Interaction of decorin with CNBr peptides from collagens I and II

Evidence for multiple binding sites and essential lysyl residues in collagen


R. Tenni, Dipartimento di Biochimica ‘A. Castellani’, University of Pavia, Via Taramelli 3b, 27100 Pavia, Italy. Fax: + 39 0382423108, Tel.: + 39 0382507228, E-mail:


Decorin is a small leucine-rich chondroitin/dermatan sulfate proteoglycan reported to interact with fibrillar collagens through its protein core and to localize at d and e bands of the collagen fibril banding pattern. Using a solid-phase assay, we have determined the interaction of peptides derived by CNBr cleavage of type I and type II collagen with decorin extracted from bovine tendon and its protein core and with a recombinant decorin preparation. At least five peptides have been found to interact with all three decorin samples. The interaction of peptides with tendon decorin has a dissociation constant in the nanomolar range. The triple helical conformation of the peptide trimeric species is a necessary requisite for the binding. All positive peptides have a region within the d and e bands of collagen fibrils. Two chemical derivatives of collagens and of positive peptides were prepared by N-acetylation and N-methylation of the primary amino group of Lys/Hyl side chains. Chemical modifications performed in mild conditions do not significantly alter the thermal stability of peptide trimeric species whereas they affect the interaction with decorin: N-acetylation eliminates both the positive charge and the binding to decorin, whereas N-methylation preserves the cationic character and modulates the binding. We conclude that decorin makes contacts with multiple sites in type I collagen and probably also in type II collagen and that some collagen Lys/Hyl residues are essential for the binding.


extracellular matrix


leucine-rich repeat




sulfosuccinimidyl acetate

T m

melting temperature.

Decorin is a member of the family of extracellular matrix (ECM) proteoglycans characterized by a protein core containing 10 tandem leucine-rich repeats, each of about 24 amino acids, flanked by cysteine clusters. The N-terminal domain carries one chondroitin/dermatan sulfate glycosaminoglycan chain and the protein core also has three consensus sites for N-linked oligosaccharides [1,2]. Leucine-rich repeats are involved in protein–protein interactions and have been found in a large number of proteins as well as small leucine-rich proteoglycans (PGs), such as biglycan, fibromodulin and lumican [1,3,4].

Decorin is considered a key regulator of the assembly and function of many ECMs. Decorin interacts with a variety of ECM proteins, e.g. with several collagen types, fibronectin and thrombospondin. Collagens have a characteristic triple helical conformation, due to the repetition of triplets Gly-X-Y. The triple helix has a high surface to volume ratio and the side chains of all X and Y residues are accessible by the solvent, X more than Y positions [5]. These geometric and molecular aspects determine the ability of many collagen types to self-associate, leading to defined supramolecular structures, and collagen propensity to interact with many ligands [6].

The specific association of decorin with collagens has been reviewed [1,2]. In particular, decorin plays a role in lateral growth of collagen fibrils, delaying the lateral assembly on the surface of the fibrils [7,8]. This might control fibril dimensions, uniformity of fibril diameter and the regular spacing of fibrils. The pathophysiological relevance of decorin–collagen interactions has been shown in decorin null mice: homozygous animals are characterized by skin with reduced tensile strength, containing collagen fibrils with irregular profiles due to lateral fusion [9]. Recent findings report the binding of decorin to collagen XIV and to the N-terminal region of collagen VI [10,11].

The interplay between ECMs and cells is mediated by integrins but recent evidence has shown that there are integrin-independent effects of decorin and collagen on cellular biological activity and proliferation. These effects are mediated by interactions with cytokines or cellular receptors, e.g. interactions between decorin and transforming growth factor β or between collagens and interleukin 2, or interactions between decorin and epidermal growth factor receptors or between fibrillar collagens and discoidin domain receptors [12–16]. Decorin–collagen interactions are thus probably able to modulate the influence of both macromolecules on cell activities.

Earlier modeling and recent evidence has shown that decorin is an arch-shaped molecule [17–19]. The convex surface is formed by α helices whereas the β strands lining the inner concavity contain several charged residues exposed to the solvent. The glycosaminoglycan chain and the N-linked oligosaccharides are on the same side of the molecule.

The main binding site for collagen within the decorin molecule appears to be located in leucine-rich repeats (LRRs) 4–5 with a glutamate (residue 180 of the protein core) playing a critical role and there are suggestions that decorin has a second binding site for collagen [20–22]. (For the human decorin sequence, we refer to Swiss-Prot, accession number P-07585, which reports the whole translated product still bearing a 16-residue signal and a 14-residue propeptide sequence.). As far as collagen fibrils are concerned, there is morphological evidence for the presence of chondroitin/dermatan sulfate PGs at the d and e bands in the gap zone of the fibrils formed by the quarter staggered array of type I collagen molecules, and the presence of keratan sulfate PGs at the a and c bands in the overlap zone [23,24]. A study using isolated type I procollagen molecules and decorin extracted from tissue has shown that the binding occurs preferentially at two sites around 50 and 100 nm from the N-terminus of the triple helical domain [25]. In a different study, the sequence GAKGDRGET, at position 853–861 of the α1(I) collagen chain, was reported as the binding site for decorin [26]. The KLER and RELH sequences within decorin were suggested as possible complementary sequences of GDRGET, allowing modelling of the position of decorin on the surface of a collagen fibril [18]. A further, theoretical model was postulated [17]: the molecular dimensions of the decorin structure (6.5 × 4.5 × 3 nm) are consistent with a space able to accommodate a single type I collagen triple helical molecule inside the concavity; this suggests that about 10 residues per collagen chain are present in the binding site of decorin. In contrast with previous findings, a very recent paper reported that recombinant decorin never subjected to the action of chaotropic agents binds near the C-terminus of the type I collagen α1(I) chain [19].

