Surfactant Protein D Binds Selectively to Klebsiella pneumoniae Lipopolysaccharides Containing Mannose-Rich O-Antigens¹ Free

The lung collectins, surfactant protein A (SP-A)³ and surfactant protein D (SP-D), are now recognized as important components of pulmonary innate immunity and the lung’s response to acute lung injury (1). These proteins, which are secreted by epithelial cells lining the respiratory tract, are ideally situated to participate in the initial interactions with inhaled microorganisms (2).

Most secreted collectins, including SP-A and SP-D, are assembled as multimers of trimeric subunits. Each subunit consists of an amino-terminal cross-linking domain, a collagen domain, a linking peptide, and a mannose-type, C-type lectin carbohydrate recognition domain (CRD) (3, 4). Each of the trimeric subunits appears “functionally univalent” with respect to binding to particulate, multivalent ligands. Multimerization of the trimeric subunits of SP-D to form dodecamers permits bridging interactions that can lead to efficient microbial agglutination and enhance the interactions of various microorganisms with phagocytic cells (1).

Klebsiella pneumoniae is an important cause of pneumonia, particularly among hospitalized patients. Pattern recognition molecules of the innate immune system are especially important for the clearance of such opportunistic pathogens. In this regard, there is growing evidence that SP-D and SP-A, and the macrophage mannose receptor (MR), another C-type lectin, participate in the host defense against this Gram-negative microorganism through interactions with cell wall glycoconjugates (5, 6, 7, 8). For example, we have previously shown that SP-A and MR preferentially interact with encapsulated strains of K. pneumoniae containing dimannose or dirhamnose sequences in their capsular polysaccharides (6).

Although the expression of capsular polysaccharides is an important mechanism for the evasion of opsonic host defenses, nonencapsulated organisms probably play particularly important roles in the initial stages of lung epithelial attachment and colonization. Nonencapsulated variants of K. pneumoniae spontaneously emerge from the colonies of encapsulated strains as sectors typical of the phase variation phenomenon (9). These nonencapsulated variants bind to and invade epithelial cells more efficiently than the parental encapsulated strains, presumably through enhanced interactions of the epithelial cells with bacterial adhesins (10, 11). It is likely that colonization of the respiratory tract is selective for these nonencapsulated variants because the nonadherent, encapsulated bacteria are expected to be cleared more efficiently by the mechanical defense systems of the lung (12). These and other observations suggest that phase variation to a nonencapsulated form is required to initiate infection in a susceptible host.

In previous studies, we found that SP-D interacts with nonencapsulated strains of K. pneumoniae, but not the corresponding encapsulated strains (8). The major surface glycoconjugate of nonencapsulated Klebsiella is the LPS. In Klebsiella, nine different LPS O-serotypes have been described based on reactivity of the antigenic O-polysaccharides with specific Abs (13, 14). Of these, the O1 serotype is the most common O-Ag found among clinical isolates (15). The molecular basis for the predominance of the O1 serotype is not clear. However, it has been suggested that O-Ags of the LPSs of pathogenic bacteria play a major role in conferring resistance of the bacteria to the bactericidal activity of human serum (16, 17).

LPS is the major SP-D ligand associated with the cell wall of strains of Escherichia coli and Salmonella minnesota (18). In addition, SP-D shows CRD-dependent binding to the LPS isolated from a number of other Gram-negative bacteria, including K. pneumoniae (19). The characterization of mutant forms of E. coli and S. minnesota LPS suggested that the glucose-containing core oligosaccharide of the LPS is necessary and sufficient for CRD-dependent binding (18). In addition, SP-D was found to selectively aggregate strains of E. coli expressing immature, “rough” forms of LPS, which lack the O-polysaccharide (O-Ag) domain (18, 20).

