Type IX collagen NC1 domain peptides can trimerize in vitro without forming a triple helix.

Synthetic peptides of the three chains of type IX collagen consisting of the carboxyl-terminal end of the COL1 domain and the complete NC1 domain were characterized by circular dichroism spectroscopy and analyzed for their ability to assemble into trimers. In vitro association and oxidation result in disulfide-linked oligomers as shown by molecular sieve chromatography and SDS-polyacrylamide electrophoresis. Whereas the individual peptides show a tendency to self-associate, when an equimolar amount of the three peptides was oxidized, a heterotrimer of the three chains was observed. This heterotrimer is recognized by a monoclonal antibody against the disulfide-linked NC1 domain of chicken type IX collagen, indicating the correct formation of the disulfide bonds. Circular dichroism measurements show that under the association conditions used, a triple helix does not form between the chains. These results indicate that these peptides contain all the necessary information for chain selection and assembly.

Synthetic peptides of the three chains of type IX collagen consisting of the carboxyl-terminal end of the COL1 domain and the complete NC1 domain were characterized by circular dichroism spectroscopy and analyzed for their ability to assemble into trimers. In vitro association and oxidation result in disulfide-linked oligomers as shown by molecular sieve chromatography and SDS-polyacrylamide electrophoresis. Whereas the individual peptides show a tendency to self-associate, when an equimolar amount of the three peptides was oxidized, a heterotrimer of the three chains was observed. This heterotrimer is recognized by a monoclonal antibody against the disulfide-linked NC1 domain of chicken type IX collagen, indicating the correct formation of the disulfide bonds. Circular dichroism measurements show that under the association conditions used, a triple helix does not form between the chains. These results indicate that these peptides contain all the necessary information for chain selection and assembly.
Vertebrate connective tissues rely on the utilization of an assortment of at least 19 types of collagen molecules (over 30 distinct gene products) to achieve their specific architectures and biomechanical properties. The biosynthesis of the collagen triple helix requires the assembly of the appropriate combination of three chains from the extensive selection of collagen polypeptides typically produced by connective tissue cells. From studies of the homologous group of "interstitial" collagens (types I, II, III, V, and XI), it was determined that this function resides in the carboxyl-terminal propeptides (1)(2)(3). For example, chondrocytes in fetal cartilage synthesize three interstitial collagenous polypeptides, ␣1(II), ␣1(XI), and ␣2(XI). Of the 10 possible trimers, the carboxyl-terminal propeptides direct the formation of only two predominant species, which are observed in this tissue, type II collagen, [␣1(II)] 3 , and type XI collagen, [␣1(XI)␣2(XI)␣1(II)] (4). Association of the chains via their carboxyl-terminal propeptides prior to triple helix formation satisfies a further requirement, the alignment or registration of the three collagenous domains in the appropriate one residue stagger.
It is not known whether the carboxyl-terminal propeptides direct the assembly of other collagens, some of which have substantially smaller carboxyl-terminal propeptide domains than the interstitial collagens. Types IX, XII, and XIV comprise the fibril-associated collagens with interrupted triple helices, which are a subgroup of the collagen family (5). Type IX collagen, the paradigm of this subgroup, is assembled from three genetically distinct chains, each of which contains three collagenous domains (COL1-COL3) and four noncollagenous domains (NC1-NC4). The molecule is covalently associated with the surface of interstitial collagen fibrils in cartilage (6), where it plays a role in maintaining the long term structural integrity of this tissue (7). Of the 10 possible molecules that could be formed by these three chains, only one, the [␣1(IX)␣2(IX)␣3(IX)] heterotrimer is found in cartilage (8). The carboxyl-terminal noncollagenous domains (NC1) of these chains are only 20 -28 amino acid residues long, including two cysteine residues that form three interchain disulfide bonds. In order to further test the model, that triple helix formation is mediated by an association of the carboxyl-terminal noncollagenous domains, the NC1 domains of each chain of type IX collagen were synthesized and tested for the ability to direct the association of the appropriate heterotrimer.

