Light and X-ray Scattering Show Decorin to Be a Dimer in Solution*

Decorin is a widely distributed member of the extracellular matrix small leucine-rich repeat glycoprotein/proteoglycan family. For investigation of its physical properties, decorin from two sources (young steer skin and a recombinant adenovirus) was used. The first sample was extracted into 7 m urea and purified, while the second was isolated from medium conditioned by 293A cells infected with adenovirus and purified without chaotropes. The only chemical differences detected between these materials were a slightly shorter glycosaminoglycan chain and the retention of the propeptide on the latter. Circular dichroism spectra of the two samples were virtually identical, showing a high proportion of β-sheet and β-turn and little α-helix. The protein cores were completely denatured in 2.25 m guanidine HCl (GdnHCl) but recovered their secondary structure on removal of chaotrope. Light scattering of material eluted from gel-filtration columns in Tris-buffered saline, pH 7.0, gave molecular mass values of 165 ± 1 kDa and 84.6 ± 4 kDa for intact decorin and the glycoprotein core produced by digestion with chondroitin ABC lyase, respectively. Intact recombinant prodecorin had a mass of 148 ± 18 kDa. These values, which are double those estimated from SDS gel electrophoresis or from the known sequences and compositions, were halved in 2.5 m GdnHCl. Data from solution x-ray scattering of intact decorin and its core in Tris-buffered saline are consistent with a dimeric particle whose protein component has a radius of gyration of 31.6 ± 0.4 Å, a maximum diameter of 98 ± 5 Å, and approximates two intertwined C shapes.

The small proteoglycan decorin is an extracellular matrix protein that has been identified in many connective tissues. Its functions are believed to include regulation of the formation and/or organization of collagen fibrils (1), inhibition of the calcification of soft connective tissues (2), modulation of the activity of growth factors such as transforming growth factor-␤ (3), and other effects on cell proliferation and behavior (4,5).
Indeed, recent evidence suggests that decorin plays an important role in suppressing tumor growth by two independent mechanisms (6,7). Decorin consists of a protein core of 329 (8,9) or 330 (10) amino acids, to which are attached a single glycosaminoglycan chain of dermatan or chondroitin sulfate (at Ser-4, Ref. 11) and three N-linked oligosaccharides (12). In chicken there is a form of decorin that carries two glycosaminoglycan chains (13). The amino acid sequence of decorin includes a series of at least 10 leucine-rich repeats (LRRs), 1 averaging 24 residues in length (14). Decorin is a member of the family of extracellular matrix small leucine-rich repeat glycoproteins/proteoglycans (15) but the LRR motif is more widely distributed and has been found in many intracellular, cell surface, and extracellular proteins (16).
There have been a number of studies of the interactions of decorin with other extracellular proteins, with cytokines such as transforming growth factor-␤ (3) and with cell-surface receptors (4,6). Nevertheless, the physical properties of this small proteoglycan, which are important for the proper interpretation and understanding of these interactions, have not been well characterized. Early light scattering studies on decorin (then called proteodermatan sulfate) extracted from pig skin suggested a very high particle mass in solution: 3.4 ϫ 10 3 kDa (17). A later study of material extracted from cow skin, as described here, gave molecular masses for the intact proteoglycan and the glycoprotein core produced by digestion with chondroitin ABC lyase in 0.15 M NaCl of 610 and 650 kDa, respectively (18). In 4 M GdnHCl the molecular masses reduced to 62 and 39 kDa; values much closer to those estimated from the mobility in SDS-PAGE or calculated from the amino acid composition (36,333 Da) plus the single dermatan sulfate chain (ϳ24 kDa, Ref. 18) and the three N-linked oligosaccharides (ϳ6 kDa). 2 The anomalously high molecular weights measured in solution were attributed to self-association that has also been reported to be a property of dermatan sulfate chains in NaCl (19). However, the exponential dependence of light scattering intensity on particle size can lead to erroneous results for samples containing small proportions of aggregated material incompletely removed during sample preparation or formed in situ. The investigation described here avoided this pitfall by using continuous monitoring of the refractive index and light scattering intensity of the effluent from gel chromatography columns. The contribution of any aggregate present in the solution is then readily detected and can be corrected for, or, if necessary, it can be eliminated by re-chromatography.
