lable at ScienceDirect

Biomaterials 84 (2016) 167e183

Contents lists avai

Biomaterials

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

Guided bone regeneration is promoted by the molecular events in themembrane compartment

Alberto Turri a, b, d, 1, Ibrahim Elgali a, d, 1, Forugh Vazirisani a, d, Anna Johansson a, d,Lena Emanuelsson a, d, Christer Dahlin a, c, d, Peter Thomsen a, d, *, Omar Omar a, d

a Department of Biomaterials, Institute of Clinical Sciences, Sahlgrenska Academy at the University of Gothenburg, Swedenb The Brånemark Clinic, Institute of Odontology, Public Dental Health Care, Gothenburg, Swedenc Department of Oral, Maxillofacial Surgery and Research and Development, NU-Hospital Organization, Trollh€attan, Swedend BIOMATCELL VINN Excellence Center of Biomaterials and Cell Therapy, Gothenburg, Sweden

a r t i c l e i n f o

Article history:Received 4 January 2016Accepted 18 January 2016Available online 19 January 2016

Keywords:Guided bone regenerationMembraneIn vivoGene expressionWestern blotHistomorphometry

* Corresponding author. Department of Biomateriences, Sahlgrenska Academy at the University of GothGothenburg, Sweden.

E-mail address: peter.thomsen@biomaterials.gu.se1 Equal contribution.

http://dx.doi.org/10.1016/j.biomaterials.2016.01.0340142-9612/© 2016 The Authors. Published by Elsevier

a b s t r a c t

The working hypothesis of guided bone regeneration (GBR) is that the membrane physically excludesnon-osteogenic tissues from interfering with bone healing. However, the underlying mechanisms areinsufficiently explained. This study aimed to investigate the molecular and structural pattern of bonehealing in trabecular bone defects, with and without naturally derived resorbable membrane. Defectswere created in rat femurs and treated with the membrane or left empty (sham). After 3d, 6d and 28d,the defect sites and membranes were harvested and analyzed with histology, histomorphometry,quantitative-polymerase chain reaction (qPCR), Western blot (WB) and immunohistochemistry (IHC).Histomorphometry demonstrated that the presence of the membrane promoted bone formation in earlyand late periods. This was in parallel with upregulation of cell recruitment and coupled bone remodelinggenes in the defect. Cells recruited into the membrane expressed signals for bone regeneration (BMP-2,FGF-2, TGF-b1 and VEGF). Whereas the native membrane contained FGF-2 but not BMP-2, an accumu-lation of FGF-2 and BMP-2 proteins and immunoreactive cells were demonstrated by WB and IHC in thein vivo implanted membrane. The results provide cellular and molecular evidence suggesting a novel rolefor the membrane during GBR, by acting as a bioactive compartment rather than a passive barrier.© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Guided bone regeneration (GBR) is an established treatmentmodality to achieve bone regeneration, especially in the maxillo-facial region. GBR treatment is based on the application of a barriermembrane to cover an osseous defect. The concept of GBR wasdeveloped on the hypothesis that the membrane excludes non-osteogenic tissues from interfering with bone healing for efficientbone formation [1,2]. Although the GBR concept is generallyaccepted, the underlying biological mechanisms are as yet insuffi-ciently explained.

The first generation of the barrier membrane comprised non-

als, Institute of Clinical Sci-enburg, Box 412, SE-405 30,

(P. Thomsen).

Ltd. This is an open access article u

resorbable materials, e.g. expanded polytetrafluoroethylene (e-PTFE). This type of membrane has shown good biocompatibilityand the maintenance of structural integrity during healing. Thenon-resorbable membranes have generally shown positive resultswhen investigated in experimental studies and when used inclinical GBR procedures [3,4]. However, the need for additionalsurgery, for the removal of the membrane, has been regarded as adisadvantage [5]. A second generation of resorbable, natural orsynthetic, membranes has therefore been introduced, aiming toeliminate the need for the second surgical procedure.

Naturally derived collagen membranes have attracted a greatdeal of attention, since collagen is a principal component of con-nective tissue andmay provide structural support. As a biomaterial,collagen has a number of properties, including resorbability andlow immunogenicity [6]. Although some clinical reviews haveindicated comparable clinical outcomes between collagen mem-branes and non-resorbable membranes [7,8], other studies havesuggested that the collagen membranes may promote even better

nder the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

A. Turri et al. / Biomaterials 84 (2016) 167e183168

wound healing and bone regeneration [7]. One example is acollagen-based membrane derived from the extracellular matrix(ECM) of the porcine small intestinal submucosa [9]. This mem-brane consists of approximately 90% collagen (mainly Type I) [9],with smaller amounts of glycosaminoglycans [10], glycoproteinsand some growth factors [11]. Experimental studies have suggestedthat this ECM material may influence the regenerative capacity ofthe host and repair in several tissues and organs, including theurinary bladder [12], the abdominal wall [13], tendon [14], bloodvessels [15], articular cartilage [16] and bone [17]. Nevertheless, theunderlying mechanisms explaining how the ECM-derived collagenmembrane promotes tissue regeneration are largely unknown.

The main questions to be addressed in this study are: (i) Doesthe presence of a collagenous membrane modulate the expressionof factors crucial for inflammation, bone formation and remodelingin an underlying bone defect? (ii) Does the membrane host cells,which express factors crucial for bone regeneration? (iii) Is theexpression of these factors in the membrane related to the mo-lecular and structural development of the bone in the defect?

Using a rat experimental model, the present study employed aset of correlative analytical techniques (histology, histo-morphometry, quantitative-polymerase chain reaction (qPCR),Western blot and immunohistochemistry) to determine whetherthe membrane acts as a bioactive compartment, in contrast to theprevailing view that a membrane used for guided bone regenera-tion mainly serves as a passive barrier.

2. Materials and methods

2.1. Material

The extracellular matrix membrane used in this study wasextracted from porcine small intestine (DynaMatrix®; KeystoneDental, Boston, USA).

2.2. Scanning electron microscopy analysis of the membrane

The native membrane was cut into 5 � 5 mm pieces and soakedin 1% osmium tetroxide (OsO4) solution for 2 h at 4 �C. The sampleswere rinsed twice with 0.3% sodium cacodylate buffer and wereallowed to air-dry for 24 h. The samples were gold (Au) sputtercoated (approximately 10 nm) from both sides. Scanning electronmicroscopy was performed at high-vacuum, 5 kV acceleratingvoltage and 5e10 mm working distance in the secondary electronmode (Ultra 55 FEG SEM; Leo Electron Microscopy Ltd, UK). TheSEM characterization of the membrane is presented in (Fig. 1AeG).

2.3. Animal surgery

The animal experiments were approved by the Local EthicsCommittee for Laboratory Animals at the University of Gothenburg(dnr 279/2011). A total of 52 male SpragueeDawley rats were usedfor the in vivo experiments. The animals were anesthetized usingisoflurane (Isoba Vet, Schering-Plough, Uxbridge, UK) gas inhala-tion with the Univentor 400 anesthesia unit (Univentor, Zejtun,Malta). Anesthesia was maintained by the continuous administra-tion of isoflurane via a mask. The surgical site was shaved andcleaned with 5 mg/mL chlorhexidine and 70% ethanol. A linearincisionwasmade at the distal aspect of the femur, followed by skinand periosteal reflection. Bilateral femoral defects were created ineach animal. A defect was created in each femoral epiphysis(trabecular bone region) using a trephine with a 2.3 mm internaldiameter and 2.5 mm penetration depth under generous irrigationwith NaCl 0.9%. The bone harvested from the defect site wascollected from the trephine and preserved for the determination of

steady-state gene expression (baseline (BL); n ¼ 8). One defect wascovered with a membrane and the other defect was left uncovered(Fig. 1H, I). The subcutaneous and skin layers were closed, usingresorbable polyglactin (Vicryl, 4-0; Ethicon, USA) and resorbableMonocryl (4-0; Ethicon, USA) sutures respectively. The retrievalprocedure was performed at 3d, 6d and 28d (16 rats at each timepoint), when the animals were sacrificed using an overdose ofbarbiturate (Mebumal, ACO L€akemedel AB, Solna, Sweden). Theretrieval procedure was chosen depending on the subsequentanalytical technique (Fig. 1J, K). For histology and histo-morphometry at 6d and 28d (n ¼ 8), the skin was carefully reop-ened and the bone defect site with the overlying membrane andsoft tissue were harvested en bloc and preserved in formalin. Forgene expression and Western blot protein analyses at 3d, 6d and28d (n ¼ 8), the skin and soft tissues were carefully reopened andthe membrane was gently retrieved using tweezers. Subsequently,the defect site was retrieved using a 2.3 mm trephine. The mem-brane and the defect tissue samples were immediately and sepa-rately preserved in tubes containing RNAlater solution. Forimmunohistochemistry at 3d and 6d (n ¼ 6), the skin was carefullyreopened and the bone defect site with the overlying membraneand soft tissue were harvested en bloc and preserved in formalin.

2.4. Histology and histomorphometry

The formalin-fixed blocks were dehydrated by a graded series ofethanol and embedded in acrylic resin (LR White) (London ResinCompany Ltd, Berkshire, UK). The long axis of the defect was cutusing a diamond saw. Ground sections were prepared using sawingand grinding (Exakt Apparatebau GmbH & Co, Norderstedt, Ger-many), according to a previously published protocol [18]. Sectionswith a final thickness of 10e20 mmwere stained with 1% toluidineblue. All sections were coded and evaluated blindly for histologyand histomorphometry using a light microscope (Nikon EclipseE600). Histomorphometry was performed on each section using a10� objective. In each defect, the area percentages of newly formedbone were calculated. The measurements were performed on thetotal defect area level and by dividing the total defect area into top,middle and bottom regions or central and peripheral regions. Thiswas achieved by performing the measurements using a softwaregrid consisting of twelve zones, which covered the total area of thedefect. The top region was the area close to the membrane and wasrepresented by the sum of the top four zones of the grid. Each of themiddle and bottom regions also consisted of four zones. Further-more, the total area of the zones adjacent to the old bone wasregarded as the peripheral region of the defect, whereas the centralregion was the sum of the remaining zones. The area of newlyformed bonewas determined separately in every zone and the areapercentage of bone was then calculated with respect to the totaldefect area or to the area of the respective region.