MATERIALS AND METHODS
Peptide Synthesis and Purification-Peptides were synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on a Milligen/Biosearch 9050 peptide synthesizer. Cleavage and deprotection were performed in trifluoroacetic acid, thioanisole, ethanedithiol, and anisole (90: 5:3:2, v/v, 2 h, room temperature), which was thoroughly degassed with nitrogen before mixing with the peptide resin to ensure that cysteine residues remained reduced after the removal of the trityl protection groups. The cleaved peptide was separated from the resin by filtration through a medium sintered glass filter. The filtrate was precipitated overnight in ethylether at 4°C and recovered by centrifugation for 10 min at 3,000 ϫ g. The soft pellet was dissolved in 20 volumes of water, degassed, and lyophilized. The dried peptide was suspended in 1% trifluoroacetic acid and purified on a Vydac C-18 reversed-phase HPLC 1 column with a mobile phase of 0.1% trifluoroacetic acid and a linear 60-min gradient of 0 -75% acetonitrile. Chromatography runs were recorded with a diode array detector and analyzed with the Millennium 2000 software (Waters). Peaks that eluted between 30 and 50% acetonitrile were pooled and lyophilized. The peptide was dissolved (2 mg/ml) in 50 mM N-ethylmorpholineacetate, pH 8.3, and cysteine residues were reduced with 50 mM dithiothreitol for 2 h at 55°C under nitrogen. The pooled products were rechromatographed and characterized by amino acid analysis, Edman sequencing, and mass spectrometry. From a theoretical yield of 500 mg, approximately 150 mg of each peptide were pure and used for association experiments.
Association Conditions-Freshly reduced peptide was suspended in 20 mM Hepes buffer, pH 7.5, containing 150 mM NaCl to a final concentration of 6 mM. Of the three peptides ␣2 was the least soluble, setting this as the maximum concentration. Oxygen was purged from * This work was supported by grants from Shriners Hospital for Crippled Children (to N. P. M., L. Y. S., D. R. K., and H. P. B.) and by Grants AR30481 and EY09908 (to R. M.) from the United States Public Health Service. The electron microscope facility at the Shriners Hospital was supported in part by the R. Blaine Bramble Medical Research Foundation and the Fred Meyer Charitable Trust Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. Shriners Hospital for Crippled Children, 3101 SW Sam Jackson Park Rd., Portland, OR 97201. Tel: 503-221-3433; Fax: 503-221-3451; E-mail: hpb@shcc.org. the association buffer with nitrogen, and all peptide manipulations were performed in a glove box under a constant flow of nitrogen. The peptides were allowed to fold into their own inherent structure for 30 min and then were mixed in equimolar amounts to final concentrations ranging from 0.1 to 2.0 mM. Lipid vesicles were prepared by suspending phosphatidylcholine dissolved in chloroform (5 mg/ml) and purged under nitrogen into a conical flask until dry. Association buffer was added, and the vesicles were formed with gentle agitation for 20 min with 2-mm glass beads. These vesicles were used to mimic membrane characteristics to the associative environment, but they also increased the reaction surface area. Therefore as a negative control for surface area, we carried out separate association reactions with 150 -212-m glass beads (Sigma). To test the association of the individual peptides, selfassociation reactions were performed using a homologous suspension under the same conditions as mentioned above. The peptides were allowed to interact for time intervals varying from 30 min to 72 h.
Disulfide Bond Formation-Multiple strategies were employed to optimize the yield of cross-linked association products. The most direct approach was to raise the pH to slightly alkaline and introduce atmospheric oxygen to the solution. This was done by removing the reactions from the nitrogen box, adding 3 M Taps buffer, pH 8.5, to a final concentration of 50 mM and vortexing gently for 2 to 24 h. Alternatively, attempts were made to mimic natural oxidation mechanisms. The oxidizing environment characteristic of the lumen of the endoplasmic reticulum is maintained by a ratio of approximately 1 mM oxidized to 2 mM reduced glutathione (9). Therefore, after raising the pH of the mixture, similar ratios of glutathione from freshly prepared stock solutions were added to final concentrations ranging from 0.1 to 5.0 mM. Protein disulfide isomerase purified from 15-day-old chick embryos (10) was added to some reactions at a final concentration of 0.2-2.0 M in the presence of reduced and oxidized glutathione to catalyze the formation of correct disulfides. After oxidation the free SH groups were alkylated with 50 mM iodoacetamide for 10 min. The samples were frozen, lyophilized, and suspended in 100 l of water for HPLC and SDS-polyacrylamide gel analysis.
Circular Dichroism-Circular dichroism measurements were performed on a Jasco J-500 instrument. After suspension in the association buffer for 30 min, spectra of each peptide were recorded in a 0.01-cm cell from 260 to 180 nm at 25°C. Concentration values for each peptide solution were determined by amino acid analysis. Analysis of the secondary structure was performed with the variable selection method (11).
Molecular Sieve Chromatography-The formation of disulfide--bonded trimers was monitored by analytical molecular sieve HPLC. 25 l of the reaction mixture were run over tandem Spherogel TSK-2000SW columns (7.5 mm ϫ 60 cm) with a mobile phase of 0.1% trifluoroacetic acid at a flow rate of 1 ml/min. The separation was monitored with a diode array detector. Chromatographic traces were extracted, and the peaks were integrated with the Millennium 2000 chromatography management system.

