Acta Biomaterialia 9 (2013) 7775–7786

Contents lists available at SciVerse ScienceDirect

Acta Biomaterialia

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

Artificial extracellular matrix composed of collagen I and highly sulfatedhyaluronan interferes with TGFb1 signaling and prevents TGFb1-inducedmyofibroblast differentiation

1742-7061/$ - see front matter � 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.actbio.2013.04.023

⇑ Corresponding author at: Department of Dermatology, Venereology and Aller-gology, Leipzig University, 04103 Leipzig, Germany. Tel.: +49 341 9725854; fax: +49341 9725878.

E-mail address: anja.vandersmissen@medizin.uni-leipzig.de (A. van derSmissen).

Anja van der Smissen a,e,⇑, Sergey Samsonov b,e, Vera Hintze c,e, Dieter Scharnweber c,e,Stephanie Moeller d,e, Matthias Schnabelrauch d,e, M. Teresa Pisabarro b,e, Ulf Anderegg a,e

a Department of Dermatology, Venereology and Allergology, Leipzig University, 04103 Leipzig, Germanyb Structural Bioinformatics, BIOTEC, Technische Universität Dresden, 01307 Dresden, Germanyc Max Bergmann Center of Biomaterials, Technische Universität Dresden, 01069 Dresden, Germanyd Biomaterials Department, INNOVENT e.V. Jena, 07745 Jena, Germanye Collaborative Research Center (SFB-TR67), Matrixengineering Leipzig and Dresden, Germany

a r t i c l e i n f o

Article history:Received 16 October 2012Received in revised form 27 March 2013Accepted 9 April 2013Available online 18 April 2013

Keywords:Wound healingMyofibroblastTransforming growth factor b1Collagen ISulfated hyaluronan

a b s t r a c t

Sulfated glycosaminoglycans are promising components for functional biomaterials since sulfate groupsmodulate the binding of growth factors and thereby influence wound healing. Here, we have investigatedthe influence of an artificial extracellular matrix (aECM) consisting of collagen I (coll) and hyaluronan(HA) or highly sulfated HA (hsHA) on dermal fibroblasts (dFb) with respect to their differentiation intomyofibroblasts (MFb). Fibroblasts were cultured on aECM in the presence of aECM-adsorbed or solubletransforming growth factor b1 (TGFb1). The synthesis of a-smooth muscle actin (aSMA), collagen andthe ED-A splice variant of fibronectin (ED-A FN) were analyzed at the mRNA and protein levels. Further-more, we investigated the bioactivity and signal transduction of TGFb1 in the presence of aECM andfinally made interaction studies of soluble HA or hsHA with TGFb1. Artificial ECM composed of colland hsHA prevents TGFb1-stimulated aSMA, collagen and ED-A FN expression. Our data suggest animpaired TGFb1 bioactivity and downstream signaling in the presence of aECM containing hsHA, shownby massively reduced Smad2/3 translocation to the nucleus. These data are explained by in silico dockingexperiments demonstrating the occupation of the TGFb-receptor I binding site by hsHA. Possibly, HA sul-fation has a strong impact on TGFb1-driven differentiation of dFb and thus could be used to modulate theproperties of biomaterials.

� 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction

The development and application of functional biomaterials isnecessary to support dermal wound healing in acute and chronicskin wounds, like trauma or chronic ulcerations [1]. Cutaneouswound healing is a multistep process that involves several cell typesand events, beginning with the inflammatory response, migrationand proliferation of dermal and epidermal cells and ending with ma-trix synthesis in order to fill the wound gap and re-establish themechanical barrier of the skin [2–4]. During these processes dermalfibroblasts (dFb) are crucially involved, rapidly adhering to thewound matrix, proliferating and subsequently synthesizing extra-

cellular matrix (ECM). To obtain an ECM synthesizing phenotypedFb have to differentiate into myofibroblasts (MFb) [5,6].

Myofibroblasts are key players in the physiological reconstruc-tion of connective tissue after injury [5]. They synthesize highamounts of ECM components like collagen and during non-patho-logical wound healing they are removed by apoptosis when epithe-lialization occurs [7]. The transition from dFb to MFb requiresstimulation with transforming growth factor b1 (TGFb1) andmechanical tension [5]. Myofibroblasts express the ED-A splicevariant of fibronectin (ED-A FN) and a-smooth muscle actin(aSMA) [8], which is integrated into stress fibers and thereby gen-erates contractile forces. Despite the synthesis of collagen and theexpression of ED-A FN, aSMA expression and fibrous organizationrepresent the most reliable marker of differentiated MFb [6]. Theformation and contraction of granulation tissue is essential tomaintain tissue integrity, to reduce the size of the wound and closethe defect, resulting in a scar. But when apoptosis of MFb is lacking,pathological changes result in hypertrophic scars or fibrosis [7].

7776 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

TGFb1 is the key cytokine in the evolution of lesions characterizedby MFb formation [6].

TGFb1 is a disulfide bonded homodimeric cytokine [9] that is se-creted as a part of the large latent complex, which consists of thelatency-associated protein, latent TGFb1 binding protein 1 andthe growth factor itself [10,11] and, thereby, can be stored in theECM [12,13]. The protease-independent release and activation oflatent TGFb1 requires integrin-mediated MFb contraction in thepresence of a stiff growth substrate [14]. Binding of active TGFb1

to TGFb receptor II (TGFbRII) leads to the recruitment and phos-phorylation of TGFb receptor I (TGFbRI). This heteromeric complexphosphorylates cytoplasmic Smad2 and Smad3, resulting in theheteromeric Smad2/3 complex that binds to Smad4 and is translo-cated to the nucleus to regulate gene transcription of a variety ofgenes [5,12]. However, alternative signaling pathways are known,as discussed later.

A promising approach for the development of functional bioma-terials is the combination of native ECM components with chemi-cally modified substances to selectively modulate physiologicalprocesses like wound healing. Here we have investigated an artifi-cial ECM (aECM) consisting of collagen I (coll) as a structural pro-tein and several hyaluronan (HA) derivatives. Hyaluronan is anatural, polyanionic, linear heteropolysaccharide, consisting ofrepeating non-sulfated disaccharide units of D-glucuronic acid (b-1–3) and N-acetyl-D-glucosamine (b-1–4) [15,16]. In skin this gly-cosaminoglycan (GAG) represents a major component of the ECMand is mainly produced by fibroblasts [17]. For this study HAwas chemically modified by the introduction of sulfate groups toobtain highly sulfated derivatives in comparison with native HA.This chemical modification aims to increase the binding of cell-derived growth factors like TGFb1. GAGs are interesting moleculesfor artificial biomaterial design because of their low immunogenic-ity and their property to bind, concentrate and prevent the diffu-sion of growth factors. Furthermore, they can improve andstabilize the presentation to relevant receptors and protect growthfactors from proteolytic degradation [18–20]. The main GAG typesdiffer only slightly in the basic sugar backbone of the GAG chain.However, subsequent sulfation, deacetylation and epimerizationmodifications distinguish individual GAG chains and are criticalfor their specific activity [21]. Protein–GAG recognition is drivenby ionic interactions because of the high negative charge providedmainly by sulfate but also by carboxylate groups of the GAG andbasic amino acids of the protein. The topology and distribution ofbasic amino acids in the GAG-binding site influence the specificityof molecular recognition of GAG sequences [20,22,23]. Despite thedegree of sulfation, GAG–mediator interactions are also affected bythe position of the sulfate groups within the anhydrosugar repeat-ing unit and by the linkage of sulfate group to the sugar ring (O- orN-sulfation) [24].

The introduction of sulfate groups along the carbohydrate back-bone of GAGs has been generated by non-template driven pro-cesses and thereby provides heterogeneity in the fine structure ofthese molecules. It is known that the interaction of different GAGswith different proteins is dictated by the pattern and orientation ofthe sulfate groups along the carbohydrate backbone [25,26]. A

Table 1Nomenclature and characteristics of native, LMW and sulfated HA.

