Sulfated hyaluronan and chondroitin sulfate derivatives interact differently with human transforming growth factor-β1 (TGF-β1)

Acta Biomaterialia 8 (2012) 2144–2152

Contents lists available at SciVerse ScienceDirect

Acta Biomaterialia journal homepage: www.elsevier.com/locate/actabiomat

Sulfated hyaluronan and chondroitin sulfate derivatives interact differently with human transforming growth factor-b1 (TGF-b1) V. Hintze a,⇑, A. Miron a, S. Moeller b, M. Schnabelrauch b, H.-P. Wiesmann a, H. Worch a, D. Scharnweber a a b

Institute of Materials Science, Max Bergmann Center of Biomaterials, Technische Universität Dresden, Budapester Str. 27, 01069 Dresden, Germany Biomaterials Department, INNOVENT e.V., Prüssing Str. 27 B, 07745 Jena, Germany

a r t i c l e

i n f o

Article history: Received 28 October 2011 Received in revised form 7 February 2012 Accepted 8 March 2012 Available online 13 March 2012 Keywords: Transforming growth factor-b1 (TGF-b1) Chondroitin sulfate derivatives Hyaluronic acid/hyaluronan derivatives Surface plasmon resonance ELISA

a b s t r a c t This study demonstrates that the modification of hyaluronan (hyaluronic acid; Hya) and chondroitin sulfate (CS) with sulfate groups leads to different binding affinities for recombinant human transforming growth factor-b1 (TGF-b1) for comparable average degrees of sulfation (DS). In general, Hya derivates showed higher binding strength than CS derivatives. In either case, a higher degree of sulfation leads to a stronger interaction. The high-sulfated hyaluronan sHya3 (average DS  3) exhibited the tightest interaction with TGF-b1, as determined by surface plasmon resonance and enzyme-linked immunosorbent assay. The binding strength was significantly weakened by carboxymethylation. Unmodified Hya and low-sulfated, native CS showed weak or no binding affinity. The interaction characteristics of the different sulfated glycosaminoglycans are promising for incorporation into bioengineered coatings of biomaterials to modulate growth factor binding in medical applications. Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction Transforming growth factor-b1 (TGF-b1) is the founding member of a superfamily of secreted polypeptides [1]. It includes three different forms of TGF-b, the bone morphogenetic proteins (BMP), the Nodals, the Activins, the anti-Müllerian hormone and other structurally related factors in vertebrates, insects and nematodes [1,2]. They are produced by diverse cell types and reported to regulate cell migration, adhesion, multiplication, differentiation and death [1]. The predominant serum and plasma form is an inactive complex with a-macroglobulin [3,4]. High concentrations of TGF-b1 can be found in platelets [5] and bone [6]. In fact, TGF-b1 is one of the most abundant cytokines in the bone matrix (200 lg kg1) [7]. TGF-b1 is a disulfide-bonded dimeric protein, synthesized and secreted as a large precursor molecule [8,9]. It is enzymatically cleaved into active TGF-b1 and latency-associated protein (LAP). LAP remains non-covalently linked to active TGF-b1, masking the receptor-binding domain and keeping it inactive. Osteoblasts are unique in producing two latent forms of TGF-b1: a 100 kDa precursor latent complex and another one that also contains a 190 kDa binding protein [10,11]. TGF-b1 is thus deposited in the bone matrix as an inactive, latent complex [7]. Active TGF-b1 is released during bone resorption, possibly owing to the acidic osteoclastic microenvironment [12] or secreted proteases degrading LAP [13]. ⇑ Corresponding author. Tel.: +49 463 39 39389; fax: +49 463 39 39401. E-mail address: [email protected] (V. Hintze).

It then coordinates bone formation by inducing migration of bone mesenchymal stem cells (MSC) [7] as well as recruitment and proliferation of osteoblasts [14–16]. However, the effect of TGF-b1 on bone cell replication is biphasic and depends on both the growth factor concentration and cell density in monolayer culture [14]. Therefore, TGF-b1 functions to couple bone resorption and formation [7] and is a powerful stimulator of bone formation in vivo [11,17]. In MSC and osteoblasts, TGF-b1 is also very important in matrix formation. It stimulates the synthesis of proteoglycans [18] and is the most prominent inducer of procollagen and fibronectin expression [14,19,20]. In bone ECM, the concentration of free TGF-b1 is regulated by its interaction with the proteoglycans biglycan and decorin [21,22]. Via binding, decorin acts as a negative regulator of the growth factor [22]. In biglycan/decorin-deficient mice, the lack of these binding partners resulted in excess of free TGF-b1, causing accelerated apoptosis of bone MSC, decreased numbers of mature osteoblasts and subsequently reduced bone formation [23]. The active 26 kDa dimer is known to also interact with other extracellular matrix components such as fibronectin [24], thrombospondin [25] and type IV collagen [26]. In contrast to decorin, thrombospondin and type IV collagen were shown to maintain the activity of TGF-b1 [25–27]. Furthermore, human TGF-b1 and TGF-b2, but not TGF-b3, bind to heparin and high-sulfated heperan sulfate [28,29]. They potentiate the activity of TGF-b1 via protection from proteolytic degradation [30] and dissociation of the a-macroglobulin/TGF-b1 inactive complex, while having no effect on the activities of TGF-b2 and

1742-7061/$ - see front matter Ó 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.actbio.2012.03.021

