Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Surfactant protein (SP)-D belongs to the collectin class of C-type lectins (1). Collectins are all large, extended oligomeric proteins characterized by the presence of a collagen-like sequence (CL domain) linked to a calcium-dependent carbohydrate-recognition domain (CRD) by a short trimeric coiled-coil region (2). Complementary DNA (cDNA) analyses of human (3), bovine (4), rat (5), and mouse (6) SP-D reveal strongly conserved protein sequences. The predominant secreted form of SP-D, a covalently linked tetramer of trimeric units or 12 monomeric subunits, appears as an X-like structure in rotary-shadowed electron micrographs with four spokelike arms extending from a central hub formed by the amino-terminal domains (7).

SP-D was originally termed a surfactant protein because it is secreted by type II pulmonary epithelial cells and has striking structural and biochemical similarities with the surfactant apoprotein SP-A (8). Despite these similarities, the distribution of SP-A and SP-D in bronchoalveolar lavage (BAL) fluid is quite distinct and the lipid-binding properties of SP-A and SP-D differ markedly. More than 90% of the total SP-A recovered by BAL is associated with the dense, aggregated form of surfactant enriched in tubular myelin subtype (11, 12). Reconstitution studies have shown that SP-A, in combination with SP-B, is a critical component of the classic square lattice of tubular myelin (13).

In contrast, SP-D is distributed between both the sedimentable and nonsedimentable forms of surfactant (11). Approximately 70% of the total BAL SP-D (11) and 10 to 15% of the total BAL phospholipid (17) is not sedimented by prolonged high-speed centrifugation. Potential interactions between SP-D and surfactant in either the sedimentable or nonsedimentable fractions might be mediated through a lectin-mediated binding to phosphatidylinositol (Pl) (18, 19), a component of most mammalian surfactants. Although these data suggest that interactions between surfactant and SP-D might occur, there has been no report on the potential structure or properties of surfactant subtypes containing SP-D. The present article reports on a study of the ultrastructure and surface properties of complexes formed between certain phospholipids and SP-D, with and without SP-B.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

cDNA Isolation

A partial-length rat SP-D cDNA (rD-1), from nucleotides 806 to 1,150, was amplified from reverse-transcribed rat type II cell RNA by polymerase chain reaction (PCR), using sense and antisense 21-mer oligonucleotide primers designed from the published rat SP-D cDNA sequence (5). The RNA for this reaction was isolated by the guanidinium isothiocyanate method (20) from > 90% pure rat type II cells (21). The 335-base pair (bp) product was ligated into a TA cloning vector (Invitrogen, San Diego, CA) and the identity of the insert checked by nucleotide sequencing. rD-1 was purified from low-melt agarose and labeled with [³²P] cytidine triphosphate for use as a probe in cDNA library screening.

Recombinant clones from an oligo (dT) and random primed B6/CBA female mouse lung cDNA library in the Lambda ZAP II vector (Stratagene, La Jolla, CA) were grown in XL1-Blue Escherichia coli on NZCYM plates at a density of approximately 50 × 10³ plaques/15-cm plate. The library was screened on nitrocellulose filter replicas using in situ hybridization with ³²P-rD-1 as a probe. Duplicate filter-lifts were prehybridized in 5× saline sodium citrate (SSC), 50 mM NaH₂PO₄ (pH 6.8), 5× Denhardt's solution, 10% Dextran sulfate, 25% formamide, and 20 µg/ml calf thymus DNA, and hybridized in the same solution containing 1 × 10⁶ counts per minute/ml ³²P-rD-1 at 42°C overnight. The filters were washed three times in 2× SSC at room temperature and once at 37°C before autoradiography on Kodak XAR film (Sigma, St. Louis, MO). Positive clones identified on the duplicate filters were isolated by two to three further rounds of screening using identical conditions. In vivo excision of the pBluescript plasmid from the Lambda ZAP II vector was performed using the manufacturer's protocol (Stratagene). The inserts were sequenced in both directions using an Applied Biosystems Automated DNA Sequencer (PEApplied Biosystems, Foster City, CA).