Sample Description S content (%) DSS

HA Native HAhsHA Highly sulfated HA 13.4/13.5 3.1/3.4LMW-HA Thermally degraded HA

DSS, average number of sulfate groups per disaccharide repeating unit (possible DSS raccording to Salek-Ardakani et al. [24]; PD, polydispersity index (molecular weight dist

certain protein can also interact with different sulfation patternswith different binding affinities and kinetics, whereby the kineticsaffect the functions of cytokines or chemokines and the biologicaloutcomes [27]. Moreover, the site on the protein recognised by theGAG structure contributes to whether GAG binding supports orinhibits a signaling process [28]. Thus, guided modifications toGAG structures may support or inhibit ligand–receptor interac-tions, providing a special mechanism for regulatory control.

Our previous work with sulfated GAGs integrated in collage-nous matrices showed that initial adhesion and cell proliferationof dFb progressively increased with the degree of GAG sulfation,while synthesis of the ECM components coll and HA was decreasedon aECM containing highly sulfated GAG. The MFb differentiationmarker aSMA was not significantly affected by aECM within 48 hat the mRNA level in resting fibroblasts [29].

Since we hypothesize that TGFb1 interactions with sulfategroups could influence its bioavailability and thereby could regu-late MFb differentiation, the present study has investigated the im-pact of aECM with highly sulfated HA (hsHA) on the TGFb1-drivendifferentiation of dFb to MFb by analyzing the induction of MFbmarker genes aSMA, coll and ED-A FN in dFb within 72 h in thepresence of aECM-adsorbed or soluble TGFb1. Furthermore, weprovide data on the underlying mechanisms concerning the bioac-tivity and signal transduction of TGFb1 in the presence of highlysulfated aECM and, finally, we provide an in silico interaction mod-el of non-sulfated HA and highly sulfated derivatives with TGFb1,which supports our data on the influence of HA sulfation on thebioactivity of TGFb1.

2. Materials and methods

2.1. Preparation and characterization of aECM

2.1.1. MaterialsNative high molecular weight HA from Streptococcus was ob-

tained from Aqua Biochem (Dessau, Germany). The synthesis andanalysis of hsHA derivatives were described in Kunze et al. [15]and Hintze et al. [18]. Low molecular weight HA (LMW-HA) wasprepared by controlled thermal degradation of native HA accordingto Moeller et al. [30]. The characteristics of the HA derivatives usedare summarized in Table 1.

2.1.2. Artificial extracellular matricesThe preparation, analysis and stability of aECM were described

previously [29,31]. In brief, the coll/GAG matrices were producedby in vitro fibrillogenesis of rat collagen I (BD Bioscience, Heidel-berg, Germany) in the presence of different GAGs in tissue cultureplates (Nunc, Langenselbold, Germany). 1 mg ml�1 acid-solubilizedcoll was mixed 1:1 with the same concentration of HA or the appro-priate same molarity of disaccharide units for hsHA according to themolecular weight of the disaccharide units (Table 1). This approachwas chosen to compare the same number of possible binding sites,since these compounds are heterogeneous in length and differ sig-nificantly in the molecular weight of their disaccharide units.

Mw (Da) PD Mw disaccharide unit (g mol-1)

1,174,865 4.80 401.349,725/51,145 1.70/1.73 717.6/748.347,785 2.37 401.3

ange 0 6 DSS P 4.0); Mw, weight-averaged molecular weight determined by GPCributions) determined by GPC according to Salek-Ardakani et al. [24].

A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786 7777

Collagen was allowed to form fibrils for 16–18 h at 37 �C andthen dried on the plates. The coatings were dried again after twowashing steps with 0.5 ml of deionized water. All aECM coatingswere washed with phosphate-buffered saline (PBS) for 1 h at37 �C before cell seeding. Characterization of the generated coat-ings revealed a ratio of coll and HA of 7:1, LMW-HA 19:1 and hsHA24:1 [29,31].

2.2. Isolation and culture of primary human dermal fibroblasts

The study was approved by the local ethics committee (065-2009) and conducted according to the Declaration of the HelsinkiPrinciples (1975).

Primary human dFb from healthy breast skin were isolated aspreviously described [32] by dispase II (Roche Diagnostics GmbH,Mannheim, Germany)-mediated removal of the epidermal sheetand digestion of the dermal compartment with collagenase(Sigma–Aldrich Chemie GmbH, Steinheim, Germany). Cell suspen-sion was passed through 70 lM filters (BD Biosciences, Bedford,MA) to remove tissue debris. Dermal fibroblasts were culturedwith Dulbecco’s modified Eagle’s medium (DMEM) (BiochromAG, Berlin, Germany) supplemented with 10 vol.% fetal calf serum(FCS) (Biochrom) and 1 vol.% penicillin/streptomycin (P/S) (PAALaboratories GmbH, Pasching, Austria) at 37 �C, 5% CO2 until con-fluence. Cells between passages 2 and 8 were used and detachedusing 0.05% trypsin/0.02% EDTA (Biochrom). Dermal fibroblastswere cultured on aECM for 72 h with DMEM/10% FCS/1% P/S at37 �C, either untreated or with TGFb1 added to the culture mediumafter cell seeding or in the presence of aECM-adsorbed TGFb1

(10 ng ml�1, R&D Systems, Wiesbaden, Germany). 10 ng ml�1

recombinant human TGFb1 in DMEM/0.1% (w/v) bovine serumalbumin (BSA) (PAA Laboratories GmbH) were added to aECM for2 h at 37 �C to enable growth factor adsorption. The supernatantwas removed and used for TGFb1 quantification. The wells werebriefly washed with PBS and used for cell seeding.

For investigations of TGFb1–GAG interaction dFb were exposedto TGFb1 and soluble LMW-HA or soluble hsHA. To exclude theinfluence of molecular weight (MW) on the interaction with TGFb1

we decided to apply LMW-HA, which has an average MW of 48 kDaand therefore shows a similar average MW as hsHA (50 and51 kDa, respectively). 5 � 104 dFb were incubated for 72 h at37 �C, 5% CO2 in DMEM/10% FCS/1% P/S in 6-well plates. 10 ng ml�1

TGFb1, 50 lg ml�1 LMW-HA, 50 lg ml�1 hsHA or combinations ofgrowth factor with GAGs were added. Before supplementation ofdFb, all components were preincubated for 1 h at 37 �C, 5% CO2

to allow TGFb1 and GAG to interact in the case of combinedapplication.

2.3. RNA preparation and quantitative real time PCR (qRT-PCR)

aECM was provided in 6-well plates and incubated with 5�104

dFb per well for 72 h at 37 �C. Total RNA was directly preparedfrom the monolayers with a Qiagen RNeasy� Mini Kit (Qiagen, Hil-den, Germany). RNA quantity and quality were determined byspectrophotometry (ND-1000, Nano Drop Technologies, Wilming-ton, DE). First strand cDNA synthesis was performed as previouslydescribed [34] using 1 lg total RNA. GoTaq� qPCR Master Mix (Pro-mega Corporation, Madison, WI) was used for qRT-PCR, which wasperformed with a Rotor-Gene Q cycler (Qiagen): denaturation for5 min at 95 �C; 28–40 amplifications of denaturation (92 �C, 10 s),annealing under primer-specific conditions (20 s) and targetgene-specific extension (30 s at 61 �C). The primers were synthe-sized by Metabion International AG (Martinsried, Germany) (seeSupplementary Table 1). Fluorescence was measured for 20 s at80–82 �C depending on the melting temperature of the amplifiedDNA. The specificity of the PCR products was confirmed by melting

curve analysis at the end of each run. Genes were normalized tothe unregulated reference gene rps26[33]. Results are expressedas fold induction with respect to corresponding coll controls.

