Collectins are present in mammalian serum, lung secretions, and on various mucosal surfaces, where they play important roles in innate immunity against bacterial, viral, and fungal pathogens. Humans with heritable deficiencies of the serum collectin mannose-binding lectin (MBL) are at increased risk for viral, bacterial, and fungal infections (
20,
34). The lung collectin surfactant protein D (SP-D) is expressed in the respiratory tract and some other mucosal surfaces (
26). SP-D binds to a broad range of pathogens and has antimicrobial effects in vitro. Deletion of the SP-D gene in mice leads to pulmonary lipidosis, abnormal alveolar macrophage activation, deficient phagocytosis of bacteria by alveolar macrophages, and increased pulmonary inflammation after bacterial infection (
21,
23).
There is in vitro and in vivo evidence that the collectins play important roles in innate host defense against influenza viruses (
4,
29). SP-D−/− mice exhibit deficient clearance of influenza virus and markedly increased pulmonary inflammation after influenza virus infection (
24). We have demonstrated that the collectins differ significantly in their interactions with influenza virus. SP-D causes much greater viral aggregation and neutralization of viral infectivity than MBL (
11). The bovine serum collectin conglutinin causes greater neutralization of infectivity of influenza viruses than either MBL or SP-D (
10,
11). Collectins are composed of four domains, including an N-terminal domain involved in disulfide bond formation, a collagen domain, an alpha-helical neck domain involved in trimerization, and a C-terminal carbohydrate recognition domain (CRD) involved in calcium-dependent binding to carbohydrates. We have constructed chimeras containing the N-terminal and collagen domains of SP-D linked to the neck recognition domains and CRDs of conglutinin or MBL (SP-D/Cong
neck+CRD and SP-D/MBL
neck+CRD, respectively). A diagram of these constructs is shown in Fig.
1. These chimeras have greater viral neutralizing and aggregating activity than wild-type SP-D, MBL, or conglutinin (
10,
12,
36). The presence of the MBL or conglutinin CRD increased viral binding activity compared to that of SP-D, while the N-terminal and collagen domains of SP-D increased viral aggregation and neutrophil uptake of the virus compared to that of wild-type MBL or conglutinin.
MATERIALS AND METHODS
Reagents.
Formyl-methionyl-leucyl-phenylalanine, cytochalasin B, horseradish peroxidase-Type II, scopoletin, Ficoll, dextran, trypan blue, and the mannose-, glucose-, or galactose-bovine serum albumin (BSA) conjugates were purchased from Sigma Co. (St. Louis, Mo.), and Hypaque was obtained from Winthrop Pharmaceuticals (Des Plaines, Ill.). Dulbecco's phosphate-buffered saline (PBS) was purchased from Flow Laboratories (Costa Mesa, Calif.).
Collectin preparations.
Table
1 summarizes the various recombinant collectin preparations used in this paper. All preparations were expressed in CHO-K1 cells as previously described (
3,
8). Wild-type recombinant collectins included recombinant rat and human SP-Ds (RrSP-D and RhSP-D), recombinant bovine conglutinin (RbConglutinin), and recombinant human MBL (RhMBL). The detailed characterization of these preparations was reported in the references listed in Table
1. RbConglutinin, RrSP-D, and RhSP-D consisted of the purified dodecamer fraction, which was isolated by gel filtration chromatography under nondenaturing conditions. Natural bovine serum conglutinin and human serum MBL were also used in some experiments (as noted in Results), purified as described previously (
6), and graciously provided by Jens Jensenius (University of Aarhus, Aarhus, Denmark). Unlike the other collectins used in these studies, natural and recombinant human MBL are composed principally of octadecamers (
16,
33).
Two mutated recombinant forms of RrSP-D were also used in these experiments: RrSP-D
CDM and RrSP-D
ala72. RrSP-D
CDM is a trimeric form of SP-D that lacks the entirety of the collagen domain (
28). RrSP-D
CDM was the gracious gift of Dennis Voelker (National Jewish Medical and Research Center, Denver, Colo.). RrSP-D
ala72 was generated by substituting alanine for serine at position 72, thereby disrupting the consensus for N-linked glycosylation at Asn 70. RrSP-D
ala72 dodecamers were prepared as previously described (
2).
Two chimeric collectins were also employed (see Fig.
