at SciVerse ScienceDirect

Biomaterials 34 (2013) 7653e7661

Contents lists available

Biomaterials

journal homepage: www.elsevier .com/locate/biomater ia ls

Leading opinion

The promotion of osteoclastogenesis by sulfated hyaluronan throughinterference with osteoprotegerin and receptor activator of NF-kBligand/osteoprotegerin complex formation

Juliane Salbach-Hirsch a,1, Julia Kraemer b,1, Martina Rauner a, Sergey A. Samsonov c,M. Teresa Pisabarro c, Stephanie Moeller d, Matthias Schnabelrauch d,Dieter Scharnweber b,e, Lorenz C. Hofbauer a,e,1, Vera Hintze b,*,1

aDivision of Endocrinology, Diabetes and Bone Diseases, Department of Medicine III, Dresden Technical University Medical Center, D-01307 Dresden,Germanyb Institute of Materials Science, Max Bergmann Center of Biomaterials, Technische Universität Dresden, D-01069 Dresden, Germanyc Structural Bioinformatics, BIOTEC, Technische Universität Dresden, D-01307 Dresden, GermanydBiomaterials Department, INNOVENT e. V, D-07745 Jena, GermanyeCenter for Regenerative Therapies Dresden, D-01307 Dresden, Germany

a r t i c l e i n f o

Article history:Received 14 May 2013Accepted 26 June 2013Available online 17 July 2013

Keywords:Osteoprotegerin (OPG)Receptor activator of NF-kB ligand (RANKL)Hyaluronic acid/hyaluronan (HA) sulfateSurface plasmon resonance (SPR)Osteoclast (OC)

* Corresponding author. Tel.: þ49 463 39389; fax:E-mail address: Vera.Hintze@tu-dresden.de (V. Hi

1 Shared first and senior authorship.

0142-9612/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.biomaterials.2013.06.053

a b s t r a c t

In order to improve bone regeneration, in particular in aged and multimorbid patients, the developmentof new adaptive biomaterials and their characterization in terms of their impact on bone biology iswarranted. Glycosaminoglycans (GAGs) such as hyaluronan (HA) are major extracellular matrix (ECM)components in bone and may display osteogenic properties that are potentially useful for biomaterialcoatings. Using native and synthetically derived sulfate-modified HA, we evaluated how GAG sulfationmodulates the activity of two main regulators of osteoclast function: receptor activator of NF-kB ligand(RANKL) and osteoprotegerin (OPG). GAGs were tested for their capability to bind to OPG and RANKLusing surface plasmon resonance (SPR), ELISA and molecular modeling techniques. Results were vali-dated in an in vitro model of osteoclastogenesis. Sulfated GAGs bound OPG but not RANKL in a sulfate-dependent manner. Furthermore, OPG pre-incubated with different GAGs displayed a sulfate- and dose-dependent loss in bioactivity, possibly due to competition of GAGs for the RANKL/OPG binding siterevealing a potential GAG interaction site at the RANKL/OPG interface. In conclusion, high-sulfated GAGsmight significantly control osteoclastogenesis via interference with the physiological RANKL/OPG com-plex formation. Whether these properties can be utilized to improve bone regeneration and fracturehealing needs to be validated in vivo.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

To date, bone grafts are one of the most demanded implantmaterials [1]. The increasing age of implant recipients has creatednew challenges for biomaterial design, since these implants areused in defective or osteoporotic bone or in patients with comor-bidities that may impair osseointegration, such as diabetes mellitus[2]. Successful osseointegration of these implants depends on thecharacteristics and regenerative potential of the host bone as wellas the material properties.

þ49 463 39401.ntze).

All rights reserved.

Physiological bone regeneration is a coordinated process,maintained by bone-forming osteoblasts and bone-resorbing os-teoclasts (OC). The activity and differentiation of OCs are controlledby receptor activator NF-kB ligand (RANKL) and osteoprotegerin(OPG), two factors produced by bone-forming osteoblasts. RANKLbinding to its receptor RANK on monocytic/macrophagic precursorcells induces osteoclastogenesis. In this system, OPG acts as a sol-uble decoy receptor that interferes with RANKL/RANK signaling bybinding RANKL.

Recently, we reported that the sulfate modification of GAGs, akey component of the native ECM of bone, may be a promising toolto regulate biological processes at the bone/biomaterial interface[3]. The sulfate modification of GAGs led to a profound inhibition ofOC differentiation and their resorptive capacity. Therefore, GAGshold an interesting potential for the development of innovative

Table 1Synoptic table of the characteristics of HEP, HA, and synthesized sulfated HA de-rivatives (degree of sulfation: D.S.), weight-average molecular weight: Mw (asdetermined by gel permeation chromatography equipped with laser light scatteringdetection or *as specified by the distributor).

Sample-ID HA sHA1 sHA3 HEP

Mw [g mol�1] 48 000 31 000 21 000 18 000*D.S. e 1.0 2.9 2.0Sulfate position

determined by NMRe C6 (GlcNAc) C4, C6 (GlcNAc),

C3, C2 (GlcA)n.d.

Designation oftetrasaccharides forin silico modeling

HA HA6 HA4630 e

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e76617654

biomimetic materials for bone tissue engineering applications andregenerative medicine [4].

Local GAGs in bone consist of about 90% chondroitin sulfate andsmall amounts of hyaluronan (HA), and dermatan sulfate [5]. TheseGAGs are linear polymers of repetitive disaccharide units whosechemical structures differ with respect to monosaccharidecomposition and posttranslational modification. Bound to a smallcore protein sulfated GAGs are subunits of proteoglycans. Becausethe properties of GAGs largely determine the properties of theentire proteoglycan molecule [6] and the lower immunogenicity ofGAGs compared to proteoglycans, GAG composites are currentlydeveloped and evaluated for a wide range of applications in tissueengineering [6].

Previous studies have suggested that GAGs, such as heparin(HEP) and dermatan sulfate, may inhibit OC functions throughdirect interactions with osteoclastic mediators such as RANKL, OPG,and cathepsin K [7,8], though the precise mechanisms have notbeen identified. Moreover, some of the data are inconsistent, whichis in part due to different sources and formulations of the GAGsemployed [6]. To circumvent this issue we investigated the effectsof chemically sulfated HA derivatives with defined chain length andsulfation degrees in comparison to native HEP and non-sulfated HA.We decided to use chemically sulfated HA derivatives since theirinteractionwith several mediator proteins relevant to bone healingwas found to be sulfation dependent in previous studies [9,10].

