The Streptococcus pneumoniae adhesin PsrP binds to Keratin 10 on lung cells

Authors


*E-mail orihuela@uthscsa.edu; Tel. (+1) 210 567 3973; Fax (+1) 210 567 6612.

Summary

Pneumococcal serine-rich repeat protein (PsrP) is a pathogenicity island-encoded adhesin that mediates attachment to lung cells. It is a member of the serine-rich repeat protein family and the largest bacterial protein known. PsrP production by S. pneumoniae was confirmed by immunoblotting and a truncated version of the protein was determined to be glycosylated. Using isogenic psrP mutants complemented with various PsrP constructs and competitive inhibition assays with recombinant proteins, we determined that PsrP requires an extended SRR2 domain for function and that adhesion is mediated through amino acids 273–341 of its basic region (BR) domain. Affinity chromatography, immunoprecipitation, enzyme-linked immunosorbent assay (ELISA), fluorescent-activated cell sorting (FACS) and immunofluorescent colocalization studies determined that PsrP binds to Keratin 10 (K10) on the surface of lung but not nasopharyngeal epithelial cells. Unglycosylated K10 bound to wild type but not psrP deficient pneumococci; suggesting that unlike other serine-rich repeat proteins, PsrP-mediated adhesion is independent of lectin activity. Finally, mice immunized with recombinant (r)PsrPBR had significantly less bacteria in their blood and improved survival versus controls following intranasal challenge. We conclude that the BR domain of PsrP binds to K10 in a lectin-independent manner, that K10 is expressed on lung cells and that vaccination with rPsrPBR is protective against pneumococcal disease.

Introduction

Streptococcus pneumoniae (the pneumococcus) is a leading cause of community-acquired pneumonia, sepsis and meningitis. Primarily a commensal, invasive pneumococcal disease (IPD) is characterized by spread of the bacteria from the nasopharynx to normally sterile sites such as the lungs, blood and central nervous system. At risk for IPD are young children, the elderly and individuals who are immunocompromised or have underlying medical conditions such as chronic obstructive pulmonary disease. Worldwide, it is estimated that S. pneumoniae is responsible for 15 cases of IPD per 100 000 persons per year and over 1.5 million deaths annually (Anonymous, 1999; Lexau et al., 2005). Of note, the preponderance of invasive disease is the result of infection with relatively few invasive clones (Beall et al., 2006), a finding that suggests that invasive clones carry genes that facilitate disease progression that are absent in non-invasive isolates (Obert et al., 2006; Embry et al., 2007).

In 2006, comparative genomic analyses of 42 invasive and 30 non-invasive clinical isolates of S. pneumoniae determined that the presence of psrP-secY2A2, a 37 kb genomic island, was positively correlated with the ability of S. pneumoniae to cause human IPD (Obert et al., 2006). Analysis of the nucleotide sequence by blastn, and the predicted coding regions by blastx, found the operon structure and gene content of psrP-secY2A2 to be homologous to loci present in oral streptococci and Staphylococcus aureus that encode a family of proteins called serine-rich repeat proteins (SRRP). In general, SRRPs are large glycosylated proteins composed primarily of serine-rich amino acid repeats (SRR); they have been shown to be adhesins and their presence has been associated with the detection of fimbriae-like structures on the surface of Streptococcus parasanguis and Streptococcus cristatus (Wu et al., 1998; Froeliger and Fives-Taylor, 2001; Takahashi et al., 2002; Bensing et al., 2004a; Handley et al., 2005; Plummer et al., 2005; Siboo et al., 2005; Obert et al., 2006; Samen et al., 2007).

The SRRP domain structure typically consists of an amino-terminus large cleavable signal peptide, a short serine-rich repeat region (SRR1), a unique non-repeat region (NR), followed by a second extremely long serine-rich repeat area (SRR2) and cell wall anchor domain. SRRPs have been shown to be glycosylated intracellularly and to require non-canonical accessory Sec components for their transport to the bacteria surface (Bensing et al., 2004b; Chen et al., 2004; Takamatsu et al., 2005a; Zhou and Wu, 2009). Takahashi and colleagues have hypothesized that the extremely long SRR2 domain of SRRPs serves to extend the NR domain outward away from the bacteria so that they can mediate adhesion (Takahashi et al., 2002; 2004). Consistent with this model, the NR domain of several SRRPs have been shown to mediate adhesion and to bind glycoconjugates on a variety of host proteins including those in saliva and on the surface of red blood cells and platelets (Bensing et al., 2004a; Takahashi et al., 2004; Plummer et al., 2005; Takamatsu et al., 2005b; 2006). Importantly, SRRPs have been implicated in biofilm formation and the development of diseases such as bacterial endocarditis, neonatal sepsis and meningitis. For example, deletion of sraP, hsa and gspB, genes encoding SRRPs in S. aureus and Streptococcus gordonii, have been shown to decrease the ability of these bacteria to form vegetative plaques on heart valves of catheterized rats (Siboo et al., 2005; Takahashi et al., 2006; Xiong et al., 2008). Likewise, Srr-1 of Streptococcus agalactiae has been shown to bind human Keratin 4 and to promote adherence to epithelial cells and human brain microvasculature endothelial cells (Samen et al., 2007; van Sorge et al., 2009). Thus, SRRPs mediate bacteria–host cell interactions that are important for disease.

In S. pneumoniae serotype 4 strain TIGR4, psrP-secY2A2 is 37 kb in length. In addition to the SRRP called pneumococcal serine-rich repeat protein (PsrP), psrP-secY2A2 encodes 10 glycosylases and 7 components of an accessory Sec system (Tettelin et al., 2001). Based on their near-identical homology to genes within the S. gordonii locus gspB-secY2A2, these genes are putatively responsible for the glycosylation and transport of PsrP to the surface (Obert et al., 2006). PsrP is composed of 4776 amino acids (AAs) and is the largest bacterial protein known (Fig. 1A). The SRR1 and SRR2 domain of PsrP is composed of 8 and 539 repeats of the AA sequence SAS[A/E/V]SAS[T/I] respectively. The NR domain of PsrP is composed primarily of basic AAs and has a predicted pI value of 9.9. For this reason it is called the basic region domain (BR). To date three studies have shown that deletion of psrP or its accessory genes attenuates S. pneumoniae virulence. First, in a signature-tagged mutagenesis screen, Hava et al. found that mutants in four psrP-secY2A2 genes failed to passage successfully through live mice (Hava and Camilli, 2002). Second, our own studies found that a psrP mutant had reduced capacity to enter the bloodstream and kill mice following intranasal challenge (Obert et al., 2006). Third and most recent, we determined that PsrP mediates attachment to bronchial and alveolar but not nasopharyngeal epithelial or vascular endothelial cells and that mutants deficient in psrP were unable to establish lung infection in mice (Rose et al., 2008). Thus considerable evidence indicates that PsrP is a pathogenicity island-encoded adhesin that contributes towards the development of pneumococcal pneumonia.

Figure 1.