In this work we have tested the binding of decorin towards CNBr peptides derived from the α chains of type I and type II collagens, by using both decorin purified from tendon and its core as well as a recombinant decorin preparation. The results suggest that multiple binding sites for decorin are present in these collagens. We have also tested the influence on decorin binding of chemical modification of Lys and Hyl side chains of collagens and peptides. Derivatizations that eliminate the positive charge of Lys/Hyl eliminate the binding to decorin, whereas the binding is modulated by a modification that preserves the charge.

Materials and methods


Type I collagen from bovine skin and its CNBr peptides were already available and characterized by our laboratory [27–30].

Sulfosuccinimidyl acetate, p-nitrophenyl phosphate, avidin conjugated with alkaline phosphatase, o-phenylenediamine dihydrochloride and sulfosuccinimidobiotin were obtained from Pierce, avidin conjugated with horseradish peroxidase and a 30-kDa heparin-binding fragment of fibronectin were purchased from Sigma, chondroitinase ABC and AC II from Seikagaku Corporation, endoproteinase Arg-C (sequencing grade) from Roche, NaBH3CN (sodium cyanoborohydride) from Fluka, DEAE–Sephacel and PD-10 columns from Pharmacia, microtiter plates from Nunc. Fibronectin was a generous gift of L. Visai (Dipartimento di Biochimica ‘A. Castellani’, University of Pavia, Italy). All other reagents were of analytical grade.

Preparation and analysis of decorin from tendon

Decorin was purified as described previously [31,32]. Briefly, proteoglycans were extracted from bovine tendon with 4 m guanidine hydrochloride in 50 mm acetate buffer, 5 mm benzamidine, 0.1 mε-aminocaproic acid, 10 mm EDTA, 1 mm phenylmethanesulfonyl fluoride, pH 5.6, and purified by preparative ultracentrifugation (100 000 g) in a CsCl gradient in the presence of buffered 4 m guanidine hydrochloride. The fraction with density 1.5 g·mL−1 was adsorbed on DEAE–Sephacel and eluted with a linear 0–0.8␣m NaCl gradient in the presence of 4 m urea. Decorin was desalted on PD-10 columns, freeze-dried and stored at␣−80 °C.

The protein content of the decorin preparation was determined with Bradford's method [33]. Electrophoretic analysis in denaturing conditions was according to Laemmli [34], both before and after chondroitinase ABC digestion [35]. The analysis of disaccharides of the glycosaminoglycan chains was performed after digestion with chondroitinase ABC or AC II with standard methods [36]. Circular dichroism analysis is described below.

Decorin from tendon or its core were labeled with biotin as follows. The samples (1 mg·mL−1) in NaCl/Pi were incubated with a 20-fold molar excess of sulfosuccinimidobiotin for 2 h at room temperature. Concentrated Tris/HCl buffer, pH 7.5, was then added to 50 mm final concentration and the samples were incubated for 1 h, extensively dialyzed against NaCl/Pi and stored at −20 °C.

Preparation and analysis of recombinant decorin

A full-length cDNA encoding the complete human decorin was inserted into a mammalian expression vector designed for high-level expression of recombinant proteins. This construct was used for transfection of human embryonic kidney cells (American Type Culture Collection) and antibiotic resistant cells were selected. The synthesis of recombinant decorin was checked by electrophoresis and immunoblotting with an antiserum specific for human decorin (a kind gift from H. Kresse, Münster, Germany). For large scale production, decorin producing cells were cultivated in a controlled fermenter system. The culture medium was DMEM/F12 supplemented with 2% fetal bovine serum. The harvested culture supernatant was centrifuged and purified. For purification, the culture medium was adjusted to 250 mm NaCl and applied on a column packed with a DEAE Trisacryl matrix (Sigma) equilibrated in 250 mm NaCl, 20 mm Tris, pH 7.4. The column was washed with the same buffer. Elution of bound decorin was carried out in a step from 350 to 580 mm NaCl in 20 mm Tris, pH 7.4. The eluted fractions were passed over a Superdex 200 HR gel filtration column (Pharmacia) equilibrated and eluted with 250 mm NaCl, 20 mm Tris, pH 7.4. The fractions containing recombinant decorin were pooled. Identity was confirmed after electophoresis and immunoblotting with the mentioned decorin antiserum.

Recombinant decorin was analyzed and biotinylated as described for tendon decorin.

