Báo cáo khoa học: Tương tác khác nhau của decorin và các đột biến decorin với collagen loại I và collagen loại VI

doi:10.1111/j.1432-1033.2004.04273.x

Eur. J. Biochem. 271, 3389–3398 (2004) (cid:1) FEBS 2004

Differential interactions of decorin and decorin mutants with type I and type VI collagens

Gordon Nareyeck1, Daniela G. Seidler1, David Troyer2, Ju¨ rgen Rauterberg2, Hans Kresse1 and Elke Scho¨ nherr1,3 1Departement of Physiological Chemistry and Pathobiochemistry, University Hospital of Mu¨nster, Germany; 2Institute of Arteriosclerosis Research, University of Mu¨nster, Germany; 3Matrix Biology and Tissue Repair Research Unit, University of Wales College of Medicine, Dental School, Cardiﬀ, UK

E180K exhibited lower association and higher dissoci- ation constants to type I collagen, compared to wild-type decorin, deviating by at least one order of magnitude. In contrast, the aﬃnities of these mutants to type VI colla- gen were 10 times higher than the aﬃnity of wild-type decorin (KD (cid:1) 10)8 M). Further investigations veriﬁed that complexes of type VI collagen and decorin bound type I collagen and that the aﬃnity of collagen type VI to type I was increased by the presence of decorin. These data show that decorin not only can regulate collagen ﬁbril formation but that it also can act as an intermediary between type I and type VI collagen and that these two types of collagen interact via diﬀerent binding sites.

Keywords: collagen type I; collagen type VI; decorin; surface plasmon resonance measurements.

The small leucine-rich proteoglycan decorin can bind via its core protein to diﬀerent types of collagens such as type I and type VI. To test whether decorin can act as a bridging molecule between these collagens, the binding properties of wild-type decorin, two full-length decorin (DCN species with single amino acid substitutions E180K, DCN E180Q), which previously showed reduced binding to collagen type I ﬁbrils, and a truncated form of decorin (DCN Q153) to the these collagens were investi- gated. In a solid phase assay dissociation constants for wild-type decorin bound to methylated, therefore mono- meric, triple helical type I collagen were in the order of 10)10 M, while dissociation constants for ﬁbrillar type I collagen were (cid:1) 10)9 M. The dissociation constant for type VI was (cid:1) 10)7 M. Using real-time analysis for a more detailed investigation DCN E180Q and DCN

bands, within the gap region of the collagen ﬁbrils, by the ﬁlamentous beaded structures of the type VI collagen- containing network [4,5].

Collagens can be divided into several subfamilies according to their quarternary structure and their localization in tissue [1,2]. The largest subfamily is represented by the banded ﬁbril forming collagens type I, II and III, which are characterized by long, uninterrupted triple helical domains that assemble laterally to form ﬁbrils. In contrast, another subfamily, of which type VI collagen is the only member, is characterized by the formation of multimolecular, ﬁlamen- tous beaded structures [3]. Although banded ﬁbril forming and ﬁlamentous beaded collagens form independent net- works, they intermingle with each other in vivo, this association providing for mechanical stabilization of tissues. Electron microscopic studies indicate that the banded ﬁbril forming collagens are traversed speciﬁcally near their (cid:1)d(cid:2)

Collagen ﬁbrils in tissues are heteropolymers of several types of collagen and of noncollagenous components. For example, collagen ﬁbrils in skin are composed primarily of type I collagen with minor amounts of type III and type V collagen. Type III collagen is found on the ﬁbrillar surface, while type V collagen is buried mainly within the ﬁbrils [6]. Noncollagenous matrix glycoproteins are additionally associated with the surface of the collagen ﬁbrils. Such glycoproteins may in part substitute for collagen species at the ﬁbrillar surface or perform auxiliary functions [7]. Some of these matrix glycoproteins contain leucine-rich repeat structures and have been shown to modulate collagen ﬁbrillogenesis and the spacing between the mature ﬁbrils. The chondroitin/ dermatan sulphate proteoglycan decorin (DCN) is a member of this family of small leucine-rich proteoglycans (SLRP), which is composed of a core protein and a single covalently linked glycosaminoglycan chain. It binds to collagen ﬁbrils near the d bands ((cid:1)decorates(cid:2) them) and delays [8,9]. the lateral assembly of collagen ﬁbrils Consequently, targeted disruption of the decorin gene in mice leads to abnormal fusion of collagen bundles and to increased fragility of skin [10]. Recently, mice twofold

Correspondence to E. Scho¨ nherr, Matrix Biology & Tissue Repair Research Unit, University of Wales College of Medicine, Dental School, Heath Park, Cardiﬀ CF14 4XY, UK. Fax: + 44 29 2074 4509, Tel.: + 44 29 2074 2595, E-mail: schonherreh@cardiﬀ.ac.uk Abbreviations: BGN, biglycan; CS/DS, chondroitin sulphate/derma- tan sulphate; DCN, decorin; GAG, glycosaminoglycan; SLRP, small leucine-rich repeat proteoglycan. Note: G. Nareyeck and D. G. Seidler contributed equally to this work. (Received 13 May 2004, accepted 30 June 2004)

3390 G. Nareyeck et al. (Eur. J. Biochem. 271)

(cid:1) FEBS 2004

Preparation of methylated type I collagen and type VI collagen

deﬁcient in the SLRP decorin, biglycan, ﬁbromodulin and lumican have been generated [11]. Interestingly, double deﬁciency in decorin and biglycan manifests itself in extremely abnormal architecture of the collagen ﬁbrils. Thus, interactions between the collagens and decorin are of paramount importance in attaining and maintaining tissue integrity.

Type I collagen was isolated from calf skin and methylated by treatment with 0.2 M methanolic HCl for 3 days at ambient temperature as described [17]. This treatment results in the modiﬁcation of about 70% of all carboxyl residues. This modiﬁcation leads to an increase in the pH and a decrease in the hydrophilic properties. Type VI collagen was solubilized from bovine placenta by pepsin treatment and puriﬁed by salt fractionation [18].

Surface plasmon resonance analysis

In the present study we investigate the binding properties of wild-type decorin, two decorin mutants and a truncated decorin species with type I and type VI collagen to demonstrate that decorin can act as a bridging molecule. The results indicate that the tertiary structure of decorin is stabilized by the glycosaminoglycan chain. Furthermore, decorin may form a dimer which is capable of interacting concurrently with both type I and type VI collagen molecules.