Immunodot blot and Western blot analysis of PsrP production.
A. Domain structure of PsrP. Note that the NR domain of PsrP is named BR due to its predicted pI of 9.9.
B. Immunodot blot of bacteria whole-cell lysates (WCL) from exponential growth phase cultures. Numbers correspond to individual clinical isolates found to carry psrP by PCR.
C. Western blot detection of full-length native PsrP using bacteria WCL and purified cell wall fractions (CWF) from TIGR4 and T4 ΩpsrP. Proteins were resolved on 1.5% agarose gels.
D. Western blot and carbohydrate staining of WCL from T4 ΩpsrP with pfcsRK::PsrP1-734 grown in THY media supplemented with 1% fucose or 1% sucrose. Carbohydrate staining was done using a Periodic Acid-Schiff (PAS) stain. Note the detection of two fucose-inducible bands at 80 and 210 kDa in T4 ΩpsrP with pfcsRK::PsrP1-734, also that only the 210 kDa band tested positive for carbohydrates by PAS.
E. Detection of native full-length PsrP on the bacteria surface by immunofluorescent microscopy. FITC-labelled bacteria were fixed (green) and incubated with antibody against rPsrPSRR1-BR and PE-labelled secondary antibody (red). Images were captured at 1000× as described in the Experimental procedures section.

In this manuscript we sought to fully characterize the molecular basis for PsrP-mediated pneumococcal pathogenesis. To this end, we have determined that psrP-secY2A2 is present in invasive clones that are globally distributed and serotypes that are not protected against by the current conjugate vaccine. We have identified the minimal region of the PsrP BR domain that is responsible for pneumococcal adhesion and determined that an extended SRR2 domain is necessary for function in an encapsulated background. We identified the host ligand of PsrP as Keratin 10 (K10) and determined that PsrP–K10 interactions are independent of lectin activity. Importantly, we found that K10 was not present on nasopharyngeal cells helping to explain why deletion of psrP had no effect on nasopharyngeal colonization (Rose et al., 2008). Finally, we demonstrate that immunization with the BR domain of PsrP protects mice against pneumococcal challenge. Thus we conclude that recombinant (r)PsrPBR may be a suitable candidate for inclusion in a multicomponent protein vaccine designed to protect against S. pneumoniae.

Results

psrP-secY2A2 is present in globally distributed invasive clones

The presence of psrP-secY2A2 in S. pneumoniae has been positively correlated with the ability of serotype 6A and 6B to cause human IPD (Obert et al., 2006). To better understand its distribution among other invasive serotypes, we searched for its presence among the 19 S. pneumoniae isolates whose genome has been sequenced and whose nucleotide data are publicly available through the National Center for Biotechnology Information Genome Project or the S. pneumoniaeblast server. Briefly, the majority of these isolates were selected for sequencing because they are representative of globally distributed clonal complexes that cause considerable morbidity and mortality worldwide. Using blastn analysis we identified homologues (E < 1e−5) of TIGR4 psrP-secY2A2 genes in 6 of the 19 genomes examined (32%) and 6 of 14 (43%) different serotypes represented (present in serotypes 1, 4, 6A, 14, 19A and 23F; Table S1). Clinical isolates carrying psrP-secY2A2 belonged to clonal complexes known to cause IPD in the United States as well as abroad (Table S2; see Discussion) (Beall et al., 2006; Blomberg et al., 2009). Previously, we had also detected psrP-secY2A2 in serotype 19F, strain EF3030 (data not shown), thus psrP-secY2A2 was detected in half of the serotypes (8 of 16) that have been searched to date. Importantly, with exception to the number of SAS[A/E/V]SAS[T/I] repeats encoded in the SRR1 and SRR2 domains of psrP, blast analyses determined that PsrP was conserved across all the genomes carrying the pathogenicity island (Table S2).

PsrP production in TIGR4

Despite considerable evidence indicating that PsrP is an important virulence determinant, production of PsrP by S. pneumoniae had not yet been confirmed. To do this we first performed immunodot blot analysis of cell lysates from TIGR4, an isogenic mutant deficient in psrP called T4 ΩpsrP, and 12 clinical isolates from individuals with IPD found to carry psrP by PCR. Using antibodies against rPsrPSRR1-BR (Obert et al., 2006) we detected PsrP in TIGR4, 11 of the 12 clinical isolates, but not in T4 ΩpsrP (Fig. 1B). Thus, TIGR4 and the majority of clinical isolates tested produced PsrP. Subsequently using an agarose gel to resolve proteins, we detected full-length PsrP in whole-cell lysates and cell wall fractions from TIGR4 but not T4 ΩpsrP by Western blot (Fig. 1C). Briefly, a band greater than 512 kDa, the predicted size of PsrP, was detected in the TIGR4 but not T4 ΩpsrP samples, suggesting that PsrP was post-translationally modified.

The SRRPs are glycosylated and have been found to run larger on SDS-PAGE gels than their predicted molecular weight (Bensing and Sullam, 2002; Takahashi et al., 2002; Siboo et al., 2005; Seifert et al., 2006). To better test for this possibility we created a plasmid encoding a truncated version of PsrP that was under the control of the fucose promoter fcsRK (Chan et al., 2003) and examined the recombinant protein following SDS-PAGE. pfcsRK::PsrP1-734 encodes the SP, SRR1 and BR domains of PsrP as well as 31 repeats of the SAS[A/E/V]SAS[T/I] motif in the SRR2 domain (rPsrP1-734). Following transformation of T4 ΩpsrP with pfcsRK::PsrP1-734, Western blot analysis of SDS-PAGE separated proteins confirmed fucose-inducible production of two seroreactive bands in cell lysates (Fig. 1D). The first and strongest band separated at a molecular weight corresponding to 210 kDa; the second and weaker band corresponding to 80 kDa, the predicted size of PsrP1-734. To test if the 210 kDa band was glycosylated, cell lysates were separated by SDS-PAGE and tested for carbohydrates using the Periodic Acid-Schiff stain. Figure 1D demonstrates that a 210 kDa protein stained positive for carbohydrates whereas the region corresponding to 80 kDa remained clear. Thus the 210 kDa band was most likely glycosylated PsrP1-734.

Assuming that native PsrP is modified in an equivalent manner as PsrP1-734, we calculated that full-length PsrP would separate at a size corresponding to 2300 kDa. This predicted size for PsrP is consistent with the high-molecular-weight bands that were detected by Western blot (Fig. 1C). Importantly, native PsrP was also confirmed to be on the bacteria surface by immunofluorescent microscopy (Fig. 1E). Thus PsrP is produced by S. pneumoniae and is available on the bacteria surface for interactions with host cells.

PsrP adhesion is mediated by the BR domain and is dependent on SRR2 length

Antibodies against the amino terminus of PsrP have been shown to neutralize TIGR4 adhesion (Rose et al., 2008). Because other SRRPs mediate adhesion through their NR domain (Bensing et al., 2004a; Takahashi et al., 2004; Plummer et al., 2005; Takamatsu et al., 2005b; 2006), we sought to confirm this for PsrP using a genetic approach. To do this we created a plasmid encoding a truncated version of PsrP with 33 SRRs in its SRR2 region (PsrPSRR2(33)) and a version of this protein deficient solely in the BR domain (PsrPSRR2(33)-BR; Fig. 2A). A truncated version of PsrP was used because we were unable to amplify full-length psrP or stably clone any psrP fragments encoding a SRR2 domain with more than 33 SRRs. Complementation of T4 ΩpsrP with PsrPSRR2(33) failed to restore the ability of T4 ΩpsrP to adhere to A549 cells (Fig. 2B). However, complementation of T4R ΩpsrP, an unencapsulated derivative of TIGR4 (Gosink et al., 2000), restored adhesion to levels threefold greater than the mutant and 40-fold greater than the empty vector control (Fig. 2B). Importantly, no effect on adhesion was observed for T4R ΩpsrP complemented with PsrPSRR2(33)-BR; moreover, delivery of the truncated PsrPs to the bacteria surface was confirmed by immunofluorescence assay (Fig. 2B). Thus, these experimental results suggest that the BR domain mediates PsrP adhesion and that an extended SRR2 domain is necessary in an encapsulated background. Presumably, the SRR2 domain serves to extend the BR domain outward through the capsular polysaccharide.