Furthermore we observed that the regulation of osteoclasto-genesis by sulfated HA and chondroitin sulfate was depending onGAG sulfation rather than the monosaccharide composition [3]. Inaddition the inexpensive accessibility of HA makes sulfated HAshighly attractive model substances with a comparable polymerstructure to the naturally sulfated HEP consisting of repeatingdisaccharide units of uronic acid (D-glucuronic and L-iduronic acid)and D-glucosamine.

A thermally degraded HAwas used as a reference instead of full-length HA since its molecular weight is known to significantlydecrease during the sulfation process [10].

In the present study, we aimed to determine, if these HAsdirectly interact with OPG and RANKL using ELISA, SPR and mo-lecular modeling techniques. Utilizing an in vitro model of osteo-clastogenesis we then analyzedwhether this translates in increasedor decreased OC differentiation and activity that could be beneficialfor biomaterial applications.

2. Materials and methods

2.1. Materials

HA (from Streptococcus, MW ¼ 1.1 � 106 gmol�1) was purchased from AquaBiochem (Dessau, Germany). Recombinant human OPG (185-OS-025/CF), mono-clonal mouse anti-human OPG (MAB8051), biotinylated polyclonal goat anti-humanOPG (BAF805) antibodies and the recombinant human RANKL (6449-TEC-10/CF)were obtained from R&D Systems (Wiesbaden-Nordenstadt, Germany). The Series SSensor Chips C1� and CM3�, the Amine Coupling Kit and HBS-EP (10�) werepurchased from GE Healthcare Europe GmbH (Freiburg, Germany). Methyl vinylether/maleic acid copolymer and adipic acid dihydrazide were purchased fromFisher Scientific (Nidderau, Germany). HEP from porcine intestinal mucosa, sodiumcyanoborohydride, bovine serum albumin (BSA), Tween 20, sucrose, and 3,30 ,5,50-tetramethylbenzidine liquid substrate were available from SigmaeAldrich(Schnelldorf, Germany).

2.2. Preparation of HA derivatives

The low molecular weight HA and the sulfated HA derivatives (sHA1, sHA3),were synthesized and characterized as previously described [10,11]. 13C NMR in-vestigations [10] on the distribution of sulfate groups within the disacchariderepeating unit of the synthesized HAs have revealed that the low-sulfated HA (sHA1)is sulfated exclusively at the C6 position of the N-acetyl-D-glucosamine (GlcNAc)sugar ring. The high-sulfated derivative sHA3 is also completely sulfated at the C6position of the GlcNAc sugar, whereas the C4 position of the same sugar unit and theC2 as well as the C3 position of the D-glucuronic acid (GlcA) unit are partly sulfated

to a comparable extent. Characteristics of the prepared HAs are summarized inTable 1.

2.3. Covalent coupling of GAGs to MaxiSorp� 96-well ELISA plates

A 96-well microtiter plate (Nunc MaxiSorp�) was coated with GAGs via theirreducing ends as previously described [12]. In brief, each well was incubated for30 min at room temperature (RT) with methyl vinyl ether-maleic anhydridecopolymer in DMSO, which was removed and adipic acid dihydrazide was added for2.5 h. Then, 55 mg HA was dissolved in 25 mM citrate-phosphate buffer pH 5.0. Themolar concentrations of the other GAG, including sulfated HA derivatives and HEP,were the same in relation to the molecular weight of their disaccharide units. Schiffbases were obtained via reaction of the reducing ends of GAG with the previouslyintroduced hydrazino groups. 50 mL of GAG solution were applied to each well. Thepreviously untreatedwells for the negative control were filled with 50 mL TBS (10mM

Tris, 50 mM NaCl, pH 7.4)/2% BSA. The plate was incubated overnight (o/N) at RT. Thenext day 10 mL of a 1% sodium cyanoborohydride solution in methanol, which re-duces the Schiff bases to stable alkylamine bonds [12], were added to each well. Therespectivewells were thenwashed three times with TBS, followed by a blocking stepfor 2 h with 300 mL/well TBS/2% BSA. Then, 0, 50 and 100 ng/ml OPG in 1% BSA/PBSwas incubated on the prepared surfaces o/N at 4 �C.

The following day 150 mL PBS/1% BSA was added to each well. The resulting200 mL were collected into tubes, the wells were flushed with 200 mL PBS/1% BSA,which were then also transferred into the respective tube and stored frozen at�80 �C for a following sandwich ELISA.

2.4. Enzyme linked immunosorbent assay (ELISA)

All procedures were performed at RT and on an orbital shaker. In a direct ELISA,the wells, which had previously been coated with GAG derivatives and incubatedwith OPG, were blocked for 1 h with 300 mL/well blocking buffer (1% (w/v) BSA, 5%(w/v) sucrose in PBS). Afterward they were washed three times with 300 mL/wellwashing buffer (0.05% (v/v) Tween 20 in PBS) for approx.1 min each and 100 mL/welldetection antibody (goat anti-human OPG, 0.05 mg/mL PBS/1% BSA) were applied.The plate was covered with parafilm and incubated for 2 h. After several washingsteps, 100 mL/well streptavidin-HRP (R&D systems, dilution 1:400 in PBS/1% BSA)were applied. The streptavidin-HRP was aspirated 20 min later and the plate waswashed as previously described. Via addition of 100 mL/well 3,30 ,5,50-tetrame-thylbenzidine liquid substrate HRP enzyme activity was detected. The reaction wasstopped after 5 min with 50 mL/well 1 M H2SO4 and absorption was measured at450 nm (Tecan Genios).

The amount of unbound OPG was determined by sandwich ELISA using a 96-well plate (Nunc MaxiSorp�) coated with 100 mL/well capture antibody (mouseanti-human OPG, 1 mg/mL PBS) and incubated o/N at 4 �C. The plate was washedthree times with 300 mL/well washing buffer for approx. 1 min and blocked for 1 hwith 300 mL/well blocking buffer. In the meantime, the samples to be tested werethawed and further diluted with 600 mL 1% BSA/PBS, resulting in an overall dilutionof 20 times. Also a standard curve was prepared from 12.5 ng OPG/mL 1% BSA/PBS.After washing, 100 mL/well of sample or standardwere incubated in thewells for 2 h.The plates werewashed and the procedure was continued as described for the directELISA starting with the addition of detection antibody.

