α-Enolase of Streptococcus pneumoniae is a plasmin(ogen)-binding protein displayed on the bacterial cell surface

Authors


Abstract

Binding of human plasminogen to Streptococcus pneumoniae and its subsequent activation promotes penetration of bacteria through reconstituted basement membranes. In this study, we have characterized a novel pneumococcal surface protein with a molecular mass of 47 kDa, designated Eno, which specifically binds human plasmin(ogen), exhibits α-enolase activity and is necessary for viability. Using enzyme assays, we have confirmed the α-enolase activity of both pneumococcal surface-displayed Eno and purified recombinant Eno protein. Immunoelectron microscopy indicated the presence of Eno in the cytoplasm as well as on the surface of encapsulated and unencapsulated pneumococci. Plasminogen-binding activity was demonstrated with whole pneumococcal cells and purified Eno protein. Binding of activated plasminogen was also shown for Eno; however, the affinity for plasmin is significantly reduced compared with plasminogen. Results from competitive inhibition assays indicate that binding is mediated through the lysine binding sites in plasmin(ogen). Carboxypeptidase B treatment and amino acid substitutions of the C-terminal lysyl residues of Eno indicated that the C-terminal lysine is pivotal for plasmin(ogen)-binding activity. Eno is ubiquitously distributed among pneumococcal serotypes, and binding experiments suggested the reassociation of secreted Eno to the bacterial cell surface. The reassociation was also confirmed by immunoelectron microscopy. The results suggest a mechanism of plasminogen activation for human pathogens that might contribute to their virulence potential in invasive infectious processes.

Introduction

Streptococcus pneumoniae colonizes the nasopharynx and is also a common aetiological agent of upper and lower respiratory tract diseases. Disseminated infections frequently result in bacterial pneumonia and otitis media, as well as in invasive diseases such as meningitis and sepsis (Tuomanen et al., 1995). A prerequisite for invasiveness is the ability of pneumococci to colonize and transmigrate through the epithelial and endothelial layers and to breach the blood–brain barrier. Gram-positive pathogens express specific cell surface components called adhesins that mediate their adherence to host tissues, thereby facilitating not only colonization but also invasion (Molinari et al., 1997). Most of these adhesins function by recognizing and binding various components of the extracellular matrix (ECM) (Patti and Hook, 1994), which consists of many diverse structures and complex macromolecules (Chhatwal and Preissner, 2000). The interactions of bacteria with the ECM therefore represent important pathogenicity mechanisms.

Among the pneumococcal factors that contribute to colonization are the phosphorylcholine on the teichoic acid, which has been implicated in direct adherence to cytokine-stimulated human lung epithelial cells and human vascular endothelial cells via the receptor for the platelet-activating factor (PAF) (Cundell et al., 1995), and the SpsA protein (Hammerschmidt et al., 1997), also designated CbpA (Rosenow et al., 1997). Binding of the bacterial adhesin to the secretory component region of the polymeric immunoglobulin receptor, which enhances pneumococcal adherence and invasion (Zhang et al., 2000), is mediated via a highly conserved hexapeptide motif in SpsA (Hammerschmidt et al., 2000). Furthermore, pneumococci have also been demonstrated to bind to immobilized fibronectin (van der Flier et al., 1995), a major constituent in the ECM. The adherence and penetration of the basement membrane and the underlying ECM, which is required to gain access to the blood, has recently been demonstrated for pneumococci. Using a dual-chamber model with reconstituted basement membrane, bacterial migration was shown to be promoted by surface-associated plasminogen (Eberhard et al., 1999).

Plasminogen is a single-chain glycoprotein with a molecular mass of ≈ 92 kDa and represents the monomeric proenzyme of the serine protease plasmin, which plays a crucial role in fibrinolysis (Collen and Verstraete, 1975), homeostasis (Saksela and Rifkin, 1988) and the degradation of ECM (Vassalli et al., 1991). In addition to the characterized eukaryotic plasminogen activators tPA (tissue-type plasminogen activator) and uPA (urokinase-type plasminogen activator), several bacterial plasminogen activators have also been identified. Typical examples are the streptokinase of pyogenic streptococci (Tewodros et al., 1995) and the staphylokinase of Staphylococcus aureus (Matsuo et al., 1990). Both are secreted plasminogen activators responsible for clot lysis by forming a complex with plasminogen and thereby converting other plasminogen molecules to active plasmin (McClintock and Bell, 1971; Paoni and Castellino, 1979).

A wide spectrum of pathogens has been found to capture plasmin(ogen), potentially allowing the bacteria to acquire surface-associated proteolytic activity. This enzymatic activity may facilitate their invasion and dissemination in the infected host (Lottenberg et al., 1994; Lottenberg, 1997). Several plasmin(ogen) receptors have been identified in Gram-negative and Gram-positive bacteria (Lottenberg et al., 1994) including a 45 kDa protein of Streptococcus pyogenes, designated SEN (streptococcal surface enolase), which has been characterized and identified as a major plasmin(ogen)-binding protein located on the surface of group A streptococci (Pancholi and Fischetti, 1998). SEN also exhibits α-enolase activity and is distinct from the other pyogenic plasmin-binding proteins SDH (Pancholi and Fischetti, 1992), Plr (Lottenberg et al., 1992) and PAM (Berge and Sjobring, 1993). Using anti-SEN antibodies, α-enolase-like molecules have also been identified on the surface of both encapsulated and unencapsulated S. pneumoniae.

In the present study, we report, for the first time that S. pneumoniae express several plasminogen-binding proteins. Of these, a 47 kDa protein representing the glycolytic enzyme α-enolase (Eno), has been characterized and identified as a novel human plasminogen-binding protein for pneumococci. Data provide evidence that the Eno protein is present both on the pneumococcal surface and in the cytoplasmic compartment of the pathogen.

Results

Binding of human plasminogen and plasmin to pneumococcal cells

The binding of human plasminogen and plasmin to pneumococci belonging to 26 different serotypes was determined using radiolabelled ligands. The defined strains NCTC 10319 (serotype 35A), NCTC 7978 (serotype 3), ATCC 11733 (serotype 2), D39, R6x, the phenotypic variants P62, P64, P376 and P384 and a further 43 pneumococcal clinical isolates from different clinical sources were used for initial screening. None of the strains showed a detectable interaction with human plasmin. In contrast, 75% of the tested strains were positive for binding to human plasminogen (data not shown). Differences in plasminogen binding among transparent and opaque variants were not significant. The type 2 pneumococcal strain ATCC 11733 was selected to investigate the saturation of binding sites in competitive inhibition experiments using human glu-plasminogen as both radioligand and competitor. The results showed a saturation curve with a calculated concentration of ≈ 1.02 µM unlabelled plasminogen causing 50% blocking of the binding sites (data not shown). Furthermore, plotting the data according to Scatchard (1949) yielded two linear plots (data not shown). In addition, competitive inhibition experiments were also performed using lys-plasminogen, kringle 1–3, and epsilon-amino caproic acid (EACA) as competitors in concentrations of 0.1 µg, 0.5 µg, 1 µg, 5 µg and 20 µg. The binding of [125I]-plasminogen to pneumococci was inhibited in a dose-dependent matter. The results revealed that 50% inhibitory dose values for lys-plasminogen, kringle 1–3 and EACA were of the same order of magnitude as those determined for glu-plasminogen. Together, these data indicated the critical role of the lysine binding sites for plasminogen binding and that binding is most probably mediated through the kringle domains in plasminogen. Pretreatment of pneumococci with proteolytic enzymes such as trypsin and pronase E abolished plasminogen binding (data not shown).