2.5. Homogenization and total RNA extraction from the defectsamples

Total RNA from the defect site samples was extracted using anRNeasy® Mini kit (QIAGEN GmbH, Hilden, Germany). The homog-enization was performed in phenol/guanidine-based trizol lysis(QIAGEN GmbH, Hilden, Germany) using a 5 mm stainless steelbead (QIAGEN GmbH, Hilden, Germany) and Tissue Lyser (QIAGENGmbH, Hilden, Germany). Subsequently, the aqueous phase wasused for RNA extraction. The phase separation was performed bythe addition of chloroform, followed by centrifugation at 12,000 gfor 15 min. To reduce genomic DNA contamination, all sampleswere DNase treated with an RNase-free DNase during RNAextraction. The RNA quality was analyzedwith a Nano 6000 RNA kit

Fig. 1. Scanning electron microscopy (SEM) analysis of the native membrane and the surgical schedule. The SEM micrographs show the rough (A) and the smooth surfaces (B) of themembrane taken at a 30� tilt. The rough surface displays gross undulations and a fibrous texture (C, D). The smooth surface has a less exaggerated topography (E, F). Well-preserved,discrete layers are observed in cross-section (G). The micrographs (H and I) show the prepared defect (sham; H) and the prepared defect with the membrane (membrane; I). Theschedules (J and K) illustrate the different tissue components for the sham defect (J) and the defect covered with the membrane (K). The small letters indicate the differentcomponents allocated to the different analyses: a ¼ the defect site dissected en bloc, with surrounding tissue, for histology and histomorphometry analysis; b ¼ the defect siteretrieved by trephine for gene expression (qPCR) analysis; c ¼ the membrane retrieved gently by tweezers for gene expression (qPCR) and protein (WB) analyses.

A. Turri et al. / Biomaterials 84 (2016) 167e183 169

using the Bioanalyser 2100 Electrophoresis System (Agilent Tech-nologies; Santa Clara, CA, USA). The RNA concentration wasmeasured using a Nanophotometer P-36 (Implen GmbH, Munich,Germany).

2.6. Homogenization and total RNA and protein extraction from themembrane samples

For total protein and RNA extraction from the retrieved mem-brane samples, a NucleoSpin RNA/protein isolation kit (Macher-eyeNagel, Germany) was used. A similar procedure was used toextract total protein from native, unimplanted membrane samples(n ¼ 8). Prior to extraction, all samples were homogenized. Thehomogenization and the lysis of the membrane samples was per-formed in guanidinium thiocyanate lysis buffer (MachereyeNagel,

Germany), using a 5 mm stainless steel bead (QIAGEN GmbH, Hil-den, Germany) and Tissue Lyser (QIAGEN GmbH, Hilden, Germany).To reduce viscosity and clear the lysate, the lysate was filtratedthrough NucleoSpin® (MachereyeNagel, Germany) in collectiontubes and centrifuged for 1min at 11,000 g. Thereafter, NucleoSpin®

RNA/Protein columns (MachereyeNagel, Germany) were placed innew collection tubes and loaded with the lysate, followed bycentrifugation for 30 s at 11,000 g. At this step, the total RNA wasbound to the column membrane, which was then purified usingDNA digestion kit (MachereyeNagel, Germany) and subsequentlyused for reverse transcription and qPCR analysis (as describedbelow). The total protein, contained in the flow-through in thecollection tubes, was then precipitated, washed in ethanol anddried at room temperature, and was then used for Western blotanalysis (as described below).

A. Turri et al. / Biomaterials 84 (2016) 167e183170

2.7. Quantitative polymerase chain reaction (qPCR)

Before reverse transcription, the total RNA extracted from thedefect site and membrane samples was normalized to 20 ng/ml. Allreverse transcriptions were performed using a GrandScript cDNASynthesis Kit (TATAA Biocenter, Sweden). In order to select themost stable reference genes for normalization, a panel of ninereference genes was screened in five samples from the retrievedtissue, as well as from the retrieved membrane at each time period.The expression profiles of the screened reference genes wereevaluated using geNorm [19] and Normfinder [20] software. Themost stable expression was achieved by HPRT1, GAPDH and ACTB,which were selected as reference genes for the present analysis.The design of primers for fifteen target genes of interest was per-formed using Primer3 web-based software. The assays were pur-chased from TATAA Biocenter AB, Gothenburg, Sweden. The targetgenes analyzed in the defect site samples were: tumor necrosisfactor alpha (TNF-a), interleukin-6 (IL-6), chemokine receptor type4 (CXCR4), monocyte chemoattractant protein-1 (MCP-1), alkalinephosphatase (ALP), osteocalcin (OC), calcitonin receptor (CR),cathepsin K (CatK), receptor activator of NF-kappa B ligand(RANKL), osteoprotegerin (OPG) and receptor activator of NF-kappaB (RANK). The target genes analyzed in the retrieved membranesamples were bone morphogenetic protein 2 (BMP-2), fibroblastgrowth factor (FGF-2), transforming growth factor beta1 (TGF-b1)and vascular endothelial growth factor (VEGF). The analysis of thetarget genes and the best three selected reference genes was per-formed in a 10 ml reaction volume in duplicate on a CFX 96 Real timeSystem (Bio-Rad Laboratories, Inc., California, USA) with Grand-master SYBR mix (TATAA Biocenter, Gothenburg, Sweden). AnInter-plate Calibrator was used to compensate for the variationbetween runs. Quantities of the target genes were normalized us-ing the geometric mean of the Cq values of the selected referencegenes. The normalized relative quantities were calculated using thedeltaedelta Cq method and 90% PCR efficiency (k*1.9DDct) [21].

2.8. Western blot

Firstly, the total protein concentration (from native andretrieved membranes) was determined using a BCA protein assaykit (Pierce, Thermo Scientific, USA). Fifty mg from the protein extractof the native and the retrieved membrane samples was prepared inLaemmli sample buffer (Bio-Rad Laboratories, Inc., California, USA).The samples were heated at 95 �C for 5 min, cooled instantly andloaded onto a 10% TGX protean precast gel (Bio-Rad Laboratories,Inc., California, USA) for gel electrophoresis. The separated proteinbands were transferred from the gel to a nitrocellulose membrane(Bio-Rad Laboratories, Inc., California, USA). The non-specificbinding sites were blocked by rinsing the nitrocellulose mem-brane with Tris-Buffered Saline-Tween (TBST) containing 2% non-fat skimmed milk powder for 90 min at room temperature. Blotswere then probed, overnight at 4 �C, with the following primaryantibodies: rabbit polyclonal anti-BMP-2 (1:1000 dilutions)(18933-1-AP, Proteintech Group, Inc., USA) and rabbit polyclonalanti-FGF-2 (1:500 dilutions) (antibodies-online GmbH, Aachen,Germany), followed by rinsing with TBST 3 � 5 min. As the nativemembrane is obtained from the small intestinal submucosa of thepig, the primary antibodies were selected with respect to speciesreactivity for both rats and pigs. Protein from rat liver was used as apositive control for BMP-2 antibody. The blots were then incubatedwith the appropriate horseradish-peroxidase secondary goat anti-rabbit antibody (Santa Cruz Biotechnology, Inc., Texas, USA) at adilution of 1:10,000 for 1 h at room temperature. All antibodieswere diluted in 2% non-fat skimmed milk powder in TBST. Finally,the blots were washed 5 � 10 min in TBST. Band detection was

performed using the Chemiluminiscence with Clarity TM WesternECL Substrate detection kit (Bio-Rad Laboratories, Inc., USA). TheChemiDoc XRS þ system with Image Lab Software (Bio-Rad Labo-ratories, Inc., California, USA) was used for digital visualization.

2.9. Immunohistochemistry

After fixation in formalin, the dissected blocks were decalcified,for 10 days, in a 10% EDTA and subsequently embedded in paraffin.3-5-mm-thick sections were obtained using microtome andmounted on poly-L-lysine slides (Menzel GmbH and Co KG,Braunschweig, Germany). After deparaffinization, the sections werehydrated and incubated with either primary rabbit anti-rat BMP-2polyclonal antibody (18933-1-AP, Proteintech Group, Inc., USA) orprimary rabbit anti-rat FGF-2 polyclonal antibody (ABIN1582147,antibodies-online GmbH, Germany). To exclude false positivereactivity, negative control slides were prepared using the sameprotocol but with the omission of the primary antibody and theincubationwith 1% BSA in PBS. The immunoreactivity of BMP-2 andFGF-2 was detected and visualized using goat anti-rabbit IgG sec-ondary antibody with horseradish peroxidase (HRP) (sc-2004,Santa Cruz Biotechnology, USA) and Betazoid DAB Chromogen kit(Biocare Medical, USA). The qualitative analysis was performedusing a light microscope (Nikon Eclipse E600) (objectives x20 andx40).

2.10. Statistics

For histomorphometry and gene expression analyses, compari-sons were made between the sham and the membrane groups andbetween the different time periods for each group (n ¼ 8). Inaddition, the statistical differences between the steady state(baseline (BL); n¼ 8) and the different time points were calculated.A non-parametric Wilcoxon's signed rank test was used to identifydifferences between the sham and membrane groups, while a non-parametric KruskaleWallis test followed by a ManneWhitney testwere used to identify statistical differences between the time BLand the different time points. In addition, Spearman's correlationanalysis was performed on the expression levels of the analyzedgenes in the defect, with and without membrane, after 3d, 6d and28d. The correlation analysis was also performed on the differentanalyzed genes in the membrane, as well as on the genes in themembrane versus the genes in the equivalent defect samples afterpooling the expression levels for all healing time periods (3d, 6dand 28d). All analyses were performed using SPSS Version 10software (SPSS, Inc., New York, USA). The significance level was setat 0.05 for the comparative analyses and at 0.01 for the correlativeanalyses.

3. Results

3.1. Macroscopic observations

Macroscopic observations during retrieval surgery showed thatthe sites had healed uneventfully, with no visual manifestations ofsurgical complications or inflammation during the experimentalperiod. At 28d, whereas the defect margins could still be identifiedin the sham group (Fig. 2A), the union of the defect margins wasobserved in the membrane group (Fig. 2B).

3.2. Histology and histomorphometry of the defect site

At 6d, the histological examination revealed a well-defineddefect with clear borders in trabecular bone. The implantedmembrane appeared beneath the soft tissue, overlying the created

Fig. 2. Macroscopic observations after 28d of healing. The micrographs show the defect sites after 28d of healing in the sham defect (A) and in the membrane defect (B), after thedissection of the soft tissue and the removal of the membrane. In the sham group, the defect is not fully restituted, as the defect boundaries can still be observed and the entrapmentof soft tissue in the defect is frequently encountered.

Fig. 3. Histology of the defects. The micrographs show non-decalcified toluidine blue-stained ground sections of femur defects after 6d (AeH) and 28d (IeP) of healing. The defectsare sham (AeD and IeL) or covered with membrane (EeH and M�P). The membrane (m) appears underneath the soft tissue (ST), covering the defect (d). The new bone (NB) formedat 6d is primarily localized in the bottom peripheral regions in both defects. In the membrane-treated defect, more islands of NB are also observed in the top region of the defectunder the membrane. At 28d, a lower proportion of NB is observed in the sham defects compared with the membrane defects. The amount of NB is apparently higher in the top-peripheral and top-central regions of the membrane defects compared with the equivalent regions in the sham defects. The ST occupies a large proportion of the defect area in thesham group.

A. Turri et al. / Biomaterials 84 (2016) 167e183 171

A. Turri et al. / Biomaterials 84 (2016) 167e183172

defect (Fig. 3). No major inflammatory infiltrates were detected inthe defect, in either the sham or the membrane groups. Moreover,signs of intramembranous osteogenesis were observed, with theformation of osteoid tissue and new bone in both defects (Fig. 3A,E). A large proportion of newly formed bone, extending from theold bone, was observed at the bottom of the defect. In addition,solitary islands of new bone were detected along the peripheralboundaries of both defects. In the membrane-treated defect, moreislands of newly formed (woven) bone were observed in the topregion of the defect under the membrane (Fig. 3F, G).