2.5. Immobilization of OPG and RANKL on Series S Sensor Chips CM3� and C1�

Interaction analysis of OPG, RANKL and GAGs were performed using a Biacore�T100 instrument (GE Healthcare). OPG was immobilized onto a Series S Sensor ChipCM3� via its amine groups at 25 �C. Following the instructions of the manufacturer(GE Healthcare) for amine coupling, the carboxyl groups of the chips dextran matrixwere activated to succinimide esters with a freshly prepared mixture of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC) and N-hydrox-ysuccinimide (NHS), which was injected over the flow cell for 7 min at 10 mL/min.Next 2.5 mg/mL of OPG diluted in sodium acetate buffer pH 5.5 were injected at

Table 2Primer sequences for b-Actin, TRAP, OSCAR, CTSK and NFATc1.

Primer Sequence 50->30

mu b-Actin s GATCTGGCACCACACCTTCTmu b-Actin as GGGGTGTTGAAGGTCTCAAAmu TRAP s ACTTGCGACCATTGTTAGCCmu TRAP as AGAGGGATCCATGAAGTTGCmu OSCAR s TGGCGGTTTGCACTCTTCAmu OSCAR as GATCCGTTACCAGCAGTTCCAGAmu CTSK s AAGTGGTTCAGAAGATGACGGGACmu CTSK as TCTTCAGAGTCAATGCCTCCGTTCmu NFATc1 s GTTCCTTCAGCCAATCATCCmu NFATc1 as GGAGGTGATCTCGATTCTCG

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e7661 7655

10 mL/min until an immobilization level of approx. 910 RU was achieved. Unreactedgroupswere deactivated via injection of 1 M ethanolamine-HCL, pH 8.5 (7min,10 mL/min). A reference surface was created without injecting OPG.

RANKL was immobilized onto a Series S Sensor Chip C1� via amine couplingfollowing the procedure as described above. However before immobilization thechip was conditioned as recommended by the manufacturer, and 0.1 M glycin-NaOH,pH 12 containing 0.3% Triton X-100 buffers were injected twice for 60 s at 10 mL/min.For immobilization RANKL was diluted to 2.5 mg/mL in sodium acetate pH 4 andinjected at 10 mL/min until an immobilization level of approx. 500 RU was achieved.

2.6. SPR measurements of OPG interaction with GAGs

Binding analysis was performed at 37 �C and at a flow rate of 30 mL/min. EachGAG was diluted in running buffer (HBS-EP: 0.01 M HEPES (pH 7.4), 0.15 M sodiumchloride, 3 mM EDTA, 0.05% surfactant P20) to a concentration of 1 mM, 10 mM and100 mM disaccharide units. Same molar concentrations in relation to disaccharideunits were injected. After five start-up cycles with running buffer, increasing con-centrations of GAG were injected with a contact time of 5 min. The values of thebinding level were recorded 10 s before the end of injection and later referencedwith the respective baseline response as well as emended for the GAG molecularweight. After dissociation for 10 min the sensor surface was regenerated for 2 minwith 5 M NaCl in HBS-EP and a 500 s stabilization timewas given, before a next cyclesstarted.

2.7. SPR analysis of RANKL-OPG complex formation in the presence of GAGs

All investigations on the interaction of pre-formed OPG/GAG complexes withimmobilized RANKL were preceded by three starteup cycles. Samples of 20, 50,100 nM and 2 mM sHA3 disaccharide units and 25 nM OPG were pre-incubated at RTfor 1 h, before the start of the measurement. For all other GAGs 50 nM disaccharideunits were used. In order to minimize any further interaction between GAGs andOPG after this time, the temperature within the Biacore� sample compartment wasset to 12 �C. Samples were injected over the sensor surface for 2min and after 10minof dissociation the chip was regenerated with 50 mM NaOH (60 s, 300 s stabilizationtime). For comparison, 25 nM OPG in running buffer was injected at the beginningand the end of the experiment. In the interest of excluding a GAG/RANKL interaction,control samples of 2 mM sHA3 or 50 nM of all other GAGs were tested without OPG.

2.8. Computational analysis of GAG-OPG/RANKL interactions

The docking of HA, sHA1 and sHA3 was carried out using tetrasaccharides withdefined sulfation pattern referred to as HA, HA6, HA4630 , indicating the position ofsulfate groups. In this context HA6 illustrates a sulfation of the C-6 position, whileHA4630 indicates sulfation in the C-4 and C-6 as well as the sulfation of the sec-ondary hydroxy groups C-30 .

Docking of tetrasaccharides was carried out on the domain I of OPG taken fromits complex structure with RANKL (PDB ID 3URF, 2.70 Å) [13]. AutoDock 3 [14] wasusedwith the following parameters: grid step of 0.7 Å, Lamarckian genetic algorithmwith 300 individual in population, 104 generations and 9995105 maximum numberenergy evaluations was applied for 1000 runs. The receptor was minimized in MOE(Chemical Computing Group, Inc. Montreal, CA) prior to docking using default pa-rameters. Clustering of docking results was done with the following parameters: 50top solutions, ε ¼ 4 Å, each cluster containing at least 5 members. Representativeposes from the clusters overlapping with the RANKL binding site were furtheranalyzed by molecular dynamics (MD) as described before [4]. MM-GBSA free en-ergy calculations were done for the last 2 ns of 10 ns productive MD run. It should benoted that our calculations did not include the HEP-binding domain (HBD) of OPGsince its 3D structure has not been yet experimentally solved.

2.9. Cell culture

In vitro experiments were performed to confirm the biophysically and bio-chemically obtained results. Cell culture media containing GAGs were pre-incubatedwith OPG prior to RANKL addition to allow an analysis whether pre-formed OPG/GAG complexes result in altered osteoclastogenesis compared to RANKL/OPG-onlytreated cells. In a preceding experiment OPG concentration was optimized for aRANKL concentrations of 10 ng/ml.

In brief, RAW264.7 cells (ATCC, Wesel, Germany) were cultivated in MEM-alpha(Biochrom, Berlin, Germany) supplemented with 10% fetal calf serum (PAA, Cölbe,Germany), 2 mM glutamine (Biochrom) and 1% Penicillin/Streptomycin (PAA) untilthey reached 80% confluency. Cells were then either split and passaged into newflasks or used for experiments. Therefore, cells were seeded at a density of12 500 cells/cm2 in 6 well cell culture plates and stimulated with 10 ng/ml RANKL(R&D systems) and serial dilutions of OPG (R&D systems) ranging from 6.25 ng/ml to100 ng/ml. For the main experiment 25 ng/ml OPG, which translates to a concen-tration of 0.288 nM, was pre-incubated for 1 h with HEP, HA, sHA1 or sHA3respectively before RANKL was added to the medium. The applied molar GAGconcentrations were based on disaccharide units with 0.288 nM (GAG:OPG 1:1) andmultiples of this ratio (10:1, 100:1, 500:1).

