Correction of Pulmonary Abnormalities in Sftpd-/- Mice Requires the Collagenous Domain of Surfactant Protein D*

Surfactant protein D (SP-D) is a member of the collectin family of innate defense proteins. Members of this family share four distinct structural domains: an N-terminal cross-linking domain, a collagenous domain, a neck region, and a carbohydrate recognition domain. In this study, the function of the collagenous domain was evaluated by expressing a SP-D collagen deletion mutant protein (rSftpdCDM) in wild type and SP-D null mice (Sftpd-/-). rSftpdCDM formed disulfide-linked trimers that further oligomerized into higher order structures. The mutant protein effectively bound carbohydrate and aggregated bacteria in vitro. Whereas rSftpdCDM did not disrupt pulmonary morphology or surfactant phospholipid levels in wild type mice, the mutant protein failed to rescue the emphysema or enlarged foamy macrophages that are characteristic of Sftpd-/- mice. Moreover, rSftpdCDM partitioned with small aggregate surfactant in a manner similar to SP-D, but rSftpdCDM did not correct the abnormal surfactant ultrastructure or phospholipid levels observed in Sftpd-/- mice. In contrast, rSftpdCDM completely corrected viral clearance and the abnormal inflammatory response that occurs following pulmonary influenza A challenge in Sftpd-/- mice. Our findings indicate that the collagen domain of SP-D is not required for assembly of disulfide-stabilized oligomers or the innate immune response to viral pathogens. The collagen domain of SP-D is required for the regulation of pulmonary macrophage activation, airspace remodeling, and surfactant lipid homeostasis.

Surfactant protein D (SP-D) 2 is a member of the collectin family of C-type lectins. Members of this family include surfactant protein A (SP-A), SP-D, mannose-binding protein, conglu-tinin, and CL-43. SP-A and SP-D contribute to the innate immune system of the lung by binding and enhancing the clearance of a variety of viral, bacterial, and fungal pathogens (1)(2)(3). The collectins are defined by four structural domains shared by all family members: a short amino-terminal cross-linking domain, a triple helical collagenous domain, a neck domain, and a carbohydrate recognition domain (CRD) (4 -8). Three neck domains associate to form a triple coiled-coil structure that facilitates the assembly of the remaining domains into a trimer (9). Final assembly of the trimer, thought to be the minimal functional unit of collectins, occurs through disulfide bonds between the cysteine-rich amino-terminal domains (10,11). Trimers further combine into larger multimeric complexes through disulfide-stabilized, noncovalent interactions. Although larger structures are commonly observed, SP-D exists predominantly as a tetramer of trimeric subunits (dodecamer) assembled into a cruciform structure (10,12).
Animal models of SP-D deficiency have revealed a complex role for SP-D in pulmonary immunity and alveolar homeostasis. Mice with a targeted deletion of the Sftpd gene (Sftpd Ϫ/Ϫ ) survived normally but developed gradually worsening pulmonary inflammation, emphysema, and surfactant phospholipid accumulations (13,14). Sftpd Ϫ/Ϫ mice accumulate apoptotic alveolar macrophages as well as enlarged, lipid-laden, macrophages that release metalloproteinases and reactive oxygen species (15)(16)(17)(18)(19). Uptake and clearance of viral pathogens, including influenza A and respiratory syncitial virus, were deficient in Sftpd Ϫ/Ϫ mice (20,21). In contrast, clearance of group B Streptococcus and Hemophilus influenza was unaltered in the absence of SP-D (19). However, oxygen radical release and production of the proinflammatory mediators, tumor necrosis factor-␣, IL-1, and IL-6, were increased in Sftpd Ϫ/Ϫ mice when exposed to either viral or bacterial pathogens (19 -21).