2.4. Immunofluorescence staining

aECM was provided in 8-well glass Chamber Slides (Thermo Sci-entific, Rochester, NY). For the analysis of aSMA, ED-A FN and col-lagen 4 � 103 dFb per well were cultured for 72 h at 37 �C. ForSmad2/3 investigations 1 � 104 dFb per well were cultured for5 h at 37 �C. Cells were fixed with 4% (w/v) paraformaldehyde(Roth, Karlsruhe, Germany), permeabilized with 0.5 vol.% Tween20 (Roth) and blocked with 0.1 vol.% Tween 20/3% (w/v) BSA. Theapplied primary antibodies mouse monoclonal Cy3-conjugatedanti-aSMA(clone1A4, Sigma–Aldrich), rabbit polyclonal anti-colla-gen I (Abcam, Cambridge, UK), mouse monoclonal anti-fibronectin(IST-9, Abcam) and rabbit monoclonal anti-Smad2/3 (D7G7, CellSignaling Technology, Danvers, MA) were combined with thesecondary antibodies goat anti-rabbit–Alexa Fluor 568 or goatanti-mouse–Alexa Fluor 568, respectively (Invitrogen, Darmstadt,Germany). Appropriate negative controls were provided by theincubation of parallel samples with the respective isotype matchedcontrols: mouse IgG2a (BD Biosciences), mouse IgG (AntibodiesOnline GmbH, Aachen, Germany), rabbit IgG (Antibodies OnlineGmbH) (Supplementary Fig. 1). Cytoskeletal F-actin and nucleiwere stained with phalloidin–Alexa Fluor 488 (Invitrogen) and4’,6-diamidino-2-phenylindole hydrochloride (DAPI) (Merck,Darmstadt, Germany), respectively. Samples were mounted withMobiglow (MoBiTec GmbH, Göttingen, Germany). Microscopywas performed with a Keyence BZ-9000E and the correspondingsoftware BZ-II Viewer and BZ-II Analyzer (Keyence, Neu-Isenburg,Germany).

2.5. TGFb1 bioassay

To quantify biologically active TGFb1 we used mink lung epithe-lial cells (MLEC) which had been stably transfected with a fragmentof human plasminogen activator inhibitor-1 (PAI-1) fused to aluciferase reporter gene [34]. MLEC were maintained in DMEMsupplemented with 10% FCS, 1% P/S and 200 lg ml�1 G418(Sigma–Aldrich) at 37 �C, 5% CO2. aECM was provided in White-F-Bottom 96-well plates (Nunc) and incubated with eitherDMEM/2% FCS or DMEM/2% FCS/2 ng ml�1 TGFb1 for 2 h at 37 �Cto enable growth factor adsorption. After a brief washing step withPBS, 1.5 � 104 MLEC per well were seeded with DMEM/2% FCS andincubated for 24 h at 37 �C, 5% CO2. Compared with the untreatedcontrol (MLEC in contact with aECM without TGFb1) and aECM pre-incubated with TGFb1, the third experimental group was directlystimulated with 2 ng ml�1 TGFb1. Supernatant was removed andthe wells were washed three times with PBS before cooled lysiswith 1 � Cell Culture Lysis Reagent (Luciferase Assay System,Promega Corp.) for 20 min. Chemiluminescence was measuredimmediately after addition of Luciferase Assay Substrate. The rawdata were blanked against the appropriate raw data for controls(MLEC cultured on aECM without TGFb1). Fold induction was cal-culated with reference to equally treated coll controls.

2.6. Quantification of aECM-adsorbed TGFb1

The amount of aECM-adsorbed TGFb1 was calculated by sub-traction of the quantified non-adsorbed TGFb1 from the appliedamount of 10 ng ml�1 TGFb1. This quantification was made by ELI-SA with TGFb1 capture antibody MAB240 (R&D), which wasblocked with 1% BSA (w/v) and 5% (w/v) sucrose (Roth) beforeincubation with supernatants. Detection was accomplished withthe biotinylated TGFb1-specific detection antibody BAF240 (R&D)

7778 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

in combination with streptavidin–horseradish peroxidase and thesubstrate 3,30,5,50-tetramethylbenzidine (Sigma–Aldrich). Thereaction was stopped with 1 M H2SO4 and absorbance was mea-sured at 450 nm.

2.7. Molecular docking experiments

2.7.1. StructuresThe structures of HA tetrasaccharide derivatives were modeled

using the structure of a HA octasaccharide available in the ProteinData Bank (PDB i.d. 2BVK, NMR). Short GAGs are considered rea-sonable models to study GAG–protein interactions (in an analo-gous manner to small peptides for the study of protein–proteininteractions), which is reflected in the current scientific literatureand in GAG–protein complexes in the PDB in which short oligo-mers (average 4- to 8mer) have been used to study the molecularbases of GAG–protein recognition and to try to understand theirbiological activity. Besides, detailed cluster analysis of the dockingresults obtained for short GAGs may contribute to add importantinformation for the possible interpretation of putative bindingmodes of longer GAGs [35], i.e. the partial overlap (i.e. superimpo-sition of equivalent periodic units) of members within the samecluster and between clusters may be interpreted as the elongationdirection of the oligochain and may therefore be indicative of pos-sible binding modes of longer GAG molecules.

The GAG geometries were optimized with the AMBER99 forcefield implemented in the Molecular Operating Environment(MOE) (Chemical Computing Group Inc., Quebec, Canada). The X-ray structure of TGFb1 used for computational analysis was ob-tained from the tertiary complex of TGFb1 with TGFbRI and TGFbRII(PDB i.d. 3KFD, 2.99 Å).

2.7.2. Docking of GAGsBlind docking experiments with flexible HA, HA4, HA6, HA46,

HA462’ and HA463’ tetrasaccharides (Table 2) to TGFb1 and tothe complex TGFb1/TGFbRII were carried out using Autodock 3[36]. The atomic potential grid of the proteins was calculated witha spacing of 0.8 Å. The Lamarckian genetic algorithm was used as asearch algorithm with an initial population size of 300 and a termi-nation condition of 10,000 generations or 999,500,000 energy eval-uations. A total of 1000 independent runs were performed. The top50 docking solutions were taken for further analysis.

2.7.3. ClusteringSpatial clustering of the docking solutions was performed using

the DBSCAN algorithm [37] with a neighborhood search radius e of4 Å. The distance between two structures was defined as the rootmean square of atomic distances, while pairing nearest atoms ofthe same type. This distance metric is invariant under symmetrytransformations applied to GAGs. For the further analysis we con-sidered clusters in the 50 top scored solutions having at least fourmembers.

Table 2Nomenclature of HA derivatives used for molecular dockingexperiments.

GAG abbreviation Description

HA (–GlcU–GlcNAc–)n

HA4 (–GlcU–GlcNAc4S–)n

HA6 (–GlcU–GlcNAc6S–)n

HA46 (–GlcU–GlcNAc4S6S–)n

HA462’ (–GlcU2S–GlcNAc4S6S–)n

HA463’ (–GlcU3S–GlcNAc4S6S–)n

GlcU, glucuronic acid; GlcNAc, N-acetylglucosamine.

2.8. Statistical analysis

All presented data were derived from primary cells from at leastthree independent donors. Experiments were performed in tripli-cate or quadruplicate. Results are presented as means ± standarderrors of the mean. Data were analyzed using the Mann–Whitneytest.

3. Results

We investigated the impact of an aECM containing hsHA on thebiological effects of TGFb1, a key factor for MFb differentiationwhich is essential for wound closure. We focused on the well-known induction of several marker genes of MFb which are in-duced by TGFb1.

3.1. Effect of hsHA-containing aECM on TGFb1-stimulated expressionof MFb marker genes aSMA, collagen and ED-A FN

Since aSMA is an accepted marker for differentiated MFb [6] in-duced by TGFb1 we analyzed aSMA mRNA expression in the pres-ence of aECM (Fig. 1A, white bars), aECM with adsorbed TGFb1

(Fig. 1A, grey bars) and aECM with directly added soluble TGFb1

(10 ng ml�1, Fig. 1A, black bars) within 72 h. mRNA expressionanalysis by qRT-PCR revealed down-regulation of aSMA mRNA le-vel for coll/hsHA in comparison with coll and non-sulfated coll/HAwhen not stimulated with TGFb1 (Fig. 1A, white bars). Stimulationwith adsorbed or soluble TGFb1 generally shows at least a 2-foldincrease in mRNA expression in comparison with non-stimulatedcontrols. Dermal fibroblasts growing on coll/hsHA exhibit a re-duced aSMA mRNA level in comparison with the non-sulfated ma-trix coll/HA when stimulated with aECM-adsorbed TGFb1 and asignificant reduction after directly applied TGFb1 (Fig. 1A, greyand black bars). The TGFb1-induced cytoplasmic aSMA should beincorporated into stress fibers [5,38], which impart high contractileactivity due to the presence of aSMA in vivo and in vitro [39,40].Here, the presence and fibrous organization of aSMA (red) at theprotein level after 72 h was investigated by immunofluorescencestaining (Fig. 1B) and supports the mRNA data. Representative im-age sections show that stimulation with soluble TGFb1 increasesaSMA expression and incorporation into stress fibers on all inves-tigated matrices except for coll/hsHA, which shows similar levelsto unstimulated controls. aECM-adsorbed TGFb1 results in in-creased expression but less pronounced integration of aSMA intostress fibers. The comparison of fluorescence signals indicates im-paired TGFb1-driven aSMA protein expression on coll/hsHA con-taining aECM. Staining for F-actin (green, inset) and nuclei (blue)monitor regular dFb morphology.

