Trimerization of the Amino Propeptide of Type IIA Procollagen Using a 14-Amino Acid Sequence Derived from the Coiled-Coil Neck Domain of Surfactant Protein D*

The folding of a collagen triple helix usually requires the presence of additional sequences that contribute to the association and correct alignment of the collagen chains. We recently reported that the C-terminal neck and lectin domains of a collagenous C-type lectin, rat pulmonary surfactant protein D (SP-D), are sufficient to drive the trimerization of a heterologous type IIA procollagen amino propeptide sequence. However, the conformation of the resulting trimeric IIA propeptide and the specific contributions of the SP-D sequence to trimerization were not elucidated. In the present study, we show that trimerization of the fusion protein is associated with correct folding of the collagen helix within the IIA propeptide domain (as assessed by circular dichroism) and that the constituent chains are hydroxylated. Chemical cross-linking and analytical ultracentrifugation showed that the IIA amino-propeptide retains its trimeric configuration even after proteolytic removal of the SP-D domains. By contrast, IIA amino-propeptides synthesized without fusion to the neck or lectin domains are assembled exclusively as monomers. To localize the trimerization sequence, mutant chimeric cDNA constructs were designed containing premature termination codons within the coiled-coil neck domain. A short, 14-amino acid sequence corresponding to the first two heptad repeats of the neck domain was sufficient to drive the trimeric association of the IIA amino-propeptide α-chains. However, deletion of the collagen domain resulted in the secretion of monomers. These studies demonstrate that two heptad repeats are sufficient for trimeric association of the propeptide but indicate that cooperative interactions between the coiled-coil and collagen domains are required for the formation of a stable helix.

The type IIA NH 2 -propeptide is encoded by eight exons; the translated protein consists of a short globular domain, a 69amino acid von Willebrand factor type C cysteine-rich domain, a minor collagen triple-helical domain containing 26 GXY repeats, and a short telopeptide domain, which links the minor collagen domain to the major collagen triple helix (1,2). Trimerization of most fibrillar collagens is dependent on the globular COOH-propeptide for the recognition and association of the three polypeptide chains resulting in registered nucleation of triple helix formation in a zipper-like fashion from the C to N terminus (3)(4)(5)(6)(7). Functions proposed for procollagen NH 2propeptides include the regulation of collagen fibrillogenesis (8) and a feedback control of net collagen biosynthesis (9 -11). Recently, it was proposed by our laboratory (12) and by others (13) that the NH 2 -propeptide of type IIA procollagen regulates growth factor activity in the extracellular matrix. Trimeric assembly of fibrillar NH 2 -propeptides affects protein valency and stability, which appear important for function in vivo. This emphasizes the importance of a procollagen COOH-propeptide, or indeed other protein domains with similar function, to drive this trimerization process.
Pulmonary SP-D 1 is predominantly assembled as dodecamers, consisting of four trimeric subunits cross-linked by disulfide bonds (14,15). Each SP-D subunit contains an N-terminal cross-linking domain, an uninterrupted triple-helical collagen domain consisting of 59 GXY repeats, a trimeric coiled-coil neck domain, and a C-type lectin carbohydrate recognition domain (CRD). Trimerization of SP-D subunits and subsequent oligomerization of these trimeric subunits to form higher order multimers results in increased valency of the CRD, an essential prerequisite for high affinity ligand binding (16). The neck domain of SP-D is the unit responsible for driving the trimerization of the three polypeptide chains of SP-D. It was demonstrated that a 35-amino acid sequence containing the human neck sequence was sufficient to form stable, non-covalent, trimeric complexes in vitro (17). The same sequence was found to be important for the association of the three CRDs of human SP-D; CRDs synthesized in prokaryotic cells without this neck domain were assembled as monomers (18).
The sequence of coiled-coil domains is characterized by a 7-residue (heptad) repeat commonly denoted (abcdefg) n in which positions a and d are primarily occupied by hydrophobic residues, positions e and g are primarily occupied by charged residues, and positions b, c and f are primarily occupied by polar or charged residues (19). The crystal structure of the neck and lectin domain of human SP-D has been solved, and the coiled-coil region was visualized as a stretch of greater than 28 amino acids (Arg 208 -Pro 235 ) consisting of ϳ8 helical turns (20).