The 5' extension of clone mD-1 was obtained using anchor-based PCR and the 5'-AmpliFinder Rapid Amplification of cDNA Ends (RACE) kit (Clonetech, Palo Alto, CA). Briefly, nested antisense primers were constructed to anneal to bases 322 through 345 (P1) and 292 through 318 (P2), respectively. Mouse lung messenger RNA (2 µg) was transcribed using AMV reverse transcriptase and P1. The cDNA was purified on a glass matrix support after alkaline hydrolysis of the RNA. The 5' anchor oligonucleotide was ligated to the transcribed cDNA using T4 ligase, and the 5'-extension of the mouse surfactant protein (mSP)-D cDNA was amplified using the anchor primer and P2 containing 12 additional nucleotides at the 5' end of the primer to facilitate cloning of the product.

Expression of Recombinant Mouse SP-D

Protocols for the expression of mouse SP-D in mammalian cells were modified from those reported for the expression of rat SP-D by Crouch and colleagues (22). A full-length mSP-D cDNA was obtained by ligating the amplified 5'-extension product to the partial-length cDNA isolated by cDNA library screening. The full-length cDNA (1,180 bp) was then ligated into the polylinker site of the mammalian expression vector pEE14 (Celtech, Glasgow, UK).

pEE14D (10 µg/dish) was transfected by Ca₂PO₄ precipitation into CHO K1 cells (ATC no. CCL61) grown in glutamine-poor Dulbecco's modified Eagle's medium containing 10% dialyzed fetal calf serum in 5% CO₂. The day after transfection, selective pressure was applied with 25 µM methionine methosuximide (MSX), a glutamine synthase inhibitor. Surviving colonies 14 to 21 d later were subcloned and placed under additional pressure with increasing MSX concentrations, to a maximum of 1 mM. The medium from stable pools was tested for SP-D content by a dot-blot immunoassay using a monospecific polyclonal antibody (23). High-producing pools in 200 µM MSX were expanded in triple-layer T75 flasks containing an additional 50 µg/ml ascorbic acid. The medium was collected 24 h later for the isolation of recombinant mouse SP-D (rmSP-D). After spinning out cell debris at 150 × g for 10 min, CaCl₂ was added (to a final concentration of 5 mM) and the medium was passed over a maltose-sepharose affinity column at 4°C. The column was washed with 20 mM Tris, 150 mM NaCl, and 5 mM CaCl₂ (pH 7.4), and eluted with a 100 mM NaBorate buffer containing 10 mM ethylenediaminetetraacetic acid (EDTA), pH 10.5. The fractions containing SP-D were identified by sodium dodecyl sulfate- polyacrylamide gel electrophoresis (SDS-PAGE), pooled, and dialyzed against 20 mM Tris, 100 mM NaCl, and 10 mM EDTA for storage at 20°C. SP-D was quantitated by the Lowry method using bovine serum albumin as a standard (24). For electron microsopy, rmSP-D was diluted to 20 µg/ml in 10 mM ammonium bicarbonate, 50% glycerol (pH 7.5), and sprayed onto freshly cleaved mica. The samples were placed in a Balzers vacuum evaporator and shadowed with platinum at an angle of approximately 8° (25).