Experimental procedures

Expression and preparation of recombinant proteoglycans

Wild-type decorin and the decorin mutant DCN E180K were expressed in human kidney 293 cells as previously described [12]. DCN E180K harbours an amino acid exchange at amino acid E180, which is an important site for the interaction of decorin with type I collagen ﬁbrils. A plasmid harbouring the cDNA for the decorin mutant DCN E180Q was generated from the respective wild-type plasmid by a one-step site-directed mutagenesis procedure (Stratagene) using the primer pair 5¢-GGTGCCCAGTTG TATGACAATC-3¢ and 5¢-GATTGTCTACAACTGGG CACC-3¢. The amino acid at E180 in DCN E180Q is likewise substituted. A cDNA construct for the truncated decorin species DCN Q153 (M1–Q153), in which six of a total of 10 leucine-rich repeats are lacking [13], was cloned into the EcoRI/XbaI site of pcDNA 3.1 (Invitrogen) and used for transfection with the Lipofectin (Life Technologies) method. Biglycan (BGN) was expressed in 293 cells as described [14].

All measurements were performed with a BIAcore 1000 analyser (Pharmacia Biosensor). Methylated type I collagen was immobilized via its primary amino groups to a research grade CM5 sensor chip [19]. During immobilization, a ﬂow rate of 10 lLÆmin)1 of HBS was maintained. The surface of the chip was activated by injecting a mixture of equal volumes of 0.2 M N-ethyl-N¢-(3-dimethylaminopropyl)- carbodiimide and 0.05 M N-hydroxysuccinimide. There- after, 70 lL of a solution of methylated type I collagen (250 lgÆmL)1) in 20 mM sodium acetate pH 4.0 was injected followed by 1 M ethanolamine/HCl pH 8.5. Injec- tion times were chosen to achieve about 6000–7000 resonance units (6–7 ng of proteinÆmm)1 [19]. Type VI collagen was immobilized via its free sulfhydryl groups. During immobilization, the ﬂow rate of HBS was main- tained at 5 lLÆmin)1. The surface was activated as described and allowed to react with 80 mM 2-(2 pyridinyldithio)ethane amine in 0.1 M sodium borate pH 8.5. At least ﬁve coupling pulses of 240 lL type VI collagen (250 lgÆmL)1) in 0.1 M sodium formiate pH 4.3, were applied until 6000–7000 resonance units were present. The sensor surface was blocked with 50 mM L-cysteine in 0.1 M sodium formiate pH 4.3, 1 M NaCl. BSA was immobilized and used to determine the proportion of nonspeciﬁc binding. The sensor surfaces were regenerated with 1 M NaCl in running buffer for type I collagen-coated chips and with 0.3 M NaCl in the case of immobilized type VI collagen.

To form a complex consisting of decorin, type I and type VI collagen, type VI collagen was ﬁrst immobilized. After binding of decorin the sensor chip surface was not regenerated in order to maintain a high level of bound proteoglycan. Methylated type I collagen in NaCl/Pi was then added to the chip and allowed to interact with the proteoglycan.

All experiments were carried out at 25 (cid:2)C at a ﬂow rate of 10 lLÆmin)1. The response to the running buffer was deﬁned as the baseline level, and all responses were expressed relative to this baseline. Experimental procedures and conditions leading to precipitation of protein complexes in the ﬂow system and the pump of the BIAcore 1000 instrument had to be strictly avoided to protect the system from damage. For this reason only experiments without the addition of complexes were performed. For the analysis of interactions between proteoglycans and collagens, the sensograms were corrected by a modiﬁcation of the method of Roden and Myszka [20]. To correct for changes in refractive index and nonspeciﬁc binding, the responses obtained with immobilized albumin were subtracted from

All proteoglycan preparations were obtained from conditioned media of transfected 293 cells under condi- tions without denaturing and/or precipitation steps. Media, supplemented with protease inhibitors, were applied directly to a DEAE-Trisacryl M (Serva) and then to a BioGel TSK DEAE-5PW HPLC column (Bio- Rad) as described previously [15]. Proteoglycans were stored at 4 (cid:2)C in elution buffer (10 mM Tris/HCl pH 7.4, containing (cid:1) 0.6 M NaCl). Immediately prior to use the proteoglycans were dialysed against either 18 mM sodium phosphate pH 7.4, 0.15 M NaCl (NaCl/Pi) or 10 mM Hepes pH 7.4, 0.15 M NaCl, 3.4 mM EDTA, 0.005% (v/v) Tween-20 (HBS) at 4 (cid:2)C. Glycosaminoglycan-free core protein was generated by exhaustive digestion with chondroitin ABC lyase (Seikagaku Kogyo) as described previously [15]. Glycosaminoglycan chains were liberated by reductive b-elimination with 1 M sodium borohydride in 0.1 M NaOH for 24 h at 37 (cid:2)C, followed by dialysis and rechromatography on BioGel TSK DEAE-5 PW as [35S]Sulphate-labelled and [35S]methio- described above. nine-labelled decorin from 293 cells and skin ﬁbroblasts were obtained as described previously [12,16].

Decorin interacting with type I and VI collagen (Eur. J. Biochem. 271) 3391

(cid:1) FEBS 2004

those obtained with bound collagen. The experimental data were then evaluated with the BIAEVALUATION 3.0 software.

Other binding assays The binding of [35S]sulphate-labelled decorin species and [35S]sulphate-labelled biglycan to reconstituted type I colla- gen ﬁbrils was performed as described [12]. Solid phase assays on hydrophilic ELISA strips (Nunc) were used to investigate interactions with type VI collagen and methyla- ted type I collagen. Type VI collagen (4 lgÆmL)1, 50 lLÆwell)1) and methylated type I collagen (10 lgÆmL)1, 100 lLÆwell)1) in 50 mM NaHCO3 pH 9.6, were incubated for 16 h at 4 (cid:2)C. After blocking with 3% BSA in NaCl/Pi, 0.05% Tween-20 for 4 h at 37 (cid:2)C, the wells were washed twice with ice-cold NaCl/Pi. Labelled proteoglycans in NaCl/Pi (18 mM sodium phosphate pH 7.4, 0.15 M NaCl) were applied for 6 h or 3 h at 37 (cid:2)C. After extensive washing with blocking solution, bound proteoglycans were solubilized with 0.1 M NaOH and neutralized with 0.1 M HCl prior to scintillation counting. KD values were deter- mined using PRISM 3.0 (GraphPad Software).

