Vitronectin binds to the head region of Moraxella catarrhalis ubiquitous surface protein A2 and confers complement-inhibitory activity

Authors

  • Birendra Singh,

    1. Medical Microbiology, Department of Laboratory Medicine, University Hospital Malmö, Lund University, SE-205 02 Malmö, Sweden.
    Search for more papers by this author
  • Anna M. Blom,

    1. Division of Medical Protein Chemistry, The Wallenberg Laboratory, Department of Laboratory Medicine, University Hospital Malmö, Lund University, SE-205 02 Malmö, Sweden.
    Search for more papers by this author
  • Can Unal,

    1. Medical Microbiology, Department of Laboratory Medicine, University Hospital Malmö, Lund University, SE-205 02 Malmö, Sweden.
    Search for more papers by this author
  • Bo Nilson,

    1. Medical Microbiology, Department of Laboratory Medicine, University Hospital Lund, Lund University, SE-223 62 Lund, Sweden.
    Search for more papers by this author
  • Matthias Mörgelin,

    1. Section of Clinical and Experimental Infectious Medicine, Department of Clinical Sciences, Lund University, SE-221 84 Lund, Sweden.
    Search for more papers by this author
  • Kristian Riesbeck

    Corresponding author
    1. Medical Microbiology, Department of Laboratory Medicine, University Hospital Malmö, Lund University, SE-205 02 Malmö, Sweden.
    Search for more papers by this author

*E-mail kristian.riesbeck@med.lu.se; Tel. (+46) 40 331340; Fax (+46) 40 336234.

Summary

The serum resistance of the common respiratory pathogen Moraxella catarrhalis is mainly dependent on ubiquitous surface proteins (Usp) A1 and A2 that interact with complement factor 3 (C3) and complement inhibitor C4b binding protein (C4BP) preventing the alternative and classical pathways of the complement system respectively. UspA2 also has the capacity to attract vitronectin that in turn binds C9 and hereby inhibits membrane attack complex (MAC) formation. We found UspA2 as a major vitronectin binding protein and hence the UspA2/vitronectin interaction was studied in detail. The affinity constant (KD) for vitronectin binding to UspA2 was 2.3 × 10−8 M, and the N-terminal region encompassing residues UspA2 30–170 bound vitronectin with a KD of 7.9 × 10−8 M. Electron microscopy verified that the active binding domain (UspA230–177) was located at the head region of UspA2. Experiments with recombinantly expressed vitronectin also revealed that UspA230–177 bound to the C-terminal region of vitronectin residues 312–396. Finally, when human serum was pre-incubated with UspA2, bacteria showed significantly less serum resistance. Our study directly reveals the binding mode between the N-terminal domain of UspA2 and the C-terminal part of vitronectin and thus sheds light upon the mechanism of M. catarrhalis-dependent serum resistance.

Introduction

Moraxella catarrhalis is a Gram-negative human respiratory pathogen that causes acute otitis media in children and is a common cause of exacerbations in patients with chronic obstructive pulmonary disease (COPD) (Catlin, 1990; Murphy, 1996; Karalus and Campagnari, 2000). The frequency of colonization of M. catarrhalis is high among pre-school children and it is sometimes considered as a commensal in that age group. The species is, however, the third most common bacterial agent causing acute otitis media in children after Streptococcus pneumoniae and Haemophilus influenzae, and is responsible for 20% of the total number of cases.

In the past decade, M. catarrhalis surface proteins have been characterized for their role in antigenicity that could be used as an impetus for the selection of vaccine candidates (Mawas et al., 2009). Several outer membrane proteins, e.g. M. catarrhalis autotransporter (McaP) (Lipski et al., 2007), protein CD (Liu et al., 2007) and ubiquitous surface proteins (Usp) A1/A2 (Cope et al., 1999; Tan et al., 2006a; Manolov et al., 2008) have been extensively studied for functional, biological as well as immunogenic characteristics. In addition, we and others have in detail described the large adhesin Moraxella IgD binding protein (MID) (Mollenkvist et al., 2003; Cotter et al., 2005; Riesbeck and Nordström, 2006), which binds epithelial cells via its residues MID 764–913. The current research progress on M. catarrhalis outer membrane proteins and pathogenesis has recently been reviewed (Tan and Riesbeck, 2007; Ruckdeschel et al., 2008; Murphy and Parameswaran, 2009; Perez Vidakovics and Riesbeck, 2009; de Vries et al., 2009).

Among the outer membrane proteins of M. catarrhalis, the Usp family has been studied most thoroughly for its role in pathogenesis and adhesion to human respiratory epithelial cells (Cope et al., 1999; Hoiczyk et al., 2000; Tan et al., 2005; Brooks et al., 2008a,b). The Usps belong to the oligomeric coiled-coil adhesins, distributed as a dense fuzzy layer on the surface of the bacterial outer membrane (thickness 600–700 Å), and appear as lollipop-like structures (Aebi et al., 1998b). UspA1 and UspA2 (88 and 62 kDa respectively) have 43% identity and both proteins contain a common epitope of 140 aa residues with 93% identity in their C-terminals (Cope et al., 1999). The hybrid UspA2H shares domain similarity with both UspA1 and UspA2, and thus have corresponding functional characteristics based upon the respective protein (Lafontaine et al., 2000; Perez Vidakovics and Riesbeck, 2009).

UspA1 is responsible for binding to conjunctival (Chang), type II alveolar (A549), and to laryngeal (Hep-2) epithelial cells (Lafontaine et al., 2000). In binding assays, both UspA1 and UspA2 interact with extracellular matrix proteins like fibronectin and laminin, and thus more likely function as adhesins (Tan et al., 2005; Tan et al., 2006b). A UspA1-dependent interaction with carcinoembryonic antigen-related cell adhesion molecule-1 (CAECAM-1) has also been described in detail and suggests a crucial role of UspA1 in adhesion of M. catarrhalis as well as dampening of the pro-inflammatory response in the respiratory tract (Hill and Virji, 2003; Conners et al., 2008; Slevogt et al., 2008). Interestingly, UspA1 and UspA2 also bind α-1 antichymotrypsin and neutralizes this enzyme resulting in free chymotrypsin, which may cause increased and excessive inflammation and thus promotion of bacterial colonization particularly in COPD patients (Manolov et al., 2008).