Histomorphometry of the defects at 6d showed that the per-centage of total area of new bone was approximately 8.6%, in boththe sham and themembrane groups (Fig. 4A). A significantly higherproportion of newly formed bone was detected in the top level ofthe membrane-treated defect (3.2 ± 0.7) compared with the toplevel in the sham defect (1.4 ± 0.4) (Fig. 4C). Otherwise, no majordifferences in new bone formation between the membrane-treatedand sham defects were found, in either the middle and bottomlevels, or in the central or the peripheral regions (Fig. 4B, C).

At 28d, a substantial amount of bone had formed in both defects(Fig. 3I, M), especially in themembrane group (Fig. 3M). The bone atthis time point appearedmature andwell mineralized, as judged bythe toluidine blue staining showing comparable intensity betweenthe old and new bone (Fig. 3J, N). One major observation at this latetime point was the level of defect restitution and the amount ofbone formation. The sham defect showed fewer bone trabeculae ofmature bone, bordered endosteally by the bone marrow and peri-osteally by the overlying soft tissue. In addition, the contour of theoriginal bone in the sham defects was not restored and soft tissueingrowth was apparently observed inside the defect (Fig. 3IeL). Incontrast, the membrane-treated defect revealed a larger amount ofmature bone, as well as a higher degree of defect restitution after28d (Fig. 3MeP).

Histomorphometry revealed two major differences between thesham and the membrane-treated defects at 28d. Firstly, the per-centage of total bone areawas significantly higher in themembranegroup (28 ± 1.7) compared with the sham group (25 ± 1.2) (Fig. 4A).Secondly, a significantly higher proportion of bone was detected atthe top level (39.8 ± 4.8), as well as in the central region (23.6 ± 3.1)of the membrane defect, compared with the top level (27.4 ± 2.1)and the central region (16.5 ± 2.5) of the sham defect (Fig. 4B, C).

Temporally, from 6d to 28d, themeasurement of bone formationat the different topological levels of the defect (top, middle andbottom) revealed different patterns in the two experimental groups(Fig. 4C). At 6d, the highest proportion of bone was revealed at thebottom levels of the defects in both groups (sham; 17.1 ± 2.7,membrane; 15.8 ± 2.7), followed by the middle level (sham;10.7 ± 3.7, membrane; 9.5 ± 2.4), while the lowest proportion wasobserved at the top level. After 28d, the membrane-treated defectshowed the highest proportion at the top level (39.8 ± 4.8), fol-lowed by the middle level (25.3 ± 4.2), while the lowest wasdetected at the bottom level (19.4 ± 1.5). At 28d, in the sham defect,the lowest bone proportion was detected at the lower level of thedefect (18.9 ± 3.0), whereas the top and middle levels showedsimilar bone proportions (approximately 27.5).

3.3. Gene expression analysis (qPCR) of the defect

Analyses were made of the gene expression in tissue samplesretrieved during defect preparation (baseline; BL) and the tissueharvested from the sham and the membrane-covered defects after3d, 6d and 28d. The targeted genes in the defect site samples werethose involved in inflammation, cell recruitment, bone formationand resorption, as well as the coupling of bone remodeling.

3.3.1. Gene expression of inflammation and cell recruitmentmarkers

Inflammatory markers: with reference to BL, the temporalpattern of TNF-a in the membrane defect revealed a significant 2.5-fold increase after 3d, followed by a significant reduction after 6dand no major change thereafter (Fig. 5A). On the other hand, nomajor change in the temporal expression of TNF-a was observed inthe sham defect. The temporal pattern of IL-6 expression for bothmembrane and sham defects revealed an increase over time,reaching the highest level at 6d (about 10-fold in comparison withBL) (Fig. 5B). From 6d to 28d, a significant downregulation of IL-6was found for both groups.

When comparing the two experimental groups at each timepoint, the expression level of TNF-a was about 2.5-fold, signifi-cantly higher in the membrane group compared with the sham at3d (Fig. 5A). This switched at 6d, where the membrane defectrevealed about a 2-fold, significantly lower expression of TNF-acompared with the sham. Later, at 28d, no significant differencewas observed between the two groups. With respect to IL-6expression at 3d and 6d, no significant differences were observedbetween the two groups (Fig. 5B). At 28d, a 2-fold, significantlyhigher expression of IL-6 was found in the membrane defectcompared with sham.

Cell recruitment factors: for both groups, the temporal patternof MCP-1 revealed significantly higher levels at all analyzed timepoints compared with BL (Fig. 5C). A sharp increase in MCP-1expression was detected after 3d in comparison to BL (sham 6-fold; membrane 14.5-fold). The MCP-1 expression levels weremaintained at 6d, followed by a reduction at 28d, for both groups. Adifferent temporal profile was observed for the CXCR4, which, forboth groups, revealed significantly lower expression levels at 3dand 6d comparedwith BL (Fig. 5D). From 6d to 28d, a sharp increasewas demonstrated for CXCR4, attaining a comparable level to BL inthe sham group and significantly exceeding the level of BL in themembrane group.

When comparing the membrane and sham groups at each timepoint, a similar trendwas observed for the expression of MCP-1 andCXCR4 (Fig. 5C, D). At 3d, both MCP-1 and CXCR4 were expressed athigher levels (2.4-fold) in the membrane group compared with thesham. However, this was only statistically significant for the CXCR4.No difference between the two groups was observed for both genesat 6d. At 28d, both MCP-1 and CXCR4 were significantly higher inthe membrane group compared with the sham (Fig. 5C, D).

3.3.2. Gene expression of bone formation and bone resorptionmarkers

Bone formation markers: the temporal changes in the boneformation markers exhibited a comparable profile in the sham andthe membrane groups (Fig. 6A, B). This was characterized by theearly downregulation of both ALP and OC at 3d compared with BL.Whereas ALP revealed a sharp increase (peak level after 6d), fol-lowed by a marked reduction (Fig. 7A), OC showed a gradual in-crease over time, attaining the highest level at 28d (Fig. 6B). Thepeak level of ALP at 6d significantly exceeded the BL level, whereasthe peak level of OC at 28d was similar to the BL.

A comparison of the two experimental groups at each time pointfailed to reveal any significant differences in ALP expression(Fig. 6A). In contrast, the expression of OC (Fig. 6B) was 17-fold and1.5-fold significantly higher in themembrane group comparedwiththe sham, at 3d and 28d respectively.

Bone resorption markers: the temporal pattern of the osteo-clastic surface marker, CR, and the osteoclastic activity marker(CatK) revealed two distinct patterns (Fig. 6C, D). The expression ofCR (Fig. 6C) was generally lower in the sham defects comparedwithBL, whereas levels comparable to the BL were observed in the

Fig. 4. Histomorphometry analysis of bone formed in defects with or without membrane. The column graphs show the bone area percentages after 6d and 28d of healing, whereasthe schematic diagrams show the respective area of measurements. The percentage of total bone area was calculated per the total area of the defect (A). The percentage of bone areawas also calculated in different regions of the defect: central and peripheral (B), top, middle and bottom (C). The results are presented as the mean ± SEM. Statistically significantdifferences are indicated by asterisks.

A. Turri et al. / Biomaterials 84 (2016) 167e183 173

membrane-treated defect. On the other hand, the temporalexpression of CatK (Fig. 6D) revealed significantly lower levels forboth groups at 3d compared with BL. Thereafter, CatK expressiongradually increased, for both groups, attaining the highest levelsafter 28d. The peak expression of CatK at 28d was significantlyhigher than the BL in the membrane group.

When comparing the two experimental groups, the CR and CatKexpression levels were significantly higher in the membrane group

compared with the sham after 3d (5-fold and 2.5-fold respectively)(Fig. 6C, D).

3.3.3. Gene expression of bone remodeling coupling factorsThe temporal pattern of RANKL and OPG in the sham group

showed a significant reduction at 3d compared with BL (Fig. 7A, C).On the other hand, in the membrane group at 3d, both RANKL andOPG revealed levels comparable to BL (Fig. 7A, C). From 3d to 6d, for

Fig. 5. Gene expression of pro-inflammatory cytokines and cell recruitment chemokines. The analysis was performed on the tissue harvested from defects with or withoutmembrane after 3d, 6d and 28d of healing. The results are presented as the mean ± SEM. Statistically significant differences are indicated by small letters: a ¼ significant differencewith baseline (BL); b ¼ significant difference with sham; c ¼ significant difference between 3d and 6d; d ¼ significant difference between 6d and 28d.

A. Turri et al. / Biomaterials 84 (2016) 167e183174

both groups, a sharp increase in the expression of the RANKL andOPG was detected, reaching significantly higher levels comparedwith the BL (Fig. 7A, C). From 6d to 28d, the expression levels ofRANKL and OPG were significantly reduced, revealing levels com-parable to the BL in the membrane group and lower than the BL inthe sham group at 28d. RANK expression (Fig. 7B) showed a similartemporal pattern for the sham and the membrane groups. It did notchange significantly at 3d compared with BL but increased signif-icantly after 6d, followed by a significant decrease thereafter(Fig. 7B). Nomajor temporal changes were observed for the RANKL/OPG expression ratio (Fig. 7D).

The comparative analysis at 3d revealed significantly higherRANKL expression in the membrane-treated defect as compared tothe sham defect (Fig. 7A). On the other hand, at 6d, both RANKL andRANK were significantly higher in the sham defect compared withthe membrane-treated defect (Fig. 7A, B). At 28d, no major differ-ences were observed between the sham and membrane groupswith respect to any of the analyzed remodeling coupling genes. Nosignificant differences in OPG expression and RANKL/OPG expres-sion ratio were found between the two groups at any time point(Fig. 7C, D).

3.3.4. Correlation analysis of genes expressed in the defectThe results of the correlation analyses of the different genes in

the defect are presented in Table 1. Several correlations weredetected, mainly at 3d and 28d, and all revealed a positive rela-tionship. At 3d, MCP-1defect revealed significant correlations withthe pro-inflammatory cytokine, TNF-adefect, the osteoclastic re-ceptor gene, CRdefect, and the coupling gene, OPGdefect. On the otherhand, CXCR4 correlated significantly with the osteoblastic gene,OCdefect, and the osteoclastic gene, CatKdefect. A significantly posi-tive relationship was also found between the osteoblastic gene,OCdefect, and the osteoclastic gene, CatKdefect. At this early timepoint, a strong positive correlation was also detected between thepro-inflammatory cytokine gene, TNF-adefect, and the osteoclasticreceptor gene, CRdefect.

At 28d, the CXCR4defect was observed in relationships withOCdefect and CatKdefect, similar to those detected at the 3d timepoint. At this late time point (28d), CXCR4defect displayed a positivecorrelation with the coupling factor gene, RANKLdefect. The osteo-blastic genes, ALPdefect and OCdefect, were found in positive re-lationships both with one another and with the osteoclastic gene,CatKdefect. Moreover, the OCdefect revealed significant correlationswith the coupling genes, RANKLdefect and RANKdefect. In addition,the CatKdefect revealed significant correlations with the RANKLdefectand OPGdefect.

