Non-typeable Haemophilus influenzae adhere to and invade human bronchial epithelial cells via an interaction of lipooligosaccharide with the PAF receptor

Authors


Abstract

Adherence and invasion are thought to be key events in the pathogenesis of non-typeable Haemophilus influenzae (NTHi). The role of NTHi lipooligosaccharide (LOS) in adherence was examined using an LOS-coated polystyrene bead adherence assay. Beads coated with NTHi 2019 LOS adhered significantly more to 16HBE14 human bronchial epithelial cells than beads coated with truncated LOS isolated from an NTHi 2019 pgmB::ermr mutant (P = 0.037). Adherence was inhibited by preincubation of cell monolayers with NTHi 2019 LOS (P = 0.0009), but not by preincubation with NTHi 2019 pgmB::ermr LOS. Competitive inhibition studies with a panel of compounds containing structures found within NTHi LOS suggested that a phosphorylcholine (ChoP) moiety was involved in adherence. Further experiments revealed that mutations affecting the oligosaccharide region of LOS or the incorporation of ChoP therein caused significant decreases in the adherence to and invasion of bronchial cells by NTHi 2019 (P < 0.01). Analysis of infected monolayers by confocal microscopy showed that ChoP+ NTHi bacilli co-localized with the PAF receptor. Pretreatment of bronchial cells with a PAF receptor antagonist inhibited invasion by NTHi 2109 and two other NTHi strains expressing ChoP+ LOS glycoforms exhibiting high reactivity with an anti-ChoP antibody on colony immunoblots. These data suggest that a particular subset of ChoP+ LOS glycoforms could mediate NTHi invasion of bronchial cells by means of interaction with the PAF receptor.

Introduction

Haemophilus influenzae is a nutritionally fastidious, Gram-negative coccobacillus that is a commensal inhabitant of the human nasopharynx and that may be isolated from most of the human population (Murphy and Apicella, 1987). Non-typeable H. influenzae (NTHi) strains cause a variety of infections, including otitis media, sinusitis, bronchitis and conjunctivitis (Murphy and Apicella, 1987; Foxwell et al., 1998; Rao et al., 1999). The lipooligosaccharide (LOS) is a major antigenic component of the NTHi cell surface. The H. influenzae LOS lacks an O-antigen and features a variety of non-repeating oligosaccharides consisting of glucose, galactose, N-acetylglucosamine, phosphorylcholine (ChoP) or N-acetyl-neuraminic acid in a number of combinations (Masoud et al., 1997; Risberg et al., 1999). The composition and expression of the oligosaccharide region is controlled in part by phase variation via slipped-strand base pairing among tandem repeat sequences within the LOS biosynthetic genes (Weiser et al., 1989, 1990a). Immunochemical analyses have shown that NTHi LOS glycoforms mimic a number of human glycolipid antigens, including Pk, N-acetyl-lactosamine (Galβ1→4GlcNAc), i and paragloboside (Campagnari et al., 1987, 1990; Krivan et al., 1988; Virji et al., 1990; Mandrell et al., 1992; Phillips et al., 1992; Moran et al., 1996; Preston et al., 1996). In vivo carriage within both rat models and human patients appears to enrich NTHi isolates expressing particular LOS glycoforms (Weiser et al., 1990b; 1998a; Weiser, 1993).

The focus of this study was to better characterize the role of particular LOS structures in adherence to and invasion of human bronchial epithelial cells by NTHi. An insertional null mutation in the pgmB gene was generated and was shown to completely abolish phosphoglucomutase activity, resulting in the production of truncated LOS molecules lacking oligosaccharide chains external to the core region (Hep3-Kdo-lipid A). This mutant was used in experiments to test the role of the oligosaccharide region of LOS in bacterial adherence to human bronchial epithelial cells. Subsequent experiments were focused upon the identification of specific LOS structures involved in bacterial adherence and invasion. The results suggest that a particular subset of ChoP+ LOS glycoforms may be of particular importance in the adherence of NTHi to bronchial epithelial cells, and may also contribute to invasion.

Results

Construction and confirmation of a pgmB null mutant

Three putative phosphomutase genes have been identified based on sequence homology through analysis of the H. influenzae Rd genomic sequence (Fleischmann et al., 1995; Hood et al., 1996). The pgmB gene was amplified from NTHi 2019 genomic DNA and the amplified fragment was cloned and sequenced. An insertional pgmB::ermr null mutant allele was constructed and used to transform NTHi strain 2019. The pgmB mutant was confirmed by Southern blotting, examination of phosphoglucomutase enzyme activity and SDS–PAGE analysis of purified LOS (data not shown).

Previous studies have shown that the LOS from NTHi 2019 contains a conserved tri-heptose core substituted with two phosphoethanolamines (PEA) that is linked to a diphosphoryl lipid A moiety through a single phosphorylated Kdo (KdoP) (Phillips et al., 1992). As determined by NMR and mass spectrometry (MS), this core is further substituted with a variable oligosaccharide branch that emanates from the first core heptose and the region in which the most abundant glycoforms contain a lactose oligosaccharide branch, i.e. Galβ1→4Glcβ1→4Hep3(PEA)2−3KdoP-Lipid A. As shown in Fig. 1A, our analysis of the LOS of NTHi 2019 by matrix-assisted laser desorption/ionization (MALDI) mass spectrometry identified a set of LOS glycoforms with (M-H) values ranging from 2236 to 2723 Da. ESI-MS analysis yielded comparable results (data not shown, see legend for Fig. 1). These molecular ions are all consistent with O-deacylated LOS glycoforms containing either the major lactose-bearing structure (Mr = 2400 and 2523) (Phillips et al., 1992) or related glycoforms containing either one less (Mr = 2237) or one to two more hexoses (Mr = 2562 and 2724) that are partially substituted with a third PEA moiety (Mr = 2360, 2685 and 2847) (Gibson et al., 1993). Monosaccharide composition analysis also showed the expected presence of heptose, glucose, galactose and glucosamine (from lipid A).

Figure 1.

