The Group B Streptococcal C5a Peptidase Is Both a Specific Protease and an Invasin

Abstract

The group B streptococcus (GBS) is a major cause of pneumonia, sepsis, and meningitis in neonates and a serious cause of mortality or morbidity in immunocompromised adults. Although these streptococci adhere efficiently and invade a variety of tissue-specific epithelial and endothelial cells, adhesins and invasins are still unknown. All serotypes of GBS studied to date express C5a peptidase (SCPB) on their surface. This investigation addresses the possibility that this relatively large surface protein has additional activities. Rabbit anti-SCPB serum inhibited invasion of lung epithelial A549 cells by the serotype Ia strain O90R, suggesting that SCPB is an invasin. This was confirmed by inserting an in-frame 25-amino-acid deletion into the scpB gene. Invasion of HEp2 and A549 human cell lines was significantly reduced by the mutation. Enzyme-linked immunosorbent assays were used to demonstrate that purified SCPB protein binds directly to HEp2 and A549 cells and also binds the extracellular matrix protein fibronectin. Binding was dose dependent and saturable. These results suggested that SCPB is one of several potential invasins essential for GBS colonization of damaged epithelium.

Several species of beta-hemolytic streptococci are known to produce a surface-bound protease that specifically inactivates the human phagocyte chemotaxin C5a. All group A streptococcal (GAS) serotypes and group B (GBS) (16), C, and G streptococci of human origin produce the C5a peptidase (SCP). The protein is highly conserved across species boundaries. The GAS enzyme (SCPA) is 95 to 98% identical to that expressed by GBS (SCPB) (6, 7). Several lines of evidence indicate that the protease is important in the pathogenesis of GAS and GBS. It was demonstrated to retard the influx of phagocytes into subdermal sites of infection and alter the trafficking of streptococci between lymph node and spleen in mice (17). Moreover, clearance of streptococci from these sites of infection in mice is delayed when they express SCPA on their surface. Immunization of mice with SCPA enhanced elimination of GAS following intranasal infection. The role of SCPB in virulence of GBS has not been extensively investigated; however, antibody directed against recombinant SCPB was found to be opsonic and to induce macrophage killing of GBS (5). Although vaccine development has focused primarily on the capsular polysaccharide (10), these studies suggested that SCPB protein could possibly induce serotype-independent immunity to infection (5). Bohnsack et al. showed that SCPB-deficient GBS were cleared more rapidly from mice that were supplemented with human C5a following lung infection (2).

Both GAS and GBS efficiently invade human epithelial cells (11, 13). Rubens et al. suggested that the potential of GBS to enter and survive within respiratory epithelial cells may represent a mechanism by which they gain access to blood and other organs (21). Although the invasins and adhesions of GBS have not been identified, they are known to bind fibronectin (Fn) (27), laminin (23), and cytokeratin 8 (26). The relationship of these binding activities to adherence or invasion of epithelial cells is also unknown. A lipoprotein, designated LmB, which binds to both Fn and laminin was cloned from GBS, but it was not shown to be involved in adherence or invasion of epithelial cells (23).

SCP is a large protease that is comprised of over 1,100 amino acids after removal of the signal sequence. Sequence comparison predicted that the protein is a serine protease which is related to the subtilisin family of enzymes (22). Domain alignment revealed homology between the N-terminal catalytic triad of SCP and subtilisin. Replacement of Ser⁵¹², His¹⁹³, or Asp¹³⁰ with alanine confirmed that these residues are essential for proteolytic activity (24). The C-terminal end of SCP contains the peptidoglycan anchor sequence common to other gram-positive bacterial surface proteins. The large size of SCP and the finding that anti-SCPB is opsonic suggested that SCP may be a multifunctional surface protein. Experiments presented here demonstrate that SCPB functions as an invasion for epithelial cells and that both SCPA and SCPB are Fn binding proteins.

MATERIALS AND METHODS

Bacterial strains.

