activity exerted by the two domains. Taken together, these results suggest that RrgB contains multiple protective epitopes, thus confirming the potential of this vaccine candidate. Furthermore, although the existence of possible conformational epitopes involving residues from different domains cannot be excluded, their contribution to the overall protective efficacy might not be essential.
To obtain more information about the structural role of D1 and to try to correlate it with the protection data discussed above, we solved the solution structure of this domain by using NMR spectroscopy. D1 shows an Ig-like fold, does not contain any intramolecular isopeptide bond, and has many flexible regions. The observed D1 flexibility, indeed, could play a fundamental role in the specific antigen-antibody recognition process (53), thus accounting for D1 enhanced protection capability with respect to the more rigid D2-D4 domains, each one containing an intramolecular isopeptide bond. In fact, the pro-tein structural plasticity could be related to the ability of D1 to undergo local conformational changes and to adapt its structure to optimize the interactions with the antibodies and increase the affinity and the specificity of the antigen-antibody recognition process. The dynamics of D1 could therefore strongly contribute to the interface adaptation for molecular recognition such that the antibody can select an optimal conformer from a wide distribution of possible D1 conformations. The rigid structure of the D2-D4 region prevents such an effective conformational selection for these domains. The above described phenomenon has been observed for other proteinprotein or protein-DNA interaction processes (53,54).
To shed light on the molecular mechanism driving pilus polymerization in S. pneumoniae, we investigated which of the lysine residues of D1 was engaged in the intermolecular isopeptide bond formation. Site-directed mutagenesis followed by complementation identified Lys-183 as crucial for the pilus assembly. This result is consistent with the observation that the spatial position of RrgB Lys-183 can be superimposed onto Lys-190 of the C. diphtheriae pilus backbone subunit SpaA, known to be involved in the intermolecular isopeptide linkage (30). Interestingly, as shown in Fig. 6A, both lysines are not fully available to form an external bond. In particular, the average relative solvent accessibility of the Lys-183 over the D1 family of conformers is 27.2% Ϯ 4.8, as the Lys side chain projects into a cleft between the main body of the protein and the segment 40 -78 containing the mobile loop 56 -69 (27). This suggests that pilus backbone proteins, to be polymerized, might undergo conformational changes, probably involving not only the Lys residue (Lys-183) but also the flexible regions spatially close to it (49 -69, 152-167, and 183-193), to allow the formation of the covalent intermolecular isopeptide bond. NMR mobility data indicate that the C terminus of the D1, where Lys-183 is located, and other loops are highly mobile and that such dynamics could be relevant for the intermolecular isopeptide bond formation (supplemental Fig. S2). Consistently, the absence of stabilizing intramolecular isopeptide bonds renders D1, unlike the other domains, less rigid and prone to conformational rearrangements. Furthermore, the occurrence of a struc- tural rearrangement of D1 within the native pilus structure is also in line with the partial fitting of its NMR structure onto the molecular shape of the native pilus determined by cryo-EM.
Mutagenesis and complementation data were used to analyze the covalent links established among the three pilus proteins either in the absence of RrgB or in the presence of the RrgB K183A mutant, in an attempt to provide further insights into the possible organization of the native pilus. In the presence of the nonpolymerizing RrgB K183A mutant, the lack of RrgA-RrgB heterodimers provides evidence that RrgA and RrgB are unidirectionally linked only through the Lys-183 of RrgB and the C-terminal Thr of RrgA. This arrangement is in accordance with the model proposed by Hilleringmann et al., positioning RrgA at the pilus terminus, thus ruling out the alternative possibility of RrgA being incorporated along the pilus shaft (22). Conversely, an RrgA-RrgC multimer is detectable either when RrgB is not expressed, as already reported by Falker et al. (20) and Le Mieux et al. (55), or in the presence of nonpolymerizing RrgB K183A. In this case, the mutated RrgB is competing with RrgA for the linkage to RrgC, as demonstrated by the presence of RrgB-RrgC hetero-oligomers (Fig. 7, B and C). Concomitant detection of RrgA-RrgC hetero-oligomers even under these conditions further strengthens the idea that in wild-type bacteria, although not detectable by electron microscopy analysis of purified pili, a fraction of RrgA and RrgC might be directly linked to each other.
In conclusion, this study provides additional information elucidating pilus proteins features and also paves the way to the rational design of new RrgB-based molecules to implement a protein-based vaccine against pneumococcal disease. Moreover, the newly acquired structural and dynamic information on the RrgB molecule provided by this study suggests that the conformational flexibility of D1 is pivotal for the protein-antibody recognition process. These findings together with the new functional information could be used to better understand pilus functions and its role in pathogenesis.