Based on the earlier work of Harbury et al. (21), Reid and co-workers (17) suggested that the presence of valine at the d positions favors the trimeric assembly of human SP-D. It was further suggested that the unusual fourth heptad, which contains Phe 225 and Tyr 228 in the a and d positions, respectively, might serve to initiate trimerization (20). However, no valine residues are found in the neck of rat SP-D. In addition, we observed that deletion of the conserved fourth heptad repeat does not prevent trimerization of recombinant rat SP-D secreted by mammalian cells (22). On the other hand, internal deletions of residues 207-214 or 214 -221 within the neck domain were found to block trimerization and indicated that sequences N-terminal to Phe 225 were required for trimerization. These studies also demonstrated that trimeric assembly of the NH 2 -propeptide of type IIA procollagen could be initiated by the C-terminal neck and lectin domains of SP-D. In doing so, the chain association function of the type IIA procollagen COOH-propeptide was replaced by these SP-D domains. However, the conformation of the resulting trimeric IIA complex and the specific contributions of the C-terminal SP-D sequence to trimerization were not elucidated. The present study describes the molecular structure of the resulting IIA NH 2 -propeptide and establishes, for the first time, that a short, amphipathic helical sequence derived from the rat SP-D neck domain can also drive the trimerization of a fibrillar collagen NH 2 -propeptide. This sequence consists of the first two heptad repeats of the neck domain, which is in agreement with our previous deletional mutagenesis studies showing that N-terminal regions of the neck domain are important for initiating trimerization (22). There have been only a few reports on the use of oligomerization domains to trimerize heterologous collagenous sequences (23). To our knowledge, this is by far the shortest sequence found to permit trimerization of a collagenous molecule and the first to demonstrate the use of a heterologous trimerization cassette to support the normal folding of a procollagen propeptide domain. High levels of a correctly folded IIA NH 2 -propeptide were produced using this system, which will enable us to study its biological function in vitro. Establishing a minimum sequence of the SP-D neck domain that can drive trimerization without inclusion of the functional SP-D lectin domain will allow us to study the function of the trimeric IIA propeptide in vivo. Knowledge gained from these findings may be applied to produce other procollagen propeptides or indeed other collagenous proteins for functional studies.

Expression of IIA/SP-D Fusion Protein in Chinese Hamster Ovary
(CHO)-K1 cells-A chimeric construct was synthesized by linking the cDNA encoding the NH 2 -propeptide of type IIA procollagen (see Fig. 1) to the cDNA encoding the neck ϩ CRD of rat SP-D (see Fig. 2). This chimeric construct and the resulting fusion protein was named IIA/ SP-D. The cDNA encoding exons 1-8 of human type IIA procollagen NH 2 -propeptide were amplified by reverse transcription-PCR from RNA that had been isolated from articular chondrocytes in culture. Specific upstream and downstream primers were designed from the pro-␣1 type II collagen complete coding sequence (GenBank TM accession number: L10347). The IIA/SP-D chimeric construct was made by overlap extension PCR. Briefly, the complete coding sequence of the IIA NH 2 -propeptide (using oligomer A, ggtacgaattcatgattcgcctcggg; this primer sequence contains extra bases and the EcoRI site at the 5Ј end, shown in bold) and a 3Ј sequence homologous to a region of the neck domain of rat SP-D (using oligomer B, cagcactgtccattggtccttgcat) was amplified by PCR for 25 cycles at an annealing temperature of 52°C. The same conditions were used to amplify the neck ϩ CRD of rat SP-D containing a 5Ј sequence homologous to a region of the IIA cDNA (using oligomer C, aggaccaatggacagtgctgctctg) and a 3Ј-EcoRI site (using a T7-specific downstream oligonucleotide). cDNA products from the two PCR amplifications were combined, and overlap extension PCR was carried out for 30 cycles at an annealing temperature of 55°C using oligomers A and T7. The resulting chimeric construct was digested with EcoRI and subcloned into pGEM-3Z, and the orientation of the subcloned insert was confirmed by restriction mapping and DNA sequencing.
