Nucleic Acid Is a Novel Ligand for Innate, Immune Pattern Recognition Collectins Surfactant Proteins A and D and Mannose-binding Lectin*

Collectins are a family of innate immune proteins that contain fibrillar collagen-like regions and globular carbohydrate recognition domains (CRDs). The CRDs of these proteins recognize various microbial surface-spe-cific carbohydrate patterns, particularly hexoses. We hypothesized that collectins, such as pulmonary surfactant proteins (SPs) SP-A and SP-D and serum protein mannose-binding lectin, could recognize nucleic acids, pentose-based anionic phosphate polymers. Here we show that collectins bind DNA from a variety of origins, including bacteria, mice, and synthetic oligonucleotides. Pentoses, such as arabinose, ribose, and deoxyribose, inhibit the interaction between SP-D and mannan, one of the well-studied hexose ligands for SP-D, and biologically relevant D -forms of the pentoses are better competitors than the L -forms. In addition, DNA and RNA polymer-related compounds, such as nucleotide diphos-phates and triphosphates, also inhibit the carbohydrate binding ability of SP-D, or (cid:1) 60 kDa trimeric recombinant fragments of SP-D that are composed of the (cid:1) -hel-ical

Nucleic acids such as DNA and RNA of higher organisms and microbes are long, pentose sugar-based anionic phosphate polymers. Apoptotic cell death results in the fragmentation of DNA and its subsequent display on the surfaces of the cells as blebs (1). Inefficient removal of apoptotic cells leads to disintegration of their contents and the formation of necrotic cells. These leaky cells eventually release their intracellular contents, and many of these components elicit tissue inflammation. Furthermore, certain pulmonary pathogens, such as Pseudomonas aeruginosa, actively secrete DNA into the extracellular matrix to establish an active biofilm and chronic infections (2). Removal of dying cells and their components, including DNA, is therefore essential to maintain inflammation-free tissues and to prevent autoimmune diseases (1). Although the multiple pathways and proteins involved in the apoptotic process have been studied in great detail, clearance of these cells and their debris is poorly understood.
Collectins are a class of innate immune proteins that are often secreted into body fluids and extracellular cavities (3). Well-studied members of the collectin family include surfactant proteins (SPs) 1 A and D and mannose-binding lectin (MBL). These proteins, particularly SP-D, bind many of the inflammation-causing ligands derived from microbes (3), such as fungal cell wall glucans (4), lipopolysaccharides (5), peptidoglycans and lipoteichoic acids (6), and lipids (7). We have shown recently that SP-D also binds to proteoglycans with long glycosaminoglycan chains, such as decorin, via both protein-protein and protein-carbohydrate interactions, and that SP-D-decorin interactions may be related to reducing tissue-specific inflammation (8). Collectins not only bind to these purified ligands but also recognize them on cell surfaces (3). Apoptotic cells, for example, contain carbohydrate polymers such as DNA on their surfaces, and accumulation of these dying cells, as well as excess free DNA, is known to cause inflammation and septic shock (9,10). Fragments of bacterial DNA with specific CpG motifs, however, act as immunostimulatory molecules (11). Regulatory mechanisms and all the proteins that are involved in these pathways are not clearly identified.
A typical collectin is composed of polypeptide chains containing a short, interchain, disulfide bond-forming N-terminal domain, a collagen-like region with Gly-X-Y repeats (where X is any amino acid and Y is often hydroxyproline or hydroxyl-ysine), an ␣-helical hydrophobic neck region and a C-terminal globular carbohydrate recognition domain (CRD) (12,13). Three of these polypeptides form a trimeric subunit (14 -16), which assembles to form higher order structures. The oligomeric assembly of SP-D (13,17) resembles that of an "X" (4 subunits) or "asterisk" (Ͼ10 subunits) with a prominent central "hub," whereas the other collectins SP-A and MBL (2-6 subunits) and a non-collectin, the complement system protein C1q (6 subunits), appear as a "bouquet of flowers" (3,17,18). SP-A further assembles as unidimensional (19) and multidimensional (20) fibers. Although the trimeric subunits of the collectins have limited affinity (M) for several carbohydrate targets, their oligomeric assembly provides high avidity so that these proteins bind to ligands selectively and with high affinity (nM to pM). Collectins bind microbes primarily via hexose sugars and enhance their phagocytosis (16), and whether these proteins can bind pentose sugars was unknown. We hypothesized that collectins could bind free RNA and DNA, as well as nucleic acid present on the apoptotic and necrotic cells, and promote their clearance.