Negative ion MALDI-MS spectra of O-deacylated LOS preparations from (A) NTHi 2019 and (B) NTHi 2019 pgmB::ermr strains. In the MALDI-MS spectrum of NTHi 2019, two abundant molecular ions (M-H), at m/z 2399 and 2522 are present that correspond to the lactose-containing LOS glycoforms (Mr = 2400 and 2423, Hex2Hep3KdoP(PEA)2−3LA*), in addition to lower abundance LOS species with one less (m/z 2237 and 2360) or 1–2 more hexose residues (m/z 2561, 2684, 2723 and 2846). In the NTHi 2019 pgmB::ermr mutant strain, the base peak at m/z 2074 corresponds to the (M-H) ion for the glycoform without an oligosaccharide branch (Mrcalc. = 2075.8, Hep3KdoP(PEA)2LA*). LOS with a lower mass (m/z 1952) and higher mass values (m/z 2198 and 2321), contain either one less PEA or 1–2 more PEA moieties, respectively. Prompt fragmentation of the LOS species at the labile KdoP-lipid A linkage yields peaks corresponding to lipid A (m/z 952) and the oligosaccharide portion (m/z 1122, 1245 and 1368). These and other fragments show Lipid A to be present in the expected diphosphoryl form, and the PEAs to be substituted on the oligosaccharide portion of the LOS. ESI data for the O-deacylated LOS preparations from both the NTHi 2019 and NTHi 2019 pgmB::ermr strains were consistent with the MALDI analysis, yielding doubly (M-2H)2– and triply deprotonated ions (M-3H)3–, with masses (and relative percentage abundances) as follows: wild type NTHi 2019–2114 (3%) 2237.7 (30%), 2359.8 (38%) for HexHep3 Kdo(P)PEA1−3 LA*, 2275.8 (6%) 2399.2 (52%), 2522.5 (100%) for Hex2Hep3Kdo(P)PEA1−3LA*, 2437.8 (32%), 2561.3 (65%), 2686.5 (12%) for Hex3Hep3Kdo(P)PEA1−3LA*, and 2599.1 (20%) 2723.4 (31%) and 2847.3 (7%) for Hex4Hep3Kdo(P)PEA1−3LA*; NTHi 2019 pgmB::ermr– 1951.8 (30%), 2074.8 (60%), 2198.0 (100%) and 2320.8 (30%) for Hep3Kdo(P)PEA1−4LA*. LA* refers to the diphosphoryl di-N-β-hydroxymyristoyl Lipid A formed after O-deacylation. Masses listed with LOS structure insets are calculated. All other masses are experimental.

In contrast, the MALDI MS spectrum obtained from the NTHi 2019 pgmB::ermr strain revealed a set of much smaller LOS that completely lack an oligosaccharide branch external to the core heptoses (see Fig. 1B). The average molecular weights as determined by MALDI MS (Mrexpt. = 1952, 2075, 2198, and 2321) show that the pgmB::ermr mutant strain produces a mixture of LOS glycoforms, consisting of the conserved Hep3-KdoP-lipid A structure containing 1–4 PEA moieties. This result is consistent with our expectations, as a lack of phosphoglucomutase activity should stop branch extension from the heptose core with glucose and/or galactose sugars. Monosaccharide analysis also supports this assignment, as only heptose and glucosamine (from lipid A) were observed. The amount of PEA substitution in the LOS of the mutant strain (1–4 PEAs) is slightly higher than that observed in the parental strain (1–3 PEAs). The reason behind this is not known, but it is possible that an increase in the overall PEA content may compensate for a lack of branch sugars, perhaps modulating membrane stability.

Adherence of LOS-coated beads

A coated microbead model system has been previously employed to examine the contribution of bacterial surface factors to adherence (Schlesinger et al., 1996). The advantage of this model system for our studies was that it provided a quantitative means to examine the role of LOS in NTHi adherence to host epithelial cells in the absence of any other bacterial factors. For each observation, the number of adherent beads was counted in 10 independent fields of view using scanning electron microscopy, and the mean number of adherent beads was calculated. The number of adherent beads coated with NTHi 2019 LOS was significantly higher than the number of adherent beads coated with gelatin or with truncated LOS isolated from the NTHi 2019 pgmB::ermr mutant (P < 0.05), as shown in Fig. 2. It may be noteworthy that the beads coated with the wild-type LOS appeared to adhere preferentially to specific cells within the human bronchial epithelial (HBE) monolayer (data not shown). The 16HBE14 cell line was originally derived by transformation of airway cells obtained from bronchial brushings, and it thus contains a diverse population of cells (Gruenert et al., 1988). Previous data from our laboratory have shown that NTHi bacilli also adhere to a subset of the bronchial cells within a monolayer (Ketterer et al., 1999).

Figure 2.

Adherence of LOS-coated polystyrene beads to bronchial cells.Beads were coated with LOS isolated from NTHi 2019 (black bars), NTHi 2019 pgmB::ermr (hatched bars), or gelatin (white bars) as described in the experimental methods and in the amounts indicated. The coated beads were added to confluent monolayers on cover slips as described in the experimental methods. The numbers of attached beads were counted by SEM analysis, and the values reported represent means of the counts obtained from 10 independent fields of view.

Competitive inhibition of the binding of LOS-coated beads

Competitive inhibition experiments were performed to examine the contribution of specific structures found in LOS glycoforms to the adherence of the LOS-coated beads (Table 1). Preincubation of the monolayers with NTHi 2019 LOS, which contains the full complement of LOS glycoforms expressed by NTHi, resulted in significant, dose-dependent inhibition of the adherence of beads coated with NTHi 2019 LOS (P >F = 0.0009). In contrast, LOS isolated from the NTHi 2019 pgmB::ermr mutant lacking oligosaccharides external to the core region had no significant effect on the adherence of the beads (P >F = 0.4071). Similarly, no inhibition was observed in control experiments using gelatin (data not shown). These data suggest that the oligosaccharide portion of the LOS molecule contributes to the adherence of LOS-coated beads.

Table 1. . Competitive inhibition of NTHi LOS-coated bead adherence.
InhibitorDescription P > FaIC50bDFc
  • a

    . P >F-values reflect the significance of the effect of an inhibitor over a range of concentrations and were determined by the comparison of linear regression plots as described in the statistical methods. P >F-values of < 0.05 were considered significant.

  • b . IC 50 values represent the projected amount of the inhibitor that would be predicted to inhibit 50% of the observed bead binding, and were calculated using linear regression analysis.

  • c

    . Degrees of freedom, indicating the amount of replication of the inhibition experiment.

  • d

    . Inhibition relative to concentration of these compounds is non-linear and thus the projection of a meaningful 50% inhibitory concentration was not possible.