Preparation of type II collagen and its CNBr peptides

Type II collagen was purified from bovine nasal septum [37]. Briefly, the tissue was extracted at 4 °C for 24 h with 4 m guanidine hydrochloride in Tris/HCl, pH 7.4, in the presence of protease inhibitors. The residue was washed with water and resuspended at 4 °C for 48 h in 0.5 m acetic acid containing 1 mg·mL−1 pepsin and 0.2 m NaCl. The solubilized material was dialyzed against 0.9 m NaCl in 0.5 m acetic acid and the precipitate of type II collagen removed by centrifugation, dialyzed against 0.1 m acetic acid and freeze-dried.

Type II collagen CNBr peptides were purified essentially following the procedures used for peptides from type I collagen, by means of a combination of gel filtration chromatography followed by ion-exchange chromatography or by reverse-phase chromatography for the two smaller peptides [27,30].

All collagens and peptides were analyzed for purity by means of a quantitative Hyp assay [38], electrophoresis in denaturing conditions [34], N-terminal sequencing for some peptides and for conformation by means of CD spectroscopy.

Chemical modification of collagens and CNBr peptides

Chemical modifications have been performed with three different methods, all involving the primary amino group of lysine and hydroxylysine side chains. After the derivatization, the samples were exhaustively dialyzed against 0.1 m acetic acid, clarified by centrifugation, freeze-dried and stored at −80 °C. All derivatized samples have been analyzed for purity and conformation by the same methods as the underivatized ones.

N-Methylation.  The derivatization was performed with formaldehyde in the presence of NaBH3CN, essentially as described previously [39]. The incubation with HCHO/NaBH3CN was performed for 2 h at room temperature followed by 12–18 h in the cold room. The derivatized samples have been dialyzed against 0.1 m NaCl, and then against 0.1 m acetic acid.

N-Acetylation with acetic anhydride.  The derivatization was performed essentially as described previously [40] at 0 °C. Because acetic anhydride quickly hydrolyzes to acetic acid, the pH was maintained constant by additions of aliquots of 5 m NaOH. These additions, however, introduce local strong basic conditions whose consequence is the breakdown of some peptide bonds and the formation of new bonds leading to the presence of molecules both smaller and larger than a single monomeric peptide (see Results).

N-Acetylation with sulfosuccinimidyl acetate (SNHSAc). This procedure is much more mild than the previous one. All operations have been performed at 4 °C. Collagenous samples (5–15 mg) were suspended overnight in 10 mL of 0.5 m borate buffer, pH 8.5. Solid SNHSAc was quickly dissolved at 10.4 mg·mL−1 (40 mm) in 10 mm acetate buffer, pH 5.4–5.6, immediately before use. SNHSAc solution was added under vigorous stirring to the collagen samples in order to have a 10 : 1 molar ratio between SNHSAc and primary amino groups. The derivatization was allowed to proceed overnight.

The degree of Lys/Hyl modification was determined by a colorimetric method with sodium trinitrobenzenesulfonate, essentially as described [41], using Nα-acetyl-l-lysine as the standard. The extent of derivatization was found to be higher than 80% for most samples. A lower percentage was found for type I and II collagens when derivatized with SNHSAc (70 and 76%, respectively) and for two peptides from type II collagen when treated with acetic anhydride (56% for CB6 and 65% for CB8).

Binding assays

Collagenous samples were dissolved in 0.1 m acetic acid at 1–1.5 mg·mL−1 and maintained at 4 °C for ≥ 7 days, with occasional vortexing. The actual concentration was determined by means of a Hyp assay [38]. After clarification by centrifugation, working solutions were prepared by dilution with NaCl/Pi, at 25 µg·mL−1 for collagens I and II or equimolecular amounts of their CNBr peptides. Control dilutions determined the amount of sodium hydroxide needed to neutralize the decrease of pH.

96-Well microtiter plates were coated overnight at 4 °C with the solutions of collagenous samples in NaCl/Pi (200 µL per well). Control wells were coated with 200 µL containing 5 µg of BSA in NaCl/Pi. All analyses were done at least in triplicate. After rinsing with 0.15 m NaCl, 0.05% (v/v) Tween-20, the wells were incubated with 200 µL of 1% (w/v) BSA in NaCl/Pi, for 1 h at room temperature. After rinsing as above, the coated wells were incubated for 2 h at room temperature with 20 pmol of biotinylated decorin dissolved in 200 µL of NaCl/Pi, 0.05% (v/v) Tween-20. For Scatchard analysis, constant concentrations of collagen or peptides were used for coating and incubated with increasing concentrations of biotinylated decorin. For every solid-phase experiment, control for dose-dependent, nonspecific binding to coated BSA wells was performed, under identical conditions.

Bound decorin from tendon or the recombinant preparation were detected by using avidin conjugated with alkaline phosphatase diluted 1 : 1000 in 1% BSA in NaCl/Pi, 0.05% (v/v) Tween-20 (200 µL per well), followed by a rinse and by 200 µL of the substrate solution (p-nitrophenyl phosphate at 1 mg·mL−1 in 0.9 m diethanolamine/HCl buffer, 0.5 mm MgCl2, 3 mm NaN3, pH 9.5). The absorbance was measured at 405 nm before and after color development. The binding of decorin core was detected as described above but by using avidin conjugated with horseradish peroxidase: all the steps were performed in a final volume of 100 µL per well; horseradish peroxidase was diluted 1 : 1000 in 2 mg·mL−1 BSA solution, followed by a rinse and by the substrate solution (0.04% o-phenylenediamine dihydrochloride and 0.04% (v/v) hydrogen peroxide in a buffer containing 514 mm disodium hydrogen phosphate, 24.3 mm citric acid, pH 5). Color development was stopped by adding 100 µL of 3 m hydrochloric acid and the absorbance measured at 490–655 nm.