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

TGF-b3 [28,29,31]. Lyon et al. postulate that the basic amino acid residues at position 26 of each monomer are a vital binding determinant and propose a binding model in which GAG binding occurs at two distinct sites of the TGF-b dimer [28]. The position of this residue is within the consensus heparin-binding region described by McCaffrey et al. [29], and a peptide of this region was able to block the binding of TGF-b1 to heparin. Some further studies provide proof of modulatory effects of GAG on TGF-b1 activity in vitro: TGF-b1-induced signalling events in human articular chondrocytes are influenced by exogenous chondroitin sulphate (CS) [32], as well as osteoblast proliferation by the addition of an osteoblast-derived GAG mixture [33]. When combined with TGF-b1, cell-surface-derived GAG appeared to potentiate the already growth inhibitory effects of TGF-b1 at higher concentrations in cell culture [14,33]. In another study, the growth factor lost its capacity to inhibit the proliferation of thymocytes after contact with high molecular weight Hya, but not with sulfated GAG. The authors show that TGF-b1 is protected from tryptic degradation within the complex, and therefore suggest an interaction between Hya and TGF-b1 [34]. These results emphasize the importance of GAG, as part of extracellular matrices, in regulating growth factor signalling and activity. Hintze et al. [35] reported that sulfation of hyaluronan influences the interaction with human bone morphogenetic protein-4 (BMP-4). It was demonstrated that high-sulfated Hya (sHya3) exhibited the tightest interaction with BMP-4 followed by the low-sulfated sHya1. In contrast, unmodified Hya and natural CS showed significantly less binding affinity. The present study continues this approach to investigate GAG derivatives with varying types and content of anionic groups and different sugar backbones for their growth factor binding capabilities. Thereby, the authors aim to create new forms of GAG with defined growth factor interaction properties for biomedical applications. The interaction of recombinant human TGF-b1 with Hya and CS of high, middle and low degree of sulfation (DS 3, 2 or 1) was investigated in comparison with natural CS and unmodified Hya, using immunological and biophysical methods, i.e., enzyme-linked immunosorbent assay (ELISA), BioLISA and surface plasmon resonance (SPR). 2. Materials and methods 2.1. Materials Hya (from Streptococcus, MW = 1.1  106 g mol1) was obtained from Aqua Biochem (Dessau, Germany), sulfur trioxide/dimethylformamide complex (SO3–DMF, purum, P97%, active SO3 P 48%), sulfur trioxide/pyridine complex (SO3–pyridine, pract.; P45% SO3) and monochloroacetic acid (MCA, puriss., P99% T) were purchased from Fluka Chemie, (Buchs, Switzerland). Chondroitin sulphate (CS) from porcine trachea (a mixture of 70% chondroitin-4-sulfate and 30% chondroitin-6-sulfate) was purchased from Kraeber (Ellerbek, Germany). Recombinant human TGF-b1 as well as monoclonal mouse anti-human TGF-b1 (MAB240), biotinylated polyclonal chicken anti-human TGF-b1 (BAF240) antibodies and recombinant human TGF-b receptor RII (TGFR; 341-BR/CF) were obtained from R&D Systems (Wiesbaden-Nordenstadt, Germany). The Series S Sensor Chip C1, the Amine Coupling Kit and HBS-EP (10) were purchased from GE Healthcare Europe GmbH (Freiburg, Germany). Methyl vinyl ether/maleic acid copolymer and adipic acid dihydrazide were purchased from Fisher Scientific (Nidderau, Germany). Heparin from porcine intestinal mucosa, sodium cyanoborohydride, bovine serum albumin (BSA), Tween 20, sucrose, 4-nitrophenylphosphate disodium salt hexahydrate and 3,30 ,5,50 -tetramethylbenzidine liquid substrate were obtained from Sigma–Aldrich (Schnelldorf, Germany).

2145

2.2. Preparation of Hya and CS derivatives The low molecular weight hyaluronan (Hya-LMW), the lowsulfated (sHya1) and the high-sulfated Hya derivatives (sHya3-21, sHya3-50) were synthesized and characterized as described by Kunze et al. and Hintze et al. [35,36]. 2.2.1. Medium-sulfated Hya (sHya2-26, sHya2-14) The procedure for low-sulfated Hya was used as described [35] with the modification that the SO3–pyridine complex was added to a suspension of Hya (tetrabutylammonium form) in DMF in a ratio of polymer:SO3 = 1:8, followed by stirring the reaction mixture for 20 min at room temperature. Yield: 75% (related to Hya, sodium salt). 13C-NMR of sHya2-26 (D2O, d): 178.30, 175.07 (CH3–CO–, – COO), 103.64–99.99 (C-10 , C-1), 83.34–72.99 (C-3, C-40 , C-50 , C-5, C-30 , C-20 , C-4 (sulfated), 68.76 (C-4, non-sulfated), 67.62 (C-6, sulfated), 54.76 (C-2), 23.35 (CH3–CO–). 2.2.2. Carboxymethyl hyaluronan sulfate (scmHya3) At room temperature, 4.98 mmol Hya (sodium salt) was suspended in isopropanol/water (200/100 v/v). Then 230 mmol NaOH (40% in water) and 80 mmol MCA were added, and the solution was heated to 60 °C and stirred for 4 h. After cooling to room temperature, the phases were separated. The lower phase was diluted with methanol, and the precipitated product was dissolved in water. The solution was dialysed against water, followed by lyophilization of the aqueous solution and drying of the resulting polymer under vacuum. Yield: 65%, degree of carboxymethylation (DCM) = 0.4. 13C-NMR (D2O, d): 178.26, 175.83, 174.79, 174.39 (CH3–CO–, – COO), 103.66, 103.18 (C-10 ), 101.30, 101.01 (C-1), 83.53, 82.91 (C-3), 81.13, 80.68 (C-40 ), 77.20–73.29 (C-50 , C-5, C-30 , C-20 ), 72.16–71.78 (CH2), 71.26 (CH2), 70.41 (C-6, carboxymethylated), 69.44 (C-4), 61.50 (C-6, free), 55.01 (C-2), 23.21 (CH3–CO–). The carboxymethylated Hya was transformed into the tetrabutylammonium form and sulfated with SO3–DMF using the procedure as described for sHya3 [35]. Yield: 85%, DS = 3.1. 13C-NMR (D2O, d): 178.30, 175.07 (CH3–CO–, –COO), 101.95 (C-10 ), 100.64 (C-1), 78.67,-73.75- (C-3, C-40 , C-50 , C-5, C-30 , C-20 , C-4), 71.11 (–CH2– COO–), 70.07 (C-6, carboxymethylated), 68.32, (C-6, sulfated), 56.01 (C-2), 23.55 (CH3–CO–). 2.2.3. High-sulfated CS (sCS3) The sulfation was performed as reported by Smissen et al. [37]. Briefly, a solution of CS (sodium salt) dissolved in distilled water was stirred with Dowex WX 8 ion exchanger to form the tetrabutylammonium form. Sulfation was performed with the SO3–DMF complex (polymer:SO3 ratio = 1:20) at room temperature and a reaction time of 1 h. The sulfated product was isolated from the reaction mixture by precipitation into acetone and neutralization using ethanolic NaOH solution. The formed Na salt of the sulfated CS was washed with acetone and purified by dialysis against distilled water followed by lyophilization of the aqueous solution and drying of the resulting polymer under vacuum. Yield: 80% (related to CS). 13C-NMR of sCS3 (D2O, d): 175.19 (CH3–CO–, –COO), 102.38, 101.85 (C-10 , C-1), 79.82–72.34 (C-40 , C-3, C-50 , C-4 (C4S), C-5 (C4S), C-30 , C-5 (C6S), C-20 , C-4 (C6S), 66.87 (C-6, sulfated), 51.59 (C-2), 23.17 (CH3–CO–). 2.3. Polymer characterization Nuclear magnetic resonance (NMR) spectra were recorded in D2O (99.9%; Sigma–Aldrich), with a Bruker Advance 300 MHz spectrometer at a temperature of 25 °C. D2O at 4.75 ppm was used as a reference line. The KBr technique was employed for recording FTIR spectra with a FT-IR-Spektrometer FTS 175 (Bio-Rad, Krefeld,