Fig. 6. Gene expression of bone formation and resorption factors. The analysis was performed on the tissue harvested from defects with or without membrane after 3d, 6d and 28dof healing. The results are presented as the mean ± SEM. Statistically significant differences are indicated by small letters: a ¼ significant difference from baseline (BL);b ¼ significant difference from sham; c ¼ significant difference between 3d and 6d; d ¼ significant difference between 6d and 28d.

A. Turri et al. / Biomaterials 84 (2016) 167e183 175

3.4. Histology of the membrane

At 6d, the collagen strands of the membrane appeared to beintact, with a relatively uniform thickness and a unidirectionalarrangement (Fig. 8A). A large population of cells was observed atthe membrane peripheries, where many of these cells appeared tobe migrating from the surrounding tissue into the membrane,through the open spaces between the strands (Fig. 8B). Therecruited cells, around and inside themembrane, assumed differentphenotypes, with both hematopoietic (polymorphonuclear cellsand monocyte-/macrophage-like cells) and mesenchymal origins(Fig. 8B) but with a predominance of inflammatory cells. After 28dof healing, the membrane was still observed at the site of the im-plantation (Fig. 8C, D). However, the collagen strands of themembrane were thinner, disrupted and apparently in a randomarrangement (Fig. 8C). Furthermore, tissue and cells, surroundingand infiltrating themembrane, showed a relatively higher degree oforganization and homogeneity. At this stage, polymorphonuclearcells were seldom detected, within or in proximity to the mem-brane. In spite of this, cells of hematopoietic origin, monocyte- ormacrophage-like cells, were observed in conjunction with stromalcells (Fig. 8D). In addition, many osteoclast-like cells wereobserved, at 28d, in connection with resorption areas on the boneside at the boundary between themembrane and the newly formed

bone (Fig. 8D).

3.5. Gene expression (qPCR) of the retrieved membrane

The gene expression of four selected growth factors wasanalyzed in the retrieved membranes at 3d, 6d and 28d of healing.The temporal expression pattern revealed a steady increase overtime for TGF-b, FGF-2 and BMP-2 (Fig. 8E, F, G). However, thesesteady increases were statistically significant for TGF-b and FGF-2but not for BMB-2. The opposite trend was observed for VEGFexpression, where the highest expression was detected at 3d and itsteadily decreased thereafter (Fig. 8H).

3.5.1. Correlation analysis of genes expressed in the retrievedmembranes and the equivalent defects

The results of the correlation analysis of the different genes inthe membrane, as well as between the genes in the membrane andthe genes in the equivalent defects, are presented in Table 2. Theanalysis was performed between the different genes after poolingthe expression levels during the entire period of GBR.

The membrane compartment showed a significantly positivecorrelation between FGF-2membrane and TGF-bmembrane. Both FGF-2membrane and TGF-bmembrane genes exhibited a significantly nega-tive correlation with VEGFmembrane.

Fig. 7. Gene expression of bone remodeling coupling factors. The analysis was performed on the tissue harvested from defects with or without membrane after 3d, 6d and 28d ofhealing. The results are presented as the mean ± SEM. Statistically significant differences are indicated by small letters: a ¼ significant difference from baseline (BL); b ¼ significantdifference from sham; c ¼ significant difference between 3d and 6d; d ¼ significant difference between 6d and 28d.

Table 1Correlation analysis of genes expressed in the defect, with and without membrane.The data show genes that revealed significant correlations, separately for eachhealing time period (3d, 6da and 28d). Spearman correlation coefficients (r) andlevel of significance (p-values) are presented.

3d 28d

Genes (r) (p) Genes (r) (p)

MCP-1defect/TNF-adefect 0.86 0.007 CXCR4defect/OCdefect 0.80 0.001MCP-1defect/CRdefect 0.88 0.004 CXCR4defect/CatKdefect 0.73 0.01MCP-1defect/OPGdefect 0.93 0.001 CXCR4defect/RANKLdefect 0.75 0.002CXCR4defect/OCdefect 0.86 0.007 ALPdefect/OCdefect 0.67 0.009CXCR4defect/CatKdefect 0.83 0.01 ALPdefect/CatKdefect 0.75 0.002TNF-adefect/CRdefect 0.93 0.001 OCdefect/CatKdefect 0.70 0.005OCdefect/CatKdefect 0.88 0.004 OCdefect/RANKLdefect 0.74 0.003

OCdefect/RANKdefect 0.80 0.0003CatKdefect/RANKLdefect 0.80 0.001CatKdefect/OPGdefect 0.80 0.001

a No significant correlations were detected at 6d, except for one positive corre-lation: ALPdefect/OPGdefect (r ¼ 0.73; p ¼ 0.003).

A. Turri et al. / Biomaterials 84 (2016) 167e183176

Several correlations were demonstrated between the differentgrowth factor genes in the membrane and the genes denoting cellrecruitment and differentiation in the defect. Significantly positivecorrelations were shown for TGF-bmembrane, FGF-2membrane and

BMP-2membrane with OCdefect and CatKdefect. On the other hand,VEGFmembrane revealed a significantly negative correlation withOCdefect and CatKdefect. Moreover, a significantly negative correla-tion was displayed between TGF-bmembrane and the chemotacticgene, MCP-1defect.

3.6. Protein expression in the native and retrieved membrane

The presence of selected growth factor proteins (FGF-2 andBMP-2) in the native and the retrieved membranes at 3d, 6d and28d was investigated. The blot revealed a FGF-2 protein band be-tween 17 and 20 kDa for both native and retrieved membrane. Thedetected band was deeply stained at the blot sites for the retrievedmembrane samples, regardless of the retrieval time points (Fig. 8I).The band for BMP-2, between 15 and 20 kDa, was also detected onthe blotted protein from the retrieved membranes at all retrievaltime points. In contrast, BMP-2 was not detected in the nativemembrane (Fig. 8I).

The presence of BMP-2 and FGF-2 proteins, in the membranecompartment during GBR, was also analyzed by immunohisto-chemistry (IHC) of the membrane in situ overlying the defect.Negative control sections (Fig. 8J, K) excluded any possible falsepositive staining, and confirmed the specificity of the employedantibodies. IHC demonstrated immunoreactivity for both BMP-2

Fig. 8. Histology, gene expression and protein analyses of the membrane. The micrographs (AeD) show non-decalcified toluidine blue-stained ground sections of the membrane inplace after 6d (A and B) and 28d (C and D) of healing. At 6d, the collagen strands of the membrane appeared to be more intact, with relatively uniform thickness and a unidirectionalarrangement (A). At this time point, the spaces between the membrane strands appeared to be infiltrated with cells from different phenotypes, including cells with an hemato-poietic origin (polymorphonuclear and monocyte-/macrophage-like cells; some of them are indicated by black arrows in B), as well as with a mesenchymal origin (MSC- andfibroblast-like cells; some of them are indicated by red arrows in B). At 28d, the collagen strands of the membrane appeared to be more distorted, with varying thickness and arandom arrangement (C). At this time point, the spaces between the membrane strands were still infiltrated with cells from different phenotypes, including cells with an he-matopoietic origin (monocyte-/macrophage-like cells; some of them are indicated by black arrows in D), as well as with a mesenchymal origin (MSC- and fibroblast-like cells; someof them are indicated by red arrows in D). Polymorphonuclear cells were seldom detected within or in the proximity of the membrane at this time point. Furthermore, manyosteoclast-like cells were evident at this time point at the boundary between the lower surface of the membrane and the newly formed bone (some of them are indicated by blackarrowheads in D). The column graphs (EeH) show the gene expression of selected growth factors TGF-b1 (E), FGF-2 (F), BMP-2 (G) and VEGF (H) in the membrane-associated cells.The analysis was performed on the retrieved membrane after 3d, 6d and 28d of healing. The results are presented as the mean ± SEM. Statistically significant differences betweendifferent time points are indicated by asterisks. The image of the Western blot analysis of the membrane (I) shows the bands of FGF2 and BMP-2 proteins. The analysis wasperformed on both native membrane and membrane retrieved from the defect sites after 3d, 6d and 28d of healing. Rat liver protein was used as a positive control for BMP2antibody. The micrographs (JeO) show the immunohistochemical analysis of decalcified sections of the membrane in place after 3d and 6d of healing. The micrographs show theBMP-2 (L and N) and FGF-2 (M and O) immunoreactivity in cells and extracellularly in the membrane compartment at 3d (L and M) and 6d (N and O) of healing. The intracellular andextracellular staining of the two proteins is exemplified with white arrows and arrowheads, respectively, at the 3d time point (in L and M). The micrographs (J and K) show the 3dnegative control samples for the immunostaining of BMP-2 and FGF-2, respectively.

A. Turri et al. / Biomaterials 84 (2016) 167e183 177

and FGF-2 inside the membrane at the two evaluated periods, 3dand 6d (Fig. 8LeO). At both time points, the expression of BMP-2and FGF-2 proteins was detected intracellularly and interstitiallyin-between the strands of the collagen membrane (Fig. 8LeO).

4. Discussion

The present experimental study investigated the mechanisms ofguided bone regeneration (GBR), in trabecular bone defects, using

an extracellular matrix-derived membrane. The spatial and tem-poral cellular and molecular events were determined, at differenttime points of healing, in the defect, underneath themembrane andin themembrane itself. The results of the study provided structural,cellular andmolecular evidence suggesting that the membrane actsas a bioactive compartment, rather than merely acting as a passivebarrier. The main findings can be summarized as following: firstly,the membrane attracts cells of different phenotypes, whichsequentially express and secrete factors and signals for bone

Table 2Correlation analysis of genes expressed in the retrieved membranes and the equivalent defect samples. The data show genes that revealed significant correlations by poolingthe expression levels for all healing time periods (3d, 6d and 28d). Spearman correlation coefficients (r) and level of significance (p-values) are presented.

Correlations between genes in the membrane Correlation between genes in the membrane with genes in the equivalentdefects

Genes (r) (p) Genes (r) (p)

TGF-bmembrane/FGF-2membrane 0.80 0.00002 TGF-bmembrane/MCP-1defect �0.68 0.001TGF-bmembrane/VEGFmembrane �0.80 0.00008 TGF-bmembrane/OCdefect 0.83 0.00001

TGF-bmembrane/CatKdefect 0.81 0.00002FGF-2membrane/VEGFmembrane �0.87 0.000002 FGF-2membrane/OCdefect 0.78 0.00009

FGF-2membrane/CatKdefect 0.77 0.0001VEGFmembrane/OCdefect �0.80 0.00004VEGFmembrane/CatKdefect �0.73 0.0004BMP-2membrane/OCdefect 0.76 0.0001BMP-2membrane/CatKdefect 0.59 0.008

A. Turri et al. / Biomaterials 84 (2016) 167e183178

regeneration and remodeling (BMP-2, FGF2 and TGF-b) andvascularization (VEGF). Secondly, the membrane compartmentpromotes the expression of chemotactic factors and modulates theosteogenic and remodeling processes in the treated defect, pri-marily underneath the membrane. Thirdly, these membrane-induced processes result in an accelerated woven bone formationand, during the course of healing, superior defect restitution withbone.