Gram-negative pathogens are killed by the innate defence including the complement system. Activation of the innate immune system leads to a cascade of protein deposition on the bacterial surface, resulting in formation of the membrane attack complex (MAC), and opsonization of the pathogen, followed by phagocytosis (Walport, 2001). Different bacterial pathogens show a diverse array of resistance against complement and many pathogens also escape the bactericidal killing of complement (Lambris et al., 2008; Blom et al., 2009). Among M. catarrhalis, two distinct lineages have been defined and are represented by complement resistant and sensitive isolates respectively (Hol et al., 1995; Aebi et al., 1998a; Bootsma et al., 2000). We previously reported that M. catarrhalis binds to the complement regulator C4b binding protein (C4BP) by means of UspA1 and UspA2, and thus inhibits the classical pathway of complement activation (Nordstrom et al., 2004). In addition, M. catarrhalis interferes with the complement cascade by neutralizing C3 using mainly UspA2 (Nordstrom et al., 2005). The hybrid UspA2H shares domain similarity with UspA1 and UspA2, and consequently also plays a role in adhesion and serum resistance (Lafontaine et al., 2000; Perez Vidakovics and Riesbeck, 2009). Moraxella-dependent serum resistance is, however, not only limited to the Usp family, as some other proteins, like CopB (Aebi et al., 1998a), protein CD (Holm et al., 2004) and protein E (Bhushan et al., 1994), are also involved.

When the functional part of the uspA2 gene from M. catarrhalis strain O35E was transferred to the serum-sensitive M. catarrhalis strain MC317 or to serum-sensitive H. influenzae, it significantly restored serum resistance. Interestingly, vitronectin-depleted serum also restored the bactericidal activity against M. catarrhalis strains while addition of vitronectin to serum regained the serum resistance, suggesting an important role of UspA2-dependent binding of vitronectin (Attia et al., 2005). Another study from the same laboratory indicated a direct role of the UspA2/vitronectin interaction regarding serum resistance of several M. catarrhalis strains (Attia et al., 2006). The rate of C9 polymerization and MAC deposition was higher in UspA2 mutants in comparison to wild-type counterparts. In that particular study, however, other complement proteins like C4BP and C3 did not directly correlate with serum resistance (Attia et al., 2006). It was suggested that the UspA2-dependent binding of vitronectin inhibits MAC formation by interfering with C9 polymerization and that this may be one of the major mechanisms related to M. catarrhalis serum resistance.

The multifunctional complement regulator vitronectin is a 67 kDa glycoprotein consisting of several separate domains. The N-terminal region contains a signal sequence of 15 amino acids followed by a 43-amino-acid-long somatomedin B (SMB) domain, which is known to bind to the plasminogen activator inhibitor 1 (PAI-1) (Zhou et al., 2003). When the SMB domain is excluded, the remaining part of the vitronectin molecule contains four putative haemopexin-like domains and three heparin binding domains (HBD) (Gibson et al., 1999), whereas the role of the C-terminal part of vitronectin is largely unknown.

The N-terminal region (102 amino acids) of UspA2 expressed serum resistance in M. catarrhalis O35E (Attia et al., 2005). However, bioinformatics revealed that UspA1, UspA2 and UspA2H largely differ at their N-terminal ends (Brooks et al., 2008a). The N-terminus of UspA1 is highly conserved and shows homology with YadA of Yersinia enterocolitica, whereas the N-termini of UspA2 have variable amino acid sequences in different M. catarrhalis strains. The N-terminal of UspA2 can be divided into two distinct groups designated N-terminal (NTER) 2A and NTER-2B (Brooks et al., 2008a). Thus, the serum resistance of M. catarrhalis depends on the expression as well as variability of UspA2 (Aebi et al., 1998b; Attia et al., 2005; 2006).

In the present study, we analysed the UspA2/vitronectin interaction in detail. Direct binding assays suggested UspA2 as a major vitronectin binding protein. Interestingly, the N-terminal of UspA2 (residues 30–177) was characterized as the active domain that bound recombinant vitronectin (residues 312–396). Addition of recombinant UspA2 to normal human serum (NHS) caused rapid killing of M. catarrhalis in comparison to control serum, and thus the UspA2/ vitronectin interaction plays a significant role in serum resistance.

Results

UspA2 is the major vitronectin binding protein of M. catarrhalis

Moraxella catarrhalis wild-type and isogenic mutants of uspA1, uspA2 and mid were analysed for vitronectin binding capacity using an [125I]-vitronectin direct binding assay. The results showed that M. catarrhalis wild-type bound vitronectin in a concentration-dependent manner (Fig. 1A). No decreased vitronectin binding was found with the M. catarrhalisΔuspA1 mutant as compared with the isogenic wild-type Moraxella. In contrast, the ΔuspA2 mutant had a significantly decreased binding (57%) of vitronectin (Fig. 1B). The M. catarrhalisΔuspA1/ΔuspA2 double mutant showed very similar vitronectin binding capacity as compared with the ΔuspA2 mutant. The M. catarrhalisΔmid isogenic mutant was included as a negative control and also proved that the well-characterized adhesin MID did not have any effect on Moraxella-dependent binding of vitronectin.

Figure 1.

UspA2 is the major vitronectin binding protein of M. catarrhalis.
A. Direct binding assay showing dose-dependent binding of [125I]-vitronectin to UspA2-expressing M. catarrhalis RH4.
B. Vitronectin binding capacity of M. catarrhalis RH4 wild-type and isogenic mutants shown in relation to the wild-type strain (100%). Mean values of triplicates from three independent experiment are shown and error bars represent standard deviations. Statistical significance of differences was calculated using Student's t-test; ***p ≤ 0.001.
C. TEM showed that M. catarrhalis binds vitronectin at the distal end of UspA2. Gold-labelled vitronectin bound to an approximately 60-nm-long fibrillar structure of M. catarrhalis RH4 wild-type expressing both UspA1 and UspA2.
D. UspA1 is not a vitronectin binding protein in M. catarrhalis, as UspA1 mutants showed similar vitronectin binding as compared with the non-mutated parental strain (C).
E. UspA2 is the major vitronectin binding protein in Moraxella as demonstrated with a M. catarrhalisΔuspA2 mutant that is defective in vitronectin binding.
F. No vitronectin binding was detected to the M. catarrhalisΔuspA1/A2 double mutant. The scale bars represent 100 nm (overviews) and 25 nm (inserts).

The same M. catarrhalis mutants were also analysed by transmission electron microscopy (TEM). The M. catarrhalisΔuspA1 mutant did not display any difference in vitronectin binding, whereas the UspA2-deficient M. catarrhalisΔuspA2 mutant significantly lost its binding to vitronectin (Fig. 1C–F). TEM revealed that UspA2 was present in abundance seen as surface fibrils of approximately 60 nm in length. These results were in complete agreement with the UspA2/vitronectin binding properties that have been reported previously (Attia et al., 2006), and further strengthened that UspA2 is a major vitronectin binding protein of M. catarrhalis.