Previous studies have shown that some samples of decorin extracted from tissues with chaotropes do not regain native structure and activity once the denaturant has been removed (15). There have also been reports that some preparations of tissue-derived decorin have lower activity in functional assays than recombinant decorin (20). Therefore, here we used both extracted decorin and recombinant prodecorin purified entirely without chaotropes. The results clearly establish that both forms of decorin are dimers in solution, over a wide concentration range. Furthermore, the solution x-ray scattering profile gives insight into the shape of this protein dimer.

EXPERIMENTAL PROCEDURES
Isolation and Purification of Intact Decorin and Core Protein-Bovine skin decorin was extracted from the hide of a young steer of between 18 months and 2 years of age into deionized 7 M urea, 0.15 M NaCl, 0.05 M Tris, adjusted to pH 6.5 with HCl and containing sodium azide (0.02%, w/v) and proteinase inhibitors (10 mM disodium EDTA, 0.1 M 6-aminohexanoic acid, 5 mM benzamidine-HCl, 0.5 mM N-ethylmaleimide, 1 mM phenylmethane sulfonic acid, 5 mg/liter pepstatin, 5 mg/liter leupeptin). It was purified by a combination of anion-exchange chromatography on DEAE-cellulose, precipitation with 75% ethanol, and cetylpyridinium chloride and gel-filtration chromatography (21). Recombinant bovine prodecorin was isolated from the medium of 293A cells infected with an adenovirus engineered to express full-length bovine decorin. The Adeno-Quest™ kit (Quantum Biotechnologies Inc., Laval, Quebec, Canada), comprising a transfer vector (pQBI-AdBM5), ClaI-digested DNA from adenovirus type 5 with E1 and E3 deletions, and associated reagents, was used. The cDNA encoding the full protein sequence of bovine decorin was kindly provided by Dr. Larry Fisher of the Bone Research Branch, NIDR, National Institutes of Health, Bethesda, MD, as an EcoR1 insert in the plasmid PG-28 (10). For large-scale production of recombinant decorin, host 293A cells were grown to confluence in ''triple flasks'' of 500-cm 2 growth area (Nalge Nunc International), 6 ϫ 10 7 cells per flask, in DMEM supplemented with 10% fetal bovine serum (FBS). The medium was changed to DMEM, 5% FBS (90 ml/flask) and 4 ϫ 10 9 PFU of decorin adenovirus were added to each flask, which was then incubated at 37°C for 72 h. Cells and medium were harvested by centrifugation at 1000 ϫ g for 10 min at 4°C. The supernatant collected from 8 triple flasks (about 700 ml) was thawed and applied to a 2.6 ϫ 10-cm column of DEAE-Sephacel (Amersham Biosciences) in 0.2 M NaCl, buffered at pH 7.2 with sodium phosphate (buffer A) at 4°C. The column was washed with 5 volumes of buffer A, and bound material was then eluted with a linear gradient of 150 ml each of buffer A and buffer B (buffer A with 1.0 M NaCl). Fractions containing decorin, as detected by the dimethylmethylene blue assay for glycosaminoglycans (22), were pooled, diluted with water to reduce the concentration of NaCl to below 0.1 M, and reapplied to the same DEAE-Sephacel column, which had been washed extensively with buffer A. Elution of bound proteoglycan, which emerged as a single broad peak at about 0.5 M NaCl, was repeated as above. After extensive dialysis against water, the partially purified decorin (yield 20 mg) was freeze-dried. Portions (10 mg) of this material were dissolved in 10 ml of 20 mM Tris/HCl, 0.5 M NaCl, pH 7.2 (buffer C) and applied to a 1 ϫ 5-cm column of concanavalinA-Sepharose (Sigma Chemical Co.), equilibrated in buffer C at 4°C. After washing to remove any unbound material, as detected by monitoring the absorbance at 230 nm, bound proteoglycan was eluted as a single peak with a linear gradient of 15 ml each of buffer C and the same buffer containing 0.5 M ␣-methylmannoside. Fractions containing glycosaminoglycan were pooled, dialyzed exhaustively against water, and freeze-dried. The yield of purified recombinant decorin was about 14 mg from each batch of 700 ml of conditioned medium. This material was stored dessicated at Ϫ80°C.