Furthermore, the synthesis of coll is a crucial process during thelater stages of wound healing to refill the wound defect with nativeECM, which is mainly realized by TGFb1-stimulated differentiatedMFb [4,7]. We analyzed the collagen Ia1 mRNA expression of dFbcultured on aECM for 72 h in the absence of TGFb1 (Fig. 2A, whitebars) and in the presence of aECM-adsorbed (Fig. 2A, grey bars)or soluble TGFb1 (Fig. 2A, black bars). In resting dFb the cultureon coll/hsHA results in significantly decreased collagen I(a1) mRNAexpression compared with coll and the respective non-sulfatedaECM coll/HA (Fig. 2A, white bars). Also, after stimulation withaECM-adsorbed or directly added TGFb1 collagen Ia1 mRNA expres-sion is significantly decreased on coll/hsHA (Fig. 2A, grey and blackbars). Representative immunofluorescence staining for collagen I(red) after 72 h (Fig. 2B) support these mRNA data and show thatstimulation with 10 ng ml�1 TGFb1 generally leads to increased col-lagen I deposition, while collagen I staining of dFb grown on coll/hsHA is reduced in comparison with dFb on coll and coll/HA. In

Fig. 1. aECM with hsHA prevents TGFb1-stimulated aSMA expression at the mRNA and protein levels. (A) qRT-PCR data for dFb cultured on aECM for 72 h in the absence ofTGFb1 (white bars), the presence of aECM-adsorbed (grey bars) or soluble TGFb1 (10 ng ml�1) (black bars). Fold induction refers to the untreated coll control. Error barsrepresent the means ± standard errors of the mean. ⁄p < 0.05; ⁄⁄p < 0.01; n = 4. (B) Representative immunofluorescence staining for aSMA (red), nuclei (blue) and F-actin(green, insets). Bar 50 lm (magnification 60�).

A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786 7779

contrast, cells cultured on aECM with adsorbed TGFb1 show lesspronounced collagen I expression on all aECMs (Fig. 2B).

Moreover, MFb can be characterized by the expression of ED-AFN, a specific splice variant of fibronectin [6,40], which is alsoimportant in promoting further MFb differentiation [39]. As forthe other TGFb1-regulated MFb marker genes, we found down-reg-ulation of the ED-A FN mRNA level after 72 h for dFb grown on coll/hsHA in unstimulated (Fig. 3A, white bars) or TGFb1-stimulated(Fig. 3A, grey and black bars) samples in comparison with dFbgrown on coll and coll/HA. Regarding the differences from theappropriate non-sulfated matrix coll/HA, the ED-A FN mRNAexpression level is significantly reduced when dFb were culturedon coll/hsHA. The presence of ED-A FN at the protein level after72 h was investigated by immunofluorescence staining and simi-larly revealed a reduction in ED-A FN expression for dFb culturedon coll/hsHA (Fig. 3B).

These results indicate that aECM with hsHA hinders TGFb1-stimulated aSMA, collagen and ED-A FN expression at the mRNAand protein levels in comparison with non-sulfated aECMs.

3.2. Mechanism of reduced TGFb1 activity in the presence of hsHAcontaining aECM

3.2.1. Adsorption of TGFb1 to aECMSince the interaction of TGFb1 with several GAGs, like heparin

[41] and sulfated HA, has recently been published [42] we mea-sured the ability of aECM to bind TGFb1. To quantify the amountof aECM-adsorbed TGFb1 within 2 h at 37 �C non-adsorbed TGFb1

was measured by ELISA and the resulting amount of adsorbedTGFb1 per culture area calculated (Fig. 4A). Coll/hsHA bound 2.4-fold more TGFb1 than the non-sulfated aECM coll/HA, which sug-gests a critical role of HA sulfation in TGFb1 binding. Althoughcoll/hsHA bound significantly more TGFb1, the differentiation ofdFb towards the MFb phenotype was clearly inhibited, as shownfor aSMA, collagen and ED-A FN expression (Figs. 1–3).

3.2.2. TGFb1. bioactivity and downstream signalingTo test the bioactivity of aECM-adsorbed TGFb1 the growth fac-

tor-mediated signal transduction was analyzed. The quantification

Fig. 2. TGFb1-stimulated collagen I expression is reduced on aECM with hsHA at the mRNA and protein levels. (A) mRNA expression analysis for collagen I(a1) of dFb culturedon aECM for 72 h in the absence of TGFb1 (white bars), or the presence of aECM-adsorbed (grey bars) or soluble 10 ng ml�1 TGFb1 (black bars). Fold induction refers to theuntreated coll control. Error bars represent the means ± standard errors of the mean. ⁄p < 0.001; n = 4. (B) Representative immunofluorescence staining for collagen I (red),nuclei (blue) and F-actin (green, insets). Bar 100 lm (magnification 20�).

7780 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

of bioactive TGFb1 was measured by MLEC-mediated chemilumi-nescence in response to active growth factor [34]. MLEC were cul-tured on aECM for 24 h in the presence of adsorbed TGFb1 ordirectly added soluble TGFb1 (2 ng ml�1). To ensure the sameamounts of adsorbed TGFb1 on all aECMs lower amounts of TGFb1

(2 ng ml�1) were applied. After 2 h at 37 �C remaining free TGFb1

was detected by ELISA. As shown in Fig. 4B, of the initial amountof TGFb1 about 0.43 ± 0.065 ng cm�2 is equally bound to all aECMs.This concentration lies within the linear range of MLEC response toTGFb1 (Supplementary Fig. 2). The incubation of MLEC on coll/hsHA with adsorbed TGFb1 (Fig. 4C) reveals a reduced responsein comparison with non-sulfated aECMs with adsorbed TGFb1. Di-rect stimulation with soluble TGFb1 (Fig. 4D) shows similar effects.

To show that aECM with hsHA interferes with TGFb1-mediatedsignal transduction in primary dFb Smad translocation on aECMswas studied. Intracellular signal transduction of TGFb1-stimulatedcells involves the phosphorylation of Smad2 and Smad3 and the

formation of a heteromeric Smad2/3 complex in the cytoplasm.Subsequently this complex is translocated to the nucleus to regu-late gene transcription [12,43]. Immunofluorescence staining forSmad2/3 (red) of dFb cultured on aECM for 5 h shows clear trans-location of cytoplasmic Smad2/3 to the nucleus when stimulatedwith 5 ng ml�1 TGFb1 for non-sulfated aECMs coll and coll/HA(Fig. 4E). In contrast, TGFb1 stimulation of dFb cultured on coll/hsHA reveals massively reduced Smad2/3 translocation to the nu-cleus. The reduced response of MLEC to TGFb1 in combination withhsHA and the failure of nuclear translocation of Smad2/3 in dFbincubated with coll/hsHA and TGFb1 suggest impaired bioactivityof the growth factor in the presence of hsHA.

3.2.3. Interaction of TGFb1 and highly sulfated HAWe investigated the possible molecular mechanisms of inhibi-

tion of TFGb1 activity by coll/hsHA. In silico docking experimentssuggest an interaction of hsHA with TGFb1. Fig. 5A shows a

Fig. 3. aECM with hsHA prevents TGFb1-stimulated ED-A FN expression at the mRNA and protein levels. dFb were cultivated on aECM for 72 h in the absence of TGFb1, thepresence of aECM-adsorbed or soluble TGFb1 (10 ng ml�1). (A) mRNA expression of ED-A FN for untreated samples (white bars), or dFb exposed to aECM-adsorbed (grey bars)or soluble TGFb1 (black bars), respectively. Fold induction refers to the untreated coll control. Error bars represent the means ± standard errors of the mean. ⁄p < 0.001; n = 4.(B) Representative immunofluorescence staining for ED-A FN (red), nuclei (blue) and F-actin (green, insets). Bar 50 lm (magnification 60�).