More recently, we demonstrated that the binding of SP-D to the LPS of unencapsulated Klebsiella results in bacterial agglutination and opsonization, with enhanced internalization and killing of the nonencapsulated variants by alveolar macrophages (8). Consistent with earlier studies of E. coli, SP-D was observed to potently agglutinate Klebsiella that express a high proportion of rough LPS (R-LPS) (21). However, in preliminary experiments, we also observed the agglutination of organisms expressing “smooth” LPS (S-LPS), suggesting previously unrecognized modes of interaction with the LPS molecule.

In the present study, we investigated the role of the O-polysaccharide (O-Ag) domain on the interaction of SP-D with LPS purified from 16 well-characterized clinical strains of K. pneumoniae. We found that the structure of the oligosaccharide-repeating units of the O-Ag determines the ability of SP-D to efficiently react with the bacterium, and with its ability to bind to the mature smooth forms of purified LPS. These observations expand the range of potential recognition sites for SP-D on Gram-negative bacteria and bacterial LPS. We also found that SP-D dodecamers decrease the adhesion of Klebsiella expressing “reactive” O-polysaccharides to lung epithelial cells in vitro. Thus, in addition to its opsonic activity, SP-D may act as an anti-adhesion molecule to provide innate immunity against certain O-serotypes of Klebsiella. Notably, Klebsiella isolates expressing “nonreactive” O-polysaccharides were more frequently associated with pneumonia among hospitalized patients.

Clinical isolates included in the O-Ag seroepidemiology (see Table II) were from patients hospitalized at the University of Kiel Hospital (Kiel, Germany), three other affiliated teaching hospitals, and two cooperating clinical centers. All patients (n = 138) fulfilled the CDC criteria of nosocomial pneumonia, urinary tract infections (UTIs), and primary bloodstream infections (22, 23). In particular, the diagnosis of Klebsiella pneumonia was based on clinical and radiographic presentation, and bacterial cultures. In addition to the isolation of K. pneumoniae, the patients had one or more of the following: 1) rales or dullness on chest percussion and new onset of purulent sputum; or 2) new lung infiltrates, consolidation, pulmonary cavitation, or pleural effusion. Bacterial isolates were from sputum, transtracheal aspirates, bronchoalveolar lavage fluid that contained a significant number of bacteria (≥10⁴ organisms/ml), or open-lung biopsies. The origin of each strain was noted and only one strain per patient was included. In most cases, other clinical information was not available.

Two nonencapsulated strains of the K. pneumoniae K21a and K50 were described elsewhere (8). The R20/O1(−) strain, which lacks the O1 Ag but retains the lipid A and core domains was also previously described (24, 25). All other Klebsiella strains used were isolates from patients in various hospitals as follows. The isolates VA16391, VA16804, VA15485, VA18495, VA1661, VA10399, VA12973, VA1619, BK2150, and BK6672 were obtained from the University Hospital and three other affiliated teaching hospitals. The CF9 and CF45 isolates were kindly donated by Dr. D. Sirot (Laboratoire de Bactériologie, Faculté de Pharmacie, Clermont-Ferrand, France), and the Cop92/96 and Cop19 isolates by Dr. D. S. Hansen (Statens Seruminstitut, Copenhagen, Denmark). The API 20E system (bioMérieux, Nürtingen, Germany) was used to confirm the identity of the strains.

The bacteria were grown either in Luria broth or on nutrient agar (Difco, Detroit, MI) for 24 h at 37°C. The bacteria were harvested from broth by centrifugation 3000 × g or by scraping the confluent growth from agar, respectively. The bacteria were washed three times with endotoxin-free HEPES-buffered saline (HBS) (5 mM HEPES, 150 mM NaCl, pH 7.5), and then resuspended at the desired density in the same buffer. The O-serotyping of the strains was performed by Dr. J. V. Benedí (Universidad de las Islas Baleares, Palma de Mallorca, Spain) (13), and confirmed by using rabbit anti-O Ag Abs and a competitive ELISA as previously described previously (26).

Human A549 adenocarcinoma cells (CCL-185; American Type Culture Collection, Manassas, VA), which are of probable alveolar type II cell derivation, were cultivated as described (27). The cells were grown and maintained in F-12K medium with 10% (v/v) FCS and 2 mM glutamine.