2146

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

Germany). The DS was determined by estimation of the sulfur content using an automatic elemental analyser (CHNS-932, Leco, Mönchengladbach, Germany). Conventional acid–base titration was used to determine the DCM value. Molecular weight determination was performed by gel permeation chromatography (GPC) using a Jasco PU 980 pump, and a combination of three Suprema-Gel columns with dimensions 8  300 mm (diameter  length) and the specifications 10 lm–100 Å, 10 lm–1000 Å and 20 lm–30,000 Å referring to grain size and pore size, respectively. The eluent was phosphate buffered saline (PBS) buffer, and the flow rate for all columns was 0.8 ml min1. A double detection system consisting of a Postnova Analytics PN 3000 (15°) laser light scattering (LLS) detector and a Jasco RID-1531 refraction (RI) detector was used. Absolute values of number-average (Mn) and weight-average (Mw) molecular weights were determined using the LLS detection system. Calculation of polydispersity (PD = Mw/Mn) was performed on the basis of Mn and Mw values obtained from RI detection. In the latter case, calibration of the detection system was performed with commercially available pullulan standards (PSS Polymer Standards Service GmbH, Mainz, Germany). 2.4. Covalent coupling of GAG to MaxiSorp™ 96-well ELISA plates The polysaccharides were immobilized in each well of MaxiSorp™ 96-well microtitre plates from Thermo Fisher Scientific (Schwerte, Germany) via their reducing ends as described by Hintze et al. [35]. In brief, 55 lg Hya or the same molar concentrations in relation to the molecular weight of the disaccharide units for the other GAG were solubilized in 25 mM citric acid supplemented phosphate buffer pH 5.0 and incubated overnight with 96-well plates pre-treated for 30 min with methyl vinyl ether/maleic acid copolymer in DMSO, followed by 2.5 h with adipic acid dihydrazide. The resulting Schiff’s bases were reduced to stable alkylamine bonds covalently linking the GAG to the dish by incubating them for 30 min with 1% sodium cyanoborohydride in methanol. The prepared surfaces were then washed with 10 mM Tris/HCl, 50 mM sodium chloride, pH 7.4 (TBS) and blocked for 2 h with 2% BSA in TBS. For non-specific binding, untreated wells were incubated with 2% BSA in TBS, and untreated wells with no additions were used as positive control for growth factor binding. Then, 0, 200 and 400 ng ml1 (0, 10 and 20 ng) of TGF-b1 in PBS with the addition of 1% BSA was incubated with the prepared surfaces for 16 h at 4 °C. 2.5. ELISA and BioLISA The amount of bound growth factor was determined indirectly with specific antibodies (ELISA) or the TGF-b1 specific receptor (TGFR; BioLISA) in the solution containing non-bound TGF-b1 recovered from the supernatant after incubation with immobilized GAG on the MaxiSorp™ 96-well microtitre plates. The amount of desorbed growth factor over 4 days in PBS/1% BSA at 37 °C was determined similarly. A calibration curve ranging from 0 ng ml1 to 25 ng ml1 (0–1.5 ng) in 1% BSA/PBS was prepared for TGF-b1 to calculate the amount of non-bound growth factors. All binding data are the average of triplicates statistically analysed by twoway analysis of variance (ANOVA) with means comparison using Tukey test. The data were verified by a second independent experiment. In brief, 10 and 20 ng of TGF-b1 were incubated for 16 h with immobilized GAG on MaxiSorp™ 96-well microtitre plates at 4 °C for binding to reach equilibrium. The supernatants were removed, diluted with 1% BSA/PBS and transferred to another MaxiSorp™ 96-well microtitre plate pre-coated with growth factor specific capture antibodies or TGFR. The surfaces were then blocked with 1% (w/v) BSA, 5% sucrose (w/v), 0.05% (v/v) Tween 20 in PBS.

After washing with 0.05% (v/v) Tween 20 in PBS, wells were incubated with biotinylated growth factor specific antibodies diluted in PBS for 2 h. Following a washing step, streptavidin-horseradish peroxidase diluted in 1% BSA/PBS was added to the wells for 20 min. The plates were developed with 3,30 ,5,50 -tetramethylbenzidine liquid substrate. The reaction was stopped with 1 M H2SO4 and measured at A450 nm using a microtitre plate reader. 2.6. Immobilization of TGF-b1 on Series S Sensor Chip C1™ For interaction analysis of growth factor and GAG a Biacore™ T100 instrument (GE Healthcare) was used. TGF-b1 was immobilized on the surface of a Series S Sensor Chip C1 (a dextran-free planar surface with carboxymethyl group, to reduce multivalent binding events) using the amine coupling reaction as described by the manufacturer. Briefly, prior to immobilization of TGF-b1 on the C1 chip surfaces, it was conditioned with two 60 s pulses of 0.1 M Glycin–NaOH, pH 12, with 0.3% Triton-X-100 at a flow rate of 10 lL min1. The carboxyl groups were then activated using an injection pulse (7 min, 70 lL) containing a mixture of 1-ethyl-3(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS). Protein immobilization was accomplished by injecting 43 lL of TGF-b1 with a concentration of 50 lg mL1 diluted in 10 mM sodium acetate buffer at pH 5.5 at 5 lL min1. The remaining non-reacted sites on the sensor surface were blocked with a 70 lL injection of 1 M ethanolamine–HCl, pH 8.5. In this way, an average of 435 RU of TGF-b1 could be immobilized on the chip surface. One flow cell was used as a reference surface and was directly deactivated after the injection pulse of EDC/ NHS by injecting 1 M ethanolamine–HCl without the immobilization of TGF-b1. 2.7. SPR measurements of growth factor interaction with GAG The running buffer for all experiments was HBS-EP (0.01 M Hepes (pH 7.4), 0.15 M sodium chloride, 3 mM EDTA, 0.05% surfactant P20), and the interactions were investigated at 37 °C. Each GAG sample was diluted in HBS-EP buffer, and some selected samples also in HBS-EP buffer containing additional 0.05 M or 0.25 M sodium chloride (total sodium chloride: 0.2 M or 0.3 M). After running five start-up cycles with HBS-EP buffer, GAG were injected at concentrations of 1 lM, 10 lM and 100 lM of disaccharide units (DU) for 5 min at 30 lL min1, and binding levels were recorded 10 s before injection stopped (values taken from the reference subtracted sensorgrams relative to a baseline report point). Binding levels were corrected for the different molecular weights of the GAG derivatives to account for the fact that binding levels are related to mass increase at the sensor chip surface. The injection was followed by a 10 min dissociation phase in running buffer at a flow rate of 30 lL min1. It was chosen to inject the same molar concentrations in relation to disaccharide units rather than molecular weight to be able to compare the same amount of possible binding sites. The sensor surface was regenerated after each sample injection with a 30 lL injection of 5 M sodium chloride in 2.5 mM sodium hydroxide, and a 2 min stabilization phase with running buffer was used to reach a stable baseline prior to injection of the next sample. To ensure that differences in binding levels were not due to a loss in binding capacity of immobilized TGF-b1 (because of repeated regeneration), sHya3 was injected as a control (sHya3-21 C) every other third to fourth cycle in a sequence of various GAG injections. For ranking of binding levels, injections were made in the order sHya3-21 C1, sHya3-50, sCS3, scmHya3, sHya3-21 C2, sHya2-26, sHya2-14, sHya3-21 C3, sHya1, CS-A, Hya-LMW, sHya3-21 C4. In a second independent experiment, in which TGF-b1 was immobi-