A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786 7781

diagrammatic representation of the crystallographic structure ofthe complex TGFb1/TGFbRI/TGFbRII (PDB i.d. 3KFD) and the struc-tures of HA and HA463’ tetrasaccharides docked on the surface ofTGFb1. The docking results obtained show the following.

i. TGFb1 binds more strongly to HA derivatives with a higherdegree of sulfation.

ii. There is a significant difference in the distribution of thedocked GAGs on the surface of TGFb1 depending on theirsulfation pattern. In particular, we observed that thehigher the sulfation of the HA derivative, the more prob-able it is that it occupies the TGFbRI binding site onTGFb1.

iii. We did not observe different orientations for the dockedGAGs, but rather an ‘‘elongation direction’’ for the putativebinding sites described, which could be indicative of the pos-sible binding mode for longer GAGs.

For one of the clusters obtained for HA (upper cluster in Fig. 5A)we observe that residue K13 establishes key interactions. Further-more, we observe that the GAG molecules in the cluster fit into thesurface shaped by two loop regions in TFGb1 (residues 46–50 and69–73), which are not positively charged. For the other cluster ob-tained for HA (lower cluster in Fig. 5A) we observe two key inter-acting residues, R107 and K110. In the case of sulfated HA463’ thedocking solutions obtained overlap with the so-called ‘‘finger helixcavity’’ at the TFGbRI interface and establish key interactions withthe basic residues K31, K60, H68 and K97 (Supplementary Fig. 3). Itis known that before the recruitment of TGFbRI, TGFbRII is requiredto bind to TGFb1 [12]. Therefore, we also performed moleculardocking calculations for HA derivatives on the complex TGFb1/TGFbRII. The results obtained show that the most probable dockingsolutions are in the same regions found when docking withoutTGFbRII, indicating that receptor RII binding to TGFb1 has no effecton occupancy of the studied recognition sites for GAGs.

Fig. 4. aECM with hsHA adsorbs the highest amount of TGFb1 but impairs TGFb1 bioactivity and prevents downstream signaling. (A) The adsorption of TGFb1 on aECM within2 h at 37 �C was determined by quantification of non-adsorbed TGFb1 and subtraction of the initial TGFb1 concentration of 2.08 ng cm�2 (10 ng ml�1). (B) Equal adsorption of�0.43 ng cm�2 (�2 ng ml�1) TGFb1 to the aECM ensuring a TGFb1 concentration within the linear range of the MLEC reporter assay. (C, D) Chemiluminescence referring toPAI-1 promoter activation as an indicator of bioactive TGFb1 was determined. MLECs were cultured on aECM for 24 h in the presence of adsorbed TGFb1 (C) or after addition ofsoluble TGFb1 (2 ng ml�1 = 0.625 ng cm�2) (D). Fold induction refers to the equally treated coll control. Error bars represent the means ± standard errors of the mean.⁄⁄p < 0.01, n = 3. (E) Immunofluorescence staining for Smad2/3 (red) of dFb cultured on aECM for 5 h. Bar 100 lm (magnification 20�); inset bars 50 lm.

7782 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

To test this in silico model experimentally dFb were exposed toTGFb1 and either soluble LMW-HA, hsHA or combinations of thegrowth factor with these GAGs. Before supplemention of dFb thecomponents were preincubated for 1 h to allow TGFb1–GAG inter-action. To exclude any influence of the total MW on the interactionwith TGFb1 we decided to apply LMW-HA, which has a similar MWto hsHA. mRNA expression analysis for aSMA, collagen I(a1) andED-A FN (Fig. 5B) shows a reduction in mRNA levels when TGFb1-stimulated samples were simultaneously treated with hsHA(Fig. 5B, grey bars). Furthermore, preincubation of hsHA withTGFb1 impaired Smad2/3 translocation in dFb, while LMW-HAdid not (Fig. 5C). Data obtained from computational and cell bio-

logical analyses indicate that hsHA hinders dFb to MFb differenti-ation, probably by occupation of the TGFbRI binding site onTGFb1, thus preventing receptor binding, signal transduction andexpression of aSMA, collagen I(a1) and ED-A FN mRNA.

4. Discussion

Due to the synthesis of new ECM and their contractile activityMFb have an essential role during the later stages of wound healingwhen the wound defect has to be filled and contracted to close thewound [6,44]. In dermal wound healing the contractile activity of

Fig. 5. Highly sulfated HA prevents TGFb1-mediated signal transduction and induction of aSMA, collagen I(a1) and ED-A FN mRNA expression by occupation of the TGFbRIbinding site. (A) Molecular docking of HA and HA463’ tetrasaccharides on the surface of TGFb1. The structure of the complex TGFb1/TGFbRI/TGFbRII (PDB i.d. 3KFD) is shownin diagrammatic representation (TGFb1 in yellow, TGFbRI in dark grey and TGFbRII in light grey). Clustering analysis was performed for the top 50 scoring docking results(Supplementary Table S2). The two clusters obtained for HA (12 and 5 members, respectively) are shown as red sticks and the two clusters obtained for HA463’ (11 and 6members, respectively) are shown as blue sticks. (B) mRNA expression analysis for aSMA, collagen I(a1) and ED-A FN were performed for dFb cultivated for 72 h in theabsence of TGFb1 and GAGs (control, white bars) or with addition of 10 ng ml�1 TGFb1, 50 lg ml�1 LMW-HA, 50 lg ml�1 hsHA or combinations of growth factor with GAGs.Fold induction refers to the untreated control. Error bars represent the means ± standard errors of the mean. ⁄p < 0.05; ⁄⁄p < 0.01; n = 3. (C) Immunofluorescence staining ofdFb cultured for 5 h with 5 ng ml�1 TGFb1, 50 lg ml�1 LMW-HA preincubated with 5 ng ml�1 TGFb1 or 50 lg ml�1 hsHA preincubated with 5 ng ml�1 TGFb1 to analyzenuclear translocation of Smad2/3 (red). Bar 100 lm (magnification 20�).

A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786 7783

MFb is terminated when the tissue is repaired. Then aSMA expres-sion decreases and MFb are removed by apoptosis [45]. However,in pathological wound healing MFb activity persists and leads totissue deformation resulting in hypertrophic scars or even keloids[46]. TGFb1 is considered to be the major growth factor directlypromoting MFb differentiation [5,38] and is locally released by res-ident [3] and inflammatory cells, especially by macrophages [47]and platelets [48]. Although enormous advances in the develop-ment of skin substitutes have occurred in the past three decades,

there are still major obstacles to be overcome in the quest for anoptimal skin substitute, which includes controlling MFb differenti-ation and scar formation [49]. Continuing our investigations onsulfated GAGs as growth-promoting ECM substitutes for dFb [29],in this study we examined whether sulfated HA has an impacton the biological effects of TGFb1 with respect to the inductionof marker genes for MFb. Our results from qRT-PCR andimmunofluorescence staining show that hsHA impairs the TGFb1

stimulation of MFb markers at the mRNA and protein levels.

7784 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

Immunofluorescence staining does not provide quantitative dataand consequently best illustrates strong differences in proteinexpression. However, the regulation of mRNA expression can beconfirmed as a tendency.

In the absence of TGFb1 dFb grown on aECM with hsHA showdecreased basal levels of MFb marker gene expression, which isin line with our previous data [29]. Although TGFb1 stimulationcan be observed on all matrices, the aECM with hsHA hindersTGFb1-stimulated expression of MFb marker genes stronger thannon-sulfated aECM. The decrease in TGFb1-driven induction ofgene expression can be observed independently of the method ofpresentation of TGFb1 in the culture system, suggesting that aninteraction of hsHA and TGFb1 might take place both in the super-natant and on the aECM coating. Investigations of these aECM withhuman mesenchymal stem cells (hMSC) showed that coll synthesiswas promoted by hsHA containing aECM with adsorbed TGFb1,which is in contrast to our results [31]. Although, specific receptorshave been found on nearly all mammalian cells, the effects ofTGFb1 differ according to cell type, growth conditions and concen-tration of growth factor [50] and are the subject of complex regu-lation. The effects of TGFb1 may depend on the method ofrecruitment from the latent complex in vivo, on orchestration withco-stimulating proteins or on the spatio-temporal binding se-quence to different receptors on different cell types.