The roles of the various domains of SP-D in its complex functions have been studied by expressing SP-D point mutants, deletion mutants, or chimeric proteins of SP-D and other collectins. For example, whereas expression of the fulllength rat Sftpd gene (rSftpd ) fully rescues the Sftpd Ϫ/Ϫ mouse phenotype, expression of a fusion protein that included the N-terminal and collagen domains of SP-A fused to the neck and CRD of SP-D (rSftpa/d) was not sufficient to correct the emphysema or lipid accumulations characteristic of Sftpd Ϫ/Ϫ mice, indicating that the collagenous and N-terminal domains of SP-D are essential for these functions (22). Substitution of serine for cysteine at positions 15 and 20 of the N-terminal domain (rSftpd Ser15,20 ) results in a protein that corresponds to a single trimeric arm of the SP-D dodecamer but fails to correct the abnormal macrophages, emphysema, or lipid accumulations in Sftpd Ϫ/Ϫ mice, suggesting that oligomers induced by the SP-D collagenous and N-terminal domains are required for complete function (23). However, intranasal administration of a truncated SP-D trimer (consisting of only the CRD and neck domains) into Sftpd Ϫ/Ϫ mice partially corrects the apoptotic macrophages, lipid accumulations, and elevated inflammatory mediators, suggesting that the collagen domain is not essential for these activities (15,24). Since most collectins form sulfhydryl-stabilized oligomers, the relatively large collagen domain of SP-D suggests that this domain serves a purpose that is beyond facilitating oligomerization of the SP-D CRD. Comparisons of SP-D, mannose-binding protein, and conglutinin fusion proteins suggest the large collagen domain promotes proper spacing of trimeric subunits in order to facilitate cross-linking of separate microorganisms and microbial aggregation (25). Therefore, to further investigate the function of the SP-D structural domains, we genetically introduced into wild type and Sftpd Ϫ/Ϫ mice lungs an SP-D collagen deletion mutant protein (rSftpdCDM) that formed multimers via an intact N terminus. Whereas expression of rSftpdCDM elicited a protective response to influenza virus, the protein failed to correct the abnormal macrophage activation, emphysema, or lipid abnormalities in Sftpd Ϫ/Ϫ mice, indicating that the SP-D collagenous domain is critical for normal regulation of lipid homeostasis, macrophage activity, and the structural integrity of peripheral airspaces.

EXPERIMENTAL PROCEDURES
Animal Husbandry-Mice were handled in accordance with approved protocols through the Institutional Animal Care and use Committee at Cincinnati Children's Hospital Medical Center. All mice had been maintained in the vivarium in barrier containment facilities. Sentinel mice in the colony were serologically negative for common murine pathogens.
Western Blot Analysis-Animals were weighed, anesthetized by intraperitoneal injection of pentobarbital, and exsanguinated. Bronchoalveolar lavage was performed five times with 1 ml of normal saline. Typically, Ͼ90% of the instilled volume was recovered. For each lane, 40 l of the bronchoalveolar lavage fluid (BALF) was dried, reconstituted in 15 l of Laemmli buffer with or without sulfhydryl reduction with ␤-mercaptoethanol, and resolved on 10 -20% SDS/Tris/glycine/polyacrylamide gel (Novex, San Diego, CA). After separation and transfer to a nitrocellulose membrane, protein was detected with rabbit anti-mouse SP-D or guinea pig anti-rat SP-A antiserum (Seven Hills Bioreagents, Cincinnati, OH) diluted 1:5000 in Tris-buffered saline as previously described (27).
SP-D Purification-Previous studies demonstrated that inactivation of the granulocyte-macrophage colony-stimulating factor gene (Gmcsf Ϫ/Ϫ ) in mice impaired SP-D clearance and increased SP-D levels in BALF severalfold (28). Therefore, to increase the amount of starting material for protein purification, rSftpdCDM was purified from rSftpdCDM Tgϩ /Sftptd Ϫ/Ϫ / Gmcsf Ϫ/Ϫ mice. Wild type mouse SP-D was purified from Sftpd ϩ/ϩ /Gmcsf Ϫ/Ϫ mice that also carried a deletion of the two expressed Sftpa genes in order to minimize the potential of SP-A contamination. A similar Sftpa deletion was not possible in the rSftpdCDM Tgϩ /Sftptd Ϫ/Ϫ /Gmcsf Ϫ/Ϫ mice; however, contaminating SP-A was not detectable by silver stain gels in rSftpdCDM preparations. SP-D-or rSftpdCDM-containing BALF was applied to a maltosyl-Sepharose (Sigma) column and selectively eluted with manganese as previously described (29).   SP-D Sizing and Detection-The size of rSftpdCDM multimers was determined by gel filtration chromatography using Sepharose CL-6B equilibrated in 20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 5 mM EDTA, and 0.02% sodium azide. Purified rSft-pdCDM was diluted in the same buffer and applied to the column (1.5 ϫ 90 cm). rSftpdCDM protein concentration was determined in each fraction using an enzyme-linked immunosorbent assay (ELISA). Plates were washed five times between incubations, and all washes and dilutions were carried out with Tris-buffered saline, 0.1% Tween 20. A mouse anti-SP-D monoclonal antibody was developed by exposing Sftpd Ϫ/Ϫ mice to purified, full-length, human SP-D and selecting cell lines that demonstrated a high affinity for SP-D and minimal cross-reactivity with SP-A (Seven Hills Bioreagents, Cincinnati, OH). ELISA plates were coated with this monoclonal antibody (1 g/ml, 100 l/well) in 0.1 M carbonate buffer, pH 9.6, overnight at 4°C and blocked with 1% bovine serum albumin for 1 h at room temperature. Plates were washed, and appropriate dilutions of standards and protein samples were added and incubated for 1 h at room temperature. Plates were washed and incubated with rabbit anti-mouse SP-D antiserum (100 l/well diluted 1:750) for 1 h. This was followed by washing and incubation with donkey horseradish peroxidase-conjugated anti-rabbit IgG (100 l/well diluted 1:10,000) (Jackson Immunoresearch, West Grove, PA) for 1 h. After washing plates again, TMB substrate (100 l/well) (BioFx Laboratories, Owings Mills, MD) was added. The color reaction was stopped after 10 min with 2 M H 2 SO 4 , and plates were read at 450 nm. Typically, this assay results in absorbance changes that are equal, parallel, and linear for mouse SP-D, rSftpdCDM, and human SP-D concentrations between 10 and 150 ng/ml.