In this study, we used pentoses, ribonucleoside phosphates, deoxyribonucleoside phosphates, synthetic short oligonucleotide fragments, DNA isolated from bacteria and mouse cells, and the apoptotic forms of Jurkat cells and mouse alveolar macrophages. Evidence obtained by electrophoretic mobility shift assay (EMSA), electron microscopy, surface plasmon resonance (SPR), fluorescent-activated cell sorting (FACS), and confocal microscopy experiments show that collectins, particularly SP-D, bind DNA obtained from many sources. Furthermore, SP-D binds DNA both via their CRDs and collagen-like regions. We used two recombinant fragments of SP-D in this study to further determine the roles of different domains of the protein: the SP-D(n/CRD) fragment, which lacks the N-terminal segment and the entire collagen-like region; and the SP-D(GXY) 8 (n/CRD) fragment, which is similar to SP-D(n/CRD) except that it also contains eight of the Gly-X-Y repeats. Both of these proteins also recognize DNA. SP-D(GXY) 8 (n/CRD) binds nucleic acids present on apoptotic cells. Therefore, our findings establish a novel ligand for collectins and suggest that these proteins could enhance the clearance of nucleic acids and play an important role in limiting inflammation.

EXPERIMENTAL PROCEDURES
DNA-Plasmid DNAs were purified by Midiprep procedures (Qiagen, West Sussex, England) and digested with EcoRI or BamHI (Promega, Southampton, England). Genomic DNA was isolated from the mouse lung by a DNeasy DNA isolation kit (Qiagen).
Protein-SP-A (21) and SP-D (22) were purified from therapeutic lung lavage fluid obtained from alveolar proteinosis patients as described previously. MBL (23) and C1q (24) were purified from pooled human plasma (HD Supplies, Aylesbury, England). Recombinant preparations of the trimeric ϳ60-kDa fragment composed of the neck and CRDs of SP-D, with and without eight Gly-X-Y repeats derived from the collagen-like region, were expressed in Escherichia coli and Pichia pastoris, respectively. As described previously, SP-D(GXY) 8 (n/CRD) and SP-D(n/CRD) were purified from inclusion bodies and cell culture media supernatant, respectively (16,25,26).
EMSA Assay-DNA-protein complexes were prepared for gel shift analyses by incubating 0.1-0.4 g of linear plasmid DNA with protein in a 20-l reaction containing 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM CaCl 2 , or 5 mM MgCl 2 , or 5 mM MnCl 2 , or 5 mM EDTA buffer as described previously (27). After 20 -60 min at 37°C, the DNA-protein complexes were fractionated by agarose gel (0.8 or 1%, w/v) electrophoresis in 40 mM Tris acetate (pH 8) buffer in the absence or presence of 1 mM EDTA and 0.5 g/ml ethidium bromide and visualized by UV transilluminator as described previously (27,28).
Protease Assay-SP-D (1.5 g) and DNA (400 ng) were allowed to form complexes as described in EMSA and were treated either with 0.05% (w/v) SDS, 20 mM octyl glucoside, or 500 g/ml proteinase K (Promega) in a 20-l reaction for 1 h at 37°C. In the collagenase experiment, SP-D (500 ng) was first treated with enzyme (40 g, type 1 collagenase from Clostridium histolyticum, Ͼ90% pure; Calbiochem) in a 20-l reaction for 1 h at 37°C and then incubated with DNA (400 ng) for 1 h at 37°C. In another reaction, SP-D (500 ng) was first incubated with DNA (400 ng) in a 20-l volume for 1 h at 37°C and then treated with enzyme for 1 h at 37°C.