TGFb1 can induce different signal cascades: signaling throughp38 MAPK results in induction of collagen synthesis whereas Smadsignaling is required for MFb differentiation in several cell types[51–54]. Interestingly, for another protein of the TGFb family, bonemorphogenetic protein-2 (BMP-2), it has been shown that theresulting signal depends on whether the cytokine binds to RI priorto RII engagement (resulting in p38 MAPK signaling) or whetherBMP-2 binds to the preformed receptor complex of RI and RII(resulting in Smad signaling) [55].

Further, additional proteins may be differently involved ingenerating the effects on the different cell types. Thus Grotendorstet al. showed in their pioneering work that TGFb1 may initiate pro-liferative or differentiating responses in rat kidney fibroblastsdepending on the co-mediators. Insulin-like growth factor-2 wasnecessary for MFb differentiation and epidermal growth factorco-stimulated proliferation with TGFb1 [56]. However, the twoprocesses were not observed simultaneously, suggesting two alter-native, discrete signaling programs.

Other authors have demonstrated that secreted protein acidicand rich in cysteine (SPARC) can interfere with TGFb1 signaling inmesangial cells. SPARC binding to TGFbRII may block cJun-relatedpathways and direct the cells to Smad signaling [57]. Together withthe observations of Choi et al. [58] showing that SPARC expressionin hMSC may depend on the growth substrate, it is conceivablethat TGFb1 can exert different responses in dFb and hMSC on differ-ent substrates.

From the receptor site, the proteoglycan content of TGFbRIIImay determine the efficiency of TGFb1 signaling. TGFbRIII also pre-sents TGFb1 to TGFbRII and augments TGFb1 responses in myo-blasts [59]. However, its activity in supporting the TGFb1 signaldepends on the number and size of proteoglycans in the molecule.Large amounts or large proteoglycan molecules like heparan sul-fate may interfere with the presentation of TGFb1 to TGFbRII,whereas removal of heparan sulfate augmented the TGFb1 re-sponse [60]. Although these topics have not yet been addressedin the context of aECM with sulfated HA, it is conceivable thatTGFb1 could have different effects on various cell types.

When measuring the amount of aECM-adsorbed TGFb1 wecould show 2.4-fold more TGFb1 bound to aECM with hsHA in com-parison with non-sulfated aECM. These data suggest that sulfategroups significantly increase TGFb1 binding to HA. Indeed, adsorp-tion of TGFb1 to aECM and its subsequent release depend on the

degree of HA sulfation [31]. In a detailed study by Hintze et al.[42] it was shown by surface plasmon resonance and ELISA thatpure hsHA exhibits the tightest interaction with human recombi-nant TGFb1 when compared with HA derivatives with a lower de-gree of sulfation. Our data indicate that this also holds true forhsHA in collagenous matrices.

Naturally occurring sulfated GAGs play critical roles in cellularprocesses by binding and regulating a remarkable number ofgrowth factors [61]. Heparin is known for its ability to stabilizethe binding between fibroblast growth factor-2 (FGF-2) and itsreceptor and promoting receptor dimerization [28,62]. Addition-ally, heparan sulfate can bind FGF-2 on the cell surface to protectit from proteolytic degradation and generate a reservoir of growthfactor [63–65]. Furthermore, FGF-10 and dermatan sulfate (DS)markedly enhanced migration of keratinocytes in an in vitro woundscratch assay, while heparin sulfate and chondroitin sulfates (CS)did not. These data strongly suggest that specific DS structuralproperties are necessary to promote FGF-10 function [66]. Fromthese examples we conclude that the design of sulfated GAG mightbe a way to generate materials with desired biological functions.

Investigations of the native highly sulfated GAGs heparin andheparin sulfate show a potentiation of TGFb1 biological activity[41,67–69] and also chemically sulfated dextrans enhance the re-sponse of MLEC to TGFb1 [70], suggesting that highly sulfatedGAG might positively influence the activity of this growth factor.However, our data for dFb indicate that the hsHA acts in an oppo-site manner with regard to MFb differentiation. The degree of sul-fation, the position of sulfated groups and the structure of thesugar backbone have been shown to influence the GAG–TGFb1

interaction [18,42] and putatively result in various biologicaleffects.

Although we are aware that besides TGFb1 mechanical forcesare also important for MFb formation [6,14], we have here focusedon the mechanism behind the attenuated activity of TGFb1 in thepresence of hsHA. The major pathway through which TGFb1 regu-lates, for instance, the expression of aSMA is the Smad signalingpathway. Binding of active TGFb1 to TGFb1RII and TGFbRI resultsin translocation of the Smad2/3 complex to the nucleus, wheregene transcription is regulated [5,12,43]. Our data obtained withTGFb1-exposed MLEC reporter cells on aECMs showing abrogatedtranslocation of the Smad2/3 complex to the nucleus strongly sug-gest an impaired bioavailability of TGFb1 in the presence of hsHA.

Further reasons for the lack of signal transduction by TGFb1 areconceivable (e.g. modified TGFbR expression on fibroblasts grow-ing on aECM with hsHA or degradation of TGFb1 by aECM-immobi-lized proteases). However, we suppose that the accessibility ofTGFb1 to receptors might be the reason for the effects observed.Detailed atomic data obtained from docking experiments indicatedthat hsHA occupies the TGFbRI binding site on TGFb1 and thus itmight prevent functional signaling resulting in the induction ofaSMA, collagen I(a1) and ED-A FN expression in dFb. Althoughthe TGFb1 interface with TGFbRII is more basic than the interfacewith TGFbRI [71], our docking results reveal a clear preference ofHA463’ for the TGFbRI site. This preference for the RI site mightbe due to the fact that the basic residues in the RII site (i.e. R25,K31, H34 and R94) are less favorably distributed to optimally rec-ognize HA463’. In addition, residue K31, one of the anchoringpoints for HA463’ on TGFb1, is located in the interface of the ter-nary complex that is part of both the RI and RII sites. Because ofthis close proximity of the RI and RII binding sites it could be sug-gested that with elongation of the docked GAG molecules the RIIsite could be partially impaired. This model is supported by exper-imental data from blocking experiments with LMW-HA and hsHA.Preincubation of TGFb1 with hsHA but not with LMW-HA blocksTGFb1-driven Smad2/3 translocation and induction of gene expres-sion. Since this was not the case for LMW-HA, the experimental

A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786 7785

data support the in silico model and data of Hintze et al. [42] whoshowed that TGFb1 functionally interacts predominantly withhsHA molecules.

Our findings suggest that the introduction of hsHA in biomate-rial coatings could modify wound healing processes by selectedinteractions with TGFb1 and putatively also other growth factors.Furthermore, a putative use of hsHA is conceivable for alternativeanti-fibrotic strategies in order to prevent the development of tis-sue contracture and excessive collagen deposition leading tohypertrophic scars or keloids. Consequently, hsHA, either as a com-ponent of biomaterials or pharmaceutical agent, might offer apromising option to interfere with TGFb1-driven fibrosis in variouslocal fibrotic disorders.

5. Conclusion

Based on previous data showing that aECM consisting of colland HA induces proliferation of dFb in a sulfation-dependent man-ner, while ECM synthesis was reduced [31], we analyzed the im-pact of this aECM on TGFb1-driven expression of MFb markergenes. We found that the bioactivity of TGFb1 was significantly re-duced by hsHA. Our data suggest that hsHA prevents TGFb1 signal-ing by occupation of the TGFbRI binding site and thereby reducesTGFb1-driven MFb differentiation. Hence, hsHA seems to be aninteresting option to modify TGFb1-driven effects or even to inter-fere with fibrotic processes. In the light of the intended use of suchan aECM for non-healing wounds possibly adjustment of the HAsulfation grade could result in improved matrices, since prolifera-tion of dFb as well as MFb differentiation are essential steps duringdermal wound healing.

Acknowledgements

This work was supported by grants from the Deutsche Fors-chungsgemeinschaft to D.S., M.S., M.T.P., and U.A. (SFB/TR67 pro-jects A2, A3, A7 and B4) and of the European Union (ESF) to U.A.The authors like to thank Annett Majok, Sibylle Vorberg, Alina Mir-on, Aline Katzschner and Ulf Markwardt for excellent technicalassistance. Furthermore, we greatly appreciate the support ofDr. Daniel B. Rifkin (New York University Medical Center) who pro-vided the MLEC cell line. We would like to thank the ZIH at Tech-nische Universität Dresden for providing computational resourcesand Ralf Gey for IT support.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.actbio.2013.04.023.