NTHi 2019 LOSVariable oligosaccharide0.0009 1.5 µg14
NTHi 2019 pgmB::ermr LOSCore heptose with single GlcNAc0.4071NLd11
Neisseria gonorrheae 1291B LOSGalGalGlc (Pk)0.5796NL16
N-acetyl-lactosamineGalβ1→4GlcNAc0.021033.9 µg9
LactoneotetroseGalβ1→4GlcNAcβ1→3Galβ1→4Glc0.2607NL9
Galβ1→4GalStructure present in NTHi LOS0.9367NL9
Galα1→3GalStructure present in NTHi LOS0.1729NL9
Galα1→3Galβ1→4GlcNAcStructure present in NTHi LOS0.1691NL9
H. influenzae H394 LPSChoP+ phase variant0.0031 1.4 µg14
H. influenzae H395 LPSChoP phase variant0.1176NL14
Pneumococcal C polysaccharideChoP-containing cell wall teichoic acid0.0133 2.2 µg10
PhosphorylcholinePhase-variable LOS component0.2630NL14
CholineChoP structural analog0.8753NL14
EthanolamineChoP structural analog0.8238NL14

Epitopes corresponding to N-acetyl-lactosamine, lactoneotetrose and the Pk antigen are present within the LOS of most NTHi strains (Mandrell et al., 1992). Similarly, structural analyses have also revealed that Galβ1→4Gal, Galα1→3Gal and Galα1→3Galβ1→4GlcNAc moieties are commonly present within NTHi LOS glycoforms (Virji et al., 1990; Phillips et al., 1992). Glycoconjugates corresponding to these structures, as well as the LOS from a mutant of Neisseria gonorrhoeae (1291B) that exclusively produces one LOS glycoform containing a major Pk epitope (John et al., 1991), were tested for the ability to competitively inhibit the adherence of beads coated with NTHi 2019 LOS. As shown in Table 1, the results showed that, among these structures, only N-acetyl-lactosamine significantly inhibited the adherence of the beads in a dose-dependent fashion (P >F = 0.0210). The calculated IC50 value for the inhibition by N-acetyl-lactosamine (33.9 µg) was over 20 times higher than that of NTHi 2019 LOS.

Previous data have suggested that ChoP may play a role in NTHi colonization and persistence within the airways (Weiser and Pan, 1998, Weiser et al., 1998a). Therefore, experiments were performed to test the contribution of ChoP to the adherence of NTHi 2019 LOS-coated beads (Table 1). LOS from a ChoP+H. influenzae phase variant (H394) significantly inhibited the binding of LOS-coated beads (P >F = 0.0031). Significant inhibition was also observed in experiments using the C-polysaccharide of Streptococcus pneumoniae (P >F = 0.0133), which consists of ChoP and cell wall teichoic acid (Fischer et al., 1993). Moreover, the calculated IC50 values for C-polysaccharide, NTHi 2019 LOS and the ChoP+ LOS were very similar (1.5, 1.4 and 2.2 µg respectively). In contrast, LOS isolated from a ChoPH. influenzae phase variant (H395) did not significantly affect the binding of the LOS-coated beads (P >F = 0.1176). No inhibition was detected for soluble ChoP or the structurally related compounds ethanolamine and choline.

Examination of ChoP expression by colony immunoblot

The ChoP expression of NTHi was examined by colony immunoblot analysis using the monoclonal antibody (mAb) TEPC-15, which is highly specific for the ChoP epitope (Weiser et al., 1997). Positive control experiments using S. pneumoniae showed a high degree of reactivity with the antibody in all colonies examined (data not shown). NTHi colonies were observed to exhibit high reactivity, low reactivity or non-reactivity with TEPC-15 (Fig. 3), as has been reported previously (Weiser et al., 1997). The expression of the different ChoP-reactive colony types was variable between strains (Table 2). Of particular interest are the decrease in high-reactivity ChoP+ colonies in the NTHi ChoP phase variant and the complete absence of high-reactivity ChoP+ colonies in the clinical isolate NTHi 7502. However, immunoblots using NTHi 3198 colonies showed that all ChoP+ colonies from this strain had high reactivity (Fig. 3 and Table 2). ChoP+ colony sectors with differential reactivity were a common observation in all strains (Fig. 3). Immunoblot analysis of NTHi 2019 pgmB::ermr and a NTHi 2019 licD::kanr null mutant revealed no ChoP+ colonies (data not shown).

Figure 3.

Visualization of ChoP epitope by immunoblotting NTHi colonies reveals different levels of reactivity. The panels show representative quadrants of TEPC-15 colony immunoblots on NTHi strains as indicated.

Table 2. . ChoP expression by NTHi strains.
Bacterial strainHigh reactivityaLow reactivityNon reactivityTotal colonies
  1. a . ChoP expression was detected by colony immunoblot analysis with the anti-ChoP Mab TEPC-15 as described in the methods. Strains were classified by reactivity as previously described ( Weiser et al., 1997).

  2. b . Phase variant of NTHi 2019 enriched for high-reactivity (ChoP +) or low-reactivity and non-reactivity (ChoP) on colony immunoblots; retains the genetic capacity to alter ChoP expression.
    c. Counts reflect the ChoP reactivity pattern exhibited within the majority of the colony. Discrete sectors of different reactivity were observed (see Fig. 3).

NTHi 201996 (9%)826 (77%)152 (14%)1074
NTHi 2019 ChoP+b401 (52%)259 (33%)116 (15%)776
NTHi 2019 ChoPb10 (2%)171 (28%)427 (70%)608
NTHi 1479319 (26%)61 (5%)848 (69%)1228
NTHi 3198880 (78%)0 c249 (22%)1129
NTHi 75020c931 (81%)213 (18%)1144

Adherence and invasion of bronchial epithelial cells

The role of LOS in NTHi adherence and invasion was examined using an immortalized human bronchial epithelial cell line (16HBE14) in standard gentamycin-survival invasion experiments (Isberg and Falkow, 1985). As shown in Table 2, the NTHi 2019 pgmB::ermr mutant was significantly less adherent (P = 0.0015) and invasive (P = 0.0114) than NTHi 2019. No significant differences in the adherence (P = 0.4425) or invasion (P = 0.2263) were observed for an NTHi 2019 ChoP+ phase variant (52% high-reactivity ChoP+ colonies) as compared with NTHi 2019 (9% high-reactivity ChoP+ colonies). However, an NTHi 2019 ChoP phase variant (2% high-reactivity ChoP+ colonies) was significantly less adherent (P = 0.0116) and invasive (P = 0.0015) than NTHi 2019. The NTHi 2019 licD::kanr mutant was also significantly less adherent (P = 0.0011) and invasive (P = 0.0108) than NTHi 2019.

The adherence and invasion of three other clinical NTHi isolates (NTHi 1479, NTHi 3198 and NTHi 7502) with different ChoP expression phenotypes were also examined. The adherence and invasion of NTHi 1479 (26% high-reactivity ChoP+ colonies) and NTHi 3198 (78% high-reactivity ChoP+ colonies) were not significantly different from NTHi 2019 (Table 3). However, NTHi 7502 was significantly less adherent (P = 0.0311) and invasive (P < 0.0001) than NTHi 2019 (Table 3) and expressed no highly reactive ChoP+ colonies (Fig. 3, Table 2). The growth indexes of the various strains in the tissue culture medium during the course of the experiments were comparable (Table 3) and no differences in gentamycin susceptibility were observed (data not shown). These results suggest that a subset of ChoP+ LOS glycoforms mediate the adherence to and invasion of bronchial cells by NTHi.