GBS O90R and its deletion mutant were grown on blood agar plates or grown in Todd-Hewitt broth to mid-log phase by incubation at 37°C to an optical density at 560 nm of 0.5 to 0.6. Strains 78-471 and COHI are serotypes II and III, respectively. Escherichia coli strain BL21 carrying various pGEX::scp plasmids was grown in 2× YT supplemented with 2% glucose and 100 μg of ampicillin per ml (6).

Construction of a defined scpB deletion mutation in GBS.

An MfeI/SacI fragment of scpA containing a 75-bp in-frame deletion of 25 amino acids was cut from the thermosensitive plasmid pG::ΔscpA1.1 (16) and ligated between the MfeI and SacI sites of gel-purified plasmid pGEX::scpB (6). Plasmid pGEX::ΔscpB was confirmed to have the correct deletion by the location of an EcoRI site created by the deletion (17). Plasmid pGEX::ΔscpB was also transformed into E. coli BL21 for protein expression. The deletion removed Ser⁵¹², which is essential for peptidase activity (24).

The 75-bp deletion was introduced into the scpB chromosomal gene of strain O90R by gene replacement. Strain O90R cells were made competent for electroporation according the protocol described by Framson et al. (12) and transformed with plasmid pG::ΔscpA1.1 (17), the source of the deletion. Transformants were selected on Todd-Hewitt agar containing 1 μg of erythromycin per ml at 30°C. Integration of the plasmid into the chromosome was forced by incubation at 39°C. Excision of plasmid sequence by homologous recombination resulted in replacement of the wild-type sequence with the deleted form. The correct in-frame deletion in recombinants was confirmed by PCR and by sequencing the 1.1-kb replaced sequence in the chromosomal DNA. DNA sequencing was performed by the Microchemical Facility at the University of Minnesota.

Invasion of A549 and HEp2 cells by GBS.

HEp2 cells or A549 cells were cultured in minimal essential medium α or RPMI medium, which were supplemented with 10% fetal calf serum (FCS) (Life Technologies) and contained 5 μg of penicillin per ml and 100 μg of streptomycin per ml. Assays of bacterial invasion and adherence were performed as previously described (13). Assays were performed in unsupplemented medium or medium supplemented with 10% FCS or with 10 μg of Fn per ml. Monolayers (about 2 × 10⁵ cells/well) were infected with 1 × 10⁵ to 5 × 10⁵ CFU of streptococci. The number of CFU associated with the monolayer after incubation for 2 h at 37°C in a CO₂ incubator was determined by viable counts to evaluate adherence. After 2 h, infected monolayers were washed three times with phosphate-buffered saline (PBS)-Ca²⁺ before incubation in minimal essential medium α or RPMI-FCS containing 100 μg of gentamicin per ml and 5 μg of penicillin per ml for an additional 2 h to kill extracellular bacteria. The number of CFU inside epithelial cells was determined by viable counts.

Purification of SCP and measurement of SCP activity.

SCPA, SCPB, and SCPB(del) proteins were expressed from pGEX plasmids in E. coli strain BL21 and purified as previously described (24). SCP activity was measured using a fluorescent fusion protein, glutathione transferase-human C5a-green fluorescent protein, bound to Sepharose beads (24). Release of green fluorescent protein by peptidase activity was measured with a Bio-Tek FL600 fluorescence reader. Purity of proteins was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Protein concentrations were determined using a bicinchoninic acid protein assay kit (Pierce).

Measurement of SCP binding to HEp2 and A549 cells.