IIA/SP-D cDNA was excised from pGEM-3Z by EcoRI digestion and ligated into the multiple cloning site of pEE14 (24) distal to the cytomegalovirus promoter/enhancer and proximal to the glutamine synthetase gene. CHO cells (CHO-K1; ATCC CCL-61) were transfected with pEE14-IIA/SP-D using LipofectAMINE (Invitrogen) and grown in selection Glasgow's minimum essential medium (Invitrogen) containing 10% dialyzed fetal bovine serum and the glutamine synthetase inhibitor, methionine sulfoxamine (25-50 M) for 2-3 weeks. Stable clones were obtained as described by Crouch et al. (25) for the expression of recombinant rat-SPD. To assess the importance of the trimerizing neck domain, a control pEE14 construct was synthesized consisting only of cDNA encoding the full-length IIA NH 2 -propeptide, devoid of cDNA encoding the neck and lectin domains. This construct was used in transient transfections of CHO cells using LipofectAMINE reagent.
Detection and Purification of IIA/SP-D Fusion Protein-Media from transiently transfected CHO cells were screened for the presence of the fusion protein by an enzyme-linked immunoassay using rabbit antihuman exon 2 (IIA) antibody (26), chicken IgY anti-human Exon3-8 antibody (27), or rabbit anti-rat SP-D antibody (27). Immunopositive proteins labeled with rabbit-horseradish peroxidase secondary antibodies were detected by enhanced chemiluminescence using SuperSignal chemiluminescent substrate (Pierce). Clones expressing the IIA/SP-D fusion protein were selected and cultured further by exposure to 50 -100 M methionine sulfoxamine, and the resulting conditioned medium was dialyzed against Tris-buffered saline, pH 7.5, containing 10 mM EDTA. CaCl 2 (20 mM) was added to the dialyzed medium, and IIA/SP-D was subsequently purified by maltosyl-agarose chromatography (25). Because the interaction of the CRD with maltose is calcium-dependent (27), IIA/SP-D was eluted from the column with Tris-buffered saline, 10 mM EDTA, pH 7.5. Eluted fractions were analyzed by SDS-polyacrylamide gel electrophoresis, silver staining, and Western blotting.
Collagenase Digestion of IIA/SP-D-Bacterial collagenase was purified by gel filtration chromatography using crude collagenase as the starting material (Worthington Biochemical Corp.) (28). IIA/SP-D or rat SP-D (30 g) in Tris-buffered saline, 10 mM EDTA, pH 7.5, was digested with purified bacterial collagenase (1 g) containing CaCl 2 (20 mM) and N-ethylmaleimide (5 mM) overnight at 37°C. Fresh collagenase (1 g) was added for a further 3 h at 37°C followed by EDTA (4 mM) to stop the reaction. An aliquot (5 g) of digested and undigested IIA/SP-D or rat SP-D was electrophoresed through a 4 -20% SDS-polyacrylamide gel to confirm collagenase digestion. The major collagenase-resistant products were detected by Coomassie Blue staining and subjected to N-terminal amino acid sequencing. Collagenase-digested IIA/SP-D or SP-D was transferred to Sequi-Blot polyvinylidene difluoride membrane (Bio-Rad), stained with Coomassie Blue, excised, and sequenced on an ABI 473A protein sequencer equipped with model 610A data analysis software.
Purification of IIA NH 2 (27). Since MMP-9 has two cleavage sites within the telopeptide and cleavage is not always 100% efficient, we proceeded to synthesize a mutant IIA/SP-D chimeric construct containing an enterokinase cleavage site in the exon 8-encoded telopeptide. Using the QuikChange TM site-directed mutagenesis kit (Stratagene), oligonucleotide primers were designed to change the DNA sequence encoding amino acids 161-165 in exon 8 of the wild-type IIA NH 2 -propeptide ( 161 GFDEK 165 ) to one that encodes the EK cleavage site ( 161 DDDDK 165 ). Stable CHO cell lines producing this mutant fusion protein (IIA/EK/ SP-D) were produced as described above. Approximately 0.001% w/w of enterokinase (New England Biolabs) was added to purified IIA/EK/ SP-D protein overnight at room temperature.
Cleavage by MMP-9 or enterokinase was confirmed by gel electrophoresis, silver staining, and immunoblotting using antibodies specific for the IIA (exon 2) domain or the CRD of SP-D. CaCI 2 was added to the cleaved products before applying to a maltosyl-agarose column to separate the IIA NH 2 -propeptide (present in the flow-through) from the neck/CRD of SP-D (present in the EDTA eluate).