Rabbit polyclonal Abs specific for the O-Ags of LPS from O1, O3, O4, and O5 reference strains (26) were kindly supplied by Prof. M. Trautmann (Ulm University Hospital, Ulm, Germany). The preparation of mouse monoclonal anti-lipid A and anti-core Abs are described elsewhere (28). LPS was isolated and purified from the reference O-serotypes and from the clinical isolates using conventional methods (8, 29).

Recombinant human SP-D dodecamers, recombinant wild-type rat SP-D dodecamers (RrSP-D), recombinant rat SP-D trimers (RrSP-Dser15,20), and recombinant rat SP-D dodecamers lacking the consensus for asparagine-linked glycosylation at residue 70 of the mature protein (recombinant rat SP-D dodecamers lacking N-linked sugars; RrSP-Dala72), were isolated by sequential saccharide affinity chromatography and gel filtration chromatography (30, 31, 32, 33). All showed minimal endotoxin contamination as confirmed using a chromogenic assay. Purified proteins were stored at −80°C in HBS containing 10 mM EDTA.

Purified LPS (1 μg/lane) was resolved on SDS-PAGE slab gels without urea as described previously (8). One of the gels was stained with silver according to the method of Tsai and Frasch (34), while the LPS bands from the other gels were transferred to nitrocellulose sheets using a wet transfer apparatus as described earlier (8). In some experiments, strips of the blotted membranes were incubated at 100°C in 0.1 M HCl for 1 h to remove the LPS O-side chains and the core oligosaccharide (35). All strips were overlaid with either purified SP-D at indicated concentrations; or with 1/1000 dilutions of anti-lipid A, anti-core, or anti-O-sera. BSA (Sigma-Aldrich, St. Louis, MO) was used as a blocking agent. Bound SP-D was detected using rabbit polyclonal Abs to SP-D, as described previously (8). The detection of bound anti-LPS-core or anti-lipid A or anti-O-Ag Abs was performed as described elsewhere (28, 36).

Recalcified SP-D in HBS, or control buffer, was added to suspensions of Klebsiella at 0.5 OD (10⁸ CFU/ml) to obtain the final indicated concentrations. After 30 min at room temperature, the bacteria were added to wells of 96-microtiter plates flat-bottom containing confluent monolayers of A549 cells, which had been briefly washed with HBS containing 20 mM CaCl₂. After 30 min at room temperature, the nonadherent bacteria were removed by repeated washing and the cell monolayer was fixed with 2.5% glutaraldehyde in HBS for 5 min. The plates were then washed and incubated in blocking buffer overnight at 4°C. The bound bacteria were quantified using anti-Klebsiella serum at 1/1000 in ELISA as described previously (37). Each adhesion experiment shown in Figs. 4–6 was performed at least three times; statistical significance was calculated using the Student’s t test.

In our previous study, we observed that human SP-D was more effective in the agglutination of Klebsiella K50-3OF (a nonencapsulated phase variant of the O3 serotype) than K21a/3 (a nonencapsulated phase variant of the O1 serotype) (8). Determination of the minimal SP-D concentration needed to induce agglutination of the O1-serotype was ∼10-fold higher than that needed for the O3 serotype (K50-3OF) (4.2 and 0.5 μg/ml, respectively; Table I). The encapsulated parent strains K21a (O1) and K50 (O3) were not agglutinated at concentrations as high as 10 μg/ml (data not shown).

LPS extracted from the two Klebsiella strains differed markedly in their ability to inhibit SP-D-induced agglutination of Klebsiella (Table I, column 2). In particular, much lower concentrations of the O3-LPS were required to inhibit agglutination. Capsular polysaccharides isolated from the parent K21a/3 (O1) and K50-OF (O3) strains did not inhibit agglutination at concentrations as high as 150 μg/ml (data not shown).

As shown in Table I, an O1 Klebsiella mutant lacking the O1-polysaccharide in its LPS (but retaining the core region) was agglutinated at much lower concentrations of SP-D than the wild-type O1 strain. Additional experiments were performed to determine whether the differences in SP-D binding to the O-serotypes are due to the structure of their O-Ags.