Decorin core protein was prepared by digesting intact proteoglycan with chondroitin ABC lyase (EC 4.2.2.4, Seikagaku) at a ratio of 1 unit of enzyme per 6 mg of proteoglycan in 0.1 M Tris adjusted to pH 7.3 with acetic acid and containing 0.02% (w/v) sodium azide. The digest was then applied to a 0.8 ϫ 5-cm column of concanavalin A-Sepharose equilibrated in 0.05 M sodium phosphate, 0.5 M NaCl, pH 7.0 at a flow rate of 15 ml/h. After washing with about 8 column volumes of this buffer, bound proteoglycan core was eluted as a single peak with the same buffer containing 0.3 M ␣-methyl-D-mannoside (Sigma). Fractions containing protein, as monitored by absorption at 280 nm, were pooled and desalted by three successive rounds of 10-fold concentration and dilution with water in a centrifugal concentrator (Millipore Ultrafree R , nominal molecular mass cutoff 10 kDa) at 1900 ϫ g. This material was stored at 4°C.
Characterization of Decorin Preparations-Protein contents were determined by the BCA assay (Pierce) using bovine serum albumin as a standard (23). Glycosaminoglycan contents were determined by the dimethylmethylene blue dye-binding assay using pig skin dermatan sulfate (Seikagaku Corporation, Tokyo, Japan) as a standard (22). The purified proteoglycans and the protein cores produced from them by digestion with chondroitin ABC lyase or by chondroitin ABC lyase followed by N-glycosidase F (EC 3.2.2.18, Roche Molecular Biochemicals) as described (12) were examined by gel electrophoresis in 10% polyacrylamide gels containing 0.1 M Tris, 0.1 M sodium borate, 0.1% SDS, pH 8.6 (24). The reactivity of the recombinant decorin with four monoclonal antibodies (7B1, 3B3, 5D1, and 6D6) prepared against natural bovine decorin was tested by Western blotting, as described (25). The sequence of the first 27 amino acid residues of the recombinant decorin was determined in a Biolynx amino acid sequencer on a sample of the core protein produced by digestion with chondroitin ABC lyase, purified by SDS-PAGE, and blotted onto a polyvinylidene difluoride membrane (26).
Circular dichroism spectra were recorded using a Jasco J-810 spectropolarimeter on samples of proteoglycan dissolved in 25 mM sodium phosphate, pH 7.0 at protein concentrations of about 0.1 mg/ml (determined by the BCA protein assay), in thermostatted cells of 1-mm pathlength.
Gel Chromatography and Light Scattering-Samples of intact natural and recombinant decorins were dissolved at various concentrations (determined by the BCA protein assay, using bovine serum albumin as standard), in TBS (0.15 M NaCl, 20 mM Tris-HCl, 0.02% (w/v) sodium azide, pH 7.0). After centrifuging at 12,000 ϫ g for 10 min to remove particulate matter, samples of 0.25 or 0.5 ml were chromatographed on Superose TM 6 HR 10/30 and Superose TM 12 HR 10/30 columns (Amersham Biosciences) in TBS with or without 2.5 M GdnHCl. The effluent from these columns was led through a DAWN-DSP laser light scattering photometer and Optilab 903 interferometer refractometer (Wyatt Technologies, Santa Barbara, CA). Light scattering data were processed by the Zimm method, using Astra TM for Windows software, version 4.10 (Wyatt Technologies). Values of dn/dc of 0.150 and 0.170, as determined previously (18), were used for the intact skin decorin and the protein cores produced by digestion with chondroitin ABC lyase. A value of 0.160 was assumed for the recombinant prodecorin based on its lower glycosaminoglycan content compared with that of skin decorin (see below). Satisfactory performance of the refractometer and light scattering detectors was checked by chromatography of samples of bovine serum albumin.