Analysis of human plasminogen binding by blot overlay assay

Soluble radioiodinated human plasminogen was also used in an overlay assay to determine the molecular masses of the putative pneumococcal plasminogen receptors. Proteins of whole-cell lysates of seven selected pneumococcal strains of different serotypes and R6x were resolved by SDS–PAGE and transferred onto nitrocellulose. The results showed a similar binding pattern in all strains. Binding of plasminogen to eight bacterial components with estimated molecular masses of ≈ 83 kDa, 54 kDa, 47 kDa, 39 kDa, 36 kDa, 34 kDa, 31 kDa, 27 kDa and 19 kDa was observed, with the highest binding to proteins with estimated molecular masses of 53 kDa and 47 kDa (Fig. 1). Pretreatment of the pneumococcal lysates with proteolytic enzymes such as trypsin and pronase E resulted in complete loss of plasminogen binding (data not shown).

Figure 1.

Binding of 125I-radiolabelled human plasminogen to different serotypes of S. pneumoniae. Lanes: 1, R6x; 2, S. pneumoniae type 2 (ATCC 11733); 3, opaque clinical isolate P62 type 9V; 4, transparent clinical isolate P64 type 9V; 5, opaque clinical isolate P376 type 6A; 6, nasopharynx isolate SP129; 7, S. pneumoniae type 35A (NCTC 10139); 8, S. pneumoniae D39 type 2.

Reactivity of the 47 kDa pneumococcal plasminogen-binding protein with anti-SEN

Antiserum raised against the S. pyogenesα-enolase SEN also reacted with pneumococci (Pancholi and Fischetti, 1998). Therefore, binding of human plasminogen and anti-SEN polyclonal serum to pneumococci was investigated in immunoblot analysis. The results showed that both human plasminogen and anti-SEN antiserum bind to a 47 kDa protein (Fig. 2). Cell wall-associated proteins and cytoplasmic proteins were treated and fractionated to determine the subcellular location of the reactive protein. In both blot overlay analysis using human plasminogen and immunoblot analysis using anti-SEN antiserum, strong reactivity was observed to a 47 kDa cytosolic protein (Fig. 2A and C). These data suggested that the 47 kDa plasminogen-binding protein was concentrated in the cytosolic fractions. Because human plasminogen binding and anti-SEN reactivity to the 47 kDa protein was also observed in the cell wall fraction, but much less than in the cytosolic fraction, the localization of the protein could not be clearly defined using this approach.

Figure 2.

Identification of a 47 kDa α-enolase-like protein as a plasmin(ogen)-binding protein of S. pneumoniae. Western blot of cell wall fraction (lane 1), cytosol fraction (lane 2) and purified recombinant Eno (lane 3) analysed with human plasminogen (A), [125I]-plasmin (B) and anti-SEN antibodies (C).

N-terminal sequence analysis of the 47 kDa plasminogen-binding protein

The cytosolic fractions showing a high binding of human plasminogen to the 47 kDa protein were combined and precipitated with 40–60% ammonium sulphate, resolved by SDS–PAGE and subsequently transferred onto a polyvinylidene difluoride (PVDF) membrane. Plasminogen binding was confirmed by a blot overlay assay, and the corresponding band was excised and subjected to N-terminal sequencing.

The resulting peptide sequence MIITDVYAREVLDSRGNPP showed 100% homology to a hypothetical protein identified in a type 4 pneumococcal strain used for genomic sequencing (http://www.tigr.org/tdb/mdb/mdbinprogress.html). The deduced 434-amino-acid sequence of the hypothetical protein with a predicted molecular mass of 47 102 Da and pI of 4.48 was compared with known proteins revealing homology to α-enolases of other Gram-positive organisms such as Bacillus subtilis, Streptococcus intermedius and Staphylococcus aureus. The results also revealed significant similarity to other known enolase-encoding genes from Gram-negative bacteria such as Enterococcus faecalis and Neisseria meningitidis(Table 1), as well as to other open reading frames (ORFs) of Gram-negative organisms (data not shown) and to N-terminal residues of eukaryotic proteins as described recently (Pancholi and Fischetti, 1998). Furthermore, using the tblastn algorithm (http://www.sanger.ac.uk/cgi-bin/nph-Blast_Server.html), sequence analysis also revealed a 93% identity to the sequence most probably representing the α-enolase SEN (Pancholi and Fischetti, 1998). The α-enolase-like sequence of the 47 kDa plasminogen-binding protein of S. pneumoniae was designated Eno. The designation Eno has been used elsewhere by Rimini et al. (2000), who described a 2.4-fold downregulation of eno transcription during competence. The deduced amino acid sequence of Eno showed neither the hexameric LPXTGX motif typical for anchoring proteins of Gram-positive bacteria to the cell wall (Fischetti et al., 1990) nor uncharged repeats of 20 amino acids each that mediate the non-covalent anchoring of choline-binding proteins to the cell surface (Yother and White, 1994). In addition, no potential signal-peptidase cleavage site could be identified (von Heijne, 1986).

Table 1. α-Enolase-like bacterial sequences and identity to the plasminogen-binding protein of S. pneumoniae encoded by eno as revealed by sequence comparison.
OrganismProteinMatrix protein bindingIdentity (%)Accession no.
  1. ND, not determined; NA, not available.

