Monoclonal antibody-assisted structure-function analysis of the carbohydrate recognition domain of surfactant protein D

Abstract

Surfactant protein D (SP-D) plays important roles in host defense against a variety of pathogens including influenza A virus (IAV). Ligand binding by SP-D is mediated by the trimeric neck and carbohydrate recognition domain (NCRD). We used monoclonal antibodies (mAbs) against human SP-D and a panel of mutant collectin NCRD constructs to identify functionally and structurally important epitopes. The ability of SP-D to bind to IAV and mannan involved partially overlapping binding sites that are distinct from those involved in binding to the glycoprotein-340 (gp-340) scavenger receptor protein. A species-specific motif (D324,D325,R343), which has been implicated in the specific binding of several ligands, contributes to recognition by mAbs that block antiviral or mannan binding activity. D325, in particular, is involved in the epitopes of these blocking mAbs. Conversely, the interspecies substitution of arginine for Lys343 in the rat NCRD (rK343R) conferred binding to two of the mAbs. The single site substitution of alanine for R349 or E347 resulted in highly selective alterations in mAb binding and caused decreased antiviral activity. Mutations at Glu333 (E333A), Trp340 (W340F), and Phe335 (F335A), which abrogated antiviral activity, were associated with decreased binding to multiple blocking mAbs, consistent with critical structural roles. More conservative substitutions at 335, which showed a significant increase in neutralization activity, caused selective loss of binding to one mAb. The analysis reveals, for the first time, an extended binding site for IAV; calcium-dependent antiviral activity involves residues flanking the primary carbohydrate binding site as well as more remote residues displayed on the carbohydrate recognition domain surface.

surfactant protein d (SP-D) is present in lung lining fluids and a variety of other mucosal locations where it participates in binding and inhibiting a wide range of infectious organisms, including bacteria, fungi, and viruses (27). SP-D is a member of the collectin family of innate defense proteins that contain a structurally important collagen domain and trimeric neck and carbohydrate recognition domains (termed NCRDs from here on) that are involved in calcium-dependent binding to specific carbohydrate epitopes on microorganisms or mammalian cells.

We and others have studied the interactions of SP-D with influenza A viruses (IAVs) (11, 14, 16, 20, 21). Mice lacking SP-D due to gene deletion exhibit more severe illness, higher viral loads, and greater inflammatory response when infected with human strains of IAV. Inhibition of IAV by SP-D is determined mainly by the presence of high mannose oligosaccharides on the viral hemagglutinin (HA) (13, 20, 23). SP-D also plays an important role in inhibiting inflammatory responses triggered by LPS and bacteria (15, 19, 27). Finally, SP-D plays an important role in maintenance of surfactant lipid homeostasis in vivo (1).

Monoclonal antibodies (mAbs) directed against the NCRD of SP-D have proved useful in determining functionally important regions of the protein and demonstrating the role of cross-linking of NCRD trimers in antiviral activity (12, 22). We (22) have previously reported that a panel of mAbs directed against the NCRD of SP-D can be grouped into ones that inhibit antiviral activity of SP-D against IAV and others that do not. The antibodies that do not inhibit viral binding or antiviral activity include two that are directed against the neck domain of SP-D (245-01 and 245-02). The latter mAb recognizes not only the neck, but also some portion of the carbohydrate recognition domain (CRD) because it retains the ability to bind to a preparation containing only the CRD of SP-D (whereas 245-01 loses binding altogether to this preparation). Two other mAbs that do not block antiviral activity of full-length SP-D strongly increase the antiviral activity of NCRD trimer preparations of SP-D by cross-linking and enhancing binding of the NCRD to the virus. Finally, one other nonblocking antibody (246-06) does not block or increase activity of full-length SP-D or NCRDs.

In this paper, we more fully characterize the effect of the various antibodies on antiviral activity as well as binding to mannan and scavenger receptor protein glycoprotein-340 (gp-340). We also characterize epitopes of the mAbs by comparing their binding with a panel of similarly prepared wild-type and mutant NCRD trimers derived from SP-D. Finally, we compare the effect of specific mutations of the NCRD on antiviral activity. Certain NCRD sites are shown to be critical in modulating antiviral activities of SP-D and antibody recognition. Some combinations of mutations are shown to restore epitopes of human SP-D lost in rodent SP-Ds or in single site mutants of human SP-D. Other mutations are shown to cause marked structural alterations based on loss of binding to multiple mAbs. Through these studies, we demonstrate that residues adjacent to the lectin site, as well as more distant residues on the CRD surface, make up an extended binding region for viral oligosaccharides.

MATERIALS AND METHODS

Virus preparation.

IAV was grown in the chorioallantoic fluid of 10-day-old chicken eggs and purified on a discontinuous sucrose gradient as previously described (10). The virus was dialyzed against PBS to remove sucrose, aliquoted, and stored at −80°C until needed. Philippines 82/H3N2 (Phil82) strain was kindly provided by Dr. E. Margot Anders (Univ. of Melbourne, Melbourne, Australia) (7). Postthawing, the viral stocks contained ∼5 × 10⁸ plaque-forming units/ml.