Dot Blot-SP-D (370 ng) and linear pUC18 DNA (150 ng and 2-fold dilutions) were carefully dotted (1 l) onto Hybond-C Extra nitrocellulose membrane discs (Amersham Biosciences) and allowed to dry in a Petri dish. Membranes were incubated with 2% (w/v) bovine serum albumin in PBS, 0.02% (v/v) Tween (PBST) buffer for 18 h at 4°C. Remaining procedures were carried out at 23°C. Membranes were washed three times with 20 ml of PBST for 5-10 min on a shaker. Membranes were either incubated with 10 ml of PBST or PBST with SP-D (0.5 g/ml) for 1 h. The membranes were washed as above and incubated further with 10 ml of biotinylated rabbit polyclonal anti-human SP-D(GXY) 8 (n/CRD) antibody (1 g/ml) for 1 h. Antibody was washed away as described above, and the membranes were incubated with horseradish peroxidase-conjugated streptavidin (1:10,000) for 20 min. SP-D complexes were detected by ECL reagents (Amersham Biosciences).
Electron Microscopy-Mouse lung DNA and proteins were incubated in the DNA binding buffer, spread in a thin layer on carbon-coated specimen grids, negatively stained with 2% (w/v) uranyl acetate, and examined using a Jeol JEM 1010 or 100 CX electron microscope (Leica UK Ltd., Cambridge, England) (18,27).
SPR Analyses-Biotinylated oligonucleotides (2.5 g/ml) or mannan (200 g/ml) were immobilized on a streptavidin SA chip (BIAcore 2000, Herts, England). The biotinylated ligands were diluted in 10 mM NaOAc (pH 5.5) buffer and immobilized on individual streptavidin-coated chips by injecting them at 5 l/min for 5-10 min. Leaving flow cell 1 as blank, the next three flow cells were used for immobilization of ligands. Free streptavidin in all flow cells was blocked with biotin, and the flow cell surface was normalized with 50% (v/v) glycerol solution.
Binding of proteins (0 -50 g/ml) to the oligonucleotides was analyzed in 10 mM HEPES-NaOH (pH 7.4), 5 mM CaCl 2 , and 0 or 150 mM NaCl buffer at a flow rate of 10 l/min. In competition assays, SP-D (1 or 5 g/ml) or SP-D(GXY) 8 (n/CRD) (200 g/ml) was allowed to bind (10 l/min) to immobilized mannan in the presence of different concentrations of competitors (0 -80 mM). The SPR response obtained in the absence of any competitor was considered as 100%, and the relative binding was calculated for each competitor concentration.

Detection of Collectin Binding to Bacterial DNA by EMSA-
To determine whether collectins bind DNA, we incubated linear plasmid DNA with collectins and related proteins and analyzed the complexes in agarose gels. This EMSA showed that native forms of SP-D, MBL, and C1q, but not SP-A, shifted the migration pattern of the DNA in the gel (Fig. 1A). SP-D(n/ CRD), a yeast-expressed, recombinant ϳ60 kDa trimeric subunit fragment of SP-D that contains only the coiled-coil neck and the CRDs, as well as a known DNA-binding pentraxin, serum amyloid component P (SAP), did not shift the DNA bands in EMSA. Because both SAP (30) and CRDs of surfactant proteins have lectin domains and bind agarose to varying degrees, the DNA binding ability of these proteins might partly be competed by the matrix. These experiments showed that in addition to the non-collectin, subcomponent C1q of the classical complement pathway (31), innate immune collectins could bind DNA. These EMSA results were consistent with the interpretation that, similar to the collagen-like regions of C1q (31), SP-D and MBL bound DNA via their collagen-like regions, or that oligomerization of the proteins was required for the formation of large complexes.
Interestingly, SP-D bound DNA even at 1 M NaCl concentration, indicating that this collectin strongly interacted with DNA ( Fig. 1B). Because calcium and other divalent cations affect the structure and function of collectins (16), we examined the effect of these ions on collectin-DNA interactions. SP-D bound more effectively to DNA in the presence of 5 mM calcium, magnesium, and manganese than EDTA (Fig. 1C). The ion-dependent increase in the binding of SP-D to DNA was detectable only with supercoiled but not linearized DNA. Hence, these results suggest that the structure of DNA is important for SP-D-DNA interactions, and that both lectin and non-lectin types of binding may be involved in the formation of these complexes.
