Available online at www.sciencedirect.com

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

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7

http://dx.doi.org/10.10014-4827/& 2014 E

nCorresponding auE-mail addresses

heidi.holappa@helsintimo.sorsa@helsinki.fi

Research Article

Matrix metalloproteinase-8 regulates transforminggrowth factor-β1 levels in mouse tongue woundsand fibroblasts in vitro

Pirjo Åströma,g, Emma Piriläa,g, Riitta Lithoviusa, Heidi Heikkolab, Jarkko T. Korpic,Marcela Hernándezd, Timo Sorsab,h, Tuula Saloa,e,f,g,n

aDepartment of Diagnostics and Oral Medicine, Institute of Dentistry, University of Oulu, PO Box 5281, Oulu 90014, FinlandbDepartment of Oral and Maxillofacial Diseases, Institute of Dentistry, University of Helsinki, Helsinki University CentralHospital, Helsinki, FinlandcDepartment of Oral and Maxillofacial Surgery, Institute of Dentistry, University of Oulu, Oulu University Hospital, Oulu, FinlanddDepartment of Pathology and Laboratory of Periodontal Biology, Faculty of Dentistry, University of Chile, Santiago, ChileeOulu University Hospital, Oulu, Finland, Institute of Dentistry, University of Helsinki, Helsinki, FinlandfGraduate Program in Estomatopatologia, Piracicaba Dental School, University of Campinas, Piracicaba-São Paulo, BrazilgMedical Research Center (MRC), Oulu, Finland

hDivision of Periodontology, Department of Dental Medicine, Karolinska Institutet, Huddinge, Sweden

a r t i c l e i n f o r m a t i o n

Article Chronology:

Received 3 May 2014Received in revised form4 July 2014Accepted 8 July 2014Available online 15 July 2014

Keywords:

Matrix metalloproteinase-8MouseFibroblastsTransforming growth factor beta1

016/j.yexcr.2014.07.010lsevier Inc. All rights reser

thor.: pirjo.astrom@oulu.fi (P. Åki.fi (H. Heikkola), jarkko.k(T. Sorsa), tuula.salo@oul

a b s t r a c t

Matrix metalloproteinase-8 (MMP-8)-deficient mice (Mmp8-/-) exhibit delayed dermal woundhealing, but also partly contradicting results have been reported. Using the Mmp8-/- mice weinvestigated the role of MMP-8 in acute wound healing of the mobile tongue, and analyzed thefunction of tongue fibroblasts in vitro. Interestingly, in the early phase the tongue wounds ofMmp8-/- mice healed faster than those of wild type (wt) mice resulting in significant difference inwound widths (P¼0.001, 6–24 h). The Mmp8-/- wounds showed no change in myeloperoxidasepositive myeloid cell count, but the level of transforming growth factor (TGF)-β1 was significantlyincreased (P¼0.007) compared to the wt tongues. Fibroblasts cultured fromwt tongues expressedMMP-8 and TGF-β1. However, higher TGF-β1 levels were detected in Mmp8-/- fibroblasts, andMMP-8 treatment decreased phosphorylated Smad-2 levels and α-smooth muscle actin expres-

sion in these fibroblasts suggesting reduced TGF-β1 signaling. Consistently, a degradation ofrecombinant TGF-β1 by MMP-8 decreased its ability to activate the signaling cascade infibroblasts. Moreover, collagen gels with Mmp8-/- fibroblasts reduced more in size. We concludethat MMP-8 regulates tongue wound contraction rate and TGF-β1 levels. In vitro analyses suggestthat MMP-8 may also play a role in regulating TGF-β1 signaling of stromal fibroblasts.

& 2014 Elsevier Inc. All rights reserved.

ved.

ström), emma.pirila@oulu.fi (E. Pirilä), riitta.lithovius@gmail.com (R. Lithovius),orpi@helsinki.fi (J.T. Korpi), mhernandezrios@gmail.com (M. Hernández),u.fi (T. Salo).

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7218

Introduction

The principal mechanisms of wound healing in vertebrate tissuesare uniform, but also clear differences between various tissuesexist [1]. The fluid of the oral cavity contains various growthfactors, such as transforming growth factor-β1 (TGF-β1) that playsa protective role in mucosal immunity by enhancing the produc-tion of immunoglobulin A [2]. The skin wounds are characterizedby scarring in which TGF-β1 and fibroblasts play crucial roles,whereas mucosal wounds heal faster with lower inflammatoryresponse and minimal scar formation [1,3]. In addition, skinand oral buccal non-keratinized mucosal fibroblasts differ intheir MMP expressions, collagen contraction and matrix re-organization ability as well as gene profile [4–6]. However, theexact molecular mechanisms and factors that drive the differ-ences leading to enhanced healing of oral mucosal wounds arenot completely understood.Matrix metalloproteinases (MMPs) constitute a family of 23

human enzymes able to cleave numerous extracellular matrix(ECM) components and modify various bioactive non-matrixmolecules. In wound healing, individual MMPs have been shownto possess either destructive or constructive properties, depend-ing on spatial and temporal distribution as well as the phase andnature (acute versus chronic) of the healing process [7].MMP-8 (collagenase-2) effectively degrades type I collagen, but

it also modulates other ECM and non-matrix molecules. Poly-morphonuclear neutrophils are the main source of MMP-8, butalso e.g. fibroblasts have been shown to produce it [8,9]. MMP-8has been shown to be the predominant collagenase in healingskin wounds, with a 100-fold increase of MMP-8 in chronicwounds compared with acute wounds [10]. A previous studywith MMP-8-deficient (Mmp8-/-) mice reported delayed skinwound healing with altered neutrophil influx and increase inTGF-β1 signaling molecule phosphorylated Smad (PSmad)-2levels [11]. However, another study reported impaired rat skinwound healing in an excess of MMP-8 expression [12]. The partlycontradicting findings demonstrate the complex role of MMP-8 intissue repair. To further examine the role of MMP-8 in the tissuerepair process we now examined oral mucosal wound healing inthe keratinized masticatory mucosa of the tongue. Because of theunique features of inflammation and behavior of the cells of theoral cavity, we hypothesized that examining wound healing inthis site may reveal previously unknown functions of MMP-8. Weshowed, for the first time, the distinct effect of MMP-8 in skin andtongue wound healings. We also demonstrated a significant effectof MMP-8 on tongue wound TGF-β1 levels and the ability ofMMP-8 to regulate TGF-β1 activity and fibroblast behavior in vitro.

1 PE: protein extraction.

Materials and methods

Animals

Mmp8-/- mice [13] in C57BL/6J/129/SvJ background were used,with equal number of female and male mice (3–4 months old).Age-matched wild type (wt) C57BL/6J mice were used as controls.The experiments were approved by the Animal Care and UseCommittee at the University of Oulu or the National Animal Care

and Use Committee of Finland and carried out in accordance withthe principles of the Helsinki Declaration.