S. pyogenes SENPlasmin(ogen)93NA
S. intermedius EnoND95AB029313
E. faecalis EnoND82AJ402252
B. subtilis EnoND71.3L29475
S. aureus EnoLaminin81AF065394
N. meningitidis Segment 5/7ND62AL162756.2
serogroup A strain Z2491   NMA5Z2491
N. meningitidis Section 119ND62AE002477.1
serogroup B MC58

Expression of Eno and plasmin(ogen)-binding activity

The identified 1305 bp ORF in the genomic sequence of S. pneumoniae type 4 was used to design oligonucleotides for polymerase chain reaction (PCR) amplification of the eno gene of the serotype 2 strain ATCC 11733. Sequence analysis of the amplified gene revealed a 100% identical nucleotide sequence to the eno gene of the type 4 pneumococcal strain. The PCR-amplified eno was cloned into expression vector pQE and, after purification, the His-tagged Eno protein was tested for human plasminogen- and plasmin-binding activity. The results of the blot overlay assays performed under native conditions (data not shown) and denaturing conditions indicated that the Eno protein interacts with human plasminogen (Fig. 2A) and also with plasmin with relatively lower affinity (Fig. 2B). In immunoblot analysis, the cross-reactivity of Eno with the anti-SEN antiserum was confirmed (Fig. 2C). This cross-reactivity was also shown in reverse for the SEN protein of group A streptococci using purified anti-Eno antiserum (data not shown).

Purified Eno protein reassociates to the pneumococcal cell surface

Different concentrations of purified His-tagged Eno protein were used in an attempt competitively to inhibit the binding of [125-I]-human plasminogen to S. pneumoniae. However, the binding of human plasminogen, expressed as a percentage of total binding, was strongly enhanced in a dose-dependent manner when higher amounts of Eno protein were added to soluble [125-I]-human plasminogen and intact pneumococci (Fig. 3A). As binding of [125I]-plasminogen to pneumococci was measured, these results suggested that the Eno protein was precipitated together with the intact pneumococci, suggesting a reassociation of soluble Eno and/or Eno–plasminogen complex with the cell wall of pneumococci. To clarify the suspected reassociation, both radiolabelled Eno protein and a complex of equimolar amounts of Eno protein and plasminogen were assayed for binding to pneumococci. The results showed that both ligands bind to pneumococcal cells (Fig. 3B). Taken together, these results suggested that soluble Eno protein is able to reassociate to intact pneumococci and capture plasminogen, leading to enhanced plasminogen binding to pneumococci.

Figure 3.

Binding of radiolabelled human plasminogen (A) and Eno protein (B) to S. pneumoniae R6x. The binding assay with 125I-labelled plasminogen was performed in the presence of different amounts of unlabelled recombinant Eno protein and a fixed amount of 13.5 ng of radiolabelled plasminogen. Binding of purified His-tagged Eno protein to intact pneumococci was performed using increasing amounts of [125I]-Eno protein. Each value is a mean of triplicates.

Human plasminogen binding to carboxypeptidase B-treated Eno protein

Plasminogen-binding activity to proteins depends on the presence of C-terminal lysine residues. The cell wall and cytoplasmic fractions of S. pneumoniae were therefore treated with carboxypeptidase B to determine whether plasminogen binding could be abolished. Both treated and untreated protein fractions were resolved by SDS–PAGE, and binding activity was observed in blot overlay assays. A reduction in plasminogen-binding activity was observed in carboxypeptidase B-treated Eno compared with untreated protein (data not shown). These data were confirmed in binding experiments in which the pretreatment of intact pneumococci with increasing units of carboxypeptidase B resulted in a reduction in but not complete loss of binding of the radioiodinated human plasminogen (data not shown). Using purified recombinant Eno protein in the carboxypeptidase B assay, a reduction in plasminogen binding was also observed (Fig. 4A and B). Taken together, the results suggested that carboxypeptidase B treatment affects the plasminogen-binding activity of Eno, but that domains other than the C-terminal lysines may also play a significant role in the binding.

Figure 4.

Effect of carboxypeptidase B treatment on human plasminogen binding to purified Eno protein (A and B). Purified untreated Eno protein (5 µg; lane 1), with 1 IU (lane 2), 10 IU (lane 3) and 50 IU (lane 4) of carboxypeptidase B were analysed for plasminogen binding by blot overlay (A). Immunoreactivity with anti-Eno antiserum was used as a control (B).

Effect of site-directed mutagenesis of C-terminal lysyl residues of Eno on plasminogen-binding activity

To validate the critical role of the C-terminal lysyl residues for plasminogen-binding activity, both the penultimate Lys-433 and the terminal Lys-434, or the terminal Lys-434 alone, were replaced with a leucine using site-directed mutagenesis. The sequences of both mutated eno DNAs were confirmed by DNA sequencing. The mutated Eno proteins were compared with wild-type Eno for their ability to bind human plasminogen. The results of the blot overlay assay revealed that both mutated Eno433–434 and mutated Eno434 showed a marked reduction in plasminogen-binding activity (Fig. 5A and B).

Figure 5.

Effect of Leu substitutions in C-terminal lysyl residues on plasminogen-binding activity.

A. Binding of human plasminogen by blot overlay assay to mutated Eno434 with individual Leu substitution at position 434.

B. Eno433–434 with Leu substitutions at positions 433 and 434.

C. Immunoblot analysis with purified anti-Eno antiserum. Lanes: purified His-tagged fusion protein Eno (1); lysates of recombinant clones expressing Eno434 (2) and Eno433–434 (3).

α-Enolase activity of Eno and whole pneumococcal cells

To confirm that Eno has α-enolase activity, a single enzyme assay was performed (Pancholi and Fischetti, 1998). His-tagged Eno protein was purified, and different concentrations were used to test α-enolase activity by monitoring the conversion of 2-phosphoglycerate (2-PGE) to phosphoenolpyruvate (PEP) at 240 nm. The results showed a dose-dependent conversion of 2-PGE to PEP (data not shown). Viable pneumococci strains also showed a dose-dependent α-enolase activity, suggesting surface localization (data not shown). In the absence of 2-PGE in the reaction mix, no detectable enzymatic activity was measured (data not shown). Purified Eno protein (5 µg) with various concentrations of 2-PGE (0.25–12 mM) were used to study the kinetics of the reaction. The results were plotted as Michaelis–Menten and double-reciprocal Lineweaver–Burk plots (Fig. 6). A kinetics coefficient (Km) of 4.5 mmol and a Vmax of 2.792 µM min−1 were calculated from the Lineweaver–Burk plot (Lineweaver and Burk, 1934) of the pneumococcal Eno protein.

Figure 6.

α-Enolase activity of purified Eno. Enzyme kinetics of Eno were determined according to the single enzyme assay. The rate of conversion of 2-PGE [S] to PEP in the presence of various amounts of substrate (2-PGE; 0.25–12 mM) was measured at 240 nm using 5 µg of the purified Eno as enzyme, and data were plotted by the method of Michaelis–Menten (inset) and Lineweaver–Burk. Evaluation of the Lineweaver–Burk plot gave Vmax − 1 of 0.3582 (intercept on y-axis), i.e. 2.792 µM min−1 and Km − 1 of 0.23 (intercept on x-axis), i.e. 4.5 mmol of 2-PGE.