Carbohydrate Binding-Direct binding of SP-D to carbohydrate was detected by mixing BALF from 6 -8-week-old mice with maltosyl-Sepharose-linked beads at 4°C for 2 h in 4 mM Tris-HCl, pH 7.4, and 5 mM CaCl 2 (23). Binding specificity and calcium dependence were confirmed by the addition of 100 mM maltose and 10 mM EDTA to the binding reaction, respectively. The samples were centrifuged at 10,000 ϫ g for 1 min, and the supernatants were removed. The amount of unbound SP-D in the supernatant was determined by Western blot analysis.
Selective carbohydrate binding was performed as described previously (30). Briefly, microtiter plates were coated with 10 g/ml mannan at 4°C overnight, washed, and blocked with 1% bovine serum albumin. BALF was incubated with increasing concentrations of maltose, glucose, galactose, or GlcNAc. Samples were subsequently added to the coated plate and incubated with rabbit anti-SP-D antibody followed by peroxidase-conjugated goat anti-rabbit IgG antibody. After washing, o-phenylenediamine was added to each well, and the A 490 nm was measured. The concentration of sugar that inhibited 50% of SP-D binding to the mannan-coated plate was defined as the IC 50 .
Bacterial Aggregation-Bacterial aggregation was assessed by measuring light transmission through a bacterial suspension after the addition of SP-D. Purified rSftpdCDM, mouse SP-D (150 nm based on the monomer molecular weight), or a control reaction that contained protein buffer without SP-D was mixed with 600 l of Escherichia coli Y1088 (OD 700 nm ϳ 1) grown the previous night and resuspended in Tris-buffered saline plus 5 mM CaCl 2 (31). The OD 700 nm was measured every 2.5 min, and all values were reported as relative to the absorbance at t ϭ 0. The extent of aggregation was determined by the decrease in the optical density of the bacterial suspension. Calcium dependence was confirmed by the inhibition of aggregation in the presence of 10 mM EDTA.
Lung Morphology-Lungs from 12-week old mice were fixed at 25 cm of water pressure with 4% paraformaldehyde in phosphate-buffered saline (PBS) and processed into paraffin blocks. Sections (5 m) from each lobe were stained with hematoxylin and eosin. Immunohistochemistry for SP-D was performed at dilutions of 1:200 by using a rabbit polyclonal antibody generated to murine SP-D. Immune complexes were detected using an avidin-biotin-peroxidase technique (Vectastain Elite ABC kit, Vector Laboratories, Burlingame, CA).
Metalloproteinase Activity-Alveolar macrophages (5 ϫ 10 5 ) were isolated by centrifuging BALF and cultured for 24 h in AIM-V medium (Invitrogen). Proteinases in the conditioned media were assayed by zymography as described previously (32).
Isolation of Large and Small Aggregate Sufactant-Large and small aggregate surfactant lipids were isolated from rSftpdCDM Tgϩ /Sftpd ϩ/ϩ mice (n ϭ 6) as previously described (14). Briefly, bronchoalveolar lavage was performed five times with 1 ml of normal saline. BALF was centrifuged at 40,000 ϫ g over a 0.8 M sucrose cushion for 15 min. Small aggregate surfactant was collected from the supernatant. Large aggregate surfactant was collected from the interface, diluted with normal saline, and centrifuged again at 40,000 ϫ g for 15 min. The large aggregate pellet was dissolved in normal saline. An equal volume of each material was dried, diluted in Laemmli buffer, and analyzed by Western blot.
Ultrastructure of lipid aggregates was determined in pooled BALF samples (n ϭ 5/pool) by electron microscopy from 6 -8week-old mice as previously described (14). Briefly, large and small aggregate surfactant was isolated from BALF; fixed with glutaraldehyde, paraformaldehyde, and CaCl 2 in 0.1 M sodium cacodylate buffer at 4°C; and stained with osmium tetroxide, potassium ferrocyanide, and uranyl acetate. After dehydration, it was embedded in Embed812 resin (Electron Microscopy Sciences, Fort Washington, PA), and ultrathin sections (90 nm) were obtained. Random electron micrographs were taken, and ultrastructure was evaluated.