Experimental wounds and sample preparation

Tongue woundsMmp8-/- and wt mice were anesthetized by isoflurane inhalationand full thickness wounds were created by penetrating from thedorsal right side through to the ventral side of the anterior tonguewith a 1-mm biopsy punch at an equal distance from the midline.The mice were given buprenorfine (0.05 mg/kg) as an analgesic30 min before wounding and for 24 h after the wounding. Healingwas followed at 6 h, 24 h, 48 h, 4 d, and 11 d, after which animalswere sacrificed and their tongues were collected (anterior partwith the wound). The harvested tongues were fixed for histolo-gical and immunohistochemical analyses in 4% (w/v) formalin for20 h or snap-frozen in liquid nitrogen for protein extraction(indicated separately below in parentheses for each time point).Equal numbers of wt and Mmp8-/- mice were used at all timepoints. The 6-h wound experiments were implemented with 58mice (38 tongues fixed and 20 tongues frozen), 24-h woundexperiments with 60 mice (40 tongues fixed and 20 frozen), 48-hwound experiments with 40 mice (all fixed), 4-d wound experi-ments with 21 mice, and 11-d wound experiments with 21 mice(all fixed).

Serum samplesFor blood collection from the hind limb (�200 ml venous blood),the mice were anesthetized intraperitoneally with a mixture ofHypnorm (0.315 mg phentanylsitrateþ10 mg fluanisone/ml H20),Dormicum (midazolam 5 mg/ml) and aqua (1:1:2) 0.1 ml/10 g.The sera were separated by centrifugation (10 min, 5000 rpm)and the samples were immediately stored in �70 1C without anypreservatives.

Tongue sample processingThe fixed tongue samples were embedded in paraffin and cut into6-mm tissue sections. In order to detect the wounds uniformlyaside, the samples were oriented in paraffine so that the cutting ofthe tongue always occurred vertically from dorsal to ventral side.Frozen tongues were minced in liquid nitrogen, and incubated atþ4 1C for 2 h in a protein extraction (PE)1 buffer [50 mM Tris–HCl,10 mM CaCl2, 150 mM NaCl, 0.05% (v/v) Brij-35 (Sigma-Aldrich,St. Louis, MO, USA), a protease inhibitory cocktail without EDTA(Roche Diagnostics, IN, USA) and a phosphatase inhibitor (Sigma-Aldrich) as instructed, pH 7.4]. The samples were centrifuged at14,900g for 10 min and filtered with a Millexs-GV 0.22-mm filter(Millipore, Carrigtwohill, County Cork, Ireland). Protein concen-trations were determined with a DC Protein Assay kit (Bio-RadLaboratories Inc., Hercules, CA).

Skin woundsSkin wounds were performed as previously described [11] with atotal of 12 mice (Mmp8-/- and wt males). Shortly, the mice wereanesthetized with isoflurane inhalation, the dorsal hair wasshaved and the skin cleaned with 70% (v/v) ethanol. 8 mm skinwounds were created with the biopsy punch to either sides of theback skin of the mice and the wounds were photographed. After

2 ECIS: Electric Cell–Substrate Impedance Sensing.

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7 219

3 d, the wounds were photographed again and the mice weresacrificed. The mice were given buprenorphine (0.05 mg/kg) as ananalgesic 30 min before wounding and each day after thewounding.

Analysis of wound healing

Tongue woundsTongue wound healing was examined blinded by measuring theleading edges of newly formed epithelium (the widest site of thewound) from hematoxylin–eosin stained histological samples(at 50� magnification) using the standard Leica QWin program(Leica, Wetzlar, Germany). In some cases the wounding with thebiopsy needle was not completely successful leaving the residualtissue within the wound site. These failed (25%) of 6-h and 24-hwounds were excluded for further analyses.

Skin woundsThe skin wounds were photographed immediately after wound-ing and again after 3 d and the wound areas were measured fromphotographs using the standard Leica QWin program. The woundclosure was calculated as a percentage of recovery with respect tothe initial wound area.

Cell culturing

Age-matched male Mmp8-/- (n¼5) and wt mice (n¼5) weresacrificed and their tongues were cut and washed (2 min/PBS,2 min/water, 1 min/EtOH, 2 min/PBS). The tongues were sectionedonto petri dishes and cultured in DMEM (D5796, Sigma-Aldrich)supplemented with 10% (v/v) fetal bovine serum (FBS), 0.025 MHepes, 100 U/ml; 0.1 mg/ml Penicillin–Streptomycin (Sigma-Aldrich) and 0.25 mg/ml Amphotericin (Sigma-Aldrich). Fibro-blasts which grew out from the pieces of tissue (after 6–7 weeks)were transferred by trypsinization (0.05% trypsin/0.02% EDTA;Sigma) to cell culture flasks. The same passages of Mmp8-/- andwt mice fibroblasts, with the same starting number of cells, wereused in all experiments. The passage numbers of the cells variedfrom 5 to 11.

Sample collection and processing for Western blot andimmunostainingTo collect serum free media and cell samples cultured in serumfree media for Western blots, the cells were grown in Opti-MEMs

(Life technologies, CA, USA) for 3 d and media (20 ml) werecollected and concentrated by lyophilization. The cell proteinswere extracted in a PE1 buffer (described above) o/n at þ4 1C. Theprotein concentrations of media and cell extracts were deter-mined (described above). For immunostaining, 20,000 fibroblast/well were applied to the 8-well-slide and after 24 h or 48 h thecells were fixed in 4% (w/v) formalin for 10 min. In some cases thecells were first preincubated o/n and cultured with or without40 ng/ml recombinant human (rh) MMP-8 (ProteaImmun GmbH,Berlin, Germany) in medium described above, but FBS wasreplaced with 0.5% (w/v) lactalbumin.

RNA extraction and polymerase chain reaction

TRI Reagent (Sigma-Aldrich) was used to extract total RNA fromthe fibroblasts according to the manufacturer's instructions.

RNA was reverse transcribed into cDNA by using a SuperScripts

III First-Strand Synthesis System (Life technologies). The goodquality of RNA was confirmed in the agarose gel electrophoresisstained with ethidium bromide (not shown). PCR reactions wereperformed with AmpliTaq Golds DNA Polymerase (Life technol-ogies). For Mmp8 amplification, the forward primer 50-TCCAT-TACTGATCTTCCTCCACACAC-30, reverse primer 50-ATGTTGATGTCTGCTTCTCCCTGTAA-30 and an annealing temperature of 54 1Cwere used to get a product of 459 bp. For beta-actin amplification,the forward primer 50-GATATCGCTGCGCTGGTC-30, reverse primer50-ATGGGGTACTTCAGGGTCAG-30 and an annealing temperatureof 60 1C were used to get a product of 203 bp.

Cell experiments

Scratch assayFibroblasts (three wt and three Mmp8-/- cell lines) were used in ascratch assay performed in a 24-well-plate. 150,000 cells per well(six wells/cell line) were allowed to attach and a pipette tip wasused to wound the confluent cell layer. The wounds werephotographed immediately after wounding and 24 h afterwounding with an EVOS Digital Inverted Microscope (AdvancedMicroscopy Group, Bothell, WA, USA). After 24 h, wound widthswere measured blinded from 5 different points at the middle ofthe well by using the standard Leica QWin program (Leica).

Proliferation assayThe proliferation of tongue fibroblasts (three wt and threeMmp8-/- cell lines, 8000 cells per well, six wells/cell line in a96-well-plate), after 24 h of culturing, was determined with acommercial Cell Proliferation ELISA, BrdU (colorimetric) kit(Roche Diagnostics, Mannheim, Germany) according to the man-ufacturer's instructions and the plate was measured using VictorELISA reader (Perkin-Elmer, Turku, Finland) at 450 nm.