CD spectroscopy

A Jobin-Yvon CD6-Dichrograph spectropolarimeter (Yvon, France) was used to measure CD spectra at ambient temperature in NaCl/Pi in a 0.1-mm path length quartz cell. Proteoglycan concentrations of (cid:1) 1 mg proteinÆmL)1 were used. Estimations of secondary structure were per- formed with the CDNN 2.1 software (ACGT Progenomics AG, Halle (Saale), Germany).

Electron microscopy

Suspensions of type I collagen in glycerol were sprayed onto mica sheets with an air brush and rotary shadowed with platinium-carbon at an angle of about 7(cid:2), followed by pure carbon as described by Cohen et al. [21]. The replicas were placed on uncoated grids and analysed with a Philips EM 410 electron microscope.

Results

Characterization of puriﬁed type I and type VI collagens

long chain and 37% the short fragment of the a3(VI) chain [22].

Characterization of puriﬁed decorin and its mutants

Wild-type decorin, DCN E180K, DCN E180Q and DCN Q153 were puriﬁed under nondenaturing conditions from conditioned medium of 293 cells transfected with the respective cDNA. No freeze-drying or precipitation steps were performed to avoid artiﬁcial complex formation of proteoglycans [23]. CD spectroscopy was used to determine whether there are major differences in secondary structure between the wild-type decorin and the mutants. The CD spectra of wild-type decorin and DCN E180Q and DCN E180K appeared similar, whereas that of DCN Q153 differed from that of wild-type decorin (Fig. 3). The CD

Type I collagen ﬁbrils were generated by neutralization of acid soluble calf skin collagen as described previously [12]. To obtain type I collagen monomers, type I collagen was methylated which shifts the isoelectric point of the molecule to a basic pH and increases hydrophobicity. The treated collagen does not form ﬁbrils under physiological condi- tions which was conﬁrmed by rotary shadowing (Fig. 1A). However, the methylated type I collagen was still able to bind to hydrophilic ELISA strips (see below). Bovine type VI collagen containing three polypeptide chains, a1(VI), a2(VI) and a3(VI) covalently linked via disulphide bonds was produced by treatment with pepsin to remove most of the C- and N-terminal globular domains (Fig. 1B). Fig. 2 shows the composition of the collagen used in the experi- ments. Quantitative analysis of the SDS gel electrophoreses indicated that 63% of the type VI collagen contained the

Fig. 1. Rotary shadowing of isolated type I and type VI collagen mole- cules. (A) Methylated type I collagen was visualized as monomers, whereas (B) pepsin digested type VI collagen appeared as short frag- ments of beaded ﬁlaments.

3392 G. Nareyeck et al. (Eur. J. Biochem. 271)

(cid:1) FEBS 2004

Table 1. Tentative structural motifs of recombinant decorin. Theoretical calculation using the program CDNN and the data from by CD spectra measured between 195 nm to 260 nm. Decorin and the decorin mutants DCN E180Q and DCN E180K and truncated decorin DCN Q153 were puriﬁed under nondenaturing conditions from the medium of 293 cells.

Secondary structural motif Decorin mutants (%) DCN Q153 (%)

show that the point mutations have only minor effects on the general structure of the decorin core protein. For DCN Q153, which lacks most of the leucine-rich repeats, 36% a-helical motifs and only 24.1% b-sheets were observed. As shown in Fig. 4, decorin expressed in 293 cells contained no free core protein, and the length of the glycosaminoglycan chain was similar to that of decorin from dermal ﬁbroblasts.

a-Helical b-Sheet b-Turn Coil 21.7 29.1 15.4 31.1 35.9 24.1 13.8 23.5

spectra of the glycosaminoglycan chain alone, obtained by reductive b-elimination, yielded only baseline data (not shown). Evaluation of secondary structure was performed with the CDNN 2.1 software. The analysis showed that wild- type decorin and DCN E180K and DCN E180Q have 21% a-helical motifs and 29.1% b-sheets (Table 1). These results

Fig. 2. Electrophoretic comparison of the composition of pepsin digested type I collagen, acid treated type I collagen, methylated type I collagen and type VI collagen used in the experiments. Samples of the diﬀerent types of collagen were applied under reducing (+DTE) and non- reducing (–DTE) conditions to a 4–12.5% polyacrylamide gradient gel. Protein was visualized by staining with Coomassie blue.

Fig. 3. CD spectra of the recombinant decorin expressed in 293 cells and puriﬁed under nondenaturing conditions. The spectra were obtained under physiological conditions in 0.15 M NaCl. Wild-type decorin, solid line; truncated decorin DCN Q153, dotted line. Spectra for the decorin mutants DCN E180K and DCN E180Q were indistinguish- able from that of wild-type decorin (not shown). Fig. 4. Comparison of decorin synthesized in 293 cells and human skin ﬁbroblasts. 293 cells transfected with human wild-type decorin cDNA and human skin ﬁbroblasts were metabolically labelled with [35S]methionine. After immunoprecipitation with a monospeciﬁc antibody, decorin proteoglycan and core protein (obtained by diges- tion with chondroitin ABC lyase) were separated by SDS/PAGE on 12.5% polyacrylamide gel. Labelled proteins were visualized by autoradiography. Decorin transfected 293 cells do not synthesize free core protein. The molecular masses of the proteoglycan and the core protein are similar to those of the respective molecules from ﬁbroblasts.

Decorin interacting with type I and VI collagen (Eur. J. Biochem. 271) 3393

(cid:1) FEBS 2004

Interaction of different forms of decorin with type I collagen

the different

ligands

To test the hypothesis that decorin can act as a bridging molecule between type I and type VI collagen we ﬁrst performed solid phase binding assays to determine that the involved. binding properties of [35S]Sulphate-labelled decorin was incubated with reconsti- tuted type I collagen ﬁbrils and its binding was compared to the different mutants. Wild-type decorin interacted strongly with collagen ﬁbrils, DCN E180K reacted weakly and DCN E180Q moderately (Fig. 5) which agreed with earlier results [12] and conﬁrmed the suitability of the mutants.