Distribution, gene expression of eno among different serotypes and human plasminogen-binding activity of expressed Eno

The presence of the eno gene was investigated by PCR amplification with primers SB20 and SB22 using defined strains and clinical isolates among 26 different serotypes. A single PCR fragment of 1.3 kb was amplified in all assayed strains (Fig. 7A). Restriction fragment length polymorphism (RFLP) using the enzymes Sau3A, EcoRI, PstI, HpaI and TaqI confirmed the conservation of eno among different serotypes. However, Southern blot analysis of HindIII- or EcoRI-digested chromosomal DNA hybridized with a PCR-derived digoxigenin (DIG)-labelled eno probe (SB20/SB22) showed a variation in the hybridization pattern in some strains (data not shown). These data indicated that the eno gene is highly conserved among different serotypes but, in contrast, the molecular organization of the genomic locus of eno varies among different strains. In order to determine the gene expression of eno among different serotypes, Northern blot analysis was performed with mRNA isolated from pneumococcal strains NCTC 10319 (serotype 35A), NCTC 7978 (serotype 3), ATCC 11733 (serotype 2), D39 and R6x. In addition, reverse transcription of the mRNA was performed using oligonucleotide SB19 and subsequent PCR using oligonucleotide primers SB51 and SB19. Both Northern blot and semi-quantitative reverse transcription (RT)–PCR indicated similar gene expression of eno for all strains tested. A size of ≈ 1.3 kb was determined for the eno transcript, suggesting a monocistronic transcript (data not shown). In parallel, the expression of the plasminogen-binding protein Eno in the strains assayed above was also shown by immunoblot analysis with purified anti-Eno antiserum (Fig. 7B).

Figure 7.

Distribution of the eno gene and prevalence of Eno in different serotypes of S. pneumoniae.

A. PCR-amplified products comprising the complete eno gene (1305 bp) of 10 pneumococcal strains among different serotypes.

B. Immunoblot analysis with purified anti-Eno antiserum.

Eno is essential for viability of S. pneumoniae

In an attempt to generate a pneumococcal strain with a disruption in eno, a type 2 S. pneumoniae strain (ATCC 11733) and R6x were transformed with plasmid pMSB1. A chloramphenicol acetyltransferase (cat) gene (Claverys et al., 1995) was cloned into the PstI restriction site present within the eno gene. Transformation of the construct in pneumococci should result in an insertion of the cat gene by allelic exchange and, thus, mutation of the eno gene. However, no transformants were obtained with pMSB1 despite high transformation efficiency being observed with the control plasmid pJDC9::spsA (Hammerschmidt et al., 2000). These results confirm Eno as an essential enzyme of the glycolytic pathway, which might be essential for viability of S. pneumoniae.

Subcellular localization of Eno protein and plasminogen binding

Eno protein was localized in pneumococci by applying pre- and post-embedding labelling methods using unencapsulated strain R6x and encapsulated strain ATCC 11733. Pre-embedding labelling studies revealed bacterial cell surface localization of Eno protein in both strains. For the encapsulated strain, Eno was localized at the outer edge of the bacterial capsule, which was imaged as a hollow around the bacterial cell (Fig. 8A). In the unencapsulated strain, those structures were not visible, and the labelling for Eno protein was restricted to the cell wall region of the bacterial cell (Fig. 8). In an attempt to visualize reassociation of the added Eno protein to pneumococci, the same methodology was applied as for the localization of the indigenous Eno protein. A concentration-dependent reassociation of Eno protein to the pneumococcal surface was observed (Fig. 8B, C, E and F). The capability of Eno protein to bind plasminogen was also confirmed by immunoelectron microscopy. In Fig. 9, the effect of different concentrations of Eno protein added before plasminogen is depicted. Whereas 1 µg of Eno protein resulted in a nearly equivalent labelling intensity to that for untreated cells (Fig. 9A and B), higher amounts of Eno protein resulted in increased plasminogen binding (Fig. 9B and D). In contrast to added Eno protein alone (Fig. 8), plasminogen was bound in aggregates to the pneumococci preincubated with Eno protein (Fig. 9D and F). In Fig. 10A and B, the aggregation of plasminogen is depicted for samples that have been fixed, dehydrated with acetone, critical point dried and carbon coated. Aggregates of plasminogen bound to previously added Eno protein (20 µg) are clearly visible on the bacterial surface for both strains. Ultrathin sections of the same samples revealed that, in the encapsulated strain ATCC 11733, the labelled plasminogen exhibits a spatial distance to the bacterial cell wall (Fig. 10D) compared with the unencapsulated strain (Fig. 10C). Post-embedding studies also confirmed the differences between these two strains. In both strains, Eno protein was detected in the cytoplasm of the bacterial cell and in the cytoplasmic membrane–cell wall region of the cells (Fig. 10E and F). Only the encapsulated strain ATCC 11733 exhibited a spatial distance between gold labelling and the cell wall caused by the capsule (Fig. 10F).

Figure 8.

Electron microscopic localization and reassociation of Eno protein. Eno was detected on the bacterial surface by anti-Eno antibodies and 15 nm protein A–gold particles; images were taken from uncoated samples (A–C, encapsulated strain ATCC 11733; D–F, unencapsulated strain R6x). White dots represent gold particles, and white stars indicate the capsule.

A and D. Cells without Eno.

B and E. Cells with 5 µg of Eno.

C and F. Cells with 20 µg of Eno.

Figure 9.

Electron microscopic visualization of plasminogen binding to Eno protein with anti-plasminogen antibodies and 15 nm protein A/G–gold particles; images were obtained with uncoated samples. Strain R6x untreated (A) and preincubated with 1 µg of Eno protein (B), 5 µg of Eno protein (C) and 20 µg of Eno protein (D) before adding plasminogen. Labelling intensities (white dots) increased remarkably with increasing amounts of Eno protein added before plasminogen incubation. Furthermore, aggregation of added plasminogen was obvious on the bacterial surface (D). Whereas Eno protein labelling (Fig. 8) revealed only single gold particle labelling on the bacterial surface even after adding 20 µg of additional Eno protein, plasminogen labelling exhibits aggregates of gold particles, which are indicative of plasminogen aggregation on the bacterial surface. The same observation was made with the encapsulated strain ATCC 11733 (E, preincubated with 5 µg of Eno protein; F, preincubated with 20 µg of Eno protein). In the bacterial capsular region (white star in E), single gold particles and aggregated gold particles were detectable (E), whereas preincubation with higher amounts of Eno protein (20 µg) resulted in the formation of large plasminogen aggegrates on the surface (F). The addition of Eno protein and plasminogen to pneumococci together resulted in nearly no labelling on the bacterial surface (G) but, instead, large aggregates of presumably Eno protein and plasminogen were detected (H). In a control experiment (I), preincuabtion of strain ATCC 11733 with 20 µg of Eno protein followed by anti-plasminogen antibodies resulted in no labelling, demonstrating the specificity of the plasminogen labelling obtained in (A–F).