HEp2 or A549 cells were seeded into 96-well cell culture plates and cultured overnight with FCS in the culture medium at 37°C. Cells were fixed with 2.5% paraformaldehyde and blocked with PBS-Ca²⁺ containing 0.1% ovalbumin for 1 h at 37°C. Preliminary experiments were done without fixing the cells and produced similar results. Various amounts of SCPA, SCPB, and SCPB(del) proteins were added into the wells, which were then incubated for 2 h at 37°C. After thorough washing, rabbit anti-SCPB (1:1,000) was added to the wells, and then the plates were incubated for 1 h at 37°C. Horseradish peroxidase (HRP)-goat anti-rabbit immunoglobulin G (IgG) (1:10,000) was used to detect rabbit anti-SCPB IgG bound to the cells. Enzyme-linked immunosorbent assays (ELISAs) were developed using an ImmunoPure TMB Substrate Kit (Pierce). Affinity-purified SCP (0 to 20 ng) was applied to wells to generate a standard curve that was repeated with every experiment. The amount of bound SCP was calculated from the standard curve. Assays were done in triplicate, and all experiments were repeated at least three times. Dissociation constants (K_d) were calculated by Scatchard plot analysis.

SCP binding to Fn.

Fn (250 ng/well) was applied to 96-well ELISA plates. After blocking with PBS containing 0.05% Tween 20 and 0.1% ovalbumin, various amounts of SCPA, SCPB, or SCPB(del) were added to each well and incubated for 2 h at 37°C. After unbound protein was washed off, rabbit anti-SCPB was added to the wells and the plates were incubated for 1 h at 37°C. HRP-goat anti-rabbit IgG was used to detect rabbit IgG. ELISAs were developed using the ImmunoPure TMB Substrate Kit (Pierce).

GBS binding to Fn.

GBS (10⁶ CFU) were applied to wells in ELISA plates overnight at 4°C. The wells were blocked with PBS containing 0.5% ovalbumin and 0.05% Tween 20. Various amounts of Fn were added and incubated for 2 h at 37°C. Sheep or goat anti-human Fn and subsequent incubation with HRP-anti-sheep IgG was used to detect bound Fn. ELISAs were developed using the ImmunoPure TMB Substrate Kit (Pierce).

RESULTS

Antibody against SCPB inhibits invasion of epithelial cells by GBS.

Our laboratory showed that anti-SCPB opsonized GBS in phagocytosis assays (5). For this reason rabbit anti-SCPB was tested for its capacity to inhibit adherence and invasion of epithelial cells by GBS. Strains COH1, O90R, and 78-471 were grown to mid-log phase and incubated with 10% normal rabbit serum (as a control) or rabbit anti-SCPB at room temperature for 1 h. Bacteria (1 × 10⁵ to 2 × 10⁵ CFU) were then incubated with A549 lung epithelial cells for 2 h at 37°C. Adherent GBS were assayed by viable counts associated with the monolayer. Anti-SCPB did not affect adherence (Table 1) but did reduce internalization of all three strains by A549 cells from 58 to 75% (Table 2). Results with bacteria exposed to normal rabbit serum were taken as 100% invasion. Since adherence was unaffected by anti-SCPA, it was unlikely that inhibition was the result of steric hindrance. These data suggested that SCPB acts in a specific way to enhance invasion and that antibody specifically binds to SCPB to block that specific interaction.

TABLE 1.

Effect of rabbit anti-SCPB on adherence of GBS to A549 cells

GBS strain	% Adherencea (mean ± SEM) with:
	Normal rabbit serum	Rabbit anti-SCPB
O90R (serotype Ia)	35.8 ± 2.5	32.9 ± 2.9
COH1 (serotype III)	45.2 ± 1.9	44.4 ± 3.5
78-471 (serotype II)	38.9 ± 3.1	42.1 ± 2.1

^aNumber of adherent bacteria/total number of bacteria × 100.

TABLE 2.

Effect of rabbit anti-SCPB on invasion of GBS into A549 cells

GBS strain	% Invasiona (mean ± SEM) with:		% Inhibition
	Normal rabbit serum	Rabbit anti-SCPB
O90R (serotype Ia)	3.60 ± 0.15	1.10 ± 0.09	69
COH1 (serotype III)	0.36 ± 0.01	0.09 ± 0.01	75
78-471 (serotype II)	0.12 ± 0.01	0.05 ± 0.003	58

^aNumber of internalized bacteria/number of adherent bacteria × 100%. Data are from a single experiment but represent data from three experiments.