2.10. Immunofluorescence

To assess the effects of OPG and OPG/GAG complexes on OC functionalmorphology, RAW264.7 cells were differentiated on glass slides as described aboveuntil mature OCs were formed in the control wells. Cells were washed with PBS, andfixed for 15 min with 4% paraformaldehyde and permeabilized for 20 min in 0.1%Triton X-100/PBS, to ensure antigen accessibility. Following a short washing step,unspecific binding sites were blocked for 1 h with 1% BSA, 0.05% Tween in PBS. Then,paxillin antibody was used as described [15]. After a brief washing with PBS, cellnuclei were stained with 2.5 mg/ml DAPI (Sigma, Schnelldorf, Germany) for 5 minand extensively washed afterward. Finally, slides were embedded in a small amountof fluorescence-preserving mounting medium (Dako, Hamburg, Germany) andexamined and documented using digital microscopy. Using ImageJ [16], at least 5randomly taken images of each slide were analyzed for their OC number, cumulativearea covered by OCs and their individual size relative to an uncoated control.

2.11. RNA isolation, RT and quantitative real-time PCR

When mature OCs had formed after 3e4 days, RNA was extracted using theHighPure RNA isolation kit (Roche, Berlin, Germany) according to the manufacturesinstructions. 500 ng RNA was reverse transcribed using SuperScript II (Invitrogen,Darmstadt, Germany), and an SYBR green-based real-time PCR was performed un-der standard conditions. PCR primers were generated with Primer3 and reviewedusing BLAST and OligoAnalyzer 3.0 software. PCR reactions were performed as re-ported [17] using the primer sequences given in Table 2. The abundance of mRNAlevels was calculated using the DD-ct-method [18] and is presented in fold increaserelative to an RANKL-only treated control.

2.12. Statistical analysis

All experiments were performed in at least triplicate. The interaction of OPG andGAGs (Fig. 1) was evaluated using two-way ANOVA, while all other data sets werejudged using one-way ANOVA including all data sets (Figs. 2B, 4 and 5EeG). Bon-ferroni post-hoc test was applied in all cases to evaluate differences between thetreatment groups. All results are presented as mean � SD. P values < 0.05 wereconsidered statistically significant.

3. Results

3.1. Interaction analysis of GAGs with OPG

The interactionofupto5ng/wellof soluteOPGwas investigatedbyincubating it with immobilized GAG derivatives. In the ensuing directELISA, binding was found to be dependent on the sulfation degreewith the highest absorbance observed for sHA3 and the lowest for HA(Fig. 1A). HEP and sHA1 were between these two extremes with nosignificant difference. On BSA coated wells however, which served asnegative controls, OPG could not be detected. Hence, all surfaceswithimmobilized GAGs retained significantly more OPG than thosecovered with BSA. Quantification of OPG by sandwich ELISA usingspecific antibodies as capture molecules confirmed that the highestamount of OPG was bound to sHA3 and the lowest to HA (Fig. 1B), ofwhich the latter binding responsewas similar to the negative control.The amount of OPG bound to HEP and sHA1 was between those ofsHA3 and HA, which were not significantly different.

The interactions of three different concentrations of GAG de-rivatives and immobilized OPG were further analyzed by SPR.

0 1 2 3 4 5

0.0

0.1

0.2

0.3

0.4

0.5

HEP

HAsHA1

sHA3BSA a

a

a

OPG [ng]

0 1 2 3 4 5

0.0

0.5

1.0

1.5

2.0

2.5

c, d

a

a, c

a, c

OPG initially incubated [ng]

OP

Gb

ou

nd

to

GA

G[n

g]

BA

HEP

HAsHA1

sHA3BSA

C

ASILEtceriD

OD

450n

m

Ab

so

rb

an

ce

GAGs [µM D.U.]

1 10 100

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

***

**

***

**

**

********

**

***

Bin

din

gR

es

po

ns

e

[relative

to

1µ

Ms

HA

3] HA

sHA1

HEPsHA3

***

**

Sandwich ELISA

a, b

c

Fig. 1. GAGs interact with OPG depending on their sulfation degree. The interaction of immobilized GAG derivatives with different concentrations of solute OPG (0.625, 1.25, 2.5,5 ng/well) as determined by direct ELISA (A) and sandwich ELISA (B) using specific antibodies. All values represent the mean � SD of at least n ¼ 3. Interaction of immobilized OPG(910 RU) with different concentrations of solute GAG derivatives (1e100 mM disaccharide units) was analyzed via their binding levels recorded seconds before injection stop by SPR(C). All values represent the normalized mean � SD of n ¼ 4 and are given as relative to baseline response and corrected for the respective molecular weight of GAG derivatives.Two-way ANOVA: (A) a, p < 0.001 vs. BSA; b, p < 0.01 HA vs. sHA3, (B) a, p < 0.001 vs. BSA, p < 0.05 sHA1 vs. BSA; c, p < 0.001 vs. sHA3, p < 0.01 sHA1 vs. sHA3; d, p < 0.05 HA vs.HEP (C) *p < 0.05; **p < 0.01; ***p < 0.001 vs. respective treatment. Abbreviations used: D.U., disaccharide units.

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e76617656

Binding analyses revealed that binding strength is depending onthe sulfation degree of each GAG (Fig. 1C). While there was almostno detectable binding at all concentrations of HA, binding strengthof the sulfate containing GAGs increased in the order of their degreeof sulfation: sHA1 < HEP < sHA3.

3.2. RANKL/OPG complex formation in the presence of GAGs

To analyze the potential impact of GAG binding to OPG on theRANKL/OPG complex formation, their interactions were studied via

0 700

0

5

10

15

20

25

30

35

40

45 25 nM OPG

HBS-EP2 µM sHA3

OPG + 20 nM sHA3OPG + 50 nM sHA3OPG + 100 nM sHA3OPG + 2 µM sHA3

OP

Gb

in

din

g

A

[resp

on

se

un

its]

100 200 300 400 500 600

Time [s]

Fig. 2. RANKL-OPG complex formation is inhibited by GAGs in a sulfation-dependent mannepre-incubation with different concentrations of sHA3 (A) and 50 nM disaccharide units of difbinding relative to OPG binding without GAG derivatives is shown. The latter values represtreatment; ##p < 0.01; vs. OPG only.