Figure 10.

Subcellular localization of plasminogen using electron microscopic pre- and post-embedding labelling studies. Strains R6x (A) and ATCC 11733 (B) exhibited gold labelling with anti-plasminogen antibodies and protein A/G–gold particles on plasminogen aggregates on the bacterial surface of fixed, dehydrated, critical point dried and carbon-coated samples. Ultrathin sections of the pre-embedding labelled samples imaged in (A) and (B) revealed that, for the unencapsulated strain R6x (C), plasminogen is located directly on the bacterial cell wall region (black dots), whereas for the encapsulated strain (D), a clearly defined space between plasminogen label and the bacterial cell wall was visible, demonstrating that plasminogen binds to the outer edge of the bacterial capsule. Post-embedding labelling studies showed similar results for the subcellular localization of Eno protein in the two strains. Eno protein was detecable in the cytoplasm of the cell and at the cell wall–cytoplasmic membrane region for strain R6x (E). The same distribution of Eno protein was visible for the ATCC 11733 strain, with the exception that Eno protein was detectable at the outer edge of the bacterial capsule (F).

Discussion

S. pneumoniae adherence to reconstituted basement membrane is promoted by surface-associated plasminogen (Eberhard et al., 1999). Simultaneously, pneumococci are also able to acquire proteolytic activity by binding to plasminogen and its subsequent activation by the tPA (Eberhard et al., 1999). Plasmin is the key enzyme of the plasminogen system and contributes to the degradation of a variety of matrix constituents (Saksela and Rifkin, 1988). Binding to plasminogen and its conversion to plasmin (a serine protease) may contribute to pathogenicity by facilitating tissue invasion (Lottenberg et al., 1994). Recently, a glycolytic enzyme, α-enolase, was identified as a plasmin(ogen)-binding protein of group A streptococci located on the bacterial surface (Pancholi and Fischetti, 1998). This 45 kDa protein, designated SEN, was different from the previously described streptococcal surface dehydrogenase (SDH), a major protein on the surface of pyogenic streptococci capable of binding to plasmin weakly (Pancholi and Fischetti, 1998). Moreover, the presence of α-enolase-like molecules on the surface of both encapsulated and unencapsulated S. pneumoniae (Pancholi and Fischetti, 1998) and the ability of pneumococci to bind plasminogen and its subsequent activation to plasmin by tPA (Eberhard et al., 1999) also suggested that the pneumococcal α-enolase-like molecules might act as plasmin(ogen)-binding proteins.

In this study, it is shown that binding of human plasminogen to S. pneumoniae is mediated by different bacterial ligands. Of these, the pneumococcal α-enolase, designated Eno, has been identified and characterized as a surface-displayed protein that binds both plasminogen and plasmin and also exhibits glycolytic enzyme activity. The kinetics coefficients (Km) and Vmax values for the pneumococcal α-enolase were comparable with those published for purified SEN and other reported α-enolase enzymes (Spring and Wold, 1971). In addition, the plasminogen- and plasmin-binding activities of Eno are consistent with those reported for SEN, as Eno also binds plasmin with relatively lower affinity than plasminogen. The results of competitive inhibition assays further underline the pivotal role of the lysine binding sites and suggest a binding of plasmin(ogen) via the kringle domains. Eno is also cross-reactive with anti-SEN antiserum, which has previously been shown to react with α-enolase-like molecules on pneumococci, but not on staphylococci (Pancholi and Fischetti, 1998). Interestingly, Eno was present in the cytoplasm as well as located on the pneumococcal cell surface.

α-Enolases are one of the key enzymes acting as a 2-phospho-d-glycerate-hydrolase in the glycolytic cycle located in the cytoplasm of prokaryotes and eukaryotes. However, SEN of group A streptococci, showing a 93% sequence similarity to Eno, was recently reported as the first cell surface α-enolase to be expressed in prokaryotes (Pancholi and Fischetti, 1998). In addition, expression of α-enolase-related molecules on the surface has also been demonstrated for several eukaryotic cell lines (Miles et al., 1991; Lopez-Alemany et al., 1994; Redlitz et al., 1995). The presence of Eno as the sole α-enolase molecule for pneumococci, the strong conservation of its gene, its equal gene expression in all strains, its ability to bind plasminogen and the fact that disruption of the eno gene appeared to be lethal indicate that Eno is an essential glycolytic enzyme. Nevertheless, Eno also represents an α-enolase-like molecule on the pneumococcal surface. Analysis of the deduced amino acid sequence of Eno showed the presence of neither a signal sequence nor a hexameric LPXTGX motif. These domains have been shown to be crucial for translocation through the cell wall and for anchoring surface proteins to the Gram-positive bacterial cell wall (Fischetti et al., 1990). Furthermore, repeats that might anchor the protein non-covalently to the surface, as shown for choline-binding proteins of S. pneumoniae (Yother and White, 1994) and internalin B of Listeria monocytogenes (Braun et al., 1997), were not present on Eno. In spite of this, α-enolase activity and immunoelectron microscopic studies indicated that Eno is secreted and attached to the bacterial cell surface.

One possibility for surface localization is the reassociation of secreted Eno onto the pneumococcal surface. In a first attempt to elucidate whether soluble recombinant Eno protein is reassociated to the surface by itself or only in conjunction with plasminogen as an Eno–plasminogen complex, a further binding experiment was performed using increasing concentrations of the recombinant purified Eno as radiolabelled ligand. The data indicated a dose-dependent reassociation of soluble Eno protein to the pneumococcal surface. Pretreatment of pneumococci with proteolytic enzymes and glutaraldehyde, respectively, abolished binding of radiolabelled Eno to pneumococci completely (data not shown), suggesting a proteinaceous nature of the receptor for Eno protein. The data from the plasminogen-binding experiments with unlabelled Eno are in contrast to the ability of purified SEN competitively to inhibit binding of plasminogen to S. pyogenes (Pancholi and Fischetti, 1998). However, in the competitive inhibition studies of Pancholi and Fischetti (1998), the streptococci were cross-linked with glutaraldehyde before the addition of SEN to the reaction mix. The cross-linking might have abolished reassociation of SEN to the surface as a result of a loss of receptors, thus allowing free SEN to inhibit binding of plasminogen to streptococci.