Solution X-ray Scattering-X-ray scattering data were collected at station 2.1 (27) of the Synchrotron Radiation Source (SRS), Daresbury Laboratory. Solutions for the scattering experiments were prepared in TBS, pH 7.0, chromatographed on Superose-12 or Superose-6 columns to remove any aggregated material (see below), and then concentrated by ultrafiltration, as described above. Both full-length decorin and decorin core were measured at 4°C, whereas the latter was also measured at 20°C. The studies were performed in a standard cell at sample concentrations of 1 or 5 mg/ml, according to established procedures (28). This included the examination of the scattering data, which had been collected in individual time frames of 60 s, to ensure that no observable radiation damage, protein aggregation, or deposition on the cell windows, had occurred. After checking and averaging several runs, it was concluded that the samples were unaffected by the synchrotron X-rays and that the profiles obtained over a total data collection time of 90 min (120 min) for intact decorin (decorin core) were suitable for subsequent data analysis. A sample-to-detector distance of 1.25 m was used for both samples to cover the momentum transfer interval 0.03 Å Ϫ1 Յ q Յ 0.63 Å Ϫ1 . In view of the large size of intact decorin, additional measurements at a distance of 4.25 m were performed, permitting q-values between 0.006 Å Ϫ1 and 0.18 Å Ϫ1 . The latter provided a sufficiently large scattering range for merging profiles collected at short and long detector distances. The modulus of the momentum transfer is defined as q ϭ 4sin/, where 2 is the scattering angle and is the wavelength (1.54 Å). The q-range was calibrated using silver behenate powder and wet rat tail collagen (based on diffraction spacings of 58.38 and 670 Å, respectively). The radius of gyration, the forward scattering intensity, and the intraparticle distance distribution function p(r) were calculated from the experimental scattering data using the indirect Fourier transform method as implemented in the program GNOM (29). Particle shapes were restored from the experimental scattering profiles using the ab initio procedure based on the simulated annealing algorithm to a set of dummy spheres representing the amino acid chain of the protein (30). Since both intact decorin and decorin core exist as dimers in solution (see below), shape restorations were performed using a 2-fold symmetry axis. The program CRYSOL (31) was used for the simulation of scattering curves from models based on the crystal structures of YopM, the leucine-rich effector protein from Yersinia pestis (PDB codes 1G9U and 1JL5). Fig. 1 shows a comparison of the purified natural and recombinant decorin samples by SDS-PAGE. The only differences that are apparent are the slightly higher electrophoretic mobility and more diffuse character of the band obtained with intact recombinant decorin (lane 4), compared with the natural material (lane 1). No differences were seen between the core proteins produced by digestion of the intact proteoglycans with chondroitin ABC lyase (lanes 2 and 5). Therefore the higher mobility of the recombinant decorin is a result of its having a shorter dermatan sulfate chain. This was confirmed by the lower content of dermatan sulfate measured colorimetrically (27% compared with 35%). The more diffuse appearance of this band indicates that the glycosaminoglycan chain on the recombinant decorin is also more variable in length. The indistin-guishable mobilities of the bands seen after further digestion of the chondroitin ABC lyase core proteins with N-glycosidase F (lanes 3 and 6), demonstrate that both materials are substituted with N-linked oligosaccharides to the same extent. The recombinant and natural decorins are both recognized on Western blots by four monoclonal antibodies (7B1, 3B3, 5D1, and 6D6) raised against bovine skin decorin (25) (data not shown).