Fig. 5. Interaction of radioactively labelled decorin and decorin mutants with reconstituted type I collagen ﬁbrils. The proteoglycans were puri- ﬁed under nondenaturing conditions as described above. Wild-type decorin (j), DCN E180Q (h), DCN E180K (m).

M for decorin and 1.4 · 10)9

To test whether methylated type I collagen monomers which were planed to be used as ligands for a decorin/ collagen type VI complex showed the expected properties solid phase assays with methylated collagen type I and decorin or biglycan as ligands were performed. ELISA plates were coated with the collagen monomers and [35S]sulphate-labelled proteoglycans were added. These solid phase assays showed dissociation constants of 2.3 · 10)10 M for biglycan (data not shown) indicating that the methylated type I collagen monomers were suitable binding partners and could be used for further studies.

experiments with isolated glycosaminoglycan chains obtained from decorin by b-elimination showed only weak interaction with type I collagen monomers. The analysis of the interaction of DCN E180K and DCN E180Q with type I collagen yielded KD ¼ 4.1 · 10–9 M and KD ¼ 1 · 10–9 M, respectively. The truncated form of decorin, DCN Q153 also showed weak interaction with type I collagen. For comparative purposes we analysed the binding of biglycan, another proteoglycan of the SLRP family, using surface plasmon resonance. A binding afﬁnity of biglycan for type I collagen monomers of KD ¼ 2.7 · 10)9 M was obtained. To test the reliability of these data v2-values were compared between the different experi- ments. Because the v2-values ranged between 0 and 2, the data were considered to be reliable (Table 2).

Interaction of decorin and decorin mutants with type VI collagen

The further analysis was performed by surface plasmon resonance spectroscopy. Collagen monomers of methylated type I collagen were covalently immobilized to a CM5 sensor chip, and the afﬁnities of wild-type decorin, its core protein (Fig. 6A) and of the decorin mutants for the immobilized collagen were measured. Surface plasmon resonance measurements with decorin and type I collagen performed in the presence or absence of 15 nM ZnCl2 showed little inﬂuence of the Zn2+ ions on the binding properties of decorin (Fig. 6B). A single binding site between wild-type decorin and type I collagen monomers with an afﬁnity of KD ¼ 5.8 · 10)10 M was found (Table 2). In addition, a KD of 2.1 · 10)8 M was observed for glycosaminoglycan-free core protein. This value was two magnitudes higher than that for wild-type decorin. Binding

Decorin is known to interact directly with banded ﬁbril forming collagens, whereas an additional, yet undeﬁned component, is thought to be involved in the binding of

Fig. 6. Surface plasmon resonance measurements of decorin binding to immobilized methylated type I collagen. (A) Interaction with wild-type decorin core protein (obtained by chondroitin ABC lyase digestion) in HBS. Decorin core protein concentrations were as indicated. (B) Interaction with wild-type decorin in HBS buﬀer (solid lines) and in HBS buﬀer containing 15 lM ZnCl2 (dotted lines). Wild-type decorin concentrations were as indicated.

3394 G. Nareyeck et al. (Eur. J. Biochem. 271)

(cid:1) FEBS 2004

Table 3. Binding of decorin and decorin mutants to type VI collagen. Type VI collagen was digested with pepsin and immobilized on CM5 chips. Surface plasmon resonance measurements were performed with decorin, diﬀerent decorin mutants, biglycan and the CS/DS chain released by b-elimination from decorin. The samples were puriﬁed under nondenaturing conditions from the medium of 293 cells. For abbreviations see Table 2.

)1)

Proteoglycan v2 KA (M KD (M) Rmax (RU) Table 2. Binding of decorin and decorin mutants to type I collagen. Type I collagen monomers were immobilized on CM5 chips. Surface plasmon resonance measurements were performed with decorin, diﬀerent dec- orin mutants, biglycan and the chondroitin sulphate/dermatan sul- phate (CS/DS) chain released by b-elimination from decorin. The samples were puriﬁed under non-denaturing conditions from the medium of 293 cells. WT, Wild-type; core, decorin digested with chondroitin ABC lyase; CS/DS, glycosaminoglycan chain from deco- rin released by b-elimination; RU, resonance units.

)1)

Proteoglycan v2 KA (M KD (M) Rmax (RU)

DCN WT DCN core DCN E180Q DCN E180K DCN Q153 BGN WT CS/DS chain 120 91 154 147 143 116 – 0.024 0.044 0.11 0.01 0.09 0.034 0.08 2.9 · 108 2.6 · 107 2.9 · 109 3.3 · 108 7.6 · 107 4.7 · 107 5 · 102 3.6 · 10)9 3.9 · 10)8 3.4 · 10)10 2.9 · 10)9 1.3 · 10)8 2.1 · 10)8 2 · 10)3

determined. Because the glycosaminoglycan chain inﬂu- enced the binding afﬁnity of decorin to type I collagen, we also studied the binding properties of glycosaminoglycan- free core protein to type VI collagen. A KD of 3.9 · 10)8 M was obtained for the core protein alone. Isolated glycos- aminoglycan chains from decorin showed a weak inter- action with type VI collagen monomers, as reported previously [25]. These binding studies constitute further evidence that the glycosaminoglycan chain stabilizes the decorin core protein.

decorin to type VI collagen [5]. Recent data showed that a complex of decorin/matrilin-1 can act as a possible linker between type VI and type II collagen [24]. In initial experiments we studied the interaction of the various decorin forms with type VI collagen in a solid phase binding assay. Decorin bound less avidly to type VI than to type I collagen (KD ¼ 3 · 10–7 M). Compared to wild-type decorin, DCN E180Q exhibited about 10-fold lower afﬁnity for type VI collagen. Unlike wild-type decorin and DCN 180Q, DCN E180K formed multimers as inferred from the nonlinear curve for the binding of radioactively labelled DCN E180K to type VI collagen (Fig. 7). However, multimers of DCN E180K were still capable of binding to type VI collagen.

The analysis of the afﬁnity of DCN E180K for type VI collagen yielded a KD of 4.1 · 10–9 M, which is similar to that obtained for wild-type decorin. Surprisingly, DCN E180Q, which showed a moderate afﬁnity for type I collagen displayed a high afﬁnity to collagen type VI (KD of 3.4 · 10)10 M), which is one magnitude lower than that found for wild-type decorin. DCN Q153 interacted only weakly with type VI collagen. For biglycan, however, the experiments revealed a KD of 2.1 · 10)8 M, which is about one order of magnitude higher than that found with type I collagen.