Binding of rmSP-D to PI-Containing Phospholipid Mixtures

Dipalmitoylphosphatidylcholine (DPPC), egg phosphatidylglycerol (PG), and brain PI were obtained from Avanti Polar Lipids (Birmingham, AL) and used without further purification. DPPC-PI (7:3, wt/wt) containing 0.1 µCi ¹⁴C-DPPC (Amersham Co., Arlington Heights, IL) was dried under nitrogen, rehydrated in 50 mM 3(N-Morpholino)-2-hydroxy-propanesulfonic acid (MOPSO), 140 mM NaCl, 0.1 mM EDTA (pH 6.9), and briefly sonicated. A total of 250 µl of the vesicle suspension (500 µg phospholipid) was then incubated in the presence or absence of rmSP-D (100 µg) and in the presence of CaCl₂ (2 mM) or EDTA (10 mM) overnight at 37°C. After incubation, each sample (750 µl final volume) was loaded onto a 0.25- to 1.0-M continuous sucrose gradient and spun at 150,000 × g_av (SW 41; Beckman, Fullerton, CA) for 18 h at 4°C. After centrifugation, 0.5-ml fractions were collected from the top of the gradient and analyzed for ¹⁴C content in a scintillation counter and by SDS-PAGE and silver staining for rmSP-D content. Control experiments were performed with rmSP-D without lipid.

Reconstitution of Proteolipid Mixtures

Phospholipid mixtures (DPPC-PG or DPPC-PI; 7:3 wt/wt) alone or with dog SP-B (10% by weight) in chloroform were dried in a round-bottom flask under vacuum. The dry lipid or lipoprotein film (1.5 to 3.0 mg) was solubilized in 1 ml 5 mM Tris and 100 mM 1-0-octyl-B-D-glucopyranoside (OGP) at 40°C. After diluting the OGP to a final concentration of 20 mM with 10 mM Tris, 145 mM NaCl, and 1 mM EDTA (pH 7.4), the samples were dialyzed against five changes of the same buffer at 4°C to remove the nonionic detergent. The concentration of the reconstituted phospholipid mixtures in buffer was determined by phosphorus assay (26). CaCl₂ was added to a final concentration of 5 mM before the samples were incubated for 12 h at 37°C with or without dog SP-A or rmSP-D (20% by weight). Methods for the isolation of dog SP-A and SP-B have been described in detail previously (14). SP-A and SP-B were quantitated by the procedure of Lowry and coworkers (24) and fluorescamine assay (27), respectively.

Ultrastructure of Reconstituted Phospholipid Samples

Reconstituted samples were fixed in suspension in 2% glutaraldehyde and 1% osmium tetroxide in sodium cacodylate, pH 7.4, for 1 h at room temperature, then spun at 10,000 × g and stored overnight in fixative at 4°C. This fixative was replaced by 2% aqueous uranyl acetate for 24 h at 4°C. The pellet was then rapidly dehydrated in cold graded acetone solutions, infiltrated with increasing concentrations of LX-112 epoxy (Ladd Research Industries, Burlington, VT), and finally embedded in LX-112. Thin sections were stained in 5% aqueous uranyl acetate and 0.8% alkaline lead citrate, and examined in a Zeiss 10 transmission electron microscope.

Secretions of Type II Cells in Culture

Rat type II alveolar epithelial cells were maintained in long-term culture on collagen gels with an apical air surface using previously described methods (28). Briefly, freshly isolated rat type II cells were cultured at density of 1 million cells per square centimeter on collagen-coated Millicell CM membranes (Millipore, Bedford, MD) in Eagle's minimum essential medium that contained 2% Ultroser G (Biosepra, Cedex, France), 100 U penicillin/ml, and 50 µg gentamicin/ml. Cells were allowed to adhere for 48 h, at which time the medium was removed from the apical surface and the cells were left exposed to ambient air. Media were replaced in the well outside the membrane every other day for the duration of the experiment. To analyze the phospholipid composition of the secretions, cells were incubated for 22 h in basally applied medium containing ¹⁴C acetate on either Day 1 or Day 21. After 22 h the apical surface liquid was collected, lipids were extracted, and the phospholipids were separated by two-dimensional thin-layer chromatography as previously described (29).

After 21 d in culture, some membranes were fixed for 2 h in 1% paraformaldehyde, 2% glutaraldehyde, and 0.1 M sodium phosphate buffer (pH 7.4), then postfixed overnight in 1.5% osmium tetroxide in 0.1 M veronal acetate buffer, pH 7.4. Tissue blocks were stained in 1.5% uranyl acetate buffered by 0.05 M sodium maleate (pH 5.2), and quickly dehydrated in cold acetone and propylene oxide before embedding in LX-112 resin. Thin sections were stained in 5% aqueous uranyl acetate and 0.8% lead citrate, and examined in a Zeiss 10 transmission electron microscope.