2147

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

lized to another sensor chip surface, the order was sHya3-50 C1, sCS3, scmHya3, sHya3-50 C2, sHya2-26, sHya1, sHya3-50 C3, CSA, Hya-LMW, sHya3-50 C4 and sHya3-50, sHya3-21, sHya2-26, sHya2-14. In a third experiment, in which TGF-b1 was again immobilized to another sensor chip surface, heparin was injected as a naturally high-sulfated positive control and compared with the binding levels of sHya3-21. the Resulting values were then normalized to the response of sHya3-21 in previous experiments for comparison with binding levels of the other GAG derivatives. Binding parameters were evaluated using the Biacore™ T100 evaluation software 2.03. A statistical analysis was performed from both independent experiments with two-way ANOVA using the Tukey test. For statistics, values were normalized to the binding response at 10 lM sHya3-21, since the amount of immobilized TGFb1 was not exactly the same for a repeated chip surface preparation. 3. Results 3.1. Preparation and characterization of Hya and CS derivatives Various sulfated GAG derivatives were prepared, starting from non-sulfated, non-derivatized Hya and carboxymethylated Hya (cmHya) as well as from native CS containing approximately one sulfate group per disaccharide repeating unit. Preparation of lowsulfated Hya (sHya1, DS  1) and medium-sulfated Hya (sHya226, sHya2-14, all with DS  2) was performed using varying amounts of SO3–pyridine as sulfating agent, whereas high-sulfated Hya (sHya3-21, sHya3-50), high-sulfated carboxymethyl Hya (scmHya3) and high-sulfated CS (sCS3), all with DS  3 (for exact DS values, see Tables 1 and 2; for structures, see Fig. 1), were sulfated using SO3DMF. Synthesized derivatives were characterized by their average DS and, in the case of scmHya3, additionally by the average DCM. DS values are summarized in Tables 1 and 2. For heparin, a DS of 2.2 was determined. The number-average (Mn) and weight-average (Mw) molecular weights of the sulfated GAG were measured by GPC with LLS detection (see Tables 1 and 2). The LLS signal is directly proportional to the molecular weight, resulting in absolute values for the molecular weights. In addition, values for the PD are given (in brackets). Owing to the fact that the LLS signal can lead to a discrimination of low-molecular-weight fractions (5000–50,000 g mol1) of the measured polymers, inaccurately narrow molecular weight distributions may be the result. Therefore, the calculation of PD (PD = Mw/Mn) was performed on the basis of Mn and Mw values measured by RI detection. For heparin, a Mw between 17,000 and 19,000 g mol1 was assumed, as specified by the distributor. 13 C-NMR spectroscopy was employed to study the distribution of sulfate groups introduced within the disaccharide repeating units of Hya and CS. As recently shown for sHya, nearly complete sulfation of the C-6 position occurs, even in the case of the low-sulfated sHya1 derivative. In both medium- and high-sulfated Hya derivatives, besides complete sulfation of the primary hydroxy group at C-6, partial sulfation of the secondary hydroxy groups C-20 , C-30 and C-4 to a comparable extent was observed [35].

Table 2 Characteristics of CS and the high-sulfated CS derivative: DS; number-average (Mn) and weight-average (Mw) molecular weight as determined by LLS and RI (in brackets) detection; molecular weight distributions (PD) based on the values calculated from RI detection. Sample

CS

sCS3

DS Mn (g mol1) Mw (g mol1) PD

0.8 16,600 (39,800) 20,700 (62,100) 1.6

3.1 18,000 (28,500) 20,000 (41,500) 1.5

For scmHya3 (DCM = 0.4, DS = 3.1), it was found by 13C-NMR that the carboxymethylation of Hya occurred predominantly but not exclusively at the primary hydroxy position of C-6. Subsequent sulfation led to a complete substitution of the C-6 position, as revealed by the disappearance of the 13C-NMR signal for the nonsubstituted C-6 at 61 ppm and the occurrence of signals at 70.07 and 68.32 ppm, which can be assigned to the carboxymethylated and the sulfated C-6 position, respectively. In addition, a partial sulfation of the secondary hydroxy groups at C-20 , C-30 and C-4 similar to high-sulfated Hya was observed. A more detailed study of the substituent distribution in carboxymethylated hyaluronan sulfates will be published elsewhere. The native CS used with an estimated DS of 0.8 is a mixture of 70% chondroitin-4-sulfate and 30% chondroitin-6-sulfate showing the corresponding signals for both, the non-substituted and the substituted C-4 and C-6 position in the 13C-NMR spectrum. In the high-sulfated sCS3 with a DS of 3, the signal at 62 ppm for the non-substituted C-6 position in the non-modified CS [38] disappeared, confirming complete substitution of hydroxy at C-6 by sulfate groups. With the exception of C-1/C-10 and C-6, carbon signals of the sugar rings move very close together in the spectra of highsulfated sCS3, thereby complicating the distinct assignment of the signals. However, as found for high-sulfated Hya, a sulfation of the secondary hydroxy goups at C-20 , C-30 and C-4 to a similar extent without a particular preference of one of the three secondary OH groups is postulated.

3.2. ELISA and BioLISA The quantification of TGF-b1 bound to the immobilized GAG, using specific antibodies as capture molecules, revealed that the highest amount of growth factor was found for sHya3-50 followed by sHya2-26 (Fig. 2A). The presence of carboxymethyl groups in sulfated Hya (sHya 3) significantly reduced the binding capacity. Similarly, CS3-20 exhibited a significantly weaker interaction with TGF-b1 than sHya3-50 at the two growth factor concentrations, even though they have the same degree of sulfation. Compared with these four high-sulfated GAG, the binding capacity of sHya1, CS and Hya-LMW was much lower, being not significantly different from a surface covered with BSA. In general, higher sulfation of Hya or CS derivatives resulted in greater amounts of bound TGF-b1. In addition to some smaller deviations, the quantification of TGF-b1 binding and release with the specific TGFR as capture molecules was quite comparable with that with the capture anti-

Table 1 Characteristics of Hya, synthesized Hya derivatives: DS; number-average (Mn) and weight-average (Mw) molecular weight as determined by LLS and RI (in brackets) detection; molecular weight distributions (PD) based on the values calculated from RI detection. Sample

Hya-LMW

sHya1

sHya2-14

sHya2-26

sHya3-21

sHya3-50

scmHya3

DS DCM Mn (g mol1) Mw (g mol1) PD

– – 37,000 (77,300) 49,000 (182,700) 2.4

1.0 – 9000 (45,300) 31,000 (156,600) 3.5

1.9 – 8000 (22,900) 14,100 (40,600) 1.8

1.8 – 18,000 (36,800) 26,000 (65,700) 1.8

2.9 – 12,000 (28,900) 21,000 (47,700) 1.7

3.1 – 20,000 (49,600) 50,000 (83,500) 1.7

3.1 0.4 16,000 (35,000) 25,000 (54,000) 1.5

2148

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

Fig. 1. Chemical structures of sulfated GAG tested for binding to TGF-b1.