Determination of the Nature of SP-D-DNA Complexes-To determine the types of complexes formed by SP-D, we incubated SP-D (1.5 g) with linear pUC18 DNA (400 ng) in the standard DNA-binding buffer for 1 h at 37°C and added either SDS (0.05%, w/v), octyl glucoside (20 mM), or proteinase K (500 g/ml) and incubated for an additional 1 h at 37°C. DNAprotein complexes were size-fractionated in an agarose gel, and the DNA present in each complex was visualized with ethidium bromide staining and UV light. Linear DNA ( Fig. 2A, lane 2) migrated into the gel, and no noticeable amount of DNA was present in SP-D preparations (Fig. 2A, lane 3). SP-D retarded the mobility of the DNA, and the DNA complexes were detected in the wells (Fig. 2A, lane 4). Although SDS did not alter the migration of DNA in the gel (Fig. 2A, lane 5), this strong anionic detergent disassociated the DNA from the complexes (Fig. 2A, lane 6). Hence, SP-D-DNA complexes were formed by noncovalent interactions. A non-ionic mild detergent, octyl glucoside, neither affected the DNA (Fig. 2A, lane 7) nor SP-D-DNA complexes (Fig. 2A, lane 8). Therefore, SP-D-DNA complexes were strong and likely involved charge interactions. Digestion of these complexes with proteinase K resulted in the release of DNA, indicating that they were formed by protein-DNA interactions ( Fig. 2A, lanes 9 and 10).
To further show that these complexes were formed by SP-D, we treated the complexes with collagenase enzyme and analyzed the samples by agarose gels, SDS-PAGE under reducing and denaturing conditions, and Western blots to specifically visualize the DNA, proteins, and SP-D, respectively (Fig. 2, B-D). Linear pUC18 DNA contained no aggregated complexes (Fig. 2, B-D, lanes 2). The 110-kDa collagenase enzyme neither showed any aggregates (Fig. 2, B-D, lanes 3) nor affected the mobility of DNA (data not shown). SP-D itself did not contain any significant amount of DNA or DNA-protein complexes when 500 ng of protein was used (Fig. 2B, lane 4). Under reducing and denaturing conditions, SP-D migrated at about 43 kDa (Fig. 2C, lane 4) and was detected by anti-SP-D antibody (Fig. 2D, lane 4). The agarose gel showed that SP-D formed complexes and shifted the DNA to the wells (Fig. 2B,  lane 5). Under reducing and denaturing conditions, SP-D separated away from DNA without any noticeable change to its structure (Fig. 2, C and D, lanes 5). When the SP-D was digested with collagenase for 1 h and then incubated with DNA, no shift in the mobility of DNA was noticed (Fig. 2B, lane  6). SDS-PAGE and Western blots showed that the collagen-like region of SP-D was digested by the collagenase (Fig. 2, C and D,  lanes 6). The calculated molecular mass of the collagenaseresistant fragment of SP-D was ϳ16 kDa, and this was consistent with the protein detected by SDS-PAGE and Western blot after the enzyme digestion. Interestingly, when the collagenase was added after the formation of SP-D-DNA complexes, the enzyme did not completely eliminate these complexes (Fig. 2B,  lane 7) suggesting that bound DNA may protect the collagenlike region from collagenase digestion. SDS-PAGE and Western blots showed that the collagen-like region of approximately half of the SP-D molecules was intact (Fig. 2, C and D, lanes 7). In the absence of DNA, collagenase completely digested the collagen-like regions of SP-D (Fig. 2, B-D, lanes 8). These results established that SP-D bound DNA, and that the collagen-like region of the protein was important for the formation of large SP-D-DNA complexes and subsequent gel shift.