4.1. Bone formation and bone remodeling in defects covered withmembrane

Histological observations revealed a higher degree of bone for-mation, preferentially at the top level and the central region of thedefect, leading to a higher overall degree of bone restitution in thedefect. This is in contrast to a smaller amount of bone and therelative collapse of the defect in the absence of the membrane. Therestitution with bone is in line with commonly observed clinicaloutcomes of GBR, using different types of membrane in differenttypes of bone defect [22e25].

The present study shows that the increased bone formation ispromoted at the molecular level, whereby the presence of themembrane significantly induced the expression of OC, a majorosteoblastic and bone formation gene. The osteogenic activity wasconfirmed during both the early and late phases of healing. Theearly membrane-induced upregulation most likely corresponds toincreased osteogenic differentiation and the observed early wovenbone formation. On the other hand, the late-phase effect of themembrane on OC expression suggests the continuous process ofmembrane-induced bone formation and maturation.

Topologically, and irrespective of the presence or absence of themembrane, new bone formation was distributed in two differentpatterns at early (6d) and late (28d) healing time points. At theearly time point, bone formed predominantly in the bottom andperipheral regions of the defect, whereas the top and central re-gions gained a significant amount of bone in the late phase. Themembrane compartment played a major role in boosting this nat-ural response. Firstly, the microenvironment created directly underthe membrane provided osteoinductive signals that rapidly pro-moted bone formation at the top level of the defect. This prefer-ential effect of the membrane-induced microenvironment on boneformation at the top level of the defect extended to the late periodof healing. Secondly, the presence of the membrane microenvi-ronment induced high osteoclastic activity, demonstrated by theupregulated expression levels of the osteoclastic receptor (CR) andosteoclastic activity marker (CatK). The finding that the membraneprovides a local environment for coupled bone formation andresorption is further supported by the observation of themembrane-induced upregulation of RANKL, which is a major

coupling factor between osteoblasts and osteoclasts [26,27].Moreover, the multiple positive correlations demonstrated be-tween the osteoblastic and osteoclastic genes in the defect indicatetight links between the expression activities. Due to this tightcoupling and the mutual cross-talk between osteoblasts and oste-oclasts, it has been suggested that the activation of one arm of theremodeling, i.e. the coordinated bone formation and resorption,will be accompanied by the activation of the other arm [28,29].Whether the membrane-induced effects primarily influenced theosteoblastic or osteoclastic lineages, or both simultaneously, re-mains to be explored.

The present evidence of the triggered coupled bone remodelingactivities, induced by themembrane as an implanted biomaterial, isin line with recent studies of biomaterial-driven bone regenerationprocesses, osseointegration and bone defect augmentation [30,31].Titanium implants with oxidized surfaces [32] and alpha-tricalciumphosphate granules with octacalcium phosphate surfaces [30] bothpromoted higher osteoblastic, osteoclastic and coupling activities,which, in parallel, promoted a higher degree of bone regenerationand the integration of the respective materials in the recipientdefect sites. Hitherto, the exact roles of osteoclasts and their ac-tivities in bone healing and restitution around biomaterials areunknown. Recent insights suggest that the osteoclast is a majorstimulus for bone formation by different routes: (i) osteogenicfactors released from bone during active resorption and (ii) osteo-genic factors secreted directly by osteoclasts [33e38]. Nevertheless,others have postulated that the osteoclast-generated osteogenicand coupling signals are not sufficient to induce the osteogenicactivity and that a complementary prerequisite is the recruitmentof osteoprogenitors to the local site [39,40]. In the present study ofthe mechanism of GBR, it is evident that the membrane compart-ment offers conditions that promote high osteoclastic activity, aswell as inducing high recruitment activity for osteoprogenitors inthe defect.

4.2. Cell recruitment and inflammation in defects covered withmembrane

The membrane-established microenvironment was rapidlysensed by cells within the defect, responding by upregulatingfundamental components in cell recruitment axes, MCP-1 andCXCR4. MCP-1 is a chemokine which is majorly responsible formonocyte trafficking to the site of injury [41,42]. It has been alsoimplicated in the recruitment and differentiation of osteoclastprecursors at the sites of bone remodeling [43]. Mutant micelacking the receptors of MCP-1 displayed a reduced number andimpaired function of osteoclasts and eventually revealed delayedfracture healing [44,45]. CXCR4 is a surface receptor expressed byvarious cells, including leukocytes [46], MSCs and osteoprogenitors

A. Turri et al. / Biomaterials 84 (2016) 167e183 179

[47e49] and osteoclast precursors and osteoclasts [50]. In recentyears, it has first and foremost attracted attention as a majorelement in regulating MSC migration and recruitment to bonerepair sites [48,51,52]. Bone defects in mutantmice partially lackingCXCR4 (CXCR4þ/� mice) showed reduced bone formation after 14d[51]. In the present study, the early, membrane-induced upregu-lation of MCP-1 and CXCR4, at 3d, corresponded to the recruitmentof different cell phenotypes. This assumption is based not only onthe fact that CXCR4 is a recruitment element expressed by differentcells but also on the early overlap between the upregulation ofCXCR4 expression and the upregulation of genes denoting inflam-matory cells, osteoblasts and osteoclasts. At the late stage, the peakof CXCR4 corresponded mainly to the peaks of OC and CatK, sug-gesting a main involvement, at this stage, in bone maturation andremodeling. Exploration of the correlation results further supportsthe assumption that MCP-1 and CXCR4 were involved in therecruitment of different cells at different stages. In the early healingstage, whereas MCP-1 expression suggested an involvement in therecruitment of inflammatory cells and osteoclast precursors, CXCR4revealed a positive correlation mainly with osteoblastic and oste-oclastic phenotypes. Correlation analysis at the late stage indicateda strong association between CXCR4 and osteoblastic and osteo-clastic activities. The present results thus suggest that themembrane-induced microenvironment enhances chemotactic cueswhich, in a time-specific manner, augment the recruitment ofdifferent cells corresponding to the ongoing biological process.These effects include the early influx of inflammatory cells andprecursors for osteoblasts and osteoclasts, as well as the laterecruitment of these cells for bone maturation and remodeling.

In the present study, an evident histological feature in themembrane-treated defect was the frequent appearance of giantmultinucleated and osteoclast-like cells, particularly in the zonebetween the lower surface of the membrane and the upper surfaceof the newly formed bone, in the top region of the defect. At his-tological level, these osteoclast-like cells appeared to be activelyresorbing the underlying bone, but it was not possible to determinewhether these cells were also involved in a degradation process ofthe membrane. The increase in CXCR4 and MCP-1 activity may playa role in the tissue responsewithin and immediately adjacent to themembrane material, e.g. providing cells for bone regeneration inthe defect region directly beneath the membrane (i.e. the top re-gion of the defect) in parallel with a contribution to the degradationprocess of the membrane. The first possibility is supported by ob-servations that a collagen pellet containing the pro-osteogenicfactor, BMP-2, enhanced the recruitment of osteoblast progenitorsand induced ectopic bone formation at the periphery of thecollagen pellet. These effects were strongly inhibited by blockingthe CXCR4 using antibodies [53]. The second possibility is in linewith earlier studies of similar membrane materials, showing thestrong upregulation of MCP-1 and macrophage inflammatoryprotein-1 (MIP-1) [54] and the enhanced recruitment of macro-phages, associated with a higher rate of graft remodeling [55].

Bone healing and regeneration is a complex process, consistingof a sequence of biological events, including inflammation and cellrecruitment, bone formation and remodeling. Studies of bonehealing around biomaterials have demonstrated that the earlyinflammation, which is an integral part of the healing process,precedes and even overlaps the recruitment and differentiation ofthe bone-forming and -resorbing cells [31,56]. By virtue of theirearly recruitment to the site of healing and their wide secretoryprofile, monocytes/macrophages have been suggested as a potentmoderator of the healing events. These cells may sense the prop-erties of the implanted material and translate and communicatethem to the neighboring cells, around the implanted material [57].In the present study, the presence of the membrane strongly

modulated the expression of the inflammatory cytokines, TNF-aand IL-6, in the healing defect. The early upregulation of TNF-a inthe membrane-treated defect suggests its engagement in the earlyrecruitment of mesenchymal cells and osteoprogenitors and theirsubsequent differentiation into bone-forming cells. In vitro studieshave shown that recombinant TNF-a enhances the invasion ca-pacity and proliferation of MSCs via the activation of IkB kinase[58]. Other studies have revealed that recombinant TNF-a added toMSCs under osteogenic culture conditions augments the expressionof BMP-2 and ALP and promotes higher mineralization via mech-anisms that involve the activation of the NF-kB pathway [59].Further, conditioned media containing a high level of TNF-a,secreted by classically activated pro-inflammatory monocytes/macrophages, strongly upregulate the osteogenic genotype of MSCscultured in normal non-osteogenic conditions [60]. The pro-osteogenic potential of TNF-a has also been demonstrated in vivowhere the intramembranous ossification was impaired in rats afterknocking out the receptors of TNF-a (TNFR1 and TNFR2) [61]. Withrespect to osteoclasts, there is general consensus on the pro-osteoclastic effect of TNF-a. The precise roles of the membrane-induced TNF-a expression are not clear. Based on the present re-sults, both osteogenic and osteoclastic differentiation may havebeen affected. TNF-a and IL-6 may be required to promote thehigher remodeling of the defect at the late stage of healing [62,63].The correlation analysis in the present study at least partiallysupports a pro-osteoclastic role for TNF-a, based on the strongcorrelation between TNF-a and the osteoclastic receptor, CR, at theearly stage of healing. Albeit speculative, the lack of statisticalpositive or negative correlations between TNF-a and the osteo-blastic markers could indicate that the fluctuation in TNF-aexpression among the individual samples is compensated for byother factors. It is also possible that the pro-osteogenic effect ofTNF-a, in vivo, is dependent on several factors in order to induce theosteogenic response.

4.3. The role of inflammation in coupled bone remodeling duringGBR

The role of pro-inflammatory cytokines in osteoclast differen-tiation has beenwell described. Cytokines, such as TNF-a and IL-1b,strongly augment the RANK-RANKL pathway and can also directlyinduce the differentiation of osteoclasts via other pathways[64e66]. In the present study, there are four indications that themembrane-induced TNF-a expression augmented the RANKL-dependent osteoclastic differentiation. Firstly, RANKL upregula-tion corresponded with the upregulated level of osteoclastic dif-ferentiation marker, CR, at 3d, when TNF-a revealed the highestexpression level in the membrane-treated defect. Secondly, at 6d,although the sham defect demonstrated higher levels of RANKL andRANK, as compared to membrane defects, this did not correspondwith any upregulation of CR or CatK. Importantly, at this time point(6d), the upregulated expression of TNF-a in the sham defect wasmuch less than that induced at 3d in the membrane-treated defect.Thirdly, apart from the upregulated RANKL levels, in the membraneat 3d and in the sham at 6d, the RANKL/OPG ratio did not show anydifferences, either between the two groups or among the differenttime periods. Since the RANKL/OPG ratio determines the RANKLeffect on osteoclasts, it is possible that the differences in osteo-clastic activity (CR and CatK) were mediated via the modulation ofTNF-a. Finally, at 3d, it was only the TNF-a that showed a statisti-cally positive correlation with the CR, suggesting an early role forthe pro-inflammatory cytokines in triggering osteoclasticdifferentiation.