Surface Activity Measurements

The effect of SP-D on the properties of the surface film initially adsorbed from reconstituted samples was evaluated in a pressure-driven captive bubble surfactometer using an experimental protocol we described previously in detail (30). Samples were tested at a phospholipid concentration of 200 µg/ml in 10 mM Tris, 140 mM NaCl, 0.1 mM EDTA, and 5 mM CaCl₂ (pH 6.9), at 37°C. After allowing the sample to adsorb to the expanded bubble surface for 30 min, a surface tension-area isotherm was inscribed by a stepwise increase in system pressure every 60 s from 0.5 to 2.8 atm. Following this initial slow isotherm, nine compression-expansion cycles were repeated over 2 to 3 min. The surface tension was calculated from the bubble dimensions to derive the surface tension-area isotherms (31). Isotherm data were normalized to relative areas of 1.0 at onset of compression. For statistical purposes, curves were first linearly interpolated (201 points, using the Igor program by Wavemetrics, Lake Oswego, OR), then groups were averaged and standard errors computed.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Mouse SP-D cDNA Isolation

Three clones containing overlapping inserts of 1,079, 518, and 765 bp, respectively, were obtained by screening the mouse lung cDNA library with ³²P-rD-1. The combined sequences encompassed 1,230 nucleotides of mSP-D sequence, including 155 bp of 3' untranslated sequence. An additional 51 bp of translated sequence and 45 bp of 5' untranslated sequence were amplified by anchored RACE- PCR to complete the entire 1,326-bp mSP-D sequence. The 374 amino acids of the translated sequence are identical to the mouse sequence published by Motwani and colleagues (6).

rmSP-D Expression and Characterization

rmSP-D was secreted by CHO K1 cells transfected with PEE14D, accumulating in the media in concentrations of approximately 2 to 4 µg/ml/24 h. By dot-blot analysis, approximately 80% of the SP-D in the medium was recovered in the eluate fraction of a maltose-affinity column. Consistent with previous descriptions, rmSP-D had an apparent relative molecular mass (M_r) of 40 kD under reducing conditions (Figure 1a). The monomer M_r was slightly larger than the M_r of 37,680 D calculated from the primary sequence. N-glycanase digestion reduced the M_r by approximately 2 kD, consistent with a single N-linked oligosaccharide at Asn-70 in the collagen domain, and collagenase digestion reduced the M_r to 18 to 20 kD (data not shown).

View larger version (138K): [in this window] [in a new window]   Figure 1.   Purification and characterization of rmSP-D. (a) Coomassie-stained SDS-PAGE of reduced rmSP-D purified by maltose-affinity chromatography from media collected from pEE14D cells. Mr indicated in kilodaltons. (b) Electron microscopy of rmSP-D demonstrating two tetrameric (12 monomers) and one dimeric (six monomers) molecules. Bar: 100 nm. The rotary-shadowing electron microscopy was performed by Dr. Harold P. Erickson, Duke University Medical Center, Durham, NC.

Electron microscopy of rmSP-D demonstrated a mixture of forms including molecules with four rodlike arms, each about 50 nm in length, arising from a central hub and ending in a globular domain (Figure 1b) as described for human and rat SP-D (7, 22). In addition to this cruciform morphology, molecules with only two arms and poorly resolved aggregates were seen. We did not try to quantitate the relative distribution of forms, but by gel permeation chromatography, > 90% of the rmSP-D had an apparent M_r in excess of 2 × 10⁶, consistent with previous reports for rat and human SP-D (7, 9).