To study the binding of decorin and the decorin mutants to type VI collagen in a real-time experiment, we again used surface plasmon resonance measurements. The data for the binding afﬁnities to type VI collagen are summarized in Table 3. For wild-type decorin a KD of 3.6 · 10)9 M was

Considering Rmax

(where 1000 resonance units ¼ 1 ngÆmm)2) stoichiometric analysis of the surface plasmon resonance measurements revealed that a single collagen molecule binds about 0.186 decorin molecules in the presence as well as the absence of its glycosaminoglycan chain (Fig. 6A,B). The number of decorin molecules binding to type VI collagen increased from 1 : 0.042 to 1 : 0.061 by the presence of the glycosaminoglycan chain (Fig. 8A,B), indicating that the glycosaminoglycan chain does not only stabilize wild-type decorin, but can also interfere with the function of decorin. The glycosamino- glycan chain alone did not show binding properties to the collagen coated chip (data not shown). Furthermore, the amino acid exchange at position E180 resulted in a change in the binding capacity of decorin to both type I and to type VI collagen.

Formation of complexes of decorin, type I collagen and type VI collagen

In a further investigation we analysed whether the same site or similar sites on wild-type decorin, DCN E180Q and DCN E180K bind to type I and to type VI collagen.

DCN WT DCN core DCN E180Q DCN E180K DCN Q153 BGN WT CS/DS chain 386 199 349 171 282 358 – 0.92 1.9 0.02 0.34 0.24 0.12 0.94 1.7 · 109 4.7 · 107 9.8 · 108 2.4 · 108 2.7 · 107 3.7 · 108 5 · 103 5.8 · 10)10 2.1 · 10)8 1 · 10)9 4.1 · 10)9 3.9 · 10)8 2.7 · 10)9 2 · 10)4

Fig. 7. Interaction of [35S]sulphate-labelled wild-type decorin and the decorin mutants DCN E180Q and DCN E180K with pepsin digested type VI collagen in the solid phase binding assay. Wild-type decorin (d), DCN E180Q (h), DCN E180K (n).

Decorin interacting with type I and VI collagen (Eur. J. Biochem. 271) 3395

(cid:1) FEBS 2004

Table 4. Type VI collagen was digested with pepsin and immobilized on CM5 chips followed by complex formation of wild-type decorin and decorin mutants. Surface plasmon resonance measurements of the collagen/proteoglycans complexes were performed with monomers of methylated type I collagen. The proteoglycans were puriﬁed under nondenaturing condition from the medium of 293 cells. WT, Wild-type.

)1)

Type I collagen binding v 2 KA (M KD (M)

0.319 0.149 0.18 0.141 6.9 · 108 Type VI collagen 1.4 · 108 Type VI collagen/DCN WT Type VI collagen/DCN E180K 1.6 · 106 1.7 · 107 Type VI collagen/DCN E180Q 1.5 · 10)7 7.2 · 10)9 6.2 · 10)7 5.9 · 10)8

M for wild-type decorin and 6 · 10)8

Measuring the interaction of decorin with different types of collagen by surface plasmon resonance analysis we found a high afﬁnity of decorin for triple helical type I collagen compared to previously published values of 10)8)10)9 for intact and chondroitin ABC lyase-treated decorin to reconstituted type I collagen ﬁbrils [26,27]. In both of these studies the proteoglycans were treated with chaotropic agents. Our studies using decorin isolated from ﬁbroblast culture medium under nondenaturing conditions revealed two unique high afﬁnity binding sites (KD ¼ 7 · 10)10 M and KD ¼ 3 · 10)9 M) and 0.043 decorin molecules per collagen monomer [28]. The present study using methylated type I collagen revealed only one binding site with KD ¼ 5.8 · 10)10 M, a value corresponding to the measurements for the highest afﬁnity binding site in our earlier study. These relatively high values may be attributable to enhanced accessibility of the binding domain of methylated compared to nonmethylated type I collagen and/or to decorin prepared under nondenaturing conditions. The binding data of decorin to monomeric collagen type I are also in agreement with studies using type I procollagen molecules [29]. Interestingly, Tenni and coworkers [30], using a different technique, found a lower afﬁnity for the interaction of decorin with methylated type I collagen peptide frag- ments generated by CNBr cleavage. The lower afﬁnity could be due to the fact that in this study the lysyl residues were methylated, and so might interact with the amino acid E180 of decorin. Thus, compared to previously published data, variations in measurements of the afﬁnities between decorin and collagen seem to be attributable to differences in the isolation and puriﬁcation of the collagens and proteogly- cans. How important the puriﬁcation method is has recently been shown by Goldoni and coworkers [23] who demon- strated that freeze-drying and precipitation steps can lead to the formation of nonfunctional complexes of decorin and biglycan.

Type VI collagen was ﬁrst immobilized on a CM5 chip and reacted with decorin prior to adding methylated type I collagen. The results showed that the initially formed type VI collagen–decorin complexes subsequently bound methy- lated type I collagen with high afﬁnity. KD values were 7 · 10)9 M for DCN E180Q. For DCN E180K we found a KD an order of magnitude lower than for DCN E180Q. As the KD value for a complex of type VI collagen and methylated type I collagen in the absence of decorin was only 1.5 · 10)7 M, it is obvious that the presence of decorin signiﬁcantly increases the binding afﬁnity between the two collagens (Table 4).

Fig. 8. Surface plasmon resonance measurement of immobilized pepsin digested type VI collagen. (A) Interaction with wild-type decorin core protein (obtained by chondroitin ABC lyase digestion) in HBS buﬀer. Concentrations of wild-type decorin core protein used were as indi- cated. (B) Interaction with wild-type decorin in HBS buﬀer (solid lines) and in HBS buﬀer containing 15 lM ZnCl2 (dotted lines). Concen- trations of wild-type decorin used were as indicated.

Discussion

Differences in afﬁnity may also be due to other factors such as complex formation of decorin and biglycan in the presence of physiological concentrations of Zn2+ [31] or phosphate [32]. However, we found no changes in the afﬁnity of decorin to collagen type I or VI for these components (Figs 6B and 8B). This does not rule out that Zn2+ is interacting with the N terminus of decorin and may cause dimerization [31], but it did not affect the interaction with the two types of collagen.