SPR. Thus, RANKL was immobilized onto a sensor chip surface andevaluated for its interaction with 25 nM OPG pre-incubated withdifferent concentrations of sHA3, the GAG derivative identified ashaving the strongest binding to OPG in this study (Fig. 1).Increasing concentrations of sHA3 led to a significantly reducedbinding response of OPG. This blocking effect peaked at a con-centration of 2 mM disaccharide units of sHA3, a dose at which theresponse was indistinguishable from a GAG sample devoid of OPGand the negative control (HBS-EP buffer) without GAGs and OPG(Fig. 2A).

0

25

50

75

100

125

150

175

##

**

##

**

**

OP

Gb

in

din

g

[re

la

tiv

eto

co

ntro

l]

B

+ + + + +OPG 25 nM

GAG 50 nM - HA sHA1 sHA3 HEP

r. SPR analysis for the interaction of 25 nM OPG with immobilized RANKL (500 RU) afterferent GAG derivatives (B). In (A) the resulting sensorgrams are displayed while in (B) %ent the mean � SD of n ¼ 3. One-way ANOVA: **p < 0.01; ***p < 0.001 vs. respective

Fig. 3. Electrostatic potential of OPG and OPG/RANKL complex. (A) Predicted GAGbinding sites as calculated by molecular modeling and (B) positively (red) and nega-tively (blue) charged patches of the OPG domain I and RANKL proteins. The predictedGAG binding site partially overlapping with the RANKL binding site is highlighted by atransparent sphere. (For interpretation of the references to colour in this figure legend,the reader is referred to the web version of this article.)

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e7661 7657

Pre-incubation of 25 nM OPGwith 50 nMGAG derivatives, whichled to a reduction in OPG binding of 40% in the presence of sHA3,was chosen to compare the impact of other differently sulfatedGAGs on OPG/RANKL complex formation. Here, the bindinglevels decreased in a sulfate-dependent manner in the ordersHA3 > HEP > sHA1, while HA led to a slight, but not significant,increase of binding response (Fig. 2B).

We could exclude a RANKL/GAG interaction in this experiment,since 50 nM GAG derivatives alone displayed no detectable bindingresponse. Even with the use of a saturated RANKL surface(w2000 RU) and GAG concentrations of 100 mM disaccharide units,we observed only a weak interaction of GAGs with RANKL (data notshown). By contrast, control injections of 4000-fold lower doses of25 nM OPG led to an appropriate response.

3.3. Computational analysis of GAGs with RANKL/OPG

Our docking experiments revealed that HA binding to OPGformed roughly two types of clusters: “blocking” and “non-block-ing”. In the blocking clusters, the obtained HA positions partiallyoverlapped with the RANKL binding site (Fig. 3A), whereas the HApositions did not show overlap with the RANKL binding site for thenon-blocking clusters. In the first case, HAs could compete withRANKL for its binding site on OPG and, therefore, potentially blockOPG/RANKL interactions. For higher sulfated HAs, we found morepopulated clusters (Table 3), which overlapped with the bindingsite for RANKL. Moreover, mean AutoDock 3 scoring and MM-GBSAfree energy calculations with the representative HA positions fromthese clusters showed that an increase of sulfation leads to strongerinteractions with OPG in this binding region. Therefore, the po-tential ability of blocking the binding site for RANKL by theanalyzed HAs increases in the following order: HA < sHA1(HA6) < sHA3 (HA4630).

Table 3Synoptic overview of docking experiments of GAGs to OPG domain I.

GAG “Non-blocking”cluster size

“Blocking” cluster size

HA 15 9HA6 e 13HA4630 5 17 (cluster1)

12 (cluster 2)

Furthermore, when OPG was bound to HAs, the association toRANKL was hindered due to an electrostatic repulsion between theHA and a negatively charged region of RANKL close to the HAbinding site (Fig. 3B).

3.4. Osteoclastic gene expression with OPG/GAG complexes

Before the main experiment an OPG dose response analysis wasperformed to determine the effective dose in this setup (data notshown). The most pronounced effects of OPG were observed forTRAP, OSCAR and NFATc1. Their expression levels showed a pro-found decline up to 98% using 100 ng/ml OPG. On the other hand,gene expression of proteins associated with the final steps ofosteoclastogenesis such as CTSK was less regulated. Only smallinhibition could be documented for 12.5 ng/ml OPG and higherdosages failed to surpass inhibition levels of 55%. To furtherinvestigate whether OPG bioactivity is affected by prior incubationwith GAGs, an OPG concentration of 25 ng/ml was chosen based onthis experiment. The addition of non-sulfated HA in a range from anequimolar to a 500-fold higher concentration did not affect OPGpotency. These cells showed no significant differences in geneexpression of TRAP (Fig. 4A), OSCAR, NFATc1 or CTSK (data notshown) with or without pre-incubation of OPG with HA. However,other sulfated GAGs modified OPG-dependent marker gene sup-pression in a sulfation-dependent manner as shown exemplarilyfor TRAP gene expression (Fig. 4BeE). Increasing concentrations ofsHA1 (Fig. 4B), sHA3 (Fig. 4C) and HEP (Fig. 4D) partly rescuedmarker gene suppression up to 60% of the control. A direct com-parison of the gene expression levels observed when treated withthe highest GAG concentration shows a gradual increase of TRAPexpression levels in response to non-sulfated, low-sulfated andhigh-sulfated HA. HEP that differs in its carbohydrate compositionis positioned between sHA1 and sHA3 reflecting its sulfate contentwith a degree of sulfation of 2 (Fig. 4E).