Influenza A Virus-Experiments utilized influenza A virus strain H 3 N 2 A/Phillipines/82 (IAV) and were performed as previously described (34). Briefly, IAV was grown in chorioallantoic fluid of 10-day-old embryonated hen eggs. Virus was purified and stored frozen in PBS until use. Six-week-old (n ϭ 6 -8) mice were anesthetized with inhaled isoflurane and inoculated intratracheally with 5 ϫ 10 5 fluorescent foci in 80 l of PBS.
Quantitative IAV cultures were performed 3 days after inoculation. Lungs were removed and homogenized in 2 ml of PBS, and aliquots were frozen. IL-6, tumor necrosis factor-␣, and interferon-␥ concentrations were determined by ELISA kits (R&D Systems, Minneapolis, MN). IAV titers were determined by incubating lung homogenates with Madin-Darby canine kidney monolayers for 7 h at 37°C. Monolayers were washed, fixed, and incubated with monoclonal antibody against IAV nucleoprotein followed by a rhodamine-labeled goat antimouse IgG. Fluorescent foci were counted, and the resulting viral titer was expressed as fluorescent foci/g lung weight.
SP-D binding to IAV was monitored by hemagglutination inhibition assays (n ϭ 3). Purified mouse SP-D and rSftpdCDM were compared. Sample protein was serially diluted in 96-well round bottom plates with IAV in PBS with 0.5 mM CaCl 2 and 0.5 mM MgCl 2 . After incubating the protein/IAV mixture at room temperature for 10 min, fertilized chicken egg erythrocytes were added, and the samples were incubated for an additional 2 h. The minimal amount of SP-D or rSftpdCDM needed to fully inhibit agglutination was observed, and the number of hemagglutination units inhibited per pmol of SP-D or rSftpdCDM was reported.
Data Analysis-Where appropriate, either a representative experiment from one mouse line was shown, or results from each line were averaged, and data were analyzed by unpaired Student's t tests.

RESULTS
Transgenic Mouse Lines-rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ transgenic mice were produced by nuclear injection of the rSftpdCDM gene into FVB/N mice and backcrossed to Sftpd Ϫ/Ϫ mice. The rat Sftpd gene is 92% identical (based on nucleotide and mature protein amino acid sequence) to the mouse Sftpd gene, and multiple studies indicate that mouse and rat SP-D have nearly identical properties. Four founder mice were identified using transgene-specific PCR on tail clip DNA. Germ line transmission was demonstrated in all four lines, and the transgene was inherited as an autosomal gene following Mendelian inheritance. Survival and breeding were not influenced by the rSftpdCDM Tgϩ transgene. Expression and secretion of transgenic protein rSftpdCDM was confirmed in all four mouse lines by Western blot analysis of mouse BALF using a rabbit anti-mouse SP-D antibody (data not shown). Two mouse lines with similar levels of rSftpdCDM protein expression were selected for further breeding and analysis. Experiments were done in parallel with both lines, and results were similar.
Expression of rSftpdCDM Protein-As determined by ELISA assays, levels of SP-D in BALF from wild type mice were 1.2 g/ml compared with rSftpdCDM levels of 11 g/ml in rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice. Rabbit anti-mouse SP-D antibody was used to detect SP-D and rSftpdCDM in BALF (Fig. 2). The monomeric form of the mutant protein migrated under reducing conditions at the predicted molecular mass of 22 kDa. rSftpdCDM migrated slightly slower than the 60-kDa standard under nonreducing conditions, indicating that the disulfide linkages normally observed between monomeric chains in the wild type N-terminal domain also formed in the mutant protein. To determine the ability of the rSftpdCDM to form high order complexes, the protein was analyzed by Sepharose column chromotography. Under these conditions, the majority of rSftpdCDM migrated between apoferritin and ␤-amylase (molecular masses of 443 and 200 kDa, respectively) at a position equaling 270 kDa, which is the molecular mass expected if adding the N-terminal domain to the CRD trimer promoted noncovalent interactions between four trimeric subunits (dodecamer). Smaller peaks were observed at the expected positions of molecules composed of 1, 2, 30, and more trimers, suggesting that the purified mutant protein consists of a heterogeneous population of multimers. Whereas the mutant protein effectively self-assembled to higher order structures, it did not form intermediate molecular weight heteropolymers of mutant and wild type protein when expressed in a wild type SP-D background (data not shown).