SDS-Polyacrylamide Gel Electrophoresis and Western
Blotting-Aliquots of the reaction mixtures and fractions of chromatography runs were analyzed on duplicate 10 -20% Tricine SDS gels (12). One gel of each set was fixed 30 min in 50% methanol, 7% acetic acid, washed twice with 5% methanol, 7% acetic acid, fixed an additional 30 min with 10% glutardialdehyde, washed extensively with water, and then stained with Colloidal Coomassie Blue (Novex). The duplicate gel was electroeluted for 30 min onto an Immobilon polyvinylidene difluoride membrane (Millipore) in 12 mM Tris, 96 mM glycine, 10% methanol, pH 8.3 at 50 mA. Membranes were incubated overnight in 1% blocking solution (Boehringer Mannheim) and incubated for 1 h at room temperature with a monoclonal antibody specific to the cross-linked NC1 domain. The blot was rigorously washed, incubated 1 h with a 1:1000 dilution of sheep anti-mouse POD (Boehringer Mannheim), and developed with enhanced chemiluminesence Western blotting detection reagents (Amersham Corp.).
Monoclonal Antibody against Chicken Type IX Collagen-A mouse monoclonal antibody against chick type IX collagen (13) was further characterized by immunoblotting and epitope mapping. For this analysis, collagens were prepared from lathyritic chick sterna as described previously (4), except that the DEAE column buffer and the sample buffers contained 0.2 M NaCl. Most of the type XI collagen, residual type II collagen, and type IX collagen lacking chondroitin sulfate chains were eluted as an unbound fraction (sample 1). Type IX collagen containing chondroitin sulfate chains was eluted at 0.5 M NaCl (sample 2). The two samples were dialyzed against 50 mM Tris/HCl buffer containing 0.1 M NaCl. For immunoblotting the samples were resolved on a 3-7.5% gradient SDS-polyacrylamide gel with or without prior reduction with 0.2% 2-mercaptoethanol followed by transfer as described previously (14). For rotary shadowing, sample 2 was dialyzed against 0.2 M ammonium bicarbonate. Incubation with antibody, rotary shadowing and electron microscopy were carried out as described previously (15). RESULTS In initial experiments we synthesized the following peptides, which comprise the entire carboxyl-terminal noncollagenous domain (NC1) of type IX collagen: chicken NC1 ␣1(IX), GYCEPSSCRMQAGQRAAGKNMKGP (16); chicken NC1 ␣2(IX), GFCEPAACLGASAYASARLTEPGAVKGPL (17); and chicken NC1 ␣3(IX), GICDTSACMGAVGASTSKKS (18).
The ␣2(IX) NC1 peptide was not soluble in aqueous buffers, thus the dried peptide was dissolved in 100% dimethyl sulfoxide diluted 10 times with association buffer to a final concentration of 3 mM. These peptides were associated with each other and oxidized repeatedly under varying conditions. HPLC sieving showed an assortment of earlier eluting peaks when the peptides were mixed, but no one peak of significant yield reproducibly appeared. Because one of the cysteines was in the third position in these peptides, it is possible that the aminoterminal charges prevented efficient association and disulfide bond formation or that the last few residues of the COL1 domain contribute to the interaction surface. We therefore extended the sequence of our peptides toward the COL1 region. It was important not to extend the peptides too far into the COL1 domain because we did not want the strong associative tendency of (Gly-Xaa-Yaa) tripeptide units to energetically override NC1 domain self-assembly. Aiming to prevent triple helical formation, we thus designed our peptides with only five tripeptide units at the amino terminus, which is below the minimum chain length dependence for the most stable tripeptide unit (Gly-Pro-Hyp) at room temperature (19). Because the partial primary sequences of both the mouse and human genes were available, we continued our studies with peptides corresponding to these mammalian sequences (Z denotes hydroxyproline): human NC1/partialCOL1 ␣1(IX), 3783 Da, GIZGVZGPZGPZGLZGFCEPASCTMQAGQRAFNKGPD (20); mouseNC1/partialCOL1 ␣2(IX), 4202 Da, GEAGRZGRZG-PVGLZGFCEPAACLGASAYTSARLTEPGSIKGP (21); and human NC1/partialCOL1 ␣3(IX), 3678 Da, GEZGLZGA-IGAQGTZGICDTSACQGAVLGGVGEKSGSRSS (22).