In this study we investigated the interaction of wild-type decorin and decorin mutants with type I and type VI collagen and, for the ﬁrst time, we analysed the formation of triple complexes consisting of type VI collagen, decorin and type I collagen using surface plasmon resonance measurements.

It is known that the amino acid E180 in decorin is involved in type I collagen binding [12]. Therefore, the

3396 G. Nareyeck et al. (Eur. J. Biochem. 271)

(cid:1) FEBS 2004

involved in ﬁbrillogenesis of type I collagen and also in the generation of the microﬁbrillar network [37].

moderate KD value for binding of DCN E180Q to reconstituted type I collagen ﬁbrils and an even lower afﬁnity of DCN E180K was expected. In contrast to reports that the glycosaminoglycan chains have no inﬂuence on the binding of decorin to collagen ﬁbrils [26,33], we observed reduced binding afﬁnity for the glycosaminoglycan-free core protein. Evidently the glycosaminoglycan chain of decorin stabilizes the tertiary structure of the proteoglycans thereby causing difference in binding afﬁnity. As decorin is not the only SLRP that interacts with type I collagen the homo- logous proteoglycan biglycan was investigated. The afﬁnity of biglycan for methylated type I collagen was lower than the afﬁnity of decorin which corroborated previous data obtained with biglycan from bacteria and from osteosarcoma cells using ﬁbrillar type I collagen [28].

Analysis of the secondary structure of decorin and its mutants by CD spectroscopy showed that no signiﬁcant alterations were induced by the amino acid substitutions in the mutations compared to wild-type decorin. However, as CD spectra give only the overall proportion of different secondary structures, small changes in the distribution might not have been registered. DCN Q153, which lacks six of the 10 leucine-rich repeats of wild-type decorin, showed signiﬁcantly changed CD spectra as expected. Some changes were observed to previous results [39–41] which may be due to different expression and puriﬁcation procedures. In our expression system transfected 293 cells synthesize decorin with its normal pre- and propeptide sequences and have expression and secretion rates similar to those in normal skin ﬁbroblasts (Fig. 4). Therefore, this system resembles more the situation in normal ﬁbroblasts than does that in an overexpressing system [39]. Crystal structures of the core proteins of small leucine-rich repeat proteoglycans have not yet been published, although in analogy to the structure of the ribonuclease A inhibitor, a horseshoe arch structure has been proposed by computer modelling [42] and this structure is supported by electron microscopic observation [43]. In this model the a-helical motifs are located on the outer face of the horseshoe, while the parallel b-sheets are located inside [44].

These results do not clarify whether decorin can bind type I collagen and type VI collagen concurrently therefore

To investigate the interaction of decorin mutants E180K and E180Q with type VI collagen a solid phase assay was performed. One mutant, DCN E180K, which has a 10-fold lower afﬁnity to collagen type I than wild-type decorin had a similar afﬁnity to type VI collagen as wild-type decorin. DCN E180Q showed an even stronger binding to type VI collagen than wild-type decorin, while its afﬁnity to type I collagen was reduced. These data suggest that amino acid E180 may be not only important for the binding of decorin to type I collagen, but may also be involved in the binding to type VI. The interaction of type VI collagen with decorin has to be seen in the context that type VI collagen is responsible for the formation of the beaded microﬁbrillar network and interacts with a wide range of molecules including membrane components such as integrins [34], matrix molecules, proteoglycans [35] and matrilins [24]. The complexity of the interaction of decorin with type VI collagen suggests a possible role of decorin as a bridging molecule between the collagen molecules. Previous studies have shown that the binding of decorin to type VI collagen is less efﬁcient than the binding of decorin to type I collagen [36]. Therefore, we tried to optimize binding of type VI collagen to the sensor chip by using a free sulfhydryl group and not amino groups as in previous studies [37]. This method was avoided as it has been described that immo- bilization via amino groups may lead to manifold inter- actions of the collagen with the dextran matrix of the sensor chip, possibly masking important binding domains on the collagen and leading to rapid saturation of the chip surface [19].

M) than the binding of decorin (KD ¼ 3 · 10)9

Previous studies indicated that biglycan interacts with type VI collagen and is involved in the organization of type VI collagen networks [37]. Our ﬁndings indicate that the binding of biglycan to type VI collagen is weaker (KD ¼ 2 · 10)8 M) to this type of collagen. These ﬁndings may indicate that the lower afﬁnity of biglycan is necessary for the fast organiza- tion of the type VI network while decorin may have a more stabilizing function [24]. A further difference in the interac- tion of decorin and biglycan with type VI collagen was that decorin without glycosaminoglycan chain had a reduced binding afﬁnity, whereas the interaction of biglycan with type VI collagen was independent of the presence of the glycosaminoglycan chains. Nevertheless, the glycosamino- glycan chain plays a role in guiding type VI collagen into the organized structure both in vitro [37] and in tissue [38]. These ﬁndings are of biological importance, because decorin is

Fig. 9. Models of the potential interaction of decorin with type I and type VI collagens. Decorin is shown by the grey arrow, the tip of the arrow represents the C-terminus; the dashed line indicates the glycos- aminoglycan chain; type I collagen, black circle; type VI collagen, white cylinder. (A) Interaction involving dimerization of decorin and binding of both collagens to the inner surface of decorin. (B) Inter- action of type I collagen with the inner surface of the decorin core protein and type VI collagen binding with its noncollagenous domain to the outer surface of decorin.

Decorin interacting with type I and VI collagen (Eur. J. Biochem. 271) 3397

(cid:1) FEBS 2004

collagen of embryonic and adult human skin. J. Struct. Biol. 106, 82–90.

5. Takahashi, T., Cho, H.-I., Kublin, C.I. & Cintron, C. (1993) Keratan sulfate and dermatan sulfate proteoglycans associate with type VI collagen in fetal rabbit cornea. J. Histochem. Cytochem. 41, 1447–1457. 6. Keene, D.R., Marinkovich, M.P. & Sakai, L.Y.

(1997) Immunodissection of the connective tissue matrix in human skin. J. Micros. Res. Technol 38, 394–4060.

7. Roughley, P.J. (2001) Articular cartilage and changes in arthritis: noncollagenous proteins and proteoglycans in the extracellular matrix of cartilage. Arthritis Res. 3, 342–347.