Binding of SP-D to Immobilized DNA-To directly show that SP-D could bind DNA, we first immobilized linearized plasmid DNA on nitrocellulose membranes and incubated the membranes with buffer or SP-D (0.5 g/ml). After washing the membranes, SP-D was detected by anti-SP-D antibody and ECL reagents. SP-D antibody detected the 370 ng of SP-D spotted on both membranes equally well (Fig. 3, A and B, lanes  1). The polyclonal antibody did not show any significant binding to DNA in the absence of SP-D (Fig. 3A, lane 2). The dot blot showed that SP-D bound to the DNA present on the membrane in a concentration-dependent manner (Fig. 3B, lane 2). These results showed that SP-D directly interacted with DNA.
Visualization of SP-D-Genomic DNA Complexes by Electron Microscopy-To obtain further confirmatory evidence of SP-D interacting with DNA, we examined the binding of SP-D to mouse lung genomic DNA by electron microscopy (Fig. 4). DNA was detected as long strands on the carbon support film (Fig. 4,  A and E). SAP appeared as pentamers (Fig. 4B), which avidly bound DNA (Fig. 4F). When SP-D(n/CRD) was incubated with DNA, it coated the entire DNA (Fig. 4, C and G). Similar complexes were obtained when SP-D(GXY) 8 (n/CRD) was incubated with the DNA (Fig. 4, D and H).
Interestingly, these complexes were very different compared with the native SP-D-DNA assemblies. Native SP-D, which contained long collagen-like fibrillar regions, efficiently spooled the DNA into aggregates (Fig. 5, A and B). The hub and nearby collagen-like regions preferentially bound DNA (Fig. 5C), suggesting that this collectin also interacted with the ligand probably via polyionic interactions. These images (Figs. 4 and 5) suggested that both globular and collagen-like regions of the collectin interacted effectively with DNA.
SPR Analysis of Interaction between Collectins and Oligonucleotides-To characterize collectin-DNA interaction using a more sensitive system, we allowed the proteins to bind to short oligonucleotides (22-mer) that were immobilized on a BIAcore chip. The SPR response showed that although SP-A bound oligonucleotides in a low ionic strength buffer, it progressively lost DNA-binding ability as the ionic strength increased (Fig. 6,  A and B). In contrast, SP-D bound to DNA in 150 mM NaCl buffer (Fig. 6C). Hence, SP-D, but not SP-A, interacted with the oligonucleotides under physiological salt concentrations (Fig.  6). MBL showed a moderate level of binding under physiological salt conditions (data not shown). We conducted similar SPR experiments with another biotinylated oligonucleotide (20-mer). Similar levels of SP-D binding were observed for both of those oligonucleotides (data not shown). Although the SPR response of SP-D-oligonucleotide (Fig. 6) was relatively small compared with SP-D-mannan ligand (8), these results suggested that all three collectins bound single-stranded short oligonucleotides.

SPR Analyses of Competition of Collectin-Carbohydrate
Interactions-To directly investigate whether collectins could bind pentoses, the carbohydrate moieties present in nucleic acids, we allowed SP-D to bind immobilized mannan present on a BIAcore chip in the presence of different concentrations of hexoses or pentoses (Figs. 7 and 8). BIAcore assays behaved similarly to the conventional ELISA system (32), and maltose, mannose, myo-inositol, and glucose inhibited the interaction between SP-D and mannan with relatively low IC 50 values compared with those of D-fucose, D-galactose, and GlcNAc (Fig.  7). We conducted competition experiments under identical conditions for pentoses (Fig. 8). D-Arabinose and glucose inhibited binding with a similar IC 50 value (Figs. 7A and 8A). Furthermore, D-ribose was a better competitor than D-2-deoxyribose, and they competed with IC 50 values comparable with those of D-galactose, D-fucose, and GlcNAc (Figs. 7A and 8A). Interestingly, the D-forms of all three of these pentoses inhibited better than their L-forms of the respective sugars (Fig. 8, A and B). Because nucleic acids contain only D-forms of the pentoses, recognition of these forms of the sugars suggested that this binding might be biologically relevant. Arabinose is an epimer of ribose at 2Ј-OH; hence, the C-type lectin domain may interact with 2Ј-OH and 3Ј-OH groups of the pentoses. SP-D interacts with two adjacent -OH groups in hexoses and forms calcium-coordinated bonds with O3 and O4 of the glucose ring (25). Hence, these data are consistent with the interpretation that the CRDs of SP-D recognized one adjacent pair of -OH groups from the pentose 1Ј-OH, 2Ј-OH, and 3Ј-OH. Because 2Ј-deoxyribose lacks the 2Ј-OH group, the interaction between this pentose and SP-D is likely to be different from that of arabinose and ribose.