The human ␣2(IX) differs from the mouse ␣2(IX) by a substitution of the threonine residue in position 30 by an alanine. These peptides were synthesized, and the identity of the peptides was established by amino acid sequencing and mass spectrometry. Fig. 1 shows the behavior of these peptides on a molecular sieve HPLC column. Despite their similarity in molecular weight, the peptides show different retention times, and this allowed for the identification of monomers in mixtures. The peptides also eluted at different acetonitrile concentrations on a reversed-phase HPLC column (data not shown). The circular dichroism spectra of each of the three peptides in the buffer used for association experiments are shown in Fig. 2. The absence of a positive peak at 221 nm in conjunction with a negligible contribution from ␣-helices indicates that these peptides do not form a triple helix in the chosen buffer conditions at 25°C at a concentration of 2 mM. An analysis of the secondary structural content of the peptides using the variable selection method is given in Table I. The secondary structure of these peptides consists mainly of ␤-sheets, ␤-turns, and aperiodic structures with a very low content of ␣-helix.
The individual peptides at a concentration of 2 mM were allowed to associate in 20 mM Hepes buffer, pH 7.5, containing 150 mM NaCl for 2 h at 25°C under a N 2 atmosphere. The solutions were then adjusted to pH 8.5 with Taps and exposed to air. Oxidation was allowed to proceed for 2 h. The resulting association products were then analyzed by molecular sieve HPLC. Fig. 3 shows the chromatograms for the three peptides. The ␣1-peptide exhibits the strongest propensity for self-assembly by forming dimers, trimers, and higher aggregates as shown by chromatography and gel electrophoresis (not shown). The ␣2-peptide shows some trimer formation, whereas the ␣3-peptide remains mostly monomeric. These experiments serve as a control for the assembly and oxidation of a mixture of the three peptides.
An equimolar mixture of the three peptides was treated in the same manner as the individual peptides, and the chromatogram of the resulting association products is shown in Fig. 4. A new trimer peak emerges, indicating the formation of (␣1␣2␣3). This composition was established by purification on reversed-phase HPLC, followed by reduction and re-chromatography on the molecular sieve column. The peak labeled (␣1␣2␣3) is a new and unique peak and forms only when the peptides are mixed. The other peaks can be accounted for by the self-assembly of the individual peptides as shown in Fig. 3. The area of the (␣1␣2␣3) peak was determined by integration and when compared with that of the starting material consti-tutes an approximately 10% yield of correctly cross-linked heterotrimer. However, the (␣1␣2␣3) peak consists of more than 95% of the newly formed molecules.
Purified (␣1␣2␣3) obtained from HPLC sieve chromatography was analyzed by circular dichroism spectroscopy at 25°C (Fig. 5). The absence of a positive peak around 221 nm, in   conjunction with a low ␣-helical content, indicates that the triple helix has not formed under these conditions. At 5°C the spectrum indicates that at lower temperatures triple helix formation is possible. At this temperature the disulfide bonds stabilize the triple helix as indicated by the positive peak at 221 nm. The reduced heterotrimer would have a lower thermal stability of the triple helix, and therefore formation of the triple helix prior to oxidation is unlikely.
The structure of the heterotrimer formed was further investigated using a monoclonal antibody specific for the disulfide linked type IX collagen NC1 domain. The antibody reacts only with nonreduced type IX collagen on Western blots, and electron micrographs of antibody-antigen complexes after rotary shadowing indicate that the epitope is located within the NC1 domain (Fig. 6). Products from the individual peptide associations and the three chain association reaction were resolved on a 10 -20% Tricine SDS-polyacrylamide gel and analyzed on a corresponding Western blot as shown in Fig. 7. No immunoreactivity was found with the individual peptide associations, indicating that the epitope recognized by the antibody comprises more than one chain. All the peaks from Fig. 4 were also analyzed with the monoclonal antibody. Only the heterotrimer peak showed reactivity with the antibody. Similar gels were used to study the kinetics of association (data not shown). A substantial amount of the heterotrimer is already formed after 30 min. The gel in Fig. 7 shows the 2 and 20 h time points. According to immunoblots, the amount of trimer formed seems to increase up to 4 h, and then the formation of higher aggregates becomes more prevalent.