Electric Cell–Substrate Impedance Sensing (ECIS2) assayFibroblast adhesion, cell spreading, and confluence were studiedby using Electric Cell–Substrate Impedance Sensing (ECIS2)(Applied Biophysics Inc., Troy, NY, USA). The trypsinized cells(three wt and three Mmp8-/- cell lines, 140,000 cells in 400 ml)were seeded on an 8-well uncoated and BSA-coated assay plate,three wells per cell line. The coating was performed with 10 mg/ml BSA at 37 1C o/n and blocked with 1 mg/ml BSA at 37 1C for2 h. Finally, the wells were washed three times with PBS and filledwith medium prior to applying the cells. The cells were mon-itored o/n within the automated ECIS2 system that providesquantitative data by measuring impedance of the system reflect-ing the changes in cell behavior.

Collagen contraction assayCollagen gels were prepared as previously described [14]. Briefly,gels including 8 volumes of type I collagen (3.45 mg/ml; BDBioscience, Bedford, MA, USA), 1 volume of 10� DMEM (SigmaAldrich) and 1 volume of FBS with fibroblasts were prepared onice and 1 ml aliquots were added into a wells of 24 well plate. Inserum-free conditions, lactalbumin (0.5% w/v) was used with orwithout 1 ng/ml active TGF-β1 (7039B, Sigma Aldrich). Theexperiment was performed twice with four gels with Mmp8-/-

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7220

and four gels with wt cells (700,000 cells/gel) using two differentwt and Mmp8-/- cell lines. After 30 min polymerization, 0.5 mlDMEM (supplemented as described or FBS replaced with 0.5% (w/v) lactalbumin with and without 1 ng/ml TGF-β1) was added ontop of the polymerized gels. The following day the gels weredetached from the wells, and contraction was followed for up to48 h and photographed with an Epson Perfection V700 Proscanner (Seiko Epson Corporate Inc., Nagano, Japan). The averageof the vertical and horizontal diameters of the gels, measuredwith Adobe Photoshop CS5, version 12.0�32 (Adobe Systems Inc.CA, USA), was calculated.

TGF-β1 activity testMmp8-/- tongue fibroblasts were exposed to rhMMP-8 (0 ng/ml,5 ng/ml, 10 ng/ml or 50 ng/ml) (ProteaImmun GmbH) in serumfree DMEM (described above) o/n in 175 cm2 cell culture flasksand incubated for 24 h. The experiment was performed twice.Before adding to the cells, MMP-8 was activated with 1 mMp-aminophenylmercuric acetate (APMA; Sigma-Aldrich) at þ37 1Cfor 2 h. In another experiment, TGF-β1 was first treated with orwithout APMA activated MMP-8 as described later in the section‘In vitro TGF-β1 cleavage assay’. The reaction products wereincluded in serum free DMEM (final concentration of TGF-β1 inthe media was 10 ng/ml) and the media were added to the cells in75 cm2 cell culture flasks. After 6 h, the cells were harvestedand suspended into a PE1 buffer (described above). The proteinconcentrations were determined (described above) and thesamples were subjected to Western blot. The experiment wasperformed once.

Immunohistochemistry

Tissue samplesEndogenous peroxidase activity was blocked with 0.3% (v/v) H2O2

in methanol, followed by microwaving in a citrate buffer (RealTarget Retrieval Solution, Dako, Glostrup, Denmark) [for myelo-peroxidase (MPO) and α-smooth muscle actin (α-SMA) samples],or with 0.4% (w/v) pepsin [for TGF-β1 and lipopolysaccharide-induced CXC chemokine (LIX) samples]. Non-specific antibodybinding was inhibited by normal goat serum (Vector Laboratories,Burlingame, CA, USA) in 2% (w/v) BSA and then incubated witheither rabbit polyclonal MPO HP9048 (Hycult biotech, Uden, TheNetherlands; 0.5 µg/ml) or rabbit polyclonal α-SMA ab 5694(Abcam, Cambridge, UK; 0.5 mg/ml) or rabbit polyclonal GCP-2/LIX no. 500-P146 (Preprotech EC Ltd., London, UK; 1 mg/ml) ormouse monoclonal TGF-β1 clone #9016 (for latent and active TGF-β1) (R&D systems, Minneapolis, MN, USA; 12.5 mg/ml) antibodiesat 37 1C for 30 min, followed by incubation at 4 1C o/n. Thesamples were stained using Vectastatin Elite ABC kits with therespective biotinylated secondary antibodies (Vector Laboratories,Burlinghame, CA, USA) and 3-amino-9-ethylcarbazole as a chro-mogen, counterstained with Mayer's hematoxylin (Merck KGaA,Darmstadt, Germany or Sigma-Aldrich), and mounted in glycergel(Dako) or Aquamount (BDH Laboratory Supplies, Poole, UK). AnMCID™ Core 7.0 (InterFocus GmbH, Mering, Germany) imageanalysis system with a Leitz Aristoplan microscope (Leica) at100� magnification was used to determine the amount of MPOpositive myeloid cells (mainly polymorphonuclear neutrophils) inthe scanned area. The α-SMA stained wounds were divided intothree groups according to staining intensity (weak/moderate/

strong) using a Leica DMRB microscope (Leica) at 50� magnifica-tion. For TGF-β1 and LIX analysis, three randomly selected imagesof stromal cells and each wound's epithelial border were capturedusing AnalySIS software under an Olympus BX61 light microscopewith 40� /0.75 and 20� /0.50 NA objectives. The analysis wasperformed using Fiji Is Just ImageJ software, a macro writtenforthe public domain. Image preprocessing included median andwatershed filtering for total cells and maximum entropy thresh-olding for immunoreactive cells.

CellsThe formalin-fixed cells in 8-well-slides were air dried for 30 minand rinsed in PBS. The cells were treated with Dako Real™Peroxidase-Blocking Solution (Dako, Glostrup, Denmark) for30 min at room temperature and washed twice with PBS. Therabbit polyclonal antibody to α-SMA (ab 5694, Abcam, Cambridge,UK; 0.5 mg/ml) or the polyclonal SP32 antibody to type IIIprocollagen [15] (0.103 mg/ml) were applied and allowed to bindat 37 1C for 30 min and at 4 1C o/n followed by washing with PBS.Finally, an EnVision™ Peroxidase/DAB Detection System, Rabbit/Mouse was applied for 30 min at room temperature and the slideswere mounted in Aquamount (BDH Laboratory Supplies, Poole,England). Two samples for each cell line were evaluated for typeIII collagen staining intensity (weak, moderate, strong) by aresearcher. The α-SMA stained cells (three wells with and withoutrhMMP-8, four pictures/well) were photographed under LeicaDMRB microscope (Leica) at 100� magnification (average of 42cells in the view). The percentages of cells with weak, moderateand strong α-SMA expressions were analyzed. All the histologicaland immunohistochemical analyses were performed blinded.

Serum TGF-β1

Total levels of circulating TGF-β1 were determined with aQuantikine TGF-β1 kit (R&D Systems, Minneapolis, MN, USA)according to the manufacturer's instructions. Briefly, 50 ml ofstandards, controls, and activated mouse serum samples (alto-gether 24 age-matched mice; 6 mice in each group, wt/Mmp8-/-,male/female) were added into the wells. Before the analysis, thesamples of two different mice were pooled together. Finally, theplate was read at 450 nm in a Victor ELISA reader (Perkin-Elmer).The reference wavelength correction was not used.