Genetic inactivation of SCPB reduces invasion of A549 and HEp2 cells by GBS.

A deletion mutant that eliminated peptidase activity was constructed in order to confirm antibody inhibition experiments. The SCP activity of strain O90R was eliminated by introducing a defined nonpolar 75-bp deletion into the chromosomal scpB gene. The SCPB⁻ O90R(del) strain was confirmed to contain the correct in-frame deletion by sequencing the replaced chromosomal DNA segment. Sequencing also confirmed that the recombination event had not introduced transition mutations which distinguish the original scpB and scpA genes (17). The deletion mutant and wild-type cultures have comparable growth rates in Todd-Hewitt broth. Bacterial surface-bound SCP activity was tested using glutathione transferase-human C5a-green fluorescent protein as a substrate (24). Strain O90R(del) showed a 90% reduction in enzyme activity. This is expected because the catalytic Ser⁵¹² was removed by the deletion.

The capacity of strain O90R(del) to invade epithelial cells was tested by a standard invasion assay with or without FCS in the wells, using both A549 cells and HEp2 cells as targets. The parent O90R and the SCPB⁻ mutant adhered to both A549 and HEp2 cells. Generally, the cultures adhered somewhat better without FCS. Although A549 is a more relevant cell line, HEp2 cells were included in this study because GBS invade them more efficiently than A549 cells (10 versus 1%). Surprisingly, inactivation of SCPB increased the capacity of O90R to adhere to both epithelial cell lines (Fig. 1A). In contrast, invasion of these cells by O90R(del) was reduced by more than 50%, despite the fact that these bacteria adhered better than the parent streptococcus (Fig. 1B). Potential explanations for this contradiction are discussed below. In general these results support the antibody data and indicate that the SCPB protein is an epithelial cell invasin.

FIG. 1.

Comparison of adherence (A) and invasion (B) of SCPB⁺ O90R (wild type [WT]) and an SCPB⁻ deletion mutant. Assays were carried out in the absence of serum. Percent adherence is normalized for growth in wells that contain epithelial cells and equals CFU associated with the monolayer divided by total CFU in the well at 2 h postinfection. Invasion is intracellular CFU divided by initial inoculum of CFU. The data are means ± standard errors of the means from three independent experiments. The percentage is indicated above the bars.

SCP binds directly to A549 and HEp2 cells.

The experiments described above suggested that surface-bound SCPB interacts directly with either the epithelial cell surface or with another ligand that is anchored to the cell surface. Tamura and Rubens reported that GBS efficiently bind to Fn anchored to microtiter plate wells (27). To address this issue, affinity-purified recombinant SCPA and SCPB proteins were tested for their potential to bind to epithelial cells and to solid-phase Fn. Figure 2A and B show the binding curves for both the SCPA and SCPB proteins and A549 and HEp2 cells, respectively. SCPA and SCPB bound to cells with similar affinities (K_d = 54.9 nM for HEp2 cells and 49.5 nM for A549 cells) and in a dose-dependent fashion. Binding was saturable when high concentrations of SCP were used. More SCP was bound to A549 cells than to HEp2 cells. This could be explained by the fact that A549 cells were found to have 50% more Fn associated with their surface (data not shown). As expected, the SCPB and SCPA proteins, which are 98% identical in amino acid sequence (6), bound to epithelial cells with equal affinities. Recombinant SCPB(del) with the in-frame deletion was less able to bind to cells (Fig. 2A and B).

FIG. 2.

SCP binding to epithelial cells and Fn. Purified SCP protein was incubated with HEp2 cells (A), A549 cells (B), and Fn (C) for 2 h at 37°C. Bound SCP was detected by rabbit anti-SCPB followed by HRP-goat anti-rabbit IgG. Standard curves for each protein were developed to compensate for the fact that they have somewhat different affinities for the primary antibody. SCPB(del) protein has residues Ala⁴⁹² to Val⁵¹⁷ deleted. Standard errors were less than 8% for all experiments.