Chemical Cross-linking-Covalent cross-linking was performed using bis-(sulfosuccinimidyl)suberate (BS 3 ; Pierce). Increasing amounts of BS 3 (0, 0.1, 0.5, 1, and 2 mM final concentration) prepared in 5 mM sodium citrate, pH 5, were added to each recombinant protein for 1 h at room temperature. Addition of SDS-PAGE loading buffer containing Tris-HCl (0.5 M) inhibited the reaction. Samples were boiled for 5 min prior to SDS-PAGE, which was carried out in the absence of sulfhydryl reduction. Cross-linked proteins were identified by silver staining or immunolocalization using anti-IIA (exon 2) polyclonal antisera.
Circular Dichroism and Determination of IIA NH 2 -Propeptide Melting Temperature-Approximately 50 g of the IIA NH 2 -propeptide (0.2 mg/ml in PBS, pH 7.5), purified by cleavage of the mutant IIA/EK/SP-D fusion protein containing the enterokinase cleavage site, was analyzed by CD spectroscopy. A Jasco J715 spectropolarimeter with a thermostated quartz cell and a path length of 0.1 cm was used, and the spectrum was recorded at 5°C between 180 and 260 nm. To determine the melting temperature of the IIA NH 2 -propeptide, the spectrum was monitored at 225 nm from 5 to 70°C.
Analytical Ultracentrifugation-Equilibrium sedimentation experiments were performed using a Beckman Optima XL-A analytical ultracentrifuge using a six-channel centerpiece in an AN-60 Ti rotor. The IIA NH 2 -propeptide, purified from enterokinase cleavage of the IIA/EK/ SP-D mutant protein in PBS (pH 7.5), was analyzed at three concentrations: 0.2, 0.4, and 0.8 mg/ml. Experiments were performed at two speeds (20,000 and 28,000 rpm) at a temperature of 20°C and wavelength of 280 nm. Data were fitted using WinNonlin V1.035 (www. ucc.uconn.edu/ϳwwwbiotc/UAF.html), and a partial specific volume of 0.73 cm 3 /g was used for determining the molecular weight.

Synthesis of IIA/SP-D mutant Constructs Containing a Premature Termination Codon in the Coiled-Coil Neck Domain-
To determine the minimum sequence that can function as a trimerizing unit, mutant constructs were designed containing termination codons at specific locations within the heptad repeats of the coiled-coil. Using IIA/SP-D cDNA in pGEM-3Z as a substrate, four mutant constructs were synthesized using the QuikChange TM site-directed mutagenesis kit (Stratagene). The sequence of the mutant was confirmed by DNA sequencing. Mutant IIA/SP-D cDNA constructs were excised from pGEM-3Z by EcoRI digestion and subcloned into pEE14. Correct orientation of the mutant cDNA insert in pEE14 was confirmed by restriction enzyme digestion (HindIII and BglII) and agarose gel electrophoresis. CHO cells were transiently transfected with each mutant construct using FuGENE 6 reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. Proteins were precipitated from the conditioned medium overnight at 4°C with 33% ammonium sulfate. Precipitated proteins were washed three times in saturated ammonium sulfate, resuspended in PBS, and dialyzed overnight in cold PBS. Chemical cross-linking of each mutant protein was carried out as described above. Proteins were detected by SDS-PAGE and immunolocalization of Western blots using the anti-IIA polyclonal antibody.