For this purpose, the LPS isolated from K50-3OF (O3) and K21a/3 (O1) were subjected to SDS-PAGE to resolve subpopulations of LPS molecules containing the lipid A plus core region (R-LPS) from populations containing various numbers of repeating units of O-Ag linked to their core oligosaccharides (S-LPS). Silver staining of the gels showed that LPS isolated from both serotypes has a typical ladder-like banding pattern with a major polydisperse component that migrates slightly slower than the buffer front (Fig. 1, lanes 1 and 10).

To definitively identify the silver-stained components and confirm the purity of the LPS, nitrocellulose blots were incubated with monoclonal anti-lipid A, monoclonal anti-core, or anti-O-Ag, with or without prior treatment with HCl. The acid treatment selectively hydrolyzes polysaccharide chains—including the core oligosaccharide and O-polysaccharide chain (35). As shown in Fig. 1, the slowly and rapidly migrating components of the ladder both reacted with the anti-core Ab. However, reactivity was lost from both populations following removal of the carbohydrate with HCl treatment. As expected, the slowly migrating components of the ladder, but not the rapidly migrating components, reacted with the Ab to O-Ag; this reactivity was also lost following treatment with HCl. In contrast, the Abs to lipid A reacted with all components, but only following removal of the carbohydrate, which masks epitopes associated with the lipid A (38). Thus, the rapidly migrating components in our preparations are definitively shown to consist of LPS molecules consisting of lipid A conjugated to core oligosaccharide but lacking O-Ag (R-LPS), while the more slowly migrating components of the ladder consist of mature LPS molecules containing O-Ag linked to their lipid A and core domains (S-LPS).

Binding of human SP-D was then examined by incubating parallel blots with SP-D, followed by detection of the bound SP-D using the indirect immunoassay described in Materials and Methods. As shown in Fig. 1 (lanes 8–9 and 17–18), SP-D strongly and specifically reacted with the rapidly migrating components (i.e., R-LPS) of the K21a/3 (O1 serotype), but showed no reaction with the slowly migrating, O-Ag containing species (S-LPS). By contrast, SP-D reacted strongly with both populations of LPS molecules derived from the K50-3OF (O3 serotype). Thus, SP-D can bind to the smooth or O-Ag-containing LPS of an O3 serotype, but not the corresponding species from an O1 serotype. These SP-D binding data are representative of several independent experiments, and similar results were obtained with RrSP-D (data not shown).

Limited acid hydrolysis eliminated all reactivity of the LPS with the SP-D (Fig. 1, lane 18), despite retention of epitopes for the anti-lipid A Ab, as described above. This finding is consistent with previous studies showing that the CRD of SP-D specifically interacts with saccharides associated with the LPS (18). Binding was prevented by the presence of EDTA. As expected, binding was also prevented by 10 mM maltose, a competing sugar of SP-D lectin activity, but not 10 mM lactose, a noncompeting sugar (data not shown). Thus, the observed interactions involve the divalent cation-dependent binding of the CRD of SP-D to carbohydrates associated with the LPS.

In other experiments, we examined the dose-dependence of SP-D binding to the LPS. As shown in Fig. 2, there was no obvious difference in the SP-D concentration needed to visualize the R-LPS components of the two serotypes, consistent with previous studies that demonstrated binding to the core oligosaccharide. Although the dose response for SP-D binding to the R-LPS and S-LPS components was similar for the O3 serotype, a higher concentration of SP-D was needed to visualize binding of SP-D to the S-LPS forms. In fact, binding to the O1-Ag containing species was only barely discernible on the original blot at the highest concentrations.