SCP is an Fn binding protein.

Tamura and Rubens (27) showed that GBS bind to solid-phase Fn. Therefore, the possibility that SCPB is an Fn binding protein was further tested. An Fn binding assay was developed, in which 250 ng of Fn was applied to microtiter plate wells. Bound SCP was detected using the same method employed to measure binding to epithelial cells. Both SCPA and SCPB bound to immobilized Fn in a dose-dependent fashion with dissociation constants of 5.6 and 6.3 nM, respectively (Fig. 2C). High concentrations of both streptococcal proteins were required to achieve saturation. These apparent association constants are comparable to those reported for other bacterial Fn binding proteins (15, 19). The inactivated form of the SCPB(del) protein was less able to bind to immobilized Fn but retained significant binding activity. The deletion did not disrupt the Fn binding region (residues 116 to 227, which was defined by Beckmann et al. [submitted for publication]), but it may have altered the binding site conformation in a manner that reduced affinity for Fn. When the assay was reversed (i.e., wells were coated with SCP and Fn was added to the aqueous phase at high concentrations), the peptidase was able to bind soluble Fn (data not shown).

Strain O90R(del) was constructed to test SCPB's impact on Fn binding to streptococci. Wild-type and SCPB⁻ streptococci were deposited onto the surface of microtiter wells to determine their capacity to bind aqueous-phase Fn (Fig. 3). The SCPB⁻ strain bound 50% less Fn than the parent SCPB⁺ culture, again suggesting that the mutation had not completely eliminated Fn binding to SCPB protein. Residual binding to strain O90R(del) could also be due to other unidentified Fn binding proteins produced by these bacteria.

FIG. 3.

Binding of Fn to SCPA⁺ O90R and SCPA⁻ O90R(del) streptococci. Increasing amounts of human Fn were added to microtiter plates coated with 10⁶ CFU of fixed wild-type O90R (WT) or SCPB-O90R(del). Bound Fn was detected by sheep anti-human Fn followed by HRP-donkey anti-sheep IgG. Data are means ± standard errors of the means from a single experiment performed in triplicate but represent three independent experiments. Control wells were without bacteria. OD, optical density.

Does soluble Fn inhibit or enhance adherence and invasion by O90R streptococci?

The adherence and invasion experiments described above were performed in the absence of serum or added Fn; i.e., cell cultures were washed free of medium. Endogenous Fn was detected on the surfaces of both HEp2 and A549 cells by using anti-human Fn in ELISA. A549 cells were found to have 50% more Fn on their surface than HEp2 cells (data not shown). Experiments were performed to determine whether soluble Fn influenced the interaction of SCPB⁺ GBS with HEp2 or A549 cells. Exogenous Fn caused a 32% increase in invasion of A549 cells by wild-type O90R streptococci but did not affect the SCPB⁻ strain (Fig. 4A). This increase in invasion was statistically significant (P < 0.05 for three independent experiments). Exogenous Fn did not affect adherence (data not shown) to or invasion of HEp2 cells by either SCPB⁺ or SCPB⁻ streptococci (Fig. 4B). These results suggested that invasion of A549 cells by strain O90R is in part dependent on the interaction of SCPB with Fn.

FIG. 4.

Effect of soluble Fn on invasion of epithelial cells. (A) Invasion of A549 cells; (B) invasion of HEp2 cells. Assays were carried out with or without 10 μg of Fn per ml in the well. Bacteria grew at comparable rates in wells with or without Fn. Percent invasion is intracellular CFU divided by the initial inoculum. Data are from three independent experiments and are means ± standard errors of the means.