Figure 2.

The BR domain of PsrP mediates adhesion.
A. Illustration of the psrP loci in TIGR4 and T4R, T4 ΩpsrP and T4R ΩpsrP, and assorted pNE1 plasmids encoding truncated versions of psrP.
B. Relative adhesion of these strains to A549 cells. Values plotted are normalized to TIGR4 (378 ± 52 cfu/105 cells) or T4R (1030 ± 56 cfu/104 cells). Statistical analysis was performed using a one-way anova. Asterisks indicate a statistical significant difference versus TIGR4 or T4R respectively (P < 0.05), whereas a cross indicates statistical significance versus the corresponding empty vector control. Values are expressed as mean ± SEM.
C. Detection of truncated PsrP constructs on the surface of S. pneumoniae. T4R ΩpsrP carrying pNE1 with assorted PsrP constructs were FITC-labelled and fixed (green). Bacteria were subsequently incubated with antibody against rPsrPSRR1-BR and PE-labelled secondary antibody (red). Images were captured at 1000× as described in the Experimental procedures section. Note detection of PsrP on the surface of bacteria expressing PsrPSRR2(33) and PsrPSRR2(33)-BR but not the T4R ΩpsrP or the empty vector control, indicating that the inability of PsrPSRR2(33)-BR to mediate adhesion is not the result of an absence of surface expression.

Adhesion of PsrP to the cell surface is mediated by AAs 273–341 of the BR domain

We next sought to determine the minimal fragment necessary to mediate bacterial attachment to lung cells. Using overlapping rPsrP constructs, we first determined that TIGR4 adhesion could be blocked by incubation of cells with fragments of PsrP that contained AAs 273–341 of the BR domain (% WT binding: rPsrPSRR1-BR: 27%; rPsrPBR: 7%, rPsrPBR.C: 12%, rPsrPBR.C.2: 4%; Fig. 3A). rPsrP fragments not containing this AA sequence, those composed of smaller fragments, and a BR construct lacking AAs 273–341 (rPsrPBRΔC.2) failed to inhibit TIGR4 adhesion to A549 cells. Using latex microspheres coated with equivalent amounts of protein, we subsequently determined the minimal section of PsrP capable of mediating the adherence of a bacterium-sized particle. Uniform with the results obtained from the inhibition assays, 1–2 μm latex beads coated with rPsrP constructs containing AAs 273–341 bound to lung cells at levels 8–20-fold greater than the bovine serum albumin (BSA) control or beads coated with other sections of PsrP (Fig. 3B, Fig. S1). Of note, beads coated with rPsrPBR.C.2.β (AAs 291–325) attached to A549 cells the best, suggesting that the primary binding domain of BR is ≤ 34 AAs in length, although flanking areas may stabilize PsrP bead–cell interactions.

Figure 3.

Amino acids 273–341 of the BR domain mediate attachment to lung cells.
A. Adhesion of TIGR4 following preincubation of A549 cells with media containing 1.0 μM of the designated rPsrP constructs. All values were normalized against cells incubated with BSA (750 ± 59 bacteria/106 cells).
B. Adhesion of 1–2 μm diameter fluorescent latex microspheres coated with rPsrP constructs to A549 cells. Values were normalized against beads coated with BSA [16 ± 4 beads/200× field of view (FOV)]. Representative micrographs of the beads attached to the cell monolayer are shown in Fig. S1.
C. Micrographs of Detroit, LA-4 and A549 cells fixed and stained with Cy3-labelled rPsrP constructs (red). Cell nuclei were stained with DAPI (blue). For (A) and (B) statistical analyses were performed using a one-way anova. Significant differences (P < 0.05) versus the BSA controls are indicated by an asterisk. Values are expressed as mean ± SEM.

In previous studies, deletion of psrP had been shown to negatively affect TIGR4 binding to bronchial epithelial cells and type II pneumocytes but not to nasopharyngeal epithelial cells (Rose et al., 2008). To test whether the rPsrP fragments maintained their cell-type specificity, we tested the ability of Cy3-labelled rPsrPSRR1, rPsrPBR, rPsrPBR.C and rPsrPBR.C.2 to bind to the surface of non-permeabilized Detroit 562 nasopharyngeal epithelial cells, LA-4 bronchial epithelial cells and A549 cells (Fig. 3C). Cy3-labelled BR constructs containing AAs 273–341 bound to LA-4 and A549 lung cells but not to nasopharyngeal cells. This suggested that the BR fragments maintained their cell specificity and that their ligand was restricted to the surface of lung cells. Importantly, Cy3-labelled rPsrPSRR1 failed to bind any of the cell types tested. This finding, along with those from the competitive inhibition assays, and the latex bead assays, strongly suggest that the SRR1 domain of PsrP is not involved in bacterial adhesion, although this remains to be proven genetically.

PsrP binds to K10 on lung cells

To identify the PsrP ligand on human lung cells, rPsrPBR was used as bait in column chromatography experiments. A549 cell lysates were passed over an rPsrPBR column. When bound proteins were eluted, separated by SDS-PAGE, stained and analysed by maldi-tof, the proteins K1, K9 and k10 were identified. The identity of the band as K10 was subsequently determined by reactivity of the band with K10 antibodies in Western blot experiments (data not shown). The interaction of PsrP with K10 was confirmed threefold. First, by immunoprecipitation, sepharose G beads coated with K10 antibodies coimmunoprecipitated a 37 kDa protein identified to be rPsrPBR in cell lysates spiked with the recombinant protein (Fig. 4A). Second, by enzyme-linked immunosorbent assay (ELISA), wells coated with rPsrPBR and rPsrPBR.C bound to K10 in LA-4 and A549 cell lysates (Fig. 4B), but not to K1 or K9 in the same lysates (data not shown). Third, by immunofluorescence, Cy3-labelled rPsrPBR colocalized with K10 on the surface of non-permeabilized A549 and LA-4 cells, as well as on alveolar epithelial cells and bronchial epithelial cells in lung sections taken from mice (Fig. 5).

Figure 4.

PsrP binds to K10 on lung cells.
A. Co-immunoprecipitation of rPsrPBR (37 kDa) from type II pneumocyte (A549) cell lysates spiked with rPsrPBR (BR) using antibodies against K10. Immunodetection of rPsrP was done using monoclonal antibodies against the hexahistidine tag on the carboxy terminus of the recombinant protein. Note that anti-K10 antibodies immunoprecipitated (IP) rPsrP only in samples where the cell lysate and rPsrPBR have been mixed together. As a control we demonstrate that antibodies against rPsrPBR.C.2 immunoprecipitate a band corresponding to the size of rPsrPBR. The 70 kDa band is IgG heavy chain and was detected as a result of non-specific binding.
B. Binding of rPsrP constructs to K10 in assorted cell lysates as measured by ELISA. Note that rPsrP fragments containing AAs 291–325 bound to K10 in cell lysate from murine bronchial epithelial cells (LA-4) and A549 cells but not human nasopharyngeal cells (Detroit 562) cells.
C. Western blot detecting K10 in cell lysates from LA-4 and A549 cells as well as whole lungs isolated from young mice. K10 was not detected in cell lysates from Detroit 562 cells. For ELISA experiments, statistical analyses were performed using a One-Way anova. Significant differences (P < 0.05) versus the blank (BSA-coated) controls are indicated by an asterisk. Values are expressed as mean ± SEM.

Figure 5.