References

[1] Clark RA, Ghosh K, Tonnesen MG. Tissue engineering for cutaneous wounds. JInvest Dermatol 2007;127:1018–29.

[2] Hackam DJ, Ford HR. Cellular, biochemical, and clinical aspects of woundhealing. Surg Infect (Larchmt) 2002;3(Suppl. 1):S23–35.

[3] Werner S, Grose R. Regulation of wound healing by growth factors andcytokines. Physiol Rev 2003;83:835–70.

[4] Werner S, Krieg T, Smola H. Keratinocyte–fibroblast interactions in woundhealing. J Invest Dermatol 2007;127:998–1008.

[5] Hinz B. Formation and function of the myofibroblast during tissue repair. JInvest Dermatol 2007;127:526–37.

[6] Tomasek JJ, Gabbiani G, Hinz B, Chaponnier C, Brown RA. Myofibroblasts andmechano-regulation of connective tissue remodelling. Nat Rev Mol Cell Biol2002;3:349–63.

[7] Gabbiani G. The myofibroblast in wound healing and fibrocontractive diseases.J Pathol 2003;200:500–3.

[8] Darby I, Skalli O, Gabbiani G. Alpha-smooth muscle actin is transientlyexpressed by myofibroblasts during experimental wound healing. Lab Invest1990;63:21–9.

[9] Hinck AP, Archer SJ, Qian SW, Roberts AB, Sporn MB, Weatherbee JA, et al.Transforming growth factor beta 1: three-dimensional structure in solutionand comparison with the X-ray structure of transforming growth factor beta 2.Biochemistry 1996;35:8517–34.

[10] Annes JP, Munger JS, Rifkin DB. Making sense of latent TGFbeta activation. JCell Sci 2003;116:217–24.

[11] Miyazono K, Heldin CH. Latent forms of TGF-beta: molecular structure andmechanisms of activation. Ciba Found Symp 1991;157:81–9.

[12] Klass BR, Grobbelaar AO, Rolfe KJ. Transforming growth factor beta1 signalling,wound healing and repair: a multifunctional cytokine with clinicalimplications for wound repair, a delicate balance. Postgrad Med J2009;85:9–14.

[13] Shi Y, Massague J. Mechanisms of TGF-beta signaling from cell membrane tothe nucleus. Cell 2003;113:685–700.

[14] Wipff PJ, Rifkin DB, Meister JJ, Hinz B. Myofibroblast contraction activateslatent TGF-beta1 from the extracellular matrix. J Cell Biol 2007;179:1311–23.

[15] Kunze R, Rosler M, Moller S, Schnabelrauch M, Riemer T, Hempel U, et al.Sulfated hyaluronan derivatives reduce the proliferation rate of primary ratcalvarial osteoblasts. Glycoconj J 2010;27:151–8.

[16] Schiller J, Becher J, Moller S, Nimptsch K, Riemer T, Schnabelrauch M. Synthesisand characterization of chemically modified hyaluronan and chondroitinsulfate. Mini-Rev Org Chem 2010;7:290–9.

[17] Clark RA. Cutaneous tissue repair: basic biologic considerations. I. J Am AcadDermatol 1985;13:701–25.

[18] Hintze V, Moeller S, Schnabelrauch M, Bierbaum S, Viola M, Worch H, et al.Modifications of hyaluronan influence the interaction with human bonemorphogenetic protein-4 (hBMP-4). Biomacromolecules 2009;10:3290–7.

[19] Lavenus S, Pilet P, Guicheux J, Weiss P, Louarn G, Layrolle P. Behaviour ofmesenchymal stem cells, fibroblasts and osteoblasts on smooth surfaces. ActaBiomater 2011;7:1525–34.

[20] Sasisekharan R, Raman R, Prabhakar V. Glycomics approach to structure–function relationships of glycosaminoglycans. Annu Rev Biomed Eng2006;8:181–231.

[21] Taylor KR, Gallo RL. Glycosaminoglycans and their proteoglycans: host-associated molecular patterns for initiation and modulation of inflammation.FASEB J 2006;20:9–22.

[22] Gandhi NS, Mancera RL. The structure of glycosaminoglycans and theirinteractions with proteins. Chem Biol Drug Des 2008;72:455–82.

[23] Hynes RO. The extracellular matrix: not just pretty fibrils. Science2009;326:1216–9.

[24] Salek-Ardakani S, Arrand JR, Shaw D, Mackett M. Heparin and heparan sulfatebind interleukin-10 and modulate its activity. Blood 2000;96:1879–88.

[25] Gama CI, Tully SE, Sotogaku N, Clark PM, Rawat M, Vaidehi N, et al. Sulfationpatterns of glycosaminoglycans encode molecular recognition and activity.Nat Chem Biol 2006;2:467–73.

[26] Lamanna WC, Kalus I, Padva M, Baldwin RJ, Merry CL, Dierks T. Theheparanome – the enigma of encoding and decoding heparan sulfatesulfation. J Biotechnol 2007;129:290–307.

[27] Tanino Y, Coombe DR, Gill SE, Kett WC, Kajikawa O, Proudfoot AE, et al.Kinetics of chemokine–glycosaminoglycan interactions control neutrophilmigration into the airspaces of the lungs. J Immunol 2010;184:2677–85.

[28] Coombe DR. Biological implications of glycosaminoglycan interactions withhaemopoietic cytokines. Immunol Cell Biol 2008;86:598–607.

[29] van der Smissen A, Hintze V, Scharnweber D, Moeller S, Schnabelrauch M,Majok A, et al. Growth promoting substrates for human dermal fibroblastsprovided by artificial extracellular matrices composed of collagen I andsulfated glycosaminoglycans. Biomaterials 2011;32:8938–46.

[30] Moeller S, Schmidtke M, Weiss D, Schiller J, Pawlik K, Wutzler P, et al.Synthesis and antiherpetic activity of carboxymethylated and sulfatedhyaluronan derivatives. Carbohydr Polym 2012;90:608–15.

[31] Hempel U, Hintze V, Moller S, Schnabelrauch M, Scharnweber D, Dieter P.Artificial extracellular matrices composed of collagen I and sulfatedhyaluronan with adsorbed transforming growth factor beta1 promotecollagen synthesis of human mesenchymal stromal cells. Acta Biomater2012;8:659–66.

[32] Saalbach A, Klein C, Schirmer C, Briest W, Anderegg U, Simon JC. Dermalfibroblasts promote the migration of dendritic cells. J Invest Dermatol2010;130:444–54.

[33] Vincent S, Marty L, Fort P. S26 ribosomal protein RNA: an invariant control forgene regulation experiments in eucaryotic cells and tissues. Nucleic Acids Res1993;21:1498.

[34] Abe M, Harpel JG, Metz CN, Nunes I, Loskutoff DJ, Rifkin DB. An assay fortransforming growth factor-beta using cells transfected with a plasminogenactivator inhibitor-1 promoter–luciferase construct. Anal Biochem1994;216:276–84.

[35] Kadirvelraj R, Gonzalez-Outeirino J, Foley BL, Beckham ML, Jennings HJ, FooteS, et al. Understanding the bacterial polysaccharide antigenicity ofStreptococcus agalactiae versus Streptococcus pneumoniae. Proc Natl Acad SciUSA 2006;103:8149–54.

[36] Morris GM, Goodsell DS, Huey R, Olson AJ. Distributed automated docking offlexible ligands to proteins: parallel applications of AutoDock 2.4. J ComputAided Mol Des 1996;10:293–304.

7786 A. van der Smissen et al. / Acta Biomaterialia 9 (2013) 7775–7786

[37] Ester M, Kriegel HP, Sander J, Xu X. A density-based algorithm for discoveringclusters in large spatial databases with noise. In: Proceedings of 2ndInternational Conference on Knowledge Discovery and Data Mining (KDD-96), 1996.

[38] Desmouliere A, Geinoz A, Gabbiani F, Gabbiani G. Transforming growth factor-beta 1 induces alpha-smooth muscle actin expression in granulation tissuemyofibroblasts and in quiescent and growing cultured fibroblasts. J Cell Biol1993;122:103–11.