In order to determine the amount of collagen or peptides adsorbed to microtiter wells, 5 µg of each collagen type or equimolecular amounts of peptides were allowed to adsorb overnight, followed by a brief rinse as above. Then, protein was extracted from the wells with two rinses of 200 µL of 6 m HCl and subjected to hydrolysis and Hyp quantitation [38]. The percentage of protein adsorbed to the wells was found to be 15.1% ± 3.0 for CNBr peptides, 9.4% ± 0.6 for type I and II collagen.

Circular dichroism spectroscopy

Solutions of collagens and peptides were prepared by dissolving dry samples in 0.1 m acetic acid at 1–1.5 mg·mL−1. All operations were performed at 4–5 °C. The solutions were equilibrated for ≥ 7 days, with occasional vortexing. After clarification by centrifugation, the concentration was determined by means of a Hyp assay [38]. Aliquots of the acidic solution were freeze-dried and then dissolved at a concentration of 80 µg·mL−1 in 0.1 m acetic acid or in NaCl/Pi containing 1 mm EDTA and 1.5 mm NaN3[30]. These solutions were equilibrated for ≥ 7 days at 4–5 °C, with occasional vortexing. Solutions of decorin or its core were prepared in NaCl/Pi at a concentration of 4 nmol·mL−1. All solutions were clarified by centrifugation immediately before CD analysis. CD spectra were recorded with a cell of 1 mm path length thermostatted at the appropriate temperature. Scans were performed at 20 nm·min−1, collecting data points every 0.05 nm and averaging the data at least over three scans.


Analysis on decorin

Two different decorin preparations have been used: decorin extracted from tendon and a recombinant decorin, as described under Materials and methods. The electrophoretic analysis in denaturing conditions, both before and after chondroitinase ABC digestion, is present in Fig. 1A. On sequencing, tendon decorin showed a unique and correct sequence, DEAxGIGPEE, where x is the dermatan/chondroitin sulfate-bearing serine residue, unrecognized by the sequencer; the recombinant preparation showed a mixture of decorin with and without the propeptide in an about 1 : 1 ratio. CD spectra at 20 °C showed that tendon and recombinant decorin are very similar to each other, differing below 210 nm (Fig. 1B). These spectra are similar to reported spectra of a recombinant decorin purified in the absence of chaotropic agents, with the exception of the wavelength of the minimum (215–216 instead of 218 nm) and very different to the spectrum of the same preparation purified in the presence of guanidine hydrochloride [42]. For each decorin preparation, the spectra at 4–30 °C are superimposable and thermal denaturation occurs at > 40 °C with a small difference between tendon and recombinant decorin (Fig. 1C,D). The protein core of tendon decorin behaved like the whole proteoglycan (data not shown). Due to the small difference found in the literature for the wavelength of the minimum between a recombinant decorin (bearing a polyhistidine tag) in the native state and after denaturation in 10 m urea/renaturation in 1 m urea [43], our CD spectra are empirical findings that do not necessarily demonstrate a native conformation for our decorin preparations.

Figure 1.

Analysis of decorin. (A) SDS/10% PAGE of tendon decorin (lanes 1 and 2) and recombinant decorin (lanes 3–4) we have used in this work, both before (lanes 1,3) and after (lanes 2,4) chondroitinase ABC digestion. About 10 µg and 5 µg were analyzed for decorins and decorin cores, respectively. Left lane: standard protein markers and their molecular masses (in kDa). The core protein is present as two bands with apparent molecular masses of 47 and 42 kDa (arrowheads). (B) CD spectra at 20 °C of tendon and recombinant decorin (continuous and dotted lines, respectively) dissolved in NaCl/Pi at 4 nmol·mL−1. (C,D) CD spectra at 30, 40, 45, 50 °C (identifiable from top to bottom at 205 nm) for tendon (C) or recombinant decorin (D). Spectra at 4–25 °C (not shown) are superimposable with the spectrum at 30 °C. (E) Competition experiments between biotinylated decorins (20 pmol) and increasing amounts of unmodified tendon decorin (data for biotinylated tendon or recombinant decorin challenged with collagen I as the coated ligand are indicted by circles and rectangles, respectively; data for biotinylated tendon decorin with type II collagen are indicated by triangles). Lines are drawn as a visual aid.

The determination of the disaccharide composition of the glycosaminoglycan chain after chondroitinase ABC digestion of tendon decorin showed a high percentage of mono-sulfated species, the 6-sulfated one prevailing: 8% of unsulfated disaccharide, 56 and 31% of 6- and 4-sulfated disaccharides, respectively, 5% of disulfated species. After chondroitinase AC II digestion the composition was found to be 11, 71, 15 and 2%, respectively. By applying the formula of Shirk et al. [44], the percentage of iduronic acid content was found to be 31%.