Table 3. . Adherence and invasion of bronchial cells by NTHi.
Bacterial strainGrowth indexaCell association bSEMcInvasiondSEM
  • The data presented are means of 10 replicates.

  • a

    . Growth of the individual strains in the tissue culture medium during the course of the experiment was expressed as the GI (growth index), or the ratio of the counts at 4 h postinfection to the inoculum.

  • b

    . The cell associated bacteria (adherent + intracellular) are expressed as percentage of inocula and were obtained from non-gentamycin treated wells 4 h postinfection.

  • c

    . Standard error of the mean.


  • d

    . Invasion is expressed as the percentage of inoculum surviving gentamycin treatment.

  • e . Phase variant of NTHi 2019 enriched for high-reactivity (ChoP +) or low-reactivity and non-reactivity (ChoP) on colony immunoblots; retains the genetic capacity to alter ChoP expression.

  • * 

    Significantly different from NTHi 2019 (P < 0.01).

NTHi 20193.60406.381.190.18
NTHi 2019 pgmB::ermr3.8418*4.620.42*0.14
NTHi 2019 licD::kanr3.5120*9.510.38*0.12
NTHi 2019 ChoP+e2.93374.921.340.09
NTHi 2019 ChoPe3.4212*2.080.53*0.16
NTHi 14794.16368.420.970.08
NTHi 31982.81447.551.580.28
NTHi 75023.3422*6.700.14*0.06

NTHi bacilli co-localize with the PAF receptor

One significant host context of ChoP is as a component of the platelet-activating factor (PAF). PAF is an ether-linked acetyl-ChoP molecule which acts as an inflammatory signal mediator, in part via interaction of PAF with a G-protein linked receptor (Snyder, 1990). The PAF receptor (PAF-R) has been found on a variety of endothelial and epithelial cell types, and S. pneumoniae has been shown to adhere to host cells by means of interaction of ChoP with the PAF-R (Cundell et al., 1995). In order to address the possibility that ChoP+ LOS may mediate adherence and invasion via interaction with the PAF-R, bronchial cells were examined by confocal microscopy for the expression of the PAF-R and its co-localization with NTHi bacilli. The PAF-R was expressed by discrete cells within uninfected bronchial cell monolayers (data not shown) and NTHi 2019 bacilli (seen in green) were clearly co-localized with the PAF-R (seen in blue) in infected monolayers (Fig. 4A). Similar co-localization was observed in monolayers infected with NTHi 1479, NTHi 3198 and NTHi 7502. However, no co-localization was observed in cells infected with NTHi 2019 pgm::ermr or NTHi 2019 licD::kanr (data not shown).

Figure 4.

Immunofluorescent analysis of infected bronchial epithelial cell monolayers by confocal laser scanning microscopy.

A. NTHi bacilli colocalize with the PAF-R in infected bronchial cells. NTHi 2019 (seen in green) were visualized with Mab 3B9 and a secondary FITC conjugate. The PAF-R (seen in blue) was visualized with an antihuman PAF-R antibody (Alexis Biochemicals) and a secondary Texas Red conjugate. For the purpose of consistency, the red channel in A is shown in blue. Areas of co-localization appear light blue. The graphs are linear profiles derived from the analysis of numbers of pixels in the region and orientation denoted by the arrows.

B. Direct examination of infected bronchial cells reveals an enrichment for ChoP+ bacilli and co-localization of ChoP+ bacilli with the PAF-R. NTHi 2019 bacilli (seen in green as above) were visualized as above, with an additional labeling for ChoP expression (seen in red) using the anti-ChoP Mab TEPC-15 and a secondary Texas Red conjugate. The PAF-R (seen in blue as above) was visualized with a secondary Cy5 conjugate. Areas of co-localization of all 3 labels appear white. Parallel experiments with NTHi 1479, NTHi 3198, and NTHi 7502 yielded comparable results. Experiments with NTHi 2019 pgmB::ermr and NTHi 2019 licD::kanr showed diffuse adherence without co-localization with the PAF-R (data not shown).

No significant changes in TEPC-15 reactivity were observed in colony immunoblots of NTHi populations recovered from the adherence and invasion experiments (data not shown). However, this experiment may not directly address the question of whether NTHi infection of bronchial cells causes an enrichment for ChoP+ LOS glycoforms because of the number of generations the bacterial populations have undergone. Therefore, an immunohistochemical approach was used to examine directly the ChoP phenotypes of NTHi bacilli during an infection. Control experiments with NTHi bacilli revealed that both the high-activity and low-activity ChoP+ colonies were visualized (data not shown). These results reflect the sensitivity of the immunohistochemical approach. The analyses of infected bronchial cells, shown in Fig. 4B, show that the majority of NTHi bacilli (green) in infected monolayers were ChoP+ (red) and were almost exclusively found co-localized with the PAF-R. These results suggest that the infection of bronchial epithelial cells results in an enrichment for ChoP+ NTHi bacilli.

Inhibition of signaling via the PAF receptor decreases invasion

The involvement of the PAF-R in the adherence and invasion of bronchial epithelial cells by NTHi was evaluated by testing the effects of a PAF receptor antagonist (PAF-Ra). As shown in Fig. 5A, pretreatment with 100 nM and 10 nM concentrations of the PAF-Ra significantly inhibited the invasion of NTHi 2019 (P-values = 0.0039 and 0.006 respectively). Pretreatment with 1 nM and 0.1 nM PAF-Ra had no effect on invasion (P-values = 0.89 and 0.61 respectively). No inhibition of NTHi adherence was observed in PAF-Ra pretreated cells. However, adherence was significantly increased in bronchial cells pretreated with 100 nM PAF-Ra (Fig. 5A).

Figure 5.

Effect of pretreatment with a PAF-R antagonist (PAF-Ra) on NTHi adherence to and invasion of bronchial cells. The effect of the PAF-R antagonist (1-O-hexadecyl-2-acetyl-sn-glycerol-3-phospho-(N,N,N-trimethyl)-hexanolamine; PAF-Ra, Calbiochem) on adherence (open bars) and invasion (black bars) are expressed as percentage of untreated controls. Results significantly different from the controls are denoted with an asterisk (*). Comparable results were obtained using the PAF-R antagonist L659 989 (trans-2-[3-methoxy-5-methylsulfonyl-4-propoxyphenyl]-5-[3,4,5-trimethoxyphenyl] tetrahydrofuran, data not shown).