Synthesis of a Collagen Deletion Mutant Construct-To determine whether the minor collagen domain of the IIA NH 2 -propeptide contributes to the stability of the truncated fusion protein, we generated a related truncation mutant with an associated deletion of the collagen sequence (see Fig. 8, mIIA-218). One pair of oligonucleotide primers was designed to amplify exons 1 and 2 of the IIA NH 2 -propeptide (upstream oligomer A, ggtacgaattcatgattcgcctcggggct; downstream oligomer B, taaggatccaactttgctgcccag). Another pair was designed to amplify exon 8 to the 3Ј CRD region of SP-D (upstream oligomer C, aatggatccaactttgctgcccag; downstream oligomer D, gtaccgaattctcagaactcacag). The BamHI site is shown in bold. Two separate PCRs were done using the mutant chimeric cDNA construct containing a premature stop codon at the end of the second heptad repeat of SP-D (see Fig. 8, mIIA-218) as a substrate. A cDNA fragment (ϳ300 bp) was amplified using oligomers A and B for 30 cycles (95°C, 30 s; 55°C, 30 s; 72°C, 30 s), and another cDNA fragment (ϳ650 bp) was amplified using oligomers C and D for 30 cycles (95°C, 30 s; 55°C, 30 s; 72°C, 1 min, 30 s). Each DNA fragment was digested with BamHI, ligated together, and another round of PCR was done, using oligomers A and D, for 30 cycles (95°C, 30 s; 55°C, 30 s; 72°C, 2 min) to amplify the ligated fragment. The resulting cDNA fragment devoid of exons 3-7 encoding the collagen domain of the IIA NH 2 -propeptide was cloned into pEE14. Orientation of the cloned insert was confirmed by restriction mapping and DNA sequencing. CHO cells were transiently transfected with the collagen deletion mutant construct using FuGENE 6 reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. Proteins were precipitated from the conditioned medium overnight at 4°C with 33% ammonium sulfate. Precipitated proteins were washed three times in saturated ammonium sulfate, resuspended in PBS, and dialyzed overnight in cold PBS. The collagen deletion mutant protein (mIIA-coll-218) was crosslinked using BS 3 and detected by SDS-PAGE and Western blotting using the anti-IIA polyclonal antibody.  Fig. 2. IIA/SP-D was efficiently purified from all other contaminating proteins present in the conditioned medium of stably transfected CHO cells after maltosyl-agarose chromatography (Fig. 3A). The monomer protein showed an apparent molecular mass of 45 kDa in the absence of sulfhydryl reduction when compared with globular protein standards used in this gel system. Interestingly, a small population of stable trimers of IIA/SP-D, resistant to SDS treatment and boiling prior to gel electrophoresis, was also visualized. Similar stable trimers of the type I procollagen NH 2 -propeptide have been detected from bone (29).

Purification of IIA/SP-D Fusion Protein-
Immunoblotting of the EDTA-eluted protein with anti-IIA, anti-IIE3-8, or anti-SPD polyclonal antisera confirmed identification of IIA/SP-D. Results were identical with all three antibodies, and Fig. 3B shows the immunopositive IIA/SP-D bands after detection with the anti-Exon3-8 antibody. The fusion protein migrated more slowly after sulfhydryl reduction due to unfolding of the looped structure created by the formation of the two intrachain disulfide bonds present within the lectin domain of SP-D. Although the type IIA NH 2 -propeptide domain is predicted to contain five intrachain disulfide bonds, the loops are comparatively small, and disruption of these bonds did not alter the electrophoretic migration of the protein (results not shown). However, disruption of the cysteine pairs within the IIA NH 2 -propeptide altered the structure of the exon 2-encoded domain such that recognition of the epitope by the anti-IIA antibody was affected (results not shown). All 10 cysteine residues in this domain are paired because reaction of IIA/SP-D with Ellman's reagent (Pierce) showed no quantifiable yellow-colored product as would be expected in the presence of free sulfhydryl groups. This suggests the presence of a very intricately folded domain since the 10 cysteine residues within the type IIA NH 2 -propeptide are arranged in close proximity to each other (Fig. 1).
Analysis of the IIA NH 2 -Propeptide Collagen Domain-To investigate the structure of the recombinant IIA NH 2 -propeptide, IIA/SP-D fusion protein was digested with purified bacterial collagenase, and the major collagenase-resistant bands were characterized by N-terminal sequencing. SP-D, which contains its own collagen domain, was included as a control. As shown in Fig. 4A (protein bands 1-3), most of the GXY collagen domain in IIA/SP-D and SP-D was digested (Fig. 4B). In addition, amino acid analysis of IIA/SP-D showed that there were 8 hydroxyproline residues in the collagen domain of the IIA NH 2propeptide. There are 11 potential sites for proline hydroxylation (GXP), but it is not known what percentage of prolines is hydroxylated in the native type II propeptide. To further determine the trimeric configuration of the IIA NH 2 -propeptide, we chose to purify the propeptide from the neck/CRD of SP-D.
This was done by cleavage of the wild-type IIA/SP-D protein with MMP-9 or by digestion of the mutant fusion protein (IIA/ EK/SP-D) synthesized with an enterokinase cleavage site within the exon 8-encoded telopeptide domain (Fig. 5A). After cleavage, the digested protein fragments were applied to a maltosyl-agarose column, which binds to the trimeric neck/ CRD fragments. The IIA NH 2 -propeptide was present in the flow-through, and the SP-D fragments were then eluted with EDTA (Fig. 5B).