8. Vogel, K.G., Paulsson, M. & Heinega˚ rd, D. (1984) Speciﬁc inhibition of type I and type II collagen ﬁbrillogenesis by the small proteoglycan of tendon. Biochem. J. 223, 587–597.

9. Rosenberg, L., Tang, L.H., Choi, H.U. & Johnson, T. (1993) Biological functions of dermatan sulphate proteoglycans. In Dermatan Sulphate Proteoglycans: Chemistry, Biology, Chemical Pathology (Scott, J.E., ed.), pp. 225–239. Portland Press, London. 10. Danielson, K.G., Baribault, H., Holmes, D.F., Graham, H., Kadler, K.E. & Iozzo, R.V. (1997) Targeted disruption of decorin leads to abnormal ﬁbril morphology and skin fragility. J. Cell Biol. 136, 729–743.

11. Ameye, L. & Young, M.F. (2002) Mice deﬁcient in small leucine- rich proteoglycans: novel in vivo models for osteoporosis, osteoarthritis, Ehlers–Danlos syndrome, muscular dystrophy, and corneal diseases. Glycobiology 12, 107–116.

12. Kresse, H., Liszio, C., Scho¨ nherr, E. & Fisher, L.W. (1997) Critical role of glutamate-180 in a central leucine-rich repeat of decorin for interaction with type I collagen. J. Biol. Chem. 272, 18404–18410.

13. Scho¨ nherr, E., Hausser, H., Beavan, L. & Kresse, H. (1995) Decorin–type I collagen interaction. Presence of separate core protein-binding domains. J. Biol. Chem. 270, 8877–8883.

14. Kresse, H., Seidler, D.G., Mu¨ ller, M., Breuer, E., Hausser, H., Roughley, P.J. & Scho¨ nherr, E. (2001) Diﬀerent usage of the glycosaminoglycan attachment sites of biglycan. J. Biol. Chem. 276, 13411–13416.

we used a different approach to study this interaction in greater detail. We formed dimeric complexes consisting of type VI collagen and decorin on the CM5 sensor chip and applied methylated type I collagen to the complexes. As expected the complexes were still able to bind type I collagen. Surprisingly, the stability of the ternary complex was higher than that of the dimeric complex of decorin with type I or type VI collagen. The existence of an interaction of decorin with type I and type VI collagen has been shown in vivo in skin [45]. More recently a complex formation between the globular domains of collagen type VI and a decorin/matrilin-1 complex has been described which can act as a bridge between type VI and type II collagen in cartilage, whereas decorin binds to the globular N-terminal domain of type VI collagen [24]. Even though in our study type VI collagen was treated with pepsin, electron micrographs still demonstrate the presence of globular domains, so decorin could act as a bridging molecule alone, by binding to the N terminus of type VI collagen and to type I collagen. Furthermore, the dissoci- ation constants for wild-type decorin and the two decorin mutants showed a similar relation to each other in the tertiary complex compared to the interaction with the type I or type VI collagen alone. Therefore, two binding models are possible: (a) decorin forms a dimer and can interact with the same binding site either with type I collagen or type VI collagen (Fig. 9A). This agrees with ﬁndings of Scott and coworkers [40], who reported that decorin and glycosamiminoglycan-free core proteins form dimers, how- ever, the puriﬁcation described in this paper was with freeze-drying. The dimer formation described in this paper cannot result from the puriﬁcation method, because puriﬁcation under nondenaturing conditions without freeze-drying avoids artiﬁcial dimerization [23]; (b) decorin binds via a binding site on the outer surface of the molecule to globular domains of collagen type VI and via a different binding site which is affected by E180 to collagen type I (Fig. 9B). The exact type of interaction remains to be established, but our ﬁndings show that decorin may be a bridging molecule between type I and type VI collagen networks in vitro and may be also in vivo.

15. Seidler, D.G., Breuer, E., Grande-Allen, K.J., Hascall, V.C. & Kresse, H. (2002) Core protein dependence of epimerization of glucuronosyl residues in galactosaminoglycans. J. Biol. Chem. 277, 42409–42416.

16. Glo¨ ssl, J., Beck, M. & Kresse, H. (1984) Biosynthesis of proteo- dermatan sulfate in cultured human ﬁbroblasts. J. Biol. Chem. 259, 14144–14150.

Acknowledgements

17. Rauterberg, J. & Ku¨ hn, K. (1968) The renaturation behaviour of modiﬁed collagen molecules. Hoppe-Seyler’s Z. Physiol. Chem. 349, 611–622.

18. Trueb, B., Schreier, T., Bruckner, P. & Winterhalter, K.H. (1987) Type VI collagen represents a major fraction of connective tissue collagens. Eur. J. Biochem. 166, 699–703. We thank Jo¨ rg Ro¨ sgen and Hans-Ju¨ rgen Hinz for help with CD spectroscopy and Konrad Beck for a critical reading of the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft, Sonderforschungsbereich 492 (cid:1)Extracellular Matrix: Biogenesis, Assem- bly and Cellular Interaction(cid:2), Projects A6, A9.

19. Johnsson, B., Lofas, S. & Lindquist, G. (1991) Immobilization of proteins to a carboxymethldextran-modiﬁed gold surface for biospeciﬁc interaction analysis in surface plasmon resonance sen- sors. Anal. Biochem. 198, 268–277.

References

1. van der Rest, M. & Garrone, M. (1991) Collagen family of pro- teins. FASEB J. 5, 2814–2823. 20. Roden, L.D. & Myszka, D.G. (1996) Global analysis of a macromolecular interaction measured on BIAcore. Biochem. Biophys. Res. Commun. 225, 1073–1077.

2. Prockop, D.J. & Kivirikko, K.I. (1995) Collagens: molecular biology, diseases and potentials for therapy. Annu. Rev. Biochem. 64, 403–434. 21. Cohen, C.M., Tyler, J.M. & Branton, D. (1980) Spectrin–actin association studied by electron microscopy of shadowed prepa- rations. Cell 21, 875–883.