Colocalization of PsrP with K10 on the surface of lung cells. Micrographs of unpermeabilized Detroit, LA-4, A549 cells and lung tissue sections from Balb/cJ mice fixed and processed for detection of K10 (green), nuclear staining with DAPI (blue) and binding of Cy3-labelled rPsrPBR (red). Overlay images demonstrate colocalization of K10 and rPsrP binding on lung cells and lung tissue sections but not on Detroit 562 cells.

We also determined that changes in K10 production modulated PsrP-mediated S. pneumoniae attachment. When K10 was silenced in A549 cells using small interfering dsRNA (SiRNA), attachment of wild-type bacteria was reduced to < 50% the level of the scrambled control (Fig. 6A and B). Likewise, when K10 was overexpressed, pneumococcal attachment increased 1.7–2.3-fold over the empty vector control (Fig. 6C and D). Importantly, K10 was detected on the surface of A549 and LA-4 cells by flow cytometry (Fig. S2); moreover, changes in the expression of K10 had no effect on the adhesion of T4 ΩpsrP (Fig. 6B and D). Thus K10 on the surface of these cells was available for specific interactions with PsrP.

Figure 6.

Changes in K10 expression modulate S. pneumoniae adhesion.
A. Western blot demonstrating knockdown of K10 expression in A549 cells 2 days after transfection with SiRNA specific to K10.
B. Adherence of TIGR4 and T4 ΩpsrP to A549 cells transfected with SiRNA. Note that TIGR4, but not T4 ΩpsrP, adhesion was significantly diminished in cells treated with K10 SiRNA versus those treated with scrambled SiRNA.
C. Western blot demonstrating dose-dependent overexpression of K10 in A549 cells 2 days after transfection with pCK-10.
D. Adherence of TIGR4 and T4 ΩpsrP to A549 cells following transfection with pCK-10. Note that TIGR4, but not T4 ΩpsrP, adhesion is significantly enhanced in cells treated with 5, 10 and 50 ng of pCK-10 but not in cells treated with the empty vector control pCDNA (50 ng). Statistical analyses were performed using a one-way anova. Significant differences (P < 0.05) versus the scrambled SiRNA control, and the empty vector (pCDNA) controls, respectively, are indicated by an asterisk.

We subsequently tested for the presence of K10 on nasopharyngeal cells. All assays used including Western blot, ELISA, immunofluorescent imaging and cytometric analyses gave negative results for Detroit 562 cells (Figs 4B, C and 5, Fig. S2). This finding helps to explain why Cy3-labelled rPsrP products carrying AAs 273–341 failed to bind these cells (Figs 3C and 5), and why psrP mutants colonized the nasopharynx normally (Rose et al., 2008). The absence of K10 precluded the ability of PsrP to mediate bacterial attachment.

PsrP–K10 interactions are independent of lectin activity

Given that several SRRPs bind to glycoconjugates (Takahashi et al., 2002; 2004; Bensing et al., 2004a; Takamatsu et al., 2005b), we sought to identify the specific carbohydrate structure on K10 to which rPsrPBR may have bound to. Surprisingly, pretreatment of A549 cells with a diverse panel of lectins failed to identify any that inhibited bacterial attachment (Fig. S3A). Similarly, preincubation of TIGR4 with assorted carbohydrates or sialylated proteins failed to identify any inhibitors of bacterial adhesion (Fig. S3B). Importantly, no inhibition was observed despite treatment of cells with lectins that bound to neuraminic acid, αNeuNAC(2→6)gal and αNeuNAC(2→3)gal or after incubation of the bacteria with sialic acid or the sialylated protein fetuin. This is of note because other SRRPs have been demonstrated to bind to neuraminic acid on sialylated proteins (Bensing et al., 2004a; Takamatsu et al., 2005b). Unexpectedly, treatment of cells with Concanavalin A significantly increased binding of both TIGR4 and T4 ΩpsrP to A549 cells (Fig. S3A). This suggests that Concanavalin A served as a bridging molecule between the pneumococcus and the lung cells during infection. While the specific mechanism(s) responsible for this interaction were left unresolved, increased adhesion of T4 ΩpsrP indicates that it was a PsrP-independent mechanism.

To further test if PsrP–K10 interactions were the result of lectin activity, we treated A549 cells with neuraminidase, β-galactosidase and N-acetylglucosaminidase to remove carbohydrates from the cell surface. Treatment of cells with neuraminidase and N-acetylglucosaminidase increased bacterial attachment above the levels observed for the mock-treated controls (Fig. 7A). In addition, it was determined that recombinant K10 (rK10), produced in Escherichia coli, bound to TIGR4 and not T4 ΩpsrP (Fig. 7B). As proteins produced in E. coli are not glycosylated, we concluded that PsrP adhesion to K10 is not mediated through a carbohydrate moiety and the removal of glycoconjugates from the cell surface improves PsrP-mediated adhesion to K10.

Figure 7.

PsrP–K10 interactions are independent of lectin activity.
A. Adhesion of bacteria following treatment of A549 cells with neuraminidase (Neur), β-galactosidase (β-gal) and N-acetylglucsaminidase (GlcNAc). All values were normalized against bacterial adhesion to mock-treated cells (640 ± 90 bacteria/106 cells). Statistical analyses were performed using a one-way anova. Significant differences (P < 0.05) versus the controls are indicated by an asterisk. Values are expressed as mean ± SEM.
B. Immunofluorescent micrographs of TIGR4 and T4 ΩpsrP stained with FITC and incubated with rK10. K10 attached to the bacteria surface was visualized using monoclonal antibodies and Cy3-labelled secondary antibodies.

Immunization with rPsrPBR protects mice against pneumococcal challenge

Given that PsrP was detected in serotypes not covered by the current conjugate vaccine, that PsrP is conserved and that passive immunization with rPsrPSRR1-BR protected mice against challenge, we sought to determine if active immunization with rPsrP protected mice against S. pneumoniae. Three days post intranasal challenge, mice vaccinated with rPsrPSRR1-BR, rPsrPBR and rPsrPBR.C not only had a significant reduction in the number of bacteria in the blood (Fig. 8A), but also a significant reduction in mortality versus those vaccinated with the unrelated control protein Yersinia pestis Outer Membrane Protein Fraction 1 (rPsrPSRR1-BR: 23% mortality, P = 0.006; rPsrPBR: 13% mortality, P = 0.002; rPsrPBR.C: 6% mortality, P < 0.001; Fisher's exact test). For mice vaccinated with rPsrPBR and rPsrPBR.C protection extended beyond the third day with 72% and 45% of the mice surviving challenge respectively (Fig. 8B). Interestingly, mice vaccinated with rPsrPBR.C.2 were not protected against pneumococcal infection. This suggested that antibodies against the minimal binding domain of PsrP failed to neutralize bacterial adhesion or to opsonize the bacteria for phagocytosis, and that an alternate or larger PsrP structure was required to do so.

Figure 8.

Immunization with rPsrPBR is protective against pneumococcal challenge.
A. Bacterial titers of individual 5-week-old female Balb/cJ mice infected intranasally with TIGR4 (107 cfu) following immunization with assorted rPsrP constructs or the Y. pestis protein Fraction 1 as a control. Mice were administered 20 μg of protein mixed with Freund's adjuvant, boosted on days 21 and 42, and challenged intranasally on day 66. Horizontal bars indicate median bacterial value. Statistical analysis of bacterial titers was performed using a one-way anova. For statistical purposes blood titers of deceased animals at day 3 were treated as the value 8.0; for mice with undetectable levels of bacteria in the blood, titers were treated as the value 3.0.
B. Kaplan–Meier plot illustrating the survival of mice vaccinated with rPsrP constructs over time following intranasal challenge. Note that mice vaccinated with rPsrPBR and rPsrPBR.C had improved survival versus the control cohort. Statistical analyses were performed using a Kaplan–Meier Log Rank test. For each protein tested, at least three replicate experiments were performed each with four to six mice for a minimum of 12 mice sampled. All experiments were performed in compliance with an approved Institutional Animal Care and Use Committee protocols.