1). The first consisted of the N-terminal and collagen domains of RrSP-D joined to the neck domain and CRD of conglutinin (termed SP-D/Cong
neck+CRD). The SP-D/Cong
neck+CRD chimera was constructed by thermal cycling with overlap extension and was expressed in CHO-K1 cells (
10). The secreted protein was isolated by GlcNAc affinity chromatography, and the SP-D/Cong
neck+CRD dodecamers were purified by gel filtration chromatography under nondenaturing conditions. The second chimera consists of the N-terminal and collagen domains of recombinant human SP-D (RhSP-D) joined to the neck domain and CRD of human MBL (SP-D/MBL
neck+CRD) (
36). The secreted protein was isolated from the medium by saccharide affinity chromatography on mannose-agarose. The dodecameric and trimeric forms of the SP-D/MBL
neck+CRD chimera were then purified by gel filtration chromatography.
Biotinylation of collectins (or antibodies) was performed by incubation of collectins with Immunopure NHS-LC-Biotin (Pierce, Rockford, Ill.), at a ratio of 2:1 of biotin to collectin by weight, for 2 h at room temperature in the dark. Excess biotin was then removed by dialysis overnight. Biotinylation did not alter the ability of the collectins to induce bacterial aggregation or to inhibit hemagglutination activity of influenza virus (data not shown).
Bacterial preparations.
Bacterial strains were selected to facilitate comparisons with the published literature (see Table
2). Rough strains of
Escherichia coli have been extensively used in previous studies of SP-D interactions with gram-negative bacteria and LPS (
9,
22). Likewise,
Salmonella enterica serovar Typhimurium has been used in studies of MBL-bacterial interactions. Both gram-negative organisms offer the advantage of a number of well-characterized mutants that vary in the structure of their LPS molecules. Unlabeled and fluorescein isothiocyanate- or Rhodamine-labeled
E. coli (K-12 strain) cells were obtained from Molecular Probes (Eugene, Oreg.), as were Zymosan particles.
Streptococcus pneumoniae strains were the gracious gift of Alan J. Parkinson (Centers for Disease Control Arctic Investigation Laboratory, Anchorage, Alaska). The TV119, TV160, and SL1102 strains of
S. enterica serovar Typhimurium were kind gifts of H. Nikaido (University of California, Berkeley) and were grown in our laboratory. These are rough strains with known LPS oligosaccharide structure (
17). The TV119 strain is also known as the Ra strain and has an outer-core oligosaccharide terminating in GlcNAc. The TV160 and SL1102 mutants contain capsular polysaccharide terminating in galactose or inner-core polysaccharides (i.e., 2-keto-3-deoxyoctonic acids), respectively (
17). Table
2 summarizes the gram-negative bacterial strains used in this paper.
Live E. coli or S. enterica serovar Typhimurium cells for experiments were prepared by growing bacteria overnight on Luria-Bertani agar, selecting several colonies the morning of experiments, and resuspending these in Luria-Bertani medium. Bacteria were incubated until an optical density at 600 nm (OD600) of ∼0.6 was reached (usually ∼2 h at 37°C). Bacteria were then counted with a Hausser counting chamber and were resuspended in PBS. S. pneumoniae strains were grown in sheep blood agar (Remel, Lenexa, Kans.) in low oxygen and then in tryptic soy broth with 10% horse serum. Live S. pneumoniae cells were grown and counted on the day of experiments as described above.
FITC labeling of formalin-fixed bacteria was carried out as previously described (
14). The FITC stock was prepared at 1 mg/ml in 1 M sodium carbonate, pH 9.6. Concentrated bacterial or yeast stocks were incubated with FITC (10:1 mixture by volume of bacteria in PBS with FITC stock) for 1 h, followed by dialysis of the mixture for 18 h in PBS. FITC labeling of live bacteria was performed in the same manner except that the bacteria were not dialyzed but were washed in PBS by repeated centrifugation to allow their immediate use in the neutrophil uptake assays. Prior to all assays using bacteria or yeast particles, the suspensions of these organisms were gently sonicated to disrupt aggregates, as monitored by light microscopy.
Assessment of binding of collectins to bacteria.