Fig. 2. Quantification of bound TGF-b1 after incubation with immobilized GAG, as determined by ELISA with capture antibodies against (A) TGF-b1 and (B) by BioLISA with specific TGFR. Original concentration of TGF-b1: 20 ng (20) and 10 ng (10). Statistics: two-way ANOVA for (A) with (a) p < 0.001 and (b) p < 0.01 GAG vs BSA, (c) p < 0.01, (d) p < 0.001, (e) p < 0.001; for (B) with (a) p < 0.001 GAG vs BSA, (b) p < 0.01, (c) p < 0.001, (d) p < 0.05. sHya3:sHya3-50, sHya2:sHya2-26. Binding data are the average of triplicates.

bodies (Fig. 2B). Therefore, desorption analysis is shown here with TGFR only. The cumulative quantifications of the total amount of TGF-b1 desorbed from immobilized GAG over 4 days at 37 °C (Fig. 3) revealed that the significantly highest amount dissociated from sHya3-50 for the two concentrations tested. In contrast, the lowest amount of TGF-b1 desorbed from 2% BSA.

Fig. 3. Cumulative quantification of the amount of TGF-b1 desorbed from immobilized GAG over 4 days in PBS/1% BSA at 37 °C as determined by BioLISA with the TGF-b1 specific receptor TGFR. Original concentration of TGF-b1: (A) 20 ng; (B) 10 ng. Statistics: two-way ANOVA for (A) with (a) p < 0.001 sHya3-50 vs sHya2-26 and p < 0.05 vs sCS3 and scmHya3, (b) p < 0.001 Hya-LMW vs CS; for (B) with (a) p < 0.01 sHya3-50 vs sCS3 and p < 0.05 vs scmHya3, (b) p < 0.001 Hya-LMW vs CS, (c) p < 0.05 sHya2-26 vs scmHya3, (d) p < 0.05 scmHya3 vs sCS3. Values are the average of triplicates. Percentage of desorbed TGF-b1 after 4 days in relation to the amount initially bound (⁄scmHya3).

Desorption of TGF-b1 from sHya2-26, scmHya3 and sCS3 was significantly higher than from sHya1. Comparing these three GAG, sCS3 exhibited the lowest amount desorbed, but this was only significant for an original growth factor concentration of 10 ng (Fig. 3B). CS and Hya-LMW were the weakest desorbing GAG. After 4 days of desorption at 37 °C the highest total amount of TGF-b1 was retained by sHya3-50, followed by sHya2-26 compared with all other GAG coatings (Figs. 2B and 3). In both cases, 30% of the original amount bound was retained by the surface after 4 days for both concentrations. SHya3-50 retained higher total amounts of TGF-b1 than sCS3. However, the amount of

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

2149

The interactions of GAG derivatives and TGF-b1 were analysed via their binding levels recorded 10 s before injection stopped and determined from referenced signals relative to baseline response. The binding analysis revealed that consecutive control injections of sHya3-21 within a series of GAG injections followed by repeated regeneration did not lead to a decrease in binding capacity of TGF-b1 of more than 10% for the two lower concentrations and not more than 20% for the highest concentration (Fig. 4A). The relatively minor loss of binding capacity and the frequency of control injections ensured that the differences in binding levels of different GAG derivatives were not due to a loss in binding capacity of immobilized TGF-b1.

The binding levels of GAG derivatives were ranked according to their binding strength in comparison with unmodified Hya as well as naturally sulfated CS and heparin. They were corrected for the respective molecular weights of the GAG derivatives to account for the fact that binding levels are related to mass increase at the sensor chip surface. Higher sulfation of Hya and CS derivatives led to significantly higher binding levels (Fig. 4B), while there was no detectable binding level for CS and Hya-LMW. The presence of carboxymethyl groups in sulfated Hya, as shown for scmHya3, again reduced the binding level. The high-sulfated sHya3-21 exhibited significantly higher binding strength than sCS3 at comparable DS and molecular weight. Furthermore, the low-sulfated Hya derivative sHya1 exhibited higher binding levels than the CS derivatives of a comparable sulfation. Heparin with an intermediate DS of 2 between sHya1 and sHya3-21 also exhibited an intermediate binding strength, but it was not significantly different from sCS3 and smaller than for sHya2-14 of close molecular weight and DS (Fig. 5A). To evaluate the possible impact of steric hindrance and/or electrostatic repulsion of GAG from the surface in this experimental

Fig. 4. Ranking of binding levels for the interaction of GAG derivatives with TGF-b1 by SPR analysis at 37 °C. The growth factor was immobilized on the sensor chip surface. GAG samples diluted in HBS-EP buffer were injected at concentrations of 1 lM, 10 lM and 100 lM (DU) for 5 min at 30 lL min1, and binding levels were recorded 10 s before injection stop. The sensor surface was regenerated after each sample injection. (A) Binding levels of sHya3-21 control injections every other third to fourth cycle within the other GAG injections (sHya3-21 C1–C4). The binding levels were determined from referenced signals relative to baseline response. (B) Binding levels of different GAG derivatives corrected for their respective molecular weights. The data presented are derived from one measurement. A second independent measurement was performed with TGF-b1 immobilized to another sensor chip surface. Data are representative for both experiments as indicated by statistics with two-way ANOVA. Significant differences with p < 0.05 were found for groups (a) and (b) as well as for: (c) GAG compared with sHya1 and (d) with HyaLMW and CS, respectively.