Gelatin zymography

The wound protein extraction samples were analyzed for theirgelatinase levels using purified MMP-2 and -9 as controls aspreviously described [16]. Protein samples were mixed withsample buffer and separated by 10% (w/v) sodium dodecylsulfate-polyacrylamide-gel electrophoresis (SDS-PAGE) with1 mg/ml of gelatin (Invitrogen, Carlsbad, CA, USA). After washing,the gels were incubated in 50 mM Tris–HCl, 5 mM CaCl2, l mMZnCl2, and 0.02% (v/v) NaN3 (pH 7.5) overnight at 37 1C andstained with 0.5% (w/v) Coomassie Blue. The intensities of thebands were quantified using optical densitometry and QuantityOne software (Bio-Rad Model GS-700 Imaging Densitometer; Bio-Rad Laboratories Inc.).

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7 221

In vitro TGF-β1 cleavage assay

200 ng of recombinant human TGF-β1 (rhTGF-β1) (R&D Systems)was incubated with 200 ng of rhMMP-8 (ProteaImmun GmbH) ina buffer [50 mM Tris–HCl pH 7.8, 0.2 M NaCl, 0.75 mM CaCl2] withand without 0.25 mM GM6001 (Millipore, Temecula, CA) and1 mM APMA (Sigma-Aldrich) at 37 1C for 48 h (volume 10 ml). Theassay was performed three times and the samples were separatedin SDS-PAGE and stained with Coomassie Blue or subjected toWestern blotting or used in the cell experiments described above.

Western blotting

Western blotting was performed as previously described [17].Rabbit antibody to phosphorylated Smad-2 (PSmad-2) (Ser465/467) (Cell Signalling Technologies Inc., USA; 68 ng/ml) was usedwith swine anti-rabbit antibody (Dako; 0.51 ng/ml) after separat-ing wound protein extractions (40 mg) or fibroblast proteinextractions (20 mg) under reducing conditions. After separatingthe TGF-β1þMMP-8 cleavage assay samples, fibroblast culturemedia (50 mg), cell protein extractions (20 mg) or wound proteinextractions (40 mg) in SDS-PAGE under non-reducing conditions,the membranes were subjected to mouse monoclonal TGF-β1clone #9016 (R&D systems; 2 µg/ml) antibody with secondaryrabbit anti-mouse antibody (Dako; 1.2 ng/ml). The band intensitieswere quantified as described above in ‘Gelatin zymography’section.

Statistical analysis

SPSS for Windows software version 18.0 (SPSS Inc., Chicago, Ill.,USA) and Stata V10 software were used for statistical analyses.After checking the normalization of the distributions, Student'st-test or a Mann–Whitney test was used to establish statisticalsignificance of differences between the two independent groupsat one time point at a time. For analyses at various time points, two-way ANOVA was used for independent values (wound healing rate)and repeated measures ANOVA for dependent values (collagen gelcontraction). A P valueo0.05 was considered statistically significant.

Results

Tongue wound healing was altered in Mmp8-/- mice

The tongue wounds of Mmp8-/- mice showed faster early phasenarrowing than those of wt mice, and after 6 h and 24 h ofwounding a statistically significant difference was observed inwound widths (P¼0.001) (Fig. 1a and b). After 48 h, the sizes ofthe wounds were impossible to evaluate reliably, because theunsuccessful wounds could not be excluded at that time point.Therefore, the presented result at this time point may not be asreliable as the results from the earlier time points, but impliedthat there was no dramatic change in the overall wound healingtime between the groups. In total, 89 wounds (75% of 6-h and24-h wounds) were suitable for further analyses. In our pilotexperiments (n¼13, not described), we did not observe any cleardifferences in the widths of wounds collected immediately afterwounding (the average size of the wounds �600 mm, range 400–900 mm). This excluded differences in the immediate wound

retractions between the groups and the initial wound widthswere expected to be the same. A delay in healing of skin woundsof male Mmp8-/- mice (Supplement Figure 1) similar to what hasbeen published previously, [11] was measured. Minimal amountof animals were used in the experiment and the result did notreach a statistical significance, but the detected trend in theresults was enough to confirm the difference in healings of skinand tongue wounds in Mmp8-/- mice. MMP-8 has previously beenshown to participate in neutrophil recruitment [13,18,19]. Herewe did not detect any difference in the amount of tongue woundMPO positive myeloid cells (mainly neutrophils) (SupplementFigure 2a and b) or in the levels of MMP-2 or MMP-9 between thegroups (Supplement Figure 2c).

TGF-β1 levels were increased in Mmp8-/- mouse wounds

TGF-β1 plays numerous crucial roles during tissue repair and ithas variable effects on the outcome of wound healing in differentcompartments in mouse [20]. In the tongue wounds, we foundTGF-β1 in the granulation tissue (GT) and in the epithelial borders(EB) at 6 h and 24 h after wounding (Fig. 1c). The immunoshis-tological levels of total TGF-β1 were significantly higher in the24-h wound beds of Mmp8-/- compared to wt mouse both in theepithelium (P¼0.03) and in the wound bed granulation tissue(P¼0.007, Fig. 1d). To confirm that particularly the levels of active,dimer form of TGF-β1 were changed, we performed Western blotanalysis using protein extracts of wounded tongues. The levels ofdimer form of TGF-β1 were low but detectable by Western blotand Mmp8-/- tongue wounds showed significantly higher levelsalready in the 6-h wounds of Mmp8-/- compared to wt mouse(P¼0.001, Fig. 1e). The levels of TGF-β1 in the 6-h wounds ingeneral were higher than in the 24-h wounds. In the 24-h woundsthe difference of immunoreactive TGF-β1 levels between thegenotypes was significant (P¼0.011, Fig. 1e), similar to immuno-histological findings (Fig. 1d). With immunohistochemistry, the24-h Mmp8-/- wounds showed higher TGF-β1 levels than the 6-hwounds, while Western blotting indicated higher level in the 6-hwounds (both genotypes). The wound area can be more specifi-cally analyzed with immunohistochemistry while the result fromWestern blotting represents TGF-β1 levels of more distributedarea around the wound. Also, by immunohistochemistry, latentand active forms of TGF-β1 cannot be separated while by Westernblot, only the dimer form was included in the analyses. Thesevariations between the two methods can explain the difference inthe timing of a ‘peak’ value for TGF-β1 levels. To ensure that theenhanced TGF-β1 levels in Mmp8-/- mice were related to thewound beds rather than the systematic changes in TGF-β1expression in these mice, we examined the total level of serumTGF-β1 from samples collected before and after the wounding.The circulating levels of TGF-β1 showed no difference betweenthe groups (Supplement Figure 3). MMP-8 has been shown toaffect neutrophil infiltration through the processing of LIX [13,19].Here we could not detect any difference in the LIX levels betweenthe groups based on the immunohistochemical analysis (Fig. 1d).The levels of PSmad-2—an intracellular mediator of TGF-β1signaling—were not significantly increased in the total proteinextracts of the Mmp8-/- mice 6-h or 24-h tongue wounds(Supplement Figure 4a). The α-SMA immunostaining (identifyingmyofibroblast-like cells) in the 6–48-h wounds was very weakand therefore was not suitable for the reliable comparisons