DISCUSSION

Passage of GBS from the vaginal or intestinal mucosa to the amniotic fluid within the placenta or to the blood and meninges of a neonate is poorly defined. One thing that is certain, however, is that the organism must traverse several mucosal membrane barriers along the way. Although GBS do not replicate within respiratory epithelial cells, their ability to enter and survive in these cells and in macrophages (28) may be a mechanism by which they can cross mucosal membranes (8, 14). Gibson et al. (14) reported that GBS are able to invade lung endothelium in vitro and suggested that the ability of GBS to breach the endothelial barrier could provide a means for them to gain access to vascular spaces or to enter interstitial spaces from the vascular compartment. Strains from invasive infections of infants are internalized by epithelial cells more effectively than to those from age-matched healthy controls (29), suggesting that invasion of epithelial cells significantly influences the pathogenesis of GBS. Winram et al. (31) demonstrated that GBS can invade the chorionic epithelium and suggested that this could result in inflammatory damage to the amniotic membrane, which would permit streptococci to invade the amniotic fluid with subsequent infection of the fetus.

Efforts to define GBS surface macromolecules that mediate adherence have been limited and contradictory. Capsular polysaccharides of GBS are not required for adherence to or invasion of epithelial cells and may, in fact, attenuate these cellular interactions to some degree (25). Investigations of lipoteichoic acid (LTA) and adherence are contradictory. Several reports implicated LTA in adherence by showing that preincubation of epithelial cells with soluble LTA at high concentrations inhibited attachment of GBS (20). Others, however, who employed similar concentrations of LTA, did not observe an inhibitory effect (1). Surface proteins were predicted to be necessary for adherence because protease treatment of the bacteria reduced adherence and invasion (1). The discovery that GBS express a C5a-specific protease on their surface (16) and that anti-SCPB is opsonic (5) prompted us to investigate the possibility that this surface protease also functions as an adhesin and/or invasin.

GBS efficiently adhered (30 to 40% of the initial inoculum) to both A549 and HEp2 cell lines. However, they invaded HEp2 cells 10 times more frequently than A549 cells. Invasion of A549 cells by three serotypes (Ia, II, and III) of GBS was blocked by rabbit anti-SCPB. The failure of antiserum to inhibit adherence, however, suggested that antibody specifically blocks an invasin-specific interaction with cells and does not merely sterically mask adhesions. Moreover, this indicates that SCPB can function as an invasin, independent of adherence. Introduction of an internal nonpolar deletion mutation into scpB confirmed the antibody inhibition results. The mutation reduced invasion of epithelial cells between by 80 and 41%, depending on the cell line. The residual invasive capacity of mutant streptococci may mean that strain O90R produces other, unidentified invasins.

The reduction in invasion was not accompanied by a significant reduction in adherence to epithelial cells. In fact, the SCPB⁻ mutant O90R(del) adhered significantly better to A549 cells, and slightly better to HEp2 cells, than the parent culture. This increased adherence is difficult to explain, since purified protein that contains the same deletion mutation bound much more poorly than wild-type SCPB protein to both epithelial cell lines and Fn. One explanation is that inactivation of protease activity protects another adhesin, which is normally cleaved in wild-type streptococci, or that conformational changes induced by the deletion exposed another adhesin. Another explanation is that SCPB proteolytically damages a cellular receptor or coreceptor which is prevalent on A549 cells and interacts only with another streptococcal adhesin. Failure of the mutant to alter this putative receptor would be expected to increase adherence but reduce invasion frequencies. This would explain the observation that mutant recombinant protein did not show increased binding to cells or Fn. The more modest increase in adherence to HEp2 cells suggests that they may have fewer of these unidentified receptors than A549 cells. It is also possible that bacterial surface-bound SCPB protein interacts differently with cellular receptors than the recombinant, purified protein.

Experiments confirmed that SCPB is an invasin and as such triggers ingestion of streptococci at some stage subsequent to their initial attachment to target cells. The protein may also promote adherence, since it binds to epithelial cells, but this activity is probably overshadowed by the activity of another, unidentified adhesin under these experimental conditions. A relationship of Fn binding to SCPB-promoted invasion was not established by our experiments. Interaction of the protease with either matrix Fn or an unidentified receptor could promote cytoskeletal changes that lead to ingestion of streptococci. The failure of soluble Fn or anti-α₅β₁ to inhibit GBS ingestion suggests that the latter is most likely.