Because the chemical structure of the O-Ags, which also determines the serotype of Klebsiella, has been characterized (39), we sought to determine whether the differential binding of SP-D to the examined O1 and O3 strains is dependent on specific structure. For this purpose, we performed blotting experiments of LPS isolated from each of four major Klebsiella serotypes (O1, O3, O4, and O5) that were characterized as part of recent epidemiological studies (15). All isolates were smooth strains and showed similar ladders of S-LPS on silver-stained gels (data not shown). As shown in Fig. 3, LPS from all isolates of a given serotype showed comparable profiles of reactivity with SP-D, suggesting that the reactivity is related to the structural differences of the O-Ag. Although SP-D (1 μg/ml) bound to the R-LPS bands of all strains tested, binding to the S-LPS bands was restricted to the O3 and O5 serotypes, which express O-Ags containing mannose, rather than galactose, repeating units (39).

We next sought to determine whether the serotype-dependent differences in SP-D reactivity correlate with differences in the propensity of these organisms to cause infection. For this purpose, we statistically analyzed the relative frequency of isolation of strains belonging to the O1, O3, O4, and O5 LPS serotypes among Klebsiella strains isolated from patients with nosocomial pneumonia, UTI, and primary bacteremia. Nonreactive strains were operationally defined as those that express S-LPS that does not react with low concentrations of SP-D (as observed for the 16 strains studied, Figs. 1–3). Reactive strains were defined as those that express S-LPS that binds to SP-D under the same conditions. Pearson’s χ² test showed that the frequency of isolation of the SP-D reactive O-serotypes was significantly lower (28%) than that of the nonreactive serotypes (72%) (Table II; p < 0.001). Although the difference between the pneumonia and nonpneumonia groups was significant (p < 0.019), there was no significant difference among isolates from blood and urine (Table II; p0.2). These frequencies of isolation are consistent with the results of our recent European multicenter study (15) and with other studies of a number of European countries (13, 26).

Epithelial adhesion is believed to constitute a critical event in the process of infection. Given our experimental findings and the epidemiological data, we examined the effects of SP-D on the adhesion of the O1 and O3 Klebsiella serotypes to A549 lung epithelial cells as a model of epithelial colonization. Recent studies have shown that nonencapsulated phase variants are required for efficient epithelial attachment and invasion, presumably through the enhanced exposure of bacterial adhesins. For this purpose, we used a well-characterized immunological method that employs specific Klebsiella Ab to quantify adherent bacteria (37). Previous studies have shown that under these conditions of assay the number of bound bacteria correlates directly with the ELISA signal. As shown in Fig. 4, human SP-D showed dose-dependent inhibition of bacterial adhesion. Notably, the minimal concentration of SP-D needed to significantly inhibit the adhesion of nonencapsulated Klebsiella to the epithelial cells was 3 μg/ml for the O1 serotype, as compared with 0.75 μg/ml for the O3 serotype. Thus, SP-D decreases bacterial adhesion in vitro, and the potency of this inhibitory effect is greater for the SP-D reactive than for the nonreactive O-serotype.

SP-D is preferentially, but not exclusively, assembled as dodecamers consisting of four trimeric subunits. Some of the known activities of SP-D, such as particle aggregation, are dependent on multimers (2). However, trimeric subunits or recombinant trimeric lectin domains can mediate other activities, including the neutralization of certain respiratory viruses in vitro and in vivo.

To further define mechanisms of the inhibitory effect of SP-D on epithelial adhesion, we next compared the activity of RrSP-D dodecamers (four arms) with single arm, trimeric subunits (RrSP-Dser15,20). We have previously shown that the mutant protein is fully active as a lectin, but is defective as a bacterial agglutinin (40). In contrast with wild-type RrSP-D dodecamers, RrSP-Dser15,20 did not significantly inhibit the adhesion of the K50 (O3) serotype to epithelial cells, even at concentrations as high as 5 μg/ml (Fig. 5). Interestingly, 5 μg/ml of RrSP-Dser15,20 was able to significantly decrease the inhibitory effect of 3.5 μg/ml SP-D dodecamers. In microscopic slide agglutination assays, a 5-fold weight excess of RrSP-Dser15,20 markedly inhibited the agglutination by RrSP-D (data not shown).