Collectin preparations.

Dodecamers of wild-type recombinant human SP-D were used as a control and were expressed in CHO cells and purified as previously described (8). Trimeric wild-type and mutant NCRD fusion proteins were expressed in Escherichia coli (5, 6). All fusion proteins contain an identical NH₂-terminal His tag, used for initial purification, and an internal S protein binding site permits detection using S protein-horseradish peroxidase (HRP) as previously described (5). Wild-type trimeric human and rat NCRDs are designated hNCRD and rNCRD, respectively. Mutant NCRDs are identified by the species of origin (h or r) followed by the position(s) of substitution in the mature proteins.

NCRD fusion proteins were extracted from inclusion bodies, refolded and oligomerized by dialysis, and purified by nickel affinity chromatography, and the predominant trimeric species was isolated by gel filtration chromatography (5, 6). All NCRDs migrated as a single major band of the appropriate size on SDS-PAGE with the expected decrease in mobility on reduction, consistent with the formation of normal intrachain disulfide bonds (data not shown). Except where indicated, mutant NCRDs showed retention of some or all of the calcium-dependent carbohydrate binding activities of the native protein. The endotoxin level of all SP-D preparations was approximately 0.1–0.5 EU/ml (Limulus Lysate Assay; Cambrex, Walkersville, MD).

mAbs.

mAbs 245-01, 245-02, and 246-02 through 246-08 were raised against SP-D by inoculating mice with 10 μg/ml human SP-D as previously described (17).

Binding of mAbs to SP-D or NCRD.

SP-D preparations were diluted in coating buffer to a concentration of 2 μg/ml and coated on ELISA plates overnight, followed by washing and addition of mAbs. The final concentration of mAbs used for the ELISA assay was 1 μg/ml. The binding assays were carried out in PBS containing calcium and magnesium. Bound mAbs were detected with HRP-conjugated donkey anti-mouse antibodies labeled followed by 3,3′,5,5′-tetramethylbenzidine (TMB) peroxidase. Optical density at 450-nm values were measured on a POLARstar OPTIMA plate reader (BMG Labtech, Durham, NC).

Binding of NCRDs to IAV.

Binding of NCRD fusion proteins to IAV was measured as previously described (5) by use of the S protein binding site on the fusion tag of the NCRD. In brief, IAV (Phil82 strain) was coated onto the surface of ELISA plates, and, following washing, NCRDs were added. After incubation and washing, S protein-HRP was added, and peroxidase activity was measured.

HA inhibition assay.

The studies involving blood cells from volunteer donors were approved by the Boston University School of Medicine institutional review board. Hemagglutination (HA) inhibition was measured by serially diluting collectins or other host defense protein preparations in round-bottom 96-well plates (Serocluster U-Vinyl plates; Costar, Cambridge, MA) using PBS containing calcium and magnesium as a diluent (11). After adding 25 μl of IAV, giving a final concentration of 40 HA U/ml or 4 HA U/well, the IAV/protein mixture was incubated for 15 min at room temperature, followed by addition of 50 μl of a type O human erythrocyte suspension. The minimum concentration of protein required to fully inhibit the HA activity of the viral suspension was determined by noting the highest dilution of protein that still inhibited HA. Inhibition of HA activity in a given well is demonstrated by absence of formation of an erythrocyte pellet. If no inhibition of HA activity was observed at the highest protein concentration used, then the value is expressed as greater than the maximal protein concentration.

Fluorescent focus assay of IAV infectivity.

Madin-Darby canine kidney (MDCK) cell monolayers were prepared in 96-well plates and grown to confluence. These layers were then infected with diluted IAV preparations for 45 min at 37°C in PBS and tested for presence of IAV-infected cells after 7 h using a mAb directed against the influenza A viral nucleoprotein (provided by Dr. Nancy Cox, Centers for Disease Control and Prevention, Atlanta, GA) as previously described (9). IAV was preincubated for 30 min at 37°C with collectins or control buffer, followed by addition of these viral samples to the MDCK cells. Where indicated, collectins were first incubated with mAbs before addition to IAV.

Statistics.

Statistical comparisons were made using Student's paired, two-tailed t-test or ANOVA with post hoc test (Tukey). ANOVA was used for multiple comparisons with a single control.

RESULTS

Comparison of effects of mAbs on the ability of wild-type human SP-D to bind to mannan, IAV, or gp-340.