RSftpdCDM Binds Carbohydrate-To compare the carbohydrate binding properties of wild type SP-D and rSftpdCDM, binding to maltosyl-Sepharose was evaluated (Fig. 3). Protein  1 and 3) and rSftpdCDM Tgϩ / Sftpd Ϫ/Ϫ (lanes 2 and 4) mice were resolved on SDS-PAGE under reducing (Reduced) conditions. Disulfide-linked trimers were resolved under nonreducing (Non Reduced) conditions. Protein was detected with rabbit anti-mouse SP-D antibody. B, purified rSftpdCDM was resolved by column chromatography in Sepharose CL-6B. Protein levels were determined by ELISA and expressed as relative to the peak fraction. The arrows indicate the peak position of eluted standards: blue dextran (a), thyroglobulin (b), apoferritin (c), ␤-amylase (d ), alcohol dehydrogenase (e), albumin (f ), and carbonic anhydrase (g).
from both wild type and mutant BALF bound maltose. Binding was inhibited by the addition of EDTA, indicating that rSftpd-CDM maintained calcium-dependent lectin activity. In addition, binding to maltosyl-Sepharose was reversed by the addition of free maltose, confirming the specificity of maltose interactions.
The saccharide binding preference of SP-D and rSftpdCDM was assessed using BALF and inhibition of binding to yeast mannan by competing saccharides (Table 1). Under the conditions utilized, a higher binding affinity for the competing saccharide was indicated by a lower concentration of saccharide needed to disrupt protein-mannan interactions. Maltose was the preferred binding substrate of the carbohydrates tested with an IC 50 of 1 mM with BALF from wild type mice, consistent with results previously reported (27,30). The order of binding preference was maltose Ͼ glucose Ͼ galactose Ͼ GlcNAc for both SP-D and rSftpdCDM. Whereas the mutant and wild type proteins had similar relative saccharide binding preferences, the IC 50 values observed with rSftpdCDM were lower than SP-D. Assuming there are no compositional differences in the mutant and wild type BALF that might influence binding activity, these results suggest that the mutant protein may have a higher carbohydrate binding affinity.
rSftpdCDM Aggregates Bacteria-Previous studies demonstrated that whereas mutant trimeric forms of SP-D (rSftpd Ser15,20 and trimeric CRDs) bind carbohydrates, they do not effectively aggregate infectious particles (11,31). Similar results were observed in earlier work with a SP-D collagen deletion mutant (35). However, unlike the current rSftpdCDM, the earlier collagen deletion mutant did not form multimers. In addition, studies with chimeric collectins containing domains of SP-D, mannose-binding protein, and conglutinin suggest that the relatively large collagen domain of SP-D increases bacterial aggregation activity (25). To determine if rSftpdCDM induced bacterial aggregation, light transmission through a suspension of E. coli was monitored after the addition of purified rSftpdCDM (Fig. 4). The drop in absorbance in the rSftpdCDM reaction indicates that the multimeric mutant protein effectively aggregates bacteria. Aggregation by rSftpdCDM was completely inhibited by the addition of EDTA, confirming the calcium dependence for protein function. When compared with a reaction that contained an equal molar concentration of wild type SP-D based on the molecular weight of the monomer chain, the aggregation activity of rSftpdCDM was slightly better than the wild type protein. Thus, bacterial aggregation is not dependent on appropriate spacing of the CRD by the collagenous domain of SP-D.
Lung Morphology-Deletion of the mouse Sftpd gene caused several distinct alterations in lung morphology, including emphysema and accumulations of enlarged foamy macrophages (13, 16 -18). To determine if the mutant protein influenced these findings, pulmonary structure was assessed in wild type and Sftpd Ϫ/Ϫ mice expressing rSftpdCDM Tgϩ (Fig. 5). Immunostaining revealed marked expression of rSftpdCDM Tgϩ in alveolar type II cells and bronchiolar epithelial cells. However, rSftpdCDM did not alter the lung morphology in 12-week-old wild type mice. Moreover, the mutant protein did not correct the abnormal morphology typically observed in Sftpd Ϫ/Ϫ mice. Enlarged, foamy macrophage accumulations and emphysema were detected readily in both transgenic mouse lines expressing rSftpdCDM Tgϩ in a Sftpd Ϫ/Ϫ background. Therefore, whereas   targeted expression of the full-length rat Sftpd gene in Sftpd Ϫ/Ϫ mice fully rescues the Sftpd Ϫ/Ϫ phenotype (36), expression of rSftpdCDM Tgϩ does not correct the abnormal lung morphology in Sftpd Ϫ/Ϫ mice. MMP-9 and MMP-2 Activity-Proteinase activity gels were used to assess MMP-9 and MMP-2 activity in media containing alveolar macrophages from wild type, Sftpd Ϫ/Ϫ , and rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice (Fig. 6). Whereas minimal MMP-9 and MMP-2 activity were observed in wild type samples, metalloproteinase activity was markedly elevated in both Sftpd Ϫ/Ϫ and rSftpdCDM Tgϩ / Sftpd Ϫ/Ϫ mice, indicating that rSft-pdCDM does not correct the increased metalloproteinase production by alveolar macrophages from Sftpd Ϫ/Ϫ mice (32).