Binding of rmSP-D to PI-Containing Phospholipid Mixtures

rmSP-D bound to vesicles made with DPPC and PI in a calcium-dependent fashion (Figure 2). In 2 mM CaCl₂ all the detectable SP-D comigrated with the main phospholipid peak in fractions 13 through 17 of an equilibrium density sucrose gradient. In 2 mM EDTA there was no detectable rmSP-D with the main lipid peak in fractions 11 through 14; rather, the rmSP-D was all detected near the top of the gradient unassociated with phospholipid. In control gradients without any added lipid, free rmSP-D remained near the top of the gradient (fractions 4 through 7) in either 2 mM CaCl₂ or 2 mM EDTA (not shown).

View larger version (30K): [in this window] [in a new window]   Figure 2.   Sucrose-gradient isolation of DPPC-PI-SP-D recombinants showing calcium-dependent binding of rmSP-D to DPPC-PI vesicles. The phospholipid content of each fraction is expressed as a percent of the total recovered from a gradient run in 2 mM CaCl2 (closed symbols) or 2 mM EDTA (open symbols). The distribution of rmSP-D is demonstrated by silver-stained SDS-PAGE (odd-numbered fractions from 5 to 23) aligned above the phospholipid profile. The small phospholipid peak in fractions 5 through 8 in EDTA was also present in the absence of rmSP-D (not shown).

Ultrastructure of Reconstituted Proteolipid Samples

All samples for microscopy and surface activity measurements were prepared by detergent dialysis using OGP. Others have reported residual detergent of less than 0.05% after 4 to 5 changes of dialysis buffer, corresponding to approximately 1 detergent/400 lipid molecules (32). In previous studies we found the ultrastructure of reconstituted surfactant was similar with and without the use of detergent (14).

Lipids alone. After a prolonged incubation in 5 mM CaCl₂, the two lipid mixtures (DPPC-PG and DPPC-PI) formed mostly aggregates of unilamellar vesicles of various sizes (approximately 200 to 400 nm in diameter) interspersed with occasional larger multilamellar structures (Figures 3a and 3b). The precipitate in these samples is the result of salt precipitation during fixation in suspension.

View larger version (83K): [in this window] [in a new window]   Figure 3.   Recombinants of SP-A, SP-B, and SP-D with phospholipid mixtures of DPPC-PG and DPPC-PI. (a) DPPC-PI without protein. (b) DPPC-PG without protein. (c) DPPC-PI + SP-B. (d) DPPC-PG + SP-B. (e) DPPC-PI + SP-D. ( f ) DPPC-PG + SP-D. (g) DPPC-PI + SP-A. (h) DPPC-PG + SP-A. All panels are the same magnification. Bar: 1 µm. See MATERIALS AND METHODS for description of ultrastructure.

Lipids + SP-B. The addition of SP-B caused extensive alterations in the ultrastructure of the phospholipids (Figures 3c and 3d). Few vesicles remained intact, and most were rearranged into large closely packed multilamellated structures. As we described previously in greater detail (14), other structures including densely aggregated discoidal particles were also present. There was no apparent difference between the PG- and PI-containing mixtures.

Lipids + rmSP-D. The DPPC-PI vesicles were aggregated by the calcium with and without SP-D, but an amorphous particulate material, presumably rmSP-D, decorated the aggregated DPPC-PI vesicles (Figure 3e). This amorphous material was not present in the DPPC-PG- SP-D recombinants (Figure 3f), which were indistinguishable from DPPC-PG alone (Figure 3b).

Lipids + SP-A. In contrast to the lack of a clear effect of rmSP-D on the ultrastructure of DPPC-PG and DPPC-PI mixtures, the addition of SP-A generated a few large multilamellar structures consisting of multiple concentric bilayers separated by a 15 to 20-nm-wide interspace filled with particulate material (Figures 3g and 3h). There was no appreciable difference in ultrastructure seen between the two species of anionic phospholipids.