Biotinylated decorins were used in all subsequent binding experiments with collagenous samples. Control experiments showed that competitive binding to coated type I and II collagens exists between biotinylated decorins and unmodified tendon decorin (Fig. 1E).

Purification, chemical modification and analysis of collagenous samples

Type I collagen and its CNBr peptides were already available to us and well characterized. Pepsin-soluble type II collagen was prepared from bovine nasal septum and its CNBr peptides were purified by a combination of two chromatographic steps. CNBr peptides from collagens type I and type II used in this work are indicated in Fig. 2. The only peptide we have not been able to purify is the C-terminal peptide of the α1(II) chain, namely CB9,7, probably because this peptide is involved in cross-linking.

Figure 2.

CNBr peptides from type I and type II collagen alpha chains. The scheme shows the names (in bold), position along the triple helical domain, size (number of residues) and melting temperature of the trimeric species of CNBr peptides. The bottom two lines indicate the N→C direction with a length scale (in residues) and the banding pattern of type I collagen fibrils [51]. Melting temperatures have been measured in NaCl/Pi containing 1 mm EDTA and 1.5 mm NaN3 (in 0.1 m acetic acid for α2(I) CB3,5 because of its low solubility in NaCl/Pi); values for type I collagen peptides are data reported previously [27, 30]. We have determined the ability to bind decorin for all peptides reported in the scheme (positive ones are marked with an asterisk) and also for the composite peptide α1(I) CB2,4 (Tm ≈ 28° in 0.1 m acetic acid), whereas we could not use peptide α1(II) CB9,7.

Chemical modification of collagens and several of their peptides was performed by derivatizing the primary amino group of Lys and Hyl side chains: methylation with HCHO/NaBH3CN that preserves the positive charge, and acetylation, either with acetic anhydride or SNHSAc, that eliminates the positive charge.

Chemical modification of Lys/Hyl side chains causes a slower electrophoretic migration of the collagenous samples (Fig. 3A). N-Acetylated samples also have a low affinity for Coomassie Brilliant Blue R 250, the standard anionic dye we used to stain polyacrylamide gels. It should be noted that N-acetylation with acetic anhydride is to be avoided because it is artifactual: some peptide bonds are broken with the formation of interchain covalent bonds leading to molecular species larger than the original sample. This is particularly evident for peptides (Fig. 3A), and also smaller molecular species, as shown by analytical gel filtration chromatography in denaturing conditions (data not shown). All this is probably the consequence of the addition of concentrated sodium hydroxide during the derivatization in order to maintain the pH constant.

Figure 3.

Analysis of collagen samples. Representative analyses for type II collagen (left column) and two CNBr peptides (central and right columns) are reported. Lane 1 indicates underivatized samples; 2, samples derivatized with HCHO/NaBH3CN; 3, with SNHSAc; 4, with acetic anhydride. (A) SDS/PAGE pattern (6% acrylamide for type II collagen; 15% for peptides). The standard anionic dye Coomassie Brilliant Blue R250 showed a low affinity for the acetylated samples whose band intensity quickly faded during destaning. The figures reported were obtained during the very early destaining steps. (B) CD spectra at 30 °C for type II collagen and at 20 °C for the two peptides. All samples were dissolved at 80 µg·mL−1 in NaCl/Pi containing 1 mm EDTA and 1.5 mm NaN3. The figures report only the portion of the spectrum centered on the maximum of the positive peak (≈ 221 nm); this positive signal is present only for collagenous samples with triple helical conformation.

Using CD spectroscopy at increasing temperatures, we have determined that many CNBr peptides are able to form trimeric species that at room temperature prevail over the random-coil monomeric species; only some small CNBr peptide trimers have low melting temperatures (see Fig. 2 for the values of melting temperatures).

Chemical modification of Lys/Hyl side chains in collagenous samples do not significantly modify both the triple helical conformation of the trimeric species (Fig. 3B) and the thermal stability, with the relevant exception of N-acetylation with acetic anhydride for the reasons mentioned above. The greatest decrease of Tm on derivatization in mild conditions was found to be less than 3 °C. A detailed thermodynamic analysis of the melting transition of modified peptide trimers will be described elsewhere.

Binding of decorin to collagenous samples and effect of chemical modifications

Equimolecular amounts of collagen type I and type II and their CNBr peptides have been used in a solid-phase assay, challenged with a constant amount of biotinylated decorin, either from tendon (intact or the protein core) or the recombinant preparation. At 23 °C, both collagen types bind decorin, as well as some CNBr peptides (asterisked in Fig. 2), namely peptides CB8, CB7 and CB6 from the α1(I) chain, CB4 from α2(I) and only peptide CB11 from α1(II). The different decorins show the same binding pattern towards the CNBr peptides, with only some differences in the intensity for some of the peptides (Fig. 4).

Figure 4.

Binding of biotinylated decorins to collagenous samples. A constant amount of type I or II collagen (5 µg) or equimolecular amounts of their CNBr peptides were used to coat polystyrene wells. A␣constant amount of biotinylated decorin was added (20 pmol); the bound decorin was determined using avidin conjugated with alkaline phosphatase or, for tendon protein core, horseradish peroxidase. The absorbance plotted in the panels for all collagens and peptides we have tested was determined by exploiting a colorimetric reaction catalyzed by the enzyme. The absorbance is the mean of analyses performed at least in triplicate; the highest standard deviation for samples able to bind decorin was 17% of the mean. Top: analysis with tendon decorin on collagen samples in native and in denaturing conditions (white and black columns, respectively). The right panel reports the binding of tendon decorin to BSA, fibronectin and a 30-kDa heparin-binding fragment of fibronectin. Bottom: analysis on collagen samples with tendon decorin core (dark gray) and recombinant decorin (light gray). (n.d. not determined.)