Cyclic AMP and cyclic dAMP did not compete with the interaction between SP-D and mannan (Fig. 9A). The 5Ј phosphate group of these nucleotides forms a covalent bond with the 3Ј-OH of the same molecule; hence, the 3Ј-OH group would not be available to interact with SP-D. These data further suggested that the 3Ј-OH of these pentoses might be one of the -OH groups involved in the SP-D-pentose interactions. The effect of AMP and dAMP could not be tested because of the poor solubility of these compounds in the assay buffer. ADP and dADP inhibited the interaction between SP-D and mannan, and ADP was a better competitor than dADP (p Ͻ 0.05; Fig.  9A). ATP and dATP, however, competed equally well. All the NTPs and dNTPs inhibited binding to approximately similar extents (Fig. 9, B and C), and their IC 50 values fell between the high (maltose, mannose, myo-inositol, and glucose) and low (D-galactose, D-fucose, and GlcNAc) affinity hexoses (Figs. 7 and 9). These results, therefore, suggested that the phosphate group had an effect on their interactions with SP-D. Importantly, these findings indicate that SP-D can interact with these key biological molecules. Because the binding affinities of SP-D for these molecules are lower than hexoses, SP-D may bind to these compounds only under specific conditions; for example, SP-D may recognize them when their local concentrations are high.
A similar competition experiment was conducted with the SP-D(GXY) 8 (n/CRD) fragments. Similar to the interactions between SP-D and mannan, SP-D(GXY) 8 (n/CRD) fragmentsmannan binding was effectively inhibited by hexoses (Fig. 10A) and dNTPs (Fig. 10B). Because dNTPs competitively inhibited the interaction between SP-D and mannan and the recombinant SP-D(GXY) 8 (n/CRD) fragments show similar results to those obtained with the native protein, we inferred that nucleic acid subunit binding also occurs via the carbohydrate binding site within SP-D.
Interaction between SP-D(GXY) 8 (n/CRD) and Apoptotic Cells-To determine whether SP-D can bind to the DNA present on apoptotic cells, we conducted double labeling experiments. We labeled the nicked DNA present in apoptotic Jurkat cells with Cy3-dUTP by terminal deoxynucleotidyl transferasemediated nick end labeling assay and incubated them with FITC-labeled SP-D(GXY) 8  microscopy ( Fig. 11) indicated that the recombinant fragment colocalized with DNA of apoptotic cells.
To further understand the binding of the SP-D(GXY) 8 (n/ CRD) fragment to apoptotic alveolar macrophages, we incubated the recombinant fragment with apoptotic mouse alveolar macrophages followed by staining with propidium iodide. In FACS experiments, Ͼ10 FL2-H propidium iodide fluorescence was detectable in 16.8, 11.4, and 2.1% of the cells in the presence of 0, 4, and 20 g/ml of SP-D(GXY) 8 (n/CRD), respectively (Fig. 12). Because the binding of propidium iodide, a DNAbinding dye, to these cells is inhibitable by the SP-D(GXY) 8 (n/ CRD) fragment, we inferred that the recombinant fragment and propidium iodide recognize the same targets. These results suggest that the C-terminal region of SP-D, which includes the  distal (GXY) 8 , neck, and CRD domains, binds directly to the DNA present on apoptotic cells.

DISCUSSION
Our investigation establishes a new role for collectins. We show that three innate immune collectins, SP-A, SP-D, and MBL, bind DNA to varying degrees. In addition, our results show that the collectins bind single-stranded and doublestranded DNA from a variety of sources including bacterial, genomic, and synthetic oligonucleotide fragments, and the binding between collectins and DNA occurs both via the CRDs and collagen-like regions of the protein. SP-D binds DNA more strongly when compared with SP-A. SP-D(GXY) 8 (n/CRD) fragments bind apoptotic cells and colocalize with DNA, suggesting that they directly bind DNA present on these cells. Hence, our study suggests that free DNA and DNA present in the apo-ptotic cells are targets for the opsonic innate immune collectins.