Lipids + SP-B + SP-A. The addition of both SP-A and SP-B to DPPC-PI or DPPC-PG in the presence of calcium resulted in a major structural reorganization of the phospholipid, with the formation of large multilamellar aggregates and tubular myelin (Figures 4a and 4b). The tubular myelin figures in both the PG- and PI-containing samples were similar in their abundance and structure to those we described previously with DPPC-PG (14). The size of the lattice squares was about 50 nm/side, similar to the size of the tubular myelin found in the pulmonary alveolus. Residual regions of densely aggregated discoidal particles, as described for SP-B-lipid recombinants, were also observed.

View larger version (153K): [in this window] [in a new window]   Figure 4.   Electron microscopy of protein-lipid recombinants demonstrating cooperative effects between collectins and SP-B. (a) DPPC-PI + SP-B + SP-A demonstrating multilamellar aggregates like those formed with SP-A alone and square lattice of typical tubular myelin. (b) DPPC-PG + SP-B + SP-A demonstrating similar morphology to DPPC-PI + SP-B + SP-A. (c) DPPC-PI + SP-B + SP-D demonstrating atypical hexagonal tubules. A dense aggregate of SP-B discs is seen at the lower right. (d) DPPC-PG + SP-B + SP-D demonstrating lamellar sheets characteristic of SP-B containing recombinants but no significant effect of SP-D. Magnification of all panels is the same. Bar: 1 µm.

Lipids + SP-B + SP-D. rmSP-D had no effect on DPPC-PG mixtures reconstituted in the presence of calcium and SP-B (Figure 4d). The predominant structure was identical to the closely packed multilamellated form seen with the SP-B-lipid recombinants.

View larger version (101K): [in this window] [in a new window]   Figure 5.   Differences in tubular morphology of recombinants containing SP-D and SP-A. (a) DPPC-PI + SP-B + SP-D demonstrating atypical hexagonal tubules and target-like appearance on cross section. Bar: 500 nm. (b) Higher-power longitudinal section showing the crossed strands (presumably rmSP-D) spanning the 90- to 110-nm interspace between tubule walls. Bar: 250 nm. (c) DPPC-PI + SP-B + SP-A demonstrating the formation of tubular myelin with PI and the difference in geometry of the tubules compared with (a). Bar: 500 nm.

Cultured Type II Cells

Abundant secreted material was seen overlying the apical surface of type II cells cultured for 21 d with an apical air surface (Figure 6). Along with the contents of secreted lamellar bodies and varied vesicular structures, two distinct tubular forms were observed. In addition to classic tubular myelin with a 50 to 55-nm-square lattice, there were atypical tubules with diameters of 90 to 110 nm. These two tubular morphologies were similar to those reconstituted in vitro with SP-A (Figure 4b) and SP-D (Figure 4c). In cross section, the larger polygonal tubules had the same target-like appearance as the PI-SP-D recombinants (Figure 7b). In the apical air surface culture system, the PI/PG ratio in the secreted lipids was 6.25 on Day 21 (n = 2), compared with 0.31 from cells cultured on plastic for 1 d (29).

View larger version (179K): [in this window] [in a new window]   Figure 6.   Surfactant secretions overlying type II cells cultured with an air surface for 21 d. (a) Multiple ultrastructural forms demonstrated, including secreted lamellar body contents with dense cores (asterisk), atypical 90 to 110-nm hexagonal tubules (single arrow), and typical 50 to 55-nm-square lattice of tubular myelin (double arrow). Bar: 250 nm. (b) Higher power of the atypical hexagonal tubules demonstrating the target-like appearance on cross section. Bar: 250 nm.

View larger version (25K): [in this window] [in a new window]   Figure 7.   Initial compression surface tension-area isotherms without and with SP-D. Four captive bubble experiments used DPPC, PI, and SP-B alone, and four used DPPC, PI, and SP-B with added SP-D. Individual data curves were interpolated to obtain plotted average curves and standard error bands (vertical hatching). The addition of SP-D produced initial compression isotherms that were steeper (less compressible, more effective) than controls.