The triple helical conformation of collagenous samples is a necessary requisite for the interaction with decorin, because heat denaturation eliminates their binding (Fig. 4). No other peptide showed any binding also when the assay was performed at 4 °C (see Tm of peptides in Fig. 2 with respect to the temperature of the binding experiments).

It is worth noting that peptide CB10 from type II collagen does not bind decorin, regardless of the fact that it is homologous to and in the homologous region of CB7. We cannot comment on α1(II) CB9,7, because we did not find it in the chromatographic purifications of our CNBr digest of type II collagen. Peptide α2(I) CB3,5 has some binding ability but the data should be judged with caution because this peptide showed a positive CD signal at 221 nm that is typical of native collagen and trimeric peptides but it is possible that it does not form trimers with the three α chains in register[28].

As controls, we have tested the interaction of decorin with other proteins: BSA, as a negative control, showed a much lower response than collagens and positive peptides, whereas fibronectin and a 30-kDa fibronectin fragment having heparin-binding ability showed interaction with tendon decorin (Fig. 4). Fibronectin is known to interact with decorin protein core [45].

Our data suggest that decorin interacts with multiple regions of collagen. In competition experiments, we have found that CNBr peptides in solution are not able to compete with type I or type II collagen for decorin. When increasing amounts of peptide α1(I) CB7 or α1(II) CB11 (up to 50-fold excess with respect to the collagen amount) were preincubated in solution with decorin (at room temperature for 1 h) we observed no variation in binding of decorin to microwells coated with collagen type I or type II, respectively. The same null result was obtained in a competition experiment between CB11 as the coated ligand and the same peptide in solution with decorin. The reason for this is probably the interaction of collagen or peptide in solution with the coated collagenous molecules [46]. It is also possible that isolated collagen trimers in solution have no or much lower affinity for decorin and that decorin binding to collagen depends on the aggregation status of collagen itself.

The affinity between decorin and collagens and peptides was determined by using constant equimolecular amounts of the collagenous samples with increasing amounts of tendon decorin (Fig. 5). The graphs in Fig. 5C,D indicate a bimodal behaviour of decorin for collagen I and II, suggesting that decorin has two distinct binding sites for these collagens, as already indicated by others [20–22]. Scatchard-type plots, drawn according to Hedbom & Heinegard [47], allowed the calculation of the dissociation constants reported in Table 1. Because our data for collagens I and II did not allow us to obtain meaningful values for both binding sites, we performed linear interpolation on all the data points (Fig. 5C,D) obtaining a single dissociation constant that is only indicative of the range. The values of Kd are in the nanomolar range and similar to the values reported in literature for decorin from cartilage or tendon, using type I collagen as the ligand (30 and 16 nm) [47,48].

Figure 5.

Affinity of collagenous samples with decorin. Increasing amounts of biotinylated tendon decorin were added to polystyrene wells coated with a constant amount of collagen (5 µg) or equimolecular amounts of CNBr peptides. The binding was determined by using avidin conjugated with alkaline phosphatase. (A,B) Saturation curves of two collagens and two peptides reported as examples. Each data point is the average value of a determination performed at least in triplicate. The highest standard deviation was 18% of the mean. Lines are added as a visual help. (C,D) Scatchard-type plots [47] on the same samples. Lines interpolating the data have been computed with the least square method. For type I and II collagens, linear interpolation was performed taking into account all data points (see text). The resulting dissociation constants are reported in Table 1.

Table 1.  Dissociation constants of the complexes between biotinylated tendon decorin and collagenous samples.
Collagen sample K d (nm)
  1. a The value reported was obtained from the linear interpolation of all data points ( Fig. 5C,D), because it was impossible from our data to calculate meaningful values for two binding sites. bIt was impossible to calculate the dissociation constant for this peptide because a saturation level was not clearly identifiable.

Type I collagen41a
CB6 from α1(I) b
CB7 from α1(I)13
CB8 from α1(I)44
CB4 from α2(I)16
Type II collagen42a
CB11 from α1(II)22

Other experiments (not shown) indicated that ionic interactions play an important role in the binding between decorin and collagen. Whereas the presence of 50 mm NaCl in the phosphate buffer improved the interaction with respect to analysis performed in NaCl/Pi (150 mm NaCl), a higher concentration of salt (250 mm) resulted in dramatically reduced binding. On the contrary, no influence of detergents was found, as determined by the addition of 1% Triton to the binding solution.

To further characterize the nature of the interaction between decorin and collagen, we have chemically modified collagen samples using agents that either disrupt or maintain the positive charge, e.g. acetylation and methylation, respectively.