As shown in Fig. 1, mAbs 246-02, 246-03, and 246-07 strongly inhibited binding to mannan. We (22) have reported that these mAbs also inhibit antiviral activity of SP-D. mAbs 246-05 and 246-06 inhibited binding to mannan at 10 but not 1 μg/ml. mAbs 245-01 and 245-02 that recognize the neck region of SP-D did not inhibit binding to mannan, and 246-04 and 246-08 caused only slight inhibition (not dose-related for 246-08). Overall, these results indicate a correlation between ability of mAbs to inhibit binding to mannan with their inhibition of antiviral activities of SP-D: mAbs 246-02, 246-03, and 246-07 strongly inhibit both, whereas 246-04 and 246-08 and the neck mAbs do not substantially inhibit either activity. However, there were two discrepancies. First, 246-05, which strongly inhibits antiviral activity, only inhibited mannan binding at the higher antibody concentration. Second, 246-06 did not inhibit antiviral activity but inhibited binding to mannan at the higher antibody concentration.

Fig. 1.

Effect of monoclonal antibodies (mAbs) on binding of surfactant protein D (SP-D) dodecamers to mannan or glycoprotein-340 (gp-340). Mannan (A) or gp-340 (B) were bound to ELISA plates. Human SP-D dodecamers were preincubated with the indicated mAbs at either 1 or 10 μg/ml as indicated. The results are means ± SE of 2 experiments. Additional experiments (data not shown) were done in which the assay was carried out in EDTA-containing buffer, which prevented binding of SP-D to either mannan or gp-340. Preincubation of SP-D with maltose prevented binding to mannan but not to gp-340 (data not shown). *A specific concentration of mAb caused significant reduction of SP-D binding to mannan or gp-340 (P < 0.05).

The ability of mAbs to inhibit binding to gp-340 did not consistently correlate with these other properties. For instance, as reported (26), mAb 245-01 inhibits binding to gp-340 (without inhibiting antiviral activity or binding to mannan). The 246-07 mAb completely inhibited binding to mannan at low concentrations but inhibited binding to gp-340 only partially and at the higher mAb concentration. Two antibodies that strongly inhibit antiviral activity and mannan binding also inhibit binding to gp-340 (246-02 and 246-03). The higher concentration of mAb 246-06 caused stronger inhibition of binding to mannan than to gp-340. Finally, 246-04, 246-08, and 245-02 do not inhibit binding to any of the ligands. Note that preincubation of SP-D with maltose did not alter its binding to gp-340 or recognition by the mAbs of SP-D bound to gp-340 (data not shown). In contrast, preincubation eliminated binding of SP-D to mannan, and EDTA abolished binding of SP-D to either mannan or gp-340 (data not shown). Hence, binding of SP-D to gp-340 is calcium-dependent but not altered by occupancy of the lectin site with maltose. Overall, these results indicate that the mechanisms of binding to mannan and IAV are similar but not identical and that SP-D has distinct mechanisms of attachment to gp-340.

Contributions of a species-specific motif to antibody binding.

Rat SP-D NCRD resembles human SP-D NCRD in its poor binding to IAV but has greater apparent affinity for phosphatidylinositol (PI) and LPS (24). These differences are in part attributable to differences in hydrophilic residues that flank calcium at the primary carbohydrate binding site. SP-Ds of most species have N324,N325 and K343, whereas human SP-D is characterized by D324,D325 and R343. We have previously shown that the difference at 325 contributes to preferential recognition of N-acetyl-mannosamine by human SP-D (6) and that the residue at 343 contributes to enhanced recognition of LPS and PI by rat and mouse NCRDs (4, 24).

Although human SP-D was used to generate the panel of mouse mAbs, certain mAbs can bind to rodent NCRDs (22). For instance, the NCRD of rat SP-D binds to mAb 246-03 and, to a lesser degree, to 246-08. Accordingly, we used combinatorial, reciprocal mutagenesis to examine the contributions of the distinctive 324,325,343 motif to differences in antibody binding.

A rat triple mutant (rN324D+N325D+K343R, designated DDR) and rK343R, which were designed to resemble the human protein, showed a marked increase in binding to mAbs 246-05 and 246-06 (Fig. 2A). By contrast, rN324D+N325D, (designated DD) showed no increase in binding to these mAbs and lost affinity for 246-03. These findings confirm our prior report that R343 in human SP-D is involved in binding to 246-05 (2). Furthermore, they demonstrate that R343 is involved in binding to 246-06 and suggest that N324 and N325 somehow contribute to cross-reactivity of rodent proteins with 246-03.

Fig. 2.

Binding of mAbs to trimeric neck and carbohydrate recognition domains (NCRDs) derived from rat or human SP-D. The indicated NCRD preparations derived from rat SP-D (A) or human SP-D (B) were coated onto ELISA plates, and binding of mAbs were tested as described. Results are means ± SE of 3 experiments and are expressed as percentage of binding to wild-type human NCRD (hNCRD) control. In A, mAbs 246-05 and 246-06 showed significantly greater binding to rat triple mutant (rN324D+N325D+K343R; rDDR) and rK343R than to wild-type rat NCRD (rNCRD; #P < 0.05). mAb 246-03 had significantly reduced binding to rN324D+N325D (rDD). In B, hD325N had significantly decreased binding to 246-03, 246-05, 246-06, and 246-07; human double mutant (hD324N+D325N; hNN) had reduced binding to 246-02, 246-03, 246-05, and 246-07; human triple mutant with complete rodent motif (hD324N+D325N+R343K; hNNK) had reduced binding to 246-02, 246-03, 246-05, 246-06, and 246-07. *Significantly reduced mAb binding (P < 0.05).