Discussion

Streptococcus pneumoniae is an opportunistic pathogen that is often part of the normal flora of the upper respiratory tract. While not all serotypes and clones are capable of causing IPD, due to selective pressure from the serotype-based polyvalent conjugate vaccine and widespread use of antibiotics, an increase in the incidence of disease caused by non-vaccine serotypes and invasive clones that are resistant to antimicrobials has occurred during the past 5–10 years (Critchley et al., 2007; Pichichero and Casey, 2007; Moore et al., 2008; Ongkasuwan et al., 2008). In this context, the presence of psrP-secY2A2 in six widespread invasive clonal complexes, as well as in 13 serotypes, 5 of which are not covered by the current conjugate vaccine (serotype 1, 6A, 10A, 15BC and 19A), suggests that psrP-secY2A2 is globally distributed and an important mechanism by which S. pneumoniae is able to cause human disease.

Among the clonal complexes determined to carry psrP-secY2A2 (Table S2), CC199 is particularly noteworthy. Based on numerous molecular epidemiological studies, CC199 has been shown to be responsible for 50–75% of the invasive disease that is caused by serotype 15BC, and 70–80% of disease caused by serotype 19A (Beall et al., 2006). Since 1999, serotype 19A has emerged to become the leading cause of IPD in children within the United States (Pichichero and Casey, 2007; Moore et al., 2008; Ongkasuwan et al., 2008). For example, in Massachusetts, pneumococcal disease caused by serotype 19A has risen from 10% of all cases during 2001–2002, to 41% of all cases from 2005 to 2006 (Critchley et al., 2007; Pelton et al., 2007). Importantly, during the past 5 years at least six new clonal complexes of serotype 19A have been identified. This includes clones resistant to all United States Food and Drug Administration-approved antibiotics for treatment of otitis media in children (Ongkasuwan et al., 2008). In order to better understand how these new invasive clones cause disease, it becomes important to determine if these emerging clonal complexes also carry the pathogenicity island psrP-secY2A2.

While several studies have shown that psrP is required for virulence, these studies are the first to show that PsrP is produced by S. pneumoniae and that PsrP is on the bacteria surface. Subsequent detection of rPsrP1-734 at 210 kDa, accompanied by positive staining for carbohydrates, is also the first experimental evidence to indicate that PsrP is glycosylated. Unmodified, rPsrP1-734 has a predicted molecular weight of 80 kDa and was detected by Western blot in small amounts. Given that rPsrP1-734 contains 39 SRRs (8 in SRR1, 31 in SRR2) and that the glycosylated version of this protein separated at 210 kDa, post-translation modification of rPsrP1-734 added 3.3 kDa in apparent molecular weight for each repeat that was encoded. Full-length PsrP, with 539 serine-rich repeats, would thereby be predicted to separate at ∼2300 kDa (512 kDa + 539 × 3.3 kDa), a size far larger than any other SRRP studied to date, and one consistent with the size observed for PsrP on Western blots. Currently, the identity and linkage of the sugars attached to PsrP are unknown. The polypeptide backbone of SRRPs such as Fap1 and GspB are O-linked to N-acetyglucosamine, N-acetlygalactosamine, rhamnose and glucose (Bensing et al., 2005; Zhou and Wu, 2009). However, psrP-secY2A2 is the most complex SRRP loci known and encodes six and four more glycosyltransferases than fap1-secY2A2 or gspB-secY2A2 respectively. Thus the chemical composition and linkage of the sugars attached to PsrP may be more complex than that of these proteins.

Capsular polysaccharide impedes host-receptor::pathogen ligand interactions and serves to protect the cell from phagocytosis. For the same reason, capsule also inhibits pneumococcal adhesion to epithelial and endothelial cells (Hammerschmidt et al., 2005). One reason that PsrPSRR2(33) may have failed to restore adhesion in a TIGR4 background is that a long SRR2 domain is required to extend the BR through the capsular polysaccharide. This notion was supported by the observation that complementation of T4R ΩpsrP, an unencapsulated mutant of TIGR4, with PsrPSRR2(33) restored pneumococcal adhesion. Electron microscopy of S. parasanguis and S. cristatus have shown that SRRPs can form fimbriae-like structures that extend outward from the cell wall (Wu et al., 1998; Handley et al., 2005). Presumably these structures serve to extend the NR domain outward, away from the bacteria, such that it can mediate adhesion to cells (Takahashi et al., 2002; 2004). Consistent with this hypothetical model, we have determined that the BR domain of PsrP is required for adhesion and that the BR domain mediates lung cell specificity. As all clinical isolates of S. pneumoniae are encapsulated, the long SRR2 domain of PsrP presumably allows the bacteria to attach to K10 on lung cells while simultaneously remain protected from alveolar macrophages and infiltrating neutrophils.

The inability of the BR-deficient PsrP construct to restore T4R ΩpsrP adhesion along with our findings that AAs 273–341 of the BR domain competitively inhibits pneumococcal binding, mediates attachment of latex beads to lung cells and retains the cell-type specificity observed for PsrP with whole bacteria, suggests that AAs 273–341 is the minimal binding domain of PsrP. Importantly, bead adhesion, but not competitive inhibition of bacterial attachment, was observed for the 34 AA fragment rPsrPBR.C.2.β (AAs 291–325). This suggests that primary binding domain of PsrP is located within this 34 AA region, but also that PsrP–K10 interactions are stabilized by the surrounding AAs. Interestingly, structural analyses of the BR domain using a bioinformatics approach suggests that the binding domain of PsrP (AAs 273–341) consists of four short β-strands interspersed with turns, suggestive of the formation of a β-sheet (data not shown). AAs 291–325 form the two middle β-strands within this putative secondary structure. Ideally, future protein structure studies can be used to determine how K10 and PsrP interact and possibly identify antagonists.

As indicated, the NR domain of GspB binds to sialyl antigen (NeuNac2-3Galβ(1-3)GalNAc) present on platelet membrane protein 1bα as well as to carbohydrates present on salivary proteins (Bensing et al., 2004a; Takamatsu et al., 2005b). Similarly, other SRRPs bind to assorted glyconjugates. Our finding that PsrP-mediated adhesion was not through a glyconjugate intermediate is the second report that SRRPs can mediate attachment to unglycosylated proteins. Interestingly, the first report being that of Srr-1, the S. agalactiae SRRP that binds to K4 (Samen et al., 2007). While the NR domains of PsrP and Srr-1 are not homologous and have very distinct pI values (Srr1 NR domain pI = 4.63), the fact that both proteins bind to a keratin suggests a conserved pathogenic mechanism that might also be present in other SRRPs. Importantly, the fact that PsrP binds to unglycosylated K10 does not exclude the possibility that PsrP also has lectin properties that mediate adhesion to other host proteins. This would not be unprecedented as other pneumococcal adhesins, such as Choline binding protein A, bind to multiple mammalian proteins through distinct binding sites of its R1 and R2 domains (Mitchell et al., 1991; Zhang et al., 2000; Dave et al., 2004; Orihuela et al., 2009). Of note, cell type-specific expression of K10 explains why rPsrP failed to bind Detroit nasopharyngeal cells and why deletion of psrP does not affect nasopharyngeal colonization (Rose et al., 2008).