A. Dose-dependent inhibition of NTHi 2019 by PAF-Ra pretreatment. Different physiologically relevant concentrations (1 nm-100 nm) of PAF-Ra were added to monolayers 30 min prior to infection and maintained during the course of the experiment.

B. PAF-Ra inhibition of NTHi invasion correlates with ChoP. Monolayers were pretreated with PAF-Ra (20 nm) as above and infected with NTHi 2019, an NTHi 2019 ChoP+ phase variant, an NTHi ChoP phase variant, and an NTHi 2019 licD::kanr mutant.

C. The effect of the PAF-Ra on three other NTHi clinical isolates. Monolayers were pretreated as in B.

Experiments were then performed to determine the relationship between NTHi ChoP expression and the inhibitory effect of the PAF-Ra, as shown in Fig. 5B. Bronchial epithelial cells were pretreated with 20 nM PAF-Ra and the percentage invasion of NTHi 2019, ChoP+ and ChoP phase variants of NTHi 2019, and a NTHi 2019 ChoP mutant (licD::kanr) was determined. Significant inhibition of invasion was observed for NTHi 2019 (P = 0.0355) and a NTHi 2019 ChoP+ phase variant (P = 0.0129). No inhibition was observed for a NTHi 2019 licD::kanr mutant (P = 0.7403). A slight degree of inhibition was observed for a NTHi 2019 ChoP phase variant, although below the range of statistical significance (P = 0.0615). As in Fig. 5A, no significant inhibition of NTHi adherence was observed.

The effect of the PAF-Ra on the invasion of three other NTHi strains was also tested. As shown in Fig. 5C, the invasion of NTHi strains 1479 (P = 0.0384) and 3198 (P = 0.0110) was significantly inhibited by pretreatment of bronchial cells with the PAF-Ra; each of these strains expressed highly TEPC-15-reactive ChoP+ LOS glycoforms (Table 2). However, no inhibition was observed for NTHi 7502 (P = 0.6047). It may be noteworthy that NTHi 7502 also produced no exclusively high-reactivity ChoP+ colonies in the colony immunoblot experiments described above (Fig. 3, Table 2). The significantly lower adherence and invasion of NTHi 7502 may also be a reason for the lack of PAF-Ra inhibition (Table 3). As in Fig. 5A and B, PAF-Ra pretreament did not inhibit NTHi adherence.

Discussion

The ability to adhere to and invade the epithelial lining is thought to be an important step in the colonization and persistence of NTHi within the upper airways. Several studies have provided evidence suggesting that NTHi persist within an intracellular niche, where the bacilli are protected from host killing and may even be resistant to antibiotic treatment (St. Geme and Falkow, 1990; Forsgren et al., 1994; van Schilfgaarde et al., 1999). A number of factors has been associated with adherence and invasion, including several outer membrane proteins that are thought to act as adhesins (St. Geme et al., 1993; Rao et al., 1999).

In a defined LOS-coated bead model, we have demonstrated the ability of NTHi LOS to mediate the adherence of polystyrene beads to bronchial epithelial cells in the absence of any other factors. In parallel electron microscopy experiments, we have also observed that NTHi LOS-coated beads enter human bronchial epithelial cells (data not shown). The role of a panel of structures found within LOS in adherence was evaluated in competitive inhibition experiments. Significant inhibition of the binding of beads coated with NTHi 2019 LOS was observed with several different compounds containing ChoP. However, no inhibition was observed for soluble ChoP alone. In studies with the host ligand of the PAF-R, it has been shown that ChoP must be present within the context of a specific structure to bind to the receptor (Snyder, 1990). Our data support the conclusion that the ChoP moiety is similarly presented within a particular LOS structural context for receptor engagement. Significant inhibition of LOS-coated bead adherence was also observed after pretreatment with N-acetyl-lactosamine, although the calculated IC50 was over 20 times greater than that of the compounds containing ChoP. These data cannot exclude the possibility that an N-acetyl-lactosamine moiety is involved in LOS-mediated adherence to bronchial cells.

Subsequent experiments focused upon the comparison of the adherence and invasion of bronchial cells by NTHi strains that differ in their capacity to synthesize and express specific LOS glycoforms. The results of these experiments provide evidence to support the hypothesis that a subset of ChoP+ LOS glycoforms have a significant role in the adherence of NTHi and the subsequent invasion of the epithelial cells. The interaction of ChoP+ NTHi bacilli with the PAF-R was confirmed by immunohistochemical staining of infected monolayers and confocal microscopy. The results clearly show that NTHi are co-localized with the PAF-R, and that enrichment for ChoP+ bacilli occurs during the infection. Of particular interest are the significant decreases in adherence and invasion observed for NTHi 7502, which expressed highly TEPC-15-reactive LOS glycoforms only as discrete sectors within colonies (Table 2 and Fig. 3). The effect of the PAF-Ra on the invasion of this strain was not significant (Fig. 4C), although ChoP+ NTHi 7502 bacilli co-localized with the PAF-R were observed (data not shown). Our data suggest that the highly TEPC-15-reactive LOS glycoforms interact with the PAF-R to confer adherence and invasion of bronchial cells. One would therefore expect that the adherence and invasion of NTHi 1479 and NTHi 3198 might be increased as compared with NTHi 2019, as both these strains express more of the highly reactive ChoP+ LOS glycoforms. The current data cannot adequately address this question as a high multiplicity of infection was used in the adherence and invasion studies to ensure full saturation of all available receptors. We plan to address this issue in future experiments. Likewise, it is uncertain whether the low-reactivity ChoP epitope(s) expressed in some LOS glycoforms interact with the PAF-R, but the results of the adherence and invasion experiments (Table 3) and susceptibility to PAF-Ra inhibition (Fig. 5C) with NTHi 7502 suggest that, even if such an interaction does occur, invasion is not elicited.

The molecular basis of the two different ChoP+ colony phenotypes is uncertain. It has recently been shown that ChoP may be incorporated into LOS at different locations, and that differences in the position of ChoP within particular LOS glycoforms can be a determinant of the accessibility of ChoP to the binding of antibody or C-reactive protein (Lysenko et al., 2000a). It has been shown that alterations in the expression of other LOS structures could serve to partially mask ChoP, resulting in lower accessibility for antibody recognition (Weiser et al., 1997). It is interesting that a dramatic enrichment for high-reactivity ChoP+ bacilli was observed in infected cells, but not in the colonies produced by the same strains recovered from the adherence and invasion experiments. ChoP is detrimental in some host niches as it is a target for the binding of the C-reactive protein and subsequent complement-mediated killing (Weiser and Pan, 1998; Weiser et al., 1998a). Conversely, ChoP expression by H. influenzae has recently been associated with increased resistance to host antimicrobial peptide killing (Lysenko et al. 2000b). It therefore seems important in the overall perspective of pathogenesis that NTHi are able to rapidly modulate the expression and/or accessibility of ChoP in response to different host environments. The rapid variation in ChoP expression on the surface of NTHi could therefore represent an important adaptation to environmental shifts that are encountered by NTHi bacilli during the infectious process.