Gram-negative organisms, including Klebsiella, can express a variety of adhesins, some of which can interact with complex oligosaccharides expressed on host glycoproteins (12). A single site of N-linked is near the amino-terminal end of the collagen domain (Asn⁷⁰) of SP-D. Thus, each dodecamer contains up to 12 complex oligosaccharides near the amino-terminal hub of the dodecamer. To further exclude the effects of bacterial attachment to these oligosaccharides, we examined the inhibitory effect of a mutant SP-D (RrSP-Dala72) that lacks the sole consensus for N-linked glycosylation. Previous studies have shown that the molecule, which is assembled as dodecamers, is fully functional as a lectin and bacterial agglutinin (40). As shown in Fig. 6, RrSP-Dala72 showed dose-dependent inhibition of binding of K50-3OF (O3), with reproducible inhibition of adhesion at concentrations down to 0.75 μg/ml SP-D.

Relatively little is known about the role of the O-Ag in determining the innate immune response to Gram-negative bacteria. In this study, we found that the O-Ag, and the specific O-serotype of this polysaccharide, can influence the interactions of SP-D with LPS associated with the Gram-negative cell wall. Although binding of SP-D to the core oligosaccharide domain of R-LPS has been well described, we have shown for the first time that SP-D can also bind to mature, smooth forms of LPS containing mannose-rich O-polysaccharides. These interactions can in turn influence the adhesion of specific serotypes of smooth bacteria to lung epithelial cells.

When populations of LPS expressed by 16 Klebsiella isolates were separated by SDS-PAGE, the rough forms (R-LPS), which lack O-Ag, uniformly reacted with SP-D. In contrast, the reaction of SP-D with smooth forms (S-LPS) was restricted to the eight strains belonging to the O3 and O5 serotypes. There were also serotypic differences in SP-D-mediated bacterial agglutination attributable to differences in the LPS. For example, purified LPS of the O1 serotype was a much less effective inhibitor of SP-D-dependent agglutination of K50-3OF (O3) than LPS of the O3 serotype (Table I).

Our data indicate that the core oligosaccharide domain is necessary for binding to at least some components of Klebsiella LPS. The R-LPS forms of all four serotypes examined show similar interactions with SP-D ( Figs. 1–3), consistent with the conserved structure of the core domain among various serotypes. In addition, hydrolytic cleavage of the O-polysaccharide and core oligosaccharide from the S-LPS molecules abrogated SP-D binding to all components of K50-3OF (O3) and K21a/3 (O1) LPS, but unmasked epitopes for the binding of Ab to lipid A (Fig. 1). However, there are several observations that indicate that the O-polysaccharide domain determines these serotype-dependent differences in SP-D activity. For example, the dose-dependence of SP-D binding to R-LPS was similar for O3 and O1 S-LPS (Fig. 2), suggesting similar affinities for the conserved core oligosaccharides of both serotypes. In contrast, there were obvious differences in the dose-dependency of SP-D binding to the S-LPS forms, with negligible binding to the S-LPS forms of the O1 LPS. Because the size distributions of the S-LPS forms visualized by silver staining or blotting were similar and overlapping, we infer that the differential binding of SP-D to S-LPS from the O1 and O3 serotypes reflects differences in the composition of the repeating units rather than simply differences in the length of the O-polysaccharide chains.

Formally, there are at least three possible explanations for the serotype-dependent differences in the interaction of SP-D with S-LPS. First, the O-Ag of LPS from O1 and O4 (but not the O3 and O5) serotypes might somehow “repel” SP-D, or more effectively mask sites in the core. Second, the O3 and O5 Ags might directly mediate binding of SP-D to LPS in lieu of interactions with the core, which is masked by the attached O-polysaccharide. Third, the O-Ags of the O3 and O5 serotypes might bind weakly to SP-D, stabilizing residual interactions of the collectin with the subjacent core oligosaccharide. These possibilities could theoretically be distinguished by comparing the binding of SP-D to purified core and purified O-Ags devoid of contaminating core sugars. However, no biochemical or genetic methods have been described for the isolation of pure O-polysaccharide, devoid of core.