3.5. OC morphology in response to OPG/GAG

Monocytes cultivated in the presence of RANKL only (Fig. 5A)matured to large, multinucleated cells with diameters oftenexceeding several hundred micrometers. Their cytoskeleton showsthe characteristic OC specific rearrangement, where an actin ringwith associated proteins, such as paxillin, develops in the periph-ery, thus forming the sealing zone. The addition of OPG (Fig. 5B)strongly inhibited osteoclastogenesis by direct interference withRANKL and subsequent reduced activation, resulting in a profoundloss of OC number (Fig. 5E), surface (Fig. 5F), size (Fig. 5G), anddiminished paxillin expression. Co-incubation of OPG with high-sulfated HA (Fig. 5C) almost abolished this anti-osteoclastogeniceffect. A comparable number of OCs of similar size were formedthat subsequently covered similar surface areas and showed OCmorphologies indistinguishable from control cells, which wereexposed to RANKL only (Fig. 5A, EeG). The application of sHA3 ledto a weak inhibition of osteoclastogenesis compared to the anti-osteoclastogenic action of OPG (Fig. 5B, D). Number, size and

AutoDock 3 score for “blocking”cluster (kcal/mol)

DGMM-GBSA(kcal/mol)

�13.3 � 0.3 �9.1 � 4.1�14.8 � 0.5 �21.7 � 5.2�17.0 � 1.0 �29.9 � 5.1�17.0 � 0.9 �22.3 � 7.4

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

######

##

##

*

**

mR

NA

exp

ressio

n

[relative

to

co

ntro

l]

E

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

### ### ### ######

mR

NA

exp

ressio

n

[relative

to

co

ntro

l]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

### ##### ## ##

mR

NA

exp

ressio

n

[relative

to

co

ntro

l]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

###

###

####

##

#

mR

NA

exp

ressio

n

[relative

to

co

ntro

l]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

### #####

###

mR

NA

exp

ressio

n

[re

la

tiv

eto

co

ntro

l]

HA sHA1

sHA3 HEP

BA

C D

+ + + + + +

OPG 25 ng/ml

RANKL 10 ng/ml

GAG [rel. to OPG]

+

- + + + + + -

0 0 1 10 100 500

+ + + + + + +

- + + + + + -

0 0 1 10 100 500

+ + + + + +

OPG 25 ng/ml

RANKL 10 ng/ml

GAG [rel. to OPG]

+

- + + + + + -

0 0 1 10 100 500 500

+ + + + + + +

- + + + + + -

0 0 1 10 100 500 500

+ + + + + +

OPG 25 ng/ml

RANKL 10 ng/ml

GAG:OPG [500:1]

- + + + + +

- - HA sHA1 sHA3 HEP

500500

RANKL

OPG

GAG

RANKL

OPG

GAG

Fig. 4. Binding of GAGs rescuesOPGmediated gene expression suppression. Quantitative real-timePCR analysis of TRAP expression ofmonocytic RAW264.7 cells treatedwith 25ng/mlOPG pre-incubated with different concentrations of (A) HA, (B) sHA1, (C) sHA3, or (D) HEP until day 3e4 of their differentiation. The respective GAGs were used in the same molarconcentration (disaccharide units) ormultiples of 10 thereof (OPG:GAG¼ 1:1e1:1000). Cells differentiatedwithout OPG andGAG, andwith the highest GAG-dosage served as controls.(E) Cumulative graph of the highest doses of each GAG and the respective controls. n¼ 3e4. One-way ANOVA: *p< 0.05; **p< 0.01; ***p< 0.001 vs. respective treatment; #p< 0.05;##p < 0.01; ###p < 0.001 vs. RANKL only. All values represent the normalized mean � SD of n ¼ 3. Abbreviations used: TRAP, tartrate-resistant acid phosphatase.

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e76617658

subsequent OC surface area (Fig. 5EeG) were slightly decreasedand paxillin expression was not as pronounced as in the controlgroup.

4. Discussion

The homeostasis of skeletal mass and strength is maintained bybalancing bone resorption with formation. After an osteoporotic

fracture, a therapeutic shift of this equilibrium in favor of increasedbone formation or decreased resorption could be beneficial fordefect healing. Over the past decades, three molecules, RANKL, itsreceptor, receptor activator of NF-kB (RANK), and its soluble decoyreceptor OPG have been identified as key regulators of OC differ-entiation and activation [19,20]. Therefore, altering either theexpression of these factors or their bioactivity may allowmodifyingbone remodeling processes.

E

RANKL RANKL + OPG

RANKL + sHA3RANKL + OPG + sHA3

BA

D

2.0

0.0

0.5

1.0

1.5

*** ***#

###

Osteo

clast

nu

mb

er

[relative

to

co

ntro

l]

0.0

0.5

1.0

1.5

2.0

*** ***

###

##

Os

te

oc

la

st

su

rfa

ce

[re

la

tiv

eto

co

ntro

l]

0.0

0.5

1.0

1.5

2.0

*** ***###

####

Os

te

oc

la

st

size

[re

la

tiv

eto

co

ntro

l]

+ + + +

OPG 25 ng/ml

RANKL 10 ng/ml

sHA3:OPG [500:1]

- + + -

- - + +

+ + + +

OPG 25 ng/ml

RANKL 10 ng/ml

sHA3:OPG [500:1]

- + + -

- - + +

+ + + +

OPG 25 ng/ml

RANKL 10 ng/ml

sHA3:OPG [500:1]

- + + -

- - + +

C

G

F

Fig. 5. Interaction with sulfated GAGs rescues osteoclastogenesis in vitro. Monocytic RAW264.7 cells were differentiated for 3e4 days until control wells showed profoundosteoclast formation. Cell differentiation was then assessed using immunofluorescence staining of f-actin (green, phalloidin-AlexaFluor488), paxillin (red, a-paxillin) and cell nuclei(blue, DAPI). (AeD) Representative pictures of (A) RANKL only, (B) RANKL and OPG, (C) RANKL, OPG and sHA3 and (D) RANKL and sHA3 treated cells. (EeG) Quantitative analysis of 6or more images per treatment group via ImageJ of (E) osteoclast number, (F) osteoclast surface, and (G) osteoclast size. All values represent the normalized mean � SD of n ¼ 3. One-way ANOVA: *p < 0.05; **p < 0.01; ***p < 0.001 vs. respective treatment; #p < 0.05; ##p < 0.01; ###p < 0.001 vs. RANKL only. (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e7661 7659

Based on our previous work, showing that high-sulfated GAGsinhibit osteoclastogenesis, we set out to determine whether analtered RANKL/OPG balance contributed to this effect [3]. One of thebest studied GAGs in this respect is HEP, which is routinelyadministered as anti-coagulant and has been shown to induceosteoporosis in patients, when administered over long time [21e23]. HEP has been shown to promote bone resorption in organcultures and increase osteoclastic bone resorption in cell culture[24,25]. Of note, osteoclastogenesis was only supported in thepresence of different animal sera, suggesting that HEP acts throughbinding to a serum factor that regulates OC activity. Furthermore,the ability of GAGs to enhance resorption was dependent on theirdegree of sulfation and its size. High molecular weight heparansulfate had a detectable effect, while low molecular weight HEP,chondroitin-2, -4 and -6 sulfate had not.