The decoration of bacterial cell surfaces with ChoP is an emerging paradigm for respiratory pathogens. A number of phylogenetically divergent bacterial pathogens inhabiting similar host niches have been shown to express ChoP on their surfaces, often in a phase-variable manner (Gillespie et al., 1996; Kolberg et al., 1997; Weiser et al., 1997, 1998b). Early insights into the importance of ChoP in host–pathogen interactions was provided by Briles and colleagues, who demonstrated that a monoclonal antibody directed against ChoP confers protection to mice against S. pneumoniae infection (Briles et al., 1981). More recently, a provocative series of studies by Tuomanen and colleagues have shown that ChoP in the pneumococcal cell wall mediates bacterial adherence to the PAF-R on immortalized type II pneumocytes that were activated by pretreatment with inflammatory cytokines. The pneumococcal/PAF–R interaction results in a complex series of host cell signal events that are not yet fully defined (Cundell et al., 1995). The engagement of the PAF-R by the pneumococcus also affects the permeability of the blood–brain barrier, allowing the bacteria to transmigrate into the subarachnoid space (Cabellos et al., 1992; Ring et al., 1998).

It has been proposed that ChoP is a determinant of the ability of NTHi to colonize and persist within the nasopharyngeal environment, perhaps by mediating bacterial adherence to and invasion of the host epithelia. Evidence for this hypothesis includes the finding that the number of tetrameric repeats in H. influenzae isolated from human respiratory secretions are enriched for variants which would be predicted to express ChoP (Weiser et al., 1998a). Our findings suggest that a particular subset of ChoP+ LOS glycoforms, that are highly reactive with TEPC-15, mediate the interaction of NTHi bacteria with the PAF-R on bronchial epithelial cells. These ChoP+ glycoforms may stimulate PAF-R-mediated signal events resulting in invasion, as evidenced by the inhibition of NTHi invasion by the PAF-Ra in direct correlation with ChoP expression. Pretreatment of bronchial epithelial cells with high doses of PAF-Ra (100 nM) was also shown to increase the number of adherent bacteria (Fig. 5A). The reason for this finding is unclear at this time.

The expression of the PAF-R is elevated in chronically inflamed airways, such as those found in chronic bronchitic and asthmatic patients. The ChoP/PAF–R interaction may be of particular importance in the elevated incidence of NTHi opportunistic disease in these populations. The stimulation of the PAF-R by its host ligand elicits a complex, G-protein mediated host cell signal cascade (Snyder, 1990). It is particularly striking that NTHi and the pneumococcus, which are phylogenetically distant but exist within the same ecological niche, both utilize a surface-exposed ChoP moiety to interact with the PAF-R. This may represent convergent evolution within different respiratory pathogens to allow exploitation of host cell biology by means of molecular mimicry of a signal mediator.

Experimental procedures

Bacteria and culture conditions

The bacterial strains used in this study are shown in Table 4. NTHi 2019, NTHi 1479, NTHi 3198 and NTHi 7502 are clinical isolates from patients with chronic obstructive pulmonary disease (Campagnari et al., 1987). These strains possess the genes necessary for the production of the HMW1 and HMW2 adhesins, as well as fibrils and pili (Ketterer et al., 1999). The cultures were propagated from frozen stock cultures on brain-heart infusion (BHI) agar (Difco) supplemented with 10 µg ml−1 hemin (Sigma) and 1 µg ml−1 NAD (Sigma), at 37°C and 5% CO2. NTHi 2019 pgmB::ermr was grown on supplemented BHI agar containing 5 µg ml−1 erythromycin (Sigma). H. influenzae strains H394 and H395 are ChoP+ (H394) and ChoP (H395) phase variants of H. influenzae strain Sec-1, which is a capsule-deficient mutant of strain Eagan (Moxon et al., 1984). These variants were obtained by screening colony immunoblots using a monoclonal antibody against ChoP (TEPC-15), as previously described (Weiser et al., 1997). H. influenzae strains H454 and H455 are ChoP+ (H454) and ChoP (H455) phase variants of NTHi 2019 that were isolated in a similar manner. H. influenzae strain H477 (NTHi 2019 licD::kanr) is an isogenic mutant of strain H454 (Weiser et al., 1997) and was propagated on supplemented BHI agar containing 15 µg ml−1 ribostamycin.

Table 4. Bacterial strains.
Bacterial strainDescriptionDerivationReference
  • a

    . Chronic obstructive pulmonary disease.

  • b . ORF designations are as published in the H. influenzae Rd genomic database ( Fleischmann et al., 1995).

  • c . Phase variant of NTHi 2019 enriched for high-reactivity (ChoP +) or low-reactivity and non-reactivity (ChoP) on colony immunoblots (see Fig. 3 and Table 3); retains the genetic capacity to alter ChoP expression.

NTHi 2019Clinical isolateCOPDa patient Campagneri et al. (1987)
NTHi 2019 pgmB::ermrTruncated LOSInactivation of ORF HI 0740bThis work
NTHi 2019 licD::kanrChoPInactivation of ORF HI 1540b Lysenko et al. (2000a)
H454NTHi 2019 ChoP+cSee Experimental proceduresThis work
H455NTHi 2019 ChoPcSee Experimental proceduresThis work
NTHi 1479Clinical isolateCOPD patient Campagneri et al. (1987)
NTHi 3198Clinical isolateCOPD patient Campagneri et al. (1987)
NTHi 7502Clinical isolateCOPD patient Campagneri et al. (1987)

Construction of a pgmB::ermr insertional mutant

A 1.8 kb fragment containing the pgmB gene was amplified from NTHi 2019 genomic DNA using the polymerase chain reaction (PCR). The primer sequences were designed by analysis of the sequence data in ORF HI0740 of the Haemophilus influenzae strain Rd genomic sequence data base (Fleischmann et al., 1995). The primers (HI1740U-5′GGAGGTTGCACAACATTGG; HI1740D-5′CCTACAACACTGTTCCGC) were purchased from IDT and pgmB was amplified from a preparation of NTHi 2019 genomic DNA. The resulting amplicon was submitted for analysis at the University of Iowa DNA sequencing core facility and was confirmed to contain the expected region of pgmB. A null allele containing an erythromycin resistance cassette inserted within an internal BamHI site was constructed and transformed into NTHi 2019 using an established procedure (Herriott et al., 1970). The transformants were screened for loss of reactivity with the monoclonal antibody 6E4, which recognizes a conserved terminal structure within the branched oligosaccharide region, and by Southern blotting.