RESULTS
Decorin from cow skin has been sequenced in its entirety in our laboratory by Edman degradation (9). The sequence of the first 27 amino acids of the recombinant decorin was determined in the present study by Edman degradation on a sample of the protein core produced by digestion with chondroitin ABC lyase. The major sequence (accounting for about 75% of the yield of phenylthiohydantoin-derivatized amino acid at early cycles) was GPFQQKGLFDFMLEDEA-GIGPEEHFP. This is the sequence of the pro-form of bovine decorin. A second, minor, sequence (DEA-GI-PE), corresponding to the N-terminal sequence of mature bovine decorin, was apparent at cycles 1 to 9. We therefore conclude that proteolytic processing of the recombinant protein by the host 293A cells was inefficient and that this product is predominantly in the pro-form. The gap in the sequence at cycle 18 (or 4) corresponds to the serine residue that is substituted with the single dermatan sulfate chain in bovine decorin (11). namely 2% ␣-helix, 43% ␤-sheet, and 14% ␤-turn (32). Circular dichroism spectra recorded in the presence of 2.25 M GdnHCl (Fig. 2) show essentially complete abolition of the negative Cotton effect near 217 nm. Removal of GdnHCl by dialysis resulted in complete recovery of the original secondary struc-ture, as previously reported by us for urea (32). Fig. 3A shows the results of chromatography on Superose-12 of a sample of bovine skin decorin dissolved in TBS. The refractive index signal shows a major peak, accounting for more than 90% of the total mass of recovered material, with a small shoulder on the leading edge accounting for about 7%. The 90°l ight scattering detector shows three peaks: one at V 0 not corresponding to any detectable refractive index signal and therefore presumably caused by a small amount of fine particulate material, which was not removed by centrifugation at 12,000 ϫ g, a second associated with the shoulder on the leading edge of the main peak, and a third, corresponding to the main refractive index peak. Results for recombinant prodecorin were identical (data not shown). Light scattering by the small amount of material associated with the shoulder, which was always present in freshly dissolved samples of intact natural decorin or recombinant prodecorin, is strong and interferes with the estimation of the molecular weight of the material in the major peak. Therefore fractions corresponding to the major peak, pooled as indicated, were re-chromatographed on the same column at least once (Fig. 3B), prior to the calculation of molecular weights. Duplicate analyses gave molecular masses for natural and recombinant decorin of 165 Ϯ 1 kDa (average and range) and 148 Ϯ 18 kDa, respectively. The effects of concentration on chromatography and mass were investigated by re-running material from the main peak on the same column. No decrease in average mass was seen for intact decorin samples at peak concentrations of between 2.53 M and 59.5 nM. Chromatographic behavior, as assessed by monitoring the refractive index, was unchanged even at 2.7 nM, although values for mass could not be estimated reliably at such low protein concentrations. The dimer is therefore stable at very high dilutions and does not have a measurable dissociation constant under the conditions used here. Fig. 4A shows gel chromatography profiles and plots of molecular weight against elution volume for intact decorin and the protein core in the presence and absence of 2.5 M GdnHCl. Denaturation of intact decorin by GdnHCl caused a slight increase in elution volume but an ϳ2-fold decrease in mass (to 88.3 Ϯ 2.8 kDa). The native core protein gave only a single peak  5. Solution x-ray scattering profiles (A) and distance distribution functions (B). Results for intact decorin are shown in red and those for the glycoprotein core in blue. The smooth curves were derived from p(r) analysis. The scattering profile for the decorin core particle has been used to restore the low-resolution shape shown in Fig. 6. Scattering profiles are scaled according to molecular weight and p(r) functions are normalized to unit area. The insets in B show the Guinier plots for intact decorin and glycoprotein core (upper and lower panels, respectively). Data points that satisfy the Guinier condition q Յ 1/R g (for which the value of R g was taken from the p(r) calculation) are shown as filled symbols. Measurements were performed at concentrations of both 1 and 5 mg/ml but since no significant concentration-dependent effects were observed only the data for 5 mg/ml are shown. of mass 84.6 Ϯ 4 kDa when chromatographed in TBS but in 2.5 M GdnHCl it eluted earlier, with a 2-fold reduction in mass (to 48.2 Ϯ 3.6 kDa). This value is in reasonable agreement with the molecular mass of 42.3 kDa that can be calculated for the decorin monomer from its composition (36.3 kDa for the protein and 6 kDa for the N-linked oligosaccharides). Fig. 4B compares the chromatographic behavior and mass of samples dissolved in TBS containing 2.5 M GdnHCl and then chromatographed in TBS alone, with those dissolved directly in TBS. Material under the peaks obtained in the first case gave mass values of 112.8 kDa (for intact decorin) and 68.9 kDa (for the protein core). These values are intermediate between those determined above for the denatured monomers (in 2.5 M GdnHCl) and the native dimers (in TBS), indicating that they are mixtures of monomer and dimer. Core protein recovered from these columns gave CD spectra that were indistinguishable from those of the native protein shown in Fig. 2. Taken together, these results show that denatured monomeric decorin can reassemble into native dimers in solution rapidly when separated from chaotrope. Solution x-ray scattering profiles of intact natural decorin and decorin core and their distance distribution functions are shown in Fig. 5, A and B, respectively. The insets to the p(r) plots in Fig. 5B show the low angle regions in the form of Guinier plots (i.e. ln(I) versus q 2 ), from which the radius of gyration (R g ) can be extracted from the slope (R g 2 /3) of the straight line. Here the curves confirm the expected linearity for q Յ 1/R g based on the R g values derived from the p(r) analysis for each protein form, emphasizing that p(r) analysis and Guinier approximation are in good agreement. When calculating the radius of gyration via the p(r) function, one has to bear in mind that the smallest measured scattering angle q min Յ /D max (see e.g. Ref. 33), a condition which has been met in our experiments.