Vitronectin binds at the head region of UspA2 comprising residues 30–177

The UspA structure can be divided into a distant head, followed by a stalk and membrane spanning regions for anchorage (Brooks et al., 2008a). In electron microscopy, gold-labelled vitronectin selectively bound to the distal N-terminal head region of UspA2 (Fig. 1C and D; inserts). To in detail examine the vitronectin binding region of UspA2, a series of recombinant proteins spanning residues 30–539 was expressed in Escherichia coli (Fig. 2A). The purified truncated recombinant proteins are shown in Fig. 2B and were used in a direct binding assay, where equimolar concentrations of purified UspA2 fragments were coated on microtitre plates followed by detection of vitronectin using anti-vitronectin polyclonal antibodies (pAb). A dose-dependent binding and a saturated binding curve was observed when full-length UspA2 devoid of the signal peptide (UspA230–539) was coated on plates (Fig. 2C). The comparative assay of different UspA2 fragments suggested that UspA230–177, UspA2101–318 and UspA2302–458 harboured the vitronectin binding regions (Fig. 2D). Further truncation of UspA230–177 into UspA230–100 and UspA2100–180 did not show any specific binding to vitronectin, which indicated that the vitronectin binding region was longer and situated within UspA230–177 (Fig. 2D).

Figure 2.

The N-terminal region of UspA2 binds to vitronectin.
A. The multifunctional protein UspA2 binds complement protein 3 (C3) (Nordstrom et al., 2005) fibronectin (Tan et al., 2005), laminin (Tan et al., 2006b) and α1-antichymotrypsin (Manolov et al., 2008). A schematic plan of the different truncated UspA2 fragments expressed in E. coli is also shown.
B. SDS-PAGE showing purified UspA2 fragments. All protein fragments (approximately 5 µg per lane) were subjected to 15% SDS-PAGE and stained with Coomassie blue R250.
C. The direct binding assay showing the UspA2-dependent interaction with vitronectin. UspA230–539 (0.625–10 nM) was coated in microtitre plates, and vitronectin binding was detected using specific antibodies. Curves are hyperbolic fits.
D. The comparative binding capacities of recombinantly expressed truncated UspA2 fragments. In this assay, 0.01 µM of all UspA2 fragments was coated on the plate, and vitronectin binding was compared. Mean values of triplicates are shown and error bars correspond to standard deviations.
E. The specificity of the binding of vitronectin to various UspA2 fragments was analysed by a competition experiment.
F. Concentration-dependent competition of UspA230–539, UspA230–177 and UspA2101–318 in UspA230–539-dependent vitronectin binding.

The binding specificity of the truncated UspA2 fragments to vitronectin was also tested in a competitive binding assay, where UspA230–539 (0.01 µM) was coated on the plate and vitronectin binding in the presence of different concentrations of truncated UspA2 fragments was analysed. UspA230–177 and UspA2101–318 (0.1 µM) blocked the binding of vitronectin to UspA230–539 by 70.3% and 53.3% respectively (Fig. 2E). Furthermore, UspA230–177 and UspA2101–318 reduced the vitronectin binding to UspA230–539 in a dose-dependent manner (Fig. 2F). Our data confirmed that a vitronectin binding site was located within the N-terminal residues 30–177 of UspA2.

To determine the affinity of the interaction between UspA2 and vitronectin, the kinetics of the binding of UspA2 to vitronectin isolated from human plasma (Sigma) was estimated by surface plasmon resonance (Biacore). Vitronectin was immobilized on a CM5 sensor chip and binding of UspA230–539 and UspA230–177 was recorded as response units. The binding kinetics of UspA230–539 to vitronectin at different concentrations is shown in Fig. 3A. The KD for UspA230–539 was 2.34 × 10−8 M (Fig. 3A), whereas the KD for the smallest truncated active vitronectin binding fragment (UspA230–177) was 7.92 × 10−8 M (Fig. 3B). Thus, these results indicated that UspA2 was a high affinity vitronectin binding protein.

Figure 3.

UspA2 is a high affinity vitronectin binding protein. Surface plasmon resonance (Biacore) showing the binding of UspA230–539 (A) and UspA230–177 (B) to vitronectin purified from human plasma (Sigma). Vitronectin was immobilized on a chip and UspA2 binding was analysed at different concentrations. Inserts denote the binding response (RU) of UspA2 at different concentrations ranging from 2 µM to 0.8 nM. Response units obtained for each concentration at equilibrium were plotted against concentrations of injected proteins and KD values calculated using a steady-state affinity model (BiaEvaluation 3.0). All experiments were repeated at least twice.
C. FPLC chromatograms showing gel filtration pattern obtained from UspA2 fragments. The major fraction of full-length UspA230–539 was eluted as trimer, whereas UspA230–177 showed two peaks in gel filtration and was eluted as a trimer and monomer. Importantly, UspA230–100 and UspA2100–180 were eluted as a monomer and trimer respectively. Molecular weights of all peaks were calculated by comparing (Ve/Vo) ratio with standard protein molecular weight markers. The inserts show collected fractions analysed on 12% SDS-PAGE and stained with Coomassie R250. Peaks with trimers are indicated by black arrows and peaks with monomers with white arrows.
D. UspA2 trimers obtained from gel filtration experiments bind vitronectin in ELISA. The analysis was done as described in Fig. 2C.

UspA2 is described as a trimeric autotransporter (Brooks et al., 2008a). To investigate the maximal binding capacity of vitronectin, the state of oligomerization of UspA2 fragments was confirmed by gel filtration. We found that the major part of the UspA230–539 fraction existed as a trimer (250 kDa) and was eluted at an elution volume of 9.9 ml (Ve). In contrast, UspA230–177 came out as two peaks in gel filtration and was eluted as trimeric (Ve = 12.6 ml, 78.1 kDa) and monomeric forms (Ve = 15.6 ml, 26 kDa). Importantly, UspA230–100 (Ve = 18.1 ml) was eluted as a monomer (12 kDa), whereas UspA2100–180 (Ve = 15.0 ml) was eluted as a trimer (46.4 kDa) (Fig. 3C). The fractions collected showed the presence of pure proteins in SDS-PAGE (Fig. 3C), and results obtained with ELISA demonstrated that only trimeric UspA230–539 and UspA230–177 bound vitronectin, whereas monomeric UspA230–177 did not (Fig. 3D). Thus, trimerization was required for an efficient UspA2-dependent vitronectin binding under these in vitro conditions.