22. Jander, R., Troyer, D. & Rauterberg, J. (1984) A collagen-like glycoprotein of the extracellular matrix is the undegraded form of type VI collagen. Biochemistry 23, 3675–3681. 3. Timpl, R. & Chu, M.L. (1994) Microﬁbrillar collagen type VI. In Extracellular Matrix: Assembly and Structure (Yurchenko, P.D., Birk, D.E. & Mecham, R.P., eds), pp. 208–242. Academic Press, Orlando, FL. 23. Goldoni, S., Owens, R.T., McQuillan, D.J., Shriver, Z., Sasisekharan, R., Birk, D.E., Campbell, S. & Iozzo, R.V. (2003) 4. Fleischmajer, R., Fisher, L.W., MacDonald, E.D., Jacobs, L., Perlish, J.S. & Termine, J.D. (1991) Decorin interacts with ﬁbrillar

3398 G. Nareyeck et al. (Eur. J. Biochem. 271)

(cid:1) FEBS 2004

Biologically active decorin is a monomer in solution J. Biol. Chem. December 3 [Epub ahead of print].

35. Bonaldo, P., Russo, V., Bucciotti, F., Doliana, R. & Colombatti, A. (1990) Structural and functional features of the alpha 3 chain indicate a bridging role for chicken collagen VI in connective tissues. Biochemistry 29, 1245–1254.

36. Bidanset, D.J., Guidry, C., Rosenberg, L.C., Choi, H.U., Timpl, R. & Hook, M. (1992) Binding of the proteoglycan decorin to collagen type VI. J. Biol. Chem. 267, 5250–5256.

24. Wiberg, C., Klatt, A.R., Wagener, R., Paulsson, M., Bateman, J.F., Heinegard, D. & Morgelin, M. (2003) Complexes of matrilin- 1 and biglycan or decorin connect collagen VI microﬁbrils to both collagen II and aggrecan. J. Biol. Chem. 278, 37698–37704. 25. Munakata, H., Takagaki, K., Majima, M. & Endo, M. (1999) Interaction between collagens and glycosaminoglycans investi- gated using a surface plasmon resonance biosensor. Glycobiology 9, 1023–1027. 37. Wiberg, C., Heinega˚ rd, D., Wengle´ n, C., Timpl, R. & Mo¨ rgelin, M. (2002) Biglycan organizes collagen VI into hexagonal-like networks resembling tissue stuctures. J. Biol. Chem. 277, 49120– 49126.

26. Brown, D.C. & Vogel, K.G. (1989) Characteristics of the in vitro interaction of a small proteoglycan (PG II) of bovine tendon with type I collagen. Matrix 9, 468–478.

38. Bruns, R.R., Press, W., Engvall, E., Timpl, R. & Gross, J. (1986) Type VI collagen in extracellular, 100-nm periodic ﬁlaments and ﬁbrils: identiﬁcation by immunoelectron microscopy. J. Cell Biol. 103, 393–404.

27. Carrino, D.A., Onnerfjord, P., Sandy, J.D., Cs-Szabo, G., Scott, P.G., Sorrell, J.M., Heinegard, D. & Caplan, A.I. (2003) Age- related changes in the proteoglycans of human skin. Speciﬁc cleavage of decorin to yield a major catabolic fragment in adult skin. J. Biol. Chem. 278, 17566–17572.

39. Krishnan, P., Hocking, A.M., Scholtz, J.M., Pace, C.N., Holik, K.K. & McQuillan, D.J. (1999) Distinct secondary structures of the leucine-rich repeat proteoglycans decorin and biglycan. Gly- cosylation-dependent conformational stability. J. Biol. Chem. 274, 10945–10950. 28. Scho¨ nherr, E., Witsch-Prehm, P., Harrach, B., Robenek, H., Rauterberg, J. & Kresse, H. (1995) Interaction of biglycan with type I collagen. J. Biol. Chem. 270, 2776–2783.

40. Scott, P.G., Grossmann, J.G., Dodd, C.M., Sheehan, J.K. & Bishop, P.N. (2003) Light and X-ray scattering show decorin to be a dimer in solution. J. Biol. Chem. 278, 18353–18359.

41. Ramamurthy, P., Hocking, A.M. & McQuillan, D.J. (1996) Recombinant decorin glycoforms. Puriﬁcation and structure. J. Biol. Chem. 271, 19578–19584.

42. Weber, T., Harrison, R.W. & Iozzo, R.V. (1996) Model structure of decorin and implications for collagen ﬁbrillogenesis. J. Biol. Chem. 271, 31767–31770.

29. Keene, D.R., San Antonio, J.D., Mayne, R., McQuillan, D.J., Sarris, G., Santoro, S.A. & Iozzo, R.V. (2000) Decorin binds near the C terminus of type I collagen. J. Biol. Chem. 275, 21801–21804. 30. Tenni, R., Viola, M., Welser, F., Sini, P., Giudici, C., Rossi, A. & Tira, M.E. (2002) Interaction of decorin with CNBr. peptides from collagens I and II. Evidence of multiple binding sites and essential lysyl residues in collagen. Eur. J. Biochem. 269, 1428–1437. 31. Yang, V.W., LaBrenz, S.R., Rosenberg, L.C., McQuillan, D. & Hook, M. (1999) Decorin is a Zn2+ metalloprotein. J. Biol. Chem. 274, 12454–12460.

43. Scott, J.E. (1996) Proteodermatan and proteokeratan sulfate (decorin, lumican/ﬁbromodulin) proteins are horseshoe shaped. Implications for their interactions with collagen. Biochemistry 35, 8795–8799. 32. Pogany, G., Hernandez, D.J. & Vogel, K.G. (1994) The in vitro interaction of proteoglycans with type I collagen is modulated by phosphate. Arch. Biochem. Biophys. 313, 102–111.

44. Kobe, B. & Deisenhofer, J. (1995) A structural basis of the interactions between leucine-rich repeats and protein ligands. Nature 374, 183–186. 33. Vogel, K.G. & Trotter, J.A. (1987) The eﬀect of proteoglycans on the morphology of collagen ﬁbrils formed in vitro. Coll. Relat Res. 7, 105–114.

45. Keene, D.R., Ridgway, C.C. & Iozzo, R.V. (1998) Type VI microﬁlaments interact with a speciﬁc region of banded collagen ﬁbrils in skin. J. Histochem. Cytochem. 46, 215–220.

34. Pfaﬀ, M., Aumailley, M., Specks, U., Knolle, J., Zerwes, H.G. & Timpl, R. (1993) Integrin and Arg-Gly-Asp dependence of cell adhesion to the native and unfolded triple helix of collagen type VI. Exp. Cell Res. 206, 167–176.