Our results indicate that elimination of the positive charge of the side chains of Lys/Hyl residues disrupts the interaction with decorin (Fig. 6). This does not depend on the derivatizing agent, SNHSAc or acetic anhydride, indicating that the side-effects of the treatment with acetic anhydride described above are not responsible for the loss of binding. Methylation of Lys/Hyl residues by treatment with HCHO/NaBH3CN preserved the positive charge and this resulted in a more complex effect on binding to decorin (Fig. 6). Whereas two peptides, α1(I) CB8 and α1(II) CB11, showed an increased binding, methylation of the C-terminal half of the α1(I) resulted in either reduced binding for α1(I) CB7, or a complete loss of the binding for α1(I) CB6. The variation of the binding ability for N-methylated samples with respect to the unmodified ones is not related to the percentage of Lys/Hyl side chains that did not react with the derivatizing agent (the percentage ranged from 3 to 12%).

Figure 6.

Effect of chemical modifications. A constant amount of type I or II collagen (5 µg) or equimolecular amounts of their CNBr peptides were used to coat polystyrene wells. A constant amount of biotinylated tendon decorin was added (20 pmol); the bound decorin was determined using avidin conjugated with alkaline phosphatase. The absorbance is the average value of at least three determinations; the highest standard deviation for samples able to bind decorin was 17% of the mean. For each collagenous sample used in native conditions, the results of the underivatized sample (white column) and for derivatives with SNHSAc (black) and HCHO/NaBH3CN (gray) are reported. The results obtained with samples treated with acetic anhydride (not shown) overlap those with SNHSAc.

Taken together, these data demonstrate the essential role of the positive charge of collagen Lys/Hyl residues for interaction with decorin.


The binding of decorin with fibrillar collagens has been extensively investigated (reviewed in [1,2]), but in vitro studies have not yet conclusively identified the collagen domains responsible for the specific association with decorin. In this study, we have analyzed the binding between decorin and CNBr peptides from type I and type II collagens, both unmodified and chemically derivatized.

We have recently characterized CNBr peptides from collagen type I [27–30]. The present work indicates that CNBr peptides from type II collagen have a very similar behaviour.

Our data on the interactions between type I and type II collagens, their peptides and decorin reveal the following.

(a) Type I and probably also type II collagen appear to have multiple binding sites for decorin, because several CNBr peptides are able to interact with this small proteoglycan.

(b) The side chain of Lys/Hyl residues in collagen is relevant for the binding, because the elimination of their positive charge eliminates the interaction. On the contrary, the chemical modification preserving the ionic character modulates the binding to decorin. This leads to a differential behaviour for the different peptides.

(c) Decorin might have two binding sites for collagen, as suggested by others and by the differential behaviour of collagen peptides.

Decorin is able to bind several CNBr peptides and type I and II collagens only when they are in triple helical conformation. Among the binding peptides, CB8, CB4 and CB11 are found in a homologous region of the N-terminal half of the respective α chains (residues 124–327 of the triple helical domain). On the contrary, CB7 and CB6 lie in the C-terminal half of the chain. Binding specificity is demonstrated by the following.

(a) The absence of interaction with decorin(s) of some peptides that are in triple helical conformation in our assay conditions [CB2, CB2,4 and CB3 from α1(I), CB12, CB8 and CB10 from α1(II)].

(b) All peptides able to bind decorin contain a region corresponding to the d and e bands of collagen fibrils (Fig. 2). This is in accordance with morphological findings showing that chondroitin/dermatan sulfate PGs, such as decorin, localize in these bands, whereas keratan sulfate PG are present at a and c bands [23,24]. However, not all collagen peptides that contain regions of the collagen molecule falling within the d band interact with decorin, e.g. the homologous peptides α1(I) CB3 and α1(II) CB8, or peptide α1(II) CB10. Collagen binding to decorin does not therefore depend on the clusters of charged residues responsible of the banding pattern but on specific sequences that contain ionic residues.

(c) The action on platelet adhesion and activation by only two peptides from type II collagen (data not shown) and only by peptide α1(I) CB3 from type I collagen, as already known from the literature [49].

Peptide CB10 from type II collagen is homologous to α1(I) CB7, but does not interact with decorin. One possible explanation of this discrepancy could lie in the fact that type␣II collagen is more glycosylated than type I collagen. It seems to us probable that glycosylation of hydroxylysine will block the binding. However, the glycosylation pattern of Hyl residues is known for CB7 [50] but not for CB10. Aliquots of both CB10 and CB7 have also been digested at 37 °C for 18 h with endoproteinase Arg-C, according to the manufacturer's guidelines, with an enzyme to substrate ratio of 1 : 130. None of the most abundant fragments, separated by reverse-phase HPLC with the same protocol used to separe small CNBr peptides, showed at 4 °C any binding ability to tendon decorin (data not shown). This suggests that also some Arg-containing sequences are relevant in collagen for its interaction with decorin, or that none of the fragment was present in our assay conditions as a trimeric species, or that the minor enzyme activity cleaving Lys peptide bonds had a relevant effect.

The affinity of the binding peptides for decorin is in the nanomolar range with the same magnitude reported by others for type I collagen [47,48], and the dissociation constants are within one order of magnitude (Table 1). Our determinations showed also that the binding between decorin and collagens or their CNBr peptides is quite sensitive to the ionic strength of the buffer, suggesting an ionic character of the binding.