In addition to the role of pro-inflammatory cytokines in aug-menting the RANKL pro-osteoclastic effect, they have been shown

A. Turri et al. / Biomaterials 84 (2016) 167e183180

directly to induce the osteoblastic expression of RANKL [67,68].Interestingly, it was the membrane-induced expression of TNF-a,but not IL-6, that corresponded with the upregulated expression ofRANKL. This observation is in agreement with previous in vitro datashowing that IL-1b and TNF-a, but not IL-6, stimulate RANKLexpression in human osteoblastic cells [68].

Taken together, the present and previous studies [30,32] ofbiomaterial-driven bone regeneration emphasize the early andcritical involvement of osteoclastic activity and bone remodeling.Some variation is observed in the biomaterial-induced osteoclasticand remodeling activities. During osseointegration, titanium im-plants with oxidized surfaces induced osteoclastic and boneremodeling activity directly via the increase in the RANKL/OPGexpression ratio [32]. In that study, the pro-inflammatory cyto-kines, TNF-a and IL-1b, did not appear to be involved in the oste-oclastic and bone remodeling activities, since both weresignificantly downregulated at the oxidized titanium implant. Thiscontrasts to the present study, where the RANKL/OPG ratio wasconstant, while the tuning was operated at the level of the pro-inflammatory cytokines.

4.4. Cellular and molecular events in the membrane during GBR

Since the membrane promoted bone regeneration in the defect,attention was paid to the presence of cells (demonstrated by his-tology) and the expression of selected genes and growth-promoting proteins in the membrane per se.

The analysis of cells in the membrane compartment demon-strated the ability of these cells to express BMP-2, FGF-2, TGF-b1and VEGF at the RNA level throughout the healing period. Further,the presence of two growth factors of importance, FGF-2 and BMP-2, was analyzed on the protein level. Whereas the existence of FGF-2 in the native membrane was confirmed, BMP-2 was not detectedin the membrane before implantation. The present results relatingto FGF-2 are in agreement with earlier studies of the membrane, asderived from porcine small intestine submucosa [69]. Importantly,the present WB data extends these observations, showing that themembrane not only maintains FGF-2 after implantation and duringthe constitutive phases of GBR but also accumulates the pro-osteogenic factor, BMP-2. An important question is if BMP-2 wasmainly secreted from cells in the membrane compartment orwhether it was also adsorbed from the surrounding fluids andtissues. The observations of BMP-2 and FGF-2 immunoreactive cellson the collagen membrane strands as well as interstitially in theretrieved membranes, suggest that a substantial amount of thedetected growth factors are derived from active cells afterrecruitment into the membrane compartment. Taken together, theqPCR, WB and IHC data strongly support that the BMP-2 protein inthe membrane is made de novo by cells in the membranecompartment.

Another intriguing observation was the early and strongexpression of VEGF in the membrane, with a different time coursecompared with the expression of the other growth factors.Together, these selected growth factors have potent individual andsynergistic effects on the different processes of bone healing,including inflammation and cell recruitment, bone formation andremodeling. Further studies are required in order to determine theexpression, secretion and biological effects of these factors,particularly in relation to the properties of the membrane.

One property is the collagenous structure and composition ofthe membrane, which is thought to contain cellular binding andactivation motifs, which enhance the binding, migration and acti-vation of cells with different phenotypes [70e72]. Another prop-erty, which is addressed in this study, is the ability of the presentECM-derived, collagenous membrane to maintain native and

in vivo accumulated growth factors. For instance, FGF-2 is wellknown to stimulate the migration and proliferation of differentcells, including endothelial cells, fibroblasts and osteoblasts, duringbone healing [73]. In addition, it plays a critical role in angiogenesisand mesenchymal cell mitogenesis [73,74]. Furthermore, both FGF-2 and BMP-2 are involved in the recruitment and formation ofosteoclasts, directly or indirectly via the increased angiogenesis andosteogenesis [75e77]. It is therefore likely that FGF-2 and BMP-2proteins, together with the collagenous support structure of themembrane, are important factors in the migration, proliferationand differentiation of different cells in the membrane. In the pre-sent study, different cell populations could be identified inside themembrane. For instance, at 6d, a combination of inflammatorycells, fibroblasts and stromal/mesenchymal-like cells was recog-nized, whereas, at 28d, stromal cells constituted the dominantpopulation. With respect to the early time point (6d), themembrane-infiltrated cells have recently been shown to compriseboth CD68-positive monocytes/macrophages as well as periostin-positive osteoprogenitors [78]. Moreover, in the present study,several blood vessels, as well as osteoclast-like cells, were observed,particularly on the outer surface of the membrane and along theborder between the membrane and the newly formed bone.

The inward migration of the inflammatory cells into the mem-brane could be also implicated in the membrane degradation, asthe ECM membrane is made of collagen and can be degraded bycellular activities [79e81]. Macroscopically, after 28d of implanta-tion, the retrieved membrane showed reasonable integrity, since itwas removed as one segment during the retrieval procedure.However, distortion and derangement in the collagen strands of themembrane were detected at histological level, suggesting a slowdegradation process. Furthermore, the osteoclast-like cellsobserved at the boundary between the membrane and the newlyformed bone might have contributed to the membrane degrada-tion, in addition to their remodeling activity on the underlyingbone.

Taken together, it is evident that the inherited properties of theECM membrane have synergistically contributed to the bone pro-moting capacity of the membrane. Regarding the physicochemicalproperties, although no systematic chemical and micron- andnano-scale topographical analyses of the membrane was conduct-ed, it is likely that the porous structure of the membrane togetherwith the surface features and motifs, by which cells can recognizeand bind to the membrane, play a critical role. Regarding the bio-logical properties, it is likely that the gradual degradation of theECMmembrane, assisted by the specific populations of membrane-recruited cells (e.g. macrophages and osteoclasts), is accompaniedby a slow release of membrane degradation products, which mayinclude both endogenous and de novo synthesized potent growthfactors. It is suggested that for the design of future GBRmembranes,tailoring of such functionalities may provide additional cues toenhance bone regeneration for specific clinical indications.

4.5. The role of the membrane in GBR

Based on the observation of important biological processes in-side the membrane compartment, the question arose as to whetherthese activities were related to the expression of factors in thedefect, involved in cell recruitment, inflammation, bone formationand bone remodeling. Important support for the role of factorsexpressed in the membrane was provided by the results of thecorrelation analysis between factors in the membrane and thedefect. Firstly, the three pro-osteogenic signals (FGF-2, TGF-b1 andBMP-2) in the membrane demonstrated a positive correlation withthe bone formation and bone resorption genes in the defect. This isin line with the current insights suggesting that signals that

A. Turri et al. / Biomaterials 84 (2016) 167e183 181

provoke one arm of the remodeling process subsequently stimulatethe other process. A collagen sponge loaded with recombinantBMP-2 in a rat calvarial defect enhances both osteoclastogenesisand osteoblastogenesis [82]. Furthermore, in vitro and in vivostudies have revealed regulatory and inducing effects for FGF-2,TGF-b and BMP-2 on the differentiation of both osteoblasts andosteoclasts [83e87]. Secondly, the inverse relationship betweenTGF-b, in the membrane, and MCP-1, in the defect, suggests a rolefor TGF-b in regulating the inflammatory phase in the defect. TGF-bis a multifunctional growth factor, which has dual pro-inflammatory and anti-inflammatory effects [88]. Finally, the in-verse correlation between VEGF in the membrane and OC and CatKin the defect may indicate a switch in the intensities of the angio-genic and bone remodeling activities.

4.6. Methodological considerations

One major limitation in the present study was the difficultydetermining the differential effect of the membrane on the cellularand molecular activities within the defect, at different distancesfrom the membrane. This is based on the fact that the molecularactivities were analyzed in the entire defect site, which may maskpossible differences in specific regions of the defect. Further studieswill require site-specific analyses, employing, for example, in-situhybridization, or laser micro-dissection to obtain samples fromdifferent regions of the defect for subsequent qPCR analysis. Asimilar approach could also be used in a kinetic study in which thelaser sampling for qPCR is performed, site by site, in a directiontowards the membrane. The latter strategy can be employed inparallel with histological and histomorphometric scoring of thedegree of membrane degradation, over a prolonged period of time.This will provide essential evidence relating to whether themembrane contributes to its own degradation, while establishingan optimal microenvironment for bone formation and remodelingand ultimately complete defect restitution.

5. Conclusion

The present results show and describe the sequence of biolog-ical events during GBR, encompassing the inwardmigration of cells,which acquire phenotypes that contribute to a pro-osteogenic andremodeling microenvironment, primarily underneath the mem-brane. The molecular mechanisms whereby this pro-osteogenicand remodeling microenvironment contributes to bone regenera-tion and defect healing involve the expression and accumulation ofpro-osteogenic factors within the membrane, which trigger themolecular cascade for rapid, higher bone formation and remodelingin the underlying defect. The present results provide cellular andmolecular evidence suggesting a novel role for the membraneduring GBR, by acting as a bioactive compartment rather thanmerely a passive barrier. These biological findings provide an op-portunity to intentionally tailor the composition and structure ofthe membrane to enhance GBR in different bone applications.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgments

This study was supported by the BIOMATCELL VINN ExcellenceCenter of Biomaterials and Cell Therapy, the V€astra G€otaland Re-gion, the Swedish Research Council (K2015-52X-09495-28-4), anLUA/ALF grant, the Stiftelsen Handlanden Hjalmar Svensson, theVilhelm and Martina Lundgren Vetenskapsfond, the IngaBritt and

Arne Lundberg Foundation and the Area of Advance Materials ofChalmers and GU Biomaterials within the Strategic Research Areainitiative launched by the Swedish Government. The membraneused in the present study was kindly provided by Keystone Dental,Boston, USA. IE was the recipient of an international scholarship forPhD studies provided by the Ministry of Higher Education, Libya.The grant providers were not involved in the study design, dataacquisition, interpretation, writing and submission of the article.The authors confirm that there are no known conflicts of interestassociated with this publication and there has been no significantfinancial support for this work that could have influenced itsoutcome.

References

[1] C. Dahlin, A. Linde, J. Gottlow, S. Nyman, Healing of bone defects by guidedtissue regeneration, Plast. Reconstr. Surg. 81 (1988) 672e676.

[2] M. Retzepi, N. Donos, Guided bone regeneration: biological principle andtherapeutic applications, Clin. Oral Implants Res. 21 (2010) 567e576.