Nevertheless, there is strong circumstantial evidence for direct interactions of SP-D with the reactive O-Ags. Most importantly, the common structures of the O3 and O5 Ags—but not the O1 and O4 Ags—contain potential saccharide ligands for SP-D, particularly mannose. The O3 and O5 Ags consist exclusively of repeating units of α-1,2-linked mannose (39), and D-mannose is a known, albeit relatively weak, competitor of SP-D binding to a variety of glycoconjugates including LPS (18, 41). Mannose is also highly represented in a number of other complex glycoconjugates known to bind to SP-D. These include the high mannose oligosaccharides of the influenza A virus hemagglutinin (42), commercial preparations of yeast mannan (19), the gpA glycoprotein of Pneumocystis carinii (43), and the lipoarabinomannan of Mycobacterium tuberculosis (44). In contrast, all the repeating units of the O1 and O4 Ags lack mannose, but contain galactose (39), a sugar that reacts very weakly with SP-D (41). CRD-dependent binding to yeast mannan is of particular interest (19) because these complex polymers consist of repeating units of mannose in α-1,2, α-1,3, and α-1,6 linkage similar to the reactive O-polysaccharides. Notably, the interactions are efficiently inhibited by known competing sugars such as maltose or mannose. Contributions of β-glucans, which contaminate many commercial preparations of mannan, cannot be entirely excluded. However, mutant yeast devoid of β-glucan but expressing mannan still show significant binding (43). Thus, SP-D can interact with mannose-rich polysaccharides closely related to those expressed by the reactive O-serotypes.

There are currently no available isogenic mutants with switched O-Ag genes and the selective exchange of mannose-rich and galactose-rich O-polysaccharides. Nevertheless, it is impressive that for 16 independent isolates there is complete concordance between SP-D binding and known O-Ag structure, i.e., all isolates with mannose-rich O-Ags bind efficiently to SP-D, while none of the isolates from galactose-rich O-serotypes show this interaction. This likely occurs in the face of numerous other significant clonal variations in cell wall structure among the individual isolates of a given O-serotype.

As shown in Table I, the presence of a nonreactive O1-polysaccharide can markedly inhibit bacterial agglutination. In earlier experiments, we found that the extent of O-substitution of nonencapsulated Klebsiella (in retrospect, a reactive O3 serotype) was dramatically decreased when the cells were grown under vigorously aerated vs static conditions (21). This was accompanied by a marked decrease in the minimum amount of SP-D required to cause macroscopic aggregation. Thus, there is an apparent inverse relationship between the number of repeating units of a reactive O-Ag and the efficiency of SP-D-dependent agglutination. If the reactive O-Ag were to function as an independent, high-affinity ligand, increasing the numbers of mannose-containing repeating units might be expected to enhance or have no effect on agglutination. Although agglutination is a complex binding phenomenon, these observations suggest that both the reactive and nonreactive O-polysaccharides decrease SP-D binding to the core.

Given the combined observations, we speculate that any binding of SP-D to reactive O-Ags is relatively weak, consistent with the comparatively weak interactions of SP-D with D-mannose. If the interactions of SP-D with the O3 or O5 Ags are weak, it is possible that they are stabilized by residual interactions with the core oligosaccharide, even though the latter interactions are too weak to mediate SP-D binding in the presence of a nonreactive O-Ag. In this regard, recent computer docking studies suggest that SP-D can favorably interact with internal—as well as terminal—glucose residues (45). Thus, the presence of O-Ag should not preclude interactions of SP-D with glucose (or heptose) residues in the subjacent core oligosaccharide.