We also took the reverse approach and studied the ability of the mAbs to bind to human NCRDs with interspecies substitutions at the 324, 325, and/or 343 positions. The hD324N mutation did not affect binding to any of the mAbs (Fig. 2B); however, hD325N decreased binding to all of the blocking mAbs (246-03, 246-05, 246-06, and 246-07) except for 246-02. A double mutant (hD324N+D325N, designated hNN) caused some decrease in binding to 246-02 but was otherwise similar to hD325N. However, a triple mutant with the complete rodent motif (hD324N+D325N+R343K, designated hNNK) caused more profound suppression of binding to 246-05 and some suppression of binding to 246-02. We (2) have previously reported that the single mutant, R343K, has reduced binding to 246-05. These observations further confirm the importance of R343 in binding to mAb 246-05. More broadly, the results indicate the importance of D325 in the antigenic structure of hNCRD and confirm its importance in binding to a variety of key ligands (6). D325 contributes to a distinctive ridge on the outer edge of the binding surface of hNCRD. Note that none of these changes affected binding to 246-04, 246-08, or the neck mAbs (data not shown).

Figure 3A shows the effects of mutant versions of rat and human SP-D NCRDs on binding to IAV as assessed by ELISA. All of these NCRDs contain S protein binding sites on the NH₂ terminus to allow for uniform detection of bound NCRD using S protein-HRP conjugates. Results obtained with hR343V and hR343A are shown for comparison. Whereas these mutants had markedly increased viral binding (as previously reported), hR343K, rDDR, hNN, and hNNK did not.

Fig. 3.

Binding, neutralization, and hemagglutination (HA) inhibition of influenza A virus (IAV) by wild-type and mutant NCRDs. A: the Philippines 82/H3N2 (Phil82) strain of IAV was coated onto ELISA plates, and binding of the indicated NCRDs was tested using S protein-horseradish peroxidase (HRP) for detection. Results are means ± SE of 3 experiments. Binding of R343V or R343A was significantly greater than wild-type hNCRD (**P < 0.01). OD450, optical density at 450-nm values. B: viral neutralization was tested by preincubating IAV with 20 μg/ml indicated NCRDs followed by measurement of infectious virus by the fluorescent focus assay (see materials and methods). Results are means ± SE of 3 or more experiments and are expressed as percentage of control infectious foci (control being untreated virus). hR343V and hR343A caused significant inhibition of viral infectivity (*), NNK caused significant increase in viral infectivity (#), and the other NCRDs did not significantly alter infectivity compared with untreated virus (P < 0.05 vs. control). C: HA inhibition by the NCRDs alone (white bars) or after preincubation of the NCRDs with S protein-HRP (black bars), which causes cross-linking of the NCRDs. hR343V caused significantly greater HA inhibition than hNCRD with and without S protein-HRP (*); however, hNNK caused significantly less HA inhibition than hNCRD (#) (both P < 0.05). Conc., concentration.

Figure 3B shows effects of these mutations on viral neutralizing activity. hNCRD had minimal neutralizing activity, as previously reported (6), and none of the mutants discussed in this section had increased activity. Again, hR343V and hR343A, which have strong activity, were tested in parallel as positive controls. Interestingly, hNNK caused significant increase in viral infectivity.

To amplify subtle differences in antiviral activity, we incubated NCRDs with S protein-HRP conjugates, which effectively cross-link the protein and increase binding via increased cooperative interactions among trimeric NCRDs (25). Figure 3C demonstrates that rDDR and hR343K had similar activity to hNCRD or rNCRD after cross-linking; however, hNNK lacked activity. Hence, hNNK lost antiviral activity compared with wild-type hNCRD.

Contributions of Phe335 to antibody binding.

In recent studies, we (3) have shown that specific residues that are relatively remote from the primary carbohydrate binding site, but that are exposed at the putative ligand binding interface, contribute to the recognition of specific oligosaccharides. In particular, we have shown that the aromatic ring of Phe335 can participate in stacking interactions with ligand and that certain amino acid substitutions can alter interactions with specific ligands.

hF335A appears inactive as a lectin (3) and lacked neutralizing activity (and even increased viral infectivity like hNNK; Fig. 4A). In contrast, the neutralizing activity of hF335Y and hF335W was significantly greater than that of either the wild-type hNCRD or hF335A. Furthermore, hF335Y inhibited HA activity after cross-linking with S protein-HRP (mean HA inhibitory concentration: 2,290 ± 100 ng/ml; n = 3), but hF335A did not. hF335A also showed greatly reduced binding to blocking antibodies 246-03, 246-05, and 246-07; there was partial inhibition of binding to 246-06 (Fig. 4B). By contrast, hF335Y or hF335W, which show normal or enhanced binding to mannan or maltotriose, respectively (3), showed much more selective changes in antibody reactivity (i.e., significantly reduced binding of mAb 246-07 only).