[3] L. Canullo, V.A. Malagnino, Vertical ridge augmentation around implants by e-PTFE titanium-reinforced membrane and bovine bone matrix: a 24- to 54-month study of 10 consecutive cases, Int. J. Oral Maxillofac. Implants 23(2008) 858e866.

[4] V. Thomaidis, K. Kazakos, D.N. Lyras, I. Dimitrakopoulos, N. Lazaridis,D. Karakasis, et al., Comparative study of 5 different membranes for guidedbone regeneration of rabbit mandibular defects beyond critical size, Med. Sci.Monit. 14 (2008) Br67er73.

[5] E.E. Machtei, The effect of membrane exposure on the outcome of regenera-tive procedures in humans: a meta-analysis, J. Periodontol. 72 (2001)512e516.

[6] H.L. Wang, M.J. Carroll, Guided bone regeneration using bone grafts andcollagen membranes, Quintessence Int. 32 (2001) 504e515.

[7] P. Bunyaratavej, H.L. Wang, Collagen membranes: a review, J. Periodontol. 72(2001) 215e229.

[8] L.C. Parrish, T. Miyamoto, N. Fong, J.S. Mattson, D.R. Cerutis, Non-bio-absorbable vs. bioabsorbable membrane: assessment of their clinical efficacyin guided tissue regeneration technique. A systematic review, J. Oral Sci. 51(2009) 383e400.

[9] S.F. Badylak, R. Record, K. Lindberg, J. Hodde, K. Park, Small intestinal sub-mucosa: a substrate for in vitro cell growth, J. Biomater. Sci. Polym. Ed. 9(1998) 863e878.

[10] J.P. Hodde, S.F. Badylak, A.O. Brightman, S.L. Voytik-Harbin, Glycosamino-glycan content of small intestinal submucosa: a bioscaffold for tissuereplacement, Tissue Eng. 2 (1996) 209e217.

[11] J. Hodde, A. Janis, D. Ernst, D. Zopf, D. Sherman, C. Johnson, Effects of sterili-zation on an extracellular matrix scaffold: part I. Composition and matrixarchitecture, J. Mater Sci. Mater Med. 18 (2007) 537e543.

[12] B.P. Kropp, B.L. Eppley, C.D. Prevel, M.K. Rippy, R.C. Harruff, S.F. Badylak, et al.,Experimental assessment of small intestinal submucosa as a bladder wallsubstitute, Urology 46 (1995) 396e400.

[13] C.D. Prevel, B.L. Eppley, D.J. Summerlin, J.R. Jackson, M. McCarty, S.F. Badylak,Small intestinal submucosa: utilization for repair of rodent abdominal walldefects, Ann. Plast. Surg. 35 (1995) 374e380.

[14] S.F. Badylak, R. Tullius, K. Kokini, K.D. Shelbourne, T. Klootwyk, S.L. Voytik, etal., The use of xenogeneic small intestinal submucosa as a biomaterial forAchilles tendon repair in a dog model, J. Biomed. Mater Res. 29 (1995)977e985.

[15] G.C. Lantz, S.F. Badylak, M.C. Hiles, A.C. Coffey, L.A. Geddes, K. Kokini, et al.,Small intestinal submucosa as a vascular graft: a review, J. Invest Surg. 6(1993) 297e310.

[16] S. Peel, H. Chen, R. Renlund, S. Badylak, R. Kandel, Formation of a SIS-cartilagecomposite graft in vitro and its use in the repair of articular cartilage defects,Tissue Eng. 4 (1998) 143e155.

[17] M.A. Suckow, S.L. Voytik-Harbin, L.A. Terril, S.F. Badylak, Enhanced boneregeneration using porcine small intestinal submucosa, J. Invest Surg. 12(1999) 277e287.

[18] K. Donath, G. Breuner, A method for the study of undecalcified bones andteeth with attached soft tissues. The Sage-Schliff (sawing and grinding)technique, J. Oral Pathol. 11 (1982) 318e326.

[19] J. Vandesompele, K. De Preter, F. Pattyn, B. Poppe, N. Van Roy, A. De Paepe, etal., Accurate normalization of real-time quantitative RT-PCR data by geo-metric averaging of multiple internal control genes, Genome Biol. 3 (2002).Research0034.

[20] C.L. Andersen, J.L. Jensen, T.F. Orntoft, Normalization of real-time quantitativereverse transcription-PCR data: a model-based variance estimation approachto identify genes suited for normalization, applied to bladder and colon cancerdata sets, Cancer Res. 64 (2004) 5245e5250.

[21] M.W. Pfaffl, A new mathematical model for relative quantification in real-timeRT-PCR, Nucleic Acids Res. 29 (2001) e45.

[22] P. Colangelo, A. Piattelli, S. Barrucci, P. Trisi, G. Formisano, S. Caiazza, Bone

A. Turri et al. / Biomaterials 84 (2016) 167e183182

regeneration guided by resorbable collagen membranes in rabbits: a pilotstudy, Implant Dent. 2 (1993) 101e105.

[23] H. He, J. Huang, G. Chen, Y. Dong, Application of a new bioresorbable film toguided bone regeneration in tibia defect model of the rabbits, J. Biomed. MaterRes. A 82 (2007) 256e262.

[24] J.Y. Kim, B.E. Yang, J.H. Ahn, S.O. Park, H.W. Shim, Comparable efficacy of silkfibroin with the collagen membranes for guided bone regeneration in ratcalvarial defects, J. Adv. Prosthodont 6 (2014) 539e546.

[25] R.D. Mundell, M.P. Mooney, M.I. Siegel, A. Losken, Osseous guided tissueregeneration using a collagen barrier membrane, J. Oral Maxillofac. Surg. 51(1993) 1004e1012.

[26] L.C. Hofbauer, M. Schoppet, Clinical implications of the osteoprotegerin/RANKL/RANK system for bone and vascular diseases, Jama 292 (2004)490e495.

[27] B.R. Troen, Molecular mechanisms underlying osteoclast formation and acti-vation, Exp. Gerontol. 38 (2003) 605e614.

[28] T. Suda, Hematopoiesis and bone remodeling, Blood 117 (2011) 5556e5557.[29] T.T. To, P.E. Witten, J. Renn, D. Bhattacharya, A. Huysseune, C. Winkler, Rankl-

induced osteoclastogenesis leads to loss of mineralization in a medaka oste-oporosis model, Development 139 (2012) 141e150.

[30] I. Elgali, K. Igawa, A. Palmquist, M. Lenneras, W. Xia, S. Choi, et al., Molecularand structural patterns of bone regeneration in surgically created defectscontaining bone substitutes, Biomaterials 35 (2014) 3229e3242.

[31] O. Omar, S. Svensson, N. Zoric, M. Lenneras, F. Suska, S. Wigren, et al., In vivogene expression in response to anodically oxidized versus machined titaniumimplants, J. Biomed. Mater Res. A 92 (2010) 1552e1566.

[32] M. Lenneras, A. Palmquist, B. Norlindh, L. Emanuelsson, P. Thomsen, O. Omar,Oxidized titanium implants enhance osseointegration via mechanismsinvolving RANK/RANKL/OPG regulation, Clin. Implant Dent. Relat. Res. 17(Suppl. 2) (2015 Oct) e486e500, http://dx.doi.org/10.1111/cid.12276.

[33] L. Kreja, R.E. Brenner, A. Tautzenberger, A. Liedert, B. Friemert, C. Ehrnthaller,et al., Non-resorbing osteoclasts induce migration and osteogenic differenti-ation of mesenchymal stem cells, J. Cell Biochem. 109 (2010) 347e355.

[34] L. Pederson, M. Ruan, J.J. Westendorf, S. Khosla, M.J. Oursler, Regulation ofbone formation by osteoclasts involves Wnt/BMP signaling and the chemo-kine sphingosine-1-phosphate, Proc. Natl. Acad. Sci. U. S. A. 105 (2008)20764e20769.

[35] S.S. Sinclair, K.J. Burg, Effect of osteoclast co-culture on the differentiation ofhuman mesenchymal stem cells grown on bone graft granules, J. Biomater.Sci. Polym. Ed. 22 (2011) 789e808.

[36] S. Takeshita, T. Fumoto, K. Matsuoka, K.A. Park, H. Aburatani, S. Kato, et al.,Osteoclast-secreted CTHRC1 in the coupling of bone resorption to formation,J. Clin. Invest 123 (2013) 3914e3924.

[37] A. Teti, Mechanisms of osteoclast-dependent bone formation, Bonekey Rep. 2(2013) 449.

[38] C. Zhao, N. Irie, Y. Takada, K. Shimoda, T. Miyamoto, T. Nishiwaki, et al.,Bidirectional ephrinB2-EphB4 signaling controls bone homeostasis, CellMetab. 4 (2006) 111e121.

[39] M.E. Abdelgawad, K. Soe, T.L. Andersen, D.M. Merrild, P. Christiansen,P. Kjaersgaard-Andersen, et al., Does collagen trigger the recruitment of os-teoblasts into vacated bone resorption lacunae during bone remodeling? Bone67 (2014) 181e188.

[40] J.M. Delaisse, The reversal phase of the bone-remodeling cycle: cellular pre-requisites for coupling resorption and formation, Bonekey Rep. 3 (2014) 561.

[41] P. Rahimi, C.Y. Wang, P. Stashenko, S.K. Lee, J.A. Lorenzo, D.T. Graves, Mono-cyte chemoattractant protein-1 expression and monocyte recruitment inosseous inflammation in the mouse, Endocrinology 136 (1995) 2752e2759.

[42] T. Yoshimura, E.A. Robinson, S. Tanaka, E. Appella, E.J. Leonard, Purificationand amino acid analysis of two human monocyte chemoattractants producedby phytohemagglutinin-stimulated human blood mononuclear leukocytes,J. Immunol. 142 (1989) 1956e1962.

[43] X. Li, L. Qin, M. Bergenstock, L.M. Bevelock, D.V. Novack, N.C. Partridge,Parathyroid hormone stimulates osteoblastic expression of MCP-1 to recruitand increase the fusion of pre/osteoclasts, J. Biol. Chem. 282 (2007)33098e33106.

[44] N.B. Binder, B. Niederreiter, O. Hoffmann, R. Stange, T. Pap, T.M. Stulnig, et al.,Estrogen-dependent and C-C chemokine receptor-2-dependent pathwaysdetermine osteoclast behavior in osteoporosis, Nat. Med. 15 (2009) 417e424.

[45] Z. Xing, C. Lu, D. Hu, Y.Y. Yu, X. Wang, C. Colnot, et al., Multiple roles for CCR2during fracture healing, Dis. Model Mech. 3 (2010) 451e458.

[46] C. Martin, P.C. Burdon, G. Bridger, J.C. Gutierrez-Ramos, T.J. Williams,S.M. Rankin, Chemokines acting via CXCR2 and CXCR4 control the release ofneutrophils from the bone marrow and their return following senescence,Immunity 19 (2003) 583e593.