Fig. 5. Influence of the molecular weight, sulfation and molecular backbone on binding level and dissociation of the complex of GAG derivatives and TGF-b1 as determined by SPR analysis at 37 °C. GAG samples diluted in HBS-EP buffer were injected at concentrations of 1 lM, 10 lM and 100 lM DU for 5 min at 30 lL min1, followed by a 10 min dissociation phase in running buffer. The sensor surface was regenerated after each sample injection. (A) Binding levels corrected for the different molecular weights. The data presented are derived from one measurement, but are representative for two independent experiments, as indicated by statistics with two-way ANOVA. Significant differences with p < 0.05 were found for groups (a), (b) and (c). (B) Dissociation of GAG derivatives from TGF-b1/GAG complexes (injected sample concentration: 100 lM DU). Sensorgrams were overlaid at the end of the injection, and the reduction of complexes over time was observed for 10 min. Data are representative of two independent experiments.

desorbed TGF-b1 did not change much after 2 days. After 4 days, the percentage of desorbed growth factor in relation to the amount initially bound was quite comparable for sHya3-50 and sCS3 (Fig. 3). In contrast, almost no TGF-b1 was retained by scmHya3, sHya1 and Hya-LMW after the same time. 3.3. SPR-analysis of GAG–TGF-b1 interaction

2150

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

Fig. 6. Influence of salt concentration on binding levels as determined by SPR at 37 °C. GAG samples at concentrations of 10 lM and 100 lM DU and diluted in HBSEP buffer with additional 0.05 M and 0.15 M sodium chloride (resulting total sodium chloride concentrations 0.20 M and 0.30 M) were injected for 5 min at 30 lL min1. Resulting binding levels were related to the response in the presence of 0.15 M sodium chloride (=running buffer) set as 100%. The sensor surface was regenerated after each sample injection. Data are representative for two independent experiments.

set-up, GAG derivatives with a comparable degree of sulfation, but different molecular weight were investigated for their binding levels. The binding analysis revealed that binding levels were higher at lower MW, but comparable DS for all three concentrations of the selected GAG tested (Fig. 5A). The higher sulfation of Hya derivatives again led to significantly higher binding levels at comparable molecular weights (sHya3-21 vs sHya2-26). The decay of TGF-b1/GAG complexes was analysed by examining their dissociation over 10 min. The comparison of the dissociation revealed that, in general, the decay of TGF-b1/GAG complexes was relatively slow (Fig. 5B). In terms of dissociation, the four GAG exhibiting the highest binding strength followed the order sHya321 > sHya2-26 > scmHya3 > sCS3. Furthermore, to elucidate the driving force of the interaction between GAG derivatives and TGF-b1, four selected GAG (sHya350, sCS3, heparin and sHya2-26) with comparatively high binding strength for the immobilized TGF-b1 were analysed for their binding levels at physiological and supra-physiological sodium chloride concentrations (Fig. 6). The binding analysis revealed that raising the sodium chloride concentration for 50 mM markedly reduced the binding level of GAG derivatives as well as of heparin, while the presence of double-physiological concentrations almost completely abolished the interaction of GAG with TGF-b1.

4. Discussion The major goal of this study was to investigate the growth factor binding abilities of Hya and CS derivatives modified with sulfate groups, aiming for a deeper understanding of structure– function relationships of GAG derivatives with respect to their unique binding properties for biological mediators. Furthermore, It was hoped to reveal candidates conducive to growth factor binding to biomaterial coatings. The combination of ELISA and SPR methods was used to ensure that confounding immobilization effects were minimized. In general, the ELISA experiments and the SPR binding analysis to investigate the interaction of GAG derivatives with TGF-b1 equally revealed that (i) Hya derivatives exhibited higher binding strength than CS derivatives of the same degree of sulfation (Figs. 2 and 4B), (ii) a higher binding strength could be observed for higher sulfation

degree within both GAG types (Figs. 2 and 4B), (iii) the presence of carboxymethyl groups in sHya3 significantly reduced the affinity to TGF-b1 (Figs. 2 and 4B), and (iv) the highest total amount of TGF-b1 or GAG dissociated from sHya3-21/50/TGF-b1, followed by sHya2-26/TGF-b1 and sCS3/TGF-b1 complexes (Figs. 3 and 5B). The literature provides further support for the importance of the degree of sulfation of GAG in the interaction with TGF-b1. Heparin and high-sulfated heperan sulfate potentiate the biological activity of TGF-b1, while low-sulfated mucosal heperan sulfate, dermatan sulfate and CS do not [28]. But obviously, besides the sulfation degree, other important factors have to be considered. Of particular interest was the different binding and desorption behaviour revealed by the two high-sulfated GAG variants, sHya3-21/50 and sCS3. While the interaction of sHya3 with TGFb1 exhibited the highest binding strength (Figs. 2 and 4B) and total release, sHya3-50 still retained the highest amount of protein/GAG complex after 4 days at 37 °C (Figs. 2, 3 and 5B). SCS3 showed significantly lower binding strength (Figs. 2 and 4B). Since the initial dissociation is slower for sCS3 than for sHya3 as determined by SPR (Fig. 5B), the faster association of TGF-b1 (as represented by the binding level in Fig. 4B) is the main reason for the higher binding strength of sHya3. However, after 4 days of TGF-b1 desorption at 37 °C the release of both GAG variants was quite comparable in relation to the amount initially bound (Figs. 2 and 3). There was no major release between days 2 and 4, indicating that at this time a steady state had been reached. One explanation for the differences in binding behaviour of Hya and CS derivatives observed in this study, with two independent methods, could be that the different molecular geometries in the sugar backbone (in Hya: N-acetyl-d-glucosamine; in CS: N-acetylD-galactosamine) might render the respective sulfated groups to interact differently. This is further emphasized by the observation that naturally sulfated heparin (N-acetyl-D-glucosamine) also demonstrated a comparable binding strength to sCS3, even though it has a lower DS (Fig. 4B). Another reason could be differences in sulfation pattern. For Hya derivatives, it could be shown that sulfation occurred preferentially at the primary OH-group at C-6 of the N-acetylglucosamine unit, which is completely sulfated, and to a lesser extent on the remaining secondary OH groups at C-20 and C-30 of the glucuronic acid unit and C-4 of the N-acetylglucosamine unit [35]. The native CS used is a mixture containing 70% chondroitin-4sulfate, which means that the C-4 position of the N-acetylgalactosamine unit is already highly sulfated before chemical sulfation is started. Further sulfation led to a complete substitution of the C-6 position and a subsequent partial substitution of the remaining secondary OH groups in C-20 and C-30 positions of the glucuronic acid unit, and the remaining 30% of the non-substituted hydroxy groups at C-4. However, this could lead to a higher overall sulfation of the C-4 position in the CS derivatives compared with the one in Hya. This might be of interest with regard to recent molecular modelling studies on equilibrium conformations of hyaluronan, chondroitin (the non-sulfated derivative of CS) and differently sulfated CS [39]. In these studies it was found that C-4 sulfation of the galactosamine unit has a strong influence on the torsion flexibility of the b-1,3 glycosidic linkage in related CS molecules, whereas the influence of C-6 sulfation on conformational changes is relatively low. Unfortunately, owing to signal overlapping in the spectral region of the 13C-NMR spectra where the C-4 signals appear, no exact determination of the DS for the C-4 positions was possible, either for medium- and high-sulfated CS or for Hya derivatives. Further analytical investigation to determine the DS values of each of the three secondary carbon positions must be per-