A growing number of studies indicate that surface keratins play important roles in bacterial colonization of mucosal epithelial cells. For example, Clumping Factor B of S. aureus binds to K10 on desquamated human nasal cells scraped from the anterior nares (O'Brien et al., 2002; Walsh et al., 2004); likewise, Burkholderia cepacia binds to K13 (Sajjan et al., 2000), and Porphyromonas gingivalis adheres to an as-yet-unidentified keratin (Sojar et al., 2002). Keratins are divided into two subclasses, type I and type II. Keratin heterodimers, composed of a type I keratin and a type II keratin, polymerize to form intermediate filaments. These filaments form the principal structural elements of the epithelial cell cytoskeleton. Of note, K1, a type II keratin, pairs with K10, a type I keratin, both of which were identified by our maldi-tof analysis of affinity-purified proteins from A549 cell lysate with the rPsrP column. In total, there are 54 members of the keratin family and their distribution in different cell types has been well documented (Schweizer et al., 2006). K10 has long been known to be expressed in postmitotic, terminally differentiating epidermal keratinocytes; consequently, it is primarily found in the suprabasal layers of the epidermis (Moll et al., 1982). Most recently, K10 has been shown to be expressed on the surface of squamous epithelial cells in the nares (O'Brien et al., 2002). Herein, we have shown by Western blot, immunofluoresence, fluorescent-activated cell sorting (FACS) and ELISA that K10 is also present on the surface of bronchial epithelial cells as well as alveolar epithelial cells of the lungs. Furthermore, examination of lung sections using immunohistochemistry suggests that K10 is predominantly expressed by type I pneumocytes that cover a majority of the alveolar surface and are also terminally differentiated. Importantly, the fact that K10 was not detected on Detroit 562 cells suggests that expression of K10 is not continuous from the anterior nares to the lower respiratory tract, and that the posterior nasopharynx, where the pneumococcus colonizes mice, lacks K10 expression.

The SRRPs are found in Gram-positive bacteria predominantly among the oral streptococci. The conserved nature of PsrP among a diverse set of clones suggests that psrP-secY2A2 was horizontally acquired once by the pneumococcus and has since spread to multiple serotypes and clonal complexes. Our finding that active vaccination with the BR domain of PsrP protects mice against pneumococcal challenge also suggests that immunization with the NR domain of other SRRPs might protect against disease. In general, although the domain organizations of SRRPs are well conserved between different Gram-positive bacteria, the AA sequences composing the NR domains are not (Zhou and Wu, 2009). This suggests that cross-reactivity of NR antibodies would be limited and that it would be necessary to include multiple NRs from a panel of oral streptococci and staphylococci to protect against infective endocarditis. Despite this limitation, use of SRRPs as a vaccine may have therapeutic potential among individuals who are at high risk for the development of infective endocarditis.

Active immunization of mice with rPsrPBR and rPsrPBR.C protected mice against intranasal challenge with S. pneumoniae. This finding was consistent with previous studies that have shown rabbit antibodies against rPsrPSRR1-BR inhibited the ability of the pneumococcus to attach to A549 cells in vitro and passive immunization experiments that showed antibodies against rPsrPSRR1-BR reduced the number of bacteria in the lower respiratory tract after infection (Rose et al., 2008). Currently, as the result of widespread vaccination with the 7-valent conjugate vaccine, epidemiologists are now observing ‘serotype shift’, where non-conjugate vaccine serotypes are becoming predominant. For this reason, scientific efforts are focused on identifying a conserved pneumococcal protein that elicits protection against S. pneumonia infection or a panel of proteins that collectively does the same. Ideally, these pneumococcal proteins would replace diphtheria toxoid in a 3rd-generation conjugate vaccine and would protect against serotypes whose polysaccharide is not included in the vaccine. As PsrP is present in many, but not all of the invasive clones that compose these non-vaccine serotypes, and the BR domain of PsrP was determined to be conserved, its consideration for inclusion in a multiprotein conjugate vaccine is warranted. Importantly, our studies indicate that immunization of mice with PsrPSRR1-BR, and rPsrPBR.C.2, failed to protect mice against pneumococcal challenge, suggesting that the optimal PsrP antigen was the complete BR domain and no more.

In summary, PsrP is found in a wide variety of clinical isolates and serotypes and has been shown to contribute towards the ability of the pneumococcus to cause IPD. We have demonstrated that PsrP binds to K10 on the surface of lung cells, furthermore, that adhesion to K10 is mediated by AAs 273–341 of the BR domain. We have also shown that a truncated version of PsrP fails to mediate adhesion, suggesting that the extremely long SRR2 domain serves to extend the BR domain outward through the bacterium's capsule. Antibodies against rPsrPBR are protective and the BR domain is conserved, together suggesting that the recombinant protein might be used in a vaccine to protect children and the elderly from non-vaccine serotypes that cause IPD. The latter finding also suggests that the NR domain of other SRRPs may be useful as vaccine antigens to protect against Gram-positive bacteria that cause infection.

Experimental procedures

Bacterial strains

Streptococcus pneumoniae were grown in Todd-Hewitt broth or on blood agar plates at 37°C in 5% CO2. Clinical isolates were a gift from James Jorgensen at the University Hospital in San Antonio, Texas. Serotype 4, strain TIGR4 and its unencapsulated derivative T4R have been previously described (Gosink et al., 2000; Tettelin et al., 2001). T4 ΩpsrP and T4R ΩpsrP were created by allelic exchange. DNA fragments immediately flanking psrP (TIGR4 gene designation: SP1772) were amplified and cloned upstream and downstream of an antisense ermB cassette in the vector pCR2.1 (Invitrogen, Carlsbad CA). The resulting mutagenic construct (∼3 kb) was PCR-amplified and used to transform TIGR4 and T4R. Deletion of psrP was confirmed by PCR and sequencing; an illustration of the mutated locus is shown in Fig. 2A. Polar effects caused by psrP deletion were tested for by quantitative RT-PCR analysis of the upstream transposon (SP1773), glyA (SP1771) and secA2 (SP1759) transcription and none were detected. Gyrase B (gyrB) expression was included for internal normalization. All mutant S. pneumoniae were grown in media supplemented with 1 μg ml−1 erythromycin. E. coli strain BL21(DE3) and its derivatives containing plasmids were grown in Luria–Bertani broth or agar supplemented, when appropriate, with ampicillin (100 μg ml−1) or spectinomycin (50 μg ml−1). All primers used in these studies are listed in Table S3.

Tissue culture cell lines

A549 (ATCC CRL-185) and LA-4 (ATCC CCL-196) cells were maintained in F12 media supplemented to 10% fetal calf serum. Detroit 562 (ATCC CCL-138) cells were maintained in Minimum Essential Medium supplemented to 10% fetal calf serum and 0.2% lactalbumin hydrolase.