To confirm that the IIA NH 2 -propeptide contained a correctly folded collagen triple helix, the propeptide purified by enterokinase cleavage of IIA/EK/SP-D was analyzed by CD spectroscopy. The CD spectrum of a collagen triple helix is characterized by a small positive peak at 220 -225 nm, a cross-over at 213 nm, and a trough at ϳ197 nm (30). Fig. 6A shows a large positive ellipticity at 225 nm, indicative of a collagen triple helix. The IIA propeptide was heated to 70°C ,and the CD spectrum was monitored at 225 nm. Fig. 6B shows that the mean residue ellipticity () decreased with increasing temperature and that the melting temperature of the collagen triple helix was ϳ42°C. Final confirmation that the IIA propeptide exists as a trimer in solution was achieved by analytical ultracentrifugation, using the sedimentation equilibrium approach, to calculate the molecular weight. The expected molecular weight of the trimeric propeptide was estimated using a Prot-Param program (us.expasy.org/tools/protparam.html) and was found to be 50,118 g/mol. The actual molecular weight calculated using the sedimentation equilibrium method was 50,838 g/mol.
A Trimerization Domain Is Necessary for the Production of a Correctly Folded IIA NH 2 -Propeptide-Chemical cross-linking was used to examine the state of oligomerization of the IIA collagen domain. In particular, cross-linking profiles were compared for: 1) the wild-type IIA/SP-D fusion protein, 2) the IIA NH 2 -propeptide purified after MMP-9 cleavage of the fusion protein, and 3) the IIA NH 2 -propeptide synthesized without fusion to the neck/CRD domains of SP-D. As shown in Fig. 7, cross-linking resulted in the dose-dependent appearance of IIA/ SP-D trimers (T) through a dimeric (D) intermediate. As expected, the isolated IIA NH 2 -propeptide showed a similar crosslinking pattern. By contrast, the IIA NH 2 -propeptide expressed in the absence of SP-D sequence showed no evidence of crosslinked dimers or trimers, indicating the secretion of monomers (M).

A 14-amino Acid Sequence of the Coiled-Coil Neck Domain Can Drive the Trimerization of the IIA NH 2 -Propeptide-
The trimerization domain of rat SP-D is a coiled-coil structure that consists of four heptad repeats as depicted in Fig. 8. To further assess the relative contributions of subregions of the neck domain, IIA/SP-D truncation mutants were synthesized by introducing premature stop codons within the coiled-coil neck domain to produce the IIA NH 2 -propeptide attached to one, two, or three contiguous heptad repeats. Two additional mutant IIA/SP-D proteins were generated as controls. One contained a stop codon at the first amino acid of the neck domain (Asp 203 ) or at the final residue (Gly 237 ) of the 35-amino acid sequence originally identified as the SP-D trimerization unit (17,18) (Fig. 8). Each mutant protein was covalently crosslinked, and the presence of protein monomers, dimers, or trimers was detected by immunoblotting using the anti-IIA antibody. Fig. 9 shows that the IIA NH 2 -propeptides lacking the neck domain sequence (mIIA-203) or fused to the first heptad repeat (mIIA-211) were secreted as monomers. However, trun-cated fusion proteins containing two or three heptad repeats (mIIA-218 and mIIA-225, respectively) showed trimeric assembly. The IIA NH 2 -propeptide attached to the 35-amino acid stretch of the coiled-coil neck (mIIA-237) was also secreted as a non-covalent trimer. However, lower concentrations of crosslinker (0.1-0.2 mM) were sufficient for detection of mIIA-237 trimers as compared with concentrations used to detect trimers of the other truncated mutant proteins (0.5-1 mM), and no dimeric intermediate was identified.