Fig. 4.

Viral neutralization by, and mAb binding to, NCRDs with substitutions at F335. Three mutant NCRD with substitutions for F335 were tested for their independent neutralizing activity in A (assay as described in Fig. 3). Wild-type hNCRD did not alter viral infectivity, but the F335Y and F335W mutants did (#significantly increased viral infectivity compared with control; *significantly reduced viral infectivity compared with control; P < 0.05; n = 4). The F335A mutant actually resulted in a significantly increased viral titer compared with control buffer or wild-type hNCRD (P < 0.02; n = 4). Binding of the indicated mAbs to the F335 mutants is shown in B, and results are shown as percentage of control binding to wild-type hNCRD. The F335Y and F335W mutants showed a significant reduction in binding of mAb 246-07 only, whereas, mAbs 246-03, 246-05, 246-06, and 246-07 all had significantly reduced binding to F335A (*significantly reduced mAb binding; P < 0.04; n = 4).

Contributions of other hydrophilic residues at the ligand binding surface.

We also modified selected, highly exposed charged residues displayed at the ligand binding surface of the CRD. The side chains of R349, E333, and E347 of the human NCRD participate in a distinctive network of hydrogen bonds, which also involves R343 (Fig. 5). In addition, E333 coordinates with Asn341, which, in turn, participates in calcium coordination at the primary carbohydrate binding site (3). The functional consequences of mutations at these sites are the subject of active investigation; however, only E333 appears generally required for lectin activity. The hR349A mutation selectively decreased binding of 246-05, 246-06, and 246-07 (Fig. 6A). Note the proximity of R349 to R343, mutations of which also reduce binding to mAbs 246-05 and 246-06. Human E333A caused more extensive structural alterations in that it greatly reduced binding of 246-03, 246-05, 246-06, and 246-07. Nevertheless, it did not alter binding to 246-04, 246-08, or the neck mAbs (data not shown). The hE347A substitution caused a modest decrease in binding of 246-07 only.

Fig. 5.

Hydrogen bonding among charged surface residues on hNCRD. A network of hydrogen bonds involving charged residues on the surface of the wild-type hNCRD are shown.

Fig. 6.

hNCRD mutants with removed charged residues: mAb binding, viral neutralization, and HA inhibition. A: experiments were performed as in Fig. 2. Note that all mAbs bound to the hK348Q or hK348A mutants to the same extent as to wild-type hNCRD. The hE347A mutant only showed slight reduction in binding to 246-07 mAb. The hR349A mutant showed significantly reduced binding to mAbs 246-05, 246-06, and 246-07. hE333A showed reduction of binding to these same mAbs but also to 246-03. Results are means ± SE of 4 experiments. *Instances of significantly reduced mAb binding (P < 0.05). B: neutralizing activity of indicated hNCRD mutants was tested as in Fig. 3 (n = 4). The hK348Q and hK348A mutants caused slight but statistically significant reduction in viral titers (*P < 0.05 for either); however, hE347A and hR349A did not. hE333A actually increased viral infectivity compared with control buffer (#P < 0.004). C: HA inhibition after cross-linking with S protein-HRP was tested as in Fig. 3 (n = 4). The hK348A mutant had similar activity to wild-type hNCRD on this assay, whereas activity of hE347A and R349A was reduced (#P < 0.05 vs. hNCRD), and hE333A had no activity (#P < 0.05 vs. hNCRD).

Based on crystallographic analysis, lysine 348 (K348) also extends into the ligand binding interface. Modification of this lysine to asparagine (hK348Q, as found in rat and mouse) or alanine (hK348A) did not cause obvious effects on binding to mannan or IAV (data not shown). It also did not affect binding to any of the mAbs (Fig. 6A).

Hence, mutating some of the exposed charged residues on hNCRD had greater effects on the structure than mutating others. In general, those in which structure was altered more substantially also showed greater loss in antiviral activity. As shown in Fig. 6B, hK348A and hK348Q had activity similar to (or greater than) wild-type hNCRD (compare neutralization and HA inhibition results with Fig. 3). Human R349A and E347A showed a significant reduction in activity in the HA inhibition assay and did not cause viral neutralization. hE333A entirely lacked HA inhibiting activity, and, like hNNK, hE333A significantly increased viral infectivity.

Contributions of residues coordinating with calcium.

Conserved residues coordinating with calcium at the primary carbohydrate binding site have not yet been systematically examined by mutagenesis. This is largely because of their inevitable effects on calcium affinity, with predictable alterations in ligand binding. However, hE321K, which was identified in genomic sequence, has been examined in conjunction with other studies. As anticipated, hE321K lacks binding activity in all solid-phase carbohydrate binding assays (data not shown). Notably, this mutation caused loss of binding to 246-03 and 246-07 and reduction in binding of 246-05, again without changing binding of 246-02, 246-04, 246-08, or the neck mAbs (245-01 and 245-02; Fig. 7A and data not shown). We (22) have previously reported that the E321K mutation abolishes HA inhibitory activity even after cross-linking.