The UspA2/ vitronectin interaction is non-ionic and blocked by heparin

To determine where on the vitronectin molecule UspA2 bound, the possible binding region of UspA2 to vitronectin was analysed by a direct binding assay. A fivefold higher molar concentration of PAI-1 did not block the UspA2-dependent vitronectin binding (Fig. 4A). Other purified recombinant proteins, such as the vitronectin binding Protein E (PE) from H. influenzae (Hallstrom et al., 2009) and the non-vitronectin binding MID764–913 (Nordstrom et al., 2005), were included as positive and negative controls respectively. PE at fivefold molar excess blocked 65.0% of the vitronectin binding to UspA230–539. On the other hand, UspA230–539 itself and heparin blocked the vitronectin binding 89.1% and 94.8% respectively (Fig. 4A). A dose-dependent inhibition was seen with heparin, and the binding between UspA2 and vitronectin was completely blocked at a concentration of 4–8 nM heparin (Fig. 4B). Moreover, to reveal whether ionic forces were involved in the interaction, increasing concentrations of NaCl (0.01–1.0 M) were added to the reactions. NaCl did not, however, interfere with the UspA2-dependent vitronectin binding (Fig. 4C). Taken together, these findings confirmed that UspA2 bound to the heparin binding domain(s) of the vitronectin molecule and that the mode of binding was independent of ionic strength.

Figure 4.

UspA2-dependent binding of vitronectin is inhibited by heparin and specific competitors.
A. A direct binding assay showing the effect of competitive ligands on vitronectin binding to UspA2. The y-axis shows the percentage (%) binding to UspA2 in the absence or presence of specific competitor (UspA230–539) and other various competitors as indicated on the x-axis.
B. Heparin was titrated for the inhibitory effect on the UspA2-dependent vitronectin binding and efficiently blocked the binding at low concentrations.
C. The interaction between vitronectin and UspA2 was non-ionic since increasing concentrations of NaCl had no effect on the binding. Bar diagrams show mean values of triplicate data and standard deviations are represented by error bars. Experiments were repeated twice.

The C-terminal region of vitronectin (residues 312–396) binds to UspA2

In order to in detail analyse the UspA2 binding site(s) on the vitronectin molecule, several different truncated vitronectin domains were recombinantly expressed in both mammalian and prokaryotic expression systems (Fig. 5A–C). The His-tagged vitronectin molecules were expressed in human embryo kidney (HEK293T) cells and purified using affinity chromatography. The vitronectin fragments from HEK293T cells were glycosylated and hence migrated as larger proteins than calculated from the aa sequences (Fig. 5B). To check the functional integrity of recombinantly expressed protein, vitronectin80–396 devoid of the SMB domain was also analysed for binding to UspA230–539 and UspA230–177 using surface plasmon resonance (Fig. 5D and E). The binding kinetics revealed that UspA230–539 and UspA230–177 interacted with vitronectin80–396 at KD of 2.84 × 10−8 M and 7.40 × 10−8 M respectively. This was in agreement with the KD values obtained with the commercial vitronectin purified from human plasma (Fig. 3A and B).

Figure 5.

Recombinant vitronectin molecules used in this study.
A. Schematic cartoon of the different truncated vitronectin molecules that were expressed either in a mammalian expression system or in E. coli.
B and C. SDS-PAGE showing vitronectin80–396, vitronectin80–320 and vitronectin80–229 expressed in HEK293T cells and GST–vitronectin80–396, GST–vitronectin80–229, GST–vitronectin230–320 and GST–vitronectin312–396 expressed in E. coli. All protein fragments (approximately 5 µg per lane) were subjected to 15% SDS-PAGE and stained with Coomassie blue R250.
D and E. Surface plasmon resonance (Biacore) showing the binding of UspA230–539 (D) and UspA230–177 (E) to recombinant vitronectin. Vitronectin80–396 was immobilized on a chip and UspA2 binding was analysed at different concentrations. Inserts denote the binding response (RU) of UspA2 at different concentrations ranging from 2 µM to 0.8 nM. Response units obtained for each concentration at equilibrium were plotted against concentrations of injected proteins and KD values were calculated using a steady-state affinity model (BiaEvaluation 3.0). All experiments were repeated at least twice.

The comparative binding of vitronectin fragments to UspA2 was analysed by the direct binding assay. UspA230–539 was coated on microtitre plates, and binding of different vitronectin fragments was analysed by the anti-vitronectin pAb. Interestingly, vitronectin80–396 significantly bound to UspA2, whereas vitronectin80–320 and vitronectin80–229 did not (Fig. 6A). Thus, our results suggested that vitronectin320–396 was involved in the interaction with UspA2.

Figure 6.

UspA2 binds to vitronectin irrespectively of glycosylation.
A. Binding of recombinant eukaryotic vitronectin fragments to UspA230–539. The data shown here are means of triplicates and curves shown are hyperbolic fits.
B. Analysis of vitronectin80–396, vitronectin80–229 and vitronectin80–320 binding to different UspA2 fragments.
C. Analysis of binding of recombinant GST–vitronectin to UspA230–539. The data shown are mean of triplicates and lines are hyperbolic fits.
D. Binding of GST–vitronectin fragments to different truncated UspA2.
E. Binding of vitronectin80–396 to UspA230–539 or UspA230–177 and blocking with different competitor ligands.
F. GST–vitronectin312–396 binding to UspA230–539 or UspA230–177 was blocked by various competitors as indicated in the panel. The data shown in the bar diagrams are means of triplicate data sets and standard deviations are represented as error bars. All experiments were repeated twice.

Similar domains of vitronectin that were expressed in HEK293T cells were also produced in E. coli and purified as GST fusion proteins (Fig. 5A and C). The rationale for this was to determine the effect of post-translational modifications, i.e. glycosylation, on the UspA2-dependent vitronectin binding. A comparison of the prokaryotic and eukaryotic vitronectin molecules regarding their binding capacity to UspA2 was done. Binding of the GST–vitronectin fusion proteins to UspA2 was analysed using a direct binding assay (Fig. 6C). GST–vitronectin80–396 and GST–vitronectin312–396 bound to UspA230–539, whereas GST–vitronectin80–229 and GST–vitronectin230–320 did not interact with UspA2. These comparative datasets (Fig. 6A and C) strongly suggested that post-translational modifications of vitronectin were not involved in the interaction with UspA2.

To exclude that other vitronectin binding regions existed in UspA2, equimolar concentrations (0.01 µM) of UspA230–539, UspA2200–539, UspA230–177 and UspA230–100 were coated on microtitre plates, and eukaryotic vitronectin and prokaryotic GST–vitronectin fragments were analysed for binding. Our results demonstrate that vitronectin80–396 bound to UspA230–539 and UspA230–177. In contrast, the two truncated fragments vitronectin80–229 and vitronectin80–320 failed to bind to any of the UspA2 fragments (Fig. 6B). Similarly, GST–vitronectin80–396 and GST–vitronectin312–396 were capable of binding to UspA230–539 and UspA230–177 (Fig. 6D). Taken together, these results suggested similar UspA2 binding properties regarding vitronectin from bacterial and mammalian sources, and also confirmed that UspA2 residues 30–177 and vitronectin amino acid residues 312–396 were involved in this important interaction.