Lung Phospholipids and Surfactant Structure-SP-D selectively interacts with small aggregate surfactant in adult mice and regulates the uptake and catabolism of surfactant lipids by alveolar type II cells (13,14). Consequently, mice lacking SP-D have 2-5-fold higher surfactant pool sizes (13,16). Moreover, surfactant isolated from Sftpd Ϫ/Ϫ mice has abnormally large aggregate lipid structures and small aggregates consisting of atypical multilamellated forms (14). To evaluate if rSftpdCDM corrected surfactant phospholipid pool sizes, Sat PC in BALF and lung homogenates were assessed in wild type and Sftpd Ϫ/Ϫ mice with and without rSftpdCDM Tgϩ (Fig. 7). Expression of the mutant protein did not alter alveolar, tissue, or total Sat PC levels in wild type mice, and it did not correct the elevated Sat PC levels in Sftpd Ϫ/Ϫ mice. In addition, to determine if rSftpdCDM corrected the abnormal surfactant ultrastructure characteristic of Sftpd Ϫ/Ϫ mice, large and small aggregate surfactant was examined by electron microscopy (Fig. 8). The large aggregate fraction from Sftpd Ϫ/Ϫ and rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice contained abnormal, large lamellated lipid structures. The small aggregate fraction in wild type mice consisted of single layer sheets or vesicles, whereas atypical multilayered structures predominated in the small aggregate fraction from Sftpd Ϫ/Ϫ and rSftpdCD-M Tgϩ /Sftpd Ϫ/Ϫ mice. Despite the failure of rSftpdCDM to correct surfactant lipid structure, analysis of large and small aggregate surfactant from rSftpdCDM Tgϩ /Sftpd ϩ/ϩ mice revealed that rSftpdCDM partitioned with small aggregate surfactant in a manner that was similar to SP-D (Fig. 9). This is in contrast to SP-A, which segregated primarily with large aggregate surfactant. Therefore, these results demonstrate that the collagen domain of SP-D is not required for the selective partitioning of SP-D with small aggregate surfactant, but it is required for normal surfactant ultrastructure that in turn influences uptake by alveolar type II cells (14).
Correction of Influenza A Infection by rSftpdCDM-Previous in vitro studies demonstrated that SP-D binds IAV and enhances IAV binding and uptake by neutrophils (31,37,38). Decreased viral clearance and enhanced inflammation was observed in Sftpd Ϫ/Ϫ mice exposed to intratracheal IAV (21). As  determined by hemagglutination inhibition assays, rSftpdCDM effectively bound IAV and inhibited IAV-mediated hemagglutination (Fig. 10). Moreover, hemagglutination inhibition activity of rSftpdCDM was ϳ2-fold greater than that observed with an equal molar (based on the molecular weight of a SP-D or rSftpdCDM monomer) amount of wild type SP-D. To further evaluate the anti-IAV activity of rSftpdCDM, IAV was administered to mouse lungs intratracheally, and the viral titers and cytokine response (Fig. 11) were measured 3 days later. In contrast to Sftpd Ϫ/Ϫ mice, no detectable IAV was recovered from the wild type or rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ lung homogenates. In addition, the increased IL-6, tumor necrosis factor-␣, and interferon-␥ levels observed in IAV-challenged Sftpd Ϫ/Ϫ mice were restored to wild type levels in rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice. Taken together, these results demonstrate that rSftpdCDM completely corrects viral binding, clearance, and inflammatory responses observed in Sftpd Ϫ/Ϫ mice.

DISCUSSION
SP-D plays multiple complex roles in pulmonary physiology, including binding and clearing of infectious pathogens, regulation of the innate host defense system, airspace remodeling, and surfactant phospholipid metabolism (1)(2)(3). In the present study, the function of the SP-D collagen domain was evaluated by expressing a SP-D collagen deletion mutant in the lungs of wild type and Sftpd Ϫ/Ϫ mice.