[47] M. Kucia, J. Ratajczak, R. Reca, A. Janowska-Wieczorek, M.Z. Ratajczak, Tissue-specific muscle, neural and liver stem/progenitor cells reside in the bonemarrow, respond to an SDF-1 gradient and are mobilized into peripheralblood during stress and tissue injury, Blood Cells Mol. Dis. 32 (2004) 52e57.

[48] C. Yellowley, CXCL12/CXCR4 signaling and other recruitment and homingpathways in fracture repair, Bonekey Rep. 2 (2013) 300.

[49] W. Zhu, G. Liang, Z. Huang, S.B. Doty, A.L. Boskey, Conditional inactivation ofthe CXCR4 receptor in osteoprecursors reduces postnatal bone formation dueto impaired osteoblast development, J. Biol. Chem. 286 (2011) 26794e26805.

[50] L.M. Wright, W. Maloney, X. Yu, L. Kindle, P. Collin-Osdoby, P. Osdoby, Stromalcell-derived factor-1 binding to its chemokine receptor CXCR4 on precursor

cells promotes the chemotactic recruitment, development and survival ofhuman osteoclasts, Bone 36 (2005) 840e853.

[51] T. Kitaori, H. Ito, E.M. Schwarz, R. Tsutsumi, H. Yoshitomi, S. Oishi, et al.,Stromal cell-derived factor 1/CXCR4 signaling is critical for the recruitment ofmesenchymal stem cells to the fracture site during skeletal repair in a mousemodel, Arthritis Rheum. 60 (2009) 813e823.

[52] M. Shahnazari, V. Chu, T.J. Wronski, R.A. Nissenson, B.P. Halloran, CXCL12/CXCR4 signaling in the osteoblast regulates the mesenchymal stem cell andosteoclast lineage populations, Faseb J. 27 (2013) 3505e3513.

[53] S. Otsuru, K. Tamai, T. Yamazaki, H. Yoshikawa, Y. Kaneda, Circulating bonemarrow-derived osteoblast progenitor cells are recruited to the bone-formingsite by the CXCR4/stromal cell-derived factor-1 pathway, Stem Cells 26 (2008)223e234.

[54] Y. Aachoui, S.K. Ghosh, Extracellular matrix from porcine small intestinalsubmucosa (SIS) as immune adjuvants, PLoS One 6 (2011) e27083.

[55] S.F. Badylak, J.E. Valentin, A.K. Ravindra, G.P. McCabe, A.M. Stewart-Akers,Macrophage phenotype as a determinant of biologic scaffold remodeling,Tissue Eng. Part A 14 (2008) 1835e1842.

[56] O. Omar, M. Lenneras, S. Svensson, F. Suska, L. Emanuelsson, J. Hall, et al.,Integrin and chemokine receptor gene expression in implant-adherent cellsduring early osseointegration, J. Mater Sci. Mater Med. 21 (2010) 969e980.

[57] P. Thomsen, C. Gretzer, Macrophage interactions with modified materialsurfaces, Curr. Opin. Solid State Mater. Sci. 5 (2001) 163e176.

[58] W. Bocker, D. Docheva, W.C. Prall, V. Egea, E. Pappou, O. Rossmann, et al., IKK-2 is required for TNF-alpha-induced invasion and proliferation of humanmesenchymal stem cells, J. Mol. Med. Berl. 86 (2008) 1183e1192.

[59] K. Hess, A. Ushmorov, J. Fiedler, R.E. Brenner, T. Wirth, TNFalpha promotesosteogenic differentiation of human mesenchymal stem cells by triggering theNF-kappaB signaling pathway, Bone 45 (2009) 367e376.

[60] O.M. Omar, C. Graneli, K. Ekstrom, C. Karlsson, A. Johansson, J. Lausmaa, et al.,The stimulation of an osteogenic response by classical monocyte activation,Biomaterials 32 (2011) 8190e8204.

[61] L.C. Gerstenfeld, T.J. Cho, T. Kon, T. Aizawa, J. Cruceta, B.D. Graves, et al.,Impaired intramembranous bone formation during bone repair in the absenceof tumor necrosis factor-alpha signaling, Cells Tissues Organs 169 (2001)285e294.

[62] L.C. Gerstenfeld, D.M. Cullinane, G.L. Barnes, D.T. Graves, T.A. Einhorn, Fracturehealing as a post-natal developmental process: molecular, spatial, and tem-poral aspects of its regulation, J. Cell Biochem. 88 (2003) 873e884.

[63] T. Kon, T.J. Cho, T. Aizawa, M. Yamazaki, N. Nooh, D. Graves, et al., Expressionof osteoprotegerin, receptor activator of NF-kappaB ligand (osteoprotegerinligand) and related proinflammatory cytokines during fracture healing, J. BoneMin. Res. 16 (2001) 1004e1014.

[64] Y. Azuma, K. Kaji, R. Katogi, S. Takeshita, A. Kudo, Tumor necrosis factor-alphainduces differentiation of and bone resorption by osteoclasts, J. Biol. Chem.275 (2000) 4858e4864.

[65] N. Kim, Y. Kadono, M. Takami, J. Lee, S.H. Lee, F. Okada, et al., Osteoclast dif-ferentiation independent of the TRANCE-RANK-TRAF6 axis, J. Exp. Med. 202(2005) 589e595.

[66] T. Nomura, M. Aoyama, Y. Waguri-Nagaya, Y. Goto, M. Suzuki, K. Miyazawa, etal., Tumor necrosis factor stimulates osteoclastogenesis from human bonemarrow cells under hypoxic conditions, Exp. Cell Res. 321 (2014) 167e177.

[67] L.C. Hofbauer, A.E. Heufelder, Osteoprotegerin and its cognate ligand: a newparadigm of osteoclastogenesis, Eur. J. Endocrinol. 139 (1998) 152e154.

[68] L.C. Hofbauer, D.L. Lacey, C.R. Dunstan, T.C. Spelsberg, B.L. Riggs, S. Khosla,Interleukin-1beta and tumor necrosis factor-alpha, but not interleukin-6,stimulate osteoprotegerin ligand gene expression in human osteoblasticcells, Bone 25 (1999) 255e259.

[69] S.L. Voytik-Harbin, A.O. Brightman, M.R. Kraine, B. Waisner, S.F. Badylak,Identification of extractable growth factors from small intestinal submucosa,J. Cell Biochem. 67 (1997) 478e491.

[70] S. Badylak, A. Liang, R. Record, R. Tullius, J. Hodde, Endothelial cell adherenceto small intestinal submucosa: an acellular bioscaffold, Biomaterials 20 (1999)2257e2263.

[71] J. Hodde, A. Janis, M. Hiles, Effects of sterilization on an extracellular matrixscaffold: part II. Bioactivity and matrix interaction, J. Mater Sci. Mater Med. 18(2007) 545e550.

[72] T.B. McPherson, S.F. Badylak, Characterization of fibronectin derived fromporcine small intestinal submucosa, Tissue Eng. 4 (1998) 75e83.

[73] J.R. Lieberman, A. Daluiski, T.A. Einhorn, The role of growth factors in therepair of bone. Biology and clinical applications, J. Bone Jt. Surg. Am. 84-a(2002) 1032e1044.

[74] P.B. Saadeh, B.J. Mehrara, D.S. Steinbrech, J.A. Spector, J.A. Greenwald,G.S. Chin, et al., Mechanisms of fibroblast growth factor-2 modulation ofvascular endothelial growth factor expression by osteoblastic cells, Endocri-nology 141 (2000) 2075e2083.

[75] P. Collin-Osdoby, L. Rothe, S. Bekker, F. Anderson, Y. Huang, P. Osdoby, Basicfibroblast growth factor stimulates osteoclast recruitment, development, andbone pit resorption in association with angiogenesis in vivo on the chickchorioallantoic membrane and activates isolated avian osteoclast resorptionin vitro, J. Bone Min. Res. 17 (2002) 1859e1871.

[76] C.M. Cowan, O.O. Aalami, Y.Y. Shi, Y.F. Chou, C. Mari, R. Thomas, et al., Bonemorphogenetic protein 2 and retinoic acid accelerate in vivo bone formation,osteoclast recruitment, and bone turnover, Tissue Eng. 11 (2005) 645e658.

[77] K. Irie, C. Alpaslan, K. Takahashi, Y. Kondo, N. Izumi, Y. Sakakura, et al.,

A. Turri et al. / Biomaterials 84 (2016) 167e183 183

Osteoclast differentiation in ectopic bone formation induced by recombinanthuman bone morphogenetic protein 2 (rhBMP-2), J. Bone Min. Metab. 21(2003) 363e369.

[78] I. Elgali, A. Turri, W. Xia, B. Norlindh, A. Johansson, C. Dahlin, et al., Guidedbone regeneration using resorbable membrane and different bone sub-stitutes: early histological and molecular events, Acta Biomater. 29 (2016)409e423.

[79] S.M. Krane, Collagenases and collagen degradation, J. Invest Dermatol. 79(Suppl. 1) (1982) 83se86s.

[80] K.W. Owens, R.A. Yukna, Collagen membrane resorption in dogs: a compar-ative study, Implant Dent. 10 (2001) 49e58.

[81] Q. Ye, M.C. Harmsen, Y. Ren, R.A. Bank, The role of collagen receptors Endo180and DDR-2 in the foreign body reaction against non-crosslinked collagen andgelatin, Biomaterials 32 (2011) 1339e1350.

[82] R.Y. Kim, J.H. Oh, B.S. Lee, Y.K. Seo, S.J. Hwang, I.S. Kim, The effect of dose onrhBMP-2 signaling, delivered via collagen sponge, on osteoclast activation andin vivo bone resorption, Biomaterials 35 (2014) 1869e1881.

[83] L.R. Chaudhary, A.M. Hofmeister, K.A. Hruska, Differential growth factor

control of bone formation through osteoprogenitor differentiation, Bone 34(2004) 402e411.

[84] E.D. Jensen, L. Pham, C.J. Billington Jr., K. Espe, A.E. Carlson, J.J. Westendorf, etal., Bone morphogenic protein 2 directly enhances differentiation of murineosteoclast precursors, J. Cell Biochem. 109 (2010) 672e682.

[85] Y. Okada, A. Montero, X. Zhang, T. Sobue, J. Lorenzo, T. Doetschman, et al.,Impaired osteoclast formation in bone marrow cultures of Fgf2 null mice inresponse to parathyroid hormone, J. Biol. Chem. 278 (2003) 21258e21266.

[86] W. Sriarj, K. Aoki, K. Ohya, M. Takahashi, Y. Takagi, H. Shimokawa, TGF-beta indentin matrix extract induces osteoclastogenesis in vitro, Odontology 103(2015) 9e18.

[87] M. Szpalski, R. Gunzburg, Recombinant human bone morphogenetic protein-2: a novel osteoinductive alternative to autogenous bone graft? Acta Orthop.Belg. 71 (2005) 133e148.

[88] S. Sanjabi, L.A. Zenewicz, M. Kamanaka, R.A. Flavell, Anti-inflammatory andpro-inflammatory roles of TGF-beta, IL-10, and IL-22 in immunity and auto-immunity, Curr. Opin. Pharmacol. 9 (2009) 447e453.