Localization of Eno protein as a plasminogen-binding protein on the pneumococcal cell surface and its reassociation was confirmed by immunoelectron microscopic analysis. S. pneumoniae strain type 2 and pneumococcal strain R6x were used to differentiate the reassociation of Eno to an encapsulated and an unencapsulated strain. Both pre- and post-embedding studies revealed the localization of Eno protein at the outer surface of the bacterial cells irrespective of the status of capsule expression. However, Eno was also detected in the cytoplasm, indicating its pivotal role as a glycolytic enzyme. Using plasminogen, binding was localized to the same cell surface regions of the pneumococci as those detected for anti-Eno antiserum. When using different amounts of Eno protein for preincubation of the bacteria, higher amounts of Eno protein resulted in the bacteria having a higher binding capacity for plasminogen but not other serum components (data not shown), confirming the reassociation of Eno to the pneumococcal cell surface. The precise mechanisms of how Eno and proteins such as SEN (Pancholi and Fischetti, 1998) and SDH (Pancholi and Fischetti, 1992) are transported through the cell wall and how these proteins are anchored to the bacterial cell surface needs further investigation.

Plasmin(ogen)-binding interaction is mediated by recognition of the C-terminal lysine residues of eukaryotic enolases by the lysine binding sites of plasminogen (Redlitz et al., 1995). Carboxypeptidase B treatment of whole pneumococci and purified Eno protein led to a significant reduction in plasminogen-binding activity, supporting the important role for C-terminal lysines in binding. Confirmation of the C-terminal lysyl residues as the crucial binding motif for plasmin(ogen) was obtained by site-directed mutagenesis. Leucine substitutions for either the terminal lysine or both the penultimate and terminal lysyl residues resulted in a marked reduction in plasminogen binding. Interactions of plasmin(ogen) with a C-terminal lysyl residue have been reported previously for another streptococcal plasmin(ogen)-binding protein (Winram and Lottenberg, 1998) and also for putative eukaryotic plasmin(ogen) receptors (Miles et al., 1991; Hajjar, 1993). These results also indicated that the C-terminal lysyl residues of Eno are surface exposed and accessible for the enzymatic activity of the carboxypeptidase B.

As a consequence of plasminogen activation on bacterial surfaces, bacteria become armed with the broad substrate spectrum proteolytic potential of plasmin that cannot be regulated by the host inhibitors (Lottenberg et al., 1994). The capture of plasminogen by Eno can be used to facilitate pneumococcal penetration through biological membranes by plasminogen activation during the invasive infection process and therefore represents an important determinant of virulence.

Experimental procedures

Bacterial strains, media and growth conditions

Defined strains from culture collections ATCC and NCTC [ATTC 11733 (type 2), NCTC 7978 (type 3), NCTC 10319 (type 35A)], opaque and transparent phenotypes of serotype 9V clinical isolate [P62 (opaque), P64 (opaque)] and of serotype 6A [P376 (opaque), P384 (transparent)] (kindly provided by J. Weiser) and further clinical isolates among 26 different serotypes provided by the Statens Serum Institut, Copenhagen, Denmark, and Medical University Düsseldorf, respectively, as well as the type 2 strain D39 and its unencapsulated derivative R6x (Tiraby and Fox, 1973) were used in this study. S. pneumoniae were cultured at 37°C overnight on blood agar plates or in Todd–Hewitt broth (Oxoid) supplemented with 0.5% yeast extract (THY) to mid-log or late-log phase (Becton Dickinson). Escherichia coli M15[pREP4] (Qiagen) was used as the host strain for recombinant pQE expression plasmids and cultured at 37°C on Luria–Bertani (LB) agar. Expression of the His-tagged fusion proteins was induced with 1 mM IPTG after reaching an OD600 of 0.8, and growth continued at 30°C for 4 h.

Generation of antiserum against Eno

Polyclonal antibodies against purified fusion protein Eno were raised in rabbit using purified Eno encoded by pQSB5 by routine immunogenic procedures (Eurogentec). Preimmune serum was collected before immunization. Purification of anti-Eno for immunofluorescence and immunoelectron microscopy studies was performed by affinity chromatography using protein A Sepharose 4B (Amersham Pharmacia Biotech).

Human plasminogen and plasmin

Human plasminogen was purchased commercially (Sigma). Lys-plasminogen was prepared from glu-plasminogen according to the method of Fleury and Anglés-Cano (1991). Briefly, 40 µM glu-plasminogen in 50 mM Tris-HCl, pH 8.6, was incubated for 4 h at 22°C with urokinase-free plasmin (Sigma) using an enzyme–substrate molar ratio of 1:100. Aprotinin–Sepharose was added to remove plasmin, and the suspension was shaken gently overnight at 4°C, followed by treatment with 15 mM phenylmethylsulphonyl fluoride (PMSF; Roth). The lys-plasminogen solution was dialysed extensively against 0.05 M sodium phosphate buffer, pH 7.4, containing 0.08 M NaCl. Human plasmin was generated from plasminogen by incubation with urokinase (20 IU ml−1; Sigma) for 1 h at 37°C in HBS gel buffer (50 mM HEPES–NaOH, pH 7.4, containing 1 mM MgCl2, 0.15 mM CaCl2, 0.1% gelatin) containing 40 mM lysine as described previously (Lottenberg et al., 1987). Conversion of glu-plasminogen to lys-plasminogen and plasmin was monitored under denaturing conditions on 10% SDS–polyacrylamide gels.

Binding experiments and competitive inhibition assay

Human plasminogen (Sigma) and plasmin (Sigma) were radiolabelled with 125I by a standard chloramin T method (Chhatwal et al., 1987), and binding experiments with [125I]-plasminogen and prepared [125I]-urokinase–plasmin were performed as described previously (Hammerschmidt et al., 1997). Briefly, 109 pneumococcal cells grown to mid-log phase (300 µl of 10% transmission) were incubated with 20 nCi of radiolabelled proteins for 40 min at room temperature. Pneumococci were sedimented by centrifugation, and the supernatant was aspirated by gentle suction. Pellet-bound radioactivity representing bound plasminogen and plasmin was measured in a gamma counter (Packard). Plasminogen binding was expressed as a percentage of total radioactivity added. In competitive inhibition assays, binding to viable pneumococci was measured in the presence of increasing molar excesses of glu-plasminogen, prepared lys-plasminogen, epsilon-amino caproic acid (EACA; Merck), lysine binding site 1 of plasminogen (kringle 1–3; Sigma) and purified His-tagged Eno. Both proteins were used in amounts from 0.01 ng up to 100 µg compared with a fixed amount of 60 ng of [125I]-plasminogen.