Our data also suggest a new potential mode of SP-D-mediated antibacterial host defense. In particular, SP-D can inhibit the binding of certain nonencapsulated Gram-negative bacteria to lung epithelial cells. Thus, the anti-adhesion activity of SP-D may represent another mechanism through which this important molecule functions in innate immunity. The minimum concentration of SP-D required to reproducibly inhibit the epithelial adhesion of the O1 serotype is close to some estimates of alveolar SP-D concentration (∼3 μg/ml) (46), whereas the observed minimum inhibitory concentration for the O3 serotype is severalfold lower. However, such calculations are based on estimates of hypophase volume and alveolar surface area, and reflect the contributions of many different alveolar and distal airway compartments. If normal SP-D concentrations are higher, as suggested by calculations based on some human lavage studies (46), serotypic differences might not be evident except in clinical settings where the local concentration of active SP-D is decreased (e.g., with smoking or lung injury). As indicated above, an additional complexity relates to the effects of “environmental conditions” on the length of reactive or nonreactive O-Ags expressed by bacteria at specific sites within the respiratory tract.

The effect of SP-D on epithelial adhesion is dependent on the lectin activity of the CRD and requires the multimerization of the SP-D subunits. In this regard, trimeric subunits are inactive as inhibitors of epithelial adhesion, but can block the inhibitory effect of dodecamers. Interactions of bacteria with collagenous sequences or N-linked sugars were excluded with the single-arm mutant and nonglycosylated dodecamers. Previous studies have shown that the single arm mutant is inactive as a bacterial or viral agglutinin (1, 40). In addition, the minimum concentration of SP-D required to agglutinate the O1 and O3 organisms was comparable to the minimum concentration required to significantly inhibit the adhesion K21a/3 (O1) and K50-3OF (O3) to A549 cells, respectively. Thus, we hypothesize that the inhibitory effect involves bacterial aggregation with interference of adhesin-mediated attachment of bacteria to the epithelial cell surface, a stage that is essential for the infectious process (12).

Interestingly, trimeric subunits significantly inhibited the antiadhesive activity of dodecamers at relatively low concentrations. This probably reflects competitive binding to SP-D ligands required for bridging interactions and bacterial agglutination. In any case, the phenomenon is of considerable potential clinical significance given that the relative amounts of collectin trimers can vary among individuals and in the setting of certain lung disorders (2, 47, 48). Thus, increased concentrations of trimers might decrease the effectiveness of the observed antiadhesive activity of SP-D. Recent studies have shown that recombinant, trimeric SP-D neck + CRD domains can neutralize certain fungi and respiratory viruses in animal models, suggesting its potential usefulness as a therapeutic agent (49, 50). However, our observation suggests a potential limitation of such strategies given that exogenous trimers could potentially favor colonization by other microbial pathogens.

SP-D is not the only constituent of the innate immune system that is able to recognize specific O-Ag structures. For example, the mannose-rich O-Ag of Salmonella was shown to be more potent than structures lacking the O-polysaccharide in activating the alternative pathway of complement (51, 52). Other potential mannose-binding molecules within the lung include the macrophage MR and SP-A. Although SP-A has been reported to interact with the lipid A domain, it has also been shown to interact with mannose-containing capsular glycoconjugates (6). Thus, a more complete profiling of the reactivity of various host defense lectins with specific serotypes of Gram-negative bacteria is indicated.

If SP-D or other lectins with similar specificity contribute to the lung’s defense against Klebsiella, then strains expressing SP-D reactive O-polysaccharides are expected to be isolated at lower frequency from patients with K. pneumoniae. In fact, the statistical analysis of the isolates bearing SP-D reactive and nonreactive O-Ags is entirely consistent with this possibility (Table II). Given the above, we speculate that in otherwise healthy individuals the levels of SP-D, together with other host defense lectins with similar saccharide selectivity, are sufficient for the clearance of Klebsiella from the respiratory tract. When the levels of these host defense molecules (or their functional oligomers) fall below a threshold, strains bearing nonreactive O-Ags can escape host recognition, contributing to the development of an infection. Recent evidence that SP-D is also expressed in the human upper respiratory tract (53), the usual initial site of bacterial colonization, makes this an attractive hypothesis. The finding that SP-D is expressed in low amounts at extrapulmonary sites (53), including the urinary tract, also raises the possibility of more generalized roles in antibacterial host defense.