Surface Activity

First compression surface tension-area isotherms of DPPC: PI:SP-B with and without rmSP-D are shown in Figure 7. Although there was some variability in the behavior of individual samples, in each of four experiments the presence of SP-D shifted the first compression isotherms of freshly adsorbed films to the right. The steeper decline in surface tension with decreasing area indicates a reduced compressibility of the film. With subsequent rapid compressions of the film, all samples achieved a surface tension < 5 mN/m with similar minimal compression.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Our observations indicate that SP-D can interact with vesicles formed from phospholipids normally found in the alveolar space, and that, similar to the effects of SP-A, there is an interaction between SP-D, SP-B, and phospholipids that results in the formation of highly ordered tubular arrays. The effects of SP-A and SP-D on lipid structure differ in a manner consistent with known differences in quaternary structure (7, 33). Both SP-A and SP-D are collectins assembled as homopolymers of trimeric subunits (reviewed in Ref. [2]). Each subunit consists of a collagen-like triple helix linked to a trimeric globular carbohydrate recognition domain. The six rodlike collagen arms of SP-A are aligned so that the molecule resembles a bunch of flowers about 20 nm long. The four longer rodlike arms of SP-D emanate from a central hub, giving the protein a distinctive cruciform appearance with a more extended length of approximately 100 nm. Both smaller and larger oligomeric forms of SP-D have also been observed. The recombinant form of SP-D used in this study had similar biochemical properties and quaternary structure to rat and human SP-D characterized by others (7, 22), including some heterogeneity in the degree of oligomerization. The extent of oligomerization markedly affects the activity of SP-D in virus aggregation assays (34), but we did not attempt to distinguish the relative activity of tetrameric and dimeric forms of rmSP-D in our reconstitution system.

The tubular arrays formed by SP-A and SP-D differed in both shape and size. As described previously, SP-A stabilizes a square lattice about 50 nm/side, often interspersed with mutilamellar aggregates of bilayers separated by 20 nm (13, 14, 35). The tubular arrays we observed with SP-D were more homogeneous, larger, and mostly hexagonal in shape. The larger interbilayer distance of 80 to 90 nm with SP-D is consistent with the dimensions of the isolated protein (7, 22). The target-like appearance of the SP-D recombinants (see Figures 5a and 6a) might be due to the central electron density of the amino-terminal hub and a concentric electron density representing the oligosaccharide at asn-70 in SP-D, but reconstitution studies using appropriately site-modified SP-D would be necessary to confirm this suggestion (7, 19, 33, 36).

Although both collectins bind phospholipids using the CRD (19, 34), the type of interaction and lipid selectivity varies. SP-A binds phospholipids hydrophobically with a preference for fully saturated phospholipids, including DPPC (37), whereas SP-D appears to interact in a more limited way with glycolipids, using calcium- and carbohydrate-dependent mechanisms (18, 19). PI was selected as the glycolipid for this study for two reasons. First, PI is the most abundant glycolipid in alveolar surfactant, constituting 2 to 10% of the total phospholipid, depending on the species, the developmental stage, and the presence or absence of lung injury. Second, PI carries a negative charge that is possibly important for interactions with the cationic protein SP-B (40, 41). Our ultrastructural results correlate with the different collectin lipid-binding specificities. We found little difference in the morphology of complexes with SP-A and either PG- or PI-containing lipid mixtures, consistent with the observed preference of SP-A for DPPC over other phospholipids (39, 42). The tubular myelin formed with PI was not apparently different from the complexes formed from PG-containing mixtures in this and previous studies, suggesting that SP-A was not sensitive to the nature of the anionic phospholipid headgroup (13, 14). In contrast, the SP-D-mediated conversion of lamellar sheets into tubular arrays was PI-dependent. On the basis of previous studies, this interaction is likely dependent on specific lectin interactions (18, 19), but we did not specifically test the effects of competing sugars in the present study. These results indicate that both collectins can mediate the formation and stabilization of lipid tubular arrays, but that different types of collectin-lipid interactions are involved.