Electrophoresis, Western blot and immunoblot analysis

Protein lysates of E. coli and pneumococci as well as fusion proteins were subjected to SDS–PAGE with 12% gels according to the method described by Laemmli (1970) and either stained with Coomassie brilliant blue (CBB) or subsequently transferred to a nylon membrane (Immobilon-P; Millipore) using a semi-dry blotting system. Binding of human plasminogen was performed using either ≈ 60 ng of radiolabelled plasminogen or non-radioactive human plasminogen in conjunction with peroxidase-labelled anti-plasminogen antibodies. The membranes were blocked by incubation in 10% fat-free milk in 10 mM PBS before the binding reaction. Binding of radiolabelled plasminogen and urokinase–plasmin was detected by exposure of the membrane to X-ray film. In immunoblots, plasminogen binding was detected by incubation of the membrane with a substrate solution containing 1 mg ml−1 4-chlor-1-naphthol and 0.1% H2O2 in PBS. Immunoblot analysis with anti-SEN antiserum and anti-Eno antiserum was performed using PO-conjugated second antibody and substrate solution.

Fractionation of pneumococcal cell wall proteins and cytosolic proteins for sequencing

For the fractionation of cell wall proteins and cytosolic proteins, bacteria were grown to mid-log phase, harvested and washed twice with 0.1 M Tris-HCl buffer (pH 7.2). The bacterial pellet was resuspended on ice in 10 ml of Tris-HCl buffer supplemented with 1 mM PMSF. Cell lysis was carried out by two passages through a French press with calculated internal cell pressure of 17 000 Ib in−2. Intact bacteria were removed by centrifugation for 15 min at 3000 g. The cell wall fraction was separated from the cytosolic fraction by centrifugation for 50 min at 19 000 g. The cytosolic proteins were subjected to ammonium sulphate precipitation at 20%, 40%, 60% and 80% saturation in 10 mM Tris-HCl buffer (pH 7.2). The proteins were sedimented by centrifugation for 30 min at 19 000 g. The cell wall fraction was resuspended in 0.1 M potassium phosphate buffer (pH 7.2) containing 150 mM KCl, 10 mM EDTA and 20% glycerol. Solubilization of the cell wall proteins was performed by sequential treatment for 30 min at 30°C in the presence of 5 mM, 10 mM, 20 mM and 50 mM detergent CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulphonate}. After centrifugation for 30 min at 19 000 g, the supernatants were collected, combined and precipitated with ammonium sulphate. The cell wall fractions and the protein fractions of the cytosol were dialysed against PBS overnight at 4°C and assayed for human plasminogen activity by an overlay assay. The proteins were resolved by SDS–PAGE and transferred onto a PVDF membrane. The pneumococcal protein showing reactivity with the anti-SEN antiserum and human plasminogen was excised from a duplicate membrane and subsequently subjected to N-terminal sequencing.

Recombinant DNA techniques and PCR

Transformation of E. coli with recombinant plasmids was achieved by electroporation (Calvin and Hanawalt, 1988). Plasmid DNA was isolated using the Qiagen plasmid kit, and PCR products were purified using the PCR purification kit (Qiagen). T4 DNA ligase and restriction enzymes were purchased from New England Biolabs and used according to the manufacturer's instructions. Southern blot analysis under stringent conditions and detection of DIG-labelled hybrids were performed according to the manufacturer's instructions (Roche). PCR was performed on a thermocycler (Hybaid) with 0.05 µg of chromosomal DNA as template using the AmpliTaq Gold polymerase under buffer conditions recommended by the manufacturer. Each reaction consisted of 35 cycles including 30 s of denaturation at 94°C, 30 s of annealing and an extension at 72°C. For generating DIG-labelled DNA probes, the PCR DIG labelling mix (Roche) was used in the PCR reaction. Bacterial RNA isolation was performed using the Qiagen RNA/DNA midi kit according to the manufacturer's instructions. Northern blot analysis under stringent conditions was performed according to standard protocols (Ausubel et al., 1989). The ThermoScript RT–PCR system (Life Technologies) was used for reverse transcription with oligonucleotide SB19 (5′-TTATTTTTTAAGGTTGT AGAATGATTTC-3′). The RT–PCR was performed using oligonucleotides SB19 and SB51 (5′-AGTACAACTTGTTG GTGACGA C-3′).

Expression cloning and sequencing

The peptide sequence identified by N-terminal sequencing was used for a homology search using the tblast program based on the algorithm of Lipman and Pearson (1985) against the type 4 genome sequence of S. pneumoniae (http://www.TIGR.org; tblast). Database searching revealed 100% identity of a sequence encoded by an ORF in contig 3836 from nucleotide 955 457 to nucleotide 956 738. This sequence encoding a putative α-enolase of S. pneumoniae was used to design primers for PCR amplification of the ORF. In order to amplify the eno gene among different pneumococcal serotypes, SB20 (5′-GGATCCTTGTCAATTATTACT GATGTTTACGC-3′), incorporating an in frame BamHI restriction site at the 5′ end, and SB22 (5′-AAGCTTTTATTT TTTAAGGTTGTAGAATGATTTC-3′), incorporating an in frame HindIII restriction site at the 3′ end, were used. For expression cloning into similarly digested pQE30 vector (Qiagen), the PCR product of pneumococcal strain ATCC 11733 was digested with BamHI and HindIII and ligated with the vector. The His-tagged fusion protein was purified by chromatography under native conditions on Ni-nitrilotriacetic acid resins according to the manufacturer's protocols (Qiagen). The integrity of insert DNA was verified by sequence analysis using ABI Prism dye terminator cycle sequencing (Perkin-Elmer).

Site-directed mutagenesis of the eno gene

Site-directed mutagenesis of the eno gene was performed using DNA primers containing HindIII restriction enzyme sites for cloning and modifications in the reverse primer to introduce the appropriate base substitutions at the 3′ end of eno. These include replacing bases 1300–1302 (AAA) with nucleotides TTG to replace Lys-434 with leucine and replacing bases 1297–1302 (AAAAAA) with nucleotides TTGTTG to replace both Lys-433 and Lys-434 with leucines. Plasmid DNA pQSB5 (40 ng) was used as DNA template. The reverse oligonucleotide primers used to generate these mutations were SB48 (5′-AACTTTAGTAAGATGTTGGAAA ACAACATTTTCGAA-3′) and SB47 (5′-AACTTTAGTAAGAT GTTGGAATTTAACATTTTCGAA-3′). The forward oligonucleotide primer used for both mutations was SB49 (5′-AAG CTTAATTAGCTGAGCTTGGAC-3′) specific for the vector. The PCR products were digested with HindIII, religated and transformed into E. coli M15 [pREP4]. Expression of the mutated Eno protein was visualized using Coomassie brilliant blue staining and anti-Eno antiserum in immunoblot analysis. Binding of human plasminogen to the mutated Eno proteins was performed as described. The site-specific mutations were verified by sequence analysis using ABI Prism dye terminator cycle sequencing (Perkin-Elmer).