Cooperativity Exists between the IIA NH 2 -Propeptide Collagen Domain and the 14-amino Acid Sequence of the SP-D Coiled-Coil Neck Domain-Based on published literature (31), it was shown that a two heptad repeat coiled-coil sequence cannot form an autonomous folding unit. Thus, it is highly likely that cooperative interactions exist between the collagen domain and the short, 14-amino acid sequence of the SP-D trimerization domain to stabilize the truncated fusion protein (Figs. 8 and 9, mIIA-218). A collagen deletion construct was synthesized to produce a mutant protein consisting of exons 1, 2, and 8 of the IIA NH 2 -propeptide fused to two heptad repeats of the coiled-coil domain (mIIA-coll-218; Fig. 10A). Protein from conditioned medium of transiently transfected CHO cells was The IIA NH 2 -propeptide purified from enterokinase cleavage of IIA/EK/SP-D fusion protein was analyzed by CD spectroscopy. As shown in A, the CD spectrum shows a large positive ellipticity at 225 nm, indicative of a collagen triple helix. As shown in B, the melting temperature of the collagen helix in the IIA propeptide is ϳ42°C as shown by the decrease in ellipticity with increasing temperature from 5 to 70°C. ϭ mean residue ellipticity. cross-linked with BS 3 and detected by SDS-PAGE and Western blotting using the anti IIA antiserum. Fig. 10B shows that the fusion protein is still monomeric after addition of the highest concentration of cross-linker, confirming the importance of the collagen domain in the stabilization of the protein. DISCUSSION These studies demonstrate that a short, 14-amino acid sequence derived from the first two heptad repeats of the ␣-helical coiled-coil domain of rat SP-D can drive the trimerization of a heterologous procollagen NH 2 -propeptide sequence. Although IIA propeptides alone are secreted as monomers, a IIA/SP-D chimera with a truncated SP-D neck domain terminating at residue 218 was sufficient to drive trimerization. Truncations at residue 211 or 203, containing one or no heptad repeats, respectively, were secreted as monomers. To our knowledge, this is the shortest sequence shown to support the trimerization of a collagen sequence.
In addition, we have shown that trimerization is accompanied by folding of the collagen triple helical domain and that, following cleavage from the SP-D sequence, the IIA NH 2propeptide retains its trimeric conformation. Amino acid analysis revealed that ϳ80% of the potential proline residues in the Y position of the collagen sequence are hydroxylated, consistent with the formation of a stable triple helix. These levels of hydroxylation are comparable with those reported for the ␣1 chain of the NH 2 -propeptide of type I procollagen extracted from developing bone (29). In addition, the melting temperature of the collagen helix within the recombinant propeptide was similar to other comparably hydroxylated collagens, ϳ42°C. Interestingly, we detected a subpopulation of IIA NH 2propeptide trimers that migrated as trimers on SDS-PAGE. In this regard, Fisher et al. (29) reported that the natural type I NH 2 -propeptide is not efficiently denatured by SDS treatment prior to electrophoresis. Together, these findings indicate the synthesis of a stable, trimeric IIA NH 2 -propeptide nearly identical to that found in vivo.
The requirements for collagen trimerization and folding vary with the collagen type. Generally, fibrillar collagens and type IV collagen require the presence of globular sequences C-terminal to the triple-helical domain to initiate chain registration (3-5, 32). However, trimerization of type XII collagen is dependent on specific post-translational modifications of the col-lagen domain (33), whereas chain association of the membraneassociated collagen, type XIII, occurs in the N-terminal region (34). Refolding experiments on collagen type III indicated that interchain disulfide bridges at the C terminus of the triple helix were sufficient to function as a nucleus for the refolding of the triple helix (6). These findings suggest that the sequences required for driving collagen trimerization can be manipulated as also exemplified by our ability to trimerize a procollagen amino propeptide using the ␣-helical coiled-coil domain of rat SP-D.
There is still little precedent for the use of heterologous sequences to drive the trimerization of collagen sequences. Recently, Frank et al. (23) utilized the bacteriophage T4 fibritin foldon domain to synthesize a chimeric protein consisting of a synthetic collagen peptide (Pro-Pro-Gly) 10 fused to the N terminus of the foldon. The foldon domain (35), which consists of 27 amino acids and forms a ␤-propeller-like structure with a hydrophobic interior, was sufficient to drive the trimerization Amino acids labeled with asterisks (*) indicate residues that may participate in electrostatic interactions to stabilize the coiled-coil at its N-terminal end: Arg 208 to Glu 212 (i to i ϩ 4 intrachain) and/or Asp 203 to Arg 208 (i to i ϩ 5; g-eЈ interchain). and correct folding of the synthetic collagen domain. In addition, the thermal stability of the folded collagen peptide was increased by fusion to the foldon domain. Another recent study showed that the COOH-propeptide of type III procollagen could be replaced with a transmembrane domain without affecting the folding of the collagen triple helix (36).