Fig. 1 – The tongue wound widths and TGF-β1 and LIX expression in Mmp8-/- and wt mice. Representative HE-stained tonguewounds at 24 h (a). The wound healing was analyzed by measuring the wound width from the leading edges of the healingepithelium (see example shown with a line in Mmp8-/- wound in a) using 50� magnification. Wound width is shown as anaverage value (with standard deviation) of the group (b). TGF-β1 and LIX were analyzed with immunohistochemical staining.Epithelial border (EB) (i) and granulation tissue (GT) (ii) sites of the TGF-β1 and LIX analyses, TGF-β1 expression in tongue woundsfrom wt and Mmp8-/- mice 24 h after wounding (iii–vi) (c). The results obtained from analyses of LIX and TGF-β1 stainingintensities with Image J. Results are expressed as medians (interquartile range), wt versus Mmp8-/-: Mann–Whitney test (6-hexperiments n¼38, 24-h experiments n¼40) (d). TGF-β1 was examined from 6-h and 24-h wound protein extracts with Westernblotting. Band intensities were analyzed with Quantity One software (e). (Mmp8-/-: matrix metalloproteinase-8 deficient, wt: wildtype, EB: epithelial border, GT: granulation tissue, TGF-β1: transforming growth factor-beta1, LIX: lipopolysaccharide-induced CXCchemokine.)

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7222

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7 223

between the groups. However, more myofibroblasts were presentduring the later maturating phase of wound healing at day four,and the results suggested a stronger expression of α-SMA inMmp8-/- wounds (Supplement Figure 4b). Due to low samplenumber in this time point, statistical analyses could not beperformed.

MMP-8 is able to regulate TGF-β1 signaling possibly bydirect inactivation of TGF-β1

Mmp8-/- mouse tongue wound beds showed increased levels ofTGF-β1, but did not show significant increase in PSmad-2. TGF-β1contributes to various cellular processes during wound healingand fibroblasts are one of the most important TGF-β1 target cells.Hence, increased TGF-β1 levels found in Mmp8-/- woundsprompted us to further examine the mechanism of MMP-8 actionby using mouse tongue fibroblasts. First we confirmed Mmp8expression in wt mouse tongue fibroblasts (Fig. 2a). Secondly, wefound that Mmp8-/- fibroblasts type III collagen expression wasslightly increased (Supplement Figure 5). The analyses of cellextracts and media confirmed that the tongue fibroblasts alsoexpressed TGF-β1 and suggested higher TGF-β1 levels in Mmp8-/-fibroblasts compared to wt fibroblasts (Fig. 2b). TGF-β1 wasdetected as a �25 kDa band and in addition, a larger molecularform of �40 kDa was found both in the cell extracts and in a lanewith rhTGF-β1 which was used as a control. Interestingly, asmaller molecular weight band (around 15 kDa) was detected infibroblasts that express MMP-8 (indicated as an arrow). Moreover,detectable levels of TGF-β1 were present only in the mediumderived from Mmp8-/- fibroblasts.

Since both in vivo wounds and fibroblasts in vitro suggestedthat TGF-β1 levels were altered when MMP-8 is absent, we nextanalyzed the action of MMP-8 on TGF-β1 with human recombi-nant proteins. We found that the active dimer form of rhTGF-β1 isa substrate for rhMMP-8, and it was modulated by MMP-8 in vitro(Fig. 2c). The degradation was visualized as a fading of TGF-β1monomer when reduced samples were analyzed by Coomassiestaining. In the Western blot, non-reduced samples were used (toconfirm antibody function according to the manufacturer'srecommendations) and TGF-β1 was detected as an active, dimerform. Unfortunately, despite several repeats, as shown here, wewere unable to identify visible TGF-β1 degradation products. Thismay be due to rapid proteolysis of small fragments or the levelsmay not reach the levels detectable with Western blot.

Next, we wanted to investigate the effect of MMP-8 on TGF-β1signaling in cell culture. We first confirmed a PSmad-2 signalingmolecule in the fibroblasts and consistent with the elevated levelsof TGF-β1, a 2.5-fold increase in PSmad-2 level was observed inMmp-8 fibroblasts (Fig. 2d). Then, by using Mmp8-/- fibroblasts,we wanted to further explore if PSmad-2 levels could be revertedto a lower level by MMP-8 treatment. Exogenously added MMP-8decreased the levels of PSmad-2: when 50 ng/ml of MMP-8 wasadded to fibroblasts, a diminished (0.3-fold) PSmad-2 level wasmeasured compared to that with no MMP-8 (1-fold) (Fig. 2e).Finally, we wanted to confirm that the in vitro modulation of TGF-β1 by MMP-8 truly inactivates the growth factor and its ability toactivate the signaling cascade in fibroblast cell culture. Weexposed the cells to intact and MMP-8 pretreated TGF-β1. Asexpected based on the previous experiment, PSmad-2 wasundetectable in the cells treated with MMP-8 alone. The signaling

increased when fibroblasts were exposed to TGF-β1, whereaspretreatment of TGF-β1 with MMP-8 reduced the detectedPSmad-2 to half of that (Fig. 2f). This experiment furthersupported that the decreased PSmad-2 level, after the cells wereexposed to exogenous MMP-8 (Fig. 2e), is most likely related tothe ability of MMP-8 to cleave endogenous TGF-β1.

Contraction of collagen gel with Mmp8-/- fibroblasts wasenhanced

We next wanted to compare the cell migration, proliferation, adhe-sion, spreading and confluence between the wt and Mmp8-/- tonguefibroblasts. Based on the various experiments, there were no differ-ences in any of these parameters between the groups (SupplementFigure 6a–c). Instead, in vitro collagen gel contraction (that can be aresult of remodeling of the fibrils or contraction of the collagen latticeby the cells) was significantly enhanced when Mmp8-/- tonguefibroblasts were embedded, compared to controls (P¼0.001)(Fig. 3a and b). Since the two cell lines of the same genotype showedvery similar results, we considered one gel with its cell population asan experimental unit. A total of eight gels containing two different wtand Mmp8-/- cell lines were analyzed. Adding TGF-β1 in serum-freecultures, the gels with Mmp8-/- fibroblasts were more contractedthan the gels with wt fibroblasts (P¼0.001) (Fig. 3c). In serum-freeconditions, the cells were not able to contract the gel to the sameextent as with serum (Fig. 3b and c). In addition, MMP-8 significantlychanged the α-SMA expression of the fibroblasts (Po0.001). Afterexposing the fibroblasts to exogenous MMP-8, most of the cells (88%)expressed only low amount of α-SMA (Supplement Figure 7).