The formation of tubular arrays by either SP-A or SP-D was dependent on the presence of SP-B. The transformation of both the PG- and PI-containing vesicles into large multilamellated sheets and discoidal aggregates by SP-B is consistent with our previous ultrastructural studies (14) and with the pronounced lytic and fusogenic activity of SP-B (41, 43, 44). These studies show that the membrane actions of SP-B are enhanced by the presence of anionic phospholipids, but our results suggest that PI and PG are interchangeable, indicating the importance of charge rather than headgroup structure in lamellar sheet formation. The precise structural basis for SP-B activity is unknown but is presumably associated with both the strongly amphipathic and cationic character of SP-B (45). The large lamellar sheets generated by the action of SP-B on membranes appear to be formed by the planar association of homogeneously sized discoidal membrane fragments ringed by SP-B. The dimeric nature of SP-B may be important for close association of individual discs. As previously noted, the diameter of these discs is approximately equal to one side of a tubular myelin square (14, 46) and, as found in this study, one side of a SP-D-dependent tubular hexagon (Figures 4, 5, and 6).

Although the discoidal aggregates generated by SP-B resemble some of the substructures reported in lamellar bodies (35), their existence in vivo is unproven and the apparent correlation between the size of the membrane fragments generated by SP-B and the dimensions of the tubular aggregates may be simply fortuitous. Even if SP-B ringed discs are used to assemble the tubular arrays, the SP-B may redistribute in the mature structures. It is not clear whether the structural reorganizations we observe depend on direct binding of the collectins to SP-B or on separate but cooperative effects on phospholipid structure. Our results, implying that the cooperativity is not strictly dependent on a specific amino acid sequence or quaternary structure, probably favor the latter possibility. In this study we did not examine the possible role of SP-C in the reconstituted structures. In a prior study, SP-C did not appear to have an effect on the ultrastructure of reconstituted tubular myelin (14), but the topology and function of this relatively abundant surfactant apoprotein will eventually have to be considered.

The reconstitution experiments give some insight into the interactions required for the formation of the highly ordered surfactant tubular arrays but do not directly address the role of SP-D in ordering surfactant lipids in vivo. We observed both regular square tubular myelin and the expanded atypical tubular hexagons in the secretions of cultured type II cells that had converted from predominantly PG synthesis to PI synthesis, indicating that the assembly of the SP-D-dependent tubular arrays was not simply a test-tube phenomenon. A similar switch from PG to PI synthesis has been reported in a number of clinical situations, including adult respiratory distress syndrome (47) and alveolar proteinosis (48). The latter condition is particularly interesting because high levels of SP-D are also reported (49). The ultrastructure of the abundant multilamellar structures in human alveolar proteinosis is very heterogeneous (16) but includes intricately arranged hexagonal tubules with spokelike and target-like substructures, coincidently termed type D tubular myelin by Takemura and colleagues (15). We have no proof that these structures contain SP-D and PI, but their overall organization is similar to our SP-D-PI recombinants and the hexagonal tubular arrays seen in the air surface type II cell secretions.

The functional correlate to the complex tubular reorganization induced by both SP-A and SP-D is uncertain. In vitro SP-A enhances adsorption of phospholipids containing SP-B to an air-liquid surface (50) and appears to promote sorting of lipids during the adsorption process (51). Similar to these results, we found that SP-D reduced the relative area compression required to achieve a low surface tension during the first isotherm with PI/SP-B-containing samples, suggesting a selective adsorption of DPPC to the surface in the presence of SP-D. Perhaps the cooperative action of the collectins and SP-B results in the segregation of phospholipids into domains presorted for selective adsorption. This speculation must be balanced by the recent finding that mice deficient in SP-A and lacking in tubular myelin have apparently normal lung function and only subtle alterations in surfactant function (52). Our observations nevertheless suggest that under particular conditions favoring PI synthesis and SP-D accumulation, SP-D might contribute to the assembly of surface-active tubular surfactant structures and compensate for a deficiency of SP-A.