Detection of PsrP in S. pneumoniae

Whole-cell lysates were prepared using 1% sodium deoxycholate and 1% sodium dodecyl sulphate in PBS with incubation of the bacteria pellet for 30 min at 37°C. To obtain the cell wall fraction, osmotic lysis followed by centrifugation was used. Exponential phase cultures of S. pneumoniae were pelleted and the bacteria incubated with 0.5 M NaCl for 20 min with shaking at 4°C. Osmotic treatment was repeated twice more for a total of three times. Bacteria were further disrupted using a sonicator with a 5 s pulse rate for 10 min while on ice. The cell wall fraction was collected by centrifugation of the suspension at 17 000 g for 1 h at 4°C. To visualize PsrP, proteins in these samples were separated by electrophoresis using an agarose gel. Whole-cell lysate and cell wall fraction samples were suspended in PBS and mixed with an equal volume of 2× SDS sample buffer (80 mM Tris, 2% SDS, 10% glycerol, 0.1% bromophenol blue). Samples were boiled for 7 min, then loaded onto 1.5% gels made with the same electrophresis buffer used to run the gel (40 mM Tris, 1 mM EDTA, 0.1% SDS, pH 7.8 adjusted with glacial acetic acid). Samples were run at 35 V at 4°C until the samples left the wells, and then the voltage was increased to 50 V for 4–5 h. Proteins were moved onto nitrocellulose by capillary transfer using a transfer buffer without methanol (25 mM Tris, 192 mM glycine, 0.1% SDS, adjusted to pH 8.3). Blocking and development of the Western blots followed standard laboratory protocols (Ausubel et al., 2008).

Streptococcus pneumoniae expressing a truncated version of PsrP was created by transformation of TIGR4 with the plasmid pfcsRK::PsrP1-734. This plasmid was created by cloning the first 2202 nucleotides of psrP (encodes 734 AAs: SP, SRR1, BR domains of PsrP, as well as 31 repeats of the SAS[A/E/V]SASAS[I/T] motif, ends with a histidine tag) into the shuttle vector pNE1 (Bartilson et al., 2001). Expression of psrP1-734 was under control of the S. pneumoniae fcsRK fucose operon promoter, which was cloned immediately upstream (Chan et al., 2003). During growth production of PsrP1-734 was induced by supplementing media with 1% fucose, alternatively suppressed with 1% sucrose. Production of full-length native PsrP and rPsrP1-734 was detected by Western blot analysis of whole-cell lysates. Samples were collected from exponential growth phase cultures (OD620 = 0.5). PsrP was visualized using rabbit polyclonal antiserum against rPsrPSRR1-BR (1:25 000) and goat anti-rabbit IgG conjugated to horseradish peroxidase (HRP, 1:7500; Bio-Rad) (Rose et al., 2008). pNE1 contains a spectinomycin resistance cassette, thus selection and maintenance of pfcsRK::PsrP1-734 in S. pneumoniae was done by supplementing media with spectinomycin to 250 μg ml−1.

Finally, detection of native PsrP on the surface of S. pneumoniae was done using immunofluorescent assay as described by Pracht et al. (2005) with slight modifications. Bacteria were first FITC-labelled using published protocols (Gosink et al., 2000). The bacteria were fixed with 40% methanol for 20 min and washed with PBS. Antiserum against rPsrPSRR1-BR (1:2000) was used as the primary antibody, whereas PE-conjugated goat anti-rabbit IgG (1:200) was used as the secondary antibody. After washing, the cell pellet was suspended in PBS and bacteria visualized using an AX-70 fluorescent microscope (Olympus) and the images were captured at 0.1112 and 0.8886 ms exposure time for Cy2 and Cy3 filters respectively. The magnification used for capture of digital images was 1000×. Captured images were processed using Simple PCI software (Hamamatsu).

Complementation and competitive inhibition studies with recombinant proteins

For studies using PsrP complemented mutants, DNA encoding the PsrP constructs shown in Fig. 2A were cloned into pNE1 using a two-step process. First, DNA corresponding from the psrP promoter through to the end of the BR domain was cloned between SacI and SmaI restriction sites. Second, DNA corresponding from the SRR2 domain to the cell wall anchor domain was cloned into the same SmaI site and a downstream HindIII restriction site. For production of rPsrP constructs in E. coli, DNA corresponding to each of the rPsrP fragments used in the competitive inhibition assays was amplified using pRSETSRR1-BR as the DNA template (Rose et al., 2008). PCR products were cloned into pRSET-A (Invitrogen), which adds an N-terminal hexahistidine tag to the expressed proteins. His-tagged proteins were purified from isopropyl-β-D-thiogalactoside-induced BL21(DE3) cells using a QIAexpress Ni-NTA Fast Start Kit (Qiagen). The purity of the eluted proteins was verified by SDS-PAGE, and the recombinant proteins dialysed and stored in 25 mM TRIS-HCl buffer, pH 7.4. All plasmids were confirmed by sequencing and production of the recombinant proteins in S. pneumoniae was confirmed by Western blot analysis.

Adhesion assays

Adhesion assays were performed as previously described (Mann et al., 2006). For competitive inhibition assays, A549 cells were treated for 1 h with media containing 1.0 μM of BSA, the indicated rPsrP fragments or lectins, prior to the addition of bacteria. For neutralization assays, pneumococci were incubated for 1 h in media containing 1.0 μM of the designated proteins or carbohydrates prior to their addition to cells. For adhesion assays involving the deglycosylation of surface proteins, A549 cells were incubated with F12 media containing 0.0085 units ml−1 purified Clostridium perfringens neuraminidase, 0.0005 units ml−1 of S. pneumoniaeβ-galactosidase or 0.003 units ml−1 of S. pneumoniaeβ-N-acetylglucosaminidase for 4 h at 37°C in 5% CO2 (King et al., 2006). For latex bead adhesion assays, purified fragments of rPsrP or BSA were adsorbed onto 1–2 μm Fluoresbrite YG Microspheres (Polysciences). The amount of protein on each bead was calculated by subtracting the amount of protein left in the supernatant following adsorption from the amount used initially (300 μg ml−1) and dividing by the number of beads adsorbed. Attached beads were visualized using a DFC300 FX fluorescent inverted microscope (Leica). Experiments were performed in triplicate, with three or more replicate wells tested for each experimental condition.

Detection of Cy3-labelled rPsrP bound to the surface of cells

rPsrP fragments were labelled using a FluorLink-Ab Cy3 labelling kit (Amersham). Fixed monolayers of cells or lung tissues were incubated with the Cy3-labelled proteins at a concentration of 1.4 mg ml−1, along with DAPI to stain the nucleus, for 1 h at 37°C. After washing, the cells were mounted on the glass slides using Fluorsave solution (Calbiochem). Cells were visualized using an AX-70 fluorescent microscope and the images were captured at 0.1112–0.2224 ms exposure time for Cy2 and Cy3 filters and 0.001–0.005 ms for UV filter. The magnification used for capture of digital images was 1000×. Captured images were processed using Simple PCI software.

Affinity purification with a rPsrP column

rPsrPSRR1-BR (5 mg) was conjugated to 1 ml of Affi-Gel 15 (Bio-Rad) as instructed by the manufacturer. A549 cell lysate was prepared using 1% Triton X-100 lysis buffer containing EDTA-free protease inhibitors (Sigma). The cell lysate and rPsrPSRR1-BR gel slurry were mixed in a column for 4 h at 4°C, then washed repeatedly with PBS/Ca2+/Mg2+/Tween-20. Bound proteins were eluted in 300 μl fractions using 100 mM glycine-HCl (pH 3.5). Proteins in the eluted fractions were visualized by SDS-PAGE and silver staining. Bands corresponding to eluted proteins were excised and identified by maldi-tof analysis. Protein band identification was done by the Protein Chemistry Core Laboratory, at Baylor College of Medicine, Houston, TX. Protein assignment was based on both MS and MS/MS spectra and the NCBInr database.