Discussion

Our results suggest that MMP-8 participates in the mouse tonguewound healing process by reducing the rate of wound close up atthe early moments of healing. However, as has also been shownby Gutiérrez-Fernández et al. [11], the skin wounds of Mmp8-/-mice instead healed slightly slower than those of wt mice.Temporal comparisons cannot be made between the tongue andskin wound experiments, since different size wounds wereexamined and the time points analyzed were not the same.Unlike in skin [11], MMP-8 had no effect on inflammation inkeratinized tongue wounds detected here as MPO positiveinflammatory cell influx. Interestingly, we observed increasedTGF-β1 levels in the tongue wound beds. By in vitro experimentsusing tongue fibroblast cell cultures and recombinant proteins, weshowed that MMP-8 directly modulated TGF-β1 which led toreduced TGF-β1 signaling cascade.Changes in tongue wound closure occurred during the early

phases of wound healing. This suggests enhanced contractionability rather than changes in re-epithelialization in Mmp8-/-mice. However, we could not detect changes in the amount ofwound bed myofibroblasts, although significant increase in localTGF-β1 levels was found in the Mmp8-/- mice. Mmp8-/- mouseskin wounds showed increased PSmad-2 level, suggestingenhanced TGF-β1 signaling, although in that paper after 3 d,TGF-β1 levels were down regulated in Mmp8-/- wounds [11].Here, we were not able to detect any major changes in the levels ofPSmad-2 in the wounded tongues. Schrementi et al. showed thatmucosal wounds in tongue contain lower levels of TGF-β1 than skin

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7224

wounds, and they suggested that this most likely results in fasterhealing of oral mucosal wounds [21]. However, their study did notexamine how the tongue wounds or fibroblasts are affected by

increased levels of TGF-β1. Interestingly, TGF-β1 has been shown toaffect the outcome of wound healing differently in various compart-ments. The mice overexpressing TGF-β1 showed accelerated healing

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7 225

in partial thickness ear wounds, but reduced healing in full thicknessskin wounds [20]. In addition, the response of skin and oral buccalmucosa fibroblasts to TGF-β1 varies [22,23] suggesting that overalleffect of TGF-β1 on wound healing process depends on the woundlocation, including variations in provisional matrix and the specificfeatures of the cells involved.

Mucosal wounds in mice heal very fast. If one or only few timepoints are analyzed, the healing mechanisms at different phasesmay not be revealed. Therefore, we further examined the role ofMMP-8 in the regulation of TGF-β1 and fibroblast action usingtongue fibroblasts cell cultures. In line with our in vivo finding,lack of MMP-8 increased the levels of TGF-β1 in fibroblasts.However, unlike in vivo, we were able to detect increased levelsof PSmad-2 in the Mmp8-/- cells, indicating enhanced activationof signaling cascade that could be reverted by adding exogenousMMP-8. Consistent with our previous finding showing increasedtype III collagen content in Mmp8-/- mice tooth extraction sockets[24], we detected slightly enhanced type III pro-collagen produc-tion by Mmp8-/- fibroblasts in vitro. Interestingly, TGF-β1 has beenshown to up-regulate type III collagen in non-scarring fetalfibroblasts [25]. Although we did not detect changes in myofi-broblasts levels in mouse experiments, the enhanced ability ofMmp8-/- fibroblasts to contract the collagen gel in vitro suggeststhat MMP-8 is able to regulate the action of tongue fibroblasts.We assume that the regulation might be through TGF-β1, sinceadding TGF-β1 on Mmp8-/- fibroblasts caused different effects thanon wt fibroblasts. Since in vivo systems are far more complicated, itis possible that in our wound healing model, the effects of increasedlevels of TGF-β1 in Mmp8-/- mice affected fibroblasts only momen-tary at the very early steps of healing. In order to reveal the effectsof MMP-8 on TGF-β1 function in vivo would require planning andimplementing additional animal experiments focusing on TGF-β1signaling and myofibroblast function.

We have previously reported that TGF-β1 downregulates MMP-8 expression [26], but here we evidenced vice versa regulation:MMP-8 influenced on TGF-β1 actions, such as fibroblasts action incollagen gel in vitro and α-SMA levels. In the recent study by Shiet al., it was shown that adding IL-10 to TGF-β1 stimulated humandermal fibroblasts significantly up-regulated MMP-8 (and MMP-1) mRNA levels. Simultaneously, the level of myofibroblasts werereduced. They also found that IL-10 injection into mouse dermal

Fig. 2 – Effects of MMP-8 on TGF-β1 in vitro. MMP-8 expression inMmp8-/- and wt fibroblast cell extracts and concentrated cell culturPAGE and analyzed with Western blotting using monoclonal TGF-β�40 kDa band in fibroblast cell extracts. Also a small TGF-β1 fragm(indicated as an arrow) (b). TGF-β1 and MMP-8 were incubated at 3by boiling for 5 min with an electrophoresis sample buffer [250 mglycerol; 0.0098% (w/v) Bromophenol Blue]. The reaction productsstained with Coomassie Blue or under non-reducing conditions anantibody (c). In the reduced gel, TGF-β1 is observed as a monomer (The predicted molecular masses of TGF-β1 monomer is 12.8 kDa aculture media were analyzed with Western blotting using polyclonin the fibroblasts (d). The cells were treated with recombinant MMrecombinant intact TGF-β1 or TGF-β1 treated with MMP-8 (f) and tdensitometry analyses (RDV) (Quantity One software) were comparethe value 1 for the sample to which the comparisons were made.metalloproteinase-8 deficient, wt: wild type, TGF-β1: transformingactivator, p-aminophenylmercuric acetate, RDV¼ relative densitom

wounds had beneficial effects in preventing scar formation [27].The ability of MMP-8 to regulate TGF-β1 activity is also demon-strated in the recent breast cancer cell study. The authorsdemonstrated that MMP-8 indirectly inhibited TGF-β1 activitythrough the cleavage of decorin [28].Several studies have demonstrated MMP-mediated activation of

latent TGF-β1 [29–31]. Therefore, the accumulating evidence thatMMP-8 may participate in downregulating TGF-β1 activity isfascinating, especially when keeping in mind the diversity ofMMP-8 actions in various human conditions. Previously TGF-β1has been shown to be degraded by cathepsin K [32] and inchronic wound fluids, by neutrophil elastase [33]. Here, we didnot show if the degradation in vitro could have been inhibited alsoby endogenous inhibitors (e.g. various tissue inhibitors of metal-loproteinases; TIMPs), but clearly demonstrated the inhibition ofMMP-8 action using broad spectrum MMP inhibitor GM6001.However, the occurrence and function of the MMP-8-producedTGF-β1 fragments in vivo remain to be elucidated in the future.Previous studies demonstrate differences between the skin and

oral buccal mucosa fibroblasts [4–6] and most importantly theirvariable responses to excess of TGF-β1 [22,23]. The varyingresponse of oral and dermal fibroblasts to TGF-β1 inducedmyofibroblast differentiation involves differences in induction ofhyaluronan acid synthase [22]. Hyaluronan is a matrix polysac-charide that promotes fibroblast proliferation mediated by TGF-β1[23]. These findings led us to hypothesize that also MMP-8 mayhave differential effects on fibroblasts from various origins. Ourfinding that unlike in skin wounds [11], MMP-8 deficiency doesnot affect neutrophil levels or gelatinases in tongue woundssuggests that MMP-8 most likely exerts other than only inflam-matory cell related functions during wound healing. The lowMMP-9 expression levels detected here correspond to a previousstudy showing that MMP-9 gene expression is not up-regulated inthe lateral border of the tongue mucosa upon tissue injury [34].In addition, the histological features in the skin and tongue areclearly different, which also may explain the opposite action ofMMP-8 during wound healing in these tissues. Skin epithelium isconnected to loose adipose-containing connective tissue, whereasthe tongue epithelium is attached via dense collagen fiberbundles to the underlying striated muscle. These differencesin stroma most likely also lead to variable molecular mechanisms