Fig. 7.

Binding of mAbs to hW340F and hE321K and viral neutralization by mutant NCRDs with substituted tryptophans or tyrosines. A: hW340F showed markedly reduced binding to all mAbs tested, and hE321K showed reduced binding to 246-03, 246-05, 246-06, and 246-07. *Significant reduction in mAb binding to the NCRD. B: hW340F significantly increased viral infectivity (#P < 0.05 vs. control buffer or hNCRD). Neutralizing activity of hY314F, hY306F, and hW317F was not significantly different than wild-type hNCRD control (i.e., no significant reduction in viral infectivity).

Modification of the conserved CRD core.

Tryptophan 340 (W340) is part of the highly conserved C-type lectin motif and plays critical structural roles, deep to the primary carbohydrate binding site. Even the comparatively conservative W340F substitution markedly altered binding to all of the mAbs except 246-02, including in this case 246-04 (Fig. 7A). W340F was also the only mutant version of hNCRD that we tested that nearly entirely lacked binding to mAb 246-08 (binding <5% of control; n = 3). Not surprisingly, the W340F mutation resulted in decreased mannan binding and loss of antiviral activity; in fact, like the E333A, NNK, and F335A mutations, W340F showed significantly increased viral infectivity (Fig. 7B). W340F also had no HA inhibitory activity after cross-linking with S protein-HRP (data not shown).

Modification of conserved residues remote from the ligand binding surface.

As noted previously, 246-08 and 246-04 have binding properties that suggest interactions with the neck or the lateral or “back” surfaces of the human CRD. In particular, these antibodies can enhance viral interactions via cross-linking of NCRDs (22). Although systematic mutational analysis of these regions has not been performed, we have recently examined the effects of mutations of tyrosines 228, 306, and 314; all of these residues are highly conserved in SP-D and are at least partially exposed to solvent. Through combinatorial substitution of phenylalanine for tyrosine, we (18) have shown that hY314 is a preferred site for nitration of SP-D and hY228 (in neck domain) is a preferred site for peroxynitrite-dependent cross-linking of NCRDs. Nevertheless, even a triple mutation did not significantly interfere with lectin activity. Notably, hY228F, hY306F, hY314F, and hW317F mutations (alone or in combination) did not change binding to any of the mAbs, including 246-04 and 246-08 (data not shown, Table 1). In contrast to the hE321K and hW340F mutants, hY306F, hY314F, and hW317F retained some neutralizing activity as shown in Fig. 7B. Human Y306F also inhibited HA activity after cross-linking with S protein-HRP (mean HA inhibitory concentration: 2,400 ± 700 ng/ml; n = 3).

Table 1.

Summary of mannan binding and antiviral activity of wild-type and mutant NCRDs and of binding of monoclonals to the NCRDs

Trimeric NCRD	Mannan Binding	Antiviral Activity	246-02	246-03	246-04	246-05	246-06	246-07	246-08
hNCRD	Binds*	None†	Binds‡	Binds	Binds	Binds	Binds	Binds	Binds
hD324N+D325N (NN)	Increased	Same	Binds	Binds	Binds	Decreased	Binds	—	Binds
hD324N (N324)	Binds		Binds	Binds	Binds	Binds	Binds	Binds	Binds
hD325N (N325)	Increased		Binds	Binds	Binds	—	Decreased	—	Binds
hR343K	Decreased	Same	Binds	Binds	Binds	Decreased	Binds	Binds
hD324N+D325N+R343K (NNK)	Binds	—	Decreased	Decreased	Binds	—	Decreased	—	Binds
rNCRD	Binds	Same	—	Binds	—	—	—	—	Binds
rN324D+N325D (DD)	Decreased	Same	—	Decreased	—	—	—	—
rK343R (R343)	Increased	Same	—	Decreased	—	Binds	Binds	—
rN324D+N325D+K343R (DDR)	Binds	Same	—	Binds	—	Binds	Binds	—	Binds
Phe335 mutants
hF335A	—	—	Binds	Decreased	Binds	Decreased	Binds	—	Binds
hF335Y	Increased	Increased	Binds	Binds	Binds	Binds	Binds	—	Binds
hF335W	Binds	Increased	Binds	Binds	Binds	Binds	Binds	—	Binds
Charge network mutants
hE333A	—	—	Binds	Decreased	Binds	—	Decreased	—	Binds
hE347A	Binds	Same	Binds	Binds	Binds	Binds	Binds	Binds	Binds
hK348A	Binds	Increased	Binds	Binds	Binds	Binds	Binds	Binds	Binds
hK348Q	Binds	Same	Binds	Binds	Binds	Binds	Binds	Binds	Binds
hR349A	Decreased	Decreased	Binds	Binds	Binds	—	Decreased	Decreased	Binds
Other mutants
hE321K	—	—	Binds	Decreased	Binds	Decreased	Binds	—	Binds
hW340F	Decreased	—	Binds	Decreased	—	Decreased	—	—	—
hY228F+Y306F+Y314F	Binds	Same	Binds	Binds	Binds	Binds	Binds	Binds	Binds

^*Binding of trimeric neck and carbohydrate recognition domains (NCRDs) to mannan expressed relative to binding of the wild-type human NCRD (hNCRD). Mannan binding results are summarized from prior publications (2-6, 21) or unpublished data.