Carboxypeptidase B enzyme assay

Pneumococci as well as purified Eno were treated with carboxypeptidase B (EC 3.4.17.2) (Sigma) in 10 mM PBS, pH 7.0, with 1, 10 and 50 units for 1 h at 37°C. Equal amounts of treated and untreated proteins were resolved by SDS–PAGE, and plasminogen-binding activity was determined by the blot overlay assay. To determine the role of the terminal lysines in plasminogen binding to intact pneumococci, binding of [125I]-plasminogen was measured to carboxypeptidase B-treated intact pneumococci, and the binding was expressed as a percentage of total binding.

α-Enolase activity of pneumococcal cells and Eno protein

The α-enolase activity of viable pneumococci and purified recombinant His-tagged Eno protein was measured in a single assay according to the method of Pancholi and Fischetti (1998). Briefly, different amounts of purified recombinant Eno protein and supernatants of pneumococci, both preincubated for 3 min at 37°C with 3 mM 2-phosphoglycerate (2-PGE), were used to measure the phosphoenolpyruvate (PEP) release at 240 nm in 100 mM HEPES, pH 7.0, containing 10 mM MgCl2 and 7.7 mM KCl. As a positive control, a yeast α-enolase (EC 4.2.1.11; Sigma) was used in the enzyme assay. For kinetics, involving only the conversion of 2-PGE to PEP by α-enolase, different amounts of 2-PGE (0.25–12 mM) were used in the single assay with 5 µg of purified Eno protein. The reactions were performed under the same conditions as described above, with the exception of preincubation.

Construction of an eno-deficient S. pneumoniae mutant

The plasmid pQSB5 containing the gene encoding Eno was linearized by digestion with PstI in position 816 of the insert DNA and blunt ended by T4 DNA polymerase. The purified DNA fragment was ligated with a blunt-ended chloramphenicol acetyltransferase (cat) gene generated by PCR using plasmid pR326 (Claverys et al., 1995) as template DNA and primers cat1 (5′-CGACTCACTATAGGGCGAATTGG-3′) and cat2 (5′-CCTCACTAAAGGGAACAAAAGCTG-3′) followed by Klenow treatment and PCR purification (Qiagen). The resultant plasmid, designated pMSB1, was electroporated into E. coli DH5α. Transformants were selected on LB agar containing 25 µg of chloramphenicol. S. pneumoniae mutants of the wild-type strains ATCC 11733 and R6x were obtained by transformation with plasmid pMSB1, using the competence stimulating peptide to enhance transformation efficiency (Håvarstein et al., 1995). Transformants deficient in plasminogen binding as a result of mutated eno were detected by their lack of reactivity with plasminogen.

Electron microscopy

In an attempt to visualize the subcellular localization of Eno protein and the binding of plasminogen to Eno protein, two approaches were taken. First, pre-embedding labelling studies were carried out in suspensions, followed by imaging in a field emission scanning electron microscope (FESEM) at low acceleration voltages and, secondly, post-embedding studies were performed for intracellular localization of the Eno protein.

Pre-embedding labelling. S. pneumoniae were cultured in THY broth to mid-log phase and harvested by centrifugation. Bacteria were washed twice with PBS and adjusted to 109 cfu ml−1 after blocking for 30 min using fetal calf serum (FCS). A suspension of 1 ml was incubated for 30 min with 10 µl of protein A-purified anti-Eno (1 mg ml−1) antiserum and with 20 µg of human plasminogen. Bacteria incubated with human plasminogen were washed several times with PBS and incubated for 30 min with goat anti-plasminogen IgG antibody (Affinity Biologicals). As a marker for electron microscopic studies, 15 nm protein A–gold particles were used for labelling anti-Eno antibodies and 15 nm protein A/G–gold particles for detecting bound anti-plasminogen antibodies. After washing the samples in PBS buffer, samples were post-fixed with 1% formaldehyde in PBS for 30 min at room temperature followed by washing in TE buffer (20 mM Tris, pH 6.9, 1 mM EDTA). Bacteria were then adsorbed onto carbon-coated Formvar grids, washed in TE buffer and distilled water before air drying. The grids were then mounted on aluminium stubs with conductive carbon adhesive tabs. Samples were examined in a Zeiss FESEM DSM962 Gemini at acceleration voltages between 1.5 and 2 kV using the Everhard–Thornley SE detector.

Post-embedding labelling. Bacteria were grown in THY broth to mid-log phase, centrifuged and fixed in a solution containing 0.2% glutaraldehyde and 0.5% formaldehyde in PBS for 1 h on ice. After washing in PBS containing 10 mM glycine to block free aldehyde groups, cells were embedded into 2% water agar, cut into small cubes and dehydrated with a graded series of ethanol before embedding in LR White resin (London Resin). After polymerization at 50°C for 24 h, ultrathin sections were cut and collected onto Formvar-coated 300-mesh nickel grids. Sections were then incubated on drops of the anti-Eno protein antibodies (100 µg ml−1 IgG protein) for 12 h at 4°C. After washing with PBS, sections were incubated on drops of 0.4% skim milk in water for 5 min, dry blotted onto filter paper before incubation on drops of a 1:100 diluted protein A colloidal gold stock solution (10 nm in size) for 30 min. After washing in PBS containing 0.01% Tween 20, sections were counterstained with 4% aqueous uranyl acetate for 5 min before examination in a Zeiss transmission electron microscope (TEM) EM910 at an acceleration voltage of 80 kV and calibrated magnifications.

Embedding and carbon coating of samples for FESEM of pre-embedding labelled samples. After performing the labelling experiments, samples were fixed with 2% formaldehyde and 2% glutaraldehyde in PBS for 30 min on ice, washed in PBS, embedded in 2% water agar, cut into small cubes, fixed with 1% aqueous osmium tetroxide for 1 h at room temperature and dehydrated with a graded series of acetone before embedding in Spurr's resin (Spurr, 1969). After polymerization at 70°C for 8 h, ultrathin sections were cut with a diamond knife. Sections were then post-stained with uranyl acetate and lead citrate before examination in the TEM.

For SEM preparation of coated samples, prelabelled bacteria were allowed to attach to poly l-lysine-coated coverslips for 5 min. After washing in TE buffer, samples were fixed with 3% glutaraldehyde for 15 min at room temperature, washed in TE buffer and dehydrated with a graded series of acetone before critical point drying using liquid CO2. After coating with carbon, samples were examined in the FESEM at an acceleration voltage of 10 kV.

Acknowledgements

We thank E. Müller and M. P. Tillig for excellent technical assistance, and V. Pancholi (Laboratory of Bacterial Pathogenesis and Immunology, The Rockefeller University, New York, USA) for providing anti-SEN polyclonal antiserum. The authors are also grateful to R. Getzlaff for N-terminal peptide sequence analysis, and R. Towers for critical reading of the manuscript.

Ancillary