The specificity of the UspA2/vitronectin interaction was further analysed with a competitive binding assay. UspA2 full-length and UspA230–177 were coated on microtitre plates (0.01 µM) and the effect of truncated UspA2 fragments was analysed for vitronectin binding. The results demonstrated that the interaction between UspA230–177 and vitronectin80–396 was highly specific as 0.1 µM of competitive ligands UspA230–539 and UspA230–177 blocked the binding while the other two recombinant UspA2 did not (Fig. 6E). In another set of experiments, the comparative binding to recombinant GST–vitronectin312–396 produced in E. coli was analysed and completely similar results were observed (Fig. 6F). This comparison further confirmed that the UspA230–177 and vitronectin312–396 interaction was highly specific and that glycosylation or other post-translational modifications of the vitronectin molecule were not involved.

A direct binding assay using whole bacterial cells was also used to confirm the specificity of the vitronectin binding to UspA2. Importantly, [125I]-vitronectin80–396 bound to Moraxella in a dose-dependent manner (Fig. 7A). The effect of addition of different cold ligands (1–5 µM) to the [125I]-vitronectin binding assay was also determined, and the results using 1 µM cold ligands are summarized in Fig. 7B. GST–vitronectin80–396, GST–vitronectin312–396 and vitronectin80–396 significantly inhibited [125I]-vitronectin80–396 binding to the UspA2-expressing M. catarrhalis. On the other hand, UspA230–539, UspA230–177 and finally heparin completely reduced vitronectin binding to M. catarrhalis. (Fig. 7B). These results proved that the interaction between vitronectin312–396 and UspA230–177 was specific at the bacterial surface.

Figure 7.

Vitronectin binding to whole M. catarrhalis bacteria is highly specific.
A. Standard binding curve with [125I]-vitronectin80–396 showing dose-dependent binding to M. catarrhalis RH4.
B. UspA2-dependent [125I]-vitronectin80–396 binding to M. catarrhalis and competition with different ligands. M. catarrhalis (107 bacteria) were subjected to interact with [125I]-vitronectin80–396 (50 000 cpm) and the binding was blocked with the indicated ligands. The data shown are the mean of 5 replicates and the standard deviations are shown as error bars.

M. catarrhalis survival is reduced in UspA2-treated human serum

Vitronectin is known to bind C9 and thus inhibit the insertion of C9 into the bacterial membrane during MAC assembly (Podak, 2009). To determine the serum resistance of our strains, bacterial survival was monitored in 10% NHS. The wild-type M. catarrhalis was completely resistant of 60 min, whereas the UspA2-deficient mutant M. catarrhalis RH4ΔuspA2 was highly susceptible resulting in 60% bacterial killing within 10–12 min (Fig. 8A).

Figure 8.

M. catarrhalis is susceptible to serum pre-incubated with recombinant UspA2.
A. The UspA2-deficient M. catarrhalis RH4 mutant (ΔuspA2) is serum-sensitive as compared with wild-type bacteria.
B. A significantly enhanced killing is seen with NHS (10%) pre-incubated with recombinant UspA230–539 as compared with untreated control serum.
C. NHS pre-incubated with UspA230–177 kills M. catarrhalis more efficiently in comparison to untreated control serum. The data shown are mean of triplicates obtained in three independent experiments. Standard deviations are indicated by error bars. Statistical significance of differences were calculated using Student's t-test; *p ≤ 0.05.

We hypothesized that vitronectin bound to soluble recombinant UspA2 in NHS would result in a higher C9 deposition at the bacterial surface causing an accelerated MAC formation and lysis of M. catarrhalis. Hence, 10% NHS was pre-incubated with recombinant UspA230–539 or UspA230–177 followed by exposure to the M. catarrhalis RH4 wild-type. A dose-dependent decreased bacterial survival was seen when purified UspA230–539 was added to NHS (Fig. 8B). Moreover, a similar pattern was observed when the shorter UspA230–177 was supplemented to the reaction (Fig. 8C). These results were in agreement with the hypothesis that UspA2 is the major outer membrane protein responsible for M. catarrhalis serum resistance (Attia et al., 2006).

Discussion

Vitronectin plays a crucial role in many biological processes including cell migration, adhesion, angiogenesis (Preissner and Seiffert, 1998) and microbial pathogenesis (Leroy-dudal et al., 2004; Attia et al., 2005; Hallstrom et al., 2006; 2009; Bergmann et al., 2009; Zipfel and Skerka, 2009). This intriguing protein has multi-domain structural arrangement, where the N-terminal SMB domain is well characterized for its various functions (Lossner et al., 2009), while the C-terminal HBDs, i.e. HBD-1 (aa 175–219), HBD-2 (aa 175–219) and HBD-3 (aa 348–361) (Liang et al., 1997b; Gibson et al., 1999), are almost unknown for their functional importance except for the heparin binding capacity. Interestingly, vitronectin is highly homologous among different mammalian spp. When various vitronectins are compared, rabbit, mouse and bovine vitronectin has an N-terminal variable region (76–152 amino acids) and a C-terminal variable region (320–396) (Fig. S1). The C-terminal variable region of human vitronectin contains several unique residues. We found that vitronectin312–396 is involved in UspA2-dependent binding, and that these unique residues may contribute to this important binding. Being one of the main regulators of the complement system, vitronectin has a crucial impact on MAC deposition on the membrane of respiratory Gram-negative pathogens that successfully exploit the MAC inhibitory role of vitronectin (Milis et al., 1993; Blom et al., 2009; Hallstrom et al., 2009). On the other hand, Gram-positive bacteria use a vitronectin/integrin interaction for cellular invasion and internalization (Bergmann et al., 2009).

In the present study, we focused on M. catarrhalis UspA2-dependent vitronectin binding and in detail studied the properties of the interaction. The binding constant between UspA2 and vitronectin was measured using Biacore (Figs 3 and 5). The full-length UspA2 had an KD = 2.3 × 10−8 M for vitronectin and the amino acid region UspA2 30–177 had a high affinity binding region with KD = 7.8 × 10−8 M. Furthermore, we found that heparin blocked the vitronectin binding to UspA2, which prompted us to study the heparin binding regions of vitronectin after expression in HEK293T cells and E. coli.