Deletion of the collagen domain resulted in the secretion of an ϳ22-kDa protein that migrated as a trimer under nonreducing conditions. These findings are consistent with in vitro studies describing an SP-D collagen deletion mutant that was expressed in Chinese hamster ovary cells by Ogasawara and Voelker (35,39). In contrast to these earlier studies, the majority of the present rSftpdCDM formed higher order complexes of four trimeric subunits (dodecamer) when expressed in the mouse respiratory system. The reason for this disparity is unclear, but given the close proximity of the regions deleted, it probably reflects differences in the protein expression systems utilized. In FIGURE 7. Increased lung saturated phosphatidylcholine in rSftpdCD-M Tg؉ /Sftpd ؊/؊ mice. Alveolar, tissue, and total Sat PC levels were determined in wild type (Sftpd ϩ/ϩ ) and SP-D null (Sftpd Ϫ/Ϫ ) mice with (open bars) or without (closed bars) the rSftpdCDM transgene (n ϭ 6 -8 mice for each genotype). Values were normalized for body weight. Sat PC levels in Sftpd Ϫ/Ϫ mice were significantly (p Յ 0.05) higher than wild type controls. The addition of rSftpdCDM did not significantly (p ϭ 0.48, 0.47, and 0.40 for alveolar, tissue, and total, respectively) correct this increase. The error bars indicate S.E. addition, the finding that adding the N terminus to the trimeric CRD facilitates dodecamer formation supports previous results that indicate that the structural predilection for dodecamers is contained within the N-terminal domain (40). Ogasawara and Voelker (35) demonstrated that the collagen deletion mutant bound mannose-Sepharose and phosphatidylinositol with affinities comparable with wild type SP-D, but binding to mannosyl bovine serum albumin and glucosylceramide was somewhat diminished. Although specific binding constants were not determined in the present work, when comparing equal molar quantities of SP-D and rSftpdCDM, the mutant protein binding activity for virus, bacteria, and carbohydrate was consistently equal to or greater than wild type SP-D, demonstrating that deletion of the collagen domain does not inhibit substrate binding.
Despite the considerable binding affinity of rSftpdCDM, expression of the mutant protein failed to correct the aberrant alveolar macrophage activity characteristic of Sftpd Ϫ/Ϫ mice (13, 16 -18). Foamy macrophages, increased secretion of metalloproteinases, and emphysema persisted despite expression of high levels of rSftpdCDM. In contrast, influenza virus binding, clearance, and the associated cytokine response mediated by rSftpdCDM were similar to or better than observed in wild type mice. Similar findings of abnormal base-line   macrophage activation in the setting of a normal response to influenza virus were also described in studies utilizing an SP-D conglutinin (SftpdCong) fusion protein consisting of the N terminus and collagen domains of rat SP-D and the neck and CRD of conglutinin (27). Several interesting considerations regarding SP-D are raised from these observations. First, although the rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ alveolar macrophages have an abnormal histological appearance and base-line level of activity, they still are able to mount an effective and relatively normal inflammatory response to the infectious pathogen, influenza A. Second, although limited by the fact that only one infectious pathogen was tested, the normal cytokine response observed following influenza A viral infection suggests that the elevated base-line macrophage activity in Sftpd Ϫ/Ϫ mice is not due to the inability to clear a persistent low level infection. Finally, the divergence of macrophage base-line activity and the inflammatory response to viral infection in rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice suggests that SP-D regulates these processes through two independent pathways.
Previous studies by Gardai et al. (41) described simultaneous inhibitory and stimulatory roles for SP-D in macrophage regulation that were proposed to be mediated through two competing signaling cascades. In the first, SP-D inhibited NF-B and subsequent immune cell activation through binding of the CRD to the signal-inhibitory regulatory protein ␣ (SIRP␣). In the second, binding to SIRP␣ was inhibited by the presence of an infectious particle within the CRD, thereby allowing interactions between the collagenous domain or N terminus of SP-D and the macrophage-activating receptor calreticulin/CD91. This model is supported by evidence from the collagen deletion mutant described by Ogasawara and Voelker (35) as well as data derived from a recombinant fragment of SP-D consisting of only a trimeric CRD that indicate that both proteins inhibited alveolar macrophage activation, presumably through interactions with SIRP␣ (41). However, this model also predicts that the fully functional CRD of rSftpdCDM in the present study would bind SIRP␣ and inhibit macrophage activation. Moreover, rSft-pdCDM-mediated stimulation of macrophage activation through calreticulin/CD91 would be limited by the absence of a collagen domain. Therefore, the model proposed by Gardai et al. (41) might predict that the rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mouse would display an alveolar macrophage phenotype that is predominantly anti-inflammatory. The enlarged foamy macrophages, elevated metalloproteinases, and emphysema indicate that rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice are in a proinflammatory state at base line and seemingly contradict the model proposed by Gardai et al. (41). As with any mutational study, the current findings may be a result of an unanticipated change in the structure of the CRD or neck or N-terminal domains of rSftpdCDM as a result of the collagenous domain deletion. This qualification notwithstanding, an alternative explanation exists that may resolve this discrepancy. The present study suggests that SP-D regulates activation of the pulmonary immune system through two independent pathways. The first would control activation of alveolar macrophages in the presence of infectious particles and might involve the competing activities of SIRP␣ and calreticulin/CD91. Activation of these receptors would be appropriately balanced by the CRD and oligomerized N termi-nus of rSftpdCDM and would explain the wild type response to influenza virus exhibited by the rSftpdCDM Tgϩ /Sftpd Ϫ/Ϫ mice. A second pathway would control the base-line level of alveolar macrophage activation in the lung and in the absence of appropriate regulation elicit the phenotype of enlarged foamy macrophages, increased metalloproteinases, and emphysema. The results of the present study suggest that rSftpdCDM does not effectively activate this pathway. Similar aberrant macrophage activation was reported with SftpdCong, rSftpd Ser15,20 , and rSftpa/d (22,23,27). Therefore, whereas the receptors and signaling molecules that mediate this pathway are unknown, SP-D-mediated regulation of base-line macrophage activity, alveolar remodeling, and surfactant homeostasis requires a multimeric SP-D containing the collagen domain.