The profile for intact decorin (Fig. 5A) shows a significant contribution from glycosaminoglycan and oligosaccharides (only ϳ60% of the molecular weight is due to protein), noticeable from the steep rise in the scattering signal at very low angles, which is independent of protein concentration. This is expected from an extended, random chain-like behavior of the sugar chains. Therefore it can be assumed that the sugar chains point away from the decorin core mixing with the solvent. In contrast to the result for intact decorin, for the decorin core scattering from protein dominates the curve, which therefore shows no significant increase in intensity at very low angles. Thus the results are consistent with a globular, dimeric decorin core particle. Molecular masses of intact decorin and decorin core were checked against the forward scattering from a bovine serum albumin solution (66 kDa) and were consistent with dimers in solution. Radii of gyration (Rg) for intact decorin and decorin core at 4°C were 92.8 Ϯ 0.7 Å and 31.6 Ϯ 0.4 Å, respectively. The latter was also measured at 20°C (Rg ϭ 31.8 Ϯ 0.4 Å) with identical scattering features, showing that the conformation of the core structure does not change in this temperature range. The maximum particle dimension as deduced from the p(r) function, which represents the distribution  Fig. 6. Only models 1, 6, and 7 yield goodness of fit values () below 10, as is evident from the comparison of the theoretical curves with the experimental data (drawn in blue with error bars). The dimer in model 1 has the highest buried surface area among the three possible dimers in the crystal structure for the YopM tetramer. Nevertheless, a -value of 5.5 indicates a rather poor fit to the decorin core scattering data. Even though model 6 shows a side maximum at around q ϭ 0.33 Å Ϫ1 , similar to that observed in the experimental profile, it yielded a value of only 4.4, signifying a mediocre fit to the data. On the other hand, the model of intertwined C shapes (model 7) gave an excellent fit to the experimental curve, with a -value of 1.5. of distances between atoms (i.e. scattering centers) within the intact decorin dimer, is D max ϭ 350 Ϯ 15 Å and for the decorin core dimer 98 Ϯ 5 Å.
In order to visualize conformational properties of the core protein, shape reconstructions were performed ab initio from the scattering profiles using the dummy spheres model (30), assuming a 2-fold molecular symmetry. As can be seen from Fig. 6, the shape of the decorin core can be approximated by two ''C'' shapes intertwining. With the purpose of providing a simple comparison between the reconstructed shape and a known protein structure containing leucine-rich repeats, a ribbon model of the Y. pestis cytotoxin YopM (residues 34 -330, only) (34) has been arranged and displayed in the form of an intertwined dimer (Fig. 6, right hand side). The choice of YopM for this purpose was based on the fact that it has no ␣-helices within the LRR domain, unlike most other known leucine-rich repeat protein structures, which contain a considerable proportion of ␣-helix. As noted above, the CD spectrum of decorin indicates little if any ␣-helix. The lengths of the repeats are also quite similar (20 or 22 residues for YopM and, on average, 24 for decorin), as is the overall length.
Finally, a scattering pattern simulation using different dimer arrangements based on the YopM crystal structure (see Fig. 7), demonstrates that shapes represented by a ''horseshoe'' or ''doughnut,'' among others, are not in harmony with the data presented here. We note that despite the lack of a contribution from carbohydrate chains, which have not been modeled due to their intrinsic flexibility, the model shown in Fig. 6 provides an excellent fit to the experimental scattering data of the decorin core (yielding a goodness of fit value of ϭ 1.5).