In this study we assessed the mechanisms by which GAGs inparticular sulfated HA derivatives interfere with the RANKL/OPGbalance in comparison to native HEP. To circumvent the variabilityof animal-derived sulfated GAGs, we synthesized GAG derivativeswith defined chain length and sulfation degrees and used SPR andcomputer modeling as cell free methods to minimize interferingfactors.

In our previous study, RANKL-induced osteoclastogenesis wasimpaired in the presence of GAGs in a sulfate-dependent manner[3]. Since our in vitro OC formation assays were performed in theabsence of OPG we hypothesized that the anti-osteoclastogenic

GAG-interference is a RANKL-dependent mechanism. This wassupported by several other studies showing that GAGs bind RANKLthereby resulting in decreased OC formation [7,8,26]. However,these studies could not conclusively link the inhibitory effect ofsulfated GAGs on osteoclastogenesis in vitro to its binding to RANKL,since RANK had a higher affinity for RANKL pre-incubated withGAGs than for RANKL alone [26]. In line with this, we could notdetect any interaction of RANKL with GAGs based on SPR and mo-lecular modeling in this work. Lack of a direct interaction of RANKLwith GAGs has also been reported by others [27,28]. The differentfindings on RANKL binding to GAGsmay be due to different sources,formulations and possible impurities of the employed GAGs [6],underlining the need for new synthetic GAG manufacturing ap-proaches and methods such as molecular modeling.

As already mentioned, HEP displays an anti-osteoclastogeniceffect in vivo, differing from our in vitro data [3]. However, in vivothe situation is more complex withmany other interaction partnersof RANKL present. We were therefore interested how the here usedchemically sulfated GAG derivatives might affect OPG/RANKLbinding in comparison to HEP.

We demonstrated that OPG interacts directly with GAGs in asulfate-dependent manner. Consistent with this, another SPR studyshowed OPG binding to native GAGs with high affinity [28].

When investigating the functional consequences of this inter-action, we found that the pre-incubation of OPGwith sulfated GAGsinhibited OPG binding to RANKL in a sulfation-dependent manner.

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e76617660

Similar to our results, other authors also reported an inhibitoryeffect of GAG on binding of OPG to the RANK/RANKL complex in adose-dependent manner [28]. Additionally, sulfation was essentialin the OPG-blocking function of GAGs since a completely desulfatedHEP did not bind and block OPG binding to RANK/RANKL and didnot regulate OC formation [28]. Furthermore, OPG pre-incubatedwith GAG-oligosaccharides was reported not to bind to immobi-lized RANKL, whereas an OPG variant lacking the HBDwas still ableto bind [29].

To address this in more detail, molecular docking was per-formed for differently sulfated GAG variants to predict potentialinteraction sites and determine which residues contribute to anOPG/GAG interaction. OPG possesses an HBD that may also allowfor binding of other GAGs. However, our study cannot conclusivelyattribute the observed effects to this domain since its crystalstructure is as of yet unknown and therefore not amenable to oursimulative approach. Using the known structure of OPG domain I[13] however, the in silico analysis of the interaction of OPGwith theHA derivatives suggested potential binding of GAGs at the RANKL/OPG interface. This site might be related to the already describedHBD site. Additionally, free energy calculations showed anincreasing binding strength depending on the GAG sulfation de-gree. Therefore we show that the combined forces of SPR andmolecular modeling excellently complement each other in theinteraction analysis of GAG/OPG complexes with RANKL.

To analyze the functional consequences of this GAG/OPG inter-action on OPG bioactivity, RANKL-induced osteoclastogenesis wasinvestigated. In the presence of sulfated GAGs, OPG bioactivity wasdecreased, resulting in rescued osteoclastogenic gene expressionand subsequent increased OC formation afore suppressed by OPG.

Like in our previous study we observed that the application ofsHA3 alone led to an inhibition of osteoclastogenesis in the presenceof RANKL, but we can now exclude that an altered RANKL/OPG bal-ance contributes to the GAG-mediated defect in OC formation [3].This inhibitory effect might also explain why increasing concentra-tionsof sulfatedGAGsarenot able to fully rescue theosteoclastogenicphenotype in the presence of OPG. We therefore hypothesize thatGAGs employ different mechanisms to regulate bone metabolism,not solely restricted to the RANKL/RANK/OPG axis, at various levelsand time points of the OC life cycle [26]. This includes several key OCregulators besides OPG such as cytokines, growth factors, matrixproteins and enzymes that have been identified to contain an HBDalso. Some of them, such as cathepsin K [30] and TGF-b [31,32] havebeen characterized with respect to their activity in association withGAGs. The complex formation of GAGs with these proteins alsoproved to regulate their activity dependingon thedegree of sulfationof the applied GAGs. Whether our in vitro findings result in a sub-sequent modulation of resorption activity needs to be furtherelucidated in vivo. This will be addressed in a follow up study, wherewe will apply GAGs as coatings on a poly lactic acid based scaffoldmaterial with subsequent analysis in normal and compromised an-imal models of sub-critical defect healing.

5. Conclusion

We demonstrated that GAG sulfation directly interferes withOPG bioactivity, resulting in increased OC formation. Apart from theRANKL/RANK/OPG axis, GAGs can also directly inhibit osteoclas-togenesis warranting further comprehensive in vitro and vivo vali-dations using appropriate pre-clinical models.

Acknowledgment

We wish to acknowledge support from the Deutsche For-schungsgemeinschaft [Transregio-67, projects A2, A3, A7, B2].

References

[1] Mathieu LM, Mueller TL, Bourban P-EE, Pioletti DP, Muller R, Manson JA, et al.Architecture and properties of anisotropic polymer composite scaffolds forbone tissue engineering. Biomaterials 2006;27:905e16.

[2] Hench LL. Biomaterials: a forecast for the future. Biomaterials 1998;19:1419e23.

[3] Salbach J, Kliemt S, Rauner M, Rachner TD, Goettsch C, Kalkhof S, et al. Theeffect of the degree of sulfation of glycosaminoglycans on osteoclast functionand signaling pathways. Biomaterials 2012;33:8418e29.