Previously, it has been shown that vitronectin binds to Streptococcus pneumoniae and contributes to bacterial internalization. This interaction can be efficiently blocked in the presence of heparin, indicating that the heparin binding domain of vitronectin is involved in this interaction (Bergmann et al., 2009). Neisseria meningitidis uses protein OpcA that binds to αvβ3 integrin via the serum factor vitronectin, and thus receives help regarding adhesion as well as induces signalling in the host (Virji et al. 1994; 1995). Another pathogenic organism, Pseudomonas aeruginosa, has the potential capacity to bind vitronectin, and vitronectin binding to αvβ5 plays a key role in P. aeruginosa internalization of A549 cells (Leroy-dudal et al., 2004; Leduc et al., 2007). Recently, we reported that H. influenaze binds vitronectin using Hsf and PE, which contributes to resistance against the serum-mediated killing, and that these interactions can be blocked by heparin (Hallstrom et al., 2006; Hallstrom et al., 2009). Another human pathogen, H. ducreyi, binds vitronectin via the trimeric autotransporter DsrA and thus exhibits serum resistance (Leduc et al., 2009). Many other research reports suggest that vitronectin plays potential important roles in the pathogenesis of Yersinia pseudotuberculosis (Gustavsson et al., 2002), S. pneumoniae (Liang et al., 1997a,b) and Candida albicans (Limper and Standing, 1994). Despite such important roles of vitronectin there are no clear reports available on the actual interaction at the protein level and little is known about microbial proteins which bind to vitronectin. The lack of structural data on the heparin binding domains of vitronectin also limits the clear views about the vitronectin–bacterial interactions. In this report, we identified UspA230–177 and vitronectin312–396 as binding regions, which strongly contributed to the interaction between M. catarrhalis and vitronectin. Previously, it has been shown that the N-terminal region of UspA2 confers serum resistance in M. catarrhalis and is approximately 101 residues long (Attia et al., 2005). This finding is supported by our data and confirms that the vitronectin binding region of UspA2 is located within residues 30–177.

Vitronectin binding to the N-terminal region of UspA2 was also observed by TEM. UspA2 protrudes out from the bacterial outer membrane and its N-terminal part is located at a distant point (Fig. 1C and D), where the gold-labelled vitronectin binds. Recently, a bioinformatics-based approach suggested that UspA1, UspA2 and UspA2H have homology with the structurally known Y. enterocolitica YadA, which has three distinguished structural regions, including the head, stalk and membrane spanning regions (Brooks et al., 2008a). That study also predicted that the N-terminal region of UspA2 spans at the head region. On the basis of this finding we present a schematic model of UspA2 with vitronectin and laminin binding sites at the head region along with other possible protein binding sites including fibronectin, α1-antichymotrypsin, and C3 at the stalk (Nordstrom et al., 2005; Tan et al., 2005; 2006b; Manolov et al., 2008) (Fig. 9). Importantly, the N-terminal region of UspA2 is variable and can be divided into two different groups, N-terminal (NTER) 2A and NTER-2B (Brooks et al., 2008a). NTER-2A sequences are present in 66% isolates and NTER-2B in 33% isolates (Fig. S2). The full UspA2 sequence used for the manufacture of different recombinant proteins included in this study is shown in Fig. S3. Vitronectin bound to the head region of UspA2 amino acids 30–177 and the sequence used belongs to NTER-2B (Fig. S2; B5L5X3).

Figure 9.

Schematic representation of vitronectin and UspA2.
A. Human vitronectin is a multifunctional glycoprotein and has a UspA2 binding region on one of the heparin binding domains. The N-terminal SMB domain (comprising 20–63 aa) binds to uPAR and is inhibited by PAI-1. An RGD binding motif located after the SMB domain interacts with cellular integrin receptors. Vitronectin has three heparin binding domains, located between residues 82–137 (HBD-1), residues 175–219 (HBD-2) and finally residues 360-385 (HBD-3). The haemopexin-like domains are predicted as putative haem binding motifs (http://pfam.sanger.ac.uk/) with unknown function.
B. Schematic model of UspA2 showing functional binding regions. The C-terminal has a membrane spanning region embedded in the outer membrane of M. catarrhalis, followed by a stalk region and a terminal head, similar to YadA of Y. enterocolitica. The head region has a vitronectin binding domain (this study) and laminin binding specificity. The C3 binding region is situated at the stalk region, whereas fibronectin and α1-antichymotrypsin binding regions are present at the upper stalk region, or partially at the head region.

Lysis by the MAC is the final event in the complement cascade, and is preceded by C5b–C8 complex formation. The C5b–C8 complex is recognized by C9 and when MAC polymerization is initiated it results in barrel shape pores at the bacterial surface (Podak, 2009). Vitronectin binds to the C5b–7 complex and C9 and thus inhibits MAC formation (Milis et al., 1993). The C9 blocking property of vitronectin is successfully exploited by many bacterial pathogens (Gustavsson et al., 2002; Hallstrom et al., 2009; Leduc et al., 2009). The mode of C9 binding to vitronectin is not known in detail, although heparin blocks the interaction (Milis et al., 1993). We recently showed that H. influenzae PE binds to vitronectin and contributes to serum resistance. Importantly, the PE/vitronectin complex was still active at the bacterial surface and effectively bound C9 and thus blocked MAC formation (Hallstrom et al., 2009). It has previously been shown that vitronectin-depleted serum restored the bactericidal activity against M. catarrhalis, whereas addition of vitronectin to serum regained the serum resistance (Attia et al., 2005). In support of this finding, addition of UspA2 to serum in our experiments indicated quenching of vitronectin by UspA2, and hence a high MAC activity might be responsible for bacterial killing (Fig. 8B–C).

In conclusion, this study suggests that the multifunctional protein UspA2 is highly important for vitronectin binding and Moraxella serum resistance, besides having crucial roles in adhesion and interaction with proteins like fibronectin, laminin, C4BP as well as C3. The glycoprotein vitronectin harbours the UspA2 binding region between 312–396 residues, while 30–177 residues of UspA2 are the major vitronectin binding region involved in this important protein–protein interaction. Finally, UspA2 has a surface fibril-like structure with the N-terminal at the head region that is completely accessible for interaction with vitronectin.

Experimental procedures

Bacterial strains and culture conditions

Moraxella catarrhalis RH4 and mutants (Mollenkvist et al., 2003; Nordstrom et al., 2005) were grown on chocolate agar plates or in brain heart infusion broth (BHI) (Difco/Becton Dickinson, Sparks, MD, USA) at 37°C in a humid atmosphere containing 5% CO2 (Tan et al., 2005; Tan et al., 2006b). The M. catarrhalis mutants RH4ΔuspA1, RH4ΔuspA2, RH4ΔuspA1/A2 and RH4ΔuspA1/A2/mid were grown in the presence of appropriate selection antibiotics, i.e. 17 µg ml−1 kanamycin, 1.5 µg ml−1 chloramphenicol (Sigma, Saint Louis, MO, USA) and 7 µg ml−1 zeocin (Invitrogen Life Technologies, Carlsbad, CA, USA). E. coli BL21 and DH5α were cultured in Luria–Bertani (LB) broth or on LB agar plates at 37°C in a humid atmosphere containing 5% CO2 (Tan et al., 2005; Tan et al., 2006b). E. coli strains harbouring the expression vector (pET26b) coding for recombinant UspA2 or other proteins were grown in LB medium supplemented with 50 µg ml−1 kanamycin.