Binding of collectins to bacteria or influenza virus was tested by enzyme-linked immunosorbent assay (ELISA) as described previously (
9). Suspensions of formalin-fixed bacteria (∼1 μg [dry weight] of bacteria/ml) were allowed to dry onto 96-well plates, followed by washing and incubation with biotinylated collectins. The presence of bound collectins was detected by using streptavidin conjugated to horseradish peroxidase and 3,3′,5,5′-tetramethylbenzidine substrate (Kierkegaard and Perry, Gaithersburg, Md.). The reaction was stopped by using 1 N H
2SO
4. Optical density was measured with an ELISA reader. A modification of the ELISA was also employed to exclude artifacts resulting from formalin treatment or drying of bacteria. Freshly isolated, live bacteria (∼3 × 10
6 bacteria/ml) were allowed to settle on ELISA plates overnight at 4°C in liquid buffer, followed by centrifugation of plates at 2,500 rpm in a Sorvall RT7 centrifuge prior to assay to maximize bacterial adhesion. Each experimental condition was tested in duplicate or triplicate wells and was averaged for analysis. The background binding of collectins to wells coated with BSA alone was subtracted from the experimental values. The background binding levels were in the range of 0.03 to 0.2 at OD
450. More than three separate ELISAs were performed for each variable tested.
Assessment of bacterial aggregation.
Bacterial aggregation was examined by several methods. First we assessed aggregation by measuring changes in light transmission through suspensions of bacteria after addition of various concentrations of collectins as previously described (
9) by using an SLM/Aminco 8000C (SLM Instruments, Urbana, Ill.) spectrofluorometer. Viable bacteria were grown to an OD
600 of ∼0.6 on the day of assay and then were counted and resuspended in PBS at a final concentration of 3 × 10
6 bacteria/ml. Another method for assessing bacterial aggregation involved fluorescence microscopy by using FITC- or Rhodamine-labeled bacteria (
9). For this assay, higher concentrations of bacteria (3.3 × 10
8 bacteria/ml) were used in order to maximize the detection of aggregates and to optimize the binding of the bacteria to neutrophils (see below).
Bacterial colony formation was measured after incubation of live bacteria for aggregation experiments. This was done by taking samples of bacteria that had been incubated with collectins for 30 min at 37°C during the aggregation assay, diluting these bacterial samples in fresh, sterile PBS at 1:1,000 dilution, and plating the samples on LB agar. Plating was performed by taking 50 μl of samples and dispersing on the agar by using sterile glass beads. Colonies were counted after incubation for 24 h at 37°C. Results were expressed as percentages of colonies formed in collectin-treated samples compared to those of bacteria treated in an identical manner with control buffer (e.g., incubated in aggregation assay and plated in identical manner).
Neutrophil preparation.
Neutrophils from healthy volunteer donors were isolated to >95% purity as previously described by using dextran precipitation followed by a Ficoll-Hypaque gradient separation for removal of mononuclear cells and hypotonic lysis to eliminate contaminating erythrocytes (
13). Cell viability was >98% as determined by trypan blue staining, and cells were used within 2 h of isolation. Cells were subsequently washed three times and resuspended in PBS.
Measurement of bacterial uptake by neutrophils.
Bacterial uptake by neutrophils was measured by incubating FITC- or Rhodamine-labeled bacteria with neutrophils, followed by evaluation of cell-associated fluorescence by using a flow cytometer as described previously (
9). Bacteria were preincubated with collectins in the manner described above for evaluation of bacterial aggregates under fluorescent microscopy (i.e., 72 μg of bacteria/ml and various concentrations of collectins for 30 min at 37°C). For experiments using live bacteria, bacteria were grown up on the day of assay as described above, incubated with FITC as described for 1 h followed by washing, and then directly incubated with neutrophils. Neutrophils were then incubated with these bacterial samples (5 × 10
5 neutrophils and 1.2 × 10
7 bacteria at a ratio of 1 neutrophil to 24 bacteria in 135 μl) for 1 h at 37°C and washed twice in fresh PBS buffer and fixed with 2% paraformaldehyde. The mean cell-associated fluorescence of 2,000 cells was measured by using flow cytometry. These assays were performed after incubation with 0.2 mg of trypan blue/ml for 3 min to eliminate extracellular fluorescence (
9).
Statistics.
Statistical comparisons were made using Student's paired t test.
DISCUSSION
In these studies, SP-D was found to have significantly greater ability than MBL or conglutinin to aggregate and/or bind to several bacterial strains. Initial experiments were carried out with formalin-fixed bacterial strains. The use of formalin-fixed strains allowed standardization and also excluded possible effects of bacterial proteases on collectin activities (e.g., proteolytic degradation of conglutinin [
19]). However, important observations were confirmed by using freshly isolated, live bacteria to exclude artifacts resulting from fixation.