[4] Pichert A, Samsonov SA, Theisgen S, Thomas L, Baumann L, Schiller J, et al.Characterization of the interaction of interleukin-8 with hyaluronan, chon-droitin sulfate, dermatan sulfate and their sulfated derivatives by spectros-copy and molecular modeling. Glycobiology 2012;22:134e45.

[5] Prince CW, Navia JM. Glycosaminoglycan alterations in rat bone due to growthand fluorosis. J Nutr 1983;113:1576e82.

[6] Salbach J, Rachner TD, Rauner M, Hempel U, Anderegg U, Franz S, et al.Regenerative potential of glycosaminoglycans for skin and bone. J Mol Med2012;90:625e35.

[7] Ariyoshi W, Takahashi T, Kanno T, Ichimiya H, Shinmyouzu K, Takano H, et al.Heparin inhibits osteoclastic differentiation and function. J Cell Biol 2008;103:1707e17.

[8] Shinmyouzu K, Takahashi T, Ariyoshi W, Ichimiya H, Kanzaki S, Nishihara T.Dermatan sulfate inhibits osteoclast formation by binding to receptor acti-vator of NF-kappa B ligand. Biochem Bioph Res Co 2007;354:447e52.

[9] Hintze V, Miron A, Moeller S, Schnabelrauch M, Wiesmann H-P, Worch H,et al. Sulfated hyaluronan and chondroitin sulfate derivatives interact differ-ently with human transforming growth factor-b1 (TGF-b1). Acta Biomaterialia2012;8:2144e52.

[10] 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;10:3290e7.

[11] Kunze R, Rösler M, Möller S, Schnabelrauch M, Riemer T, Hempel U, et al.Sulfated hyaluronan derivatives reduce the proliferation rate of primary ratcalvarial osteoblasts. Glycoconj J 2010;27:151e8.

[12] Satoh A, Kojima K, Koyama T, Ogawa H, Matsumoto I. Immobilization ofsaccharides and peptides on 96-well microtiter plates coated with methylvinyl ether-maleic anhydride copolymer. Anal Biochem 1998;260:96e102.

[13] Luan X, Lu Q, Jiang Y, Zhang S, Wang Q, Yuan H, et al. Crystal structure ofhuman RANKL complexed with its decoy receptor osteoprotegerin. J Immunol2012;189:245e52.

[14] Morris GM, Goodsell DS, Huey R, Olson AJ. Distributed automated docking offlexible ligands to proteins: parallel applications of AutoDock 2.4. J ComputAid Mol Des 1996;10:293e304.

[15] Vives V, Laurin M, Cres G, Larrousse P, Morichaud Z, Noel D, et al. The Rac1exchange factor Dock5 is essential for bone resorption by osteoclasts. J BoneMiner Res 2011;26:1099e110.

[16] Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years ofimage analysis. Nat Methods 2012;9:671e5.

[17] Soares-Schanoski A, Gómez-Piña V, Del Fresno C, Rodríguez-Rojas A, García F,Glaría A, et al. 6-Methylprednisolone down-regulates IRAK-M in human andmurine osteoclasts and boosts bone-resorbing activity: a putative mechanismfor corticoid-induced osteoporosis. J Leukoc Biol 2007;82:700e9.

[18] Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 2001;25:402e8.

[19] Khosla S. Minireview: the OPG/RANKL/RANK system. Endocrinology2001;142:5050e5.

[20] Boyce BF, Xing L. Functions of RANKL/RANK/OPG in bone modeling andremodeling. Arch Biochem Biophys 2008;473:139e46.

[21] Griffith GC, Nichols G, Asher JD, Flanagan B. Heparin osteoporosis. JAMA1965;193:91e4.

[22] Jaffe MD, Willis PW. Multiple fractures associated with long-term sodiumheparin therapy. JAMA 1965;193:158e60.

[23] Miller WE, DeWolfe VG. Osteoporosis resulting from heparin therapy. Reportof a case. Cleveland Clinic Q 1966;33:31e4.

[24] Goldhaber P. Heparin enhancement of factors stimulating bone resorption intissue culture. Sci (N.Y.) 1965;147:407e8.

[25] Fuller K, Chambers TJ, Gallagher a C. Heparin augments osteoclast resorption-stimulating activity in serum. J Cell Physiol 1991;147:208e14.

[26] Velasco CR, Baud’huin M, Sinquin C, Maillasson M, Heymann D, Colliec-Jouault S, et al. Effects of a sulfated exopolysaccharide produced by Alter-omonas infernus on bone biology. Glycobiology 2011;21:781e95.

[27] Irie A, Takami M, Kubo H, Sekino-Suzuki N, Kasahara K, Sanai Y. Heparinenhances osteoclastic bone resorption by inhibiting osteoprotegerin activity.Bone 2007;41:165e74.

[28] Théoleyre S, Kwan Tat S, Vusio P, Blanchard F, Gallagher J, Ricard-Blum S, et al.Characterization of osteoprotegerin binding to glycosaminoglycans by surfaceplasmon resonance: role in the interactions with receptor activator of nuclearfactor kappaB ligand (RANKL) and RANK. Biochem Biophys Res Commun2006;347:460e7.

[29] Lamoureux F, Picarda G, Garrigue-Antar L, Baud’huin M, Trichet V, Vidal A,et al. Glycosaminoglycans as potential regulators of osteoprotegerin thera-peutic activity in osteosarcoma. Cancer Res 2009;69:526e36.

J. Salbach-Hirsch et al. / Biomaterials 34 (2013) 7653e7661 7661

[30] Li Z, Hou WS, Escalante-Torres CR, Gelb BD, Bromme D. Collagenase activity ofcathepsin K depends on complex formation with chondroitin sulfate. J BiolChem 2002;277:28669e76.

[31] Hempel U, Hintze V, Möller S, Schnabelrauch M, Scharnweber D, Dieter P.Artificial extracellular matrices composed of collagen I and sulfated hya-luronan with adsorbed transforming growth factor b1 promote collagen

synthesis of human mesenchymal stromal cells. Acta Biomater 2012;8:659e66.

[32] Van der Smissen A, Samsonov SA, Hintze V, Scharnweber D, Moeller S,SchnabelrauchM, et al. Artificial extracellular matrix composed of collagen I andhigh-sulfated hyaluronan interferes with TGFb1 signaling and prevents TGFb1inducedmyofibroblast differentiation. ActaBiomater 2013. [Epub aheadof print].