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

formed to validate a higher sulfation in the C-4 position of sCS3 compared with sHya3. The importance of the sulfation pattern is further supported by the fact that less regular and naturally sulfated heparin exhibited a lower binding strength than sHya2-14, even though they are of close molecular weight and DS (Figs. 4B and 5A). In the SPR analysis, some interaction could be found for sHya1, but no interaction was detected for Hya-LMW and CS (Fig. 4B). However, in ELISA studies, no major differences could be found between sHya1, CS and Hya-LMW and the control BSA (Fig. 2), probably owing to the lower sensitivity compared with SPR. This is an important difference from previous findings with BMP-4, which exhibited a significantly higher binding strength for sHya1 than for CS in SPR as well as in ELISA [35]. This might suggest that the sulfation in the C-6 position of the N-acetylglucosamine unit (sHya1) in comparison with the C-4 position in the N-acetylgalactosamine amine unit is of greater importance for the interaction of the GAG derivatives investigated with BMP-4 than with TGF-b1. It also means that, for a significant interaction with TGF-b1, a higher sulfation of GAG derivatives is necessary, i.e., at least two sulfate groups per disaccharide unit. The findings for a non-binding CS are in line with Hildebrand and co-workers, who demonstrated that the core proteins of decorin and biglycan interact with the TGF-b isoforms 1, 2 and 3, but that the GAG chains CS and dermatan sulfate seemed to hinder binding, since their removal increased the competitiveness for TGF-b binding of both proteoglycans [21]. However, the present findings for a non-binding Hya-LMW are in opposition to a suggestion by Locci et al. for a direct interaction between non-sulfated, high molecular weight Hya and TGF-b1 based on the observation that the growth factor lost its capacity to inhibit the proliferation of thymocytes when pre-incubated with Hya, but not with sulfated glycosaminoglycans [34]. But in line with their findings, the present authors could not detect any loss in specific TGFR binding activity of TGF-b1 after contact or desorption from the sulfated GAG derivatives, since the binding to the specific capture antibody was quite comparable. Interestingly, the carboxymethylation of Hya before sulfation (scmHya3) reduced the affinity in both experimental set-ups (Figs. 2 and 4B), which could be explained by an altered molecular geometry due to the incorporation of about one carboxymethyl group per tetrasaccharide unit. The conformational change could then lead to a less efficient interaction with the positive charges on the growth factor. Another explanation could be that, since the carboxymethylation preceded the sulfation of scmHya3, the C-6 positions were partially occupied by carboxymethyl groups and significantly less sulfated than in the sHya3 derivatives. This would suggest that sulfation in this position is of special importance for the interaction of GAG derivatives with TGF-b1, as already postulated for BMP-4 [35,40]. Furthermore, the secondary OH groups in the C-20 , C-30 and C-4 positions carry a higher sulfation in scmHya3 then in sHya3, which in case of C-4 could again lead to an altered torsion flexibility of the b-1,3 glycosidic linkage, as mentioned for sCS3. In SPR analysis with immobilized growth factor and solute GAG derivatives, one has to consider possible influences of steric hindrance and/or electrostatic repulsion due to highly charged molecules. These effects play an important role, since GAG derivatives with comparable DS but lower molecular weight reached higher binding levels. In case of, for example, high-sulfated Hya, this means that the derivative with the higher molecular weight contains more negatively charged groups and therefore exhibits a bigger hydrodynamic radius, leading to reduced diffusion rates and increased electrostatic repulsion among GAG molecules. Therefore, the interaction of the 50 kDa variant with immobilized TGF-b1 reached only approximately half the binding level in comparison

2151

with the 21 kDa variant (Fig. 5A). This is especially important to consider for correct ranking, since the markedly different binding levels of sHya3-21 and sCS3 at comparable molecular weight reflected the higher affinity of sHya, while the comparison with the 50 kDa variant sHya3-50 did not (Figs. 4B and 5A). Avoiding further electrostatic repulsion of sulfated GAG by a negatively charged carboxymethylated dextran matrix was also the reason why the present authors used a matrix-free C1 chip instead of a standard CM5 chip. The major driving force of the interaction between GAG derivatives and naturally sulfated heparin with TGF-b1 was revealed to be electrostatic, since the presence of double-physiological sodium chloride concentrations reduced the binding level to almost zero (Fig. 6). This is in contrast to the findings of Mooradian et al. [24] demonstrating that the interaction of fibronectin and TGF-b1 is insensitive to ionic strength over a range of 0.1–1 M sodium chloride and thus not strongly dependent on charge. This difference seems reasonable, since sulfated glycosaminoglycans do not offer hydrophobic interaction sites. The present findings demonstrate that the modified GAG investigated provide different and defined growth factor interaction profiles which are promising for adding beneficial growth factor binding features to specifically tailored implant coatings in biomedical applications. 5. Conclusion Specific interactions between Hya and CS derivatives and TGFb1 with different binding strength depending on sulfation degree and sugar backbone were demonstrated. The differentially sulfated Hya and CS derivatives are therefore promising model substances for reaching a further understanding of the structure–function relationships of GAG. They are also interesting candidates for bioengineered coatings to improve the biological properties of biomaterials used in medical applications. Acknowledgements The authors thank Aline Katzschner for excellent technical assistance and Dr. Anja Drescher for critical suggestions to the SPR part. They further gratefully acknowledge financial support by the DFG (PAK 105; TRR 67, A3). References [1] Massague J, Wotton D. Transcriptional control by the TGF-beta/Smad signaling system. EMBO J 2000;19:1745–54. [2] Massague J. TGF-beta signal transduction. Annu Rev Biochem 1998;67:753–91. [3] O’Connor-McCourt MD, Wakefield LM. Latent transforming growth factor-beta in serum. A specific complex with alpha 2-macroglobulin. J Biol Chem 1987;262:14090–9. [4] Huang SS, O’Grady P, Huang JS. Human transforming growth factor beta.alpha 2-macroglobulin complex is a latent form of transforming growth factor beta. J Biol Chem 1988;263:1535–41. [5] Assoian RK, Sporn MB. Type beta transforming growth factor in human platelets: release during platelet degranulation and action on vascular smooth muscle cells. J Cell Biol 1986;102:1217–23. [6] Carrington JL, Roberts AB, Flanders KC, Roche NS, Reddi AH. Accumulation, localization, and compartmentation of transforming growth factor beta during endochondral bone development. J Cell Biol 1988;107:1969–75. [7] Tang Y, Wu X, Lei W, Pang L, Wan C, Shi Z, et al. TGF-beta1-induced migration of bone mesenchymal stem cells couples bone resorption with formation. Nat Med 2009;15:757–65. [8] Pircher R, Jullien P, Lawrence DA. Beta-transforming growth factor is stored in human blood platelets as a latent high molecular weight complex. Biochem Biophys Res Commun 1986;136:30–7. [9] Kanzaki T, Olofsson A, Moren A, Wernstedt C, Hellman U, Miyazono K, et al. TGF-beta 1 binding protein: a component of the large latent complex of TGFbeta 1 with multiple repeat sequences. Cell 1990;61:1051–61. [10] Bonewald LF, Dallas SL. Role of active and latent transforming growth factor beta in bone formation. J Cell Biochem 1994;55:350–7.