Recombinant protein expression and purification

The E. coli strains expressing recombinant uspA2 fragments were as described (Tan et al., 2005). In addition, two more constructs were manufactured. The uspA2 gene was amplified from the regions comprising residues UspA2 30–100 and UspA2 100–180. The amplified and restriction enzyme digested fragments were cloned into pET26b. Finally, E. coli BL21 (DE3) was transformed with the resulting plasmids.

Escherichia coli BL21 (DE3) harbouring recombinant pET26b vectors were grown in 500 ml LB medium with kanamycin at 37°C until an OD600 reached 0.8–1. Expression was induced by 1 mM IPTG and cultivation continued for another 3 h at 37°C. The cells were harvested at 5000 r.p.m. for 15 min at 4°C and resuspended in His-tag binding buffer 50 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 50 mM imidazole. Cellular lysis was performed by sonication and the supernatant was loaded on Ni-NTA column (Histrap FF Crude, GE Healthcare Biosciences), followed by washing step and elution with His-tag elution buffer 50 mM Tris-HCl, pH 7.5, containing 500 mM NaCl and 200 mM imidazole. The periplasmic fraction was separated according to a protocol described by Singh and Rohm (2008).

Some recombinant proteins enriched in inclusion bodies and were therefore purified by urea solubilization followed by refolding (Tsumoto et al., 2003). The bacterial cells (1 l of expressed culture) were lysed by sonication and centrifuged at 10 000 r.p.m. for 20 min at 4°C. Pellets were dissolved into 10 ml of 8 M urea solution for 3 h at 4°C. The refolding was performed by dialysis against His-tag protein binding buffer. The dialysed protein solutions were cleared by centrifugation and supernatants were loaded on Ni-NTA columns and purified as described previously. For Biacore binding analysis, purified proteins were dialysed by several exchanges of HNET buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 3 mM EDTA and 0.005% Tween-20), and concentrated by using Centricon cartridge (3 kDa molecular weight cut-off; Millipore, Bedford, MA, USA).

Expression of vitronectin in HEK293T cells and E. coli

cDNA encoding vitronectin was amplified by PCR from a human liver cDNA library (Stratagene, La Jolla, CA, USA), and the vitronectin domain excluding the N-terminal signal sequence and SMB domain were expressed in HEK293T cells and in E. coli BL21(DE3). Vitronectin domains harbouring heparin binding sites were expressed in HEK293T cell lines as vitronectin residues 80–396, 80–320 and 80–229 with C-terminal 6x-His tags (Fig. 5A and B). The forward primer was inserted with BshT1 (AgeI) and the reverse primer with Acc65I (KpnI) restriction enzyme sites. Amplified and digested inserts were ligated into the pHLsec vector (Aricescu et al., 2006). The vector pHLsec was a modified vector with a pLEX backbone (Open Biosystems, Huntsville, AL, USA) containing an N-terminal Kozak sequence and signal sequence for secretion of expressed recombinant proteins to the culture medium. The C-terminus of vector pHLsec contains sequences encoding a 6x-His tag for purification of the recombinantly expressed proteins. The sequenced vectors were transformed into HEK293T cells using polyethyleneimine (PEI) ‘Max’ (Polysciences, Warrington, PA, USA) as described elsewhere (Durocher et al., 2002). In brief, HEK293T cells were grown in three triple flasks (Nunc) to > 80% confluency using advanced Dulbecco's modified Eagle's medium (DMEM) (Gibco) medium supplemented with 1 µg ml−1 gentamicin and 1% FCS in 37°C with 5% CO2. For transfection, 600 µg of plasmid DNA was mixed with 1.6 mg of PEI in 1 ml ddH2O and incubated for 10 min at RT. The DNA–PEI complex mixture was added in 500 ml of advanced DMEM and was used for replacement of the culture medium in confluent cell cultures. The cells were incubated for 3 days at 37°C with 5% CO2 followed by harvest of the supernatant. Similar volume of advanced DMEM was once again added to the cells and the procedure was repeated after 3 days. His-tagged vitronectin was secreted into the medium that was purified by Ni-NTA chromatography. Medium was diluted twice with addition of 50 mM Tris-HCl, pH 7.5, 500 mM NaCl and 30 mM imidazole. Five millilitres of Ni-NTA resin (Novagen) was mixed with culture and purification was done as recommended by Novagen.

The vitronectin domains were also expressed as N-terminal GST-tagged fusion proteins in E. coli BL21 (DE3) using pET41b (Novagen, Darmstadt, Germany). Vitronectin residues 80–396, 80–229, 230–320, 312–396 were cloned from the human liver cDNA library by using specific primers. A BamHI restriction site was included in the forward primer, whereas a HindIII site was inserted into the reverse primer. Thereafter, amplified and digested inserts were cloned into pET41b. The sequenced vectors were finally transformed into E. coli BL21 (DE3) for expression. The GST-tagged recombinant proteins were also expressed similarly as described for the His-tagged proteins. The induced E. coli were resuspended into GST-tagged binding buffer with 10 mM Na2HPO4, 1.8 mM KH2PO4, pH 7.2, containing 140 mM NaCl and 2.7 mM KCl. The cells were lysed by sonication and centrifuged at 10 000 r.p.m. for 20 min at 4°C. Supernatants were loaded on a GSTrap FF column (GE Healthcare Biosciences), and after 10 volumes of column wash with binding buffer, GST-tagged proteins were eluted with the GST-tagged elution buffer 50 mM Tris-HCl, pH 7.5, containing 10 mM reduced glutathione. Proteins were dialysed against appropriate buffers before experiments and concentrations were measured by UV absorbance (280 nm) using a Nano-drop spectrophotometer (Thermo Scientific, Wilmington, DE). The concentrations were also verified by using a BCA protein estimation method (Pierce, Rockford, IL, USA). The purity of expressed proteins was judged on 15% SDS-PAGE stained with Coomassie blue R250.

Gel filtration

The gel filtration experiments were performed by using Superdex 200™ column (GE Healthcare Biosciences) attached to AKTAprime plus (GE Healthcare Biosciences). The column was equilibrated in 50 mM Tris-HCl, pH 7.5, containing 100 mM NaCl. Approximately 0.5–1.0 mg of UspA2 fragments (purified by Ni-NTA columns) was injected and the separation was performed at a 0.5 ml min−1 flow rate of buffer. The fractions were collected and followed by measurement of protein concentration as well as SDS-PAGE analysis for estimation of the purity. The elution volume/void volume ratio (Ve/Vo) of all peaks were compared with standard protein molecular weight markers (Sigma) to calculate the molecular weight.