The failure of rSftpdCDM to correct the enlarged foamy macrophages, emphysema, and pulmonary phospholipid accumulations observed in Sftpd Ϫ/Ϫ mice is in contrast to earlier work with a recombinant fragment of SP-D consisting of a trimeric CRD and neck domain, which partially corrected the abnormal macrophages and surfactant pool sizes in Sftpd Ϫ/Ϫ mice (15). The reason why a single trimeric CRD would partially resolve the abnormalities that an oligomerized trimeric CRD fails to correct is uncertain, but it may reflect unanticipated changes that sometimes occur in mutant proteins or differences in the concentration of the mutant SP-Ds utilized in each study. Alternatively, this inconsistency may suggest a complex interplay between the N-terminal domain and the CRD in macrophage regulation and control of phospholipid pool sizes.
Although alveolar macrophages and type II epithelial cells equally contribute to surfactant phospholipid catabolism, previous studies demonstrated that SP-D regulates surfactant pool size by enhancing surfactant uptake by alveolar type II cells (14). Specifically, SP-D maintains normal surfactant ultrastructure by selectively interacting with small aggregate surfactant, thereby facilitating surfactant uptake by type II cells. The lipid binding specificity of collectins is mediated by the CRD in vitro (42)(43)(44)(45) and intranasal administration of trimeric CRDs decreased intra-alveolar lipid accumulations in Sftpd Ϫ/Ϫ mice (15,24). In contrast, the SP-D CRD and neck domain in the rSftpa/d fusion protein failed to correct surfactant lipid ultrastructure or lower the phospholipid levels when expressed in Sftpd Ϫ/Ϫ mice (22). In the present study, the addition of the SP-D N terminus to the CRD failed to correct surfactant ultrastructure or increased phospholipid levels. Therefore, normal surfactant ultrastructure and the uptake of surfactant by type II cells depend on the collagen domain.
Although rSftpa/d included the neck and CRD of SP-D, analysis of large and small aggregate surfactant lipids demonstrated that rSftpa/d partitioned with large aggregate surfactant in a manner similar to SP-A (22). In contrast, rSftpdCDM, which included the neck, CRD, and N-terminal domains of SP-D, partitioned with small aggregate surfactant like the full-length SP-D protein. Taken together, these studies suggest that whereas phospholipid binding in vitro may be mediated by the CRD, the structural features of SP-D (and SP-A) that influence partitioning between small versus large aggregate surfactant are contained within the N-terminal domain.
In summary, our study provides further support for the importance of the distinct molecular domains of SP-D in mediating the complex functions of this protein. The finding that the collagen deletion mutant fails to correct the abnormal lung morphology or surfactant lipid homeostasis of Sftpd Ϫ/Ϫ mice supports a role of the collagen domain that is beyond proper spacing of trimeric subunits for bacterial aggregation. Assimilating the results of this study with those of prior reports on SP-D reveals that whereas our understanding of SP-D is improving, inconsistencies still exist. The CRD is critical for binding to lipopolysaccharide, viruses, bacteria, fungus, and lipids. Nonetheless, SP-D interactions with small aggregate surfactant and uptake of surfactant by alveolar type II cells in vivo do not depend on the CRD alone. The collagen domain is required for surfactant lipid structure and metabolism, but it is not needed to effectively aggregate bacteria or to suppress inflammatory responses to influenza A virus. The N-terminal domain is critical for partitioning with large versus small aggregate surfactant lipids in adult mice, and interactions between interchain N-terminal domains are essential for oligomerization, which, in turn, influences CRD binding affinity, phospholipid catabolism, and the inflammatory response to infectious pathogens. Finally, at the time of this writing, all SP-D mutant proteins that delete or substitute even a single domain of SP-D fail to fully correct the enlarged, foamy macrophages that are characteristic of Sftpd Ϫ/Ϫ mice, suggesting that this function of SP-D is dependent on a full-length, multimeric protein.