[39] Serini G, Gabbiani G. Mechanisms of myofibroblast activity and phenotypicmodulation. Exp Cell Res 1999;250:273–83.

[40] Vaughan MB, Howard EW, Tomasek JJ. Transforming growth factor-beta1promotes the morphological and functional differentiation of themyofibroblast. Exp Cell Res 2000;257:180–9.

[41] McCaffrey TA, Falcone DJ, Du B. Transforming growth factor-beta 1 is aheparin-binding protein: identification of putative heparin-binding regionsand isolation of heparins with varying affinity for TGF-beta 1. J Cell Physiol1992;152:430–40.

[42] Hintze V, Miron A, Moeller S, Schnabelrauch M, Wiesmann HP, Worch H, et al.Sulfated hyaluronan and chondroitin sulfate derivatives interact differentlywith human transforming growth factor-beta1 (TGF-beta1). Acta Biomater2012;8:2144–52.

[43] Heldin CH, Miyazono K, ten Dijke P. TGF-beta signalling from cell membrane tonucleus through SMAD proteins. Nature 1997;390:465–71.

[44] Sarrazy V, Billet F, Micallef L, Coulomb B, Desmouliere A. Mechanisms ofpathological scarring: role of myofibroblasts and current developments.Wound Repair Regen 2011;19(Suppl. 1):s10–5.

[45] Desmouliere A, Redard M, Darby I, Gabbiani G. Apoptosis mediates thedecrease in cellularity during the transition between granulation tissue andscar. Am J Pathol 1995;146:56–66.

[46] Schurch W, Seemayer TA, Hinz B, Gabbiani G. The Myofibroblast. In: Mills SE,editor. Histology for Pathologists, Philadelphia: Lippincott-Williams & WilkinsPublishers; 2007. p. 123–164.

[47] Sindrilaru A, Peters T, Wieschalka S, Baican C, Baican A, Peter H, et al. Anunrestrained proinflammatory M1 macrophage population induced by ironimpairs wound healing in humans and mice. J Clin Invest 2011;121:985–97.

[48] Assoian RK, Komoriya A, Meyers CA, Miller DM, Sporn MB. Transforminggrowth factor-beta in human platelets. Identification of a major storage site,purification, and characterization. J Biol Chem 1983;258:7155–60.

[49] Nolte SV, Xu W, Rennekampff HO, Rodemann HP. Diversity of fibroblasts – areview on implications for skin tissue engineering. Cells Tissues Organs2008;187:165–76.

[50] Massague J. TGF-beta signal transduction. Annu Rev Biochem1998;67:753–91.

[51] Gu L, Zhu YJ, Yang X, Guo ZJ, Xu WB, Tian XL. Effect of TGF-beta/Smad signalingpathway on lung myofibroblast differentiation. Acta Pharmacol Sin2007;28:382–91.

[52] Hu C, Dandapat A, Sun L, Khan JA, Liu Y, Hermonat PL, et al. Regulation ofTGFbeta1-mediated collagen formation by LOX-1: studies based on forcedoverexpression of TGFbeta1 in wild-type and lox-1 knock-out mouse cardiacfibroblasts. J Biol Chem 2008;283:10226–31.

[53] Kim SI, Kwak JH, Zachariah M, He Y, Wang L, Choi ME. TGF-beta-activatedkinase 1 and TAK1-binding protein 1 cooperate to mediate TGF-beta1-inducedMKK3-p38 MAPK activation and stimulation of type I collagen. Am J PhysiolRenal Physiol 2007;292:F1471–8.

[54] Urtasun R, Lopategi A, George J, Leung TM, Lu Y, Wang X, et al. Osteopontin, anoxidant stress sensitive cytokine, up-regulates collagen-I via integrinalpha(V)beta(3) engagement and PI3K/pAkt/NFkappaB signaling. Hepatology2012;55:594–608.

[55] Hauburger A, von ES, Schwaerzer GK, Buttstedt A, Zebisch M, Schraml M, et al.The pro-form of BMP-2 interferes with BMP-2 signalling by competing withBMP-2 for IA receptor binding. FEBS J 2009;276:6386–98.

[56] Grotendorst GR, Rahmanie H, Duncan MR. Combinatorial signaling pathwaysdetermine fibroblast proliferation and myofibroblast differentiation. FASEB J2004;18:469–79.

[57] Francki A, McClure TD, Brekken RA, Motamed K, Murri C, Wang T, et al. SPARCregulates TGF-beta1-dependent signaling in primary glomerular mesangialcells. J Cell Biochem 2004;91:915–25.

[58] Choi J, Park EJ, Shin HS, Shin IS, Ra JC, Koh KS. In vivo differentiation ofundifferentiated human adipose tissue-derived mesenchymal stem cells incritical-sized calvarial bone defects. Ann Plast Surg 2012.

[59] Blobe GC, Liu X, Fang SJ, How T, Lodish HF. A novel mechanism for regulatingtransforming growth factor beta (TGF-beta) signaling. Functional modulationof type III TGF-beta receptor expression through interaction with the PDZdomain protein, GIPC. J Biol Chem 2001;276:39608–17.

[60] Chen CL, Huang SS, Huang JS. Cellular heparan sulfate negatively modulatestransforming growth factor-beta1 (TGF-beta1) responsiveness in epithelialcells. J Biol Chem 2006;281:11506–14.

[61] Rogers CJ, Clark PM, Tully SE, Abrol R, Garcia KC, Goddard WA, et al.Elucidating glycosaminoglycan–protein–protein interactions usingcarbohydrate microarray and computational approaches. Proc Natl Acad SciUSA 2011;108:9747–52.

[62] Rusnati M, Coltrini D, Caccia P, Dell’Era P, Zoppetti G, Oreste P, et al. Distinctrole of 2-O-, N-, and 6-O-sulfate groups of heparin in the formation of theternary complex with basic fibroblast growth factor and soluble FGF receptor-1. Biochem Biophys Res Commun 1994;203:450–8.

[63] Gallagher JT, Turnbull JE. Heparan sulphate in the binding and activation ofbasic fibroblast growth factor. Glycobiology 1992;2:523–8.

[64] Nissen NN, Shankar R, Gamelli RL, Singh A, DiPietro LA. Heparin and heparansulphate protect basic fibroblast growth factor from non-enzymicglycosylation. Biochem J 1999;338(3):637–42.

[65] Richardson TP, Trinkaus-Randall V, Nugent MA. Regulation of basic fibroblastgrowth factor binding and activity by cell density and heparan sulfate. J BiolChem 1999;274:13534–40.

[66] Radek KA, Taylor KR, Gallo RL. FGF-10 and specific structural elements ofdermatan sulfate size and sulfation promote maximal keratinocyte migrationand cellular proliferation. Wound Repair Regen 2009;17:118–26.

[67] Chen Y, Shi-Wen X, van BJ, Kennedy L, McLeod M, Renzoni EA, et al. Matrixcontraction by dermal fibroblasts requires transforming growth factor-beta/activin-linked kinase 5, heparan sulfate-containing proteoglycans, and MEK/ERK: insights into pathological scarring in chronic fibrotic disease. Am J Pathol2005;167:1699–711.

[68] McCaffrey TA, Falcone DJ, Brayton CF, Agarwal LA, Welt FG, Weksler BB.Transforming growth factor-beta activity is potentiated by heparin viadissociation of the transforming growth factor-beta/alpha 2-macroglobulininactive complex. J Cell Biol 1989;109:441–8.

[69] McCaffrey TA, Falcone DJ, Vicente D, Du B, Consigli S, Borth W. Protection oftransforming growth factor-beta 1 activity by heparin and fucoidan. J CellPhysiol 1994;159:51–9.

[70] Logeart-Avramoglou D, Huynh R, Chaubet F, Sedel L, Meunier A. Interaction ofspecifically chemically modified dextrans with transforming growth factorbeta1: potentiation of its biological activity. Biochem Pharmacol2002;63:129–37.

[71] Radaev S, Zou Z, Huang T, Lafer EM, Hinck AP, Sun PD. Ternary complex oftransforming growth factor-beta1 reveals isoform-specific ligand recognitionand receptor recruitment in the superfamily. J Biol Chem 2010;285:14806–14.