The process was shown to be concentration-dependent over a range from 0.5 to 2 mM peptide concentration. Increasing concentrations of peptides leads to increasing amounts of trimers but also an increased amount of higher aggregates. Under these conditions, protein disulfide isomerase did not facilitate heterotrimer formation but clearly decreased the formation of higher aggregates (data not shown). To test whether trimer formation of these peptides is facilitated by the presence of lipids, association reactions were carried out in the presence of phosphatidyl choline vesicles. Phosphatidyl choline vesicles were chosen because 60% of the naturally occurring lipids in the rough endoplasmic reticulum membrane are phosphatidyl choline. As a negative control glass beads were used to account  for general surface effects. At shorter time points the amount of heterotrimer was reduced in the presence of phosphatidyl choline vesicles, but at longer time points there was an increased amount of heterotrimer found. The monoclonal antibody identifies a band with a molecular mass of approximately 12 kDa, which is the correct theoretical molecular mass for (␣1␣2␣3). The antibody does not recognize other higher molecular mass bands or oxidation products formed by individual peptides. Because of the specificity of the monoclonal antibody, these experiments demonstrate that the correct disulfide bonds are formed in this in vitro association. DISCUSSION A single cell will simultaneously synthesize different types of collagens. How does the cell select the three appropriate chains for assembly into a triple helical molecule? The mechanism of chain selection and association has been the focus of many studies (1)(2)(3)(23)(24)(25). The information for chain selection and association of interstitial collagens was shown to reside in the carboxyl-terminal propeptides. During biosynthesis the nascent constituent chains of these collagens become partially hydroxylated at proline and lysine residues. Only after synthesis of the carboxyl-terminal propeptides do chains associate. The carboxyl-terminal propeptides become interchain disulfide-linked, and triple helix formation proceeds concurrently from the carboxyl-terminal end toward the amino-terminal end. Hydroxylation of proline residues continues until the triple helix is formed. There is evidence that folding of the triple helix is catalyzed by peptidyl-prolyl cis-trans isomerases (3, 26 -28). The carboxyl-terminal propeptides of the interstitial collagens are highly conserved, but further studies are required to determine which residues determine the selection of the appropriate chains. In this study, we investigated whether this mechanism of association also applies to fibril-associated collagens with interrupted triple helices. The small size of the carboxyl-terminal noncollagenous domains of the constituent chains of type IX collagen makes it possible to investigate this question with synthetic peptides. Our results indicate that the chain selection and association of type IX is determined by the NC1 domain. The results with the peptides that were synthesized without the triple helical segments indicate that these adjacent sequences are important for assembly. Our results indicate, however, that these sequences do not have to be in a triple helical conformation. This is in contrast to experiments with type XII collagen, which contains a homologous cysteine containing portion of the NC1 domain. In expression studies with this homotrimer disulfide bonds only formed if triple helix formation was not inhibited by inhibition of prolylhydroxylase (29). This result was interpreted to show that the triple helix needs to form for the NC1 domains to come together and/or become disulfide-linked. It is, however, not clear that a triple helical region can have enough information for the selection of the appropriate chains or the correct staggering. Another explanation could be that the proline residues that are in close proximity to the cysteine residues are required to be hydroxylated and are part of the recognition surface.
In reoxidation experiments with pepsinized type IX collagen (LMW fragment, comprising the Col1 domain and the cysteine containing region of NC1), it was shown that the ␣1and ␣2-chains formed a low amount of homotrimers, whereas the ␣3-chain did not. Dimer bands of the ␣1and ␣2-chains were observed in addition to a very weak heterotrimer band (30). In our experiments the ␣1-chain readily formed aggregates at higher concentrations and fragmented at concentrations around 0.1 mM. Both the ␣1and ␣2-chains formed homotrimers when incubated individually. Regardless of this ability to self-associate, the only new product seen upon mixing is the correct heterotrimer. The product formed is recognized by a monoclonal antibody to the native NC1 domain. The epitope for this monoclonal antibody must comprise at least two chains. None of the products formed by the individual peptide associations is recognized by the antibody. Because the chains of the collagen triple helix are staggered by one residue, this antibody is sensitive to the correct arrangement of the chains or the structure of the disulfide bonds. Therefore immunoreactivity is indicating that the heterotrimer contains the correct arrangement of disulfide bonds and folding of the NC1 domains. The yield of this heterotrimer is only about 10% when analyzed by HPLC chromatography. On SDS gels the yield would be estimated to be higher. The heterotrimer however accounts for more than 95% of the new molecules formed when equimolar mixtures were oxidized.
A lipid environment in the form of phosphatidylcholine vesicles apparently does not increase the yield, nor is there an influence with glass beads. The low affinity for trimer formation has also been found with the carboxyl-terminal domains of type IV collagen (31). We suspect that additional molecules such as molecular chaperones help this association to occur in vivo.
In summary, these experiments show that the information for chain selection in type IX collagen is found in a region comprised of the carboxyl-terminal domain of the COL1 region and in the NC1 domain. Triple helix formation however is not the driving force for this interaction.