wt fibroblasts was confirmed by PCR (a). The proteins frome media were separated under non-reducing conditions in SDS-1 antibody (b). TGF-β1 was detected as a �25 kDa band andent (o25 kDa) was observed in fibroblasts that express MMP-87 1C in a Tris/NaCl/CaCl2-buffer for 48 h. The reaction was haltedM Tris–HCl pH 6.8; 8% (w/v) sodium dodecyl sulfate; 40% (v/v)were separated under reducing conditions in SDS-PAGE andd analyzed with Western blotting using monoclonal TGF-β1�12–14 kDa) and in the non-reduced gel, as a dimer (�23 kDa).nd of dimer, 25 kDa (c). The Mmp8-/- and wt fibroblast cellal PSmad-2 antibody. PSmad-2 was observed as a 60 kDa bandP-8 (the results from the two experiments were similar) (e) orhe changes in PSmad-2 levels were examined. The values fromd to each other by relating the values to each other after giving(MMP-8: matrix metalloproteinase-8, Mmp8-/-: matrixgrowth factor-beta1, GM 6001: MMP inhibitor, APMA: MMP-etry value.)

Fig. 3 – Effects of MMP-8 and TGF-β1 on collagen contractionby fibroblasts. Cells (700,000/ml) were embedded into type Icollagen and the gel size was followed for up to 48 h (a). Gelcontraction was detected by measuring the diameters of thegels (b). Recombinant TGF-β1 was added within the gels andinto the serum-free culture media and gel contraction wasfollowed by measuring the diameters of the gels (c). (Mmp8-/-:matrix metalloproteinase-8 deficient, wt: wild type, TGF-β1:transforming growth factor-beta1.)

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7226

during wound healing, including divergent outcomes of actions ofproteases, such as MMP-8.We conclude that MMP-8 contributes to the mouse tongue

wound healing cascade at early stage and locally regulates thelevel of TGF-β1 during this process. However, at this point wewere unable to demonstrate if TGF-β1 is responsible for thechanges in the early wound healing rate observed in theMmp8-/- wounds. According to our in vitro experiments, MMP-8

participates in regulating TGF-β1 levels and signaling in fibro-blasts, and this may be related to its ability to inactivate thisgrowth factor. Since the expression of MMP-8 has been stronglyconnected to e.g. human chronic wounds and periodontal diseases[10,35], our proposed role of MMP-8 in the modulation of TGF-β1action should be further studied in clinical settings aiming toimprove the diagnostics and treatment of these diseases.

Competing interests

The authors state no conflicts of interest.

Acknowledgments

This study was supported by grant from the Academy of Finland(No. 135573) and salary support for P. Åström by the FinnishDoctoral Program in Oral Sciences (FINDOS). We thank Mrs. Eeva-Maija Kiljander and Mrs. Maija-Leena Lehtonen and Ms. SannaJuntunen for technical assistance, Dr. Tuire Salonurmi for herexpertise with the Mmp8-/- mice, Dr. Mika Hukkanen and theBiomedicum Imaging Unit, Institute of Biomedicine (University ofHelsinki) for designing the macro for the ImageJ software, MrHannu Vähänikkilä for his help in the statistical analyses, Prof.Carlos López-Otín for providing the Mmp8-/- mice and Prof. JuhaRisteli for providing the type III procollagen antibody, and Dr.Lauri Eklund for providing the PSmad-2 antibody.

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.yexcr.2014.07.010.

r e f e r e n c e s

[1] L. Häkkinen, V.J. Uitto, H. Larjava, Cell biology of gingival woundhealing, Periodontol 24 (2000) (2000) 127–152.

[2] S. Borsutzky, B.B. Cazac, J. Roes, C.A. Guzmán, TGF-beta receptorsignaling is critical for mucosal IgA responses, J. Immunol. 173(2004) 3305–3309.

[3] A.M. Szpaderska, J.D. Zuckerman, L.A. DiPietro, Differential injuryresponses in oral mucosal and cutaneous wounds, J. Dent. Res. 82(2003) 621–626.

[4] P. Stephens, K.J. Davies, T. al-Khateeb, J.P. Shepherd, D.W. Thomas,A comparison of the ability of intra-oral and extra-oral fibroblaststo stimulate extracellular matrix reorganization in a model ofwound contraction, J. Dent. Res. 75 (1996) 1358–1364.

[5] P. Stephens, K.J. Davies, N. Occleston, R.D. Pleass, C. Kon,J. Daniels, P.T. Khaw, D.W. Thomas, Skin and oral fibroblastsexhibit phenotypic differences in extracellular matrix reorgani-zation and matrix metalloproteinase activity, Br. J. Dermatol. 144(2001) 229–237.

[6] S. Enoch, M.A. Peake, I. Wall, L. Davies, J. Farrier, P. Giles,D. Kipling, P. Price, R. Moseley, D. Thomas, P. Stephens, ‘Young’oral fibroblasts are geno/phenotypically distinct, J. Dent. Res. 89(2010) 1407–1413.

[7] S.E. Gill, W.C. Parks, Metalloproteinases and their inhibitors:regulators of wound healing, Int. J. Biochem. Cell Biol. 40 (2008)1334–1347.

E X P E R I M E N T A L C E L L R E S E A R C H 3 2 8 ( 2 0 1 4 ) 2 1 7 – 2 2 7 227

[8] E. Dejonckheere, R.E. Vandenbroucke, C. Libert, Matrix metallo-proteinase8 has a central role in inflammatory disorders andcancer progression, Cytokine Growth Factor Rev. 22 (2011) 73–81.

[9] R. Hanemaaijer, T. Sorsa, Y.T. Konttinen, Y. Ding, M. Sutinen,H. Visser, V.W. van Hinsbergh, T. Helaakoski, T. Kainulainen,H. Rönkä, H. Tschesche, T. Salo, Matrix metalloproteinase-8 isexpressed in rheumatoid synovial fibroblasts and endothelialcells. Regulation by tumor necrosis factor-alpha and doxycycline,J. Biol. Chem. 272 (1997) 31504–31509.

[10] B.C. Nwomeh, H.X. Liang, I.K. Cohen, D.R. Yager, MMP-8 is thepredominant collagenase in healing wounds and nonhealingulcers, J. Surg. Res. 81 (1999) 189–195.

[11] A. Gutiérrez-Fernández, M. Inada, M. Balbín, A. Fueyo, A.S. Pitiot,A. Astudillo, K. Hirose, M. Hirata, S.D. Shapiro, A. Noël, Z. Werb, S.M. Krane, C. López-Otín, X.S. Puente, Increased inflammationdelays wound healing in mice deficient in collagenase-2 (MMP-8), FASEB J. 21 (2007) 2580–2591.

[12] P.L. Danielsen, A.V. Holst, H.R. Maltesen, M.R. Bassi, P.J. Holst, K.M.Heinemeier, J. Olsen, C.C. Danielsen, S.S. Poulsen, L.N. Jorgensen,M.S. Ågren, Matrix metalloproteinase-8 overexpression preventsproper tissue repair, Surgery 150 (2011) 897–906.