Phosphoglucomutase assays

Cell-free extracts were prepared from bacteria harvested from supplemented BHI agar plates as described previously and disrupted by sonication (Joshi, 1982; Pilkis et al., 1982; Sa-Correia et al., 1987). The phosphoglucomutase activity was measured using a colorimetric assay (Joshi, 1982; Pilkis et al., 1982). For negative control reactions, the substrates were omitted from the reaction mixture.

Isolation and analysis of LOS

LOS was isolated from NTHi 2019 and NTHi 2019 pgmB::ermr either by the phenol–chloroform–petroleum ether method (Galanos et al., 1979) or by the method of Johnson (Johnson and Perry, 1976). The LOS samples were examined by Tricine–SDS–PAGE according to previously described methods (Lesse et al., 1990). The LOS bands were visualized by silver staining. The ChoP content of the extracted H394 and H395 LOS was confirmed by Western blotting with mAb TEPC-15 according to previously described methods (Weiser et al., 1997).

Mass spectrometric analysis of LOS

LOS preparations from NTHi 2019 and NTHi 2019 pgmB::ermr were analysed by matrix-assisted laser desorption ionization (MALDI) and electrospray ionization (ESI) mass spectrometry after conversion to their corresponding O-deacylated derivatives (anhydrous hydrazine, 37°C for 30 min) (Phillips et al., 1990). The resulting O-deacylated LOS were precipitated with cold acetone, taken up in H20, lyophilized and re-dissolved in H20/acetonitrile (1 : 1, v/v). For MALDI analysis (Gibson et al., 1997), 1 µl of analyte containing between 0.1 and 1 µg of O-deacylated LOS was mixed with 1 µl of matrix solution (320 mM 2,5 dihydroxybenzoic acid in acetonitrile/water), desalted with cation-exchange resin beads (DOWEX 50X, NH4+) and air dried on the MALDI target. Mass spectra were obtained in the negative ion mode on a Voyager DE (PE Biosystems) equipped with a nitrogen laser (337 nm) and delayed extraction optics (Vestal et al., 1995). An equimolar mixture of angiotensin II, bradykinin, LHRH, bombesin α-MSH and ACTH 1–25 (CZE mixture, Bio-Rad) was separately prepared and used for external mass calibration. For ESI analysis (Gibson et al., 1993), O-deacylated LOS samples were dissolved in H2O/acetonitrile containing 1% acetic acid at a concentration of 1 µg ml−1 and 5 ml was injected onto an API-300 Sciex triple quadrupole operating in the negative ion mode. Spectra were externally calibrated.

Composition analysis

Monosaccharide composition analyses were carried out on O-deacylated LOS preparations from NTHi 2019 and NTHi 2019 pgmB::ermr. Neutral sugars were prepared by the hydrolysis of each LOS sample in 2 M trifluoroacetic acid and amino sugars and by hydrolysis in 6 N HCl, both at 100°C for 3 h. The resulting monosaccharides were separated and analysed by high pH anion exchange chromatography (HPEAC), using a Dionex HPLC system equipped with a PA1 column as previously described (Phillips et al., 1990).

Culture of immortalized bronchial epithelial cells

The 16HBE14 cell line was originally derived by SV-40 transformation of a polyclonal population of cells obtained from human bronchial brushings and thus contains a number of different cell types present within the upper airways (Gruenert et al., 1988). Seed cultures of 16HBE14 were kindly provided by Dr D. C. Gruenert (University of California, San Francisco, USA). The bronchial cells were propagated in Eagle's MEM culture medium (Gibco), supplemented with 10% fetal calf serum and 1% l-glutamine. The cells were seeded into individual wells of 24-well culture dishes at an approximate density of 105 cells well−1. The cells were incubated at 37°C and 5% CO2 until reaching confluence, then used for the infection studies.

Invasion experiments

Bacterial invasion was estimated using a standard gentamycin-survival assay. Fresh cultures of the bacterial strains were prepared on prewarmed supplemented BHI plates and incubated overnight. The bacteria were harvested with a sterile cotton-tipped swab and suspended in sterile PBS to a density of 50 Klett units (approximately 108 bacteria ml−1). Confluent monolayers were washed twice with fresh prewarmed culture medium and infected with a 1:10 dilution of each bacterial suspension in culture medium (1 ml well−1). The infected monolayers were incubated for 4 h, washed twice and fresh medium containing gentamycin (50 µg ml−1) was added to kill the extracellular bacteria. After 1.5 h of incubation, the monolayers were washed three times with fresh medium and lysed by the addition of sterile 1% saponin. The lysed cells were mixed thoroughly by vigorous pipetting and serially diluted in warm Morse's buffered saline. The dilutions were plated on prewarmed supplemented BHI agar and incubated overnight, and the numbers of colonies were counted and used to estimate the numbers of colony-forming units (cfu) well−1. For the inhibition of the PAF-R, cell monolayers were preincubated for 30 min in the presence of a PAF-R antagonist (1-O-hexadecyl-2-acetyl-sn-glycerol-3-phospho-[N,N,N-trimethyl]-hexanolamine; PAF-Ra, Calbiochem), which binds competitively to the active site of the PAF-R, but does not elicit cell signalling (Takumura et al., 1985). The inhibitory concentrations in these experiments are within physiologically relevant ranges, as indicated by biochemical studies concerning PAF-R/ligand interactions (Takumura et al., 1985). PAF-Ra concentrations were maintained throughout the course of the infection and invasion was determined as described above. Similar experiments were performed using comparable concentrations of the PAF-R antagonist L659 989 [trans-2(3-methoxy-5-methylsulfonyl-4-propoxyphenyl)-5-(3,4,5-trimethoxyphenyl) tetrahydrofuran], which was kindly provided by Merck, Sharp, and Dohme Research Laboratories.

For the determination of bacterial adherence, monolayers were infected as above, washed vigorously four times with fresh media to remove nonadherent bacilli and then lysed with 1% saponin. The lysates were serially diluted in warm Morse's buffered saline and plated onto sBHI agar plates as above.