The ability of a 14-amino acid sequence to direct trimerization is surprising. Previous studies have shown that a classical two heptad repeat coiled-coil sequence is unable to form an autonomous folding unit (31). Even the complete neck domain of SP-D is short as compared with many coiled-coil domains, which average 7 repeats or 14 helical turns for three-stranded coiled-coils (37). As indicated in the Introduction, the potential importance of ␤-branched side chains for determining the assembly of coiled-coils was emphasized by Harbury et al. (21). In that study, the occurrence of ␤-branched residues at the d position disfavored dimers, whereas these residues at the a position disfavored tetramers, and the presence of branched residues at both positions favored trimers. Given the occur-rence of valine residues in the first three a positions of the human SP-D neck sequence (Fig. 8), Hoppe et al. (17) suggested that this feature contributes to trimeric assembly. However, no ␤-branched amino acids occur in these positions in the rat sequence, and none of the other known SP-Ds or related collectins show a similar conservation of ␤-branched residues in this position. Studies using model peptides and surveys of known coiled-coils have identified residues that favor various oligomeric states (38 -40). Interestingly, residues found in the a and d positions of SP-D are usually non-discriminatory with respect to oligomerization or favor dimer formation. For example, leucine, which is present in the d position of the first three heptad repeats of SP-D, marginally favors dimers over trimers. Consistent with these observations, analysis of both human and rat ␣-helical coiled-coil sequences using MultiCoil (41) predicted a dimeric association. For example, dimer formation probability for the human SP-D coiled-coil sequence was ϳ90%, or 70% for the rat sequence, using the available windows of 21 residues.
Thus, it seems likely that other interactions contribute to the stability or oligomerization of the 14-amino acid sequence. In this regard, g-eЈ ionic interactions can contribute to the stability and oligomerization of some ␣-helical coiled-coils (42,43). Although most discussions emphasize the effects of electrostatic interactions on stability, Beck et al. (44) recently showed that specific electrostatic interactions were required for trimerization of the considerably longer coiled-coil domain of the cartilage matrix protein. Inspection of the neck sequence of rat SP-D suggests the possible occurrence of an intrahelical ionic interaction (i to i ϩ 4 spacing between Arg 208 and Glu 212 ) and/or an interchain ionic interaction (i to i ϩ 5 spacing between Asp 203 and Arg 208 ; g-eЈ) (Fig. 8). Studies are underway to replace the Arg 208 residue with Gln to investigate subsequent effects on trimeric assembly.
In any case, our finding that mIIA-218 is secreted as monomers, whereas IIA-218 is secreted as trimers, is consistent with contributions of the collagen domain to trimer stability and the findings of Su et al. (31). Thus, both the N-terminal heptad repeats of the neck of SPD and the IIA collagen sequence are required to form stable chimeric trimers. To our knowledge, this is the direct demonstration of a cooperative and mutually stabilizing interaction between a collagen domain and its noncollagenous trimerization domain.
We were intrigued by the finding that the mIIA-237 fusion protein reproducibly trimerizes but without a detectable dimeric intermediate. Trimerization was also more efficient, requiring less cross-linker than for the other truncation mutants. We speculate that this "all-or-none" cross-linking of mIIA-237 results from the presence of two contiguous sites for BS 3 crosslinking at Lys 229 -Lys 230 within the fourth heptad repeat. Although this seems at odds with the observation that crosslinking of IIA/SP-D also proceeds through a dimeric intermediate, the three chains may not be within an equivalent environment as compared with the context of the intact neck ϩ CRD domain. The crystal structure of the human SP-D neck ϩ CRD shows a striking deviation from 3-fold symmetry involving the fourth heptad repeat with one of the three tyrosines at position 228 buried and the other two partially exposed (20). Thus, our findings are consistent with the possibility that asymmetry is imposed on the neck by the presence of the CRD domain. Another potential implication is that the observed asymmetry exists in solution and is not simply an artifact of crystallization.
This study describes an efficient system for producing high levels of an apparently correctly folded NH 2 -propeptide of type IIA procollagen. This approach could likely be applied to the synthesis of other procollagen NH 2 -propeptides, which are difficult to isolate from tissues. Given that the propeptide is trimeric and correctly folded, it will be possible to examine the contributions of valency to the biological function of this peptide. The ability to express a secreted trimeric propeptide without inclusion of the functional lectin domain of SP-D will also enable us to investigate the effects of the propeptide in in vivo models of tissue development and repair.