^†Antiviral activity was assessed by direct neutralizing activity or hemagglutinin (HA)-inhibitory activity (after cross-linking of the NCRDs with S protein-horseradish peroxidase). Without cross-linking, the wild-type hNCRD had no neutralizing or HA-inhibiting activity, but it did have HA-inhibitory activity after cross-linking. The other NCRDs are compared with wild-type hNCRD (see Figs. 3, 4, 6, and 7 for specific results of antiviral activity assays).

^‡Monoclonal antibody (mAb) binding results are summarized from Figs. 2, 4, 6, and 7. An antibody “binds” if binding is >50% of that observed for the hNCRD. Binding is described as “decreased” if binding is <50%. Binding (or antiviral activity) of <20% of control was not considered significant and is indicated by a minus sign. Where spaces are left blank, no value was determined. IAV, influenza A virus; rNCRD, wild-type trimeric rat NCRD; charge network mutants, hNCRD mutants with removed charged residues.

DISCUSSION

mAbs facilitate the identification of ligand binding sites.

Our results establish that sites for binding of gp-340 and mannan are distinct and that the mannan binding region overlaps with that for binding to IAV. In general, mAbs that block binding to IAV or mannan have epitopes that include hydrophilic residues that reside on the ridges that flank the primary carbohydrate binding site. mAbs binding to the neck or those that bind neither to the ligand binding surface nor the neck (246-04 and 246-08) do not block antiviral activity and in some cases enhance it.

We cannot precisely localize the site(s) of binding for gp-340; however, it is “near” the lectin site because mAbs 246-02 and 246-03, which recognize residues flanking this site and block binding to mannan or IAV, strongly inhibit binding to gp-340. The finding that 245-01 inhibits binding of gp-340 suggests that gp-340 also interacts with more distant sites. Notably, binding to gp-340 is calcium-dependent but not sensitive to competing maltose, suggesting that it recognizes a calcium-dependent conformation of the NCRD. In addition, porcine SP-D has a glycan on the lateral surface of the NCRD that inhibits binding to gp-340 (26). In sum, these findings support the concept that the binding site for gp-340 is distinct from the lectin site. We speculate that the calcium dependence reflects structural roles of the secondary calcium ions, which stabilize loop structures at the lateral surface of the CRD, near the glycan attachment in porcine SP-D.

mAbs identify key antigenic sites on the NCRD.

Table 1 summarizes our current findings regarding the functional activities and mAb binding profile of the panel of NCRDs used in this paper. Our studies indicate that D325, which is the most prominent residue on one of the ridges alongside the primary carbohydrate binding site of SP-D, is both a key epitope and critical for the ligand recognition properties of SP-D. Substitutions at this site not only change saccharide binding selectivity (6), but also interfere with binding to most of the blocking mAbs we tested. It is of interest that rat SP-D NCRD does not bind to most blocking mAb, whereas binding to one of the nonblocking mAbs (246-08) is relatively conserved. R343 defines the other ridge adjacent to the lectin site and also contributes to antibody binding but is apparently less important than D325; substitution of hydrophobic residues at 343 strongly affected binding of one antibody, 246-05, and affected to a lesser extent 246-06. Notably, both of these mAbs also recognize D325. It is of interest that the single site substitution of arginine for Lys343 in rat SP-D confers binding to these antibodies. Changes in other residues that interact with R343 based on structural analysis (e.g., E333) also cause loss of binding of 246-05 and reduction of binding to 246-06, suggesting that the orientation of R343 is altered in these mutants.

mAbs detect structural changes in NCRD that correlate with loss of antiviral activity.

mAbs can readily detect major structural changes in NCRD as illustrated by the case of hW340F. W340 is a highly conserved residue in the core of the NCRD and more generally in C-type lectins. Even the seemingly conservative substitution of phenylalanine led to loss of antiviral activity with altered binding of nearly all tested antibodies; this is despite retention of normal intrachain disulfide bond formation of the CRD (data not shown). The fact that hW340F even caused loss of the highly conserved 246-08 and 246-04 epitopes suggests major structural alteration in the CRD.

Arg349 plays an important role in tethering Arg343 in position, as does Glu333, but substitution of the latter residue with alanine causes more extensive changes. This probably reflects important interactions of this residue with Asn341, which coordinates with calcium at the primary carbohydrate binding site. As noted, the hE333A mutant loses all antiviral activity, whereas hR349A has a much smaller but significant reduction in activity.