Preparation of coated polystyrene beads

The coated polystyrene microbeads were prepared by a modification of the method described by Schlesinger et al. (1996). A suspension of 25 µl of 1 µm polystyrene beads was diluted in 0.05 M carbonate–bicarbonate buffer (CB buffer, pH = 9.6) in a silicon-coated tube. The beads were washed twice (2 min wash−1) in CB buffer, resuspended in 1 ml of NTHi 2019 LOS (0.05, 0.5, 5, 25 and 50 µg ml−1), NTHi 2019 pgmB::ermr LOS (5, 25, or 50 µg ml−1) or CB buffer alone, and rotated for 2 h at room temperature (RT). The beads were washed twice (2 min wash−1) in CB, resuspended in 1 ml of a 0.1% gelatin solution and rotated for 2 h at RT, then washed once with 0.1% gelatin, twice with CB (2 min wash−1) and resuspended in 0.5 ml CB for storage at 4°C. The binding of LOS to the beads was confirmed by ELISA and FACScan analysis using polyclonal rabbit antiserum to NTHi 2019.

Bead adherence experiments

Bronchial cells were seeded onto collagen-coated glass cover slips (12 mm2) and grown to confluence, then washed twice in medium. The bead suspensions (10 µl) were diluted 1: 10 in medium, added to the surface of the monolayers and incubated for 4 h. The monolayers were then washed twice using an orbital shaker with Dulbecco's phosphate-buffered saline (PBS, Gibco) and fixed in 2% paraformaldehyde–PBS for 30 min. The cover slips were processed for scanning electron microscope (SEM) analysis and viewed on a S-4000 Hitachi scanning electron microscope. Ten random fields from each sample were analysed to determine the mean number of beads field−1 using the NIH image program (available at http://rsb.info.nih.gov/nih-image). In the competitive inhibition experiments, the cover slips were preincubated with the various treatments (0.1–10 µg ml−1) for 30 min, then washed in PBS and the bead adherence was assessed as described above.

Confocal microscopy

Bronchial cells were seeded into four- or eight-chamber slides (Nunc) and grown to confluence, then infected for 4 h essentially as described above. The monolayers were washed twice with PBS and fixed for 30 min in 2% paraformaldehyde/PBS, then washed again with PBS and stored at 4°C. The monolayers were incubated in 2% normal goat serum in PBS for 30 min as a blocking step, and immunochemical staining was performed using the monoclonal antibody 3B9, which recognizes the NTHi outer membrane protein P6, the monoclonal anti-ChoP antibody TEPC-15 (Sigma) and/or rabbit antiserum against the human PAF-R (Alexis Biochemicals). Fluorescent antibody conjugates were added according to standard methodology and as indicated in the figure legends. The fluorescent secondary antibody conjugates were obtained from Molecular Probes (goat anti-mouse IgG/FITC, goat anti-rabbit IgG/Texas Red) or Cortex Biochemical (rabbit anti-mouse IgA/Texas Red, goat anti-rabbit IgG/Cy5). Positive control infections were performed using the acapsular Streptococcus pneumoniae strain Rx1 and visualized using a monoclonal antibody against the human PAF-R (Alexis Biochemicals), MAb TEPC-15, and a rabbit antibody against the surface-exposed pneumococcal protein PspA (kindly provided by David Briles, University of Alabama at Birmingham, USA). The fluorescent conjugates used in the positive control experiments were goat anti-rabbit IgG/FITC (Molecular Probes), goat anti-mouse IgG/Cy5 (Cortex Biochemical) and rabbit anti-mouse IgA/Texas Red (Cortex Biochemical). Negative controls, consisting of uninfected and infected monolayers incubated with the fluorescent conjugates in the absence of any primary antibody, were performed for each bacterial strain and conjugate. Cover slips were mounted using the ProLong Antifade reagent (Molecular Probes) according to the manufacturer's instructions, and stored in the dark at room temperature. The slides were viewed on a Zeiss 510 confocal laser scanning microscope.

Colony immunoblots

ChoP expression by the NTHi strains used in this study was assessed by colony immunoblotting, essentially as described previously (Weiser et al., 1997). NTHi colonies from plates with 100–1000 colonies were transferred onto nitrocellulose membranes and washed twice in TSBB (0.5 M NaCl, 0.5% Tween-20, 10 mM Tris-HCl, pH = 8.0) for 15 min wash−1. The membranes were incubated with mAb TEPC-15 in TSBB overnight, then washed five times in fresh TSBB (5 min wash−1). The membranes were then incubated overnight in a goat anti-mouse IgA/alkaline phosphatase conjugate (1:10,000, Sigma) in TSBB, washed five times and incubated for 10 min in AP buffer (100 mM NaCl, 5 mM MgCl2, 100 mM Tris-HCl, pH = 9.5). ChoP+ colonies were visualized using BCIP (Sigma, 0.16 mg ml−1) and nitroblue tetrazolium (Sigma, 0.33 mg ml−1) in AP buffer. Colonies were classified as having high, low or no reactivity with the antibody as described previously (Weiser et al., 1997).

Statistical analysis

Statistical analyses were performed using the interactive versions of sas (SAS Institute) and statview (Abacus Software). The P-values reported for the adherence and invasion experiments were determined by paired T-test analysis. P-values of < 0.05 were considered to indicate statistically significant differences. The P >F-values given in the bead adherence experiments represent the significance of the effect of an inhibitor over a range of concentrations and were determined by comparison of linear regression plots, using the analysis of variance (anova) method. As with P-values derived from T-tests, P >F-values of < 0.05 were considered to be significant. To facilitate comparisons of the efficacy of different compounds in the inhibition experiments, the 50% inhibitory concentration (IC50) was determined for each structure tested using formulas derived from simple linear regression analysis on a semilogarithmic scale. Because the calculation of a meaningful IC50 value requires a significant dose-dependent inhibition, these values were only reported for those structures for which significant inhibition was observed (as determined by P >F-value as indicated above). All experiments were performed in sufficient replication to ensure statistical validity; however, in some cases limiting reagents were an issue and fewer replicates were performed. The degrees of freedom (DF) are therefore included for all experiments.

Acknowledgements

W.E.S. and B.A.B. contributed equally to this work. The authors gratefully acknowledge expert assistance and helpful discussions with Bridget Zimmerman concerning the statistical analysis of the data. Meg Ketterer, Randy Nessler, Jian Shao and other members of the University of Iowa Central Microscopy Research Facility staff provided expert assistance with microscopy and in the preparation of figures. Thanks also to Peter Giardina, Hillery Harvey, Paul Jones, Theresa Miller and Deb Post for critical reviews of the manuscript and helpful discussions. W.E.S. was supported by a NIH training program in Mechanisms of Parasitism (AI07511). B.A.B. was supported by NSF REU grant 9605122. This work was supported by NIH research grants awarded to M.A.A. (AI24616, AI65298), J.N.W. (AI38446, AI44231) and B.W.G (AI31254). The mass spectrometry was performed at the UCSF Mass Spectrometry Facility which is partially supported by a grant from the National Center for Research Resources (NCRR BRTP 01614).

Ancillary