DISCUSSION
By carrying out the light scattering analysis of intact decorin and its glycoprotein core on the continuously flowing effluent from gel filtration columns, we demonstrate here that the predominant species in solution at room temperature are, in both cases, dimers. A small amount of aggregated material with strong light scattering elutes as a shoulder ahead of the main peak when freshly dissolved samples of either natural decorin or recombinant prodecorin are chromatographed. This shoulder corresponds to about 7% of the total mass, as measured by refractive index. Once removed by chromatography it does not reform in solution, even after 2 weeks at 4°C. 2 Further characterization of this minor fraction, which appears to be generated when salt-free solutions are freeze-dried, will be described elsewhere. Failure to remove aggregates such as this in earlier light scattering studies may have led to the erroneously high estimates of particle mass.
Repeated re-chromatography on the same column of the material from the major peak, demonstrated that the decorin dimer is stable over a wide range of concentrations: no dissociation could be detected even at peak concentrations as low as 2.7 nM. Exposure to 2.5 M GdnHCl, however, converted both the intact decorin and glycoprotein core dimers to monomers with estimated values for M w close to those that can be calculated from the compositions. This concentration of GdnHCl is sufficient to destroy all secondary structure (Fig. 2), and the dimerization therefore depends on the existence of a native protein core.
Removal of GdnHCl from solutions of decorin by dialysis or by dilution during gel filtration restores the secondary structure and facilitates the rapid reformation of dimers. Since the sample of natural decorin used here had been isolated from skin using a chaotropic agent (7 M urea), it might be argued that the reversible denaturation and dimerization are artifactual. However, identical behavior was seen for a sample of recombinant prodecorin that had been isolated and purified without the use of chaotropes. The tendency to form a stable dimer in solution would therefore seem to be an intrinsic property of decorin.
Decorin belongs to a family of 12 extracellular matrix proteins that contain leucine-rich repeats. Biglycan has been shown to form dimers reversibly (35) but the aggregation states of the other extracellular matrix LRR proteins have not been characterized. However these are likely to adopt similar structural folds, resembling those in ribonuclease inhibitor (36) and other related proteins of known structure. With one exception (an Azotobacter vinelandii protein, PDB code: 1LRV) these all take the form of a partial ''solenoid'' with an inner face made up of parallel ␤-strands, a structural feature that has been suggested to provide sites for interaction with other proteins. In the case of decorin we suggest that this feature facilitates the dimerization.
Electron-microscopic images of decorin preparations have been interpreted as showing horseshoe-or arch-shaped structures, such as might be expected for a molecule with the same overall topology as ribonuclease inhibitor (37,38). The resolution available in the electron microscope does not, however, permit the detailed shape to be deduced, nor does it allow the unambiguous distinction of monomers and dimers. The appearance of the specimens may also be influenced by the preparative techniques used. Solution x-ray scattering can give a higher (but still low) resolution structure of macromolecules directly in solution. We have applied this technique to both intact decorin and its glycoprotein core. Satisfactory reconstruction of a shape for the intact decorin dimer is difficult in view of the significant contribution to the scattering from the two dermatan sulfate chains. These are presumably very flexible and their large contribution to the total molecular mass makes model building for scattering profile simulations complicated and beyond the scope of the present report. However the x-ray scattering by the decorin core particle can be modeled satisfactorily and the resulting model fitted with a dimer of YopM monomers arranged as 2 intertwined C shapes. Interestingly, YopM carries two long ␣-helices at the extremities. It is tempting to suggest that in decorin there are sugar chains contributing equivalent mass in this location.
We do not present evidence here relating directly to the aggregation state of decorin in tissues. However, the finding that it is a highly stable dimer in solution, and that dimerization probably results from interactions between the ␤-sheet surfaces, has important implications for understanding interactions with other proteins both in vivo and in vitro. It has been proposed that a horseshoe-shaped monomeric decorin, modeled on the ␣-helix rich ribonuclease inhibitor, binds to monomers of collagen through its inner ␤-sheet face (39). However the x-ray scattering data presented here are not consistent with a horseshoe shape and further imply that part or all of the ␤-sheet face is unavailable for binding. Our data suggest that in functional studies it should not be assumed that decorin is a monomer when calculating affinity constants etc.; moreover, as a dimer it may bind more than one ligand. In the light of the results described here, re-examination of these earlier assumptions and models would appear necessary.