Fn has been shown to bridge GAS (9), staphylococci (4), and gonococci (30) to integrins on the surface of epithelial cells in a manner that triggers internalization of the bacteria. Extracellular matrix proteins that serve as substrates for bacterial adherence include Fn, collagens, laminin, thrombospondin, and integrins (18, 19). These interactions may be particularly important for bacterial colonization of damaged tissues. GBS were previously shown to adhere to immobilized Fn (27), laminin (23), and cytokeratin 8 (26). However, the importance of these interactions for adherence to epithelial cells was not investigated. Moreover, streptococcal receptors for Fn and cytokleratin 8 were not identified.

Beckmann et al. identified SCPB while scanning a phage display library for proteins that bind Fn (submitted for publication). They also found that inactivation of SCPB by mutation reduced Fn binding by only 50%, consistent with the suggestion that GBS express multiple Fn binding proteins. Beckmann et al. discovered a fragment of SCPB, residues 116 to 227, that was sufficient for Fn binding. This region of SCPB also contains amino acids Asp¹³⁰ and His¹⁹³, which form part of the charge transfer system that is required for protease activity. Location of the binding site to this region of the protease conflicts with our finding that deletion of 25 residues around Ser⁵¹² also significantly reduced epithelial cell and Fn binding, by 50%. This difference can be explained, however, if the physical Fn binding site is conformation dependent and overlaps the catalytic domain of the protease. The deletion could alter the conformation of this region. An alternative explanation is that the Fn binding site is masked by a propeptide and that deletion of Ser⁵¹², which is required for autocatalytic activation of C5ase activity (unpublished results), leaves the binding site inaccessible to Fn or receptors on cells.

Endogenous Fn could be detected on surfaces of monolayers of both A549 and HEp2 cells. A549 cells produce twofold more Fn than HEp2 cells. As note above, O90R adheres efficiently to these cells in the absence of soluble Fn. Fn significantly stimulated invasion of A549 cells, by 10 to 20%, but failed to increase invasion by O90R(del). Fn also failed to increase adherence to and invasion of HEp2 cells by the wild type parent strain. It is possible that GBS associate with epithelial cells by SCPB binding to the solid-phase isoform of Fn and that this interaction triggers cytoskeletal changes which result in ingestion of the bacteria. This suggestion is based on the fact that inactivation of SCPB reduced binding to both Fn and epithelial cells. It is also possible that HEp2 cells produce an isoform of Fn that has a higher affinity for SCPB than the purified serum Fn used in these experiments. This would explain the failure of exogenous Fn to enhance or inhibit interactions of streptococci with epithelial cells and could also account for the observation that HEp2 cells more efficiently ingest strain O90R (Fig. 1B). Our experiments do not eliminate the possibility, however, that SCPB also binds directly to epithelial cells via some unknown receptor and that invasion is independent of its potential to bind Fn.

Since GBS were first discovered to produce a C5a-specific protease (16), considerable evidence has accumulated to implicate it in the pathogenesis of GBS infections. SCPB was shown to decrease recruitment of neutrophils to the lungs of mice which were supplemented with human C5a and infected with GBS (2). Moreover, the O90R(del) strain is cleared from lungs of infected mice much faster than the SCPB⁺ parent culture (unpublished data). SCPB is highly immunogenic, and antibodies against the protein facilitate killing of GBS by primary cultures of bone marrow-derived macrophages and by neutrophils (5). The experiments described here demonstrate that SCPB is also an invasin, suggesting yet another mechanism by which the SCP can contribute to the virulence of GBS. Bohnsack et al. (3) were unable to demonstrate cleavage of rodent C5a by SCP; therefore, the potential of the enzyme to act as an invasin may be more important than protease activity in mouse models of infection. However, the high degree of specificity for human C5a and the exquisite sensitivity of this chemotaxin to the enzyme suggest that peptidase activity is important in human infections.