The hF335A substitution also markedly changes structure and leads to a loss of antiviral activity; however, the conservative mutations hF335Y and hF335W did not similarly alter structure and were associated with a statistically significant increase in neutralizing activity. We (3) have shown that F335 forms part of an extended binding surface (i.e., by forming a secondary binding site for the 3rd hexose ring in maltotriose and probably other ligands). Our findings suggest that the aromatic residue in this position is important for viral binding. It is unclear at this point why the hF335A, hNNK, and hW340F mutant NCRDs actually promoted viral infectivity, and this will be investigated in future studies.

Characterization of an extended binding site for IAV.

Figure 8 depicts the ligand binding surface of the SP-D NCRD trimer, indicating key amino acids mutated for this study. Based on findings presented in this and previous papers, the ridges on either side of the saccharide binding site of SP-D are involved in viral recognition and antiviral activity. These ridges are made up principally by the hydrophilic residues D325 and R343 (yellow and blue in Fig. 8). The effect of R343 on viral binding is particularly evident, since hydrophobic substitutions in this area markedly increase viral binding and antiviral activity. In contrast, replacing R343 with another basic residue (lysine) does not significantly alter antiviral activity (despite significantly increasing binding to LPS and PI). The conservative substitution, D325N, does not alter antiviral activity, but a RAK insertion in that area does. Studies of other substitutions for D325 are underway. Note that D324 is not located on the binding surface but rather on the side surface of the NCRD. This may explain why alteration of this residue did not alter antibody binding or antiviral activity. The approximate locations of antibody binding sites are also indicated in Fig. 8.

Fig. 8.

Diagram of the extended binding surface of human SP-D NCRD. The figure was rendered in PovRay from DeepView. The figure shows the molecular surface of the SP-D-maltotriose complex [Research Collaboratory for Structural Bioinformatics Protein Data Bank (PDB) code 2GGU]. The ligand is shown in stick figure. In A, the approximate sites for antibodies 246-02 (red), 246-03 (orange), and 246-07 (green) are shown on different heads of the trimer. In B, the sites for antibodies 246-05 (cyan) and 246-06 (blue) are shown. The bottom monomer in B shows the contribution to the molecular surface from the key residues Glu321 (red), Asp325 (yellow), Glu333 (orange), Phe335 (green), Arg343 (blue), E347 (cyan), and Arg349 (magenta). The orientation of the figure is looking down from above on the planar binding surface of the trimer.

In this paper, we demonstrate that multiple additional residues on the NCRD surface contribute to antiviral activity. F335 (shown in green in Fig. 8) contributes since the F335Y and F335W increase viral neutralizing activity slightly, whereas the F335A substitution ablates viral inhibition. It is possible that the conservative substitutions increase stacking interactions with viral oligosaccharides (note that the 3rd ring in maltotriose overlies and has binding interactions with F335; see Fig. 8) (3). The F335A substitution, in contrast, may decrease viral binding by either eliminating stacking interactions or by causing more extensive alterations in the surrounding NCRD structure (as suggested by alterations in binding of several mAbs).

Some of the charged residues on the NCRD surface also affect viral interactions (Fig. 8), including E333 (orange), E347 (cyan), R349 (magenta), and K348. Substitutions at K348 slightly increased antiviral activity, whereas the R349A and E347A substitutions slightly reduced, and the E333A substitution eliminated, antiviral activity. Antiviral activity can be further increased by cross-linking of NCRDs through mAbs that bind to other areas on the CRD that have yet to be defined. Overall, the colored portions of the binding surface of the human SP-D NCRD define an extended area that is involved in binding of blocking mAbs and viral binding. It should be noted that the effects of mutating some of these residues could relate to structural changes in surrounding residues as well as (or instead of) direct interaction of the mutated residues with ligands. For instance, the E333A, E347A, and R349A substitutions might affect the orientation of R343V or calcium chelating residues. The F335A mutation appears to cause a major change in NCRD structure. W340 is clearly very important for the overall structure of the NCRD, and this probably explains the loss of antiviral activity of W340F.

Summary.

The mAbs used in this study all have distinct epitopes on the hNCRD and are, therefore, useful in comparing mechanisms of binding to specific ligands. We provide evidence for partially distinct mechanisms of binding of hNCRD to gp-340, IAV, and mannan. The mAbs also proved useful in monitoring structural changes resulting from mutation of key residues in the CRD (e.g., R343 or F335). Through correlation of mutational and antibody recognition analysis with crystallographic findings and antiviral activity, we develop a detailed picture of the surface features of the SP-D CRD involved in viral binding and inhibition. The analysis reveals, for the first time, an extended binding site for IAV; calcium-dependent antiviral activity involves residues flanking the primary carbohydrate binding site as well as more remote residues on the ligand binding interface.

GRANTS

This work was supported by National Institutes of Health Grants AI-83222 (K. L. Hartshorn, E. C. Crouch, and J. Head) and HL-069031 (K. L. Hartshorn).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).