[13] M. Balbín, A. Fueyo, A.M. Tester, A.M. Pendas, A.S. Pitiot, A.Astudillo, C.M. Overall, S.D. Shapiro, C. López-Otín, Loss ofcollagenase-2 confers increased skin tumor susceptibility to malemice, Nat. Genet. 35 (2003) 252–257.

[14] S. Nurmenniemi, T. Sinikumpu, I. Alahuhta, S. Salo, M. Sutinen, M.Santala, J. Risteli, P. Nyberg, T. Salo, A novel organotypic modelmimics the tumor microenvironment, Am. J. Pathol. 175 (2009)1281–1291.

[15] J. Risteli, S. Niemi, P. Trivedi, O. Mäentausta, A.P. Mowat, L. Risteli,Rapid equilibrium radioimmunoassay for the amino-terminalpropeptide of human type III procollagen, Clin. Chem. 34 (1988)715–718.

[16] E. Pirilä, J.T. Korpi, T. Korkiamäki, T. Jahkola, A. Gutiérrez-Fernández, C. López-Otín, U. Saarialho-Kere, T. Salo, T. Sorsa,Collagenase-2 (MMP-8) and matrilysin-2 (MMP-26) expressionin human wounds of different etiologies, Wound Repair Regen.15 (2007) 47–57.

[17] J.T. Korpi, V. Kervinen, H. Mäklin, A. Väänänen, M. Lahtinen,E. Läärä, A. Ristimäki, G. Thomas, M. Ylipalosaari, P. Åström,C. López-Otín, T. Sorsa, S. Kantola, E. Pirilä, T. Salo, Collagenase-2(matrix metalloproteinase-8) plays a protective role in tonguecancer, Br. J. Cancer 98 (2008) 766–775.

[18] M. Lin, P. Jackson, A.M. Tester, E. Diaconu, C.M. Overall, J.E.Blalock, E. Pearlman, Matrix metalloproteinase-8 facilitates neu-trophil migration through the corneal stromal matrix by collagendegradation and production of the chemotactic peptide pro-gly-pro, Am. J. Pathol. 173 (2008) 144–153.

[19] A.M. Tester, J.H. Cox, A.R. Connor, A.E. Starr, R.A. Dean, X.S.Puente, C. López-Otín, C.M. Overall, LPS responsiveness andneutrophil chemotaxis in vivo require PMN MMP-8 activity, PLoSOne 2 (2007) e312.

[20] E.B. Tredget, J. Demare, G. Chandran, E.E. Tredget, L. Yang, A.Ghahary, Transforming growth factor-beta and its effect onreepithelialization of partial-thickness ear wounds in transgenicmice, Wound Repair Regen. 13 (2005) 61–67.

[21] M.E. Schrementi, A.M. Ferreira, C. Zender, L.A. DiPietro, Site-specific production of TGF-beta in oral mucosal and cutaneouswounds, Wound Repair Regen. 16 (2008) 80–86.

[22] S. Meran, D. Thomas, P. Stephens, J. Martin, T. Bowen, A. Phillips,R. Steadman, Involvement of hyaluronan in regulation of fibro-blast phenotype, J. Biol. Chem. 282 (2007) 25687–25697.

[23] S. Meran, D.W. Thomas, P. Stephens, S. Enoch, J. Martin, R.Steadman, A.O. Phillips, Hyaluronan facilitates transforminggrowth factor-beta1-mediated fibroblast proliferation, J. Biol.Chem. 283 (2008) 6530–6545.

[24] J.T. Korpi, P. Åström, N. Lehtonen, L. Tjäderhane, S. Kallio-Pulkkinen, M. Siponen, T. Sorsa, E. Pirilä, T. Salo, Healing ofextraction sockets in collagenase-2 (matrix metalloproteinase-8)-deficient mice, Eur. J. Oral Sci. 117 (2009) 248–254.

[25] K.J. Rolfe, L.M. Irvine, A.O. Grobbelaar, C. Linge, Differential geneexpression in response to transforming growth factor-beta1 byfetal and postnatal dermal fibroblasts, Wound Repair Regen. 15(2007) 897–906.

[26] H. Palosaari, J. Wahlgren, M. Larmas, H. Rönkä, T. Sorsa, T. Salo, L.Tjäderhane, The expression of MMP-8 in human odontoblastsand dental pulp cells is down-regulated by TGF-beta1, J. Dent.Res. 79 (2000) 77–84.

[27] J.H. Shi, H. Guan, S. Shi, W.X. Cai, X.Z. Bai, X.L. Hu, X.B. Fang, J.Q.Liu, K. Tao, X.X. Zhu, C.W. Tang, D.H. Hu, Protection against TGF-beta1-induced fibrosis effects of IL-10 on dermal fibroblasts andits potential therapeutics for the reduction of skin scarring, Arch.Dermatol. Res. 305 (2013) 341–352.

[28] C. Soria-Valles, A. Gutiérrez-Fernández, M. Guiu, B. Mari, A.Fueyo, R.R. Gomis, C. López-Otín, The anti-metastatic activity ofcollagenase-2 in breast cancer cells is mediated by a signalingpathway involving decorin and miR-21, Oncogene 33 (2014)3054–3063.

[29] Q. Yu, I. Stamenkovic, Cell surface-localized matrixmetalloproteinase-9 proteolytically activates TGF-beta and pro-motes tumor invasion and angiogenesis, Genes Dev. 14 (2000)163–176.

[30] M.A. Karsdal, L. Larsen, M.T. Engsig, H. Lou, M. Ferreras, A.Lochter, J.M. Delaissé, N.T. Foged, Matrix metalloproteinase-dependent activation of latent transforming growth factor-betacontrols the conversion of osteoblasts into osteocytes by block-ing osteoblast apoptosis, J. Biol. Chem. 277 (2002) 44061–44067.

[31] D. Mu, S. Cambier, L. Fjellbirkeland, J.L. Baron, J.S. Munger, H.Kawakatsu, D. Sheppard, V.C. Broaddus, S.L. Nishimura, Theintegrin alpha(v)beta8 mediates epithelial homeostasis throughMT1-MMP-dependent activation of TGF-beta1, J. Cell Biol. 157(2002) 493–507.

[32] D. Zhang, N. Leung, E. Weber, P. Saftig, D. Brömme, The effect ofcathepsin K deficiency on airway development and TGF-beta1degradation, Respir. Res. 12 (2011) 72.

[33] S.M. Chen, S.I. Ward, O.O. Olutoye, R.F. Diegelmann, I. KelmanCohen, Ability of chronic wound fluids to degrade peptidegrowth factors is associated with increased levels of elastaseactivity and diminished levels of proteinase inhibitors, WoundRepair Regen. 5 (1997) 23–32.

[34] L. Chen, Z.H. Arbieva, S. Guo, P.T. Marucha, T.A. Mustoe, L.A.DiPietro, Positional differences in the wound transcriptome ofskin and oral mucosa, BMC Genomics 11 (2010) 471.

[35] T. Sorsa, L. Tjäderhane, Y.T. Konttinen, A. Lauhio, T. Salo, H.M. Lee,L.M. Golub, D.L. Brown, P. Mäntylä, Matrix metalloproteinases:contribution to pathogenesis, diagnosis and treatment of peri-odontal inflammation, Ann. Med. 38 (2006) 306–321.