[125I]-vitronectin binding to M. catarrhalis and isogenic mutants

The labelling of vitronectin from human plasma (Sigma) or recombinantly expressed vitronectin was performed by the Chloramine-T method. For labelling, 0.05 M of [125I] (Amersham Biosciences, Buckinghamshire, UK) was used per mole of protein. The labelled proteins were separated from unincorporated [125I] using PD10 columns (GE Healthcare Biosciences, Uppsala, Sweden) and were used within 2 weeks.

To evaluate the binding of vitronectin to M. catarrhalis and isogenic mutants, a direct binding assay using labelled [125I]-vitronectin was performed. Bacteria were grown overnight on solid chocolate agar or BHI agar with appropriate antibiotics. A new inoculation was made in BHI and grown until optical density (OD) at 600 nm reached 0.8. Moraxella were harvested and washed with PBS and resuspended in PBS containing 2.5% BSA. Approximately, 107 bacteria per well were dispensed in a microtitre plate and incubated with different concentrations of [125I]-labelled vitronectin at 37°C for 1 h. In blocking experiments, different protein ligands (0.01–10 µM) were added together with [125I]-labelled vitronectin. Subsequently, three washing steps with PBS containing 2.5% BSA were performed and bacteria were harvested in 96 square glass fibre filtermats (Wallac) by using a microtitre plate harvester (Tomtec) followed by measurement of [125I]-vitronectin bound to bacteria in a gamma counter (Trilux, Wallac).

Direct binding assay

Purified UspA2 fragments or other proteins were coated on 96 wells PolySorb microtitre plates (Nunc-Immuno, Roskilde, Denmark) (Hallstrom et al., 2009). Coating of UspA2 (0.01 µM) or equimolar amount of other proteins was performed in 100 µl of 100 mM Tris-HCl, pH 9.0 for 15 h at 4°C. Plates were washed 3 times with PBS to remove excess unbound protein and blocked with PBS containing 2.5% BSA for 1 h at 25°C. The vitronectin (Sigma) or expressed vitronectin fragments was added to the wells in PBS, 2.5% BSA and allowed to bind for 1 h at 25°C. In a competition assay, increasing amounts of competitor ligands were added to the binding ligands and allowed to bind for 1 h at RT. The unbound vitronectin fraction was removed by washing with PBS containing 0.005% Tween-20 and bound vitronectin was detected by polyclonal sheep anti-human vitronectin antibodies (AbD Serotec, Kidlington, Oxford) and HRP-conjugated donkey anti-sheep secondary antibodies (AbD Serotec). The binding of GST–vitronectin fragments were analysed by anti-GST antibodies conjugated with HRP (GE healthcare Biosciences). Plates were developed by addition of 100 µl HRP substrate consisting of 20 mM tetramethylbenzidine and 0.1 M potassium citrate. The reactions were terminated by 50 µl of 1 M H2SO4, and finally plates were read at 450 nm in a microplate reader.

Serum resistance assay

Normal human serum was pooled from five healthy volunteers and stored at −80°C as aliquots. M. catarrhalis RH4 and RH4ΔuspA2 were grown until mid log phase (A600 = 1.0), washed once in PBS and resuspended into DGVB++ buffer (2.5 mM Veronal buffer, pH 7.3), containing 1 mM MgCl2, 0.15 mM CaCl2, 2.5% glucose and 0.1% (w/v) gelatin. Bacterial cells (7 × 104) were incubated in 100 µl DGVB++ containing 10% NHS at 37°C. At different time points, 10 µl of solution was plated out in chocolate agar plates. Colony-forming units were counted after 18 h incubation of plates at 37°C. The effect of UspA2 on serum resistance of M. catarrhalis was evaluated by preincubation of serum with different concentrations of UspA2 fragments at 37°C for 30 min.

Surface plasmon resonance (Biacore)

The interaction between vitronectin and UspA2 fragments was analysed using surface plasmon resonance (Biacore 2000, Uppsala, Sweden). Three flow cells of a CM5 sensor chip were activated, each with 20 µl of a mixture of 0.2 M 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide and 0.05 M N-hydroxy-sulphosuccinimide at a flow rate of 10 µl min−1, after which vitronectin (Sigma; 10 µl ml−1 in 10 mM sodium acetate buffer, pH 4.0) was injected over flow cell 2, while recombinant vitronectin was injected under the same conditions over flow cell 3, both to reach 4000 resonance units (RU). Unreacted groups were blocked with 20 µl of 1 M ethanolamine (pH 8.5). A negative control was prepared by activating and subsequently blocking the surface of flow cell 1. The comparative association kinetics was studied for various concentrations of the UspA2 fragments. The flow buffer (50 mM HEPES, pH 7.5 containing 150 mM NaCl, 3 mM EDTA and 0.005% Tween-20) was used for binding. Protein solutions (160 µl of serial twofold dilutions ranging from 2 µM to 0.8 nM) were injected during the association phase at a constant flow rate of 5 µl min−1. The sample was first injected over the negative control surface and then over immobilized vitronectin. The signal from the control surface was subtracted. The dissociation was followed for 400 s at the same flow rate. In all experiments, two consecutive injections of 12 µl of 2 M NaCl were used to remove bound ligands during a regeneration step. BiaEvaluation 3.0 software (Biacore) was used for data analysis and Graph pad for generating final figures. The equilibrium affinity constants (KD) were calculated using signals obtained for each concentration of injected protein at equilibrium employing steady-state affinity model (BiaEvaluation 3.0).

Transmission electron microscopy

The wild-type M. catarrhalis and mutants were grown in BHI for 3 h at 37°C. Bacterial solutions were incubated with gold-labelled vitronectin (100 µg ml−1), fixed in PBS containing 4% paraformaldehyde and 0.1% glutaraldehyde and prepared for electron microscopy as described (Carlemalm, 1990). TEM was thereafter performed as described (Bengtson et al., 2008), and specimens were examined in a JEOL JEM 1230 transmission electron microscope (JEOL, Peabody, MA, USA) at 60 kV accelerating voltage. Images were recorded with a Gatan Multiscan 791 CCD camera (Gatan, Pleasanton, CA, USA).

Acknowledgements

This work was supported by grants from the Alfred Österlund, the Anna and Edwin Berger, the Marianne and Marcus Wallenberg, Knut and Alice Wallenberg, Krapperup, Inga-Britt and Arne Lundberg, the Söderberg, and the Greta and Johan Kock Foundations, the Swedish Medical Research Council, the Swedish Foundation for Strategic Research, the Cancer Foundation at the University Hospital in Malmö, and Skane county council's research and development foundation.

Ancillary