2152

V. Hintze et al. / Acta Biomaterialia 8 (2012) 2144–2152

[11] Dallas SL, Park-Snyder S, Miyazono K, Twardzik D, Mundy GR, Bonewald LF. Characterization and autoregulation of latent transforming growth factor beta (TGF beta) complexes in osteoblast-like cell lines. Production of a latent complex lacking the latent TGF beta-binding protein. J Biol Chem 1994;269: 6815–21. [12] Pfeilschifter J, Bonewald L, Mundy GR. Characterization of the latent transforming growth factor beta complex in bone. J Bone Miner Res 1990;5: 49–58. [13] Oursler MJ. Osteoclast synthesis and secretion and activation of latent transforming growth factor beta. J Bone Miner Res 1994;9:443–52. [14] Centrella M, McCarthy TL, Canalis E. Transforming growth factor beta is a bifunctional regulator of replication and collagen synthesis in osteoblastenriched cell cultures from fetal rat bone. J Biol Chem 1987;262:2869–74. [15] Hock JM, Canalis E, Centrella M. Transforming growth factor-beta stimulates bone matrix apposition and bone cell replication in cultured fetal rat calvariae. Endocrinology 1990;126:421–6. [16] Janssens K, ten Dijke P, Janssens S, Van Hul W. Transforming growth factorbeta1 to the bone. Endocr Rev 2005;26:743–74. [17] Noda M, Camilliere JJ. In vivo stimulation of bone formation by transforming growth factor-beta. Endocrinology 1989;124:2991–4. [18] Takeuchi Y, Matsumoto T, Ogata E, Shishiba Y. Effects of transforming growth factor beta 1 and L-ascorbate on synthesis and distribution of proteoglycans in murine osteoblast-like cells. J Bone Miner Res 1993;8:823–30. [19] Wrana JL, Maeno M, Hawrylyshyn B, Yao KL, Domenicucci C, Sodek J. Differential effects of transforming growth factor-beta on the synthesis of extracellular matrix proteins by normal fetal rat calvarial bone cell populations. J Cell Biol 1988;106:915–24. [20] Ignotz RA, Massague J. Transforming growth factor-beta stimulates the expression of fibronectin and collagen and their incorporation into the extracellular matrix. J Biol Chem 1986;261:4337–45. [21] Hildebrand A, Romaris M, Rasmussen LM, Heinegard D, Twardzik DR, Border WA, et al. Interaction of the small interstitial proteoglycans biglycan, decorin and fibromodulin with transforming growth factor beta. Biochem J 1994;302(Pt 2):527–34. [22] Yamaguchi Y, Mann DM, Ruoslahti E. Negative regulation of transforming growth factor-beta by the proteoglycan decorin. Nature 1990;346:281–4. [23] Bi Y, Stuelten CH, Kilts T, Wadhwa S, Iozzo RV, Robey PG, et al. Extracellular matrix proteoglycans control the fate of bone marrow stromal cells. J Biol Chem 2005;280:30481–9. [24] Mooradian DL, Lucas RC, Weatherbee JA, Furcht LT. Transforming growth factor-beta 1 binds to immobilized fibronectin. J Cell Biochem 1989;41: 189–200. [25] Murphy-Ullrich JE, Schultz-Cherry S, Hook M. Transforming growth factorbeta complexes with thrombospondin. Mol Biol Cell 1992;3:181–8. [26] Paralkar VM, Vukicevic S, Reddi AH. Transforming growth factor beta type 1 binds to collagen IV of basement membrane matrix: implications for development. Dev Biol 1991;143:303–8.

[27] Schultz-Cherry S, Murphy-Ullrich JE. Thrombospondin causes activation of latent transforming growth factor-beta secreted by endothelial cells by a novel mechanism. J Cell Biol 1993;122:923–32. [28] Lyon M, Rushton G, Gallagher JT. The interaction of the transforming growth factor-betas with heparin/heparan sulfate is isoform-specific. J Biol Chem 1997;272:18000–6. [29] McCaffrey TA, Falcone DJ, Du B. Transforming growth factor-beta 1 is a heparin-binding protein: identification of putative heparin-binding regions and isolation of heparins with varying affinity for TGF-beta 1. J Cell Physiol 1992;152:430–40. [30] McCaffrey TA, Falcone DJ, Vicente D, Du B, Consigli S, Borth W. Protection of transforming growth factor-beta 1 activity by heparin and fucoidan. J Cell Physiol 1994;159:51–9. [31] McCaffrey TA, Falcone DJ, Brayton CF, Agarwal LA, Welt FG, Weksler BB. Transforming growth factor-beta activity is potentiated by heparin via dissociation of the transforming growth factor-beta/alpha 2-macroglobulin inactive complex. J Cell Biol 1989;109:441–8. [32] Holzmann J, Brandl N, Zemann A, Schabus R, Marlovits S, Cowburn R, et al. Assorted effects of TGFbeta and chondroitinsulfate on p38 and ERK1/2 activation levels in human articular chondrocytes stimulated with LPS. Osteoarthr Cartil 2006;14:519–25. [33] Manton KJ, Haupt LM, Vengadasalam K, Nurcombe V, Cool SM. Glycosaminoglycan and growth factor mediated murine calvarial cell proliferation. J Mol Histol 2007;38:415–24. [34] Locci P, Marinucci L, Lilli C, Martinese D, Becchetti E. Transforming growth factor beta 1-hyaluronic acid interaction. Cell Tissue Res 1995;281: 317–24. [35] Hintze V, Moeller S, Schnabelrauch M, Bierbaum S, Viola M, Worch H, et al. Modifications of hyaluronan influence the interaction with human bone morphogenetic protein-4 (hBMP-4). Biomacromolecules 2009;10:3290–7. [36] Kunze R, Roesler M, Moeller S, Schnabelrauch M, Riemer T, Hempel U, et al. Sulfated hyaluronan derivatives reduce the proliferation rate of primary rat calvarial osteoblasts. Glycoconj J 2010;27:151–8. [37] van der Smissen A, Hintze V, Scharnweber D, Moeller S, Schnabelrauch M, Majok A, et al. Growth promoting substrates for human dermal fibroblasts provided by artificial extracellular matrices composed of collagen I and sulfated glycosaminoglycans. Biomaterials 2011;32:8938–46. [38] Bociek SM, Darke AH, Welti D, Rees DA. The 13C-NMR spectra of hyaluronate and chondroitin sulphates. Further evidence on an alkali-induced conformation change. Eur J Biochem 1980;109:447–56. [39] Bathe M, Rutledge GC, Grodzinsky AJ, Tidor B. A coarse-grained molecular model for glycosaminoglycans: application to chondroitin, chondroitin sulfate, and hyaluronic acid. Biophys J 2005;88:3870–87. [40] Miyazaki T, Miyauchi S, Tawada A, Anada T, Matsuzaka S, Suzuki O. Oversulfated chondroitin sulfate-E binds to BMP-4 and enhances osteoblast differentiation. J Cell Physiol 2008;217:769–77.