We examined interactions of these collectins with gram-negative bacteria that have known LPS carbohydrate structure and interactions with serum collectins and/or SP-D (see Table
2). These strains are useful because they allow comparisons with published literature and assessment of the contribution of LPS-associated carbohydrates. Both the
E. coli K-12 strain and
S. enterica serovar Typhimurium TV119 strain have been shown to be susceptible to complement-dependent lysis by MBL and are among the bacterial strains to which MBL binds most strongly (
5,
17,
18). It is notable, therefore, that SP-D aggregated and bound to these strains more strongly than MBL.
It is likely that the monosaccharide-binding affinities of the collectins play an important role in their attachment to these gram-negative bacteria (e.g., note the failure of SP-D or MBL to bind to the TV160 or SL1002 strains of S. enterica serovar Typhimurium); however, other factors may be important as well (see below). Because MBL has higher affinity for GlcNAc than SP-D, our initial hypothesis was that the reduced ability of MBL to aggregate or bind to E. coli K-12 and S. enterica serovar Typhimurium TV119 resulted from differences in the N-terminal and collagen domains of the two molecules. Both of these bacterial strains contain LPS-associated sugars (e.g., GlcNAc) that should allow strong attachment of the MBL CRD. SP-D/MBLneck+CRD dodecamers did bind to and aggregate the bacteria to a greater extent than comparable concentrations of MBL. These results were dependent on formation of dodecamers, since the trimeric fraction of SP-D/MBLneck+CRD did not have increased bacterial binding or aggregating activity compared to that of MBL. SP-D/MBLneck+CRD dodecamers also suppressed colony formation of live bacteria to a greater extent than RhMBL. Although further studies are needed to determine the mechanism of this effect (i.e., aggregation versus other effects on bacterial viability), it provides further evidence of enhanced interaction of the chimera compared to that of MBL.
On the basis of prior studies with influenza virus (
35,
36), we anticipated that SP-D/MBL
neck+CRD dodecamers would have enhanced bacterial aggregating and binding activity compared to that of wild-type SP-D as well. However, SP-D/MBL
neck+CRD dodecamers had approximately the same bacterial aggregating activity as SP-D and bound the bacteria less extensively than SP-D. It is unclear why SP-D/MBL
neck+CRD dodecamers bound bacteria less strongly than RhSP-D dodecamers while having similar aggregating activity, although it is possible that different interactions can occur in suspension during the aggregation assay compared to that for the ELISA, which was in solid phase.
Recombinant or native conglutinin and SP-D/Cong
neck+CRD had markedly reduced ability to bind to
S. enterica serovar Typhimurium TV119 and other bacteria compared to that of SP-D. Despite strong homologies between conglutinin and SP-D (
25), the CRDs of these proteins appear to have significant differences in biological activity. The CRD of conglutinin binds to iC3b and mediates conglutination (
37). Also, the CRD of conglutinin has significantly greater ability to bind to and neutralize influenza viruses than does of SP-D (
10,
12). In vivo, replacement of wild-type SP-D in SP-D-gene-deleted (SP-D−/−) mice with the SP-D/Cong
neck+CRD chimera through transgenic expression in the lung does not restore all normal activities of SP-D (
37). These SP-D−/−, SP-D/Cong
neck+CRD overexpressing mice have restored ability to process pulmonary phospholipids and to inhibit pulmonary influenza infection but develop macrophage abnormalities and emphysema, like SP-D−/− mice. Results of the present paper provide further evidence of functional differences between the CRD of conglutinin and SP-D. Of note,
S. enterica serovar Typhimurium TV119 contains LPS with an O-polysaccharide that terminates in a monosaccharide (GlcNAc), for which MBL and conglutinin have higher affinity than SP-D. Hence, the reduced ability of conglutinin, SP-D/MBL
neck+CRD, or SP-D/Cong
neck+CRD to bind to these bacteria compared to binding of SP-D is striking and suggests that the identity of the terminal LPS monosaccharide does not alone determine collectin binding.