Immunoprecipitation of K10

Protein G Sepharose beads (50 μl; Amersham) were incubated overnight at 4°C with 10 μg of rabbit monoclonal anti-human K10 antibody (Epitomics) or mouse polyclonal rPsrPBR.C.2 antibody in 500 μl of F12 media supplemented with 10% FBS. Beads were incubated with: 400 μl of A549 cell lysates alone, lysate + 200 μg rPsrPBR, 200 μg rPsrPBR alone, for 4 h at 4°C with constant agitation. The beads were washed with RIPA buffer (Ausubel et al., 2008), then boiled in SDS sample buffer for 10 min. Western blots were carried out using mouse monoclonal Penta-His Antibody (1:2000; Qiagen), which recognizes the His tag on the recombinant protein as the primary antibody, and goat anti-mouse IgG antibody conjugated to HRP (1:7500; Bio-Rad) as the secondary antibody.

Western blot and ELISA for K10

Cell lysates were prepared using RIPA buffer. Proteins in the lysates (10 μg) were separated by SDS-PAGE and electophoretically transferred to nitrocellulose membranes. Membranes were blocked with PBS containing 5% non-fat milk and 0.1% Tween-20 for 1 h. Membranes were incubated overnight at 4°C with monoclonal K10 antibody (clone ID# EP1607IHCY; Cat# 2210-1; Epitomics, CA) at a dilution of 1:2000. HRP-conjugated goat anti-rabbit Immunoglobulin G was used as the secondary antibody. The immunoblots were stripped and probed with rabbit anti-actin antibodies (Sigma) at a 1:5000 dilution. For ELISA, rPsrP proteins were coated onto Nunc Maxisorp plates and left overnight at 4°C. The plates were blocked with 1% BSA for 1 h at 25°C. Cell lysates were added to the wells and bound K10 detected with monoclonal antibodies. The HRP-conjugated complexes were developed using ABTS substrate solution (Sigma) and the colour intensity measured at 405 nm using an ELISA plate reader.

Detection of K10 on cells and tissue sections

Cells grown on glass coverslips were fixed and blocked for 1 h with F12 media containing 10% FBS. K10 antibody (1:100) and Cy3-labelled rPsrPBR.C (1.4 mg ml−1) were added to the cells and incubated overnight at 4°C. The next day, cells were washed and incubated with fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (1:1000) along with DAPI for nuclear staining for 1 h. Cells were washed and mounted on glass slides using Fluorsave solution. Fluorescent images were captured using the AX-70 microscope as described. Mouse lung sections were used to create 5-μm-thick tissue sections on glass slides. For immunofluorescence experiments, tissue sections were treated in the same manner as that described for cells on coverslips, with the exception that DAPI was not used. Exposure times and magnification were the same as described above.

To detect K10 on the surface of cells, Detroit, A549 and LA-4 cells grown in monolayers were harvested by scraping and fixed with 4% paraformaldehyde for 20 min. Cells were incubated with either rabbit monoclonal anti-human K10 antibody (1:100) or rabbit mAb IgG isotype control (1:100; Clone ID# DA1E; cat# 3900S; Cell Signalling Technology, MA). Incubation of cells with antibodies was carried out for 1 h at 4°C. Cells were washed and suspended in MACS buffer (Miltenyi Biotec, Bergisch Gladbach, Germany) containing FITC-labelled goat anti-rabbit IgG (1:250; Invitrogen) and incubated for an additional 30 min at 4°C. Cells were washed to remove any unbound secondary antibody and suspended in MACS buffer. Fluorescently labelled cells were detected using an A2-Laser BD FACSCaliber analyser (Becton Dickinson, NJ; Institutional Flow Cytometry Core Facility at the Health Science Center) and the data were processed and normalized against the area of the isotype control using CellQuest software.

Modification of K10 expression

The SiRNAs were constructed using the SiRNA Construction Kit (Ambion). The target sequence was selected from exon 4 of the K10 mRNA sequence (NCBI Accession No.: NM_000421.2). Both K10 and the corresponding scrambled sense and antisense primers (Table S3) were designed following the protocols outlined by the kit. SiRNAs were used to transfect 24-well plate monolayers of A549 cells at a concentration of 50 nm using the vehicle Lipofectamine 2000 (Invitrogen). For overexpression of K10, pCDNAK10 and the control plasmid pCDNA were used to transfect A549 cells (Paramio et al., 1999). Transient transfection was done with 50 ng of pCDNA and 5–50 ng of pCDNAK10 using lipofectamine. For both siRNA and overexpression experiments, Western blot analysis of K10 expression and bacterial adhesion assays were performed 2 days post transfection.

Visualization of rK10 bound to TIGR4

Total RNA was isolated from A549 cells, DNAsed and single-strand cDNA synthesis carried out using the Thermoscript RT-PCR system (Invitrogen). Full-length K10 was amplified using specific primers for the three fragments (Table S3). Amplified DNA fragments were cloned into the bacterial expression vector pRSET-A, between the restriction sites EcoR1 and HindIII, and the plasmids used to transform E. coli. Protein expression was induced during exponential growth with 200 μm IPTG for 3 h at 37°C. rK10 was isolated and purified using QIAexpress Ni-NTA Fast Start Kit. TIGR4 and T4 ΩpsrP were pelleted and suspended in 1 ml of carbonate buffer (pH 9.0) containing FITC (1 mg ml−1). Incubation was carried in the dark at room temperature with constant end-to-end tumbling. FITC-labelled bacteria were washed with PBS (pH 7.4) and centrifuged, repeatedly, until the supernatant became clear. Labelled bacteria were suspended in serum-free F-12 media containing human rK10 at a 1 μm concentration and incubated for 1 h at 37°C in 5% CO2. Subsequently, pneumococci were washed then fixed with 4% paraformaldehyde for 20 min, washed and suspended F-12 medium. Bound rK10 was detected using monoclonal antibodies to K10 (1:100) and goat anti-rabbit IgG conjugated with Alexafluor 568. Labelled bacteria were visualized using AX-70 Fluorescence microscope using the previously described exposure times. Images were processed using SimplePCI software.

Genomic comparisons and clonal complex determination

Homologues of TIGR4 psrP-secY2A2 genes were identified by blastn analyses (Altschul et al., 1997). Nucleotide sequence data were obtained through the National Center for Biotechnology Information Genome Project (http://www.ncbi.nlm.nih.gov) and the S. pneumoniaeblast server (http://www.sanger.ac.uk/cgi-bin/blast/submitblast/s_pneumoniae). Clonal identity was determined by multi-locus sequence typing using software made available through the S. pneumoniae multi-locus sequence typing website (http://spneumoniae.mlst.net).

Immunization and pneumococcal challenge

Mice were subcutaneously vaccinated with 20 μg of protein emulsified with Fruend's complete adjuvant. On days 21 and 42, animals were boosted with the same protein emulsified with Fruend's incomplete adjuvant. Two weeks later, mice were challenged intranasally with 107 cfu of TIGR4 in 25 μl PBS. On day 3, blood was collected from the tail vein, serially diluted and spread on blood agar plates. The numbers of colonies on the plates were subsequently used to extrapolate the bacterial titer in the blood. Survival of all experimental animals was tracked for up to 7 days after infection.

Statistical analysis

Statistical analyses were made using a one-way anova, Fisher's exact test, Student's t-test or a Kaplan–Meier Log Rank test. The test used is indicated in the corresponding text or figure legend.

Acknowledgements

This work was supported by National Institutes of Health Grant AI078972. The authors would also like to thank Jose L. Jocarno for his generous gift of pCDNAK10, Peter Dube for Outer Membrane Protein Fraction 1 antigen from Y. pestis, James Jorgensen for the clinical isolates, and last but not least, Caroline Obert for assistance with multilocus sequence typing. The authors have no conflicting financial interests.

Ancillary