The main decorin region implicated in the binding to collagens was hypothesized to lie inside the concave area of the arch-shaped protein core [17]. Residues in LRR 4 and 5 were considered responsible for the binding [21]. The concave surface, formed by β strands, is lined by many charged residues and several hydrophobic side chains. Charged residues probably make ionic contacts; in particular, carboxylate ions might bridge two positive residues, and/or Lys ammonium ions or Arg guanidinium ions might bridge two negative groups. One of the relevant residues is glutamate-180 found by Kresse and coworkers to be relevant for the collagen binding [22]. It should however, be noted that the constructs lacking LRR 5 or bearing the substitution Glu180 to Lys [22] bring several positive charges close to each other. This might have a direct influence on the conformation of decorin core, and only an indirect one on the collagen binding. However, this remains a hypothesis, as, to our knowledge, no conformational analysis was reported on these constructs.

The presence in the decorin molecule of a second binding site for collagen was suggested previously [20–22]. The results we have obtained from the Scatchard-type plots for type I and type II collagens (Fig. 5) and the different behavior of the N-terminal collagen peptides with respect to CB7 and CB6 might be a further support to this hypothesis.

Chemical modification of collagens and their CNBr peptides demonstrated that acetylation eliminates their binding to decorin. Lys/Hyl side chains are therefore present at, or very near to the binding site(s) and their positive charge is a stringent requisite for the binding. This is not surprising, if indeed collagen binds inside the concave surface of decorin, owing to the presence of an elevated number of ionic residues. On the contrary, reductive methylation modulates the binding of all peptides to decorin, the largest decrease being shown by α1(I) CB7 and CB6 (Fig. 6), suggesting a different specificity of these peptides. All these effects are direct, because all peptides we have derivatized in mild conditions maintain the ability to form trimeric species that are the major species in our binding assays.

A previous study reported that decorin binding occurs preferentially at about 50 and 100 nm from the N-terminus of type I collagen [25]. We found the region ≈ 50 nm from the N-terminus falls within peptides CB8, CB4 and CB11 and was able to bind decorin. Apart from the presence of Lys, we are not able to compare our data with the suggestion of a collagen sequence able to bind decorin, namely GAKGDRGET, at position 853–861 of the triple helical domain of the α1(I) chain, within peptide CB7 [26]. A similar sequence is present in the homologous region of type␣II and III collagens. Without GAK, the sequence G-D/E-R-G-E-Hyp/T is present also at position 623–628 of the same chain (in peptide CB7) and of homologous sequences of other collagen alpha chains. The collagen sequence DRGE might have KLER and RELH as possible complementary sequences in decorin [18], at position 130–133 and 272–275, in the LRR 3 and 9, respectively. The model proposed on the basis of these complementary sequences in the two interacting proteins showed a double contact between decorin and two collagen molecules. However, this is discordant with the decorin model [3,4,17] where the ionic residues of KLER/RELH fall inside the concave surface of decorin, with the exception of K-130.

It is not possible to reconcile our findings with most results recently reported by Keene et al. [19] which are in disagreement with many previous results, as widely discussed in the paper. On one side, a periodicity was noticed by these authors in aggregates of decorin and type I pC-collagen seen in electron micrographs of rotary shadowed molecules; this was due to the presence of decorin, as pC-collagen alone did not show a similar pattern. CNBr peptides of the α1(I) chain that we have found to bind decorin are positioned along the chain in a manner that periodicity of binding is the natural outcome, even if our data do not allow a determination of the size of the period and even if peptide CB3, unable to bind decorin, interrupts the periodicity. On the other side, the relevance of Lys/Hyl residues both in collagens and peptides for interaction with decorin is in contrast with the findings that the binding site for decorin is located in a sequence within the peptide α1(I) CB6 devoid of any Lys/Hyl residue and containing, as ionic amino acids, only one Glu and one Arg, 13 residues apart. It is interesting to note that the same region of the α2(I) chain contains the dipeptide HH. The triplet GHH is unique in the triple helical domain of all collagen chains, as determined by a search in Swiss-Prot. One can thus hypothesize that the polyhistidine tag present in the recombinant decorin preparation used by Keene et al. [19] is able, in the presence of minute amounts of proper cations, to interact with GHH in α2(I) and direct the binding of decorin to the collagen C-terminus in CB6. However, this cannot be deduced because no control experiments are reported with decorin lacking the polyhistidine tag or with decorin purified in the presence of chaotropic agents to compare with conditions used in previous determinations.

On this basis, we can conclude the precise location and the relative orientation of the binding sites in decorin and collagen are not yet known. Our findings on multiple binding sites in collagen and on the relevance of Lys/Hyl residues set some limitations, as do the fact that decorin might have a second binding site for collagen. Because decorin physiologically interacts with collagens when they are in their specific aggregation states, multiple contacts are probably essential for the strength and the specificity of the interaction.


We thank Antonella Forlino for helpful suggestion and criticism, Elena Campari and Luigi Corazza for technical assistance, ‘Centro Grandi Strumenti’, University of Pavia, for peptide sequencing and free access to the spectropolarimeter. This work was supported by grants from Italian MURST (grant MM05148132-3) and University of Pavia (FAR and Progetto Giovani Ricercatori 2000/2001).