Although collectins are known to bind several microbial cell wall ligands, whether they bind other key ligands that are released during chronic infection and cell death is unknown. Several other laboratories that study the binding of these innate immune molecules to microbes focus upon the interaction between the collectins and arrays of hexose sugars present on the microbes. We considered that collectins might also bind nucleic acids that contain repeating pentameric sugars 2 (33,34). Our first set of EMSA experiments detected the binding of SP-D and MBL to bacterial DNA (Fig. 1A). This assay typically requires tight binding of protein and DNA. SP-A did not shift the electrophoretic mobility of DNA bands, suggesting that the binding between these two molecules is weak or that they do not interact with each other. This assay easily detects the interaction between the collagen-like domain of a protein and DNA, as has been seen for C1q (Fig. 1A), which is known to bind to this ligand via its collagen-like region (31). Because the interaction between SAP or SP-D(n/CRD) and DNA was not easily detectable by this method, lectin-nucleic acid binding may not be determined by EMSA. A less stringent requirement for calcium ions in these reactions further suggests an existence of non-C-type lectin interaction between DNA and collectins. Because SP-D and MBL, but not SP-A, contain charged stretches of amino acid residues in their collagen-like regions (35), the result is consistent with the presence of collagen-like, region-dependent interaction between these two proteins with DNA. Scavenger receptors present on macrophages also bind DNA via their collagen-like domains (36). Hence, our findings place the soluble opsonic collectins as another class of proteins that may enhance the clearance of free DNA. The SP-D used in the study was purified from lung washings that contained lipids, cell debris, and other proteins. Hence, to verify the nature of DNA complexes formed by the purified SP-D preparations, we analyzed the complexes in the presence of various compounds or enzymes by EMSA (Fig. 2). The SP-D-DNA complexes were tightly and noncovalently held together by charge interactions but were not formed by molecules such as cationic lipids. Collagenase treatment of these com-plexes further revealed that SP-D mediates the formation of large aggregates, which are mediated in part by the collagenlike regions.
Dot blots (Fig. 3) and electron microscopy (Figs. 4 and 5) confirmed the results obtained by EMSA (Figs. 1 and 2), and the latter suggested that the N-terminal/proximal part of the collagen-like region of the SP-D interacts with DNA. Because this region of the SP-D contains stretches of charged amino acids (35), binding of DNA at this location is consistent with the presence of a charge interaction between these two molecules. The structural appearance of the complexes between DNA and SP-D(n/CRD) or SP-D(GXY) 8 (n/CRD) fragments is striking and shows that these ϳ60-kDa protein fragments of SP-D coat the DNA effectively. These images therefore indicate that the globular domains of SP-D also interact with DNA. The crystal structures of the SP-D(n/CRD) and SP-D(GXY) 8 (n/CRD), but not SP-A (15), show the presence of a highly positively charged region in the center of the trimers (16,25,35). Hence, this region of SP-D may directly interact with the negatively charged array of phosphates in the nucleic acid polymers. Detection of the interactions between DNA and SP-D(n/CRD) or SP-D(GXY) 8 (n/CRD) in the electron microscopy experiments, but not in gel shift assays, may be attributable to the differences in the support matrices used in these two experiments and/or to any increase in local concentrations of protein occurring during staining of the sample for microscopy.
Data from competition experiments suggest that SP-D may bind to pentoses in a similar manner as it does to hexoses (Figs. 7 and 8). Furthermore, MBL recognizes pentoses 3 ; hence, our findings place this group of sugars as an important class of ligands for collectins. Because the D-forms of the pentoses are better ligands than the L-forms, C-type lectin domains may have evolved to recognize biologically relevant pentose sugars. Interestingly, efficient competition required pentoses with adjacent carbons containing -OH groups. Binding of collectins to hexoses is dependent on the presence of two adjacent equatorial -OH groups (25,37). Hence, recognition of pentoses by collectins may also occur via similar calcium-coordinated bonds.