A variety of factors could account for differences in the abilities of the collectins to bind to specific bacterial strains apart from the terminal sugar present on the bacterial coat proteins. First, it has been demonstrated that collectins can bind to nonterminal sugar residues in a polysaccharide (
1). Furthermore, the orientation of a polysaccharide relative to the protein backbone can markedly affect binding. Conglutinin can bind to oligosaccharides on RNase only when the protein is reduced (
30). The ability of conglutinin to bind to a high mannose oligosaccharide on iC3b, whereas MBL cannot, also appears to depend on the relationship of the oligosaccharide to the underlying protein backbone (
31). The density or topographical arrangement of LPS on the surface of some bacteria may strongly affect collectin binding, as suggested by Devyatyarova-Johnson et al. (
5) with respect to binding of MBL to
Salmonella and
Neisseria strains. The ability of a collectin molecule to bind to many sugar moieties simultaneously on the surface of a pathogen can greatly affect binding affinity. Finally, it is also possible that the CRD of SP-D contains ancillary binding sites (apart from the principle calcium and carbohydrate binding pocket) that contribute to its enhanced ability to bind to certain bacteria, as suggested by comparisons of the crystal structures of the SP-D and MBL CRDs (
7). The fact that binding of SP-D to
S. enterica serovar Typhimurium TV119 was less readily inhibited by monosaccharides than binding of SP-D/MBL
neck+CRD dodecamers (Fig.
4C) supports the hypothesis that ancillary binding interactions may contribute to binding of SP-D to bacteria.
In any case, our studies indicate that significant differences exist in the ability of neck domains and/or CRDs of SP-D, MBL, and conglutinin to mediate binding to some strains of bacteria. Prior studies demonstrated differences in the ability of SP-D/MBL
neck+CRD and SP-D/Cong
neck+CRD to interact with influenza virus compared to that of wild-type SP-D, as summarized in Table
6 (
10,
35,
36). It is notable, however, that these prior studies showed clearly increased ability of the chimeras to bind to, aggregate, or neutralize influenza viruses compared to that of SP-D (or compared to that of wild-type MBL and conglutinin). Hence, the requirements for optimal binding of a collectin to influenza virus or bacteria appear to differ. Further studies with chimeric SP-D molecules containing smaller portions of the CRDs of serum collectins should clarify whether or not specific segments of the CRD are responsible for differences in bacterial or viral binding activity of SP-D and serum collectins.
Our studies also demonstrate that the N terminus and collagen domains of SP-D have significant functional differences from the comparable domains of MBL or conglutinin. Dodecamers of the SP-D/MBLneck+CRD or SP-D/Congneck+CRD chimera bound to and aggregated all of the bacterial strains tested to a greater extent than MBL or conglutinin, respectively. As noted, SP-D/MBLneck+CRD caused significantly greater increases in bacterial uptake by neutrophils than RhMBL. Similar results were obtained from comparing the ability of SP-D/Congneck+CRD and conglutinin to increase neutrophil uptake of E. coli (data not shown). We hypothesize that the greater bacterial or viral aggregating activity of SP-D/MBLneck+CRD dodecamers compared with that of MBL indicates that the much larger collagen domain of SP-D allows greater opportunities for cross-linking of microbial particles than the more compact domain of MBL.
The greater ability of SP-D/Cong
neck+CRD to aggregate influenza virus (
10) or bacteria compared with that of conglutinin is less readily explained given greater similarity between the N-terminal and collagen domains of SP-D and conglutinin. Our results obtained with the RrSP-D
ala72 mutant version of SP-D indicate that the N-linked sugar on SP-D is not responsible for the enhanced interactions of SP-D/Cong
neck+CRD with influenza virus (
10,
11) or bacteria (this paper). In fact, removal of the N-linked sugar from SP-D increased its aggregating activity. This finding requires further study but suggests that the presence of the N-linked sugar in the collagen domain of SP-D may limit mobility of the arms of multimeric SP-D molecules. Other differences in the N-terminal domains of SP-D and conglutinin include the presence of interchain disulfide cross-links, the interruption in the collagen sequence, and the shorter length of the collagen domain of conglutinin (
32). Further analysis of the structural features of the N terminus and collagen domains of SP-D, MBL, and conglutinin that are responsible for enhanced microbial aggregating and opsonizing activities of SP-D, or chimeras containing the N terminus and collagen domains of SP-D, are warranted.
In summary, we demonstrate that the serum collectins and SP-D differ significantly in their interactions with bacteria. Conglutinin and MBL show significantly less bacterial binding and aggregating activity than SP-D. These differences are, in part, secondary to differences in binding properties of the SP-D CRD compared to those of the CRDs of MBL or conglutinin but also reflect differences in the N-terminal and collagen domains. Although most of the experiments described in this paper involved laboratory-derived gram-negative bacterial strains, similar differences in bacterial binding activity were found with clinically relevant strains of S. pneumoniae. The enhanced binding activity of SP-D/MBLneck+CRD dodecamers compared to that of RhMBL may also allow this molecule to interfere with complement activation by MBL. Further studies will address this question, since such complement activation could be deleterious in certain settings.