Because the pentose-based compounds such as ADP, dADP, NTPs, and dNTPs inhibit the interaction between SP-D, or SP-D(GXY) 8 (n/CRD) fragments, and mannan, another highaffinity carbohydrate ligand for SP-D, the CRDs of the protein appear to directly bind these molecules (Figs. 9 and 10). Because intracellular collectins have been described recently (38) and many other soluble proteins and receptors also contain C-type lectin domains (3,39,40), recognition of these compounds by the CRDs may be of physiological importance. Both DNA and RNA contain repeating arrays of pentoses that are connected by 5Ј-phosphate and 3Ј-OH groups, and collectins may recognize these arrays under certain conditions.
All of these collectins SP-A (41, 42), SP-D (41,42), and MBL (43) have been shown to bind apoptotic cells; however, the ligands that they bind have not yet been clearly identified. Our previous experiments had shown that apoptotic/necrotic macrophages accumulate in the lungs of SP-D (Ϫ/Ϫ) mice, and that this defect could be corrected by treating these animals with recombinant SP-D(GXY) 8 (n/CRD) fragment (29). Here we show that the same recombinant fragment colocalizes with DNA present on the apoptotic cells (Fig. 11). Other in vivo studies show that only SP-D gene deficiency, but not deficiency of SP-A or C1q, leads to a significant defect in the clearance of apoptotic cells from the lung (42). Interestingly, SP-D can bind DNA even above physiological salt concentrations (Fig. 1B), which suggests that the interaction between these two molecules is strong. Analysis of collectin-DNA interactions by a more sensitive system further confirmed this finding. Although low compared with binding to oligomeric mannan (8), SP-D, but not SP-A, binds short synthetic oligonucleotides at physiological salt conditions (Fig. 6). Considering that the SP-D(GXY) 8 (n/ CRD) fragment also binds apoptotic cells and colocalizes to DNA (Figs. 11 and 12), opsonization of SP-D to apoptotic cell DNA may be an important signal for phagocytosis. We have determined that collectins enhance the pinocytosis and phagocytosis of DNA by macrophages (33,34), and these findings will be described in detail elsewhere. 2 Common inflammatory disease/conditions such as septic shock (44), chronic obstructive pulmonary disease, and asthma contain impaired surfactant and increased amounts of apoptotic cells, cellular debris, and secretions in their airways. Interestingly, Pseudomonas aeruginosa secretes DNA, particularly into the lung environment of cystic fibrosis (CF) patients to form a coat of biofilm to establish persistent infection (2), and innate immune proteins are primarily responsible for clearing these pulmonary infections (3). Pulmonary diseases with chronic infection such as cystic fibrosis results in the accumulation of free DNA in the lungs (2,10), and these patients also have low concentrations of SP-A and SP-D (45). Furthermore, concentrations of collectins, SP-A and SP-D, in the lungs are inversely related to the severity of bacterial infection in these patients (46). Hence, collectins are important proteins that minimize microbial infections and inflammation, and recombinant fragments of SP-D may be useful in treating cystic fibrosis patients.
It is also known that increased concentrations of free DNA in the lung cause inflammation and septic shock during gene therapy experiments (47) in addition to the disease conditions (9,44,48). Because our results show that the innate immune collectins bind both bacterial and host genomic DNA (Figs. 1-5), these new findings could have significant implications for the understanding of the pathology and treatment of these diseases. Furthermore, collectins regulate multiple functions, but the regulatory mechanisms are often not clear (49,50). Collectins may bind signaling molecules such as ATP and related compounds (Figs. 6 -10) and, hence, could potentially regulate diverse cellular functions.
It is interesting to note that most of the previously described proteins implicated in DNA clearance, such as scavenger receptors (36) and pentraxins (30), contain either collagen-like regions or lectin/DNA-binding domains. Although the domain organization is different, scavenger receptors (36) as well as collectins contain both of these domains in the same protein, and are capable of enhancing phagocytosis (3,50). This may represent an evolutionary process where an important biological function, the clearance of DNA, becomes mediated by two distinct types of protein structures that are combined within a single molecule. Thus, collectins are particularly well suited for opsonizing and promoting the clearance of free nucleic acid and apoptotic cells.